Packman Build Service PMBS

Changes of Revision 28

x265.changes Changed

@@ -1,4 +1,36 @@
 -------------------------------------------------------------------
+Thu Mar  1 23:14:47 UTC 2018 - zaitor@opensuse.org
+
+- Update to version 2.7:
+  * New features:
+    - option:`--gop-lookahead` can be used to extend the gop
+      boundary(set by `--keyint`). The GOP will be extended, if a
+      scene-cut frame is found within this many number of frames.
+    - Support for RADL pictures added in x265.
+    - option:`--radl` can be used to decide number of RADL pictures
+      preceding the IDR picture.
+  * Encoder enhancements:
+    - Moved from YASM to NASM assembler. Supports NASM assembler
+      version 2.13 and greater.
+    - Enable analysis save and load in a single run. Introduces two
+      new cli options `--analysis-save <filename>` and
+      `--analysis-load <filename>`.
+    - Comply to HDR10+ LLC specification.
+    - Reduced x265 build time by more than 50% by re-factoring
+      ipfilter.asm.
+  * Bug fixes:
+    - Fixed inconsistent output issue in deblock filter and
+      --const-vbv.
+    - Fixed Mac OS build warnings.
+    - Fixed inconsistency in pass-2 when weightp and cutree are
+      enabled.
+    - Fixed deadlock issue due to dropping of BREF frames, while
+      forcing slice types through qp file.
+- Bump soname to 151, also in baselibs.conf following upstream
+  changes.
+- Replace yasm with nasm BuildRequires following upstreams changes.
+
+-------------------------------------------------------------------
 Fri Dec 01 16:40:13 UTC 2017 - joerg.lorenzen@ki.tng.de
 
 - Update to version 2.6

x265.spec Changed

@@ -1,10 +1,10 @@
 # based on the spec file from https://build.opensuse.org/package/view_file/home:Simmphonie/libx265/
 
 Name:           x265
-%define soname  146
+%define soname  151
 %define libname lib%{name}
 %define libsoname %{libname}-%{soname}
-Version:        2.6
+Version:        2.7
 Release:        0
 License:        GPL-2.0+
 Summary:        A free h265/HEVC encoder - encoder binary
@@ -13,16 +13,19 @@
 Source0:        https://bitbucket.org/multicoreware/x265/downloads/%{name}_%{version}.tar.gz
 Patch0:         arm.patch
 Patch1:         x265.pkgconfig.patch
-BuildRequires:  gcc gcc-c++
+BuildRequires:  gcc
+BuildRequires:  gcc-c++
 BuildRequires:  cmake >= 2.8.8
 BuildRequires:  pkg-config
-BuildRequires:  yasm >= 1.2.0
-BuildRoot:      %{_tmppath}/%{name}-%{version}-build
+%if 0%{?suse_version} >=1500
+BuildRequires:  nasm >= 2.13
+%endif
 %if 0%{?suse_version} > 1310
 %ifarch x86_64
 BuildRequires:  libnuma-devel >= 2.0.9
 %endif
 %endif
+BuildRoot:      %{_tmppath}/%{name}-%{version}-build
 
 %description
 x265 is a free library for encoding next-generation H265/HEVC video
@@ -49,7 +52,7 @@
 streams. 
 
 %prep
-%setup -q -n %{name}_v%{version}
+%setup -q -n %{name}_%{version}
 %patch0 -p1
 %patch1 -p1

baselibs.conf Changed

x265_2.6.tar.gz/source/cmake/CMakeASM_YASMInformation.cmake Deleted

@@ -1,68 +0,0 @@
-set(ASM_DIALECT "_YASM")
-set(CMAKE_ASM${ASM_DIALECT}_SOURCE_FILE_EXTENSIONS asm)
-
-if(X64)
-    list(APPEND ASM_FLAGS -DARCH_X86_64=1)
-    if(ENABLE_PIC)
-        list(APPEND ASM_FLAGS -DPIC)
-    endif()
-    if(APPLE)
-        set(ARGS -f macho64 -m amd64 -DPREFIX)
-    elseif(UNIX AND NOT CYGWIN)
-        set(ARGS -f elf64 -m amd64)
-    else()
-        set(ARGS -f win64 -m amd64)
-    endif()
-else()
-    list(APPEND ASM_FLAGS -DARCH_X86_64=0)
-    if(APPLE)
-        set(ARGS -f macho -DPREFIX)
-    elseif(UNIX AND NOT CYGWIN)
-        set(ARGS -f elf32)
-    else()
-        set(ARGS -f win32 -DPREFIX)
-    endif()
-endif()
-
-if(GCC)
-    list(APPEND ASM_FLAGS -DHAVE_ALIGNED_STACK=1)
-else()
-    list(APPEND ASM_FLAGS -DHAVE_ALIGNED_STACK=0)
-endif()
-
-if(HIGH_BIT_DEPTH)
-    if(MAIN12)
-        list(APPEND ASM_FLAGS -DHIGH_BIT_DEPTH=1 -DBIT_DEPTH=12 -DX265_NS=${X265_NS})
-    else()
-        list(APPEND ASM_FLAGS -DHIGH_BIT_DEPTH=1 -DBIT_DEPTH=10 -DX265_NS=${X265_NS})
-    endif()
-else()
-    list(APPEND ASM_FLAGS -DHIGH_BIT_DEPTH=0 -DBIT_DEPTH=8 -DX265_NS=${X265_NS})
-endif()
-
-list(APPEND ASM_FLAGS "${CMAKE_ASM_YASM_FLAGS}")
-
-if(CMAKE_BUILD_TYPE MATCHES Release)
-    list(APPEND ASM_FLAGS "${CMAKE_ASM_YASM_FLAGS_RELEASE}")
-elseif(CMAKE_BUILD_TYPE MATCHES Debug)
-    list(APPEND ASM_FLAGS "${CMAKE_ASM_YASM_FLAGS_DEBUG}")
-elseif(CMAKE_BUILD_TYPE MATCHES MinSizeRel)
-    list(APPEND ASM_FLAGS "${CMAKE_ASM_YASM_FLAGS_MINSIZEREL}")
-elseif(CMAKE_BUILD_TYPE MATCHES RelWithDebInfo)
-    list(APPEND ASM_FLAGS "${CMAKE_ASM_YASM_FLAGS_RELWITHDEBINFO}")
-endif()
-
-set(YASM_FLAGS ${ARGS} ${ASM_FLAGS} PARENT_SCOPE)
-string(REPLACE ";" " " CMAKE_ASM_YASM_COMPILER_ARG1 "${ARGS}")
-
-# This section exists to override the one in CMakeASMInformation.cmake
-# (the default Information file). This removes the <FLAGS>
-# thing so that your C compiler flags that have been set via
-# set_target_properties don't get passed to yasm and confuse it.
-if(NOT CMAKE_ASM${ASM_DIALECT}_COMPILE_OBJECT)
-    string(REPLACE ";" " " STR_ASM_FLAGS "${ASM_FLAGS}")
-    set(CMAKE_ASM${ASM_DIALECT}_COMPILE_OBJECT "<CMAKE_ASM${ASM_DIALECT}_COMPILER> ${STR_ASM_FLAGS} -o <OBJECT> <SOURCE>")
-endif()
-
-include(CMakeASMInformation)
-set(ASM_DIALECT)

x265_2.6.tar.gz/source/cmake/CMakeDetermineASM_YASMCompiler.cmake Deleted

x265_2.6.tar.gz/source/cmake/CMakeTestASM_YASMCompiler.cmake Deleted

x265_2.6.tar.gz/source/cmake/FindYasm.cmake Deleted

x265_2.6.tar.gz/.hg_archival.txt -> x265_2.7.tar.gz/.hg_archival.txt Changed

x265_2.6.tar.gz/.hgtags -> x265_2.7.tar.gz/.hgtags Changed

x265_2.6.tar.gz/build/README.txt -> x265_2.7.tar.gz/build/README.txt Changed

@@ -9,7 +9,8 @@
 
 = Optional Prerequisites =
 
-1. Yasm 1.2.0 or later, to compile assembly primitives (performance)
+1. To compile assembly primitives (performance) 
+   a) If you are using release 2.6 or older, download and install Yasm 1.2.0 or later, 
 
    For Windows, download the latest yasm executable
    http://yasm.tortall.net/Download.html and copy the EXE into
@@ -33,6 +34,24 @@
    If cpu capabilities line says 'none!', then the encoder was built
    without yasm.
 
+   b) If you are building from the default branch after release 2.6, download and install nasm 2.13 or newer
+   
+   For windows and linux, you can download the nasm installer from http://www.nasm.us/pub/nasm/releasebuilds/?C=M;O=D.
+   Make sure that it is in your PATH environment variable (%PATH% in windows, and $PATH in linux) so that cmake
+   can find it.
+
+   Once NASM is properly installed, run cmake to regenerate projects. If you
+   do not see the below line in the cmake output, NASM is not in the PATH.
+
+   -- Found Nasm 2.13 to build assembly primitives
+
+   Now build the encoder and run x265 -V:
+
+   x265 [info]: using cpu capabilities: MMX, SSE2, ...
+
+   If cpu capabilities line says 'none!', then the encoder was built
+   without nasm and will be considerably slower for performance.
+
 2. VisualLeakDetector (Windows Only)
 
    Download from https://vld.codeplex.com/releases and install. May need

x265_2.6.tar.gz/doc/reST/api.rst -> x265_2.7.tar.gz/doc/reST/api.rst Changed

x265_2.6.tar.gz/doc/reST/cli.rst -> x265_2.7.tar.gz/doc/reST/cli.rst Changed

@@ -863,21 +863,22 @@
 sequence multiple times (presumably at varying bitrates). The encoder
 will not reuse analysis if slice type parameters do not match.
 
-.. option:: --analysis-reuse-mode <string|int>
+.. option:: --analysis-save <filename>
 
-	This option allows reuse of analysis information from first pass to second pass.
-	:option:`--analysis-reuse-mode save` specifies that encoder outputs analysis information of each frame.
-	:option:`--analysis-reuse-mode load` specifies that encoder reuses analysis information from first pass.
-	There is no benefit using load mode without running encoder in save mode. Analysis data from save mode is
-	written to a file specified by :option:`--analysis-reuse-file`. The amount of analysis data stored/reused
-	is determined by :option:`--analysis-reuse-level`. By reading the analysis data writen by an earlier encode
-	of the same sequence, substantial redundant work may be avoided. Requires cutree, pmode to be off. Default 0.
+	Encoder outputs analysis information of each frame. Analysis data from save mode is
+	written to the file specified. Requires cutree, pmode to be off. Default disabled.
+	
+.. option:: --analysis-load <filename>
+
+	Encoder reuses analysis information from the file specified. By reading the analysis data writen by
+	an earlier encode of the same sequence, substantial redundant work may be avoided. Requires cutree, pmode
+	to be off. Default disabled.
 
-	**Values:** off(0), save(1): dump analysis data, load(2): read analysis data
+	The amount of analysis data stored/reused is determined by :option:`--analysis-reuse-level`.
 
 .. option:: --analysis-reuse-file <filename>
 
-	Specify a filename for analysis data (see :option:`--analysis-reuse-mode`)
+	Specify a filename for `multi-pass-opt-analysis` and `multi-pass-opt-distortion`.
 	If no filename is specified, x265_analysis.dat is used.
 
 .. option:: --analysis-reuse-level <1..10>
@@ -1028,7 +1029,13 @@
 	Level 4 - uses the depth of the neighbouring/ co-located CUs TU depth 
 	to limit the 1st subTU depth. The 1st subTU depth is taken as the 
 	limiting depth for the other subTUs.
-
+	
+	Enabling levels 3 or 4 may cause a mismatch in the output bitstreams 
+	between option:`--analysis-save` and option:`--analysis-load`
+	as all neighbouring CUs TU depth may not be available in the 
+	option:`--analysis-load` run as only the best mode's information is 
+	available to it.
+	
 	Default: 0
 
 .. option:: --nr-intra <integer>, --nr-inter <integer>
@@ -1344,7 +1351,14 @@
 	This value represents the percentage difference between the inter cost and
 	intra cost of a frame used in scenecut detection. For example, a value of 5 indicates,
 	if the inter cost of a frame is greater than or equal to 95 percent of the intra cost of the frame,
-	then detect this frame as scenecut. Values between 5 and 15 are recommended. Default 5.	
+	then detect this frame as scenecut. Values between 5 and 15 are recommended. Default 5.
+	
+.. option:: --radl <integer>
+	
+	Number of RADL pictures allowed infront of IDR. Requires fixed keyframe interval.
+	Recommended value is 2-3. Default 0 (disabled).
+	
+	**Range of values: Between 0 and `--bframes`
 
 .. option:: --ctu-info <0, 1, 2, 4, 6>
 
@@ -1373,6 +1387,16 @@
 	Default 20
 
 	**Range of values:** Between the maximum consecutive bframe count (:option:`--bframes`) and 250
+.. option:: --gop-lookahead <integer>
+
+        Number of frames for GOP boundary decision lookahead. If a scenecut frame is found
+        within this from the gop boundary set by `--keyint`, the GOP will be extented until such a point,
+        otherwise the GOP will be terminated as set by `--keyint`. Default 0.
+
+        **Range of values:** Between 0 and (`--rc-lookahead` - mini-GOP length)
+
+        It is recommended to have `--gop-lookahaed` less than `--min-keyint` as scenecuts beyond
+        `--min-keyint` are already being coded as keyframes.
 
 .. option:: --lookahead-slices <0..16>
 
@@ -2040,7 +2064,7 @@
 	Example for MaxCLL=1000 candela per square meter, MaxFALL=400
 	candela per square meter:
 
-		--max-cll 1000,400
+		--max-cll "1000,400"
 
 	Note that this string value will need to be escaped or quoted to
 	protect against shell expansion on many platforms. No default.

x265_2.6.tar.gz/doc/reST/releasenotes.rst -> x265_2.7.tar.gz/doc/reST/releasenotes.rst Changed

@@ -2,6 +2,32 @@
 Release Notes
 *************
 
+Version 2.7
+===========
+
+Release date - 21st Feb, 2018.
+
+New features
+------------
+1. :option:`--gop-lookahead` can be used to extend the gop boundary(set by `--keyint`). The GOP will be extended, if a scene-cut frame is found within this many number of frames. 
+2. Support for RADL pictures added in x265.
+   :option:`--radl` can be used to decide number of RADL pictures preceding the IDR picture.
+
+Encoder enhancements
+--------------------
+1. Moved from YASM to NASM assembler. Supports NASM assembler version 2.13 and greater.
+2. Enable analysis save and load in a single run. Introduces two new cli options `--analysis-save <filename>` and `--analysis-load <filename>`.
+3. Comply to HDR10+ LLC specification.
+4. Reduced x265 build time by more than 50% by re-factoring ipfilter.asm.  
+
+Bug fixes
+---------
+1. Fixed inconsistent output issue in deblock filter and --const-vbv.
+2. Fixed Mac OS build warnings.
+3. Fixed inconsistency in pass-2 when weightp and cutree are enabled.
+4. Fixed deadlock issue due to dropping of BREF frames, while forcing slice types through qp file.
+
+
 Version 2.6
 ===========

x265_2.6.tar.gz/source/CMakeLists.txt -> x265_2.7.tar.gz/source/CMakeLists.txt Changed

@@ -29,7 +29,7 @@
 option(STATIC_LINK_CRT "Statically link C runtime for release builds" OFF)
 mark_as_advanced(FPROFILE_USE FPROFILE_GENERATE NATIVE_BUILD)
 # X265_BUILD must be incremented each time the public API is changed
-set(X265_BUILD 146)
+set(X265_BUILD 151)
 configure_file("${PROJECT_SOURCE_DIR}/x265.def.in"
                "${PROJECT_BINARY_DIR}/x265.def")
 configure_file("${PROJECT_SOURCE_DIR}/x265_config.h.in"
@@ -323,15 +323,15 @@
     execute_process(COMMAND ${CMAKE_CXX_COMPILER} -dumpversion OUTPUT_VARIABLE CC_VERSION)
 endif(GCC)
 
-find_package(Yasm)
+find_package(Nasm)
 if(ARM OR CROSS_COMPILE_ARM)
     option(ENABLE_ASSEMBLY "Enable use of assembly coded primitives" ON)
-elseif(YASM_FOUND AND X86)
-    if (YASM_VERSION_STRING VERSION_LESS "1.2.0")
-        message(STATUS "Yasm version ${YASM_VERSION_STRING} is too old. 1.2.0 or later required")
+elseif(NASM_FOUND AND X86)
+    if (NASM_VERSION_STRING VERSION_LESS "2.13.0")
+        message(STATUS "Nasm version ${NASM_VERSION_STRING} is too old. 2.13.0 or later required")
         option(ENABLE_ASSEMBLY "Enable use of assembly coded primitives" OFF)
     else()
-        message(STATUS "Found Yasm ${YASM_VERSION_STRING} to build assembly primitives")
+        message(STATUS "Found Nasm ${NASM_VERSION_STRING} to build assembly primitives")
         option(ENABLE_ASSEMBLY "Enable use of assembly coded primitives" ON)
     endif()
 else()
@@ -517,18 +517,18 @@
             list(APPEND ASM_OBJS ${ASM}.${SUFFIX})
             add_custom_command(
                 OUTPUT ${ASM}.${SUFFIX}
-                COMMAND ${YASM_EXECUTABLE} ARGS ${YASM_FLAGS} ${ASM_SRC} -o ${ASM}.${SUFFIX}
+                COMMAND ${NASM_EXECUTABLE} ARGS ${NASM_FLAGS} ${ASM_SRC} -o ${ASM}.${SUFFIX}
                 DEPENDS ${ASM_SRC})
         endforeach()
     endif()
 endif()
 source_group(ASM FILES ${ASM_SRCS})
 if(ENABLE_HDR10_PLUS)
-    add_library(x265-static STATIC $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> $<TARGET_OBJECTS:dynamicHDR10> ${ASM_OBJS} ${ASM_SRCS})
+    add_library(x265-static STATIC $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> $<TARGET_OBJECTS:dynamicHDR10> ${ASM_OBJS})
     add_library(hdr10plus-static STATIC $<TARGET_OBJECTS:dynamicHDR10>)
     set_target_properties(hdr10plus-static PROPERTIES OUTPUT_NAME hdr10plus)
 else()
-    add_library(x265-static STATIC $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> ${ASM_OBJS} ${ASM_SRCS})
+    add_library(x265-static STATIC $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> ${ASM_OBJS})
 endif()
 if(NOT MSVC)
     set_target_properties(x265-static PROPERTIES OUTPUT_NAME x265)
@@ -546,14 +546,19 @@
         ARCHIVE DESTINATION ${LIB_INSTALL_DIR})
 endif()
 install(FILES x265.h "${PROJECT_BINARY_DIR}/x265_config.h" DESTINATION include)
-
 if(WIN32)
-    install(FILES "${PROJECT_BINARY_DIR}/Debug/x265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS Debug)
-    install(FILES "${PROJECT_BINARY_DIR}/RelWithDebInfo/x265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS RelWithDebInfo)
-    install(FILES "${PROJECT_BINARY_DIR}/Debug/libx265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS Debug OPTIONAL NAMELINK_ONLY)
-    install(FILES "${PROJECT_BINARY_DIR}/RelWithDebInfo/libx265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS RelWithDebInfo OPTIONAL NAMELINK_ONLY)
+    if(MSVC_IDE)
+        install(FILES "${PROJECT_BINARY_DIR}/Debug/x265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS Debug)
+        install(FILES "${PROJECT_BINARY_DIR}/RelWithDebInfo/x265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS RelWithDebInfo)
+        install(FILES "${PROJECT_BINARY_DIR}/Debug/libx265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS Debug OPTIONAL NAMELINK_ONLY)
+        install(FILES "${PROJECT_BINARY_DIR}/RelWithDebInfo/libx265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS RelWithDebInfo OPTIONAL NAMELINK_ONLY)
+    else()
+        install(FILES "${PROJECT_BINARY_DIR}/x265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS Debug)
+        install(FILES "${PROJECT_BINARY_DIR}/x265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS RelWithDebInfo)
+        install(FILES "${PROJECT_BINARY_DIR}/libx265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS Debug OPTIONAL NAMELINK_ONLY)
+        install(FILES "${PROJECT_BINARY_DIR}/libx265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS RelWithDebInfo OPTIONAL NAMELINK_ONLY)
+    endif()
 endif()
-
 if(CMAKE_RC_COMPILER)
     # The resource compiler does not need CFLAGS or macro defines. It
     # often breaks them
@@ -642,7 +647,9 @@
     endforeach()
     if(PLIBLIST)
         # blacklist of libraries that should not be in Libs.private
-        list(REMOVE_ITEM PLIBLIST "-lc" "-lpthread")
+        list(REMOVE_ITEM PLIBLIST "-lc" "-lpthread" "-lmingwex" "-lmingwthrd"
+            "-lmingw32" "-lmoldname" "-lmsvcrt" "-ladvapi32" "-lshell32"
+            "-luser32" "-lkernel32")
         string(REPLACE ";" " " PRIVATE_LIBS "${PLIBLIST}")
     else()
         set(PRIVATE_LIBS "")
@@ -686,11 +693,11 @@
         if(ENABLE_HDR10_PLUS)
         add_executable(cli ../COPYING ${InputFiles} ${OutputFiles} ${GETOPT}
                         x265.cpp x265.h x265cli.h
-                        $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> $<TARGET_OBJECTS:dynamicHDR10> ${ASM_OBJS} ${ASM_SRCS})
+                        $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> $<TARGET_OBJECTS:dynamicHDR10> ${ASM_OBJS})
         else()
             add_executable(cli ../COPYING ${InputFiles} ${OutputFiles} ${GETOPT}
                         x265.cpp x265.h x265cli.h
-                        $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> ${ASM_OBJS} ${ASM_SRCS})
+                        $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> ${ASM_OBJS})
         endif()
     else()
         add_executable(cli ../COPYING ${InputFiles} ${OutputFiles} ${GETOPT} ${X265_RC_FILE}

x265_2.7.tar.gz/source/cmake/CMakeASM_NASMInformation.cmake Added

@@ -0,0 +1,68 @@
+set(ASM_DIALECT "_NASM")
+set(CMAKE_ASM${ASM_DIALECT}_SOURCE_FILE_EXTENSIONS asm)
+
+if(X64)
+    list(APPEND ASM_FLAGS -DARCH_X86_64=1 -I ${CMAKE_CURRENT_SOURCE_DIR}/../common/x86/)
+    if(ENABLE_PIC)
+        list(APPEND ASM_FLAGS -DPIC)
+    endif()
+    if(APPLE)
+        set(ARGS -f macho64 -DPREFIX)
+    elseif(UNIX AND NOT CYGWIN)
+        set(ARGS -f elf64)
+    else()
+        set(ARGS -f win64)
+    endif()
+else()
+    list(APPEND ASM_FLAGS -DARCH_X86_64=0 -I ${CMAKE_CURRENT_SOURCE_DIR}/../common/x86/)
+    if(APPLE)
+        set(ARGS -f macho32 -DPREFIX)
+    elseif(UNIX AND NOT CYGWIN)
+        set(ARGS -f elf32)
+    else()
+        set(ARGS -f win32 -DPREFIX)
+    endif()
+endif()
+
+if(GCC)
+    list(APPEND ASM_FLAGS -DHAVE_ALIGNED_STACK=1)
+else()
+    list(APPEND ASM_FLAGS -DHAVE_ALIGNED_STACK=0)
+endif()
+
+if(HIGH_BIT_DEPTH)
+    if(MAIN12)
+        list(APPEND ASM_FLAGS -DHIGH_BIT_DEPTH=1 -DBIT_DEPTH=12 -DX265_NS=${X265_NS})
+    else()
+        list(APPEND ASM_FLAGS -DHIGH_BIT_DEPTH=1 -DBIT_DEPTH=10 -DX265_NS=${X265_NS})
+    endif()
+else()
+    list(APPEND ASM_FLAGS -DHIGH_BIT_DEPTH=0 -DBIT_DEPTH=8 -DX265_NS=${X265_NS})
+endif()
+
+list(APPEND ASM_FLAGS "${CMAKE_ASM_NASM_FLAGS}")
+
+if(CMAKE_BUILD_TYPE MATCHES Release)
+    list(APPEND ASM_FLAGS "${CMAKE_ASM_NASM_FLAGS_RELEASE}")
+elseif(CMAKE_BUILD_TYPE MATCHES Debug)
+    list(APPEND ASM_FLAGS "${CMAKE_ASM_NASM_FLAGS_DEBUG}")
+elseif(CMAKE_BUILD_TYPE MATCHES MinSizeRel)
+    list(APPEND ASM_FLAGS "${CMAKE_ASM_NASM_FLAGS_MINSIZEREL}")
+elseif(CMAKE_BUILD_TYPE MATCHES RelWithDebInfo)
+    list(APPEND ASM_FLAGS "${CMAKE_ASM_NASM_FLAGS_RELWITHDEBINFO}")
+endif()
+
+set(NASM_FLAGS ${ARGS} ${ASM_FLAGS} PARENT_SCOPE)
+string(REPLACE ";" " " CMAKE_ASM_NASM_COMPILER_ARG1 "${ARGS}")
+
+# This section exists to override the one in CMakeASMInformation.cmake
+# (the default Information file). This removes the <FLAGS>
+# thing so that your C compiler flags that have been set via
+# set_target_properties don't get passed to nasm and confuse it.
+if(NOT CMAKE_ASM${ASM_DIALECT}_COMPILE_OBJECT)
+    string(REPLACE ";" " " STR_ASM_FLAGS "${ASM_FLAGS}")
+    set(CMAKE_ASM${ASM_DIALECT}_COMPILE_OBJECT "<CMAKE_ASM${ASM_DIALECT}_COMPILER> ${STR_ASM_FLAGS} -o <OBJECT> <SOURCE>")
+endif()
+
+include(CMakeASMInformation)
+set(ASM_DIALECT)

x265_2.7.tar.gz/source/cmake/CMakeDetermineASM_NASMCompiler.cmake Added

x265_2.7.tar.gz/source/cmake/CMakeTestASM_NASMCompiler.cmake Added

x265_2.7.tar.gz/source/cmake/FindNasm.cmake Added

x265_2.6.tar.gz/source/cmake/version.cmake -> x265_2.7.tar.gz/source/cmake/version.cmake Changed

x265_2.6.tar.gz/source/common/CMakeLists.txt -> x265_2.7.tar.gz/source/common/CMakeLists.txt Changed

@@ -56,28 +56,26 @@
     endif()
     set(VEC_PRIMITIVES vec/vec-primitives.cpp ${PRIMITIVES})
     source_group(Intrinsics FILES ${VEC_PRIMITIVES})
-
     set(C_SRCS asm-primitives.cpp pixel.h mc.h ipfilter8.h blockcopy8.h dct8.h loopfilter.h seaintegral.h)
     set(A_SRCS pixel-a.asm const-a.asm cpu-a.asm ssd-a.asm mc-a.asm
                mc-a2.asm pixel-util8.asm blockcopy8.asm
                pixeladd8.asm dct8.asm seaintegral.asm)
     if(HIGH_BIT_DEPTH)
-        set(A_SRCS ${A_SRCS} sad16-a.asm intrapred16.asm ipfilter16.asm loopfilter.asm)
+        set(A_SRCS ${A_SRCS} sad16-a.asm intrapred16.asm v4-ipfilter16.asm h4-ipfilter16.asm h-ipfilter16.asm ipfilter16.asm loopfilter.asm)
     else()
-        set(A_SRCS ${A_SRCS} sad-a.asm intrapred8.asm intrapred8_allangs.asm ipfilter8.asm loopfilter.asm)
+        set(A_SRCS ${A_SRCS} sad-a.asm intrapred8.asm intrapred8_allangs.asm v4-ipfilter8.asm h-ipfilter8.asm ipfilter8.asm loopfilter.asm)
     endif()
-
     if(NOT X64)
         set(A_SRCS ${A_SRCS} pixel-32.asm)
     endif()
 
     if(MSVC_IDE OR XCODE)
-        # MSVC requires custom build rules in the main cmake script for yasm
-        set(MSVC_ASMS "${A_SRCS}" CACHE INTERNAL "yasm sources")
+        # MSVC requires custom build rules in the main cmake script for nasm
+        set(MSVC_ASMS "${A_SRCS}" CACHE INTERNAL "nasm sources")
         set(A_SRCS)
     endif()
 
-    enable_language(ASM_YASM)
+    enable_language(ASM_NASM)
 
     foreach(SRC ${A_SRCS} ${C_SRCS})
         set(ASM_PRIMITIVES ${ASM_PRIMITIVES} x86/${SRC})

x265_2.6.tar.gz/source/common/common.h -> x265_2.7.tar.gz/source/common/common.h Changed

x265_2.6.tar.gz/source/common/cudata.cpp -> x265_2.7.tar.gz/source/common/cudata.cpp Changed

@@ -1626,7 +1626,7 @@
                 dir |= (1 << list);
                 candMvField[count][list].mv = colmv;
                 candMvField[count][list].refIdx = refIdx;
-                if (m_encData->m_param->scaleFactor && m_encData->m_param->analysisReuseMode == X265_ANALYSIS_SAVE && m_log2CUSize[0] < 4)
+                if (m_encData->m_param->scaleFactor && m_encData->m_param->analysisSave && m_log2CUSize[0] < 4)
                 {
                     MV dist(MAX_MV, MAX_MV);
                     candMvField[count][list].mv = dist;
@@ -1791,7 +1791,7 @@
             int curRefPOC = m_slice->m_refPOCList[picList][refIdx];
             int curPOC = m_slice->m_poc;
 
-            if (m_encData->m_param->scaleFactor && m_encData->m_param->analysisReuseMode == X265_ANALYSIS_SAVE && (m_log2CUSize[0] < 4))
+            if (m_encData->m_param->scaleFactor && m_encData->m_param->analysisSave && (m_log2CUSize[0] < 4))
             {
                 MV dist(MAX_MV, MAX_MV);
                 pmv[numMvc++] = amvpCand[num++] = dist;

x265_2.6.tar.gz/source/common/deblock.cpp -> x265_2.7.tar.gz/source/common/deblock.cpp Changed

@@ -207,21 +207,18 @@
     static const MV zeroMv(0, 0);
     const Slice* const sliceQ = cuQ->m_slice;
     const Slice* const sliceP = cuP->m_slice;
-
-    const Frame* refP0 = sliceP->m_refFrameList[0][cuP->m_refIdx[0][partP]];
-    const Frame* refQ0 = sliceQ->m_refFrameList[0][cuQ->m_refIdx[0][partQ]];
+    const Frame* refP0 = (cuP->m_refIdx[0][partP] >= 0) ? sliceP->m_refFrameList[0][cuP->m_refIdx[0][partP]] : NULL;
+    const Frame* refQ0 = (cuQ->m_refIdx[0][partQ] >= 0) ? sliceQ->m_refFrameList[0][cuQ->m_refIdx[0][partQ]] : NULL;
     const MV& mvP0 = refP0 ? cuP->m_mv[0][partP] : zeroMv;
     const MV& mvQ0 = refQ0 ? cuQ->m_mv[0][partQ] : zeroMv;
-
     if (sliceQ->isInterP() && sliceP->isInterP())
     {
         return ((refP0 != refQ0) ||
                 (abs(mvQ0.x - mvP0.x) >= 4) || (abs(mvQ0.y - mvP0.y) >= 4)) ? 1 : 0;
     }
-
     // (sliceQ->isInterB() || sliceP->isInterB())
-    const Frame* refP1 = sliceP->m_refFrameList[1][cuP->m_refIdx[1][partP]];
-    const Frame* refQ1 = sliceQ->m_refFrameList[1][cuQ->m_refIdx[1][partQ]];
+    const Frame* refP1 = (cuP->m_refIdx[1][partP] >= 0) ? sliceP->m_refFrameList[1][cuP->m_refIdx[1][partP]] : NULL;
+    const Frame* refQ1 = (cuQ->m_refIdx[1][partQ] >= 0) ? sliceQ->m_refFrameList[1][cuQ->m_refIdx[1][partQ]] : NULL;
     const MV& mvP1 = refP1 ? cuP->m_mv[1][partP] : zeroMv;
     const MV& mvQ1 = refQ1 ? cuQ->m_mv[1][partQ] : zeroMv;

x265_2.6.tar.gz/source/common/frame.h -> x265_2.7.tar.gz/source/common/frame.h Changed

x265_2.6.tar.gz/source/common/framedata.cpp -> x265_2.7.tar.gz/source/common/framedata.cpp Changed

@@ -40,11 +40,12 @@
     m_spsrpsIdx = -1;
     if (param.rc.bStatWrite)
         m_spsrps = const_cast<RPS*>(sps.spsrps);
-
-    m_cuMemPool.create(0, param.internalCsp, sps.numCUsInFrame, param);
-    for (uint32_t ctuAddr = 0; ctuAddr < sps.numCUsInFrame; ctuAddr++)
-        m_picCTU[ctuAddr].initialize(m_cuMemPool, 0, param, ctuAddr);
-
+    bool isallocated = m_cuMemPool.create(0, param.internalCsp, sps.numCUsInFrame, param);
+    if (isallocated)
+        for (uint32_t ctuAddr = 0; ctuAddr < sps.numCUsInFrame; ctuAddr++)
+            m_picCTU[ctuAddr].initialize(m_cuMemPool, 0, param, ctuAddr);
+    else
+        return false;
     CHECKED_MALLOC_ZERO(m_cuStat, RCStatCU, sps.numCUsInFrame);
     CHECKED_MALLOC(m_rowStat, RCStatRow, sps.numCuInHeight);
     reinit(sps);
@@ -76,16 +77,12 @@
 
     X265_FREE(m_cuStat);
     X265_FREE(m_rowStat);
-
-    if (m_meBuffer)
+    for (int i = 0; i < INTEGRAL_PLANE_NUM; i++)
     {
-        for (int i = 0; i < INTEGRAL_PLANE_NUM; i++)
+        if (m_meBuffer[i] != NULL)
         {
-            if (m_meBuffer[i] != NULL)
-            {
-                X265_FREE(m_meBuffer[i]);
-                m_meBuffer[i] = NULL;
-            }
+            X265_FREE(m_meBuffer[i]);
+            m_meBuffer[i] = NULL;
         }
     }
 }

x265_2.6.tar.gz/source/common/lowres.cpp -> x265_2.7.tar.gz/source/common/lowres.cpp Changed

x265_2.6.tar.gz/source/common/lowres.h -> x265_2.7.tar.gz/source/common/lowres.h Changed

x265_2.6.tar.gz/source/common/param.cpp -> x265_2.7.tar.gz/source/common/param.cpp Changed

@@ -144,6 +144,7 @@
     /* Coding Structure */
     param->keyframeMin = 0;
     param->keyframeMax = 250;
+    param->gopLookahead = 0;
     param->bOpenGOP = 1;
     param->bframes = 4;
     param->lookaheadDepth = 20;
@@ -153,6 +154,7 @@
     param->lookaheadSlices = 8;
     param->lookaheadThreads = 0;
     param->scenecutBias = 5.0;
+    param->radl = 0;
     /* Intra Coding Tools */
     param->bEnableConstrainedIntra = 0;
     param->bEnableStrongIntraSmoothing = 1;
@@ -196,10 +198,12 @@
     param->rdPenalty = 0;
     param->psyRd = 2.0;
     param->psyRdoq = 0.0;
-    param->analysisReuseMode = 0;
+    param->analysisReuseMode = 0; /*DEPRECATED*/
     param->analysisMultiPassRefine = 0;
     param->analysisMultiPassDistortion = 0;
     param->analysisReuseFileName = NULL;
+    param->analysisSave = NULL;
+    param->analysisLoad = NULL;
     param->bIntraInBFrames = 0;
     param->bLossless = 0;
     param->bCULossless = 0;
@@ -849,7 +853,7 @@
         p->rc.bStrictCbr = atobool(value);
         p->rc.pbFactor = 1.0;
     }
-    OPT("analysis-reuse-mode") p->analysisReuseMode = parseName(value, x265_analysis_names, bError);
+    OPT("analysis-reuse-mode") p->analysisReuseMode = parseName(value, x265_analysis_names, bError); /*DEPRECATED*/
     OPT("sar")
     {
         p->vui.aspectRatioIdc = parseName(value, x265_sar_names, bError);
@@ -1004,6 +1008,10 @@
                 bError = true;
             }
          }
+        OPT("gop-lookahead") p->gopLookahead = atoi(value);
+        OPT("analysis-save") p->analysisSave = strdup(value);
+        OPT("analysis-load") p->analysisLoad = strdup(value);
+        OPT("radl") p->radl = atoi(value);
         else
             return X265_PARAM_BAD_NAME;
     }
@@ -1310,10 +1318,14 @@
           "scenecutThreshold must be greater than 0");
     CHECK(param->scenecutBias < 0 || 100 < param->scenecutBias,
            "scenecut-bias must be between 0 and 100");
+    CHECK(param->radl < 0 || param->radl > param->bframes,
+          "radl must be between 0 and bframes");
     CHECK(param->rdPenalty < 0 || param->rdPenalty > 2,
           "Valid penalty for 32x32 intra TU in non-I slices. 0:disabled 1:RD-penalty 2:maximum");
     CHECK(param->keyframeMax < -1,
           "Invalid max IDR period in frames. value should be greater than -1");
+    CHECK(param->gopLookahead < -1,
+          "GOP lookahead must be greater than -1");
     CHECK(param->decodedPictureHashSEI < 0 || param->decodedPictureHashSEI > 3,
           "Invalid hash option. Decoded Picture Hash SEI 0: disabled, 1: MD5, 2: CRC, 3: Checksum");
     CHECK(param->rc.vbvBufferSize < 0,
@@ -1340,9 +1352,7 @@
           "Constant QP is incompatible with 2pass");
     CHECK(param->rc.bStrictCbr && (param->rc.bitrate <= 0 || param->rc.vbvBufferSize <=0),
           "Strict-cbr cannot be applied without specifying target bitrate or vbv bufsize");
-    CHECK(param->analysisReuseMode && (param->analysisReuseMode < X265_ANALYSIS_OFF || param->analysisReuseMode > X265_ANALYSIS_LOAD),
-        "Invalid analysis mode. Analysis mode 0: OFF 1: SAVE : 2 LOAD");
-    CHECK(param->analysisReuseMode && (param->analysisReuseLevel < 1 || param->analysisReuseLevel > 10),
+    CHECK((param->analysisSave || param->analysisLoad) && (param->analysisReuseLevel < 1 || param->analysisReuseLevel > 10),
         "Invalid analysis refine level. Value must be between 1 and 10 (inclusive)");
     CHECK(param->scaleFactor > 2, "Invalid scale-factor. Supports factor <= 2");
     CHECK(param->rc.qpMax < QP_MIN || param->rc.qpMax > QP_MAX_MAX,
@@ -1520,11 +1530,15 @@
 char *x265_param2string(x265_param* p, int padx, int pady)
 {
     char *buf, *s;
+    size_t bufSize = 4000 + p->rc.zoneCount * 64;
+    if (p->numaPools)
+        bufSize += strlen(p->numaPools);
+    if (p->masteringDisplayColorVolume)
+        bufSize += strlen(p->masteringDisplayColorVolume);
 
-    buf = s = X265_MALLOC(char, MAXPARAMSIZE);
+    buf = s = X265_MALLOC(char, bufSize);
     if (!buf)
         return NULL;
-
 #define BOOL(param, cliopt) \
     s += sprintf(s, " %s", (param) ? cliopt : "no-" cliopt);
 
@@ -1539,7 +1553,7 @@
     BOOL(p->bEnableSsim, "ssim");
     s += sprintf(s, " log-level=%d", p->logLevel);
     if (p->csvfn)
-        s += sprintf(s, " csvfn=%s csv-log-level=%d", p->csvfn, p->csvLogLevel);
+        s += sprintf(s, " csv csv-log-level=%d", p->csvLogLevel);
     s += sprintf(s, " bitdepth=%d", p->internalBitDepth);
     s += sprintf(s, " input-csp=%d", p->internalCsp);
     s += sprintf(s, " fps=%u/%u", p->fpsNum, p->fpsDenom);
@@ -1561,6 +1575,7 @@
     BOOL(p->bOpenGOP, "open-gop");
     s += sprintf(s, " min-keyint=%d", p->keyframeMin);
     s += sprintf(s, " keyint=%d", p->keyframeMax);
+    s += sprintf(s, " gop-lookahead=%d", p->gopLookahead);
     s += sprintf(s, " bframes=%d", p->bframes);
     s += sprintf(s, " b-adapt=%d", p->bFrameAdaptive);
     BOOL(p->bBPyramid, "b-pyramid");
@@ -1568,6 +1583,7 @@
     s += sprintf(s, " rc-lookahead=%d", p->lookaheadDepth);
     s += sprintf(s, " lookahead-slices=%d", p->lookaheadSlices);
     s += sprintf(s, " scenecut=%d", p->scenecutThreshold);
+    s += sprintf(s, " radl=%d", p->radl);
     BOOL(p->bIntraRefresh, "intra-refresh");
     s += sprintf(s, " ctu=%d", p->maxCUSize);
     s += sprintf(s, " min-cu-size=%d", p->minCUSize);
@@ -1613,7 +1629,6 @@
     s += sprintf(s, " psy-rd=%.2f", p->psyRd);
     s += sprintf(s, " psy-rdoq=%.2f", p->psyRdoq);
     BOOL(p->bEnableRdRefine, "rd-refine");
-    s += sprintf(s, " analysis-reuse-mode=%d", p->analysisReuseMode);
     BOOL(p->bLossless, "lossless");
     s += sprintf(s, " cbqpoffs=%d", p->cbQpOffset);
     s += sprintf(s, " crqpoffs=%d", p->crQpOffset);
@@ -1711,6 +1726,10 @@
     BOOL(p->bEmitHDRSEI, "hdr");
     BOOL(p->bHDROpt, "hdr-opt");
     BOOL(p->bDhdr10opt, "dhdr10-opt");
+    if (p->analysisSave)
+        s += sprintf(s, " analysis-save");
+    if (p->analysisLoad)
+        s += sprintf(s, " analysis-load");
     s += sprintf(s, " analysis-reuse-level=%d", p->analysisReuseLevel);
     s += sprintf(s, " scale-factor=%d", p->scaleFactor);
     s += sprintf(s, " refine-intra=%d", p->intraRefine);

x265_2.6.tar.gz/source/common/param.h -> x265_2.7.tar.gz/source/common/param.h Changed

x265_2.6.tar.gz/source/common/picyuv.cpp -> x265_2.7.tar.gz/source/common/picyuv.cpp Changed

@@ -358,18 +358,20 @@
     pixel *uPic = m_picOrg[1];
     pixel *vPic = m_picOrg[2];
 
-    for (int r = 0; r < height; r++)
+    if (param.csvLogLevel >= 2 || param.maxCLL || param.maxFALL)
     {
-        for (int c = 0; c < width; c++)
+        for (int r = 0; r < height; r++)
         {
-            m_maxLumaLevel = X265_MAX(yPic[c], m_maxLumaLevel);
-            m_minLumaLevel = X265_MIN(yPic[c], m_minLumaLevel);
-            lumaSum += yPic[c];
+            for (int c = 0; c < width; c++)
+            {
+                m_maxLumaLevel = X265_MAX(yPic[c], m_maxLumaLevel);
+                m_minLumaLevel = X265_MIN(yPic[c], m_minLumaLevel);
+                lumaSum += yPic[c];
+            }
+            yPic += m_stride;
         }
-        yPic += m_stride;
+        m_avgLumaLevel = (double)lumaSum / (m_picHeight * m_picWidth);
     }
-    m_avgLumaLevel = (double)lumaSum / (m_picHeight * m_picWidth);
-
     if (param.csvLogLevel >= 2)
     {
         if (param.internalCsp != X265_CSP_I400)

x265_2.6.tar.gz/source/common/x86/asm-primitives.cpp -> x265_2.7.tar.gz/source/common/x86/asm-primitives.cpp Changed

x265_2.6.tar.gz/source/common/x86/blockcopy8.asm -> x265_2.7.tar.gz/source/common/x86/blockcopy8.asm Changed

x265_2.7.tar.gz/source/common/x86/h-ipfilter16.asm Added

@@ -0,0 +1,2537 @@
+;*****************************************************************************
+;* Copyright (C) 2013-2017 MulticoreWare, Inc
+;*
+;* Authors: Nabajit Deka <nabajit@multicorewareinc.com>
+;*          Murugan Vairavel <murugan@multicorewareinc.com>
+;*          Min Chen <chenm003@163.com>
+;*
+;* This program is free software; you can redistribute it and/or modify
+;* it under the terms of the GNU General Public License as published by
+;* the Free Software Foundation; either version 2 of the License, or
+;* (at your option) any later version.
+;*
+;* This program is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+;* GNU General Public License for more details.
+;*
+;* You should have received a copy of the GNU General Public License
+;* along with this program; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
+;*
+;* This program is also available under a commercial proprietary license.
+;* For more information, contact us at license @ x265.com.
+;*****************************************************************************/
+%include "x86inc.asm"
+%include "x86util.asm"
+
+
+%define INTERP_OFFSET_PP        pd_32
+%define INTERP_SHIFT_PP         6
+
+%if BIT_DEPTH == 10
+    %define INTERP_SHIFT_PS         2
+    %define INTERP_OFFSET_PS        pd_n32768
+    %define INTERP_SHIFT_SP         10
+    %define INTERP_OFFSET_SP        h_pd_524800
+%elif BIT_DEPTH == 12
+    %define INTERP_SHIFT_PS         4
+    %define INTERP_OFFSET_PS        pd_n131072
+    %define INTERP_SHIFT_SP         8
+    %define INTERP_OFFSET_SP        pd_524416
+%else
+    %error Unsupport bit depth!
+%endif
+
+SECTION_RODATA 32
+
+h_pd_524800:        times 8 dd 524800
+                                    
+tab_LumaCoeff:    dw   0, 0,  0,  64,  0,   0,  0,  0
+                  dw  -1, 4, -10, 58,  17, -5,  1,  0
+                  dw  -1, 4, -11, 40,  40, -11, 4, -1
+                  dw   0, 1, -5,  17,  58, -10, 4, -1
+                  
+ALIGN 32
+h_tab_LumaCoeffV:   times 4 dw 0, 0
+                  times 4 dw 0, 64
+                  times 4 dw 0, 0
+                  times 4 dw 0, 0
+
+                  times 4 dw -1, 4
+                  times 4 dw -10, 58
+                  times 4 dw 17, -5
+                  times 4 dw 1, 0
+
+                  times 4 dw -1, 4
+                  times 4 dw -11, 40
+                  times 4 dw 40, -11
+                  times 4 dw 4, -1
+
+                  times 4 dw 0, 1
+                  times 4 dw -5, 17
+                  times 4 dw 58, -10
+                  times 4 dw 4, -1
+                  
+const interp8_hps_shuf,     dd 0, 4, 1, 5, 2, 6, 3, 7                 
+
+const interp8_hpp_shuf,     db 0, 1, 2, 3, 4, 5, 6, 7, 2, 3, 4, 5, 6, 7, 8, 9
+                            db 4, 5, 6, 7, 8, 9, 10, 11, 6, 7, 8, 9, 10, 11, 12, 13 
+
+const interp8_hpp_shuf_new, db 0, 1, 2, 3, 2, 3, 4, 5, 4, 5, 6, 7, 6, 7, 8, 9
+                            db 4, 5, 6, 7, 6, 7, 8, 9, 8, 9, 10, 11, 10, 11, 12, 13                         
+                            
+SECTION .text
+cextern pd_8
+cextern pd_32
+cextern pw_pixel_max
+cextern pd_524416
+cextern pd_n32768
+cextern pd_n131072
+cextern pw_2000
+cextern idct8_shuf2
+
+%macro FILTER_LUMA_HOR_4_sse2 1
+    movu        m4,     [r0 + %1]       ; m4 = src[0-7]
+    movu        m5,     [r0 + %1 + 2]   ; m5 = src[1-8]
+    pmaddwd     m4,     m0
+    pmaddwd     m5,     m0
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m4,     m4,     q3120
+    pshufd      m5,     m5,     q3120
+    punpcklqdq  m4,     m5
+
+    movu        m5,     [r0 + %1 + 4]   ; m5 = src[2-9]
+    movu        m3,     [r0 + %1 + 6]   ; m3 = src[3-10]
+    pmaddwd     m5,     m0
+    pmaddwd     m3,     m0
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m2,     m3,     q2301
+    paddd       m3,     m2
+    pshufd      m5,     m5,     q3120
+    pshufd      m3,     m3,     q3120
+    punpcklqdq  m5,     m3
+
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m4,     m4,     q3120
+    pshufd      m5,     m5,     q3120
+    punpcklqdq  m4,     m5
+    paddd       m4,     m1
+%endmacro
+
+%macro FILTER_LUMA_HOR_8_sse2 1
+    movu        m4,     [r0 + %1]       ; m4 = src[0-7]
+    movu        m5,     [r0 + %1 + 2]   ; m5 = src[1-8]
+    pmaddwd     m4,     m0
+    pmaddwd     m5,     m0
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m4,     m4,     q3120
+    pshufd      m5,     m5,     q3120
+    punpcklqdq  m4,     m5
+
+    movu        m5,     [r0 + %1 + 4]   ; m5 = src[2-9]
+    movu        m3,     [r0 + %1 + 6]   ; m3 = src[3-10]
+    pmaddwd     m5,     m0
+    pmaddwd     m3,     m0
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m2,     m3,     q2301
+    paddd       m3,     m2
+    pshufd      m5,     m5,     q3120
+    pshufd      m3,     m3,     q3120
+    punpcklqdq  m5,     m3
+
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m4,     m4,     q3120
+    pshufd      m5,     m5,     q3120
+    punpcklqdq  m4,     m5
+    paddd       m4,     m1
+
+    movu        m5,     [r0 + %1 + 8]   ; m5 = src[4-11]
+    movu        m6,     [r0 + %1 + 10]  ; m6 = src[5-12]
+    pmaddwd     m5,     m0
+    pmaddwd     m6,     m0
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m2,     m6,     q2301
+    paddd       m6,     m2
+    pshufd      m5,     m5,     q3120
+    pshufd      m6,     m6,     q3120
+    punpcklqdq  m5,     m6
+
+    movu        m6,     [r0 + %1 + 12]  ; m6 = src[6-13]
+    movu        m3,     [r0 + %1 + 14]  ; m3 = src[7-14]
+    pmaddwd     m6,     m0
+    pmaddwd     m3,     m0
+    pshufd      m2,     m6,     q2301
+    paddd       m6,     m2
+    pshufd      m2,     m3,     q2301
+    paddd       m3,     m2
+    pshufd      m6,     m6,     q3120
+    pshufd      m3,     m3,     q3120
+    punpcklqdq  m6,     m3
+
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m2,     m6,     q2301
+    paddd       m6,     m2
+    pshufd      m5,     m5,     q3120
+    pshufd      m6,     m6,     q3120
+    punpcklqdq  m5,     m6
+    paddd       m5,     m1
+%endmacro
+
+;------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_p%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_sse2 3
+INIT_XMM sse2
+cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
+    mov         r4d,    r4m
+    sub         r0,     6
+    shl         r4d,    4
+    add         r1d,    r1d
+    add         r3d,    r3d
+
+%ifdef PIC
+    lea         r6,     [tab_LumaCoeff]
+    mova        m0,     [r6 + r4]
+%else
+    mova        m0,     [tab_LumaCoeff + r4]
+%endif
+
+%ifidn %3, pp
+    mova        m1,     [pd_32]
+    pxor        m7,     m7
+%else
+    mova        m1,     [INTERP_OFFSET_PS]
+%endif
+
+    mov         r4d,    %2
+%ifidn %3, ps
+    cmp         r5m,    byte 0
+    je          .loopH
+    lea         r6,     [r1 + 2 * r1]
+    sub         r0,     r6
+    add         r4d,    7
+%endif
+
+.loopH:
+%assign x 0
+%rep %1/8
+    FILTER_LUMA_HOR_8_sse2 x
+
+%ifidn %3, pp
+    psrad       m4,     6
+    psrad       m5,     6
+    packssdw    m4,     m5
+    CLIPW       m4,     m7,     [pw_pixel_max]
+%else
+  %if BIT_DEPTH == 10
+    psrad       m4,     2
+    psrad       m5,     2
+  %elif BIT_DEPTH == 12
+    psrad       m4,     4
+    psrad       m5,     4
+  %endif
+    packssdw    m4,     m5
+%endif
+
+    movu        [r2 + x], m4
+%assign x x+16
+%endrep
+
+%rep (%1 % 8)/4
+    FILTER_LUMA_HOR_4_sse2 x
+
+%ifidn %3, pp
+    psrad       m4,     6
+    packssdw    m4,     m4
+    CLIPW       m4,     m7,     [pw_pixel_max]
+%else
+  %if BIT_DEPTH == 10
+    psrad       m4,     2
+  %elif BIT_DEPTH == 12
+    psrad       m4,     4
+  %endif
+    packssdw    m4,     m4
+%endif
+
+    movh        [r2 + x], m4
+%endrep
+
+    add         r0,     r1
+    add         r2,     r3
+
+    dec         r4d
+    jnz         .loopH
+    RET
+
+%endmacro
+
+;------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;------------------------------------------------------------------------------------------------------------
+    FILTER_HOR_LUMA_sse2 4, 4, pp
+    FILTER_HOR_LUMA_sse2 4, 8, pp
+    FILTER_HOR_LUMA_sse2 4, 16, pp
+    FILTER_HOR_LUMA_sse2 8, 4, pp
+    FILTER_HOR_LUMA_sse2 8, 8, pp
+    FILTER_HOR_LUMA_sse2 8, 16, pp
+    FILTER_HOR_LUMA_sse2 8, 32, pp
+    FILTER_HOR_LUMA_sse2 12, 16, pp
+    FILTER_HOR_LUMA_sse2 16, 4, pp
+    FILTER_HOR_LUMA_sse2 16, 8, pp
+    FILTER_HOR_LUMA_sse2 16, 12, pp
+    FILTER_HOR_LUMA_sse2 16, 16, pp
+    FILTER_HOR_LUMA_sse2 16, 32, pp
+    FILTER_HOR_LUMA_sse2 16, 64, pp
+    FILTER_HOR_LUMA_sse2 24, 32, pp
+    FILTER_HOR_LUMA_sse2 32, 8, pp
+    FILTER_HOR_LUMA_sse2 32, 16, pp
+    FILTER_HOR_LUMA_sse2 32, 24, pp
+    FILTER_HOR_LUMA_sse2 32, 32, pp
+    FILTER_HOR_LUMA_sse2 32, 64, pp
+    FILTER_HOR_LUMA_sse2 48, 64, pp
+    FILTER_HOR_LUMA_sse2 64, 16, pp
+    FILTER_HOR_LUMA_sse2 64, 32, pp
+    FILTER_HOR_LUMA_sse2 64, 48, pp
+    FILTER_HOR_LUMA_sse2 64, 64, pp
+
+;---------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;---------------------------------------------------------------------------------------------------------------------------
+    FILTER_HOR_LUMA_sse2 4, 4, ps
+    FILTER_HOR_LUMA_sse2 4, 8, ps
+    FILTER_HOR_LUMA_sse2 4, 16, ps
+    FILTER_HOR_LUMA_sse2 8, 4, ps
+    FILTER_HOR_LUMA_sse2 8, 8, ps
+    FILTER_HOR_LUMA_sse2 8, 16, ps
+    FILTER_HOR_LUMA_sse2 8, 32, ps
+    FILTER_HOR_LUMA_sse2 12, 16, ps
+    FILTER_HOR_LUMA_sse2 16, 4, ps
+    FILTER_HOR_LUMA_sse2 16, 8, ps
+    FILTER_HOR_LUMA_sse2 16, 12, ps
+    FILTER_HOR_LUMA_sse2 16, 16, ps
+    FILTER_HOR_LUMA_sse2 16, 32, ps
+    FILTER_HOR_LUMA_sse2 16, 64, ps
+    FILTER_HOR_LUMA_sse2 24, 32, ps
+    FILTER_HOR_LUMA_sse2 32, 8, ps
+    FILTER_HOR_LUMA_sse2 32, 16, ps
+    FILTER_HOR_LUMA_sse2 32, 24, ps
+    FILTER_HOR_LUMA_sse2 32, 32, ps
+    FILTER_HOR_LUMA_sse2 32, 64, ps
+    FILTER_HOR_LUMA_sse2 48, 64, ps
+    FILTER_HOR_LUMA_sse2 64, 16, ps
+    FILTER_HOR_LUMA_sse2 64, 32, ps
+    FILTER_HOR_LUMA_sse2 64, 48, ps
+    FILTER_HOR_LUMA_sse2 64, 64, ps
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_HOR_CHROMA_sse3 3
+INIT_XMM sse3
+cglobal interp_4tap_horiz_%3_%1x%2, 4, 7, 8
+    add         r3,     r3
+    add         r1,     r1
+    sub         r0,     2
+    mov         r4d,    r4m
+    add         r4d,    r4d
+
+%ifdef PIC
+    lea         r6,     [tab_ChromaCoeff]
+    movddup     m0,     [r6 + r4 * 4]
+%else
+    movddup     m0,     [tab_ChromaCoeff + r4 * 4]
+%endif
+
+%ifidn %3, ps
+    mova        m1,     [INTERP_OFFSET_PS]
+    cmp         r5m,    byte 0
+%if %1 <= 6
+    lea         r4,     [r1 * 3]
+    lea         r5,     [r3 * 3]
+%endif
+    je          .skip
+    sub         r0,     r1
+%if %1 <= 6
+%assign y 1
+%else
+%assign y 3
+%endif
+%assign z 0
+%rep y
+%assign x 0
+%rep %1/8
+    FILTERH_W8_1_sse3 x, %3
+%assign x x+16
+%endrep
+%if %1 == 4 || (%1 == 6 && z == 0) || (%1 == 12 && z == 0)
+    FILTERH_W4_2_sse3 x, %3
+    FILTERH_W4_1_sse3 x
+%assign x x+8
+%endif
+%if %1 == 2 || (%1 == 6 && z == 0)
+    FILTERH_W2_3_sse3 x
+%endif
+%if %1 <= 6
+    lea         r0,     [r0 + r4]
+    lea         r2,     [r2 + r5]
+%else
+    lea         r0,     [r0 + r1]
+    lea         r2,     [r2 + r3]
+%endif
+%assign z z+1
+%endrep
+.skip:
+%elifidn %3, pp
+    pxor        m7,     m7
+    mova        m6,     [pw_pixel_max]
+    mova        m1,     [tab_c_32]
+%if %1 == 2 || %1 == 6
+    lea         r4,     [r1 * 3]
+    lea         r5,     [r3 * 3]
+%endif
+%endif
+
+%if %1 == 2
+%assign y %2/4
+%elif %1 <= 6
+%assign y %2/2
+%else
+%assign y %2
+%endif
+%assign z 0
+%rep y
+%assign x 0
+%rep %1/8
+    FILTERH_W8_1_sse3 x, %3
+%assign x x+16
+%endrep
+%if %1 == 4 || %1 == 6 || (%1 == 12 && (z % 2) == 0)
+    FILTERH_W4_2_sse3 x, %3
+%assign x x+8
+%endif
+%if %1 == 2 || (%1 == 6 && (z % 2) == 0)
+    FILTERH_W2_4_sse3 x, %3
+%endif
+%assign z z+1
+%if z < y
+%if %1 == 2
+    lea         r0,     [r0 + 4 * r1]
+    lea         r2,     [r2 + 4 * r3]
+%elif %1 <= 6
+    lea         r0,     [r0 + 2 * r1]
+    lea         r2,     [r2 + 2 * r3]
+%else
+    lea         r0,     [r0 + r1]
+    lea         r2,     [r2 + r3]
+%endif
+%endif ;z < y
+%endrep
+
+    RET
+%endmacro
+
+%macro FILTER_P2S_2_4_sse2 1
+    movd        m0,     [r0 + %1]
+    movd        m2,     [r0 + r1 * 2 + %1]
+    movhps      m0,     [r0 + r1 + %1]
+    movhps      m2,     [r0 + r4 + %1]
+    psllw       m0,     (14 - BIT_DEPTH)
+    psllw       m2,     (14 - BIT_DEPTH)
+    psubw       m0,     m1
+    psubw       m2,     m1
+
+    movd        [r2 + r3 * 0 + %1], m0
+    movd        [r2 + r3 * 2 + %1], m2
+    movhlps     m0,     m0
+    movhlps     m2,     m2
+    movd        [r2 + r3 * 1 + %1], m0
+    movd        [r2 + r5 + %1], m2
+%endmacro
+
+%macro FILTER_P2S_4_4_sse2 1
+    movh        m0,     [r0 + %1]
+    movhps      m0,     [r0 + r1 + %1]
+    psllw       m0,     (14 - BIT_DEPTH)
+    psubw       m0,     m1
+    movh        [r2 + r3 * 0 + %1], m0
+    movhps      [r2 + r3 * 1 + %1], m0
+
+    movh        m2,     [r0 + r1 * 2 + %1]
+    movhps      m2,     [r0 + r4 + %1]
+    psllw       m2,     (14 - BIT_DEPTH)
+    psubw       m2,     m1
+    movh        [r2 + r3 * 2 + %1], m2
+    movhps      [r2 + r5 + %1], m2
+%endmacro
+
+%macro FILTER_P2S_4_2_sse2 0
+    movh        m0,     [r0]
+    movhps      m0,     [r0 + r1 * 2]
+    psllw       m0,     (14 - BIT_DEPTH)
+    psubw       m0,     [pw_2000]
+    movh        [r2 + r3 * 0], m0
+    movhps      [r2 + r3 * 2], m0
+%endmacro
+
+%macro FILTER_P2S_8_4_sse2 1
+    movu        m0,     [r0 + %1]
+    movu        m2,     [r0 + r1 + %1]
+    psllw       m0,     (14 - BIT_DEPTH)
+    psllw       m2,     (14 - BIT_DEPTH)
+    psubw       m0,     m1
+    psubw       m2,     m1
+    movu        [r2 + r3 * 0 + %1], m0
+    movu        [r2 + r3 * 1 + %1], m2
+
+    movu        m3,     [r0 + r1 * 2 + %1]
+    movu        m4,     [r0 + r4 + %1]
+    psllw       m3,     (14 - BIT_DEPTH)
+    psllw       m4,     (14 - BIT_DEPTH)
+    psubw       m3,     m1
+    psubw       m4,     m1
+    movu        [r2 + r3 * 2 + %1], m3
+    movu        [r2 + r5 + %1], m4
+%endmacro
+
+%macro FILTER_P2S_8_2_sse2 1
+    movu        m0,     [r0 + %1]
+    movu        m2,     [r0 + r1 + %1]
+    psllw       m0,     (14 - BIT_DEPTH)
+    psllw       m2,     (14 - BIT_DEPTH)
+    psubw       m0,     m1
+    psubw       m2,     m1
+    movu        [r2 + r3 * 0 + %1], m0
+    movu        [r2 + r3 * 1 + %1], m2
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro FILTER_PIX_TO_SHORT_sse2 2
+INIT_XMM sse2
+cglobal filterPixelToShort_%1x%2, 4, 6, 3
+%if %2 == 2
+%if %1 == 4
+    FILTER_P2S_4_2_sse2
+%elif %1 == 8
+    add        r1d, r1d
+    add        r3d, r3d
+    mova       m1, [pw_2000]
+    FILTER_P2S_8_2_sse2 0
+%endif
+%else
+    add        r1d, r1d
+    add        r3d, r3d
+    mova       m1, [pw_2000]
+    lea        r4, [r1 * 3]
+    lea        r5, [r3 * 3]
+%assign y 1
+%rep %2/4
+%assign x 0
+%rep %1/8
+    FILTER_P2S_8_4_sse2 x
+%if %2 == 6
+    lea         r0,     [r0 + 4 * r1]
+    lea         r2,     [r2 + 4 * r3]
+    FILTER_P2S_8_2_sse2 x
+%endif
+%assign x x+16
+%endrep
+%rep (%1 % 8)/4
+    FILTER_P2S_4_4_sse2 x
+%assign x x+8
+%endrep
+%rep (%1 % 4)/2
+    FILTER_P2S_2_4_sse2 x
+%endrep
+%if y < %2/4
+    lea         r0,     [r0 + 4 * r1]
+    lea         r2,     [r2 + 4 * r3]
+%assign y y+1
+%endif
+%endrep
+%endif
+RET
+%endmacro
+
+    FILTER_PIX_TO_SHORT_sse2 2, 4
+    FILTER_PIX_TO_SHORT_sse2 2, 8
+    FILTER_PIX_TO_SHORT_sse2 2, 16
+    FILTER_PIX_TO_SHORT_sse2 4, 2
+    FILTER_PIX_TO_SHORT_sse2 4, 4
+    FILTER_PIX_TO_SHORT_sse2 4, 8
+    FILTER_PIX_TO_SHORT_sse2 4, 16
+    FILTER_PIX_TO_SHORT_sse2 4, 32
+    FILTER_PIX_TO_SHORT_sse2 6, 8
+    FILTER_PIX_TO_SHORT_sse2 6, 16
+    FILTER_PIX_TO_SHORT_sse2 8, 2
+    FILTER_PIX_TO_SHORT_sse2 8, 4
+    FILTER_PIX_TO_SHORT_sse2 8, 6
+    FILTER_PIX_TO_SHORT_sse2 8, 8
+    FILTER_PIX_TO_SHORT_sse2 8, 12
+    FILTER_PIX_TO_SHORT_sse2 8, 16
+    FILTER_PIX_TO_SHORT_sse2 8, 32
+    FILTER_PIX_TO_SHORT_sse2 8, 64
+    FILTER_PIX_TO_SHORT_sse2 12, 16
+    FILTER_PIX_TO_SHORT_sse2 12, 32
+    FILTER_PIX_TO_SHORT_sse2 16, 4
+    FILTER_PIX_TO_SHORT_sse2 16, 8
+    FILTER_PIX_TO_SHORT_sse2 16, 12
+    FILTER_PIX_TO_SHORT_sse2 16, 16
+    FILTER_PIX_TO_SHORT_sse2 16, 24
+    FILTER_PIX_TO_SHORT_sse2 16, 32
+    FILTER_PIX_TO_SHORT_sse2 16, 64
+    FILTER_PIX_TO_SHORT_sse2 24, 32
+    FILTER_PIX_TO_SHORT_sse2 24, 64
+    FILTER_PIX_TO_SHORT_sse2 32, 8
+    FILTER_PIX_TO_SHORT_sse2 32, 16
+    FILTER_PIX_TO_SHORT_sse2 32, 24
+    FILTER_PIX_TO_SHORT_sse2 32, 32
+    FILTER_PIX_TO_SHORT_sse2 32, 48
+    FILTER_PIX_TO_SHORT_sse2 32, 64
+    FILTER_PIX_TO_SHORT_sse2 48, 64
+    FILTER_PIX_TO_SHORT_sse2 64, 16
+    FILTER_PIX_TO_SHORT_sse2 64, 32
+    FILTER_PIX_TO_SHORT_sse2 64, 48
+    FILTER_PIX_TO_SHORT_sse2 64, 64
+
+;------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W4 3
+INIT_XMM sse4
+cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
+    mov         r4d, r4m
+    sub         r0, 6
+    shl         r4d, 4
+    add         r1, r1
+    add         r3, r3
+
+%ifdef PIC
+    lea         r6, [tab_LumaCoeff]
+    mova        m0, [r6 + r4]
+%else
+    mova        m0, [tab_LumaCoeff + r4]
+%endif
+
+%ifidn %3, pp
+    mova        m1, [pd_32]
+    pxor        m6, m6
+    mova        m7, [pw_pixel_max]
+%else
+    mova        m1, [INTERP_OFFSET_PS]
+%endif
+
+    mov         r4d, %2
+%ifidn %3, ps
+    cmp         r5m, byte 0
+    je          .loopH
+    lea         r6, [r1 + 2 * r1]
+    sub         r0, r6
+    add         r4d, 7
+%endif
+
+.loopH:
+    movu        m2, [r0]                     ; m2 = src[0-7]
+    movu        m3, [r0 + 16]                ; m3 = src[8-15]
+
+    pmaddwd     m4, m2, m0
+    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
+    pmaddwd     m5, m0
+    palignr     m3, m2, 6                    ; m3 = src[3-10]
+    pmaddwd     m3, m0
+    phaddd      m5, m3
+
+    phaddd      m4, m5
+    paddd       m4, m1
+%ifidn %3, pp
+    psrad       m4, 6
+    packusdw    m4, m4
+    CLIPW       m4, m6, m7
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    packssdw    m4, m4
+%endif
+
+    movh        [r2], m4
+
+    add         r0, r1
+    add         r2, r3
+
+    dec         r4d
+    jnz         .loopH
+    RET
+%endmacro
+
+;------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_4x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W4 4, 4, pp
+FILTER_HOR_LUMA_W4 4, 8, pp
+FILTER_HOR_LUMA_W4 4, 16, pp
+
+;---------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_4x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;---------------------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W4 4, 4, ps
+FILTER_HOR_LUMA_W4 4, 8, ps
+FILTER_HOR_LUMA_W4 4, 16, ps
+
+;------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W8 3
+INIT_XMM sse4
+cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
+
+    add         r1, r1
+    add         r3, r3
+    mov         r4d, r4m
+    sub         r0, 6
+    shl         r4d, 4
+
+%ifdef PIC
+    lea         r6, [tab_LumaCoeff]
+    mova        m0, [r6 + r4]
+%else
+    mova        m0, [tab_LumaCoeff + r4]
+%endif
+
+%ifidn %3, pp
+    mova        m1, [pd_32]
+    pxor        m7, m7
+%else
+    mova        m1, [INTERP_OFFSET_PS]
+%endif
+
+    mov         r4d, %2
+%ifidn %3, ps
+    cmp         r5m, byte 0
+    je         .loopH
+    lea         r6, [r1 + 2 * r1]
+    sub         r0, r6
+    add         r4d, 7
+%endif
+
+.loopH:
+    movu        m2, [r0]                     ; m2 = src[0-7]
+    movu        m3, [r0 + 16]                ; m3 = src[8-15]
+
+    pmaddwd     m4, m2, m0
+    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+    phaddd      m4, m5
+    paddd       m4, m1
+
+    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+
+    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
+    pmaddwd     m6, m0
+    palignr     m3, m2, 14                   ; m3 = src[7-14]
+    pmaddwd     m3, m0
+    phaddd      m6, m3
+    phaddd      m5, m6
+    paddd       m5, m1
+%ifidn %3, pp
+    psrad       m4, 6
+    psrad       m5, 6
+    packusdw    m4, m5
+    CLIPW       m4, m7, [pw_pixel_max]
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    psrad       m5, INTERP_SHIFT_PS
+    packssdw    m4, m5
+%endif
+
+    movu        [r2], m4
+
+    add         r0, r1
+    add         r2, r3
+
+    dec         r4d
+    jnz        .loopH
+    RET
+%endmacro
+
+;------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W8 8, 4, pp
+FILTER_HOR_LUMA_W8 8, 8, pp
+FILTER_HOR_LUMA_W8 8, 16, pp
+FILTER_HOR_LUMA_W8 8, 32, pp
+
+;---------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;---------------------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W8 8, 4, ps
+FILTER_HOR_LUMA_W8 8, 8, ps
+FILTER_HOR_LUMA_W8 8, 16, ps
+FILTER_HOR_LUMA_W8 8, 32, ps
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W12 3
+INIT_XMM sse4
+cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
+
+    add         r1, r1
+    add         r3, r3
+    mov         r4d, r4m
+    sub         r0, 6
+    shl         r4d, 4
+
+%ifdef PIC
+    lea         r6, [tab_LumaCoeff]
+    mova        m0, [r6 + r4]
+%else
+    mova        m0, [tab_LumaCoeff + r4]
+%endif
+%ifidn %3, pp
+    mova        m1, [INTERP_OFFSET_PP]
+%else
+    mova        m1, [INTERP_OFFSET_PS]
+%endif
+
+    mov         r4d, %2
+%ifidn %3, ps
+    cmp         r5m, byte 0
+    je          .loopH
+    lea         r6, [r1 + 2 * r1]
+    sub         r0, r6
+    add         r4d, 7
+%endif
+
+.loopH:
+    movu        m2, [r0]                     ; m2 = src[0-7]
+    movu        m3, [r0 + 16]                ; m3 = src[8-15]
+
+    pmaddwd     m4, m2, m0
+    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+    phaddd      m4, m5
+    paddd       m4, m1
+
+    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+
+    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
+    pmaddwd     m6, m0
+    palignr     m7, m3, m2, 14               ; m2 = src[7-14]
+    pmaddwd     m7, m0
+    phaddd      m6, m7
+    phaddd      m5, m6
+    paddd       m5, m1
+%ifidn %3, pp
+    psrad       m4, INTERP_SHIFT_PP
+    psrad       m5, INTERP_SHIFT_PP
+    packusdw    m4, m5
+    pxor        m5, m5
+    CLIPW       m4, m5, [pw_pixel_max]
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    psrad       m5, INTERP_SHIFT_PS
+    packssdw    m4, m5
+%endif
+
+    movu        [r2], m4
+
+    movu        m2, [r0 + 32]                ; m2 = src[16-23]
+
+    pmaddwd     m4, m3, m0                   ; m3 = src[8-15]
+    palignr     m5, m2, m3, 2                ; m5 = src[9-16]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m2, m3, 4                ; m5 = src[10-17]
+    pmaddwd     m5, m0
+    palignr     m2, m3, 6                    ; m2 = src[11-18]
+    pmaddwd     m2, m0
+    phaddd      m5, m2
+    phaddd      m4, m5
+    paddd       m4, m1
+%ifidn %3, pp
+    psrad       m4, INTERP_SHIFT_PP
+    packusdw    m4, m4
+    pxor        m5, m5
+    CLIPW       m4, m5, [pw_pixel_max]
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    packssdw    m4, m4
+%endif
+
+    movh        [r2 + 16], m4
+
+    add         r0, r1
+    add         r2, r3
+
+    dec         r4d
+    jnz         .loopH
+    RET
+%endmacro
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_12x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W12 12, 16, pp
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_12x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W12 12, 16, ps
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W16 3
+INIT_XMM sse4
+cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
+
+    add         r1, r1
+    add         r3, r3
+    mov         r4d, r4m
+    sub         r0, 6
+    shl         r4d, 4
+
+%ifdef PIC
+    lea         r6, [tab_LumaCoeff]
+    mova        m0, [r6 + r4]
+%else
+    mova        m0, [tab_LumaCoeff + r4]
+%endif
+
+%ifidn %3, pp
+    mova        m1, [pd_32]
+%else
+    mova        m1, [INTERP_OFFSET_PS]
+%endif
+
+    mov         r4d, %2
+%ifidn %3, ps
+    cmp         r5m, byte 0
+    je          .loopH
+    lea         r6, [r1 + 2 * r1]
+    sub         r0, r6
+    add         r4d, 7
+%endif
+
+.loopH:
+%assign x 0
+%rep %1 / 16
+    movu        m2, [r0 + x]                 ; m2 = src[0-7]
+    movu        m3, [r0 + 16 + x]            ; m3 = src[8-15]
+
+    pmaddwd     m4, m2, m0
+    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+    phaddd      m4, m5
+    paddd       m4, m1
+
+    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+
+    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
+    pmaddwd     m6, m0
+    palignr     m7, m3, m2, 14               ; m2 = src[7-14]
+    pmaddwd     m7, m0
+    phaddd      m6, m7
+    phaddd      m5, m6
+    paddd       m5, m1
+%ifidn %3, pp
+    psrad       m4, INTERP_SHIFT_PP
+    psrad       m5, INTERP_SHIFT_PP
+    packusdw    m4, m5
+    pxor        m5, m5
+    CLIPW       m4, m5, [pw_pixel_max]
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    psrad       m5, INTERP_SHIFT_PS
+    packssdw    m4, m5
+%endif
+    movu        [r2 + x], m4
+
+    movu        m2, [r0 + 32 + x]            ; m2 = src[16-23]
+
+    pmaddwd     m4, m3, m0                   ; m3 = src[8-15]
+    palignr     m5, m2, m3, 2                ; m5 = src[9-16]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m2, m3, 4                ; m5 = src[10-17]
+    pmaddwd     m5, m0
+    palignr     m6, m2, m3, 6                ; m6 = src[11-18]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+    phaddd      m4, m5
+    paddd       m4, m1
+
+    palignr     m5, m2, m3, 8                ; m5 = src[12-19]
+    pmaddwd     m5, m0
+    palignr     m6, m2, m3, 10               ; m6 = src[13-20]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+
+    palignr     m6, m2, m3, 12               ; m6 = src[14-21]
+    pmaddwd     m6, m0
+    palignr     m2, m3, 14                   ; m3 = src[15-22]
+    pmaddwd     m2, m0
+    phaddd      m6, m2
+    phaddd      m5, m6
+    paddd       m5, m1
+%ifidn %3, pp
+    psrad       m4, INTERP_SHIFT_PP
+    psrad       m5, INTERP_SHIFT_PP
+    packusdw    m4, m5
+    pxor        m5, m5
+    CLIPW       m4, m5, [pw_pixel_max]
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    psrad       m5, INTERP_SHIFT_PS
+    packssdw    m4, m5
+%endif
+    movu        [r2 + 16 + x], m4
+
+%assign x x+32
+%endrep
+
+    add         r0, r1
+    add         r2, r3
+
+    dec         r4d
+    jnz         .loopH
+    RET
+%endmacro
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W16 16, 4, pp
+FILTER_HOR_LUMA_W16 16, 8, pp
+FILTER_HOR_LUMA_W16 16, 12, pp
+FILTER_HOR_LUMA_W16 16, 16, pp
+FILTER_HOR_LUMA_W16 16, 32, pp
+FILTER_HOR_LUMA_W16 16, 64, pp
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W16 16, 4, ps
+FILTER_HOR_LUMA_W16 16, 8, ps
+FILTER_HOR_LUMA_W16 16, 12, ps
+FILTER_HOR_LUMA_W16 16, 16, ps
+FILTER_HOR_LUMA_W16 16, 32, ps
+FILTER_HOR_LUMA_W16 16, 64, ps
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W16 32, 8, pp
+FILTER_HOR_LUMA_W16 32, 16, pp
+FILTER_HOR_LUMA_W16 32, 24, pp
+FILTER_HOR_LUMA_W16 32, 32, pp
+FILTER_HOR_LUMA_W16 32, 64, pp
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W16 32, 8, ps
+FILTER_HOR_LUMA_W16 32, 16, ps
+FILTER_HOR_LUMA_W16 32, 24, ps
+FILTER_HOR_LUMA_W16 32, 32, ps
+FILTER_HOR_LUMA_W16 32, 64, ps
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_48x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W16 48, 64, pp
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_48x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W16 48, 64, ps
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_64x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W16 64, 16, pp
+FILTER_HOR_LUMA_W16 64, 32, pp
+FILTER_HOR_LUMA_W16 64, 48, pp
+FILTER_HOR_LUMA_W16 64, 64, pp
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_64x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W16 64, 16, ps
+FILTER_HOR_LUMA_W16 64, 32, ps
+FILTER_HOR_LUMA_W16 64, 48, ps
+FILTER_HOR_LUMA_W16 64, 64, ps
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W24 3
+INIT_XMM sse4
+cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
+
+    add         r1, r1
+    add         r3, r3
+    mov         r4d, r4m
+    sub         r0, 6
+    shl         r4d, 4
+
+%ifdef PIC
+    lea         r6, [tab_LumaCoeff]
+    mova        m0, [r6 + r4]
+%else
+    mova        m0, [tab_LumaCoeff + r4]
+%endif
+%ifidn %3, pp
+    mova        m1, [pd_32]
+%else
+    mova        m1, [INTERP_OFFSET_PS]
+%endif
+
+    mov         r4d, %2
+%ifidn %3, ps
+    cmp         r5m, byte 0
+    je          .loopH
+    lea         r6, [r1 + 2 * r1]
+    sub         r0, r6
+    add         r4d, 7
+%endif
+
+.loopH:
+    movu        m2, [r0]                     ; m2 = src[0-7]
+    movu        m3, [r0 + 16]                ; m3 = src[8-15]
+
+    pmaddwd     m4, m2, m0
+    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+    phaddd      m4, m5
+    paddd       m4, m1
+
+    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+
+    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
+    pmaddwd     m6, m0
+    palignr     m7, m3, m2, 14               ; m7 = src[7-14]
+    pmaddwd     m7, m0
+    phaddd      m6, m7
+    phaddd      m5, m6
+    paddd       m5, m1
+%ifidn %3, pp
+    psrad       m4, INTERP_SHIFT_PP
+    psrad       m5, INTERP_SHIFT_PP
+    packusdw    m4, m5
+    pxor        m5, m5
+    CLIPW       m4, m5, [pw_pixel_max]
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    psrad       m5, INTERP_SHIFT_PS
+    packssdw    m4, m5
+%endif
+    movu        [r2], m4
+
+    movu        m2, [r0 + 32]                ; m2 = src[16-23]
+
+    pmaddwd     m4, m3, m0                   ; m3 = src[8-15]
+    palignr     m5, m2, m3, 2                ; m5 = src[1-8]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m2, m3, 4                ; m5 = src[2-9]
+    pmaddwd     m5, m0
+    palignr     m6, m2, m3, 6                ; m6 = src[3-10]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+    phaddd      m4, m5
+    paddd       m4, m1
+
+    palignr     m5, m2, m3, 8                ; m5 = src[4-11]
+    pmaddwd     m5, m0
+    palignr     m6, m2, m3, 10               ; m6 = src[5-12]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+
+    palignr     m6, m2, m3, 12               ; m6 = src[6-13]
+    pmaddwd     m6, m0
+    palignr     m7, m2, m3, 14               ; m7 = src[7-14]
+    pmaddwd     m7, m0
+    phaddd      m6, m7
+    phaddd      m5, m6
+    paddd       m5, m1
+%ifidn %3, pp
+    psrad       m4, INTERP_SHIFT_PP
+    psrad       m5, INTERP_SHIFT_PP
+    packusdw    m4, m5
+    pxor        m5, m5
+    CLIPW       m4, m5, [pw_pixel_max]
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    psrad       m5, INTERP_SHIFT_PS
+    packssdw    m4, m5
+%endif
+    movu        [r2 + 16], m4
+
+    movu        m3, [r0 + 48]                ; m3 = src[24-31]
+
+    pmaddwd     m4, m2, m0                   ; m2 = src[16-23]
+    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
+    pmaddwd     m5, m0
+    phaddd      m4, m5
+
+    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+    phaddd      m4, m5
+    paddd       m4, m1
+
+    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
+    pmaddwd     m5, m0
+    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
+    pmaddwd     m6, m0
+    phaddd      m5, m6
+
+    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
+    pmaddwd     m6, m0
+    palignr     m7, m3, m2, 14               ; m7 = src[7-14]
+    pmaddwd     m7, m0
+    phaddd      m6, m7
+    phaddd      m5, m6
+    paddd       m5, m1
+%ifidn %3, pp
+    psrad       m4, INTERP_SHIFT_PP
+    psrad       m5, INTERP_SHIFT_PP
+    packusdw    m4, m5
+    pxor        m5, m5
+    CLIPW       m4, m5, [pw_pixel_max]
+%else
+    psrad       m4, INTERP_SHIFT_PS
+    psrad       m5, INTERP_SHIFT_PS
+    packssdw    m4, m5
+%endif
+    movu        [r2 + 32], m4
+
+    add         r0, r1
+    add         r2, r3
+
+    dec         r4d
+    jnz         .loopH
+    RET
+%endmacro
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_24x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W24 24, 32, pp
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_24x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+FILTER_HOR_LUMA_W24 24, 32, ps
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W4_avx2 1
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_4x%1, 4,7,7
+    add              r1d, r1d
+    add              r3d, r3d
+    sub              r0, 6
+    mov              r4d, r4m
+    shl              r4d, 4
+%ifdef PIC
+    lea              r5, [tab_LumaCoeff]
+    vpbroadcastq     m0, [r5 + r4]
+    vpbroadcastq     m1, [r5 + r4 + 8]
+%else
+    vpbroadcastq     m0, [tab_LumaCoeff + r4]
+    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    lea              r6, [pw_pixel_max]
+    mova             m3, [interp8_hpp_shuf]
+    mova             m6, [pd_32]
+    pxor             m2, m2
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    mov              r4d, %1/2
+
+.loop:
+    vbroadcasti128   m4, [r0]
+    vbroadcasti128   m5, [r0 + 8]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    phaddd           m4, m4
+    vpermq           m4, m4, q3120
+    paddd            m4, m6
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [r6]
+    movq             [r2], xm4
+
+    vbroadcasti128   m4, [r0 + r1]
+    vbroadcasti128   m5, [r0 + r1 + 8]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    phaddd           m4, m4
+    vpermq           m4, m4, q3120
+    paddd            m4, m6
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [r6]
+    movq             [r2 + r3], xm4
+
+    lea              r2, [r2 + 2 * r3]
+    lea              r0, [r0 + 2 * r1]
+    dec              r4d
+    jnz              .loop
+    RET
+%endmacro
+FILTER_HOR_LUMA_W4_avx2 4
+FILTER_HOR_LUMA_W4_avx2 8
+FILTER_HOR_LUMA_W4_avx2 16
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W8 1
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_8x%1, 4,6,8
+    add              r1d, r1d
+    add              r3d, r3d
+    sub              r0, 6
+    mov              r4d, r4m
+    shl              r4d, 4
+%ifdef PIC
+    lea              r5, [tab_LumaCoeff]
+    vpbroadcastq     m0, [r5 + r4]
+    vpbroadcastq     m1, [r5 + r4 + 8]
+%else
+    vpbroadcastq     m0, [tab_LumaCoeff + r4]
+    vpbroadcastq     m1, [h_ab_LumaCoeff + r4 + 8]
+%endif
+    mova             m3, [interp8_hpp_shuf]
+    mova             m7, [pd_32]
+    pxor             m2, m2
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    mov              r4d, %1/2
+
+.loop:
+    vbroadcasti128   m4, [r0]
+    vbroadcasti128   m5, [r0 + 8]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 8]
+    vbroadcasti128   m6, [r0 + 16]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2], xm4
+
+    vbroadcasti128   m4, [r0 + r1]
+    vbroadcasti128   m5, [r0 + r1 + 8]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + r1 + 8]
+    vbroadcasti128   m6, [r0 + r1 + 16]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + r3], xm4
+
+    lea              r2, [r2 + 2 * r3]
+    lea              r0, [r0 + 2 * r1]
+    dec              r4d
+    jnz              .loop
+    RET
+%endmacro
+FILTER_HOR_LUMA_W8 4
+FILTER_HOR_LUMA_W8 8
+FILTER_HOR_LUMA_W8 16
+FILTER_HOR_LUMA_W8 32
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W16 1
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_16x%1, 4,6,8
+    add              r1d, r1d
+    add              r3d, r3d
+    sub              r0, 6
+    mov              r4d, r4m
+    shl              r4d, 4
+%ifdef PIC
+    lea              r5, [tab_LumaCoeff]
+    vpbroadcastq     m0, [r5 + r4]
+    vpbroadcastq     m1, [r5 + r4 + 8]
+%else
+    vpbroadcastq     m0, [tab_LumaCoeff + r4]
+    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova             m3, [interp8_hpp_shuf]
+    mova             m7, [pd_32]
+    pxor             m2, m2
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    mov              r4d, %1
+
+.loop:
+    vbroadcasti128   m4, [r0]
+    vbroadcasti128   m5, [r0 + 8]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 8]
+    vbroadcasti128   m6, [r0 + 16]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2], xm4
+
+    vbroadcasti128   m4, [r0 + 16]
+    vbroadcasti128   m5, [r0 + 24]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 24]
+    vbroadcasti128   m6, [r0 + 32]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + 16], xm4
+
+    add              r2, r3
+    add              r0, r1
+    dec              r4d
+    jnz              .loop
+    RET
+%endmacro
+FILTER_HOR_LUMA_W16 4
+FILTER_HOR_LUMA_W16 8
+FILTER_HOR_LUMA_W16 12
+FILTER_HOR_LUMA_W16 16
+FILTER_HOR_LUMA_W16 32
+FILTER_HOR_LUMA_W16 64
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro FILTER_HOR_LUMA_W32 2
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
+    add              r1d, r1d
+    add              r3d, r3d
+    sub              r0, 6
+    mov              r4d, r4m
+    shl              r4d, 4
+%ifdef PIC
+    lea              r5, [tab_LumaCoeff]
+    vpbroadcastq     m0, [r5 + r4]
+    vpbroadcastq     m1, [r5 + r4 + 8]
+%else
+    vpbroadcastq     m0, [tab_LumaCoeff + r4]
+    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova             m3, [interp8_hpp_shuf]
+    mova             m7, [pd_32]
+    pxor             m2, m2
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    mov              r4d, %2
+
+.loop:
+%assign x 0
+%rep %1/16
+    vbroadcasti128   m4, [r0 + x]
+    vbroadcasti128   m5, [r0 + 8 + x]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 8 + x]
+    vbroadcasti128   m6, [r0 + 16 + x]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + x], xm4
+
+    vbroadcasti128   m4, [r0 + 16 + x]
+    vbroadcasti128   m5, [r0 + 24 + x]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 24 + x]
+    vbroadcasti128   m6, [r0 + 32 + x]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + 16 + x], xm4
+
+%assign x x+32
+%endrep
+
+    add              r2, r3
+    add              r0, r1
+    dec              r4d
+    jnz              .loop
+    RET
+%endmacro
+FILTER_HOR_LUMA_W32 32, 8
+FILTER_HOR_LUMA_W32 32, 16
+FILTER_HOR_LUMA_W32 32, 24
+FILTER_HOR_LUMA_W32 32, 32
+FILTER_HOR_LUMA_W32 32, 64
+FILTER_HOR_LUMA_W32 64, 16
+FILTER_HOR_LUMA_W32 64, 32
+FILTER_HOR_LUMA_W32 64, 48
+FILTER_HOR_LUMA_W32 64, 64
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_12x16, 4,6,8
+    add              r1d, r1d
+    add              r3d, r3d
+    sub              r0, 6
+    mov              r4d, r4m
+    shl              r4d, 4
+%ifdef PIC
+    lea              r5, [tab_LumaCoeff]
+    vpbroadcastq     m0, [r5 + r4]
+    vpbroadcastq     m1, [r5 + r4 + 8]
+%else
+    vpbroadcastq     m0, [tab_LumaCoeff + r4]
+    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova             m3, [interp8_hpp_shuf]
+    mova             m7, [pd_32]
+    pxor             m2, m2
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    mov              r4d, 16
+
+.loop:
+    vbroadcasti128   m4, [r0]
+    vbroadcasti128   m5, [r0 + 8]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 8]
+    vbroadcasti128   m6, [r0 + 16]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2], xm4
+
+    vbroadcasti128   m4, [r0 + 16]
+    vbroadcasti128   m5, [r0 + 24]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 24]
+    vbroadcasti128   m6, [r0 + 32]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movq             [r2 + 16], xm4
+
+    add              r2, r3
+    add              r0, r1
+    dec              r4d
+    jnz              .loop
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_24x32, 4,6,8
+    add              r1d, r1d
+    add              r3d, r3d
+    sub              r0, 6
+    mov              r4d, r4m
+    shl              r4d, 4
+%ifdef PIC
+    lea              r5, [tab_LumaCoeff]
+    vpbroadcastq     m0, [r5 + r4]
+    vpbroadcastq     m1, [r5 + r4 + 8]
+%else
+    vpbroadcastq     m0, [tab_LumaCoeff + r4]
+    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova             m3, [interp8_hpp_shuf]
+    mova             m7, [pd_32]
+    pxor             m2, m2
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    mov              r4d, 32
+
+.loop:
+    vbroadcasti128   m4, [r0]
+    vbroadcasti128   m5, [r0 + 8]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 8]
+    vbroadcasti128   m6, [r0 + 16]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2], xm4
+
+    vbroadcasti128   m4, [r0 + 16]
+    vbroadcasti128   m5, [r0 + 24]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 24]
+    vbroadcasti128   m6, [r0 + 32]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + 16], xm4
+
+    vbroadcasti128   m4, [r0 + 32]
+    vbroadcasti128   m5, [r0 + 40]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 40]
+    vbroadcasti128   m6, [r0 + 48]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + 32], xm4
+
+    add              r2, r3
+    add              r0, r1
+    dec              r4d
+    jnz              .loop
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_48x64, 4,6,8
+    add              r1d, r1d
+    add              r3d, r3d
+    sub              r0, 6
+    mov              r4d, r4m
+    shl              r4d, 4
+%ifdef PIC
+    lea              r5, [tab_LumaCoeff]
+    vpbroadcastq     m0, [r5 + r4]
+    vpbroadcastq     m1, [r5 + r4 + 8]
+%else
+    vpbroadcastq     m0, [tab_LumaCoeff + r4]
+    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova             m3, [interp8_hpp_shuf]
+    mova             m7, [pd_32]
+    pxor             m2, m2
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    mov              r4d, 64
+
+.loop:
+%assign x 0
+%rep 2
+    vbroadcasti128   m4, [r0 + x]
+    vbroadcasti128   m5, [r0 + 8 + x]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 8 + x]
+    vbroadcasti128   m6, [r0 + 16 + x]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + x], xm4
+
+    vbroadcasti128   m4, [r0 + 16 + x]
+    vbroadcasti128   m5, [r0 + 24 + x]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 24 + x]
+    vbroadcasti128   m6, [r0 + 32 + x]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + 16 + x], xm4
+
+    vbroadcasti128   m4, [r0 + 32 + x]
+    vbroadcasti128   m5, [r0 + 40 + x]
+    pshufb           m4, m3
+    pshufb           m5, m3
+
+    pmaddwd          m4, m0
+    pmaddwd          m5, m1
+    paddd            m4, m5
+
+    vbroadcasti128   m5, [r0 + 40 + x]
+    vbroadcasti128   m6, [r0 + 48 + x]
+    pshufb           m5, m3
+    pshufb           m6, m3
+
+    pmaddwd          m5, m0
+    pmaddwd          m6, m1
+    paddd            m5, m6
+
+    phaddd           m4, m5
+    vpermq           m4, m4, q3120
+    paddd            m4, m7
+    psrad            m4, INTERP_SHIFT_PP
+
+    packusdw         m4, m4
+    vpermq           m4, m4, q2020
+    CLIPW            m4, m2, [pw_pixel_max]
+    movu             [r2 + 32 + x], xm4
+
+%assign x x+48
+%endrep
+
+    add              r2, r3
+    add              r0, r1
+    dec              r4d
+    jnz              .loop
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+;void interp_horiz_ps_c(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+
+%macro IPFILTER_LUMA_PS_4xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_ps_4x%1, 6,8,7
+    mov                         r5d,               r5m
+    mov                         r4d,               r4m
+    add                         r1d,               r1d
+    add                         r3d,               r3d
+
+%ifdef PIC
+    lea                         r6,                [tab_LumaCoeff]
+    lea                         r4,                [r4 * 8]
+    vbroadcasti128              m0,                [r6 + r4 * 2]
+%else
+    lea                         r4,                [r4 * 8]
+    vbroadcasti128              m0,                [tab_LumaCoeff + r4 * 2]
+%endif
+
+    vbroadcasti128              m2,                [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - pw_2000
+
+    sub                         r0,                6
+    test                        r5d,               r5d
+    mov                         r7d,               %1                                    ; loop count variable - height
+    jz                         .preloop
+    lea                         r6,                [r1 * 3]                              ; r6 = (N / 2 - 1) * srcStride
+    sub                         r0,                r6                                    ; r0(src) - 3 * srcStride
+    add                         r7d,               6                                     ;7 - 1(since last row not in loop)                            ; need extra 7 rows, just set a specially flag here, blkheight += N - 1  (7 - 3 = 4 ; since the last three rows not in loop)
+
+.preloop:
+    lea                         r6,                [r3 * 3]
+.loop:
+    ; Row 0
+    movu                        xm3,                [r0]                                 ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    movu                        xm4,                [r0 + 2]                             ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    vinserti128                 m3,                 m3,                xm4,       1
+    movu                        xm4,                [r0 + 4]
+    movu                        xm5,                [r0 + 6]
+    vinserti128                 m4,                 m4,                xm5,       1
+    pmaddwd                     m3,                m0
+    pmaddwd                     m4,                m0
+    phaddd                      m3,                m4                                    ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
+
+    ; Row 1
+    movu                        xm4,                [r0 + r1]                            ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    movu                        xm5,                [r0 + r1 + 2]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    vinserti128                 m4,                 m4,                xm5,       1
+    movu                        xm5,                [r0 + r1 + 4]
+    movu                        xm6,                [r0 + r1 + 6]
+    vinserti128                 m5,                 m5,                xm6,       1
+    pmaddwd                     m4,                m0
+    pmaddwd                     m5,                m0
+    phaddd                      m4,                m5                                     ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
+    phaddd                      m3,                m4                                     ; all rows and col completed.
+
+    mova                        m5,                [interp8_hps_shuf]
+    vpermd                      m3,                m5,                  m3
+    paddd                       m3,                m2
+    vextracti128                xm4,               m3,                  1
+    psrad                       xm3,               INTERP_SHIFT_PS
+    psrad                       xm4,               INTERP_SHIFT_PS
+    packssdw                    xm3,               xm3
+    packssdw                    xm4,               xm4
+
+    movq                        [r2],              xm3                                   ;row 0
+    movq                        [r2 + r3],         xm4                                   ;row 1
+    lea                         r0,                [r0 + r1 * 2]                         ; first loop src ->5th row(i.e 4)
+    lea                         r2,                [r2 + r3 * 2]                         ; first loop dst ->5th row(i.e 4)
+
+    sub                         r7d,               2
+    jg                          .loop
+    test                        r5d,               r5d
+    jz                          .end
+
+    ; Row 10
+    movu                        xm3,                [r0]
+    movu                        xm4,                [r0 + 2]
+    vinserti128                 m3,                 m3,                 xm4,      1
+    movu                        xm4,                [r0 + 4]
+    movu                        xm5,                [r0 + 6]
+    vinserti128                 m4,                 m4,                 xm5,      1
+    pmaddwd                     m3,                m0
+    pmaddwd                     m4,                m0
+    phaddd                      m3,                m4
+
+    ; Row11
+    phaddd                      m3,                m4                                    ; all rows and col completed.
+
+    mova                        m5,                [interp8_hps_shuf]
+    vpermd                      m3,                m5,                  m3
+    paddd                       m3,                m2
+    vextracti128                xm4,               m3,                  1
+    psrad                       xm3,               INTERP_SHIFT_PS
+    psrad                       xm4,               INTERP_SHIFT_PS
+    packssdw                    xm3,               xm3
+    packssdw                    xm4,               xm4
+
+    movq                        [r2],              xm3                                   ;row 0
+.end:
+    RET
+%endif
+%endmacro
+
+    IPFILTER_LUMA_PS_4xN_AVX2 4
+    IPFILTER_LUMA_PS_4xN_AVX2 8
+    IPFILTER_LUMA_PS_4xN_AVX2 16
+
+%macro IPFILTER_LUMA_PS_8xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_ps_8x%1, 4, 6, 8
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+    shl                 r4d, 4
+%ifdef PIC
+    lea                 r6, [tab_LumaCoeff]
+    vpbroadcastq        m0, [r6 + r4]
+    vpbroadcastq        m1, [r6 + r4 + 8]
+%else
+    vpbroadcastq        m0, [tab_LumaCoeff + r4]
+    vpbroadcastq        m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova                m3, [interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 6
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    lea                 r6, [r1*3]
+    sub                 r0, r6
+    add                 r4d, 7
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m7, m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m7, m1
+    paddd               m4, m7
+
+    vbroadcasti128      m6, [r0 + 16]
+    pshufb              m5, m3
+    pshufb              m6, m3
+    pmaddwd             m5, m0
+    pmaddwd             m6, m1
+    paddd               m5, m6
+
+    phaddd              m4, m5
+    vpermq              m4, m4, q3120
+    paddd               m4, m2
+    vextracti128        xm5,m4, 1
+    psrad               xm4, INTERP_SHIFT_PS
+    psrad               xm5, INTERP_SHIFT_PS
+    packssdw            xm4, xm5
+
+    movu                [r2], xm4
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+    IPFILTER_LUMA_PS_8xN_AVX2 4
+    IPFILTER_LUMA_PS_8xN_AVX2 8
+    IPFILTER_LUMA_PS_8xN_AVX2 16
+    IPFILTER_LUMA_PS_8xN_AVX2 32
+
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_ps_24x32, 4, 6, 8
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+    shl                 r4d, 4
+%ifdef PIC
+    lea                 r6, [tab_LumaCoeff]
+    vpbroadcastq        m0, [r6 + r4]
+    vpbroadcastq        m1, [r6 + r4 + 8]
+%else
+    vpbroadcastq        m0, [tab_LumaCoeff + r4]
+    vpbroadcastq        m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova                m3, [interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 6
+    test                r5d, r5d
+    mov                 r4d, 32
+    jz                  .loop0
+    lea                 r6, [r1*3]
+    sub                 r0, r6
+    add                 r4d, 7
+
+.loop0:
+%assign x 0
+%rep 24/8
+    vbroadcasti128      m4, [r0 + x]
+    vbroadcasti128      m5, [r0 + 8 + x]
+    pshufb              m4, m3
+    pshufb              m7, m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m7, m1
+    paddd               m4, m7
+
+    vbroadcasti128      m6, [r0 + 16 + x]
+    pshufb              m5, m3
+    pshufb              m6, m3
+    pmaddwd             m5, m0
+    pmaddwd             m6, m1
+    paddd               m5, m6
+
+    phaddd              m4, m5
+    vpermq              m4, m4, q3120
+    paddd               m4, m2
+    vextracti128        xm5,m4, 1
+    psrad               xm4, INTERP_SHIFT_PS
+    psrad               xm5, INTERP_SHIFT_PS
+    packssdw            xm4, xm5
+
+    movu                [r2 + x], xm4
+    %assign x x+16
+    %endrep
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%macro IPFILTER_LUMA_PS_32_64_AVX2 2
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_ps_%1x%2, 4, 6, 8
+
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+    shl                 r4d, 6
+%ifdef PIC
+    lea                 r6, [h_tab_LumaCoeffV]
+    movu                m0, [r6 + r4]
+    movu                m1, [r6 + r4 + mmsize]
+%else
+    movu                m0, [h_tab_LumaCoeffV + r4]
+    movu                m1, [h_tab_LumaCoeffV + r4 + mmsize]
+%endif
+    mova                m3, [interp8_hpp_shuf_new]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 6
+    test                r5d, r5d
+    mov                 r4d, %2
+    jz                 .loop0
+    lea                 r6, [r1*3]
+    sub                 r0, r6
+    add                 r4d, 7
+
+.loop0:
+%assign x 0
+%rep %1/16
+    vbroadcasti128      m4, [r0 + x]
+    vbroadcasti128      m5, [r0 + 4 * SIZEOF_PIXEL + x]
+    pshufb              m4, m3
+    pshufb              m5, m3
+
+    pmaddwd             m4, m0
+    pmaddwd             m7, m5, m1
+    paddd               m4, m7
+    vextracti128        xm7, m4, 1
+    paddd               xm4, xm7
+    paddd               xm4, xm2
+    psrad               xm4, INTERP_SHIFT_PS
+
+    vbroadcasti128      m6, [r0 + 16 + x]
+    pshufb              m6, m3
+
+    pmaddwd             m5, m0
+    pmaddwd             m7, m6, m1
+    paddd               m5, m7
+    vextracti128        xm7, m5, 1
+    paddd               xm5, xm7
+    paddd               xm5, xm2
+    psrad               xm5, INTERP_SHIFT_PS
+
+    packssdw            xm4, xm5
+    movu                [r2 + x], xm4
+
+    vbroadcasti128      m5, [r0 + 24 + x]
+    pshufb              m5, m3
+
+    pmaddwd             m6, m0
+    pmaddwd             m7, m5, m1
+    paddd               m6, m7
+    vextracti128        xm7, m6, 1
+    paddd               xm6, xm7
+    paddd               xm6, xm2
+    psrad               xm6, INTERP_SHIFT_PS
+
+    vbroadcasti128      m7, [r0 + 32 + x]
+    pshufb              m7, m3
+
+    pmaddwd             m5, m0
+    pmaddwd             m7, m1
+    paddd               m5, m7
+    vextracti128        xm7, m5, 1
+    paddd               xm5, xm7
+    paddd               xm5, xm2
+    psrad               xm5, INTERP_SHIFT_PS
+
+    packssdw            xm6, xm5
+    movu                [r2 + 16 + x], xm6
+
+%assign x x+32
+%endrep
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                .loop0
+    RET
+%endif
+%endmacro
+
+    IPFILTER_LUMA_PS_32_64_AVX2 32, 8
+    IPFILTER_LUMA_PS_32_64_AVX2 32, 16
+    IPFILTER_LUMA_PS_32_64_AVX2 32, 24
+    IPFILTER_LUMA_PS_32_64_AVX2 32, 32
+    IPFILTER_LUMA_PS_32_64_AVX2 32, 64
+
+    IPFILTER_LUMA_PS_32_64_AVX2 64, 16
+    IPFILTER_LUMA_PS_32_64_AVX2 64, 32
+    IPFILTER_LUMA_PS_32_64_AVX2 64, 48
+    IPFILTER_LUMA_PS_32_64_AVX2 64, 64
+
+    IPFILTER_LUMA_PS_32_64_AVX2 48, 64
+
+%macro IPFILTER_LUMA_PS_16xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_ps_16x%1, 4, 6, 8
+
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+    shl                 r4d, 4
+%ifdef PIC
+    lea                 r6, [tab_LumaCoeff]
+    vpbroadcastq        m0, [r6 + r4]
+    vpbroadcastq        m1, [r6 + r4 + 8]
+%else
+    vpbroadcastq        m0, [tab_LumaCoeff + r4]
+    vpbroadcastq        m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova                m3, [interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 6
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    lea                 r6, [r1*3]
+    sub                 r0, r6
+    add                 r4d, 7
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m7, m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m7, m1
+    paddd               m4, m7
+
+    vbroadcasti128      m6, [r0 + 16]
+    pshufb              m5, m3
+    pshufb              m7, m6, m3
+    pmaddwd             m5, m0
+    pmaddwd             m7, m1
+    paddd               m5, m7
+
+    phaddd              m4, m5
+    vpermq              m4, m4, q3120
+    paddd               m4, m2
+    vextracti128        xm5, m4, 1
+    psrad               xm4, INTERP_SHIFT_PS
+    psrad               xm5, INTERP_SHIFT_PS
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+
+    vbroadcasti128      m5, [r0 + 24]
+    pshufb              m6, m3
+    pshufb              m7, m5, m3
+    pmaddwd             m6, m0
+    pmaddwd             m7, m1
+    paddd               m6, m7
+
+    vbroadcasti128      m7, [r0 + 32]
+    pshufb              m5, m3
+    pshufb              m7, m3
+    pmaddwd             m5, m0
+    pmaddwd             m7, m1
+    paddd               m5, m7
+
+    phaddd              m6, m5
+    vpermq              m6, m6, q3120
+    paddd               m6, m2
+    vextracti128        xm5,m6, 1
+    psrad               xm6, INTERP_SHIFT_PS
+    psrad               xm5, INTERP_SHIFT_PS
+    packssdw            xm6, xm5
+    movu                [r2 + 16], xm6
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+    IPFILTER_LUMA_PS_16xN_AVX2 4
+    IPFILTER_LUMA_PS_16xN_AVX2 8
+    IPFILTER_LUMA_PS_16xN_AVX2 12
+    IPFILTER_LUMA_PS_16xN_AVX2 16
+    IPFILTER_LUMA_PS_16xN_AVX2 32
+    IPFILTER_LUMA_PS_16xN_AVX2 64
+
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_ps_12x16, 4, 6, 8
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+    shl                 r4d, 4
+%ifdef PIC
+    lea                 r6, [tab_LumaCoeff]
+    vpbroadcastq        m0, [r6 + r4]
+    vpbroadcastq        m1, [r6 + r4 + 8]
+%else
+    vpbroadcastq        m0, [tab_LumaCoeff + r4]
+    vpbroadcastq        m1, [tab_LumaCoeff + r4 + 8]
+%endif
+    mova                m3, [interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 6
+    test                r5d, r5d
+    mov                 r4d, 16
+    jz                  .loop0
+    lea                 r6, [r1*3]
+    sub                 r0, r6
+    add                 r4d, 7
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m7, m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m7, m1
+    paddd               m4, m7
+
+    vbroadcasti128      m6, [r0 + 16]
+    pshufb              m5, m3
+    pshufb              m7, m6, m3
+    pmaddwd             m5, m0
+    pmaddwd             m7, m1
+    paddd               m5, m7
+
+    phaddd              m4, m5
+    vpermq              m4, m4, q3120
+    paddd               m4, m2
+    vextracti128        xm5,m4, 1
+    psrad               xm4, INTERP_SHIFT_PS
+    psrad               xm5, INTERP_SHIFT_PS
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+
+    vbroadcasti128      m5, [r0 + 24]
+    pshufb              m6, m3
+    pshufb              m5, m3
+    pmaddwd             m6, m0
+    pmaddwd             m5, m1
+    paddd               m6, m5
+
+    phaddd              m6, m6
+    vpermq              m6, m6, q3120
+    paddd               xm6, xm2
+    psrad               xm6, INTERP_SHIFT_PS
+    packssdw            xm6, xm6
+    movq                [r2 + 16], xm6
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif

x265_2.7.tar.gz/source/common/x86/h-ipfilter8.asm Added

@@ -0,0 +1,6736 @@
+;*****************************************************************************
+;* Copyright (C) 2013-2017 MulticoreWare, Inc
+;*
+;* Authors: Min Chen <chenm003@163.com>
+;*          Nabajit Deka <nabajit@multicorewareinc.com>
+;*          Praveen Kumar Tiwari <praveen@multicorewareinc.com>
+;*
+;* This program is free software; you can redistribute it and/or modify
+;* it under the terms of the GNU General Public License as published by
+;* the Free Software Foundation; either version 2 of the License, or
+;* (at your option) any later version.
+;*
+;* This program is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+;* GNU General Public License for more details.
+;*
+;* You should have received a copy of the GNU General Public License
+;* along with this program; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
+;*
+;* This program is also available under a commercial proprietary license.
+;* For more information, contact us at license @ x265.com.
+;*****************************************************************************/
+
+%include "x86inc.asm"
+%include "x86util.asm"
+
+SECTION_RODATA 32
+
+const h_tabw_LumaCoeff,  dw   0, 0,  0,  64,  0,   0,  0,  0
+                       dw  -1, 4, -10, 58,  17, -5,  1,  0
+                       dw  -1, 4, -11, 40,  40, -11, 4, -1
+                       dw   0, 1, -5,  17,  58, -10, 4, -1
+
+const h_tabw_ChromaCoeff, dw  0, 64,  0,  0
+                        dw -2, 58, 10, -2
+                        dw -4, 54, 16, -2
+                        dw -6, 46, 28, -4
+                        dw -4, 36, 36, -4
+                        dw -4, 28, 46, -6
+                        dw -2, 16, 54, -4
+                        dw -2, 10, 58, -2
+
+const h_tab_ChromaCoeff, db  0, 64,  0,  0
+                       db -2, 58, 10, -2
+                       db -4, 54, 16, -2
+                       db -6, 46, 28, -4
+                       db -4, 36, 36, -4
+                       db -4, 28, 46, -6
+                       db -2, 16, 54, -4
+                       db -2, 10, 58, -2
+
+const h_tab_LumaCoeff,   db   0, 0,  0,  64,  0,   0,  0,  0
+                       db  -1, 4, -10, 58,  17, -5,  1,  0
+                       db  -1, 4, -11, 40,  40, -11, 4, -1
+                       db   0, 1, -5,  17,  58, -10, 4, -1
+
+const h_tab_Tm,    db 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
+                 db 4, 5, 6, 7, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10
+                 db 8, 9,10,11, 9,10,11,12,10,11,12,13,11,12,13, 14
+
+
+const h_tab_Lm,    db 0, 1, 2, 3, 4,  5,  6,  7,  1, 2, 3, 4,  5,  6,  7,  8
+                 db 2, 3, 4, 5, 6,  7,  8,  9,  3, 4, 5, 6,  7,  8,  9,  10
+                 db 4, 5, 6, 7, 8,  9,  10, 11, 5, 6, 7, 8,  9,  10, 11, 12
+                 db 6, 7, 8, 9, 10, 11, 12, 13, 7, 8, 9, 10, 11, 12, 13, 14
+
+const interp4_shuf, times 2 db 0, 1, 8, 9, 4, 5, 12, 13, 2, 3, 10, 11, 6, 7, 14, 15
+
+const h_interp8_hps_shuf,     dd 0, 4, 1, 5, 2, 6, 3, 7
+
+const interp4_hpp_shuf,     times 2 db 0, 1, 2, 3, 1, 2, 3, 4, 8, 9, 10, 11, 9, 10, 11, 12
+
+const interp4_horiz_shuf1,  db 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
+                            db 8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 11, 12, 13, 14
+
+const h_pd_526336, times 8 dd 8192*64+2048
+
+const pb_LumaCoeffVer,  times 16 db 0, 0
+                        times 16 db 0, 64
+                        times 16 db 0, 0
+                        times 16 db 0, 0
+
+                        times 16 db -1, 4
+                        times 16 db -10, 58
+                        times 16 db 17, -5
+                        times 16 db 1, 0
+
+                        times 16 db -1, 4
+                        times 16 db -11, 40
+                        times 16 db 40, -11
+                        times 16 db 4, -1
+
+                        times 16 db 0, 1
+                        times 16 db -5, 17
+                        times 16 db 58, -10
+                        times 16 db 4, -1
+
+const h_pw_LumaCoeffVer,  times 8 dw 0, 0
+                        times 8 dw 0, 64
+                        times 8 dw 0, 0
+                        times 8 dw 0, 0
+
+                        times 8 dw -1, 4
+                        times 8 dw -10, 58
+                        times 8 dw 17, -5
+                        times 8 dw 1, 0
+
+                        times 8 dw -1, 4
+                        times 8 dw -11, 40
+                        times 8 dw 40, -11
+                        times 8 dw 4, -1
+
+                        times 8 dw 0, 1
+                        times 8 dw -5, 17
+                        times 8 dw 58, -10
+                        times 8 dw 4, -1
+
+const pb_8tap_hps_0, times 2 db 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8
+                     times 2 db 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,10
+                     times 2 db 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,10,10,11,11,12
+                     times 2 db 6, 7, 7, 8, 8, 9, 9,10,10,11,11,12,12,13,13,14
+
+ALIGN 32
+interp4_hps_shuf: times 2 db 0, 1, 2, 3, 1, 2, 3, 4, 8, 9, 10, 11, 9, 10, 11, 12
+
+SECTION .text
+
+cextern pw_1
+cextern pw_32
+cextern pw_2000
+cextern pw_512
+
+%macro PROCESS_LUMA_AVX2_W8_16R 1
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r7, [r0 + r1 * 4]
+    movu            xm4, [r7]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+    movu            xm5, [r7 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 2 * mmsize]
+    paddd           m0, m6
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm7, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddwd         m7, m5, [r5 + 2 * mmsize]
+    paddd           m1, m7
+    pmaddwd         m7, m5, [r5 + 1 * mmsize]
+    paddd           m3, m7
+    pmaddwd         m5, [r5]
+    movu            xm7, [r7 + r4]                  ; m7 = row 7
+    punpckhwd       xm8, xm6, xm7
+    punpcklwd       xm6, xm7
+    vinserti128     m6, m6, xm8, 1
+    pmaddwd         m8, m6, [r5 + 3 * mmsize]
+    paddd           m0, m8
+    pmaddwd         m8, m6, [r5 + 2 * mmsize]
+    paddd           m2, m8
+    pmaddwd         m8, m6, [r5 + 1 * mmsize]
+    paddd           m4, m8
+    pmaddwd         m6, [r5]
+    lea             r7, [r7 + r1 * 4]
+    movu            xm8, [r7]                       ; m8 = row 8
+    punpckhwd       xm9, xm7, xm8
+    punpcklwd       xm7, xm8
+    vinserti128     m7, m7, xm9, 1
+    pmaddwd         m9, m7, [r5 + 3 * mmsize]
+    paddd           m1, m9
+    pmaddwd         m9, m7, [r5 + 2 * mmsize]
+    paddd           m3, m9
+    pmaddwd         m9, m7, [r5 + 1 * mmsize]
+    paddd           m5, m9
+    pmaddwd         m7, [r5]
+    movu            xm9, [r7 + r1]                  ; m9 = row 9
+    punpckhwd       xm10, xm8, xm9
+    punpcklwd       xm8, xm9
+    vinserti128     m8, m8, xm10, 1
+    pmaddwd         m10, m8, [r5 + 3 * mmsize]
+    paddd           m2, m10
+    pmaddwd         m10, m8, [r5 + 2 * mmsize]
+    paddd           m4, m10
+    pmaddwd         m10, m8, [r5 + 1 * mmsize]
+    paddd           m6, m10
+    pmaddwd         m8, [r5]
+    movu            xm10, [r7 + r1 * 2]             ; m10 = row 10
+    punpckhwd       xm11, xm9, xm10
+    punpcklwd       xm9, xm10
+    vinserti128     m9, m9, xm11, 1
+    pmaddwd         m11, m9, [r5 + 3 * mmsize]
+    paddd           m3, m11
+    pmaddwd         m11, m9, [r5 + 2 * mmsize]
+    paddd           m5, m11
+    pmaddwd         m11, m9, [r5 + 1 * mmsize]
+    paddd           m7, m11
+    pmaddwd         m9, [r5]
+    movu            xm11, [r7 + r4]                 ; m11 = row 11
+    punpckhwd       xm12, xm10, xm11
+    punpcklwd       xm10, xm11
+    vinserti128     m10, m10, xm12, 1
+    pmaddwd         m12, m10, [r5 + 3 * mmsize]
+    paddd           m4, m12
+    pmaddwd         m12, m10, [r5 + 2 * mmsize]
+    paddd           m6, m12
+    pmaddwd         m12, m10, [r5 + 1 * mmsize]
+    paddd           m8, m12
+    pmaddwd         m10, [r5]
+    lea             r7, [r7 + r1 * 4]
+    movu            xm12, [r7]                      ; m12 = row 12
+    punpckhwd       xm13, xm11, xm12
+    punpcklwd       xm11, xm12
+    vinserti128     m11, m11, xm13, 1
+    pmaddwd         m13, m11, [r5 + 3 * mmsize]
+    paddd           m5, m13
+    pmaddwd         m13, m11, [r5 + 2 * mmsize]
+    paddd           m7, m13
+    pmaddwd         m13, m11, [r5 + 1 * mmsize]
+    paddd           m9, m13
+    pmaddwd         m11, [r5]
+
+%ifidn %1,sp
+    paddd           m0, m14
+    paddd           m1, m14
+    paddd           m2, m14
+    paddd           m3, m14
+    paddd           m4, m14
+    paddd           m5, m14
+    psrad           m0, 12
+    psrad           m1, 12
+    psrad           m2, 12
+    psrad           m3, 12
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+    psrad           m2, 6
+    psrad           m3, 6
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m0, m1
+    packssdw        m2, m3
+    packssdw        m4, m5
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m5, [h_interp8_hps_shuf]
+    vpermd          m0, m5, m0
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+%endif
+
+    movu            xm13, [r7 + r1]                 ; m13 = row 13
+    punpckhwd       xm0, xm12, xm13
+    punpcklwd       xm12, xm13
+    vinserti128     m12, m12, xm0, 1
+    pmaddwd         m0, m12, [r5 + 3 * mmsize]
+    paddd           m6, m0
+    pmaddwd         m0, m12, [r5 + 2 * mmsize]
+    paddd           m8, m0
+    pmaddwd         m0, m12, [r5 + 1 * mmsize]
+    paddd           m10, m0
+    pmaddwd         m12, [r5]
+    movu            xm0, [r7 + r1 * 2]              ; m0 = row 14
+    punpckhwd       xm1, xm13, xm0
+    punpcklwd       xm13, xm0
+    vinserti128     m13, m13, xm1, 1
+    pmaddwd         m1, m13, [r5 + 3 * mmsize]
+    paddd           m7, m1
+    pmaddwd         m1, m13, [r5 + 2 * mmsize]
+    paddd           m9, m1
+    pmaddwd         m1, m13, [r5 + 1 * mmsize]
+    paddd           m11, m1
+    pmaddwd         m13, [r5]
+
+%ifidn %1,sp
+    paddd           m6, m14
+    paddd           m7, m14
+    psrad           m6, 12
+    psrad           m7, 12
+%else
+    psrad           m6, 6
+    psrad           m7, 6
+%endif
+    packssdw        m6, m7
+    lea             r8, [r2 + r3 * 4]
+
+%ifidn %1,sp
+    packuswb        m4, m6
+    vpermd          m4, m5, m4
+    vextracti128    xm6, m4, 1
+    movq            [r8], xm4
+    movhps          [r8 + r3], xm4
+    movq            [r8 + r3 * 2], xm6
+    movhps          [r8 + r6], xm6
+%else
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm1, m4, 1
+    vextracti128    xm7, m6, 1
+    movu            [r8], xm4
+    movu            [r8 + r3], xm1
+    movu            [r8 + r3 * 2], xm6
+    movu            [r8 + r6], xm7
+%endif
+
+    movu            xm1, [r7 + r4]                  ; m1 = row 15
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m2, m0, [r5 + 3 * mmsize]
+    paddd           m8, m2
+    pmaddwd         m2, m0, [r5 + 2 * mmsize]
+    paddd           m10, m2
+    pmaddwd         m2, m0, [r5 + 1 * mmsize]
+    paddd           m12, m2
+    pmaddwd         m0, [r5]
+    lea             r7, [r7 + r1 * 4]
+    movu            xm2, [r7]                       ; m2 = row 16
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m3, m1, [r5 + 3 * mmsize]
+    paddd           m9, m3
+    pmaddwd         m3, m1, [r5 + 2 * mmsize]
+    paddd           m11, m3
+    pmaddwd         m3, m1, [r5 + 1 * mmsize]
+    paddd           m13, m3
+    pmaddwd         m1, [r5]
+    movu            xm3, [r7 + r1]                  ; m3 = row 17
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 3 * mmsize]
+    paddd           m10, m4
+    pmaddwd         m4, m2, [r5 + 2 * mmsize]
+    paddd           m12, m4
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+    movu            xm4, [r7 + r1 * 2]              ; m4 = row 18
+    punpckhwd       xm2, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm2, 1
+    pmaddwd         m2, m3, [r5 + 3 * mmsize]
+    paddd           m11, m2
+    pmaddwd         m2, m3, [r5 + 2 * mmsize]
+    paddd           m13, m2
+    pmaddwd         m3, [r5 + 1 * mmsize]
+    paddd           m1, m3
+    movu            xm2, [r7 + r4]                  ; m2 = row 19
+    punpckhwd       xm6, xm4, xm2
+    punpcklwd       xm4, xm2
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 3 * mmsize]
+    paddd           m12, m6
+    pmaddwd         m4, [r5 + 2 * mmsize]
+    paddd           m0, m4
+    lea             r7, [r7 + r1 * 4]
+    movu            xm6, [r7]                       ; m6 = row 20
+    punpckhwd       xm7, xm2, xm6
+    punpcklwd       xm2, xm6
+    vinserti128     m2, m2, xm7, 1
+    pmaddwd         m7, m2, [r5 + 3 * mmsize]
+    paddd           m13, m7
+    pmaddwd         m2, [r5 + 2 * mmsize]
+    paddd           m1, m2
+    movu            xm7, [r7 + r1]                  ; m7 = row 21
+    punpckhwd       xm2, xm6, xm7
+    punpcklwd       xm6, xm7
+    vinserti128     m6, m6, xm2, 1
+    pmaddwd         m6, [r5 + 3 * mmsize]
+    paddd           m0, m6
+    movu            xm2, [r7 + r1 * 2]              ; m2 = row 22
+    punpckhwd       xm3, xm7, xm2
+    punpcklwd       xm7, xm2
+    vinserti128     m7, m7, xm3, 1
+    pmaddwd         m7, [r5 + 3 * mmsize]
+    paddd           m1, m7
+
+%ifidn %1,sp
+    paddd           m8, m14
+    paddd           m9, m14
+    paddd           m10, m14
+    paddd           m11, m14
+    paddd           m12, m14
+    paddd           m13, m14
+    paddd           m0, m14
+    paddd           m1, m14
+    psrad           m8, 12
+    psrad           m9, 12
+    psrad           m10, 12
+    psrad           m11, 12
+    psrad           m12, 12
+    psrad           m13, 12
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m8, 6
+    psrad           m9, 6
+    psrad           m10, 6
+    psrad           m11, 6
+    psrad           m12, 6
+    psrad           m13, 6
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m8, m9
+    packssdw        m10, m11
+    packssdw        m12, m13
+    packssdw        m0, m1
+    lea             r8, [r8 + r3 * 4]
+
+%ifidn %1,sp
+    packuswb        m8, m10
+    packuswb        m12, m0
+    vpermd          m8, m5, m8
+    vpermd          m12, m5, m12
+    vextracti128    xm10, m8, 1
+    vextracti128    xm0, m12, 1
+    movq            [r8], xm8
+    movhps          [r8 + r3], xm8
+    movq            [r8 + r3 * 2], xm10
+    movhps          [r8 + r6], xm10
+    lea             r8, [r8 + r3 * 4]
+    movq            [r8], xm12
+    movhps          [r8 + r3], xm12
+    movq            [r8 + r3 * 2], xm0
+    movhps          [r8 + r6], xm0
+%else
+    vpermq          m8, m8, 11011000b
+    vpermq          m10, m10, 11011000b
+    vpermq          m12, m12, 11011000b
+    vpermq          m0, m0, 11011000b
+    vextracti128    xm9, m8, 1
+    vextracti128    xm11, m10, 1
+    vextracti128    xm13, m12, 1
+    vextracti128    xm1, m0, 1
+    movu            [r8], xm8
+    movu            [r8 + r3], xm9
+    movu            [r8 + r3 * 2], xm10
+    movu            [r8 + r6], xm11
+    lea             r8, [r8 + r3 * 4]
+    movu            [r8], xm12
+    movu            [r8 + r3], xm13
+    movu            [r8 + r3 * 2], xm0
+    movu            [r8 + r6], xm1
+%endif
+%endmacro
+
+%macro FILTER_H8_W8_sse2 0
+    movh        m1, [r0 + x - 3]
+    movh        m4, [r0 + x - 2]
+    punpcklbw   m1, m6
+    punpcklbw   m4, m6
+    movh        m5, [r0 + x - 1]
+    movh        m0, [r0 + x]
+    punpcklbw   m5, m6
+    punpcklbw   m0, m6
+    pmaddwd     m1, m3
+    pmaddwd     m4, m3
+    pmaddwd     m5, m3
+    pmaddwd     m0, m3
+    packssdw    m1, m4
+    packssdw    m5, m0
+    pshuflw     m4, m1, q2301
+    pshufhw     m4, m4, q2301
+    pshuflw     m0, m5, q2301
+    pshufhw     m0, m0, q2301
+    paddw       m1, m4
+    paddw       m5, m0
+    psrldq      m1, 2
+    psrldq      m5, 2
+    pshufd      m1, m1, q3120
+    pshufd      m5, m5, q3120
+    punpcklqdq  m1, m5
+    movh        m7, [r0 + x + 1]
+    movh        m4, [r0 + x + 2]
+    punpcklbw   m7, m6
+    punpcklbw   m4, m6
+    movh        m5, [r0 + x + 3]
+    movh        m0, [r0 + x + 4]
+    punpcklbw   m5, m6
+    punpcklbw   m0, m6
+    pmaddwd     m7, m3
+    pmaddwd     m4, m3
+    pmaddwd     m5, m3
+    pmaddwd     m0, m3
+    packssdw    m7, m4
+    packssdw    m5, m0
+    pshuflw     m4, m7, q2301
+    pshufhw     m4, m4, q2301
+    pshuflw     m0, m5, q2301
+    pshufhw     m0, m0, q2301
+    paddw       m7, m4
+    paddw       m5, m0
+    psrldq      m7, 2
+    psrldq      m5, 2
+    pshufd      m7, m7, q3120
+    pshufd      m5, m5, q3120
+    punpcklqdq  m7, m5
+    pshuflw     m4, m1, q2301
+    pshufhw     m4, m4, q2301
+    pshuflw     m0, m7, q2301
+    pshufhw     m0, m0, q2301
+    paddw       m1, m4
+    paddw       m7, m0
+    psrldq      m1, 2
+    psrldq      m7, 2
+    pshufd      m1, m1, q3120
+    pshufd      m7, m7, q3120
+    punpcklqdq  m1, m7
+%endmacro
+
+%macro FILTER_H8_W4_sse2 0
+    movh        m1, [r0 + x - 3]
+    movh        m0, [r0 + x - 2]
+    punpcklbw   m1, m6
+    punpcklbw   m0, m6
+    movh        m4, [r0 + x - 1]
+    movh        m5, [r0 + x]
+    punpcklbw   m4, m6
+    punpcklbw   m5, m6
+    pmaddwd     m1, m3
+    pmaddwd     m0, m3
+    pmaddwd     m4, m3
+    pmaddwd     m5, m3
+    packssdw    m1, m0
+    packssdw    m4, m5
+    pshuflw     m0, m1, q2301
+    pshufhw     m0, m0, q2301
+    pshuflw     m5, m4, q2301
+    pshufhw     m5, m5, q2301
+    paddw       m1, m0
+    paddw       m4, m5
+    psrldq      m1, 2
+    psrldq      m4, 2
+    pshufd      m1, m1, q3120
+    pshufd      m4, m4, q3120
+    punpcklqdq  m1, m4
+    pshuflw     m0, m1, q2301
+    pshufhw     m0, m0, q2301
+    paddw       m1, m0
+    psrldq      m1, 2
+    pshufd      m1, m1, q3120
+%endmacro
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_LUMA_sse2 3
+INIT_XMM sse2
+cglobal interp_8tap_horiz_%3_%1x%2, 4,6,8
+    mov       r4d, r4m
+    add       r4d, r4d
+    pxor      m6, m6
+
+%ifidn %3, ps
+    add       r3d, r3d
+    cmp       r5m, byte 0
+%endif
+
+%ifdef PIC
+    lea       r5, [h_tabw_LumaCoeff]
+    movu      m3, [r5 + r4 * 8]
+%else
+    movu      m3, [h_tabw_LumaCoeff + r4 * 8]
+%endif
+
+    mov       r4d, %2
+
+%ifidn %3, pp
+    mova      m2, [pw_32]
+%else
+    mova      m2, [pw_2000]
+    je        .loopH
+    lea       r5, [r1 + 2 * r1]
+    sub       r0, r5
+    add       r4d, 7
+%endif
+
+.loopH:
+%assign x 0
+%rep %1 / 8
+    FILTER_H8_W8_sse2
+  %ifidn %3, pp
+    paddw     m1, m2
+    psraw     m1, 6
+    packuswb  m1, m1
+    movh      [r2 + x], m1
+  %else
+    psubw     m1, m2
+    movu      [r2 + 2 * x], m1
+  %endif
+%assign x x+8
+%endrep
+
+%rep (%1 % 8) / 4
+    FILTER_H8_W4_sse2
+  %ifidn %3, pp
+    paddw     m1, m2
+    psraw     m1, 6
+    packuswb  m1, m1
+    movd      [r2 + x], m1
+  %else
+    psubw     m1, m2
+    movh      [r2 + 2 * x], m1
+  %endif
+%endrep
+
+    add       r0, r1
+    add       r2, r3
+
+    dec       r4d
+    jnz       .loopH
+    RET
+
+%endmacro
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+    IPFILTER_LUMA_sse2 4, 4, pp
+    IPFILTER_LUMA_sse2 4, 8, pp
+    IPFILTER_LUMA_sse2 8, 4, pp
+    IPFILTER_LUMA_sse2 8, 8, pp
+    IPFILTER_LUMA_sse2 16, 16, pp
+    IPFILTER_LUMA_sse2 16, 8, pp
+    IPFILTER_LUMA_sse2 8, 16, pp
+    IPFILTER_LUMA_sse2 16, 12, pp
+    IPFILTER_LUMA_sse2 12, 16, pp
+    IPFILTER_LUMA_sse2 16, 4, pp
+    IPFILTER_LUMA_sse2 4, 16, pp
+    IPFILTER_LUMA_sse2 32, 32, pp
+    IPFILTER_LUMA_sse2 32, 16, pp
+    IPFILTER_LUMA_sse2 16, 32, pp
+    IPFILTER_LUMA_sse2 32, 24, pp
+    IPFILTER_LUMA_sse2 24, 32, pp
+    IPFILTER_LUMA_sse2 32, 8, pp
+    IPFILTER_LUMA_sse2 8, 32, pp
+    IPFILTER_LUMA_sse2 64, 64, pp
+    IPFILTER_LUMA_sse2 64, 32, pp
+    IPFILTER_LUMA_sse2 32, 64, pp
+    IPFILTER_LUMA_sse2 64, 48, pp
+    IPFILTER_LUMA_sse2 48, 64, pp
+    IPFILTER_LUMA_sse2 64, 16, pp
+    IPFILTER_LUMA_sse2 16, 64, pp
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+    IPFILTER_LUMA_sse2 4, 4, ps
+    IPFILTER_LUMA_sse2 8, 8, ps
+    IPFILTER_LUMA_sse2 8, 4, ps
+    IPFILTER_LUMA_sse2 4, 8, ps
+    IPFILTER_LUMA_sse2 16, 16, ps
+    IPFILTER_LUMA_sse2 16, 8, ps
+    IPFILTER_LUMA_sse2 8, 16, ps
+    IPFILTER_LUMA_sse2 16, 12, ps
+    IPFILTER_LUMA_sse2 12, 16, ps
+    IPFILTER_LUMA_sse2 16, 4, ps
+    IPFILTER_LUMA_sse2 4, 16, ps
+    IPFILTER_LUMA_sse2 32, 32, ps
+    IPFILTER_LUMA_sse2 32, 16, ps
+    IPFILTER_LUMA_sse2 16, 32, ps
+    IPFILTER_LUMA_sse2 32, 24, ps
+    IPFILTER_LUMA_sse2 24, 32, ps
+    IPFILTER_LUMA_sse2 32, 8, ps
+    IPFILTER_LUMA_sse2 8, 32, ps
+    IPFILTER_LUMA_sse2 64, 64, ps
+    IPFILTER_LUMA_sse2 64, 32, ps
+    IPFILTER_LUMA_sse2 32, 64, ps
+    IPFILTER_LUMA_sse2 64, 48, ps
+    IPFILTER_LUMA_sse2 48, 64, ps
+    IPFILTER_LUMA_sse2 64, 16, ps
+    IPFILTER_LUMA_sse2 16, 64, ps
+
+%macro FILTER_H4_w2_2_sse2 0
+    pxor        m3, m3
+    movd        m0, [srcq - 1]
+    movd        m2, [srcq]
+    punpckldq   m0, m2
+    punpcklbw   m0, m3
+    movd        m1, [srcq + srcstrideq - 1]
+    movd        m2, [srcq + srcstrideq]
+    punpckldq   m1, m2
+    punpcklbw   m1, m3
+    pmaddwd     m0, m4
+    pmaddwd     m1, m4
+    packssdw    m0, m1
+    pshuflw     m1, m0, q2301
+    pshufhw     m1, m1, q2301
+    paddw       m0, m1
+    psrld       m0, 16
+    packssdw    m0, m0
+    paddw       m0, m5
+    psraw       m0, 6
+    packuswb    m0, m0
+    movd        r4d, m0
+    mov         [dstq], r4w
+    shr         r4, 16
+    mov         [dstq + dststrideq], r4w
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_2xN(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_H4_W2xN_sse3 1
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_2x%1, 4, 6, 6, src, srcstride, dst, dststride
+    mov         r4d,    r4m
+    mova        m5,     [pw_32]
+
+%ifdef PIC
+    lea         r5,     [h_tabw_ChromaCoeff]
+    movddup     m4,     [r5 + r4 * 8]
+%else
+    movddup     m4,     [h_tabw_ChromaCoeff + r4 * 8]
+%endif
+
+%assign x 1
+%rep %1/2
+    FILTER_H4_w2_2_sse2
+%if x < %1/2
+    lea         srcq,   [srcq + srcstrideq * 2]
+    lea         dstq,   [dstq + dststrideq * 2]
+%endif
+%assign x x+1
+%endrep
+
+    RET
+
+%endmacro
+
+    FILTER_H4_W2xN_sse3 4
+    FILTER_H4_W2xN_sse3 8
+    FILTER_H4_W2xN_sse3 16
+
+%macro FILTER_H4_w4_2_sse2 0
+    pxor        m5, m5
+    movd        m0, [srcq - 1]
+    movd        m6, [srcq]
+    punpckldq   m0, m6
+    punpcklbw   m0, m5
+    movd        m1, [srcq + 1]
+    movd        m6, [srcq + 2]
+    punpckldq   m1, m6
+    punpcklbw   m1, m5
+    movd        m2, [srcq + srcstrideq - 1]
+    movd        m6, [srcq + srcstrideq]
+    punpckldq   m2, m6
+    punpcklbw   m2, m5
+    movd        m3, [srcq + srcstrideq + 1]
+    movd        m6, [srcq + srcstrideq + 2]
+    punpckldq   m3, m6
+    punpcklbw   m3, m5
+    pmaddwd     m0, m4
+    pmaddwd     m1, m4
+    pmaddwd     m2, m4
+    pmaddwd     m3, m4
+    packssdw    m0, m1
+    packssdw    m2, m3
+    pshuflw     m1, m0, q2301
+    pshufhw     m1, m1, q2301
+    pshuflw     m3, m2, q2301
+    pshufhw     m3, m3, q2301
+    paddw       m0, m1
+    paddw       m2, m3
+    psrld       m0, 16
+    psrld       m2, 16
+    packssdw    m0, m2
+    paddw       m0, m7
+    psraw       m0, 6
+    packuswb    m0, m2
+    movd        [dstq], m0
+    psrldq      m0, 4
+    movd        [dstq + dststrideq], m0
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_H4_W4xN_sse3 1
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_4x%1, 4, 6, 8, src, srcstride, dst, dststride
+    mov         r4d,    r4m
+    mova        m7,     [pw_32]
+
+%ifdef PIC
+    lea         r5,     [h_tabw_ChromaCoeff]
+    movddup     m4,     [r5 + r4 * 8]
+%else
+    movddup     m4,     [h_tabw_ChromaCoeff + r4 * 8]
+%endif
+
+%assign x 1
+%rep %1/2
+    FILTER_H4_w4_2_sse2
+%if x < %1/2
+    lea         srcq,   [srcq + srcstrideq * 2]
+    lea         dstq,   [dstq + dststrideq * 2]
+%endif
+%assign x x+1
+%endrep
+
+    RET
+
+%endmacro
+
+    FILTER_H4_W4xN_sse3 2
+    FILTER_H4_W4xN_sse3 4
+    FILTER_H4_W4xN_sse3 8
+    FILTER_H4_W4xN_sse3 16
+    FILTER_H4_W4xN_sse3 32
+
+%macro FILTER_H4_w6_sse2 0
+    pxor        m4, m4
+    movh        m0, [srcq - 1]
+    movh        m5, [srcq]
+    punpckldq   m0, m5
+    movhlps     m2, m0
+    punpcklbw   m0, m4
+    punpcklbw   m2, m4
+    movd        m1, [srcq + 1]
+    movd        m5, [srcq + 2]
+    punpckldq   m1, m5
+    punpcklbw   m1, m4
+    pmaddwd     m0, m6
+    pmaddwd     m1, m6
+    pmaddwd     m2, m6
+    packssdw    m0, m1
+    packssdw    m2, m2
+    pshuflw     m1, m0, q2301
+    pshufhw     m1, m1, q2301
+    pshuflw     m3, m2, q2301
+    paddw       m0, m1
+    paddw       m2, m3
+    psrld       m0, 16
+    psrld       m2, 16
+    packssdw    m0, m2
+    paddw       m0, m7
+    psraw       m0, 6
+    packuswb    m0, m0
+    movd        [dstq], m0
+    pextrw      r4d, m0, 2
+    mov         [dstq + 4], r4w
+%endmacro
+
+%macro FILH4W8_sse2 1
+    movh        m0, [srcq - 1 + %1]
+    movh        m5, [srcq + %1]
+    punpckldq   m0, m5
+    movhlps     m2, m0
+    punpcklbw   m0, m4
+    punpcklbw   m2, m4
+    movh        m1, [srcq + 1 + %1]
+    movh        m5, [srcq + 2 + %1]
+    punpckldq   m1, m5
+    movhlps     m3, m1
+    punpcklbw   m1, m4
+    punpcklbw   m3, m4
+    pmaddwd     m0, m6
+    pmaddwd     m1, m6
+    pmaddwd     m2, m6
+    pmaddwd     m3, m6
+    packssdw    m0, m1
+    packssdw    m2, m3
+    pshuflw     m1, m0, q2301
+    pshufhw     m1, m1, q2301
+    pshuflw     m3, m2, q2301
+    pshufhw     m3, m3, q2301
+    paddw       m0, m1
+    paddw       m2, m3
+    psrld       m0, 16
+    psrld       m2, 16
+    packssdw    m0, m2
+    paddw       m0, m7
+    psraw       m0, 6
+    packuswb    m0, m0
+    movh        [dstq + %1], m0
+%endmacro
+
+%macro FILTER_H4_w8_sse2 0
+    FILH4W8_sse2 0
+%endmacro
+
+%macro FILTER_H4_w12_sse2 0
+    FILH4W8_sse2 0
+    movd        m1, [srcq - 1 + 8]
+    movd        m3, [srcq + 8]
+    punpckldq   m1, m3
+    punpcklbw   m1, m4
+    movd        m2, [srcq + 1 + 8]
+    movd        m3, [srcq + 2 + 8]
+    punpckldq   m2, m3
+    punpcklbw   m2, m4
+    pmaddwd     m1, m6
+    pmaddwd     m2, m6
+    packssdw    m1, m2
+    pshuflw     m2, m1, q2301
+    pshufhw     m2, m2, q2301
+    paddw       m1, m2
+    psrld       m1, 16
+    packssdw    m1, m1
+    paddw       m1, m7
+    psraw       m1, 6
+    packuswb    m1, m1
+    movd        [dstq + 8], m1
+%endmacro
+
+%macro FILTER_H4_w16_sse2 0
+    FILH4W8_sse2 0
+    FILH4W8_sse2 8
+%endmacro
+
+%macro FILTER_H4_w24_sse2 0
+    FILH4W8_sse2 0
+    FILH4W8_sse2 8
+    FILH4W8_sse2 16
+%endmacro
+
+%macro FILTER_H4_w32_sse2 0
+    FILH4W8_sse2 0
+    FILH4W8_sse2 8
+    FILH4W8_sse2 16
+    FILH4W8_sse2 24
+%endmacro
+
+%macro FILTER_H4_w48_sse2 0
+    FILH4W8_sse2 0
+    FILH4W8_sse2 8
+    FILH4W8_sse2 16
+    FILH4W8_sse2 24
+    FILH4W8_sse2 32
+    FILH4W8_sse2 40
+%endmacro
+
+%macro FILTER_H4_w64_sse2 0
+    FILH4W8_sse2 0
+    FILH4W8_sse2 8
+    FILH4W8_sse2 16
+    FILH4W8_sse2 24
+    FILH4W8_sse2 32
+    FILH4W8_sse2 40
+    FILH4W8_sse2 48
+    FILH4W8_sse2 56
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_sse3 2
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 8, src, srcstride, dst, dststride
+    mov         r4d,        r4m
+    mova        m7,         [pw_32]
+    pxor        m4,         m4
+
+%ifdef PIC
+    lea         r5,          [h_tabw_ChromaCoeff]
+    movddup     m6,       [r5 + r4 * 8]
+%else
+    movddup     m6,       [h_tabw_ChromaCoeff + r4 * 8]
+%endif
+
+%assign x 1
+%rep %2
+    FILTER_H4_w%1_sse2
+%if x < %2
+    add         srcq,        srcstrideq
+    add         dstq,        dststrideq
+%endif
+%assign x x+1
+%endrep
+
+    RET
+
+%endmacro
+
+    IPFILTER_CHROMA_sse3 6,   8
+    IPFILTER_CHROMA_sse3 8,   2
+    IPFILTER_CHROMA_sse3 8,   4
+    IPFILTER_CHROMA_sse3 8,   6
+    IPFILTER_CHROMA_sse3 8,   8
+    IPFILTER_CHROMA_sse3 8,  16
+    IPFILTER_CHROMA_sse3 8,  32
+    IPFILTER_CHROMA_sse3 12, 16
+
+    IPFILTER_CHROMA_sse3 6,  16
+    IPFILTER_CHROMA_sse3 8,  12
+    IPFILTER_CHROMA_sse3 8,  64
+    IPFILTER_CHROMA_sse3 12, 32
+
+    IPFILTER_CHROMA_sse3 16,  4
+    IPFILTER_CHROMA_sse3 16,  8
+    IPFILTER_CHROMA_sse3 16, 12
+    IPFILTER_CHROMA_sse3 16, 16
+    IPFILTER_CHROMA_sse3 16, 32
+    IPFILTER_CHROMA_sse3 32,  8
+    IPFILTER_CHROMA_sse3 32, 16
+    IPFILTER_CHROMA_sse3 32, 24
+    IPFILTER_CHROMA_sse3 24, 32
+    IPFILTER_CHROMA_sse3 32, 32
+
+    IPFILTER_CHROMA_sse3 16, 24
+    IPFILTER_CHROMA_sse3 16, 64
+    IPFILTER_CHROMA_sse3 32, 48
+    IPFILTER_CHROMA_sse3 24, 64
+    IPFILTER_CHROMA_sse3 32, 64
+
+    IPFILTER_CHROMA_sse3 64, 64
+    IPFILTER_CHROMA_sse3 64, 32
+    IPFILTER_CHROMA_sse3 64, 48
+    IPFILTER_CHROMA_sse3 48, 64
+    IPFILTER_CHROMA_sse3 64, 16
+
+%macro FILTER_2 2
+    movd        m3,     [srcq + %1]
+    movd        m4,     [srcq + 1 + %1]
+    punpckldq   m3,     m4
+    punpcklbw   m3,     m0
+    pmaddwd     m3,     m1
+    packssdw    m3,     m3
+    pshuflw     m4,     m3, q2301
+    paddw       m3,     m4
+    psrldq      m3,     2
+    psubw       m3,     m2
+    movd        [dstq + %2], m3
+%endmacro
+
+%macro FILTER_4 2
+    movd        m3,     [srcq + %1]
+    movd        m4,     [srcq + 1 + %1]
+    punpckldq   m3,     m4
+    punpcklbw   m3,     m0
+    pmaddwd     m3,     m1
+    movd        m4,     [srcq + 2 + %1]
+    movd        m5,     [srcq + 3 + %1]
+    punpckldq   m4,     m5
+    punpcklbw   m4,     m0
+    pmaddwd     m4,     m1
+    packssdw    m3,     m4
+    pshuflw     m4,     m3, q2301
+    pshufhw     m4,     m4, q2301
+    paddw       m3,     m4
+    psrldq      m3,     2
+    pshufd      m3,     m3,     q3120
+    psubw       m3,     m2
+    movh        [dstq + %2], m3
+%endmacro
+
+%macro FILTER_4TAP_HPS_sse3 2
+INIT_XMM sse3
+cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
+    mov         r4d,    r4m
+    add         dststrided, dststrided
+    mova        m2,     [pw_2000]
+    pxor        m0,     m0
+
+%ifdef PIC
+    lea         r6,     [h_tabw_ChromaCoeff]
+    movddup     m1,     [r6 + r4 * 8]
+%else
+    movddup     m1,     [h_tabw_ChromaCoeff + r4 * 8]
+%endif
+
+    mov        r4d,     %2
+    cmp        r5m,     byte 0
+    je         .loopH
+    sub        srcq,    srcstrideq
+    add        r4d,     3
+
+.loopH:
+%assign x -1
+%assign y 0
+%rep %1/4
+    FILTER_4 x,y
+%assign x x+4
+%assign y y+8
+%endrep
+%rep (%1 % 4)/2
+    FILTER_2 x,y
+%endrep
+    add         srcq,   srcstrideq
+    add         dstq,   dststrideq
+
+    dec         r4d
+    jnz         .loopH
+    RET
+
+%endmacro
+
+    FILTER_4TAP_HPS_sse3 2, 4
+    FILTER_4TAP_HPS_sse3 2, 8
+    FILTER_4TAP_HPS_sse3 2, 16
+    FILTER_4TAP_HPS_sse3 4, 2
+    FILTER_4TAP_HPS_sse3 4, 4
+    FILTER_4TAP_HPS_sse3 4, 8
+    FILTER_4TAP_HPS_sse3 4, 16
+    FILTER_4TAP_HPS_sse3 4, 32
+    FILTER_4TAP_HPS_sse3 6, 8
+    FILTER_4TAP_HPS_sse3 6, 16
+    FILTER_4TAP_HPS_sse3 8, 2
+    FILTER_4TAP_HPS_sse3 8, 4
+    FILTER_4TAP_HPS_sse3 8, 6
+    FILTER_4TAP_HPS_sse3 8, 8
+    FILTER_4TAP_HPS_sse3 8, 12
+    FILTER_4TAP_HPS_sse3 8, 16
+    FILTER_4TAP_HPS_sse3 8, 32
+    FILTER_4TAP_HPS_sse3 8, 64
+    FILTER_4TAP_HPS_sse3 12, 16
+    FILTER_4TAP_HPS_sse3 12, 32
+    FILTER_4TAP_HPS_sse3 16, 4
+    FILTER_4TAP_HPS_sse3 16, 8
+    FILTER_4TAP_HPS_sse3 16, 12
+    FILTER_4TAP_HPS_sse3 16, 16
+    FILTER_4TAP_HPS_sse3 16, 24
+    FILTER_4TAP_HPS_sse3 16, 32
+    FILTER_4TAP_HPS_sse3 16, 64
+    FILTER_4TAP_HPS_sse3 24, 32
+    FILTER_4TAP_HPS_sse3 24, 64
+    FILTER_4TAP_HPS_sse3 32,  8
+    FILTER_4TAP_HPS_sse3 32, 16
+    FILTER_4TAP_HPS_sse3 32, 24
+    FILTER_4TAP_HPS_sse3 32, 32
+    FILTER_4TAP_HPS_sse3 32, 48
+    FILTER_4TAP_HPS_sse3 32, 64
+    FILTER_4TAP_HPS_sse3 48, 64
+    FILTER_4TAP_HPS_sse3 64, 16
+    FILTER_4TAP_HPS_sse3 64, 32
+    FILTER_4TAP_HPS_sse3 64, 48
+    FILTER_4TAP_HPS_sse3 64, 64
+
+%macro FILTER_H4_w2_2 3
+    movh        %2, [srcq - 1]
+    pshufb      %2, %2, Tm0
+    movh        %1, [srcq + srcstrideq - 1]
+    pshufb      %1, %1, Tm0
+    punpcklqdq  %2, %1
+    pmaddubsw   %2, coef2
+    phaddw      %2, %2
+    pmulhrsw    %2, %3
+    packuswb    %2, %2
+    movd        r4d, %2
+    mov         [dstq], r4w
+    shr         r4, 16
+    mov         [dstq + dststrideq], r4w
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_2x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_2x4, 4, 6, 5, src, srcstride, dst, dststride
+%define coef2       m4
+%define Tm0         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+
+%rep 2
+    FILTER_H4_w2_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endrep
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_2x8, 4, 6, 5, src, srcstride, dst, dststride
+%define coef2       m4
+%define Tm0         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+
+%rep 4
+    FILTER_H4_w2_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endrep
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_2x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_2x16, 4, 6, 5, src, srcstride, dst, dststride
+%define coef2       m4
+%define Tm0         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+
+    mov         r5d,        16/2
+
+.loop:
+    FILTER_H4_w2_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+    dec         r5d
+    jnz         .loop
+
+    RET
+
+%macro FILTER_H4_w4_2 3
+    movh        %2, [srcq - 1]
+    pshufb      %2, %2, Tm0
+    pmaddubsw   %2, coef2
+    movh        %1, [srcq + srcstrideq - 1]
+    pshufb      %1, %1, Tm0
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    pmulhrsw    %2, %3
+    packuswb    %2, %2
+    movd        [dstq], %2
+    palignr     %2, %2, 4
+    movd        [dstq + dststrideq], %2
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_4x2, 4, 6, 5, src, srcstride, dst, dststride
+%define coef2       m4
+%define Tm0         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+
+    FILTER_H4_w4_2   t0, t1, t2
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_4x4, 4, 6, 5, src, srcstride, dst, dststride
+%define coef2       m4
+%define Tm0         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+
+%rep 2
+    FILTER_H4_w4_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endrep
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_4x8, 4, 6, 5, src, srcstride, dst, dststride
+%define coef2       m4
+%define Tm0         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+
+%rep 4
+    FILTER_H4_w4_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endrep
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_4x16, 4, 6, 5, src, srcstride, dst, dststride
+%define coef2       m4
+%define Tm0         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+
+%rep 8
+    FILTER_H4_w4_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endrep
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_4x32, 4, 6, 5, src, srcstride, dst, dststride
+%define coef2       m4
+%define Tm0         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+
+    mov         r5d,        32/2
+
+.loop:
+    FILTER_H4_w4_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+    dec         r5d
+    jnz         .loop
+
+    RET
+
+ALIGN 32
+const interp_4tap_8x8_horiz_shuf,   dd 0, 4, 1, 5, 2, 6, 3, 7
+
+%macro FILTER_H4_w6 3
+    movu        %1, [srcq - 1]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    pmulhrsw    %2, %3
+    packuswb    %2, %2
+    movd        [dstq],      %2
+    pextrw      [dstq + 4], %2, 2
+%endmacro
+
+%macro FILTER_H4_w8 3
+    movu        %1, [srcq - 1]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    pmulhrsw    %2, %3
+    packuswb    %2, %2
+    movh        [dstq],      %2
+%endmacro
+
+%macro FILTER_H4_w12 3
+    movu        %1, [srcq - 1]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    pmulhrsw    %2, %3
+    movu        %1, [srcq - 1 + 8]
+    pshufb      %1, %1, Tm0
+    pmaddubsw   %1, coef2
+    phaddw      %1, %1
+    pmulhrsw    %1, %3
+    packuswb    %2, %1
+    movh        [dstq],      %2
+    pextrd      [dstq + 8], %2, 2
+%endmacro
+
+%macro FILTER_H4_w16 4
+    movu        %1, [srcq - 1]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq - 1 + 8]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    pmulhrsw    %2, %3
+    pmulhrsw    %4, %3
+    packuswb    %2, %4
+    movu        [dstq],      %2
+%endmacro
+
+%macro FILTER_H4_w24 4
+    movu        %1, [srcq - 1]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq - 1 + 8]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    pmulhrsw    %2, %3
+    pmulhrsw    %4, %3
+    packuswb    %2, %4
+    movu        [dstq],          %2
+    movu        %1, [srcq - 1 + 16]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    pmulhrsw    %2, %3
+    packuswb    %2, %2
+    movh        [dstq + 16],     %2
+%endmacro
+
+%macro FILTER_H4_w32 4
+    movu        %1, [srcq - 1]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq - 1 + 8]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    pmulhrsw    %2, %3
+    pmulhrsw    %4, %3
+    packuswb    %2, %4
+    movu        [dstq],      %2
+    movu        %1, [srcq - 1 + 16]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq - 1 + 24]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    pmulhrsw    %2, %3
+    pmulhrsw    %4, %3
+    packuswb    %2, %4
+    movu        [dstq + 16],      %2
+%endmacro
+
+%macro FILTER_H4_w16o 5
+    movu        %1, [srcq + %5 - 1]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq + %5 - 1 + 8]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    pmulhrsw    %2, %3
+    pmulhrsw    %4, %3
+    packuswb    %2, %4
+    movu        [dstq + %5],      %2
+%endmacro
+
+%macro FILTER_H4_w48 4
+    FILTER_H4_w16o %1, %2, %3, %4, 0
+    FILTER_H4_w16o %1, %2, %3, %4, 16
+    FILTER_H4_w16o %1, %2, %3, %4, 32
+%endmacro
+
+%macro FILTER_H4_w64 4
+    FILTER_H4_w16o %1, %2, %3, %4, 0
+    FILTER_H4_w16o %1, %2, %3, %4, 16
+    FILTER_H4_w16o %1, %2, %3, %4, 32
+    FILTER_H4_w16o %1, %2, %3, %4, 48
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro IPFILTER_CHROMA 2
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 6, src, srcstride, dst, dststride
+%define coef2       m5
+%define Tm0         m4
+%define Tm1         m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,        r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mov           r5d,       %2
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+    mova        Tm1,         [h_tab_Tm + 16]
+
+.loop:
+    FILTER_H4_w%1   t0, t1, t2
+    add         srcq,        srcstrideq
+    add         dstq,        dststrideq
+
+    dec         r5d
+    jnz        .loop
+
+    RET
+%endmacro
+
+
+    IPFILTER_CHROMA 6,   8
+    IPFILTER_CHROMA 8,   2
+    IPFILTER_CHROMA 8,   4
+    IPFILTER_CHROMA 8,   6
+    IPFILTER_CHROMA 8,   8
+    IPFILTER_CHROMA 8,  16
+    IPFILTER_CHROMA 8,  32
+    IPFILTER_CHROMA 12, 16
+
+    IPFILTER_CHROMA 6,  16
+    IPFILTER_CHROMA 8,  12
+    IPFILTER_CHROMA 8,  64
+    IPFILTER_CHROMA 12, 32
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_W 2
+INIT_XMM sse4
+cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 7, src, srcstride, dst, dststride
+%define coef2       m6
+%define Tm0         m5
+%define Tm1         m4
+%define t3          m3
+%define t2          m2
+%define t1          m1
+%define t0          m0
+
+    mov         r4d,         r4m
+
+%ifdef PIC
+    lea         r5,          [h_tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
+%else
+    movd        coef2,       [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mov         r5d,          %2
+
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [h_tab_Tm]
+    mova        Tm1,         [h_tab_Tm + 16]
+
+.loop:
+    FILTER_H4_w%1   t0, t1, t2, t3
+    add         srcq,        srcstrideq
+    add         dstq,        dststrideq
+
+    dec         r5d
+    jnz        .loop
+
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_W 16,  4
+    IPFILTER_CHROMA_W 16,  8
+    IPFILTER_CHROMA_W 16, 12
+    IPFILTER_CHROMA_W 16, 16
+    IPFILTER_CHROMA_W 16, 32
+    IPFILTER_CHROMA_W 32,  8
+    IPFILTER_CHROMA_W 32, 16
+    IPFILTER_CHROMA_W 32, 24
+    IPFILTER_CHROMA_W 24, 32
+    IPFILTER_CHROMA_W 32, 32
+
+    IPFILTER_CHROMA_W 16, 24
+    IPFILTER_CHROMA_W 16, 64
+    IPFILTER_CHROMA_W 32, 48
+    IPFILTER_CHROMA_W 24, 64
+    IPFILTER_CHROMA_W 32, 64
+
+    IPFILTER_CHROMA_W 64, 64
+    IPFILTER_CHROMA_W 64, 32
+    IPFILTER_CHROMA_W 64, 48
+    IPFILTER_CHROMA_W 48, 64
+    IPFILTER_CHROMA_W 64, 16
+
+%macro FILTER_H8_W8 7-8   ; t0, t1, t2, t3, coef, c512, src, dst
+    movu        %1, %7
+    pshufb      %2, %1, [h_tab_Lm +  0]
+    pmaddubsw   %2, %5
+    pshufb      %3, %1, [h_tab_Lm + 16]
+    pmaddubsw   %3, %5
+    phaddw      %2, %3
+    pshufb      %4, %1, [h_tab_Lm + 32]
+    pmaddubsw   %4, %5
+    pshufb      %1, %1, [h_tab_Lm + 48]
+    pmaddubsw   %1, %5
+    phaddw      %4, %1
+    phaddw      %2, %4
+  %if %0 == 8
+    pmulhrsw    %2, %6
+    packuswb    %2, %2
+    movh        %8, %2
+  %endif
+%endmacro
+
+%macro FILTER_H8_W4 2
+    movu        %1, [r0 - 3 + r5]
+    pshufb      %2, %1, [h_tab_Lm]
+    pmaddubsw   %2, m3
+    pshufb      m7, %1, [h_tab_Lm + 16]
+    pmaddubsw   m7, m3
+    phaddw      %2, m7
+    phaddw      %2, %2
+%endmacro
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_LUMA 3
+INIT_XMM sse4
+cglobal interp_8tap_horiz_%3_%1x%2, 4,7,8
+
+    mov       r4d, r4m
+
+%ifdef PIC
+    lea       r6, [h_tab_LumaCoeff]
+    movh      m3, [r6 + r4 * 8]
+%else
+    movh      m3, [h_tab_LumaCoeff + r4 * 8]
+%endif
+    punpcklqdq  m3, m3
+
+%ifidn %3, pp
+    mova      m2, [pw_512]
+%else
+    mova      m2, [pw_2000]
+%endif
+
+    mov       r4d, %2
+%ifidn %3, ps
+    add       r3, r3
+    cmp       r5m, byte 0
+    je        .loopH
+    lea       r6, [r1 + 2 * r1]
+    sub       r0, r6
+    add       r4d, 7
+%endif
+
+.loopH:
+    xor       r5, r5
+%rep %1 / 8
+  %ifidn %3, pp
+    FILTER_H8_W8  m0, m1, m4, m5, m3, m2, [r0 - 3 + r5], [r2 + r5]
+  %else
+    FILTER_H8_W8  m0, m1, m4, m5, m3, UNUSED, [r0 - 3 + r5]
+    psubw     m1, m2
+    movu      [r2 + 2 * r5], m1
+  %endif
+    add       r5, 8
+%endrep
+
+%rep (%1 % 8) / 4
+    FILTER_H8_W4  m0, m1
+  %ifidn %3, pp
+    pmulhrsw  m1, m2
+    packuswb  m1, m1
+    movd      [r2 + r5], m1
+  %else
+    psubw     m1, m2
+    movh      [r2 + 2 * r5], m1
+  %endif
+%endrep
+
+    add       r0, r1
+    add       r2, r3
+
+    dec       r4d
+    jnz       .loopH
+    RET
+%endmacro
+
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_4x4, 4,6,6
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea             r5, [h_tab_LumaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [h_tab_LumaCoeff + r4 * 8]
+%endif
+
+    mova            m1, [h_tab_Lm]
+    vpbroadcastd    m2, [pw_1]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    sub             r0, 3
+    ; Row 0-1
+    vbroadcasti128  m3, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m3, m1
+    pmaddubsw       m3, m0
+    pmaddwd         m3, m2
+    vbroadcasti128  m4, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddwd         m4, m2
+    phaddd          m3, m4                          ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
+
+    ; Row 2-3
+    lea             r0, [r0 + r1 * 2]
+    vbroadcasti128  m4, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddwd         m4, m2
+    vbroadcasti128  m5, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m5, m1
+    pmaddubsw       m5, m0
+    pmaddwd         m5, m2
+    phaddd          m4, m5                          ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
+
+    packssdw        m3, m4                          ; WORD [R3D R3C R2D R2C R1D R1C R0D R0C R3B R3A R2B R2A R1B R1A R0B R0A]
+    pmulhrsw        m3, [pw_512]
+    vextracti128    xm4, m3, 1
+    packuswb        xm3, xm4                        ; BYTE [R3D R3C R2D R2C R1D R1C R0D R0C R3B R3A R2B R2A R1B R1A R0B R0A]
+    pshufb          xm3, [interp4_shuf]             ; [row3 row1 row2 row0]
+
+    lea             r0, [r3 * 3]
+    movd            [r2], xm3
+    pextrd          [r2+r3], xm3, 2
+    pextrd          [r2+r3*2], xm3, 1
+    pextrd          [r2+r0], xm3, 3
+    RET
+
+%macro FILTER_HORIZ_LUMA_AVX2_4xN 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_pp_4x%1, 4, 6, 9
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea             r5, [h_tab_LumaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [h_tab_LumaCoeff + r4 * 8]
+%endif
+
+    mova            m1, [h_tab_Lm]
+    mova            m2, [pw_1]
+    mova            m7, [h_interp8_hps_shuf]
+    mova            m8, [pw_512]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+    lea             r4, [r1 * 3]
+    lea             r5, [r3 * 3]
+    sub             r0, 3
+%rep %1 / 8
+    ; Row 0-1
+    vbroadcasti128  m3, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m3, m1
+    pmaddubsw       m3, m0
+    pmaddwd         m3, m2
+    vbroadcasti128  m4, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddwd         m4, m2
+    phaddd          m3, m4                          ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
+
+    ; Row 2-3
+    vbroadcasti128  m4, [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddwd         m4, m2
+    vbroadcasti128  m5, [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m5, m1
+    pmaddubsw       m5, m0
+    pmaddwd         m5, m2
+    phaddd          m4, m5                          ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
+
+    packssdw        m3, m4                          ; WORD [R3D R3C R2D R2C R1D R1C R0D R0C R3B R3A R2B R2A R1B R1A R0B R0A]
+    lea             r0, [r0 + r1 * 4]
+    ; Row 4-5
+    vbroadcasti128  m5, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m5, m1
+    pmaddubsw       m5, m0
+    pmaddwd         m5, m2
+    vbroadcasti128  m4, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddwd         m4, m2
+    phaddd          m5, m4                          ; DWORD [R5D R5C R4D R4C R5B R5A R4B R4A]
+
+    ; Row 6-7
+    vbroadcasti128  m4, [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddwd         m4, m2
+    vbroadcasti128  m6, [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m6, m1
+    pmaddubsw       m6, m0
+    pmaddwd         m6, m2
+    phaddd          m4, m6                          ; DWORD [R7D R7C R6D R6C R7B R7A R6B R6A]
+
+    packssdw        m5, m4                          ; WORD [R7D R7C R6D R6C R5D R5C R4D R4C R7B R7A R6B R6A R5B R5A R4B R4A]
+    vpermd          m3, m7, m3
+    vpermd          m5, m7, m5
+    pmulhrsw        m3, m8
+    pmulhrsw        m5, m8
+    packuswb        m3, m5
+    vextracti128    xm5, m3, 1
+
+    movd            [r2], xm3
+    pextrd          [r2 + r3], xm3, 1
+    movd            [r2 + r3 * 2], xm5
+    pextrd          [r2 + r5], xm5, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm3, 2
+    pextrd          [r2 + r3], xm3, 3
+    pextrd          [r2 + r3 * 2], xm5, 2
+    pextrd          [r2 + r5], xm5, 3
+    lea             r0, [r0 + r1 * 4]
+    lea             r2, [r2 + r3 * 4]
+%endrep
+    RET
+%endif
+%endmacro
+
+    FILTER_HORIZ_LUMA_AVX2_4xN 8
+    FILTER_HORIZ_LUMA_AVX2_4xN 16
+
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_8x4, 4, 6, 7
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea             r5, [h_tab_LumaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [h_tab_LumaCoeff + r4 * 8]
+%endif
+
+    mova            m1, [h_tab_Lm]
+    mova            m2, [h_tab_Lm + 32]
+
+    ; register map
+    ; m0     - interpolate coeff
+    ; m1, m2 - shuffle order table
+
+    sub             r0, 3
+    lea             r5, [r1 * 3]
+    lea             r4, [r3 * 3]
+
+    ; Row 0
+    vbroadcasti128  m3, [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m4, m3, m2
+    pshufb          m3, m1
+    pmaddubsw       m3, m0
+    pmaddubsw       m4, m0
+    phaddw          m3, m4
+    ; Row 1
+    vbroadcasti128  m4, [r0 + r1]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m5, m4, m2
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddubsw       m5, m0
+    phaddw          m4, m5
+
+    phaddw          m3, m4                          ; WORD [R1H R1G R1D R1C R0H R0G R0D R0C R1F R1E R1B R1A R0F R0E R0B R0A]
+    pmulhrsw        m3, [pw_512]
+
+    ; Row 2
+    vbroadcasti128  m4, [r0 + r1 * 2]               ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m5, m4, m2
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddubsw       m5, m0
+    phaddw          m4, m5
+    ; Row 3
+    vbroadcasti128  m5, [r0 + r5]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m6, m5, m2
+    pshufb          m5, m1
+    pmaddubsw       m5, m0
+    pmaddubsw       m6, m0
+    phaddw          m5, m6
+
+    phaddw          m4, m5                          ; WORD [R3H R3G R3D R3C R2H R2G R2D R2C R3F R3E R3B R3A R2F R2E R2B R2A]
+    pmulhrsw        m4, [pw_512]
+
+    packuswb        m3, m4
+    vextracti128    xm4, m3, 1
+    punpcklwd       xm5, xm3, xm4
+
+    movq            [r2], xm5
+    movhps          [r2 + r3], xm5
+
+    punpckhwd       xm5, xm3, xm4
+    movq            [r2 + r3 * 2], xm5
+    movhps          [r2 + r4], xm5
+    RET
+
+%macro IPFILTER_LUMA_AVX2_8xN 2
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_%1x%2, 4, 7, 7
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea             r5, [h_tab_LumaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [h_tab_LumaCoeff + r4 * 8]
+%endif
+
+    mova            m1, [h_tab_Lm]
+    mova            m2, [h_tab_Lm + 32]
+
+    ; register map
+    ; m0     - interpolate coeff
+    ; m1, m2 - shuffle order table
+
+    sub             r0, 3
+    lea             r5, [r1 * 3]
+    lea             r6, [r3 * 3]
+    mov             r4d, %2 / 4
+.loop:
+    ; Row 0
+    vbroadcasti128  m3, [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m4, m3, m2
+    pshufb          m3, m1
+    pmaddubsw       m3, m0
+    pmaddubsw       m4, m0
+    phaddw          m3, m4
+    ; Row 1
+    vbroadcasti128  m4, [r0 + r1]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m5, m4, m2
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddubsw       m5, m0
+    phaddw          m4, m5
+
+    phaddw          m3, m4                          ; WORD [R1H R1G R1D R1C R0H R0G R0D R0C R1F R1E R1B R1A R0F R0E R0B R0A]
+    pmulhrsw        m3, [pw_512]
+
+    ; Row 2
+    vbroadcasti128  m4, [r0 + r1 * 2]               ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m5, m4, m2
+    pshufb          m4, m1
+    pmaddubsw       m4, m0
+    pmaddubsw       m5, m0
+    phaddw          m4, m5
+    ; Row 3
+    vbroadcasti128  m5, [r0 + r5]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb          m6, m5, m2
+    pshufb          m5, m1
+    pmaddubsw       m5, m0
+    pmaddubsw       m6, m0
+    phaddw          m5, m6
+
+    phaddw          m4, m5                          ; WORD [R3H R3G R3D R3C R2H R2G R2D R2C R3F R3E R3B R3A R2F R2E R2B R2A]
+    pmulhrsw        m4, [pw_512]
+
+    packuswb        m3, m4
+    vextracti128    xm4, m3, 1
+    punpcklwd       xm5, xm3, xm4
+
+    movq            [r2], xm5
+    movhps          [r2 + r3], xm5
+
+    punpckhwd       xm5, xm3, xm4
+    movq            [r2 + r3 * 2], xm5
+    movhps          [r2 + r6], xm5
+
+    lea             r0, [r0 + r1 * 4]
+    lea             r2, [r2 + r3 * 4]
+    dec             r4d
+    jnz             .loop
+    RET
+%endmacro
+
+    IPFILTER_LUMA_AVX2_8xN 8, 8
+    IPFILTER_LUMA_AVX2_8xN 8, 16
+    IPFILTER_LUMA_AVX2_8xN 8, 32
+
+%macro IPFILTER_LUMA_AVX2 2
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
+    sub               r0,        3
+    mov               r4d,       r4m
+%ifdef PIC
+    lea               r5,        [h_tab_LumaCoeff]
+    vpbroadcastd      m0,        [r5 + r4 * 8]
+    vpbroadcastd      m1,        [r5 + r4 * 8 + 4]
+%else
+    vpbroadcastd      m0,         [h_tab_LumaCoeff + r4 * 8]
+    vpbroadcastd      m1,         [h_tab_LumaCoeff + r4 * 8 + 4]
+%endif
+    movu              m3,         [h_tab_Tm + 16]
+    vpbroadcastd      m7,         [pw_1]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+    ; m2 , m2  shuffle order table
+    ; m7 - pw_1
+
+    mov               r4d,        %2/2
+.loop:
+    ; Row 0
+    vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,         m4,     m3
+    pshufb            m4,         [h_tab_Tm]
+    pmaddubsw         m4,         m0
+    pmaddubsw         m5,         m1
+    paddw             m4,         m5
+    pmaddwd           m4,         m7
+    vbroadcasti128    m5,         [r0 + 8]                    ; second 8 elements in Row0
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [h_tab_Tm]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
+    pmulhrsw          m4,         [pw_512]
+    vbroadcasti128    m2,         [r0 + r1]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,         m2,     m3
+    pshufb            m2,         [h_tab_Tm]
+    pmaddubsw         m2,         m0
+    pmaddubsw         m5,         m1
+    paddw             m2,         m5
+    pmaddwd           m2,         m7
+    vbroadcasti128    m5,         [r0 + r1 + 8]                    ; second 8 elements in Row0
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [h_tab_Tm]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+    packssdw          m2,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
+    pmulhrsw          m2,         [pw_512]
+    packuswb          m4,         m2
+    vpermq            m4,         m4,     11011000b
+    vextracti128      xm5,        m4,     1
+    pshufd            xm4,        xm4,    11011000b
+    pshufd            xm5,        xm5,    11011000b
+    movu              [r2],       xm4
+    movu              [r2+r3],    xm5
+    lea               r0,         [r0 + r1 * 2]
+    lea               r2,         [r2 + r3 * 2]
+    dec               r4d
+    jnz              .loop
+    RET
+%endmacro
+
+%macro IPFILTER_LUMA_32x_avx2 2
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
+    sub               r0,         3
+    mov               r4d,        r4m
+%ifdef PIC
+    lea               r5,         [h_tab_LumaCoeff]
+    vpbroadcastd      m0,         [r5 + r4 * 8]
+    vpbroadcastd      m1,         [r5 + r4 * 8 + 4]
+%else
+    vpbroadcastd      m0,         [h_tab_LumaCoeff + r4 * 8]
+    vpbroadcastd      m1,         [h_tab_LumaCoeff + r4 * 8 + 4]
+%endif
+    movu              m3,         [h_tab_Tm + 16]
+    vpbroadcastd      m7,         [pw_1]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+    ; m2 , m2  shuffle order table
+    ; m7 - pw_1
+
+    mov               r4d,        %2
+.loop:
+    ; Row 0
+    vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,         m4,     m3
+    pshufb            m4,         [h_tab_Tm]
+    pmaddubsw         m4,         m0
+    pmaddubsw         m5,         m1
+    paddw             m4,         m5
+    pmaddwd           m4,         m7
+    vbroadcasti128    m5,         [r0 + 8]
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [h_tab_Tm]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
+    pmulhrsw          m4,         [pw_512]
+    vbroadcasti128    m2,         [r0 + 16]
+    pshufb            m5,         m2,     m3
+    pshufb            m2,         [h_tab_Tm]
+    pmaddubsw         m2,         m0
+    pmaddubsw         m5,         m1
+    paddw             m2,         m5
+    pmaddwd           m2,         m7
+    vbroadcasti128    m5,         [r0 + 24]
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [h_tab_Tm]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+    packssdw          m2,         m5
+    pmulhrsw          m2,         [pw_512]
+    packuswb          m4,         m2
+    vpermq            m4,         m4,     11011000b
+    vextracti128      xm5,        m4,     1
+    pshufd            xm4,        xm4,    11011000b
+    pshufd            xm5,        xm5,    11011000b
+    movu              [r2],       xm4
+    movu              [r2 + 16],  xm5
+    lea               r0,         [r0 + r1]
+    lea               r2,         [r2 + r3]
+    dec               r4d
+    jnz               .loop
+    RET
+%endmacro
+%macro IPFILTER_LUMA_64x_avx2 2
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
+    sub               r0,    3
+    mov               r4d,   r4m
+%ifdef PIC
+    lea               r5,        [h_tab_LumaCoeff]
+    vpbroadcastd      m0,        [r5 + r4 * 8]
+    vpbroadcastd      m1,        [r5 + r4 * 8 + 4]
+%else
+    vpbroadcastd      m0,        [h_tab_LumaCoeff + r4 * 8]
+    vpbroadcastd      m1,        [h_tab_LumaCoeff + r4 * 8 + 4]
+%endif
+    movu              m3,        [h_tab_Tm + 16]
+    vpbroadcastd      m7,        [pw_1]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+    ; m2 , m2  shuffle order table
+    ; m7 - pw_1
+
+    mov               r4d,   %2
+.loop:
+    ; Row 0
+    vbroadcasti128    m4,        [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,        m4,    m3
+    pshufb            m4,        [h_tab_Tm]
+    pmaddubsw         m4,        m0
+    pmaddubsw         m5,        m1
+    paddw             m4,        m5
+    pmaddwd           m4,        m7
+    vbroadcasti128    m5,        [r0 + 8]
+    pshufb            m6,        m5,    m3
+    pshufb            m5,        [h_tab_Tm]
+    pmaddubsw         m5,        m0
+    pmaddubsw         m6,        m1
+    paddw             m5,        m6
+    pmaddwd           m5,        m7
+    packssdw          m4,        m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
+    pmulhrsw          m4,        [pw_512]
+    vbroadcasti128    m2,        [r0 + 16]
+    pshufb            m5,        m2,    m3
+    pshufb            m2,        [h_tab_Tm]
+    pmaddubsw         m2,        m0
+    pmaddubsw         m5,        m1
+    paddw             m2,        m5
+    pmaddwd           m2,        m7
+    vbroadcasti128    m5,        [r0 + 24]
+    pshufb            m6,        m5,    m3
+    pshufb            m5,        [h_tab_Tm]
+    pmaddubsw         m5,        m0
+    pmaddubsw         m6,        m1
+    paddw             m5,        m6
+    pmaddwd           m5,        m7
+    packssdw          m2,        m5
+    pmulhrsw          m2,        [pw_512]
+    packuswb          m4,        m2
+    vpermq            m4,        m4,    11011000b
+    vextracti128      xm5,       m4,    1
+    pshufd            xm4,       xm4,   11011000b
+    pshufd            xm5,       xm5,   11011000b
+    movu              [r2],      xm4
+    movu              [r2 + 16], xm5
+
+    vbroadcasti128    m4,        [r0 + 32]
+    pshufb            m5,        m4,    m3
+    pshufb            m4,        [h_tab_Tm]
+    pmaddubsw         m4,        m0
+    pmaddubsw         m5,        m1
+    paddw             m4,        m5
+    pmaddwd           m4,        m7
+    vbroadcasti128    m5,        [r0 + 40]
+    pshufb            m6,        m5,    m3
+    pshufb            m5,        [h_tab_Tm]
+    pmaddubsw         m5,        m0
+    pmaddubsw         m6,        m1
+    paddw             m5,        m6
+    pmaddwd           m5,        m7
+    packssdw          m4,        m5
+    pmulhrsw          m4,        [pw_512]
+    vbroadcasti128    m2,        [r0 + 48]
+    pshufb            m5,        m2,    m3
+    pshufb            m2,        [h_tab_Tm]
+    pmaddubsw         m2,        m0
+    pmaddubsw         m5,        m1
+    paddw             m2,        m5
+    pmaddwd           m2,        m7
+    vbroadcasti128    m5,        [r0 + 56]
+    pshufb            m6,        m5,    m3
+    pshufb            m5,        [h_tab_Tm]
+    pmaddubsw         m5,        m0
+    pmaddubsw         m6,        m1
+    paddw             m5,        m6
+    pmaddwd           m5,        m7
+    packssdw          m2,        m5
+    pmulhrsw          m2,        [pw_512]
+    packuswb          m4,        m2
+    vpermq            m4,        m4,    11011000b
+    vextracti128      xm5,       m4,    1
+    pshufd            xm4,       xm4,   11011000b
+    pshufd            xm5,       xm5,   11011000b
+    movu              [r2 +32],  xm4
+    movu              [r2 + 48], xm5
+
+    lea               r0,        [r0 + r1]
+    lea               r2,        [r2 + r3]
+    dec               r4d
+    jnz               .loop
+    RET
+%endmacro
+
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_48x64, 4,6,8
+    sub               r0,         3
+    mov               r4d,        r4m
+%ifdef PIC
+    lea               r5,         [h_tab_LumaCoeff]
+    vpbroadcastd      m0,         [r5 + r4 * 8]
+    vpbroadcastd      m1,         [r5 + r4 * 8 + 4]
+%else
+    vpbroadcastd      m0,         [h_tab_LumaCoeff + r4 * 8]
+    vpbroadcastd      m1,         [h_tab_LumaCoeff + r4 * 8 + 4]
+%endif
+    movu              m3,         [h_tab_Tm + 16]
+    vpbroadcastd      m7,         [pw_1]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+    ; m2 , m2  shuffle order table
+    ; m7 - pw_1
+
+    mov               r4d,        64
+.loop:
+    ; Row 0
+    vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,         m4,     m3
+    pshufb            m4,         [h_tab_Tm]
+    pmaddubsw         m4,         m0
+    pmaddubsw         m5,         m1
+    paddw             m4,         m5
+    pmaddwd           m4,         m7
+    vbroadcasti128    m5,         [r0 + 8]
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [h_tab_Tm]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
+    pmulhrsw          m4,         [pw_512]
+
+    vbroadcasti128    m2,         [r0 + 16]
+    pshufb            m5,         m2,     m3
+    pshufb            m2,         [h_tab_Tm]
+    pmaddubsw         m2,         m0
+    pmaddubsw         m5,         m1
+    paddw             m2,         m5
+    pmaddwd           m2,         m7
+    vbroadcasti128    m5,         [r0 + 24]
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [h_tab_Tm]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+    packssdw          m2,         m5
+    pmulhrsw          m2,         [pw_512]
+    packuswb          m4,         m2
+    vpermq            m4,         m4,     11011000b
+    vextracti128      xm5,        m4,     1
+    pshufd            xm4,        xm4,    11011000b
+    pshufd            xm5,        xm5,    11011000b
+    movu              [r2],       xm4
+    movu              [r2 + 16],  xm5
+
+    vbroadcasti128    m4,         [r0 + 32]
+    pshufb            m5,         m4,     m3
+    pshufb            m4,         [h_tab_Tm]
+    pmaddubsw         m4,         m0
+    pmaddubsw         m5,         m1
+    paddw             m4,         m5
+    pmaddwd           m4,         m7
+    vbroadcasti128    m5,         [r0 + 40]
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [h_tab_Tm]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+    packssdw          m4,         m5
+    pmulhrsw          m4,         [pw_512]
+    packuswb          m4,         m4
+    vpermq            m4,         m4,     11011000b
+    pshufd            xm4,        xm4,    11011000b
+    movu              [r2 + 32],  xm4
+
+    lea               r0,         [r0 + r1]
+    lea               r2,         [r2 + r3]
+    dec               r4d
+    jnz               .loop
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_4x4, 4,6,6
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vpbroadcastd      m2,           [pw_1]
+    vbroadcasti128    m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec                r0
+
+    ; Row 0-1
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    vinserti128       m3,           m3,      [r0 + r1],     1
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 2-3
+    lea               r0,           [r0 + r1 * 2]
+    vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    vinserti128       m4,           m4,      [r0 + r1],     1
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    pmulhrsw          m3,           [pw_512]
+    vextracti128      xm4,          m3,     1
+    packuswb          xm3,          xm4
+
+    lea               r0,           [r3 * 3]
+    movd              [r2],         xm3
+    pextrd            [r2+r3],      xm3,     2
+    pextrd            [r2+r3*2],    xm3,     1
+    pextrd            [r2+r0],      xm3,     3
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_2x4, 4, 6, 3
+    mov               r4d,           r4m
+
+%ifdef PIC
+    lea               r5,            [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,            [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,            [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    dec               r0
+    lea               r4,            [r1 * 3]
+    movq              xm1,           [r0]
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m1,            m1,          xm2,          1
+    pshufb            m1,            [interp4_hpp_shuf]
+    pmaddubsw         m1,            m0
+    pmaddwd           m1,            [pw_1]
+    vextracti128      xm2,           m1,          1
+    packssdw          xm1,           xm2
+    pmulhrsw          xm1,           [pw_512]
+    packuswb          xm1,           xm1
+
+    lea               r4,            [r3 * 3]
+    pextrw            [r2],          xm1,         0
+    pextrw            [r2 + r3],     xm1,         1
+    pextrw            [r2 + r3 * 2], xm1,         2
+    pextrw            [r2 + r4],     xm1,         3
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_2x8, 4, 6, 6
+    mov               r4d,           r4m
+
+%ifdef PIC
+    lea               r5,            [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,            [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,            [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m4,            [interp4_hpp_shuf]
+    mova              m5,            [pw_1]
+    dec               r0
+    lea               r4,            [r1 * 3]
+    movq              xm1,           [r0]
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m1,            m1,          xm2,          1
+    lea               r0,            [r0 + r1 * 4]
+    movq              xm3,           [r0]
+    movhps            xm3,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m3,            m3,          xm2,          1
+
+    pshufb            m1,            m4
+    pshufb            m3,            m4
+    pmaddubsw         m1,            m0
+    pmaddubsw         m3,            m0
+    pmaddwd           m1,            m5
+    pmaddwd           m3,            m5
+    packssdw          m1,            m3
+    pmulhrsw          m1,            [pw_512]
+    vextracti128      xm2,           m1,          1
+    packuswb          xm1,           xm2
+
+    lea               r4,            [r3 * 3]
+    pextrw            [r2],          xm1,         0
+    pextrw            [r2 + r3],     xm1,         1
+    pextrw            [r2 + r3 * 2], xm1,         4
+    pextrw            [r2 + r4],     xm1,         5
+    lea               r2,            [r2 + r3 * 4]
+    pextrw            [r2],          xm1,         2
+    pextrw            [r2 + r3],     xm1,         3
+    pextrw            [r2 + r3 * 2], xm1,         6
+    pextrw            [r2 + r4],     xm1,         7
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_32x32, 4,6,7
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+    mova              m6,           [pw_512]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          32
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    vbroadcasti128    m4,           [r0 + 16]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + 20]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+
+    movu              [r2],         m3
+    lea               r2,           [r2 + r3]
+    lea               r0,           [r0 + r1]
+    dec               r4d
+    jnz               .loop
+    RET
+
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_16x16, 4, 6, 7
+    mov               r4d,          r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m6,           [pw_512]
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          8
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    ; Row 1
+    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+
+    vextracti128      xm4,          m3,       1
+    movu              [r2],         xm3
+    movu              [r2 + r3],    xm4
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1 * 2]
+    dec               r4d
+    jnz               .loop
+    RET
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+    IPFILTER_LUMA 4, 4, pp
+    IPFILTER_LUMA 4, 8, pp
+    IPFILTER_LUMA 12, 16, pp
+    IPFILTER_LUMA 4, 16, pp
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_8x8, 4,6,6
+    mov               r4d,    r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    movu              m1,           [h_tab_Tm]
+    vpbroadcastd      m2,           [pw_1]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    sub               r0,           1
+    mov               r4d,          2
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 1
+    vbroadcasti128    m4,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           [pw_512]
+    lea               r0,           [r0 + r1 * 2]
+
+    ; Row 2
+    vbroadcasti128    m4,           [r0 ]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    ; Row 3
+    vbroadcasti128    m5,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           [pw_512]
+
+    packuswb          m3,           m4
+    mova              m5,           [interp_4tap_8x8_horiz_shuf]
+    vpermd            m3,           m5,     m3
+    vextracti128      xm4,          m3,     1
+    movq              [r2],         xm3
+    movhps            [r2 + r3],    xm3
+    lea               r2,           [r2 + r3 * 2]
+    movq              [r2],         xm4
+    movhps            [r2 + r3],    xm4
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1*2]
+    dec               r4d
+    jnz               .loop
+    RET
+
+    IPFILTER_LUMA_AVX2 16, 4
+    IPFILTER_LUMA_AVX2 16, 8
+    IPFILTER_LUMA_AVX2 16, 12
+    IPFILTER_LUMA_AVX2 16, 16
+    IPFILTER_LUMA_AVX2 16, 32
+    IPFILTER_LUMA_AVX2 16, 64
+
+    IPFILTER_LUMA_32x_avx2 32 , 8
+    IPFILTER_LUMA_32x_avx2 32 , 16
+    IPFILTER_LUMA_32x_avx2 32 , 24
+    IPFILTER_LUMA_32x_avx2 32 , 32
+    IPFILTER_LUMA_32x_avx2 32 , 64
+
+    IPFILTER_LUMA_64x_avx2 64 , 64
+    IPFILTER_LUMA_64x_avx2 64 , 48
+    IPFILTER_LUMA_64x_avx2 64 , 32
+    IPFILTER_LUMA_64x_avx2 64 , 16
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_8x2, 4, 6, 5
+    mov               r4d,          r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,           [h_tab_Tm]
+    mova              m2,           [pw_1]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 1
+    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           [pw_512]
+    vextracti128      xm4,          m3,          1
+    packuswb          xm3,          xm4
+    pshufd            xm3,          xm3,         11011000b
+    movq              [r2],         xm3
+    movhps            [r2 + r3],    xm3
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_8x6, 4, 6, 7
+    mov               r4d,           r4m
+
+%ifdef PIC
+    lea               r5,            [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,            [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,            [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,            [h_tab_Tm]
+    mova              m2,            [pw_1]
+    mova              m6,            [pw_512]
+    lea               r4,            [r1 * 3]
+    lea               r5,            [r3 * 3]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    ; Row 0
+    vbroadcasti128    m3,            [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,            m1
+    pmaddubsw         m3,            m0
+    pmaddwd           m3,            m2
+
+    ; Row 1
+    vbroadcasti128    m4,            [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,            m1
+    pmaddubsw         m4,            m0
+    pmaddwd           m4,            m2
+    packssdw          m3,            m4
+    pmulhrsw          m3,            m6
+
+    ; Row 2
+    vbroadcasti128    m4,            [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,            m1
+    pmaddubsw         m4,            m0
+    pmaddwd           m4,            m2
+
+    ; Row 3
+    vbroadcasti128    m5,            [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,            m1
+    pmaddubsw         m5,            m0
+    pmaddwd           m5,            m2
+    packssdw          m4,            m5
+    pmulhrsw          m4,            m6
+
+    packuswb          m3,            m4
+    mova              m5,            [h_interp8_hps_shuf]
+    vpermd            m3,            m5,          m3
+    vextracti128      xm4,           m3,          1
+    movq              [r2],          xm3
+    movhps            [r2 + r3],     xm3
+    movq              [r2 + r3 * 2], xm4
+    movhps            [r2 + r5],     xm4
+    lea               r2,            [r2 + r3 * 4]
+    lea               r0,            [r0 + r1 * 4]
+    ; Row 4
+    vbroadcasti128    m3,            [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,            m1
+    pmaddubsw         m3,            m0
+    pmaddwd           m3,            m2
+
+    ; Row 5
+    vbroadcasti128    m4,            [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,            m1
+    pmaddubsw         m4,            m0
+    pmaddwd           m4,            m2
+    packssdw          m3,            m4
+    pmulhrsw          m3,            m6
+    vextracti128      xm4,           m3,          1
+    packuswb          xm3,           xm4
+    pshufd            xm3,           xm3,         11011000b
+    movq              [r2],          xm3
+    movhps            [r2 + r3],     xm3
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_6x8, 4, 6, 7
+    mov               r4d,               r4m
+
+%ifdef PIC
+    lea               r5,                [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,                [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,                [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,                [h_tab_Tm]
+    mova              m2,                [pw_1]
+    mova              m6,                [pw_512]
+    lea               r4,                [r1 * 3]
+    lea               r5,                [r3 * 3]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+%rep 2
+    ; Row 0
+    vbroadcasti128    m3,                [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,                m1
+    pmaddubsw         m3,                m0
+    pmaddwd           m3,                m2
+
+    ; Row 1
+    vbroadcasti128    m4,                [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,                m1
+    pmaddubsw         m4,                m0
+    pmaddwd           m4,                m2
+    packssdw          m3,                m4
+    pmulhrsw          m3,                m6
+
+    ; Row 2
+    vbroadcasti128    m4,                [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,                m1
+    pmaddubsw         m4,                m0
+    pmaddwd           m4,                m2
+
+    ; Row 3
+    vbroadcasti128    m5,                [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,                m1
+    pmaddubsw         m5,                m0
+    pmaddwd           m5,                m2
+    packssdw          m4,                m5
+    pmulhrsw          m4,                m6
+
+    packuswb          m3,                m4
+    vextracti128      xm4,               m3,          1
+    movd              [r2],              xm3
+    pextrw            [r2 + 4],          xm4,         0
+    pextrd            [r2 + r3],         xm3,         1
+    pextrw            [r2 + r3 + 4],     xm4,         2
+    pextrd            [r2 + r3 * 2],     xm3,         2
+    pextrw            [r2 + r3 * 2 + 4], xm4,         4
+    pextrd            [r2 + r5],         xm3,         3
+    pextrw            [r2 + r5 + 4],     xm4,         6
+    lea               r2,                [r2 + r3 * 4]
+    lea               r0,                [r0 + r1 * 4]
+%endrep
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_64xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------;
+%macro IPFILTER_CHROMA_HPS_64xN 1
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_64x%1, 4,7,6
+    mov             r4d, r4m
+    mov             r5d, r5m
+    add             r3d, r3d
+
+%ifdef PIC
+    lea               r6,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,           [pw_1]
+    vbroadcasti128     m5,           [pw_2000]
+    mova               m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+    mov                r6d,         %1
+    dec                r0
+    test                r5d,      r5d
+    je                 .loop
+    sub                r0 ,         r1
+    add                r6d ,        3
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 8]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+    vpermq            m3,           m3,          11011000b
+    movu              [r2],         m3
+
+    vbroadcasti128    m3,           [r0 + 16]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 24]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+    vpermq            m3,           m3,          11011000b
+    movu              [r2 + 32],    m3
+
+    vbroadcasti128    m3,           [r0 + 32]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 40]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+    vpermq            m3,           m3,          11011000b
+    movu              [r2 + 64],    m3
+
+    vbroadcasti128    m3,           [r0 + 48]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 56]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+    vpermq            m3,           m3,          11011000b
+    movu              [r2 + 96],    m3
+
+    add                r2,           r3
+    add                r0,           r1
+    dec                r6d
+    jnz                .loop
+    RET
+%endmacro
+
+   IPFILTER_CHROMA_HPS_64xN 64
+   IPFILTER_CHROMA_HPS_64xN 32
+   IPFILTER_CHROMA_HPS_64xN 48
+   IPFILTER_CHROMA_HPS_64xN 16
+
+;-----------------------------------------------------------------------------------------------------------------------------
+;void interp_horiz_ps_c(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+
+%macro IPFILTER_LUMA_PS_4xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_ps_4x%1, 6,7,6
+    mov                         r5d,               r5m
+    mov                         r4d,               r4m
+%ifdef PIC
+    lea                         r6,                [h_tab_LumaCoeff]
+    vpbroadcastq                m0,                [r6 + r4 * 8]
+%else
+    vpbroadcastq                m0,                [h_tab_LumaCoeff + r4 * 8]
+%endif
+    mova                        m1,                [h_tab_Lm]
+    add                         r3d,               r3d
+    vbroadcasti128              m2,                [pw_2000]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - pw_2000
+
+    sub                         r0,                3
+    test                        r5d,               r5d
+    mov                         r5d,               %1                           ; loop count variable - height
+    jz                         .preloop
+    lea                         r6,                [r1 * 3]                     ; r8 = (N / 2 - 1) * srcStride
+    sub                         r0,                r6                           ; r0(src) - 3 * srcStride
+    add                         r5d,               7                            ; need extra 7 rows, just set a specially flag here, blkheight += N - 1  (7 - 3 = 4 ; since the last three rows not in loop)
+
+.preloop:
+    lea                         r6,                [r3 * 3]
+.loop:
+    ; Row 0-1
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm
+    pmaddubsw                   m3,                m0
+    vbroadcasti128              m4,                [r0 + r1]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m1
+    pmaddubsw                   m4,                m0
+    phaddw                      m3,                m4                           ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
+
+    ; Row 2-3
+    lea                         r0,                [r0 + r1 * 2]                ;3rd row(i.e 2nd row)
+    vbroadcasti128              m4,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m1
+    pmaddubsw                   m4,                m0
+    vbroadcasti128              m5,                [r0 + r1]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m5,                m1
+    pmaddubsw                   m5,                m0
+    phaddw                      m4,                m5                           ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
+    phaddw                      m3,                m4                           ; all rows and col completed.
+
+    mova                        m5,                [h_interp8_hps_shuf]
+    vpermd                      m3,                m5,               m3
+    psubw                       m3,                m2
+
+    vextracti128                xm4,               m3,               1
+    movq                        [r2],              xm3                          ;row 0
+    movhps                      [r2 + r3],         xm3                          ;row 1
+    movq                        [r2 + r3 * 2],     xm4                          ;row 2
+    movhps                      [r2 + r6],         xm4                          ;row 3
+
+    lea                         r0,                [r0 + r1 * 2]                ; first loop src ->5th row(i.e 4)
+    lea                         r2,                [r2 + r3 * 4]                ; first loop dst ->5th row(i.e 4)
+    sub                         r5d,               4
+    jz                         .end
+    cmp                         r5d,               4
+    jge                        .loop
+
+    ; Row 8-9
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m3,                m1
+    pmaddubsw                   m3,                m0
+    vbroadcasti128              m4,                [r0 + r1]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m1
+    pmaddubsw                   m4,                m0
+    phaddw                      m3,                m4                           ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
+
+    ; Row 10
+    vbroadcasti128              m4,                [r0 + r1 * 2]                ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m1
+    pmaddubsw                   m4,                m0
+    phaddw                      m4,                m4                           ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
+    phaddw                      m3,                m4
+
+    vpermd                      m3,                m5,            m3            ; m5 don't broken in above
+    psubw                       m3,                m2
+
+    vextracti128                xm4,               m3,            1
+    movq                        [r2],              xm3
+    movhps                      [r2 + r3],         xm3
+    movq                        [r2 + r3 * 2],     xm4
+.end:
+    RET
+%endif
+%endmacro
+
+    IPFILTER_LUMA_PS_4xN_AVX2 4
+    IPFILTER_LUMA_PS_4xN_AVX2 8
+    IPFILTER_LUMA_PS_4xN_AVX2 16
+
+%macro IPFILTER_LUMA_PS_8xN_AVX2 1
+; TODO: verify and enable on X86 mode
+%if ARCH_X86_64 == 1
+; void filter_hps(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+INIT_YMM avx2
+cglobal interp_8tap_horiz_ps_8x%1, 4,7,6
+    mov                         r5d,        r5m
+    mov                         r4d,        r4m
+    shl                         r4d,        7
+%ifdef PIC
+    lea                         r6,         [pb_LumaCoeffVer]
+    add                         r6,         r4
+%else
+    lea                         r6,         [pb_LumaCoeffVer + r4]
+%endif
+    add                         r3d,        r3d
+    vpbroadcastd                m0,         [pw_2000]
+    sub                         r0,         3
+    lea                         r4,         [pb_8tap_hps_0]
+    vbroadcasti128              m5,         [r4 + 0 * mmsize]
+
+    ; check row count extend for interpolateHV
+    test                        r5d,        r5d;
+    mov                         r5d,        %1
+    jz                         .enter_loop
+    lea                         r4,         [r1 * 3]                        ; r8 = (N / 2 - 1) * srcStride
+    sub                         r0,         r4                              ; r0(src)-r8
+    add                         r5d,        8-1-2                           ; blkheight += N - 1  (7 - 3 = 4 ; since the last three rows not in loop)
+
+.enter_loop:
+    lea                         r4,         [pb_8tap_hps_0]
+
+    ; ***** register map *****
+    ; m0 - pw_2000
+    ; r4 - base pointer of shuffle order table
+    ; r5 - count of loop
+    ; r6 - point to LumaCoeff
+.loop:
+
+    ; Row 0-1
+    movu                        xm1,        [r0]
+    movu                        xm2,        [r0 + r1]
+    vinserti128                 m1,         m1,         xm2, 1
+    pshufb                      m2,         m1,         m5                  ; [0 1 1 2 2 3 3 4 ...]
+    pshufb                      m3,         m1,         [r4 + 1 * mmsize]   ; [2 3 3 4 4 5 5 6 ...]
+    pshufb                      m4,         m1,         [r4 + 2 * mmsize]   ; [4 5 5 6 6 7 7 8 ...]
+    pshufb                      m1,         m1,         [r4 + 3 * mmsize]   ; [6 7 7 8 8 9 9 A ...]
+    pmaddubsw                   m2,         [r6 + 0 * mmsize]
+    pmaddubsw                   m3,         [r6 + 1 * mmsize]
+    pmaddubsw                   m4,         [r6 + 2 * mmsize]
+    pmaddubsw                   m1,         [r6 + 3 * mmsize]
+    paddw                       m2,         m3
+    paddw                       m1,         m4
+    paddw                       m1,         m2
+    psubw                       m1,         m0
+
+    vextracti128                xm2,        m1,         1
+    movu                        [r2],       xm1                             ; row 0
+    movu                        [r2 + r3],  xm2                             ; row 1
+
+    lea                         r0,         [r0 + r1 * 2]                   ; first loop src ->5th row(i.e 4)
+    lea                         r2,         [r2 + r3 * 2]                   ; first loop dst ->5th row(i.e 4)
+    sub                         r5d,        2
+    jg                         .loop
+    jz                         .end
+
+    ; last row
+    movu                        xm1,        [r0]
+    pshufb                      xm2,        xm1,         xm5                ; [0 1 1 2 2 3 3 4 ...]
+    pshufb                      xm3,        xm1,         [r4 + 1 * mmsize]  ; [2 3 3 4 4 5 5 6 ...]
+    pshufb                      xm4,        xm1,         [r4 + 2 * mmsize]  ; [4 5 5 6 6 7 7 8 ...]
+    pshufb                      xm1,        xm1,         [r4 + 3 * mmsize]  ; [6 7 7 8 8 9 9 A ...]
+    pmaddubsw                   xm2,        [r6 + 0 * mmsize]
+    pmaddubsw                   xm3,        [r6 + 1 * mmsize]
+    pmaddubsw                   xm4,        [r6 + 2 * mmsize]
+    pmaddubsw                   xm1,        [r6 + 3 * mmsize]
+    paddw                       xm2,        xm3
+    paddw                       xm1,        xm4
+    paddw                       xm1,        xm2
+    psubw                       xm1,        xm0
+    movu                        [r2],       xm1                          ;row 0
+.end:
+    RET
+%endif
+%endmacro ; IPFILTER_LUMA_PS_8xN_AVX2
+
+    IPFILTER_LUMA_PS_8xN_AVX2  4
+    IPFILTER_LUMA_PS_8xN_AVX2  8
+    IPFILTER_LUMA_PS_8xN_AVX2 16
+    IPFILTER_LUMA_PS_8xN_AVX2 32
+
+
+%macro IPFILTER_LUMA_PS_16x_AVX2 2
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_horiz_ps_%1x%2, 6, 10, 7
+    mov                         r5d,               r5m
+    mov                         r4d,               r4m
+%ifdef PIC
+    lea                         r6,                [h_tab_LumaCoeff]
+    vpbroadcastq                m0,                [r6 + r4 * 8]
+%else
+    vpbroadcastq                m0,                [h_tab_LumaCoeff + r4 * 8]
+%endif
+    mova                        m6,                [h_tab_Lm + 32]
+    mova                        m1,                [h_tab_Lm]
+    mov                         r9,                %2                           ;height
+    add                         r3d,               r3d
+    vbroadcasti128              m2,                [pw_2000]
+
+    ; register map
+    ; m0      - interpolate coeff
+    ; m1 , m6 - shuffle order table
+    ; m2      - pw_2000
+
+    xor                         r7,                r7                          ; loop count variable
+    sub                         r0,                3
+    test                        r5d,               r5d
+    jz                          .label
+    lea                         r8,                [r1 * 3]                     ; r8 = (N / 2 - 1) * srcStride
+    sub                         r0,                r8                           ; r0(src)-r8
+    add                         r9,                7                            ; blkheight += N - 1  (7 - 1 = 6 ; since the last one row not in loop)
+
+.label:
+    ; Row 0
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
+    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    phaddw                      m3,                m4                           ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
+
+    vbroadcasti128              m4,                [r0 + 8]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m5,                m4,            m6            ;row 1 (col 4 to 7)
+    pshufb                      m4,                m1                           ;row 1 (col 0 to 3)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    phaddw                      m4,                m5                           ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
+    phaddw                      m3,                m4                           ; all rows and col completed.
+
+    mova                        m5,                [h_interp8_hps_shuf]
+    vpermd                      m3,                m5,               m3
+    psubw                       m3,                m2
+
+    movu                        [r2],              m3                          ;row 0
+
+    lea                         r0,                [r0 + r1]                ; first loop src ->5th row(i.e 4)
+    lea                         r2,                [r2 + r3]                ; first loop dst ->5th row(i.e 4)
+    dec                         r9d
+    jnz                         .label
+
+    RET
+%endif
+%endmacro
+
+
+    IPFILTER_LUMA_PS_16x_AVX2 16 , 16
+    IPFILTER_LUMA_PS_16x_AVX2 16 , 8
+    IPFILTER_LUMA_PS_16x_AVX2 16 , 12
+    IPFILTER_LUMA_PS_16x_AVX2 16 , 4
+    IPFILTER_LUMA_PS_16x_AVX2 16 , 32
+    IPFILTER_LUMA_PS_16x_AVX2 16 , 64
+
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_LUMA_PP_W8 2
+INIT_XMM sse4
+cglobal interp_8tap_horiz_pp_%1x%2, 4,6,7
+    mov         r4d, r4m
+
+%ifdef PIC
+    lea         r5, [h_tab_LumaCoeff]
+    movh        m3, [r5 + r4 * 8]
+%else
+    movh        m3, [h_tab_LumaCoeff + r4 * 8]
+%endif
+    pshufd      m0, m3, 0                       ; m0 = coeff-L
+    pshufd      m1, m3, 0x55                    ; m1 = coeff-H
+    lea         r5, [h_tab_Tm]                    ; r5 = shuffle
+    mova        m2, [pw_512]                    ; m2 = 512
+
+    mov         r4d, %2
+.loopH:
+%assign x 0
+%rep %1 / 8
+    movu        m3, [r0 - 3 + x]                ; m3 = [F E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb      m4, m3, [r5 + 0*16]             ; m4 = [6 5 4 3 5 4 3 2 4 3 2 1 3 2 1 0]
+    pshufb      m5, m3, [r5 + 1*16]             ; m5 = [A 9 8 7 9 8 7 6 8 7 6 5 7 6 5 4]
+    pshufb          m3, [r5 + 2*16]             ; m3 = [E D C B D C B A C B A 9 B A 9 8]
+    pmaddubsw   m4, m0
+    pmaddubsw   m6, m5, m1
+    pmaddubsw   m5, m0
+    pmaddubsw   m3, m1
+    paddw       m4, m6
+    paddw       m5, m3
+    phaddw      m4, m5
+    pmulhrsw    m4, m2
+    packuswb    m4, m4
+    movh        [r2 + x], m4
+%assign x x+8
+%endrep
+
+    add       r0, r1
+    add       r2, r3
+
+    dec       r4d
+    jnz      .loopH
+    RET
+%endmacro
+
+    IPFILTER_LUMA_PP_W8      8,  4
+    IPFILTER_LUMA_PP_W8      8,  8
+    IPFILTER_LUMA_PP_W8      8, 16
+    IPFILTER_LUMA_PP_W8      8, 32
+    IPFILTER_LUMA_PP_W8     16,  4
+    IPFILTER_LUMA_PP_W8     16,  8
+    IPFILTER_LUMA_PP_W8     16, 12
+    IPFILTER_LUMA_PP_W8     16, 16
+    IPFILTER_LUMA_PP_W8     16, 32
+    IPFILTER_LUMA_PP_W8     16, 64
+    IPFILTER_LUMA_PP_W8     24, 32
+    IPFILTER_LUMA_PP_W8     32,  8
+    IPFILTER_LUMA_PP_W8     32, 16
+    IPFILTER_LUMA_PP_W8     32, 24
+    IPFILTER_LUMA_PP_W8     32, 32
+    IPFILTER_LUMA_PP_W8     32, 64
+    IPFILTER_LUMA_PP_W8     48, 64
+    IPFILTER_LUMA_PP_W8     64, 16
+    IPFILTER_LUMA_PP_W8     64, 32
+    IPFILTER_LUMA_PP_W8     64, 48
+    IPFILTER_LUMA_PP_W8     64, 64
+
+;----------------------------------------------------------------------------------------------------------------------------
+; void interp_8tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;----------------------------------------------------------------------------------------------------------------------------
+    IPFILTER_LUMA 4, 4, ps
+    IPFILTER_LUMA 8, 8, ps
+    IPFILTER_LUMA 8, 4, ps
+    IPFILTER_LUMA 4, 8, ps
+    IPFILTER_LUMA 16, 16, ps
+    IPFILTER_LUMA 16, 8, ps
+    IPFILTER_LUMA 8, 16, ps
+    IPFILTER_LUMA 16, 12, ps
+    IPFILTER_LUMA 12, 16, ps
+    IPFILTER_LUMA 16, 4, ps
+    IPFILTER_LUMA 4, 16, ps
+    IPFILTER_LUMA 32, 32, ps
+    IPFILTER_LUMA 32, 16, ps
+    IPFILTER_LUMA 16, 32, ps
+    IPFILTER_LUMA 32, 24, ps
+    IPFILTER_LUMA 24, 32, ps
+    IPFILTER_LUMA 32, 8, ps
+    IPFILTER_LUMA 8, 32, ps
+    IPFILTER_LUMA 64, 64, ps
+    IPFILTER_LUMA 64, 32, ps
+    IPFILTER_LUMA 32, 64, ps
+    IPFILTER_LUMA 64, 48, ps
+    IPFILTER_LUMA 48, 64, ps
+    IPFILTER_LUMA 64, 16, ps
+    IPFILTER_LUMA 16, 64, ps
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_2x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+%macro FILTER_HORIZ_CHROMA_2xN 2
+INIT_XMM sse4
+cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 4, src, srcstride, dst, dststride
+%define coef2  m3
+%define Tm0    m2
+%define t1     m1
+%define t0     m0
+
+    dec        srcq
+    mov        r4d, r4m
+    add        dststrided, dststrided
+
+%ifdef PIC
+    lea        r6, [h_tab_ChromaCoeff]
+    movd       coef2, [r6 + r4 * 4]
+%else
+    movd       coef2, [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd     coef2, coef2, 0
+    mova       t1, [pw_2000]
+    mova       Tm0, [h_tab_Tm]
+
+    mov        r4d, %2
+    cmp        r5m, byte 0
+    je         .loopH
+    sub        srcq, srcstrideq
+    add        r4d, 3
+
+.loopH:
+    movh       t0, [srcq]
+    pshufb     t0, t0, Tm0
+    pmaddubsw  t0, coef2
+    phaddw     t0, t0
+    psubw      t0, t1
+    movd       [dstq], t0
+
+    lea        srcq, [srcq + srcstrideq]
+    lea        dstq, [dstq + dststrideq]
+
+    dec        r4d
+    jnz        .loopH
+
+    RET
+%endmacro
+
+    FILTER_HORIZ_CHROMA_2xN 2, 4
+    FILTER_HORIZ_CHROMA_2xN 2, 8
+
+    FILTER_HORIZ_CHROMA_2xN 2, 16
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_4x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+%macro FILTER_HORIZ_CHROMA_4xN 2
+INIT_XMM sse4
+cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 4, src, srcstride, dst, dststride
+%define coef2  m3
+%define Tm0    m2
+%define t1     m1
+%define t0     m0
+
+    dec        srcq
+    mov        r4d, r4m
+    add        dststrided, dststrided
+
+%ifdef PIC
+    lea        r6, [h_tab_ChromaCoeff]
+    movd       coef2, [r6 + r4 * 4]
+%else
+    movd       coef2, [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd     coef2, coef2, 0
+    mova       t1, [pw_2000]
+    mova       Tm0, [h_tab_Tm]
+
+    mov        r4d, %2
+    cmp        r5m, byte 0
+    je         .loopH
+    sub        srcq, srcstrideq
+    add        r4d, 3
+
+.loopH:
+    movh       t0, [srcq]
+    pshufb     t0, t0, Tm0
+    pmaddubsw  t0, coef2
+    phaddw     t0, t0
+    psubw      t0, t1
+    movlps     [dstq], t0
+
+    lea        srcq, [srcq + srcstrideq]
+    lea        dstq, [dstq + dststrideq]
+
+    dec        r4d
+    jnz        .loopH
+    RET
+%endmacro
+
+    FILTER_HORIZ_CHROMA_4xN 4, 2
+    FILTER_HORIZ_CHROMA_4xN 4, 4
+    FILTER_HORIZ_CHROMA_4xN 4, 8
+    FILTER_HORIZ_CHROMA_4xN 4, 16
+
+    FILTER_HORIZ_CHROMA_4xN 4, 32
+
+%macro PROCESS_CHROMA_W6 3
+    movu       %1, [srcq]
+    pshufb     %2, %1, Tm0
+    pmaddubsw  %2, coef2
+    pshufb     %1, %1, Tm1
+    pmaddubsw  %1, coef2
+    phaddw     %2, %1
+    psubw      %2, %3
+    movh       [dstq], %2
+    pshufd     %2, %2, 2
+    movd       [dstq + 8], %2
+%endmacro
+
+%macro PROCESS_CHROMA_W12 3
+    movu       %1, [srcq]
+    pshufb     %2, %1, Tm0
+    pmaddubsw  %2, coef2
+    pshufb     %1, %1, Tm1
+    pmaddubsw  %1, coef2
+    phaddw     %2, %1
+    psubw      %2, %3
+    movu       [dstq], %2
+    movu       %1, [srcq + 8]
+    pshufb     %1, %1, Tm0
+    pmaddubsw  %1, coef2
+    phaddw     %1, %1
+    psubw      %1, %3
+    movh       [dstq + 16], %1
+%endmacro
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_6x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+%macro FILTER_HORIZ_CHROMA 2
+INIT_XMM sse4
+cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
+%define coef2    m5
+%define Tm0      m4
+%define Tm1      m3
+%define t2       m2
+%define t1       m1
+%define t0       m0
+
+    dec     srcq
+    mov     r4d, r4m
+    add     dststrided, dststrided
+
+%ifdef PIC
+    lea     r6, [h_tab_ChromaCoeff]
+    movd    coef2, [r6 + r4 * 4]
+%else
+    movd    coef2, [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd  coef2, coef2, 0
+    mova    t2, [pw_2000]
+    mova    Tm0, [h_tab_Tm]
+    mova    Tm1, [h_tab_Tm + 16]
+
+    mov     r4d, %2
+    cmp     r5m, byte 0
+    je      .loopH
+    sub     srcq, srcstrideq
+    add     r4d, 3
+
+.loopH:
+    PROCESS_CHROMA_W%1  t0, t1, t2
+    add     srcq, srcstrideq
+    add     dstq, dststrideq
+
+    dec     r4d
+    jnz     .loopH
+
+    RET
+%endmacro
+
+    FILTER_HORIZ_CHROMA 6, 8
+    FILTER_HORIZ_CHROMA 12, 16
+
+    FILTER_HORIZ_CHROMA 6, 16
+    FILTER_HORIZ_CHROMA 12, 32
+
+%macro PROCESS_CHROMA_W8 3
+    movu        %1, [srcq]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    psubw       %2, %3
+    movu        [dstq], %2
+%endmacro
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+%macro FILTER_HORIZ_CHROMA_8xN 2
+INIT_XMM sse4
+cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
+%define coef2    m5
+%define Tm0      m4
+%define Tm1      m3
+%define t2       m2
+%define t1       m1
+%define t0       m0
+
+    dec     srcq
+    mov     r4d, r4m
+    add     dststrided, dststrided
+
+%ifdef PIC
+    lea     r6, [h_tab_ChromaCoeff]
+    movd    coef2, [r6 + r4 * 4]
+%else
+    movd    coef2, [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd  coef2, coef2, 0
+    mova    t2, [pw_2000]
+    mova    Tm0, [h_tab_Tm]
+    mova    Tm1, [h_tab_Tm + 16]
+
+    mov     r4d, %2
+    cmp     r5m, byte 0
+    je      .loopH
+    sub     srcq, srcstrideq
+    add     r4d, 3
+
+.loopH:
+    PROCESS_CHROMA_W8  t0, t1, t2
+    add     srcq, srcstrideq
+    add     dstq, dststrideq
+
+    dec     r4d
+    jnz     .loopH
+
+    RET
+%endmacro
+
+    FILTER_HORIZ_CHROMA_8xN 8, 2
+    FILTER_HORIZ_CHROMA_8xN 8, 4
+    FILTER_HORIZ_CHROMA_8xN 8, 6
+    FILTER_HORIZ_CHROMA_8xN 8, 8
+    FILTER_HORIZ_CHROMA_8xN 8, 16
+    FILTER_HORIZ_CHROMA_8xN 8, 32
+
+    FILTER_HORIZ_CHROMA_8xN 8, 12
+    FILTER_HORIZ_CHROMA_8xN 8, 64
+
+%macro PROCESS_CHROMA_W16 4
+    movu        %1, [srcq]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq + 8]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    psubw       %2, %3
+    psubw       %4, %3
+    movu        [dstq], %2
+    movu        [dstq + 16], %4
+%endmacro
+
+%macro PROCESS_CHROMA_W24 4
+    movu        %1, [srcq]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq + 8]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    psubw       %2, %3
+    psubw       %4, %3
+    movu        [dstq], %2
+    movu        [dstq + 16], %4
+    movu        %1, [srcq + 16]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    psubw       %2, %3
+    movu        [dstq + 32], %2
+%endmacro
+
+%macro PROCESS_CHROMA_W32 4
+    movu        %1, [srcq]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq + 8]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    psubw       %2, %3
+    psubw       %4, %3
+    movu        [dstq], %2
+    movu        [dstq + 16], %4
+    movu        %1, [srcq + 16]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq + 24]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    psubw       %2, %3
+    psubw       %4, %3
+    movu        [dstq + 32], %2
+    movu        [dstq + 48], %4
+%endmacro
+
+%macro PROCESS_CHROMA_W16o 5
+    movu        %1, [srcq + %5]
+    pshufb      %2, %1, Tm0
+    pmaddubsw   %2, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %2, %1
+    movu        %1, [srcq + %5 + 8]
+    pshufb      %4, %1, Tm0
+    pmaddubsw   %4, coef2
+    pshufb      %1, %1, Tm1
+    pmaddubsw   %1, coef2
+    phaddw      %4, %1
+    psubw       %2, %3
+    psubw       %4, %3
+    movu        [dstq + %5 * 2], %2
+    movu        [dstq + %5 * 2 + 16], %4
+%endmacro
+
+%macro PROCESS_CHROMA_W48 4
+    PROCESS_CHROMA_W16o %1, %2, %3, %4, 0
+    PROCESS_CHROMA_W16o %1, %2, %3, %4, 16
+    PROCESS_CHROMA_W16o %1, %2, %3, %4, 32
+%endmacro
+
+%macro PROCESS_CHROMA_W64 4
+    PROCESS_CHROMA_W16o %1, %2, %3, %4, 0
+    PROCESS_CHROMA_W16o %1, %2, %3, %4, 16
+    PROCESS_CHROMA_W16o %1, %2, %3, %4, 32
+    PROCESS_CHROMA_W16o %1, %2, %3, %4, 48
+%endmacro
+
+;------------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;------------------------------------------------------------------------------------------------------------------------------
+%macro FILTER_HORIZ_CHROMA_WxN 2
+INIT_XMM sse4
+cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 7, src, srcstride, dst, dststride
+%define coef2    m6
+%define Tm0      m5
+%define Tm1      m4
+%define t3       m3
+%define t2       m2
+%define t1       m1
+%define t0       m0
+
+    dec     srcq
+    mov     r4d, r4m
+    add     dststrided, dststrided
+
+%ifdef PIC
+    lea     r6, [h_tab_ChromaCoeff]
+    movd    coef2, [r6 + r4 * 4]
+%else
+    movd    coef2, [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufd  coef2, coef2, 0
+    mova    t2, [pw_2000]
+    mova    Tm0, [h_tab_Tm]
+    mova    Tm1, [h_tab_Tm + 16]
+
+    mov     r4d, %2
+    cmp     r5m, byte 0
+    je      .loopH
+    sub     srcq, srcstrideq
+    add     r4d, 3
+
+.loopH:
+    PROCESS_CHROMA_W%1   t0, t1, t2, t3
+    add     srcq, srcstrideq
+    add     dstq, dststrideq
+
+    dec     r4d
+    jnz     .loopH
+
+    RET
+%endmacro
+
+    FILTER_HORIZ_CHROMA_WxN 16, 4
+    FILTER_HORIZ_CHROMA_WxN 16, 8
+    FILTER_HORIZ_CHROMA_WxN 16, 12
+    FILTER_HORIZ_CHROMA_WxN 16, 16
+    FILTER_HORIZ_CHROMA_WxN 16, 32
+    FILTER_HORIZ_CHROMA_WxN 24, 32
+    FILTER_HORIZ_CHROMA_WxN 32,  8
+    FILTER_HORIZ_CHROMA_WxN 32, 16
+    FILTER_HORIZ_CHROMA_WxN 32, 24
+    FILTER_HORIZ_CHROMA_WxN 32, 32
+
+    FILTER_HORIZ_CHROMA_WxN 16, 24
+    FILTER_HORIZ_CHROMA_WxN 16, 64
+    FILTER_HORIZ_CHROMA_WxN 24, 64
+    FILTER_HORIZ_CHROMA_WxN 32, 48
+    FILTER_HORIZ_CHROMA_WxN 32, 64
+
+    FILTER_HORIZ_CHROMA_WxN 64, 64
+    FILTER_HORIZ_CHROMA_WxN 64, 32
+    FILTER_HORIZ_CHROMA_WxN 64, 48
+    FILTER_HORIZ_CHROMA_WxN 48, 64
+    FILTER_HORIZ_CHROMA_WxN 64, 16
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_32x32(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------;
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_32x32, 4,6,8
+    mov             r4d, r4m
+    add             r3d, r3d
+    dec             r0
+
+    ; check isRowExt
+    cmp             r5m, byte 0
+
+    lea             r5, [h_tab_ChromaCoeff]
+    vpbroadcastw    m0, [r5 + r4 * 4 + 0]
+    vpbroadcastw    m1, [r5 + r4 * 4 + 2]
+    mova            m7, [pw_2000]
+
+    ; register map
+    ; m0 - interpolate coeff Low
+    ; m1 - interpolate coeff High
+    ; m7 - constant pw_2000
+    mov             r4d, 32
+    je             .loop
+    sub             r0, r1
+    add             r4d, 3
+
+.loop:
+    ; Row 0
+    movu            m2, [r0]
+    movu            m3, [r0 + 1]
+    punpckhbw       m4, m2, m3
+    punpcklbw       m2, m3
+    pmaddubsw       m4, m0
+    pmaddubsw       m2, m0
+
+    movu            m3, [r0 + 2]
+    movu            m5, [r0 + 3]
+    punpckhbw       m6, m3, m5
+    punpcklbw       m3, m5
+    pmaddubsw       m6, m1
+    pmaddubsw       m3, m1
+
+    paddw           m4, m6
+    paddw           m2, m3
+    psubw           m4, m7
+    psubw           m2, m7
+    vperm2i128      m3, m2, m4, 0x20
+    vperm2i128      m5, m2, m4, 0x31
+    movu            [r2], m3
+    movu            [r2 + mmsize], m5
+
+    add             r2, r3
+    add             r0, r1
+    dec             r4d
+    jnz            .loop
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_16x16(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------;
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_16x16, 4,7,6
+    mov             r4d, r4m
+    mov             r5d, r5m
+    add             r3d, r3d
+
+%ifdef PIC
+    lea               r6,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,           [pw_1]
+    vbroadcasti128     m5,           [pw_2000]
+    mova               m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+    mov                r6d,         16
+    dec                r0
+    test                r5d,        r5d
+    je                 .loop
+    sub                r0 ,         r1
+    add                r6d ,        3
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 8]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+    vpermq            m3,           m3,          11011000b
+    movu              [r2],         m3
+
+    add                r2,          r3
+    add                r0,          r1
+    dec                r6d
+    jnz                .loop
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_16xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_PS_16xN_AVX2 2
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_%1x%2, 4,7,6
+    mov                    r4d,        r4m
+    mov                    r5d,        r5m
+    add                    r3d,        r3d
+
+%ifdef PIC
+    lea                    r6,         [h_tab_ChromaCoeff]
+    vpbroadcastd           m0,         [r6 + r4 * 4]
+%else
+    vpbroadcastd           m0,         [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128         m2,         [pw_1]
+    vbroadcasti128         m5,         [pw_2000]
+    mova                   m1,         [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+    mov                    r6d,        %2
+    dec                    r0
+    test                   r5d,        r5d
+    je                     .loop
+    sub                    r0 ,        r1
+    add                    r6d ,       3
+
+.loop:
+    ; Row 0
+    vbroadcasti128         m3,         [r0]
+    pshufb                 m3,         m1
+    pmaddubsw              m3,         m0
+    pmaddwd                m3,         m2
+    vbroadcasti128         m4,         [r0 + 8]
+    pshufb                 m4,         m1
+    pmaddubsw              m4,         m0
+    pmaddwd                m4,         m2
+
+    packssdw               m3,         m4
+    psubw                  m3,         m5
+
+    vpermq                 m3,         m3,          11011000b
+    movu                   [r2],       m3
+
+    add                    r2,         r3
+    add                    r0,         r1
+    dec                    r6d
+    jnz                    .loop
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 32
+    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 12
+    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 8
+    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 4
+    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 24
+    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 64
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_32xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_PS_32xN_AVX2 2
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_%1x%2, 4,7,6
+    mov                r4d,          r4m
+    mov                r5d,          r5m
+    add                r3d,          r3d
+
+%ifdef PIC
+    lea                r6,           [h_tab_ChromaCoeff]
+    vpbroadcastd       m0,           [r6 + r4 * 4]
+%else
+    vpbroadcastd       m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,           [pw_1]
+    vbroadcasti128     m5,           [pw_2000]
+    mova               m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+    mov                r6d,          %2
+    dec                r0
+    test               r5d,          r5d
+    je                 .loop
+    sub                r0 ,          r1
+    add                r6d ,         3
+
+.loop:
+    ; Row 0
+    vbroadcasti128     m3,           [r0]
+    pshufb             m3,           m1
+    pmaddubsw          m3,           m0
+    pmaddwd            m3,           m2
+    vbroadcasti128     m4,           [r0 + 8]
+    pshufb             m4,           m1
+    pmaddubsw          m4,           m0
+    pmaddwd            m4,           m2
+
+    packssdw           m3,           m4
+    psubw              m3,           m5
+
+    vpermq             m3,           m3,          11011000b
+    movu              [r2],          m3
+
+    vbroadcasti128     m3,           [r0 + 16]
+    pshufb             m3,           m1
+    pmaddubsw          m3,           m0
+    pmaddwd            m3,           m2
+    vbroadcasti128     m4,           [r0 + 24]
+    pshufb             m4,           m1
+    pmaddubsw          m4,           m0
+    pmaddwd            m4,           m2
+
+    packssdw           m3,           m4
+    psubw              m3,           m5
+
+    vpermq             m3,           m3,          11011000b
+    movu               [r2 + 32],    m3
+
+    add                r2,           r3
+    add                r0,           r1
+    dec                r6d
+    jnz                .loop
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 16
+    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 24
+    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 8
+    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 64
+    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 48
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_4x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_4x4, 4,7,5
+    mov             r4d, r4m
+    mov             r5d, r5m
+    add             r3d, r3d
+
+%ifdef PIC
+    lea               r6,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,           [pw_1]
+    vbroadcasti128     m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec                r0
+    test                r5d,       r5d
+    je                 .label
+    sub                r0 , r1
+
+.label:
+    ; Row 0-1
+    movu              xm3,           [r0]
+    vinserti128       m3,           m3,      [r0 + r1],     1
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 2-3
+    lea               r0,           [r0 + r1 * 2]
+    movu              xm4,           [r0]
+    vinserti128       m4,           m4,      [r0 + r1],     1
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           [pw_2000]
+    vextracti128      xm4,          m3,     1
+    movq              [r2],         xm3
+    movq              [r2+r3],      xm4
+    lea               r2,           [r2 + r3 * 2]
+    movhps            [r2],         xm3
+    movhps            [r2 + r3],    xm4
+
+    test                r5d,        r5d
+    jz                .end
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1 * 2]
+
+    ;Row 5-6
+    movu              xm3,          [r0]
+    vinserti128       m3,           m3,      [r0 + r1],     1
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 7
+    lea               r0,           [r0 + r1 * 2]
+    vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           [pw_2000]
+
+    vextracti128      xm4,          m3,     1
+    movq              [r2],         xm3
+    movq              [r2+r3],      xm4
+    lea               r2,           [r2 + r3 * 2]
+    movhps            [r2],         xm3
+.end:
+    RET
+
+cglobal interp_4tap_horiz_ps_4x2, 4,7,5
+    mov             r4d, r4m
+    mov             r5d, r5m
+    add             r3d, r3d
+
+%ifdef PIC
+    lea               r6,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,           [pw_1]
+    vbroadcasti128     m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec                r0
+    test                r5d,       r5d
+    je                 .label
+    sub                r0 , r1
+
+.label:
+    ; Row 0-1
+    movu              xm3,           [r0]
+    vinserti128       m3,           m3,      [r0 + r1],     1
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    packssdw          m3,           m3
+    psubw             m3,           [pw_2000]
+    vextracti128      xm4,          m3,     1
+    movq              [r2],         xm3
+    movq              [r2+r3],      xm4
+
+    test              r5d,          r5d
+    jz                .end
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1 * 2]
+
+    ;Row 2-3
+    movu              xm3,          [r0]
+    vinserti128       m3,           m3,      [r0 + r1],     1
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 5
+    lea               r0,           [r0 + r1 * 2]
+    vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           [pw_2000]
+
+    vextracti128      xm4,          m3,     1
+    movq              [r2],         xm3
+    movq              [r2+r3],      xm4
+    lea               r2,           [r2 + r3 * 2]
+    movhps            [r2],         xm3
+.end:
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_4xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------;
+%macro IPFILTER_CHROMA_PS_4xN_AVX2 2
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_%1x%2, 4,7,5
+    mov             r4d, r4m
+    mov             r5d, r5m
+    add             r3d, r3d
+
+%ifdef PIC
+    lea               r6,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,           [pw_1]
+    vbroadcasti128     m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    mov              r4,                %2
+    dec              r0
+    test             r5d,       r5d
+    je               .loop
+    sub              r0 ,               r1
+
+
+.loop:
+    sub              r4d,           4
+    ; Row 0-1
+    movu              xm3,          [r0]
+    vinserti128       m3,           m3,      [r0 + r1],     1
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 2-3
+    lea               r0,           [r0 + r1 * 2]
+    movu              xm4,          [r0]
+    vinserti128       m4,           m4,      [r0 + r1],     1
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           [pw_2000]
+    vextracti128      xm4,          m3,     1
+    movq              [r2],         xm3
+    movq              [r2+r3],      xm4
+    lea               r2,           [r2 + r3 * 2]
+    movhps            [r2],         xm3
+    movhps            [r2 + r3],    xm4
+
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1 * 2]
+
+    test              r4d,          r4d
+    jnz               .loop
+    test                r5d,        r5d
+    jz                .end
+
+    ;Row 5-6
+    movu              xm3,          [r0]
+    vinserti128       m3,           m3,      [r0 + r1],     1
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 7
+    lea               r0,           [r0 + r1 * 2]
+    vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           [pw_2000]
+
+    vextracti128      xm4,          m3,     1
+    movq              [r2],         xm3
+    movq              [r2+r3],      xm4
+    lea               r2,           [r2 + r3 * 2]
+    movhps            [r2],         xm3
+.end:
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PS_4xN_AVX2  4 , 8
+    IPFILTER_CHROMA_PS_4xN_AVX2  4 , 16
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_8x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------;
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_8x8, 4,7,6
+    mov             r4d, r4m
+    mov             r5d, r5m
+    add             r3d, r3d
+
+%ifdef PIC
+    lea               r6,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,           [pw_1]
+    vbroadcasti128     m5,           [pw_2000]
+    mova               m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    mov                r6d,      4
+    dec                r0
+    test                r5d,     r5d
+    je                 .loop
+    sub                r0 ,      r1
+    add                r6d ,     1
+
+.loop:
+     dec               r6d
+    ; Row 0
+    vbroadcasti128    m3,           [r0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 1
+    vbroadcasti128    m4,           [r0 + r1]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+
+    vpermq            m3,           m3,          11011000b
+    vextracti128      xm4,          m3,     1
+    movu             [r2],         xm3
+    movu             [r2 + r3],    xm4
+
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1 * 2]
+    test               r6d,          r6d
+    jnz               .loop
+    test              r5d,         r5d
+    je                .end
+
+    ;Row 11
+    vbroadcasti128    m3,           [r0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    packssdw          m3,           m3
+    psubw             m3,           m5
+    vpermq            m3,           m3,          11011000b
+    movu             [r2],         xm3
+.end:
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_4x2, 4,6,4
+    mov             r4d, r4m
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128    m1,           [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+
+    ; Row 0-1
+    movu              xm2,          [r0 - 1]
+    vinserti128       m2,           m2,      [r0 + r1 - 1],     1
+    pshufb            m2,           m1
+    pmaddubsw         m2,           m0
+    pmaddwd           m2,           [pw_1]
+
+    packssdw          m2,           m2
+    pmulhrsw          m2,           [pw_512]
+    vextracti128      xm3,          m2,     1
+    packuswb          xm2,          xm3
+
+    movd              [r2],         xm2
+    pextrd            [r2+r3],      xm2,     2
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_32xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_PP_32xN_AVX2 2
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_%1x%2, 4,6,7
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+    mova              m6,           [pw_512]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          %2
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    vbroadcasti128    m4,           [r0 + 16]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + 20]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+
+    movu              [r2],         m3
+    add               r2,           r3
+    add               r0,           r1
+    dec               r4d
+    jnz               .loop
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PP_32xN_AVX2 32, 16
+    IPFILTER_CHROMA_PP_32xN_AVX2 32, 24
+    IPFILTER_CHROMA_PP_32xN_AVX2 32, 8
+    IPFILTER_CHROMA_PP_32xN_AVX2 32, 64
+    IPFILTER_CHROMA_PP_32xN_AVX2 32, 48
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_8xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_PP_8xN_AVX2 2
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_%1x%2, 4,6,6
+    mov               r4d,    r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    movu              m1,           [h_tab_Tm]
+    vpbroadcastd      m2,           [pw_1]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    sub               r0,           1
+    mov               r4d,          %2
+
+.loop:
+    sub               r4d,          4
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 1
+    vbroadcasti128    m4,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           [pw_512]
+    lea               r0,           [r0 + r1 * 2]
+
+    ; Row 2
+    vbroadcasti128    m4,           [r0 ]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    ; Row 3
+    vbroadcasti128    m5,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           [pw_512]
+
+    packuswb          m3,           m4
+    mova              m5,           [interp_4tap_8x8_horiz_shuf]
+    vpermd            m3,           m5,     m3
+    vextracti128      xm4,          m3,     1
+    movq              [r2],         xm3
+    movhps            [r2 + r3],    xm3
+    lea               r2,           [r2 + r3 * 2]
+    movq              [r2],         xm4
+    movhps            [r2 + r3],    xm4
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1*2]
+    test              r4d,          r4d
+    jnz               .loop
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 16
+    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 32
+    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 4
+    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 64
+    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 12
+    
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_PP_4xN_AVX2 2
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_%1x%2, 4,6,6
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vpbroadcastd      m2,           [pw_1]
+    vbroadcasti128    m1,           [h_tab_Tm]
+    mov               r4d,          %2
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec                r0
+
+.loop:
+    sub               r4d,          4
+    ; Row 0-1
+    movu              xm3,          [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    vinserti128       m3,           m3,      [r0 + r1],     1
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+
+    ; Row 2-3
+    lea               r0,           [r0 + r1 * 2]
+    movu              xm4,          [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    vinserti128       m4,           m4,      [r0 + r1],     1
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    pmulhrsw          m3,           [pw_512]
+    vextracti128      xm4,          m3,                     1
+    packuswb          xm3,          xm4
+
+    movd              [r2],         xm3
+    pextrd            [r2+r3],      xm3,                    2
+    lea               r2,           [r2 + r3 * 2]
+    pextrd            [r2],         xm3,                    1
+    pextrd            [r2+r3],      xm3,                    3
+
+    lea               r0,           [r0 + r1 * 2]
+    lea               r2,           [r2 + r3 * 2]
+    test              r4d,          r4d
+    jnz               .loop
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PP_4xN_AVX2  4 , 8
+    IPFILTER_CHROMA_PP_4xN_AVX2  4 , 16
+
+%macro IPFILTER_LUMA_PS_32xN_AVX2 2
+INIT_YMM avx2
+cglobal interp_8tap_horiz_ps_%1x%2, 4, 7, 8
+    mov                         r5d,               r5m
+    mov                         r4d,               r4m
+%ifdef PIC
+    lea                         r6,                [h_tab_LumaCoeff]
+    vpbroadcastq                m0,                [r6 + r4 * 8]
+%else
+    vpbroadcastq                m0,                [h_tab_LumaCoeff + r4 * 8]
+%endif
+    mova                        m6,                [h_tab_Lm + 32]
+    mova                        m1,                [h_tab_Lm]
+    mov                         r4d,                %2                           ;height
+    add                         r3d,               r3d
+    vbroadcasti128              m2,                [pw_1]
+    mova                        m7,                [h_interp8_hps_shuf]
+
+    ; register map
+    ; m0      - interpolate coeff
+    ; m1 , m6 - shuffle order table
+    ; m2      - pw_1
+
+
+    sub                         r0,                3
+    test                        r5d,               r5d
+    jz                          .label
+    lea                         r6,                [r1 * 3]                     ; r8 = (N / 2 - 1) * srcStride
+    sub                         r0,                r6
+    add                         r4d,                7
+
+.label:
+    lea                         r6,                 [pw_2000]
+.loop:
+    ; Row 0
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
+    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+
+
+    vbroadcasti128              m4,                [r0 + 8]
+    pshufb                      m5,                m4,            m6            ;row 0 (col 12 to 15)
+    pshufb                      m4,                m1                           ;row 0 (col 8 to 11)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m2
+    pmaddwd                     m5,                m2
+    packssdw                    m4,                m5
+
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+    vpermd                      m3,                m7,               m3
+    psubw                       m3,                [r6]
+
+    movu                        [r2],              m3                          ;row 0
+
+    vbroadcasti128              m3,                [r0 + 16]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 20 to 23)
+    pshufb                      m3,                m1                           ; row 0 (col 16 to 19)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 24]
+    pshufb                      m5,                m4,            m6            ;row 0 (col 28 to 31)
+    pshufb                      m4,                m1                           ;row 0 (col 24 to 27)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m2
+    pmaddwd                     m5,                m2
+    packssdw                    m4,                m5
+
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+    vpermd                      m3,                m7,               m3
+    psubw                       m3,                [r6]
+
+    movu                        [r2 + 32],         m3                          ;row 0
+
+    add                         r0,                r1
+    add                         r2,                r3
+    dec                         r4d
+    jnz                         .loop
+    RET
+%endmacro
+
+    IPFILTER_LUMA_PS_32xN_AVX2 32 , 32
+    IPFILTER_LUMA_PS_32xN_AVX2 32 , 16
+    IPFILTER_LUMA_PS_32xN_AVX2 32 , 24
+    IPFILTER_LUMA_PS_32xN_AVX2 32 , 8
+    IPFILTER_LUMA_PS_32xN_AVX2 32 , 64
+
+INIT_YMM avx2
+cglobal interp_8tap_horiz_ps_48x64, 4, 7, 8
+    mov                         r5d,               r5m
+    mov                         r4d,               r4m
+%ifdef PIC
+    lea                         r6,                [h_tab_LumaCoeff]
+    vpbroadcastq                m0,                [r6 + r4 * 8]
+%else
+    vpbroadcastq                m0,                [h_tab_LumaCoeff + r4 * 8]
+%endif
+    mova                        m6,                [h_tab_Lm + 32]
+    mova                        m1,                [h_tab_Lm]
+    mov                         r4d,               64                           ;height
+    add                         r3d,               r3d
+    vbroadcasti128              m2,                [pw_2000]
+    mova                        m7,                [pw_1]
+
+    ; register map
+    ; m0      - interpolate coeff
+    ; m1 , m6 - shuffle order table
+    ; m2      - pw_2000
+
+    sub                         r0,                3
+    test                        r5d,               r5d
+    jz                          .label
+    lea                         r6,                [r1 * 3]                     ; r6 = (N / 2 - 1) * srcStride
+    sub                         r0,                r6                           ; r0(src)-r6
+    add                         r4d,                7                            ; blkheight += N - 1  (7 - 1 = 6 ; since the last one row not in loop)
+
+.label:
+    lea                         r6,                [h_interp8_hps_shuf]
+.loop:
+    ; Row 0
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
+    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 8]
+    pshufb                      m5,                m4,             m6            ;row 0 (col 12 to 15)
+    pshufb                      m4,                m1                           ;row 0 (col 8 to 11)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m7
+    pmaddwd                     m5,                m7
+    packssdw                    m4,                m5
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+    mova                        m5,                [r6]
+    vpermd                      m3,                m5,             m3
+    psubw                       m3,                m2
+    movu                        [r2],              m3                          ;row 0
+
+    vbroadcasti128              m3,                [r0 + 16]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 20 to 23)
+    pshufb                      m3,                m1                           ; row 0 (col 16 to 19)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 24]
+    pshufb                      m5,                m4,             m6            ;row 0 (col 28 to 31)
+    pshufb                      m4,                m1                           ;row 0 (col 24 to 27)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m7
+    pmaddwd                     m5,                m7
+    packssdw                    m4,                m5
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+    mova                        m5,                [r6]
+    vpermd                      m3,                m5,               m3
+    psubw                       m3,                m2
+    movu                        [r2 + 32],         m3                          ;row 0
+
+    vbroadcasti128              m3,                [r0 + 32]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 36 to 39)
+    pshufb                      m3,                m1                           ; row 0 (col 32 to 35)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 40]
+    pshufb                      m5,                m4,            m6            ;row 0 (col 44 to 47)
+    pshufb                      m4,                m1                           ;row 0 (col 40 to 43)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m7
+    pmaddwd                     m5,                m7
+    packssdw                    m4,                m5
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+    mova                        m5,                [r6]
+    vpermd                      m3,                m5,               m3
+    psubw                       m3,                m2
+    movu                        [r2 + 64],         m3                          ;row 0
+
+    add                         r0,                r1
+    add                         r2,                r3
+    dec                         r4d
+    jnz                         .loop
+    RET
+
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_24x32, 4,6,8
+    sub               r0,         3
+    mov               r4d,        r4m
+%ifdef PIC
+    lea               r5,         [h_tab_LumaCoeff]
+    vpbroadcastd      m0,         [r5 + r4 * 8]
+    vpbroadcastd      m1,         [r5 + r4 * 8 + 4]
+%else
+    vpbroadcastd      m0,         [h_tab_LumaCoeff + r4 * 8]
+    vpbroadcastd      m1,         [h_tab_LumaCoeff + r4 * 8 + 4]
+%endif
+    movu              m3,         [h_tab_Tm + 16]
+    vpbroadcastd      m7,         [pw_1]
+    lea               r5,         [h_tab_Tm]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+    ; m2 , m2  shuffle order table
+    ; m7 - pw_1
+
+    mov               r4d,        32
+.loop:
+    ; Row 0
+    vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,         m4,     m3
+    pshufb            m4,         [r5]
+    pmaddubsw         m4,         m0
+    pmaddubsw         m5,         m1
+    paddw             m4,         m5
+    pmaddwd           m4,         m7
+
+    vbroadcasti128    m5,         [r0 + 8]
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [r5]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
+    pmulhrsw          m4,         [pw_512]
+
+    vbroadcasti128    m2,         [r0 + 16]
+    pshufb            m5,         m2,     m3
+    pshufb            m2,         [r5]
+    pmaddubsw         m2,         m0
+    pmaddubsw         m5,         m1
+    paddw             m2,         m5
+    pmaddwd           m2,         m7
+
+    packssdw          m2,         m2
+    pmulhrsw          m2,         [pw_512]
+    packuswb          m4,         m2
+    vpermq            m4,         m4,     11011000b
+    vextracti128      xm5,        m4,     1
+    pshufd            xm4,        xm4,    11011000b
+    pshufd            xm5,        xm5,    11011000b
+
+    movu              [r2],       xm4
+    movq              [r2 + 16],  xm5
+    add               r0,         r1
+    add               r2,         r3
+    dec               r4d
+    jnz               .loop
+    RET
+
+INIT_YMM avx2
+cglobal interp_8tap_horiz_pp_12x16, 4,6,8
+    sub               r0,        3
+    mov               r4d,       r4m
+%ifdef PIC
+    lea               r5,        [h_tab_LumaCoeff]
+    vpbroadcastd      m0,        [r5 + r4 * 8]
+    vpbroadcastd      m1,        [r5 + r4 * 8 + 4]
+%else
+    vpbroadcastd      m0,         [h_tab_LumaCoeff + r4 * 8]
+    vpbroadcastd      m1,         [h_tab_LumaCoeff + r4 * 8 + 4]
+%endif
+    movu              m3,         [h_tab_Tm + 16]
+    vpbroadcastd      m7,         [pw_1]
+    lea               r5,         [h_tab_Tm]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+    ; m2 , m2  shuffle order table
+    ; m7 - pw_1
+
+    mov               r4d,        8
+.loop:
+    ; Row 0
+    vbroadcasti128    m4,         [r0]                        ;first 8 element
+    pshufb            m5,         m4,     m3
+    pshufb            m4,         [r5]
+    pmaddubsw         m4,         m0
+    pmaddubsw         m5,         m1
+    paddw             m4,         m5
+    pmaddwd           m4,         m7
+
+    vbroadcasti128    m5,         [r0 + 8]                    ; element 8 to 11
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [r5]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+
+    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
+    pmulhrsw          m4,         [pw_512]
+
+    ;Row 1
+    vbroadcasti128    m2,         [r0 + r1]
+    pshufb            m5,         m2,     m3
+    pshufb            m2,         [r5]
+    pmaddubsw         m2,         m0
+    pmaddubsw         m5,         m1
+    paddw             m2,         m5
+    pmaddwd           m2,         m7
+
+    vbroadcasti128    m5,         [r0 + r1 + 8]
+    pshufb            m6,         m5,     m3
+    pshufb            m5,         [r5]
+    pmaddubsw         m5,         m0
+    pmaddubsw         m6,         m1
+    paddw             m5,         m6
+    pmaddwd           m5,         m7
+
+    packssdw          m2,         m5
+    pmulhrsw          m2,         [pw_512]
+    packuswb          m4,         m2
+    vpermq            m4,         m4,     11011000b
+    vextracti128      xm5,        m4,     1
+    pshufd            xm4,        xm4,    11011000b
+    pshufd            xm5,        xm5,    11011000b
+
+    movq              [r2],       xm4
+    pextrd            [r2+8],     xm4,    2
+    movq              [r2 + r3],  xm5
+    pextrd            [r2+r3+8],  xm5,    2
+    lea               r0,         [r0 + r1 * 2]
+    lea               r2,         [r2 + r3 * 2]
+    dec               r4d
+    jnz              .loop
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_16xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_PP_16xN_AVX2 2
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 7
+    mov               r4d,          r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m6,           [pw_512]
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          %2/2
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    ; Row 1
+    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+
+    vextracti128      xm4,          m3,       1
+    movu              [r2],         xm3
+    movu              [r2 + r3],    xm4
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1 * 2]
+    dec               r4d
+    jnz               .loop
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 8
+    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 32
+    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 12
+    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 4
+    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 64
+    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 24
+
+%macro IPFILTER_LUMA_PS_64xN_AVX2 1
+INIT_YMM avx2
+cglobal interp_8tap_horiz_ps_64x%1, 4, 7, 8
+    mov                         r5d,               r5m
+    mov                         r4d,               r4m
+%ifdef PIC
+    lea                         r6,                [h_tab_LumaCoeff]
+    vpbroadcastq                m0,                [r6 + r4 * 8]
+%else
+    vpbroadcastq                m0,                [h_tab_LumaCoeff + r4 * 8]
+%endif
+    mova                        m6,                [h_tab_Lm + 32]
+    mova                        m1,                [h_tab_Lm]
+    mov                         r4d,               %1                           ;height
+    add                         r3d,               r3d
+    vbroadcasti128              m2,                [pw_1]
+    mova                        m7,                [h_interp8_hps_shuf]
+
+    ; register map
+    ; m0      - interpolate coeff
+    ; m1 , m6 - shuffle order table
+    ; m2      - pw_2000
+
+    sub                         r0,                3
+    test                        r5d,               r5d
+    jz                          .label
+    lea                         r6,                [r1 * 3]
+    sub                         r0,                r6                           ; r0(src)-r6
+    add                         r4d,               7                            ; blkheight += N - 1
+
+.label:
+    lea                         r6,                [pw_2000]
+.loop:
+    ; Row 0
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
+    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 8]
+    pshufb                      m5,                m4,            m6            ;row 0 (col 12 to 15)
+    pshufb                      m4,                m1                           ;row 0 (col 8 to 11)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m2
+    pmaddwd                     m5,                m2
+    packssdw                    m4,                m5
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+    vpermd                      m3,                m7,               m3
+    psubw                       m3,                [r6]
+    movu                        [r2],              m3                          ;row 0
+
+    vbroadcasti128              m3,                [r0 + 16]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 20 to 23)
+    pshufb                      m3,                m1                           ; row 0 (col 16 to 19)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 24]
+    pshufb                      m5,                m4,            m6            ;row 0 (col 28 to 31)
+    pshufb                      m4,                m1                           ;row 0 (col 24 to 27)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m2
+    pmaddwd                     m5,                m2
+    packssdw                    m4,                m5
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+    vpermd                      m3,                m7,               m3
+    psubw                       m3,                [r6]
+    movu                        [r2 + 32],         m3                          ;row 0
+
+    vbroadcasti128              m3,                [r0 + 32]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 36 to 39)
+    pshufb                      m3,                m1                           ; row 0 (col 32 to 35)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 40]
+    pshufb                      m5,                m4,            m6            ;row 0 (col 44 to 47)
+    pshufb                      m4,                m1                           ;row 0 (col 40 to 43)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m2
+    pmaddwd                     m5,                m2
+    packssdw                    m4,                m5
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+    vpermd                      m3,                m7,               m3
+    psubw                       m3,                [r6]
+    movu                        [r2 + 64],         m3                          ;row 0
+    vbroadcasti128              m3,                [r0 + 48]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 52 to 55)
+    pshufb                      m3,                m1                           ; row 0 (col 48 to 51)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 56]
+    pshufb                      m5,                m4,            m6            ;row 0 (col 60 to 63)
+    pshufb                      m4,                m1                           ;row 0 (col 56 to 59)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m2
+    pmaddwd                     m5,                m2
+    packssdw                    m4,                m5
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4
+    vpermd                      m3,                m7,               m3
+    psubw                       m3,                [r6]
+    movu                        [r2 + 96],         m3                          ;row 0
+
+    add                          r0,                r1
+    add                          r2,                r3
+    dec                          r4d
+    jnz                         .loop
+    RET
+%endmacro
+
+    IPFILTER_LUMA_PS_64xN_AVX2 64
+    IPFILTER_LUMA_PS_64xN_AVX2 48
+    IPFILTER_LUMA_PS_64xN_AVX2 32
+    IPFILTER_LUMA_PS_64xN_AVX2 16
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_8xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_PS_8xN_AVX2 1
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_8x%1, 4,7,6
+    mov                r4d,             r4m
+    mov                r5d,             r5m
+    add                r3d,             r3d
+
+%ifdef PIC
+    lea                r6,              [h_tab_ChromaCoeff]
+    vpbroadcastd       m0,              [r6 + r4 * 4]
+%else
+    vpbroadcastd       m0,              [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,              [pw_1]
+    vbroadcasti128     m5,              [pw_2000]
+    mova               m1,              [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    mov                r6d,             %1/2
+    dec                r0
+    test               r5d,             r5d
+    jz                 .loop
+    sub                r0 ,             r1
+    inc                r6d
+
+.loop:
+    ; Row 0
+    vbroadcasti128     m3,              [r0]
+    pshufb             m3,              m1
+    pmaddubsw          m3,              m0
+    pmaddwd            m3,              m2
+
+    ; Row 1
+    vbroadcasti128     m4,              [r0 + r1]
+    pshufb             m4,              m1
+    pmaddubsw          m4,              m0
+    pmaddwd            m4,              m2
+    packssdw           m3,              m4
+    psubw              m3,              m5
+    vpermq             m3,              m3,          11011000b
+    vextracti128       xm4,             m3,          1
+    movu               [r2],            xm3
+    movu               [r2 + r3],       xm4
+
+    lea                r2,              [r2 + r3 * 2]
+    lea                r0,              [r0 + r1 * 2]
+    dec                r6d
+    jnz                .loop
+    test               r5d,             r5d
+    jz                 .end
+
+    ;Row 11
+    vbroadcasti128     m3,              [r0]
+    pshufb             m3,              m1
+    pmaddubsw          m3,              m0
+    pmaddwd            m3,              m2
+    packssdw           m3,              m3
+    psubw              m3,              m5
+    vpermq             m3,              m3,          11011000b
+    movu               [r2],            xm3
+.end:
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PS_8xN_AVX2  2
+    IPFILTER_CHROMA_PS_8xN_AVX2  32
+    IPFILTER_CHROMA_PS_8xN_AVX2  16
+    IPFILTER_CHROMA_PS_8xN_AVX2  6
+    IPFILTER_CHROMA_PS_8xN_AVX2  4
+    IPFILTER_CHROMA_PS_8xN_AVX2  12
+    IPFILTER_CHROMA_PS_8xN_AVX2  64
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_2x4, 4, 7, 3
+    mov                r4d,            r4m
+    mov                r5d,            r5m
+    add                r3d,            r3d
+%ifdef PIC
+    lea                r6,             [h_tab_ChromaCoeff]
+    vpbroadcastd       m0,             [r6 + r4 * 4]
+%else
+    vpbroadcastd       m0,             [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova               xm3,            [pw_2000]
+    dec                r0
+    test               r5d,            r5d
+    jz                 .label
+    sub                r0,             r1
+
+.label:
+    lea                r6,             [r1 * 3]
+    movq               xm1,            [r0]
+    movhps             xm1,            [r0 + r1]
+    movq               xm2,            [r0 + r1 * 2]
+    movhps             xm2,            [r0 + r6]
+
+    vinserti128        m1,             m1,          xm2,          1
+    pshufb             m1,             [interp4_hpp_shuf]
+    pmaddubsw          m1,             m0
+    pmaddwd            m1,             [pw_1]
+    vextracti128       xm2,            m1,          1
+    packssdw           xm1,            xm2
+    psubw              xm1,            xm3
+
+    lea                r4,             [r3 * 3]
+    movd               [r2],           xm1
+    pextrd             [r2 + r3],      xm1,         1
+    pextrd             [r2 + r3 * 2],  xm1,         2
+    pextrd             [r2 + r4],      xm1,         3
+
+    test               r5d,            r5d
+    jz                .end
+    lea                r2,             [r2 + r3 * 4]
+    lea                r0,             [r0 + r1 * 4]
+
+    movq               xm1,            [r0]
+    movhps             xm1,            [r0 + r1]
+    movq               xm2,            [r0 + r1 * 2]
+    vinserti128        m1,             m1,          xm2,          1
+    pshufb             m1,             [interp4_hpp_shuf]
+    pmaddubsw          m1,             m0
+    pmaddwd            m1,             [pw_1]
+    vextracti128       xm2,            m1,          1
+    packssdw           xm1,            xm2
+    psubw              xm1,            xm3
+
+    movd               [r2],           xm1
+    pextrd             [r2 + r3],      xm1,         1
+    pextrd             [r2 + r3 * 2],  xm1,         2
+.end:
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_2x8, 4, 7, 7
+    mov               r4d,           r4m
+    mov               r5d,           r5m
+    add               r3d,           r3d
+
+%ifdef PIC
+    lea               r6,            [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,            [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,            [h_tab_ChromaCoeff + r4 * 4]
+%endif
+    vbroadcasti128    m6,            [pw_2000]
+    test              r5d,            r5d
+    jz                .label
+    sub               r0,             r1
+
+.label:
+    mova              m4,            [interp4_hpp_shuf]
+    mova              m5,            [pw_1]
+    dec               r0
+    lea               r4,            [r1 * 3]
+    movq              xm1,           [r0]                                   ;row 0
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m1,            m1,          xm2,          1
+    lea               r0,            [r0 + r1 * 4]
+    movq              xm3,           [r0]
+    movhps            xm3,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m3,            m3,          xm2,          1
+
+    pshufb            m1,            m4
+    pshufb            m3,            m4
+    pmaddubsw         m1,            m0
+    pmaddubsw         m3,            m0
+    pmaddwd           m1,            m5
+    pmaddwd           m3,            m5
+    packssdw          m1,            m3
+    psubw             m1,            m6
+
+    lea               r4,            [r3 * 3]
+    vextracti128      xm2,           m1,          1
+
+    movd              [r2],          xm1
+    pextrd            [r2 + r3],     xm1,         1
+    movd              [r2 + r3 * 2], xm2
+    pextrd            [r2 + r4],     xm2,         1
+    lea               r2,            [r2 + r3 * 4]
+    pextrd            [r2],          xm1,         2
+    pextrd            [r2 + r3],     xm1,         3
+    pextrd            [r2 + r3 * 2], xm2,         2
+    pextrd            [r2 + r4],     xm2,         3
+    test              r5d,            r5d
+    jz                .end
+
+    lea               r0,            [r0 + r1 * 4]
+    lea               r2,            [r2 + r3 * 4]
+    movq              xm1,           [r0]                                   ;row 0
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    vinserti128       m1,            m1,          xm2,          1
+    pshufb            m1,            m4
+    pmaddubsw         m1,            m0
+    pmaddwd           m1,            m5
+    packssdw          m1,            m1
+    psubw             m1,            m6
+    vextracti128      xm2,           m1,          1
+
+    movd              [r2],          xm1
+    pextrd            [r2 + r3],     xm1,         1
+    movd              [r2 + r3 * 2], xm2
+.end:
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_12x16, 4, 6, 7
+    mov               r4d,          r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m6,           [pw_512]
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          8
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    ; Row 1
+    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+
+    vextracti128      xm4,          m3,       1
+    movq              [r2],         xm3
+    pextrd            [r2+8],       xm3,      2
+    movq              [r2 + r3],    xm4
+    pextrd            [r2 + r3 + 8],xm4,      2
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1 * 2]
+    dec               r4d
+    jnz               .loop
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_24x32, 4,6,7
+    mov              r4d,           r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+    mova              m6,           [pw_512]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          32
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    vbroadcasti128    m4,           [r0 + 16]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + 20]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+
+    vextracti128      xm4,          m3,       1
+    movu              [r2],         xm3
+    movq              [r2 + 16],    xm4
+    add               r2,           r3
+    add               r0,           r1
+    dec               r4d
+    jnz               .loop
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_6x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------;
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_6x8, 4,7,6
+    mov                r4d,            r4m
+    mov                r5d,            r5m
+    add                r3d,            r3d
+
+%ifdef PIC
+    lea                r6,             [h_tab_ChromaCoeff]
+    vpbroadcastd       m0,             [r6 + r4 * 4]
+%else
+    vpbroadcastd       m0,             [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,             [pw_1]
+    vbroadcasti128     m5,             [pw_2000]
+    mova               m1,             [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    mov               r6d,             8/2
+    dec               r0
+    test              r5d,             r5d
+    jz                .loop
+    sub               r0 ,             r1
+    inc               r6d
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,              [r0]
+    pshufb            m3,              m1
+    pmaddubsw         m3,              m0
+    pmaddwd           m3,              m2
+
+    ; Row 1
+    vbroadcasti128    m4,              [r0 + r1]
+    pshufb            m4,              m1
+    pmaddubsw         m4,              m0
+    pmaddwd           m4,              m2
+    packssdw          m3,              m4
+    psubw             m3,              m5
+    vpermq            m3,              m3,          11011000b
+    vextracti128      xm4,             m3,          1
+    movq              [r2],            xm3
+    pextrd            [r2 + 8],        xm3,         2
+    movq              [r2 + r3],       xm4
+    pextrd            [r2 + r3 + 8],   xm4,         2
+    lea               r2,              [r2 + r3 * 2]
+    lea               r0,              [r0 + r1 * 2]
+    dec               r6d
+    jnz              .loop
+    test              r5d,             r5d
+    jz               .end
+
+    ;Row 11
+    vbroadcasti128    m3,              [r0]
+    pshufb            m3,              m1
+    pmaddubsw         m3,              m0
+    pmaddwd           m3,              m2
+    packssdw          m3,              m3
+    psubw             m3,              m5
+    vextracti128      xm4,             m3,          1
+    movq              [r2],            xm3
+    movd              [r2+8],          xm4
+.end:
+    RET
+
+INIT_YMM avx2
+cglobal interp_8tap_horiz_ps_12x16, 6, 7, 8
+    mov                         r5d,               r5m
+    mov                         r4d,               r4m
+%ifdef PIC
+    lea                         r6,                [h_tab_LumaCoeff]
+    vpbroadcastq                m0,                [r6 + r4 * 8]
+%else
+    vpbroadcastq                m0,                [h_tab_LumaCoeff + r4 * 8]
+%endif
+    mova                        m6,                [h_tab_Lm + 32]
+    mova                        m1,                [h_tab_Lm]
+    add                         r3d,               r3d
+    vbroadcasti128              m2,                [pw_2000]
+    mov                         r4d,                16
+    vbroadcasti128              m7,                [pw_1]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - pw_2000
+
+    mova                        m5,                [h_interp8_hps_shuf]
+    sub                         r0,                3
+    test                        r5d,               r5d
+    jz                          .loop
+    lea                         r6,                [r1 * 3]                     ; r6 = (N / 2 - 1) * srcStride
+    sub                         r0,                r6                           ; r0(src)-r6
+    add                         r4d,                7
+.loop:
+
+    ; Row 0
+
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m3,        m6
+    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 8]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m1
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m4,                m7
+    packssdw                    m4,                m4
+
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+
+    vpermd                      m3,                m5,               m3
+    psubw                       m3,                m2
+
+    vextracti128                xm4,               m3,               1
+    movu                        [r2],              xm3                          ;row 0
+    movq                        [r2 + 16],         xm4                          ;row 1
+
+    add                         r0,                r1
+    add                         r2,                r3
+    dec                         r4d
+    jnz                         .loop
+    RET
+
+INIT_YMM avx2
+cglobal interp_8tap_horiz_ps_24x32, 4, 7, 8
+    mov                         r5d,               r5m
+    mov                         r4d,               r4m
+%ifdef PIC
+    lea                         r6,                [h_tab_LumaCoeff]
+    vpbroadcastq                m0,                [r6 + r4 * 8]
+%else
+    vpbroadcastq                m0,                [h_tab_LumaCoeff + r4 * 8]
+%endif
+    mova                        m6,                [h_tab_Lm + 32]
+    mova                        m1,                [h_tab_Lm]
+    mov                         r4d,               32                           ;height
+    add                         r3d,               r3d
+    vbroadcasti128              m2,                [pw_2000]
+    vbroadcasti128              m7,                [pw_1]
+
+    ; register map
+    ; m0      - interpolate coeff
+    ; m1 , m6 - shuffle order table
+    ; m2      - pw_2000
+
+    sub                         r0,                3
+    test                        r5d,               r5d
+    jz                          .label
+    lea                         r6,                [r1 * 3]                     ; r6 = (N / 2 - 1) * srcStride
+    sub                         r0,                r6                           ; r0(src)-r6
+    add                         r4d,               7                            ; blkheight += N - 1  (7 - 1 = 6 ; since the last one row not in loop)
+
+.label:
+    lea                         r6,                [h_interp8_hps_shuf]
+.loop:
+    ; Row 0
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
+    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+
+    vbroadcasti128              m4,                [r0 + 8]                     ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m5,                m4,            m6            ;row 1 (col 4 to 7)
+    pshufb                      m4,                m1                           ;row 1 (col 0 to 3)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m7
+    pmaddwd                     m5,                m7
+    packssdw                    m4,                m5
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+    mova                        m5,                [r6]
+    vpermd                      m3,                m5,               m3
+    psubw                       m3,                m2
+    movu                        [r2],              m3                          ;row 0
+
+    vbroadcasti128              m3,                [r0 + 16]
+    pshufb                      m4,                m3,          m6
+    pshufb                      m3,                m1
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+    pmaddwd                     m3,                m7
+    pmaddwd                     m4,                m7
+    packssdw                    m3,                m4
+    mova                        m4,                [r6]
+    vpermd                      m3,                m4,            m3
+    psubw                       m3,                m2
+    movu                        [r2 + 32],         xm3                          ;row 0
+
+    add                         r0,                r1
+    add                         r2,                r3
+    dec                         r4d
+    jnz                         .loop
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_24x32(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_24x32, 4,7,6
+    mov                r4d,            r4m
+    mov                r5d,            r5m
+    add                r3d,            r3d
+%ifdef PIC
+    lea                r6,             [h_tab_ChromaCoeff]
+    vpbroadcastd       m0,             [r6 + r4 * 4]
+%else
+    vpbroadcastd       m0,             [h_tab_ChromaCoeff + r4 * 4]
+%endif
+    vbroadcasti128     m2,             [pw_1]
+    vbroadcasti128     m5,             [pw_2000]
+    mova               m1,             [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+    mov                r6d,            32
+    dec                r0
+    test               r5d,            r5d
+    je                 .loop
+    sub                r0 ,            r1
+    add                r6d ,           3
+
+.loop:
+    ; Row 0
+    vbroadcasti128     m3,             [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb             m3,             m1
+    pmaddubsw          m3,             m0
+    pmaddwd            m3,             m2
+    vbroadcasti128     m4,             [r0 + 8]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb             m4,             m1
+    pmaddubsw          m4,             m0
+    pmaddwd            m4,             m2
+    packssdw           m3,             m4
+    psubw              m3,             m5
+    vpermq             m3,             m3,          11011000b
+    movu               [r2],           m3
+
+    vbroadcasti128     m3,             [r0 + 16]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb             m3,             m1
+    pmaddubsw          m3,             m0
+    pmaddwd            m3,             m2
+    packssdw           m3,             m3
+    psubw              m3,             m5
+    vpermq             m3,             m3,          11011000b
+    movu               [r2 + 32],      xm3
+
+    add                r2,             r3
+    add                r0,             r1
+    dec                r6d
+    jnz                .loop
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------
+;macro FILTER_H8_W8_16N_AVX2
+;-----------------------------------------------------------------------------------------------------------------------
+%macro  FILTER_H8_W8_16N_AVX2 0
+    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
+    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
+    pmaddubsw                   m3,                m0
+    pmaddubsw                   m4,                m0
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4                         ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
+
+    vbroadcasti128              m4,                [r0 + 8]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb                      m5,                m4,            m6            ;row 1 (col 4 to 7)
+    pshufb                      m4,                m1                           ;row 1 (col 0 to 3)
+    pmaddubsw                   m4,                m0
+    pmaddubsw                   m5,                m0
+    pmaddwd                     m4,                m2
+    pmaddwd                     m5,                m2
+    packssdw                    m4,                m5                         ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
+
+    pmaddwd                     m3,                m2
+    pmaddwd                     m4,                m2
+    packssdw                    m3,                m4                         ; all rows and col completed.
+
+    mova                        m5,                [h_interp8_hps_shuf]
+    vpermd                      m3,                m5,               m3
+    psubw                       m3,                m8
+
+    vextracti128                xm4,               m3,               1
+    mova                        [r4],              xm3
+    mova                        [r4 + 16],         xm4
+    %endmacro
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_64xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_PP_64xN_AVX2 1
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_64x%1, 4,6,7
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+    mova              m6,           [pw_512]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          %1
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    vbroadcasti128    m4,           [r0 + 16]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + 20]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+    movu              [r2],         m3
+
+    vbroadcasti128    m3,           [r0 + 32]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 36]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    vbroadcasti128    m4,           [r0 + 48]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + 52]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+    movu              [r2 + 32],         m3
+
+    add               r2,           r3
+    add               r0,           r1
+    dec               r4d
+    jnz               .loop
+    RET
+%endmacro
+
+    IPFILTER_CHROMA_PP_64xN_AVX2  64
+    IPFILTER_CHROMA_PP_64xN_AVX2  32
+    IPFILTER_CHROMA_PP_64xN_AVX2  48
+    IPFILTER_CHROMA_PP_64xN_AVX2  16
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_48x64(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_48x64, 4,6,7
+    mov             r4d, r4m
+
+%ifdef PIC
+    lea               r5,            [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,            [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,            [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,            [interp4_horiz_shuf1]
+    vpbroadcastd      m2,            [pw_1]
+    mova              m6,            [pw_512]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,           64
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,            [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,            m1
+    pmaddubsw         m3,            m0
+    pmaddwd           m3,            m2
+    vbroadcasti128    m4,            [r0 + 4]
+    pshufb            m4,            m1
+    pmaddubsw         m4,            m0
+    pmaddwd           m4,            m2
+    packssdw          m3,            m4
+    pmulhrsw          m3,            m6
+
+    vbroadcasti128    m4,            [r0 + 16]
+    pshufb            m4,            m1
+    pmaddubsw         m4,            m0
+    pmaddwd           m4,            m2
+    vbroadcasti128    m5,            [r0 + 20]
+    pshufb            m5,            m1
+    pmaddubsw         m5,            m0
+    pmaddwd           m5,            m2
+    packssdw          m4,            m5
+    pmulhrsw          m4,            m6
+
+    packuswb          m3,            m4
+    vpermq            m3,            m3,      q3120
+
+    movu              [r2],          m3
+
+    vbroadcasti128    m3,            [r0 + mmsize]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,            m1
+    pmaddubsw         m3,            m0
+    pmaddwd           m3,            m2
+    vbroadcasti128    m4,            [r0 + mmsize + 4]
+    pshufb            m4,            m1
+    pmaddubsw         m4,            m0
+    pmaddwd           m4,            m2
+    packssdw          m3,            m4
+    pmulhrsw          m3,            m6
+
+    vbroadcasti128    m4,            [r0 + mmsize + 16]
+    pshufb            m4,            m1
+    pmaddubsw         m4,            m0
+    pmaddwd           m4,            m2
+    vbroadcasti128    m5,            [r0 + mmsize + 20]
+    pshufb            m5,            m1
+    pmaddubsw         m5,            m0
+    pmaddwd           m5,            m2
+    packssdw          m4,            m5
+    pmulhrsw          m4,            m6
+
+    packuswb          m3,            m4
+    vpermq            m3,            m3,      q3120
+    movu              [r2 + mmsize], xm3
+
+    add               r2,            r3
+    add               r0,            r1
+    dec               r4d
+    jnz               .loop
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_48x64(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------;
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_48x64, 4,7,6
+    mov             r4d, r4m
+    mov             r5d, r5m
+    add             r3d, r3d
+
+%ifdef PIC
+    lea               r6,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    vbroadcasti128     m2,          [pw_1]
+    vbroadcasti128     m5,          [pw_2000]
+    mova               m1,          [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+    mov               r6d,          64
+    dec               r0
+    test              r5d,          r5d
+    je                .loop
+    sub               r0 ,          r1
+    add               r6d ,         3
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                           ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 8]                       ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+    vpermq            m3,           m3,          q3120
+    movu              [r2],         m3
+
+    vbroadcasti128    m3,           [r0 + 16]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 24]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+    vpermq            m3,           m3,          q3120
+    movu              [r2 + 32],    m3
+
+    vbroadcasti128    m3,           [r0 + 32]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 40]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    psubw             m3,           m5
+    vpermq            m3,           m3,          q3120
+    movu              [r2 + 64],    m3
+
+    add               r2,          r3
+    add               r0,          r1
+    dec               r6d
+    jnz               .loop
+    RET
+
+;-----------------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_24x64(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
+;-----------------------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_24x64, 4,7,6
+    mov                r4d,            r4m
+    mov                r5d,            r5m
+    add                r3d,            r3d
+%ifdef PIC
+    lea                r6,             [h_tab_ChromaCoeff]
+    vpbroadcastd       m0,             [r6 + r4 * 4]
+%else
+    vpbroadcastd       m0,             [h_tab_ChromaCoeff + r4 * 4]
+%endif
+    vbroadcasti128     m2,             [pw_1]
+    vbroadcasti128     m5,             [pw_2000]
+    mova               m1,             [h_tab_Tm]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+    mov                r6d,            64
+    dec                r0
+    test               r5d,            r5d
+    je                 .loop
+    sub                r0 ,            r1
+    add                r6d ,           3
+
+.loop:
+    ; Row 0
+    vbroadcasti128     m3,             [r0]                          ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb             m3,             m1
+    pmaddubsw          m3,             m0
+    pmaddwd            m3,             m2
+    vbroadcasti128     m4,             [r0 + 8]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb             m4,             m1
+    pmaddubsw          m4,             m0
+    pmaddwd            m4,             m2
+    packssdw           m3,             m4
+    psubw              m3,             m5
+    vpermq             m3,             m3,          q3120
+    movu               [r2],           m3
+
+    vbroadcasti128     m3,             [r0 + 16]                     ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb             m3,             m1
+    pmaddubsw          m3,             m0
+    pmaddwd            m3,             m2
+    packssdw           m3,             m3
+    psubw              m3,             m5
+    vpermq             m3,             m3,          q3120
+    movu               [r2 + 32],      xm3
+
+    add                r2,             r3
+    add                r0,             r1
+    dec                r6d
+    jnz                .loop
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_ps_2x16, 4, 7, 7
+    mov               r4d,           r4m
+    mov               r5d,           r5m
+    add               r3d,           r3d
+
+%ifdef PIC
+    lea               r6,            [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,            [r6 + r4 * 4]
+%else
+    vpbroadcastd      m0,            [h_tab_ChromaCoeff + r4 * 4]
+%endif
+    vbroadcasti128    m6,            [pw_2000]
+    test              r5d,            r5d
+    jz                .label
+    sub               r0,             r1
+
+.label:
+    mova              m4,            [interp4_hps_shuf]
+    mova              m5,            [pw_1]
+    dec               r0
+    lea               r4,            [r1 * 3]
+    movq              xm1,           [r0]                                   ;row 0
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m1,            m1,           xm2,          1
+    lea               r0,            [r0 + r1 * 4]
+    movq              xm3,           [r0]
+    movhps            xm3,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m3,            m3,           xm2,          1
+
+    pshufb            m1,            m4
+    pshufb            m3,            m4
+    pmaddubsw         m1,            m0
+    pmaddubsw         m3,            m0
+    pmaddwd           m1,            m5
+    pmaddwd           m3,            m5
+    packssdw          m1,            m3
+    psubw             m1,            m6
+
+    lea               r4,            [r3 * 3]
+    vextracti128      xm2,           m1,           1
+
+    movd              [r2],          xm1
+    pextrd            [r2 + r3],     xm1,          1
+    movd              [r2 + r3 * 2], xm2
+    pextrd            [r2 + r4],     xm2,          1
+    lea               r2,            [r2 + r3 * 4]
+    pextrd            [r2],          xm1,          2
+    pextrd            [r2 + r3],     xm1,          3
+    pextrd            [r2 + r3 * 2], xm2,          2
+    pextrd            [r2 + r4],     xm2,          3
+
+    lea               r0,            [r0 + r1 * 4]
+    lea               r2,            [r2 + r3 * 4]
+    lea               r4,            [r1 * 3]
+    movq              xm1,           [r0]
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m1,            m1,          xm2,           1
+    lea               r0,            [r0 + r1 * 4]
+    movq              xm3,           [r0]
+    movhps            xm3,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m3,            m3,          xm2,           1
+
+    pshufb            m1,            m4
+    pshufb            m3,            m4
+    pmaddubsw         m1,            m0
+    pmaddubsw         m3,            m0
+    pmaddwd           m1,            m5
+    pmaddwd           m3,            m5
+    packssdw          m1,            m3
+    psubw             m1,            m6
+
+    lea               r4,            [r3 * 3]
+    vextracti128      xm2,           m1,           1
+
+    movd              [r2],          xm1
+    pextrd            [r2 + r3],     xm1,          1
+    movd              [r2 + r3 * 2], xm2
+    pextrd            [r2 + r4],     xm2,          1
+    lea               r2,            [r2 + r3 * 4]
+    pextrd            [r2],          xm1,          2
+    pextrd            [r2 + r3],     xm1,          3
+    pextrd            [r2 + r3 * 2], xm2,          2
+    pextrd            [r2 + r4],     xm2,          3
+
+    test              r5d,            r5d
+    jz                .end
+
+    lea               r0,            [r0 + r1 * 4]
+    lea               r2,            [r2 + r3 * 4]
+    movq              xm1,           [r0]
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    vinserti128       m1,            m1,          xm2,           1
+    pshufb            m1,            m4
+    pmaddubsw         m1,            m0
+    pmaddwd           m1,            m5
+    packssdw          m1,            m1
+    psubw             m1,            m6
+    vextracti128      xm2,           m1,           1
+
+    movd              [r2],          xm1
+    pextrd            [r2 + r3],     xm1,          1
+    movd              [r2 + r3 * 2], xm2
+.end:
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_6x16, 4, 6, 7
+    mov               r4d,               r4m
+
+%ifdef PIC
+    lea               r5,                [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,                [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,                [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,                [h_tab_Tm]
+    mova              m2,                [pw_1]
+    mova              m6,                [pw_512]
+    lea               r4,                [r1 * 3]
+    lea               r5,                [r3 * 3]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+%rep 4
+    ; Row 0
+    vbroadcasti128    m3,                [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,                m1
+    pmaddubsw         m3,                m0
+    pmaddwd           m3,                m2
+
+    ; Row 1
+    vbroadcasti128    m4,                [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,                m1
+    pmaddubsw         m4,                m0
+    pmaddwd           m4,                m2
+    packssdw          m3,                m4
+    pmulhrsw          m3,                m6
+
+    ; Row 2
+    vbroadcasti128    m4,                [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,                m1
+    pmaddubsw         m4,                m0
+    pmaddwd           m4,                m2
+
+    ; Row 3
+    vbroadcasti128    m5,                [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,                m1
+    pmaddubsw         m5,                m0
+    pmaddwd           m5,                m2
+    packssdw          m4,                m5
+    pmulhrsw          m4,                m6
+
+    packuswb          m3,                m4
+    vextracti128      xm4,               m3,          1
+    movd              [r2],              xm3
+    pextrw            [r2 + 4],          xm4,         0
+    pextrd            [r2 + r3],         xm3,         1
+    pextrw            [r2 + r3 + 4],     xm4,         2
+    pextrd            [r2 + r3 * 2],     xm3,         2
+    pextrw            [r2 + r3 * 2 + 4], xm4,         4
+    pextrd            [r2 + r5],         xm3,         3
+    pextrw            [r2 + r5 + 4],     xm4,         6
+    lea               r2,                [r2 + r3 * 4]
+    lea               r0,                [r0 + r1 * 4]
+%endrep
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_8tap_hv_pp_16x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int idxX, int idxY)
+;-----------------------------------------------------------------------------
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_8tap_hv_pp_16x16, 4, 10, 15, 0-31*32
+%define stk_buf1    rsp
+    mov                         r4d,               r4m
+    mov                         r5d,               r5m
+%ifdef PIC
+    lea                         r6,                [h_tab_LumaCoeff]
+    vpbroadcastq                m0,                [r6 + r4 * 8]
+%else
+    vpbroadcastq                m0,                [h_tab_LumaCoeff + r4 * 8]
+%endif
+
+    xor                         r6,                 r6
+    mov                         r4,                 rsp
+    mova                        m6,                [h_tab_Lm + 32]
+    mova                        m1,                [h_tab_Lm]
+    mov                         r8,                16                           ;height
+    vbroadcasti128              m8,                [pw_2000]
+    vbroadcasti128              m2,                [pw_1]
+    sub                         r0,                3
+    lea                         r7,                [r1 * 3]                     ; r7 = (N / 2 - 1) * srcStride
+    sub                         r0,                r7                           ; r0(src)-r7
+    add                         r8,                7
+
+.loopH:
+    FILTER_H8_W8_16N_AVX2
+    add                         r0,                r1
+    add                         r4,                32
+    inc                         r6
+    cmp                         r6,                16+7
+    jnz                        .loopH
+
+; vertical phase
+    xor                         r6,                r6
+    xor                         r1,                r1
+.loopV:
+
+;load necessary variables
+    mov                         r4d,               r5d          ;coeff here for vertical is r5m
+    shl                         r4d,               7
+    mov                         r1d,               16
+    add                         r1d,               r1d
+
+ ; load intermedia buffer
+    mov                         r0,                stk_buf1
+
+    ; register mapping
+    ; r0 - src
+    ; r5 - coeff
+    ; r6 - loop_i
+
+; load coeff table
+%ifdef PIC
+    lea                          r5,                [h_pw_LumaCoeffVer]
+    add                          r5,                r4
+%else
+    lea                          r5,                [h_pw_LumaCoeffVer + r4]
+%endif
+
+    lea                          r4,                [r1*3]
+    mova                         m14,               [h_pd_526336]
+    lea                          r6,                [r3 * 3]
+    mov                          r9d,               16 / 8
+
+.loopW:
+    PROCESS_LUMA_AVX2_W8_16R sp
+    add                          r2,                 8
+    add                          r0,                 16
+    dec                          r9d
+    jnz                          .loopW
+    RET
+%endif
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_12x32, 4, 6, 7
+    mov               r4d,          r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m6,           [pw_512]
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          16
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    ; Row 1
+    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+
+    vextracti128      xm4,          m3,       1
+    movq              [r2],         xm3
+    pextrd            [r2+8],       xm3,      2
+    movq              [r2 + r3],    xm4
+    pextrd            [r2 + r3 + 8],xm4,      2
+    lea               r2,           [r2 + r3 * 2]
+    lea               r0,           [r0 + r1 * 2]
+    dec               r4d
+    jnz               .loop
+    RET
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_24x64, 4,6,7
+    mov              r4d,           r4m
+
+%ifdef PIC
+    lea               r5,           [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m1,           [interp4_horiz_shuf1]
+    vpbroadcastd      m2,           [pw_1]
+    mova              m6,           [pw_512]
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    dec               r0
+    mov               r4d,          64
+
+.loop:
+    ; Row 0
+    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 4]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    packssdw          m3,           m4
+    pmulhrsw          m3,           m6
+
+    vbroadcasti128    m4,           [r0 + 16]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+    vbroadcasti128    m5,           [r0 + 20]
+    pshufb            m5,           m1
+    pmaddubsw         m5,           m0
+    pmaddwd           m5,           m2
+    packssdw          m4,           m5
+    pmulhrsw          m4,           m6
+
+    packuswb          m3,           m4
+    vpermq            m3,           m3,      11011000b
+
+    vextracti128      xm4,          m3,       1
+    movu              [r2],         xm3
+    movq              [r2 + 16],    xm4
+    add               r2,           r3
+    add               r0,           r1
+    dec               r4d
+    jnz               .loop
+    RET
+
+
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_2x16, 4, 6, 6
+    mov               r4d,           r4m
+
+%ifdef PIC
+    lea               r5,            [h_tab_ChromaCoeff]
+    vpbroadcastd      m0,            [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,            [h_tab_ChromaCoeff + r4 * 4]
+%endif
+
+    mova              m4,            [interp4_hpp_shuf]
+    mova              m5,            [pw_1]
+    dec               r0
+    lea               r4,            [r1 * 3]
+    movq              xm1,           [r0]
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m1,            m1,          xm2,          1
+    lea               r0,            [r0 + r1 * 4]
+    movq              xm3,           [r0]
+    movhps            xm3,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m3,            m3,          xm2,          1
+
+    pshufb            m1,            m4
+    pshufb            m3,            m4
+    pmaddubsw         m1,            m0
+    pmaddubsw         m3,            m0
+    pmaddwd           m1,            m5
+    pmaddwd           m3,            m5
+    packssdw          m1,            m3
+    pmulhrsw          m1,            [pw_512]
+    vextracti128      xm2,           m1,          1
+    packuswb          xm1,           xm2
+
+    lea               r4,            [r3 * 3]
+    pextrw            [r2],          xm1,         0
+    pextrw            [r2 + r3],     xm1,         1
+    pextrw            [r2 + r3 * 2], xm1,         4
+    pextrw            [r2 + r4],     xm1,         5
+    lea               r2,            [r2 + r3 * 4]
+    pextrw            [r2],          xm1,         2
+    pextrw            [r2 + r3],     xm1,         3
+    pextrw            [r2 + r3 * 2], xm1,         6
+    pextrw            [r2 + r4],     xm1,         7
+    lea               r2,            [r2 + r3 * 4]
+    lea               r0,            [r0 + r1 * 4]
+
+    lea               r4,            [r1 * 3]
+    movq              xm1,           [r0]
+    movhps            xm1,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m1,            m1,          xm2,          1
+    lea               r0,            [r0 + r1 * 4]
+    movq              xm3,           [r0]
+    movhps            xm3,           [r0 + r1]
+    movq              xm2,           [r0 + r1 * 2]
+    movhps            xm2,           [r0 + r4]
+    vinserti128       m3,            m3,          xm2,          1
+
+    pshufb            m1,            m4
+    pshufb            m3,            m4
+    pmaddubsw         m1,            m0
+    pmaddubsw         m3,            m0
+    pmaddwd           m1,            m5
+    pmaddwd           m3,            m5
+    packssdw          m1,            m3
+    pmulhrsw          m1,            [pw_512]
+    vextracti128      xm2,           m1,          1
+    packuswb          xm1,           xm2
+
+    lea               r4,            [r3 * 3]
+    pextrw            [r2],          xm1,         0
+    pextrw            [r2 + r3],     xm1,         1
+    pextrw            [r2 + r3 * 2], xm1,         4
+    pextrw            [r2 + r4],     xm1,         5
+    lea               r2,            [r2 + r3 * 4]
+    pextrw            [r2],          xm1,         2
+    pextrw            [r2 + r3],     xm1,         3
+    pextrw            [r2 + r3 * 2], xm1,         6
+    pextrw            [r2 + r4],     xm1,         7
+    RET

x265_2.7.tar.gz/source/common/x86/h4-ipfilter16.asm Added

@@ -0,0 +1,2632 @@
+;*****************************************************************************
+;* Copyright (C) 2013-2017 MulticoreWare, Inc
+;*
+;* Authors: Nabajit Deka <nabajit@multicorewareinc.com>
+;*          Murugan Vairavel <murugan@multicorewareinc.com>
+;*          Min Chen <chenm003@163.com>
+;*
+;* This program is free software; you can redistribute it and/or modify
+;* it under the terms of the GNU General Public License as published by
+;* the Free Software Foundation; either version 2 of the License, or
+;* (at your option) any later version.
+;*
+;* This program is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+;* GNU General Public License for more details.
+;*
+;* You should have received a copy of the GNU General Public License
+;* along with this program; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
+;*
+;* This program is also available under a commercial proprietary license.
+;* For more information, contact us at license @ x265.com.
+;*****************************************************************************/
+
+%include "x86inc.asm"
+%include "x86util.asm"
+
+
+%define INTERP_OFFSET_PP        pd_32
+%define INTERP_SHIFT_PP         6
+
+%if BIT_DEPTH == 10
+    %define INTERP_SHIFT_PS         2
+    %define INTERP_OFFSET_PS        pd_n32768
+    %define INTERP_SHIFT_SP         10
+    %define INTERP_OFFSET_SP        h4_pd_524800
+%elif BIT_DEPTH == 12
+    %define INTERP_SHIFT_PS         4
+    %define INTERP_OFFSET_PS        pd_n131072
+    %define INTERP_SHIFT_SP         8
+    %define INTERP_OFFSET_SP        pd_524416
+%else
+    %error Unsupport bit depth!
+%endif
+
+
+SECTION_RODATA 32
+
+tab_c_32:         times 8 dd 32
+h4_pd_524800:        times 8 dd 524800
+
+tab_Tm16:         db 0, 1, 2, 3, 4,  5,  6, 7, 2, 3, 4,  5, 6, 7, 8, 9
+
+tab_ChromaCoeff:  dw  0, 64,  0,  0
+                  dw -2, 58, 10, -2
+                  dw -4, 54, 16, -2
+                  dw -6, 46, 28, -4
+                  dw -4, 36, 36, -4
+                  dw -4, 28, 46, -6
+                  dw -2, 16, 54, -4
+                  dw -2, 10, 58, -2
+
+const h4_interp8_hpp_shuf,     db 0, 1, 2, 3, 4, 5, 6, 7, 2, 3, 4, 5, 6, 7, 8, 9
+                            db 4, 5, 6, 7, 8, 9, 10, 11, 6, 7, 8, 9, 10, 11, 12, 13
+
+SECTION .text
+cextern pd_8
+cextern pd_32
+cextern pw_pixel_max
+cextern pd_524416
+cextern pd_n32768
+cextern pd_n131072
+cextern pw_2000
+cextern idct8_shuf2
+
+%macro FILTERH_W2_4_sse3 2
+    movh        m3,     [r0 + %1]
+    movhps      m3,     [r0 + %1 + 2]
+    pmaddwd     m3,     m0
+    movh        m4,     [r0 + r1 + %1]
+    movhps      m4,     [r0 + r1 + %1 + 2]
+    pmaddwd     m4,     m0
+    pshufd      m2,     m3,     q2301
+    paddd       m3,     m2
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m3,     m3,     q3120
+    pshufd      m4,     m4,     q3120
+    punpcklqdq  m3,     m4
+    paddd       m3,     m1
+    movh        m5,     [r0 + 2 * r1 + %1]
+    movhps      m5,     [r0 + 2 * r1 + %1 + 2]
+    pmaddwd     m5,     m0
+    movh        m4,     [r0 + r4 + %1]
+    movhps      m4,     [r0 + r4 + %1 + 2]
+    pmaddwd     m4,     m0
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m5,     m5,     q3120
+    pshufd      m4,     m4,     q3120
+    punpcklqdq  m5,     m4
+    paddd       m5,     m1
+%ifidn %2, pp
+    psrad       m3,     6
+    psrad       m5,     6
+    packssdw    m3,     m5
+    CLIPW       m3,     m7,     m6
+%else
+    psrad       m3,     INTERP_SHIFT_PS
+    psrad       m5,     INTERP_SHIFT_PS
+    packssdw    m3,     m5
+%endif
+    movd        [r2 + %1], m3
+    psrldq      m3,     4
+    movd        [r2 + r3 + %1], m3
+    psrldq      m3,     4
+    movd        [r2 + r3 * 2 + %1], m3
+    psrldq      m3,     4
+    movd        [r2 + r5 + %1], m3
+%endmacro
+
+%macro FILTERH_W2_3_sse3 1
+    movh        m3,     [r0 + %1]
+    movhps      m3,     [r0 + %1 + 2]
+    pmaddwd     m3,     m0
+    movh        m4,     [r0 + r1 + %1]
+    movhps      m4,     [r0 + r1 + %1 + 2]
+    pmaddwd     m4,     m0
+    pshufd      m2,     m3,     q2301
+    paddd       m3,     m2
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m3,     m3,     q3120
+    pshufd      m4,     m4,     q3120
+    punpcklqdq  m3,     m4
+    paddd       m3,     m1
+
+    movh        m5,     [r0 + 2 * r1 + %1]
+    movhps      m5,     [r0 + 2 * r1 + %1 + 2]
+    pmaddwd     m5,     m0
+
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m5,     m5,     q3120
+    paddd       m5,     m1
+
+    psrad       m3,     INTERP_SHIFT_PS
+    psrad       m5,     INTERP_SHIFT_PS
+    packssdw    m3,     m5
+
+    movd        [r2 + %1], m3
+    psrldq      m3,     4
+    movd        [r2 + r3 + %1], m3
+    psrldq      m3,     4
+    movd        [r2 + r3 * 2 + %1], m3
+%endmacro
+
+%macro FILTERH_W4_2_sse3 2
+    movh        m3,     [r0 + %1]
+    movhps      m3,     [r0 + %1 + 2]
+    pmaddwd     m3,     m0
+    movh        m4,     [r0 + %1 + 4]
+    movhps      m4,     [r0 + %1 + 6]
+    pmaddwd     m4,     m0
+    pshufd      m2,     m3,     q2301
+    paddd       m3,     m2
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m3,     m3,     q3120
+    pshufd      m4,     m4,     q3120
+    punpcklqdq  m3,     m4
+    paddd       m3,     m1
+
+    movh        m5,     [r0 + r1 + %1]
+    movhps      m5,     [r0 + r1 + %1 + 2]
+    pmaddwd     m5,     m0
+    movh        m4,     [r0 + r1 + %1 + 4]
+    movhps      m4,     [r0 + r1 + %1 + 6]
+    pmaddwd     m4,     m0
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m5,     m5,     q3120
+    pshufd      m4,     m4,     q3120
+    punpcklqdq  m5,     m4
+    paddd       m5,     m1
+%ifidn %2, pp
+    psrad       m3,     6
+    psrad       m5,     6
+    packssdw    m3,     m5
+    CLIPW       m3,     m7,     m6
+%else
+    psrad       m3,     INTERP_SHIFT_PS
+    psrad       m5,     INTERP_SHIFT_PS
+    packssdw    m3,     m5
+%endif
+    movh        [r2 + %1], m3
+    movhps      [r2 + r3 + %1], m3
+%endmacro
+
+%macro FILTERH_W4_1_sse3 1
+    movh        m3,     [r0 + 2 * r1 + %1]
+    movhps      m3,     [r0 + 2 * r1 + %1 + 2]
+    pmaddwd     m3,     m0
+    movh        m4,     [r0 + 2 * r1 + %1 + 4]
+    movhps      m4,     [r0 + 2 * r1 + %1 + 6]
+    pmaddwd     m4,     m0
+    pshufd      m2,     m3,     q2301
+    paddd       m3,     m2
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m3,     m3,     q3120
+    pshufd      m4,     m4,     q3120
+    punpcklqdq  m3,     m4
+    paddd       m3,     m1
+
+    psrad       m3,     INTERP_SHIFT_PS
+    packssdw    m3,     m3
+    movh        [r2 + r3 * 2 + %1], m3
+%endmacro
+
+%macro FILTERH_W8_1_sse3 2
+    movh        m3,     [r0 + %1]
+    movhps      m3,     [r0 + %1 + 2]
+    pmaddwd     m3,     m0
+    movh        m4,     [r0 + %1 + 4]
+    movhps      m4,     [r0 + %1 + 6]
+    pmaddwd     m4,     m0
+    pshufd      m2,     m3,     q2301
+    paddd       m3,     m2
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m3,     m3,     q3120
+    pshufd      m4,     m4,     q3120
+    punpcklqdq  m3,     m4
+    paddd       m3,     m1
+
+    movh        m5,     [r0 + %1 + 8]
+    movhps      m5,     [r0 + %1 + 10]
+    pmaddwd     m5,     m0
+    movh        m4,     [r0 + %1 + 12]
+    movhps      m4,     [r0 + %1 + 14]
+    pmaddwd     m4,     m0
+    pshufd      m2,     m5,     q2301
+    paddd       m5,     m2
+    pshufd      m2,     m4,     q2301
+    paddd       m4,     m2
+    pshufd      m5,     m5,     q3120
+    pshufd      m4,     m4,     q3120
+    punpcklqdq  m5,     m4
+    paddd       m5,     m1
+%ifidn %2, pp
+    psrad       m3,     6
+    psrad       m5,     6
+    packssdw    m3,     m5
+    CLIPW       m3,     m7,     m6
+%else
+    psrad       m3,     INTERP_SHIFT_PS
+    psrad       m5,     INTERP_SHIFT_PS
+    packssdw    m3,     m5
+%endif
+    movdqu      [r2 + %1], m3
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_HOR_CHROMA_sse3 3
+INIT_XMM sse3
+cglobal interp_4tap_horiz_%3_%1x%2, 4, 7, 8
+    add         r3,     r3
+    add         r1,     r1
+    sub         r0,     2
+    mov         r4d,    r4m
+    add         r4d,    r4d
+
+%ifdef PIC
+    lea         r6,     [tab_ChromaCoeff]
+    movddup     m0,     [r6 + r4 * 4]
+%else
+    movddup     m0,     [tab_ChromaCoeff + r4 * 4]
+%endif
+
+%ifidn %3, ps
+    mova        m1,     [INTERP_OFFSET_PS]
+    cmp         r5m,    byte 0
+%if %1 <= 6
+    lea         r4,     [r1 * 3]
+    lea         r5,     [r3 * 3]
+%endif
+    je          .skip
+    sub         r0,     r1
+%if %1 <= 6
+%assign y 1
+%else
+%assign y 3
+%endif
+%assign z 0
+%rep y
+%assign x 0
+%rep %1/8
+    FILTERH_W8_1_sse3 x, %3
+%assign x x+16
+%endrep
+%if %1 == 4 || (%1 == 6 && z == 0) || (%1 == 12 && z == 0)
+    FILTERH_W4_2_sse3 x, %3
+    FILTERH_W4_1_sse3 x
+%assign x x+8
+%endif
+%if %1 == 2 || (%1 == 6 && z == 0)
+    FILTERH_W2_3_sse3 x
+%endif
+%if %1 <= 6
+    lea         r0,     [r0 + r4]
+    lea         r2,     [r2 + r5]
+%else
+    lea         r0,     [r0 + r1]
+    lea         r2,     [r2 + r3]
+%endif
+%assign z z+1
+%endrep
+.skip:
+%elifidn %3, pp
+    pxor        m7,     m7
+    mova        m6,     [pw_pixel_max]
+    mova        m1,     [tab_c_32]
+%if %1 == 2 || %1 == 6
+    lea         r4,     [r1 * 3]
+    lea         r5,     [r3 * 3]
+%endif
+%endif
+
+%if %1 == 2
+%assign y %2/4
+%elif %1 <= 6
+%assign y %2/2
+%else
+%assign y %2
+%endif
+%assign z 0
+%rep y
+%assign x 0
+%rep %1/8
+    FILTERH_W8_1_sse3 x, %3
+%assign x x+16
+%endrep
+%if %1 == 4 || %1 == 6 || (%1 == 12 && (z % 2) == 0)
+    FILTERH_W4_2_sse3 x, %3
+%assign x x+8
+%endif
+%if %1 == 2 || (%1 == 6 && (z % 2) == 0)
+    FILTERH_W2_4_sse3 x, %3
+%endif
+%assign z z+1
+%if z < y
+%if %1 == 2
+    lea         r0,     [r0 + 4 * r1]
+    lea         r2,     [r2 + 4 * r3]
+%elif %1 <= 6
+    lea         r0,     [r0 + 2 * r1]
+    lea         r2,     [r2 + 2 * r3]
+%else
+    lea         r0,     [r0 + r1]
+    lea         r2,     [r2 + r3]
+%endif
+%endif ;z < y
+%endrep
+
+    RET
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+
+FILTER_HOR_CHROMA_sse3 2, 4, pp
+FILTER_HOR_CHROMA_sse3 2, 8, pp
+FILTER_HOR_CHROMA_sse3 2, 16, pp
+FILTER_HOR_CHROMA_sse3 4, 2, pp
+FILTER_HOR_CHROMA_sse3 4, 4, pp
+FILTER_HOR_CHROMA_sse3 4, 8, pp
+FILTER_HOR_CHROMA_sse3 4, 16, pp
+FILTER_HOR_CHROMA_sse3 4, 32, pp
+FILTER_HOR_CHROMA_sse3 6, 8, pp
+FILTER_HOR_CHROMA_sse3 6, 16, pp
+FILTER_HOR_CHROMA_sse3 8, 2, pp
+FILTER_HOR_CHROMA_sse3 8, 4, pp
+FILTER_HOR_CHROMA_sse3 8, 6, pp
+FILTER_HOR_CHROMA_sse3 8, 8, pp
+FILTER_HOR_CHROMA_sse3 8, 12, pp
+FILTER_HOR_CHROMA_sse3 8, 16, pp
+FILTER_HOR_CHROMA_sse3 8, 32, pp
+FILTER_HOR_CHROMA_sse3 8, 64, pp
+FILTER_HOR_CHROMA_sse3 12, 16, pp
+FILTER_HOR_CHROMA_sse3 12, 32, pp
+FILTER_HOR_CHROMA_sse3 16, 4, pp
+FILTER_HOR_CHROMA_sse3 16, 8, pp
+FILTER_HOR_CHROMA_sse3 16, 12, pp
+FILTER_HOR_CHROMA_sse3 16, 16, pp
+FILTER_HOR_CHROMA_sse3 16, 24, pp
+FILTER_HOR_CHROMA_sse3 16, 32, pp
+FILTER_HOR_CHROMA_sse3 16, 64, pp
+FILTER_HOR_CHROMA_sse3 24, 32, pp
+FILTER_HOR_CHROMA_sse3 24, 64, pp
+FILTER_HOR_CHROMA_sse3 32, 8, pp
+FILTER_HOR_CHROMA_sse3 32, 16, pp
+FILTER_HOR_CHROMA_sse3 32, 24, pp
+FILTER_HOR_CHROMA_sse3 32, 32, pp
+FILTER_HOR_CHROMA_sse3 32, 48, pp
+FILTER_HOR_CHROMA_sse3 32, 64, pp
+FILTER_HOR_CHROMA_sse3 48, 64, pp
+FILTER_HOR_CHROMA_sse3 64, 16, pp
+FILTER_HOR_CHROMA_sse3 64, 32, pp
+FILTER_HOR_CHROMA_sse3 64, 48, pp
+FILTER_HOR_CHROMA_sse3 64, 64, pp
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+
+FILTER_HOR_CHROMA_sse3 2, 4, ps
+FILTER_HOR_CHROMA_sse3 2, 8, ps
+FILTER_HOR_CHROMA_sse3 2, 16, ps
+FILTER_HOR_CHROMA_sse3 4, 2, ps
+FILTER_HOR_CHROMA_sse3 4, 4, ps
+FILTER_HOR_CHROMA_sse3 4, 8, ps
+FILTER_HOR_CHROMA_sse3 4, 16, ps
+FILTER_HOR_CHROMA_sse3 4, 32, ps
+FILTER_HOR_CHROMA_sse3 6, 8, ps
+FILTER_HOR_CHROMA_sse3 6, 16, ps
+FILTER_HOR_CHROMA_sse3 8, 2, ps
+FILTER_HOR_CHROMA_sse3 8, 4, ps
+FILTER_HOR_CHROMA_sse3 8, 6, ps
+FILTER_HOR_CHROMA_sse3 8, 8, ps
+FILTER_HOR_CHROMA_sse3 8, 12, ps
+FILTER_HOR_CHROMA_sse3 8, 16, ps
+FILTER_HOR_CHROMA_sse3 8, 32, ps
+FILTER_HOR_CHROMA_sse3 8, 64, ps
+FILTER_HOR_CHROMA_sse3 12, 16, ps
+FILTER_HOR_CHROMA_sse3 12, 32, ps
+FILTER_HOR_CHROMA_sse3 16, 4, ps
+FILTER_HOR_CHROMA_sse3 16, 8, ps
+FILTER_HOR_CHROMA_sse3 16, 12, ps
+FILTER_HOR_CHROMA_sse3 16, 16, ps
+FILTER_HOR_CHROMA_sse3 16, 24, ps
+FILTER_HOR_CHROMA_sse3 16, 32, ps
+FILTER_HOR_CHROMA_sse3 16, 64, ps
+FILTER_HOR_CHROMA_sse3 24, 32, ps
+FILTER_HOR_CHROMA_sse3 24, 64, ps
+FILTER_HOR_CHROMA_sse3 32, 8, ps
+FILTER_HOR_CHROMA_sse3 32, 16, ps
+FILTER_HOR_CHROMA_sse3 32, 24, ps
+FILTER_HOR_CHROMA_sse3 32, 32, ps
+FILTER_HOR_CHROMA_sse3 32, 48, ps
+FILTER_HOR_CHROMA_sse3 32, 64, ps
+FILTER_HOR_CHROMA_sse3 48, 64, ps
+FILTER_HOR_CHROMA_sse3 64, 16, ps
+FILTER_HOR_CHROMA_sse3 64, 32, ps
+FILTER_HOR_CHROMA_sse3 64, 48, ps
+FILTER_HOR_CHROMA_sse3 64, 64, ps
+
+%macro FILTER_W2_2 1
+    movu        m3,         [r0]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + r1]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    packusdw    m3,         m3
+    CLIPW       m3,         m7,    m6
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    packssdw    m3,         m3
+%endif
+    movd        [r2],       m3
+    pextrd      [r2 + r3],  m3, 1
+%endmacro
+
+%macro FILTER_W4_2 1
+    movu        m3,         [r0]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + r1]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + r1 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m7,    m6
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2],       m3
+    movhps      [r2 + r3],  m3
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_CHROMA_H 6
+INIT_XMM sse4
+cglobal interp_4tap_horiz_%3_%1x%2, 4, %4, %5
+
+    add         r3,       r3
+    add         r1,       r1
+    sub         r0,       2
+    mov         r4d,      r4m
+    add         r4d,      r4d
+
+%ifdef PIC
+    lea         r%6,      [tab_ChromaCoeff]
+    movh        m0,       [r%6 + r4 * 4]
+%else
+    movh        m0,       [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    punpcklqdq  m0,       m0
+    mova        m2,       [tab_Tm16]
+
+%ifidn %3, ps
+    mova        m1,       [INTERP_OFFSET_PS]
+    cmp         r5m, byte 0
+    je          .skip
+    sub         r0,       r1
+    movu        m3,       [r0]
+    pshufb      m3,       m3, m2
+    pmaddwd     m3,       m0
+
+  %if %1 == 4
+    movu        m4,       [r0 + 4]
+    pshufb      m4,       m4, m2
+    pmaddwd     m4,       m0
+    phaddd      m3,       m4
+  %else
+    phaddd      m3,       m3
+  %endif
+
+    paddd       m3,       m1
+    psrad       m3,       INTERP_SHIFT_PS
+    packssdw    m3,       m3
+
+  %if %1 == 2
+    movd        [r2],     m3
+  %else
+    movh        [r2],     m3
+  %endif
+
+    add         r0,       r1
+    add         r2,       r3
+    FILTER_W%1_2 %3
+    lea         r0,       [r0 + 2 * r1]
+    lea         r2,       [r2 + 2 * r3]
+
+.skip:
+
+%else     ;%ifidn %3, ps
+    pxor        m7,       m7
+    mova        m6,       [pw_pixel_max]
+    mova        m1,       [tab_c_32]
+%endif    ;%ifidn %3, ps
+
+    FILTER_W%1_2 %3
+
+%rep (%2/2) - 1
+    lea         r0,       [r0 + 2 * r1]
+    lea         r2,       [r2 + 2 * r3]
+    FILTER_W%1_2 %3
+%endrep
+    RET
+%endmacro
+
+FILTER_CHROMA_H 2, 4, pp, 6, 8, 5
+FILTER_CHROMA_H 2, 8, pp, 6, 8, 5
+FILTER_CHROMA_H 4, 2, pp, 6, 8, 5
+FILTER_CHROMA_H 4, 4, pp, 6, 8, 5
+FILTER_CHROMA_H 4, 8, pp, 6, 8, 5
+FILTER_CHROMA_H 4, 16, pp, 6, 8, 5
+
+FILTER_CHROMA_H 2, 4, ps, 7, 5, 6
+FILTER_CHROMA_H 2, 8, ps, 7, 5, 6
+FILTER_CHROMA_H 4, 2, ps, 7, 6, 6
+FILTER_CHROMA_H 4, 4, ps, 7, 6, 6
+FILTER_CHROMA_H 4, 8, ps, 7, 6, 6
+FILTER_CHROMA_H 4, 16, ps, 7, 6, 6
+
+FILTER_CHROMA_H 2, 16, pp, 6, 8, 5
+FILTER_CHROMA_H 4, 32, pp, 6, 8, 5
+FILTER_CHROMA_H 2, 16, ps, 7, 5, 6
+FILTER_CHROMA_H 4, 32, ps, 7, 6, 6
+
+
+%macro FILTER_W6_1 1
+    movu        m3,         [r0]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m4,         [r0 + 8]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m4,         m4
+    paddd       m4,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m4,         INTERP_SHIFT_PP
+    packusdw    m3,         m4
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m4,         INTERP_SHIFT_PS
+    packssdw    m3,         m4
+%endif
+    movh        [r2],       m3
+    pextrd      [r2 + 8],   m3, 2
+%endmacro
+
+cglobal chroma_filter_pp_6x1_internal
+    FILTER_W6_1 pp
+    ret
+
+cglobal chroma_filter_ps_6x1_internal
+    FILTER_W6_1 ps
+    ret
+
+%macro FILTER_W8_1 1
+    movu        m3,         [r0]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 8]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 12]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2],       m3
+    movhps      [r2 + 8],   m3
+%endmacro
+
+cglobal chroma_filter_pp_8x1_internal
+    FILTER_W8_1 pp
+    ret
+
+cglobal chroma_filter_ps_8x1_internal
+    FILTER_W8_1 ps
+    ret
+
+%macro FILTER_W12_1 1
+    movu        m3,         [r0]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 8]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 12]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2],       m3
+    movhps      [r2 + 8],   m3
+
+    movu        m3,         [r0 + 16]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 20]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    packusdw    m3,         m3
+    CLIPW       m3,         m6, m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    packssdw    m3,         m3
+%endif
+    movh        [r2 + 16],  m3
+%endmacro
+
+cglobal chroma_filter_pp_12x1_internal
+    FILTER_W12_1 pp
+    ret
+
+cglobal chroma_filter_ps_12x1_internal
+    FILTER_W12_1 ps
+    ret
+
+%macro FILTER_W16_1 1
+    movu        m3,         [r0]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 8]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 12]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2],       m3
+    movhps      [r2 + 8],   m3
+
+    movu        m3,         [r0 + 16]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 20]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 24]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 28]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2 + 16],  m3
+    movhps      [r2 + 24],  m3
+%endmacro
+
+cglobal chroma_filter_pp_16x1_internal
+    FILTER_W16_1 pp
+    ret
+
+cglobal chroma_filter_ps_16x1_internal
+    FILTER_W16_1 ps
+    ret
+
+%macro FILTER_W24_1 1
+    movu        m3,         [r0]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 8]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 12]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2],       m3
+    movhps      [r2 + 8],   m3
+
+    movu        m3,         [r0 + 16]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 20]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 24]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 28]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2 + 16],  m3
+    movhps      [r2 + 24],  m3
+
+    movu        m3,         [r0 + 32]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 36]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 40]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 44]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2 + 32],  m3
+    movhps      [r2 + 40],  m3
+%endmacro
+
+cglobal chroma_filter_pp_24x1_internal
+    FILTER_W24_1 pp
+    ret
+
+cglobal chroma_filter_ps_24x1_internal
+    FILTER_W24_1 ps
+    ret
+
+%macro FILTER_W32_1 1
+    movu        m3,         [r0]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 8]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 12]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2],       m3
+    movhps      [r2 + 8],   m3
+
+    movu        m3,         [r0 + 16]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 20]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 24]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 28]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2 + 16],  m3
+    movhps      [r2 + 24],  m3
+
+    movu        m3,         [r0 + 32]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 36]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 40]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 44]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2 + 32],  m3
+    movhps      [r2 + 40],  m3
+
+    movu        m3,         [r0 + 48]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + 52]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + 56]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + 60]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2 + 48],  m3
+    movhps      [r2 + 56],  m3
+%endmacro
+
+cglobal chroma_filter_pp_32x1_internal
+    FILTER_W32_1 pp
+    ret
+
+cglobal chroma_filter_ps_32x1_internal
+    FILTER_W32_1 ps
+    ret
+
+%macro FILTER_W8o_1 2
+    movu        m3,         [r0 + %2]
+    pshufb      m3,         m3, m2
+    pmaddwd     m3,         m0
+    movu        m4,         [r0 + %2 + 4]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m3,         m4
+    paddd       m3,         m1
+
+    movu        m5,         [r0 + %2 + 8]
+    pshufb      m5,         m5, m2
+    pmaddwd     m5,         m0
+    movu        m4,         [r0 + %2 + 12]
+    pshufb      m4,         m4, m2
+    pmaddwd     m4,         m0
+    phaddd      m5,         m4
+    paddd       m5,         m1
+%ifidn %1, pp
+    psrad       m3,         INTERP_SHIFT_PP
+    psrad       m5,         INTERP_SHIFT_PP
+    packusdw    m3,         m5
+    CLIPW       m3,         m6,    m7
+%else
+    psrad       m3,         INTERP_SHIFT_PS
+    psrad       m5,         INTERP_SHIFT_PS
+    packssdw    m3,         m5
+%endif
+    movh        [r2 + %2],       m3
+    movhps      [r2 + %2 + 8],   m3
+%endmacro
+
+%macro FILTER_W48_1 1
+    FILTER_W8o_1 %1, 0
+    FILTER_W8o_1 %1, 16
+    FILTER_W8o_1 %1, 32
+    FILTER_W8o_1 %1, 48
+    FILTER_W8o_1 %1, 64
+    FILTER_W8o_1 %1, 80
+%endmacro
+
+cglobal chroma_filter_pp_48x1_internal
+    FILTER_W48_1 pp
+    ret
+
+cglobal chroma_filter_ps_48x1_internal
+    FILTER_W48_1 ps
+    ret
+
+%macro FILTER_W64_1 1
+    FILTER_W8o_1 %1, 0
+    FILTER_W8o_1 %1, 16
+    FILTER_W8o_1 %1, 32
+    FILTER_W8o_1 %1, 48
+    FILTER_W8o_1 %1, 64
+    FILTER_W8o_1 %1, 80
+    FILTER_W8o_1 %1, 96
+    FILTER_W8o_1 %1, 112
+%endmacro
+
+cglobal chroma_filter_pp_64x1_internal
+    FILTER_W64_1 pp
+    ret
+
+cglobal chroma_filter_ps_64x1_internal
+    FILTER_W64_1 ps
+    ret
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+
+INIT_XMM sse4
+%macro IPFILTER_CHROMA 6
+cglobal interp_4tap_horiz_%3_%1x%2, 4, %5, %6
+
+    add         r3,        r3
+    add         r1,        r1
+    sub         r0,         2
+    mov         r4d,        r4m
+    add         r4d,        r4d
+
+%ifdef PIC
+    lea         r%4,       [tab_ChromaCoeff]
+    movh        m0,       [r%4 + r4 * 4]
+%else
+    movh        m0,       [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    punpcklqdq  m0,       m0
+    mova        m2,       [tab_Tm16]
+
+%ifidn %3, ps
+    mova        m1,       [INTERP_OFFSET_PS]
+    cmp         r5m, byte 0
+    je          .skip
+    sub         r0, r1
+    call chroma_filter_%3_%1x1_internal
+    add         r0, r1
+    add         r2, r3
+    call chroma_filter_%3_%1x1_internal
+    add         r0, r1
+    add         r2, r3
+    call chroma_filter_%3_%1x1_internal
+    add         r0, r1
+    add         r2, r3
+.skip:
+%else
+    mova        m1,         [tab_c_32]
+    pxor        m6,         m6
+    mova        m7,         [pw_pixel_max]
+%endif
+
+    call chroma_filter_%3_%1x1_internal
+%rep %2 - 1
+    add         r0,       r1
+    add         r2,       r3
+    call chroma_filter_%3_%1x1_internal
+%endrep
+RET
+%endmacro
+IPFILTER_CHROMA 6, 8, pp, 5, 6, 8
+IPFILTER_CHROMA 8, 2, pp, 5, 6, 8
+IPFILTER_CHROMA 8, 4, pp, 5, 6, 8
+IPFILTER_CHROMA 8, 6, pp, 5, 6, 8
+IPFILTER_CHROMA 8, 8, pp, 5, 6, 8
+IPFILTER_CHROMA 8, 16, pp, 5, 6, 8
+IPFILTER_CHROMA 8, 32, pp, 5, 6, 8
+IPFILTER_CHROMA 12, 16, pp, 5, 6, 8
+IPFILTER_CHROMA 16, 4, pp, 5, 6, 8
+IPFILTER_CHROMA 16, 8, pp, 5, 6, 8
+IPFILTER_CHROMA 16, 12, pp, 5, 6, 8
+IPFILTER_CHROMA 16, 16, pp, 5, 6, 8
+IPFILTER_CHROMA 16, 32, pp, 5, 6, 8
+IPFILTER_CHROMA 24, 32, pp, 5, 6, 8
+IPFILTER_CHROMA 32, 8, pp, 5, 6, 8
+IPFILTER_CHROMA 32, 16, pp, 5, 6, 8
+IPFILTER_CHROMA 32, 24, pp, 5, 6, 8
+IPFILTER_CHROMA 32, 32, pp, 5, 6, 8
+
+IPFILTER_CHROMA 6, 8, ps, 6, 7, 6
+IPFILTER_CHROMA 8, 2, ps, 6, 7, 6
+IPFILTER_CHROMA 8, 4, ps, 6, 7, 6
+IPFILTER_CHROMA 8, 6, ps, 6, 7, 6
+IPFILTER_CHROMA 8, 8, ps, 6, 7, 6
+IPFILTER_CHROMA 8, 16, ps, 6, 7, 6
+IPFILTER_CHROMA 8, 32, ps, 6, 7, 6
+IPFILTER_CHROMA 12, 16, ps, 6, 7, 6
+IPFILTER_CHROMA 16, 4, ps, 6, 7, 6
+IPFILTER_CHROMA 16, 8, ps, 6, 7, 6
+IPFILTER_CHROMA 16, 12, ps, 6, 7, 6
+IPFILTER_CHROMA 16, 16, ps, 6, 7, 6
+IPFILTER_CHROMA 16, 32, ps, 6, 7, 6
+IPFILTER_CHROMA 24, 32, ps, 6, 7, 6
+IPFILTER_CHROMA 32, 8, ps, 6, 7, 6
+IPFILTER_CHROMA 32, 16, ps, 6, 7, 6
+IPFILTER_CHROMA 32, 24, ps, 6, 7, 6
+IPFILTER_CHROMA 32, 32, ps, 6, 7, 6
+
+IPFILTER_CHROMA 6, 16, pp, 5, 6, 8
+IPFILTER_CHROMA 8, 12, pp, 5, 6, 8
+IPFILTER_CHROMA 8, 64, pp, 5, 6, 8
+IPFILTER_CHROMA 12, 32, pp, 5, 6, 8
+IPFILTER_CHROMA 16, 24, pp, 5, 6, 8
+IPFILTER_CHROMA 16, 64, pp, 5, 6, 8
+IPFILTER_CHROMA 24, 64, pp, 5, 6, 8
+IPFILTER_CHROMA 32, 48, pp, 5, 6, 8
+IPFILTER_CHROMA 32, 64, pp, 5, 6, 8
+IPFILTER_CHROMA 6, 16, ps, 6, 7, 6
+IPFILTER_CHROMA 8, 12, ps, 6, 7, 6
+IPFILTER_CHROMA 8, 64, ps, 6, 7, 6
+IPFILTER_CHROMA 12, 32, ps, 6, 7, 6
+IPFILTER_CHROMA 16, 24, ps, 6, 7, 6
+IPFILTER_CHROMA 16, 64, ps, 6, 7, 6
+IPFILTER_CHROMA 24, 64, ps, 6, 7, 6
+IPFILTER_CHROMA 32, 48, ps, 6, 7, 6
+IPFILTER_CHROMA 32, 64, ps, 6, 7, 6
+
+IPFILTER_CHROMA 48, 64, pp, 5, 6, 8
+IPFILTER_CHROMA 64, 48, pp, 5, 6, 8
+IPFILTER_CHROMA 64, 64, pp, 5, 6, 8
+IPFILTER_CHROMA 64, 32, pp, 5, 6, 8
+IPFILTER_CHROMA 64, 16, pp, 5, 6, 8
+IPFILTER_CHROMA 48, 64, ps, 6, 7, 6
+IPFILTER_CHROMA 64, 48, ps, 6, 7, 6
+IPFILTER_CHROMA 64, 64, ps, 6, 7, 6
+IPFILTER_CHROMA 64, 32, ps, 6, 7, 6
+IPFILTER_CHROMA 64, 16, ps, 6, 7, 6
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+%macro IPFILTER_CHROMA_avx2_6xN 1
+cglobal interp_4tap_horiz_pp_6x%1, 5,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    mov             r4d, %1/2
+.loop:
+    vbroadcasti128  m3, [r0]
+    vbroadcasti128  m4, [r0 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, INTERP_SHIFT_PP           ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movq            [r2], xm3
+    pextrd          [r2 + 8], xm3, 2
+
+    vbroadcasti128  m3, [r0 + r1]
+    vbroadcasti128  m4, [r0 + r1 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, INTERP_SHIFT_PP           ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movq            [r2 + r3], xm3
+    pextrd          [r2 + r3 + 8], xm3, 2
+
+    lea             r0, [r0 + r1 * 2]
+    lea             r2, [r2 + r3 * 2]
+    dec             r4d
+    jnz             .loop
+    RET
+%endmacro
+IPFILTER_CHROMA_avx2_6xN 8
+IPFILTER_CHROMA_avx2_6xN 16
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_8x2, 5,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    vbroadcasti128  m3, [r0]
+    vbroadcasti128  m4, [r0 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, INTERP_SHIFT_PP          ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3,q2020
+    pshufb          xm3, xm6                     ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movu            [r2], xm3
+
+    vbroadcasti128  m3, [r0 + r1]
+    vbroadcasti128  m4, [r0 + r1 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, INTERP_SHIFT_PP           ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3,q2020
+    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movu            [r2 + r3], xm3
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_8x4, 5,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+%rep 2
+    vbroadcasti128  m3, [r0]
+    vbroadcasti128  m4, [r0 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3,q2020
+    pshufb          xm3, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movu            [r2], xm3
+
+    vbroadcasti128  m3, [r0 + r1]
+    vbroadcasti128  m4, [r0 + r1 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3,q2020
+    pshufb          xm3, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movu            [r2 + r3], xm3
+
+    lea             r0, [r0 + r1 * 2]
+    lea             r2, [r2 + r3 * 2]
+%endrep
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+%macro IPFILTER_CHROMA_avx2_8xN 1
+cglobal interp_4tap_horiz_pp_8x%1, 5,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    mov             r4d, %1/2
+.loop:
+    vbroadcasti128  m3, [r0]
+    vbroadcasti128  m4, [r0 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6                         ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movu            [r2], xm3
+
+    vbroadcasti128  m3, [r0 + r1]
+    vbroadcasti128  m4, [r0 + r1 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6                         ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movu            [r2 + r3], xm3
+
+    lea             r0, [r0 + r1 * 2]
+    lea             r2, [r2 + r3 * 2]
+    dec             r4d
+    jnz             .loop
+    RET
+%endmacro
+IPFILTER_CHROMA_avx2_8xN 6
+IPFILTER_CHROMA_avx2_8xN 8
+IPFILTER_CHROMA_avx2_8xN 12
+IPFILTER_CHROMA_avx2_8xN 16
+IPFILTER_CHROMA_avx2_8xN 32
+IPFILTER_CHROMA_avx2_8xN 64
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+%macro IPFILTER_CHROMA_avx2_16xN 1
+%if ARCH_X86_64
+cglobal interp_4tap_horiz_pp_16x%1, 5,6,9
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    mov             r4d, %1
+.loop:
+    vbroadcasti128  m3, [r0]
+    vbroadcasti128  m4, [r0 + 8]
+
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6                     ; m3 = WORD[7 6 5 4 3 2 1 0]
+
+    vbroadcasti128  m4, [r0 + 16]
+    vbroadcasti128  m8, [r0 + 24]
+
+    pshufb          m4, m1
+    pshufb          m8, m1
+
+    pmaddwd         m4, m0
+    pmaddwd         m8, m0
+    phaddd          m4, m8
+    paddd           m4, m2
+    psrad           m4, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m4, m4
+    vpermq          m4, m4, q2020
+    pshufb          xm4, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
+    vinserti128     m3, m3, xm4, 1
+    CLIPW           m3, m5, m7
+    movu            [r2], m3
+
+    add             r0, r1
+    add             r2, r3
+    dec             r4d
+    jnz             .loop
+    RET
+%endif
+%endmacro
+IPFILTER_CHROMA_avx2_16xN 4
+IPFILTER_CHROMA_avx2_16xN 8
+IPFILTER_CHROMA_avx2_16xN 12
+IPFILTER_CHROMA_avx2_16xN 16
+IPFILTER_CHROMA_avx2_16xN 24
+IPFILTER_CHROMA_avx2_16xN 32
+IPFILTER_CHROMA_avx2_16xN 64
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+%macro IPFILTER_CHROMA_avx2_32xN 1
+%if ARCH_X86_64
+cglobal interp_4tap_horiz_pp_32x%1, 5,6,9
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    mov             r6d, %1
+.loop:
+%assign x 0
+%rep 2
+    vbroadcasti128  m3, [r0 + x]
+    vbroadcasti128  m4, [r0 + 8 + x]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6                     ; m3 = WORD[7 6 5 4 3 2 1 0]
+
+    vbroadcasti128  m4, [r0 + 16 + x]
+    vbroadcasti128  m8, [r0 + 24 + x]
+    pshufb          m4, m1
+    pshufb          m8, m1
+
+    pmaddwd         m4, m0
+    pmaddwd         m8, m0
+    phaddd          m4, m8
+    paddd           m4, m2
+    psrad           m4, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m4, m4
+    vpermq          m4, m4, q2020
+    pshufb          xm4, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
+    vinserti128     m3, m3, xm4, 1
+    CLIPW           m3, m5, m7
+    movu            [r2 + x], m3
+    %assign x x+32
+    %endrep
+
+    add             r0, r1
+    add             r2, r3
+    dec             r6d
+    jnz             .loop
+    RET
+%endif
+%endmacro
+IPFILTER_CHROMA_avx2_32xN 8
+IPFILTER_CHROMA_avx2_32xN 16
+IPFILTER_CHROMA_avx2_32xN 24
+IPFILTER_CHROMA_avx2_32xN 32
+IPFILTER_CHROMA_avx2_32xN 48
+IPFILTER_CHROMA_avx2_32xN 64
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+%macro IPFILTER_CHROMA_avx2_12xN 1
+%if ARCH_X86_64
+cglobal interp_4tap_horiz_pp_12x%1, 5,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    mov             r4d, %1
+.loop:
+    vbroadcasti128  m3, [r0]
+    vbroadcasti128  m4, [r0 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6                     ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movu            [r2], xm3
+
+    vbroadcasti128  m3, [r0 + 16]
+    vbroadcasti128  m4, [r0 + 24]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
+    CLIPW           xm3, xm5, xm7
+    movq            [r2 + 16], xm3
+
+    add             r0, r1
+    add             r2, r3
+    dec             r4d
+    jnz             .loop
+    RET
+%endif
+%endmacro
+IPFILTER_CHROMA_avx2_12xN 16
+IPFILTER_CHROMA_avx2_12xN 32
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+%macro IPFILTER_CHROMA_avx2_24xN 1
+%if ARCH_X86_64
+cglobal interp_4tap_horiz_pp_24x%1, 5,6,9
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    mov             r4d, %1
+.loop:
+    vbroadcasti128  m3, [r0]
+    vbroadcasti128  m4, [r0 + 8]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6
+
+    vbroadcasti128  m4, [r0 + 16]
+    vbroadcasti128  m8, [r0 + 24]
+    pshufb          m4, m1
+    pshufb          m8, m1
+
+    pmaddwd         m4, m0
+    pmaddwd         m8, m0
+    phaddd          m4, m8
+    paddd           m4, m2
+    psrad           m4, 6
+
+    packusdw        m3, m4
+    vpermq          m3, m3, q3120
+    pshufb          m3, m6
+    CLIPW           m3, m5, m7
+    movu            [r2], m3
+
+    vbroadcasti128  m3, [r0 + 32]
+    vbroadcasti128  m4, [r0 + 40]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6
+
+    packusdw        m3, m3
+    vpermq          m3, m3, q2020
+    pshufb          xm3, xm6
+    CLIPW           xm3, xm5, xm7
+    movu            [r2 + 32], xm3
+
+    add             r0, r1
+    add             r2, r3
+    dec             r4d
+    jnz             .loop
+    RET
+%endif
+%endmacro
+IPFILTER_CHROMA_avx2_24xN 32
+IPFILTER_CHROMA_avx2_24xN 64
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+%macro IPFILTER_CHROMA_avx2_64xN 1
+%if ARCH_X86_64
+cglobal interp_4tap_horiz_pp_64x%1, 5,6,9
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    mov             r6d, %1
+.loop:
+%assign x 0
+%rep 4
+    vbroadcasti128  m3, [r0 + x]
+    vbroadcasti128  m4, [r0 + 8 + x]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6
+
+    vbroadcasti128  m4, [r0 + 16 + x]
+    vbroadcasti128  m8, [r0 + 24 + x]
+    pshufb          m4, m1
+    pshufb          m8, m1
+
+    pmaddwd         m4, m0
+    pmaddwd         m8, m0
+    phaddd          m4, m8
+    paddd           m4, m2
+    psrad           m4, 6
+
+    packusdw        m3, m4
+    vpermq          m3, m3, q3120
+    pshufb          m3, m6
+    CLIPW           m3, m5, m7
+    movu            [r2 + x], m3
+    %assign x x+32
+    %endrep
+
+    add             r0, r1
+    add             r2, r3
+    dec             r6d
+    jnz             .loop
+    RET
+%endif
+%endmacro
+IPFILTER_CHROMA_avx2_64xN 16
+IPFILTER_CHROMA_avx2_64xN 32
+IPFILTER_CHROMA_avx2_64xN 48
+IPFILTER_CHROMA_avx2_64xN 64
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
+;-------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+%if ARCH_X86_64
+cglobal interp_4tap_horiz_pp_48x64, 5,6,9
+    add             r1d, r1d
+    add             r3d, r3d
+    sub             r0, 2
+    mov             r4d, r4m
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff]
+    vpbroadcastq    m0, [r5 + r4 * 8]
+%else
+    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova            m1, [h4_interp8_hpp_shuf]
+    vpbroadcastd    m2, [pd_32]
+    pxor            m5, m5
+    mova            m6, [idct8_shuf2]
+    mova            m7, [pw_pixel_max]
+
+    mov             r4d, 64
+.loop:
+%assign x 0
+%rep 3
+    vbroadcasti128  m3, [r0 + x]
+    vbroadcasti128  m4, [r0 + 8 + x]
+    pshufb          m3, m1
+    pshufb          m4, m1
+
+    pmaddwd         m3, m0
+    pmaddwd         m4, m0
+    phaddd          m3, m4
+    paddd           m3, m2
+    psrad           m3, 6
+
+    vbroadcasti128  m4, [r0 + 16 + x]
+    vbroadcasti128  m8, [r0 + 24 + x]
+    pshufb          m4, m1
+    pshufb          m8, m1
+
+    pmaddwd         m4, m0
+    pmaddwd         m8, m0
+    phaddd          m4, m8
+    paddd           m4, m2
+    psrad           m4, 6
+
+    packusdw        m3, m4
+    vpermq          m3, m3, q3120
+    pshufb          m3, m6
+    CLIPW           m3, m5, m7
+    movu            [r2 + x], m3
+%assign x x+32
+%endrep
+
+    add             r0, r1
+    add             r2, r3
+    dec             r4d
+    jnz             .loop
+    RET
+%endif
+
+%macro IPFILTER_CHROMA_PS_8xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_horiz_ps_8x%1, 4, 7, 6
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+
+%ifdef PIC
+    lea                 r6, [tab_ChromaCoeff]
+    vpbroadcastq        m0, [r6 + r4 * 8]
+%else
+    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova                m3, [h4_interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 2
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    sub                 r0, r1
+    add                 r4d, 3
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+    IPFILTER_CHROMA_PS_8xN_AVX2 4
+    IPFILTER_CHROMA_PS_8xN_AVX2 8
+    IPFILTER_CHROMA_PS_8xN_AVX2 16
+    IPFILTER_CHROMA_PS_8xN_AVX2 32
+    IPFILTER_CHROMA_PS_8xN_AVX2 6
+    IPFILTER_CHROMA_PS_8xN_AVX2 2
+    IPFILTER_CHROMA_PS_8xN_AVX2 12
+    IPFILTER_CHROMA_PS_8xN_AVX2 64
+
+%macro IPFILTER_CHROMA_PS_16xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_horiz_ps_16x%1, 4, 7, 6
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+
+%ifdef PIC
+    lea                 r6, [tab_ChromaCoeff]
+    vpbroadcastq        m0, [r6 + r4 * 8]
+%else
+    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova                m3, [h4_interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 2
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    sub                 r0, r1
+    add                 r4d, 3
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+
+    vbroadcasti128      m4, [r0 + 16]
+    vbroadcasti128      m5, [r0 + 24]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 16], xm4
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+IPFILTER_CHROMA_PS_16xN_AVX2 16
+IPFILTER_CHROMA_PS_16xN_AVX2 8
+IPFILTER_CHROMA_PS_16xN_AVX2 32
+IPFILTER_CHROMA_PS_16xN_AVX2 12
+IPFILTER_CHROMA_PS_16xN_AVX2 4
+IPFILTER_CHROMA_PS_16xN_AVX2 64
+IPFILTER_CHROMA_PS_16xN_AVX2 24
+
+%macro IPFILTER_CHROMA_PS_24xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_horiz_ps_24x%1, 4, 7, 6
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+
+%ifdef PIC
+    lea                 r6, [tab_ChromaCoeff]
+    vpbroadcastq        m0, [r6 + r4 * 8]
+%else
+    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova                m3, [h4_interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 2
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    sub                 r0, r1
+    add                 r4d, 3
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+
+    vbroadcasti128      m4, [r0 + 16]
+    vbroadcasti128      m5, [r0 + 24]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 16], xm4
+
+    vbroadcasti128      m4, [r0 + 32]
+    vbroadcasti128      m5, [r0 + 40]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 32], xm4
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+IPFILTER_CHROMA_PS_24xN_AVX2 32
+IPFILTER_CHROMA_PS_24xN_AVX2 64
+
+%macro IPFILTER_CHROMA_PS_12xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_horiz_ps_12x%1, 4, 7, 6
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+
+%ifdef PIC
+    lea                 r6, [tab_ChromaCoeff]
+    vpbroadcastq        m0, [r6 + r4 * 8]
+%else
+    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova                m3, [h4_interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 2
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    sub                 r0, r1
+    add                 r4d, 3
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+
+    vbroadcasti128      m4, [r0 + 16]
+    pshufb              m4, m3
+    pmaddwd             m4, m0
+    phaddd              m4, m4
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movq                [r2 + 16], xm4
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+IPFILTER_CHROMA_PS_12xN_AVX2 16
+IPFILTER_CHROMA_PS_12xN_AVX2 32
+
+%macro IPFILTER_CHROMA_PS_32xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_horiz_ps_32x%1, 4, 7, 6
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+
+%ifdef PIC
+    lea                 r6, [tab_ChromaCoeff]
+    vpbroadcastq        m0, [r6 + r4 * 8]
+%else
+    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova                m3, [h4_interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 2
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    sub                 r0, r1
+    add                 r4d, 3
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+
+    vbroadcasti128      m4, [r0 + 16]
+    vbroadcasti128      m5, [r0 + 24]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 16], xm4
+
+    vbroadcasti128      m4, [r0 + 32]
+    vbroadcasti128      m5, [r0 + 40]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 32], xm4
+
+    vbroadcasti128      m4, [r0 + 48]
+    vbroadcasti128      m5, [r0 + 56]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 48], xm4
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+IPFILTER_CHROMA_PS_32xN_AVX2 32
+IPFILTER_CHROMA_PS_32xN_AVX2 16
+IPFILTER_CHROMA_PS_32xN_AVX2 24
+IPFILTER_CHROMA_PS_32xN_AVX2 8
+IPFILTER_CHROMA_PS_32xN_AVX2 64
+IPFILTER_CHROMA_PS_32xN_AVX2 48
+
+
+%macro IPFILTER_CHROMA_PS_64xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_horiz_ps_64x%1, 4, 7, 6
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+
+%ifdef PIC
+    lea                 r6, [tab_ChromaCoeff]
+    vpbroadcastq        m0, [r6 + r4 * 8]
+%else
+    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova                m3, [h4_interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 2
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    sub                 r0, r1
+    add                 r4d, 3
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+
+    vbroadcasti128      m4, [r0 + 16]
+    vbroadcasti128      m5, [r0 + 24]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 16], xm4
+
+    vbroadcasti128      m4, [r0 + 32]
+    vbroadcasti128      m5, [r0 + 40]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 32], xm4
+
+    vbroadcasti128      m4, [r0 + 48]
+    vbroadcasti128      m5, [r0 + 56]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 48], xm4
+
+    vbroadcasti128      m4, [r0 + 64]
+    vbroadcasti128      m5, [r0 + 72]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 64], xm4
+
+    vbroadcasti128      m4, [r0 + 80]
+    vbroadcasti128      m5, [r0 + 88]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 80], xm4
+
+    vbroadcasti128      m4, [r0 + 96]
+    vbroadcasti128      m5, [r0 + 104]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 96], xm4
+
+    vbroadcasti128      m4, [r0 + 112]
+    vbroadcasti128      m5, [r0 + 120]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 112], xm4
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+IPFILTER_CHROMA_PS_64xN_AVX2 64
+IPFILTER_CHROMA_PS_64xN_AVX2 48
+IPFILTER_CHROMA_PS_64xN_AVX2 32
+IPFILTER_CHROMA_PS_64xN_AVX2 16
+
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_horiz_ps_48x64, 4, 7, 6
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+
+%ifdef PIC
+    lea                 r6, [tab_ChromaCoeff]
+    vpbroadcastq        m0, [r6 + r4 * 8]
+%else
+    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova                m3, [h4_interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 2
+    test                r5d, r5d
+    mov                 r4d, 64
+    jz                  .loop0
+    sub                 r0, r1
+    add                 r4d, 3
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2], xm4
+
+    vbroadcasti128      m4, [r0 + 16]
+    vbroadcasti128      m5, [r0 + 24]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 16], xm4
+
+    vbroadcasti128      m4, [r0 + 32]
+    vbroadcasti128      m5, [r0 + 40]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 32], xm4
+
+    vbroadcasti128      m4, [r0 + 48]
+    vbroadcasti128      m5, [r0 + 56]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 48], xm4
+
+    vbroadcasti128      m4, [r0 + 64]
+    vbroadcasti128      m5, [r0 + 72]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 64], xm4
+
+    vbroadcasti128      m4, [r0 + 80]
+    vbroadcasti128      m5, [r0 + 88]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movu                [r2 + 80], xm4
+
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+
+%macro IPFILTER_CHROMA_PS_6xN_AVX2 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_horiz_ps_6x%1, 4, 7, 6
+    add                 r1d, r1d
+    add                 r3d, r3d
+    mov                 r4d, r4m
+    mov                 r5d, r5m
+
+%ifdef PIC
+    lea                 r6, [tab_ChromaCoeff]
+    vpbroadcastq        m0, [r6 + r4 * 8]
+%else
+    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
+%endif
+    mova                m3, [h4_interp8_hpp_shuf]
+    vbroadcasti128      m2, [INTERP_OFFSET_PS]
+
+    ; register map
+    ; m0 , m1 interpolate coeff
+
+    sub                 r0, 2
+    test                r5d, r5d
+    mov                 r4d, %1
+    jz                  .loop0
+    sub                 r0, r1
+    add                 r4d, 3
+
+.loop0:
+    vbroadcasti128      m4, [r0]
+    vbroadcasti128      m5, [r0 + 8]
+    pshufb              m4, m3
+    pshufb              m5, m3
+    pmaddwd             m4, m0
+    pmaddwd             m5, m0
+    phaddd              m4, m5
+    paddd               m4, m2
+    vpermq              m4, m4, q3120
+    psrad               m4, INTERP_SHIFT_PS
+    vextracti128        xm5, m4, 1
+    packssdw            xm4, xm5
+    movq                [r2], xm4
+    pextrd              [r2 + 8], xm4, 2
+    add                 r2, r3
+    add                 r0, r1
+    dec                 r4d
+    jnz                 .loop0
+    RET
+%endif
+%endmacro
+
+    IPFILTER_CHROMA_PS_6xN_AVX2 8
+    IPFILTER_CHROMA_PS_6xN_AVX2 16

x265_2.6.tar.gz/source/common/x86/intrapred8.asm -> x265_2.7.tar.gz/source/common/x86/intrapred8.asm Changed

x265_2.6.tar.gz/source/common/x86/ipfilter16.asm -> x265_2.7.tar.gz/source/common/x86/ipfilter16.asm Changed

@@ -47,75 +47,10 @@
 
 SECTION_RODATA 32
 
-tab_c_32:         times 8 dd 32
 tab_c_524800:     times 4 dd 524800
 tab_c_n8192:      times 8 dw -8192
 pd_524800:        times 8 dd 524800
 
-tab_Tm16:         db 0, 1, 2, 3, 4,  5,  6, 7, 2, 3, 4,  5, 6, 7, 8, 9
-
-tab_ChromaCoeff:  dw  0, 64,  0,  0
-                  dw -2, 58, 10, -2
-                  dw -4, 54, 16, -2
-                  dw -6, 46, 28, -4
-                  dw -4, 36, 36, -4
-                  dw -4, 28, 46, -6
-                  dw -2, 16, 54, -4
-                  dw -2, 10, 58, -2
-
-const tab_ChromaCoeffV,  times 8 dw 0, 64
-                         times 8 dw 0, 0
-
-                         times 8 dw -2, 58
-                         times 8 dw 10, -2
-
-                         times 8 dw -4, 54
-                         times 8 dw 16, -2
-
-                         times 8 dw -6, 46
-                         times 8 dw 28, -4
-
-                         times 8 dw -4, 36
-                         times 8 dw 36, -4
-
-                         times 8 dw -4, 28
-                         times 8 dw 46, -6
-
-                         times 8 dw -2, 16
-                         times 8 dw 54, -4
-
-                         times 8 dw -2, 10
-                         times 8 dw 58, -2
-
-tab_ChromaCoeffVer: times 8 dw 0, 64
-                    times 8 dw 0, 0
-
-                    times 8 dw -2, 58
-                    times 8 dw 10, -2
-
-                    times 8 dw -4, 54
-                    times 8 dw 16, -2
-
-                    times 8 dw -6, 46
-                    times 8 dw 28, -4
-
-                    times 8 dw -4, 36
-                    times 8 dw 36, -4
-
-                    times 8 dw -4, 28
-                    times 8 dw 46, -6
-
-                    times 8 dw -2, 16
-                    times 8 dw 54, -4
-
-                    times 8 dw -2, 10
-                    times 8 dw 58, -2
-
-tab_LumaCoeff:    dw   0, 0,  0,  64,  0,   0,  0,  0
-                  dw  -1, 4, -10, 58,  17, -5,  1,  0
-                  dw  -1, 4, -11, 40,  40, -11, 4, -1
-                  dw   0, 1, -5,  17,  58, -10, 4, -1
-
 ALIGN 32
 tab_LumaCoeffV:   times 4 dw 0, 0
                   times 4 dw 0, 64
@@ -157,14 +92,6 @@
                   times 8 dw 58, -10
                   times 8 dw 4, -1
 
-const interp8_hps_shuf,     dd 0, 4, 1, 5, 2, 6, 3, 7
-
-const interp8_hpp_shuf,     db 0, 1, 2, 3, 4, 5, 6, 7, 2, 3, 4, 5, 6, 7, 8, 9
-                            db 4, 5, 6, 7, 8, 9, 10, 11, 6, 7, 8, 9, 10, 11, 12, 13
-
-const interp8_hpp_shuf_new, db 0, 1, 2, 3, 2, 3, 4, 5, 4, 5, 6, 7, 6, 7, 8, 9
-                            db 4, 5, 6, 7, 6, 7, 8, 9, 8, 9, 10, 11, 10, 11, 12, 13
-
 SECTION .text
 cextern pd_8
 cextern pd_32
@@ -175,255 +102,6 @@
 cextern pw_2000
 cextern idct8_shuf2
 
-%macro FILTER_LUMA_HOR_4_sse2 1
-    movu        m4,     [r0 + %1]       ; m4 = src[0-7]
-    movu        m5,     [r0 + %1 + 2]   ; m5 = src[1-8]
-    pmaddwd     m4,     m0
-    pmaddwd     m5,     m0
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m4,     m4,     q3120
-    pshufd      m5,     m5,     q3120
-    punpcklqdq  m4,     m5
-
-    movu        m5,     [r0 + %1 + 4]   ; m5 = src[2-9]
-    movu        m3,     [r0 + %1 + 6]   ; m3 = src[3-10]
-    pmaddwd     m5,     m0
-    pmaddwd     m3,     m0
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m2,     m3,     q2301
-    paddd       m3,     m2
-    pshufd      m5,     m5,     q3120
-    pshufd      m3,     m3,     q3120
-    punpcklqdq  m5,     m3
-
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m4,     m4,     q3120
-    pshufd      m5,     m5,     q3120
-    punpcklqdq  m4,     m5
-    paddd       m4,     m1
-%endmacro
-
-%macro FILTER_LUMA_HOR_8_sse2 1
-    movu        m4,     [r0 + %1]       ; m4 = src[0-7]
-    movu        m5,     [r0 + %1 + 2]   ; m5 = src[1-8]
-    pmaddwd     m4,     m0
-    pmaddwd     m5,     m0
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m4,     m4,     q3120
-    pshufd      m5,     m5,     q3120
-    punpcklqdq  m4,     m5
-
-    movu        m5,     [r0 + %1 + 4]   ; m5 = src[2-9]
-    movu        m3,     [r0 + %1 + 6]   ; m3 = src[3-10]
-    pmaddwd     m5,     m0
-    pmaddwd     m3,     m0
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m2,     m3,     q2301
-    paddd       m3,     m2
-    pshufd      m5,     m5,     q3120
-    pshufd      m3,     m3,     q3120
-    punpcklqdq  m5,     m3
-
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m4,     m4,     q3120
-    pshufd      m5,     m5,     q3120
-    punpcklqdq  m4,     m5
-    paddd       m4,     m1
-
-    movu        m5,     [r0 + %1 + 8]   ; m5 = src[4-11]
-    movu        m6,     [r0 + %1 + 10]  ; m6 = src[5-12]
-    pmaddwd     m5,     m0
-    pmaddwd     m6,     m0
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m2,     m6,     q2301
-    paddd       m6,     m2
-    pshufd      m5,     m5,     q3120
-    pshufd      m6,     m6,     q3120
-    punpcklqdq  m5,     m6
-
-    movu        m6,     [r0 + %1 + 12]  ; m6 = src[6-13]
-    movu        m3,     [r0 + %1 + 14]  ; m3 = src[7-14]
-    pmaddwd     m6,     m0
-    pmaddwd     m3,     m0
-    pshufd      m2,     m6,     q2301
-    paddd       m6,     m2
-    pshufd      m2,     m3,     q2301
-    paddd       m3,     m2
-    pshufd      m6,     m6,     q3120
-    pshufd      m3,     m3,     q3120
-    punpcklqdq  m6,     m3
-
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m2,     m6,     q2301
-    paddd       m6,     m2
-    pshufd      m5,     m5,     q3120
-    pshufd      m6,     m6,     q3120
-    punpcklqdq  m5,     m6
-    paddd       m5,     m1
-%endmacro
-
-;------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_p%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_sse2 3
-INIT_XMM sse2
-cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
-    mov         r4d,    r4m
-    sub         r0,     6
-    shl         r4d,    4
-    add         r1d,    r1d
-    add         r3d,    r3d
-
-%ifdef PIC
-    lea         r6,     [tab_LumaCoeff]
-    mova        m0,     [r6 + r4]
-%else
-    mova        m0,     [tab_LumaCoeff + r4]
-%endif
-
-%ifidn %3, pp
-    mova        m1,     [pd_32]
-    pxor        m7,     m7
-%else
-    mova        m1,     [INTERP_OFFSET_PS]
-%endif
-
-    mov         r4d,    %2
-%ifidn %3, ps
-    cmp         r5m,    byte 0
-    je          .loopH
-    lea         r6,     [r1 + 2 * r1]
-    sub         r0,     r6
-    add         r4d,    7
-%endif
-
-.loopH:
-%assign x 0
-%rep %1/8
-    FILTER_LUMA_HOR_8_sse2 x
-
-%ifidn %3, pp
-    psrad       m4,     6
-    psrad       m5,     6
-    packssdw    m4,     m5
-    CLIPW       m4,     m7,     [pw_pixel_max]
-%else
-  %if BIT_DEPTH == 10
-    psrad       m4,     2
-    psrad       m5,     2
-  %elif BIT_DEPTH == 12
-    psrad       m4,     4
-    psrad       m5,     4
-  %endif
-    packssdw    m4,     m5
-%endif
-
-    movu        [r2 + x], m4
-%assign x x+16
-%endrep
-
-%rep (%1 % 8)/4
-    FILTER_LUMA_HOR_4_sse2 x
-
-%ifidn %3, pp
-    psrad       m4,     6
-    packssdw    m4,     m4
-    CLIPW       m4,     m7,     [pw_pixel_max]
-%else
-  %if BIT_DEPTH == 10
-    psrad       m4,     2
-  %elif BIT_DEPTH == 12
-    psrad       m4,     4
-  %endif
-    packssdw    m4,     m4
-%endif
-
-    movh        [r2 + x], m4
-%endrep
-
-    add         r0,     r1
-    add         r2,     r3
-
-    dec         r4d
-    jnz         .loopH
-    RET
-
-%endmacro
-
-;------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;------------------------------------------------------------------------------------------------------------
-    FILTER_HOR_LUMA_sse2 4, 4, pp
-    FILTER_HOR_LUMA_sse2 4, 8, pp
-    FILTER_HOR_LUMA_sse2 4, 16, pp
-    FILTER_HOR_LUMA_sse2 8, 4, pp
-    FILTER_HOR_LUMA_sse2 8, 8, pp
-    FILTER_HOR_LUMA_sse2 8, 16, pp
-    FILTER_HOR_LUMA_sse2 8, 32, pp
-    FILTER_HOR_LUMA_sse2 12, 16, pp
-    FILTER_HOR_LUMA_sse2 16, 4, pp
-    FILTER_HOR_LUMA_sse2 16, 8, pp
-    FILTER_HOR_LUMA_sse2 16, 12, pp
-    FILTER_HOR_LUMA_sse2 16, 16, pp
-    FILTER_HOR_LUMA_sse2 16, 32, pp
-    FILTER_HOR_LUMA_sse2 16, 64, pp
-    FILTER_HOR_LUMA_sse2 24, 32, pp
-    FILTER_HOR_LUMA_sse2 32, 8, pp
-    FILTER_HOR_LUMA_sse2 32, 16, pp
-    FILTER_HOR_LUMA_sse2 32, 24, pp
-    FILTER_HOR_LUMA_sse2 32, 32, pp
-    FILTER_HOR_LUMA_sse2 32, 64, pp
-    FILTER_HOR_LUMA_sse2 48, 64, pp
-    FILTER_HOR_LUMA_sse2 64, 16, pp
-    FILTER_HOR_LUMA_sse2 64, 32, pp
-    FILTER_HOR_LUMA_sse2 64, 48, pp
-    FILTER_HOR_LUMA_sse2 64, 64, pp
-
-;---------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;---------------------------------------------------------------------------------------------------------------------------
-    FILTER_HOR_LUMA_sse2 4, 4, ps
-    FILTER_HOR_LUMA_sse2 4, 8, ps
-    FILTER_HOR_LUMA_sse2 4, 16, ps
-    FILTER_HOR_LUMA_sse2 8, 4, ps
-    FILTER_HOR_LUMA_sse2 8, 8, ps
-    FILTER_HOR_LUMA_sse2 8, 16, ps
-    FILTER_HOR_LUMA_sse2 8, 32, ps
-    FILTER_HOR_LUMA_sse2 12, 16, ps
-    FILTER_HOR_LUMA_sse2 16, 4, ps
-    FILTER_HOR_LUMA_sse2 16, 8, ps
-    FILTER_HOR_LUMA_sse2 16, 12, ps
-    FILTER_HOR_LUMA_sse2 16, 16, ps
-    FILTER_HOR_LUMA_sse2 16, 32, ps
-    FILTER_HOR_LUMA_sse2 16, 64, ps
-    FILTER_HOR_LUMA_sse2 24, 32, ps
-    FILTER_HOR_LUMA_sse2 32, 8, ps
-    FILTER_HOR_LUMA_sse2 32, 16, ps
-    FILTER_HOR_LUMA_sse2 32, 24, ps
-    FILTER_HOR_LUMA_sse2 32, 32, ps
-    FILTER_HOR_LUMA_sse2 32, 64, ps
-    FILTER_HOR_LUMA_sse2 48, 64, ps
-    FILTER_HOR_LUMA_sse2 64, 16, ps
-    FILTER_HOR_LUMA_sse2 64, 32, ps
-    FILTER_HOR_LUMA_sse2 64, 48, ps
-    FILTER_HOR_LUMA_sse2 64, 64, ps
-
 %macro PROCESS_LUMA_VER_W4_4R_sse2 0
     movq       m0, [r0]
     movq       m1, [r0 + r1]
@@ -623,5270 +301,6 @@
     FILTER_VER_LUMA_sse2 ps, 64, 16
     FILTER_VER_LUMA_sse2 ps, 16, 64
 
-%macro FILTERH_W2_4_sse3 2
-    movh        m3,     [r0 + %1]
-    movhps      m3,     [r0 + %1 + 2]
-    pmaddwd     m3,     m0
-    movh        m4,     [r0 + r1 + %1]
-    movhps      m4,     [r0 + r1 + %1 + 2]
-    pmaddwd     m4,     m0
-    pshufd      m2,     m3,     q2301
-    paddd       m3,     m2
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m3,     m3,     q3120
-    pshufd      m4,     m4,     q3120
-    punpcklqdq  m3,     m4
-    paddd       m3,     m1
-    movh        m5,     [r0 + 2 * r1 + %1]
-    movhps      m5,     [r0 + 2 * r1 + %1 + 2]
-    pmaddwd     m5,     m0
-    movh        m4,     [r0 + r4 + %1]
-    movhps      m4,     [r0 + r4 + %1 + 2]
-    pmaddwd     m4,     m0
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m5,     m5,     q3120
-    pshufd      m4,     m4,     q3120
-    punpcklqdq  m5,     m4
-    paddd       m5,     m1
-%ifidn %2, pp
-    psrad       m3,     6
-    psrad       m5,     6
-    packssdw    m3,     m5
-    CLIPW       m3,     m7,     m6
-%else
-    psrad       m3,     INTERP_SHIFT_PS
-    psrad       m5,     INTERP_SHIFT_PS
-    packssdw    m3,     m5
-%endif
-    movd        [r2 + %1], m3
-    psrldq      m3,     4
-    movd        [r2 + r3 + %1], m3
-    psrldq      m3,     4
-    movd        [r2 + r3 * 2 + %1], m3
-    psrldq      m3,     4
-    movd        [r2 + r5 + %1], m3
-%endmacro
-
-%macro FILTERH_W2_3_sse3 1
-    movh        m3,     [r0 + %1]
-    movhps      m3,     [r0 + %1 + 2]
-    pmaddwd     m3,     m0
-    movh        m4,     [r0 + r1 + %1]
-    movhps      m4,     [r0 + r1 + %1 + 2]
-    pmaddwd     m4,     m0
-    pshufd      m2,     m3,     q2301
-    paddd       m3,     m2
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m3,     m3,     q3120
-    pshufd      m4,     m4,     q3120
-    punpcklqdq  m3,     m4
-    paddd       m3,     m1
-
-    movh        m5,     [r0 + 2 * r1 + %1]
-    movhps      m5,     [r0 + 2 * r1 + %1 + 2]
-    pmaddwd     m5,     m0
-
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m5,     m5,     q3120
-    paddd       m5,     m1
-
-    psrad       m3,     INTERP_SHIFT_PS
-    psrad       m5,     INTERP_SHIFT_PS
-    packssdw    m3,     m5
-
-    movd        [r2 + %1], m3
-    psrldq      m3,     4
-    movd        [r2 + r3 + %1], m3
-    psrldq      m3,     4
-    movd        [r2 + r3 * 2 + %1], m3
-%endmacro
-
-%macro FILTERH_W4_2_sse3 2
-    movh        m3,     [r0 + %1]
-    movhps      m3,     [r0 + %1 + 2]
-    pmaddwd     m3,     m0
-    movh        m4,     [r0 + %1 + 4]
-    movhps      m4,     [r0 + %1 + 6]
-    pmaddwd     m4,     m0
-    pshufd      m2,     m3,     q2301
-    paddd       m3,     m2
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m3,     m3,     q3120
-    pshufd      m4,     m4,     q3120
-    punpcklqdq  m3,     m4
-    paddd       m3,     m1
-
-    movh        m5,     [r0 + r1 + %1]
-    movhps      m5,     [r0 + r1 + %1 + 2]
-    pmaddwd     m5,     m0
-    movh        m4,     [r0 + r1 + %1 + 4]
-    movhps      m4,     [r0 + r1 + %1 + 6]
-    pmaddwd     m4,     m0
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m5,     m5,     q3120
-    pshufd      m4,     m4,     q3120
-    punpcklqdq  m5,     m4
-    paddd       m5,     m1
-%ifidn %2, pp
-    psrad       m3,     6
-    psrad       m5,     6
-    packssdw    m3,     m5
-    CLIPW       m3,     m7,     m6
-%else
-    psrad       m3,     INTERP_SHIFT_PS
-    psrad       m5,     INTERP_SHIFT_PS
-    packssdw    m3,     m5
-%endif
-    movh        [r2 + %1], m3
-    movhps      [r2 + r3 + %1], m3
-%endmacro
-
-%macro FILTERH_W4_1_sse3 1
-    movh        m3,     [r0 + 2 * r1 + %1]
-    movhps      m3,     [r0 + 2 * r1 + %1 + 2]
-    pmaddwd     m3,     m0
-    movh        m4,     [r0 + 2 * r1 + %1 + 4]
-    movhps      m4,     [r0 + 2 * r1 + %1 + 6]
-    pmaddwd     m4,     m0
-    pshufd      m2,     m3,     q2301
-    paddd       m3,     m2
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m3,     m3,     q3120
-    pshufd      m4,     m4,     q3120
-    punpcklqdq  m3,     m4
-    paddd       m3,     m1
-
-    psrad       m3,     INTERP_SHIFT_PS
-    packssdw    m3,     m3
-    movh        [r2 + r3 * 2 + %1], m3
-%endmacro
-
-%macro FILTERH_W8_1_sse3 2
-    movh        m3,     [r0 + %1]
-    movhps      m3,     [r0 + %1 + 2]
-    pmaddwd     m3,     m0
-    movh        m4,     [r0 + %1 + 4]
-    movhps      m4,     [r0 + %1 + 6]
-    pmaddwd     m4,     m0
-    pshufd      m2,     m3,     q2301
-    paddd       m3,     m2
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m3,     m3,     q3120
-    pshufd      m4,     m4,     q3120
-    punpcklqdq  m3,     m4
-    paddd       m3,     m1
-
-    movh        m5,     [r0 + %1 + 8]
-    movhps      m5,     [r0 + %1 + 10]
-    pmaddwd     m5,     m0
-    movh        m4,     [r0 + %1 + 12]
-    movhps      m4,     [r0 + %1 + 14]
-    pmaddwd     m4,     m0
-    pshufd      m2,     m5,     q2301
-    paddd       m5,     m2
-    pshufd      m2,     m4,     q2301
-    paddd       m4,     m2
-    pshufd      m5,     m5,     q3120
-    pshufd      m4,     m4,     q3120
-    punpcklqdq  m5,     m4
-    paddd       m5,     m1
-%ifidn %2, pp
-    psrad       m3,     6
-    psrad       m5,     6
-    packssdw    m3,     m5
-    CLIPW       m3,     m7,     m6
-%else
-    psrad       m3,     INTERP_SHIFT_PS
-    psrad       m5,     INTERP_SHIFT_PS
-    packssdw    m3,     m5
-%endif
-    movdqu      [r2 + %1], m3
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_HOR_CHROMA_sse3 3
-INIT_XMM sse3
-cglobal interp_4tap_horiz_%3_%1x%2, 4, 7, 8
-    add         r3,     r3
-    add         r1,     r1
-    sub         r0,     2
-    mov         r4d,    r4m
-    add         r4d,    r4d
-
-%ifdef PIC
-    lea         r6,     [tab_ChromaCoeff]
-    movddup     m0,     [r6 + r4 * 4]
-%else
-    movddup     m0,     [tab_ChromaCoeff + r4 * 4]
-%endif
-
-%ifidn %3, ps
-    mova        m1,     [INTERP_OFFSET_PS]
-    cmp         r5m,    byte 0
-%if %1 <= 6
-    lea         r4,     [r1 * 3]
-    lea         r5,     [r3 * 3]
-%endif
-    je          .skip
-    sub         r0,     r1
-%if %1 <= 6
-%assign y 1
-%else
-%assign y 3
-%endif
-%assign z 0
-%rep y
-%assign x 0
-%rep %1/8
-    FILTERH_W8_1_sse3 x, %3
-%assign x x+16
-%endrep
-%if %1 == 4 || (%1 == 6 && z == 0) || (%1 == 12 && z == 0)
-    FILTERH_W4_2_sse3 x, %3
-    FILTERH_W4_1_sse3 x
-%assign x x+8
-%endif
-%if %1 == 2 || (%1 == 6 && z == 0)
-    FILTERH_W2_3_sse3 x
-%endif
-%if %1 <= 6
-    lea         r0,     [r0 + r4]
-    lea         r2,     [r2 + r5]
-%else
-    lea         r0,     [r0 + r1]
-    lea         r2,     [r2 + r3]
-%endif
-%assign z z+1
-%endrep
-.skip:
-%elifidn %3, pp
-    pxor        m7,     m7
-    mova        m6,     [pw_pixel_max]
-    mova        m1,     [tab_c_32]
-%if %1 == 2 || %1 == 6
-    lea         r4,     [r1 * 3]
-    lea         r5,     [r3 * 3]
-%endif
-%endif
-
-%if %1 == 2
-%assign y %2/4
-%elif %1 <= 6
-%assign y %2/2
-%else
-%assign y %2
-%endif
-%assign z 0
-%rep y
-%assign x 0
-%rep %1/8
-    FILTERH_W8_1_sse3 x, %3
-%assign x x+16
-%endrep
-%if %1 == 4 || %1 == 6 || (%1 == 12 && (z % 2) == 0)
-    FILTERH_W4_2_sse3 x, %3
-%assign x x+8
-%endif
-%if %1 == 2 || (%1 == 6 && (z % 2) == 0)
-    FILTERH_W2_4_sse3 x, %3
-%endif
-%assign z z+1
-%if z < y
-%if %1 == 2
-    lea         r0,     [r0 + 4 * r1]
-    lea         r2,     [r2 + 4 * r3]
-%elif %1 <= 6
-    lea         r0,     [r0 + 2 * r1]
-    lea         r2,     [r2 + 2 * r3]
-%else
-    lea         r0,     [r0 + r1]
-    lea         r2,     [r2 + r3]
-%endif
-%endif ;z < y
-%endrep
-
-    RET
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-
-FILTER_HOR_CHROMA_sse3 2, 4, pp
-FILTER_HOR_CHROMA_sse3 2, 8, pp
-FILTER_HOR_CHROMA_sse3 2, 16, pp
-FILTER_HOR_CHROMA_sse3 4, 2, pp
-FILTER_HOR_CHROMA_sse3 4, 4, pp
-FILTER_HOR_CHROMA_sse3 4, 8, pp
-FILTER_HOR_CHROMA_sse3 4, 16, pp
-FILTER_HOR_CHROMA_sse3 4, 32, pp
-FILTER_HOR_CHROMA_sse3 6, 8, pp
-FILTER_HOR_CHROMA_sse3 6, 16, pp
-FILTER_HOR_CHROMA_sse3 8, 2, pp
-FILTER_HOR_CHROMA_sse3 8, 4, pp
-FILTER_HOR_CHROMA_sse3 8, 6, pp
-FILTER_HOR_CHROMA_sse3 8, 8, pp
-FILTER_HOR_CHROMA_sse3 8, 12, pp
-FILTER_HOR_CHROMA_sse3 8, 16, pp
-FILTER_HOR_CHROMA_sse3 8, 32, pp
-FILTER_HOR_CHROMA_sse3 8, 64, pp
-FILTER_HOR_CHROMA_sse3 12, 16, pp
-FILTER_HOR_CHROMA_sse3 12, 32, pp
-FILTER_HOR_CHROMA_sse3 16, 4, pp
-FILTER_HOR_CHROMA_sse3 16, 8, pp
-FILTER_HOR_CHROMA_sse3 16, 12, pp
-FILTER_HOR_CHROMA_sse3 16, 16, pp
-FILTER_HOR_CHROMA_sse3 16, 24, pp
-FILTER_HOR_CHROMA_sse3 16, 32, pp
-FILTER_HOR_CHROMA_sse3 16, 64, pp
-FILTER_HOR_CHROMA_sse3 24, 32, pp
-FILTER_HOR_CHROMA_sse3 24, 64, pp
-FILTER_HOR_CHROMA_sse3 32, 8, pp
-FILTER_HOR_CHROMA_sse3 32, 16, pp
-FILTER_HOR_CHROMA_sse3 32, 24, pp
-FILTER_HOR_CHROMA_sse3 32, 32, pp
-FILTER_HOR_CHROMA_sse3 32, 48, pp
-FILTER_HOR_CHROMA_sse3 32, 64, pp
-FILTER_HOR_CHROMA_sse3 48, 64, pp
-FILTER_HOR_CHROMA_sse3 64, 16, pp
-FILTER_HOR_CHROMA_sse3 64, 32, pp
-FILTER_HOR_CHROMA_sse3 64, 48, pp
-FILTER_HOR_CHROMA_sse3 64, 64, pp
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-
-FILTER_HOR_CHROMA_sse3 2, 4, ps
-FILTER_HOR_CHROMA_sse3 2, 8, ps
-FILTER_HOR_CHROMA_sse3 2, 16, ps
-FILTER_HOR_CHROMA_sse3 4, 2, ps
-FILTER_HOR_CHROMA_sse3 4, 4, ps
-FILTER_HOR_CHROMA_sse3 4, 8, ps
-FILTER_HOR_CHROMA_sse3 4, 16, ps
-FILTER_HOR_CHROMA_sse3 4, 32, ps
-FILTER_HOR_CHROMA_sse3 6, 8, ps
-FILTER_HOR_CHROMA_sse3 6, 16, ps
-FILTER_HOR_CHROMA_sse3 8, 2, ps
-FILTER_HOR_CHROMA_sse3 8, 4, ps
-FILTER_HOR_CHROMA_sse3 8, 6, ps
-FILTER_HOR_CHROMA_sse3 8, 8, ps
-FILTER_HOR_CHROMA_sse3 8, 12, ps
-FILTER_HOR_CHROMA_sse3 8, 16, ps
-FILTER_HOR_CHROMA_sse3 8, 32, ps
-FILTER_HOR_CHROMA_sse3 8, 64, ps
-FILTER_HOR_CHROMA_sse3 12, 16, ps
-FILTER_HOR_CHROMA_sse3 12, 32, ps
-FILTER_HOR_CHROMA_sse3 16, 4, ps
-FILTER_HOR_CHROMA_sse3 16, 8, ps
-FILTER_HOR_CHROMA_sse3 16, 12, ps
-FILTER_HOR_CHROMA_sse3 16, 16, ps
-FILTER_HOR_CHROMA_sse3 16, 24, ps
-FILTER_HOR_CHROMA_sse3 16, 32, ps
-FILTER_HOR_CHROMA_sse3 16, 64, ps
-FILTER_HOR_CHROMA_sse3 24, 32, ps
-FILTER_HOR_CHROMA_sse3 24, 64, ps
-FILTER_HOR_CHROMA_sse3 32, 8, ps
-FILTER_HOR_CHROMA_sse3 32, 16, ps
-FILTER_HOR_CHROMA_sse3 32, 24, ps
-FILTER_HOR_CHROMA_sse3 32, 32, ps
-FILTER_HOR_CHROMA_sse3 32, 48, ps
-FILTER_HOR_CHROMA_sse3 32, 64, ps
-FILTER_HOR_CHROMA_sse3 48, 64, ps
-FILTER_HOR_CHROMA_sse3 64, 16, ps
-FILTER_HOR_CHROMA_sse3 64, 32, ps
-FILTER_HOR_CHROMA_sse3 64, 48, ps
-FILTER_HOR_CHROMA_sse3 64, 64, ps
-
-%macro FILTER_P2S_2_4_sse2 1
-    movd        m0,     [r0 + %1]
-    movd        m2,     [r0 + r1 * 2 + %1]
-    movhps      m0,     [r0 + r1 + %1]
-    movhps      m2,     [r0 + r4 + %1]
-    psllw       m0,     (14 - BIT_DEPTH)
-    psllw       m2,     (14 - BIT_DEPTH)
-    psubw       m0,     m1
-    psubw       m2,     m1
-
-    movd        [r2 + r3 * 0 + %1], m0
-    movd        [r2 + r3 * 2 + %1], m2
-    movhlps     m0,     m0
-    movhlps     m2,     m2
-    movd        [r2 + r3 * 1 + %1], m0
-    movd        [r2 + r5 + %1], m2
-%endmacro
-
-%macro FILTER_P2S_4_4_sse2 1
-    movh        m0,     [r0 + %1]
-    movhps      m0,     [r0 + r1 + %1]
-    psllw       m0,     (14 - BIT_DEPTH)
-    psubw       m0,     m1
-    movh        [r2 + r3 * 0 + %1], m0
-    movhps      [r2 + r3 * 1 + %1], m0
-
-    movh        m2,     [r0 + r1 * 2 + %1]
-    movhps      m2,     [r0 + r4 + %1]
-    psllw       m2,     (14 - BIT_DEPTH)
-    psubw       m2,     m1
-    movh        [r2 + r3 * 2 + %1], m2
-    movhps      [r2 + r5 + %1], m2
-%endmacro
-
-%macro FILTER_P2S_4_2_sse2 0
-    movh        m0,     [r0]
-    movhps      m0,     [r0 + r1 * 2]
-    psllw       m0,     (14 - BIT_DEPTH)
-    psubw       m0,     [pw_2000]
-    movh        [r2 + r3 * 0], m0
-    movhps      [r2 + r3 * 2], m0
-%endmacro
-
-%macro FILTER_P2S_8_4_sse2 1
-    movu        m0,     [r0 + %1]
-    movu        m2,     [r0 + r1 + %1]
-    psllw       m0,     (14 - BIT_DEPTH)
-    psllw       m2,     (14 - BIT_DEPTH)
-    psubw       m0,     m1
-    psubw       m2,     m1
-    movu        [r2 + r3 * 0 + %1], m0
-    movu        [r2 + r3 * 1 + %1], m2
-
-    movu        m3,     [r0 + r1 * 2 + %1]
-    movu        m4,     [r0 + r4 + %1]
-    psllw       m3,     (14 - BIT_DEPTH)
-    psllw       m4,     (14 - BIT_DEPTH)
-    psubw       m3,     m1
-    psubw       m4,     m1
-    movu        [r2 + r3 * 2 + %1], m3
-    movu        [r2 + r5 + %1], m4
-%endmacro
-
-%macro FILTER_P2S_8_2_sse2 1
-    movu        m0,     [r0 + %1]
-    movu        m2,     [r0 + r1 + %1]
-    psllw       m0,     (14 - BIT_DEPTH)
-    psllw       m2,     (14 - BIT_DEPTH)
-    psubw       m0,     m1
-    psubw       m2,     m1
-    movu        [r2 + r3 * 0 + %1], m0
-    movu        [r2 + r3 * 1 + %1], m2
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
-;-----------------------------------------------------------------------------
-%macro FILTER_PIX_TO_SHORT_sse2 2
-INIT_XMM sse2
-cglobal filterPixelToShort_%1x%2, 4, 6, 3
-%if %2 == 2
-%if %1 == 4
-    FILTER_P2S_4_2_sse2
-%elif %1 == 8
-    add        r1d, r1d
-    add        r3d, r3d
-    mova       m1, [pw_2000]
-    FILTER_P2S_8_2_sse2 0
-%endif
-%else
-    add        r1d, r1d
-    add        r3d, r3d
-    mova       m1, [pw_2000]
-    lea        r4, [r1 * 3]
-    lea        r5, [r3 * 3]
-%assign y 1
-%rep %2/4
-%assign x 0
-%rep %1/8
-    FILTER_P2S_8_4_sse2 x
-%if %2 == 6
-    lea         r0,     [r0 + 4 * r1]
-    lea         r2,     [r2 + 4 * r3]
-    FILTER_P2S_8_2_sse2 x
-%endif
-%assign x x+16
-%endrep
-%rep (%1 % 8)/4
-    FILTER_P2S_4_4_sse2 x
-%assign x x+8
-%endrep
-%rep (%1 % 4)/2
-    FILTER_P2S_2_4_sse2 x
-%endrep
-%if y < %2/4
-    lea         r0,     [r0 + 4 * r1]
-    lea         r2,     [r2 + 4 * r3]
-%assign y y+1
-%endif
-%endrep
-%endif
-RET
-%endmacro
-
-    FILTER_PIX_TO_SHORT_sse2 2, 4
-    FILTER_PIX_TO_SHORT_sse2 2, 8
-    FILTER_PIX_TO_SHORT_sse2 2, 16
-    FILTER_PIX_TO_SHORT_sse2 4, 2
-    FILTER_PIX_TO_SHORT_sse2 4, 4
-    FILTER_PIX_TO_SHORT_sse2 4, 8
-    FILTER_PIX_TO_SHORT_sse2 4, 16
-    FILTER_PIX_TO_SHORT_sse2 4, 32
-    FILTER_PIX_TO_SHORT_sse2 6, 8
-    FILTER_PIX_TO_SHORT_sse2 6, 16
-    FILTER_PIX_TO_SHORT_sse2 8, 2
-    FILTER_PIX_TO_SHORT_sse2 8, 4
-    FILTER_PIX_TO_SHORT_sse2 8, 6
-    FILTER_PIX_TO_SHORT_sse2 8, 8
-    FILTER_PIX_TO_SHORT_sse2 8, 12
-    FILTER_PIX_TO_SHORT_sse2 8, 16
-    FILTER_PIX_TO_SHORT_sse2 8, 32
-    FILTER_PIX_TO_SHORT_sse2 8, 64
-    FILTER_PIX_TO_SHORT_sse2 12, 16
-    FILTER_PIX_TO_SHORT_sse2 12, 32
-    FILTER_PIX_TO_SHORT_sse2 16, 4
-    FILTER_PIX_TO_SHORT_sse2 16, 8
-    FILTER_PIX_TO_SHORT_sse2 16, 12
-    FILTER_PIX_TO_SHORT_sse2 16, 16
-    FILTER_PIX_TO_SHORT_sse2 16, 24
-    FILTER_PIX_TO_SHORT_sse2 16, 32
-    FILTER_PIX_TO_SHORT_sse2 16, 64
-    FILTER_PIX_TO_SHORT_sse2 24, 32
-    FILTER_PIX_TO_SHORT_sse2 24, 64
-    FILTER_PIX_TO_SHORT_sse2 32, 8
-    FILTER_PIX_TO_SHORT_sse2 32, 16
-    FILTER_PIX_TO_SHORT_sse2 32, 24
-    FILTER_PIX_TO_SHORT_sse2 32, 32
-    FILTER_PIX_TO_SHORT_sse2 32, 48
-    FILTER_PIX_TO_SHORT_sse2 32, 64
-    FILTER_PIX_TO_SHORT_sse2 48, 64
-    FILTER_PIX_TO_SHORT_sse2 64, 16
-    FILTER_PIX_TO_SHORT_sse2 64, 32
-    FILTER_PIX_TO_SHORT_sse2 64, 48
-    FILTER_PIX_TO_SHORT_sse2 64, 64
-
-;------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W4 3
-INIT_XMM sse4
-cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
-    mov         r4d, r4m
-    sub         r0, 6
-    shl         r4d, 4
-    add         r1, r1
-    add         r3, r3
-
-%ifdef PIC
-    lea         r6, [tab_LumaCoeff]
-    mova        m0, [r6 + r4]
-%else
-    mova        m0, [tab_LumaCoeff + r4]
-%endif
-
-%ifidn %3, pp
-    mova        m1, [pd_32]
-    pxor        m6, m6
-    mova        m7, [pw_pixel_max]
-%else
-    mova        m1, [INTERP_OFFSET_PS]
-%endif
-
-    mov         r4d, %2
-%ifidn %3, ps
-    cmp         r5m, byte 0
-    je          .loopH
-    lea         r6, [r1 + 2 * r1]
-    sub         r0, r6
-    add         r4d, 7
-%endif
-
-.loopH:
-    movu        m2, [r0]                     ; m2 = src[0-7]
-    movu        m3, [r0 + 16]                ; m3 = src[8-15]
-
-    pmaddwd     m4, m2, m0
-    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
-    pmaddwd     m5, m0
-    palignr     m3, m2, 6                    ; m3 = src[3-10]
-    pmaddwd     m3, m0
-    phaddd      m5, m3
-
-    phaddd      m4, m5
-    paddd       m4, m1
-%ifidn %3, pp
-    psrad       m4, 6
-    packusdw    m4, m4
-    CLIPW       m4, m6, m7
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    packssdw    m4, m4
-%endif
-
-    movh        [r2], m4
-
-    add         r0, r1
-    add         r2, r3
-
-    dec         r4d
-    jnz         .loopH
-    RET
-%endmacro
-
-;------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_4x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W4 4, 4, pp
-FILTER_HOR_LUMA_W4 4, 8, pp
-FILTER_HOR_LUMA_W4 4, 16, pp
-
-;---------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_4x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;---------------------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W4 4, 4, ps
-FILTER_HOR_LUMA_W4 4, 8, ps
-FILTER_HOR_LUMA_W4 4, 16, ps
-
-;------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W8 3
-INIT_XMM sse4
-cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
-
-    add         r1, r1
-    add         r3, r3
-    mov         r4d, r4m
-    sub         r0, 6
-    shl         r4d, 4
-
-%ifdef PIC
-    lea         r6, [tab_LumaCoeff]
-    mova        m0, [r6 + r4]
-%else
-    mova        m0, [tab_LumaCoeff + r4]
-%endif
-
-%ifidn %3, pp
-    mova        m1, [pd_32]
-    pxor        m7, m7
-%else
-    mova        m1, [INTERP_OFFSET_PS]
-%endif
-
-    mov         r4d, %2
-%ifidn %3, ps
-    cmp         r5m, byte 0
-    je         .loopH
-    lea         r6, [r1 + 2 * r1]
-    sub         r0, r6
-    add         r4d, 7
-%endif
-
-.loopH:
-    movu        m2, [r0]                     ; m2 = src[0-7]
-    movu        m3, [r0 + 16]                ; m3 = src[8-15]
-
-    pmaddwd     m4, m2, m0
-    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-    phaddd      m4, m5
-    paddd       m4, m1
-
-    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-
-    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
-    pmaddwd     m6, m0
-    palignr     m3, m2, 14                   ; m3 = src[7-14]
-    pmaddwd     m3, m0
-    phaddd      m6, m3
-    phaddd      m5, m6
-    paddd       m5, m1
-%ifidn %3, pp
-    psrad       m4, 6
-    psrad       m5, 6
-    packusdw    m4, m5
-    CLIPW       m4, m7, [pw_pixel_max]
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    psrad       m5, INTERP_SHIFT_PS
-    packssdw    m4, m5
-%endif
-
-    movu        [r2], m4
-
-    add         r0, r1
-    add         r2, r3
-
-    dec         r4d
-    jnz        .loopH
-    RET
-%endmacro
-
-;------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W8 8, 4, pp
-FILTER_HOR_LUMA_W8 8, 8, pp
-FILTER_HOR_LUMA_W8 8, 16, pp
-FILTER_HOR_LUMA_W8 8, 32, pp
-
-;---------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;---------------------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W8 8, 4, ps
-FILTER_HOR_LUMA_W8 8, 8, ps
-FILTER_HOR_LUMA_W8 8, 16, ps
-FILTER_HOR_LUMA_W8 8, 32, ps
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W12 3
-INIT_XMM sse4
-cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
-
-    add         r1, r1
-    add         r3, r3
-    mov         r4d, r4m
-    sub         r0, 6
-    shl         r4d, 4
-
-%ifdef PIC
-    lea         r6, [tab_LumaCoeff]
-    mova        m0, [r6 + r4]
-%else
-    mova        m0, [tab_LumaCoeff + r4]
-%endif
-%ifidn %3, pp
-    mova        m1, [INTERP_OFFSET_PP]
-%else
-    mova        m1, [INTERP_OFFSET_PS]
-%endif
-
-    mov         r4d, %2
-%ifidn %3, ps
-    cmp         r5m, byte 0
-    je          .loopH
-    lea         r6, [r1 + 2 * r1]
-    sub         r0, r6
-    add         r4d, 7
-%endif
-
-.loopH:
-    movu        m2, [r0]                     ; m2 = src[0-7]
-    movu        m3, [r0 + 16]                ; m3 = src[8-15]
-
-    pmaddwd     m4, m2, m0
-    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-    phaddd      m4, m5
-    paddd       m4, m1
-
-    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-
-    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
-    pmaddwd     m6, m0
-    palignr     m7, m3, m2, 14               ; m2 = src[7-14]
-    pmaddwd     m7, m0
-    phaddd      m6, m7
-    phaddd      m5, m6
-    paddd       m5, m1
-%ifidn %3, pp
-    psrad       m4, INTERP_SHIFT_PP
-    psrad       m5, INTERP_SHIFT_PP
-    packusdw    m4, m5
-    pxor        m5, m5
-    CLIPW       m4, m5, [pw_pixel_max]
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    psrad       m5, INTERP_SHIFT_PS
-    packssdw    m4, m5
-%endif
-
-    movu        [r2], m4
-
-    movu        m2, [r0 + 32]                ; m2 = src[16-23]
-
-    pmaddwd     m4, m3, m0                   ; m3 = src[8-15]
-    palignr     m5, m2, m3, 2                ; m5 = src[9-16]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m2, m3, 4                ; m5 = src[10-17]
-    pmaddwd     m5, m0
-    palignr     m2, m3, 6                    ; m2 = src[11-18]
-    pmaddwd     m2, m0
-    phaddd      m5, m2
-    phaddd      m4, m5
-    paddd       m4, m1
-%ifidn %3, pp
-    psrad       m4, INTERP_SHIFT_PP
-    packusdw    m4, m4
-    pxor        m5, m5
-    CLIPW       m4, m5, [pw_pixel_max]
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    packssdw    m4, m4
-%endif
-
-    movh        [r2 + 16], m4
-
-    add         r0, r1
-    add         r2, r3
-
-    dec         r4d
-    jnz         .loopH
-    RET
-%endmacro
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_12x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W12 12, 16, pp
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_12x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W12 12, 16, ps
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W16 3
-INIT_XMM sse4
-cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
-
-    add         r1, r1
-    add         r3, r3
-    mov         r4d, r4m
-    sub         r0, 6
-    shl         r4d, 4
-
-%ifdef PIC
-    lea         r6, [tab_LumaCoeff]
-    mova        m0, [r6 + r4]
-%else
-    mova        m0, [tab_LumaCoeff + r4]
-%endif
-
-%ifidn %3, pp
-    mova        m1, [pd_32]
-%else
-    mova        m1, [INTERP_OFFSET_PS]
-%endif
-
-    mov         r4d, %2
-%ifidn %3, ps
-    cmp         r5m, byte 0
-    je          .loopH
-    lea         r6, [r1 + 2 * r1]
-    sub         r0, r6
-    add         r4d, 7
-%endif
-
-.loopH:
-%assign x 0
-%rep %1 / 16
-    movu        m2, [r0 + x]                 ; m2 = src[0-7]
-    movu        m3, [r0 + 16 + x]            ; m3 = src[8-15]
-
-    pmaddwd     m4, m2, m0
-    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-    phaddd      m4, m5
-    paddd       m4, m1
-
-    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-
-    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
-    pmaddwd     m6, m0
-    palignr     m7, m3, m2, 14               ; m2 = src[7-14]
-    pmaddwd     m7, m0
-    phaddd      m6, m7
-    phaddd      m5, m6
-    paddd       m5, m1
-%ifidn %3, pp
-    psrad       m4, INTERP_SHIFT_PP
-    psrad       m5, INTERP_SHIFT_PP
-    packusdw    m4, m5
-    pxor        m5, m5
-    CLIPW       m4, m5, [pw_pixel_max]
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    psrad       m5, INTERP_SHIFT_PS
-    packssdw    m4, m5
-%endif
-    movu        [r2 + x], m4
-
-    movu        m2, [r0 + 32 + x]            ; m2 = src[16-23]
-
-    pmaddwd     m4, m3, m0                   ; m3 = src[8-15]
-    palignr     m5, m2, m3, 2                ; m5 = src[9-16]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m2, m3, 4                ; m5 = src[10-17]
-    pmaddwd     m5, m0
-    palignr     m6, m2, m3, 6                ; m6 = src[11-18]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-    phaddd      m4, m5
-    paddd       m4, m1
-
-    palignr     m5, m2, m3, 8                ; m5 = src[12-19]
-    pmaddwd     m5, m0
-    palignr     m6, m2, m3, 10               ; m6 = src[13-20]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-
-    palignr     m6, m2, m3, 12               ; m6 = src[14-21]
-    pmaddwd     m6, m0
-    palignr     m2, m3, 14                   ; m3 = src[15-22]
-    pmaddwd     m2, m0
-    phaddd      m6, m2
-    phaddd      m5, m6
-    paddd       m5, m1
-%ifidn %3, pp
-    psrad       m4, INTERP_SHIFT_PP
-    psrad       m5, INTERP_SHIFT_PP
-    packusdw    m4, m5
-    pxor        m5, m5
-    CLIPW       m4, m5, [pw_pixel_max]
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    psrad       m5, INTERP_SHIFT_PS
-    packssdw    m4, m5
-%endif
-    movu        [r2 + 16 + x], m4
-
-%assign x x+32
-%endrep
-
-    add         r0, r1
-    add         r2, r3
-
-    dec         r4d
-    jnz         .loopH
-    RET
-%endmacro
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W16 16, 4, pp
-FILTER_HOR_LUMA_W16 16, 8, pp
-FILTER_HOR_LUMA_W16 16, 12, pp
-FILTER_HOR_LUMA_W16 16, 16, pp
-FILTER_HOR_LUMA_W16 16, 32, pp
-FILTER_HOR_LUMA_W16 16, 64, pp
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W16 16, 4, ps
-FILTER_HOR_LUMA_W16 16, 8, ps
-FILTER_HOR_LUMA_W16 16, 12, ps
-FILTER_HOR_LUMA_W16 16, 16, ps
-FILTER_HOR_LUMA_W16 16, 32, ps
-FILTER_HOR_LUMA_W16 16, 64, ps
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W16 32, 8, pp
-FILTER_HOR_LUMA_W16 32, 16, pp
-FILTER_HOR_LUMA_W16 32, 24, pp
-FILTER_HOR_LUMA_W16 32, 32, pp
-FILTER_HOR_LUMA_W16 32, 64, pp
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W16 32, 8, ps
-FILTER_HOR_LUMA_W16 32, 16, ps
-FILTER_HOR_LUMA_W16 32, 24, ps
-FILTER_HOR_LUMA_W16 32, 32, ps
-FILTER_HOR_LUMA_W16 32, 64, ps
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_48x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W16 48, 64, pp
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_48x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W16 48, 64, ps
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_64x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W16 64, 16, pp
-FILTER_HOR_LUMA_W16 64, 32, pp
-FILTER_HOR_LUMA_W16 64, 48, pp
-FILTER_HOR_LUMA_W16 64, 64, pp
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_64x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W16 64, 16, ps
-FILTER_HOR_LUMA_W16 64, 32, ps
-FILTER_HOR_LUMA_W16 64, 48, ps
-FILTER_HOR_LUMA_W16 64, 64, ps
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W24 3
-INIT_XMM sse4
-cglobal interp_8tap_horiz_%3_%1x%2, 4, 7, 8
-
-    add         r1, r1
-    add         r3, r3
-    mov         r4d, r4m
-    sub         r0, 6
-    shl         r4d, 4
-
-%ifdef PIC
-    lea         r6, [tab_LumaCoeff]
-    mova        m0, [r6 + r4]
-%else
-    mova        m0, [tab_LumaCoeff + r4]
-%endif
-%ifidn %3, pp
-    mova        m1, [pd_32]
-%else
-    mova        m1, [INTERP_OFFSET_PS]
-%endif
-
-    mov         r4d, %2
-%ifidn %3, ps
-    cmp         r5m, byte 0
-    je          .loopH
-    lea         r6, [r1 + 2 * r1]
-    sub         r0, r6
-    add         r4d, 7
-%endif
-
-.loopH:
-    movu        m2, [r0]                     ; m2 = src[0-7]
-    movu        m3, [r0 + 16]                ; m3 = src[8-15]
-
-    pmaddwd     m4, m2, m0
-    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-    phaddd      m4, m5
-    paddd       m4, m1
-
-    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-
-    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
-    pmaddwd     m6, m0
-    palignr     m7, m3, m2, 14               ; m7 = src[7-14]
-    pmaddwd     m7, m0
-    phaddd      m6, m7
-    phaddd      m5, m6
-    paddd       m5, m1
-%ifidn %3, pp
-    psrad       m4, INTERP_SHIFT_PP
-    psrad       m5, INTERP_SHIFT_PP
-    packusdw    m4, m5
-    pxor        m5, m5
-    CLIPW       m4, m5, [pw_pixel_max]
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    psrad       m5, INTERP_SHIFT_PS
-    packssdw    m4, m5
-%endif
-    movu        [r2], m4
-
-    movu        m2, [r0 + 32]                ; m2 = src[16-23]
-
-    pmaddwd     m4, m3, m0                   ; m3 = src[8-15]
-    palignr     m5, m2, m3, 2                ; m5 = src[1-8]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m2, m3, 4                ; m5 = src[2-9]
-    pmaddwd     m5, m0
-    palignr     m6, m2, m3, 6                ; m6 = src[3-10]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-    phaddd      m4, m5
-    paddd       m4, m1
-
-    palignr     m5, m2, m3, 8                ; m5 = src[4-11]
-    pmaddwd     m5, m0
-    palignr     m6, m2, m3, 10               ; m6 = src[5-12]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-
-    palignr     m6, m2, m3, 12               ; m6 = src[6-13]
-    pmaddwd     m6, m0
-    palignr     m7, m2, m3, 14               ; m7 = src[7-14]
-    pmaddwd     m7, m0
-    phaddd      m6, m7
-    phaddd      m5, m6
-    paddd       m5, m1
-%ifidn %3, pp
-    psrad       m4, INTERP_SHIFT_PP
-    psrad       m5, INTERP_SHIFT_PP
-    packusdw    m4, m5
-    pxor        m5, m5
-    CLIPW       m4, m5, [pw_pixel_max]
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    psrad       m5, INTERP_SHIFT_PS
-    packssdw    m4, m5
-%endif
-    movu        [r2 + 16], m4
-
-    movu        m3, [r0 + 48]                ; m3 = src[24-31]
-
-    pmaddwd     m4, m2, m0                   ; m2 = src[16-23]
-    palignr     m5, m3, m2, 2                ; m5 = src[1-8]
-    pmaddwd     m5, m0
-    phaddd      m4, m5
-
-    palignr     m5, m3, m2, 4                ; m5 = src[2-9]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 6                ; m6 = src[3-10]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-    phaddd      m4, m5
-    paddd       m4, m1
-
-    palignr     m5, m3, m2, 8                ; m5 = src[4-11]
-    pmaddwd     m5, m0
-    palignr     m6, m3, m2, 10               ; m6 = src[5-12]
-    pmaddwd     m6, m0
-    phaddd      m5, m6
-
-    palignr     m6, m3, m2, 12               ; m6 = src[6-13]
-    pmaddwd     m6, m0
-    palignr     m7, m3, m2, 14               ; m7 = src[7-14]
-    pmaddwd     m7, m0
-    phaddd      m6, m7
-    phaddd      m5, m6
-    paddd       m5, m1
-%ifidn %3, pp
-    psrad       m4, INTERP_SHIFT_PP
-    psrad       m5, INTERP_SHIFT_PP
-    packusdw    m4, m5
-    pxor        m5, m5
-    CLIPW       m4, m5, [pw_pixel_max]
-%else
-    psrad       m4, INTERP_SHIFT_PS
-    psrad       m5, INTERP_SHIFT_PS
-    packssdw    m4, m5
-%endif
-    movu        [r2 + 32], m4
-
-    add         r0, r1
-    add         r2, r3
-
-    dec         r4d
-    jnz         .loopH
-    RET
-%endmacro
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_24x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W24 24, 32, pp
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_24x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-FILTER_HOR_LUMA_W24 24, 32, ps
-
-%macro FILTER_W2_2 1
-    movu        m3,         [r0]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + r1]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    packusdw    m3,         m3
-    CLIPW       m3,         m7,    m6
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    packssdw    m3,         m3
-%endif
-    movd        [r2],       m3
-    pextrd      [r2 + r3],  m3, 1
-%endmacro
-
-%macro FILTER_W4_2 1
-    movu        m3,         [r0]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + r1]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + r1 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m7,    m6
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2],       m3
-    movhps      [r2 + r3],  m3
-%endmacro
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W4_avx2 1
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_4x%1, 4,7,7
-    add              r1d, r1d
-    add              r3d, r3d
-    sub              r0, 6
-    mov              r4d, r4m
-    shl              r4d, 4
-%ifdef PIC
-    lea              r5, [tab_LumaCoeff]
-    vpbroadcastq     m0, [r5 + r4]
-    vpbroadcastq     m1, [r5 + r4 + 8]
-%else
-    vpbroadcastq     m0, [tab_LumaCoeff + r4]
-    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    lea              r6, [pw_pixel_max]
-    mova             m3, [interp8_hpp_shuf]
-    mova             m6, [pd_32]
-    pxor             m2, m2
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    mov              r4d, %1/2
-
-.loop:
-    vbroadcasti128   m4, [r0]
-    vbroadcasti128   m5, [r0 + 8]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    phaddd           m4, m4
-    vpermq           m4, m4, q3120
-    paddd            m4, m6
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [r6]
-    movq             [r2], xm4
-
-    vbroadcasti128   m4, [r0 + r1]
-    vbroadcasti128   m5, [r0 + r1 + 8]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    phaddd           m4, m4
-    vpermq           m4, m4, q3120
-    paddd            m4, m6
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [r6]
-    movq             [r2 + r3], xm4
-
-    lea              r2, [r2 + 2 * r3]
-    lea              r0, [r0 + 2 * r1]
-    dec              r4d
-    jnz              .loop
-    RET
-%endmacro
-FILTER_HOR_LUMA_W4_avx2 4
-FILTER_HOR_LUMA_W4_avx2 8
-FILTER_HOR_LUMA_W4_avx2 16
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W8 1
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_8x%1, 4,6,8
-    add              r1d, r1d
-    add              r3d, r3d
-    sub              r0, 6
-    mov              r4d, r4m
-    shl              r4d, 4
-%ifdef PIC
-    lea              r5, [tab_LumaCoeff]
-    vpbroadcastq     m0, [r5 + r4]
-    vpbroadcastq     m1, [r5 + r4 + 8]
-%else
-    vpbroadcastq     m0, [tab_LumaCoeff + r4]
-    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova             m3, [interp8_hpp_shuf]
-    mova             m7, [pd_32]
-    pxor             m2, m2
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    mov              r4d, %1/2
-
-.loop:
-    vbroadcasti128   m4, [r0]
-    vbroadcasti128   m5, [r0 + 8]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 8]
-    vbroadcasti128   m6, [r0 + 16]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2], xm4
-
-    vbroadcasti128   m4, [r0 + r1]
-    vbroadcasti128   m5, [r0 + r1 + 8]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + r1 + 8]
-    vbroadcasti128   m6, [r0 + r1 + 16]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + r3], xm4
-
-    lea              r2, [r2 + 2 * r3]
-    lea              r0, [r0 + 2 * r1]
-    dec              r4d
-    jnz              .loop
-    RET
-%endmacro
-FILTER_HOR_LUMA_W8 4
-FILTER_HOR_LUMA_W8 8
-FILTER_HOR_LUMA_W8 16
-FILTER_HOR_LUMA_W8 32
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W16 1
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_16x%1, 4,6,8
-    add              r1d, r1d
-    add              r3d, r3d
-    sub              r0, 6
-    mov              r4d, r4m
-    shl              r4d, 4
-%ifdef PIC
-    lea              r5, [tab_LumaCoeff]
-    vpbroadcastq     m0, [r5 + r4]
-    vpbroadcastq     m1, [r5 + r4 + 8]
-%else
-    vpbroadcastq     m0, [tab_LumaCoeff + r4]
-    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova             m3, [interp8_hpp_shuf]
-    mova             m7, [pd_32]
-    pxor             m2, m2
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    mov              r4d, %1
-
-.loop:
-    vbroadcasti128   m4, [r0]
-    vbroadcasti128   m5, [r0 + 8]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 8]
-    vbroadcasti128   m6, [r0 + 16]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2], xm4
-
-    vbroadcasti128   m4, [r0 + 16]
-    vbroadcasti128   m5, [r0 + 24]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 24]
-    vbroadcasti128   m6, [r0 + 32]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + 16], xm4
-
-    add              r2, r3
-    add              r0, r1
-    dec              r4d
-    jnz              .loop
-    RET
-%endmacro
-FILTER_HOR_LUMA_W16 4
-FILTER_HOR_LUMA_W16 8
-FILTER_HOR_LUMA_W16 12
-FILTER_HOR_LUMA_W16 16
-FILTER_HOR_LUMA_W16 32
-FILTER_HOR_LUMA_W16 64
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro FILTER_HOR_LUMA_W32 2
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
-    add              r1d, r1d
-    add              r3d, r3d
-    sub              r0, 6
-    mov              r4d, r4m
-    shl              r4d, 4
-%ifdef PIC
-    lea              r5, [tab_LumaCoeff]
-    vpbroadcastq     m0, [r5 + r4]
-    vpbroadcastq     m1, [r5 + r4 + 8]
-%else
-    vpbroadcastq     m0, [tab_LumaCoeff + r4]
-    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova             m3, [interp8_hpp_shuf]
-    mova             m7, [pd_32]
-    pxor             m2, m2
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    mov              r4d, %2
-
-.loop:
-%assign x 0
-%rep %1/16
-    vbroadcasti128   m4, [r0 + x]
-    vbroadcasti128   m5, [r0 + 8 + x]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 8 + x]
-    vbroadcasti128   m6, [r0 + 16 + x]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + x], xm4
-
-    vbroadcasti128   m4, [r0 + 16 + x]
-    vbroadcasti128   m5, [r0 + 24 + x]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 24 + x]
-    vbroadcasti128   m6, [r0 + 32 + x]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + 16 + x], xm4
-
-%assign x x+32
-%endrep
-
-    add              r2, r3
-    add              r0, r1
-    dec              r4d
-    jnz              .loop
-    RET
-%endmacro
-FILTER_HOR_LUMA_W32 32, 8
-FILTER_HOR_LUMA_W32 32, 16
-FILTER_HOR_LUMA_W32 32, 24
-FILTER_HOR_LUMA_W32 32, 32
-FILTER_HOR_LUMA_W32 32, 64
-FILTER_HOR_LUMA_W32 64, 16
-FILTER_HOR_LUMA_W32 64, 32
-FILTER_HOR_LUMA_W32 64, 48
-FILTER_HOR_LUMA_W32 64, 64
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_12x16, 4,6,8
-    add              r1d, r1d
-    add              r3d, r3d
-    sub              r0, 6
-    mov              r4d, r4m
-    shl              r4d, 4
-%ifdef PIC
-    lea              r5, [tab_LumaCoeff]
-    vpbroadcastq     m0, [r5 + r4]
-    vpbroadcastq     m1, [r5 + r4 + 8]
-%else
-    vpbroadcastq     m0, [tab_LumaCoeff + r4]
-    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova             m3, [interp8_hpp_shuf]
-    mova             m7, [pd_32]
-    pxor             m2, m2
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    mov              r4d, 16
-
-.loop:
-    vbroadcasti128   m4, [r0]
-    vbroadcasti128   m5, [r0 + 8]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 8]
-    vbroadcasti128   m6, [r0 + 16]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2], xm4
-
-    vbroadcasti128   m4, [r0 + 16]
-    vbroadcasti128   m5, [r0 + 24]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 24]
-    vbroadcasti128   m6, [r0 + 32]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movq             [r2 + 16], xm4
-
-    add              r2, r3
-    add              r0, r1
-    dec              r4d
-    jnz              .loop
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_24x32, 4,6,8
-    add              r1d, r1d
-    add              r3d, r3d
-    sub              r0, 6
-    mov              r4d, r4m
-    shl              r4d, 4
-%ifdef PIC
-    lea              r5, [tab_LumaCoeff]
-    vpbroadcastq     m0, [r5 + r4]
-    vpbroadcastq     m1, [r5 + r4 + 8]
-%else
-    vpbroadcastq     m0, [tab_LumaCoeff + r4]
-    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova             m3, [interp8_hpp_shuf]
-    mova             m7, [pd_32]
-    pxor             m2, m2
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    mov              r4d, 32
-
-.loop:
-    vbroadcasti128   m4, [r0]
-    vbroadcasti128   m5, [r0 + 8]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 8]
-    vbroadcasti128   m6, [r0 + 16]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2], xm4
-
-    vbroadcasti128   m4, [r0 + 16]
-    vbroadcasti128   m5, [r0 + 24]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 24]
-    vbroadcasti128   m6, [r0 + 32]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + 16], xm4
-
-    vbroadcasti128   m4, [r0 + 32]
-    vbroadcasti128   m5, [r0 + 40]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 40]
-    vbroadcasti128   m6, [r0 + 48]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + 32], xm4
-
-    add              r2, r3
-    add              r0, r1
-    dec              r4d
-    jnz              .loop
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_48x64, 4,6,8
-    add              r1d, r1d
-    add              r3d, r3d
-    sub              r0, 6
-    mov              r4d, r4m
-    shl              r4d, 4
-%ifdef PIC
-    lea              r5, [tab_LumaCoeff]
-    vpbroadcastq     m0, [r5 + r4]
-    vpbroadcastq     m1, [r5 + r4 + 8]
-%else
-    vpbroadcastq     m0, [tab_LumaCoeff + r4]
-    vpbroadcastq     m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova             m3, [interp8_hpp_shuf]
-    mova             m7, [pd_32]
-    pxor             m2, m2
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    mov              r4d, 64
-
-.loop:
-%assign x 0
-%rep 2
-    vbroadcasti128   m4, [r0 + x]
-    vbroadcasti128   m5, [r0 + 8 + x]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 8 + x]
-    vbroadcasti128   m6, [r0 + 16 + x]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + x], xm4
-
-    vbroadcasti128   m4, [r0 + 16 + x]
-    vbroadcasti128   m5, [r0 + 24 + x]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 24 + x]
-    vbroadcasti128   m6, [r0 + 32 + x]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + 16 + x], xm4
-
-    vbroadcasti128   m4, [r0 + 32 + x]
-    vbroadcasti128   m5, [r0 + 40 + x]
-    pshufb           m4, m3
-    pshufb           m5, m3
-
-    pmaddwd          m4, m0
-    pmaddwd          m5, m1
-    paddd            m4, m5
-
-    vbroadcasti128   m5, [r0 + 40 + x]
-    vbroadcasti128   m6, [r0 + 48 + x]
-    pshufb           m5, m3
-    pshufb           m6, m3
-
-    pmaddwd          m5, m0
-    pmaddwd          m6, m1
-    paddd            m5, m6
-
-    phaddd           m4, m5
-    vpermq           m4, m4, q3120
-    paddd            m4, m7
-    psrad            m4, INTERP_SHIFT_PP
-
-    packusdw         m4, m4
-    vpermq           m4, m4, q2020
-    CLIPW            m4, m2, [pw_pixel_max]
-    movu             [r2 + 32 + x], xm4
-
-%assign x x+48
-%endrep
-
-    add              r2, r3
-    add              r0, r1
-    dec              r4d
-    jnz              .loop
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_CHROMA_H 6
-INIT_XMM sse4
-cglobal interp_4tap_horiz_%3_%1x%2, 4, %4, %5
-
-    add         r3,       r3
-    add         r1,       r1
-    sub         r0,       2
-    mov         r4d,      r4m
-    add         r4d,      r4d
-
-%ifdef PIC
-    lea         r%6,      [tab_ChromaCoeff]
-    movh        m0,       [r%6 + r4 * 4]
-%else
-    movh        m0,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    punpcklqdq  m0,       m0
-    mova        m2,       [tab_Tm16]
-
-%ifidn %3, ps
-    mova        m1,       [INTERP_OFFSET_PS]
-    cmp         r5m, byte 0
-    je          .skip
-    sub         r0,       r1
-    movu        m3,       [r0]
-    pshufb      m3,       m3, m2
-    pmaddwd     m3,       m0
-
-  %if %1 == 4
-    movu        m4,       [r0 + 4]
-    pshufb      m4,       m4, m2
-    pmaddwd     m4,       m0
-    phaddd      m3,       m4
-  %else
-    phaddd      m3,       m3
-  %endif
-
-    paddd       m3,       m1
-    psrad       m3,       INTERP_SHIFT_PS
-    packssdw    m3,       m3
-
-  %if %1 == 2
-    movd        [r2],     m3
-  %else
-    movh        [r2],     m3
-  %endif
-
-    add         r0,       r1
-    add         r2,       r3
-    FILTER_W%1_2 %3
-    lea         r0,       [r0 + 2 * r1]
-    lea         r2,       [r2 + 2 * r3]
-
-.skip:
-
-%else     ;%ifidn %3, ps
-    pxor        m7,       m7
-    mova        m6,       [pw_pixel_max]
-    mova        m1,       [tab_c_32]
-%endif    ;%ifidn %3, ps
-
-    FILTER_W%1_2 %3
-
-%rep (%2/2) - 1
-    lea         r0,       [r0 + 2 * r1]
-    lea         r2,       [r2 + 2 * r3]
-    FILTER_W%1_2 %3
-%endrep
-    RET
-%endmacro
-
-FILTER_CHROMA_H 2, 4, pp, 6, 8, 5
-FILTER_CHROMA_H 2, 8, pp, 6, 8, 5
-FILTER_CHROMA_H 4, 2, pp, 6, 8, 5
-FILTER_CHROMA_H 4, 4, pp, 6, 8, 5
-FILTER_CHROMA_H 4, 8, pp, 6, 8, 5
-FILTER_CHROMA_H 4, 16, pp, 6, 8, 5
-
-FILTER_CHROMA_H 2, 4, ps, 7, 5, 6
-FILTER_CHROMA_H 2, 8, ps, 7, 5, 6
-FILTER_CHROMA_H 4, 2, ps, 7, 6, 6
-FILTER_CHROMA_H 4, 4, ps, 7, 6, 6
-FILTER_CHROMA_H 4, 8, ps, 7, 6, 6
-FILTER_CHROMA_H 4, 16, ps, 7, 6, 6
-
-FILTER_CHROMA_H 2, 16, pp, 6, 8, 5
-FILTER_CHROMA_H 4, 32, pp, 6, 8, 5
-FILTER_CHROMA_H 2, 16, ps, 7, 5, 6
-FILTER_CHROMA_H 4, 32, ps, 7, 6, 6
-
-
-%macro FILTER_W6_1 1
-    movu        m3,         [r0]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m4,         [r0 + 8]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m4,         m4
-    paddd       m4,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m4,         INTERP_SHIFT_PP
-    packusdw    m3,         m4
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m4,         INTERP_SHIFT_PS
-    packssdw    m3,         m4
-%endif
-    movh        [r2],       m3
-    pextrd      [r2 + 8],   m3, 2
-%endmacro
-
-cglobal chroma_filter_pp_6x1_internal
-    FILTER_W6_1 pp
-    ret
-
-cglobal chroma_filter_ps_6x1_internal
-    FILTER_W6_1 ps
-    ret
-
-%macro FILTER_W8_1 1
-    movu        m3,         [r0]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 8]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 12]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2],       m3
-    movhps      [r2 + 8],   m3
-%endmacro
-
-cglobal chroma_filter_pp_8x1_internal
-    FILTER_W8_1 pp
-    ret
-
-cglobal chroma_filter_ps_8x1_internal
-    FILTER_W8_1 ps
-    ret
-
-%macro FILTER_W12_1 1
-    movu        m3,         [r0]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 8]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 12]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2],       m3
-    movhps      [r2 + 8],   m3
-
-    movu        m3,         [r0 + 16]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 20]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    packusdw    m3,         m3
-    CLIPW       m3,         m6, m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    packssdw    m3,         m3
-%endif
-    movh        [r2 + 16],  m3
-%endmacro
-
-cglobal chroma_filter_pp_12x1_internal
-    FILTER_W12_1 pp
-    ret
-
-cglobal chroma_filter_ps_12x1_internal
-    FILTER_W12_1 ps
-    ret
-
-%macro FILTER_W16_1 1
-    movu        m3,         [r0]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 8]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 12]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2],       m3
-    movhps      [r2 + 8],   m3
-
-    movu        m3,         [r0 + 16]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 20]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 24]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 28]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2 + 16],  m3
-    movhps      [r2 + 24],  m3
-%endmacro
-
-cglobal chroma_filter_pp_16x1_internal
-    FILTER_W16_1 pp
-    ret
-
-cglobal chroma_filter_ps_16x1_internal
-    FILTER_W16_1 ps
-    ret
-
-%macro FILTER_W24_1 1
-    movu        m3,         [r0]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 8]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 12]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2],       m3
-    movhps      [r2 + 8],   m3
-
-    movu        m3,         [r0 + 16]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 20]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 24]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 28]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2 + 16],  m3
-    movhps      [r2 + 24],  m3
-
-    movu        m3,         [r0 + 32]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 36]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 40]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 44]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2 + 32],  m3
-    movhps      [r2 + 40],  m3
-%endmacro
-
-cglobal chroma_filter_pp_24x1_internal
-    FILTER_W24_1 pp
-    ret
-
-cglobal chroma_filter_ps_24x1_internal
-    FILTER_W24_1 ps
-    ret
-
-%macro FILTER_W32_1 1
-    movu        m3,         [r0]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 8]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 12]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2],       m3
-    movhps      [r2 + 8],   m3
-
-    movu        m3,         [r0 + 16]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 20]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 24]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 28]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2 + 16],  m3
-    movhps      [r2 + 24],  m3
-
-    movu        m3,         [r0 + 32]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 36]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 40]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 44]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2 + 32],  m3
-    movhps      [r2 + 40],  m3
-
-    movu        m3,         [r0 + 48]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + 52]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + 56]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + 60]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2 + 48],  m3
-    movhps      [r2 + 56],  m3
-%endmacro
-
-cglobal chroma_filter_pp_32x1_internal
-    FILTER_W32_1 pp
-    ret
-
-cglobal chroma_filter_ps_32x1_internal
-    FILTER_W32_1 ps
-    ret
-
-%macro FILTER_W8o_1 2
-    movu        m3,         [r0 + %2]
-    pshufb      m3,         m3, m2
-    pmaddwd     m3,         m0
-    movu        m4,         [r0 + %2 + 4]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m3,         m4
-    paddd       m3,         m1
-
-    movu        m5,         [r0 + %2 + 8]
-    pshufb      m5,         m5, m2
-    pmaddwd     m5,         m0
-    movu        m4,         [r0 + %2 + 12]
-    pshufb      m4,         m4, m2
-    pmaddwd     m4,         m0
-    phaddd      m5,         m4
-    paddd       m5,         m1
-%ifidn %1, pp
-    psrad       m3,         INTERP_SHIFT_PP
-    psrad       m5,         INTERP_SHIFT_PP
-    packusdw    m3,         m5
-    CLIPW       m3,         m6,    m7
-%else
-    psrad       m3,         INTERP_SHIFT_PS
-    psrad       m5,         INTERP_SHIFT_PS
-    packssdw    m3,         m5
-%endif
-    movh        [r2 + %2],       m3
-    movhps      [r2 + %2 + 8],   m3
-%endmacro
-
-%macro FILTER_W48_1 1
-    FILTER_W8o_1 %1, 0
-    FILTER_W8o_1 %1, 16
-    FILTER_W8o_1 %1, 32
-    FILTER_W8o_1 %1, 48
-    FILTER_W8o_1 %1, 64
-    FILTER_W8o_1 %1, 80
-%endmacro
-
-cglobal chroma_filter_pp_48x1_internal
-    FILTER_W48_1 pp
-    ret
-
-cglobal chroma_filter_ps_48x1_internal
-    FILTER_W48_1 ps
-    ret
-
-%macro FILTER_W64_1 1
-    FILTER_W8o_1 %1, 0
-    FILTER_W8o_1 %1, 16
-    FILTER_W8o_1 %1, 32
-    FILTER_W8o_1 %1, 48
-    FILTER_W8o_1 %1, 64
-    FILTER_W8o_1 %1, 80
-    FILTER_W8o_1 %1, 96
-    FILTER_W8o_1 %1, 112
-%endmacro
-
-cglobal chroma_filter_pp_64x1_internal
-    FILTER_W64_1 pp
-    ret
-
-cglobal chroma_filter_ps_64x1_internal
-    FILTER_W64_1 ps
-    ret
-
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-
-INIT_XMM sse4
-%macro IPFILTER_CHROMA 6
-cglobal interp_4tap_horiz_%3_%1x%2, 4, %5, %6
-
-    add         r3,        r3
-    add         r1,        r1
-    sub         r0,         2
-    mov         r4d,        r4m
-    add         r4d,        r4d
-
-%ifdef PIC
-    lea         r%4,       [tab_ChromaCoeff]
-    movh        m0,       [r%4 + r4 * 4]
-%else
-    movh        m0,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    punpcklqdq  m0,       m0
-    mova        m2,       [tab_Tm16]
-
-%ifidn %3, ps
-    mova        m1,       [INTERP_OFFSET_PS]
-    cmp         r5m, byte 0
-    je          .skip
-    sub         r0, r1
-    call chroma_filter_%3_%1x1_internal
-    add         r0, r1
-    add         r2, r3
-    call chroma_filter_%3_%1x1_internal
-    add         r0, r1
-    add         r2, r3
-    call chroma_filter_%3_%1x1_internal
-    add         r0, r1
-    add         r2, r3
-.skip:
-%else
-    mova        m1,         [tab_c_32]
-    pxor        m6,         m6
-    mova        m7,         [pw_pixel_max]
-%endif
-
-    call chroma_filter_%3_%1x1_internal
-%rep %2 - 1
-    add         r0,       r1
-    add         r2,       r3
-    call chroma_filter_%3_%1x1_internal
-%endrep
-RET
-%endmacro
-IPFILTER_CHROMA 6, 8, pp, 5, 6, 8
-IPFILTER_CHROMA 8, 2, pp, 5, 6, 8
-IPFILTER_CHROMA 8, 4, pp, 5, 6, 8
-IPFILTER_CHROMA 8, 6, pp, 5, 6, 8
-IPFILTER_CHROMA 8, 8, pp, 5, 6, 8
-IPFILTER_CHROMA 8, 16, pp, 5, 6, 8
-IPFILTER_CHROMA 8, 32, pp, 5, 6, 8
-IPFILTER_CHROMA 12, 16, pp, 5, 6, 8
-IPFILTER_CHROMA 16, 4, pp, 5, 6, 8
-IPFILTER_CHROMA 16, 8, pp, 5, 6, 8
-IPFILTER_CHROMA 16, 12, pp, 5, 6, 8
-IPFILTER_CHROMA 16, 16, pp, 5, 6, 8
-IPFILTER_CHROMA 16, 32, pp, 5, 6, 8
-IPFILTER_CHROMA 24, 32, pp, 5, 6, 8
-IPFILTER_CHROMA 32, 8, pp, 5, 6, 8
-IPFILTER_CHROMA 32, 16, pp, 5, 6, 8
-IPFILTER_CHROMA 32, 24, pp, 5, 6, 8
-IPFILTER_CHROMA 32, 32, pp, 5, 6, 8
-
-IPFILTER_CHROMA 6, 8, ps, 6, 7, 6
-IPFILTER_CHROMA 8, 2, ps, 6, 7, 6
-IPFILTER_CHROMA 8, 4, ps, 6, 7, 6
-IPFILTER_CHROMA 8, 6, ps, 6, 7, 6
-IPFILTER_CHROMA 8, 8, ps, 6, 7, 6
-IPFILTER_CHROMA 8, 16, ps, 6, 7, 6
-IPFILTER_CHROMA 8, 32, ps, 6, 7, 6
-IPFILTER_CHROMA 12, 16, ps, 6, 7, 6
-IPFILTER_CHROMA 16, 4, ps, 6, 7, 6
-IPFILTER_CHROMA 16, 8, ps, 6, 7, 6
-IPFILTER_CHROMA 16, 12, ps, 6, 7, 6
-IPFILTER_CHROMA 16, 16, ps, 6, 7, 6
-IPFILTER_CHROMA 16, 32, ps, 6, 7, 6
-IPFILTER_CHROMA 24, 32, ps, 6, 7, 6
-IPFILTER_CHROMA 32, 8, ps, 6, 7, 6
-IPFILTER_CHROMA 32, 16, ps, 6, 7, 6
-IPFILTER_CHROMA 32, 24, ps, 6, 7, 6
-IPFILTER_CHROMA 32, 32, ps, 6, 7, 6
-
-IPFILTER_CHROMA 6, 16, pp, 5, 6, 8
-IPFILTER_CHROMA 8, 12, pp, 5, 6, 8
-IPFILTER_CHROMA 8, 64, pp, 5, 6, 8
-IPFILTER_CHROMA 12, 32, pp, 5, 6, 8
-IPFILTER_CHROMA 16, 24, pp, 5, 6, 8
-IPFILTER_CHROMA 16, 64, pp, 5, 6, 8
-IPFILTER_CHROMA 24, 64, pp, 5, 6, 8
-IPFILTER_CHROMA 32, 48, pp, 5, 6, 8
-IPFILTER_CHROMA 32, 64, pp, 5, 6, 8
-IPFILTER_CHROMA 6, 16, ps, 6, 7, 6
-IPFILTER_CHROMA 8, 12, ps, 6, 7, 6
-IPFILTER_CHROMA 8, 64, ps, 6, 7, 6
-IPFILTER_CHROMA 12, 32, ps, 6, 7, 6
-IPFILTER_CHROMA 16, 24, ps, 6, 7, 6
-IPFILTER_CHROMA 16, 64, ps, 6, 7, 6
-IPFILTER_CHROMA 24, 64, ps, 6, 7, 6
-IPFILTER_CHROMA 32, 48, ps, 6, 7, 6
-IPFILTER_CHROMA 32, 64, ps, 6, 7, 6
-
-IPFILTER_CHROMA 48, 64, pp, 5, 6, 8
-IPFILTER_CHROMA 64, 48, pp, 5, 6, 8
-IPFILTER_CHROMA 64, 64, pp, 5, 6, 8
-IPFILTER_CHROMA 64, 32, pp, 5, 6, 8
-IPFILTER_CHROMA 64, 16, pp, 5, 6, 8
-IPFILTER_CHROMA 48, 64, ps, 6, 7, 6
-IPFILTER_CHROMA 64, 48, ps, 6, 7, 6
-IPFILTER_CHROMA 64, 64, ps, 6, 7, 6
-IPFILTER_CHROMA 64, 32, ps, 6, 7, 6
-IPFILTER_CHROMA 64, 16, ps, 6, 7, 6
-
-
-%macro PROCESS_CHROMA_SP_W4_4R 0
-    movq       m0, [r0]
-    movq       m1, [r0 + r1]
-    punpcklwd  m0, m1                          ;m0=[0 1]
-    pmaddwd    m0, [r6 + 0 *32]                ;m0=[0+1]         Row1
-
-    lea        r0, [r0 + 2 * r1]
-    movq       m4, [r0]
-    punpcklwd  m1, m4                          ;m1=[1 2]
-    pmaddwd    m1, [r6 + 0 *32]                ;m1=[1+2]         Row2
-
-    movq       m5, [r0 + r1]
-    punpcklwd  m4, m5                          ;m4=[2 3]
-    pmaddwd    m2, m4, [r6 + 0 *32]            ;m2=[2+3]         Row3
-    pmaddwd    m4, [r6 + 1 * 32]
-    paddd      m0, m4                          ;m0=[0+1+2+3]     Row1 done
-
-    lea        r0, [r0 + 2 * r1]
-    movq       m4, [r0]
-    punpcklwd  m5, m4                          ;m5=[3 4]
-    pmaddwd    m3, m5, [r6 + 0 *32]            ;m3=[3+4]         Row4
-    pmaddwd    m5, [r6 + 1 * 32]
-    paddd      m1, m5                          ;m1 = [1+2+3+4]   Row2
-
-    movq       m5, [r0 + r1]
-    punpcklwd  m4, m5                          ;m4=[4 5]
-    pmaddwd    m4, [r6 + 1 * 32]
-    paddd      m2, m4                          ;m2=[2+3+4+5]     Row3
-
-    movq       m4, [r0 + 2 * r1]
-    punpcklwd  m5, m4                          ;m5=[5 6]
-    pmaddwd    m5, [r6 + 1 * 32]
-    paddd      m3, m5                          ;m3=[3+4+5+6]     Row4
-%endmacro
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-%macro IPFILTER_CHROMA_avx2_6xN 1
-cglobal interp_4tap_horiz_pp_6x%1, 5,6,8
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    mov             r4d, %1/2
-.loop:
-    vbroadcasti128  m3, [r0]
-    vbroadcasti128  m4, [r0 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, INTERP_SHIFT_PP           ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movq            [r2], xm3
-    pextrd          [r2 + 8], xm3, 2
-
-    vbroadcasti128  m3, [r0 + r1]
-    vbroadcasti128  m4, [r0 + r1 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, INTERP_SHIFT_PP           ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movq            [r2 + r3], xm3
-    pextrd          [r2 + r3 + 8], xm3, 2
-
-    lea             r0, [r0 + r1 * 2]
-    lea             r2, [r2 + r3 * 2]
-    dec             r4d
-    jnz             .loop
-    RET
-%endmacro
-IPFILTER_CHROMA_avx2_6xN 8
-IPFILTER_CHROMA_avx2_6xN 16
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_8x2, 5,6,8
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    vbroadcasti128  m3, [r0]
-    vbroadcasti128  m4, [r0 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, INTERP_SHIFT_PP          ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3,q2020
-    pshufb          xm3, xm6                     ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movu            [r2], xm3
-
-    vbroadcasti128  m3, [r0 + r1]
-    vbroadcasti128  m4, [r0 + r1 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, INTERP_SHIFT_PP           ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3,q2020
-    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movu            [r2 + r3], xm3
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_8x4, 5,6,8
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-%rep 2
-    vbroadcasti128  m3, [r0]
-    vbroadcasti128  m4, [r0 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3,q2020
-    pshufb          xm3, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movu            [r2], xm3
-
-    vbroadcasti128  m3, [r0 + r1]
-    vbroadcasti128  m4, [r0 + r1 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3,q2020
-    pshufb          xm3, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movu            [r2 + r3], xm3
-
-    lea             r0, [r0 + r1 * 2]
-    lea             r2, [r2 + r3 * 2]
-%endrep
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-%macro IPFILTER_CHROMA_avx2_8xN 1
-cglobal interp_4tap_horiz_pp_8x%1, 5,6,8
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    mov             r4d, %1/2
-.loop:
-    vbroadcasti128  m3, [r0]
-    vbroadcasti128  m4, [r0 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6                         ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movu            [r2], xm3
-
-    vbroadcasti128  m3, [r0 + r1]
-    vbroadcasti128  m4, [r0 + r1 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6                         ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6                      ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movu            [r2 + r3], xm3
-
-    lea             r0, [r0 + r1 * 2]
-    lea             r2, [r2 + r3 * 2]
-    dec             r4d
-    jnz             .loop
-    RET
-%endmacro
-IPFILTER_CHROMA_avx2_8xN 6
-IPFILTER_CHROMA_avx2_8xN 8
-IPFILTER_CHROMA_avx2_8xN 12
-IPFILTER_CHROMA_avx2_8xN 16
-IPFILTER_CHROMA_avx2_8xN 32
-IPFILTER_CHROMA_avx2_8xN 64
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-%macro IPFILTER_CHROMA_avx2_16xN 1
-%if ARCH_X86_64
-cglobal interp_4tap_horiz_pp_16x%1, 5,6,9
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    mov             r4d, %1
-.loop:
-    vbroadcasti128  m3, [r0]
-    vbroadcasti128  m4, [r0 + 8]
-
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6                     ; m3 = WORD[7 6 5 4 3 2 1 0]
-
-    vbroadcasti128  m4, [r0 + 16]
-    vbroadcasti128  m8, [r0 + 24]
-
-    pshufb          m4, m1
-    pshufb          m8, m1
-
-    pmaddwd         m4, m0
-    pmaddwd         m8, m0
-    phaddd          m4, m8
-    paddd           m4, m2
-    psrad           m4, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m4, m4
-    vpermq          m4, m4, q2020
-    pshufb          xm4, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
-    vinserti128     m3, m3, xm4, 1
-    CLIPW           m3, m5, m7
-    movu            [r2], m3
-
-    add             r0, r1
-    add             r2, r3
-    dec             r4d
-    jnz             .loop
-    RET
-%endif
-%endmacro
-IPFILTER_CHROMA_avx2_16xN 4
-IPFILTER_CHROMA_avx2_16xN 8
-IPFILTER_CHROMA_avx2_16xN 12
-IPFILTER_CHROMA_avx2_16xN 16
-IPFILTER_CHROMA_avx2_16xN 24
-IPFILTER_CHROMA_avx2_16xN 32
-IPFILTER_CHROMA_avx2_16xN 64
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-%macro IPFILTER_CHROMA_avx2_32xN 1
-%if ARCH_X86_64
-cglobal interp_4tap_horiz_pp_32x%1, 5,6,9
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    mov             r6d, %1
-.loop:
-%assign x 0
-%rep 2
-    vbroadcasti128  m3, [r0 + x]
-    vbroadcasti128  m4, [r0 + 8 + x]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6                     ; m3 = WORD[7 6 5 4 3 2 1 0]
-
-    vbroadcasti128  m4, [r0 + 16 + x]
-    vbroadcasti128  m8, [r0 + 24 + x]
-    pshufb          m4, m1
-    pshufb          m8, m1
-
-    pmaddwd         m4, m0
-    pmaddwd         m8, m0
-    phaddd          m4, m8
-    paddd           m4, m2
-    psrad           m4, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m4, m4
-    vpermq          m4, m4, q2020
-    pshufb          xm4, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
-    vinserti128     m3, m3, xm4, 1
-    CLIPW           m3, m5, m7
-    movu            [r2 + x], m3
-    %assign x x+32
-    %endrep
-
-    add             r0, r1
-    add             r2, r3
-    dec             r6d
-    jnz             .loop
-    RET
-%endif
-%endmacro
-IPFILTER_CHROMA_avx2_32xN 8
-IPFILTER_CHROMA_avx2_32xN 16
-IPFILTER_CHROMA_avx2_32xN 24
-IPFILTER_CHROMA_avx2_32xN 32
-IPFILTER_CHROMA_avx2_32xN 48
-IPFILTER_CHROMA_avx2_32xN 64
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-%macro IPFILTER_CHROMA_avx2_12xN 1
-%if ARCH_X86_64
-cglobal interp_4tap_horiz_pp_12x%1, 5,6,8
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    mov             r4d, %1
-.loop:
-    vbroadcasti128  m3, [r0]
-    vbroadcasti128  m4, [r0 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6                     ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movu            [r2], xm3
-
-    vbroadcasti128  m3, [r0 + 16]
-    vbroadcasti128  m4, [r0 + 24]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6                       ; m3 = DWORD[7 6 3 2 5 4 1 0]
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6                    ; m3 = WORD[7 6 5 4 3 2 1 0]
-    CLIPW           xm3, xm5, xm7
-    movq            [r2 + 16], xm3
-
-    add             r0, r1
-    add             r2, r3
-    dec             r4d
-    jnz             .loop
-    RET
-%endif
-%endmacro
-IPFILTER_CHROMA_avx2_12xN 16
-IPFILTER_CHROMA_avx2_12xN 32
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-%macro IPFILTER_CHROMA_avx2_24xN 1
-%if ARCH_X86_64
-cglobal interp_4tap_horiz_pp_24x%1, 5,6,9
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    mov             r4d, %1
-.loop:
-    vbroadcasti128  m3, [r0]
-    vbroadcasti128  m4, [r0 + 8]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6
-
-    vbroadcasti128  m4, [r0 + 16]
-    vbroadcasti128  m8, [r0 + 24]
-    pshufb          m4, m1
-    pshufb          m8, m1
-
-    pmaddwd         m4, m0
-    pmaddwd         m8, m0
-    phaddd          m4, m8
-    paddd           m4, m2
-    psrad           m4, 6
-
-    packusdw        m3, m4
-    vpermq          m3, m3, q3120
-    pshufb          m3, m6
-    CLIPW           m3, m5, m7
-    movu            [r2], m3
-
-    vbroadcasti128  m3, [r0 + 32]
-    vbroadcasti128  m4, [r0 + 40]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6
-
-    packusdw        m3, m3
-    vpermq          m3, m3, q2020
-    pshufb          xm3, xm6
-    CLIPW           xm3, xm5, xm7
-    movu            [r2 + 32], xm3
-
-    add             r0, r1
-    add             r2, r3
-    dec             r4d
-    jnz             .loop
-    RET
-%endif
-%endmacro
-IPFILTER_CHROMA_avx2_24xN 32
-IPFILTER_CHROMA_avx2_24xN 64
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-%macro IPFILTER_CHROMA_avx2_64xN 1
-%if ARCH_X86_64
-cglobal interp_4tap_horiz_pp_64x%1, 5,6,9
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    mov             r6d, %1
-.loop:
-%assign x 0
-%rep 4
-    vbroadcasti128  m3, [r0 + x]
-    vbroadcasti128  m4, [r0 + 8 + x]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6
-
-    vbroadcasti128  m4, [r0 + 16 + x]
-    vbroadcasti128  m8, [r0 + 24 + x]
-    pshufb          m4, m1
-    pshufb          m8, m1
-
-    pmaddwd         m4, m0
-    pmaddwd         m8, m0
-    phaddd          m4, m8
-    paddd           m4, m2
-    psrad           m4, 6
-
-    packusdw        m3, m4
-    vpermq          m3, m3, q3120
-    pshufb          m3, m6
-    CLIPW           m3, m5, m7
-    movu            [r2 + x], m3
-    %assign x x+32
-    %endrep
-
-    add             r0, r1
-    add             r2, r3
-    dec             r6d
-    jnz             .loop
-    RET
-%endif
-%endmacro
-IPFILTER_CHROMA_avx2_64xN 16
-IPFILTER_CHROMA_avx2_64xN 32
-IPFILTER_CHROMA_avx2_64xN 48
-IPFILTER_CHROMA_avx2_64xN 64
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-%if ARCH_X86_64
-cglobal interp_4tap_horiz_pp_48x64, 5,6,9
-    add             r1d, r1d
-    add             r3d, r3d
-    sub             r0, 2
-    mov             r4d, r4m
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova            m1, [interp8_hpp_shuf]
-    vpbroadcastd    m2, [pd_32]
-    pxor            m5, m5
-    mova            m6, [idct8_shuf2]
-    mova            m7, [pw_pixel_max]
-
-    mov             r4d, 64
-.loop:
-%assign x 0
-%rep 3
-    vbroadcasti128  m3, [r0 + x]
-    vbroadcasti128  m4, [r0 + 8 + x]
-    pshufb          m3, m1
-    pshufb          m4, m1
-
-    pmaddwd         m3, m0
-    pmaddwd         m4, m0
-    phaddd          m3, m4
-    paddd           m3, m2
-    psrad           m3, 6
-
-    vbroadcasti128  m4, [r0 + 16 + x]
-    vbroadcasti128  m8, [r0 + 24 + x]
-    pshufb          m4, m1
-    pshufb          m8, m1
-
-    pmaddwd         m4, m0
-    pmaddwd         m8, m0
-    phaddd          m4, m8
-    paddd           m4, m2
-    psrad           m4, 6
-
-    packusdw        m3, m4
-    vpermq          m3, m3, q3120
-    pshufb          m3, m6
-    CLIPW           m3, m5, m7
-    movu            [r2 + x], m3
-%assign x x+32
-%endrep
-
-    add             r0, r1
-    add             r2, r3
-    dec             r4d
-    jnz             .loop
-    RET
-%endif
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_%3_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SS 4
-INIT_XMM sse2
-cglobal interp_4tap_vert_%3_%1x%2, 5, 7, %4 ,0-gprsize
-
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r6, [r5 + r4]
-%else
-    lea       r6, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       dword [rsp], %2/4
-
-%ifnidn %3, ss
-    %ifnidn %3, ps
-        mova      m7, [pw_pixel_max]
-        %ifidn %3, pp
-            mova      m6, [INTERP_OFFSET_PP]
-        %else
-            mova      m6, [INTERP_OFFSET_SP]
-        %endif
-    %else
-        mova      m6, [INTERP_OFFSET_PS]
-    %endif
-%endif
-
-.loopH:
-    mov       r4d, (%1/4)
-.loopW:
-    PROCESS_CHROMA_SP_W4_4R
-
-%ifidn %3, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%elifidn %3, ps
-    paddd     m0, m6
-    paddd     m1, m6
-    paddd     m2, m6
-    paddd     m3, m6
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%else
-    paddd     m0, m6
-    paddd     m1, m6
-    paddd     m2, m6
-    paddd     m3, m6
-    %ifidn %3, pp
-        psrad     m0, INTERP_SHIFT_PP
-        psrad     m1, INTERP_SHIFT_PP
-        psrad     m2, INTERP_SHIFT_PP
-        psrad     m3, INTERP_SHIFT_PP
-    %else
-        psrad     m0, INTERP_SHIFT_SP
-        psrad     m1, INTERP_SHIFT_SP
-        psrad     m2, INTERP_SHIFT_SP
-        psrad     m3, INTERP_SHIFT_SP
-    %endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    pxor      m5, m5
-    CLIPW2    m0, m2, m5, m7
-%endif
-
-    movh      [r2], m0
-    movhps    [r2 + r3], m0
-    lea       r5, [r2 + 2 * r3]
-    movh      [r5], m2
-    movhps    [r5 + r3], m2
-
-    lea       r5, [4 * r1 - 2 * 4]
-    sub       r0, r5
-    add       r2, 2 * 4
-
-    dec       r4d
-    jnz       .loopW
-
-    lea       r0, [r0 + 4 * r1 - 2 * %1]
-    lea       r2, [r2 + 4 * r3 - 2 * %1]
-
-    dec       dword [rsp]
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SS 4, 4, ss, 6
-    FILTER_VER_CHROMA_SS 4, 8, ss, 6
-    FILTER_VER_CHROMA_SS 16, 16, ss, 6
-    FILTER_VER_CHROMA_SS 16, 8, ss, 6
-    FILTER_VER_CHROMA_SS 16, 12, ss, 6
-    FILTER_VER_CHROMA_SS 12, 16, ss, 6
-    FILTER_VER_CHROMA_SS 16, 4, ss, 6
-    FILTER_VER_CHROMA_SS 4, 16, ss, 6
-    FILTER_VER_CHROMA_SS 32, 32, ss, 6
-    FILTER_VER_CHROMA_SS 32, 16, ss, 6
-    FILTER_VER_CHROMA_SS 16, 32, ss, 6
-    FILTER_VER_CHROMA_SS 32, 24, ss, 6
-    FILTER_VER_CHROMA_SS 24, 32, ss, 6
-    FILTER_VER_CHROMA_SS 32, 8, ss, 6
-
-    FILTER_VER_CHROMA_SS 4, 4, ps, 7
-    FILTER_VER_CHROMA_SS 4, 8, ps, 7
-    FILTER_VER_CHROMA_SS 16, 16, ps, 7
-    FILTER_VER_CHROMA_SS 16, 8, ps, 7
-    FILTER_VER_CHROMA_SS 16, 12, ps, 7
-    FILTER_VER_CHROMA_SS 12, 16, ps, 7
-    FILTER_VER_CHROMA_SS 16, 4, ps, 7
-    FILTER_VER_CHROMA_SS 4, 16, ps, 7
-    FILTER_VER_CHROMA_SS 32, 32, ps, 7
-    FILTER_VER_CHROMA_SS 32, 16, ps, 7
-    FILTER_VER_CHROMA_SS 16, 32, ps, 7
-    FILTER_VER_CHROMA_SS 32, 24, ps, 7
-    FILTER_VER_CHROMA_SS 24, 32, ps, 7
-    FILTER_VER_CHROMA_SS 32, 8, ps, 7
-
-    FILTER_VER_CHROMA_SS 4, 4, sp, 8
-    FILTER_VER_CHROMA_SS 4, 8, sp, 8
-    FILTER_VER_CHROMA_SS 16, 16, sp, 8
-    FILTER_VER_CHROMA_SS 16, 8, sp, 8
-    FILTER_VER_CHROMA_SS 16, 12, sp, 8
-    FILTER_VER_CHROMA_SS 12, 16, sp, 8
-    FILTER_VER_CHROMA_SS 16, 4, sp, 8
-    FILTER_VER_CHROMA_SS 4, 16, sp, 8
-    FILTER_VER_CHROMA_SS 32, 32, sp, 8
-    FILTER_VER_CHROMA_SS 32, 16, sp, 8
-    FILTER_VER_CHROMA_SS 16, 32, sp, 8
-    FILTER_VER_CHROMA_SS 32, 24, sp, 8
-    FILTER_VER_CHROMA_SS 24, 32, sp, 8
-    FILTER_VER_CHROMA_SS 32, 8, sp, 8
-
-    FILTER_VER_CHROMA_SS 4, 4, pp, 8
-    FILTER_VER_CHROMA_SS 4, 8, pp, 8
-    FILTER_VER_CHROMA_SS 16, 16, pp, 8
-    FILTER_VER_CHROMA_SS 16, 8, pp, 8
-    FILTER_VER_CHROMA_SS 16, 12, pp, 8
-    FILTER_VER_CHROMA_SS 12, 16, pp, 8
-    FILTER_VER_CHROMA_SS 16, 4, pp, 8
-    FILTER_VER_CHROMA_SS 4, 16, pp, 8
-    FILTER_VER_CHROMA_SS 32, 32, pp, 8
-    FILTER_VER_CHROMA_SS 32, 16, pp, 8
-    FILTER_VER_CHROMA_SS 16, 32, pp, 8
-    FILTER_VER_CHROMA_SS 32, 24, pp, 8
-    FILTER_VER_CHROMA_SS 24, 32, pp, 8
-    FILTER_VER_CHROMA_SS 32, 8, pp, 8
-
-
-    FILTER_VER_CHROMA_SS 16, 24, ss, 6
-    FILTER_VER_CHROMA_SS 12, 32, ss, 6
-    FILTER_VER_CHROMA_SS 4, 32, ss, 6
-    FILTER_VER_CHROMA_SS 32, 64, ss, 6
-    FILTER_VER_CHROMA_SS 16, 64, ss, 6
-    FILTER_VER_CHROMA_SS 32, 48, ss, 6
-    FILTER_VER_CHROMA_SS 24, 64, ss, 6
-
-    FILTER_VER_CHROMA_SS 16, 24, ps, 7
-    FILTER_VER_CHROMA_SS 12, 32, ps, 7
-    FILTER_VER_CHROMA_SS 4, 32, ps, 7
-    FILTER_VER_CHROMA_SS 32, 64, ps, 7
-    FILTER_VER_CHROMA_SS 16, 64, ps, 7
-    FILTER_VER_CHROMA_SS 32, 48, ps, 7
-    FILTER_VER_CHROMA_SS 24, 64, ps, 7
-
-    FILTER_VER_CHROMA_SS 16, 24, sp, 8
-    FILTER_VER_CHROMA_SS 12, 32, sp, 8
-    FILTER_VER_CHROMA_SS 4, 32, sp, 8
-    FILTER_VER_CHROMA_SS 32, 64, sp, 8
-    FILTER_VER_CHROMA_SS 16, 64, sp, 8
-    FILTER_VER_CHROMA_SS 32, 48, sp, 8
-    FILTER_VER_CHROMA_SS 24, 64, sp, 8
-
-    FILTER_VER_CHROMA_SS 16, 24, pp, 8
-    FILTER_VER_CHROMA_SS 12, 32, pp, 8
-    FILTER_VER_CHROMA_SS 4, 32, pp, 8
-    FILTER_VER_CHROMA_SS 32, 64, pp, 8
-    FILTER_VER_CHROMA_SS 16, 64, pp, 8
-    FILTER_VER_CHROMA_SS 32, 48, pp, 8
-    FILTER_VER_CHROMA_SS 24, 64, pp, 8
-
-
-    FILTER_VER_CHROMA_SS 48, 64, ss, 6
-    FILTER_VER_CHROMA_SS 64, 48, ss, 6
-    FILTER_VER_CHROMA_SS 64, 64, ss, 6
-    FILTER_VER_CHROMA_SS 64, 32, ss, 6
-    FILTER_VER_CHROMA_SS 64, 16, ss, 6
-
-    FILTER_VER_CHROMA_SS 48, 64, ps, 7
-    FILTER_VER_CHROMA_SS 64, 48, ps, 7
-    FILTER_VER_CHROMA_SS 64, 64, ps, 7
-    FILTER_VER_CHROMA_SS 64, 32, ps, 7
-    FILTER_VER_CHROMA_SS 64, 16, ps, 7
-
-    FILTER_VER_CHROMA_SS 48, 64, sp, 8
-    FILTER_VER_CHROMA_SS 64, 48, sp, 8
-    FILTER_VER_CHROMA_SS 64, 64, sp, 8
-    FILTER_VER_CHROMA_SS 64, 32, sp, 8
-    FILTER_VER_CHROMA_SS 64, 16, sp, 8
-
-    FILTER_VER_CHROMA_SS 48, 64, pp, 8
-    FILTER_VER_CHROMA_SS 64, 48, pp, 8
-    FILTER_VER_CHROMA_SS 64, 64, pp, 8
-    FILTER_VER_CHROMA_SS 64, 32, pp, 8
-    FILTER_VER_CHROMA_SS 64, 16, pp, 8
-
-
-%macro PROCESS_CHROMA_SP_W2_4R 1
-    movd       m0, [r0]
-    movd       m1, [r0 + r1]
-    punpcklwd  m0, m1                          ;m0=[0 1]
-
-    lea        r0, [r0 + 2 * r1]
-    movd       m2, [r0]
-    punpcklwd  m1, m2                          ;m1=[1 2]
-    punpcklqdq m0, m1                          ;m0=[0 1 1 2]
-    pmaddwd    m0, [%1 + 0 *32]                ;m0=[0+1 1+2] Row 1-2
-
-    movd       m1, [r0 + r1]
-    punpcklwd  m2, m1                          ;m2=[2 3]
-
-    lea        r0, [r0 + 2 * r1]
-    movd       m3, [r0]
-    punpcklwd  m1, m3                          ;m2=[3 4]
-    punpcklqdq m2, m1                          ;m2=[2 3 3 4]
-
-    pmaddwd    m4, m2, [%1 + 1 * 32]           ;m4=[2+3 3+4] Row 1-2
-    pmaddwd    m2, [%1 + 0 * 32]               ;m2=[2+3 3+4] Row 3-4
-    paddd      m0, m4                          ;m0=[0+1+2+3 1+2+3+4] Row 1-2
-
-    movd       m1, [r0 + r1]
-    punpcklwd  m3, m1                          ;m3=[4 5]
-
-    movd       m4, [r0 + 2 * r1]
-    punpcklwd  m1, m4                          ;m1=[5 6]
-    punpcklqdq m3, m1                          ;m2=[4 5 5 6]
-    pmaddwd    m3, [%1 + 1 * 32]               ;m3=[4+5 5+6] Row 3-4
-    paddd      m2, m3                          ;m2=[2+3+4+5 3+4+5+6] Row 3-4
-%endmacro
-
-;---------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vertical_%2_2x%1(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W2 3
-INIT_XMM sse4
-cglobal interp_4tap_vert_%2_2x%1, 5, 6, %3
-
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       r4d, (%1/4)
-%ifnidn %2, ss
-    %ifnidn %2, ps
-        pxor      m7, m7
-        mova      m6, [pw_pixel_max]
-        %ifidn %2, pp
-            mova      m5, [INTERP_OFFSET_PP]
-        %else
-            mova      m5, [INTERP_OFFSET_SP]
-        %endif
-    %else
-        mova      m5, [INTERP_OFFSET_PS]
-    %endif
-%endif
-
-.loopH:
-    PROCESS_CHROMA_SP_W2_4R r5
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m2, 6
-    packssdw  m0, m2
-%elifidn %2, ps
-    paddd     m0, m5
-    paddd     m2, m5
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    packssdw  m0, m2
-%else
-    paddd     m0, m5
-    paddd     m2, m5
-    %ifidn %2, pp
-        psrad     m0, INTERP_SHIFT_PP
-        psrad     m2, INTERP_SHIFT_PP
-    %else
-        psrad     m0, INTERP_SHIFT_SP
-        psrad     m2, INTERP_SHIFT_SP
-    %endif
-    packusdw  m0, m2
-    CLIPW     m0, m7,    m6
-%endif
-
-    movd      [r2], m0
-    pextrd    [r2 + r3], m0, 1
-    lea       r2, [r2 + 2 * r3]
-    pextrd    [r2], m0, 2
-    pextrd    [r2 + r3], m0, 3
-
-    lea       r2, [r2 + 2 * r3]
-
-    dec       r4d
-    jnz       .loopH
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_W2 4, ss, 5
-FILTER_VER_CHROMA_W2 8, ss, 5
-
-FILTER_VER_CHROMA_W2 4, pp, 8
-FILTER_VER_CHROMA_W2 8, pp, 8
-
-FILTER_VER_CHROMA_W2 4, ps, 6
-FILTER_VER_CHROMA_W2 8, ps, 6
-
-FILTER_VER_CHROMA_W2 4, sp, 8
-FILTER_VER_CHROMA_W2 8, sp, 8
-
-FILTER_VER_CHROMA_W2 16, ss, 5
-FILTER_VER_CHROMA_W2 16, pp, 8
-FILTER_VER_CHROMA_W2 16, ps, 6
-FILTER_VER_CHROMA_W2 16, sp, 8
-
-
-;---------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_%1_4x2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W4 3
-INIT_XMM sse4
-cglobal interp_4tap_vert_%2_4x%1, 5, 6, %3
-    add        r1d, r1d
-    add        r3d, r3d
-    sub        r0, r1
-    shl        r4d, 6
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeffV]
-    lea        r5, [r5 + r4]
-%else
-    lea        r5, [tab_ChromaCoeffV + r4]
-%endif
-
-%ifnidn %2, 2
-    mov        r4d, %1/2
-%endif
-
-%ifnidn %2, ss
-    %ifnidn %2, ps
-        pxor      m6, m6
-        mova      m5, [pw_pixel_max]
-        %ifidn %2, pp
-            mova      m4, [INTERP_OFFSET_PP]
-        %else
-            mova      m4, [INTERP_OFFSET_SP]
-        %endif
-    %else
-        mova      m4, [INTERP_OFFSET_PS]
-    %endif
-%endif
-
-%ifnidn %2, 2
-.loop:
-%endif
-
-    movh       m0, [r0]
-    movh       m1, [r0 + r1]
-    punpcklwd  m0, m1                          ;m0=[0 1]
-    pmaddwd    m0, [r5 + 0 *32]                ;m0=[0+1]  Row1
-
-    lea        r0, [r0 + 2 * r1]
-    movh       m2, [r0]
-    punpcklwd  m1, m2                          ;m1=[1 2]
-    pmaddwd    m1, [r5 + 0 *32]                ;m1=[1+2]  Row2
-
-    movh       m3, [r0 + r1]
-    punpcklwd  m2, m3                          ;m4=[2 3]
-    pmaddwd    m2, [r5 + 1 * 32]
-    paddd      m0, m2                          ;m0=[0+1+2+3]  Row1 done
-
-    movh       m2, [r0 + 2 * r1]
-    punpcklwd  m3, m2                          ;m5=[3 4]
-    pmaddwd    m3, [r5 + 1 * 32]
-    paddd      m1, m3                          ;m1=[1+2+3+4]  Row2 done
-
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    packssdw  m0, m1
-%elifidn %2, ps
-    paddd     m0, m4
-    paddd     m1, m4
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    packssdw  m0, m1
-%else
-    paddd     m0, m4
-    paddd     m1, m4
-    %ifidn %2, pp
-        psrad     m0, INTERP_SHIFT_PP
-        psrad     m1, INTERP_SHIFT_PP
-    %else
-        psrad     m0, INTERP_SHIFT_SP
-        psrad     m1, INTERP_SHIFT_SP
-    %endif
-    packusdw  m0, m1
-    CLIPW     m0, m6,    m5
-%endif
-
-    movh       [r2], m0
-    movhps     [r2 + r3], m0
-
-%ifnidn %2, 2
-    lea        r2, [r2 + r3 * 2]
-    dec        r4d
-    jnz        .loop
-%endif
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_W4 2, ss, 4
-FILTER_VER_CHROMA_W4 2, pp, 7
-FILTER_VER_CHROMA_W4 2, ps, 5
-FILTER_VER_CHROMA_W4 2, sp, 7
-
-FILTER_VER_CHROMA_W4 4, ss, 4
-FILTER_VER_CHROMA_W4 4, pp, 7
-FILTER_VER_CHROMA_W4 4, ps, 5
-FILTER_VER_CHROMA_W4 4, sp, 7
-
-;-------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vertical_%1_6x8(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-------------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W6 3
-INIT_XMM sse4
-cglobal interp_4tap_vert_%2_6x%1, 5, 7, %3
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r6, [r5 + r4]
-%else
-    lea       r6, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       r4d, %1/4
-
-%ifnidn %2, ss
-    %ifnidn %2, ps
-        mova      m7, [pw_pixel_max]
-        %ifidn %2, pp
-            mova      m6, [INTERP_OFFSET_PP]
-        %else
-            mova      m6, [INTERP_OFFSET_SP]
-        %endif
-    %else
-        mova      m6, [INTERP_OFFSET_PS]
-    %endif
-%endif
-
-.loopH:
-    PROCESS_CHROMA_SP_W4_4R
-
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%elifidn %2, ps
-    paddd     m0, m6
-    paddd     m1, m6
-    paddd     m2, m6
-    paddd     m3, m6
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%else
-    paddd     m0, m6
-    paddd     m1, m6
-    paddd     m2, m6
-    paddd     m3, m6
-    %ifidn %2, pp
-        psrad     m0, INTERP_SHIFT_PP
-        psrad     m1, INTERP_SHIFT_PP
-        psrad     m2, INTERP_SHIFT_PP
-        psrad     m3, INTERP_SHIFT_PP
-    %else
-        psrad     m0, INTERP_SHIFT_SP
-        psrad     m1, INTERP_SHIFT_SP
-        psrad     m2, INTERP_SHIFT_SP
-        psrad     m3, INTERP_SHIFT_SP
-    %endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    pxor      m5, m5
-    CLIPW2    m0, m2, m5, m7
-%endif
-
-    movh      [r2], m0
-    movhps    [r2 + r3], m0
-    lea       r5, [r2 + 2 * r3]
-    movh      [r5], m2
-    movhps    [r5 + r3], m2
-
-    lea       r5, [4 * r1 - 2 * 4]
-    sub       r0, r5
-    add       r2, 2 * 4
-
-    PROCESS_CHROMA_SP_W2_4R r6
-
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m2, 6
-    packssdw  m0, m2
-%elifidn %2, ps
-    paddd     m0, m6
-    paddd     m2, m6
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    packssdw  m0, m2
-%else
-    paddd     m0, m6
-    paddd     m2, m6
-    %ifidn %2, pp
-        psrad     m0, INTERP_SHIFT_PP
-        psrad     m2, INTERP_SHIFT_PP
-    %else
-        psrad     m0, INTERP_SHIFT_SP
-        psrad     m2, INTERP_SHIFT_SP
-    %endif
-    packusdw  m0, m2
-    CLIPW     m0, m5,    m7
-%endif
-
-    movd      [r2], m0
-    pextrd    [r2 + r3], m0, 1
-    lea       r2, [r2 + 2 * r3]
-    pextrd    [r2], m0, 2
-    pextrd    [r2 + r3], m0, 3
-
-    sub       r0, 2 * 4
-    lea       r2, [r2 + 2 * r3 - 2 * 4]
-
-    dec       r4d
-    jnz       .loopH
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_W6 8, ss, 6
-FILTER_VER_CHROMA_W6 8, ps, 7
-FILTER_VER_CHROMA_W6 8, sp, 8
-FILTER_VER_CHROMA_W6 8, pp, 8
-
-FILTER_VER_CHROMA_W6 16, ss, 6
-FILTER_VER_CHROMA_W6 16, ps, 7
-FILTER_VER_CHROMA_W6 16, sp, 8
-FILTER_VER_CHROMA_W6 16, pp, 8
-
-%macro PROCESS_CHROMA_SP_W8_2R 0
-    movu       m1, [r0]
-    movu       m3, [r0 + r1]
-    punpcklwd  m0, m1, m3
-    pmaddwd    m0, [r5 + 0 * 32]                ;m0 = [0l+1l]  Row1l
-    punpckhwd  m1, m3
-    pmaddwd    m1, [r5 + 0 * 32]                ;m1 = [0h+1h]  Row1h
-
-    movu       m4, [r0 + 2 * r1]
-    punpcklwd  m2, m3, m4
-    pmaddwd    m2, [r5 + 0 * 32]                ;m2 = [1l+2l]  Row2l
-    punpckhwd  m3, m4
-    pmaddwd    m3, [r5 + 0 * 32]                ;m3 = [1h+2h]  Row2h
-
-    lea        r0, [r0 + 2 * r1]
-    movu       m5, [r0 + r1]
-    punpcklwd  m6, m4, m5
-    pmaddwd    m6, [r5 + 1 * 32]                ;m6 = [2l+3l]  Row1l
-    paddd      m0, m6                           ;m0 = [0l+1l+2l+3l]  Row1l sum
-    punpckhwd  m4, m5
-    pmaddwd    m4, [r5 + 1 * 32]                ;m6 = [2h+3h]  Row1h
-    paddd      m1, m4                           ;m1 = [0h+1h+2h+3h]  Row1h sum
-
-    movu       m4, [r0 + 2 * r1]
-    punpcklwd  m6, m5, m4
-    pmaddwd    m6, [r5 + 1 * 32]                ;m6 = [3l+4l]  Row2l
-    paddd      m2, m6                           ;m2 = [1l+2l+3l+4l]  Row2l sum
-    punpckhwd  m5, m4
-    pmaddwd    m5, [r5 + 1 * 32]                ;m1 = [3h+4h]  Row2h
-    paddd      m3, m5                           ;m3 = [1h+2h+3h+4h]  Row2h sum
-%endmacro
-
-;----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_%3_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W8 4
-INIT_XMM sse2
-cglobal interp_4tap_vert_%3_%1x%2, 5, 6, %4
-
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       r4d, %2/2
-
-%ifidn %3, pp
-    mova      m7, [INTERP_OFFSET_PP]
-%elifidn %3, sp
-    mova      m7, [INTERP_OFFSET_SP]
-%elifidn %3, ps
-    mova      m7, [INTERP_OFFSET_PS]
-%endif
-
-.loopH:
-    PROCESS_CHROMA_SP_W8_2R
-
-%ifidn %3, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%elifidn %3, ps
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%else
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    %ifidn %3, pp
-        psrad     m0, INTERP_SHIFT_PP
-        psrad     m1, INTERP_SHIFT_PP
-        psrad     m2, INTERP_SHIFT_PP
-        psrad     m3, INTERP_SHIFT_PP
-    %else
-        psrad     m0, INTERP_SHIFT_SP
-        psrad     m1, INTERP_SHIFT_SP
-        psrad     m2, INTERP_SHIFT_SP
-        psrad     m3, INTERP_SHIFT_SP
-    %endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    pxor      m5, m5
-    mova      m6, [pw_pixel_max]
-    CLIPW2    m0, m2, m5, m6
-%endif
-
-    movu      [r2], m0
-    movu      [r2 + r3], m2
-
-    lea       r2, [r2 + 2 * r3]
-
-    dec       r4d
-    jnz       .loopH
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_W8 8, 2, ss, 7
-FILTER_VER_CHROMA_W8 8, 4, ss, 7
-FILTER_VER_CHROMA_W8 8, 6, ss, 7
-FILTER_VER_CHROMA_W8 8, 8, ss, 7
-FILTER_VER_CHROMA_W8 8, 16, ss, 7
-FILTER_VER_CHROMA_W8 8, 32, ss, 7
-
-FILTER_VER_CHROMA_W8 8, 2, sp, 8
-FILTER_VER_CHROMA_W8 8, 4, sp, 8
-FILTER_VER_CHROMA_W8 8, 6, sp, 8
-FILTER_VER_CHROMA_W8 8, 8, sp, 8
-FILTER_VER_CHROMA_W8 8, 16, sp, 8
-FILTER_VER_CHROMA_W8 8, 32, sp, 8
-
-FILTER_VER_CHROMA_W8 8, 2, ps, 8
-FILTER_VER_CHROMA_W8 8, 4, ps, 8
-FILTER_VER_CHROMA_W8 8, 6, ps, 8
-FILTER_VER_CHROMA_W8 8, 8, ps, 8
-FILTER_VER_CHROMA_W8 8, 16, ps, 8
-FILTER_VER_CHROMA_W8 8, 32, ps, 8
-
-FILTER_VER_CHROMA_W8 8, 2, pp, 8
-FILTER_VER_CHROMA_W8 8, 4, pp, 8
-FILTER_VER_CHROMA_W8 8, 6, pp, 8
-FILTER_VER_CHROMA_W8 8, 8, pp, 8
-FILTER_VER_CHROMA_W8 8, 16, pp, 8
-FILTER_VER_CHROMA_W8 8, 32, pp, 8
-
-FILTER_VER_CHROMA_W8 8, 12, ss, 7
-FILTER_VER_CHROMA_W8 8, 64, ss, 7
-FILTER_VER_CHROMA_W8 8, 12, sp, 8
-FILTER_VER_CHROMA_W8 8, 64, sp, 8
-FILTER_VER_CHROMA_W8 8, 12, ps, 8
-FILTER_VER_CHROMA_W8 8, 64, ps, 8
-FILTER_VER_CHROMA_W8 8, 12, pp, 8
-FILTER_VER_CHROMA_W8 8, 64, pp, 8
-
-%macro PROCESS_CHROMA_VERT_W16_2R 0
-    movu       m1, [r0]
-    movu       m3, [r0 + r1]
-    punpcklwd  m0, m1, m3
-    pmaddwd    m0, [r5 + 0 * 32]
-    punpckhwd  m1, m3
-    pmaddwd    m1, [r5 + 0 * 32]
-
-    movu       m4, [r0 + 2 * r1]
-    punpcklwd  m2, m3, m4
-    pmaddwd    m2, [r5 + 0 * 32]
-    punpckhwd  m3, m4
-    pmaddwd    m3, [r5 + 0 * 32]
-
-    lea        r0, [r0 + 2 * r1]
-    movu       m5, [r0 + r1]
-    punpcklwd  m6, m4, m5
-    pmaddwd    m6, [r5 + 1 * 32]
-    paddd      m0, m6
-    punpckhwd  m4, m5
-    pmaddwd    m4, [r5 + 1 * 32]
-    paddd      m1, m4
-
-    movu       m4, [r0 + 2 * r1]
-    punpcklwd  m6, m5, m4
-    pmaddwd    m6, [r5 + 1 * 32]
-    paddd      m2, m6
-    punpckhwd  m5, m4
-    pmaddwd    m5, [r5 + 1 * 32]
-    paddd      m3, m5
-%endmacro
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_AVX2_6xN 2
-INIT_YMM avx2
-%if ARCH_X86_64
-cglobal interp_4tap_vert_%2_6x%1, 4, 7, 10
-    mov             r4d, r4m
-    add             r1d, r1d
-    add             r3d, r3d
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    sub             r0, r1
-    mov             r6d, %1/4
-
-%ifidn %2,pp
-    vbroadcasti128  m8, [INTERP_OFFSET_PP]
-%elifidn %2, sp
-    vbroadcasti128  m8, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m8, [INTERP_OFFSET_PS]
-%endif
-
-.loopH:
-    movu            xm0, [r0]
-    movu            xm1, [r0 + r1]
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-
-    movu            xm2, [r0 + r1 * 2]
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-
-    lea             r4, [r1 * 3]
-    movu            xm3, [r0 + r4]
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m4
-
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    pmaddwd         m3, [r5]
-    paddd           m1, m5
-
-    movu            xm5, [r0 + r1]
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    pmaddwd         m4, [r5]
-    paddd           m2, m6
-
-    movu            xm6, [r0 + r1 * 2]
-    punpckhwd       xm7, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddwd         m7, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m7
-    lea             r4, [r3 * 3]
-%ifidn %2,ss
-    psrad           m0, 6
-    psrad           m1, 6
-    psrad           m2, 6
-    psrad           m3, 6
-%else
-    paddd           m0, m8
-    paddd           m1, m8
-    paddd           m2, m8
-    paddd           m3, m8
-%ifidn %2,pp
-    psrad           m0, INTERP_SHIFT_PP
-    psrad           m1, INTERP_SHIFT_PP
-    psrad           m2, INTERP_SHIFT_PP
-    psrad           m3, INTERP_SHIFT_PP
-%elifidn %2, sp
-    psrad           m0, INTERP_SHIFT_SP
-    psrad           m1, INTERP_SHIFT_SP
-    psrad           m2, INTERP_SHIFT_SP
-    psrad           m3, INTERP_SHIFT_SP
-%else
-    psrad           m0, INTERP_SHIFT_PS
-    psrad           m1, INTERP_SHIFT_PS
-    psrad           m2, INTERP_SHIFT_PS
-    psrad           m3, INTERP_SHIFT_PS
-%endif
-%endif
-
-    packssdw        m0, m1
-    packssdw        m2, m3
-    vpermq          m0, m0, q3120
-    vpermq          m2, m2, q3120
-    pxor            m5, m5
-    mova            m9, [pw_pixel_max]
-%ifidn %2,pp
-    CLIPW           m0, m5, m9
-    CLIPW           m2, m5, m9
-%elifidn %2, sp
-    CLIPW           m0, m5, m9
-    CLIPW           m2, m5, m9
-%endif
-
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m2, 1
-    movq            [r2], xm0
-    pextrd          [r2 + 8], xm0, 2
-    movq            [r2 + r3], xm1
-    pextrd          [r2 + r3 + 8], xm1, 2
-    movq            [r2 + r3 * 2], xm2
-    pextrd          [r2 + r3 * 2 + 8], xm2, 2
-    movq            [r2 + r4], xm3
-    pextrd          [r2 + r4 + 8], xm3, 2
-
-    lea             r2, [r2 + r3 * 4]
-    dec r6d
-    jnz .loopH
-    RET
-%endif
-%endmacro
-FILTER_VER_CHROMA_AVX2_6xN 8, pp
-FILTER_VER_CHROMA_AVX2_6xN 8, ps
-FILTER_VER_CHROMA_AVX2_6xN 8, ss
-FILTER_VER_CHROMA_AVX2_6xN 8, sp
-FILTER_VER_CHROMA_AVX2_6xN 16, pp
-FILTER_VER_CHROMA_AVX2_6xN 16, ps
-FILTER_VER_CHROMA_AVX2_6xN 16, ss
-FILTER_VER_CHROMA_AVX2_6xN 16, sp
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W16_16xN_avx2 3
-INIT_YMM avx2
-cglobal interp_4tap_vert_%2_16x%1, 5, 6, %3
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       r4d, %1/2
-
-%ifidn %2, pp
-    vbroadcasti128  m7, [INTERP_OFFSET_PP]
-%elifidn %2, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%elifidn %2, ps
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-
-.loopH:
-    PROCESS_CHROMA_VERT_W16_2R
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%elifidn %2, ps
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%else
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
- %ifidn %2, pp
-    psrad     m0, INTERP_SHIFT_PP
-    psrad     m1, INTERP_SHIFT_PP
-    psrad     m2, INTERP_SHIFT_PP
-    psrad     m3, INTERP_SHIFT_PP
-%else
-    psrad     m0, INTERP_SHIFT_SP
-    psrad     m1, INTERP_SHIFT_SP
-    psrad     m2, INTERP_SHIFT_SP
-    psrad     m3, INTERP_SHIFT_SP
-%endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    pxor      m5, m5
-    CLIPW2    m0, m2, m5, [pw_pixel_max]
-%endif
-
-    movu      [r2], m0
-    movu      [r2 + r3], m2
-    lea       r2, [r2 + 2 * r3]
-    dec       r4d
-    jnz       .loopH
-    RET
-%endmacro
-    FILTER_VER_CHROMA_W16_16xN_avx2 4, pp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 8, pp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 12, pp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 24, pp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 16, pp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 32, pp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 64, pp, 8
-
-    FILTER_VER_CHROMA_W16_16xN_avx2 4, ps, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 8, ps, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 12, ps, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 24, ps, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 16, ps, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 32, ps, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 64, ps, 8
-
-    FILTER_VER_CHROMA_W16_16xN_avx2 4, ss, 7
-    FILTER_VER_CHROMA_W16_16xN_avx2 8, ss, 7
-    FILTER_VER_CHROMA_W16_16xN_avx2 12, ss, 7
-    FILTER_VER_CHROMA_W16_16xN_avx2 24, ss, 7
-    FILTER_VER_CHROMA_W16_16xN_avx2 16, ss, 7
-    FILTER_VER_CHROMA_W16_16xN_avx2 32, ss, 7
-    FILTER_VER_CHROMA_W16_16xN_avx2 64, ss, 7
-
-    FILTER_VER_CHROMA_W16_16xN_avx2 4, sp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 8, sp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 12, sp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 24, sp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 16, sp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 32, sp, 8
-    FILTER_VER_CHROMA_W16_16xN_avx2 64, sp, 8
-
-%macro PROCESS_CHROMA_VERT_W32_2R 0
-    movu       m1, [r0]
-    movu       m3, [r0 + r1]
-    punpcklwd  m0, m1, m3
-    pmaddwd    m0, [r5 + 0 * mmsize]
-    punpckhwd  m1, m3
-    pmaddwd    m1, [r5 + 0 * mmsize]
-
-    movu       m9, [r0 + mmsize]
-    movu       m11, [r0 + r1 + mmsize]
-    punpcklwd  m8, m9, m11
-    pmaddwd    m8, [r5 + 0 * mmsize]
-    punpckhwd  m9, m11
-    pmaddwd    m9, [r5 + 0 * mmsize]
-
-    movu       m4, [r0 + 2 * r1]
-    punpcklwd  m2, m3, m4
-    pmaddwd    m2, [r5 + 0 * mmsize]
-    punpckhwd  m3, m4
-    pmaddwd    m3, [r5 + 0 * mmsize]
-
-    movu       m12, [r0 + 2 * r1 + mmsize]
-    punpcklwd  m10, m11, m12
-    pmaddwd    m10, [r5 + 0 * mmsize]
-    punpckhwd  m11, m12
-    pmaddwd    m11, [r5 + 0 * mmsize]
-
-    lea        r6, [r0 + 2 * r1]
-    movu       m5, [r6 + r1]
-    punpcklwd  m6, m4, m5
-    pmaddwd    m6, [r5 + 1 * mmsize]
-    paddd      m0, m6
-    punpckhwd  m4, m5
-    pmaddwd    m4, [r5 + 1 * mmsize]
-    paddd      m1, m4
-
-    movu       m13, [r6 + r1 + mmsize]
-    punpcklwd  m14, m12, m13
-    pmaddwd    m14, [r5 + 1 * mmsize]
-    paddd      m8, m14
-    punpckhwd  m12, m13
-    pmaddwd    m12, [r5 + 1 * mmsize]
-    paddd      m9, m12
-
-    movu       m4, [r6 + 2 * r1]
-    punpcklwd  m6, m5, m4
-    pmaddwd    m6, [r5 + 1 * mmsize]
-    paddd      m2, m6
-    punpckhwd  m5, m4
-    pmaddwd    m5, [r5 + 1 * mmsize]
-    paddd      m3, m5
-
-    movu       m12, [r6 + 2 * r1 + mmsize]
-    punpcklwd  m14, m13, m12
-    pmaddwd    m14, [r5 + 1 * mmsize]
-    paddd      m10, m14
-    punpckhwd  m13, m12
-    pmaddwd    m13, [r5 + 1 * mmsize]
-    paddd      m11, m13
-%endmacro
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W16_32xN_avx2 3
-INIT_YMM avx2
-%if ARCH_X86_64
-cglobal interp_4tap_vert_%2_32x%1, 5, 7, %3
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-    mov       r4d, %1/2
-
-%ifidn %2, pp
-    vbroadcasti128  m7, [INTERP_OFFSET_PP]
-%elifidn %2, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%elifidn %2, ps
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-
-.loopH:
-    PROCESS_CHROMA_VERT_W32_2R
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    psrad     m8, 6
-    psrad     m9, 6
-    psrad     m10, 6
-    psrad     m11, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-    packssdw  m8, m9
-    packssdw  m10, m11
-%elifidn %2, ps
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-    paddd     m8, m7
-    paddd     m9, m7
-    paddd     m10, m7
-    paddd     m11, m7
-    psrad     m8, INTERP_SHIFT_PS
-    psrad     m9, INTERP_SHIFT_PS
-    psrad     m10, INTERP_SHIFT_PS
-    psrad     m11, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-    packssdw  m8, m9
-    packssdw  m10, m11
-%else
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    paddd     m8, m7
-    paddd     m9, m7
-    paddd     m10, m7
-    paddd     m11, m7
- %ifidn %2, pp
-    psrad     m0, INTERP_SHIFT_PP
-    psrad     m1, INTERP_SHIFT_PP
-    psrad     m2, INTERP_SHIFT_PP
-    psrad     m3, INTERP_SHIFT_PP
-    psrad     m8, INTERP_SHIFT_PP
-    psrad     m9, INTERP_SHIFT_PP
-    psrad     m10, INTERP_SHIFT_PP
-    psrad     m11, INTERP_SHIFT_PP
-%else
-    psrad     m0, INTERP_SHIFT_SP
-    psrad     m1, INTERP_SHIFT_SP
-    psrad     m2, INTERP_SHIFT_SP
-    psrad     m3, INTERP_SHIFT_SP
-    psrad     m8, INTERP_SHIFT_SP
-    psrad     m9, INTERP_SHIFT_SP
-    psrad     m10, INTERP_SHIFT_SP
-    psrad     m11, INTERP_SHIFT_SP
-%endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    packssdw  m8, m9
-    packssdw  m10, m11
-    pxor      m5, m5
-    CLIPW2    m0, m2, m5, [pw_pixel_max]
-    CLIPW2    m8, m10, m5, [pw_pixel_max]
-%endif
-
-    movu      [r2], m0
-    movu      [r2 + r3], m2
-    movu      [r2 + mmsize], m8
-    movu      [r2 + r3 + mmsize], m10
-    lea       r2, [r2 + 2 * r3]
-    lea       r0, [r0 + 2 * r1]
-    dec       r4d
-    jnz       .loopH
-    RET
-%endif
-%endmacro
-    FILTER_VER_CHROMA_W16_32xN_avx2 8, pp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 16, pp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 24, pp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 32, pp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 48, pp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 64, pp, 15
-
-    FILTER_VER_CHROMA_W16_32xN_avx2 8, ps, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 16, ps, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 24, ps, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 32, ps, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 48, ps, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 64, ps, 15
-
-    FILTER_VER_CHROMA_W16_32xN_avx2 8, ss, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 16, ss, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 24, ss, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 32, ss, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 48, ss, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 64, ss, 15
-
-    FILTER_VER_CHROMA_W16_32xN_avx2 8, sp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 16, sp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 24, sp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 32, sp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 48, sp, 15
-    FILTER_VER_CHROMA_W16_32xN_avx2 64, sp, 15
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W16_64xN_avx2 3
-INIT_YMM avx2
-cglobal interp_4tap_vert_%2_64x%1, 5, 7, %3
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-    mov       r4d, %1/2
-
-%ifidn %2, pp
-    vbroadcasti128  m7, [INTERP_OFFSET_PP]
-%elifidn %2, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%elifidn %2, ps
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-
-.loopH:
-%assign x 0
-%rep 4
-    movu       m1, [r0 + x]
-    movu       m3, [r0 + r1 + x]
-    movu       m5, [r5 + 0 * mmsize]
-    punpcklwd  m0, m1, m3
-    pmaddwd    m0, m5
-    punpckhwd  m1, m3
-    pmaddwd    m1, m5
-
-    movu       m4, [r0 + 2 * r1 + x]
-    punpcklwd  m2, m3, m4
-    pmaddwd    m2, m5
-    punpckhwd  m3, m4
-    pmaddwd    m3, m5
-
-    lea        r6, [r0 + 2 * r1]
-    movu       m5, [r6 + r1 + x]
-    punpcklwd  m6, m4, m5
-    pmaddwd    m6, [r5 + 1 * mmsize]
-    paddd      m0, m6
-    punpckhwd  m4, m5
-    pmaddwd    m4, [r5 + 1 * mmsize]
-    paddd      m1, m4
-
-    movu       m4, [r6 + 2 * r1 + x]
-    punpcklwd  m6, m5, m4
-    pmaddwd    m6, [r5 + 1 * mmsize]
-    paddd      m2, m6
-    punpckhwd  m5, m4
-    pmaddwd    m5, [r5 + 1 * mmsize]
-    paddd      m3, m5
-
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%elifidn %2, ps
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%else
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-%ifidn %2, pp
-    psrad     m0, INTERP_SHIFT_PP
-    psrad     m1, INTERP_SHIFT_PP
-    psrad     m2, INTERP_SHIFT_PP
-    psrad     m3, INTERP_SHIFT_PP
-%else
-    psrad     m0, INTERP_SHIFT_SP
-    psrad     m1, INTERP_SHIFT_SP
-    psrad     m2, INTERP_SHIFT_SP
-    psrad     m3, INTERP_SHIFT_SP
-%endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    pxor      m5, m5
-    CLIPW2    m0, m2, m5, [pw_pixel_max]
-%endif
-
-    movu      [r2 + x], m0
-    movu      [r2 + r3 + x], m2
-%assign x x+mmsize
-%endrep
-
-    lea       r2, [r2 + 2 * r3]
-    lea       r0, [r0 + 2 * r1]
-    dec       r4d
-    jnz       .loopH
-    RET
-%endmacro
-    FILTER_VER_CHROMA_W16_64xN_avx2 16, ss, 7
-    FILTER_VER_CHROMA_W16_64xN_avx2 32, ss, 7
-    FILTER_VER_CHROMA_W16_64xN_avx2 48, ss, 7
-    FILTER_VER_CHROMA_W16_64xN_avx2 64, ss, 7
-    FILTER_VER_CHROMA_W16_64xN_avx2 16, sp, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 32, sp, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 48, sp, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 64, sp, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 16, ps, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 32, ps, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 48, ps, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 64, ps, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 16, pp, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 32, pp, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 48, pp, 8
-    FILTER_VER_CHROMA_W16_64xN_avx2 64, pp, 8
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W16_12xN_avx2 3
-INIT_YMM avx2
-cglobal interp_4tap_vert_%2_12x%1, 5, 8, %3
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-    mov       r4d, %1/2
-
-%ifidn %2, pp
-    vbroadcasti128  m7, [INTERP_OFFSET_PP]
-%elifidn %2, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%elifidn %2, ps
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-
-.loopH:
-    PROCESS_CHROMA_VERT_W16_2R
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%elifidn %2, ps
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%else
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
- %ifidn %2, pp
-    psrad     m0, INTERP_SHIFT_PP
-    psrad     m1, INTERP_SHIFT_PP
-    psrad     m2, INTERP_SHIFT_PP
-    psrad     m3, INTERP_SHIFT_PP
-%else
-    psrad     m0, INTERP_SHIFT_SP
-    psrad     m1, INTERP_SHIFT_SP
-    psrad     m2, INTERP_SHIFT_SP
-    psrad     m3, INTERP_SHIFT_SP
-%endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    pxor      m5, m5
-    CLIPW2    m0, m2, m5, [pw_pixel_max]
-%endif
-
-    movu      [r2], xm0
-    movu      [r2 + r3], xm2
-    vextracti128 xm0, m0, 1
-    vextracti128 xm2, m2, 1
-    movq      [r2 + 16], xm0
-    movq      [r2 + r3 + 16], xm2
-    lea       r2, [r2 + 2 * r3]
-    dec       r4d
-    jnz       .loopH
-    RET
-%endmacro
-    FILTER_VER_CHROMA_W16_12xN_avx2 16, ss, 7
-    FILTER_VER_CHROMA_W16_12xN_avx2 16, sp, 8
-    FILTER_VER_CHROMA_W16_12xN_avx2 16, ps, 8
-    FILTER_VER_CHROMA_W16_12xN_avx2 16, pp, 8
-    FILTER_VER_CHROMA_W16_12xN_avx2 32, ss, 7
-    FILTER_VER_CHROMA_W16_12xN_avx2 32, sp, 8
-    FILTER_VER_CHROMA_W16_12xN_avx2 32, ps, 8
-    FILTER_VER_CHROMA_W16_12xN_avx2 32, pp, 8
-
-%macro PROCESS_CHROMA_VERT_W24_2R 0
-    movu       m1, [r0]
-    movu       m3, [r0 + r1]
-    punpcklwd  m0, m1, m3
-    pmaddwd    m0, [r5 + 0 * mmsize]
-    punpckhwd  m1, m3
-    pmaddwd    m1, [r5 + 0 * mmsize]
-
-    movu       xm9, [r0 + mmsize]
-    movu       xm11, [r0 + r1 + mmsize]
-    punpcklwd  xm8, xm9, xm11
-    pmaddwd    xm8, [r5 + 0 * mmsize]
-    punpckhwd  xm9, xm11
-    pmaddwd    xm9, [r5 + 0 * mmsize]
-
-    movu       m4, [r0 + 2 * r1]
-    punpcklwd  m2, m3, m4
-    pmaddwd    m2, [r5 + 0 * mmsize]
-    punpckhwd  m3, m4
-    pmaddwd    m3, [r5 + 0 * mmsize]
-
-    movu       xm12, [r0 + 2 * r1 + mmsize]
-    punpcklwd  xm10, xm11, xm12
-    pmaddwd    xm10, [r5 + 0 * mmsize]
-    punpckhwd  xm11, xm12
-    pmaddwd    xm11, [r5 + 0 * mmsize]
-
-    lea        r6, [r0 + 2 * r1]
-    movu       m5, [r6 + r1]
-    punpcklwd  m6, m4, m5
-    pmaddwd    m6, [r5 + 1 * mmsize]
-    paddd      m0, m6
-    punpckhwd  m4, m5
-    pmaddwd    m4, [r5 + 1 * mmsize]
-    paddd      m1, m4
-
-    movu       xm13, [r6 + r1 + mmsize]
-    punpcklwd  xm14, xm12, xm13
-    pmaddwd    xm14, [r5 + 1 * mmsize]
-    paddd      xm8, xm14
-    punpckhwd  xm12, xm13
-    pmaddwd    xm12, [r5 + 1 * mmsize]
-    paddd      xm9, xm12
-
-    movu       m4, [r6 + 2 * r1]
-    punpcklwd  m6, m5, m4
-    pmaddwd    m6, [r5 + 1 * mmsize]
-    paddd      m2, m6
-    punpckhwd  m5, m4
-    pmaddwd    m5, [r5 + 1 * mmsize]
-    paddd      m3, m5
-
-    movu       xm12, [r6 + 2 * r1 + mmsize]
-    punpcklwd  xm14, xm13, xm12
-    pmaddwd    xm14, [r5 + 1 * mmsize]
-    paddd      xm10, xm14
-    punpckhwd  xm13, xm12
-    pmaddwd    xm13, [r5 + 1 * mmsize]
-    paddd      xm11, xm13
-%endmacro
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W16_24xN_avx2 3
-INIT_YMM avx2
-%if ARCH_X86_64
-cglobal interp_4tap_vert_%2_24x%1, 5, 7, %3
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-    mov       r4d, %1/2
-
-%ifidn %2, pp
-    vbroadcasti128  m7, [INTERP_OFFSET_PP]
-%elifidn %2, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%elifidn %2, ps
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-
-.loopH:
-    PROCESS_CHROMA_VERT_W24_2R
-%ifidn %2, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    psrad     m8, 6
-    psrad     m9, 6
-    psrad     m10, 6
-    psrad     m11, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-    packssdw  m8, m9
-    packssdw  m10, m11
-%elifidn %2, ps
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-    paddd     m8, m7
-    paddd     m9, m7
-    paddd     m10, m7
-    paddd     m11, m7
-    psrad     m8, INTERP_SHIFT_PS
-    psrad     m9, INTERP_SHIFT_PS
-    psrad     m10, INTERP_SHIFT_PS
-    psrad     m11, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-    packssdw  m8, m9
-    packssdw  m10, m11
-%else
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    paddd     m8, m7
-    paddd     m9, m7
-    paddd     m10, m7
-    paddd     m11, m7
- %ifidn %2, pp
-    psrad     m0, INTERP_SHIFT_PP
-    psrad     m1, INTERP_SHIFT_PP
-    psrad     m2, INTERP_SHIFT_PP
-    psrad     m3, INTERP_SHIFT_PP
-    psrad     m8, INTERP_SHIFT_PP
-    psrad     m9, INTERP_SHIFT_PP
-    psrad     m10, INTERP_SHIFT_PP
-    psrad     m11, INTERP_SHIFT_PP
-%else
-    psrad     m0, INTERP_SHIFT_SP
-    psrad     m1, INTERP_SHIFT_SP
-    psrad     m2, INTERP_SHIFT_SP
-    psrad     m3, INTERP_SHIFT_SP
-    psrad     m8, INTERP_SHIFT_SP
-    psrad     m9, INTERP_SHIFT_SP
-    psrad     m10, INTERP_SHIFT_SP
-    psrad     m11, INTERP_SHIFT_SP
-%endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    packssdw  m8, m9
-    packssdw  m10, m11
-    pxor      m5, m5
-    CLIPW2    m0, m2, m5, [pw_pixel_max]
-    CLIPW2    m8, m10, m5, [pw_pixel_max]
-%endif
-
-    movu      [r2], m0
-    movu      [r2 + r3], m2
-    movu      [r2 + mmsize], xm8
-    movu      [r2 + r3 + mmsize], xm10
-    lea       r2, [r2 + 2 * r3]
-    lea       r0, [r0 + 2 * r1]
-    dec       r4d
-    jnz       .loopH
-    RET
-%endif
-%endmacro
-    FILTER_VER_CHROMA_W16_24xN_avx2 32, ss, 15
-    FILTER_VER_CHROMA_W16_24xN_avx2 32, sp, 15
-    FILTER_VER_CHROMA_W16_24xN_avx2 32, ps, 15
-    FILTER_VER_CHROMA_W16_24xN_avx2 32, pp, 15
-    FILTER_VER_CHROMA_W16_24xN_avx2 64, ss, 15
-    FILTER_VER_CHROMA_W16_24xN_avx2 64, sp, 15
-    FILTER_VER_CHROMA_W16_24xN_avx2 64, ps, 15
-    FILTER_VER_CHROMA_W16_24xN_avx2 64, pp, 15
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_W16_48x64_avx2 2
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_48x64, 5, 7, %2
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 6
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-    mov       r4d, 32
-
-%ifidn %1, pp
-    vbroadcasti128  m7, [INTERP_OFFSET_PP]
-%elifidn %1, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%elifidn %1, ps
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-
-.loopH:
-%assign x 0
-%rep 3
-    movu       m1, [r0 + x]
-    movu       m3, [r0 + r1 + x]
-    movu       m5, [r5 + 0 * mmsize]
-    punpcklwd  m0, m1, m3
-    pmaddwd    m0, m5
-    punpckhwd  m1, m3
-    pmaddwd    m1, m5
-
-    movu       m4, [r0 + 2 * r1 + x]
-    punpcklwd  m2, m3, m4
-    pmaddwd    m2, m5
-    punpckhwd  m3, m4
-    pmaddwd    m3, m5
-
-    lea        r6, [r0 + 2 * r1]
-    movu       m5, [r6 + r1 + x]
-    punpcklwd  m6, m4, m5
-    pmaddwd    m6, [r5 + 1 * mmsize]
-    paddd      m0, m6
-    punpckhwd  m4, m5
-    pmaddwd    m4, [r5 + 1 * mmsize]
-    paddd      m1, m4
-
-    movu       m4, [r6 + 2 * r1 + x]
-    punpcklwd  m6, m5, m4
-    pmaddwd    m6, [r5 + 1 * mmsize]
-    paddd      m2, m6
-    punpckhwd  m5, m4
-    pmaddwd    m5, [r5 + 1 * mmsize]
-    paddd      m3, m5
-
-%ifidn %1, ss
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%elifidn %1, ps
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-    psrad     m0, INTERP_SHIFT_PS
-    psrad     m1, INTERP_SHIFT_PS
-    psrad     m2, INTERP_SHIFT_PS
-    psrad     m3, INTERP_SHIFT_PS
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-%else
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-%ifidn %1, pp
-    psrad     m0, INTERP_SHIFT_PP
-    psrad     m1, INTERP_SHIFT_PP
-    psrad     m2, INTERP_SHIFT_PP
-    psrad     m3, INTERP_SHIFT_PP
-%else
-    psrad     m0, INTERP_SHIFT_SP
-    psrad     m1, INTERP_SHIFT_SP
-    psrad     m2, INTERP_SHIFT_SP
-    psrad     m3, INTERP_SHIFT_SP
-%endif
-    packssdw  m0, m1
-    packssdw  m2, m3
-    pxor      m5, m5
-    CLIPW2    m0, m2, m5, [pw_pixel_max]
-%endif
-
-    movu      [r2 + x], m0
-    movu      [r2 + r3 + x], m2
-%assign x x+mmsize
-%endrep
-
-    lea       r2, [r2 + 2 * r3]
-    lea       r0, [r0 + 2 * r1]
-    dec       r4d
-    jnz       .loopH
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_W16_48x64_avx2 pp, 8
-    FILTER_VER_CHROMA_W16_48x64_avx2 ps, 8
-    FILTER_VER_CHROMA_W16_48x64_avx2 ss, 7
-    FILTER_VER_CHROMA_W16_48x64_avx2 sp, 8
-
-INIT_XMM sse2
-cglobal chroma_p2s, 3, 7, 3
-    ; load width and height
-    mov         r3d, r3m
-    mov         r4d, r4m
-    add         r1, r1
-
-    ; load constant
-    mova        m2, [tab_c_n8192]
-
-.loopH:
-
-    xor         r5d, r5d
-.loopW:
-    lea         r6, [r0 + r5 * 2]
-
-    movu        m0, [r6]
-    psllw       m0, (14 - BIT_DEPTH)
-    paddw       m0, m2
-
-    movu        m1, [r6 + r1]
-    psllw       m1, (14 - BIT_DEPTH)
-    paddw       m1, m2
-
-    add         r5d, 8
-    cmp         r5d, r3d
-    lea         r6, [r2 + r5 * 2]
-    jg          .width4
-    movu        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
-    movu        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
-    je          .nextH
-    jmp         .loopW
-
-.width4:
-    test        r3d, 4
-    jz          .width2
-    test        r3d, 2
-    movh        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
-    movh        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
-    lea         r6, [r6 + 8]
-    pshufd      m0, m0, 2
-    pshufd      m1, m1, 2
-    jz          .nextH
-
-.width2:
-    movd        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
-    movd        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
-
-.nextH:
-    lea         r0, [r0 + r1 * 2]
-    add         r2, FENC_STRIDE / 2 * 4
-
-    sub         r4d, 2
-    jnz         .loopH
-    RET
 
 %macro PROCESS_LUMA_VER_W4_4R 0
     movq       m0, [r0]
@@ -9103,7 +3517,7 @@
     ; load constant
     mova       m2, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     psllw      m0, (14 - BIT_DEPTH)
@@ -9156,7 +3570,7 @@
     ; load constant
     mova       m1, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     psllw      m0, (14 - BIT_DEPTH)
     psubw      m0, m1
@@ -9277,7 +3691,7 @@
     ; load constant
     mova       m2, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     psllw      m0, (14 - BIT_DEPTH)
@@ -9351,7 +3765,7 @@
     ; load constant
     mova       m2, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     psllw      m0, (14 - BIT_DEPTH)
@@ -9405,7 +3819,7 @@
     ; load constant
     mova       m4, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     movu       m2, [r0 + r1 * 2]
@@ -9510,7 +3924,7 @@
     ; load constant
     mova       m2, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     psllw      m0, (14 - BIT_DEPTH)
@@ -9583,7 +3997,7 @@
     ; load constant
     mova       m4, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     movu       m2, [r0 + r1 * 2]
@@ -9758,7 +4172,7 @@
     ; load constant
     mova       m2, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     psllw      m0, (14 - BIT_DEPTH)
@@ -9869,7 +4283,7 @@
     ; load constant
     mova       m4, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     movu       m2, [r0 + r1 * 2]
@@ -9952,7 +4366,7 @@
     ; load constant
     mova       m2, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + 32]
     psllw      m0, (14 - BIT_DEPTH)
@@ -10017,7 +4431,7 @@
     ; load constant
     mova       m2, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     psllw      m0, (14 - BIT_DEPTH)
@@ -10081,7 +4495,7 @@
     ; load constant
     mova       m4, [pw_2000]
 
-.loop
+.loop:
     movu       m0, [r0]
     movu       m1, [r0 + r1]
     movu       m2, [r0 + r1 * 2]
@@ -10197,2811 +4611,3 @@
     jnz        .loop
     RET
 
-;-----------------------------------------------------------------------------
-; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
-;-----------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal filterPixelToShort_48x64, 3, 7, 4
-    add        r1d, r1d
-    mov        r3d, r3m
-    add        r3d, r3d
-    lea        r4, [r3 * 3]
-    lea        r5, [r1 * 3]
-
-    ; load height
-    mov        r6d, 16
-
-    ; load constant
-    mova       m3, [pw_2000]
-
-.loop
-    movu       m0, [r0]
-    movu       m1, [r0 + 32]
-    movu       m2, [r0 + 64]
-    psllw      m0, (14 - BIT_DEPTH)
-    psllw      m1, (14 - BIT_DEPTH)
-    psllw      m2, (14 - BIT_DEPTH)
-    psubw      m0, m3
-    psubw      m1, m3
-    psubw      m2, m3
-    movu       [r2 + r3 * 0], m0
-    movu       [r2 + r3 * 0 + 32], m1
-    movu       [r2 + r3 * 0 + 64], m2
-
-    movu       m0, [r0 + r1]
-    movu       m1, [r0 + r1 + 32]
-    movu       m2, [r0 + r1 + 64]
-    psllw      m0, (14 - BIT_DEPTH)
-    psllw      m1, (14 - BIT_DEPTH)
-    psllw      m2, (14 - BIT_DEPTH)
-    psubw      m0, m3
-    psubw      m1, m3
-    psubw      m2, m3
-    movu       [r2 + r3 * 1], m0
-    movu       [r2 + r3 * 1 + 32], m1
-    movu       [r2 + r3 * 1 + 64], m2
-
-    movu       m0, [r0 + r1 * 2]
-    movu       m1, [r0 + r1 * 2 + 32]
-    movu       m2, [r0 + r1 * 2 + 64]
-    psllw      m0, (14 - BIT_DEPTH)
-    psllw      m1, (14 - BIT_DEPTH)
-    psllw      m2, (14 - BIT_DEPTH)
-    psubw      m0, m3
-    psubw      m1, m3
-    psubw      m2, m3
-    movu       [r2 + r3 * 2], m0
-    movu       [r2 + r3 * 2 + 32], m1
-    movu       [r2 + r3 * 2 + 64], m2
-
-    movu       m0, [r0 + r5]
-    movu       m1, [r0 + r5 + 32]
-    movu       m2, [r0 + r5 + 64]
-    psllw      m0, (14 - BIT_DEPTH)
-    psllw      m1, (14 - BIT_DEPTH)
-    psllw      m2, (14 - BIT_DEPTH)
-    psubw      m0, m3
-    psubw      m1, m3
-    psubw      m2, m3
-    movu       [r2 + r4], m0
-    movu       [r2 + r4 + 32], m1
-    movu       [r2 + r4 + 64], m2
-
-    lea        r0, [r0 + r1 * 4]
-    lea        r2, [r2 + r3 * 4]
-
-    dec        r6d
-    jnz        .loop
-    RET
-
-
-;-----------------------------------------------------------------------------------------------------------------------------
-;void interp_horiz_ps_c(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-
-%macro IPFILTER_LUMA_PS_4xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_ps_4x%1, 6,8,7
-    mov                         r5d,               r5m
-    mov                         r4d,               r4m
-    add                         r1d,               r1d
-    add                         r3d,               r3d
-
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    lea                         r4,                [r4 * 8]
-    vbroadcasti128              m0,                [r6 + r4 * 2]
-%else
-    lea                         r4,                [r4 * 8]
-    vbroadcasti128              m0,                [tab_LumaCoeff + r4 * 2]
-%endif
-
-    vbroadcasti128              m2,                [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - pw_2000
-
-    sub                         r0,                6
-    test                        r5d,               r5d
-    mov                         r7d,               %1                                    ; loop count variable - height
-    jz                         .preloop
-    lea                         r6,                [r1 * 3]                              ; r6 = (N / 2 - 1) * srcStride
-    sub                         r0,                r6                                    ; r0(src) - 3 * srcStride
-    add                         r7d,               6                                     ;7 - 1(since last row not in loop)                            ; need extra 7 rows, just set a specially flag here, blkheight += N - 1  (7 - 3 = 4 ; since the last three rows not in loop)
-
-.preloop:
-    lea                         r6,                [r3 * 3]
-.loop
-    ; Row 0
-    movu                        xm3,                [r0]                                 ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    movu                        xm4,                [r0 + 2]                             ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    vinserti128                 m3,                 m3,                xm4,       1
-    movu                        xm4,                [r0 + 4]
-    movu                        xm5,                [r0 + 6]
-    vinserti128                 m4,                 m4,                xm5,       1
-    pmaddwd                     m3,                m0
-    pmaddwd                     m4,                m0
-    phaddd                      m3,                m4                                    ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
-
-    ; Row 1
-    movu                        xm4,                [r0 + r1]                            ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    movu                        xm5,                [r0 + r1 + 2]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    vinserti128                 m4,                 m4,                xm5,       1
-    movu                        xm5,                [r0 + r1 + 4]
-    movu                        xm6,                [r0 + r1 + 6]
-    vinserti128                 m5,                 m5,                xm6,       1
-    pmaddwd                     m4,                m0
-    pmaddwd                     m5,                m0
-    phaddd                      m4,                m5                                     ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
-    phaddd                      m3,                m4                                     ; all rows and col completed.
-
-    mova                        m5,                [interp8_hps_shuf]
-    vpermd                      m3,                m5,                  m3
-    paddd                       m3,                m2
-    vextracti128                xm4,               m3,                  1
-    psrad                       xm3,               INTERP_SHIFT_PS
-    psrad                       xm4,               INTERP_SHIFT_PS
-    packssdw                    xm3,               xm3
-    packssdw                    xm4,               xm4
-
-    movq                        [r2],              xm3                                   ;row 0
-    movq                        [r2 + r3],         xm4                                   ;row 1
-    lea                         r0,                [r0 + r1 * 2]                         ; first loop src ->5th row(i.e 4)
-    lea                         r2,                [r2 + r3 * 2]                         ; first loop dst ->5th row(i.e 4)
-
-    sub                         r7d,               2
-    jg                          .loop
-    test                        r5d,               r5d
-    jz                          .end
-
-    ; Row 10
-    movu                        xm3,                [r0]
-    movu                        xm4,                [r0 + 2]
-    vinserti128                 m3,                 m3,                 xm4,      1
-    movu                        xm4,                [r0 + 4]
-    movu                        xm5,                [r0 + 6]
-    vinserti128                 m4,                 m4,                 xm5,      1
-    pmaddwd                     m3,                m0
-    pmaddwd                     m4,                m0
-    phaddd                      m3,                m4
-
-    ; Row11
-    phaddd                      m3,                m4                                    ; all rows and col completed.
-
-    mova                        m5,                [interp8_hps_shuf]
-    vpermd                      m3,                m5,                  m3
-    paddd                       m3,                m2
-    vextracti128                xm4,               m3,                  1
-    psrad                       xm3,               INTERP_SHIFT_PS
-    psrad                       xm4,               INTERP_SHIFT_PS
-    packssdw                    xm3,               xm3
-    packssdw                    xm4,               xm4
-
-    movq                        [r2],              xm3                                   ;row 0
-.end
-    RET
-%endif
-%endmacro
-
-    IPFILTER_LUMA_PS_4xN_AVX2 4
-    IPFILTER_LUMA_PS_4xN_AVX2 8
-    IPFILTER_LUMA_PS_4xN_AVX2 16
-
-%macro IPFILTER_LUMA_PS_8xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_ps_8x%1, 4, 6, 8
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-    shl                 r4d, 4
-%ifdef PIC
-    lea                 r6, [tab_LumaCoeff]
-    vpbroadcastq        m0, [r6 + r4]
-    vpbroadcastq        m1, [r6 + r4 + 8]
-%else
-    vpbroadcastq        m0, [tab_LumaCoeff + r4]
-    vpbroadcastq        m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 6
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    lea                 r6, [r1*3]
-    sub                 r0, r6
-    add                 r4d, 7
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m7, m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m7, m1
-    paddd               m4, m7
-
-    vbroadcasti128      m6, [r0 + 16]
-    pshufb              m5, m3
-    pshufb              m6, m3
-    pmaddwd             m5, m0
-    pmaddwd             m6, m1
-    paddd               m5, m6
-
-    phaddd              m4, m5
-    vpermq              m4, m4, q3120
-    paddd               m4, m2
-    vextracti128        xm5,m4, 1
-    psrad               xm4, INTERP_SHIFT_PS
-    psrad               xm5, INTERP_SHIFT_PS
-    packssdw            xm4, xm5
-
-    movu                [r2], xm4
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-    IPFILTER_LUMA_PS_8xN_AVX2 4
-    IPFILTER_LUMA_PS_8xN_AVX2 8
-    IPFILTER_LUMA_PS_8xN_AVX2 16
-    IPFILTER_LUMA_PS_8xN_AVX2 32
-
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_ps_24x32, 4, 6, 8
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-    shl                 r4d, 4
-%ifdef PIC
-    lea                 r6, [tab_LumaCoeff]
-    vpbroadcastq        m0, [r6 + r4]
-    vpbroadcastq        m1, [r6 + r4 + 8]
-%else
-    vpbroadcastq        m0, [tab_LumaCoeff + r4]
-    vpbroadcastq        m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 6
-    test                r5d, r5d
-    mov                 r4d, 32
-    jz                  .loop0
-    lea                 r6, [r1*3]
-    sub                 r0, r6
-    add                 r4d, 7
-
-.loop0:
-%assign x 0
-%rep 24/8
-    vbroadcasti128      m4, [r0 + x]
-    vbroadcasti128      m5, [r0 + 8 + x]
-    pshufb              m4, m3
-    pshufb              m7, m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m7, m1
-    paddd               m4, m7
-
-    vbroadcasti128      m6, [r0 + 16 + x]
-    pshufb              m5, m3
-    pshufb              m6, m3
-    pmaddwd             m5, m0
-    pmaddwd             m6, m1
-    paddd               m5, m6
-
-    phaddd              m4, m5
-    vpermq              m4, m4, q3120
-    paddd               m4, m2
-    vextracti128        xm5,m4, 1
-    psrad               xm4, INTERP_SHIFT_PS
-    psrad               xm5, INTERP_SHIFT_PS
-    packssdw            xm4, xm5
-
-    movu                [r2 + x], xm4
-    %assign x x+16
-    %endrep
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-
-
-%macro IPFILTER_LUMA_PS_32_64_AVX2 2
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_ps_%1x%2, 4, 6, 8
-
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-    shl                 r4d, 6
-%ifdef PIC
-    lea                 r6, [tab_LumaCoeffV]
-    movu                m0, [r6 + r4]
-    movu                m1, [r6 + r4 + mmsize]
-%else
-    movu                m0, [tab_LumaCoeffV + r4]
-    movu                m1, [tab_LumaCoeffV + r4 + mmsize]
-%endif
-    mova                m3, [interp8_hpp_shuf_new]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 6
-    test                r5d, r5d
-    mov                 r4d, %2
-    jz                 .loop0
-    lea                 r6, [r1*3]
-    sub                 r0, r6
-    add                 r4d, 7
-
-.loop0:
-%assign x 0
-%rep %1/16
-    vbroadcasti128      m4, [r0 + x]
-    vbroadcasti128      m5, [r0 + 4 * SIZEOF_PIXEL + x]
-    pshufb              m4, m3
-    pshufb              m5, m3
-
-    pmaddwd             m4, m0
-    pmaddwd             m7, m5, m1
-    paddd               m4, m7
-    vextracti128        xm7, m4, 1
-    paddd               xm4, xm7
-    paddd               xm4, xm2
-    psrad               xm4, INTERP_SHIFT_PS
-
-    vbroadcasti128      m6, [r0 + 16 + x]
-    pshufb              m6, m3
-
-    pmaddwd             m5, m0
-    pmaddwd             m7, m6, m1
-    paddd               m5, m7
-    vextracti128        xm7, m5, 1
-    paddd               xm5, xm7
-    paddd               xm5, xm2
-    psrad               xm5, INTERP_SHIFT_PS
-
-    packssdw            xm4, xm5
-    movu                [r2 + x], xm4
-
-    vbroadcasti128      m5, [r0 + 24 + x]
-    pshufb              m5, m3
-
-    pmaddwd             m6, m0
-    pmaddwd             m7, m5, m1
-    paddd               m6, m7
-    vextracti128        xm7, m6, 1
-    paddd               xm6, xm7
-    paddd               xm6, xm2
-    psrad               xm6, INTERP_SHIFT_PS
-
-    vbroadcasti128      m7, [r0 + 32 + x]
-    pshufb              m7, m3
-
-    pmaddwd             m5, m0
-    pmaddwd             m7, m1
-    paddd               m5, m7
-    vextracti128        xm7, m5, 1
-    paddd               xm5, xm7
-    paddd               xm5, xm2
-    psrad               xm5, INTERP_SHIFT_PS
-
-    packssdw            xm6, xm5
-    movu                [r2 + 16 + x], xm6
-
-%assign x x+32
-%endrep
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                .loop0
-    RET
-%endif
-%endmacro
-
-    IPFILTER_LUMA_PS_32_64_AVX2 32, 8
-    IPFILTER_LUMA_PS_32_64_AVX2 32, 16
-    IPFILTER_LUMA_PS_32_64_AVX2 32, 24
-    IPFILTER_LUMA_PS_32_64_AVX2 32, 32
-    IPFILTER_LUMA_PS_32_64_AVX2 32, 64
-
-    IPFILTER_LUMA_PS_32_64_AVX2 64, 16
-    IPFILTER_LUMA_PS_32_64_AVX2 64, 32
-    IPFILTER_LUMA_PS_32_64_AVX2 64, 48
-    IPFILTER_LUMA_PS_32_64_AVX2 64, 64
-
-    IPFILTER_LUMA_PS_32_64_AVX2 48, 64
-
-%macro IPFILTER_LUMA_PS_16xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_ps_16x%1, 4, 6, 8
-
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-    shl                 r4d, 4
-%ifdef PIC
-    lea                 r6, [tab_LumaCoeff]
-    vpbroadcastq        m0, [r6 + r4]
-    vpbroadcastq        m1, [r6 + r4 + 8]
-%else
-    vpbroadcastq        m0, [tab_LumaCoeff + r4]
-    vpbroadcastq        m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 6
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    lea                 r6, [r1*3]
-    sub                 r0, r6
-    add                 r4d, 7
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m7, m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m7, m1
-    paddd               m4, m7
-
-    vbroadcasti128      m6, [r0 + 16]
-    pshufb              m5, m3
-    pshufb              m7, m6, m3
-    pmaddwd             m5, m0
-    pmaddwd             m7, m1
-    paddd               m5, m7
-
-    phaddd              m4, m5
-    vpermq              m4, m4, q3120
-    paddd               m4, m2
-    vextracti128        xm5, m4, 1
-    psrad               xm4, INTERP_SHIFT_PS
-    psrad               xm5, INTERP_SHIFT_PS
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-
-    vbroadcasti128      m5, [r0 + 24]
-    pshufb              m6, m3
-    pshufb              m7, m5, m3
-    pmaddwd             m6, m0
-    pmaddwd             m7, m1
-    paddd               m6, m7
-
-    vbroadcasti128      m7, [r0 + 32]
-    pshufb              m5, m3
-    pshufb              m7, m3
-    pmaddwd             m5, m0
-    pmaddwd             m7, m1
-    paddd               m5, m7
-
-    phaddd              m6, m5
-    vpermq              m6, m6, q3120
-    paddd               m6, m2
-    vextracti128        xm5,m6, 1
-    psrad               xm6, INTERP_SHIFT_PS
-    psrad               xm5, INTERP_SHIFT_PS
-    packssdw            xm6, xm5
-    movu                [r2 + 16], xm6
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-    IPFILTER_LUMA_PS_16xN_AVX2 4
-    IPFILTER_LUMA_PS_16xN_AVX2 8
-    IPFILTER_LUMA_PS_16xN_AVX2 12
-    IPFILTER_LUMA_PS_16xN_AVX2 16
-    IPFILTER_LUMA_PS_16xN_AVX2 32
-    IPFILTER_LUMA_PS_16xN_AVX2 64
-
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_ps_12x16, 4, 6, 8
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-    shl                 r4d, 4
-%ifdef PIC
-    lea                 r6, [tab_LumaCoeff]
-    vpbroadcastq        m0, [r6 + r4]
-    vpbroadcastq        m1, [r6 + r4 + 8]
-%else
-    vpbroadcastq        m0, [tab_LumaCoeff + r4]
-    vpbroadcastq        m1, [tab_LumaCoeff + r4 + 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 6
-    test                r5d, r5d
-    mov                 r4d, 16
-    jz                  .loop0
-    lea                 r6, [r1*3]
-    sub                 r0, r6
-    add                 r4d, 7
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m7, m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m7, m1
-    paddd               m4, m7
-
-    vbroadcasti128      m6, [r0 + 16]
-    pshufb              m5, m3
-    pshufb              m7, m6, m3
-    pmaddwd             m5, m0
-    pmaddwd             m7, m1
-    paddd               m5, m7
-
-    phaddd              m4, m5
-    vpermq              m4, m4, q3120
-    paddd               m4, m2
-    vextracti128        xm5,m4, 1
-    psrad               xm4, INTERP_SHIFT_PS
-    psrad               xm5, INTERP_SHIFT_PS
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-
-    vbroadcasti128      m5, [r0 + 24]
-    pshufb              m6, m3
-    pshufb              m5, m3
-    pmaddwd             m6, m0
-    pmaddwd             m5, m1
-    paddd               m6, m5
-
-    phaddd              m6, m6
-    vpermq              m6, m6, q3120
-    paddd               xm6, xm2
-    psrad               xm6, INTERP_SHIFT_PS
-    packssdw            xm6, xm6
-    movq                [r2 + 16], xm6
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-
-%macro IPFILTER_CHROMA_PS_8xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_horiz_ps_8x%1, 4, 7, 6
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-
-%ifdef PIC
-    lea                 r6, [tab_ChromaCoeff]
-    vpbroadcastq        m0, [r6 + r4 * 8]
-%else
-    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 2
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    sub                 r0, r1
-    add                 r4d, 3
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-    IPFILTER_CHROMA_PS_8xN_AVX2 4
-    IPFILTER_CHROMA_PS_8xN_AVX2 8
-    IPFILTER_CHROMA_PS_8xN_AVX2 16
-    IPFILTER_CHROMA_PS_8xN_AVX2 32
-    IPFILTER_CHROMA_PS_8xN_AVX2 6
-    IPFILTER_CHROMA_PS_8xN_AVX2 2
-    IPFILTER_CHROMA_PS_8xN_AVX2 12
-    IPFILTER_CHROMA_PS_8xN_AVX2 64
-
-%macro IPFILTER_CHROMA_PS_16xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_horiz_ps_16x%1, 4, 7, 6
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-
-%ifdef PIC
-    lea                 r6, [tab_ChromaCoeff]
-    vpbroadcastq        m0, [r6 + r4 * 8]
-%else
-    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 2
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    sub                 r0, r1
-    add                 r4d, 3
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-
-    vbroadcasti128      m4, [r0 + 16]
-    vbroadcasti128      m5, [r0 + 24]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 16], xm4
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-IPFILTER_CHROMA_PS_16xN_AVX2 16
-IPFILTER_CHROMA_PS_16xN_AVX2 8
-IPFILTER_CHROMA_PS_16xN_AVX2 32
-IPFILTER_CHROMA_PS_16xN_AVX2 12
-IPFILTER_CHROMA_PS_16xN_AVX2 4
-IPFILTER_CHROMA_PS_16xN_AVX2 64
-IPFILTER_CHROMA_PS_16xN_AVX2 24
-
-%macro IPFILTER_CHROMA_PS_24xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_horiz_ps_24x%1, 4, 7, 6
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-
-%ifdef PIC
-    lea                 r6, [tab_ChromaCoeff]
-    vpbroadcastq        m0, [r6 + r4 * 8]
-%else
-    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 2
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    sub                 r0, r1
-    add                 r4d, 3
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-
-    vbroadcasti128      m4, [r0 + 16]
-    vbroadcasti128      m5, [r0 + 24]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 16], xm4
-
-    vbroadcasti128      m4, [r0 + 32]
-    vbroadcasti128      m5, [r0 + 40]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 32], xm4
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-IPFILTER_CHROMA_PS_24xN_AVX2 32
-IPFILTER_CHROMA_PS_24xN_AVX2 64
-
-%macro IPFILTER_CHROMA_PS_12xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_horiz_ps_12x%1, 4, 7, 6
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-
-%ifdef PIC
-    lea                 r6, [tab_ChromaCoeff]
-    vpbroadcastq        m0, [r6 + r4 * 8]
-%else
-    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 2
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    sub                 r0, r1
-    add                 r4d, 3
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-
-    vbroadcasti128      m4, [r0 + 16]
-    pshufb              m4, m3
-    pmaddwd             m4, m0
-    phaddd              m4, m4
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movq                [r2 + 16], xm4
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-IPFILTER_CHROMA_PS_12xN_AVX2 16
-IPFILTER_CHROMA_PS_12xN_AVX2 32
-
-%macro IPFILTER_CHROMA_PS_32xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_horiz_ps_32x%1, 4, 7, 6
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-
-%ifdef PIC
-    lea                 r6, [tab_ChromaCoeff]
-    vpbroadcastq        m0, [r6 + r4 * 8]
-%else
-    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 2
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    sub                 r0, r1
-    add                 r4d, 3
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-
-    vbroadcasti128      m4, [r0 + 16]
-    vbroadcasti128      m5, [r0 + 24]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 16], xm4
-
-    vbroadcasti128      m4, [r0 + 32]
-    vbroadcasti128      m5, [r0 + 40]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 32], xm4
-
-    vbroadcasti128      m4, [r0 + 48]
-    vbroadcasti128      m5, [r0 + 56]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 48], xm4
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-IPFILTER_CHROMA_PS_32xN_AVX2 32
-IPFILTER_CHROMA_PS_32xN_AVX2 16
-IPFILTER_CHROMA_PS_32xN_AVX2 24
-IPFILTER_CHROMA_PS_32xN_AVX2 8
-IPFILTER_CHROMA_PS_32xN_AVX2 64
-IPFILTER_CHROMA_PS_32xN_AVX2 48
-
-
-%macro IPFILTER_CHROMA_PS_64xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_horiz_ps_64x%1, 4, 7, 6
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-
-%ifdef PIC
-    lea                 r6, [tab_ChromaCoeff]
-    vpbroadcastq        m0, [r6 + r4 * 8]
-%else
-    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 2
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    sub                 r0, r1
-    add                 r4d, 3
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-
-    vbroadcasti128      m4, [r0 + 16]
-    vbroadcasti128      m5, [r0 + 24]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 16], xm4
-
-    vbroadcasti128      m4, [r0 + 32]
-    vbroadcasti128      m5, [r0 + 40]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 32], xm4
-
-    vbroadcasti128      m4, [r0 + 48]
-    vbroadcasti128      m5, [r0 + 56]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 48], xm4
-
-    vbroadcasti128      m4, [r0 + 64]
-    vbroadcasti128      m5, [r0 + 72]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 64], xm4
-
-    vbroadcasti128      m4, [r0 + 80]
-    vbroadcasti128      m5, [r0 + 88]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 80], xm4
-
-    vbroadcasti128      m4, [r0 + 96]
-    vbroadcasti128      m5, [r0 + 104]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 96], xm4
-
-    vbroadcasti128      m4, [r0 + 112]
-    vbroadcasti128      m5, [r0 + 120]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 112], xm4
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-IPFILTER_CHROMA_PS_64xN_AVX2 64
-IPFILTER_CHROMA_PS_64xN_AVX2 48
-IPFILTER_CHROMA_PS_64xN_AVX2 32
-IPFILTER_CHROMA_PS_64xN_AVX2 16
-
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_horiz_ps_48x64, 4, 7, 6
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-
-%ifdef PIC
-    lea                 r6, [tab_ChromaCoeff]
-    vpbroadcastq        m0, [r6 + r4 * 8]
-%else
-    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 2
-    test                r5d, r5d
-    mov                 r4d, 64
-    jz                  .loop0
-    sub                 r0, r1
-    add                 r4d, 3
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2], xm4
-
-    vbroadcasti128      m4, [r0 + 16]
-    vbroadcasti128      m5, [r0 + 24]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 16], xm4
-
-    vbroadcasti128      m4, [r0 + 32]
-    vbroadcasti128      m5, [r0 + 40]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 32], xm4
-
-    vbroadcasti128      m4, [r0 + 48]
-    vbroadcasti128      m5, [r0 + 56]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 48], xm4
-
-    vbroadcasti128      m4, [r0 + 64]
-    vbroadcasti128      m5, [r0 + 72]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 64], xm4
-
-    vbroadcasti128      m4, [r0 + 80]
-    vbroadcasti128      m5, [r0 + 88]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movu                [r2 + 80], xm4
-
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-
-%macro IPFILTER_CHROMA_PS_6xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_horiz_ps_6x%1, 4, 7, 6
-    add                 r1d, r1d
-    add                 r3d, r3d
-    mov                 r4d, r4m
-    mov                 r5d, r5m
-
-%ifdef PIC
-    lea                 r6, [tab_ChromaCoeff]
-    vpbroadcastq        m0, [r6 + r4 * 8]
-%else
-    vpbroadcastq        m0, [tab_ChromaCoeff + r4 * 8]
-%endif
-    mova                m3, [interp8_hpp_shuf]
-    vbroadcasti128      m2, [INTERP_OFFSET_PS]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-
-    sub                 r0, 2
-    test                r5d, r5d
-    mov                 r4d, %1
-    jz                  .loop0
-    sub                 r0, r1
-    add                 r4d, 3
-
-.loop0:
-    vbroadcasti128      m4, [r0]
-    vbroadcasti128      m5, [r0 + 8]
-    pshufb              m4, m3
-    pshufb              m5, m3
-    pmaddwd             m4, m0
-    pmaddwd             m5, m0
-    phaddd              m4, m5
-    paddd               m4, m2
-    vpermq              m4, m4, q3120
-    psrad               m4, INTERP_SHIFT_PS
-    vextracti128        xm5, m4, 1
-    packssdw            xm4, xm5
-    movq                [r2], xm4
-    pextrd              [r2 + 8], xm4, 2
-    add                 r2, r3
-    add                 r0, r1
-    dec                 r4d
-    jnz                 .loop0
-    RET
-%endif
-%endmacro
-
-    IPFILTER_CHROMA_PS_6xN_AVX2 8
-    IPFILTER_CHROMA_PS_6xN_AVX2 16
-
-%macro FILTER_VER_CHROMA_AVX2_8xN 2
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_8x%2, 4, 9, 15
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    add             r3d, r3d
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    vbroadcasti128  m14, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m14, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m14, [INTERP_OFFSET_PS]
-%endif
-    lea             r6, [r3 * 3]
-    lea             r7, [r1 * 4]
-    mov             r8d, %2 / 16
-.loopH:
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm7, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddwd         m7, m5, [r5 + 1 * mmsize]
-    paddd           m3, m7
-    pmaddwd         m5, [r5]
-
-    movu            xm7, [r0 + r4]                  ; m7 = row 7
-    punpckhwd       xm8, xm6, xm7
-    punpcklwd       xm6, xm7
-    vinserti128     m6, m6, xm8, 1
-    pmaddwd         m8, m6, [r5 + 1 * mmsize]
-    paddd           m4, m8
-    pmaddwd         m6, [r5]
-
-    lea             r0, [r0 + r1 * 4]
-    movu            xm8, [r0]                       ; m8 = row 8
-    punpckhwd       xm9, xm7, xm8
-    punpcklwd       xm7, xm8
-    vinserti128     m7, m7, xm9, 1
-    pmaddwd         m9, m7, [r5 + 1 * mmsize]
-    paddd           m5, m9
-    pmaddwd         m7, [r5]
-
-
-    movu            xm9, [r0 + r1]                  ; m9 = row 9
-    punpckhwd       xm10, xm8, xm9
-    punpcklwd       xm8, xm9
-    vinserti128     m8, m8, xm10, 1
-    pmaddwd         m10, m8, [r5 + 1 * mmsize]
-    paddd           m6, m10
-    pmaddwd         m8, [r5]
-
-
-    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
-    punpckhwd       xm11, xm9, xm10
-    punpcklwd       xm9, xm10
-    vinserti128     m9, m9, xm11, 1
-    pmaddwd         m11, m9, [r5 + 1 * mmsize]
-    paddd           m7, m11
-    pmaddwd         m9, [r5]
-
-    movu            xm11, [r0 + r4]                 ; m11 = row 11
-    punpckhwd       xm12, xm10, xm11
-    punpcklwd       xm10, xm11
-    vinserti128     m10, m10, xm12, 1
-    pmaddwd         m12, m10, [r5 + 1 * mmsize]
-    paddd           m8, m12
-    pmaddwd         m10, [r5]
-
-    lea             r0, [r0 + r1 * 4]
-    movu            xm12, [r0]                      ; m12 = row 12
-    punpckhwd       xm13, xm11, xm12
-    punpcklwd       xm11, xm12
-    vinserti128     m11, m11, xm13, 1
-    pmaddwd         m13, m11, [r5 + 1 * mmsize]
-    paddd           m9, m13
-    pmaddwd         m11, [r5]
-
-%ifidn %1,ss
-    psrad           m0, 6
-    psrad           m1, 6
-    psrad           m2, 6
-    psrad           m3, 6
-    psrad           m4, 6
-    psrad           m5, 6
-%else
-    paddd           m0, m14
-    paddd           m1, m14
-    paddd           m2, m14
-    paddd           m3, m14
-    paddd           m4, m14
-    paddd           m5, m14
-%ifidn %1,pp
-    psrad           m0, 6
-    psrad           m1, 6
-    psrad           m2, 6
-    psrad           m3, 6
-    psrad           m4, 6
-    psrad           m5, 6
-%elifidn %1, sp
-    psrad           m0, INTERP_SHIFT_SP
-    psrad           m1, INTERP_SHIFT_SP
-    psrad           m2, INTERP_SHIFT_SP
-    psrad           m3, INTERP_SHIFT_SP
-    psrad           m4, INTERP_SHIFT_SP
-    psrad           m5, INTERP_SHIFT_SP
-%else
-    psrad           m0, INTERP_SHIFT_PS
-    psrad           m1, INTERP_SHIFT_PS
-    psrad           m2, INTERP_SHIFT_PS
-    psrad           m3, INTERP_SHIFT_PS
-    psrad           m4, INTERP_SHIFT_PS
-    psrad           m5, INTERP_SHIFT_PS
-%endif
-%endif
-
-    packssdw        m0, m1
-    packssdw        m2, m3
-    packssdw        m4, m5
-    vpermq          m0, m0, q3120
-    vpermq          m2, m2, q3120
-    vpermq          m4, m4, q3120
-    pxor            m5, m5
-    mova            m3, [pw_pixel_max]
-%ifidn %1,pp
-    CLIPW           m0, m5, m3
-    CLIPW           m2, m5, m3
-    CLIPW           m4, m5, m3
-%elifidn %1, sp
-    CLIPW           m0, m5, m3
-    CLIPW           m2, m5, m3
-    CLIPW           m4, m5, m3
-%endif
-
-    vextracti128    xm1, m0, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    vextracti128    xm1, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm1
-    lea             r2, [r2 + r3 * 4]
-    vextracti128    xm1, m4, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm1
-
-    movu            xm13, [r0 + r1]                 ; m13 = row 13
-    punpckhwd       xm0, xm12, xm13
-    punpcklwd       xm12, xm13
-    vinserti128     m12, m12, xm0, 1
-    pmaddwd         m0, m12, [r5 + 1 * mmsize]
-    paddd           m10, m0
-    pmaddwd         m12, [r5]
-
-    movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
-    punpckhwd       xm1, xm13, xm0
-    punpcklwd       xm13, xm0
-    vinserti128     m13, m13, xm1, 1
-    pmaddwd         m1, m13, [r5 + 1 * mmsize]
-    paddd           m11, m1
-    pmaddwd         m13, [r5]
-
-%ifidn %1,ss
-    psrad           m6, 6
-    psrad           m7, 6
-%else
-    paddd           m6, m14
-    paddd           m7, m14
-%ifidn %1,pp
-    psrad           m6, 6
-    psrad           m7, 6
-%elifidn %1, sp
-    psrad           m6, INTERP_SHIFT_SP
-    psrad           m7, INTERP_SHIFT_SP
-%else
-    psrad           m6, INTERP_SHIFT_PS
-    psrad           m7, INTERP_SHIFT_PS
-%endif
-%endif
-
-    packssdw        m6, m7
-    vpermq          m6, m6, q3120
-%ifidn %1,pp
-    CLIPW           m6, m5, m3
-%elifidn %1, sp
-    CLIPW           m6, m5, m3
-%endif
-    vextracti128    xm7, m6, 1
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r6], xm7
-
-    movu            xm1, [r0 + r4]                  ; m1 = row 15
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m2, m0, [r5 + 1 * mmsize]
-    paddd           m12, m2
-    pmaddwd         m0, [r5]
-
-    lea             r0, [r0 + r1 * 4]
-    movu            xm2, [r0]                       ; m2 = row 16
-    punpckhwd       xm6, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm6, 1
-    pmaddwd         m6, m1, [r5 + 1 * mmsize]
-    paddd           m13, m6
-    pmaddwd         m1, [r5]
-
-    movu            xm6, [r0 + r1]                  ; m6 = row 17
-    punpckhwd       xm4, xm2, xm6
-    punpcklwd       xm2, xm6
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m0, m2
-
-    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
-    punpckhwd       xm2, xm6, xm4
-    punpcklwd       xm6, xm4
-    vinserti128     m6, m6, xm2, 1
-    pmaddwd         m6, [r5 + 1 * mmsize]
-    paddd           m1, m6
-
-%ifidn %1,ss
-    psrad           m8, 6
-    psrad           m9, 6
-    psrad           m10, 6
-    psrad           m11, 6
-    psrad           m12, 6
-    psrad           m13, 6
-    psrad           m0, 6
-    psrad           m1, 6
-%else
-    paddd           m8, m14
-    paddd           m9, m14
-    paddd           m10, m14
-    paddd           m11, m14
-    paddd           m12, m14
-    paddd           m13, m14
-    paddd           m0, m14
-    paddd           m1, m14
-%ifidn %1,pp
-    psrad           m8, 6
-    psrad           m9, 6
-    psrad           m10, 6
-    psrad           m11, 6
-    psrad           m12, 6
-    psrad           m13, 6
-    psrad           m0, 6
-    psrad           m1, 6
-%elifidn %1, sp
-    psrad           m8, INTERP_SHIFT_SP
-    psrad           m9, INTERP_SHIFT_SP
-    psrad           m10, INTERP_SHIFT_SP
-    psrad           m11, INTERP_SHIFT_SP
-    psrad           m12, INTERP_SHIFT_SP
-    psrad           m13, INTERP_SHIFT_SP
-    psrad           m0, INTERP_SHIFT_SP
-    psrad           m1, INTERP_SHIFT_SP
-%else
-    psrad           m8, INTERP_SHIFT_PS
-    psrad           m9, INTERP_SHIFT_PS
-    psrad           m10, INTERP_SHIFT_PS
-    psrad           m11, INTERP_SHIFT_PS
-    psrad           m12, INTERP_SHIFT_PS
-    psrad           m13, INTERP_SHIFT_PS
-    psrad           m0, INTERP_SHIFT_PS
-    psrad           m1, INTERP_SHIFT_PS
-%endif
-%endif
-
-    packssdw        m8, m9
-    packssdw        m10, m11
-    packssdw        m12, m13
-    packssdw        m0, m1
-    vpermq          m8, m8, q3120
-    vpermq          m10, m10, q3120
-    vpermq          m12, m12, q3120
-    vpermq          m0, m0, q3120
-%ifidn %1,pp
-    CLIPW           m8, m5, m3
-    CLIPW           m10, m5, m3
-    CLIPW           m12, m5, m3
-    CLIPW           m0, m5, m3
-%elifidn %1, sp
-    CLIPW           m8, m5, m3
-    CLIPW           m10, m5, m3
-    CLIPW           m12, m5, m3
-    CLIPW           m0, m5, m3
-%endif
-    vextracti128    xm9, m8, 1
-    vextracti128    xm11, m10, 1
-    vextracti128    xm13, m12, 1
-    vextracti128    xm1, m0, 1
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm8
-    movu            [r2 + r3], xm9
-    movu            [r2 + r3 * 2], xm10
-    movu            [r2 + r6], xm11
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm12
-    movu            [r2 + r3], xm13
-    movu            [r2 + r3 * 2], xm0
-    movu            [r2 + r6], xm1
-    lea             r2, [r2 + r3 * 4]
-    dec             r8d
-    jnz             .loopH
-    RET
-%endif
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_8xN pp, 16
-FILTER_VER_CHROMA_AVX2_8xN ps, 16
-FILTER_VER_CHROMA_AVX2_8xN ss, 16
-FILTER_VER_CHROMA_AVX2_8xN sp, 16
-FILTER_VER_CHROMA_AVX2_8xN pp, 32
-FILTER_VER_CHROMA_AVX2_8xN ps, 32
-FILTER_VER_CHROMA_AVX2_8xN sp, 32
-FILTER_VER_CHROMA_AVX2_8xN ss, 32
-FILTER_VER_CHROMA_AVX2_8xN pp, 64
-FILTER_VER_CHROMA_AVX2_8xN ps, 64
-FILTER_VER_CHROMA_AVX2_8xN sp, 64
-FILTER_VER_CHROMA_AVX2_8xN ss, 64
-
-%macro PROCESS_CHROMA_AVX2_8x2 3
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m2, m2, [r5 + 1 * mmsize]
-    paddd           m0, m2
-
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m3, m3, [r5 + 1 * mmsize]
-    paddd           m1, m3
-
-%ifnidn %1,ss
-    paddd           m0, m7
-    paddd           m1, m7
-%endif
-    psrad           m0, %3
-    psrad           m1, %3
-
-    packssdw        m0, m1
-    vpermq          m0, m0, q3120
-    pxor            m4, m4
-
-%if %2
-    CLIPW           m0, m4, [pw_pixel_max]
-%endif
-    vextracti128    xm1, m0, 1
-%endmacro
-
-
-%macro FILTER_VER_CHROMA_AVX2_8x2 3
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x2, 4, 6, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    add             r3d, r3d
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    vbroadcasti128  m7, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-
-    PROCESS_CHROMA_AVX2_8x2 %1, %2, %3
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_8x2 pp, 1, 6
-FILTER_VER_CHROMA_AVX2_8x2 ps, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_8x2 sp, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_8x2 ss, 0, 6
-
-%macro FILTER_VER_CHROMA_AVX2_4x2 3
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x2, 4, 6, 7
-    mov             r4d, r4m
-    add             r1d, r1d
-    add             r3d, r3d
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-
-%ifidn %1,pp
-    vbroadcasti128  m6, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m6, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m6, [INTERP_OFFSET_PS]
-%endif
-
-    movq            xm0, [r0]                       ; row 0
-    movq            xm1, [r0 + r1]                  ; row 1
-    punpcklwd       xm0, xm1
-
-    movq            xm2, [r0 + r1 * 2]              ; row 2
-    punpcklwd       xm1, xm2
-    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
-    pmaddwd         m0, [r5]
-
-    movq            xm3, [r0 + r4]                  ; row 3
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm4, [r0]                       ; row 4
-    punpcklwd       xm3, xm4
-    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    paddd           m0, m5
-
-%ifnidn %1, ss
-    paddd           m0, m6
-%endif
-    psrad           m0, %3
-    packssdw        m0, m0
-    pxor            m1, m1
-
-%if %2
-    CLIPW           m0, m1, [pw_pixel_max]
-%endif
-
-    vextracti128    xm2, m0, 1
-    lea             r4, [r3 * 3]
-    movq            [r2], xm0
-    movq            [r2 + r3], xm2
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_4x2 pp, 1, 6
-FILTER_VER_CHROMA_AVX2_4x2 ps, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_4x2 sp, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_4x2 ss, 0, 6
-
-%macro FILTER_VER_CHROMA_AVX2_4x4 3
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x4, 4, 6, 7
-    mov             r4d, r4m
-    add             r1d, r1d
-    add             r3d, r3d
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-
-%ifidn %1,pp
-   vbroadcasti128  m6, [pd_32]
-%elifidn %1, sp
-   vbroadcasti128  m6, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m6, [INTERP_OFFSET_PS]
-%endif
-    movq            xm0, [r0]                       ; row 0
-    movq            xm1, [r0 + r1]                  ; row 1
-    punpcklwd       xm0, xm1
-
-    movq            xm2, [r0 + r1 * 2]              ; row 2
-    punpcklwd       xm1, xm2
-    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
-    pmaddwd         m0, [r5]
-
-    movq            xm3, [r0 + r4]                  ; row 3
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm4, [r0]                       ; row 4
-    punpcklwd       xm3, xm4
-    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m5
-
-    movq            xm3, [r0 + r1]                  ; row 5
-    punpcklwd       xm4, xm3
-    movq            xm1, [r0 + r1 * 2]              ; row 6
-    punpcklwd       xm3, xm1
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
-    pmaddwd         m4, [r5 + 1 * mmsize]
-    paddd           m2, m4
-
-%ifnidn %1,ss
-    paddd           m0, m6
-    paddd           m2, m6
-%endif
-    psrad           m0, %3
-    psrad           m2, %3
-
-    packssdw        m0, m2
-    pxor            m1, m1
-%if %2
-    CLIPW           m0, m1, [pw_pixel_max]
-%endif
-
-    vextracti128    xm2, m0, 1
-    lea             r4, [r3 * 3]
-    movq            [r2], xm0
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r4], xm2
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_4x4 pp, 1, 6
-FILTER_VER_CHROMA_AVX2_4x4 ps, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_4x4 sp, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_4x4 ss, 0, 6
-
-
-%macro FILTER_VER_CHROMA_AVX2_4x8 3
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x8, 4, 7, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    add             r3d, r3d
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-
-%ifidn %1,pp
-    vbroadcasti128  m7, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-    lea             r6, [r3 * 3]
-
-    movq            xm0, [r0]                       ; row 0
-    movq            xm1, [r0 + r1]                  ; row 1
-    punpcklwd       xm0, xm1
-    movq            xm2, [r0 + r1 * 2]              ; row 2
-    punpcklwd       xm1, xm2
-    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
-    pmaddwd         m0, [r5]
-
-    movq            xm3, [r0 + r4]                  ; row 3
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm4, [r0]                       ; row 4
-    punpcklwd       xm3, xm4
-    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m5
-
-    movq            xm3, [r0 + r1]                  ; row 5
-    punpcklwd       xm4, xm3
-    movq            xm1, [r0 + r1 * 2]              ; row 6
-    punpcklwd       xm3, xm1
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
-    pmaddwd         m5, m4, [r5 + 1 * mmsize]
-    paddd           m2, m5
-    pmaddwd         m4, [r5]
-
-    movq            xm3, [r0 + r4]                  ; row 7
-    punpcklwd       xm1, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm6, [r0]                       ; row 8
-    punpcklwd       xm3, xm6
-    vinserti128     m1, m1, xm3, 1                  ; m1 = [8 7 7 6]
-    pmaddwd         m5, m1, [r5 + 1 * mmsize]
-    paddd           m4, m5
-    pmaddwd         m1, [r5]
-
-    movq            xm3, [r0 + r1]                  ; row 9
-    punpcklwd       xm6, xm3
-    movq            xm5, [r0 + 2 * r1]              ; row 10
-    punpcklwd       xm3, xm5
-    vinserti128     m6, m6, xm3, 1                  ; m6 = [A 9 9 8]
-    pmaddwd         m6, [r5 + 1 * mmsize]
-    paddd           m1, m6
-%ifnidn %1,ss
-    paddd           m0, m7
-    paddd           m2, m7
-%endif
-    psrad           m0, %3
-    psrad           m2, %3
-    packssdw        m0, m2
-    pxor            m6, m6
-    mova            m3, [pw_pixel_max]
-%if %2
-    CLIPW           m0, m6, m3
-%endif
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r6], xm2
-%ifnidn %1,ss
-    paddd           m4, m7
-    paddd           m1, m7
-%endif
-    psrad           m4, %3
-    psrad           m1, %3
-    packssdw        m4, m1
-%if %2
-    CLIPW           m4, m6, m3
-%endif
-    vextracti128    xm1, m4, 1
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm4
-    movq            [r2 + r3], xm1
-    movhps          [r2 + r3 * 2], xm4
-    movhps          [r2 + r6], xm1
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_4x8 pp, 1, 6
-FILTER_VER_CHROMA_AVX2_4x8 ps, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_4x8 sp, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_4x8 ss, 0 , 6
-
-%macro PROCESS_LUMA_AVX2_W4_16R_4TAP 3
-    movq            xm0, [r0]                       ; row 0
-    movq            xm1, [r0 + r1]                  ; row 1
-    punpcklwd       xm0, xm1
-    movq            xm2, [r0 + r1 * 2]              ; row 2
-    punpcklwd       xm1, xm2
-    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
-    pmaddwd         m0, [r5]
-    movq            xm3, [r0 + r4]                  ; row 3
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm4, [r0]                       ; row 4
-    punpcklwd       xm3, xm4
-    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m5
-    movq            xm3, [r0 + r1]                  ; row 5
-    punpcklwd       xm4, xm3
-    movq            xm1, [r0 + r1 * 2]              ; row 6
-    punpcklwd       xm3, xm1
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
-    pmaddwd         m5, m4, [r5 + 1 * mmsize]
-    paddd           m2, m5
-    pmaddwd         m4, [r5]
-    movq            xm3, [r0 + r4]                  ; row 7
-    punpcklwd       xm1, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm6, [r0]                       ; row 8
-    punpcklwd       xm3, xm6
-    vinserti128     m1, m1, xm3, 1                  ; m1 = [8 7 7 6]
-    pmaddwd         m5, m1, [r5 + 1 * mmsize]
-    paddd           m4, m5
-    pmaddwd         m1, [r5]
-    movq            xm3, [r0 + r1]                  ; row 9
-    punpcklwd       xm6, xm3
-    movq            xm5, [r0 + 2 * r1]              ; row 10
-    punpcklwd       xm3, xm5
-    vinserti128     m6, m6, xm3, 1                  ; m6 = [10 9 9 8]
-    pmaddwd         m3, m6, [r5 + 1 * mmsize]
-    paddd           m1, m3
-    pmaddwd         m6, [r5]
-%ifnidn %1,ss
-    paddd           m0, m7
-    paddd           m2, m7
-%endif
-    psrad           m0, %3
-    psrad           m2, %3
-    packssdw        m0, m2
-    pxor            m3, m3
-%if %2
-    CLIPW           m0, m3, [pw_pixel_max]
-%endif
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r6], xm2
-    movq            xm2, [r0 + r4]                  ;row 11
-    punpcklwd       xm5, xm2
-    lea             r0, [r0 + 4 * r1]
-    movq            xm0, [r0]                       ; row 12
-    punpcklwd       xm2, xm0
-    vinserti128     m5, m5, xm2, 1                  ; m5 = [12 11 11 10]
-    pmaddwd         m2, m5, [r5 + 1 * mmsize]
-    paddd           m6, m2
-    pmaddwd         m5, [r5]
-    movq            xm2, [r0 + r1]                  ; row 13
-    punpcklwd       xm0, xm2
-    movq            xm3, [r0 + 2 * r1]              ; row 14
-    punpcklwd       xm2, xm3
-    vinserti128     m0, m0, xm2, 1                  ; m0 = [14 13 13 12]
-    pmaddwd         m2, m0, [r5 + 1 * mmsize]
-    paddd           m5, m2
-    pmaddwd         m0, [r5]
-%ifnidn %1,ss
-    paddd           m4, m7
-    paddd           m1, m7
-%endif
-    psrad           m4, %3
-    psrad           m1, %3
-    packssdw        m4, m1
-    pxor            m2, m2
-%if %2
-    CLIPW           m4, m2, [pw_pixel_max]
-%endif
-
-    vextracti128    xm1, m4, 1
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm4
-    movq            [r2 + r3], xm1
-    movhps          [r2 + r3 * 2], xm4
-    movhps          [r2 + r6], xm1
-    movq            xm4, [r0 + r4]                  ; row 15
-    punpcklwd       xm3, xm4
-    lea             r0, [r0 + 4 * r1]
-    movq            xm1, [r0]                       ; row 16
-    punpcklwd       xm4, xm1
-    vinserti128     m3, m3, xm4, 1                  ; m3 = [16 15 15 14]
-    pmaddwd         m4, m3, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m3, [r5]
-    movq            xm4, [r0 + r1]                  ; row 17
-    punpcklwd       xm1, xm4
-    movq            xm2, [r0 + 2 * r1]              ; row 18
-    punpcklwd       xm4, xm2
-    vinserti128     m1, m1, xm4, 1                  ; m1 = [18 17 17 16]
-    pmaddwd         m1, [r5 + 1 * mmsize]
-    paddd           m3, m1
-
-%ifnidn %1,ss
-    paddd           m6, m7
-    paddd           m5, m7
-%endif
-    psrad           m6, %3
-    psrad           m5, %3
-    packssdw        m6, m5
-    pxor            m1, m1
-%if %2
-    CLIPW           m6, m1, [pw_pixel_max]
-%endif
-    vextracti128    xm5, m6, 1
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm6
-    movq            [r2 + r3], xm5
-    movhps          [r2 + r3 * 2], xm6
-    movhps          [r2 + r6], xm5
-%ifnidn %1,ss
-    paddd           m0, m7
-    paddd           m3, m7
-%endif
-    psrad           m0, %3
-    psrad           m3, %3
-    packssdw        m0, m3
-%if %2
-    CLIPW           m0, m1, [pw_pixel_max]
-%endif
-    vextracti128    xm3, m0, 1
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm0
-    movq            [r2 + r3], xm3
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r6], xm3
-%endmacro
-
-
-%macro FILTER_VER_CHROMA_AVX2_4xN 4
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x%2, 4, 8, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    add             r3d, r3d
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-    mov             r7d, %2 / 16
-%ifidn %1,pp
-    vbroadcasti128  m7, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-    lea             r6, [r3 * 3]
-.loopH:
-    PROCESS_LUMA_AVX2_W4_16R_4TAP %1, %3, %4
-    lea             r2, [r2 + r3 * 4]
-    dec             r7d
-    jnz             .loopH
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_4xN pp, 16, 1, 6
-FILTER_VER_CHROMA_AVX2_4xN ps, 16, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_4xN sp, 16, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_4xN ss, 16, 0, 6
-FILTER_VER_CHROMA_AVX2_4xN pp, 32, 1, 6
-FILTER_VER_CHROMA_AVX2_4xN ps, 32, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_4xN sp, 32, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_4xN ss, 32, 0, 6
-
-%macro FILTER_VER_CHROMA_AVX2_8x8 3
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_8x8, 4, 6, 12
-    mov             r4d, r4m
-    add             r1d, r1d
-    add             r3d, r3d
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-
-%ifidn %1,pp
-    vbroadcasti128  m11, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m11, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m11, [INTERP_OFFSET_PS]
-%endif
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m4                          ; res row0 done(0,1,2,3)
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    pmaddwd         m3, [r5]
-    paddd           m1, m5                          ;res row1 done(1, 2, 3, 4)
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    pmaddwd         m4, [r5]
-    paddd           m2, m6                          ;res row2 done(2,3,4,5)
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm7, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddwd         m7, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m7                          ;res row3 done(3,4,5,6)
-    movu            xm7, [r0 + r4]                  ; m7 = row 7
-    punpckhwd       xm8, xm6, xm7
-    punpcklwd       xm6, xm7
-    vinserti128     m6, m6, xm8, 1
-    pmaddwd         m8, m6, [r5 + 1 * mmsize]
-    pmaddwd         m6, [r5]
-    paddd           m4, m8                          ;res row4 done(4,5,6,7)
-    lea             r0, [r0 + r1 * 4]
-    movu            xm8, [r0]                       ; m8 = row 8
-    punpckhwd       xm9, xm7, xm8
-    punpcklwd       xm7, xm8
-    vinserti128     m7, m7, xm9, 1
-    pmaddwd         m9, m7, [r5 + 1 * mmsize]
-    pmaddwd         m7, [r5]
-    paddd           m5, m9                          ;res row5 done(5,6,7,8)
-    movu            xm9, [r0 + r1]                  ; m9 = row 9
-    punpckhwd       xm10, xm8, xm9
-    punpcklwd       xm8, xm9
-    vinserti128     m8, m8, xm10, 1
-    pmaddwd         m8, [r5 + 1 * mmsize]
-    paddd           m6, m8                          ;res row6 done(6,7,8,9)
-    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
-    punpckhwd       xm8, xm9, xm10
-    punpcklwd       xm9, xm10
-    vinserti128     m9, m9, xm8, 1
-    pmaddwd         m9, [r5 + 1 * mmsize]
-    paddd           m7, m9                          ;res row7 done 7,8,9,10
-    lea             r4, [r3 * 3]
-%ifnidn %1,ss
-    paddd           m0, m11
-    paddd           m1, m11
-    paddd           m2, m11
-    paddd           m3, m11
-%endif
-    psrad           m0, %3
-    psrad           m1, %3
-    psrad           m2, %3
-    psrad           m3, %3
-    packssdw        m0, m1
-    packssdw        m2, m3
-    vpermq          m0, m0, q3120
-    vpermq          m2, m2, q3120
-    pxor            m1, m1
-    mova            m3, [pw_pixel_max]
-%if %2
-    CLIPW           m0, m1, m3
-    CLIPW           m2, m1, m3
-%endif
-    vextracti128    xm9, m0, 1
-    vextracti128    xm8, m2, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm9
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r4], xm8
-%ifnidn %1,ss
-    paddd           m4, m11
-    paddd           m5, m11
-    paddd           m6, m11
-    paddd           m7, m11
-%endif
-    psrad           m4, %3
-    psrad           m5, %3
-    psrad           m6, %3
-    psrad           m7, %3
-    packssdw        m4, m5
-    packssdw        m6, m7
-    vpermq          m4, m4, q3120
-    vpermq          m6, m6, q3120
-%if %2
-    CLIPW           m4, m1, m3
-    CLIPW           m6, m1, m3
-%endif
-    vextracti128    xm5, m4, 1
-    vextracti128    xm7, m6, 1
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm4
-    movu            [r2 + r3], xm5
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r4], xm7
-    RET
-%endif
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_8x8 pp, 1, 6
-FILTER_VER_CHROMA_AVX2_8x8 ps, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_8x8 sp, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_8x8 ss, 0, 6
-
-%macro FILTER_VER_CHROMA_AVX2_8x6 3
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_8x6, 4, 6, 12
-    mov             r4d, r4m
-    add             r1d, r1d
-    add             r3d, r3d
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-
-%ifidn %1,pp
-    vbroadcasti128  m11, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m11, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m11, [INTERP_OFFSET_PS]
-%endif
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m4                          ; r0 done(0,1,2,3)
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    pmaddwd         m3, [r5]
-    paddd           m1, m5                          ;r1 done(1, 2, 3, 4)
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    pmaddwd         m4, [r5]
-    paddd           m2, m6                          ;r2 done(2,3,4,5)
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm7, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddwd         m7, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m7                          ;r3 done(3,4,5,6)
-    movu            xm7, [r0 + r4]                  ; m7 = row 7
-    punpckhwd       xm8, xm6, xm7
-    punpcklwd       xm6, xm7
-    vinserti128     m6, m6, xm8, 1
-    pmaddwd         m8, m6, [r5 + 1 * mmsize]
-    paddd           m4, m8                          ;r4 done(4,5,6,7)
-    lea             r0, [r0 + r1 * 4]
-    movu            xm8, [r0]                       ; m8 = row 8
-    punpckhwd       xm9, xm7, xm8
-    punpcklwd       xm7, xm8
-    vinserti128     m7, m7, xm9, 1
-    pmaddwd         m7, m7, [r5 + 1 * mmsize]
-    paddd           m5, m7                          ;r5 done(5,6,7,8)
-    lea             r4, [r3 * 3]
-%ifnidn %1,ss
-    paddd           m0, m11
-    paddd           m1, m11
-    paddd           m2, m11
-    paddd           m3, m11
-%endif
-    psrad           m0, %3
-    psrad           m1, %3
-    psrad           m2, %3
-    psrad           m3, %3
-    packssdw        m0, m1
-    packssdw        m2, m3
-    vpermq          m0, m0, q3120
-    vpermq          m2, m2, q3120
-    pxor            m10, m10
-    mova            m9, [pw_pixel_max]
-%if %2
-    CLIPW           m0, m10, m9
-    CLIPW           m2, m10, m9
-%endif
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r4], xm3
-%ifnidn %1,ss
-    paddd           m4, m11
-    paddd           m5, m11
-%endif
-    psrad           m4, %3
-    psrad           m5, %3
-    packssdw        m4, m5
-    vpermq          m4, m4, 11011000b
-%if %2
-    CLIPW           m4, m10, m9
-%endif
-    vextracti128    xm5, m4, 1
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm4
-    movu            [r2 + r3], xm5
-    RET
-%endif
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_8x6 pp, 1, 6
-FILTER_VER_CHROMA_AVX2_8x6 ps, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_8x6 sp, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_8x6 ss, 0, 6
-
-%macro PROCESS_CHROMA_AVX2 3
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m4, [r5 + 1 * mmsize]
-    paddd           m2, m4
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm4, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm4, 1
-    pmaddwd         m5, [r5 + 1 * mmsize]
-    paddd           m3, m5
-%ifnidn %1,ss
-    paddd           m0, m7
-    paddd           m1, m7
-    paddd           m2, m7
-    paddd           m3, m7
-%endif
-    psrad           m0, %3
-    psrad           m1, %3
-    psrad           m2, %3
-    psrad           m3, %3
-    packssdw        m0, m1
-    packssdw        m2, m3
-    vpermq          m0, m0, q3120
-    vpermq          m2, m2, q3120
-    pxor            m4, m4
-%if %2
-    CLIPW           m0, m4, [pw_pixel_max]
-    CLIPW           m2, m4, [pw_pixel_max]
-%endif
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m2, 1
-%endmacro
-
-
-%macro FILTER_VER_CHROMA_AVX2_8x4 3
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x4, 4, 6, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    add             r3d, r3d
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    vbroadcasti128  m7, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m7, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m7, [INTERP_OFFSET_PS]
-%endif
-    PROCESS_CHROMA_AVX2 %1, %2, %3
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    lea             r4, [r3 * 3]
-    movu            [r2 + r4], xm3
-    RET
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_8x4 pp, 1, 6
-FILTER_VER_CHROMA_AVX2_8x4 ps, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_8x4 sp, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_8x4 ss, 0, 6
-
-%macro FILTER_VER_CHROMA_AVX2_8x12 3
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_8x12, 4, 7, 15
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    add             r3d, r3d
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    vbroadcasti128  m14, [pd_32]
-%elifidn %1, sp
-    vbroadcasti128  m14, [INTERP_OFFSET_SP]
-%else
-    vbroadcasti128  m14, [INTERP_OFFSET_PS]
-%endif
-    lea             r6, [r3 * 3]
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm7, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddwd         m7, m5, [r5 + 1 * mmsize]
-    paddd           m3, m7
-    pmaddwd         m5, [r5]
-    movu            xm7, [r0 + r4]                  ; m7 = row 7
-    punpckhwd       xm8, xm6, xm7
-    punpcklwd       xm6, xm7
-    vinserti128     m6, m6, xm8, 1
-    pmaddwd         m8, m6, [r5 + 1 * mmsize]
-    paddd           m4, m8
-    pmaddwd         m6, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm8, [r0]                       ; m8 = row 8
-    punpckhwd       xm9, xm7, xm8
-    punpcklwd       xm7, xm8
-    vinserti128     m7, m7, xm9, 1
-    pmaddwd         m9, m7, [r5 + 1 * mmsize]
-    paddd           m5, m9
-    pmaddwd         m7, [r5]
-    movu            xm9, [r0 + r1]                  ; m9 = row 9
-    punpckhwd       xm10, xm8, xm9
-    punpcklwd       xm8, xm9
-    vinserti128     m8, m8, xm10, 1
-    pmaddwd         m10, m8, [r5 + 1 * mmsize]
-    paddd           m6, m10
-    pmaddwd         m8, [r5]
-    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
-    punpckhwd       xm11, xm9, xm10
-    punpcklwd       xm9, xm10
-    vinserti128     m9, m9, xm11, 1
-    pmaddwd         m11, m9, [r5 + 1 * mmsize]
-    paddd           m7, m11
-    pmaddwd         m9, [r5]
-    movu            xm11, [r0 + r4]                 ; m11 = row 11
-    punpckhwd       xm12, xm10, xm11
-    punpcklwd       xm10, xm11
-    vinserti128     m10, m10, xm12, 1
-    pmaddwd         m12, m10, [r5 + 1 * mmsize]
-    paddd           m8, m12
-    pmaddwd         m10, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm12, [r0]                      ; m12 = row 12
-    punpckhwd       xm13, xm11, xm12
-    punpcklwd       xm11, xm12
-    vinserti128     m11, m11, xm13, 1
-    pmaddwd         m13, m11, [r5 + 1 * mmsize]
-    paddd           m9, m13
-    pmaddwd         m11, [r5]
-%ifnidn %1,ss
-    paddd           m0, m14
-    paddd           m1, m14
-    paddd           m2, m14
-    paddd           m3, m14
-    paddd           m4, m14
-    paddd           m5, m14
-%endif
-    psrad           m0, %3
-    psrad           m1, %3
-    psrad           m2, %3
-    psrad           m3, %3
-    psrad           m4, %3
-    psrad           m5, %3
-    packssdw        m0, m1
-    packssdw        m2, m3
-    packssdw        m4, m5
-    vpermq          m0, m0, q3120
-    vpermq          m2, m2, q3120
-    vpermq          m4, m4, q3120
-    pxor            m5, m5
-    mova            m3, [pw_pixel_max]
-%if %2
-    CLIPW           m0, m5, m3
-    CLIPW           m2, m5, m3
-    CLIPW           m4, m5, m3
-%endif
-    vextracti128    xm1, m0, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    vextracti128    xm1, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm1
-    lea             r2, [r2 + r3 * 4]
-    vextracti128    xm1, m4, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm1
-    movu            xm13, [r0 + r1]                 ; m13 = row 13
-    punpckhwd       xm0, xm12, xm13
-    punpcklwd       xm12, xm13
-    vinserti128     m12, m12, xm0, 1
-    pmaddwd         m12, m12, [r5 + 1 * mmsize]
-    paddd           m10, m12
-    movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
-    punpckhwd       xm1, xm13, xm0
-    punpcklwd       xm13, xm0
-    vinserti128     m13, m13, xm1, 1
-    pmaddwd         m13, m13, [r5 + 1 * mmsize]
-    paddd           m11, m13
-%ifnidn %1,ss
-    paddd           m6, m14
-    paddd           m7, m14
-    paddd           m8, m14
-    paddd           m9, m14
-    paddd           m10, m14
-    paddd           m11, m14
-%endif
-    psrad           m6, %3
-    psrad           m7, %3
-    psrad           m8, %3
-    psrad           m9, %3
-    psrad           m10, %3
-    psrad           m11, %3
-    packssdw        m6, m7
-    packssdw        m8, m9
-    packssdw        m10, m11
-    vpermq          m6, m6, q3120
-    vpermq          m8, m8, q3120
-    vpermq          m10, m10, q3120
-%if %2
-    CLIPW           m6, m5, m3
-    CLIPW           m8, m5, m3
-    CLIPW           m10, m5, m3
-%endif
-    vextracti128    xm7, m6, 1
-    vextracti128    xm9, m8, 1
-    vextracti128    xm11, m10, 1
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r6], xm7
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm8
-    movu            [r2 + r3], xm9
-    movu            [r2 + r3 * 2], xm10
-    movu            [r2 + r6], xm11
-    RET
-%endif
-%endmacro
-
-FILTER_VER_CHROMA_AVX2_8x12 pp, 1, 6
-FILTER_VER_CHROMA_AVX2_8x12 ps, 0, INTERP_SHIFT_PS
-FILTER_VER_CHROMA_AVX2_8x12 sp, 1, INTERP_SHIFT_SP
-FILTER_VER_CHROMA_AVX2_8x12 ss, 0, 6

x265_2.6.tar.gz/source/common/x86/ipfilter8.asm -> x265_2.7.tar.gz/source/common/x86/ipfilter8.asm Changed

@@ -33,119 +33,16 @@
 
 const interp4_vpp_shuf, times 2 db 0, 4, 1, 5, 2, 6, 3, 7, 8, 12, 9, 13, 10, 14, 11, 15
 
-const interp_vert_shuf, times 2 db 0, 2, 1, 3, 2, 4, 3, 5, 4, 6, 5, 7, 6, 8, 7, 9
-                        times 2 db 4, 6, 5, 7, 6, 8, 7, 9, 8, 10, 9, 11, 10, 12, 11, 13
-
 const interp4_vpp_shuf1, dd 0, 1, 1, 2, 2, 3, 3, 4
                          dd 2, 3, 3, 4, 4, 5, 5, 6
 
-const pb_8tap_hps_0, times 2 db 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8
-                     times 2 db 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,10
-                     times 2 db 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,10,10,11,11,12
-                     times 2 db 6, 7, 7, 8, 8, 9, 9,10,10,11,11,12,12,13,13,14
-
 const tab_Lm,    db 0, 1, 2, 3, 4,  5,  6,  7,  1, 2, 3, 4,  5,  6,  7,  8
                  db 2, 3, 4, 5, 6,  7,  8,  9,  3, 4, 5, 6,  7,  8,  9,  10
                  db 4, 5, 6, 7, 8,  9,  10, 11, 5, 6, 7, 8,  9,  10, 11, 12
                  db 6, 7, 8, 9, 10, 11, 12, 13, 7, 8, 9, 10, 11, 12, 13, 14
 
-const tab_Vm,    db 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1
-                 db 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3
-
-const tab_Cm,    db 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3
-
 const pd_526336, times 8 dd 8192*64+2048
 
-const tab_ChromaCoeff, db  0, 64,  0,  0
-                       db -2, 58, 10, -2
-                       db -4, 54, 16, -2
-                       db -6, 46, 28, -4
-                       db -4, 36, 36, -4
-                       db -4, 28, 46, -6
-                       db -2, 16, 54, -4
-                       db -2, 10, 58, -2
-
-const tabw_ChromaCoeff, dw  0, 64,  0,  0
-                        dw -2, 58, 10, -2
-                        dw -4, 54, 16, -2
-                        dw -6, 46, 28, -4
-                        dw -4, 36, 36, -4
-                        dw -4, 28, 46, -6
-                        dw -2, 16, 54, -4
-                        dw -2, 10, 58, -2
-
-const tab_ChromaCoeff_V, times 8 db 0, 64
-                         times 8 db 0,  0
-
-                         times 8 db -2, 58
-                         times 8 db 10, -2
-
-                         times 8 db -4, 54
-                         times 8 db 16, -2
-
-                         times 8 db -6, 46
-                         times 8 db 28, -4
-
-                         times 8 db -4, 36
-                         times 8 db 36, -4
-
-                         times 8 db -4, 28
-                         times 8 db 46, -6
-
-                         times 8 db -2, 16
-                         times 8 db 54, -4
-
-                         times 8 db -2, 10
-                         times 8 db 58, -2
-
-const tab_ChromaCoeffV, times 4 dw 0, 64
-                        times 4 dw 0, 0
-
-                        times 4 dw -2, 58
-                        times 4 dw 10, -2
-
-                        times 4 dw -4, 54
-                        times 4 dw 16, -2
-
-                        times 4 dw -6, 46
-                        times 4 dw 28, -4
-
-                        times 4 dw -4, 36
-                        times 4 dw 36, -4
-
-                        times 4 dw -4, 28
-                        times 4 dw 46, -6
-
-                        times 4 dw -2, 16
-                        times 4 dw 54, -4
-
-                        times 4 dw -2, 10
-                        times 4 dw 58, -2
-
-const pw_ChromaCoeffV,  times 8 dw 0, 64
-                        times 8 dw 0, 0
-
-                        times 8 dw -2, 58
-                        times 8 dw 10, -2
-
-                        times 8 dw -4, 54
-                        times 8 dw 16, -2
-
-                        times 8 dw -6, 46
-                        times 8 dw 28, -4
-
-                        times 8 dw -4, 36
-                        times 8 dw 36, -4
-
-                        times 8 dw -4, 28
-                        times 8 dw 46, -6
-
-                        times 8 dw -2, 16
-                        times 8 dw 54, -4
-
-                        times 8 dw -2, 10
-                        times 8 dw 58, -2
-
 const tab_LumaCoeff,   db   0, 0,  0,  64,  0,   0,  0,  0
                        db  -1, 4, -10, 58,  17, -5,  1,  0
                        db  -1, 4, -11, 40,  40, -11, 4, -1
@@ -196,26 +93,6 @@
                         times 8 dw 58, -10
                         times 8 dw 4, -1
 
-const pb_LumaCoeffVer,  times 16 db 0, 0
-                        times 16 db 0, 64
-                        times 16 db 0, 0
-                        times 16 db 0, 0
-
-                        times 16 db -1, 4
-                        times 16 db -10, 58
-                        times 16 db 17, -5
-                        times 16 db 1, 0
-
-                        times 16 db -1, 4
-                        times 16 db -11, 40
-                        times 16 db 40, -11
-                        times 16 db 4, -1
-
-                        times 16 db 0, 1
-                        times 16 db -5, 17
-                        times 16 db 58, -10
-                        times 16 db 4, -1
-
 const tab_LumaCoeffVer, times 8 db 0, 0
                         times 8 db 0, 64
                         times 8 db 0, 0
@@ -256,44 +133,10 @@
                             times 16 db 58, -10
                             times 16 db 4, -1
 
-const tab_ChromaCoeffVer_32,    times 16 db 0, 64
-                                times 16 db 0, 0
-
-                                times 16 db -2, 58
-                                times 16 db 10, -2
-
-                                times 16 db -4, 54
-                                times 16 db 16, -2
-
-                                times 16 db -6, 46
-                                times 16 db 28, -4
-
-                                times 16 db -4, 36
-                                times 16 db 36, -4
-
-                                times 16 db -4, 28
-                                times 16 db 46, -6
-
-                                times 16 db -2, 16
-                                times 16 db 54, -4
-
-                                times 16 db -2, 10
-                                times 16 db 58, -2
-
 const tab_c_64_n64, times 8 db 64, -64
 
-const interp4_shuf, times 2 db 0, 1, 8, 9, 4, 5, 12, 13, 2, 3, 10, 11, 6, 7, 14, 15
-
-const interp4_horiz_shuf1,  db 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
-                            db 8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 11, 12, 13, 14
-
-const interp4_hpp_shuf,     times 2 db 0, 1, 2, 3, 1, 2, 3, 4, 8, 9, 10, 11, 9, 10, 11, 12
-
 const interp8_hps_shuf,     dd 0, 4, 1, 5, 2, 6, 3, 7
 
-ALIGN 32
-interp4_hps_shuf: times 2 db 0, 1, 2, 3, 1, 2, 3, 4, 8, 9, 10, 11, 9, 10, 11, 12
-
 SECTION .text
 
 cextern pb_128
@@ -303,462 +146,6 @@
 cextern pw_2000
 cextern pw_8192
 
-%macro FILTER_H4_w2_2_sse2 0
-    pxor        m3, m3
-    movd        m0, [srcq - 1]
-    movd        m2, [srcq]
-    punpckldq   m0, m2
-    punpcklbw   m0, m3
-    movd        m1, [srcq + srcstrideq - 1]
-    movd        m2, [srcq + srcstrideq]
-    punpckldq   m1, m2
-    punpcklbw   m1, m3
-    pmaddwd     m0, m4
-    pmaddwd     m1, m4
-    packssdw    m0, m1
-    pshuflw     m1, m0, q2301
-    pshufhw     m1, m1, q2301
-    paddw       m0, m1
-    psrld       m0, 16
-    packssdw    m0, m0
-    paddw       m0, m5
-    psraw       m0, 6
-    packuswb    m0, m0
-    movd        r4, m0
-    mov         [dstq], r4w
-    shr         r4, 16
-    mov         [dstq + dststrideq], r4w
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_2xN(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_H4_W2xN_sse3 1
-INIT_XMM sse3
-cglobal interp_4tap_horiz_pp_2x%1, 4, 6, 6, src, srcstride, dst, dststride
-    mov         r4d,    r4m
-    mova        m5,     [pw_32]
-
-%ifdef PIC
-    lea         r5,     [tabw_ChromaCoeff]
-    movddup     m4,     [r5 + r4 * 8]
-%else
-    movddup     m4,     [tabw_ChromaCoeff + r4 * 8]
-%endif
-
-%assign x 1
-%rep %1/2
-    FILTER_H4_w2_2_sse2
-%if x < %1/2
-    lea         srcq,   [srcq + srcstrideq * 2]
-    lea         dstq,   [dstq + dststrideq * 2]
-%endif
-%assign x x+1
-%endrep
-
-    RET
-
-%endmacro
-
-    FILTER_H4_W2xN_sse3 4
-    FILTER_H4_W2xN_sse3 8
-    FILTER_H4_W2xN_sse3 16
-
-%macro FILTER_H4_w4_2_sse2 0
-    pxor        m5, m5
-    movd        m0, [srcq - 1]
-    movd        m6, [srcq]
-    punpckldq   m0, m6
-    punpcklbw   m0, m5
-    movd        m1, [srcq + 1]
-    movd        m6, [srcq + 2]
-    punpckldq   m1, m6
-    punpcklbw   m1, m5
-    movd        m2, [srcq + srcstrideq - 1]
-    movd        m6, [srcq + srcstrideq]
-    punpckldq   m2, m6
-    punpcklbw   m2, m5
-    movd        m3, [srcq + srcstrideq + 1]
-    movd        m6, [srcq + srcstrideq + 2]
-    punpckldq   m3, m6
-    punpcklbw   m3, m5
-    pmaddwd     m0, m4
-    pmaddwd     m1, m4
-    pmaddwd     m2, m4
-    pmaddwd     m3, m4
-    packssdw    m0, m1
-    packssdw    m2, m3
-    pshuflw     m1, m0, q2301
-    pshufhw     m1, m1, q2301
-    pshuflw     m3, m2, q2301
-    pshufhw     m3, m3, q2301
-    paddw       m0, m1
-    paddw       m2, m3
-    psrld       m0, 16
-    psrld       m2, 16
-    packssdw    m0, m2
-    paddw       m0, m7
-    psraw       m0, 6
-    packuswb    m0, m2
-    movd        [dstq], m0
-    psrldq      m0, 4
-    movd        [dstq + dststrideq], m0
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_4x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_H4_W4xN_sse3 1
-INIT_XMM sse3
-cglobal interp_4tap_horiz_pp_4x%1, 4, 6, 8, src, srcstride, dst, dststride
-    mov         r4d,    r4m
-    mova        m7,     [pw_32]
-
-%ifdef PIC
-    lea         r5,     [tabw_ChromaCoeff]
-    movddup     m4,     [r5 + r4 * 8]
-%else
-    movddup     m4,     [tabw_ChromaCoeff + r4 * 8]
-%endif
-
-%assign x 1
-%rep %1/2
-    FILTER_H4_w4_2_sse2
-%if x < %1/2
-    lea         srcq,   [srcq + srcstrideq * 2]
-    lea         dstq,   [dstq + dststrideq * 2]
-%endif
-%assign x x+1
-%endrep
-
-    RET
-
-%endmacro
-
-    FILTER_H4_W4xN_sse3 2
-    FILTER_H4_W4xN_sse3 4
-    FILTER_H4_W4xN_sse3 8
-    FILTER_H4_W4xN_sse3 16
-    FILTER_H4_W4xN_sse3 32
-
-%macro FILTER_H4_w6_sse2 0
-    pxor        m4, m4
-    movh        m0, [srcq - 1]
-    movh        m5, [srcq]
-    punpckldq   m0, m5
-    movhlps     m2, m0
-    punpcklbw   m0, m4
-    punpcklbw   m2, m4
-    movd        m1, [srcq + 1]
-    movd        m5, [srcq + 2]
-    punpckldq   m1, m5
-    punpcklbw   m1, m4
-    pmaddwd     m0, m6
-    pmaddwd     m1, m6
-    pmaddwd     m2, m6
-    packssdw    m0, m1
-    packssdw    m2, m2
-    pshuflw     m1, m0, q2301
-    pshufhw     m1, m1, q2301
-    pshuflw     m3, m2, q2301
-    paddw       m0, m1
-    paddw       m2, m3
-    psrld       m0, 16
-    psrld       m2, 16
-    packssdw    m0, m2
-    paddw       m0, m7
-    psraw       m0, 6
-    packuswb    m0, m0
-    movd        [dstq], m0
-    pextrw      r4d, m0, 2
-    mov         [dstq + 4], r4w
-%endmacro
-
-%macro FILH4W8_sse2 1
-    movh        m0, [srcq - 1 + %1]
-    movh        m5, [srcq + %1]
-    punpckldq   m0, m5
-    movhlps     m2, m0
-    punpcklbw   m0, m4
-    punpcklbw   m2, m4
-    movh        m1, [srcq + 1 + %1]
-    movh        m5, [srcq + 2 + %1]
-    punpckldq   m1, m5
-    movhlps     m3, m1
-    punpcklbw   m1, m4
-    punpcklbw   m3, m4
-    pmaddwd     m0, m6
-    pmaddwd     m1, m6
-    pmaddwd     m2, m6
-    pmaddwd     m3, m6
-    packssdw    m0, m1
-    packssdw    m2, m3
-    pshuflw     m1, m0, q2301
-    pshufhw     m1, m1, q2301
-    pshuflw     m3, m2, q2301
-    pshufhw     m3, m3, q2301
-    paddw       m0, m1
-    paddw       m2, m3
-    psrld       m0, 16
-    psrld       m2, 16
-    packssdw    m0, m2
-    paddw       m0, m7
-    psraw       m0, 6
-    packuswb    m0, m0
-    movh        [dstq + %1], m0
-%endmacro
-
-%macro FILTER_H4_w8_sse2 0
-    FILH4W8_sse2 0
-%endmacro
-
-%macro FILTER_H4_w12_sse2 0
-    FILH4W8_sse2 0
-    movd        m1, [srcq - 1 + 8]
-    movd        m3, [srcq + 8]
-    punpckldq   m1, m3
-    punpcklbw   m1, m4
-    movd        m2, [srcq + 1 + 8]
-    movd        m3, [srcq + 2 + 8]
-    punpckldq   m2, m3
-    punpcklbw   m2, m4
-    pmaddwd     m1, m6
-    pmaddwd     m2, m6
-    packssdw    m1, m2
-    pshuflw     m2, m1, q2301
-    pshufhw     m2, m2, q2301
-    paddw       m1, m2
-    psrld       m1, 16
-    packssdw    m1, m1
-    paddw       m1, m7
-    psraw       m1, 6
-    packuswb    m1, m1
-    movd        [dstq + 8], m1
-%endmacro
-
-%macro FILTER_H4_w16_sse2 0
-    FILH4W8_sse2 0
-    FILH4W8_sse2 8
-%endmacro
-
-%macro FILTER_H4_w24_sse2 0
-    FILH4W8_sse2 0
-    FILH4W8_sse2 8
-    FILH4W8_sse2 16
-%endmacro
-
-%macro FILTER_H4_w32_sse2 0
-    FILH4W8_sse2 0
-    FILH4W8_sse2 8
-    FILH4W8_sse2 16
-    FILH4W8_sse2 24
-%endmacro
-
-%macro FILTER_H4_w48_sse2 0
-    FILH4W8_sse2 0
-    FILH4W8_sse2 8
-    FILH4W8_sse2 16
-    FILH4W8_sse2 24
-    FILH4W8_sse2 32
-    FILH4W8_sse2 40
-%endmacro
-
-%macro FILTER_H4_w64_sse2 0
-    FILH4W8_sse2 0
-    FILH4W8_sse2 8
-    FILH4W8_sse2 16
-    FILH4W8_sse2 24
-    FILH4W8_sse2 32
-    FILH4W8_sse2 40
-    FILH4W8_sse2 48
-    FILH4W8_sse2 56
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_sse3 2
-INIT_XMM sse3
-cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 8, src, srcstride, dst, dststride
-    mov         r4d,        r4m
-    mova        m7,         [pw_32]
-    pxor        m4,         m4
-
-%ifdef PIC
-    lea         r5,          [tabw_ChromaCoeff]
-    movddup     m6,       [r5 + r4 * 8]
-%else
-    movddup     m6,       [tabw_ChromaCoeff + r4 * 8]
-%endif
-
-%assign x 1
-%rep %2
-    FILTER_H4_w%1_sse2
-%if x < %2
-    add         srcq,        srcstrideq
-    add         dstq,        dststrideq
-%endif
-%assign x x+1
-%endrep
-
-    RET
-
-%endmacro
-
-    IPFILTER_CHROMA_sse3 6,   8
-    IPFILTER_CHROMA_sse3 8,   2
-    IPFILTER_CHROMA_sse3 8,   4
-    IPFILTER_CHROMA_sse3 8,   6
-    IPFILTER_CHROMA_sse3 8,   8
-    IPFILTER_CHROMA_sse3 8,  16
-    IPFILTER_CHROMA_sse3 8,  32
-    IPFILTER_CHROMA_sse3 12, 16
-
-    IPFILTER_CHROMA_sse3 6,  16
-    IPFILTER_CHROMA_sse3 8,  12
-    IPFILTER_CHROMA_sse3 8,  64
-    IPFILTER_CHROMA_sse3 12, 32
-
-    IPFILTER_CHROMA_sse3 16,  4
-    IPFILTER_CHROMA_sse3 16,  8
-    IPFILTER_CHROMA_sse3 16, 12
-    IPFILTER_CHROMA_sse3 16, 16
-    IPFILTER_CHROMA_sse3 16, 32
-    IPFILTER_CHROMA_sse3 32,  8
-    IPFILTER_CHROMA_sse3 32, 16
-    IPFILTER_CHROMA_sse3 32, 24
-    IPFILTER_CHROMA_sse3 24, 32
-    IPFILTER_CHROMA_sse3 32, 32
-
-    IPFILTER_CHROMA_sse3 16, 24
-    IPFILTER_CHROMA_sse3 16, 64
-    IPFILTER_CHROMA_sse3 32, 48
-    IPFILTER_CHROMA_sse3 24, 64
-    IPFILTER_CHROMA_sse3 32, 64
-
-    IPFILTER_CHROMA_sse3 64, 64
-    IPFILTER_CHROMA_sse3 64, 32
-    IPFILTER_CHROMA_sse3 64, 48
-    IPFILTER_CHROMA_sse3 48, 64
-    IPFILTER_CHROMA_sse3 64, 16
-
-%macro FILTER_2 2
-    movd        m3,     [srcq + %1]
-    movd        m4,     [srcq + 1 + %1]
-    punpckldq   m3,     m4
-    punpcklbw   m3,     m0
-    pmaddwd     m3,     m1
-    packssdw    m3,     m3
-    pshuflw     m4,     m3, q2301
-    paddw       m3,     m4
-    psrldq      m3,     2
-    psubw       m3,     m2
-    movd        [dstq + %2], m3
-%endmacro
-
-%macro FILTER_4 2
-    movd        m3,     [srcq + %1]
-    movd        m4,     [srcq + 1 + %1]
-    punpckldq   m3,     m4
-    punpcklbw   m3,     m0
-    pmaddwd     m3,     m1
-    movd        m4,     [srcq + 2 + %1]
-    movd        m5,     [srcq + 3 + %1]
-    punpckldq   m4,     m5
-    punpcklbw   m4,     m0
-    pmaddwd     m4,     m1
-    packssdw    m3,     m4
-    pshuflw     m4,     m3, q2301
-    pshufhw     m4,     m4, q2301
-    paddw       m3,     m4
-    psrldq      m3,     2
-    pshufd      m3,     m3,     q3120
-    psubw       m3,     m2
-    movh        [dstq + %2], m3
-%endmacro
-
-%macro FILTER_4TAP_HPS_sse3 2
-INIT_XMM sse3
-cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
-    mov         r4d,    r4m
-    add         dststrided, dststrided
-    mova        m2,     [pw_2000]
-    pxor        m0,     m0
-
-%ifdef PIC
-    lea         r6,     [tabw_ChromaCoeff]
-    movddup     m1,     [r6 + r4 * 8]
-%else
-    movddup     m1,     [tabw_ChromaCoeff + r4 * 8]
-%endif
-
-    mov        r4d,     %2
-    cmp        r5m,     byte 0
-    je         .loopH
-    sub        srcq,    srcstrideq
-    add        r4d,     3
-
-.loopH:
-%assign x -1
-%assign y 0
-%rep %1/4
-    FILTER_4 x,y
-%assign x x+4
-%assign y y+8
-%endrep
-%rep (%1 % 4)/2
-    FILTER_2 x,y
-%endrep
-    add         srcq,   srcstrideq
-    add         dstq,   dststrideq
-
-    dec         r4d
-    jnz         .loopH
-    RET
-
-%endmacro
-
-    FILTER_4TAP_HPS_sse3 2, 4
-    FILTER_4TAP_HPS_sse3 2, 8
-    FILTER_4TAP_HPS_sse3 2, 16
-    FILTER_4TAP_HPS_sse3 4, 2
-    FILTER_4TAP_HPS_sse3 4, 4
-    FILTER_4TAP_HPS_sse3 4, 8
-    FILTER_4TAP_HPS_sse3 4, 16
-    FILTER_4TAP_HPS_sse3 4, 32
-    FILTER_4TAP_HPS_sse3 6, 8
-    FILTER_4TAP_HPS_sse3 6, 16
-    FILTER_4TAP_HPS_sse3 8, 2
-    FILTER_4TAP_HPS_sse3 8, 4
-    FILTER_4TAP_HPS_sse3 8, 6
-    FILTER_4TAP_HPS_sse3 8, 8
-    FILTER_4TAP_HPS_sse3 8, 12
-    FILTER_4TAP_HPS_sse3 8, 16
-    FILTER_4TAP_HPS_sse3 8, 32
-    FILTER_4TAP_HPS_sse3 8, 64
-    FILTER_4TAP_HPS_sse3 12, 16
-    FILTER_4TAP_HPS_sse3 12, 32
-    FILTER_4TAP_HPS_sse3 16, 4
-    FILTER_4TAP_HPS_sse3 16, 8
-    FILTER_4TAP_HPS_sse3 16, 12
-    FILTER_4TAP_HPS_sse3 16, 16
-    FILTER_4TAP_HPS_sse3 16, 24
-    FILTER_4TAP_HPS_sse3 16, 32
-    FILTER_4TAP_HPS_sse3 16, 64
-    FILTER_4TAP_HPS_sse3 24, 32
-    FILTER_4TAP_HPS_sse3 24, 64
-    FILTER_4TAP_HPS_sse3 32,  8
-    FILTER_4TAP_HPS_sse3 32, 16
-    FILTER_4TAP_HPS_sse3 32, 24
-    FILTER_4TAP_HPS_sse3 32, 32
-    FILTER_4TAP_HPS_sse3 32, 48
-    FILTER_4TAP_HPS_sse3 32, 64
-    FILTER_4TAP_HPS_sse3 48, 64
-    FILTER_4TAP_HPS_sse3 64, 16
-    FILTER_4TAP_HPS_sse3 64, 32
-    FILTER_4TAP_HPS_sse3 64, 48
-    FILTER_4TAP_HPS_sse3 64, 64
-
 %macro FILTER_H8_W8_sse2 0
     movh        m1, [r0 + x - 3]
     movh        m4, [r0 + x - 2]
@@ -855,137 +242,6 @@
     psrldq      m1, 2
     pshufd      m1, m1, q3120
 %endmacro
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_LUMA_sse2 3
-INIT_XMM sse2
-cglobal interp_8tap_horiz_%3_%1x%2, 4,6,8
-    mov       r4d, r4m
-    add       r4d, r4d
-    pxor      m6, m6
-
-%ifidn %3, ps
-    add       r3d, r3d
-    cmp       r5m, byte 0
-%endif
-
-%ifdef PIC
-    lea       r5, [tabw_LumaCoeff]
-    movu      m3, [r5 + r4 * 8]
-%else
-    movu      m3, [tabw_LumaCoeff + r4 * 8]
-%endif
-
-    mov       r4d, %2
-
-%ifidn %3, pp
-    mova      m2, [pw_32]
-%else
-    mova      m2, [pw_2000]
-    je        .loopH
-    lea       r5, [r1 + 2 * r1]
-    sub       r0, r5
-    add       r4d, 7
-%endif
-
-.loopH:
-%assign x 0
-%rep %1 / 8
-    FILTER_H8_W8_sse2
-  %ifidn %3, pp
-    paddw     m1, m2
-    psraw     m1, 6
-    packuswb  m1, m1
-    movh      [r2 + x], m1
-  %else
-    psubw     m1, m2
-    movu      [r2 + 2 * x], m1
-  %endif
-%assign x x+8
-%endrep
-
-%rep (%1 % 8) / 4
-    FILTER_H8_W4_sse2
-  %ifidn %3, pp
-    paddw     m1, m2
-    psraw     m1, 6
-    packuswb  m1, m1
-    movd      [r2 + x], m1
-  %else
-    psubw     m1, m2
-    movh      [r2 + 2 * x], m1
-  %endif
-%endrep
-
-    add       r0, r1
-    add       r2, r3
-
-    dec       r4d
-    jnz       .loopH
-    RET
-
-%endmacro
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-    IPFILTER_LUMA_sse2 4, 4, pp
-    IPFILTER_LUMA_sse2 4, 8, pp
-    IPFILTER_LUMA_sse2 8, 4, pp
-    IPFILTER_LUMA_sse2 8, 8, pp
-    IPFILTER_LUMA_sse2 16, 16, pp
-    IPFILTER_LUMA_sse2 16, 8, pp
-    IPFILTER_LUMA_sse2 8, 16, pp
-    IPFILTER_LUMA_sse2 16, 12, pp
-    IPFILTER_LUMA_sse2 12, 16, pp
-    IPFILTER_LUMA_sse2 16, 4, pp
-    IPFILTER_LUMA_sse2 4, 16, pp
-    IPFILTER_LUMA_sse2 32, 32, pp
-    IPFILTER_LUMA_sse2 32, 16, pp
-    IPFILTER_LUMA_sse2 16, 32, pp
-    IPFILTER_LUMA_sse2 32, 24, pp
-    IPFILTER_LUMA_sse2 24, 32, pp
-    IPFILTER_LUMA_sse2 32, 8, pp
-    IPFILTER_LUMA_sse2 8, 32, pp
-    IPFILTER_LUMA_sse2 64, 64, pp
-    IPFILTER_LUMA_sse2 64, 32, pp
-    IPFILTER_LUMA_sse2 32, 64, pp
-    IPFILTER_LUMA_sse2 64, 48, pp
-    IPFILTER_LUMA_sse2 48, 64, pp
-    IPFILTER_LUMA_sse2 64, 16, pp
-    IPFILTER_LUMA_sse2 16, 64, pp
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-    IPFILTER_LUMA_sse2 4, 4, ps
-    IPFILTER_LUMA_sse2 8, 8, ps
-    IPFILTER_LUMA_sse2 8, 4, ps
-    IPFILTER_LUMA_sse2 4, 8, ps
-    IPFILTER_LUMA_sse2 16, 16, ps
-    IPFILTER_LUMA_sse2 16, 8, ps
-    IPFILTER_LUMA_sse2 8, 16, ps
-    IPFILTER_LUMA_sse2 16, 12, ps
-    IPFILTER_LUMA_sse2 12, 16, ps
-    IPFILTER_LUMA_sse2 16, 4, ps
-    IPFILTER_LUMA_sse2 4, 16, ps
-    IPFILTER_LUMA_sse2 32, 32, ps
-    IPFILTER_LUMA_sse2 32, 16, ps
-    IPFILTER_LUMA_sse2 16, 32, ps
-    IPFILTER_LUMA_sse2 32, 24, ps
-    IPFILTER_LUMA_sse2 24, 32, ps
-    IPFILTER_LUMA_sse2 32, 8, ps
-    IPFILTER_LUMA_sse2 8, 32, ps
-    IPFILTER_LUMA_sse2 64, 64, ps
-    IPFILTER_LUMA_sse2 64, 32, ps
-    IPFILTER_LUMA_sse2 32, 64, ps
-    IPFILTER_LUMA_sse2 64, 48, ps
-    IPFILTER_LUMA_sse2 48, 64, ps
-    IPFILTER_LUMA_sse2 64, 16, ps
-    IPFILTER_LUMA_sse2 16, 64, ps
-
 %macro PROCESS_LUMA_W4_4R_sse2 0
     movd        m2,     [r0]
     movd        m7,     [r0 + r1]
@@ -1345,1945 +601,6 @@
     FILTER_VER_LUMA_sse2 64, 64, ps
 %endif
 
-%macro  WORD_TO_DOUBLE 1
-%if ARCH_X86_64
-    punpcklbw   %1,     m8
-%else
-    punpcklbw   %1,     %1
-    psrlw       %1,     8
-%endif
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_2x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W2_H4_sse2 2
-INIT_XMM sse2
-%if ARCH_X86_64
-cglobal interp_4tap_vert_%1_2x%2, 4, 6, 9
-    pxor        m8,        m8
-%else
-cglobal interp_4tap_vert_%1_2x%2, 4, 6, 8
-%endif
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifidn %1,pp
-    mova        m1,        [pw_32]
-%elifidn %1,ps
-    mova        m1,        [pw_2000]
-    add         r3d,       r3d
-%endif
-
-%ifdef PIC
-    lea         r5,        [tabw_ChromaCoeff]
-    movh        m0,        [r5 + r4 * 8]
-%else
-    movh        m0,        [tabw_ChromaCoeff + r4 * 8]
-%endif
-
-    punpcklqdq  m0,        m0
-    lea         r5,        [3 * r1]
-
-%assign x 1
-%rep %2/4
-    movd        m2,        [r0]
-    movd        m3,        [r0 + r1]
-    movd        m4,        [r0 + 2 * r1]
-    movd        m5,        [r0 + r5]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m6,        m4,        m5
-    punpcklwd   m2,        m6
-
-    WORD_TO_DOUBLE         m2
-    pmaddwd     m2,        m0
-
-    lea         r0,        [r0 + 4 * r1]
-    movd        m6,        [r0]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m7,        m5,        m6
-    punpcklwd   m3,        m7
-
-    WORD_TO_DOUBLE         m3
-    pmaddwd     m3,        m0
-
-    packssdw    m2,        m3
-    pshuflw     m3,        m2,          q2301
-    pshufhw     m3,        m3,          q2301
-    paddw       m2,        m3
-
-    movd        m7,        [r0 + r1]
-
-    punpcklbw   m4,        m5
-    punpcklbw   m3,        m6,        m7
-    punpcklwd   m4,        m3
-
-    WORD_TO_DOUBLE         m4
-    pmaddwd     m4,        m0
-
-    movd        m3,        [r0 + 2 * r1]
-
-    punpcklbw   m5,        m6
-    punpcklbw   m7,        m3
-    punpcklwd   m5,        m7
-
-    WORD_TO_DOUBLE         m5
-    pmaddwd     m5,        m0
-
-    packssdw    m4,        m5
-    pshuflw     m5,        m4,          q2301
-    pshufhw     m5,        m5,          q2301
-    paddw       m4,        m5
-
-%ifidn %1,pp
-    psrld       m2,        16
-    psrld       m4,        16
-    packssdw    m2,        m4
-    paddw       m2,        m1
-    psraw       m2,        6
-    packuswb    m2,        m2
-
-%if ARCH_X86_64
-    movq        r4,        m2
-    mov         [r2],      r4w
-    shr         r4,        16
-    mov         [r2 + r3], r4w
-    lea         r2,        [r2 + 2 * r3]
-    shr         r4,        16
-    mov         [r2],      r4w
-    shr         r4,        16
-    mov         [r2 + r3], r4w
-%else
-    movd        r4,        m2
-    mov         [r2],      r4w
-    shr         r4,        16
-    mov         [r2 + r3], r4w
-    lea         r2,        [r2 + 2 * r3]
-    psrldq      m2,        4
-    movd        r4,        m2
-    mov         [r2],      r4w
-    shr         r4,        16
-    mov         [r2 + r3], r4w
-%endif
-%elifidn %1,ps
-    psrldq      m2,        2
-    psrldq      m4,        2
-    pshufd      m2,        m2, q3120
-    pshufd      m4,        m4, q3120
-    psubw       m4,        m1
-    psubw       m2,        m1
-
-    movd        [r2],      m2
-    psrldq      m2,        4
-    movd        [r2 + r3], m2
-    lea         r2,        [r2 + 2 * r3]
-    movd        [r2],      m4
-    psrldq      m4,        4
-    movd        [r2 + r3], m4
-%endif
-
-%if x < %2/4
-    lea         r2,        [r2 + 2 * r3]
-%endif
-%assign x x+1
-%endrep
-    RET
-
-%endmacro
-
-    FILTER_V4_W2_H4_sse2 pp, 4
-    FILTER_V4_W2_H4_sse2 pp, 8
-    FILTER_V4_W2_H4_sse2 pp, 16
-
-    FILTER_V4_W2_H4_sse2 ps, 4
-    FILTER_V4_W2_H4_sse2 ps, 8
-    FILTER_V4_W2_H4_sse2 ps, 16
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro  FILTER_V2_W4_H4_sse2 1
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_4x2, 4, 6, 8
-    mov         r4d,       r4m
-    sub         r0,        r1
-    pxor        m7,        m7
-
-%ifdef PIC
-    lea         r5,        [tabw_ChromaCoeff]
-    movh        m0,        [r5 + r4 * 8]
-%else
-    movh        m0,        [tabw_ChromaCoeff + r4 * 8]
-%endif
-
-    lea         r5,        [r0 + 2 * r1]
-    punpcklqdq  m0,        m0
-    movd        m2,        [r0]
-    movd        m3,        [r0 + r1]
-    movd        m4,        [r5]
-    movd        m5,        [r5 + r1]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m1,        m4,        m5
-    punpcklwd   m2,        m1
-
-    movhlps     m6,        m2
-    punpcklbw   m2,        m7
-    punpcklbw   m6,        m7
-    pmaddwd     m2,        m0
-    pmaddwd     m6,        m0
-    packssdw    m2,        m6
-
-    movd        m1,        [r0 + 4 * r1]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m5,        m1
-    punpcklwd   m3,        m5
-
-    movhlps     m6,        m3
-    punpcklbw   m3,        m7
-    punpcklbw   m6,        m7
-    pmaddwd     m3,        m0
-    pmaddwd     m6,        m0
-    packssdw    m3,        m6
-
-    pshuflw     m4,        m2,        q2301
-    pshufhw     m4,        m4,        q2301
-    paddw       m2,        m4
-    pshuflw     m5,        m3,        q2301
-    pshufhw     m5,        m5,        q2301
-    paddw       m3,        m5
-
-%ifidn %1, pp
-    psrld       m2,        16
-    psrld       m3,        16
-    packssdw    m2,        m3
-
-    paddw       m2,        [pw_32]
-    psraw       m2,        6
-    packuswb    m2,        m2
-
-    movd        [r2],      m2
-    psrldq      m2,        4
-    movd        [r2 + r3], m2
-%elifidn %1, ps
-    psrldq      m2,        2
-    psrldq      m3,        2
-    pshufd      m2,        m2, q3120
-    pshufd      m3,        m3, q3120
-    punpcklqdq  m2, m3
-
-    add         r3d,       r3d
-    psubw       m2,        [pw_2000]
-    movh        [r2],      m2
-    movhps      [r2 + r3], m2
-%endif
-    RET
-
-%endmacro
-
-    FILTER_V2_W4_H4_sse2 pp
-    FILTER_V2_W4_H4_sse2 ps
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_4x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W4_H4_sse2 2
-INIT_XMM sse2
-%if ARCH_X86_64
-cglobal interp_4tap_vert_%1_4x%2, 4, 6, 9
-    pxor        m8,        m8
-%else
-cglobal interp_4tap_vert_%1_4x%2, 4, 6, 8
-%endif
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tabw_ChromaCoeff]
-    movh        m0,        [r5 + r4 * 8]
-%else
-    movh        m0,        [tabw_ChromaCoeff + r4 * 8]
-%endif
-
-%ifidn %1,pp
-    mova        m1,        [pw_32]
-%elifidn %1,ps
-    add         r3d,       r3d
-    mova        m1,        [pw_2000]
-%endif
-
-    lea         r5,        [3 * r1]
-    lea         r4,        [3 * r3]
-    punpcklqdq  m0,        m0
-
-%assign x 1
-%rep %2/4
-    movd        m2,        [r0]
-    movd        m3,        [r0 + r1]
-    movd        m4,        [r0 + 2 * r1]
-    movd        m5,        [r0 + r5]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m6,        m4,        m5
-    punpcklwd   m2,        m6
-
-    movhlps     m6,        m2
-    WORD_TO_DOUBLE         m2
-    WORD_TO_DOUBLE         m6
-    pmaddwd     m2,        m0
-    pmaddwd     m6,        m0
-    packssdw    m2,        m6
-
-    lea         r0,        [r0 + 4 * r1]
-    movd        m6,        [r0]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m7,        m5,        m6
-    punpcklwd   m3,        m7
-
-    movhlps     m7,        m3
-    WORD_TO_DOUBLE         m3
-    WORD_TO_DOUBLE         m7
-    pmaddwd     m3,        m0
-    pmaddwd     m7,        m0
-    packssdw    m3,        m7
-
-    pshuflw     m7,        m2,        q2301
-    pshufhw     m7,        m7,        q2301
-    paddw       m2,        m7
-    pshuflw     m7,        m3,        q2301
-    pshufhw     m7,        m7,        q2301
-    paddw       m3,        m7
-
-%ifidn %1,pp
-    psrld       m2,        16
-    psrld       m3,        16
-    packssdw    m2,        m3
-    paddw       m2,        m1
-    psraw       m2,        6
-%elifidn %1,ps
-    psrldq      m2,        2
-    psrldq      m3,        2
-    pshufd      m2,        m2, q3120
-    pshufd      m3,        m3, q3120
-    punpcklqdq  m2,        m3
-
-    psubw       m2,        m1
-    movh        [r2],      m2
-    movhps      [r2 + r3], m2
-%endif
-
-    movd        m7,        [r0 + r1]
-
-    punpcklbw   m4,        m5
-    punpcklbw   m3,        m6,        m7
-    punpcklwd   m4,        m3
-
-    movhlps     m3,        m4
-    WORD_TO_DOUBLE         m4
-    WORD_TO_DOUBLE         m3
-    pmaddwd     m4,        m0
-    pmaddwd     m3,        m0
-    packssdw    m4,        m3
-
-    movd        m3,        [r0 + 2 * r1]
-
-    punpcklbw   m5,        m6
-    punpcklbw   m7,        m3
-    punpcklwd   m5,        m7
-
-    movhlps     m3,        m5
-    WORD_TO_DOUBLE         m5
-    WORD_TO_DOUBLE         m3
-    pmaddwd     m5,        m0
-    pmaddwd     m3,        m0
-    packssdw    m5,        m3
-
-    pshuflw     m7,        m4,        q2301
-    pshufhw     m7,        m7,        q2301
-    paddw       m4,        m7
-    pshuflw     m7,        m5,        q2301
-    pshufhw     m7,        m7,        q2301
-    paddw       m5,        m7
-
-%ifidn %1,pp
-    psrld       m4,        16
-    psrld       m5,        16
-    packssdw    m4,        m5
-
-    paddw       m4,        m1
-    psraw       m4,        6
-    packuswb    m2,        m4
-
-    movd        [r2],      m2
-    psrldq      m2,        4
-    movd        [r2 + r3], m2
-    psrldq      m2,        4
-    movd        [r2 + 2 * r3],      m2
-    psrldq      m2,        4
-    movd        [r2 + r4], m2
-%elifidn %1,ps
-    psrldq      m4,        2
-    psrldq      m5,        2
-    pshufd      m4,        m4, q3120
-    pshufd      m5,        m5, q3120
-    punpcklqdq  m4,        m5
-    psubw       m4,        m1
-    movh        [r2 + 2 * r3],      m4
-    movhps      [r2 + r4], m4
-%endif
-
-%if x < %2/4
-    lea         r2,        [r2 + 4 * r3]
-%endif
-
-%assign x x+1
-%endrep
-    RET
-
-%endmacro
-
-    FILTER_V4_W4_H4_sse2 pp, 4
-    FILTER_V4_W4_H4_sse2 pp, 8
-    FILTER_V4_W4_H4_sse2 pp, 16
-    FILTER_V4_W4_H4_sse2 pp, 32
-
-    FILTER_V4_W4_H4_sse2 ps, 4
-    FILTER_V4_W4_H4_sse2 ps, 8
-    FILTER_V4_W4_H4_sse2 ps, 16
-    FILTER_V4_W4_H4_sse2 ps, 32
-
-;-----------------------------------------------------------------------------
-;void interp_4tap_vert_%1_6x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W6_H4_sse2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_6x%2, 4, 7, 10
-    mov         r4d,       r4m
-    sub         r0,        r1
-    shl         r4d,       5
-    pxor        m9,        m9
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeffV]
-    mova        m6,        [r5 + r4]
-    mova        m5,        [r5 + r4 + 16]
-%else
-    mova        m6,        [tab_ChromaCoeffV + r4]
-    mova        m5,        [tab_ChromaCoeffV + r4 + 16]
-%endif
-
-%ifidn %1,pp
-    mova        m4,        [pw_32]
-%elifidn %1,ps
-    mova        m4,        [pw_2000]
-    add         r3d,       r3d
-%endif
-    lea         r5,        [3 * r1]
-
-%assign x 1
-%rep %2/4
-    movq        m0,        [r0]
-    movq        m1,        [r0 + r1]
-    movq        m2,        [r0 + 2 * r1]
-    movq        m3,        [r0 + r5]
-
-    punpcklbw   m0,        m1
-    punpcklbw   m1,        m2
-    punpcklbw   m2,        m3
-
-    movhlps     m7,        m0
-    punpcklbw   m0,        m9
-    punpcklbw   m7,        m9
-    pmaddwd     m0,        m6
-    pmaddwd     m7,        m6
-    packssdw    m0,        m7
-
-    movhlps     m8,        m2
-    movq        m7,        m2
-    punpcklbw   m8,        m9
-    punpcklbw   m7,        m9
-    pmaddwd     m8,        m5
-    pmaddwd     m7,        m5
-    packssdw    m7,        m8
-
-    paddw       m0,        m7
-
-%ifidn %1,pp
-    paddw       m0,        m4
-    psraw       m0,        6
-    packuswb    m0,        m0
-
-    movd        [r2],      m0
-    pextrw      r6d,       m0,        2
-    mov         [r2 + 4],  r6w
-%elifidn %1,ps
-    psubw       m0,        m4
-    movh        [r2],      m0
-    pshufd      m0,        m0,        2
-    movd        [r2 + 8],  m0
-%endif
-
-    lea         r0,        [r0 + 4 * r1]
-
-    movq        m0,        [r0]
-    punpcklbw   m3,        m0
-
-    movhlps     m8,        m1
-    punpcklbw   m1,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m1,        m6
-    pmaddwd     m8,        m6
-    packssdw    m1,        m8
-
-    movhlps     m8,        m3
-    movq        m7,        m3
-    punpcklbw   m8,        m9
-    punpcklbw   m7,        m9
-    pmaddwd     m8,        m5
-    pmaddwd     m7,        m5
-    packssdw    m7,        m8
-
-    paddw       m1,        m7
-
-%ifidn %1,pp
-    paddw       m1,        m4
-    psraw       m1,        6
-    packuswb    m1,        m1
-
-    movd        [r2 + r3], m1
-    pextrw      r6d,       m1,        2
-    mov         [r2 + r3 + 4], r6w
-%elifidn %1,ps
-    psubw       m1,        m4
-    movh        [r2 + r3], m1
-    pshufd      m1,        m1,        2
-    movd        [r2 + r3 + 8],  m1
-%endif
-
-    movq        m1,        [r0 + r1]
-    punpcklbw   m7,        m0,        m1
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m6
-    pmaddwd     m8,        m6
-    packssdw    m2,        m8
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m5
-    pmaddwd     m8,        m5
-    packssdw    m7,        m8
-
-    paddw       m2,        m7
-    lea         r2,        [r2 + 2 * r3]
-
-%ifidn %1,pp
-    paddw       m2,        m4
-    psraw       m2,        6
-    packuswb    m2,        m2
-    movd        [r2],      m2
-    pextrw      r6d,       m2,    2
-    mov         [r2 + 4],  r6w
-%elifidn %1,ps
-    psubw       m2,        m4
-    movh        [r2],      m2
-    pshufd      m2,        m2,        2
-    movd        [r2 + 8],  m2
-%endif
-
-    movq        m2,        [r0 + 2 * r1]
-    punpcklbw   m1,        m2
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m6
-    pmaddwd     m8,        m6
-    packssdw    m3,        m8
-
-    movhlps     m8,        m1
-    punpcklbw   m1,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m1,        m5
-    pmaddwd     m8,        m5
-    packssdw    m1,        m8
-
-    paddw       m3,        m1
-
-%ifidn %1,pp
-    paddw       m3,        m4
-    psraw       m3,        6
-    packuswb    m3,        m3
-
-    movd        [r2 + r3], m3
-    pextrw      r6d,       m3,    2
-    mov         [r2 + r3 + 4], r6w
-%elifidn %1,ps
-    psubw       m3,        m4
-    movh        [r2 + r3], m3
-    pshufd      m3,        m3,        2
-    movd        [r2 + r3 + 8],  m3
-%endif
-
-%if x < %2/4
-    lea         r2,        [r2 + 2 * r3]
-%endif
-
-%assign x x+1
-%endrep
-    RET
-
-%endmacro
-
-%if ARCH_X86_64
-    FILTER_V4_W6_H4_sse2 pp, 8
-    FILTER_V4_W6_H4_sse2 pp, 16
-    FILTER_V4_W6_H4_sse2 ps, 8
-    FILTER_V4_W6_H4_sse2 ps, 16
-%endif
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W8_sse2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_8x%2, 4, 7, 12
-    mov         r4d,       r4m
-    sub         r0,        r1
-    shl         r4d,       5
-    pxor        m9,        m9
-
-%ifidn %1,pp
-    mova        m4,        [pw_32]
-%elifidn %1,ps
-    mova        m4,        [pw_2000]
-    add         r3d,       r3d
-%endif
-
-%ifdef PIC
-    lea         r6,        [tab_ChromaCoeffV]
-    mova        m6,        [r6 + r4]
-    mova        m5,        [r6 + r4 + 16]
-%else
-    mova        m6,        [tab_ChromaCoeffV + r4]
-    mova        m5,        [tab_ChromaCoeffV + r4 + 16]
-%endif
-
-    movq        m0,        [r0]
-    movq        m1,        [r0 + r1]
-    movq        m2,        [r0 + 2 * r1]
-    lea         r5,        [r0 + 2 * r1]
-    movq        m3,        [r5 + r1]
-
-    punpcklbw   m0,        m1
-    punpcklbw   m7,        m2,          m3
-
-    movhlps     m8,        m0
-    punpcklbw   m0,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m0,        m6
-    pmaddwd     m8,        m6
-    packssdw    m0,        m8
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m5
-    pmaddwd     m8,        m5
-    packssdw    m7,        m8
-
-    paddw       m0,        m7
-
-%ifidn %1,pp
-    paddw       m0,        m4
-    psraw       m0,        6
-%elifidn %1,ps
-    psubw       m0,        m4
-    movu        [r2],      m0
-%endif
-
-    movq        m11,        [r0 + 4 * r1]
-
-    punpcklbw   m1,        m2
-    punpcklbw   m7,        m3,        m11
-
-    movhlps     m8,        m1
-    punpcklbw   m1,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m1,        m6
-    pmaddwd     m8,        m6
-    packssdw    m1,        m8
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m5
-    pmaddwd     m8,        m5
-    packssdw    m7,        m8
-
-    paddw       m1,        m7
-
-%ifidn %1,pp
-    paddw       m1,        m4
-    psraw       m1,        6
-    packuswb    m1,        m0
-
-    movhps      [r2],      m1
-    movh        [r2 + r3], m1
-%elifidn %1,ps
-    psubw       m1,        m4
-    movu        [r2 + r3], m1
-%endif
-%if %2 == 2     ;end of 8x2
-    RET
-
-%else
-    lea         r6,        [r0 + 4 * r1]
-    movq        m1,        [r6 + r1]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m7,        m11,        m1
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m6
-    pmaddwd     m8,        m6
-    packssdw    m2,        m8
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m5
-    pmaddwd     m8,        m5
-    packssdw    m7,        m8
-
-    paddw       m2,        m7
-
-%ifidn %1,pp
-    paddw       m2,        m4
-    psraw       m2,        6
-%elifidn %1,ps
-    psubw       m2,        m4
-    movu        [r2 + 2 * r3], m2
-%endif
-
-    movq        m10,        [r6 + 2 * r1]
-
-    punpcklbw   m3,        m11
-    punpcklbw   m7,        m1,        m10
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m6
-    pmaddwd     m8,        m6
-    packssdw    m3,        m8
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m5
-    pmaddwd     m8,        m5
-    packssdw    m7,        m8
-
-    paddw       m3,        m7
-    lea         r5,        [r2 + 2 * r3]
-
-%ifidn %1,pp
-    paddw       m3,        m4
-    psraw       m3,        6
-    packuswb    m3,        m2
-
-    movhps      [r2 + 2 * r3], m3
-    movh        [r5 + r3], m3
-%elifidn %1,ps
-    psubw       m3,        m4
-    movu        [r5 + r3], m3
-%endif
-%if %2 == 4     ;end of 8x4
-    RET
-
-%else
-    lea         r6,        [r6 + 2 * r1]
-    movq        m3,        [r6 + r1]
-
-    punpcklbw   m11,        m1
-    punpcklbw   m7,        m10,        m3
-
-    movhlps     m8,        m11
-    punpcklbw   m11,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m11,        m6
-    pmaddwd     m8,        m6
-    packssdw    m11,        m8
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m5
-    pmaddwd     m8,        m5
-    packssdw    m7,        m8
-
-    paddw       m11,       m7
-
-%ifidn %1, pp
-    paddw       m11,       m4
-    psraw       m11,       6
-%elifidn %1,ps
-    psubw       m11,       m4
-    movu        [r2 + 4 * r3], m11
-%endif
-
-    movq        m7,        [r0 + 8 * r1]
-
-    punpcklbw   m1,        m10
-    punpcklbw   m3,        m7
-
-    movhlps     m8,        m1
-    punpcklbw   m1,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m1,        m6
-    pmaddwd     m8,        m6
-    packssdw    m1,        m8
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m5
-    pmaddwd     m8,        m5
-    packssdw    m3,        m8
-
-    paddw       m1,        m3
-    lea         r5,        [r2 + 4 * r3]
-
-%ifidn %1,pp
-    paddw       m1,        m4
-    psraw       m1,        6
-    packuswb    m1,        m11
-
-    movhps      [r2 + 4 * r3], m1
-    movh        [r5 + r3], m1
-%elifidn %1,ps
-    psubw       m1,        m4
-    movu        [r5 + r3], m1
-%endif
-%if %2 == 6
-    RET
-
-%else
-  %error INVALID macro argument, only 2, 4 or 6!
-%endif
-%endif
-%endif
-%endmacro
-
-%if ARCH_X86_64
-    FILTER_V4_W8_sse2 pp, 2
-    FILTER_V4_W8_sse2 pp, 4
-    FILTER_V4_W8_sse2 pp, 6
-    FILTER_V4_W8_sse2 ps, 2
-    FILTER_V4_W8_sse2 ps, 4
-    FILTER_V4_W8_sse2 ps, 6
-%endif
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W8_H8_H16_H32_sse2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_8x%2, 4, 6, 11
-    mov         r4d,       r4m
-    sub         r0,        r1
-    shl         r4d,       5
-    pxor        m9,        m9
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeffV]
-    mova        m6,        [r5 + r4]
-    mova        m5,        [r5 + r4 + 16]
-%else
-    mova        m6,        [tab_ChromaCoeff + r4]
-    mova        m5,        [tab_ChromaCoeff + r4 + 16]
-%endif
-
-%ifidn %1,pp
-    mova        m4,        [pw_32]
-%elifidn %1,ps
-    mova        m4,        [pw_2000]
-    add         r3d,       r3d
-%endif
-
-    lea         r5,        [r1 * 3]
-
-%assign x 1
-%rep %2/4
-    movq        m0,        [r0]
-    movq        m1,        [r0 + r1]
-    movq        m2,        [r0 + 2 * r1]
-    movq        m3,        [r0 + r5]
-
-    punpcklbw   m0,        m1
-    punpcklbw   m1,        m2
-    punpcklbw   m2,        m3
-
-    movhlps     m7,        m0
-    punpcklbw   m0,        m9
-    punpcklbw   m7,        m9
-    pmaddwd     m0,        m6
-    pmaddwd     m7,        m6
-    packssdw    m0,        m7
-
-    movhlps     m8,        m2
-    movq        m7,        m2
-    punpcklbw   m8,        m9
-    punpcklbw   m7,        m9
-    pmaddwd     m8,        m5
-    pmaddwd     m7,        m5
-    packssdw    m7,        m8
-
-    paddw       m0,        m7
-
-%ifidn %1,pp
-    paddw       m0,        m4
-    psraw       m0,        6
-%elifidn %1,ps
-    psubw       m0,        m4
-    movu        [r2],      m0
-%endif
-
-    lea         r0,        [r0 + 4 * r1]
-    movq        m10,       [r0]
-    punpcklbw   m3,        m10
-
-    movhlps     m8,        m1
-    punpcklbw   m1,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m1,        m6
-    pmaddwd     m8,        m6
-    packssdw    m1,        m8
-
-    movhlps     m8,        m3
-    movq        m7,        m3
-    punpcklbw   m8,        m9
-    punpcklbw   m7,        m9
-    pmaddwd     m8,        m5
-    pmaddwd     m7,        m5
-    packssdw    m7,        m8
-
-    paddw       m1,        m7
-
-%ifidn %1,pp
-    paddw       m1,        m4
-    psraw       m1,        6
-
-    packuswb    m0,        m1
-    movh        [r2],      m0
-    movhps      [r2 + r3], m0
-%elifidn %1,ps
-    psubw       m1,        m4
-    movu        [r2 + r3], m1
-%endif
-
-    movq        m1,        [r0 + r1]
-    punpcklbw   m10,       m1
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m6
-    pmaddwd     m8,        m6
-    packssdw    m2,        m8
-
-    movhlps     m8,        m10
-    punpcklbw   m10,       m9
-    punpcklbw   m8,        m9
-    pmaddwd     m10,       m5
-    pmaddwd     m8,        m5
-    packssdw    m10,       m8
-
-    paddw       m2,        m10
-    lea         r2,        [r2 + 2 * r3]
-
-%ifidn %1,pp
-    paddw       m2,        m4
-    psraw       m2,        6
-%elifidn %1,ps
-    psubw       m2,        m4
-    movu        [r2],      m2
-%endif
-
-    movq        m7,        [r0 + 2 * r1]
-    punpcklbw   m1,        m7
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m6
-    pmaddwd     m8,        m6
-    packssdw    m3,        m8
-
-    movhlps     m8,        m1
-    punpcklbw   m1,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m1,        m5
-    pmaddwd     m8,        m5
-    packssdw    m1,        m8
-
-    paddw       m3,        m1
-
-%ifidn %1,pp
-    paddw       m3,        m4
-    psraw       m3,        6
-
-    packuswb    m2,        m3
-    movh        [r2],      m2
-    movhps      [r2 + r3], m2
-%elifidn %1,ps
-    psubw       m3,        m4
-    movu        [r2 + r3], m3
-%endif
-
-%if x < %2/4
-    lea         r2,        [r2 + 2 * r3]
-%endif
-%endrep
-    RET
-%endmacro
-
-%if ARCH_X86_64
-    FILTER_V4_W8_H8_H16_H32_sse2 pp,  8
-    FILTER_V4_W8_H8_H16_H32_sse2 pp, 16
-    FILTER_V4_W8_H8_H16_H32_sse2 pp, 32
-
-    FILTER_V4_W8_H8_H16_H32_sse2 pp, 12
-    FILTER_V4_W8_H8_H16_H32_sse2 pp, 64
-
-    FILTER_V4_W8_H8_H16_H32_sse2 ps,  8
-    FILTER_V4_W8_H8_H16_H32_sse2 ps, 16
-    FILTER_V4_W8_H8_H16_H32_sse2 ps, 32
-
-    FILTER_V4_W8_H8_H16_H32_sse2 ps, 12
-    FILTER_V4_W8_H8_H16_H32_sse2 ps, 64
-%endif
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_12x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W12_H2_sse2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_12x%2, 4, 6, 11
-    mov         r4d,       r4m
-    sub         r0,        r1
-    shl         r4d,       5
-    pxor        m9,        m9
-
-%ifidn %1,pp
-    mova        m6,        [pw_32]
-%elifidn %1,ps
-    mova        m6,        [pw_2000]
-    add         r3d,       r3d
-%endif
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeffV]
-    mova        m1,        [r5 + r4]
-    mova        m0,        [r5 + r4 + 16]
-%else
-    mova        m1,        [tab_ChromaCoeffV + r4]
-    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
-%endif
-
-%assign x 1
-%rep %2/2
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m1
-    pmaddwd     m8,        m1
-    packssdw    m2,        m8
-
-    lea         r0,        [r0 + 2 * r1]
-    movu        m5,        [r0]
-    movu        m7,        [r0 + r1]
-
-    punpcklbw   m10,       m5,        m7
-    movhlps     m8,        m10
-    punpcklbw   m10,       m9
-    punpcklbw   m8,        m9
-    pmaddwd     m10,       m0
-    pmaddwd     m8,        m0
-    packssdw    m10,       m8
-
-    paddw       m4,        m10
-
-    punpckhbw   m10,       m5,        m7
-    movhlps     m8,        m10
-    punpcklbw   m10,       m9
-    punpcklbw   m8,        m9
-    pmaddwd     m10,       m0
-    pmaddwd     m8,        m0
-    packssdw    m10,       m8
-
-    paddw       m2,        m10
-
-%ifidn %1,pp
-    paddw       m4,        m6
-    psraw       m4,        6
-    paddw       m2,        m6
-    psraw       m2,        6
-
-    packuswb    m4,        m2
-    movh        [r2],      m4
-    psrldq      m4,        8
-    movd        [r2 + 8],  m4
-%elifidn %1,ps
-    psubw       m4,        m6
-    psubw       m2,        m6
-    movu        [r2],      m4
-    movh        [r2 + 16], m2
-%endif
-
-    punpcklbw   m4,        m3,        m5
-    punpckhbw   m3,        m5
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m4
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m1
-    pmaddwd     m8,        m1
-    packssdw    m3,        m8
-
-    movu        m5,        [r0 + 2 * r1]
-    punpcklbw   m2,        m7,        m5
-    punpckhbw   m7,        m5
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m0
-    pmaddwd     m8,        m0
-    packssdw    m2,        m8
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m0
-    pmaddwd     m8,        m0
-    packssdw    m7,        m8
-
-    paddw       m4,        m2
-    paddw       m3,        m7
-
-%ifidn %1,pp
-    paddw       m4,        m6
-    psraw       m4,        6
-    paddw       m3,        m6
-    psraw       m3,        6
-
-    packuswb    m4,        m3
-    movh        [r2 + r3], m4
-    psrldq      m4,        8
-    movd        [r2 + r3 + 8], m4
-%elifidn %1,ps
-    psubw       m4,        m6
-    psubw       m3,        m6
-    movu        [r2 + r3], m4
-    movh        [r2 + r3 + 16], m3
-%endif
-
-%if x < %2/2
-    lea         r2,        [r2 + 2 * r3]
-%endif
-%assign x x+1
-%endrep
-    RET
-
-%endmacro
-
-%if ARCH_X86_64
-    FILTER_V4_W12_H2_sse2 pp, 16
-    FILTER_V4_W12_H2_sse2 pp, 32
-    FILTER_V4_W12_H2_sse2 ps, 16
-    FILTER_V4_W12_H2_sse2 ps, 32
-%endif
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W16_H2_sse2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_16x%2, 4, 6, 11
-    mov         r4d,       r4m
-    sub         r0,        r1
-    shl         r4d,       5
-    pxor        m9,        m9
-
-%ifidn %1,pp
-    mova        m6,        [pw_32]
-%elifidn %1,ps
-    mova        m6,        [pw_2000]
-    add         r3d,       r3d
-%endif
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeffV]
-    mova        m1,        [r5 + r4]
-    mova        m0,        [r5 + r4 + 16]
-%else
-    mova        m1,        [tab_ChromaCoeffV + r4]
-    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
-%endif
-
-%assign x 1
-%rep %2/2
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m1
-    pmaddwd     m8,        m1
-    packssdw    m2,        m8
-
-    lea         r0,        [r0 + 2 * r1]
-    movu        m5,        [r0]
-    movu        m10,       [r0 + r1]
-
-    punpckhbw   m7,        m5,        m10
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m0
-    pmaddwd     m8,        m0
-    packssdw    m7,        m8
-    paddw       m2,        m7
-
-    punpcklbw   m7,        m5,        m10
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m0
-    pmaddwd     m8,        m0
-    packssdw    m7,        m8
-    paddw       m4,        m7
-
-%ifidn %1,pp
-    paddw       m4,        m6
-    psraw       m4,        6
-    paddw       m2,        m6
-    psraw       m2,        6
-
-    packuswb    m4,        m2
-    movu        [r2],      m4
-%elifidn %1,ps
-    psubw       m4,        m6
-    psubw       m2,        m6
-    movu        [r2],      m4
-    movu        [r2 + 16], m2
-%endif
-
-    punpcklbw   m4,        m3,        m5
-    punpckhbw   m3,        m5
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m1
-    pmaddwd     m8,        m1
-    packssdw    m3,        m8
-
-    movu        m5,        [r0 + 2 * r1]
-
-    punpcklbw   m2,        m10,       m5
-    punpckhbw   m10,       m5
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m0
-    pmaddwd     m8,        m0
-    packssdw    m2,        m8
-
-    movhlps     m8,        m10
-    punpcklbw   m10,       m9
-    punpcklbw   m8,        m9
-    pmaddwd     m10,       m0
-    pmaddwd     m8,        m0
-    packssdw    m10,       m8
-
-    paddw       m4,        m2
-    paddw       m3,        m10
-
-%ifidn %1,pp
-    paddw       m4,        m6
-    psraw       m4,        6
-    paddw       m3,        m6
-    psraw       m3,        6
-
-    packuswb    m4,        m3
-    movu        [r2 + r3], m4
-%elifidn %1,ps
-    psubw       m4,        m6
-    psubw       m3,        m6
-    movu        [r2 + r3], m4
-    movu        [r2 + r3 + 16], m3
-%endif
-
-%if x < %2/2
-    lea         r2,        [r2 + 2 * r3]
-%endif
-%assign x x+1
-%endrep
-    RET
-
-%endmacro
-
-%if ARCH_X86_64
-    FILTER_V4_W16_H2_sse2 pp, 4
-    FILTER_V4_W16_H2_sse2 pp, 8
-    FILTER_V4_W16_H2_sse2 pp, 12
-    FILTER_V4_W16_H2_sse2 pp, 16
-    FILTER_V4_W16_H2_sse2 pp, 32
-
-    FILTER_V4_W16_H2_sse2 pp, 24
-    FILTER_V4_W16_H2_sse2 pp, 64
-
-    FILTER_V4_W16_H2_sse2 ps, 4
-    FILTER_V4_W16_H2_sse2 ps, 8
-    FILTER_V4_W16_H2_sse2 ps, 12
-    FILTER_V4_W16_H2_sse2 ps, 16
-    FILTER_V4_W16_H2_sse2 ps, 32
-
-    FILTER_V4_W16_H2_sse2 ps, 24
-    FILTER_V4_W16_H2_sse2 ps, 64
-%endif
-
-;-----------------------------------------------------------------------------
-;void interp_4tap_vert_%1_24%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W24_sse2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_24x%2, 4, 6, 11
-    mov         r4d,       r4m
-    sub         r0,        r1
-    shl         r4d,       5
-    pxor        m9,        m9
-
-%ifidn %1,pp
-    mova        m6,        [pw_32]
-%elifidn %1,ps
-    mova        m6,        [pw_2000]
-    add         r3d,       r3d
-%endif
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeffV]
-    mova        m1,        [r5 + r4]
-    mova        m0,        [r5 + r4 + 16]
-%else
-    mova        m1,        [tab_ChromaCoeffV + r4]
-    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
-%endif
-
-%assign x 1
-%rep %2/2
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m1
-    pmaddwd     m8,        m1
-    packssdw    m2,        m8
-
-    lea         r5,        [r0 + 2 * r1]
-    movu        m5,        [r5]
-    movu        m10,       [r5 + r1]
-    punpcklbw   m7,        m5,        m10
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m0
-    pmaddwd     m8,        m0
-    packssdw    m7,        m8
-    paddw       m4,        m7
-
-    punpckhbw   m7,        m5,        m10
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m0
-    pmaddwd     m8,        m0
-    packssdw    m7,        m8
-
-    paddw       m2,        m7
-
-%ifidn %1,pp
-    paddw       m4,        m6
-    psraw       m4,        6
-    paddw       m2,        m6
-    psraw       m2,        6
-
-    packuswb    m4,        m2
-    movu        [r2],      m4
-%elifidn %1,ps
-    psubw       m4,        m6
-    psubw       m2,        m6
-    movu        [r2],      m4
-    movu        [r2 + 16], m2
-%endif
-
-    punpcklbw   m4,        m3,        m5
-    punpckhbw   m3,        m5
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m1
-    pmaddwd     m8,        m1
-    packssdw    m3,        m8
-
-    movu        m2,        [r5 + 2 * r1]
-
-    punpcklbw   m5,        m10,        m2
-    punpckhbw   m10,       m2
-
-    movhlps     m8,        m5
-    punpcklbw   m5,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m5,        m0
-    pmaddwd     m8,        m0
-    packssdw    m5,        m8
-
-    movhlps     m8,        m10
-    punpcklbw   m10,       m9
-    punpcklbw   m8,        m9
-    pmaddwd     m10,       m0
-    pmaddwd     m8,        m0
-    packssdw    m10,       m8
-
-    paddw       m4,        m5
-    paddw       m3,        m10
-
-%ifidn %1,pp
-    paddw       m4,        m6
-    psraw       m4,        6
-    paddw       m3,        m6
-    psraw       m3,        6
-
-    packuswb    m4,        m3
-    movu        [r2 + r3], m4
-%elifidn %1,ps
-    psubw       m4,        m6
-    psubw       m3,        m6
-    movu        [r2 + r3], m4
-    movu        [r2 + r3 + 16], m3
-%endif
-
-    movq        m2,        [r0 + 16]
-    movq        m3,        [r0 + r1 + 16]
-    movq        m4,        [r5 + 16]
-    movq        m5,        [r5 + r1 + 16]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m4,        m5
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m0
-    pmaddwd     m8,        m0
-    packssdw    m4,        m8
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m1
-    pmaddwd     m8,        m1
-    packssdw    m2,        m8
-
-    paddw       m2,        m4
-
-%ifidn %1,pp
-    paddw       m2,        m6
-    psraw       m2,        6
-%elifidn %1,ps
-    psubw       m2,        m6
-    movu        [r2 + 32], m2
-%endif
-
-    movq        m3,        [r0 + r1 + 16]
-    movq        m4,        [r5 + 16]
-    movq        m5,        [r5 + r1 + 16]
-    movq        m7,        [r5 + 2 * r1 + 16]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m5,        m7
-
-    movhlps     m8,        m5
-    punpcklbw   m5,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m5,        m0
-    pmaddwd     m8,        m0
-    packssdw    m5,        m8
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m1
-    pmaddwd     m8,        m1
-    packssdw    m3,        m8
-
-    paddw       m3,        m5
-
-%ifidn %1,pp
-    paddw       m3,        m6
-    psraw       m3,        6
-
-    packuswb    m2,        m3
-    movh        [r2 + 16], m2
-    movhps      [r2 + r3 + 16], m2
-%elifidn %1,ps
-    psubw       m3,        m6
-    movu        [r2 + r3 + 32], m3
-%endif
-
-%if x < %2/2
-    mov         r0,        r5
-    lea         r2,        [r2 + 2 * r3]
-%endif
-%assign x x+1
-%endrep
-    RET
-
-%endmacro
-
-%if ARCH_X86_64
-    FILTER_V4_W24_sse2 pp, 32
-    FILTER_V4_W24_sse2 pp, 64
-    FILTER_V4_W24_sse2 ps, 32
-    FILTER_V4_W24_sse2 ps, 64
-%endif
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W32_sse2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_32x%2, 4, 6, 10
-    mov         r4d,       r4m
-    sub         r0,        r1
-    shl         r4d,       5
-    pxor        m9,        m9
-
-%ifidn %1,pp
-    mova        m6,        [pw_32]
-%elifidn %1,ps
-    mova        m6,        [pw_2000]
-    add         r3d,       r3d
-%endif
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeffV]
-    mova        m1,        [r5 + r4]
-    mova        m0,        [r5 + r4 + 16]
-%else
-    mova        m1,        [tab_ChromaCoeffV + r4]
-    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
-%endif
-
-    mov         r4d,       %2
-
-.loop:
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m1
-    pmaddwd     m8,        m1
-    packssdw    m2,        m8
-
-    lea         r5,        [r0 + 2 * r1]
-    movu        m3,        [r5]
-    movu        m5,        [r5 + r1]
-
-    punpcklbw   m7,        m3,        m5
-    punpckhbw   m3,        m5
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m0
-    pmaddwd     m8,        m0
-    packssdw    m7,        m8
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m0
-    pmaddwd     m8,        m0
-    packssdw    m3,        m8
-
-    paddw       m4,        m7
-    paddw       m2,        m3
-
-%ifidn %1,pp
-    paddw       m4,        m6
-    psraw       m4,        6
-    paddw       m2,        m6
-    psraw       m2,        6
-
-    packuswb    m4,        m2
-    movu        [r2],      m4
-%elifidn %1,ps
-    psubw       m4,        m6
-    psubw       m2,        m6
-    movu        [r2],      m4
-    movu        [r2 + 16], m2
-%endif
-
-    movu        m2,        [r0 + 16]
-    movu        m3,        [r0 + r1 + 16]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m1
-    pmaddwd     m8,        m1
-    packssdw    m2,        m8
-
-    movu        m3,        [r5 + 16]
-    movu        m5,        [r5 + r1 + 16]
-
-    punpcklbw   m7,        m3,        m5
-    punpckhbw   m3,        m5
-
-    movhlps     m8,        m7
-    punpcklbw   m7,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m7,        m0
-    pmaddwd     m8,        m0
-    packssdw    m7,        m8
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m0
-    pmaddwd     m8,        m0
-    packssdw    m3,        m8
-
-    paddw       m4,        m7
-    paddw       m2,        m3
-
-%ifidn %1,pp
-    paddw       m4,        m6
-    psraw       m4,        6
-    paddw       m2,        m6
-    psraw       m2,        6
-
-    packuswb    m4,        m2
-    movu        [r2 + 16], m4
-%elifidn %1,ps
-    psubw       m4,        m6
-    psubw       m2,        m6
-    movu        [r2 + 32], m4
-    movu        [r2 + 48], m2
-%endif
-
-    lea         r0,        [r0 + r1]
-    lea         r2,        [r2 + r3]
-    dec         r4
-    jnz        .loop
-    RET
-
-%endmacro
-
-%if ARCH_X86_64
-    FILTER_V4_W32_sse2 pp, 8
-    FILTER_V4_W32_sse2 pp, 16
-    FILTER_V4_W32_sse2 pp, 24
-    FILTER_V4_W32_sse2 pp, 32
-
-    FILTER_V4_W32_sse2 pp, 48
-    FILTER_V4_W32_sse2 pp, 64
-
-    FILTER_V4_W32_sse2 ps, 8
-    FILTER_V4_W32_sse2 ps, 16
-    FILTER_V4_W32_sse2 ps, 24
-    FILTER_V4_W32_sse2 ps, 32
-
-    FILTER_V4_W32_sse2 ps, 48
-    FILTER_V4_W32_sse2 ps, 64
-%endif
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_%1_%2x%3(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W16n_H2_sse2 3
-INIT_XMM sse2
-cglobal interp_4tap_vert_%1_%2x%3, 4, 7, 11
-    mov         r4d,       r4m
-    sub         r0,        r1
-    shl         r4d,       5
-    pxor        m9,        m9
-
-%ifidn %1,pp
-    mova        m7,        [pw_32]
-%elifidn %1,ps
-    mova        m7,        [pw_2000]
-    add         r3d,       r3d
-%endif
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeffV]
-    mova        m1,        [r5 + r4]
-    mova        m0,        [r5 + r4 + 16]
-%else
-    mova        m1,        [tab_ChromaCoeffV + r4]
-    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
-%endif
-
-    mov         r4d,       %3/2
-
-.loop:
-
-    mov         r6d,       %2/16
-
-.loopW:
-
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m1
-    pmaddwd     m8,        m1
-    packssdw    m2,        m8
-
-    lea         r5,        [r0 + 2 * r1]
-    movu        m5,        [r5]
-    movu        m6,        [r5 + r1]
-
-    punpckhbw   m10,        m5,        m6
-    movhlps     m8,        m10
-    punpcklbw   m10,       m9
-    punpcklbw   m8,        m9
-    pmaddwd     m10,       m0
-    pmaddwd     m8,        m0
-    packssdw    m10,       m8
-    paddw       m2,        m10
-
-    punpcklbw   m10,        m5,        m6
-    movhlps     m8,        m10
-    punpcklbw   m10,       m9
-    punpcklbw   m8,        m9
-    pmaddwd     m10,       m0
-    pmaddwd     m8,        m0
-    packssdw    m10,       m8
-    paddw       m4,        m10
-
-%ifidn %1,pp
-    paddw       m4,        m7
-    psraw       m4,        6
-    paddw       m2,        m7
-    psraw       m2,        6
-
-    packuswb    m4,        m2
-    movu        [r2],      m4
-%elifidn %1,ps
-    psubw       m4,        m7
-    psubw       m2,        m7
-    movu        [r2],      m4
-    movu        [r2 + 16], m2
-%endif
-
-    punpcklbw   m4,        m3,        m5
-    punpckhbw   m3,        m5
-
-    movhlps     m8,        m4
-    punpcklbw   m4,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m4,        m1
-    pmaddwd     m8,        m1
-    packssdw    m4,        m8
-
-    movhlps     m8,        m3
-    punpcklbw   m3,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m3,        m1
-    pmaddwd     m8,        m1
-    packssdw    m3,        m8
-
-    movu        m5,        [r5 + 2 * r1]
-
-    punpcklbw   m2,        m6,        m5
-    punpckhbw   m6,        m5
-
-    movhlps     m8,        m2
-    punpcklbw   m2,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m2,        m0
-    pmaddwd     m8,        m0
-    packssdw    m2,        m8
-
-    movhlps     m8,        m6
-    punpcklbw   m6,        m9
-    punpcklbw   m8,        m9
-    pmaddwd     m6,        m0
-    pmaddwd     m8,        m0
-    packssdw    m6,        m8
-
-    paddw       m4,        m2
-    paddw       m3,        m6
-
-%ifidn %1,pp
-    paddw       m4,        m7
-    psraw       m4,        6
-    paddw       m3,        m7
-    psraw       m3,        6
-
-    packuswb    m4,        m3
-    movu        [r2 + r3], m4
-    add         r2,        16
-%elifidn %1,ps
-    psubw       m4,        m7
-    psubw       m3,        m7
-    movu        [r2 + r3], m4
-    movu        [r2 + r3 + 16], m3
-    add         r2,        32
-%endif
-
-    add         r0,        16
-    dec         r6d
-    jnz         .loopW
-
-    lea         r0,        [r0 + r1 * 2 - %2]
-
-%ifidn %1,pp
-    lea         r2,        [r2 + r3 * 2 - %2]
-%elifidn %1,ps
-    lea         r2,        [r2 + r3 * 2 - (%2 * 2)]
-%endif
-
-    dec         r4d
-    jnz        .loop
-    RET
-
-%endmacro
-
-%if ARCH_X86_64
-    FILTER_V4_W16n_H2_sse2 pp, 64, 64
-    FILTER_V4_W16n_H2_sse2 pp, 64, 32
-    FILTER_V4_W16n_H2_sse2 pp, 64, 48
-    FILTER_V4_W16n_H2_sse2 pp, 48, 64
-    FILTER_V4_W16n_H2_sse2 pp, 64, 16
-    FILTER_V4_W16n_H2_sse2 ps, 64, 64
-    FILTER_V4_W16n_H2_sse2 ps, 64, 32
-    FILTER_V4_W16n_H2_sse2 ps, 64, 48
-    FILTER_V4_W16n_H2_sse2 ps, 48, 64
-    FILTER_V4_W16n_H2_sse2 ps, 64, 16
-%endif
-
 %macro FILTER_P2S_2_4_sse2 1
     movd        m2,     [r0 + %1]
     movd        m3,     [r0 + r1 + %1]
@@ -3461,577 +778,6 @@
     FILTER_PIX_TO_SHORT_sse2 64, 48
     FILTER_PIX_TO_SHORT_sse2 64, 64
 
-%macro FILTER_H4_w2_2 3
-    movh        %2, [srcq - 1]
-    pshufb      %2, %2, Tm0
-    movh        %1, [srcq + srcstrideq - 1]
-    pshufb      %1, %1, Tm0
-    punpcklqdq  %2, %1
-    pmaddubsw   %2, coef2
-    phaddw      %2, %2
-    pmulhrsw    %2, %3
-    packuswb    %2, %2
-    movd        r4, %2
-    mov         [dstq], r4w
-    shr         r4, 16
-    mov         [dstq + dststrideq], r4w
-%endmacro
-
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_2x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_2x4, 4, 6, 5, src, srcstride, dst, dststride
-%define coef2       m4
-%define Tm0         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-
-%rep 2
-    FILTER_H4_w2_2   t0, t1, t2
-    lea         srcq,       [srcq + srcstrideq * 2]
-    lea         dstq,       [dstq + dststrideq * 2]
-%endrep
-
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_2x8, 4, 6, 5, src, srcstride, dst, dststride
-%define coef2       m4
-%define Tm0         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-
-%rep 4
-    FILTER_H4_w2_2   t0, t1, t2
-    lea         srcq,       [srcq + srcstrideq * 2]
-    lea         dstq,       [dstq + dststrideq * 2]
-%endrep
-
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_2x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_2x16, 4, 6, 5, src, srcstride, dst, dststride
-%define coef2       m4
-%define Tm0         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-
-    mov         r5d,        16/2
-
-.loop:
-    FILTER_H4_w2_2   t0, t1, t2
-    lea         srcq,       [srcq + srcstrideq * 2]
-    lea         dstq,       [dstq + dststrideq * 2]
-    dec         r5d
-    jnz         .loop
-
-    RET
-
-%macro FILTER_H4_w4_2 3
-    movh        %2, [srcq - 1]
-    pshufb      %2, %2, Tm0
-    pmaddubsw   %2, coef2
-    movh        %1, [srcq + srcstrideq - 1]
-    pshufb      %1, %1, Tm0
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    pmulhrsw    %2, %3
-    packuswb    %2, %2
-    movd        [dstq], %2
-    palignr     %2, %2, 4
-    movd        [dstq + dststrideq], %2
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_4x2, 4, 6, 5, src, srcstride, dst, dststride
-%define coef2       m4
-%define Tm0         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-
-    FILTER_H4_w4_2   t0, t1, t2
-
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_4x4, 4, 6, 5, src, srcstride, dst, dststride
-%define coef2       m4
-%define Tm0         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-
-%rep 2
-    FILTER_H4_w4_2   t0, t1, t2
-    lea         srcq,       [srcq + srcstrideq * 2]
-    lea         dstq,       [dstq + dststrideq * 2]
-%endrep
-
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_4x8, 4, 6, 5, src, srcstride, dst, dststride
-%define coef2       m4
-%define Tm0         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-
-%rep 4
-    FILTER_H4_w4_2   t0, t1, t2
-    lea         srcq,       [srcq + srcstrideq * 2]
-    lea         dstq,       [dstq + dststrideq * 2]
-%endrep
-
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_4x16, 4, 6, 5, src, srcstride, dst, dststride
-%define coef2       m4
-%define Tm0         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-
-%rep 8
-    FILTER_H4_w4_2   t0, t1, t2
-    lea         srcq,       [srcq + srcstrideq * 2]
-    lea         dstq,       [dstq + dststrideq * 2]
-%endrep
-
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_4x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_4x32, 4, 6, 5, src, srcstride, dst, dststride
-%define coef2       m4
-%define Tm0         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-
-    mov         r5d,        32/2
-
-.loop:
-    FILTER_H4_w4_2   t0, t1, t2
-    lea         srcq,       [srcq + srcstrideq * 2]
-    lea         dstq,       [dstq + dststrideq * 2]
-    dec         r5d
-    jnz         .loop
-
-    RET
-
-ALIGN 32
-const interp_4tap_8x8_horiz_shuf,   dd 0, 4, 1, 5, 2, 6, 3, 7
-
-
-%macro FILTER_H4_w6 3
-    movu        %1, [srcq - 1]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    pmulhrsw    %2, %3
-    packuswb    %2, %2
-    movd        [dstq],      %2
-    pextrw      [dstq + 4], %2, 2
-%endmacro
-
-%macro FILTER_H4_w8 3
-    movu        %1, [srcq - 1]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    pmulhrsw    %2, %3
-    packuswb    %2, %2
-    movh        [dstq],      %2
-%endmacro
-
-%macro FILTER_H4_w12 3
-    movu        %1, [srcq - 1]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    pmulhrsw    %2, %3
-    movu        %1, [srcq - 1 + 8]
-    pshufb      %1, %1, Tm0
-    pmaddubsw   %1, coef2
-    phaddw      %1, %1
-    pmulhrsw    %1, %3
-    packuswb    %2, %1
-    movh        [dstq],      %2
-    pextrd      [dstq + 8], %2, 2
-%endmacro
-
-%macro FILTER_H4_w16 4
-    movu        %1, [srcq - 1]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq - 1 + 8]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    pmulhrsw    %2, %3
-    pmulhrsw    %4, %3
-    packuswb    %2, %4
-    movu        [dstq],      %2
-%endmacro
-
-%macro FILTER_H4_w24 4
-    movu        %1, [srcq - 1]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq - 1 + 8]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    pmulhrsw    %2, %3
-    pmulhrsw    %4, %3
-    packuswb    %2, %4
-    movu        [dstq],          %2
-    movu        %1, [srcq - 1 + 16]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    pmulhrsw    %2, %3
-    packuswb    %2, %2
-    movh        [dstq + 16],     %2
-%endmacro
-
-%macro FILTER_H4_w32 4
-    movu        %1, [srcq - 1]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq - 1 + 8]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    pmulhrsw    %2, %3
-    pmulhrsw    %4, %3
-    packuswb    %2, %4
-    movu        [dstq],      %2
-    movu        %1, [srcq - 1 + 16]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq - 1 + 24]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    pmulhrsw    %2, %3
-    pmulhrsw    %4, %3
-    packuswb    %2, %4
-    movu        [dstq + 16],      %2
-%endmacro
-
-%macro FILTER_H4_w16o 5
-    movu        %1, [srcq + %5 - 1]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq + %5 - 1 + 8]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    pmulhrsw    %2, %3
-    pmulhrsw    %4, %3
-    packuswb    %2, %4
-    movu        [dstq + %5],      %2
-%endmacro
-
-%macro FILTER_H4_w48 4
-    FILTER_H4_w16o %1, %2, %3, %4, 0
-    FILTER_H4_w16o %1, %2, %3, %4, 16
-    FILTER_H4_w16o %1, %2, %3, %4, 32
-%endmacro
-
-%macro FILTER_H4_w64 4
-    FILTER_H4_w16o %1, %2, %3, %4, 0
-    FILTER_H4_w16o %1, %2, %3, %4, 16
-    FILTER_H4_w16o %1, %2, %3, %4, 32
-    FILTER_H4_w16o %1, %2, %3, %4, 48
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro IPFILTER_CHROMA 2
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 6, src, srcstride, dst, dststride
-%define coef2       m5
-%define Tm0         m4
-%define Tm1         m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,        r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mov           r5d,       %2
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-    mova        Tm1,         [tab_Tm + 16]
-
-.loop:
-    FILTER_H4_w%1   t0, t1, t2
-    add         srcq,        srcstrideq
-    add         dstq,        dststrideq
-
-    dec         r5d
-    jnz        .loop
-
-    RET
-%endmacro
-
-
-    IPFILTER_CHROMA 6,   8
-    IPFILTER_CHROMA 8,   2
-    IPFILTER_CHROMA 8,   4
-    IPFILTER_CHROMA 8,   6
-    IPFILTER_CHROMA 8,   8
-    IPFILTER_CHROMA 8,  16
-    IPFILTER_CHROMA 8,  32
-    IPFILTER_CHROMA 12, 16
-
-    IPFILTER_CHROMA 6,  16
-    IPFILTER_CHROMA 8,  12
-    IPFILTER_CHROMA 8,  64
-    IPFILTER_CHROMA 12, 32
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_W 2
-INIT_XMM sse4
-cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 7, src, srcstride, dst, dststride
-%define coef2       m6
-%define Tm0         m5
-%define Tm1         m4
-%define t3          m3
-%define t2          m2
-%define t1          m1
-%define t0          m0
-
-    mov         r4d,         r4m
-
-%ifdef PIC
-    lea         r5,          [tab_ChromaCoeff]
-    movd        coef2,       [r5 + r4 * 4]
-%else
-    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mov         r5d,          %2
-
-    pshufd      coef2,       coef2,      0
-    mova        t2,          [pw_512]
-    mova        Tm0,         [tab_Tm]
-    mova        Tm1,         [tab_Tm + 16]
-
-.loop:
-    FILTER_H4_w%1   t0, t1, t2, t3
-    add         srcq,        srcstrideq
-    add         dstq,        dststrideq
-
-    dec         r5d
-    jnz        .loop
-
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_W 16,  4
-    IPFILTER_CHROMA_W 16,  8
-    IPFILTER_CHROMA_W 16, 12
-    IPFILTER_CHROMA_W 16, 16
-    IPFILTER_CHROMA_W 16, 32
-    IPFILTER_CHROMA_W 32,  8
-    IPFILTER_CHROMA_W 32, 16
-    IPFILTER_CHROMA_W 32, 24
-    IPFILTER_CHROMA_W 24, 32
-    IPFILTER_CHROMA_W 32, 32
-
-    IPFILTER_CHROMA_W 16, 24
-    IPFILTER_CHROMA_W 16, 64
-    IPFILTER_CHROMA_W 32, 48
-    IPFILTER_CHROMA_W 24, 64
-    IPFILTER_CHROMA_W 32, 64
-
-    IPFILTER_CHROMA_W 64, 64
-    IPFILTER_CHROMA_W 64, 32
-    IPFILTER_CHROMA_W 64, 48
-    IPFILTER_CHROMA_W 48, 64
-    IPFILTER_CHROMA_W 64, 16
-
-
 %macro FILTER_H8_W8 7-8   ; t0, t1, t2, t3, coef, c512, src, dst
     movu        %1, %7
     pshufb      %2, %1, [tab_Lm +  0]
@@ -4052,1728 +798,6 @@
   %endif
 %endmacro
 
-%macro FILTER_H8_W4 2
-    movu        %1, [r0 - 3 + r5]
-    pshufb      %2, %1, [tab_Lm]
-    pmaddubsw   %2, m3
-    pshufb      m7, %1, [tab_Lm + 16]
-    pmaddubsw   m7, m3
-    phaddw      %2, m7
-    phaddw      %2, %2
-%endmacro
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_LUMA 3
-INIT_XMM sse4
-cglobal interp_8tap_horiz_%3_%1x%2, 4,7,8
-
-    mov       r4d, r4m
-
-%ifdef PIC
-    lea       r6, [tab_LumaCoeff]
-    movh      m3, [r6 + r4 * 8]
-%else
-    movh      m3, [tab_LumaCoeff + r4 * 8]
-%endif
-    punpcklqdq  m3, m3
-
-%ifidn %3, pp
-    mova      m2, [pw_512]
-%else
-    mova      m2, [pw_2000]
-%endif
-
-    mov       r4d, %2
-%ifidn %3, ps
-    add       r3, r3
-    cmp       r5m, byte 0
-    je        .loopH
-    lea       r6, [r1 + 2 * r1]
-    sub       r0, r6
-    add       r4d, 7
-%endif
-
-.loopH:
-    xor       r5, r5
-%rep %1 / 8
-  %ifidn %3, pp
-    FILTER_H8_W8  m0, m1, m4, m5, m3, m2, [r0 - 3 + r5], [r2 + r5]
-  %else
-    FILTER_H8_W8  m0, m1, m4, m5, m3, UNUSED, [r0 - 3 + r5]
-    psubw     m1, m2
-    movu      [r2 + 2 * r5], m1
-  %endif
-    add       r5, 8
-%endrep
-
-%rep (%1 % 8) / 4
-    FILTER_H8_W4  m0, m1
-  %ifidn %3, pp
-    pmulhrsw  m1, m2
-    packuswb  m1, m1
-    movd      [r2 + r5], m1
-  %else
-    psubw     m1, m2
-    movh      [r2 + 2 * r5], m1
-  %endif
-%endrep
-
-    add       r0, r1
-    add       r2, r3
-
-    dec       r4d
-    jnz       .loopH
-    RET
-%endmacro
-
-
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_4x4, 4,6,6
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea             r5, [tab_LumaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_LumaCoeff + r4 * 8]
-%endif
-
-    mova            m1, [tab_Lm]
-    vpbroadcastd    m2, [pw_1]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    sub             r0, 3
-    ; Row 0-1
-    vbroadcasti128  m3, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m3, m1
-    pmaddubsw       m3, m0
-    pmaddwd         m3, m2
-    vbroadcasti128  m4, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddwd         m4, m2
-    phaddd          m3, m4                          ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
-
-    ; Row 2-3
-    lea             r0, [r0 + r1 * 2]
-    vbroadcasti128  m4, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddwd         m4, m2
-    vbroadcasti128  m5, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m5, m1
-    pmaddubsw       m5, m0
-    pmaddwd         m5, m2
-    phaddd          m4, m5                          ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
-
-    packssdw        m3, m4                          ; WORD [R3D R3C R2D R2C R1D R1C R0D R0C R3B R3A R2B R2A R1B R1A R0B R0A]
-    pmulhrsw        m3, [pw_512]
-    vextracti128    xm4, m3, 1
-    packuswb        xm3, xm4                        ; BYTE [R3D R3C R2D R2C R1D R1C R0D R0C R3B R3A R2B R2A R1B R1A R0B R0A]
-    pshufb          xm3, [interp4_shuf]             ; [row3 row1 row2 row0]
-
-    lea             r0, [r3 * 3]
-    movd            [r2], xm3
-    pextrd          [r2+r3], xm3, 2
-    pextrd          [r2+r3*2], xm3, 1
-    pextrd          [r2+r0], xm3, 3
-    RET
-
-%macro FILTER_HORIZ_LUMA_AVX2_4xN 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_pp_4x%1, 4, 6, 9
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea             r5, [tab_LumaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_LumaCoeff + r4 * 8]
-%endif
-
-    mova            m1, [tab_Lm]
-    mova            m2, [pw_1]
-    mova            m7, [interp8_hps_shuf]
-    mova            m8, [pw_512]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-    lea             r4, [r1 * 3]
-    lea             r5, [r3 * 3]
-    sub             r0, 3
-%rep %1 / 8
-    ; Row 0-1
-    vbroadcasti128  m3, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m3, m1
-    pmaddubsw       m3, m0
-    pmaddwd         m3, m2
-    vbroadcasti128  m4, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddwd         m4, m2
-    phaddd          m3, m4                          ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
-
-    ; Row 2-3
-    vbroadcasti128  m4, [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddwd         m4, m2
-    vbroadcasti128  m5, [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m5, m1
-    pmaddubsw       m5, m0
-    pmaddwd         m5, m2
-    phaddd          m4, m5                          ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
-
-    packssdw        m3, m4                          ; WORD [R3D R3C R2D R2C R1D R1C R0D R0C R3B R3A R2B R2A R1B R1A R0B R0A]
-    lea             r0, [r0 + r1 * 4]
-    ; Row 4-5
-    vbroadcasti128  m5, [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m5, m1
-    pmaddubsw       m5, m0
-    pmaddwd         m5, m2
-    vbroadcasti128  m4, [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddwd         m4, m2
-    phaddd          m5, m4                          ; DWORD [R5D R5C R4D R4C R5B R5A R4B R4A]
-
-    ; Row 6-7
-    vbroadcasti128  m4, [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddwd         m4, m2
-    vbroadcasti128  m6, [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m6, m1
-    pmaddubsw       m6, m0
-    pmaddwd         m6, m2
-    phaddd          m4, m6                          ; DWORD [R7D R7C R6D R6C R7B R7A R6B R6A]
-
-    packssdw        m5, m4                          ; WORD [R7D R7C R6D R6C R5D R5C R4D R4C R7B R7A R6B R6A R5B R5A R4B R4A]
-    vpermd          m3, m7, m3
-    vpermd          m5, m7, m5
-    pmulhrsw        m3, m8
-    pmulhrsw        m5, m8
-    packuswb        m3, m5
-    vextracti128    xm5, m3, 1
-
-    movd            [r2], xm3
-    pextrd          [r2 + r3], xm3, 1
-    movd            [r2 + r3 * 2], xm5
-    pextrd          [r2 + r5], xm5, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm3, 2
-    pextrd          [r2 + r3], xm3, 3
-    pextrd          [r2 + r3 * 2], xm5, 2
-    pextrd          [r2 + r5], xm5, 3
-    lea             r0, [r0 + r1 * 4]
-    lea             r2, [r2 + r3 * 4]
-%endrep
-    RET
-%endif
-%endmacro
-
-    FILTER_HORIZ_LUMA_AVX2_4xN 8
-    FILTER_HORIZ_LUMA_AVX2_4xN 16
-
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_8x4, 4, 6, 7
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea             r5, [tab_LumaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_LumaCoeff + r4 * 8]
-%endif
-
-    mova            m1, [tab_Lm]
-    mova            m2, [tab_Lm + 32]
-
-    ; register map
-    ; m0     - interpolate coeff
-    ; m1, m2 - shuffle order table
-
-    sub             r0, 3
-    lea             r5, [r1 * 3]
-    lea             r4, [r3 * 3]
-
-    ; Row 0
-    vbroadcasti128  m3, [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m4, m3, m2
-    pshufb          m3, m1
-    pmaddubsw       m3, m0
-    pmaddubsw       m4, m0
-    phaddw          m3, m4
-    ; Row 1
-    vbroadcasti128  m4, [r0 + r1]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m5, m4, m2
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddubsw       m5, m0
-    phaddw          m4, m5
-
-    phaddw          m3, m4                          ; WORD [R1H R1G R1D R1C R0H R0G R0D R0C R1F R1E R1B R1A R0F R0E R0B R0A]
-    pmulhrsw        m3, [pw_512]
-
-    ; Row 2
-    vbroadcasti128  m4, [r0 + r1 * 2]               ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m5, m4, m2
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddubsw       m5, m0
-    phaddw          m4, m5
-    ; Row 3
-    vbroadcasti128  m5, [r0 + r5]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m6, m5, m2
-    pshufb          m5, m1
-    pmaddubsw       m5, m0
-    pmaddubsw       m6, m0
-    phaddw          m5, m6
-
-    phaddw          m4, m5                          ; WORD [R3H R3G R3D R3C R2H R2G R2D R2C R3F R3E R3B R3A R2F R2E R2B R2A]
-    pmulhrsw        m4, [pw_512]
-
-    packuswb        m3, m4
-    vextracti128    xm4, m3, 1
-    punpcklwd       xm5, xm3, xm4
-
-    movq            [r2], xm5
-    movhps          [r2 + r3], xm5
-
-    punpckhwd       xm5, xm3, xm4
-    movq            [r2 + r3 * 2], xm5
-    movhps          [r2 + r4], xm5
-    RET
-
-%macro IPFILTER_LUMA_AVX2_8xN 2
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_%1x%2, 4, 7, 7
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea             r5, [tab_LumaCoeff]
-    vpbroadcastq    m0, [r5 + r4 * 8]
-%else
-    vpbroadcastq    m0, [tab_LumaCoeff + r4 * 8]
-%endif
-
-    mova            m1, [tab_Lm]
-    mova            m2, [tab_Lm + 32]
-
-    ; register map
-    ; m0     - interpolate coeff
-    ; m1, m2 - shuffle order table
-
-    sub             r0, 3
-    lea             r5, [r1 * 3]
-    lea             r6, [r3 * 3]
-    mov             r4d, %2 / 4
-.loop:
-    ; Row 0
-    vbroadcasti128  m3, [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m4, m3, m2
-    pshufb          m3, m1
-    pmaddubsw       m3, m0
-    pmaddubsw       m4, m0
-    phaddw          m3, m4
-    ; Row 1
-    vbroadcasti128  m4, [r0 + r1]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m5, m4, m2
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddubsw       m5, m0
-    phaddw          m4, m5
-
-    phaddw          m3, m4                          ; WORD [R1H R1G R1D R1C R0H R0G R0D R0C R1F R1E R1B R1A R0F R0E R0B R0A]
-    pmulhrsw        m3, [pw_512]
-
-    ; Row 2
-    vbroadcasti128  m4, [r0 + r1 * 2]               ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m5, m4, m2
-    pshufb          m4, m1
-    pmaddubsw       m4, m0
-    pmaddubsw       m5, m0
-    phaddw          m4, m5
-    ; Row 3
-    vbroadcasti128  m5, [r0 + r5]                   ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb          m6, m5, m2
-    pshufb          m5, m1
-    pmaddubsw       m5, m0
-    pmaddubsw       m6, m0
-    phaddw          m5, m6
-
-    phaddw          m4, m5                          ; WORD [R3H R3G R3D R3C R2H R2G R2D R2C R3F R3E R3B R3A R2F R2E R2B R2A]
-    pmulhrsw        m4, [pw_512]
-
-    packuswb        m3, m4
-    vextracti128    xm4, m3, 1
-    punpcklwd       xm5, xm3, xm4
-
-    movq            [r2], xm5
-    movhps          [r2 + r3], xm5
-
-    punpckhwd       xm5, xm3, xm4
-    movq            [r2 + r3 * 2], xm5
-    movhps          [r2 + r6], xm5
-
-    lea             r0, [r0 + r1 * 4]
-    lea             r2, [r2 + r3 * 4]
-    dec             r4d
-    jnz             .loop
-    RET
-%endmacro
-
-    IPFILTER_LUMA_AVX2_8xN 8, 8
-    IPFILTER_LUMA_AVX2_8xN 8, 16
-    IPFILTER_LUMA_AVX2_8xN 8, 32
-
-%macro IPFILTER_LUMA_AVX2 2
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
-    sub               r0,        3
-    mov               r4d,       r4m
-%ifdef PIC
-    lea               r5,        [tab_LumaCoeff]
-    vpbroadcastd      m0,        [r5 + r4 * 8]
-    vpbroadcastd      m1,        [r5 + r4 * 8 + 4]
-%else
-    vpbroadcastd      m0,         [tab_LumaCoeff + r4 * 8]
-    vpbroadcastd      m1,         [tab_LumaCoeff + r4 * 8 + 4]
-%endif
-    movu              m3,         [tab_Tm + 16]
-    vpbroadcastd      m7,         [pw_1]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-    ; m2 , m2  shuffle order table
-    ; m7 - pw_1
-
-    mov               r4d,        %2/2
-.loop:
-    ; Row 0
-    vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,         m4,     m3
-    pshufb            m4,         [tab_Tm]
-    pmaddubsw         m4,         m0
-    pmaddubsw         m5,         m1
-    paddw             m4,         m5
-    pmaddwd           m4,         m7
-    vbroadcasti128    m5,         [r0 + 8]                    ; second 8 elements in Row0
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [tab_Tm]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
-    pmulhrsw          m4,         [pw_512]
-    vbroadcasti128    m2,         [r0 + r1]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,         m2,     m3
-    pshufb            m2,         [tab_Tm]
-    pmaddubsw         m2,         m0
-    pmaddubsw         m5,         m1
-    paddw             m2,         m5
-    pmaddwd           m2,         m7
-    vbroadcasti128    m5,         [r0 + r1 + 8]                    ; second 8 elements in Row0
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [tab_Tm]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-    packssdw          m2,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
-    pmulhrsw          m2,         [pw_512]
-    packuswb          m4,         m2
-    vpermq            m4,         m4,     11011000b
-    vextracti128      xm5,        m4,     1
-    pshufd            xm4,        xm4,    11011000b
-    pshufd            xm5,        xm5,    11011000b
-    movu              [r2],       xm4
-    movu              [r2+r3],    xm5
-    lea               r0,         [r0 + r1 * 2]
-    lea               r2,         [r2 + r3 * 2]
-    dec               r4d
-    jnz              .loop
-    RET
-%endmacro
-
-%macro IPFILTER_LUMA_32x_avx2 2
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
-    sub               r0,         3
-    mov               r4d,        r4m
-%ifdef PIC
-    lea               r5,         [tab_LumaCoeff]
-    vpbroadcastd      m0,         [r5 + r4 * 8]
-    vpbroadcastd      m1,         [r5 + r4 * 8 + 4]
-%else
-    vpbroadcastd      m0,         [tab_LumaCoeff + r4 * 8]
-    vpbroadcastd      m1,         [tab_LumaCoeff + r4 * 8 + 4]
-%endif
-    movu              m3,         [tab_Tm + 16]
-    vpbroadcastd      m7,         [pw_1]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-    ; m2 , m2  shuffle order table
-    ; m7 - pw_1
-
-    mov               r4d,        %2
-.loop:
-    ; Row 0
-    vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,         m4,     m3
-    pshufb            m4,         [tab_Tm]
-    pmaddubsw         m4,         m0
-    pmaddubsw         m5,         m1
-    paddw             m4,         m5
-    pmaddwd           m4,         m7
-    vbroadcasti128    m5,         [r0 + 8]
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [tab_Tm]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
-    pmulhrsw          m4,         [pw_512]
-    vbroadcasti128    m2,         [r0 + 16]
-    pshufb            m5,         m2,     m3
-    pshufb            m2,         [tab_Tm]
-    pmaddubsw         m2,         m0
-    pmaddubsw         m5,         m1
-    paddw             m2,         m5
-    pmaddwd           m2,         m7
-    vbroadcasti128    m5,         [r0 + 24]
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [tab_Tm]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-    packssdw          m2,         m5
-    pmulhrsw          m2,         [pw_512]
-    packuswb          m4,         m2
-    vpermq            m4,         m4,     11011000b
-    vextracti128      xm5,        m4,     1
-    pshufd            xm4,        xm4,    11011000b
-    pshufd            xm5,        xm5,    11011000b
-    movu              [r2],       xm4
-    movu              [r2 + 16],  xm5
-    lea               r0,         [r0 + r1]
-    lea               r2,         [r2 + r3]
-    dec               r4d
-    jnz               .loop
-    RET
-%endmacro
-
-%macro IPFILTER_LUMA_64x_avx2 2
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_%1x%2, 4,6,8
-    sub               r0,    3
-    mov               r4d,   r4m
-%ifdef PIC
-    lea               r5,        [tab_LumaCoeff]
-    vpbroadcastd      m0,        [r5 + r4 * 8]
-    vpbroadcastd      m1,        [r5 + r4 * 8 + 4]
-%else
-    vpbroadcastd      m0,        [tab_LumaCoeff + r4 * 8]
-    vpbroadcastd      m1,        [tab_LumaCoeff + r4 * 8 + 4]
-%endif
-    movu              m3,        [tab_Tm + 16]
-    vpbroadcastd      m7,        [pw_1]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-    ; m2 , m2  shuffle order table
-    ; m7 - pw_1
-
-    mov               r4d,   %2
-.loop:
-    ; Row 0
-    vbroadcasti128    m4,        [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,        m4,    m3
-    pshufb            m4,        [tab_Tm]
-    pmaddubsw         m4,        m0
-    pmaddubsw         m5,        m1
-    paddw             m4,        m5
-    pmaddwd           m4,        m7
-    vbroadcasti128    m5,        [r0 + 8]
-    pshufb            m6,        m5,    m3
-    pshufb            m5,        [tab_Tm]
-    pmaddubsw         m5,        m0
-    pmaddubsw         m6,        m1
-    paddw             m5,        m6
-    pmaddwd           m5,        m7
-    packssdw          m4,        m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
-    pmulhrsw          m4,        [pw_512]
-    vbroadcasti128    m2,        [r0 + 16]
-    pshufb            m5,        m2,    m3
-    pshufb            m2,        [tab_Tm]
-    pmaddubsw         m2,        m0
-    pmaddubsw         m5,        m1
-    paddw             m2,        m5
-    pmaddwd           m2,        m7
-    vbroadcasti128    m5,        [r0 + 24]
-    pshufb            m6,        m5,    m3
-    pshufb            m5,        [tab_Tm]
-    pmaddubsw         m5,        m0
-    pmaddubsw         m6,        m1
-    paddw             m5,        m6
-    pmaddwd           m5,        m7
-    packssdw          m2,        m5
-    pmulhrsw          m2,        [pw_512]
-    packuswb          m4,        m2
-    vpermq            m4,        m4,    11011000b
-    vextracti128      xm5,       m4,    1
-    pshufd            xm4,       xm4,   11011000b
-    pshufd            xm5,       xm5,   11011000b
-    movu              [r2],      xm4
-    movu              [r2 + 16], xm5
-
-    vbroadcasti128    m4,        [r0 + 32]
-    pshufb            m5,        m4,    m3
-    pshufb            m4,        [tab_Tm]
-    pmaddubsw         m4,        m0
-    pmaddubsw         m5,        m1
-    paddw             m4,        m5
-    pmaddwd           m4,        m7
-    vbroadcasti128    m5,        [r0 + 40]
-    pshufb            m6,        m5,    m3
-    pshufb            m5,        [tab_Tm]
-    pmaddubsw         m5,        m0
-    pmaddubsw         m6,        m1
-    paddw             m5,        m6
-    pmaddwd           m5,        m7
-    packssdw          m4,        m5
-    pmulhrsw          m4,        [pw_512]
-    vbroadcasti128    m2,        [r0 + 48]
-    pshufb            m5,        m2,    m3
-    pshufb            m2,        [tab_Tm]
-    pmaddubsw         m2,        m0
-    pmaddubsw         m5,        m1
-    paddw             m2,        m5
-    pmaddwd           m2,        m7
-    vbroadcasti128    m5,        [r0 + 56]
-    pshufb            m6,        m5,    m3
-    pshufb            m5,        [tab_Tm]
-    pmaddubsw         m5,        m0
-    pmaddubsw         m6,        m1
-    paddw             m5,        m6
-    pmaddwd           m5,        m7
-    packssdw          m2,        m5
-    pmulhrsw          m2,        [pw_512]
-    packuswb          m4,        m2
-    vpermq            m4,        m4,    11011000b
-    vextracti128      xm5,       m4,    1
-    pshufd            xm4,       xm4,   11011000b
-    pshufd            xm5,       xm5,   11011000b
-    movu              [r2 +32],  xm4
-    movu              [r2 + 48], xm5
-
-    lea               r0,        [r0 + r1]
-    lea               r2,        [r2 + r3]
-    dec               r4d
-    jnz               .loop
-    RET
-%endmacro
-
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_48x64, 4,6,8
-    sub               r0,         3
-    mov               r4d,        r4m
-%ifdef PIC
-    lea               r5,         [tab_LumaCoeff]
-    vpbroadcastd      m0,         [r5 + r4 * 8]
-    vpbroadcastd      m1,         [r5 + r4 * 8 + 4]
-%else
-    vpbroadcastd      m0,         [tab_LumaCoeff + r4 * 8]
-    vpbroadcastd      m1,         [tab_LumaCoeff + r4 * 8 + 4]
-%endif
-    movu              m3,         [tab_Tm + 16]
-    vpbroadcastd      m7,         [pw_1]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-    ; m2 , m2  shuffle order table
-    ; m7 - pw_1
-
-    mov               r4d,        64
-.loop:
-    ; Row 0
-    vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,         m4,     m3
-    pshufb            m4,         [tab_Tm]
-    pmaddubsw         m4,         m0
-    pmaddubsw         m5,         m1
-    paddw             m4,         m5
-    pmaddwd           m4,         m7
-    vbroadcasti128    m5,         [r0 + 8]
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [tab_Tm]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
-    pmulhrsw          m4,         [pw_512]
-
-    vbroadcasti128    m2,         [r0 + 16]
-    pshufb            m5,         m2,     m3
-    pshufb            m2,         [tab_Tm]
-    pmaddubsw         m2,         m0
-    pmaddubsw         m5,         m1
-    paddw             m2,         m5
-    pmaddwd           m2,         m7
-    vbroadcasti128    m5,         [r0 + 24]
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [tab_Tm]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-    packssdw          m2,         m5
-    pmulhrsw          m2,         [pw_512]
-    packuswb          m4,         m2
-    vpermq            m4,         m4,     11011000b
-    vextracti128      xm5,        m4,     1
-    pshufd            xm4,        xm4,    11011000b
-    pshufd            xm5,        xm5,    11011000b
-    movu              [r2],       xm4
-    movu              [r2 + 16],  xm5
-
-    vbroadcasti128    m4,         [r0 + 32]
-    pshufb            m5,         m4,     m3
-    pshufb            m4,         [tab_Tm]
-    pmaddubsw         m4,         m0
-    pmaddubsw         m5,         m1
-    paddw             m4,         m5
-    pmaddwd           m4,         m7
-    vbroadcasti128    m5,         [r0 + 40]
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [tab_Tm]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-    packssdw          m4,         m5
-    pmulhrsw          m4,         [pw_512]
-    packuswb          m4,         m4
-    vpermq            m4,         m4,     11011000b
-    pshufd            xm4,        xm4,    11011000b
-    movu              [r2 + 32],  xm4
-
-    lea               r0,         [r0 + r1]
-    lea               r2,         [r2 + r3]
-    dec               r4d
-    jnz               .loop
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_4x4, 4,6,6
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vpbroadcastd      m2,           [pw_1]
-    vbroadcasti128    m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec                r0
-
-    ; Row 0-1
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    vinserti128       m3,           m3,      [r0 + r1],     1
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 2-3
-    lea               r0,           [r0 + r1 * 2]
-    vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    vinserti128       m4,           m4,      [r0 + r1],     1
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    pmulhrsw          m3,           [pw_512]
-    vextracti128      xm4,          m3,     1
-    packuswb          xm3,          xm4
-
-    lea               r0,           [r3 * 3]
-    movd              [r2],         xm3
-    pextrd            [r2+r3],      xm3,     2
-    pextrd            [r2+r3*2],    xm3,     1
-    pextrd            [r2+r0],      xm3,     3
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_2x4, 4, 6, 3
-    mov               r4d,           r4m
-
-%ifdef PIC
-    lea               r5,            [tab_ChromaCoeff]
-    vpbroadcastd      m0,            [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,            [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    dec               r0
-    lea               r4,            [r1 * 3]
-    movq              xm1,           [r0]
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m1,            m1,          xm2,          1
-    pshufb            m1,            [interp4_hpp_shuf]
-    pmaddubsw         m1,            m0
-    pmaddwd           m1,            [pw_1]
-    vextracti128      xm2,           m1,          1
-    packssdw          xm1,           xm2
-    pmulhrsw          xm1,           [pw_512]
-    packuswb          xm1,           xm1
-
-    lea               r4,            [r3 * 3]
-    pextrw            [r2],          xm1,         0
-    pextrw            [r2 + r3],     xm1,         1
-    pextrw            [r2 + r3 * 2], xm1,         2
-    pextrw            [r2 + r4],     xm1,         3
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_2x8, 4, 6, 6
-    mov               r4d,           r4m
-
-%ifdef PIC
-    lea               r5,            [tab_ChromaCoeff]
-    vpbroadcastd      m0,            [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,            [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m4,            [interp4_hpp_shuf]
-    mova              m5,            [pw_1]
-    dec               r0
-    lea               r4,            [r1 * 3]
-    movq              xm1,           [r0]
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m1,            m1,          xm2,          1
-    lea               r0,            [r0 + r1 * 4]
-    movq              xm3,           [r0]
-    movhps            xm3,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m3,            m3,          xm2,          1
-
-    pshufb            m1,            m4
-    pshufb            m3,            m4
-    pmaddubsw         m1,            m0
-    pmaddubsw         m3,            m0
-    pmaddwd           m1,            m5
-    pmaddwd           m3,            m5
-    packssdw          m1,            m3
-    pmulhrsw          m1,            [pw_512]
-    vextracti128      xm2,           m1,          1
-    packuswb          xm1,           xm2
-
-    lea               r4,            [r3 * 3]
-    pextrw            [r2],          xm1,         0
-    pextrw            [r2 + r3],     xm1,         1
-    pextrw            [r2 + r3 * 2], xm1,         4
-    pextrw            [r2 + r4],     xm1,         5
-    lea               r2,            [r2 + r3 * 4]
-    pextrw            [r2],          xm1,         2
-    pextrw            [r2 + r3],     xm1,         3
-    pextrw            [r2 + r3 * 2], xm1,         6
-    pextrw            [r2 + r4],     xm1,         7
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_32x32, 4,6,7
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-    mova              m6,           [pw_512]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          32
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    vbroadcasti128    m4,           [r0 + 16]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + 20]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-
-    movu              [r2],         m3
-    lea               r2,           [r2 + r3]
-    lea               r0,           [r0 + r1]
-    dec               r4d
-    jnz               .loop
-    RET
-
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_16x16, 4, 6, 7
-    mov               r4d,          r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m6,           [pw_512]
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          8
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    ; Row 1
-    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-
-    vextracti128      xm4,          m3,       1
-    movu              [r2],         xm3
-    movu              [r2 + r3],    xm4
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1 * 2]
-    dec               r4d
-    jnz               .loop
-    RET
-;--------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-    IPFILTER_LUMA 4, 4, pp
-    IPFILTER_LUMA 4, 8, pp
-    IPFILTER_LUMA 12, 16, pp
-    IPFILTER_LUMA 4, 16, pp
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_8x8, 4,6,6
-    mov               r4d,    r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    movu              m1,           [tab_Tm]
-    vpbroadcastd      m2,           [pw_1]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    sub               r0,           1
-    mov               r4d,          2
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 1
-    vbroadcasti128    m4,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           [pw_512]
-    lea               r0,           [r0 + r1 * 2]
-
-    ; Row 2
-    vbroadcasti128    m4,           [r0 ]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    ; Row 3
-    vbroadcasti128    m5,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           [pw_512]
-
-    packuswb          m3,           m4
-    mova              m5,           [interp_4tap_8x8_horiz_shuf]
-    vpermd            m3,           m5,     m3
-    vextracti128      xm4,          m3,     1
-    movq              [r2],         xm3
-    movhps            [r2 + r3],    xm3
-    lea               r2,           [r2 + r3 * 2]
-    movq              [r2],         xm4
-    movhps            [r2 + r3],    xm4
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1*2]
-    dec               r4d
-    jnz               .loop
-    RET
-
-    IPFILTER_LUMA_AVX2 16, 4
-    IPFILTER_LUMA_AVX2 16, 8
-    IPFILTER_LUMA_AVX2 16, 12
-    IPFILTER_LUMA_AVX2 16, 16
-    IPFILTER_LUMA_AVX2 16, 32
-    IPFILTER_LUMA_AVX2 16, 64
-
-    IPFILTER_LUMA_32x_avx2 32 , 8
-    IPFILTER_LUMA_32x_avx2 32 , 16
-    IPFILTER_LUMA_32x_avx2 32 , 24
-    IPFILTER_LUMA_32x_avx2 32 , 32
-    IPFILTER_LUMA_32x_avx2 32 , 64
-
-    IPFILTER_LUMA_64x_avx2 64 , 64
-    IPFILTER_LUMA_64x_avx2 64 , 48
-    IPFILTER_LUMA_64x_avx2 64 , 32
-    IPFILTER_LUMA_64x_avx2 64 , 16
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_8x2, 4, 6, 5
-    mov               r4d,          r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,           [tab_Tm]
-    mova              m2,           [pw_1]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 1
-    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           [pw_512]
-    vextracti128      xm4,          m3,          1
-    packuswb          xm3,          xm4
-    pshufd            xm3,          xm3,         11011000b
-    movq              [r2],         xm3
-    movhps            [r2 + r3],    xm3
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_8x6, 4, 6, 7
-    mov               r4d,           r4m
-
-%ifdef PIC
-    lea               r5,            [tab_ChromaCoeff]
-    vpbroadcastd      m0,            [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,            [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,            [tab_Tm]
-    mova              m2,            [pw_1]
-    mova              m6,            [pw_512]
-    lea               r4,            [r1 * 3]
-    lea               r5,            [r3 * 3]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    ; Row 0
-    vbroadcasti128    m3,            [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,            m1
-    pmaddubsw         m3,            m0
-    pmaddwd           m3,            m2
-
-    ; Row 1
-    vbroadcasti128    m4,            [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,            m1
-    pmaddubsw         m4,            m0
-    pmaddwd           m4,            m2
-    packssdw          m3,            m4
-    pmulhrsw          m3,            m6
-
-    ; Row 2
-    vbroadcasti128    m4,            [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,            m1
-    pmaddubsw         m4,            m0
-    pmaddwd           m4,            m2
-
-    ; Row 3
-    vbroadcasti128    m5,            [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,            m1
-    pmaddubsw         m5,            m0
-    pmaddwd           m5,            m2
-    packssdw          m4,            m5
-    pmulhrsw          m4,            m6
-
-    packuswb          m3,            m4
-    mova              m5,            [interp8_hps_shuf]
-    vpermd            m3,            m5,          m3
-    vextracti128      xm4,           m3,          1
-    movq              [r2],          xm3
-    movhps            [r2 + r3],     xm3
-    movq              [r2 + r3 * 2], xm4
-    movhps            [r2 + r5],     xm4
-    lea               r2,            [r2 + r3 * 4]
-    lea               r0,            [r0 + r1 * 4]
-    ; Row 4
-    vbroadcasti128    m3,            [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,            m1
-    pmaddubsw         m3,            m0
-    pmaddwd           m3,            m2
-
-    ; Row 5
-    vbroadcasti128    m4,            [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,            m1
-    pmaddubsw         m4,            m0
-    pmaddwd           m4,            m2
-    packssdw          m3,            m4
-    pmulhrsw          m3,            m6
-    vextracti128      xm4,           m3,          1
-    packuswb          xm3,           xm4
-    pshufd            xm3,           xm3,         11011000b
-    movq              [r2],          xm3
-    movhps            [r2 + r3],     xm3
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_6x8, 4, 6, 7
-    mov               r4d,               r4m
-
-%ifdef PIC
-    lea               r5,                [tab_ChromaCoeff]
-    vpbroadcastd      m0,                [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,                [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,                [tab_Tm]
-    mova              m2,                [pw_1]
-    mova              m6,                [pw_512]
-    lea               r4,                [r1 * 3]
-    lea               r5,                [r3 * 3]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-%rep 2
-    ; Row 0
-    vbroadcasti128    m3,                [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,                m1
-    pmaddubsw         m3,                m0
-    pmaddwd           m3,                m2
-
-    ; Row 1
-    vbroadcasti128    m4,                [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,                m1
-    pmaddubsw         m4,                m0
-    pmaddwd           m4,                m2
-    packssdw          m3,                m4
-    pmulhrsw          m3,                m6
-
-    ; Row 2
-    vbroadcasti128    m4,                [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,                m1
-    pmaddubsw         m4,                m0
-    pmaddwd           m4,                m2
-
-    ; Row 3
-    vbroadcasti128    m5,                [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,                m1
-    pmaddubsw         m5,                m0
-    pmaddwd           m5,                m2
-    packssdw          m4,                m5
-    pmulhrsw          m4,                m6
-
-    packuswb          m3,                m4
-    vextracti128      xm4,               m3,          1
-    movd              [r2],              xm3
-    pextrw            [r2 + 4],          xm4,         0
-    pextrd            [r2 + r3],         xm3,         1
-    pextrw            [r2 + r3 + 4],     xm4,         2
-    pextrd            [r2 + r3 * 2],     xm3,         2
-    pextrw            [r2 + r3 * 2 + 4], xm4,         4
-    pextrd            [r2 + r5],         xm3,         3
-    pextrw            [r2 + r5 + 4],     xm4,         6
-    lea               r2,                [r2 + r3 * 4]
-    lea               r0,                [r0 + r1 * 4]
-%endrep
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_64xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------;
-%macro IPFILTER_CHROMA_HPS_64xN 1
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_64x%1, 4,7,6
-    mov             r4d, r4m
-    mov             r5d, r5m
-    add             r3d, r3d
-
-%ifdef PIC
-    lea               r6,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,           [pw_1]
-    vbroadcasti128     m5,           [pw_2000]
-    mova               m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-    mov                r6d,         %1
-    dec                r0
-    test                r5d,      r5d
-    je                 .loop
-    sub                r0 ,         r1
-    add                r6d ,        3
-
-.loop
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 8]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-    vpermq            m3,           m3,          11011000b
-    movu              [r2],         m3
-
-    vbroadcasti128    m3,           [r0 + 16]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 24]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-    vpermq            m3,           m3,          11011000b
-    movu              [r2 + 32],    m3
-
-    vbroadcasti128    m3,           [r0 + 32]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 40]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-    vpermq            m3,           m3,          11011000b
-    movu              [r2 + 64],    m3
-
-    vbroadcasti128    m3,           [r0 + 48]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 56]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-    vpermq            m3,           m3,          11011000b
-    movu              [r2 + 96],    m3
-
-    add                r2,           r3
-    add                r0,           r1
-    dec                r6d
-    jnz                .loop
-    RET
-%endmacro
-
-   IPFILTER_CHROMA_HPS_64xN 64
-   IPFILTER_CHROMA_HPS_64xN 32
-   IPFILTER_CHROMA_HPS_64xN 48
-   IPFILTER_CHROMA_HPS_64xN 16
-
-;-----------------------------------------------------------------------------------------------------------------------------
-;void interp_horiz_ps_c(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-
-%macro IPFILTER_LUMA_PS_4xN_AVX2 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_ps_4x%1, 6,7,6
-    mov                         r5d,               r5m
-    mov                         r4d,               r4m
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    vpbroadcastq                m0,                [r6 + r4 * 8]
-%else
-    vpbroadcastq                m0,                [tab_LumaCoeff + r4 * 8]
-%endif
-    mova                        m1,                [tab_Lm]
-    add                         r3d,               r3d
-    vbroadcasti128              m2,                [pw_2000]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - pw_2000
-
-    sub                         r0,                3
-    test                        r5d,               r5d
-    mov                         r5d,               %1                           ; loop count variable - height
-    jz                         .preloop
-    lea                         r6,                [r1 * 3]                     ; r8 = (N / 2 - 1) * srcStride
-    sub                         r0,                r6                           ; r0(src) - 3 * srcStride
-    add                         r5d,               7                            ; need extra 7 rows, just set a specially flag here, blkheight += N - 1  (7 - 3 = 4 ; since the last three rows not in loop)
-
-.preloop:
-    lea                         r6,                [r3 * 3]
-.loop
-    ; Row 0-1
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm
-    pmaddubsw                   m3,                m0
-    vbroadcasti128              m4,                [r0 + r1]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m1
-    pmaddubsw                   m4,                m0
-    phaddw                      m3,                m4                           ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
-
-    ; Row 2-3
-    lea                         r0,                [r0 + r1 * 2]                ;3rd row(i.e 2nd row)
-    vbroadcasti128              m4,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m1
-    pmaddubsw                   m4,                m0
-    vbroadcasti128              m5,                [r0 + r1]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m5,                m1
-    pmaddubsw                   m5,                m0
-    phaddw                      m4,                m5                           ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
-    phaddw                      m3,                m4                           ; all rows and col completed.
-
-    mova                        m5,                [interp8_hps_shuf]
-    vpermd                      m3,                m5,               m3
-    psubw                       m3,                m2
-
-    vextracti128                xm4,               m3,               1
-    movq                        [r2],              xm3                          ;row 0
-    movhps                      [r2 + r3],         xm3                          ;row 1
-    movq                        [r2 + r3 * 2],     xm4                          ;row 2
-    movhps                      [r2 + r6],         xm4                          ;row 3
-
-    lea                         r0,                [r0 + r1 * 2]                ; first loop src ->5th row(i.e 4)
-    lea                         r2,                [r2 + r3 * 4]                ; first loop dst ->5th row(i.e 4)
-    sub                         r5d,               4
-    jz                         .end
-    cmp                         r5d,               4
-    jge                        .loop
-
-    ; Row 8-9
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m3,                m1
-    pmaddubsw                   m3,                m0
-    vbroadcasti128              m4,                [r0 + r1]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m1
-    pmaddubsw                   m4,                m0
-    phaddw                      m3,                m4                           ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
-
-    ; Row 10
-    vbroadcasti128              m4,                [r0 + r1 * 2]                ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m1
-    pmaddubsw                   m4,                m0
-    phaddw                      m4,                m4                           ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
-    phaddw                      m3,                m4
-
-    vpermd                      m3,                m5,            m3            ; m5 don't broken in above
-    psubw                       m3,                m2
-
-    vextracti128                xm4,               m3,            1
-    movq                        [r2],              xm3
-    movhps                      [r2 + r3],         xm3
-    movq                        [r2 + r3 * 2],     xm4
-.end
-    RET
-%endif
-%endmacro
-
-    IPFILTER_LUMA_PS_4xN_AVX2 4
-    IPFILTER_LUMA_PS_4xN_AVX2 8
-    IPFILTER_LUMA_PS_4xN_AVX2 16
-
-%macro IPFILTER_LUMA_PS_8xN_AVX2 1
-; TODO: verify and enable on X86 mode
-%if ARCH_X86_64 == 1
-; void filter_hps(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-INIT_YMM avx2
-cglobal interp_8tap_horiz_ps_8x%1, 4,7,6
-    mov                         r5d,        r5m
-    mov                         r4d,        r4m
-    shl                         r4d,        7
-%ifdef PIC
-    lea                         r6,         [pb_LumaCoeffVer]
-    add                         r6,         r4
-%else
-    lea                         r6,         [pb_LumaCoeffVer + r4]
-%endif
-    add                         r3d,        r3d
-    vpbroadcastd                m0,         [pw_2000]
-    sub                         r0,         3
-    lea                         r4,         [pb_8tap_hps_0]
-    vbroadcasti128              m5,         [r4 + 0 * mmsize]
-
-    ; check row count extend for interpolateHV
-    test                        r5d,        r5d;
-    mov                         r5d,        %1
-    jz                         .enter_loop
-    lea                         r4,         [r1 * 3]                        ; r8 = (N / 2 - 1) * srcStride
-    sub                         r0,         r4                              ; r0(src)-r8
-    add                         r5d,        8-1-2                           ; blkheight += N - 1  (7 - 3 = 4 ; since the last three rows not in loop)
-
-.enter_loop:
-    lea                         r4,         [pb_8tap_hps_0]
-
-    ; ***** register map *****
-    ; m0 - pw_2000
-    ; r4 - base pointer of shuffle order table
-    ; r5 - count of loop
-    ; r6 - point to LumaCoeff
-.loop:
-
-    ; Row 0-1
-    movu                        xm1,        [r0]
-    movu                        xm2,        [r0 + r1]
-    vinserti128                 m1,         m1,         xm2, 1
-    pshufb                      m2,         m1,         m5                  ; [0 1 1 2 2 3 3 4 ...]
-    pshufb                      m3,         m1,         [r4 + 1 * mmsize]   ; [2 3 3 4 4 5 5 6 ...]
-    pshufb                      m4,         m1,         [r4 + 2 * mmsize]   ; [4 5 5 6 6 7 7 8 ...]
-    pshufb                      m1,         m1,         [r4 + 3 * mmsize]   ; [6 7 7 8 8 9 9 A ...]
-    pmaddubsw                   m2,         [r6 + 0 * mmsize]
-    pmaddubsw                   m3,         [r6 + 1 * mmsize]
-    pmaddubsw                   m4,         [r6 + 2 * mmsize]
-    pmaddubsw                   m1,         [r6 + 3 * mmsize]
-    paddw                       m2,         m3
-    paddw                       m1,         m4
-    paddw                       m1,         m2
-    psubw                       m1,         m0
-
-    vextracti128                xm2,        m1,         1
-    movu                        [r2],       xm1                             ; row 0
-    movu                        [r2 + r3],  xm2                             ; row 1
-
-    lea                         r0,         [r0 + r1 * 2]                   ; first loop src ->5th row(i.e 4)
-    lea                         r2,         [r2 + r3 * 2]                   ; first loop dst ->5th row(i.e 4)
-    sub                         r5d,        2
-    jg                         .loop
-    jz                         .end
-
-    ; last row
-    movu                        xm1,        [r0]
-    pshufb                      xm2,        xm1,         xm5                ; [0 1 1 2 2 3 3 4 ...]
-    pshufb                      xm3,        xm1,         [r4 + 1 * mmsize]  ; [2 3 3 4 4 5 5 6 ...]
-    pshufb                      xm4,        xm1,         [r4 + 2 * mmsize]  ; [4 5 5 6 6 7 7 8 ...]
-    pshufb                      xm1,        xm1,         [r4 + 3 * mmsize]  ; [6 7 7 8 8 9 9 A ...]
-    pmaddubsw                   xm2,        [r6 + 0 * mmsize]
-    pmaddubsw                   xm3,        [r6 + 1 * mmsize]
-    pmaddubsw                   xm4,        [r6 + 2 * mmsize]
-    pmaddubsw                   xm1,        [r6 + 3 * mmsize]
-    paddw                       xm2,        xm3
-    paddw                       xm1,        xm4
-    paddw                       xm1,        xm2
-    psubw                       xm1,        xm0
-    movu                        [r2],       xm1                          ;row 0
-.end
-    RET
-%endif
-%endmacro ; IPFILTER_LUMA_PS_8xN_AVX2
-
-    IPFILTER_LUMA_PS_8xN_AVX2  4
-    IPFILTER_LUMA_PS_8xN_AVX2  8
-    IPFILTER_LUMA_PS_8xN_AVX2 16
-    IPFILTER_LUMA_PS_8xN_AVX2 32
-
-
-%macro IPFILTER_LUMA_PS_16x_AVX2 2
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_horiz_ps_%1x%2, 6, 10, 7
-    mov                         r5d,               r5m
-    mov                         r4d,               r4m
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    vpbroadcastq                m0,                [r6 + r4 * 8]
-%else
-    vpbroadcastq                m0,                [tab_LumaCoeff + r4 * 8]
-%endif
-    mova                        m6,                [tab_Lm + 32]
-    mova                        m1,                [tab_Lm]
-    mov                         r9,                %2                           ;height
-    add                         r3d,               r3d
-    vbroadcasti128              m2,                [pw_2000]
-
-    ; register map
-    ; m0      - interpolate coeff
-    ; m1 , m6 - shuffle order table
-    ; m2      - pw_2000
-
-    xor                         r7,                r7                          ; loop count variable
-    sub                         r0,                3
-    test                        r5d,               r5d
-    jz                          .label
-    lea                         r8,                [r1 * 3]                     ; r8 = (N / 2 - 1) * srcStride
-    sub                         r0,                r8                           ; r0(src)-r8
-    add                         r9,                7                            ; blkheight += N - 1  (7 - 1 = 6 ; since the last one row not in loop)
-
-.label
-    ; Row 0
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
-    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    phaddw                      m3,                m4                           ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
-
-    vbroadcasti128              m4,                [r0 + 8]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m5,                m4,            m6            ;row 1 (col 4 to 7)
-    pshufb                      m4,                m1                           ;row 1 (col 0 to 3)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    phaddw                      m4,                m5                           ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
-    phaddw                      m3,                m4                           ; all rows and col completed.
-
-    mova                        m5,                [interp8_hps_shuf]
-    vpermd                      m3,                m5,               m3
-    psubw                       m3,                m2
-
-    movu                        [r2],              m3                          ;row 0
-
-    lea                         r0,                [r0 + r1]                ; first loop src ->5th row(i.e 4)
-    lea                         r2,                [r2 + r3]                ; first loop dst ->5th row(i.e 4)
-    dec                         r9d
-    jnz                         .label
-
-    RET
-%endif
-%endmacro
-
-
-    IPFILTER_LUMA_PS_16x_AVX2 16 , 16
-    IPFILTER_LUMA_PS_16x_AVX2 16 , 8
-    IPFILTER_LUMA_PS_16x_AVX2 16 , 12
-    IPFILTER_LUMA_PS_16x_AVX2 16 , 4
-    IPFILTER_LUMA_PS_16x_AVX2 16 , 32
-    IPFILTER_LUMA_PS_16x_AVX2 16 , 64
-
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_LUMA_PP_W8 2
-INIT_XMM sse4
-cglobal interp_8tap_horiz_pp_%1x%2, 4,6,7
-    mov         r4d, r4m
-
-%ifdef PIC
-    lea         r5, [tab_LumaCoeff]
-    movh        m3, [r5 + r4 * 8]
-%else
-    movh        m3, [tab_LumaCoeff + r4 * 8]
-%endif
-    pshufd      m0, m3, 0                       ; m0 = coeff-L
-    pshufd      m1, m3, 0x55                    ; m1 = coeff-H
-    lea         r5, [tab_Tm]                    ; r5 = shuffle
-    mova        m2, [pw_512]                    ; m2 = 512
-
-    mov         r4d, %2
-.loopH:
-%assign x 0
-%rep %1 / 8
-    movu        m3, [r0 - 3 + x]                ; m3 = [F E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb      m4, m3, [r5 + 0*16]             ; m4 = [6 5 4 3 5 4 3 2 4 3 2 1 3 2 1 0]
-    pshufb      m5, m3, [r5 + 1*16]             ; m5 = [A 9 8 7 9 8 7 6 8 7 6 5 7 6 5 4]
-    pshufb          m3, [r5 + 2*16]             ; m3 = [E D C B D C B A C B A 9 B A 9 8]
-    pmaddubsw   m4, m0
-    pmaddubsw   m6, m5, m1
-    pmaddubsw   m5, m0
-    pmaddubsw   m3, m1
-    paddw       m4, m6
-    paddw       m5, m3
-    phaddw      m4, m5
-    pmulhrsw    m4, m2
-    packuswb    m4, m4
-    movh        [r2 + x], m4
-%assign x x+8
-%endrep
-
-    add       r0, r1
-    add       r2, r3
-
-    dec       r4d
-    jnz      .loopH
-    RET
-%endmacro
-
-    IPFILTER_LUMA_PP_W8      8,  4
-    IPFILTER_LUMA_PP_W8      8,  8
-    IPFILTER_LUMA_PP_W8      8, 16
-    IPFILTER_LUMA_PP_W8      8, 32
-    IPFILTER_LUMA_PP_W8     16,  4
-    IPFILTER_LUMA_PP_W8     16,  8
-    IPFILTER_LUMA_PP_W8     16, 12
-    IPFILTER_LUMA_PP_W8     16, 16
-    IPFILTER_LUMA_PP_W8     16, 32
-    IPFILTER_LUMA_PP_W8     16, 64
-    IPFILTER_LUMA_PP_W8     24, 32
-    IPFILTER_LUMA_PP_W8     32,  8
-    IPFILTER_LUMA_PP_W8     32, 16
-    IPFILTER_LUMA_PP_W8     32, 24
-    IPFILTER_LUMA_PP_W8     32, 32
-    IPFILTER_LUMA_PP_W8     32, 64
-    IPFILTER_LUMA_PP_W8     48, 64
-    IPFILTER_LUMA_PP_W8     64, 16
-    IPFILTER_LUMA_PP_W8     64, 32
-    IPFILTER_LUMA_PP_W8     64, 48
-    IPFILTER_LUMA_PP_W8     64, 64
-
-;----------------------------------------------------------------------------------------------------------------------------
-; void interp_8tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;----------------------------------------------------------------------------------------------------------------------------
-    IPFILTER_LUMA 4, 4, ps
-    IPFILTER_LUMA 8, 8, ps
-    IPFILTER_LUMA 8, 4, ps
-    IPFILTER_LUMA 4, 8, ps
-    IPFILTER_LUMA 16, 16, ps
-    IPFILTER_LUMA 16, 8, ps
-    IPFILTER_LUMA 8, 16, ps
-    IPFILTER_LUMA 16, 12, ps
-    IPFILTER_LUMA 12, 16, ps
-    IPFILTER_LUMA 16, 4, ps
-    IPFILTER_LUMA 4, 16, ps
-    IPFILTER_LUMA 32, 32, ps
-    IPFILTER_LUMA 32, 16, ps
-    IPFILTER_LUMA 16, 32, ps
-    IPFILTER_LUMA 32, 24, ps
-    IPFILTER_LUMA 24, 32, ps
-    IPFILTER_LUMA 32, 8, ps
-    IPFILTER_LUMA 8, 32, ps
-    IPFILTER_LUMA 64, 64, ps
-    IPFILTER_LUMA 64, 32, ps
-    IPFILTER_LUMA 32, 64, ps
-    IPFILTER_LUMA 64, 48, ps
-    IPFILTER_LUMA 48, 64, ps
-    IPFILTER_LUMA 64, 16, ps
-    IPFILTER_LUMA 16, 64, ps
-
 ;-----------------------------------------------------------------------------
 ; Interpolate HV
 ;-----------------------------------------------------------------------------
@@ -5975,6259 +999,6 @@
     RET
 
 ;-----------------------------------------------------------------------------
-;void interp_4tap_vert_pp_2x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_2x4, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-    lea         r4,        [r1 * 3]
-    lea         r5,        [r0 + 4 * r1]
-    pshufb      m0,        [tab_Cm]
-    mova        m1,        [pw_512]
-
-    movd        m2,        [r0]
-    movd        m3,        [r0 + r1]
-    movd        m4,        [r0 + 2 * r1]
-    movd        m5,        [r0 + r4]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m6,        m4,        m5
-    punpcklbw   m2,        m6
-
-    pmaddubsw   m2,        m0
-
-    movd        m6,        [r5]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m7,        m5,        m6
-    punpcklbw   m3,        m7
-
-    pmaddubsw   m3,        m0
-
-    phaddw      m2,        m3
-
-    pmulhrsw    m2,        m1
-
-    movd        m7,        [r5 + r1]
-
-    punpcklbw   m4,        m5
-    punpcklbw   m3,        m6,        m7
-    punpcklbw   m4,        m3
-
-    pmaddubsw   m4,        m0
-
-    movd        m3,        [r5 + 2 * r1]
-
-    punpcklbw   m5,        m6
-    punpcklbw   m7,        m3
-    punpcklbw   m5,        m7
-
-    pmaddubsw   m5,        m0
-
-    phaddw      m4,        m5
-
-    pmulhrsw    m4,        m1
-    packuswb    m2,        m4
-
-    pextrw      [r2],      m2, 0
-    pextrw      [r2 + r3], m2, 2
-    lea         r2,        [r2 + 2 * r3]
-    pextrw      [r2],      m2, 4
-    pextrw      [r2 + r3], m2, 6
-
-    RET
-
-%macro FILTER_VER_CHROMA_AVX2_2x4 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_2x4, 4, 6, 2
-    mov             r4d, r4m
-    shl             r4d, 5
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff_V]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeff_V + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-
-    pinsrw          xm1, [r0], 0
-    pinsrw          xm1, [r0 + r1], 1
-    pinsrw          xm1, [r0 + r1 * 2], 2
-    pinsrw          xm1, [r0 + r4], 3
-    lea             r0, [r0 + r1 * 4]
-    pinsrw          xm1, [r0], 4
-    pinsrw          xm1, [r0 + r1], 5
-    pinsrw          xm1, [r0 + r1 * 2], 6
-
-    pshufb          xm0, xm1, [interp_vert_shuf]
-    pshufb          xm1, [interp_vert_shuf + 32]
-    vinserti128     m0, m0, xm1, 1
-    pmaddubsw       m0, [r5]
-    vextracti128    xm1, m0, 1
-    paddw           xm0, xm1
-%ifidn %1,pp
-    pmulhrsw        xm0, [pw_512]
-    packuswb        xm0, xm0
-    lea             r4, [r3 * 3]
-    pextrw          [r2], xm0, 0
-    pextrw          [r2 + r3], xm0, 1
-    pextrw          [r2 + r3 * 2], xm0, 2
-    pextrw          [r2 + r4], xm0, 3
-%else
-    add             r3d, r3d
-    lea             r4, [r3 * 3]
-    psubw           xm0, [pw_2000]
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    pextrd          [r2 + r3 * 2], xm0, 2
-    pextrd          [r2 + r4], xm0, 3
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_2x4 pp
-    FILTER_VER_CHROMA_AVX2_2x4 ps
-
-%macro FILTER_VER_CHROMA_AVX2_2x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_2x8, 4, 6, 2
-    mov             r4d, r4m
-    shl             r4d, 6
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-
-    pinsrw          xm1, [r0], 0
-    pinsrw          xm1, [r0 + r1], 1
-    pinsrw          xm1, [r0 + r1 * 2], 2
-    pinsrw          xm1, [r0 + r4], 3
-    lea             r0, [r0 + r1 * 4]
-    pinsrw          xm1, [r0], 4
-    pinsrw          xm1, [r0 + r1], 5
-    pinsrw          xm1, [r0 + r1 * 2], 6
-    pinsrw          xm1, [r0 + r4], 7
-    movhlps         xm0, xm1
-    lea             r0, [r0 + r1 * 4]
-    pinsrw          xm0, [r0], 4
-    pinsrw          xm0, [r0 + r1], 5
-    pinsrw          xm0, [r0 + r1 * 2], 6
-    vinserti128     m1, m1, xm0, 1
-
-    pshufb          m0, m1, [interp_vert_shuf]
-    pshufb          m1, [interp_vert_shuf + 32]
-    pmaddubsw       m0, [r5]
-    pmaddubsw       m1, [r5 + 1 * mmsize]
-    paddw           m0, m1
-%ifidn %1,pp
-    pmulhrsw        m0, [pw_512]
-    vextracti128    xm1, m0, 1
-    packuswb        xm0, xm1
-    lea             r4, [r3 * 3]
-    pextrw          [r2], xm0, 0
-    pextrw          [r2 + r3], xm0, 1
-    pextrw          [r2 + r3 * 2], xm0, 2
-    pextrw          [r2 + r4], xm0, 3
-    lea             r2, [r2 + r3 * 4]
-    pextrw          [r2], xm0, 4
-    pextrw          [r2 + r3], xm0, 5
-    pextrw          [r2 + r3 * 2], xm0, 6
-    pextrw          [r2 + r4], xm0, 7
-%else
-    add             r3d, r3d
-    lea             r4, [r3 * 3]
-    psubw           m0, [pw_2000]
-    vextracti128    xm1, m0, 1
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    pextrd          [r2 + r3 * 2], xm0, 2
-    pextrd          [r2 + r4], xm0, 3
-    lea             r2, [r2 + r3 * 4]
-    movd            [r2], xm1
-    pextrd          [r2 + r3], xm1, 1
-    pextrd          [r2 + r3 * 2], xm1, 2
-    pextrd          [r2 + r4], xm1, 3
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_2x8 pp
-    FILTER_VER_CHROMA_AVX2_2x8 ps
-
-%macro FILTER_VER_CHROMA_AVX2_2x16 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_2x16, 4, 6, 3
-    mov             r4d, r4m
-    shl             r4d, 6
-    sub             r0,  r1
-
-%ifdef PIC
-    lea             r5,  [tab_ChromaCoeffVer_32]
-    add             r5,  r4
-%else
-    lea             r5,  [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4,  [r1 * 3]
-
-    movd            xm1, [r0]
-    pinsrw          xm1, [r0 + r1], 1
-    pinsrw          xm1, [r0 + r1 * 2], 2
-    pinsrw          xm1, [r0 + r4], 3
-    lea             r0,  [r0 + r1 * 4]
-    pinsrw          xm1, [r0], 4
-    pinsrw          xm1, [r0 + r1], 5
-    pinsrw          xm1, [r0 + r1 * 2], 6
-    pinsrw          xm1, [r0 + r4], 7
-    lea             r0,  [r0 + r1 * 4]
-    pinsrw          xm0, [r0], 4
-    pinsrw          xm0, [r0 + r1], 5
-    pinsrw          xm0, [r0 + r1 * 2], 6
-    pinsrw          xm0, [r0 + r4], 7
-    punpckhqdq      xm0, xm1, xm0
-    vinserti128     m1,  m1,  xm0,  1
-
-    pshufb          m2,  m1,  [interp_vert_shuf]
-    pshufb          m1,  [interp_vert_shuf + 32]
-    pmaddubsw       m2,  [r5]
-    pmaddubsw       m1,  [r5 + 1 * mmsize]
-    paddw           m2,  m1
-
-    lea             r0,  [r0 + r1 * 4]
-    pinsrw          xm1, [r0], 4
-    pinsrw          xm1, [r0 + r1], 5
-    pinsrw          xm1, [r0 + r1 * 2], 6
-    pinsrw          xm1, [r0 + r4], 7
-    punpckhqdq      xm1, xm0, xm1
-    lea             r0,  [r0 + r1 * 4]
-    pinsrw          xm0, [r0], 4
-    pinsrw          xm0, [r0 + r1], 5
-    pinsrw          xm0, [r0 + r1 * 2], 6
-    punpckhqdq      xm0, xm1, xm0
-    vinserti128     m1,  m1,  xm0,  1
-
-    pshufb          m0,  m1,  [interp_vert_shuf]
-    pshufb          m1,  [interp_vert_shuf + 32]
-    pmaddubsw       m0,  [r5]
-    pmaddubsw       m1,  [r5 + 1 * mmsize]
-    paddw           m0,  m1
-%ifidn %1,pp
-    mova            m1,  [pw_512]
-    pmulhrsw        m2,  m1
-    pmulhrsw        m0,  m1
-    packuswb        m2,  m0
-    lea             r4,  [r3 * 3]
-    pextrw          [r2], xm2, 0
-    pextrw          [r2 + r3], xm2, 1
-    pextrw          [r2 + r3 * 2], xm2, 2
-    pextrw          [r2 + r4], xm2, 3
-    vextracti128    xm0, m2, 1
-    lea             r2,  [r2 + r3 * 4]
-    pextrw          [r2], xm0, 0
-    pextrw          [r2 + r3], xm0, 1
-    pextrw          [r2 + r3 * 2], xm0, 2
-    pextrw          [r2 + r4], xm0, 3
-    lea             r2,  [r2 + r3 * 4]
-    pextrw          [r2], xm2, 4
-    pextrw          [r2 + r3], xm2, 5
-    pextrw          [r2 + r3 * 2], xm2, 6
-    pextrw          [r2 + r4], xm2, 7
-    lea             r2,  [r2 + r3 * 4]
-    pextrw          [r2], xm0, 4
-    pextrw          [r2 + r3], xm0, 5
-    pextrw          [r2 + r3 * 2], xm0, 6
-    pextrw          [r2 + r4], xm0, 7
-%else
-    add             r3d, r3d
-    lea             r4,  [r3 * 3]
-    vbroadcasti128  m1,  [pw_2000]
-    psubw           m2,  m1
-    psubw           m0,  m1
-    vextracti128    xm1, m2, 1
-    movd            [r2], xm2
-    pextrd          [r2 + r3], xm2, 1
-    pextrd          [r2 + r3 * 2], xm2, 2
-    pextrd          [r2 + r4], xm2, 3
-    lea             r2, [r2 + r3 * 4]
-    movd            [r2], xm1
-    pextrd          [r2 + r3], xm1, 1
-    pextrd          [r2 + r3 * 2], xm1, 2
-    pextrd          [r2 + r4], xm1, 3
-    vextracti128    xm1, m0, 1
-    lea             r2,  [r2 + r3 * 4]
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    pextrd          [r2 + r3 * 2], xm0, 2
-    pextrd          [r2 + r4], xm0, 3
-    lea             r2,  [r2 + r3 * 4]
-    movd            [r2], xm1
-    pextrd          [r2 + r3], xm1, 1
-    pextrd          [r2 + r3 * 2], xm1, 2
-    pextrd          [r2 + r4], xm1, 3
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_2x16 pp
-    FILTER_VER_CHROMA_AVX2_2x16 ps
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W2_H4 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_2x%2, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m0,        [tab_Cm]
-
-    mova        m1,        [pw_512]
-
-    mov         r4d,       %2
-    lea         r5,        [3 * r1]
-
-.loop:
-    movd        m2,        [r0]
-    movd        m3,        [r0 + r1]
-    movd        m4,        [r0 + 2 * r1]
-    movd        m5,        [r0 + r5]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m6,        m4,        m5
-    punpcklbw   m2,        m6
-
-    pmaddubsw   m2,        m0
-
-    lea         r0,        [r0 + 4 * r1]
-    movd        m6,        [r0]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m7,        m5,        m6
-    punpcklbw   m3,        m7
-
-    pmaddubsw   m3,        m0
-
-    phaddw      m2,        m3
-
-    pmulhrsw    m2,        m1
-
-    movd        m7,        [r0 + r1]
-
-    punpcklbw   m4,        m5
-    punpcklbw   m3,        m6,        m7
-    punpcklbw   m4,        m3
-
-    pmaddubsw   m4,        m0
-
-    movd        m3,        [r0 + 2 * r1]
-
-    punpcklbw   m5,        m6
-    punpcklbw   m7,        m3
-    punpcklbw   m5,        m7
-
-    pmaddubsw   m5,        m0
-
-    phaddw      m4,        m5
-
-    pmulhrsw    m4,        m1
-    packuswb    m2,        m4
-
-    pextrw      [r2],      m2, 0
-    pextrw      [r2 + r3], m2, 2
-    lea         r2,        [r2 + 2 * r3]
-    pextrw      [r2],      m2, 4
-    pextrw      [r2 + r3], m2, 6
-
-    lea         r2,        [r2 + 2 * r3]
-
-    sub         r4,        4
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W2_H4 2, 8
-
-    FILTER_V4_W2_H4 2, 16
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_4x2, 4, 6, 6
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m0,        [tab_Cm]
-    lea         r5,        [r0 + 2 * r1]
-
-    movd        m2,        [r0]
-    movd        m3,        [r0 + r1]
-    movd        m4,        [r5]
-    movd        m5,        [r5 + r1]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m1,        m4,        m5
-    punpcklbw   m2,        m1
-
-    pmaddubsw   m2,        m0
-
-    movd        m1,        [r0 + 4 * r1]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m5,        m1
-    punpcklbw   m3,        m5
-
-    pmaddubsw   m3,        m0
-
-    phaddw      m2,        m3
-
-    pmulhrsw    m2,        [pw_512]
-    packuswb    m2,        m2
-    movd        [r2],      m2
-    pextrd      [r2 + r3], m2,  1
-
-    RET
-
-%macro FILTER_VER_CHROMA_AVX2_4x2 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x2, 4, 6, 4
-    mov             r4d, r4m
-    shl             r4d, 5
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeff_V]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeff_V + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-
-    movd            xm1, [r0]
-    movd            xm2, [r0 + r1]
-    punpcklbw       xm1, xm2
-    movd            xm3, [r0 + r1 * 2]
-    punpcklbw       xm2, xm3
-    movlhps         xm1, xm2
-    movd            xm0, [r0 + r4]
-    punpcklbw       xm3, xm0
-    movd            xm2, [r0 + r1 * 4]
-    punpcklbw       xm0, xm2
-    movlhps         xm3, xm0
-    vinserti128     m1, m1, xm3, 1                          ; m1 = row[x x x 4 3 2 1 0]
-
-    pmaddubsw       m1, [r5]
-    vextracti128    xm3, m1, 1
-    paddw           xm1, xm3
-%ifidn %1,pp
-    pmulhrsw        xm1, [pw_512]
-    packuswb        xm1, xm1
-    movd            [r2], xm1
-    pextrd          [r2 + r3], xm1, 1
-%else
-    add             r3d, r3d
-    psubw           xm1, [pw_2000]
-    movq            [r2], xm1
-    movhps          [r2 + r3], xm1
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_4x2 pp
-    FILTER_VER_CHROMA_AVX2_4x2 ps
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_4x4, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m0,        [tab_Cm]
-    mova        m1,        [pw_512]
-    lea         r5,        [r0 + 4 * r1]
-    lea         r4,        [r1 * 3]
-
-    movd        m2,        [r0]
-    movd        m3,        [r0 + r1]
-    movd        m4,        [r0 + 2 * r1]
-    movd        m5,        [r0 + r4]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m6,        m4,        m5
-    punpcklbw   m2,        m6
-
-    pmaddubsw   m2,        m0
-
-    movd        m6,        [r5]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m7,        m5,        m6
-    punpcklbw   m3,        m7
-
-    pmaddubsw   m3,        m0
-
-    phaddw      m2,        m3
-
-    pmulhrsw    m2,        m1
-
-    movd        m7,        [r5 + r1]
-
-    punpcklbw   m4,        m5
-    punpcklbw   m3,        m6,        m7
-    punpcklbw   m4,        m3
-
-    pmaddubsw   m4,        m0
-
-    movd        m3,        [r5 + 2 * r1]
-
-    punpcklbw   m5,        m6
-    punpcklbw   m7,        m3
-    punpcklbw   m5,        m7
-
-    pmaddubsw   m5,        m0
-
-    phaddw      m4,        m5
-
-    pmulhrsw    m4,        m1
-
-    packuswb    m2,        m4
-    movd        [r2],      m2
-    pextrd      [r2 + r3], m2, 1
-    lea         r2,        [r2 + 2 * r3]
-    pextrd      [r2],      m2, 2
-    pextrd      [r2 + r3], m2, 3
-    RET
-%macro FILTER_VER_CHROMA_AVX2_4x4 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x4, 4, 6, 3
-    mov             r4d, r4m
-    shl             r4d, 6
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-
-    movd            xm1, [r0]
-    pinsrd          xm1, [r0 + r1], 1
-    pinsrd          xm1, [r0 + r1 * 2], 2
-    pinsrd          xm1, [r0 + r4], 3                       ; m1 = row[3 2 1 0]
-    lea             r0, [r0 + r1 * 4]
-    movd            xm2, [r0]
-    pinsrd          xm2, [r0 + r1], 1
-    pinsrd          xm2, [r0 + r1 * 2], 2                   ; m2 = row[x 6 5 4]
-    vinserti128     m1, m1, xm2, 1                          ; m1 = row[x 6 5 4 3 2 1 0]
-    mova            m2, [interp4_vpp_shuf1]
-    vpermd          m0, m2, m1                              ; m0 = row[4 3 3 2 2 1 1 0]
-    mova            m2, [interp4_vpp_shuf1 + mmsize]
-    vpermd          m1, m2, m1                              ; m1 = row[6 5 5 4 4 3 3 2]
-
-    mova            m2, [interp4_vpp_shuf]
-    pshufb          m0, m0, m2
-    pshufb          m1, m1, m2
-    pmaddubsw       m0, [r5]
-    pmaddubsw       m1, [r5 + mmsize]
-    paddw           m0, m1                                  ; m0 = WORD ROW[3 2 1 0]
-%ifidn %1,pp
-    pmulhrsw        m0, [pw_512]
-    vextracti128    xm1, m0, 1
-    packuswb        xm0, xm1
-    lea             r5, [r3 * 3]
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    pextrd          [r2 + r3 * 2], xm0, 2
-    pextrd          [r2 + r5], xm0, 3
-%else
-    add             r3d, r3d
-    psubw           m0, [pw_2000]
-    vextracti128    xm1, m0, 1
-    lea             r5, [r3 * 3]
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm1
-    movhps          [r2 + r5], xm1
-%endif
-    RET
-%endmacro
-    FILTER_VER_CHROMA_AVX2_4x4 pp
-    FILTER_VER_CHROMA_AVX2_4x4 ps
-
-%macro FILTER_VER_CHROMA_AVX2_4x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x8, 4, 6, 5
-    mov             r4d, r4m
-    shl             r4d, 6
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-
-    movd            xm1, [r0]
-    pinsrd          xm1, [r0 + r1], 1
-    pinsrd          xm1, [r0 + r1 * 2], 2
-    pinsrd          xm1, [r0 + r4], 3                       ; m1 = row[3 2 1 0]
-    lea             r0, [r0 + r1 * 4]
-    movd            xm2, [r0]
-    pinsrd          xm2, [r0 + r1], 1
-    pinsrd          xm2, [r0 + r1 * 2], 2
-    pinsrd          xm2, [r0 + r4], 3                       ; m2 = row[7 6 5 4]
-    vinserti128     m1, m1, xm2, 1                          ; m1 = row[7 6 5 4 3 2 1 0]
-    lea             r0, [r0 + r1 * 4]
-    movd            xm3, [r0]
-    pinsrd          xm3, [r0 + r1], 1
-    pinsrd          xm3, [r0 + r1 * 2], 2                   ; m3 = row[x 10 9 8]
-    vinserti128     m2, m2, xm3, 1                          ; m2 = row[x 10 9 8 7 6 5 4]
-    mova            m3, [interp4_vpp_shuf1]
-    vpermd          m0, m3, m1                              ; m0 = row[4 3 3 2 2 1 1 0]
-    vpermd          m4, m3, m2                              ; m4 = row[8 7 7 6 6 5 5 4]
-    mova            m3, [interp4_vpp_shuf1 + mmsize]
-    vpermd          m1, m3, m1                              ; m1 = row[6 5 5 4 4 3 3 2]
-    vpermd          m2, m3, m2                              ; m2 = row[10 9 9 8 8 7 7 6]
-
-    mova            m3, [interp4_vpp_shuf]
-    pshufb          m0, m0, m3
-    pshufb          m1, m1, m3
-    pshufb          m2, m2, m3
-    pshufb          m4, m4, m3
-    pmaddubsw       m0, [r5]
-    pmaddubsw       m4, [r5]
-    pmaddubsw       m1, [r5 + mmsize]
-    pmaddubsw       m2, [r5 + mmsize]
-    paddw           m0, m1                                  ; m0 = WORD ROW[3 2 1 0]
-    paddw           m4, m2                                  ; m4 = WORD ROW[7 6 5 4]
-%ifidn %1,pp
-    pmulhrsw        m0, [pw_512]
-    pmulhrsw        m4, [pw_512]
-    packuswb        m0, m4
-    vextracti128    xm1, m0, 1
-    lea             r5, [r3 * 3]
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    movd            [r2 + r3 * 2], xm1
-    pextrd          [r2 + r5], xm1, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm0, 2
-    pextrd          [r2 + r3], xm0, 3
-    pextrd          [r2 + r3 * 2], xm1, 2
-    pextrd          [r2 + r5], xm1, 3
-%else
-    add             r3d, r3d
-    psubw           m0, [pw_2000]
-    psubw           m4, [pw_2000]
-    vextracti128    xm1, m0, 1
-    vextracti128    xm2, m4, 1
-    lea             r5, [r3 * 3]
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm1
-    movhps          [r2 + r5], xm1
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm4
-    movhps          [r2 + r3], xm4
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r5], xm2
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_4x8 pp
-    FILTER_VER_CHROMA_AVX2_4x8 ps
-
-%macro FILTER_VER_CHROMA_AVX2_4xN 2
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x%2, 4, 6, 12
-    mov             r4d, r4m
-    shl             r4d, 6
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    mova            m10, [r5]
-    mova            m11, [r5 + mmsize]
-%ifidn %1,pp
-    mova            m9, [pw_512]
-%else
-    add             r3d, r3d
-    mova            m9, [pw_2000]
-%endif
-    lea             r5, [r3 * 3]
-%rep %2 / 16
-    movd            xm1, [r0]
-    pinsrd          xm1, [r0 + r1], 1
-    pinsrd          xm1, [r0 + r1 * 2], 2
-    pinsrd          xm1, [r0 + r4], 3                       ; m1 = row[3 2 1 0]
-    lea             r0, [r0 + r1 * 4]
-    movd            xm2, [r0]
-    pinsrd          xm2, [r0 + r1], 1
-    pinsrd          xm2, [r0 + r1 * 2], 2
-    pinsrd          xm2, [r0 + r4], 3                       ; m2 = row[7 6 5 4]
-    vinserti128     m1, m1, xm2, 1                          ; m1 = row[7 6 5 4 3 2 1 0]
-    lea             r0, [r0 + r1 * 4]
-    movd            xm3, [r0]
-    pinsrd          xm3, [r0 + r1], 1
-    pinsrd          xm3, [r0 + r1 * 2], 2
-    pinsrd          xm3, [r0 + r4], 3                       ; m3 = row[11 10 9 8]
-    vinserti128     m2, m2, xm3, 1                          ; m2 = row[11 10 9 8 7 6 5 4]
-    lea             r0, [r0 + r1 * 4]
-    movd            xm4, [r0]
-    pinsrd          xm4, [r0 + r1], 1
-    pinsrd          xm4, [r0 + r1 * 2], 2
-    pinsrd          xm4, [r0 + r4], 3                       ; m4 = row[15 14 13 12]
-    vinserti128     m3, m3, xm4, 1                          ; m3 = row[15 14 13 12 11 10 9 8]
-    lea             r0, [r0 + r1 * 4]
-    movd            xm5, [r0]
-    pinsrd          xm5, [r0 + r1], 1
-    pinsrd          xm5, [r0 + r1 * 2], 2                   ; m5 = row[x 18 17 16]
-    vinserti128     m4, m4, xm5, 1                          ; m4 = row[x 18 17 16 15 14 13 12]
-    mova            m5, [interp4_vpp_shuf1]
-    vpermd          m0, m5, m1                              ; m0 = row[4 3 3 2 2 1 1 0]
-    vpermd          m6, m5, m2                              ; m6 = row[8 7 7 6 6 5 5 4]
-    vpermd          m7, m5, m3                              ; m7 = row[12 11 11 10 10 9 9 8]
-    vpermd          m8, m5, m4                              ; m8 = row[16 15 15 14 14 13 13 12]
-    mova            m5, [interp4_vpp_shuf1 + mmsize]
-    vpermd          m1, m5, m1                              ; m1 = row[6 5 5 4 4 3 3 2]
-    vpermd          m2, m5, m2                              ; m2 = row[10 9 9 8 8 7 7 6]
-    vpermd          m3, m5, m3                              ; m3 = row[14 13 13 12 12 11 11 10]
-    vpermd          m4, m5, m4                              ; m4 = row[18 17 17 16 16 15 15 14]
-
-    mova            m5, [interp4_vpp_shuf]
-    pshufb          m0, m0, m5
-    pshufb          m1, m1, m5
-    pshufb          m2, m2, m5
-    pshufb          m4, m4, m5
-    pshufb          m3, m3, m5
-    pshufb          m6, m6, m5
-    pshufb          m7, m7, m5
-    pshufb          m8, m8, m5
-    pmaddubsw       m0, m10
-    pmaddubsw       m6, m10
-    pmaddubsw       m7, m10
-    pmaddubsw       m8, m10
-    pmaddubsw       m1, m11
-    pmaddubsw       m2, m11
-    pmaddubsw       m3, m11
-    pmaddubsw       m4, m11
-    paddw           m0, m1                                  ; m0 = WORD ROW[3 2 1 0]
-    paddw           m6, m2                                  ; m6 = WORD ROW[7 6 5 4]
-    paddw           m7, m3                                  ; m7 = WORD ROW[11 10 9 8]
-    paddw           m8, m4                                  ; m8 = WORD ROW[15 14 13 12]
-%ifidn %1,pp
-    pmulhrsw        m0, m9
-    pmulhrsw        m6, m9
-    pmulhrsw        m7, m9
-    pmulhrsw        m8, m9
-    packuswb        m0, m6
-    packuswb        m7, m8
-    vextracti128    xm1, m0, 1
-    vextracti128    xm2, m7, 1
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    movd            [r2 + r3 * 2], xm1
-    pextrd          [r2 + r5], xm1, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm0, 2
-    pextrd          [r2 + r3], xm0, 3
-    pextrd          [r2 + r3 * 2], xm1, 2
-    pextrd          [r2 + r5], xm1, 3
-    lea             r2, [r2 + r3 * 4]
-    movd            [r2], xm7
-    pextrd          [r2 + r3], xm7, 1
-    movd            [r2 + r3 * 2], xm2
-    pextrd          [r2 + r5], xm2, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm7, 2
-    pextrd          [r2 + r3], xm7, 3
-    pextrd          [r2 + r3 * 2], xm2, 2
-    pextrd          [r2 + r5], xm2, 3
-%else
-    psubw           m0, m9
-    psubw           m6, m9
-    psubw           m7, m9
-    psubw           m8, m9
-    vextracti128    xm1, m0, 1
-    vextracti128    xm2, m6, 1
-    vextracti128    xm3, m7, 1
-    vextracti128    xm4, m8, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm1
-    movhps          [r2 + r5], xm1
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm6
-    movhps          [r2 + r3], xm6
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r5], xm2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm7
-    movhps          [r2 + r3], xm7
-    movq            [r2 + r3 * 2], xm3
-    movhps          [r2 + r5], xm3
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm8
-    movhps          [r2 + r3], xm8
-    movq            [r2 + r3 * 2], xm4
-    movhps          [r2 + r5], xm4
-%endif
-    lea             r2, [r2 + r3 * 4]
-%endrep
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_4xN pp, 16
-    FILTER_VER_CHROMA_AVX2_4xN ps, 16
-    FILTER_VER_CHROMA_AVX2_4xN pp, 32
-    FILTER_VER_CHROMA_AVX2_4xN ps, 32
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W4_H4 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m0,        [tab_Cm]
-
-    mova        m1,        [pw_512]
-
-    mov         r4d,       %2
-
-    lea         r5,        [3 * r1]
-
-.loop:
-    movd        m2,        [r0]
-    movd        m3,        [r0 + r1]
-    movd        m4,        [r0 + 2 * r1]
-    movd        m5,        [r0 + r5]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m6,        m4,        m5
-    punpcklbw   m2,        m6
-
-    pmaddubsw   m2,        m0
-
-    lea         r0,        [r0 + 4 * r1]
-    movd        m6,        [r0]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m7,        m5,        m6
-    punpcklbw   m3,        m7
-
-    pmaddubsw   m3,        m0
-
-    phaddw      m2,        m3
-
-    pmulhrsw    m2,        m1
-
-    movd        m7,        [r0 + r1]
-
-    punpcklbw   m4,        m5
-    punpcklbw   m3,        m6,        m7
-    punpcklbw   m4,        m3
-
-    pmaddubsw   m4,        m0
-
-    movd        m3,        [r0 + 2 * r1]
-
-    punpcklbw   m5,        m6
-    punpcklbw   m7,        m3
-    punpcklbw   m5,        m7
-
-    pmaddubsw   m5,        m0
-
-    phaddw      m4,        m5
-
-    pmulhrsw    m4,        m1
-    packuswb    m2,        m4
-    movd        [r2],      m2
-    pextrd      [r2 + r3], m2,  1
-    lea         r2,        [r2 + 2 * r3]
-    pextrd      [r2],      m2, 2
-    pextrd      [r2 + r3], m2, 3
-
-    lea         r2,        [r2 + 2 * r3]
-
-    sub         r4,        4
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W4_H4 4,  8
-    FILTER_V4_W4_H4 4, 16
-
-    FILTER_V4_W4_H4 4, 32
-
-%macro FILTER_V4_W8_H2 0
-    punpcklbw   m1,        m2
-    punpcklbw   m7,        m3,        m0
-
-    pmaddubsw   m1,        m6
-    pmaddubsw   m7,        m5
-
-    paddw       m1,        m7
-
-    pmulhrsw    m1,        m4
-    packuswb    m1,        m1
-%endmacro
-
-%macro FILTER_V4_W8_H3 0
-    punpcklbw   m2,        m3
-    punpcklbw   m7,        m0,        m1
-
-    pmaddubsw   m2,        m6
-    pmaddubsw   m7,        m5
-
-    paddw       m2,        m7
-
-    pmulhrsw    m2,        m4
-    packuswb    m2,        m2
-%endmacro
-
-%macro FILTER_V4_W8_H4 0
-    punpcklbw   m3,        m0
-    punpcklbw   m7,        m1,        m2
-
-    pmaddubsw   m3,        m6
-    pmaddubsw   m7,        m5
-
-    paddw       m3,        m7
-
-    pmulhrsw    m3,        m4
-    packuswb    m3,        m3
-%endmacro
-
-%macro FILTER_V4_W8_H5 0
-    punpcklbw   m0,        m1
-    punpcklbw   m7,        m2,        m3
-
-    pmaddubsw   m0,        m6
-    pmaddubsw   m7,        m5
-
-    paddw       m0,        m7
-
-    pmulhrsw    m0,        m4
-    packuswb    m0,        m0
-%endmacro
-
-%macro FILTER_V4_W8_8x2 2
-    FILTER_V4_W8 %1, %2
-    movq        m0,        [r0 + 4 * r1]
-
-    FILTER_V4_W8_H2
-
-    movh        [r2 + r3], m1
-%endmacro
-
-%macro FILTER_V4_W8_8x4 2
-    FILTER_V4_W8_8x2 %1, %2
-;8x3
-    lea         r6,        [r0 + 4 * r1]
-    movq        m1,        [r6 + r1]
-
-    FILTER_V4_W8_H3
-
-    movh        [r2 + 2 * r3], m2
-
-;8x4
-    movq        m2,        [r6 + 2 * r1]
-
-    FILTER_V4_W8_H4
-
-    lea         r5,        [r2 + 2 * r3]
-    movh        [r5 + r3], m3
-%endmacro
-
-%macro FILTER_V4_W8_8x6 2
-    FILTER_V4_W8_8x4 %1, %2
-;8x5
-    lea         r6,        [r6 + 2 * r1]
-    movq        m3,        [r6 + r1]
-
-    FILTER_V4_W8_H5
-
-    movh        [r2 + 4 * r3], m0
-
-;8x6
-    movq        m0,        [r0 + 8 * r1]
-
-    FILTER_V4_W8_H2
-
-    lea         r5,        [r2 + 4 * r3]
-    movh        [r5 + r3], m1
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W8 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_%1x%2, 4, 7, 8
-
-    mov         r4d,       r4m
-
-    sub         r0,        r1
-    movq        m0,        [r0]
-    movq        m1,        [r0 + r1]
-    movq        m2,        [r0 + 2 * r1]
-    lea         r5,        [r0 + 2 * r1]
-    movq        m3,        [r5 + r1]
-
-    punpcklbw   m0,        m1
-    punpcklbw   m4,        m2,          m3
-
-%ifdef PIC
-    lea         r6,        [tab_ChromaCoeff]
-    movd        m5,        [r6 + r4 * 4]
-%else
-    movd        m5,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m6,        m5,       [tab_Vm]
-    pmaddubsw   m0,        m6
-
-    pshufb      m5,        [tab_Vm + 16]
-    pmaddubsw   m4,        m5
-
-    paddw       m0,        m4
-
-    mova        m4,        [pw_512]
-
-    pmulhrsw    m0,        m4
-    packuswb    m0,        m0
-    movh        [r2],      m0
-%endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_8x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-    FILTER_V4_W8_8x2 8, 2
-
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-    FILTER_V4_W8_8x4 8, 4
-
-    RET
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_8x6(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-    FILTER_V4_W8_8x6 8, 6
-
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_4x2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-------------------------------------------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_4x2, 4, 6, 6
-
-    mov         r4d, r4m
-    sub         r0, r1
-    add         r3d, r3d
-
-%ifdef PIC
-    lea         r5, [tab_ChromaCoeff]
-    movd        m0, [r5 + r4 * 4]
-%else
-    movd        m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m0, [tab_Cm]
-
-    movd        m2, [r0]
-    movd        m3, [r0 + r1]
-    lea         r5, [r0 + 2 * r1]
-    movd        m4, [r5]
-    movd        m5, [r5 + r1]
-
-    punpcklbw   m2, m3
-    punpcklbw   m1, m4, m5
-    punpcklbw   m2, m1
-
-    pmaddubsw   m2, m0
-
-    movd        m1, [r0 + 4 * r1]
-
-    punpcklbw   m3, m4
-    punpcklbw   m5, m1
-    punpcklbw   m3, m5
-
-    pmaddubsw   m3, m0
-
-    phaddw      m2, m3
-
-    psubw       m2, [pw_2000]
-    movh        [r2], m2
-    movhps      [r2 + r3], m2
-
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_4x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-------------------------------------------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_4x4, 4, 6, 7
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m0, [r5 + r4 * 4]
-%else
-    movd       m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m0, [tab_Cm]
-
-    lea        r4, [r1 * 3]
-    lea        r5, [r0 + 4 * r1]
-
-    movd       m2, [r0]
-    movd       m3, [r0 + r1]
-    movd       m4, [r0 + 2 * r1]
-    movd       m5, [r0 + r4]
-
-    punpcklbw  m2, m3
-    punpcklbw  m6, m4, m5
-    punpcklbw  m2, m6
-
-    pmaddubsw  m2, m0
-
-    movd       m6, [r5]
-
-    punpcklbw  m3, m4
-    punpcklbw  m1, m5, m6
-    punpcklbw  m3, m1
-
-    pmaddubsw  m3, m0
-
-    phaddw     m2, m3
-
-    mova       m1, [pw_2000]
-
-    psubw      m2, m1
-    movh       [r2], m2
-    movhps     [r2 + r3], m2
-
-    movd       m2, [r5 + r1]
-
-    punpcklbw  m4, m5
-    punpcklbw  m3, m6, m2
-    punpcklbw  m4, m3
-
-    pmaddubsw  m4, m0
-
-    movd       m3, [r5 + 2 * r1]
-
-    punpcklbw  m5, m6
-    punpcklbw  m2, m3
-    punpcklbw  m5, m2
-
-    pmaddubsw  m5, m0
-
-    phaddw     m4, m5
-
-    psubw      m4, m1
-    lea        r2, [r2 + 2 * r3]
-    movh       [r2], m4
-    movhps     [r2 + r3], m4
-
-    RET
-
-;---------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W4_H4 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m0, [r5 + r4 * 4]
-%else
-    movd       m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m0, [tab_Cm]
-
-    mova       m1, [pw_2000]
-
-    mov        r4d, %2/4
-    lea        r5, [3 * r1]
-
-.loop:
-    movd       m2, [r0]
-    movd       m3, [r0 + r1]
-    movd       m4, [r0 + 2 * r1]
-    movd       m5, [r0 + r5]
-
-    punpcklbw  m2, m3
-    punpcklbw  m6, m4, m5
-    punpcklbw  m2, m6
-
-    pmaddubsw  m2, m0
-
-    lea        r0, [r0 + 4 * r1]
-    movd       m6, [r0]
-
-    punpcklbw  m3, m4
-    punpcklbw  m7, m5, m6
-    punpcklbw  m3, m7
-
-    pmaddubsw  m3, m0
-
-    phaddw     m2, m3
-
-    psubw      m2, m1
-    movh       [r2], m2
-    movhps     [r2 + r3], m2
-
-    movd       m2, [r0 + r1]
-
-    punpcklbw  m4, m5
-    punpcklbw  m3, m6, m2
-    punpcklbw  m4, m3
-
-    pmaddubsw  m4, m0
-
-    movd       m3, [r0 + 2 * r1]
-
-    punpcklbw  m5, m6
-    punpcklbw  m2, m3
-    punpcklbw  m5, m2
-
-    pmaddubsw  m5, m0
-
-    phaddw     m4, m5
-
-    psubw      m4, m1
-    lea        r2, [r2 + 2 * r3]
-    movh       [r2], m4
-    movhps     [r2 + r3], m4
-
-    lea        r2, [r2 + 2 * r3]
-
-    dec        r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V_PS_W4_H4 4, 8
-    FILTER_V_PS_W4_H4 4, 16
-
-    FILTER_V_PS_W4_H4 4, 32
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W8_H8_H16_H2 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 7
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m5, [r5 + r4 * 4]
-%else
-    movd       m5, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m6, m5, [tab_Vm]
-    pshufb     m5, [tab_Vm + 16]
-    mova       m4, [pw_2000]
-
-    mov        r4d, %2/2
-    lea        r5, [3 * r1]
-
-.loopH:
-    movq       m0, [r0]
-    movq       m1, [r0 + r1]
-    movq       m2, [r0 + 2 * r1]
-    movq       m3, [r0 + r5]
-
-    punpcklbw  m0, m1
-    punpcklbw  m1, m2
-    punpcklbw  m2, m3
-
-    pmaddubsw  m0, m6
-    pmaddubsw  m2, m5
-
-    paddw      m0, m2
-
-    psubw      m0, m4
-    movu       [r2], m0
-
-    movq       m0, [r0 + 4 * r1]
-
-    punpcklbw  m3, m0
-
-    pmaddubsw  m1, m6
-    pmaddubsw  m3, m5
-
-    paddw      m1, m3
-    psubw      m1, m4
-
-    movu       [r2 + r3], m1
-
-    lea        r0, [r0 + 2 * r1]
-    lea        r2, [r2 + 2 * r3]
-
-    dec        r4d
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_V_PS_W8_H8_H16_H2 8, 2
-    FILTER_V_PS_W8_H8_H16_H2 8, 4
-    FILTER_V_PS_W8_H8_H16_H2 8, 6
-
-    FILTER_V_PS_W8_H8_H16_H2 8, 12
-    FILTER_V_PS_W8_H8_H16_H2 8, 64
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W8_H8_H16_H32 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m5, [r5 + r4 * 4]
-%else
-    movd       m5, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m6, m5, [tab_Vm]
-    pshufb     m5, [tab_Vm + 16]
-    mova       m4, [pw_2000]
-
-    mov        r4d, %2/4
-    lea        r5, [3 * r1]
-
-.loop:
-    movq       m0, [r0]
-    movq       m1, [r0 + r1]
-    movq       m2, [r0 + 2 * r1]
-    movq       m3, [r0 + r5]
-
-    punpcklbw  m0, m1
-    punpcklbw  m1, m2
-    punpcklbw  m2, m3
-
-    pmaddubsw  m0, m6
-    pmaddubsw  m7, m2, m5
-
-    paddw      m0, m7
-
-    psubw       m0, m4
-    movu       [r2], m0
-
-    lea        r0, [r0 + 4 * r1]
-    movq       m0, [r0]
-
-    punpcklbw  m3, m0
-
-    pmaddubsw  m1, m6
-    pmaddubsw  m7, m3, m5
-
-    paddw      m1, m7
-
-    psubw      m1, m4
-    movu       [r2 + r3], m1
-
-    movq       m1, [r0 + r1]
-
-    punpcklbw  m0, m1
-
-    pmaddubsw  m2, m6
-    pmaddubsw  m0, m5
-
-    paddw      m2, m0
-
-    psubw      m2, m4
-    lea        r2, [r2 + 2 * r3]
-    movu       [r2], m2
-
-    movq       m2, [r0 + 2 * r1]
-
-    punpcklbw  m1, m2
-
-    pmaddubsw  m3, m6
-    pmaddubsw  m1, m5
-
-    paddw      m3, m1
-    psubw      m3, m4
-
-    movu       [r2 + r3], m3
-
-    lea        r2, [r2 + 2 * r3]
-
-    dec        r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V_PS_W8_H8_H16_H32 8,  8
-    FILTER_V_PS_W8_H8_H16_H32 8, 16
-    FILTER_V_PS_W8_H8_H16_H32 8, 32
-
-;------------------------------------------------------------------------------------------------------------
-;void interp_4tap_vert_ps_6x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W6 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_6x%2, 4, 6, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m5, [r5 + r4 * 4]
-%else
-    movd       m5, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m6, m5, [tab_Vm]
-    pshufb     m5, [tab_Vm + 16]
-    mova       m4, [pw_2000]
-    lea        r5, [3 * r1]
-    mov        r4d, %2/4
-
-.loop:
-    movq       m0, [r0]
-    movq       m1, [r0 + r1]
-    movq       m2, [r0 + 2 * r1]
-    movq       m3, [r0 + r5]
-
-    punpcklbw  m0, m1
-    punpcklbw  m1, m2
-    punpcklbw  m2, m3
-
-    pmaddubsw  m0, m6
-    pmaddubsw  m7, m2, m5
-
-    paddw      m0, m7
-    psubw      m0, m4
-
-    movh       [r2], m0
-    pshufd     m0, m0, 2
-    movd       [r2 + 8], m0
-
-    lea        r0, [r0 + 4 * r1]
-    movq       m0, [r0]
-    punpcklbw  m3, m0
-
-    pmaddubsw  m1, m6
-    pmaddubsw  m7, m3, m5
-
-    paddw      m1, m7
-    psubw      m1, m4
-
-    movh       [r2 + r3], m1
-    pshufd     m1, m1, 2
-    movd       [r2 + r3 + 8], m1
-
-    movq       m1, [r0 + r1]
-    punpcklbw  m0, m1
-
-    pmaddubsw  m2, m6
-    pmaddubsw  m0, m5
-
-    paddw      m2, m0
-    psubw      m2, m4
-
-    lea        r2,[r2 + 2 * r3]
-    movh       [r2], m2
-    pshufd     m2, m2, 2
-    movd       [r2 + 8], m2
-
-    movq       m2,[r0 + 2 * r1]
-    punpcklbw  m1, m2
-
-    pmaddubsw  m3, m6
-    pmaddubsw  m1, m5
-
-    paddw      m3, m1
-    psubw      m3, m4
-
-    movh       [r2 + r3], m3
-    pshufd     m3, m3, 2
-    movd       [r2 + r3 + 8], m3
-
-    lea        r2, [r2 + 2 * r3]
-
-    dec        r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V_PS_W6 6, 8
-    FILTER_V_PS_W6 6, 16
-
-;---------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_12x16(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W12 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_12x%2, 4, 6, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m0, [r5 + r4 * 4]
-%else
-    movd       m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m1, m0, [tab_Vm]
-    pshufb     m0, [tab_Vm + 16]
-
-    mov        r4d, %2/2
-
-.loop:
-    movu       m2, [r0]
-    movu       m3, [r0 + r1]
-
-    punpcklbw  m4, m2, m3
-    punpckhbw  m2, m3
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m2, m1
-
-    lea        r0, [r0 + 2 * r1]
-    movu       m5, [r0]
-    movu       m7, [r0 + r1]
-
-    punpcklbw  m6, m5, m7
-    pmaddubsw  m6, m0
-    paddw      m4, m6
-
-    punpckhbw  m6, m5, m7
-    pmaddubsw  m6, m0
-    paddw      m2, m6
-
-    mova       m6, [pw_2000]
-
-    psubw      m4, m6
-    psubw      m2, m6
-
-    movu       [r2], m4
-    movh       [r2 + 16], m2
-
-    punpcklbw  m4, m3, m5
-    punpckhbw  m3, m5
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m3, m1
-
-    movu       m2, [r0 + 2 * r1]
-
-    punpcklbw  m5, m7, m2
-    punpckhbw  m7, m2
-
-    pmaddubsw  m5, m0
-    pmaddubsw  m7, m0
-
-    paddw      m4, m5
-    paddw      m3, m7
-
-    psubw      m4, m6
-    psubw      m3, m6
-
-    movu       [r2 + r3], m4
-    movh       [r2 + r3 + 16], m3
-
-    lea        r2, [r2 + 2 * r3]
-
-    dec        r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V_PS_W12 12, 16
-    FILTER_V_PS_W12 12, 32
-
-;---------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_16x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W16 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m0, [r5 + r4 * 4]
-%else
-    movd       m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m1, m0, [tab_Vm]
-    pshufb     m0, [tab_Vm + 16]
-    mov        r4d, %2/2
-
-.loop:
-    movu       m2, [r0]
-    movu       m3, [r0 + r1]
-
-    punpcklbw  m4, m2, m3
-    punpckhbw  m2, m3
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m2, m1
-
-    lea        r0, [r0 + 2 * r1]
-    movu       m5, [r0]
-    movu       m7, [r0 + r1]
-
-    punpcklbw  m6, m5, m7
-    pmaddubsw  m6, m0
-    paddw      m4, m6
-
-    punpckhbw  m6, m5, m7
-    pmaddubsw  m6, m0
-    paddw      m2, m6
-
-    mova       m6, [pw_2000]
-
-    psubw      m4, m6
-    psubw      m2, m6
-
-    movu       [r2], m4
-    movu       [r2 + 16], m2
-
-    punpcklbw  m4, m3, m5
-    punpckhbw  m3, m5
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m3, m1
-
-    movu       m5, [r0 + 2 * r1]
-
-    punpcklbw  m2, m7, m5
-    punpckhbw  m7, m5
-
-    pmaddubsw  m2, m0
-    pmaddubsw  m7, m0
-
-    paddw      m4, m2
-    paddw      m3, m7
-
-    psubw      m4, m6
-    psubw      m3, m6
-
-    movu       [r2 + r3], m4
-    movu       [r2 + r3 + 16], m3
-
-    lea        r2, [r2 + 2 * r3]
-
-    dec        r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V_PS_W16 16,  4
-    FILTER_V_PS_W16 16,  8
-    FILTER_V_PS_W16 16, 12
-    FILTER_V_PS_W16 16, 16
-    FILTER_V_PS_W16 16, 32
-
-    FILTER_V_PS_W16 16, 24
-    FILTER_V_PS_W16 16, 64
-
-;--------------------------------------------------------------------------------------------------------------
-;void interp_4tap_vert_ps_24x32(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro FILTER_V4_PS_W24 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_24x%2, 4, 6, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m0, [r5 + r4 * 4]
-%else
-    movd       m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m1, m0, [tab_Vm]
-    pshufb     m0, [tab_Vm + 16]
-
-    mov        r4d, %2/2
-
-.loop:
-    movu       m2, [r0]
-    movu       m3, [r0 + r1]
-
-    punpcklbw  m4, m2, m3
-    punpckhbw  m2, m3
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m2, m1
-
-    lea        r5, [r0 + 2 * r1]
-
-    movu       m5, [r5]
-    movu       m7, [r5 + r1]
-
-    punpcklbw  m6, m5, m7
-    pmaddubsw  m6, m0
-    paddw      m4, m6
-
-    punpckhbw  m6, m5, m7
-    pmaddubsw  m6, m0
-    paddw      m2, m6
-
-    mova       m6, [pw_2000]
-
-    psubw      m4, m6
-    psubw      m2, m6
-
-    movu       [r2], m4
-    movu       [r2 + 16], m2
-
-    punpcklbw  m4, m3, m5
-    punpckhbw  m3, m5
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m3, m1
-
-    movu       m2, [r5 + 2 * r1]
-
-    punpcklbw  m5, m7, m2
-    punpckhbw  m7, m2
-
-    pmaddubsw  m5, m0
-    pmaddubsw  m7, m0
-
-    paddw      m4, m5
-    paddw      m3, m7
-
-    psubw      m4, m6
-    psubw      m3, m6
-
-    movu       [r2 + r3], m4
-    movu       [r2 + r3 + 16], m3
-
-    movq       m2, [r0 + 16]
-    movq       m3, [r0 + r1 + 16]
-    movq       m4, [r5 + 16]
-    movq       m5, [r5 + r1 + 16]
-
-    punpcklbw  m2, m3
-    punpcklbw  m7, m4, m5
-
-    pmaddubsw  m2, m1
-    pmaddubsw  m7, m0
-
-    paddw      m2, m7
-    psubw      m2, m6
-
-    movu       [r2 + 32], m2
-
-    movq       m2, [r5 + 2 * r1 + 16]
-
-    punpcklbw  m3, m4
-    punpcklbw  m5, m2
-
-    pmaddubsw  m3, m1
-    pmaddubsw  m5, m0
-
-    paddw      m3, m5
-    psubw      m3,  m6
-
-    movu       [r2 + r3 + 32], m3
-
-    mov        r0, r5
-    lea        r2, [r2 + 2 * r3]
-
-    dec        r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_PS_W24 24, 32
-
-    FILTER_V4_PS_W24 24, 64
-
-;---------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_32x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W32 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m0, [r5 + r4 * 4]
-%else
-    movd       m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m1, m0, [tab_Vm]
-    pshufb     m0, [tab_Vm + 16]
-
-    mova       m7, [pw_2000]
-
-    mov        r4d, %2
-
-.loop:
-    movu       m2, [r0]
-    movu       m3, [r0 + r1]
-
-    punpcklbw  m4, m2, m3
-    punpckhbw  m2, m3
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m2, m1
-
-    lea        r5, [r0 + 2 * r1]
-    movu       m3, [r5]
-    movu       m5, [r5 + r1]
-
-    punpcklbw  m6, m3, m5
-    punpckhbw  m3, m5
-
-    pmaddubsw  m6, m0
-    pmaddubsw  m3, m0
-
-    paddw      m4, m6
-    paddw      m2, m3
-
-    psubw      m4, m7
-    psubw      m2, m7
-
-    movu       [r2], m4
-    movu       [r2 + 16], m2
-
-    movu       m2, [r0 + 16]
-    movu       m3, [r0 + r1 + 16]
-
-    punpcklbw  m4, m2, m3
-    punpckhbw  m2, m3
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m2, m1
-
-    movu       m3, [r5 + 16]
-    movu       m5, [r5 + r1 + 16]
-
-    punpcklbw  m6, m3, m5
-    punpckhbw  m3, m5
-
-    pmaddubsw  m6, m0
-    pmaddubsw  m3, m0
-
-    paddw      m4, m6
-    paddw      m2, m3
-
-    psubw      m4, m7
-    psubw      m2, m7
-
-    movu       [r2 + 32], m4
-    movu       [r2 + 48], m2
-
-    lea        r0, [r0 + r1]
-    lea        r2, [r2 + r3]
-
-    dec        r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V_PS_W32 32,  8
-    FILTER_V_PS_W32 32, 16
-    FILTER_V_PS_W32 32, 24
-    FILTER_V_PS_W32 32, 32
-
-    FILTER_V_PS_W32 32, 48
-    FILTER_V_PS_W32 32, 64
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W8_H8_H16_H32 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m5,        [r5 + r4 * 4]
-%else
-    movd        m5,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m6,        m5,       [tab_Vm]
-    pshufb      m5,        [tab_Vm + 16]
-    mova        m4,        [pw_512]
-    lea         r5,        [r1 * 3]
-
-    mov         r4d,       %2
-
-.loop:
-    movq        m0,        [r0]
-    movq        m1,        [r0 + r1]
-    movq        m2,        [r0 + 2 * r1]
-    movq        m3,        [r0 + r5]
-
-    punpcklbw   m0,        m1
-    punpcklbw   m1,        m2
-    punpcklbw   m2,        m3
-
-    pmaddubsw   m0,        m6
-    pmaddubsw   m7,        m2, m5
-
-    paddw       m0,        m7
-
-    pmulhrsw    m0,        m4
-    packuswb    m0,        m0
-    movh        [r2],      m0
-
-    lea         r0,        [r0 + 4 * r1]
-    movq        m0,        [r0]
-
-    punpcklbw   m3,        m0
-
-    pmaddubsw   m1,        m6
-    pmaddubsw   m7,        m3, m5
-
-    paddw       m1,        m7
-
-    pmulhrsw    m1,        m4
-    packuswb    m1,        m1
-    movh        [r2 + r3], m1
-
-    movq        m1,        [r0 + r1]
-
-    punpcklbw   m0,        m1
-
-    pmaddubsw   m2,        m6
-    pmaddubsw   m0,        m5
-
-    paddw       m2,        m0
-
-    pmulhrsw    m2,        m4
-
-    movq        m7,        [r0 + 2 * r1]
-    punpcklbw   m1,        m7
-
-    pmaddubsw   m3,        m6
-    pmaddubsw   m1,        m5
-
-    paddw       m3,        m1
-
-    pmulhrsw    m3,        m4
-    packuswb    m2,        m3
-
-    lea         r2,        [r2 + 2 * r3]
-    movh        [r2],      m2
-    movhps      [r2 + r3], m2
-
-    lea         r2,        [r2 + 2 * r3]
-
-    sub         r4,         4
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W8_H8_H16_H32 8,  8
-    FILTER_V4_W8_H8_H16_H32 8, 16
-    FILTER_V4_W8_H8_H16_H32 8, 32
-
-    FILTER_V4_W8_H8_H16_H32 8, 12
-    FILTER_V4_W8_H8_H16_H32 8, 64
-
-%macro PROCESS_CHROMA_AVX2_W8_8R 0
-    movq            xm1, [r0]                       ; m1 = row 0
-    movq            xm2, [r0 + r1]                  ; m2 = row 1
-    punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
-    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
-    punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
-    vinserti128     m5, m1, xm2, 1                  ; m5 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
-    pmaddubsw       m5, [r5]
-    movq            xm4, [r0 + r4]                  ; m4 = row 3
-    punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
-    lea             r0, [r0 + r1 * 4]
-    movq            xm1, [r0]                       ; m1 = row 4
-    punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
-    vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
-    pmaddubsw       m0, m2, [r5 + 1 * mmsize]
-    paddw           m5, m0
-    pmaddubsw       m2, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 5
-    punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
-    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
-    punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
-    vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
-    pmaddubsw       m0, m1, [r5 + 1 * mmsize]
-    paddw           m2, m0
-    pmaddubsw       m1, [r5]
-    movq            xm3, [r0 + r4]                  ; m3 = row 7
-    punpcklbw       xm4, xm3                        ; m4 = [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
-    lea             r0, [r0 + r1 * 4]
-    movq            xm0, [r0]                       ; m0 = row 8
-    punpcklbw       xm3, xm0                        ; m3 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70]
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70] - [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
-    pmaddubsw       m3, m4, [r5 + 1 * mmsize]
-    paddw           m1, m3
-    pmaddubsw       m4, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 9
-    punpcklbw       xm0, xm3                        ; m0 = [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
-    movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
-    punpcklbw       xm3, xm6                        ; m3 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90]
-    vinserti128     m0, m0, xm3, 1                  ; m0 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90] - [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
-    pmaddubsw       m0, [r5 + 1 * mmsize]
-    paddw           m4, m0
-%endmacro
-
-%macro FILTER_VER_CHROMA_AVX2_8x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x8, 4, 6, 7
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-    PROCESS_CHROMA_AVX2_W8_8R
-%ifidn %1,pp
-    lea             r4, [r3 * 3]
-    mova            m3, [pw_512]
-    pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
-    pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
-    pmulhrsw        m1, m3                          ; m1 = word: row 4, row 5
-    pmulhrsw        m4, m3                          ; m4 = word: row 6, row 7
-    packuswb        m5, m2
-    packuswb        m1, m4
-    vextracti128    xm2, m5, 1
-    vextracti128    xm4, m1, 1
-    movq            [r2], xm5
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm5
-    movhps          [r2 + r4], xm2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm1
-    movq            [r2 + r3], xm4
-    movhps          [r2 + r3 * 2], xm1
-    movhps          [r2 + r4], xm4
-%else
-    add             r3d, r3d
-    vbroadcasti128  m3, [pw_2000]
-    lea             r4, [r3 * 3]
-    psubw           m5, m3                          ; m5 = word: row 0, row 1
-    psubw           m2, m3                          ; m2 = word: row 2, row 3
-    psubw           m1, m3                          ; m1 = word: row 4, row 5
-    psubw           m4, m3                          ; m4 = word: row 6, row 7
-    vextracti128    xm6, m5, 1
-    vextracti128    xm3, m2, 1
-    vextracti128    xm0, m1, 1
-    movu            [r2], xm5
-    movu            [r2 + r3], xm6
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r4], xm3
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm1
-    movu            [r2 + r3], xm0
-    movu            [r2 + r3 * 2], xm4
-    vextracti128    xm4, m4, 1
-    movu            [r2 + r4], xm4
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_8x8 pp
-    FILTER_VER_CHROMA_AVX2_8x8 ps
-
-%macro FILTER_VER_CHROMA_AVX2_8x6 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x6, 4, 6, 6
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-
-    movq            xm1, [r0]                       ; m1 = row 0
-    movq            xm2, [r0 + r1]                  ; m2 = row 1
-    punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
-    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
-    punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
-    vinserti128     m5, m1, xm2, 1                  ; m5 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
-    pmaddubsw       m5, [r5]
-    movq            xm4, [r0 + r4]                  ; m4 = row 3
-    punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
-    lea             r0, [r0 + r1 * 4]
-    movq            xm1, [r0]                       ; m1 = row 4
-    punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
-    vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
-    pmaddubsw       m0, m2, [r5 + 1 * mmsize]
-    paddw           m5, m0
-    pmaddubsw       m2, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 5
-    punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
-    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
-    punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
-    vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
-    pmaddubsw       m0, m1, [r5 + 1 * mmsize]
-    paddw           m2, m0
-    pmaddubsw       m1, [r5]
-    movq            xm3, [r0 + r4]                  ; m3 = row 7
-    punpcklbw       xm4, xm3                        ; m4 = [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
-    lea             r0, [r0 + r1 * 4]
-    movq            xm0, [r0]                       ; m0 = row 8
-    punpcklbw       xm3, xm0                        ; m3 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70]
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70] - [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
-    pmaddubsw       m4, [r5 + 1 * mmsize]
-    paddw           m1, m4
-%ifidn %1,pp
-    lea             r4, [r3 * 3]
-    mova            m3, [pw_512]
-    pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
-    pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
-    pmulhrsw        m1, m3                          ; m1 = word: row 4, row 5
-    packuswb        m5, m2
-    packuswb        m1, m1
-    vextracti128    xm2, m5, 1
-    vextracti128    xm4, m1, 1
-    movq            [r2], xm5
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm5
-    movhps          [r2 + r4], xm2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm1
-    movq            [r2 + r3], xm4
-%else
-    add             r3d, r3d
-    mova            m3, [pw_2000]
-    lea             r4, [r3 * 3]
-    psubw           m5, m3                          ; m5 = word: row 0, row 1
-    psubw           m2, m3                          ; m2 = word: row 2, row 3
-    psubw           m1, m3                          ; m1 = word: row 4, row 5
-    vextracti128    xm4, m5, 1
-    vextracti128    xm3, m2, 1
-    vextracti128    xm0, m1, 1
-    movu            [r2], xm5
-    movu            [r2 + r3], xm4
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r4], xm3
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm1
-    movu            [r2 + r3], xm0
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_8x6 pp
-    FILTER_VER_CHROMA_AVX2_8x6 ps
-
-%macro PROCESS_CHROMA_AVX2_W8_16R 1
-    movq            xm1, [r0]                       ; m1 = row 0
-    movq            xm2, [r0 + r1]                  ; m2 = row 1
-    punpcklbw       xm1, xm2
-    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
-    punpcklbw       xm2, xm3
-    vinserti128     m5, m1, xm2, 1
-    pmaddubsw       m5, [r5]
-    movq            xm4, [r0 + r4]                  ; m4 = row 3
-    punpcklbw       xm3, xm4
-    lea             r0, [r0 + r1 * 4]
-    movq            xm1, [r0]                       ; m1 = row 4
-    punpcklbw       xm4, xm1
-    vinserti128     m2, m3, xm4, 1
-    pmaddubsw       m0, m2, [r5 + 1 * mmsize]
-    paddw           m5, m0
-    pmaddubsw       m2, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 5
-    punpcklbw       xm1, xm3
-    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
-    punpcklbw       xm3, xm4
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m0, m1, [r5 + 1 * mmsize]
-    paddw           m2, m0
-    pmaddubsw       m1, [r5]
-    movq            xm3, [r0 + r4]                  ; m3 = row 7
-    punpcklbw       xm4, xm3
-    lea             r0, [r0 + r1 * 4]
-    movq            xm0, [r0]                       ; m0 = row 8
-    punpcklbw       xm3, xm0
-    vinserti128     m4, m4, xm3, 1
-    pmaddubsw       m3, m4, [r5 + 1 * mmsize]
-    paddw           m1, m3
-    pmaddubsw       m4, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 9
-    punpcklbw       xm0, xm3
-    movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
-    punpcklbw       xm3, xm6
-    vinserti128     m0, m0, xm3, 1
-    pmaddubsw       m3, m0, [r5 + 1 * mmsize]
-    paddw           m4, m3
-    pmaddubsw       m0, [r5]
-%ifidn %1,pp
-    pmulhrsw        m5, m7                          ; m5 = word: row 0, row 1
-    pmulhrsw        m2, m7                          ; m2 = word: row 2, row 3
-    pmulhrsw        m1, m7                          ; m1 = word: row 4, row 5
-    pmulhrsw        m4, m7                          ; m4 = word: row 6, row 7
-    packuswb        m5, m2
-    packuswb        m1, m4
-    vextracti128    xm2, m5, 1
-    vextracti128    xm4, m1, 1
-    movq            [r2], xm5
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm5
-    movhps          [r2 + r6], xm2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm1
-    movq            [r2 + r3], xm4
-    movhps          [r2 + r3 * 2], xm1
-    movhps          [r2 + r6], xm4
-%else
-    psubw           m5, m7                          ; m5 = word: row 0, row 1
-    psubw           m2, m7                          ; m2 = word: row 2, row 3
-    psubw           m1, m7                          ; m1 = word: row 4, row 5
-    psubw           m4, m7                          ; m4 = word: row 6, row 7
-    vextracti128    xm3, m5, 1
-    movu            [r2], xm5
-    movu            [r2 + r3], xm3
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-    lea             r2, [r2 + r3 * 4]
-    vextracti128    xm5, m1, 1
-    vextracti128    xm3, m4, 1
-    movu            [r2], xm1
-    movu            [r2 + r3], xm5
-    movu            [r2 + r3 * 2], xm4
-    movu            [r2 + r6], xm3
-%endif
-    movq            xm3, [r0 + r4]                  ; m3 = row 11
-    punpcklbw       xm6, xm3
-    lea             r0, [r0 + r1 * 4]
-    movq            xm5, [r0]                       ; m5 = row 12
-    punpcklbw       xm3, xm5
-    vinserti128     m6, m6, xm3, 1
-    pmaddubsw       m3, m6, [r5 + 1 * mmsize]
-    paddw           m0, m3
-    pmaddubsw       m6, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 13
-    punpcklbw       xm5, xm3
-    movq            xm2, [r0 + r1 * 2]              ; m2 = row 14
-    punpcklbw       xm3, xm2
-    vinserti128     m5, m5, xm3, 1
-    pmaddubsw       m3, m5, [r5 + 1 * mmsize]
-    paddw           m6, m3
-    pmaddubsw       m5, [r5]
-    movq            xm3, [r0 + r4]                  ; m3 = row 15
-    punpcklbw       xm2, xm3
-    lea             r0, [r0 + r1 * 4]
-    movq            xm1, [r0]                       ; m1 = row 16
-    punpcklbw       xm3, xm1
-    vinserti128     m2, m2, xm3, 1
-    pmaddubsw       m3, m2, [r5 + 1 * mmsize]
-    paddw           m5, m3
-    pmaddubsw       m2, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 17
-    punpcklbw       xm1, xm3
-    movq            xm4, [r0 + r1 * 2]              ; m4 = row 18
-    punpcklbw       xm3, xm4
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m1, [r5 + 1 * mmsize]
-    paddw           m2, m1
-    lea             r2, [r2 + r3 * 4]
-%ifidn %1,pp
-    pmulhrsw        m0, m7                          ; m0 = word: row 8, row 9
-    pmulhrsw        m6, m7                          ; m6 = word: row 10, row 11
-    pmulhrsw        m5, m7                          ; m5 = word: row 12, row 13
-    pmulhrsw        m2, m7                          ; m2 = word: row 14, row 15
-    packuswb        m0, m6
-    packuswb        m5, m2
-    vextracti128    xm6, m0, 1
-    vextracti128    xm2, m5, 1
-    movq            [r2], xm0
-    movq            [r2 + r3], xm6
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r6], xm6
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm5
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm5
-    movhps          [r2 + r6], xm2
-%else
-    psubw           m0, m7                          ; m0 = word: row 8, row 9
-    psubw           m6, m7                          ; m6 = word: row 10, row 11
-    psubw           m5, m7                          ; m5 = word: row 12, row 13
-    psubw           m2, m7                          ; m2 = word: row 14, row 15
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m6, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r6], xm3
-    lea             r2, [r2 + r3 * 4]
-    vextracti128    xm1, m5, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2], xm5
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-%endif
-%endmacro
-
-%macro FILTER_VER_CHROMA_AVX2_8x16 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x16, 4, 7, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m7, [pw_512]
-%else
-    add             r3d, r3d
-    mova            m7, [pw_2000]
-%endif
-    lea             r6, [r3 * 3]
-    PROCESS_CHROMA_AVX2_W8_16R %1
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_8x16 pp
-    FILTER_VER_CHROMA_AVX2_8x16 ps
-
-%macro FILTER_VER_CHROMA_AVX2_8x12 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x12, 4, 7, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1, pp
-    mova            m7, [pw_512]
-%else
-    add             r3d, r3d
-    mova            m7, [pw_2000]
-%endif
-    lea             r6, [r3 * 3]
-    movq            xm1, [r0]                       ; m1 = row 0
-    movq            xm2, [r0 + r1]                  ; m2 = row 1
-    punpcklbw       xm1, xm2
-    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
-    punpcklbw       xm2, xm3
-    vinserti128     m5, m1, xm2, 1
-    pmaddubsw       m5, [r5]
-    movq            xm4, [r0 + r4]                  ; m4 = row 3
-    punpcklbw       xm3, xm4
-    lea             r0, [r0 + r1 * 4]
-    movq            xm1, [r0]                       ; m1 = row 4
-    punpcklbw       xm4, xm1
-    vinserti128     m2, m3, xm4, 1
-    pmaddubsw       m0, m2, [r5 + 1 * mmsize]
-    paddw           m5, m0
-    pmaddubsw       m2, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 5
-    punpcklbw       xm1, xm3
-    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
-    punpcklbw       xm3, xm4
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m0, m1, [r5 + 1 * mmsize]
-    paddw           m2, m0
-    pmaddubsw       m1, [r5]
-    movq            xm3, [r0 + r4]                  ; m3 = row 7
-    punpcklbw       xm4, xm3
-    lea             r0, [r0 + r1 * 4]
-    movq            xm0, [r0]                       ; m0 = row 8
-    punpcklbw       xm3, xm0
-    vinserti128     m4, m4, xm3, 1
-    pmaddubsw       m3, m4, [r5 + 1 * mmsize]
-    paddw           m1, m3
-    pmaddubsw       m4, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 9
-    punpcklbw       xm0, xm3
-    movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
-    punpcklbw       xm3, xm6
-    vinserti128     m0, m0, xm3, 1
-    pmaddubsw       m3, m0, [r5 + 1 * mmsize]
-    paddw           m4, m3
-    pmaddubsw       m0, [r5]
-%ifidn %1, pp
-    pmulhrsw        m5, m7                          ; m5 = word: row 0, row 1
-    pmulhrsw        m2, m7                          ; m2 = word: row 2, row 3
-    pmulhrsw        m1, m7                          ; m1 = word: row 4, row 5
-    pmulhrsw        m4, m7                          ; m4 = word: row 6, row 7
-    packuswb        m5, m2
-    packuswb        m1, m4
-    vextracti128    xm2, m5, 1
-    vextracti128    xm4, m1, 1
-    movq            [r2], xm5
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm5
-    movhps          [r2 + r6], xm2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm1
-    movq            [r2 + r3], xm4
-    movhps          [r2 + r3 * 2], xm1
-    movhps          [r2 + r6], xm4
-%else
-    psubw           m5, m7                          ; m5 = word: row 0, row 1
-    psubw           m2, m7                          ; m2 = word: row 2, row 3
-    psubw           m1, m7                          ; m1 = word: row 4, row 5
-    psubw           m4, m7                          ; m4 = word: row 6, row 7
-    vextracti128    xm3, m5, 1
-    movu            [r2], xm5
-    movu            [r2 + r3], xm3
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-    lea             r2, [r2 + r3 * 4]
-    vextracti128    xm5, m1, 1
-    vextracti128    xm3, m4, 1
-    movu            [r2], xm1
-    movu            [r2 + r3], xm5
-    movu            [r2 + r3 * 2], xm4
-    movu            [r2 + r6], xm3
-%endif
-    movq            xm3, [r0 + r4]                  ; m3 = row 11
-    punpcklbw       xm6, xm3
-    lea             r0, [r0 + r1 * 4]
-    movq            xm5, [r0]                       ; m5 = row 12
-    punpcklbw       xm3, xm5
-    vinserti128     m6, m6, xm3, 1
-    pmaddubsw       m3, m6, [r5 + 1 * mmsize]
-    paddw           m0, m3
-    pmaddubsw       m6, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 13
-    punpcklbw       xm5, xm3
-    movq            xm2, [r0 + r1 * 2]              ; m2 = row 14
-    punpcklbw       xm3, xm2
-    vinserti128     m5, m5, xm3, 1
-    pmaddubsw       m3, m5, [r5 + 1 * mmsize]
-    paddw           m6, m3
-    lea             r2, [r2 + r3 * 4]
-%ifidn %1, pp
-    pmulhrsw        m0, m7                          ; m0 = word: row 8, row 9
-    pmulhrsw        m6, m7                          ; m6 = word: row 10, row 11
-    packuswb        m0, m6
-    vextracti128    xm6, m0, 1
-    movq            [r2], xm0
-    movq            [r2 + r3], xm6
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r6], xm6
-%else
-    psubw           m0, m7                          ; m0 = word: row 8, row 9
-    psubw           m6, m7                          ; m6 = word: row 10, row 11
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m6, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r6], xm3
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_8x12 pp
-    FILTER_VER_CHROMA_AVX2_8x12 ps
-
-%macro FILTER_VER_CHROMA_AVX2_8xN 2
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x%2, 4, 7, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m7, [pw_512]
-%else
-    add             r3d, r3d
-    mova            m7, [pw_2000]
-%endif
-    lea             r6, [r3 * 3]
-%rep %2 / 16
-    PROCESS_CHROMA_AVX2_W8_16R %1
-    lea             r2, [r2 + r3 * 4]
-%endrep
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_8xN pp, 32
-    FILTER_VER_CHROMA_AVX2_8xN ps, 32
-    FILTER_VER_CHROMA_AVX2_8xN pp, 64
-    FILTER_VER_CHROMA_AVX2_8xN ps, 64
-
-%macro PROCESS_CHROMA_AVX2_W8_4R 0
-    movq            xm1, [r0]                       ; m1 = row 0
-    movq            xm2, [r0 + r1]                  ; m2 = row 1
-    punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
-    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
-    punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
-    vinserti128     m0, m1, xm2, 1                  ; m0 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
-    pmaddubsw       m0, [r5]
-    movq            xm4, [r0 + r4]                  ; m4 = row 3
-    punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
-    lea             r0, [r0 + r1 * 4]
-    movq            xm1, [r0]                       ; m1 = row 4
-    punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
-    vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
-    pmaddubsw       m4, m2, [r5 + 1 * mmsize]
-    paddw           m0, m4
-    pmaddubsw       m2, [r5]
-    movq            xm3, [r0 + r1]                  ; m3 = row 5
-    punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
-    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
-    punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
-    vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
-    pmaddubsw       m1, [r5 + 1 * mmsize]
-    paddw           m2, m1
-%endmacro
-
-%macro FILTER_VER_CHROMA_AVX2_8x4 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x4, 4, 6, 5
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-    PROCESS_CHROMA_AVX2_W8_4R
-%ifidn %1,pp
-    lea             r4, [r3 * 3]
-    mova            m3, [pw_512]
-    pmulhrsw        m0, m3                          ; m0 = word: row 0, row 1
-    pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
-    packuswb        m0, m2
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r4], xm2
-%else
-    add             r3d, r3d
-    vbroadcasti128  m3, [pw_2000]
-    lea             r4, [r3 * 3]
-    psubw           m0, m3                          ; m0 = word: row 0, row 1
-    psubw           m2, m3                          ; m2 = word: row 2, row 3
-    vextracti128    xm1, m0, 1
-    vextracti128    xm4, m2, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r4], xm4
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_8x4 pp
-    FILTER_VER_CHROMA_AVX2_8x4 ps
-
-%macro FILTER_VER_CHROMA_AVX2_8x2 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x2, 4, 6, 4
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-
-    movq            xm1, [r0]                       ; m1 = row 0
-    movq            xm2, [r0 + r1]                  ; m2 = row 1
-    punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
-    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
-    punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
-    vinserti128     m1, m1, xm2, 1                  ; m1 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
-    pmaddubsw       m1, [r5]
-    movq            xm2, [r0 + r4]                  ; m2 = row 3
-    punpcklbw       xm3, xm2                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
-    movq            xm0, [r0 + r1 * 4]              ; m0 = row 4
-    punpcklbw       xm2, xm0                        ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
-    vinserti128     m3, m3, xm2, 1                  ; m3 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
-    pmaddubsw       m3, [r5 + 1 * mmsize]
-    paddw           m1, m3
-%ifidn %1,pp
-    pmulhrsw        m1, [pw_512]                    ; m1 = word: row 0, row 1
-    packuswb        m1, m1
-    vextracti128    xm0, m1, 1
-    movq            [r2], xm1
-    movq            [r2 + r3], xm0
-%else
-    add             r3d, r3d
-    psubw           m1, [pw_2000]                   ; m1 = word: row 0, row 1
-    vextracti128    xm0, m1, 1
-    movu            [r2], xm1
-    movu            [r2 + r3], xm0
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_8x2 pp
-    FILTER_VER_CHROMA_AVX2_8x2 ps
-
-%macro FILTER_VER_CHROMA_AVX2_6x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_6x8, 4, 6, 7
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-    PROCESS_CHROMA_AVX2_W8_8R
-%ifidn %1,pp
-    lea             r4, [r3 * 3]
-    mova            m3, [pw_512]
-    pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
-    pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
-    pmulhrsw        m1, m3                          ; m1 = word: row 4, row 5
-    pmulhrsw        m4, m3                          ; m4 = word: row 6, row 7
-    packuswb        m5, m2
-    packuswb        m1, m4
-    vextracti128    xm2, m5, 1
-    vextracti128    xm4, m1, 1
-    movd            [r2], xm5
-    pextrw          [r2 + 4], xm5, 2
-    movd            [r2 + r3], xm2
-    pextrw          [r2 + r3 + 4], xm2, 2
-    pextrd          [r2 + r3 * 2], xm5, 2
-    pextrw          [r2 + r3 * 2 + 4], xm5, 6
-    pextrd          [r2 + r4], xm2, 2
-    pextrw          [r2 + r4 + 4], xm2, 6
-    lea             r2, [r2 + r3 * 4]
-    movd            [r2], xm1
-    pextrw          [r2 + 4], xm1, 2
-    movd            [r2 + r3], xm4
-    pextrw          [r2 + r3 + 4], xm4, 2
-    pextrd          [r2 + r3 * 2], xm1, 2
-    pextrw          [r2 + r3 * 2 + 4], xm1, 6
-    pextrd          [r2 + r4], xm4, 2
-    pextrw          [r2 + r4 + 4], xm4, 6
-%else
-    add             r3d, r3d
-    vbroadcasti128  m3, [pw_2000]
-    lea             r4, [r3 * 3]
-    psubw           m5, m3                          ; m5 = word: row 0, row 1
-    psubw           m2, m3                          ; m2 = word: row 2, row 3
-    psubw           m1, m3                          ; m1 = word: row 4, row 5
-    psubw           m4, m3                          ; m4 = word: row 6, row 7
-    vextracti128    xm6, m5, 1
-    vextracti128    xm3, m2, 1
-    vextracti128    xm0, m1, 1
-    movq            [r2], xm5
-    pextrd          [r2 + 8], xm5, 2
-    movq            [r2 + r3], xm6
-    pextrd          [r2 + r3 + 8], xm6, 2
-    movq            [r2 + r3 * 2], xm2
-    pextrd          [r2 + r3 * 2 + 8], xm2, 2
-    movq            [r2 + r4], xm3
-    pextrd          [r2 + r4 + 8], xm3, 2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm1
-    pextrd          [r2 + 8], xm1, 2
-    movq            [r2 + r3], xm0
-    pextrd          [r2 + r3 + 8], xm0, 2
-    movq            [r2 + r3 * 2], xm4
-    pextrd          [r2 + r3 * 2 + 8], xm4, 2
-    vextracti128    xm4, m4, 1
-    movq            [r2 + r4], xm4
-    pextrd          [r2 + r4 + 8], xm4, 2
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_6x8 pp
-    FILTER_VER_CHROMA_AVX2_6x8 ps
-
-;-----------------------------------------------------------------------------
-;void interp_4tap_vert_pp_6x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W6_H4 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_6x%2, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m5,        [r5 + r4 * 4]
-%else
-    movd        m5,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m6,        m5,       [tab_Vm]
-    pshufb      m5,        [tab_Vm + 16]
-    mova        m4,        [pw_512]
-
-    mov         r4d,       %2
-    lea         r5,        [3 * r1]
-
-.loop:
-    movq        m0,        [r0]
-    movq        m1,        [r0 + r1]
-    movq        m2,        [r0 + 2 * r1]
-    movq        m3,        [r0 + r5]
-
-    punpcklbw   m0,        m1
-    punpcklbw   m1,        m2
-    punpcklbw   m2,        m3
-
-    pmaddubsw   m0,        m6
-    pmaddubsw   m7,        m2, m5
-
-    paddw       m0,        m7
-
-    pmulhrsw    m0,        m4
-    packuswb    m0,        m0
-    movd        [r2],      m0
-    pextrw      [r2 + 4],  m0,    2
-
-    lea         r0,        [r0 + 4 * r1]
-
-    movq        m0,        [r0]
-    punpcklbw   m3,        m0
-
-    pmaddubsw   m1,        m6
-    pmaddubsw   m7,        m3, m5
-
-    paddw       m1,        m7
-
-    pmulhrsw    m1,        m4
-    packuswb    m1,        m1
-    movd        [r2 + r3],      m1
-    pextrw      [r2 + r3 + 4],  m1,    2
-
-    movq        m1,        [r0 + r1]
-    punpcklbw   m7,        m0,        m1
-
-    pmaddubsw   m2,        m6
-    pmaddubsw   m7,        m5
-
-    paddw       m2,        m7
-
-    pmulhrsw    m2,        m4
-    packuswb    m2,        m2
-    lea         r2,        [r2 + 2 * r3]
-    movd        [r2],      m2
-    pextrw      [r2 + 4],  m2,    2
-
-    movq        m2,        [r0 + 2 * r1]
-    punpcklbw   m1,        m2
-
-    pmaddubsw   m3,        m6
-    pmaddubsw   m1,        m5
-
-    paddw       m3,        m1
-
-    pmulhrsw    m3,        m4
-    packuswb    m3,        m3
-
-    movd        [r2 + r3],        m3
-    pextrw      [r2 + r3 + 4],    m3,    2
-
-    lea         r2,        [r2 + 2 * r3]
-
-    sub         r4,         4
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W6_H4 6, 8
-
-    FILTER_V4_W6_H4 6, 16
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_12x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W12_H2 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_12x%2, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m1,        m0,       [tab_Vm]
-    pshufb      m0,        [tab_Vm + 16]
-
-    mov         r4d,       %2
-
-.loop:
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m2,        m1
-
-    lea         r0,        [r0 + 2 * r1]
-    movu        m5,        [r0]
-    movu        m7,        [r0 + r1]
-
-    punpcklbw   m6,        m5,        m7
-    pmaddubsw   m6,        m0
-    paddw       m4,        m6
-
-    punpckhbw   m6,        m5,        m7
-    pmaddubsw   m6,        m0
-    paddw       m2,        m6
-
-    mova        m6,        [pw_512]
-
-    pmulhrsw    m4,        m6
-    pmulhrsw    m2,        m6
-
-    packuswb    m4,        m2
-
-    movh         [r2],     m4
-    pextrd       [r2 + 8], m4,  2
-
-    punpcklbw   m4,        m3,        m5
-    punpckhbw   m3,        m5
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m3,        m1
-
-    movu        m5,        [r0 + 2 * r1]
-
-    punpcklbw   m2,        m7,        m5
-    punpckhbw   m7,        m5
-
-    pmaddubsw   m2,        m0
-    pmaddubsw   m7,        m0
-
-    paddw       m4,        m2
-    paddw       m3,        m7
-
-    pmulhrsw    m4,        m6
-    pmulhrsw    m3,        m6
-
-    packuswb    m4,        m3
-
-    movh        [r2 + r3],      m4
-    pextrd      [r2 + r3 + 8],  m4,  2
-
-    lea         r2,        [r2 + 2 * r3]
-
-    sub         r4,        2
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W12_H2 12, 16
-
-    FILTER_V4_W12_H2 12, 32
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W16_H2 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_16x%2, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m1,        m0,       [tab_Vm]
-    pshufb      m0,        [tab_Vm + 16]
-
-    mov         r4d,       %2/2
-
-.loop:
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m2,        m1
-
-    lea         r0,        [r0 + 2 * r1]
-    movu        m5,        [r0]
-    movu        m6,        [r0 + r1]
-
-    punpckhbw   m7,        m5,        m6
-    pmaddubsw   m7,        m0
-    paddw       m2,        m7
-
-    punpcklbw   m7,        m5,        m6
-    pmaddubsw   m7,        m0
-    paddw       m4,        m7
-
-    mova        m7,        [pw_512]
-
-    pmulhrsw    m4,        m7
-    pmulhrsw    m2,        m7
-
-    packuswb    m4,        m2
-
-    movu        [r2],      m4
-
-    punpcklbw   m4,        m3,        m5
-    punpckhbw   m3,        m5
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m3,        m1
-
-    movu        m5,        [r0 + 2 * r1]
-
-    punpcklbw   m2,        m6,        m5
-    punpckhbw   m6,        m5
-
-    pmaddubsw   m2,        m0
-    pmaddubsw   m6,        m0
-
-    paddw       m4,        m2
-    paddw       m3,        m6
-
-    pmulhrsw    m4,        m7
-    pmulhrsw    m3,        m7
-
-    packuswb    m4,        m3
-
-    movu        [r2 + r3],      m4
-
-    lea         r2,        [r2 + 2 * r3]
-
-    dec         r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W16_H2 16,  4
-    FILTER_V4_W16_H2 16,  8
-    FILTER_V4_W16_H2 16, 12
-    FILTER_V4_W16_H2 16, 16
-    FILTER_V4_W16_H2 16, 32
-
-    FILTER_V4_W16_H2 16, 24
-    FILTER_V4_W16_H2 16, 64
-
-%macro FILTER_VER_CHROMA_AVX2_16x16 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_16x16, 4, 6, 15
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    mova            m12, [r5]
-    mova            m13, [r5 + mmsize]
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m14, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m14, [pw_2000]
-%endif
-    lea             r5, [r3 * 3]
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhbw       xm2, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddubsw       m0, m12
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhbw       xm3, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m1, m12
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhbw       xm4, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddubsw       m4, m2, m13
-    paddw           m0, m4
-    pmaddubsw       m2, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhbw       xm5, xm3, xm4
-    punpcklbw       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddubsw       m5, m3, m13
-    paddw           m1, m5
-    pmaddubsw       m3, m12
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhbw       xm6, xm4, xm5
-    punpcklbw       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddubsw       m6, m4, m13
-    paddw           m2, m6
-    pmaddubsw       m4, m12
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhbw       xm7, xm5, xm6
-    punpcklbw       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddubsw       m7, m5, m13
-    paddw           m3, m7
-    pmaddubsw       m5, m12
-    movu            xm7, [r0 + r4]                  ; m7 = row 7
-    punpckhbw       xm8, xm6, xm7
-    punpcklbw       xm6, xm7
-    vinserti128     m6, m6, xm8, 1
-    pmaddubsw       m8, m6, m13
-    paddw           m4, m8
-    pmaddubsw       m6, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm8, [r0]                       ; m8 = row 8
-    punpckhbw       xm9, xm7, xm8
-    punpcklbw       xm7, xm8
-    vinserti128     m7, m7, xm9, 1
-    pmaddubsw       m9, m7, m13
-    paddw           m5, m9
-    pmaddubsw       m7, m12
-    movu            xm9, [r0 + r1]                  ; m9 = row 9
-    punpckhbw       xm10, xm8, xm9
-    punpcklbw       xm8, xm9
-    vinserti128     m8, m8, xm10, 1
-    pmaddubsw       m10, m8, m13
-    paddw           m6, m10
-    pmaddubsw       m8, m12
-    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
-    punpckhbw       xm11, xm9, xm10
-    punpcklbw       xm9, xm10
-    vinserti128     m9, m9, xm11, 1
-    pmaddubsw       m11, m9, m13
-    paddw           m7, m11
-    pmaddubsw       m9, m12
-
-%ifidn %1,pp
-    pmulhrsw        m0, m14                         ; m0 = word: row 0
-    pmulhrsw        m1, m14                         ; m1 = word: row 1
-    pmulhrsw        m2, m14                         ; m2 = word: row 2
-    pmulhrsw        m3, m14                         ; m3 = word: row 3
-    pmulhrsw        m4, m14                         ; m4 = word: row 4
-    pmulhrsw        m5, m14                         ; m5 = word: row 5
-    pmulhrsw        m6, m14                         ; m6 = word: row 6
-    pmulhrsw        m7, m14                         ; m7 = word: row 7
-    packuswb        m0, m1
-    packuswb        m2, m3
-    packuswb        m4, m5
-    packuswb        m6, m7
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    vpermq          m4, m4, 11011000b
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m2, 1
-    vextracti128    xm5, m4, 1
-    vextracti128    xm7, m6, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r5], xm3
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm4
-    movu            [r2 + r3], xm5
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r5], xm7
-%else
-    psubw           m0, m14                         ; m0 = word: row 0
-    psubw           m1, m14                         ; m1 = word: row 1
-    psubw           m2, m14                         ; m2 = word: row 2
-    psubw           m3, m14                         ; m3 = word: row 3
-    psubw           m4, m14                         ; m4 = word: row 4
-    psubw           m5, m14                         ; m5 = word: row 5
-    psubw           m6, m14                         ; m6 = word: row 6
-    psubw           m7, m14                         ; m7 = word: row 7
-    movu            [r2], m0
-    movu            [r2 + r3], m1
-    movu            [r2 + r3 * 2], m2
-    movu            [r2 + r5], m3
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], m4
-    movu            [r2 + r3], m5
-    movu            [r2 + r3 * 2], m6
-    movu            [r2 + r5], m7
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm11, [r0 + r4]                 ; m11 = row 11
-    punpckhbw       xm6, xm10, xm11
-    punpcklbw       xm10, xm11
-    vinserti128     m10, m10, xm6, 1
-    pmaddubsw       m6, m10, m13
-    paddw           m8, m6
-    pmaddubsw       m10, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm6, [r0]                       ; m6 = row 12
-    punpckhbw       xm7, xm11, xm6
-    punpcklbw       xm11, xm6
-    vinserti128     m11, m11, xm7, 1
-    pmaddubsw       m7, m11, m13
-    paddw           m9, m7
-    pmaddubsw       m11, m12
-
-    movu            xm7, [r0 + r1]                  ; m7 = row 13
-    punpckhbw       xm0, xm6, xm7
-    punpcklbw       xm6, xm7
-    vinserti128     m6, m6, xm0, 1
-    pmaddubsw       m0, m6, m13
-    paddw           m10, m0
-    pmaddubsw       m6, m12
-    movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
-    punpckhbw       xm1, xm7, xm0
-    punpcklbw       xm7, xm0
-    vinserti128     m7, m7, xm1, 1
-    pmaddubsw       m1, m7, m13
-    paddw           m11, m1
-    pmaddubsw       m7, m12
-    movu            xm1, [r0 + r4]                  ; m1 = row 15
-    punpckhbw       xm2, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddubsw       m2, m0, m13
-    paddw           m6, m2
-    pmaddubsw       m0, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm2, [r0]                       ; m2 = row 16
-    punpckhbw       xm3, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m3, m1, m13
-    paddw           m7, m3
-    pmaddubsw       m1, m12
-    movu            xm3, [r0 + r1]                  ; m3 = row 17
-    punpckhbw       xm4, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddubsw       m2, m13
-    paddw           m0, m2
-    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
-    punpckhbw       xm5, xm3, xm4
-    punpcklbw       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddubsw       m3, m13
-    paddw           m1, m3
-
-%ifidn %1,pp
-    pmulhrsw        m8, m14                         ; m8 = word: row 8
-    pmulhrsw        m9, m14                         ; m9 = word: row 9
-    pmulhrsw        m10, m14                        ; m10 = word: row 10
-    pmulhrsw        m11, m14                        ; m11 = word: row 11
-    pmulhrsw        m6, m14                         ; m6 = word: row 12
-    pmulhrsw        m7, m14                         ; m7 = word: row 13
-    pmulhrsw        m0, m14                         ; m0 = word: row 14
-    pmulhrsw        m1, m14                         ; m1 = word: row 15
-    packuswb        m8, m9
-    packuswb        m10, m11
-    packuswb        m6, m7
-    packuswb        m0, m1
-    vpermq          m8, m8, 11011000b
-    vpermq          m10, m10, 11011000b
-    vpermq          m6, m6, 11011000b
-    vpermq          m0, m0, 11011000b
-    vextracti128    xm9, m8, 1
-    vextracti128    xm11, m10, 1
-    vextracti128    xm7, m6, 1
-    vextracti128    xm1, m0, 1
-    movu            [r2], xm8
-    movu            [r2 + r3], xm9
-    movu            [r2 + r3 * 2], xm10
-    movu            [r2 + r5], xm11
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm6
-    movu            [r2 + r3], xm7
-    movu            [r2 + r3 * 2], xm0
-    movu            [r2 + r5], xm1
-%else
-    psubw           m8, m14                         ; m8 = word: row 8
-    psubw           m9, m14                         ; m9 = word: row 9
-    psubw           m10, m14                        ; m10 = word: row 10
-    psubw           m11, m14                        ; m11 = word: row 11
-    psubw           m6, m14                         ; m6 = word: row 12
-    psubw           m7, m14                         ; m7 = word: row 13
-    psubw           m0, m14                         ; m0 = word: row 14
-    psubw           m1, m14                         ; m1 = word: row 15
-    movu            [r2], m8
-    movu            [r2 + r3], m9
-    movu            [r2 + r3 * 2], m10
-    movu            [r2 + r5], m11
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], m6
-    movu            [r2 + r3], m7
-    movu            [r2 + r3 * 2], m0
-    movu            [r2 + r5], m1
-%endif
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_16x16 pp
-    FILTER_VER_CHROMA_AVX2_16x16 ps
-%macro FILTER_VER_CHROMA_AVX2_16x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_16x8, 4, 7, 7
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m6, [pw_512]
-%else
-    add             r3d, r3d
-    mova            m6, [pw_2000]
-%endif
-    lea             r6, [r3 * 3]
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhbw       xm2, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddubsw       m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhbw       xm3, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhbw       xm4, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddubsw       m4, m2, [r5 + mmsize]
-    paddw           m0, m4
-    pmaddubsw       m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhbw       xm5, xm3, xm4
-    punpcklbw       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddubsw       m5, m3, [r5 + mmsize]
-    paddw           m1, m5
-    pmaddubsw       m3, [r5]
-%ifidn %1,pp
-    pmulhrsw        m0, m6                          ; m0 = word: row 0
-    pmulhrsw        m1, m6                          ; m1 = word: row 1
-    packuswb        m0, m1
-    vpermq          m0, m0, 11011000b
-    vextracti128    xm1, m0, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-%else
-    psubw           m0, m6                          ; m0 = word: row 0
-    psubw           m1, m6                          ; m1 = word: row 1
-    movu            [r2], m0
-    movu            [r2 + r3], m1
-%endif
-
-    movu            xm0, [r0 + r1]                  ; m0 = row 5
-    punpckhbw       xm1, xm4, xm0
-    punpcklbw       xm4, xm0
-    vinserti128     m4, m4, xm1, 1
-    pmaddubsw       m1, m4, [r5 + mmsize]
-    paddw           m2, m1
-    pmaddubsw       m4, [r5]
-    movu            xm1, [r0 + r1 * 2]              ; m1 = row 6
-    punpckhbw       xm5, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm5, 1
-    pmaddubsw       m5, m0, [r5 + mmsize]
-    paddw           m3, m5
-    pmaddubsw       m0, [r5]
-%ifidn %1,pp
-    pmulhrsw        m2, m6                          ; m2 = word: row 2
-    pmulhrsw        m3, m6                          ; m3 = word: row 3
-    packuswb        m2, m3
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-%else
-    psubw           m2, m6                          ; m2 = word: row 2
-    psubw           m3, m6                          ; m3 = word: row 3
-    movu            [r2 + r3 * 2], m2
-    movu            [r2 + r6], m3
-%endif
-
-    movu            xm2, [r0 + r4]                  ; m2 = row 7
-    punpckhbw       xm3, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m3, m1, [r5 + mmsize]
-    paddw           m4, m3
-    pmaddubsw       m1, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm3, [r0]                       ; m3 = row 8
-    punpckhbw       xm5, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm5, 1
-    pmaddubsw       m5, m2, [r5 + mmsize]
-    paddw           m0, m5
-    pmaddubsw       m2, [r5]
-    lea             r2, [r2 + r3 * 4]
-%ifidn %1,pp
-    pmulhrsw        m4, m6                          ; m4 = word: row 4
-    pmulhrsw        m0, m6                          ; m0 = word: row 5
-    packuswb        m4, m0
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm0, m4, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm0
-%else
-    psubw           m4, m6                          ; m4 = word: row 4
-    psubw           m0, m6                          ; m0 = word: row 5
-    movu            [r2], m4
-    movu            [r2 + r3], m0
-%endif
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 9
-    punpckhbw       xm4, xm3, xm5
-    punpcklbw       xm3, xm5
-    vinserti128     m3, m3, xm4, 1
-    pmaddubsw       m3, [r5 + mmsize]
-    paddw           m1, m3
-    movu            xm4, [r0 + r1 * 2]              ; m4 = row 10
-    punpckhbw       xm0, xm5, xm4
-    punpcklbw       xm5, xm4
-    vinserti128     m5, m5, xm0, 1
-    pmaddubsw       m5, [r5 + mmsize]
-    paddw           m2, m5
-%ifidn %1,pp
-    pmulhrsw        m1, m6                          ; m1 = word: row 6
-    pmulhrsw        m2, m6                          ; m2 = word: row 7
-    packuswb        m1, m2
-    vpermq          m1, m1, 11011000b
-    vextracti128    xm2, m1, 1
-    movu            [r2 + r3 * 2], xm1
-    movu            [r2 + r6], xm2
-%else
-    psubw           m1, m6                          ; m1 = word: row 6
-    psubw           m2, m6                          ; m2 = word: row 7
-    movu            [r2 + r3 * 2], m1
-    movu            [r2 + r6], m2
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_16x8 pp
-    FILTER_VER_CHROMA_AVX2_16x8 ps
-
-%macro FILTER_VER_CHROMA_AVX2_16x12 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_16x12, 4, 6, 10
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    mova            m8, [r5]
-    mova            m9, [r5 + mmsize]
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m7, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m7, [pw_2000]
-%endif
-    lea             r5, [r3 * 3]
-
-    movu            xm0, [r0]
-    vinserti128     m0, m0, [r0 + r1 * 2], 1
-    movu            xm1, [r0 + r1]
-    vinserti128     m1, m1, [r0 + r4], 1
-
-    punpcklbw       m2, m0, m1
-    punpckhbw       m3, m0, m1
-    vperm2i128      m4, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    pmaddubsw       m4, m8
-    pmaddubsw       m3, m2, m9
-    paddw           m4, m3
-    pmaddubsw       m2, m8
-
-    vextracti128    xm0, m0, 1
-    lea             r0, [r0 + r1 * 4]
-    vinserti128     m0, m0, [r0], 1
-
-    punpcklbw       m5, m1, m0
-    punpckhbw       m3, m1, m0
-    vperm2i128      m6, m5, m3, 0x20
-    vperm2i128      m5, m5, m3, 0x31
-    pmaddubsw       m6, m8
-    pmaddubsw       m3, m5, m9
-    paddw           m6, m3
-    pmaddubsw       m5, m8
-%ifidn %1,pp
-    pmulhrsw        m4, m7                         ; m4 = word: row 0
-    pmulhrsw        m6, m7                         ; m6 = word: row 1
-    packuswb        m4, m6
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm6, m4, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm6
-%else
-    psubw           m4, m7                         ; m4 = word: row 0
-    psubw           m6, m7                         ; m6 = word: row 1
-    movu            [r2], m4
-    movu            [r2 + r3], m6
-%endif
-
-    movu            xm4, [r0 + r1 * 2]
-    vinserti128     m4, m4, [r0 + r1], 1
-    vextracti128    xm1, m4, 1
-    vinserti128     m0, m0, xm1, 0
-
-    punpcklbw       m6, m0, m4
-    punpckhbw       m1, m0, m4
-    vperm2i128      m0, m6, m1, 0x20
-    vperm2i128      m6, m6, m1, 0x31
-    pmaddubsw       m1, m0, m9
-    paddw           m5, m1
-    pmaddubsw       m0, m8
-    pmaddubsw       m1, m6, m9
-    paddw           m2, m1
-    pmaddubsw       m6, m8
-
-%ifidn %1,pp
-    pmulhrsw        m2, m7                         ; m2 = word: row 2
-    pmulhrsw        m5, m7                         ; m5 = word: row 3
-    packuswb        m2, m5
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm5, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r5], xm5
-%else
-    psubw           m2, m7                         ; m2 = word: row 2
-    psubw           m5, m7                         ; m5 = word: row 3
-    movu            [r2 + r3 * 2], m2
-    movu            [r2 + r5], m5
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm1, [r0 + r4]
-    lea             r0, [r0 + r1 * 4]
-    vinserti128     m1, m1, [r0], 1
-    vinserti128     m4, m4, xm1, 1
-
-    punpcklbw       m2, m4, m1
-    punpckhbw       m5, m4, m1
-    vperm2i128      m3, m2, m5, 0x20
-    vperm2i128      m2, m2, m5, 0x31
-    pmaddubsw       m5, m3, m9
-    paddw           m6, m5
-    pmaddubsw       m3, m8
-    pmaddubsw       m5, m2, m9
-    paddw           m0, m5
-    pmaddubsw       m2, m8
-
-%ifidn %1,pp
-    pmulhrsw        m6, m7                         ; m6 = word: row 4
-    pmulhrsw        m0, m7                         ; m0 = word: row 5
-    packuswb        m6, m0
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm0, m6, 1
-    movu            [r2], xm6
-    movu            [r2 + r3], xm0
-%else
-    psubw           m6, m7                         ; m6 = word: row 4
-    psubw           m0, m7                         ; m0 = word: row 5
-    movu            [r2], m6
-    movu            [r2 + r3], m0
-%endif
-
-    movu            xm6, [r0 + r1 * 2]
-    vinserti128     m6, m6, [r0 + r1], 1
-    vextracti128    xm0, m6, 1
-    vinserti128     m1, m1, xm0, 0
-
-    punpcklbw       m4, m1, m6
-    punpckhbw       m5, m1, m6
-    vperm2i128      m0, m4, m5, 0x20
-    vperm2i128      m5, m4, m5, 0x31
-    pmaddubsw       m4, m0, m9
-    paddw           m2, m4
-    pmaddubsw       m0, m8
-    pmaddubsw       m4, m5, m9
-    paddw           m3, m4
-    pmaddubsw       m5, m8
-
-%ifidn %1,pp
-    pmulhrsw        m3, m7                         ; m3 = word: row 6
-    pmulhrsw        m2, m7                         ; m2 = word: row 7
-    packuswb        m3, m2
-    vpermq          m3, m3, 11011000b
-    vextracti128    xm2, m3, 1
-    movu            [r2 + r3 * 2], xm3
-    movu            [r2 + r5], xm2
-%else
-    psubw           m3, m7                         ; m3 = word: row 6
-    psubw           m2, m7                         ; m2 = word: row 7
-    movu            [r2 + r3 * 2], m3
-    movu            [r2 + r5], m2
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm3, [r0 + r4]
-    lea             r0, [r0 + r1 * 4]
-    vinserti128     m3, m3, [r0], 1
-    vinserti128     m6, m6, xm3, 1
-
-    punpcklbw       m2, m6, m3
-    punpckhbw       m1, m6, m3
-    vperm2i128      m4, m2, m1, 0x20
-    vperm2i128      m2, m2, m1, 0x31
-    pmaddubsw       m1, m4, m9
-    paddw           m5, m1
-    pmaddubsw       m4, m8
-    pmaddubsw       m1, m2, m9
-    paddw           m0, m1
-    pmaddubsw       m2, m8
-
-%ifidn %1,pp
-    pmulhrsw        m5, m7                         ; m5 = word: row 8
-    pmulhrsw        m0, m7                         ; m0 = word: row 9
-    packuswb        m5, m0
-    vpermq          m5, m5, 11011000b
-    vextracti128    xm0, m5, 1
-    movu            [r2], xm5
-    movu            [r2 + r3], xm0
-%else
-    psubw           m5, m7                         ; m5 = word: row 8
-    psubw           m0, m7                         ; m0 = word: row 9
-    movu            [r2], m5
-    movu            [r2 + r3], m0
-%endif
-
-    movu            xm5, [r0 + r1 * 2]
-    vinserti128     m5, m5, [r0 + r1], 1
-    vextracti128    xm0, m5, 1
-    vinserti128     m3, m3, xm0, 0
-
-    punpcklbw       m1, m3, m5
-    punpckhbw       m0, m3, m5
-    vperm2i128      m6, m1, m0, 0x20
-    vperm2i128      m0, m1, m0, 0x31
-    pmaddubsw       m1, m6, m9
-    paddw           m2, m1
-    pmaddubsw       m1, m0, m9
-    paddw           m4, m1
-
-%ifidn %1,pp
-    pmulhrsw        m4, m7                         ; m4 = word: row 10
-    pmulhrsw        m2, m7                         ; m2 = word: row 11
-    packuswb        m4, m2
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm2, m4, 1
-    movu            [r2 + r3 * 2], xm4
-    movu            [r2 + r5], xm2
-%else
-    psubw           m4, m7                         ; m4 = word: row 10
-    psubw           m2, m7                         ; m2 = word: row 11
-    movu            [r2 + r3 * 2], m4
-    movu            [r2 + r5], m2
-%endif
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_16x12 pp
-    FILTER_VER_CHROMA_AVX2_16x12 ps
-
-%macro FILTER_VER_CHROMA_AVX2_16xN 2
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_16x%2, 4, 8, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m7, [pw_512]
-%else
-    add             r3d, r3d
-    mova            m7, [pw_2000]
-%endif
-    lea             r6, [r3 * 3]
-    mov             r7d, %2 / 16
-.loopH:
-    movu            xm0, [r0]
-    vinserti128     m0, m0, [r0 + r1 * 2], 1
-    movu            xm1, [r0 + r1]
-    vinserti128     m1, m1, [r0 + r4], 1
-
-    punpcklbw       m2, m0, m1
-    punpckhbw       m3, m0, m1
-    vperm2i128      m4, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    pmaddubsw       m4, [r5]
-    pmaddubsw       m3, m2, [r5 + mmsize]
-    paddw           m4, m3
-    pmaddubsw       m2, [r5]
-
-    vextracti128    xm0, m0, 1
-    lea             r0, [r0 + r1 * 4]
-    vinserti128     m0, m0, [r0], 1
-
-    punpcklbw       m5, m1, m0
-    punpckhbw       m3, m1, m0
-    vperm2i128      m6, m5, m3, 0x20
-    vperm2i128      m5, m5, m3, 0x31
-    pmaddubsw       m6, [r5]
-    pmaddubsw       m3, m5, [r5 + mmsize]
-    paddw           m6, m3
-    pmaddubsw       m5, [r5]
-%ifidn %1,pp
-    pmulhrsw        m4, m7                         ; m4 = word: row 0
-    pmulhrsw        m6, m7                         ; m6 = word: row 1
-    packuswb        m4, m6
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm6, m4, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm6
-%else
-    psubw           m4, m7                         ; m4 = word: row 0
-    psubw           m6, m7                         ; m6 = word: row 1
-    movu            [r2], m4
-    movu            [r2 + r3], m6
-%endif
-
-    movu            xm4, [r0 + r1 * 2]
-    vinserti128     m4, m4, [r0 + r1], 1
-    vextracti128    xm1, m4, 1
-    vinserti128     m0, m0, xm1, 0
-
-    punpcklbw       m6, m0, m4
-    punpckhbw       m1, m0, m4
-    vperm2i128      m0, m6, m1, 0x20
-    vperm2i128      m6, m6, m1, 0x31
-    pmaddubsw       m1, m0, [r5 + mmsize]
-    paddw           m5, m1
-    pmaddubsw       m0, [r5]
-    pmaddubsw       m1, m6, [r5 + mmsize]
-    paddw           m2, m1
-    pmaddubsw       m6, [r5]
-
-%ifidn %1,pp
-    pmulhrsw        m2, m7                         ; m2 = word: row 2
-    pmulhrsw        m5, m7                         ; m5 = word: row 3
-    packuswb        m2, m5
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm5, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm5
-%else
-    psubw           m2, m7                         ; m2 = word: row 2
-    psubw           m5, m7                         ; m5 = word: row 3
-    movu            [r2 + r3 * 2], m2
-    movu            [r2 + r6], m5
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm1, [r0 + r4]
-    lea             r0, [r0 + r1 * 4]
-    vinserti128     m1, m1, [r0], 1
-    vinserti128     m4, m4, xm1, 1
-
-    punpcklbw       m2, m4, m1
-    punpckhbw       m5, m4, m1
-    vperm2i128      m3, m2, m5, 0x20
-    vperm2i128      m2, m2, m5, 0x31
-    pmaddubsw       m5, m3, [r5 + mmsize]
-    paddw           m6, m5
-    pmaddubsw       m3, [r5]
-    pmaddubsw       m5, m2, [r5 + mmsize]
-    paddw           m0, m5
-    pmaddubsw       m2, [r5]
-
-%ifidn %1,pp
-    pmulhrsw        m6, m7                         ; m6 = word: row 4
-    pmulhrsw        m0, m7                         ; m0 = word: row 5
-    packuswb        m6, m0
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm0, m6, 1
-    movu            [r2], xm6
-    movu            [r2 + r3], xm0
-%else
-    psubw           m6, m7                         ; m6 = word: row 4
-    psubw           m0, m7                         ; m0 = word: row 5
-    movu            [r2], m6
-    movu            [r2 + r3], m0
-%endif
-
-    movu            xm6, [r0 + r1 * 2]
-    vinserti128     m6, m6, [r0 + r1], 1
-    vextracti128    xm0, m6, 1
-    vinserti128     m1, m1, xm0, 0
-
-    punpcklbw       m4, m1, m6
-    punpckhbw       m5, m1, m6
-    vperm2i128      m0, m4, m5, 0x20
-    vperm2i128      m5, m4, m5, 0x31
-    pmaddubsw       m4, m0, [r5 + mmsize]
-    paddw           m2, m4
-    pmaddubsw       m0, [r5]
-    pmaddubsw       m4, m5, [r5 + mmsize]
-    paddw           m3, m4
-    pmaddubsw       m5, [r5]
-
-%ifidn %1,pp
-    pmulhrsw        m3, m7                         ; m3 = word: row 6
-    pmulhrsw        m2, m7                         ; m2 = word: row 7
-    packuswb        m3, m2
-    vpermq          m3, m3, 11011000b
-    vextracti128    xm2, m3, 1
-    movu            [r2 + r3 * 2], xm3
-    movu            [r2 + r6], xm2
-%else
-    psubw           m3, m7                         ; m3 = word: row 6
-    psubw           m2, m7                         ; m2 = word: row 7
-    movu            [r2 + r3 * 2], m3
-    movu            [r2 + r6], m2
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm3, [r0 + r4]
-    lea             r0, [r0 + r1 * 4]
-    vinserti128     m3, m3, [r0], 1
-    vinserti128     m6, m6, xm3, 1
-
-    punpcklbw       m2, m6, m3
-    punpckhbw       m1, m6, m3
-    vperm2i128      m4, m2, m1, 0x20
-    vperm2i128      m2, m2, m1, 0x31
-    pmaddubsw       m1, m4, [r5 + mmsize]
-    paddw           m5, m1
-    pmaddubsw       m4, [r5]
-    pmaddubsw       m1, m2, [r5 + mmsize]
-    paddw           m0, m1
-    pmaddubsw       m2, [r5]
-
-%ifidn %1,pp
-    pmulhrsw        m5, m7                         ; m5 = word: row 8
-    pmulhrsw        m0, m7                         ; m0 = word: row 9
-    packuswb        m5, m0
-    vpermq          m5, m5, 11011000b
-    vextracti128    xm0, m5, 1
-    movu            [r2], xm5
-    movu            [r2 + r3], xm0
-%else
-    psubw           m5, m7                         ; m5 = word: row 8
-    psubw           m0, m7                         ; m0 = word: row 9
-    movu            [r2], m5
-    movu            [r2 + r3], m0
-%endif
-
-    movu            xm5, [r0 + r1 * 2]
-    vinserti128     m5, m5, [r0 + r1], 1
-    vextracti128    xm0, m5, 1
-    vinserti128     m3, m3, xm0, 0
-
-    punpcklbw       m1, m3, m5
-    punpckhbw       m0, m3, m5
-    vperm2i128      m6, m1, m0, 0x20
-    vperm2i128      m0, m1, m0, 0x31
-    pmaddubsw       m1, m6, [r5 + mmsize]
-    paddw           m2, m1
-    pmaddubsw       m6, [r5]
-    pmaddubsw       m1, m0, [r5 + mmsize]
-    paddw           m4, m1
-    pmaddubsw       m0, [r5]
-
-%ifidn %1,pp
-    pmulhrsw        m4, m7                         ; m4 = word: row 10
-    pmulhrsw        m2, m7                         ; m2 = word: row 11
-    packuswb        m4, m2
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm2, m4, 1
-    movu            [r2 + r3 * 2], xm4
-    movu            [r2 + r6], xm2
-%else
-    psubw           m4, m7                         ; m4 = word: row 10
-    psubw           m2, m7                         ; m2 = word: row 11
-    movu            [r2 + r3 * 2], m4
-    movu            [r2 + r6], m2
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm3, [r0 + r4]
-    lea             r0, [r0 + r1 * 4]
-    vinserti128     m3, m3, [r0], 1
-    vinserti128     m5, m5, xm3, 1
-
-    punpcklbw       m2, m5, m3
-    punpckhbw       m1, m5, m3
-    vperm2i128      m4, m2, m1, 0x20
-    vperm2i128      m2, m2, m1, 0x31
-    pmaddubsw       m1, m4, [r5 + mmsize]
-    paddw           m0, m1
-    pmaddubsw       m4, [r5]
-    pmaddubsw       m1, m2, [r5 + mmsize]
-    paddw           m6, m1
-    pmaddubsw       m2, [r5]
-
-%ifidn %1,pp
-    pmulhrsw        m0, m7                         ; m0 = word: row 12
-    pmulhrsw        m6, m7                         ; m6 = word: row 13
-    packuswb        m0, m6
-    vpermq          m0, m0, 11011000b
-    vextracti128    xm6, m0, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm6
-%else
-    psubw           m0, m7                         ; m0 = word: row 12
-    psubw           m6, m7                         ; m6 = word: row 13
-    movu            [r2], m0
-    movu            [r2 + r3], m6
-%endif
-
-    movu            xm5, [r0 + r1 * 2]
-    vinserti128     m5, m5, [r0 + r1], 1
-    vextracti128    xm0, m5, 1
-    vinserti128     m3, m3, xm0, 0
-
-    punpcklbw       m1, m3, m5
-    punpckhbw       m0, m3, m5
-    vperm2i128      m6, m1, m0, 0x20
-    vperm2i128      m0, m1, m0, 0x31
-    pmaddubsw       m6, [r5 + mmsize]
-    paddw           m2, m6
-    pmaddubsw       m0, [r5 + mmsize]
-    paddw           m4, m0
-
-%ifidn %1,pp
-    pmulhrsw        m4, m7                         ; m4 = word: row 14
-    pmulhrsw        m2, m7                         ; m2 = word: row 15
-    packuswb        m4, m2
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm2, m4, 1
-    movu            [r2 + r3 * 2], xm4
-    movu            [r2 + r6], xm2
-%else
-    psubw           m4, m7                         ; m4 = word: row 14
-    psubw           m2, m7                         ; m2 = word: row 15
-    movu            [r2 + r3 * 2], m4
-    movu            [r2 + r6], m2
-%endif
-    lea             r2, [r2 + r3 * 4]
-    dec             r7d
-    jnz             .loopH
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_16xN pp, 32
-    FILTER_VER_CHROMA_AVX2_16xN ps, 32
-    FILTER_VER_CHROMA_AVX2_16xN pp, 64
-    FILTER_VER_CHROMA_AVX2_16xN ps, 64
-
-%macro FILTER_VER_CHROMA_AVX2_16x24 1
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_16x24, 4, 6, 15
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    mova            m12, [r5]
-    mova            m13, [r5 + mmsize]
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m14, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m14, [pw_2000]
-%endif
-    lea             r5, [r3 * 3]
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhbw       xm2, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddubsw       m0, m12
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhbw       xm3, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m1, m12
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhbw       xm4, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddubsw       m4, m2, m13
-    paddw           m0, m4
-    pmaddubsw       m2, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhbw       xm5, xm3, xm4
-    punpcklbw       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddubsw       m5, m3, m13
-    paddw           m1, m5
-    pmaddubsw       m3, m12
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhbw       xm6, xm4, xm5
-    punpcklbw       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddubsw       m6, m4, m13
-    paddw           m2, m6
-    pmaddubsw       m4, m12
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhbw       xm7, xm5, xm6
-    punpcklbw       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddubsw       m7, m5, m13
-    paddw           m3, m7
-    pmaddubsw       m5, m12
-    movu            xm7, [r0 + r4]                  ; m7 = row 7
-    punpckhbw       xm8, xm6, xm7
-    punpcklbw       xm6, xm7
-    vinserti128     m6, m6, xm8, 1
-    pmaddubsw       m8, m6, m13
-    paddw           m4, m8
-    pmaddubsw       m6, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm8, [r0]                       ; m8 = row 8
-    punpckhbw       xm9, xm7, xm8
-    punpcklbw       xm7, xm8
-    vinserti128     m7, m7, xm9, 1
-    pmaddubsw       m9, m7, m13
-    paddw           m5, m9
-    pmaddubsw       m7, m12
-    movu            xm9, [r0 + r1]                  ; m9 = row 9
-    punpckhbw       xm10, xm8, xm9
-    punpcklbw       xm8, xm9
-    vinserti128     m8, m8, xm10, 1
-    pmaddubsw       m10, m8, m13
-    paddw           m6, m10
-    pmaddubsw       m8, m12
-    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
-    punpckhbw       xm11, xm9, xm10
-    punpcklbw       xm9, xm10
-    vinserti128     m9, m9, xm11, 1
-    pmaddubsw       m11, m9, m13
-    paddw           m7, m11
-    pmaddubsw       m9, m12
-
-%ifidn %1,pp
-    pmulhrsw        m0, m14                         ; m0 = word: row 0
-    pmulhrsw        m1, m14                         ; m1 = word: row 1
-    pmulhrsw        m2, m14                         ; m2 = word: row 2
-    pmulhrsw        m3, m14                         ; m3 = word: row 3
-    pmulhrsw        m4, m14                         ; m4 = word: row 4
-    pmulhrsw        m5, m14                         ; m5 = word: row 5
-    pmulhrsw        m6, m14                         ; m6 = word: row 6
-    pmulhrsw        m7, m14                         ; m7 = word: row 7
-    packuswb        m0, m1
-    packuswb        m2, m3
-    packuswb        m4, m5
-    packuswb        m6, m7
-    vpermq          m0, m0, q3120
-    vpermq          m2, m2, q3120
-    vpermq          m4, m4, q3120
-    vpermq          m6, m6, q3120
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m2, 1
-    vextracti128    xm5, m4, 1
-    vextracti128    xm7, m6, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r5], xm3
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm4
-    movu            [r2 + r3], xm5
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r5], xm7
-%else
-    psubw           m0, m14                         ; m0 = word: row 0
-    psubw           m1, m14                         ; m1 = word: row 1
-    psubw           m2, m14                         ; m2 = word: row 2
-    psubw           m3, m14                         ; m3 = word: row 3
-    psubw           m4, m14                         ; m4 = word: row 4
-    psubw           m5, m14                         ; m5 = word: row 5
-    psubw           m6, m14                         ; m6 = word: row 6
-    psubw           m7, m14                         ; m7 = word: row 7
-    movu            [r2], m0
-    movu            [r2 + r3], m1
-    movu            [r2 + r3 * 2], m2
-    movu            [r2 + r5], m3
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], m4
-    movu            [r2 + r3], m5
-    movu            [r2 + r3 * 2], m6
-    movu            [r2 + r5], m7
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm11, [r0 + r4]                 ; m11 = row 11
-    punpckhbw       xm6, xm10, xm11
-    punpcklbw       xm10, xm11
-    vinserti128     m10, m10, xm6, 1
-    pmaddubsw       m6, m10, m13
-    paddw           m8, m6
-    pmaddubsw       m10, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm6, [r0]                       ; m6 = row 12
-    punpckhbw       xm7, xm11, xm6
-    punpcklbw       xm11, xm6
-    vinserti128     m11, m11, xm7, 1
-    pmaddubsw       m7, m11, m13
-    paddw           m9, m7
-    pmaddubsw       m11, m12
-
-    movu            xm7, [r0 + r1]                  ; m7 = row 13
-    punpckhbw       xm0, xm6, xm7
-    punpcklbw       xm6, xm7
-    vinserti128     m6, m6, xm0, 1
-    pmaddubsw       m0, m6, m13
-    paddw           m10, m0
-    pmaddubsw       m6, m12
-    movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
-    punpckhbw       xm1, xm7, xm0
-    punpcklbw       xm7, xm0
-    vinserti128     m7, m7, xm1, 1
-    pmaddubsw       m1, m7, m13
-    paddw           m11, m1
-    pmaddubsw       m7, m12
-    movu            xm1, [r0 + r4]                  ; m1 = row 15
-    punpckhbw       xm2, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddubsw       m2, m0, m13
-    paddw           m6, m2
-    pmaddubsw       m0, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm2, [r0]                       ; m2 = row 16
-    punpckhbw       xm3, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m3, m1, m13
-    paddw           m7, m3
-    pmaddubsw       m1, m12
-    movu            xm3, [r0 + r1]                  ; m3 = row 17
-    punpckhbw       xm4, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddubsw       m4, m2, m13
-    paddw           m0, m4
-    pmaddubsw       m2, m12
-    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
-    punpckhbw       xm5, xm3, xm4
-    punpcklbw       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddubsw       m5, m3, m13
-    paddw           m1, m5
-    pmaddubsw       m3, m12
-
-%ifidn %1,pp
-    pmulhrsw        m8, m14                         ; m8 = word: row 8
-    pmulhrsw        m9, m14                         ; m9 = word: row 9
-    pmulhrsw        m10, m14                        ; m10 = word: row 10
-    pmulhrsw        m11, m14                        ; m11 = word: row 11
-    pmulhrsw        m6, m14                         ; m6 = word: row 12
-    pmulhrsw        m7, m14                         ; m7 = word: row 13
-    pmulhrsw        m0, m14                         ; m0 = word: row 14
-    pmulhrsw        m1, m14                         ; m1 = word: row 15
-    packuswb        m8, m9
-    packuswb        m10, m11
-    packuswb        m6, m7
-    packuswb        m0, m1
-    vpermq          m8, m8, q3120
-    vpermq          m10, m10, q3120
-    vpermq          m6, m6, q3120
-    vpermq          m0, m0, q3120
-    vextracti128    xm9, m8, 1
-    vextracti128    xm11, m10, 1
-    vextracti128    xm7, m6, 1
-    vextracti128    xm1, m0, 1
-    movu            [r2], xm8
-    movu            [r2 + r3], xm9
-    movu            [r2 + r3 * 2], xm10
-    movu            [r2 + r5], xm11
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm6
-    movu            [r2 + r3], xm7
-    movu            [r2 + r3 * 2], xm0
-    movu            [r2 + r5], xm1
-%else
-    psubw           m8, m14                         ; m8 = word: row 8
-    psubw           m9, m14                         ; m9 = word: row 9
-    psubw           m10, m14                        ; m10 = word: row 10
-    psubw           m11, m14                        ; m11 = word: row 11
-    psubw           m6, m14                         ; m6 = word: row 12
-    psubw           m7, m14                         ; m7 = word: row 13
-    psubw           m0, m14                         ; m0 = word: row 14
-    psubw           m1, m14                         ; m1 = word: row 15
-    movu            [r2], m8
-    movu            [r2 + r3], m9
-    movu            [r2 + r3 * 2], m10
-    movu            [r2 + r5], m11
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], m6
-    movu            [r2 + r3], m7
-    movu            [r2 + r3 * 2], m0
-    movu            [r2 + r5], m1
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm5, [r0 + r4]                  ; m5 = row 19
-    punpckhbw       xm6, xm4, xm5
-    punpcklbw       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddubsw       m6, m4, m13
-    paddw           m2, m6
-    pmaddubsw       m4, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm6, [r0]                       ; m6 = row 20
-    punpckhbw       xm7, xm5, xm6
-    punpcklbw       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddubsw       m7, m5, m13
-    paddw           m3, m7
-    pmaddubsw       m5, m12
-    movu            xm7, [r0 + r1]                  ; m7 = row 21
-    punpckhbw       xm0, xm6, xm7
-    punpcklbw       xm6, xm7
-    vinserti128     m6, m6, xm0, 1
-    pmaddubsw       m0, m6, m13
-    paddw           m4, m0
-    pmaddubsw       m6, m12
-    movu            xm0, [r0 + r1 * 2]              ; m0 = row 22
-    punpckhbw       xm1, xm7, xm0
-    punpcklbw       xm7, xm0
-    vinserti128     m7, m7, xm1, 1
-    pmaddubsw       m1, m7, m13
-    paddw           m5, m1
-    pmaddubsw       m7, m12
-    movu            xm1, [r0 + r4]                  ; m1 = row 23
-    punpckhbw       xm8, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm8, 1
-    pmaddubsw       m8, m0, m13
-    paddw           m6, m8
-    pmaddubsw       m0, m12
-    lea             r0, [r0 + r1 * 4]
-    movu            xm8, [r0]                       ; m8 = row 24
-    punpckhbw       xm9, xm1, xm8
-    punpcklbw       xm1, xm8
-    vinserti128     m1, m1, xm9, 1
-    pmaddubsw       m9, m1, m13
-    paddw           m7, m9
-    pmaddubsw       m1, m12
-    movu            xm9, [r0 + r1]                  ; m9 = row 25
-    punpckhbw       xm10, xm8, xm9
-    punpcklbw       xm8, xm9
-    vinserti128     m8, m8, xm10, 1
-    pmaddubsw       m8, m13
-    paddw           m0, m8
-    movu            xm10, [r0 + r1 * 2]             ; m10 = row 26
-    punpckhbw       xm11, xm9, xm10
-    punpcklbw       xm9, xm10
-    vinserti128     m9, m9, xm11, 1
-    pmaddubsw       m9, m13
-    paddw           m1, m9
-
-%ifidn %1,pp
-    pmulhrsw        m2, m14                         ; m2 = word: row 16
-    pmulhrsw        m3, m14                         ; m3 = word: row 17
-    pmulhrsw        m4, m14                         ; m4 = word: row 18
-    pmulhrsw        m5, m14                         ; m5 = word: row 19
-    pmulhrsw        m6, m14                         ; m6 = word: row 20
-    pmulhrsw        m7, m14                         ; m7 = word: row 21
-    pmulhrsw        m0, m14                         ; m0 = word: row 22
-    pmulhrsw        m1, m14                         ; m1 = word: row 23
-    packuswb        m2, m3
-    packuswb        m4, m5
-    packuswb        m6, m7
-    packuswb        m0, m1
-    vpermq          m2, m2, q3120
-    vpermq          m4, m4, q3120
-    vpermq          m6, m6, q3120
-    vpermq          m0, m0, q3120
-    vextracti128    xm3, m2, 1
-    vextracti128    xm5, m4, 1
-    vextracti128    xm7, m6, 1
-    vextracti128    xm1, m0, 1
-    movu            [r2], xm2
-    movu            [r2 + r3], xm3
-    movu            [r2 + r3 * 2], xm4
-    movu            [r2 + r5], xm5
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm6
-    movu            [r2 + r3], xm7
-    movu            [r2 + r3 * 2], xm0
-    movu            [r2 + r5], xm1
-%else
-    psubw           m2, m14                         ; m2 = word: row 16
-    psubw           m3, m14                         ; m3 = word: row 17
-    psubw           m4, m14                         ; m4 = word: row 18
-    psubw           m5, m14                         ; m5 = word: row 19
-    psubw           m6, m14                         ; m6 = word: row 20
-    psubw           m7, m14                         ; m7 = word: row 21
-    psubw           m0, m14                         ; m0 = word: row 22
-    psubw           m1, m14                         ; m1 = word: row 23
-    movu            [r2], m2
-    movu            [r2 + r3], m3
-    movu            [r2 + r3 * 2], m4
-    movu            [r2 + r5], m5
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], m6
-    movu            [r2 + r3], m7
-    movu            [r2 + r3 * 2], m0
-    movu            [r2 + r5], m1
-%endif
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_16x24 pp
-    FILTER_VER_CHROMA_AVX2_16x24 ps
-
-%macro FILTER_VER_CHROMA_AVX2_24x32 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_24x32, 4, 9, 10
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    mova            m8, [r5]
-    mova            m9, [r5 + mmsize]
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m7, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m7, [pw_2000]
-%endif
-    lea             r6, [r3 * 3]
-    mov             r5d, 2
-.loopH:
-    movu            xm0, [r0]
-    vinserti128     m0, m0, [r0 + r1 * 2], 1
-    movu            xm1, [r0 + r1]
-    vinserti128     m1, m1, [r0 + r4], 1
-
-    punpcklbw       m2, m0, m1
-    punpckhbw       m3, m0, m1
-    vperm2i128      m4, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    pmaddubsw       m4, m8
-    pmaddubsw       m3, m2, m9
-    paddw           m4, m3
-    pmaddubsw       m2, m8
-
-    vextracti128    xm0, m0, 1
-    lea             r7, [r0 + r1 * 4]
-    vinserti128     m0, m0, [r7], 1
-
-    punpcklbw       m5, m1, m0
-    punpckhbw       m3, m1, m0
-    vperm2i128      m6, m5, m3, 0x20
-    vperm2i128      m5, m5, m3, 0x31
-    pmaddubsw       m6, m8
-    pmaddubsw       m3, m5, m9
-    paddw           m6, m3
-    pmaddubsw       m5, m8
-%ifidn %1,pp
-    pmulhrsw        m4, m7                         ; m4 = word: row 0
-    pmulhrsw        m6, m7                         ; m6 = word: row 1
-    packuswb        m4, m6
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm6, m4, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm6
-%else
-    psubw           m4, m7                         ; m4 = word: row 0
-    psubw           m6, m7                         ; m6 = word: row 1
-    movu            [r2], m4
-    movu            [r2 + r3], m6
-%endif
-
-    movu            xm4, [r7 + r1 * 2]
-    vinserti128     m4, m4, [r7 + r1], 1
-    vextracti128    xm1, m4, 1
-    vinserti128     m0, m0, xm1, 0
-
-    punpcklbw       m6, m0, m4
-    punpckhbw       m1, m0, m4
-    vperm2i128      m0, m6, m1, 0x20
-    vperm2i128      m6, m6, m1, 0x31
-    pmaddubsw       m1, m0, m9
-    paddw           m5, m1
-    pmaddubsw       m0, m8
-    pmaddubsw       m1, m6, m9
-    paddw           m2, m1
-    pmaddubsw       m6, m8
-
-%ifidn %1,pp
-    pmulhrsw        m2, m7                         ; m2 = word: row 2
-    pmulhrsw        m5, m7                         ; m5 = word: row 3
-    packuswb        m2, m5
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm5, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm5
-%else
-    psubw           m2, m7                         ; m2 = word: row 2
-    psubw           m5, m7                         ; m5 = word: row 3
-    movu            [r2 + r3 * 2], m2
-    movu            [r2 + r6], m5
-%endif
-    lea             r8, [r2 + r3 * 4]
-
-    movu            xm1, [r7 + r4]
-    lea             r7, [r7 + r1 * 4]
-    vinserti128     m1, m1, [r7], 1
-    vinserti128     m4, m4, xm1, 1
-
-    punpcklbw       m2, m4, m1
-    punpckhbw       m5, m4, m1
-    vperm2i128      m3, m2, m5, 0x20
-    vperm2i128      m2, m2, m5, 0x31
-    pmaddubsw       m5, m3, m9
-    paddw           m6, m5
-    pmaddubsw       m3, m8
-    pmaddubsw       m5, m2, m9
-    paddw           m0, m5
-    pmaddubsw       m2, m8
-
-%ifidn %1,pp
-    pmulhrsw        m6, m7                         ; m6 = word: row 4
-    pmulhrsw        m0, m7                         ; m0 = word: row 5
-    packuswb        m6, m0
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm0, m6, 1
-    movu            [r8], xm6
-    movu            [r8 + r3], xm0
-%else
-    psubw           m6, m7                         ; m6 = word: row 4
-    psubw           m0, m7                         ; m0 = word: row 5
-    movu            [r8], m6
-    movu            [r8 + r3], m0
-%endif
-
-    movu            xm6, [r7 + r1 * 2]
-    vinserti128     m6, m6, [r7 + r1], 1
-    vextracti128    xm0, m6, 1
-    vinserti128     m1, m1, xm0, 0
-
-    punpcklbw       m4, m1, m6
-    punpckhbw       m5, m1, m6
-    vperm2i128      m0, m4, m5, 0x20
-    vperm2i128      m5, m4, m5, 0x31
-    pmaddubsw       m4, m0, m9
-    paddw           m2, m4
-    pmaddubsw       m0, m8
-    pmaddubsw       m4, m5, m9
-    paddw           m3, m4
-    pmaddubsw       m5, m8
-
-%ifidn %1,pp
-    pmulhrsw        m3, m7                         ; m3 = word: row 6
-    pmulhrsw        m2, m7                         ; m2 = word: row 7
-    packuswb        m3, m2
-    vpermq          m3, m3, 11011000b
-    vextracti128    xm2, m3, 1
-    movu            [r8 + r3 * 2], xm3
-    movu            [r8 + r6], xm2
-%else
-    psubw           m3, m7                         ; m3 = word: row 6
-    psubw           m2, m7                         ; m2 = word: row 7
-    movu            [r8 + r3 * 2], m3
-    movu            [r8 + r6], m2
-%endif
-    lea             r8, [r8 + r3 * 4]
-
-    movu            xm3, [r7 + r4]
-    lea             r7, [r7 + r1 * 4]
-    vinserti128     m3, m3, [r7], 1
-    vinserti128     m6, m6, xm3, 1
-
-    punpcklbw       m2, m6, m3
-    punpckhbw       m1, m6, m3
-    vperm2i128      m4, m2, m1, 0x20
-    vperm2i128      m2, m2, m1, 0x31
-    pmaddubsw       m1, m4, m9
-    paddw           m5, m1
-    pmaddubsw       m4, m8
-    pmaddubsw       m1, m2, m9
-    paddw           m0, m1
-    pmaddubsw       m2, m8
-
-%ifidn %1,pp
-    pmulhrsw        m5, m7                         ; m5 = word: row 8
-    pmulhrsw        m0, m7                         ; m0 = word: row 9
-    packuswb        m5, m0
-    vpermq          m5, m5, 11011000b
-    vextracti128    xm0, m5, 1
-    movu            [r8], xm5
-    movu            [r8 + r3], xm0
-%else
-    psubw           m5, m7                         ; m5 = word: row 8
-    psubw           m0, m7                         ; m0 = word: row 9
-    movu            [r8], m5
-    movu            [r8 + r3], m0
-%endif
-
-    movu            xm5, [r7 + r1 * 2]
-    vinserti128     m5, m5, [r7 + r1], 1
-    vextracti128    xm0, m5, 1
-    vinserti128     m3, m3, xm0, 0
-
-    punpcklbw       m1, m3, m5
-    punpckhbw       m0, m3, m5
-    vperm2i128      m6, m1, m0, 0x20
-    vperm2i128      m0, m1, m0, 0x31
-    pmaddubsw       m1, m6, m9
-    paddw           m2, m1
-    pmaddubsw       m6, m8
-    pmaddubsw       m1, m0, m9
-    paddw           m4, m1
-    pmaddubsw       m0, m8
-
-%ifidn %1,pp
-    pmulhrsw        m4, m7                         ; m4 = word: row 10
-    pmulhrsw        m2, m7                         ; m2 = word: row 11
-    packuswb        m4, m2
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm2, m4, 1
-    movu            [r8 + r3 * 2], xm4
-    movu            [r8 + r6], xm2
-%else
-    psubw           m4, m7                         ; m4 = word: row 10
-    psubw           m2, m7                         ; m2 = word: row 11
-    movu            [r8 + r3 * 2], m4
-    movu            [r8 + r6], m2
-%endif
-    lea             r8, [r8 + r3 * 4]
-
-    movu            xm3, [r7 + r4]
-    lea             r7, [r7 + r1 * 4]
-    vinserti128     m3, m3, [r7], 1
-    vinserti128     m5, m5, xm3, 1
-
-    punpcklbw       m2, m5, m3
-    punpckhbw       m1, m5, m3
-    vperm2i128      m4, m2, m1, 0x20
-    vperm2i128      m2, m2, m1, 0x31
-    pmaddubsw       m1, m4, m9
-    paddw           m0, m1
-    pmaddubsw       m4, m8
-    pmaddubsw       m1, m2, m9
-    paddw           m6, m1
-    pmaddubsw       m2, m8
-
-%ifidn %1,pp
-    pmulhrsw        m0, m7                         ; m0 = word: row 12
-    pmulhrsw        m6, m7                         ; m6 = word: row 13
-    packuswb        m0, m6
-    vpermq          m0, m0, 11011000b
-    vextracti128    xm6, m0, 1
-    movu            [r8], xm0
-    movu            [r8 + r3], xm6
-%else
-    psubw           m0, m7                         ; m0 = word: row 12
-    psubw           m6, m7                         ; m6 = word: row 13
-    movu            [r8], m0
-    movu            [r8 + r3], m6
-%endif
-
-    movu            xm5, [r7 + r1 * 2]
-    vinserti128     m5, m5, [r7 + r1], 1
-    vextracti128    xm0, m5, 1
-    vinserti128     m3, m3, xm0, 0
-
-    punpcklbw       m1, m3, m5
-    punpckhbw       m0, m3, m5
-    vperm2i128      m6, m1, m0, 0x20
-    vperm2i128      m0, m1, m0, 0x31
-    pmaddubsw       m6, m9
-    paddw           m2, m6
-    pmaddubsw       m0, m9
-    paddw           m4, m0
-
-%ifidn %1,pp
-    pmulhrsw        m4, m7                         ; m4 = word: row 14
-    pmulhrsw        m2, m7                         ; m2 = word: row 15
-    packuswb        m4, m2
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm2, m4, 1
-    movu            [r8 + r3 * 2], xm4
-    movu            [r8 + r6], xm2
-    add             r2, 16
-%else
-    psubw           m4, m7                         ; m4 = word: row 14
-    psubw           m2, m7                         ; m2 = word: row 15
-    movu            [r8 + r3 * 2], m4
-    movu            [r8 + r6], m2
-    add             r2, 32
-%endif
-    add             r0, 16
-    movq            xm1, [r0]                       ; m1 = row 0
-    movq            xm2, [r0 + r1]                  ; m2 = row 1
-    punpcklbw       xm1, xm2
-    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
-    punpcklbw       xm2, xm3
-    vinserti128     m5, m1, xm2, 1
-    pmaddubsw       m5, m8
-    movq            xm4, [r0 + r4]                  ; m4 = row 3
-    punpcklbw       xm3, xm4
-    lea             r7, [r0 + r1 * 4]
-    movq            xm1, [r7]                       ; m1 = row 4
-    punpcklbw       xm4, xm1
-    vinserti128     m2, m3, xm4, 1
-    pmaddubsw       m0, m2, m9
-    paddw           m5, m0
-    pmaddubsw       m2, m8
-    movq            xm3, [r7 + r1]                  ; m3 = row 5
-    punpcklbw       xm1, xm3
-    movq            xm4, [r7 + r1 * 2]              ; m4 = row 6
-    punpcklbw       xm3, xm4
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m0, m1, m9
-    paddw           m2, m0
-    pmaddubsw       m1, m8
-    movq            xm3, [r7 + r4]                  ; m3 = row 7
-    punpcklbw       xm4, xm3
-    lea             r7, [r7 + r1 * 4]
-    movq            xm0, [r7]                       ; m0 = row 8
-    punpcklbw       xm3, xm0
-    vinserti128     m4, m4, xm3, 1
-    pmaddubsw       m3, m4, m9
-    paddw           m1, m3
-    pmaddubsw       m4, m8
-    movq            xm3, [r7 + r1]                  ; m3 = row 9
-    punpcklbw       xm0, xm3
-    movq            xm6, [r7 + r1 * 2]              ; m6 = row 10
-    punpcklbw       xm3, xm6
-    vinserti128     m0, m0, xm3, 1
-    pmaddubsw       m3, m0, m9
-    paddw           m4, m3
-    pmaddubsw       m0, m8
-
-%ifidn %1,pp
-    pmulhrsw        m5, m7                          ; m5 = word: row 0, row 1
-    pmulhrsw        m2, m7                          ; m2 = word: row 2, row 3
-    pmulhrsw        m1, m7                          ; m1 = word: row 4, row 5
-    pmulhrsw        m4, m7                          ; m4 = word: row 6, row 7
-    packuswb        m5, m2
-    packuswb        m1, m4
-    vextracti128    xm2, m5, 1
-    vextracti128    xm4, m1, 1
-    movq            [r2], xm5
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm5
-    movhps          [r2 + r6], xm2
-    lea             r8, [r2 + r3 * 4]
-    movq            [r8], xm1
-    movq            [r8 + r3], xm4
-    movhps          [r8 + r3 * 2], xm1
-    movhps          [r8 + r6], xm4
-%else
-    psubw           m5, m7                          ; m5 = word: row 0, row 1
-    psubw           m2, m7                          ; m2 = word: row 2, row 3
-    psubw           m1, m7                          ; m1 = word: row 4, row 5
-    psubw           m4, m7                          ; m4 = word: row 6, row 7
-    vextracti128    xm3, m5, 1
-    movu            [r2], xm5
-    movu            [r2 + r3], xm3
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-    vextracti128    xm3, m1, 1
-    lea             r8, [r2 + r3 * 4]
-    movu            [r8], xm1
-    movu            [r8 + r3], xm3
-    vextracti128    xm3, m4, 1
-    movu            [r8 + r3 * 2], xm4
-    movu            [r8 + r6], xm3
-%endif
-    lea             r8, [r8 + r3 * 4]
-
-    movq            xm3, [r7 + r4]                  ; m3 = row 11
-    punpcklbw       xm6, xm3
-    lea             r7, [r7 + r1 * 4]
-    movq            xm5, [r7]                       ; m5 = row 12
-    punpcklbw       xm3, xm5
-    vinserti128     m6, m6, xm3, 1
-    pmaddubsw       m3, m6, m9
-    paddw           m0, m3
-    pmaddubsw       m6, m8
-    movq            xm3, [r7 + r1]                  ; m3 = row 13
-    punpcklbw       xm5, xm3
-    movq            xm2, [r7 + r1 * 2]              ; m2 = row 14
-    punpcklbw       xm3, xm2
-    vinserti128     m5, m5, xm3, 1
-    pmaddubsw       m3, m5, m9
-    paddw           m6, m3
-    pmaddubsw       m5, m8
-    movq            xm3, [r7 + r4]                  ; m3 = row 15
-    punpcklbw       xm2, xm3
-    lea             r7, [r7 + r1 * 4]
-    movq            xm1, [r7]                       ; m1 = row 16
-    punpcklbw       xm3, xm1
-    vinserti128     m2, m2, xm3, 1
-    pmaddubsw       m3, m2, m9
-    paddw           m5, m3
-    pmaddubsw       m2, m8
-    movq            xm3, [r7 + r1]                  ; m3 = row 17
-    punpcklbw       xm1, xm3
-    movq            xm4, [r7 + r1 * 2]              ; m4 = row 18
-    punpcklbw       xm3, xm4
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m3, m1, m9
-    paddw           m2, m3
-%ifidn %1,pp
-    pmulhrsw        m0, m7                          ; m0 = word: row 8, row 9
-    pmulhrsw        m6, m7                          ; m6 = word: row 10, row 11
-    pmulhrsw        m5, m7                          ; m5 = word: row 12, row 13
-    pmulhrsw        m2, m7                          ; m2 = word: row 14, row 15
-    packuswb        m0, m6
-    packuswb        m5, m2
-    vextracti128    xm6, m0, 1
-    vextracti128    xm2, m5, 1
-    movq            [r8], xm0
-    movq            [r8 + r3], xm6
-    movhps          [r8 + r3 * 2], xm0
-    movhps          [r8 + r6], xm6
-    lea             r8, [r8 + r3 * 4]
-    movq            [r8], xm5
-    movq            [r8 + r3], xm2
-    movhps          [r8 + r3 * 2], xm5
-    movhps          [r8 + r6], xm2
-    lea             r2, [r8 + r3 * 4 - 16]
-%else
-    psubw           m0, m7                          ; m0 = word: row 8, row 9
-    psubw           m6, m7                          ; m6 = word: row 10, row 11
-    psubw           m5, m7                          ; m5 = word: row 12, row 13
-    psubw           m2, m7                          ; m2 = word: row 14, row 15
-    vextracti128    xm3, m0, 1
-    movu            [r8], xm0
-    movu            [r8 + r3], xm3
-    vextracti128    xm3, m6, 1
-    movu            [r8 + r3 * 2], xm6
-    movu            [r8 + r6], xm3
-    vextracti128    xm3, m5, 1
-    lea             r8, [r8 + r3 * 4]
-    movu            [r8], xm5
-    movu            [r8 + r3], xm3
-    vextracti128    xm3, m2, 1
-    movu            [r8 + r3 * 2], xm2
-    movu            [r8 + r6], xm3
-    lea             r2, [r8 + r3 * 4 - 32]
-%endif
-    lea             r0, [r7 - 16]
-    dec             r5d
-    jnz             .loopH
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_24x32 pp
-    FILTER_VER_CHROMA_AVX2_24x32 ps
-
-%macro FILTER_VER_CHROMA_AVX2_24x64 1
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_24x64, 4, 7, 13
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    mova            m10, [r5]
-    mova            m11, [r5 + mmsize]
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m12, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m12, [pw_2000]
-%endif
-    lea             r5, [r3 * 3]
-    mov             r6d, 16
-.loopH:
-    movu            m0, [r0]                        ; m0 = row 0
-    movu            m1, [r0 + r1]                   ; m1 = row 1
-    punpcklbw       m2, m0, m1
-    punpckhbw       m3, m0, m1
-    pmaddubsw       m2, m10
-    pmaddubsw       m3, m10
-    movu            m0, [r0 + r1 * 2]               ; m0 = row 2
-    punpcklbw       m4, m1, m0
-    punpckhbw       m5, m1, m0
-    pmaddubsw       m4, m10
-    pmaddubsw       m5, m10
-    movu            m1, [r0 + r4]                   ; m1 = row 3
-    punpcklbw       m6, m0, m1
-    punpckhbw       m7, m0, m1
-    pmaddubsw       m8, m6, m11
-    pmaddubsw       m9, m7, m11
-    pmaddubsw       m6, m10
-    pmaddubsw       m7, m10
-    paddw           m2, m8
-    paddw           m3, m9
-%ifidn %1,pp
-    pmulhrsw        m2, m12
-    pmulhrsw        m3, m12
-    packuswb        m2, m3
-    movu            [r2], xm2
-    vextracti128    xm2, m2, 1
-    movq            [r2 + 16], xm2
-%else
-    psubw           m2, m12
-    psubw           m3, m12
-    vperm2i128      m0, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    movu            [r2], m0
-    movu            [r2 + mmsize], xm2
-%endif
-    lea             r0, [r0 + r1 * 4]
-    movu            m0, [r0]                        ; m0 = row 4
-    punpcklbw       m2, m1, m0
-    punpckhbw       m3, m1, m0
-    pmaddubsw       m8, m2, m11
-    pmaddubsw       m9, m3, m11
-    pmaddubsw       m2, m10
-    pmaddubsw       m3, m10
-    paddw           m4, m8
-    paddw           m5, m9
-%ifidn %1,pp
-    pmulhrsw        m4, m12
-    pmulhrsw        m5, m12
-    packuswb        m4, m5
-    movu            [r2 + r3], xm4
-    vextracti128    xm4, m4, 1
-    movq            [r2 + r3 + 16], xm4
-%else
-    psubw           m4, m12
-    psubw           m5, m12
-    vperm2i128      m1, m4, m5, 0x20
-    vperm2i128      m4, m4, m5, 0x31
-    movu            [r2 + r3], m1
-    movu            [r2 + r3 + mmsize], xm4
-%endif
-
-    movu            m1, [r0 + r1]                   ; m1 = row 5
-    punpcklbw       m4, m0, m1
-    punpckhbw       m5, m0, m1
-    pmaddubsw       m4, m11
-    pmaddubsw       m5, m11
-    paddw           m6, m4
-    paddw           m7, m5
-%ifidn %1,pp
-    pmulhrsw        m6, m12
-    pmulhrsw        m7, m12
-    packuswb        m6, m7
-    movu            [r2 + r3 * 2], xm6
-    vextracti128    xm6, m6, 1
-    movq            [r2 + r3 * 2 + 16], xm6
-%else
-    psubw           m6, m12
-    psubw           m7, m12
-    vperm2i128      m0, m6, m7, 0x20
-    vperm2i128      m6, m6, m7, 0x31
-    movu            [r2 + r3 * 2], m0
-    movu            [r2 + r3 * 2 + mmsize], xm6
-%endif
-
-    movu            m0, [r0 + r1 * 2]               ; m0 = row 6
-    punpcklbw       m6, m1, m0
-    punpckhbw       m7, m1, m0
-    pmaddubsw       m6, m11
-    pmaddubsw       m7, m11
-    paddw           m2, m6
-    paddw           m3, m7
-%ifidn %1,pp
-    pmulhrsw        m2, m12
-    pmulhrsw        m3, m12
-    packuswb        m2, m3
-    movu            [r2 + r5], xm2
-    vextracti128    xm2, m2, 1
-    movq            [r2 + r5 + 16], xm2
-%else
-    psubw           m2, m12
-    psubw           m3, m12
-    vperm2i128      m0, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    movu            [r2 + r5], m0
-    movu            [r2 + r5 + mmsize], xm2
-%endif
-    lea             r2, [r2 + r3 * 4]
-    dec             r6d
-    jnz             .loopH
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_24x64 pp
-    FILTER_VER_CHROMA_AVX2_24x64 ps
-
-%macro FILTER_VER_CHROMA_AVX2_16x4 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_16x4, 4, 6, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m7, [pw_512]
-%else
-    add             r3d, r3d
-    mova            m7, [pw_2000]
-%endif
-
-    movu            xm0, [r0]
-    vinserti128     m0, m0, [r0 + r1 * 2], 1
-    movu            xm1, [r0 + r1]
-    vinserti128     m1, m1, [r0 + r4], 1
-
-    punpcklbw       m2, m0, m1
-    punpckhbw       m3, m0, m1
-    vperm2i128      m4, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    pmaddubsw       m4, [r5]
-    pmaddubsw       m3, m2, [r5 + mmsize]
-    paddw           m4, m3
-    pmaddubsw       m2, [r5]
-
-    vextracti128    xm0, m0, 1
-    lea             r0, [r0 + r1 * 4]
-    vinserti128     m0, m0, [r0], 1
-
-    punpcklbw       m5, m1, m0
-    punpckhbw       m3, m1, m0
-    vperm2i128      m6, m5, m3, 0x20
-    vperm2i128      m5, m5, m3, 0x31
-    pmaddubsw       m6, [r5]
-    pmaddubsw       m3, m5, [r5 + mmsize]
-    paddw           m6, m3
-    pmaddubsw       m5, [r5]
-%ifidn %1,pp
-    pmulhrsw        m4, m7                          ; m4 = word: row 0
-    pmulhrsw        m6, m7                          ; m6 = word: row 1
-    packuswb        m4, m6
-    vpermq          m4, m4, 11011000b
-    vextracti128    xm6, m4, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm6
-%else
-    psubw           m4, m7                          ; m4 = word: row 0
-    psubw           m6, m7                          ; m6 = word: row 1
-    movu            [r2], m4
-    movu            [r2 + r3], m6
-%endif
-    lea             r2, [r2 + r3 * 2]
-
-    movu            xm4, [r0 + r1 * 2]
-    vinserti128     m4, m4, [r0 + r1], 1
-    vextracti128    xm1, m4, 1
-    vinserti128     m0, m0, xm1, 0
-
-    punpcklbw       m6, m0, m4
-    punpckhbw       m1, m0, m4
-    vperm2i128      m0, m6, m1, 0x20
-    vperm2i128      m6, m6, m1, 0x31
-    pmaddubsw       m0, [r5 + mmsize]
-    paddw           m5, m0
-    pmaddubsw       m6, [r5 + mmsize]
-    paddw           m2, m6
-
-%ifidn %1,pp
-    pmulhrsw        m2, m7                          ; m2 = word: row 2
-    pmulhrsw        m5, m7                          ; m5 = word: row 3
-    packuswb        m2, m5
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm5, m2, 1
-    movu            [r2], xm2
-    movu            [r2 + r3], xm5
-%else
-    psubw           m2, m7                          ; m2 = word: row 2
-    psubw           m5, m7                          ; m5 = word: row 3
-    movu            [r2], m2
-    movu            [r2 + r3], m5
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_16x4 pp
-    FILTER_VER_CHROMA_AVX2_16x4 ps
-
-%macro FILTER_VER_CHROMA_AVX2_12xN 2
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_12x%2, 4, 7, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m7, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m7, [pw_2000]
-%endif
-    lea             r6, [r3 * 3]
-%rep %2 / 16
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhbw       xm2, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddubsw       m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhbw       xm3, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhbw       xm4, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddubsw       m4, m2, [r5 + 1 * mmsize]
-    paddw           m0, m4
-    pmaddubsw       m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhbw       xm5, xm3, xm4
-    punpcklbw       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddubsw       m5, m3, [r5 + 1 * mmsize]
-    paddw           m1, m5
-    pmaddubsw       m3, [r5]
-%ifidn %1,pp
-    pmulhrsw        m0, m7                          ; m0 = word: row 0
-    pmulhrsw        m1, m7                          ; m1 = word: row 1
-    packuswb        m0, m1
-    vextracti128    xm1, m0, 1
-    movq            [r2], xm0
-    movd            [r2 + 8], xm1
-    movhps          [r2 + r3], xm0
-    pextrd          [r2 + r3 + 8], xm1, 2
-%else
-    psubw           m0, m7                          ; m0 = word: row 0
-    psubw           m1, m7                          ; m1 = word: row 1
-    movu            [r2], xm0
-    vextracti128    xm0, m0, 1
-    movq            [r2 + 16], xm0
-    movu            [r2 + r3], xm1
-    vextracti128    xm1, m1, 1
-    movq            [r2 + r3 + 16], xm1
-%endif
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhbw       xm6, xm4, xm5
-    punpcklbw       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddubsw       m6, m4, [r5 + 1 * mmsize]
-    paddw           m2, m6
-    pmaddubsw       m4, [r5]
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhbw       xm0, xm5, xm6
-    punpcklbw       xm5, xm6
-    vinserti128     m5, m5, xm0, 1
-    pmaddubsw       m0, m5, [r5 + 1 * mmsize]
-    paddw           m3, m0
-    pmaddubsw       m5, [r5]
-%ifidn %1,pp
-    pmulhrsw        m2, m7                          ; m2 = word: row 2
-    pmulhrsw        m3, m7                          ; m3 = word: row 3
-    packuswb        m2, m3
-    vextracti128    xm3, m2, 1
-    movq            [r2 + r3 * 2], xm2
-    movd            [r2 + r3 * 2 + 8], xm3
-    movhps          [r2 + r6], xm2
-    pextrd          [r2 + r6 + 8], xm3, 2
-%else
-    psubw           m2, m7                          ; m2 = word: row 2
-    psubw           m3, m7                          ; m3 = word: row 3
-    movu            [r2 + r3 * 2], xm2
-    vextracti128    xm2, m2, 1
-    movq            [r2 + r3 * 2 + 16], xm2
-    movu            [r2 + r6], xm3
-    vextracti128    xm3, m3, 1
-    movq            [r2 + r6 + 16], xm3
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm0, [r0 + r4]                  ; m0 = row 7
-    punpckhbw       xm3, xm6, xm0
-    punpcklbw       xm6, xm0
-    vinserti128     m6, m6, xm3, 1
-    pmaddubsw       m3, m6, [r5 + 1 * mmsize]
-    paddw           m4, m3
-    pmaddubsw       m6, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm3, [r0]                       ; m3 = row 8
-    punpckhbw       xm1, xm0, xm3
-    punpcklbw       xm0, xm3
-    vinserti128     m0, m0, xm1, 1
-    pmaddubsw       m1, m0, [r5 + 1 * mmsize]
-    paddw           m5, m1
-    pmaddubsw       m0, [r5]
-%ifidn %1,pp
-    pmulhrsw        m4, m7                          ; m4 = word: row 4
-    pmulhrsw        m5, m7                          ; m5 = word: row 5
-    packuswb        m4, m5
-    vextracti128    xm5, m4, 1
-    movq            [r2], xm4
-    movd            [r2 + 8], xm5
-    movhps          [r2 + r3], xm4
-    pextrd          [r2 + r3 + 8], xm5, 2
-%else
-    psubw           m4, m7                          ; m4 = word: row 4
-    psubw           m5, m7                          ; m5 = word: row 5
-    movu            [r2], xm4
-    vextracti128    xm4, m4, 1
-    movq            [r2 + 16], xm4
-    movu            [r2 + r3], xm5
-    vextracti128    xm5, m5, 1
-    movq            [r2 + r3 + 16], xm5
-%endif
-
-    movu            xm1, [r0 + r1]                  ; m1 = row 9
-    punpckhbw       xm2, xm3, xm1
-    punpcklbw       xm3, xm1
-    vinserti128     m3, m3, xm2, 1
-    pmaddubsw       m2, m3, [r5 + 1 * mmsize]
-    paddw           m6, m2
-    pmaddubsw       m3, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 10
-    punpckhbw       xm4, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm4, 1
-    pmaddubsw       m4, m1, [r5 + 1 * mmsize]
-    paddw           m0, m4
-    pmaddubsw       m1, [r5]
-
-%ifidn %1,pp
-    pmulhrsw        m6, m7                          ; m6 = word: row 6
-    pmulhrsw        m0, m7                          ; m0 = word: row 7
-    packuswb        m6, m0
-    vextracti128    xm0, m6, 1
-    movq            [r2 + r3 * 2], xm6
-    movd            [r2 + r3 * 2 + 8], xm0
-    movhps          [r2 + r6], xm6
-    pextrd          [r2 + r6 + 8], xm0, 2
-%else
-    psubw           m6, m7                          ; m6 = word: row 6
-    psubw           m0, m7                          ; m0 = word: row 7
-    movu            [r2 + r3 * 2], xm6
-    vextracti128    xm6, m6, 1
-    movq            [r2 + r3 * 2 + 16], xm6
-    movu            [r2 + r6], xm0
-    vextracti128    xm0, m0, 1
-    movq            [r2 + r6 + 16], xm0
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm4, [r0 + r4]                  ; m4 = row 11
-    punpckhbw       xm6, xm2, xm4
-    punpcklbw       xm2, xm4
-    vinserti128     m2, m2, xm6, 1
-    pmaddubsw       m6, m2, [r5 + 1 * mmsize]
-    paddw           m3, m6
-    pmaddubsw       m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm6, [r0]                       ; m6 = row 12
-    punpckhbw       xm0, xm4, xm6
-    punpcklbw       xm4, xm6
-    vinserti128     m4, m4, xm0, 1
-    pmaddubsw       m0, m4, [r5 + 1 * mmsize]
-    paddw           m1, m0
-    pmaddubsw       m4, [r5]
-%ifidn %1,pp
-    pmulhrsw        m3, m7                          ; m3 = word: row 8
-    pmulhrsw        m1, m7                          ; m1 = word: row 9
-    packuswb        m3, m1
-    vextracti128    xm1, m3, 1
-    movq            [r2], xm3
-    movd            [r2 + 8], xm1
-    movhps          [r2 + r3], xm3
-    pextrd          [r2 + r3 + 8], xm1, 2
-%else
-    psubw           m3, m7                          ; m3 = word: row 8
-    psubw           m1, m7                          ; m1 = word: row 9
-    movu            [r2], xm3
-    vextracti128    xm3, m3, 1
-    movq            [r2 + 16], xm3
-    movu            [r2 + r3], xm1
-    vextracti128    xm1, m1, 1
-    movq            [r2 + r3 + 16], xm1
-%endif
-
-    movu            xm0, [r0 + r1]                  ; m0 = row 13
-    punpckhbw       xm1, xm6, xm0
-    punpcklbw       xm6, xm0
-    vinserti128     m6, m6, xm1, 1
-    pmaddubsw       m1, m6, [r5 + 1 * mmsize]
-    paddw           m2, m1
-    pmaddubsw       m6, [r5]
-    movu            xm1, [r0 + r1 * 2]              ; m1 = row 14
-    punpckhbw       xm5, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm5, 1
-    pmaddubsw       m5, m0, [r5 + 1 * mmsize]
-    paddw           m4, m5
-    pmaddubsw       m0, [r5]
-%ifidn %1,pp
-    pmulhrsw        m2, m7                          ; m2 = word: row 10
-    pmulhrsw        m4, m7                          ; m4 = word: row 11
-    packuswb        m2, m4
-    vextracti128    xm4, m2, 1
-    movq            [r2 + r3 * 2], xm2
-    movd            [r2 + r3 * 2 + 8], xm4
-    movhps          [r2 + r6], xm2
-    pextrd          [r2 + r6 + 8], xm4, 2
-%else
-    psubw           m2, m7                          ; m2 = word: row 10
-    psubw           m4, m7                          ; m4 = word: row 11
-    movu            [r2 + r3 * 2], xm2
-    vextracti128    xm2, m2, 1
-    movq            [r2 + r3 * 2 + 16], xm2
-    movu            [r2 + r6], xm4
-    vextracti128    xm4, m4, 1
-    movq            [r2 + r6 + 16], xm4
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm5, [r0 + r4]                  ; m5 = row 15
-    punpckhbw       xm2, xm1, xm5
-    punpcklbw       xm1, xm5
-    vinserti128     m1, m1, xm2, 1
-    pmaddubsw       m2, m1, [r5 + 1 * mmsize]
-    paddw           m6, m2
-    pmaddubsw       m1, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm2, [r0]                       ; m2 = row 16
-    punpckhbw       xm3, xm5, xm2
-    punpcklbw       xm5, xm2
-    vinserti128     m5, m5, xm3, 1
-    pmaddubsw       m3, m5, [r5 + 1 * mmsize]
-    paddw           m0, m3
-    pmaddubsw       m5, [r5]
-    movu            xm3, [r0 + r1]                  ; m3 = row 17
-    punpckhbw       xm4, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddubsw       m2, [r5 + 1 * mmsize]
-    paddw           m1, m2
-    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
-    punpckhbw       xm2, xm3, xm4
-    punpcklbw       xm3, xm4
-    vinserti128     m3, m3, xm2, 1
-    pmaddubsw       m3, [r5 + 1 * mmsize]
-    paddw           m5, m3
-
-%ifidn %1,pp
-    pmulhrsw        m6, m7                          ; m6 = word: row 12
-    pmulhrsw        m0, m7                          ; m0 = word: row 13
-    pmulhrsw        m1, m7                          ; m1 = word: row 14
-    pmulhrsw        m5, m7                          ; m5 = word: row 15
-    packuswb        m6, m0
-    packuswb        m1, m5
-    vextracti128    xm0, m6, 1
-    vextracti128    xm5, m1, 1
-    movq            [r2], xm6
-    movd            [r2 + 8], xm0
-    movhps          [r2 + r3], xm6
-    pextrd          [r2 + r3 + 8], xm0, 2
-    movq            [r2 + r3 * 2], xm1
-    movd            [r2 + r3 * 2 + 8], xm5
-    movhps          [r2 + r6], xm1
-    pextrd          [r2 + r6 + 8], xm5, 2
-%else
-    psubw           m6, m7                          ; m6 = word: row 12
-    psubw           m0, m7                          ; m0 = word: row 13
-    psubw           m1, m7                          ; m1 = word: row 14
-    psubw           m5, m7                          ; m5 = word: row 15
-    movu            [r2], xm6
-    vextracti128    xm6, m6, 1
-    movq            [r2 + 16], xm6
-    movu            [r2 + r3], xm0
-    vextracti128    xm0, m0, 1
-    movq            [r2 + r3 + 16], xm0
-    movu            [r2 + r3 * 2], xm1
-    vextracti128    xm1, m1, 1
-    movq            [r2 + r3 * 2 + 16], xm1
-    movu            [r2 + r6], xm5
-    vextracti128    xm5, m5, 1
-    movq            [r2 + r6 + 16], xm5
-%endif
-    lea             r2, [r2 + r3 * 4]
-%endrep
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_12xN pp, 16
-    FILTER_VER_CHROMA_AVX2_12xN ps, 16
-    FILTER_VER_CHROMA_AVX2_12xN pp, 32
-    FILTER_VER_CHROMA_AVX2_12xN ps, 32
-
-;-----------------------------------------------------------------------------
-;void interp_4tap_vert_pp_24x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W24 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_24x%2, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m1,        m0,       [tab_Vm]
-    pshufb      m0,        [tab_Vm + 16]
-
-    mov         r4d,       %2
-
-.loop:
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m2,        m1
-
-    lea         r5,        [r0 + 2 * r1]
-    movu        m5,        [r5]
-    movu        m7,        [r5 + r1]
-
-    punpcklbw   m6,        m5,        m7
-    pmaddubsw   m6,        m0
-    paddw       m4,        m6
-
-    punpckhbw   m6,        m5,        m7
-    pmaddubsw   m6,        m0
-    paddw       m2,        m6
-
-    mova        m6,        [pw_512]
-
-    pmulhrsw    m4,        m6
-    pmulhrsw    m2,        m6
-
-    packuswb    m4,        m2
-
-    movu        [r2],      m4
-
-    punpcklbw   m4,        m3,        m5
-    punpckhbw   m3,        m5
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m3,        m1
-
-    movu        m2,        [r5 + 2 * r1]
-
-    punpcklbw   m5,        m7,        m2
-    punpckhbw   m7,        m2
-
-    pmaddubsw   m5,        m0
-    pmaddubsw   m7,        m0
-
-    paddw       m4,        m5
-    paddw       m3,        m7
-
-    pmulhrsw    m4,        m6
-    pmulhrsw    m3,        m6
-
-    packuswb    m4,        m3
-
-    movu        [r2 + r3],      m4
-
-    movq        m2,        [r0 + 16]
-    movq        m3,        [r0 + r1 + 16]
-    movq        m4,        [r5 + 16]
-    movq        m5,        [r5 + r1 + 16]
-
-    punpcklbw   m2,        m3
-    punpcklbw   m4,        m5
-
-    pmaddubsw   m2,        m1
-    pmaddubsw   m4,        m0
-
-    paddw       m2,        m4
-
-    pmulhrsw    m2,        m6
-
-    movq        m3,        [r0 + r1 + 16]
-    movq        m4,        [r5 + 16]
-    movq        m5,        [r5 + r1 + 16]
-    movq        m7,        [r5 + 2 * r1 + 16]
-
-    punpcklbw   m3,        m4
-    punpcklbw   m5,        m7
-
-    pmaddubsw   m3,        m1
-    pmaddubsw   m5,        m0
-
-    paddw       m3,        m5
-
-    pmulhrsw    m3,        m6
-    packuswb    m2,        m3
-
-    movh        [r2 + 16], m2
-    movhps      [r2 + r3 + 16], m2
-
-    mov         r0,        r5
-    lea         r2,        [r2 + 2 * r3]
-
-    sub         r4,        2
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W24 24, 32
-
-    FILTER_V4_W24 24, 64
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W32 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m1,        m0,       [tab_Vm]
-    pshufb      m0,        [tab_Vm + 16]
-
-    mova        m7,        [pw_512]
-
-    mov         r4d,       %2
-
-.loop:
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m2,        m1
-
-    lea         r5,        [r0 + 2 * r1]
-    movu        m3,        [r5]
-    movu        m5,        [r5 + r1]
-
-    punpcklbw   m6,        m3,        m5
-    punpckhbw   m3,        m5
-
-    pmaddubsw   m6,        m0
-    pmaddubsw   m3,        m0
-
-    paddw       m4,        m6
-    paddw       m2,        m3
-
-    pmulhrsw    m4,        m7
-    pmulhrsw    m2,        m7
-
-    packuswb    m4,        m2
-
-    movu        [r2],      m4
-
-    movu        m2,        [r0 + 16]
-    movu        m3,        [r0 + r1 + 16]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m2,        m1
-
-    movu        m3,        [r5 + 16]
-    movu        m5,        [r5 + r1 + 16]
-
-    punpcklbw   m6,        m3,        m5
-    punpckhbw   m3,        m5
-
-    pmaddubsw   m6,        m0
-    pmaddubsw   m3,        m0
-
-    paddw       m4,        m6
-    paddw       m2,        m3
-
-    pmulhrsw    m4,        m7
-    pmulhrsw    m2,        m7
-
-    packuswb    m4,        m2
-
-    movu        [r2 + 16], m4
-
-    lea         r0,        [r0 + r1]
-    lea         r2,        [r2 + r3]
-
-    dec         r4
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W32 32,  8
-    FILTER_V4_W32 32, 16
-    FILTER_V4_W32 32, 24
-    FILTER_V4_W32 32, 32
-
-    FILTER_V4_W32 32, 48
-    FILTER_V4_W32 32, 64
-
-%macro FILTER_VER_CHROMA_AVX2_32xN 2
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_32x%2, 4, 7, 13
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    mova            m10, [r5]
-    mova            m11, [r5 + mmsize]
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m12, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m12, [pw_2000]
-%endif
-    lea             r5, [r3 * 3]
-    mov             r6d, %2 / 4
-.loopW:
-    movu            m0, [r0]                        ; m0 = row 0
-    movu            m1, [r0 + r1]                   ; m1 = row 1
-    punpcklbw       m2, m0, m1
-    punpckhbw       m3, m0, m1
-    pmaddubsw       m2, m10
-    pmaddubsw       m3, m10
-    movu            m0, [r0 + r1 * 2]               ; m0 = row 2
-    punpcklbw       m4, m1, m0
-    punpckhbw       m5, m1, m0
-    pmaddubsw       m4, m10
-    pmaddubsw       m5, m10
-    movu            m1, [r0 + r4]                   ; m1 = row 3
-    punpcklbw       m6, m0, m1
-    punpckhbw       m7, m0, m1
-    pmaddubsw       m8, m6, m11
-    pmaddubsw       m9, m7, m11
-    pmaddubsw       m6, m10
-    pmaddubsw       m7, m10
-    paddw           m2, m8
-    paddw           m3, m9
-%ifidn %1,pp
-    pmulhrsw        m2, m12
-    pmulhrsw        m3, m12
-    packuswb        m2, m3
-    movu            [r2], m2
-%else
-    psubw           m2, m12
-    psubw           m3, m12
-    vperm2i128      m0, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    movu            [r2], m0
-    movu            [r2 + mmsize], m2
-%endif
-    lea             r0, [r0 + r1 * 4]
-    movu            m0, [r0]                        ; m0 = row 4
-    punpcklbw       m2, m1, m0
-    punpckhbw       m3, m1, m0
-    pmaddubsw       m8, m2, m11
-    pmaddubsw       m9, m3, m11
-    pmaddubsw       m2, m10
-    pmaddubsw       m3, m10
-    paddw           m4, m8
-    paddw           m5, m9
-%ifidn %1,pp
-    pmulhrsw        m4, m12
-    pmulhrsw        m5, m12
-    packuswb        m4, m5
-    movu            [r2 + r3], m4
-%else
-    psubw           m4, m12
-    psubw           m5, m12
-    vperm2i128      m1, m4, m5, 0x20
-    vperm2i128      m4, m4, m5, 0x31
-    movu            [r2 + r3], m1
-    movu            [r2 + r3 + mmsize], m4
-%endif
-
-    movu            m1, [r0 + r1]                   ; m1 = row 5
-    punpcklbw       m4, m0, m1
-    punpckhbw       m5, m0, m1
-    pmaddubsw       m4, m11
-    pmaddubsw       m5, m11
-    paddw           m6, m4
-    paddw           m7, m5
-%ifidn %1,pp
-    pmulhrsw        m6, m12
-    pmulhrsw        m7, m12
-    packuswb        m6, m7
-    movu            [r2 + r3 * 2], m6
-%else
-    psubw           m6, m12
-    psubw           m7, m12
-    vperm2i128      m0, m6, m7, 0x20
-    vperm2i128      m6, m6, m7, 0x31
-    movu            [r2 + r3 * 2], m0
-    movu            [r2 + r3 * 2 + mmsize], m6
-%endif
-
-    movu            m0, [r0 + r1 * 2]               ; m0 = row 6
-    punpcklbw       m6, m1, m0
-    punpckhbw       m7, m1, m0
-    pmaddubsw       m6, m11
-    pmaddubsw       m7, m11
-    paddw           m2, m6
-    paddw           m3, m7
-%ifidn %1,pp
-    pmulhrsw        m2, m12
-    pmulhrsw        m3, m12
-    packuswb        m2, m3
-    movu            [r2 + r5], m2
-%else
-    psubw           m2, m12
-    psubw           m3, m12
-    vperm2i128      m0, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    movu            [r2 + r5], m0
-    movu            [r2 + r5 + mmsize], m2
-%endif
-    lea             r2, [r2 + r3 * 4]
-    dec             r6d
-    jnz             .loopW
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_32xN pp, 64
-    FILTER_VER_CHROMA_AVX2_32xN pp, 48
-    FILTER_VER_CHROMA_AVX2_32xN pp, 32
-    FILTER_VER_CHROMA_AVX2_32xN pp, 24
-    FILTER_VER_CHROMA_AVX2_32xN pp, 16
-    FILTER_VER_CHROMA_AVX2_32xN pp, 8
-    FILTER_VER_CHROMA_AVX2_32xN ps, 64
-    FILTER_VER_CHROMA_AVX2_32xN ps, 48
-    FILTER_VER_CHROMA_AVX2_32xN ps, 32
-    FILTER_VER_CHROMA_AVX2_32xN ps, 24
-    FILTER_VER_CHROMA_AVX2_32xN ps, 16
-    FILTER_VER_CHROMA_AVX2_32xN ps, 8
-
-%macro FILTER_VER_CHROMA_AVX2_48x64 1
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_48x64, 4, 8, 13
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    mova            m10, [r5]
-    mova            m11, [r5 + mmsize]
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m12, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m12, [pw_2000]
-%endif
-    lea             r5, [r3 * 3]
-    lea             r7, [r1 * 4]
-    mov             r6d, 16
-.loopH:
-    movu            m0, [r0]                        ; m0 = row 0
-    movu            m1, [r0 + r1]                   ; m1 = row 1
-    punpcklbw       m2, m0, m1
-    punpckhbw       m3, m0, m1
-    pmaddubsw       m2, m10
-    pmaddubsw       m3, m10
-    movu            m0, [r0 + r1 * 2]               ; m0 = row 2
-    punpcklbw       m4, m1, m0
-    punpckhbw       m5, m1, m0
-    pmaddubsw       m4, m10
-    pmaddubsw       m5, m10
-    movu            m1, [r0 + r4]                   ; m1 = row 3
-    punpcklbw       m6, m0, m1
-    punpckhbw       m7, m0, m1
-    pmaddubsw       m8, m6, m11
-    pmaddubsw       m9, m7, m11
-    pmaddubsw       m6, m10
-    pmaddubsw       m7, m10
-    paddw           m2, m8
-    paddw           m3, m9
-%ifidn %1,pp
-    pmulhrsw        m2, m12
-    pmulhrsw        m3, m12
-    packuswb        m2, m3
-    movu            [r2], m2
-%else
-    psubw           m2, m12
-    psubw           m3, m12
-    vperm2i128      m0, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    movu            [r2], m0
-    movu            [r2 + mmsize], m2
-%endif
-    lea             r0, [r0 + r1 * 4]
-    movu            m0, [r0]                        ; m0 = row 4
-    punpcklbw       m2, m1, m0
-    punpckhbw       m3, m1, m0
-    pmaddubsw       m8, m2, m11
-    pmaddubsw       m9, m3, m11
-    pmaddubsw       m2, m10
-    pmaddubsw       m3, m10
-    paddw           m4, m8
-    paddw           m5, m9
-%ifidn %1,pp
-    pmulhrsw        m4, m12
-    pmulhrsw        m5, m12
-    packuswb        m4, m5
-    movu            [r2 + r3], m4
-%else
-    psubw           m4, m12
-    psubw           m5, m12
-    vperm2i128      m1, m4, m5, 0x20
-    vperm2i128      m4, m4, m5, 0x31
-    movu            [r2 + r3], m1
-    movu            [r2 + r3 + mmsize], m4
-%endif
-
-    movu            m1, [r0 + r1]                   ; m1 = row 5
-    punpcklbw       m4, m0, m1
-    punpckhbw       m5, m0, m1
-    pmaddubsw       m4, m11
-    pmaddubsw       m5, m11
-    paddw           m6, m4
-    paddw           m7, m5
-%ifidn %1,pp
-    pmulhrsw        m6, m12
-    pmulhrsw        m7, m12
-    packuswb        m6, m7
-    movu            [r2 + r3 * 2], m6
-%else
-    psubw           m6, m12
-    psubw           m7, m12
-    vperm2i128      m0, m6, m7, 0x20
-    vperm2i128      m6, m6, m7, 0x31
-    movu            [r2 + r3 * 2], m0
-    movu            [r2 + r3 * 2 + mmsize], m6
-%endif
-
-    movu            m0, [r0 + r1 * 2]               ; m0 = row 6
-    punpcklbw       m6, m1, m0
-    punpckhbw       m7, m1, m0
-    pmaddubsw       m6, m11
-    pmaddubsw       m7, m11
-    paddw           m2, m6
-    paddw           m3, m7
-%ifidn %1,pp
-    pmulhrsw        m2, m12
-    pmulhrsw        m3, m12
-    packuswb        m2, m3
-    movu            [r2 + r5], m2
-    add             r2, 32
-%else
-    psubw           m2, m12
-    psubw           m3, m12
-    vperm2i128      m0, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    movu            [r2 + r5], m0
-    movu            [r2 + r5 + mmsize], m2
-    add             r2, 64
-%endif
-    sub             r0, r7
-
-    movu            xm0, [r0 + 32]                  ; m0 = row 0
-    movu            xm1, [r0 + r1 + 32]             ; m1 = row 1
-    punpckhbw       xm2, xm0, xm1
-    punpcklbw       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddubsw       m0, m10
-    movu            xm2, [r0 + r1 * 2 + 32]         ; m2 = row 2
-    punpckhbw       xm3, xm1, xm2
-    punpcklbw       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddubsw       m1, m10
-    movu            xm3, [r0 + r4 + 32]             ; m3 = row 3
-    punpckhbw       xm4, xm2, xm3
-    punpcklbw       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddubsw       m4, m2, m11
-    paddw           m0, m4
-    pmaddubsw       m2, m10
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0 + 32]                  ; m4 = row 4
-    punpckhbw       xm5, xm3, xm4
-    punpcklbw       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddubsw       m5, m3, m11
-    paddw           m1, m5
-    pmaddubsw       m3, m10
-    movu            xm5, [r0 + r1 + 32]             ; m5 = row 5
-    punpckhbw       xm6, xm4, xm5
-    punpcklbw       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddubsw       m4, m11
-    paddw           m2, m4
-    movu            xm6, [r0 + r1 * 2 + 32]         ; m6 = row 6
-    punpckhbw       xm7, xm5, xm6
-    punpcklbw       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddubsw       m5, m11
-    paddw           m3, m5
-%ifidn %1,pp
-    pmulhrsw        m0, m12                         ; m0 = word: row 0
-    pmulhrsw        m1, m12                         ; m1 = word: row 1
-    pmulhrsw        m2, m12                         ; m2 = word: row 2
-    pmulhrsw        m3, m12                         ; m3 = word: row 3
-    packuswb        m0, m1
-    packuswb        m2, m3
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r5], xm3
-    lea             r2, [r2 + r3 * 4 - 32]
-%else
-    psubw           m0, m12                         ; m0 = word: row 0
-    psubw           m1, m12                         ; m1 = word: row 1
-    psubw           m2, m12                         ; m2 = word: row 2
-    psubw           m3, m12                         ; m3 = word: row 3
-    movu            [r2], m0
-    movu            [r2 + r3], m1
-    movu            [r2 + r3 * 2], m2
-    movu            [r2 + r5], m3
-    lea             r2, [r2 + r3 * 4 - 64]
-%endif
-    dec             r6d
-    jnz             .loopH
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_48x64 pp
-    FILTER_VER_CHROMA_AVX2_48x64 ps
-
-%macro FILTER_VER_CHROMA_AVX2_64xN 2
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_64x%2, 4, 8, 13
-    mov             r4d, r4m
-    shl             r4d, 6
-
-%ifdef PIC
-    lea             r5, [tab_ChromaCoeffVer_32]
-    add             r5, r4
-%else
-    lea             r5, [tab_ChromaCoeffVer_32 + r4]
-%endif
-
-    mova            m10, [r5]
-    mova            m11, [r5 + mmsize]
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,pp
-    mova            m12, [pw_512]
-%else
-    add             r3d, r3d
-    vbroadcasti128  m12, [pw_2000]
-%endif
-    lea             r5, [r3 * 3]
-    lea             r7, [r1 * 4]
-    mov             r6d, %2 / 4
-.loopH:
-%assign x 0
-%rep 2
-    movu            m0, [r0 + x]                    ; m0 = row 0
-    movu            m1, [r0 + r1 + x]               ; m1 = row 1
-    punpcklbw       m2, m0, m1
-    punpckhbw       m3, m0, m1
-    pmaddubsw       m2, m10
-    pmaddubsw       m3, m10
-    movu            m0, [r0 + r1 * 2 + x]           ; m0 = row 2
-    punpcklbw       m4, m1, m0
-    punpckhbw       m5, m1, m0
-    pmaddubsw       m4, m10
-    pmaddubsw       m5, m10
-    movu            m1, [r0 + r4 + x]               ; m1 = row 3
-    punpcklbw       m6, m0, m1
-    punpckhbw       m7, m0, m1
-    pmaddubsw       m8, m6, m11
-    pmaddubsw       m9, m7, m11
-    pmaddubsw       m6, m10
-    pmaddubsw       m7, m10
-    paddw           m2, m8
-    paddw           m3, m9
-%ifidn %1,pp
-    pmulhrsw        m2, m12
-    pmulhrsw        m3, m12
-    packuswb        m2, m3
-    movu            [r2], m2
-%else
-    psubw           m2, m12
-    psubw           m3, m12
-    vperm2i128      m0, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    movu            [r2], m0
-    movu            [r2 + mmsize], m2
-%endif
-    lea             r0, [r0 + r1 * 4]
-    movu            m0, [r0 + x]                    ; m0 = row 4
-    punpcklbw       m2, m1, m0
-    punpckhbw       m3, m1, m0
-    pmaddubsw       m8, m2, m11
-    pmaddubsw       m9, m3, m11
-    pmaddubsw       m2, m10
-    pmaddubsw       m3, m10
-    paddw           m4, m8
-    paddw           m5, m9
-%ifidn %1,pp
-    pmulhrsw        m4, m12
-    pmulhrsw        m5, m12
-    packuswb        m4, m5
-    movu            [r2 + r3], m4
-%else
-    psubw           m4, m12
-    psubw           m5, m12
-    vperm2i128      m1, m4, m5, 0x20
-    vperm2i128      m4, m4, m5, 0x31
-    movu            [r2 + r3], m1
-    movu            [r2 + r3 + mmsize], m4
-%endif
-
-    movu            m1, [r0 + r1 + x]               ; m1 = row 5
-    punpcklbw       m4, m0, m1
-    punpckhbw       m5, m0, m1
-    pmaddubsw       m4, m11
-    pmaddubsw       m5, m11
-    paddw           m6, m4
-    paddw           m7, m5
-%ifidn %1,pp
-    pmulhrsw        m6, m12
-    pmulhrsw        m7, m12
-    packuswb        m6, m7
-    movu            [r2 + r3 * 2], m6
-%else
-    psubw           m6, m12
-    psubw           m7, m12
-    vperm2i128      m0, m6, m7, 0x20
-    vperm2i128      m6, m6, m7, 0x31
-    movu            [r2 + r3 * 2], m0
-    movu            [r2 + r3 * 2 + mmsize], m6
-%endif
-
-    movu            m0, [r0 + r1 * 2 + x]           ; m0 = row 6
-    punpcklbw       m6, m1, m0
-    punpckhbw       m7, m1, m0
-    pmaddubsw       m6, m11
-    pmaddubsw       m7, m11
-    paddw           m2, m6
-    paddw           m3, m7
-%ifidn %1,pp
-    pmulhrsw        m2, m12
-    pmulhrsw        m3, m12
-    packuswb        m2, m3
-    movu            [r2 + r5], m2
-    add             r2, 32
-%else
-    psubw           m2, m12
-    psubw           m3, m12
-    vperm2i128      m0, m2, m3, 0x20
-    vperm2i128      m2, m2, m3, 0x31
-    movu            [r2 + r5], m0
-    movu            [r2 + r5 + mmsize], m2
-    add             r2, 64
-%endif
-    sub             r0, r7
-%assign x x+32
-%endrep
-%ifidn %1,pp
-    lea             r2, [r2 + r3 * 4 - 64]
-%else
-    lea             r2, [r2 + r3 * 4 - 128]
-%endif
-    add             r0, r7
-    dec             r6d
-    jnz             .loopH
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_AVX2_64xN pp, 64
-    FILTER_VER_CHROMA_AVX2_64xN pp, 48
-    FILTER_VER_CHROMA_AVX2_64xN pp, 32
-    FILTER_VER_CHROMA_AVX2_64xN pp, 16
-    FILTER_VER_CHROMA_AVX2_64xN ps, 64
-    FILTER_VER_CHROMA_AVX2_64xN ps, 48
-    FILTER_VER_CHROMA_AVX2_64xN ps, 32
-    FILTER_VER_CHROMA_AVX2_64xN ps, 16
-
-;-----------------------------------------------------------------------------
-; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------
-%macro FILTER_V4_W16n_H2 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_pp_%1x%2, 4, 7, 8
-
-    mov         r4d,       r4m
-    sub         r0,        r1
-
-%ifdef PIC
-    lea         r5,        [tab_ChromaCoeff]
-    movd        m0,        [r5 + r4 * 4]
-%else
-    movd        m0,        [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m1,        m0,       [tab_Vm]
-    pshufb      m0,        [tab_Vm + 16]
-
-    mov         r4d,       %2/2
-
-.loop:
-
-    mov         r6d,       %1/16
-
-.loopW:
-
-    movu        m2,        [r0]
-    movu        m3,        [r0 + r1]
-
-    punpcklbw   m4,        m2,        m3
-    punpckhbw   m2,        m3
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m2,        m1
-
-    lea         r5,        [r0 + 2 * r1]
-    movu        m5,        [r5]
-    movu        m6,        [r5 + r1]
-
-    punpckhbw   m7,        m5,        m6
-    pmaddubsw   m7,        m0
-    paddw       m2,        m7
-
-    punpcklbw   m7,        m5,        m6
-    pmaddubsw   m7,        m0
-    paddw       m4,        m7
-
-    mova        m7,        [pw_512]
-
-    pmulhrsw    m4,        m7
-    pmulhrsw    m2,        m7
-
-    packuswb    m4,        m2
-
-    movu        [r2],      m4
-
-    punpcklbw   m4,        m3,        m5
-    punpckhbw   m3,        m5
-
-    pmaddubsw   m4,        m1
-    pmaddubsw   m3,        m1
-
-    movu        m5,        [r5 + 2 * r1]
-
-    punpcklbw   m2,        m6,        m5
-    punpckhbw   m6,        m5
-
-    pmaddubsw   m2,        m0
-    pmaddubsw   m6,        m0
-
-    paddw       m4,        m2
-    paddw       m3,        m6
-
-    pmulhrsw    m4,        m7
-    pmulhrsw    m3,        m7
-
-    packuswb    m4,        m3
-
-    movu        [r2 + r3],      m4
-
-    add         r0,        16
-    add         r2,        16
-    dec         r6d
-    jnz         .loopW
-
-    lea         r0,        [r0 + r1 * 2 - %1]
-    lea         r2,        [r2 + r3 * 2 - %1]
-
-    dec         r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V4_W16n_H2 64, 64
-    FILTER_V4_W16n_H2 64, 32
-    FILTER_V4_W16n_H2 64, 48
-    FILTER_V4_W16n_H2 48, 64
-    FILTER_V4_W16n_H2 64, 16
-
-;-----------------------------------------------------------------------------
 ; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
 ;-----------------------------------------------------------------------------
 %macro P2S_H_2xN 1
@@ -12374,7 +1145,7 @@
     mova        m4, [pb_128]
     mova        m5, [tab_c_64_n64]
 
-.loop
+.loop:
     movh        m0, [r0]
     punpcklbw   m0, m4
     pmaddubsw   m0, m5
@@ -18020,4888 +6791,6 @@
 
     RET
 
-%macro PROCESS_CHROMA_SP_W4_4R 0
-    movq       m0, [r0]
-    movq       m1, [r0 + r1]
-    punpcklwd  m0, m1                          ;m0=[0 1]
-    pmaddwd    m0, [r6 + 0 *16]                ;m0=[0+1]         Row1
-
-    lea        r0, [r0 + 2 * r1]
-    movq       m4, [r0]
-    punpcklwd  m1, m4                          ;m1=[1 2]
-    pmaddwd    m1, [r6 + 0 *16]                ;m1=[1+2]         Row2
-
-    movq       m5, [r0 + r1]
-    punpcklwd  m4, m5                          ;m4=[2 3]
-    pmaddwd    m2, m4, [r6 + 0 *16]            ;m2=[2+3]         Row3
-    pmaddwd    m4, [r6 + 1 * 16]
-    paddd      m0, m4                          ;m0=[0+1+2+3]     Row1 done
-
-    lea        r0, [r0 + 2 * r1]
-    movq       m4, [r0]
-    punpcklwd  m5, m4                          ;m5=[3 4]
-    pmaddwd    m3, m5, [r6 + 0 *16]            ;m3=[3+4]         Row4
-    pmaddwd    m5, [r6 + 1 * 16]
-    paddd      m1, m5                          ;m1 = [1+2+3+4]   Row2
-
-    movq       m5, [r0 + r1]
-    punpcklwd  m4, m5                          ;m4=[4 5]
-    pmaddwd    m4, [r6 + 1 * 16]
-    paddd      m2, m4                          ;m2=[2+3+4+5]     Row3
-
-    movq       m4, [r0 + 2 * r1]
-    punpcklwd  m5, m4                          ;m5=[5 6]
-    pmaddwd    m5, [r6 + 1 * 16]
-    paddd      m3, m5                          ;m3=[3+4+5+6]     Row4
-%endmacro
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_sp_%1x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SP 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_sp_%1x%2, 5, 7, 7 ,0-gprsize
-
-    add       r1d, r1d
-    sub       r0, r1
-    shl       r4d, 5
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r6, [r5 + r4]
-%else
-    lea       r6, [tab_ChromaCoeffV + r4]
-%endif
-
-    mova      m6, [pd_526336]
-
-    mov       dword [rsp], %2/4
-
-.loopH:
-    mov       r4d, (%1/4)
-.loopW:
-    PROCESS_CHROMA_SP_W4_4R
-
-    paddd     m0, m6
-    paddd     m1, m6
-    paddd     m2, m6
-    paddd     m3, m6
-
-    psrad     m0, 12
-    psrad     m1, 12
-    psrad     m2, 12
-    psrad     m3, 12
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-
-    packuswb  m0, m2
-
-    movd      [r2], m0
-    pextrd    [r2 + r3], m0, 1
-    lea       r5, [r2 + 2 * r3]
-    pextrd    [r5], m0, 2
-    pextrd    [r5 + r3], m0, 3
-
-    lea       r5, [4 * r1 - 2 * 4]
-    sub       r0, r5
-    add       r2, 4
-
-    dec       r4d
-    jnz       .loopW
-
-    lea       r0, [r0 + 4 * r1 - 2 * %1]
-    lea       r2, [r2 + 4 * r3 - %1]
-
-    dec       dword [rsp]
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SP 4, 4
-    FILTER_VER_CHROMA_SP 4, 8
-    FILTER_VER_CHROMA_SP 16, 16
-    FILTER_VER_CHROMA_SP 16, 8
-    FILTER_VER_CHROMA_SP 16, 12
-    FILTER_VER_CHROMA_SP 12, 16
-    FILTER_VER_CHROMA_SP 16, 4
-    FILTER_VER_CHROMA_SP 4, 16
-    FILTER_VER_CHROMA_SP 32, 32
-    FILTER_VER_CHROMA_SP 32, 16
-    FILTER_VER_CHROMA_SP 16, 32
-    FILTER_VER_CHROMA_SP 32, 24
-    FILTER_VER_CHROMA_SP 24, 32
-    FILTER_VER_CHROMA_SP 32, 8
-
-    FILTER_VER_CHROMA_SP 16, 24
-    FILTER_VER_CHROMA_SP 16, 64
-    FILTER_VER_CHROMA_SP 12, 32
-    FILTER_VER_CHROMA_SP 4, 32
-    FILTER_VER_CHROMA_SP 32, 64
-    FILTER_VER_CHROMA_SP 32, 48
-    FILTER_VER_CHROMA_SP 24, 64
-
-    FILTER_VER_CHROMA_SP 64, 64
-    FILTER_VER_CHROMA_SP 64, 32
-    FILTER_VER_CHROMA_SP 64, 48
-    FILTER_VER_CHROMA_SP 48, 64
-    FILTER_VER_CHROMA_SP 64, 16
-
-
-%macro PROCESS_CHROMA_SP_W2_4R 1
-    movd       m0, [r0]
-    movd       m1, [r0 + r1]
-    punpcklwd  m0, m1                          ;m0=[0 1]
-
-    lea        r0, [r0 + 2 * r1]
-    movd       m2, [r0]
-    punpcklwd  m1, m2                          ;m1=[1 2]
-    punpcklqdq m0, m1                          ;m0=[0 1 1 2]
-    pmaddwd    m0, [%1 + 0 *16]                ;m0=[0+1 1+2] Row 1-2
-
-    movd       m1, [r0 + r1]
-    punpcklwd  m2, m1                          ;m2=[2 3]
-
-    lea        r0, [r0 + 2 * r1]
-    movd       m3, [r0]
-    punpcklwd  m1, m3                          ;m2=[3 4]
-    punpcklqdq m2, m1                          ;m2=[2 3 3 4]
-
-    pmaddwd    m4, m2, [%1 + 1 * 16]           ;m4=[2+3 3+4] Row 1-2
-    pmaddwd    m2, [%1 + 0 * 16]               ;m2=[2+3 3+4] Row 3-4
-    paddd      m0, m4                          ;m0=[0+1+2+3 1+2+3+4] Row 1-2
-
-    movd       m1, [r0 + r1]
-    punpcklwd  m3, m1                          ;m3=[4 5]
-
-    movd       m4, [r0 + 2 * r1]
-    punpcklwd  m1, m4                          ;m1=[5 6]
-    punpcklqdq m3, m1                          ;m2=[4 5 5 6]
-    pmaddwd    m3, [%1 + 1 * 16]               ;m3=[4+5 5+6] Row 3-4
-    paddd      m2, m3                          ;m2=[2+3+4+5 3+4+5+6] Row 3-4
-%endmacro
-
-;-------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vertical_sp_%1x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-------------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SP_W2_4R 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_sp_%1x%2, 5, 6, 6
-
-    add       r1d, r1d
-    sub       r0, r1
-    shl       r4d, 5
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    mova      m5, [pd_526336]
-
-    mov       r4d, (%2/4)
-
-.loopH:
-    PROCESS_CHROMA_SP_W2_4R r5
-
-    paddd     m0, m5
-    paddd     m2, m5
-
-    psrad     m0, 12
-    psrad     m2, 12
-
-    packssdw  m0, m2
-    packuswb  m0, m0
-
-    pextrw    [r2], m0, 0
-    pextrw    [r2 + r3], m0, 1
-    lea       r2, [r2 + 2 * r3]
-    pextrw    [r2], m0, 2
-    pextrw    [r2 + r3], m0, 3
-
-    lea       r2, [r2 + 2 * r3]
-
-    dec       r4d
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SP_W2_4R 2, 4
-    FILTER_VER_CHROMA_SP_W2_4R 2, 8
-
-    FILTER_VER_CHROMA_SP_W2_4R 2, 16
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_sp_4x2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_vert_sp_4x2, 5, 6, 5
-
-    add        r1d, r1d
-    sub        r0, r1
-    shl        r4d, 5
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeffV]
-    lea        r5, [r5 + r4]
-%else
-    lea        r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    mova       m4, [pd_526336]
-
-    movq       m0, [r0]
-    movq       m1, [r0 + r1]
-    punpcklwd  m0, m1                          ;m0=[0 1]
-    pmaddwd    m0, [r5 + 0 *16]                ;m0=[0+1]  Row1
-
-    lea        r0, [r0 + 2 * r1]
-    movq       m2, [r0]
-    punpcklwd  m1, m2                          ;m1=[1 2]
-    pmaddwd    m1, [r5 + 0 *16]                ;m1=[1+2]  Row2
-
-    movq       m3, [r0 + r1]
-    punpcklwd  m2, m3                          ;m4=[2 3]
-    pmaddwd    m2, [r5 + 1 * 16]
-    paddd      m0, m2                          ;m0=[0+1+2+3]  Row1 done
-    paddd      m0, m4
-    psrad      m0, 12
-
-    movq       m2, [r0 + 2 * r1]
-    punpcklwd  m3, m2                          ;m5=[3 4]
-    pmaddwd    m3, [r5 + 1 * 16]
-    paddd      m1, m3                          ;m1 = [1+2+3+4]  Row2 done
-    paddd      m1, m4
-    psrad      m1, 12
-
-    packssdw   m0, m1
-    packuswb   m0, m0
-
-    movd       [r2], m0
-    pextrd     [r2 + r3], m0, 1
-
-    RET
-
-;-------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vertical_sp_6x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;-------------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SP_W6_H4 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_sp_6x%2, 5, 7, 7
-
-    add       r1d, r1d
-    sub       r0, r1
-    shl       r4d, 5
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r6, [r5 + r4]
-%else
-    lea       r6, [tab_ChromaCoeffV + r4]
-%endif
-
-    mova      m6, [pd_526336]
-
-    mov       r4d, %2/4
-
-.loopH:
-    PROCESS_CHROMA_SP_W4_4R
-
-    paddd     m0, m6
-    paddd     m1, m6
-    paddd     m2, m6
-    paddd     m3, m6
-
-    psrad     m0, 12
-    psrad     m1, 12
-    psrad     m2, 12
-    psrad     m3, 12
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-
-    packuswb  m0, m2
-
-    movd      [r2], m0
-    pextrd    [r2 + r3], m0, 1
-    lea       r5, [r2 + 2 * r3]
-    pextrd    [r5], m0, 2
-    pextrd    [r5 + r3], m0, 3
-
-    lea       r5, [4 * r1 - 2 * 4]
-    sub       r0, r5
-    add       r2, 4
-
-    PROCESS_CHROMA_SP_W2_4R r6
-
-    paddd     m0, m6
-    paddd     m2, m6
-
-    psrad     m0, 12
-    psrad     m2, 12
-
-    packssdw  m0, m2
-    packuswb  m0, m0
-
-    pextrw    [r2], m0, 0
-    pextrw    [r2 + r3], m0, 1
-    lea       r2, [r2 + 2 * r3]
-    pextrw    [r2], m0, 2
-    pextrw    [r2 + r3], m0, 3
-
-    sub       r0, 2 * 4
-    lea       r2, [r2 + 2 * r3 - 4]
-
-    dec       r4d
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SP_W6_H4 6, 8
-
-    FILTER_VER_CHROMA_SP_W6_H4 6, 16
-
-%macro PROCESS_CHROMA_SP_W8_2R 0
-    movu       m1, [r0]
-    movu       m3, [r0 + r1]
-    punpcklwd  m0, m1, m3
-    pmaddwd    m0, [r5 + 0 * 16]                ;m0 = [0l+1l]  Row1l
-    punpckhwd  m1, m3
-    pmaddwd    m1, [r5 + 0 * 16]                ;m1 = [0h+1h]  Row1h
-
-    movu       m4, [r0 + 2 * r1]
-    punpcklwd  m2, m3, m4
-    pmaddwd    m2, [r5 + 0 * 16]                ;m2 = [1l+2l]  Row2l
-    punpckhwd  m3, m4
-    pmaddwd    m3, [r5 + 0 * 16]                ;m3 = [1h+2h]  Row2h
-
-    lea        r0, [r0 + 2 * r1]
-    movu       m5, [r0 + r1]
-    punpcklwd  m6, m4, m5
-    pmaddwd    m6, [r5 + 1 * 16]                ;m6 = [2l+3l]  Row1l
-    paddd      m0, m6                           ;m0 = [0l+1l+2l+3l]  Row1l sum
-    punpckhwd  m4, m5
-    pmaddwd    m4, [r5 + 1 * 16]                ;m6 = [2h+3h]  Row1h
-    paddd      m1, m4                           ;m1 = [0h+1h+2h+3h]  Row1h sum
-
-    movu       m4, [r0 + 2 * r1]
-    punpcklwd  m6, m5, m4
-    pmaddwd    m6, [r5 + 1 * 16]                ;m6 = [3l+4l]  Row2l
-    paddd      m2, m6                           ;m2 = [1l+2l+3l+4l]  Row2l sum
-    punpckhwd  m5, m4
-    pmaddwd    m5, [r5 + 1 * 16]                ;m1 = [3h+4h]  Row2h
-    paddd      m3, m5                           ;m3 = [1h+2h+3h+4h]  Row2h sum
-%endmacro
-
-;--------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_sp_8x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
-;--------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SP_W8_H2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_sp_%1x%2, 5, 6, 8
-
-    add       r1d, r1d
-    sub       r0, r1
-    shl       r4d, 5
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    mova      m7, [pd_526336]
-
-    mov       r4d, %2/2
-.loopH:
-    PROCESS_CHROMA_SP_W8_2R
-
-    paddd     m0, m7
-    paddd     m1, m7
-    paddd     m2, m7
-    paddd     m3, m7
-
-    psrad     m0, 12
-    psrad     m1, 12
-    psrad     m2, 12
-    psrad     m3, 12
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-
-    packuswb  m0, m2
-
-    movlps    [r2], m0
-    movhps    [r2 + r3], m0
-
-    lea       r2, [r2 + 2 * r3]
-
-    dec r4d
-    jnz .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SP_W8_H2 8, 2
-    FILTER_VER_CHROMA_SP_W8_H2 8, 4
-    FILTER_VER_CHROMA_SP_W8_H2 8, 6
-    FILTER_VER_CHROMA_SP_W8_H2 8, 8
-    FILTER_VER_CHROMA_SP_W8_H2 8, 16
-    FILTER_VER_CHROMA_SP_W8_H2 8, 32
-
-    FILTER_VER_CHROMA_SP_W8_H2 8, 12
-    FILTER_VER_CHROMA_SP_W8_H2 8, 64
-
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_2x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-%macro FILTER_HORIZ_CHROMA_2xN 2
-INIT_XMM sse4
-cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 4, src, srcstride, dst, dststride
-%define coef2  m3
-%define Tm0    m2
-%define t1     m1
-%define t0     m0
-
-    dec        srcq
-    mov        r4d, r4m
-    add        dststrided, dststrided
-
-%ifdef PIC
-    lea        r6, [tab_ChromaCoeff]
-    movd       coef2, [r6 + r4 * 4]
-%else
-    movd       coef2, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd     coef2, coef2, 0
-    mova       t1, [pw_2000]
-    mova       Tm0, [tab_Tm]
-
-    mov        r4d, %2
-    cmp        r5m, byte 0
-    je         .loopH
-    sub        srcq, srcstrideq
-    add        r4d, 3
-
-.loopH:
-    movh       t0, [srcq]
-    pshufb     t0, t0, Tm0
-    pmaddubsw  t0, coef2
-    phaddw     t0, t0
-    psubw      t0, t1
-    movd       [dstq], t0
-
-    lea        srcq, [srcq + srcstrideq]
-    lea        dstq, [dstq + dststrideq]
-
-    dec        r4d
-    jnz        .loopH
-
-    RET
-%endmacro
-
-    FILTER_HORIZ_CHROMA_2xN 2, 4
-    FILTER_HORIZ_CHROMA_2xN 2, 8
-
-    FILTER_HORIZ_CHROMA_2xN 2, 16
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_4x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-%macro FILTER_HORIZ_CHROMA_4xN 2
-INIT_XMM sse4
-cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 4, src, srcstride, dst, dststride
-%define coef2  m3
-%define Tm0    m2
-%define t1     m1
-%define t0     m0
-
-    dec        srcq
-    mov        r4d, r4m
-    add        dststrided, dststrided
-
-%ifdef PIC
-    lea        r6, [tab_ChromaCoeff]
-    movd       coef2, [r6 + r4 * 4]
-%else
-    movd       coef2, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd     coef2, coef2, 0
-    mova       t1, [pw_2000]
-    mova       Tm0, [tab_Tm]
-
-    mov        r4d, %2
-    cmp        r5m, byte 0
-    je         .loopH
-    sub        srcq, srcstrideq
-    add        r4d, 3
-
-.loopH:
-    movh       t0, [srcq]
-    pshufb     t0, t0, Tm0
-    pmaddubsw  t0, coef2
-    phaddw     t0, t0
-    psubw      t0, t1
-    movlps     [dstq], t0
-
-    lea        srcq, [srcq + srcstrideq]
-    lea        dstq, [dstq + dststrideq]
-
-    dec        r4d
-    jnz        .loopH
-    RET
-%endmacro
-
-    FILTER_HORIZ_CHROMA_4xN 4, 2
-    FILTER_HORIZ_CHROMA_4xN 4, 4
-    FILTER_HORIZ_CHROMA_4xN 4, 8
-    FILTER_HORIZ_CHROMA_4xN 4, 16
-
-    FILTER_HORIZ_CHROMA_4xN 4, 32
-
-%macro PROCESS_CHROMA_W6 3
-    movu       %1, [srcq]
-    pshufb     %2, %1, Tm0
-    pmaddubsw  %2, coef2
-    pshufb     %1, %1, Tm1
-    pmaddubsw  %1, coef2
-    phaddw     %2, %1
-    psubw      %2, %3
-    movh       [dstq], %2
-    pshufd     %2, %2, 2
-    movd       [dstq + 8], %2
-%endmacro
-
-%macro PROCESS_CHROMA_W12 3
-    movu       %1, [srcq]
-    pshufb     %2, %1, Tm0
-    pmaddubsw  %2, coef2
-    pshufb     %1, %1, Tm1
-    pmaddubsw  %1, coef2
-    phaddw     %2, %1
-    psubw      %2, %3
-    movu       [dstq], %2
-    movu       %1, [srcq + 8]
-    pshufb     %1, %1, Tm0
-    pmaddubsw  %1, coef2
-    phaddw     %1, %1
-    psubw      %1, %3
-    movh       [dstq + 16], %1
-%endmacro
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_6x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-%macro FILTER_HORIZ_CHROMA 2
-INIT_XMM sse4
-cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
-%define coef2    m5
-%define Tm0      m4
-%define Tm1      m3
-%define t2       m2
-%define t1       m1
-%define t0       m0
-
-    dec     srcq
-    mov     r4d, r4m
-    add     dststrided, dststrided
-
-%ifdef PIC
-    lea     r6, [tab_ChromaCoeff]
-    movd    coef2, [r6 + r4 * 4]
-%else
-    movd    coef2, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd  coef2, coef2, 0
-    mova    t2, [pw_2000]
-    mova    Tm0, [tab_Tm]
-    mova    Tm1, [tab_Tm + 16]
-
-    mov     r4d, %2
-    cmp     r5m, byte 0
-    je      .loopH
-    sub     srcq, srcstrideq
-    add     r4d, 3
-
-.loopH:
-    PROCESS_CHROMA_W%1  t0, t1, t2
-    add     srcq, srcstrideq
-    add     dstq, dststrideq
-
-    dec     r4d
-    jnz     .loopH
-
-    RET
-%endmacro
-
-    FILTER_HORIZ_CHROMA 6, 8
-    FILTER_HORIZ_CHROMA 12, 16
-
-    FILTER_HORIZ_CHROMA 6, 16
-    FILTER_HORIZ_CHROMA 12, 32
-
-%macro PROCESS_CHROMA_W8 3
-    movu        %1, [srcq]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    psubw       %2, %3
-    movu        [dstq], %2
-%endmacro
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-%macro FILTER_HORIZ_CHROMA_8xN 2
-INIT_XMM sse4
-cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
-%define coef2    m5
-%define Tm0      m4
-%define Tm1      m3
-%define t2       m2
-%define t1       m1
-%define t0       m0
-
-    dec     srcq
-    mov     r4d, r4m
-    add     dststrided, dststrided
-
-%ifdef PIC
-    lea     r6, [tab_ChromaCoeff]
-    movd    coef2, [r6 + r4 * 4]
-%else
-    movd    coef2, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd  coef2, coef2, 0
-    mova    t2, [pw_2000]
-    mova    Tm0, [tab_Tm]
-    mova    Tm1, [tab_Tm + 16]
-
-    mov     r4d, %2
-    cmp     r5m, byte 0
-    je      .loopH
-    sub     srcq, srcstrideq
-    add     r4d, 3
-
-.loopH:
-    PROCESS_CHROMA_W8  t0, t1, t2
-    add     srcq, srcstrideq
-    add     dstq, dststrideq
-
-    dec     r4d
-    jnz     .loopH
-
-    RET
-%endmacro
-
-    FILTER_HORIZ_CHROMA_8xN 8, 2
-    FILTER_HORIZ_CHROMA_8xN 8, 4
-    FILTER_HORIZ_CHROMA_8xN 8, 6
-    FILTER_HORIZ_CHROMA_8xN 8, 8
-    FILTER_HORIZ_CHROMA_8xN 8, 16
-    FILTER_HORIZ_CHROMA_8xN 8, 32
-
-    FILTER_HORIZ_CHROMA_8xN 8, 12
-    FILTER_HORIZ_CHROMA_8xN 8, 64
-
-%macro PROCESS_CHROMA_W16 4
-    movu        %1, [srcq]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq + 8]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    psubw       %2, %3
-    psubw       %4, %3
-    movu        [dstq], %2
-    movu        [dstq + 16], %4
-%endmacro
-
-%macro PROCESS_CHROMA_W24 4
-    movu        %1, [srcq]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq + 8]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    psubw       %2, %3
-    psubw       %4, %3
-    movu        [dstq], %2
-    movu        [dstq + 16], %4
-    movu        %1, [srcq + 16]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    psubw       %2, %3
-    movu        [dstq + 32], %2
-%endmacro
-
-%macro PROCESS_CHROMA_W32 4
-    movu        %1, [srcq]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq + 8]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    psubw       %2, %3
-    psubw       %4, %3
-    movu        [dstq], %2
-    movu        [dstq + 16], %4
-    movu        %1, [srcq + 16]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq + 24]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    psubw       %2, %3
-    psubw       %4, %3
-    movu        [dstq + 32], %2
-    movu        [dstq + 48], %4
-%endmacro
-
-%macro PROCESS_CHROMA_W16o 5
-    movu        %1, [srcq + %5]
-    pshufb      %2, %1, Tm0
-    pmaddubsw   %2, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %2, %1
-    movu        %1, [srcq + %5 + 8]
-    pshufb      %4, %1, Tm0
-    pmaddubsw   %4, coef2
-    pshufb      %1, %1, Tm1
-    pmaddubsw   %1, coef2
-    phaddw      %4, %1
-    psubw       %2, %3
-    psubw       %4, %3
-    movu        [dstq + %5 * 2], %2
-    movu        [dstq + %5 * 2 + 16], %4
-%endmacro
-
-%macro PROCESS_CHROMA_W48 4
-    PROCESS_CHROMA_W16o %1, %2, %3, %4, 0
-    PROCESS_CHROMA_W16o %1, %2, %3, %4, 16
-    PROCESS_CHROMA_W16o %1, %2, %3, %4, 32
-%endmacro
-
-%macro PROCESS_CHROMA_W64 4
-    PROCESS_CHROMA_W16o %1, %2, %3, %4, 0
-    PROCESS_CHROMA_W16o %1, %2, %3, %4, 16
-    PROCESS_CHROMA_W16o %1, %2, %3, %4, 32
-    PROCESS_CHROMA_W16o %1, %2, %3, %4, 48
-%endmacro
-
-;------------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;------------------------------------------------------------------------------------------------------------------------------
-%macro FILTER_HORIZ_CHROMA_WxN 2
-INIT_XMM sse4
-cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 7, src, srcstride, dst, dststride
-%define coef2    m6
-%define Tm0      m5
-%define Tm1      m4
-%define t3       m3
-%define t2       m2
-%define t1       m1
-%define t0       m0
-
-    dec     srcq
-    mov     r4d, r4m
-    add     dststrided, dststrided
-
-%ifdef PIC
-    lea     r6, [tab_ChromaCoeff]
-    movd    coef2, [r6 + r4 * 4]
-%else
-    movd    coef2, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufd  coef2, coef2, 0
-    mova    t2, [pw_2000]
-    mova    Tm0, [tab_Tm]
-    mova    Tm1, [tab_Tm + 16]
-
-    mov     r4d, %2
-    cmp     r5m, byte 0
-    je      .loopH
-    sub     srcq, srcstrideq
-    add     r4d, 3
-
-.loopH:
-    PROCESS_CHROMA_W%1   t0, t1, t2, t3
-    add     srcq, srcstrideq
-    add     dstq, dststrideq
-
-    dec     r4d
-    jnz     .loopH
-
-    RET
-%endmacro
-
-    FILTER_HORIZ_CHROMA_WxN 16, 4
-    FILTER_HORIZ_CHROMA_WxN 16, 8
-    FILTER_HORIZ_CHROMA_WxN 16, 12
-    FILTER_HORIZ_CHROMA_WxN 16, 16
-    FILTER_HORIZ_CHROMA_WxN 16, 32
-    FILTER_HORIZ_CHROMA_WxN 24, 32
-    FILTER_HORIZ_CHROMA_WxN 32,  8
-    FILTER_HORIZ_CHROMA_WxN 32, 16
-    FILTER_HORIZ_CHROMA_WxN 32, 24
-    FILTER_HORIZ_CHROMA_WxN 32, 32
-
-    FILTER_HORIZ_CHROMA_WxN 16, 24
-    FILTER_HORIZ_CHROMA_WxN 16, 64
-    FILTER_HORIZ_CHROMA_WxN 24, 64
-    FILTER_HORIZ_CHROMA_WxN 32, 48
-    FILTER_HORIZ_CHROMA_WxN 32, 64
-
-    FILTER_HORIZ_CHROMA_WxN 64, 64
-    FILTER_HORIZ_CHROMA_WxN 64, 32
-    FILTER_HORIZ_CHROMA_WxN 64, 48
-    FILTER_HORIZ_CHROMA_WxN 48, 64
-    FILTER_HORIZ_CHROMA_WxN 64, 16
-
-
-;---------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W16n 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_%1x%2, 4, 7, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m0, [r5 + r4 * 4]
-%else
-    movd       m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m1, m0, [tab_Vm]
-    pshufb     m0, [tab_Vm + 16]
-    mov        r4d, %2/2
-
-.loop:
-
-    mov         r6d,       %1/16
-
-.loopW:
-
-    movu       m2, [r0]
-    movu       m3, [r0 + r1]
-
-    punpcklbw  m4, m2, m3
-    punpckhbw  m2, m3
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m2, m1
-
-    lea        r5, [r0 + 2 * r1]
-    movu       m5, [r5]
-    movu       m7, [r5 + r1]
-
-    punpcklbw  m6, m5, m7
-    pmaddubsw  m6, m0
-    paddw      m4, m6
-
-    punpckhbw  m6, m5, m7
-    pmaddubsw  m6, m0
-    paddw      m2, m6
-
-    mova       m6, [pw_2000]
-
-    psubw      m4, m6
-    psubw      m2, m6
-
-    movu       [r2], m4
-    movu       [r2 + 16], m2
-
-    punpcklbw  m4, m3, m5
-    punpckhbw  m3, m5
-
-    pmaddubsw  m4, m1
-    pmaddubsw  m3, m1
-
-    movu       m5, [r5 + 2 * r1]
-
-    punpcklbw  m2, m7, m5
-    punpckhbw  m7, m5
-
-    pmaddubsw  m2, m0
-    pmaddubsw  m7, m0
-
-    paddw      m4, m2
-    paddw      m3, m7
-
-    psubw      m4, m6
-    psubw      m3, m6
-
-    movu       [r2 + r3], m4
-    movu       [r2 + r3 + 16], m3
-
-    add         r0,        16
-    add         r2,        32
-    dec         r6d
-    jnz         .loopW
-
-    lea         r0,        [r0 + r1 * 2 - %1]
-    lea         r2,        [r2 + r3 * 2 - %1 * 2]
-
-    dec        r4d
-    jnz        .loop
-    RET
-%endmacro
-
-    FILTER_V_PS_W16n 64, 64
-    FILTER_V_PS_W16n 64, 32
-    FILTER_V_PS_W16n 64, 48
-    FILTER_V_PS_W16n 48, 64
-    FILTER_V_PS_W16n 64, 16
-
-
-;------------------------------------------------------------------------------------------------------------
-;void interp_4tap_vert_ps_2x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;------------------------------------------------------------------------------------------------------------
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_2x4, 4, 6, 7
-
-    mov         r4d, r4m
-    sub         r0, r1
-    add         r3d, r3d
-
-%ifdef PIC
-    lea         r5, [tab_ChromaCoeff]
-    movd        m0, [r5 + r4 * 4]
-%else
-    movd        m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb      m0, [tab_Cm]
-
-    lea         r5, [3 * r1]
-
-    movd        m2, [r0]
-    movd        m3, [r0 + r1]
-    movd        m4, [r0 + 2 * r1]
-    movd        m5, [r0 + r5]
-
-    punpcklbw   m2, m3
-    punpcklbw   m6, m4, m5
-    punpcklbw   m2, m6
-
-    pmaddubsw   m2, m0
-
-    lea         r0, [r0 + 4 * r1]
-    movd        m6, [r0]
-
-    punpcklbw   m3, m4
-    punpcklbw   m1, m5, m6
-    punpcklbw   m3, m1
-
-    pmaddubsw   m3, m0
-    phaddw      m2, m3
-
-    mova        m1, [pw_2000]
-
-    psubw       m2, m1
-
-    movd        [r2], m2
-    pextrd      [r2 + r3], m2, 2
-
-    movd        m2, [r0 + r1]
-
-    punpcklbw   m4, m5
-    punpcklbw   m3, m6, m2
-    punpcklbw   m4, m3
-
-    pmaddubsw   m4, m0
-
-    movd        m3, [r0 + 2 * r1]
-
-    punpcklbw   m5, m6
-    punpcklbw   m2, m3
-    punpcklbw   m5, m2
-
-    pmaddubsw   m5, m0
-    phaddw      m4, m5
-    psubw       m4, m1
-
-    lea         r2, [r2 + 2 * r3]
-    movd        [r2], m4
-    pextrd      [r2 + r3], m4, 2
-
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ps_2x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-------------------------------------------------------------------------------------------------------------
-%macro FILTER_V_PS_W2 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ps_2x%2, 4, 6, 8
-
-    mov        r4d, r4m
-    sub        r0, r1
-    add        r3d, r3d
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeff]
-    movd       m0, [r5 + r4 * 4]
-%else
-    movd       m0, [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    pshufb     m0, [tab_Cm]
-
-    mova       m1, [pw_2000]
-    lea        r5, [3 * r1]
-    mov        r4d, %2/4
-.loop:
-    movd       m2, [r0]
-    movd       m3, [r0 + r1]
-    movd       m4, [r0 + 2 * r1]
-    movd       m5, [r0 + r5]
-
-    punpcklbw  m2, m3
-    punpcklbw  m6, m4, m5
-    punpcklbw  m2, m6
-
-    pmaddubsw  m2, m0
-
-    lea        r0, [r0 + 4 * r1]
-    movd       m6, [r0]
-
-    punpcklbw  m3, m4
-    punpcklbw  m7, m5, m6
-    punpcklbw  m3, m7
-
-    pmaddubsw  m3, m0
-
-    phaddw     m2, m3
-    psubw      m2, m1
-
-
-    movd       [r2], m2
-    pshufd     m2, m2, 2
-    movd       [r2 + r3], m2
-
-    movd       m2, [r0 + r1]
-
-    punpcklbw  m4, m5
-    punpcklbw  m3, m6, m2
-    punpcklbw  m4, m3
-
-    pmaddubsw  m4, m0
-
-    movd       m3, [r0 + 2 * r1]
-
-    punpcklbw  m5, m6
-    punpcklbw  m2, m3
-    punpcklbw  m5, m2
-
-    pmaddubsw  m5, m0
-
-    phaddw     m4, m5
-
-    psubw      m4, m1
-
-    lea        r2, [r2 + 2 * r3]
-    movd       [r2], m4
-    pshufd     m4 , m4 ,2
-    movd       [r2 + r3], m4
-
-    lea        r2, [r2 + 2 * r3]
-
-    dec        r4d
-    jnz        .loop
-
-    RET
-%endmacro
-
-    FILTER_V_PS_W2 2, 8
-
-    FILTER_V_PS_W2 2, 16
-
-;-----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SS 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_ss_%1x%2, 5, 7, 6 ,0-gprsize
-
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 5
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r6, [r5 + r4]
-%else
-    lea       r6, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       dword [rsp], %2/4
-
-.loopH:
-    mov       r4d, (%1/4)
-.loopW:
-    PROCESS_CHROMA_SP_W4_4R
-
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-
-    movlps    [r2], m0
-    movhps    [r2 + r3], m0
-    lea       r5, [r2 + 2 * r3]
-    movlps    [r5], m2
-    movhps    [r5 + r3], m2
-
-    lea       r5, [4 * r1 - 2 * 4]
-    sub       r0, r5
-    add       r2, 2 * 4
-
-    dec       r4d
-    jnz       .loopW
-
-    lea       r0, [r0 + 4 * r1 - 2 * %1]
-    lea       r2, [r2 + 4 * r3 - 2 * %1]
-
-    dec       dword [rsp]
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SS 4, 4
-    FILTER_VER_CHROMA_SS 4, 8
-    FILTER_VER_CHROMA_SS 16, 16
-    FILTER_VER_CHROMA_SS 16, 8
-    FILTER_VER_CHROMA_SS 16, 12
-    FILTER_VER_CHROMA_SS 12, 16
-    FILTER_VER_CHROMA_SS 16, 4
-    FILTER_VER_CHROMA_SS 4, 16
-    FILTER_VER_CHROMA_SS 32, 32
-    FILTER_VER_CHROMA_SS 32, 16
-    FILTER_VER_CHROMA_SS 16, 32
-    FILTER_VER_CHROMA_SS 32, 24
-    FILTER_VER_CHROMA_SS 24, 32
-    FILTER_VER_CHROMA_SS 32, 8
-
-    FILTER_VER_CHROMA_SS 16, 24
-    FILTER_VER_CHROMA_SS 12, 32
-    FILTER_VER_CHROMA_SS 4, 32
-    FILTER_VER_CHROMA_SS 32, 64
-    FILTER_VER_CHROMA_SS 16, 64
-    FILTER_VER_CHROMA_SS 32, 48
-    FILTER_VER_CHROMA_SS 24, 64
-
-    FILTER_VER_CHROMA_SS 64, 64
-    FILTER_VER_CHROMA_SS 64, 32
-    FILTER_VER_CHROMA_SS 64, 48
-    FILTER_VER_CHROMA_SS 48, 64
-    FILTER_VER_CHROMA_SS 64, 16
-
-%macro FILTER_VER_CHROMA_S_AVX2_4x4 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x4, 4, 6, 7
-    mov             r4d, r4m
-    add             r1d, r1d
-    shl             r4d, 6
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-%ifidn %1,sp
-    mova            m6, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-
-    movq            xm0, [r0]
-    movq            xm1, [r0 + r1]
-    punpcklwd       xm0, xm1
-    movq            xm2, [r0 + r1 * 2]
-    punpcklwd       xm1, xm2
-    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
-    pmaddwd         m0, [r5]
-    movq            xm3, [r0 + r4]
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm4, [r0]
-    punpcklwd       xm3, xm4
-    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m5
-    movq            xm3, [r0 + r1]
-    punpcklwd       xm4, xm3
-    movq            xm1, [r0 + r1 * 2]
-    punpcklwd       xm3, xm1
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
-    pmaddwd         m4, [r5 + 1 * mmsize]
-    paddd           m2, m4
-
-%ifidn %1,sp
-    paddd           m0, m6
-    paddd           m2, m6
-    psrad           m0, 12
-    psrad           m2, 12
-%else
-    psrad           m0, 6
-    psrad           m2, 6
-%endif
-    packssdw        m0, m2
-    vextracti128    xm2, m0, 1
-    lea             r4, [r3 * 3]
-
-%ifidn %1,sp
-    packuswb        xm0, xm2
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 2
-    pextrd          [r2 + r3 * 2], xm0, 1
-    pextrd          [r2 + r4], xm0, 3
-%else
-    movq            [r2], xm0
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r4], xm2
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_4x4 sp
-    FILTER_VER_CHROMA_S_AVX2_4x4 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_4x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x8, 4, 6, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-
-    movq            xm0, [r0]
-    movq            xm1, [r0 + r1]
-    punpcklwd       xm0, xm1
-    movq            xm2, [r0 + r1 * 2]
-    punpcklwd       xm1, xm2
-    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
-    pmaddwd         m0, [r5]
-    movq            xm3, [r0 + r4]
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm4, [r0]
-    punpcklwd       xm3, xm4
-    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m5
-    movq            xm3, [r0 + r1]
-    punpcklwd       xm4, xm3
-    movq            xm1, [r0 + r1 * 2]
-    punpcklwd       xm3, xm1
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
-    pmaddwd         m5, m4, [r5 + 1 * mmsize]
-    paddd           m2, m5
-    pmaddwd         m4, [r5]
-    movq            xm3, [r0 + r4]
-    punpcklwd       xm1, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm6, [r0]
-    punpcklwd       xm3, xm6
-    vinserti128     m1, m1, xm3, 1                  ; m1 = [8 7 7 6]
-    pmaddwd         m5, m1, [r5 + 1 * mmsize]
-    paddd           m4, m5
-    pmaddwd         m1, [r5]
-    movq            xm3, [r0 + r1]
-    punpcklwd       xm6, xm3
-    movq            xm5, [r0 + 2 * r1]
-    punpcklwd       xm3, xm5
-    vinserti128     m6, m6, xm3, 1                  ; m6 = [A 9 9 8]
-    pmaddwd         m6, [r5 + 1 * mmsize]
-    paddd           m1, m6
-    lea             r4, [r3 * 3]
-
-%ifidn %1,sp
-    paddd           m0, m7
-    paddd           m2, m7
-    paddd           m4, m7
-    paddd           m1, m7
-    psrad           m0, 12
-    psrad           m2, 12
-    psrad           m4, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m2, 6
-    psrad           m4, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m2
-    packssdw        m4, m1
-%ifidn %1,sp
-    packuswb        m0, m4
-    vextracti128    xm2, m0, 1
-    movd            [r2], xm0
-    movd            [r2 + r3], xm2
-    pextrd          [r2 + r3 * 2], xm0, 1
-    pextrd          [r2 + r4], xm2, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm0, 2
-    pextrd          [r2 + r3], xm2, 2
-    pextrd          [r2 + r3 * 2], xm0, 3
-    pextrd          [r2 + r4], xm2, 3
-%else
-    vextracti128    xm2, m0, 1
-    vextracti128    xm1, m4, 1
-    movq            [r2], xm0
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r4], xm2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm4
-    movq            [r2 + r3], xm1
-    movhps          [r2 + r3 * 2], xm4
-    movhps          [r2 + r4], xm1
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_4x8 sp
-    FILTER_VER_CHROMA_S_AVX2_4x8 ss
-
-%macro PROCESS_CHROMA_AVX2_W4_16R 1
-    movq            xm0, [r0]
-    movq            xm1, [r0 + r1]
-    punpcklwd       xm0, xm1
-    movq            xm2, [r0 + r1 * 2]
-    punpcklwd       xm1, xm2
-    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
-    pmaddwd         m0, [r5]
-    movq            xm3, [r0 + r4]
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm4, [r0]
-    punpcklwd       xm3, xm4
-    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m5
-    movq            xm3, [r0 + r1]
-    punpcklwd       xm4, xm3
-    movq            xm1, [r0 + r1 * 2]
-    punpcklwd       xm3, xm1
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
-    pmaddwd         m5, m4, [r5 + 1 * mmsize]
-    paddd           m2, m5
-    pmaddwd         m4, [r5]
-    movq            xm3, [r0 + r4]
-    punpcklwd       xm1, xm3
-    lea             r0, [r0 + 4 * r1]
-    movq            xm6, [r0]
-    punpcklwd       xm3, xm6
-    vinserti128     m1, m1, xm3, 1                  ; m1 = [8 7 7 6]
-    pmaddwd         m5, m1, [r5 + 1 * mmsize]
-    paddd           m4, m5
-    pmaddwd         m1, [r5]
-    movq            xm3, [r0 + r1]
-    punpcklwd       xm6, xm3
-    movq            xm5, [r0 + 2 * r1]
-    punpcklwd       xm3, xm5
-    vinserti128     m6, m6, xm3, 1                  ; m6 = [10 9 9 8]
-    pmaddwd         m3, m6, [r5 + 1 * mmsize]
-    paddd           m1, m3
-    pmaddwd         m6, [r5]
-
-%ifidn %1,sp
-    paddd           m0, m7
-    paddd           m2, m7
-    paddd           m4, m7
-    paddd           m1, m7
-    psrad           m4, 12
-    psrad           m1, 12
-    psrad           m0, 12
-    psrad           m2, 12
-%else
-    psrad           m0, 6
-    psrad           m2, 6
-    psrad           m4, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m2
-    packssdw        m4, m1
-%ifidn %1,sp
-    packuswb        m0, m4
-    vextracti128    xm4, m0, 1
-    movd            [r2], xm0
-    movd            [r2 + r3], xm4
-    pextrd          [r2 + r3 * 2], xm0, 1
-    pextrd          [r2 + r6], xm4, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm0, 2
-    pextrd          [r2 + r3], xm4, 2
-    pextrd          [r2 + r3 * 2], xm0, 3
-    pextrd          [r2 + r6], xm4, 3
-%else
-    vextracti128    xm2, m0, 1
-    vextracti128    xm1, m4, 1
-    movq            [r2], xm0
-    movq            [r2 + r3], xm2
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r6], xm2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm4
-    movq            [r2 + r3], xm1
-    movhps          [r2 + r3 * 2], xm4
-    movhps          [r2 + r6], xm1
-%endif
-
-    movq            xm2, [r0 + r4]
-    punpcklwd       xm5, xm2
-    lea             r0, [r0 + 4 * r1]
-    movq            xm0, [r0]
-    punpcklwd       xm2, xm0
-    vinserti128     m5, m5, xm2, 1                  ; m5 = [12 11 11 10]
-    pmaddwd         m2, m5, [r5 + 1 * mmsize]
-    paddd           m6, m2
-    pmaddwd         m5, [r5]
-    movq            xm2, [r0 + r1]
-    punpcklwd       xm0, xm2
-    movq            xm3, [r0 + 2 * r1]
-    punpcklwd       xm2, xm3
-    vinserti128     m0, m0, xm2, 1                  ; m0 = [14 13 13 12]
-    pmaddwd         m2, m0, [r5 + 1 * mmsize]
-    paddd           m5, m2
-    pmaddwd         m0, [r5]
-    movq            xm4, [r0 + r4]
-    punpcklwd       xm3, xm4
-    lea             r0, [r0 + 4 * r1]
-    movq            xm1, [r0]
-    punpcklwd       xm4, xm1
-    vinserti128     m3, m3, xm4, 1                  ; m3 = [16 15 15 14]
-    pmaddwd         m4, m3, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m3, [r5]
-    movq            xm4, [r0 + r1]
-    punpcklwd       xm1, xm4
-    movq            xm2, [r0 + 2 * r1]
-    punpcklwd       xm4, xm2
-    vinserti128     m1, m1, xm4, 1                  ; m1 = [18 17 17 16]
-    pmaddwd         m1, [r5 + 1 * mmsize]
-    paddd           m3, m1
-
-%ifidn %1,sp
-    paddd           m6, m7
-    paddd           m5, m7
-    paddd           m0, m7
-    paddd           m3, m7
-    psrad           m6, 12
-    psrad           m5, 12
-    psrad           m0, 12
-    psrad           m3, 12
-%else
-    psrad           m6, 6
-    psrad           m5, 6
-    psrad           m0, 6
-    psrad           m3, 6
-%endif
-    packssdw        m6, m5
-    packssdw        m0, m3
-    lea             r2, [r2 + r3 * 4]
-
-%ifidn %1,sp
-    packuswb        m6, m0
-    vextracti128    xm0, m6, 1
-    movd            [r2], xm6
-    movd            [r2 + r3], xm0
-    pextrd          [r2 + r3 * 2], xm6, 1
-    pextrd          [r2 + r6], xm0, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm6, 2
-    pextrd          [r2 + r3], xm0, 2
-    pextrd          [r2 + r3 * 2], xm6, 3
-    pextrd          [r2 + r6], xm0, 3
-%else
-    vextracti128    xm5, m6, 1
-    vextracti128    xm3, m0, 1
-    movq            [r2], xm6
-    movq            [r2 + r3], xm5
-    movhps          [r2 + r3 * 2], xm6
-    movhps          [r2 + r6], xm5
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm0
-    movq            [r2 + r3], xm3
-    movhps          [r2 + r3 * 2], xm0
-    movhps          [r2 + r6], xm3
-%endif
-%endmacro
-
-%macro FILTER_VER_CHROMA_S_AVX2_4x16 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x16, 4, 7, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-    PROCESS_CHROMA_AVX2_W4_16R %1
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_4x16 sp
-    FILTER_VER_CHROMA_S_AVX2_4x16 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_4x32 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x32, 4, 7, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-%rep 2
-    PROCESS_CHROMA_AVX2_W4_16R %1
-    lea             r2, [r2 + r3 * 4]
-%endrep
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_4x32 sp
-    FILTER_VER_CHROMA_S_AVX2_4x32 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_4x2 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_4x2, 4, 6, 6
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-%ifidn %1,sp
-    mova            m5, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    movq            xm0, [r0]
-    movq            xm1, [r0 + r1]
-    punpcklwd       xm0, xm1
-    movq            xm2, [r0 + r1 * 2]
-    punpcklwd       xm1, xm2
-    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
-    pmaddwd         m0, [r5]
-    movq            xm3, [r0 + r4]
-    punpcklwd       xm2, xm3
-    movq            xm4, [r0 + 4 * r1]
-    punpcklwd       xm3, xm4
-    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m0, m2
-%ifidn %1,sp
-    paddd           m0, m5
-    psrad           m0, 12
-%else
-    psrad           m0, 6
-%endif
-    vextracti128    xm1, m0, 1
-    packssdw        xm0, xm1
-%ifidn %1,sp
-    packuswb        xm0, xm0
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-%else
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_4x2 sp
-    FILTER_VER_CHROMA_S_AVX2_4x2 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_2x4 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_2x4, 4, 6, 6
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-%ifidn %1,sp
-    mova            m5, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    movd            xm0, [r0]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm0, xm1
-    movd            xm2, [r0 + r1 * 2]
-    punpcklwd       xm1, xm2
-    punpcklqdq      xm0, xm1                        ; m0 = [2 1 1 0]
-    movd            xm3, [r0 + r4]
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movd            xm4, [r0]
-    punpcklwd       xm3, xm4
-    punpcklqdq      xm2, xm3                        ; m2 = [4 3 3 2]
-    vinserti128     m0, m0, xm2, 1                  ; m0 = [4 3 3 2 2 1 1 0]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm4, xm1
-    movd            xm3, [r0 + r1 * 2]
-    punpcklwd       xm1, xm3
-    punpcklqdq      xm4, xm1                        ; m4 = [6 5 5 4]
-    vinserti128     m2, m2, xm4, 1                  ; m2 = [6 5 5 4 4 3 3 2]
-    pmaddwd         m0, [r5]
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m0, m2
-%ifidn %1,sp
-    paddd           m0, m5
-    psrad           m0, 12
-%else
-    psrad           m0, 6
-%endif
-    vextracti128    xm1, m0, 1
-    packssdw        xm0, xm1
-    lea             r4, [r3 * 3]
-%ifidn %1,sp
-    packuswb        xm0, xm0
-    pextrw          [r2], xm0, 0
-    pextrw          [r2 + r3], xm0, 1
-    pextrw          [r2 + 2 * r3], xm0, 2
-    pextrw          [r2 + r4], xm0, 3
-%else
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    pextrd          [r2 + 2 * r3], xm0, 2
-    pextrd          [r2 + r4], xm0, 3
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_2x4 sp
-    FILTER_VER_CHROMA_S_AVX2_2x4 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_8x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x8, 4, 6, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m4
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    pmaddwd         m3, [r5]
-    paddd           m1, m5
-%ifidn %1,sp
-    paddd           m0, m7
-    paddd           m1, m7
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm1, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m1
-%ifidn %1,sp
-    paddd           m2, m7
-    paddd           m3, m7
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m2, m3
-
-    movu            xm1, [r0 + r4]                  ; m1 = row 7
-    punpckhwd       xm3, xm6, xm1
-    punpcklwd       xm6, xm1
-    vinserti128     m6, m6, xm3, 1
-    pmaddwd         m3, m6, [r5 + 1 * mmsize]
-    pmaddwd         m6, [r5]
-    paddd           m4, m3
-
-    lea             r4, [r3 * 3]
-%ifidn %1,sp
-    packuswb        m0, m2
-    mova            m3, [interp8_hps_shuf]
-    vpermd          m0, m3, m0
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r4], xm2
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    movu            [r2], xm0
-    vextracti128    xm0, m0, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3], xm0
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r4], xm3
-%endif
-    lea             r2, [r2 + r3 * 4]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm0, [r0]                       ; m0 = row 8
-    punpckhwd       xm2, xm1, xm0
-    punpcklwd       xm1, xm0
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    pmaddwd         m1, [r5]
-    paddd           m5, m2
-%ifidn %1,sp
-    paddd           m4, m7
-    paddd           m5, m7
-    psrad           m4, 12
-    psrad           m5, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-%endif
-    packssdw        m4, m5
-
-    movu            xm2, [r0 + r1]                  ; m2 = row 9
-    punpckhwd       xm5, xm0, xm2
-    punpcklwd       xm0, xm2
-    vinserti128     m0, m0, xm5, 1
-    pmaddwd         m0, [r5 + 1 * mmsize]
-    paddd           m6, m0
-    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
-    punpckhwd       xm0, xm2, xm5
-    punpcklwd       xm2, xm5
-    vinserti128     m2, m2, xm0, 1
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m1, m2
-
-%ifidn %1,sp
-    paddd           m6, m7
-    paddd           m1, m7
-    psrad           m6, 12
-    psrad           m1, 12
-%else
-    psrad           m6, 6
-    psrad           m1, 6
-%endif
-    packssdw        m6, m1
-%ifidn %1,sp
-    packuswb        m4, m6
-    vpermd          m4, m3, m4
-    vextracti128    xm6, m4, 1
-    movq            [r2], xm4
-    movhps          [r2 + r3], xm4
-    movq            [r2 + r3 * 2], xm6
-    movhps          [r2 + r4], xm6
-%else
-    vpermq          m4, m4, 11011000b
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm5, m4, 1
-    vextracti128    xm1, m6, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm5
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r4], xm1
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_8x8 sp
-    FILTER_VER_CHROMA_S_AVX2_8x8 ss
-
-%macro PROCESS_CHROMA_S_AVX2_W8_16R 1
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r7, [r0 + r1 * 4]
-    movu            xm4, [r7]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-    movu            xm5, [r7 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm7, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddwd         m7, m5, [r5 + 1 * mmsize]
-    paddd           m3, m7
-    pmaddwd         m5, [r5]
-%ifidn %1,sp
-    paddd           m0, m9
-    paddd           m1, m9
-    paddd           m2, m9
-    paddd           m3, m9
-    psrad           m0, 12
-    psrad           m1, 12
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m0, m1
-    packssdw        m2, m3
-%ifidn %1,sp
-    packuswb        m0, m2
-    mova            m3, [interp8_hps_shuf]
-    vpermd          m0, m3, m0
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-%endif
-
-    movu            xm7, [r7 + r4]                  ; m7 = row 7
-    punpckhwd       xm8, xm6, xm7
-    punpcklwd       xm6, xm7
-    vinserti128     m6, m6, xm8, 1
-    pmaddwd         m8, m6, [r5 + 1 * mmsize]
-    paddd           m4, m8
-    pmaddwd         m6, [r5]
-    lea             r7, [r7 + r1 * 4]
-    movu            xm8, [r7]                       ; m8 = row 8
-    punpckhwd       xm0, xm7, xm8
-    punpcklwd       xm7, xm8
-    vinserti128     m7, m7, xm0, 1
-    pmaddwd         m0, m7, [r5 + 1 * mmsize]
-    paddd           m5, m0
-    pmaddwd         m7, [r5]
-    movu            xm0, [r7 + r1]                  ; m0 = row 9
-    punpckhwd       xm1, xm8, xm0
-    punpcklwd       xm8, xm0
-    vinserti128     m8, m8, xm1, 1
-    pmaddwd         m1, m8, [r5 + 1 * mmsize]
-    paddd           m6, m1
-    pmaddwd         m8, [r5]
-    movu            xm1, [r7 + r1 * 2]              ; m1 = row 10
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m2, m0, [r5 + 1 * mmsize]
-    paddd           m7, m2
-    pmaddwd         m0, [r5]
-%ifidn %1,sp
-    paddd           m4, m9
-    paddd           m5, m9
-    psrad           m4, 12
-    psrad           m5, 12
-    paddd           m6, m9
-    paddd           m7, m9
-    psrad           m6, 12
-    psrad           m7, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-    psrad           m6, 6
-    psrad           m7, 6
-%endif
-    packssdw        m4, m5
-    packssdw        m6, m7
-    lea             r8, [r2 + r3 * 4]
-%ifidn %1,sp
-    packuswb        m4, m6
-    vpermd          m4, m3, m4
-    vextracti128    xm6, m4, 1
-    movq            [r8], xm4
-    movhps          [r8 + r3], xm4
-    movq            [r8 + r3 * 2], xm6
-    movhps          [r8 + r6], xm6
-%else
-    vpermq          m4, m4, 11011000b
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm5, m4, 1
-    vextracti128    xm7, m6, 1
-    movu            [r8], xm4
-    movu            [r8 + r3], xm5
-    movu            [r8 + r3 * 2], xm6
-    movu            [r8 + r6], xm7
-%endif
-
-    movu            xm2, [r7 + r4]                  ; m2 = row 11
-    punpckhwd       xm4, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm4, 1
-    pmaddwd         m4, m1, [r5 + 1 * mmsize]
-    paddd           m8, m4
-    pmaddwd         m1, [r5]
-    lea             r7, [r7 + r1 * 4]
-    movu            xm4, [r7]                       ; m4 = row 12
-    punpckhwd       xm5, xm2, xm4
-    punpcklwd       xm2, xm4
-    vinserti128     m2, m2, xm5, 1
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    paddd           m0, m5
-    pmaddwd         m2, [r5]
-    movu            xm5, [r7 + r1]                  ; m5 = row 13
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m1, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r7 + r1 * 2]              ; m6 = row 14
-    punpckhwd       xm7, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm7, 1
-    pmaddwd         m7, m5, [r5 + 1 * mmsize]
-    paddd           m2, m7
-    pmaddwd         m5, [r5]
-%ifidn %1,sp
-    paddd           m8, m9
-    paddd           m0, m9
-    paddd           m1, m9
-    paddd           m2, m9
-    psrad           m8, 12
-    psrad           m0, 12
-    psrad           m1, 12
-    psrad           m2, 12
-%else
-    psrad           m8, 6
-    psrad           m0, 6
-    psrad           m1, 6
-    psrad           m2, 6
-%endif
-    packssdw        m8, m0
-    packssdw        m1, m2
-    lea             r8, [r8 + r3 * 4]
-%ifidn %1,sp
-    packuswb        m8, m1
-    vpermd          m8, m3, m8
-    vextracti128    xm1, m8, 1
-    movq            [r8], xm8
-    movhps          [r8 + r3], xm8
-    movq            [r8 + r3 * 2], xm1
-    movhps          [r8 + r6], xm1
-%else
-    vpermq          m8, m8, 11011000b
-    vpermq          m1, m1, 11011000b
-    vextracti128    xm0, m8, 1
-    vextracti128    xm2, m1, 1
-    movu            [r8], xm8
-    movu            [r8 + r3], xm0
-    movu            [r8 + r3 * 2], xm1
-    movu            [r8 + r6], xm2
-%endif
-    lea             r8, [r8 + r3 * 4]
-
-    movu            xm7, [r7 + r4]                  ; m7 = row 15
-    punpckhwd       xm2, xm6, xm7
-    punpcklwd       xm6, xm7
-    vinserti128     m6, m6, xm2, 1
-    pmaddwd         m2, m6, [r5 + 1 * mmsize]
-    paddd           m4, m2
-    pmaddwd         m6, [r5]
-    lea             r7, [r7 + r1 * 4]
-    movu            xm2, [r7]                       ; m2 = row 16
-    punpckhwd       xm1, xm7, xm2
-    punpcklwd       xm7, xm2
-    vinserti128     m7, m7, xm1, 1
-    pmaddwd         m1, m7, [r5 + 1 * mmsize]
-    paddd           m5, m1
-    pmaddwd         m7, [r5]
-    movu            xm1, [r7 + r1]                  ; m1 = row 17
-    punpckhwd       xm0, xm2, xm1
-    punpcklwd       xm2, xm1
-    vinserti128     m2, m2, xm0, 1
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m6, m2
-    movu            xm0, [r7 + r1 * 2]              ; m0 = row 18
-    punpckhwd       xm2, xm1, xm0
-    punpcklwd       xm1, xm0
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m1, [r5 + 1 * mmsize]
-    paddd           m7, m1
-
-%ifidn %1,sp
-    paddd           m4, m9
-    paddd           m5, m9
-    paddd           m6, m9
-    paddd           m7, m9
-    psrad           m4, 12
-    psrad           m5, 12
-    psrad           m6, 12
-    psrad           m7, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-    psrad           m6, 6
-    psrad           m7, 6
-%endif
-    packssdw        m4, m5
-    packssdw        m6, m7
-%ifidn %1,sp
-    packuswb        m4, m6
-    vpermd          m4, m3, m4
-    vextracti128    xm6, m4, 1
-    movq            [r8], xm4
-    movhps          [r8 + r3], xm4
-    movq            [r8 + r3 * 2], xm6
-    movhps          [r8 + r6], xm6
-%else
-    vpermq          m4, m4, 11011000b
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm5, m4, 1
-    vextracti128    xm7, m6, 1
-    movu            [r8], xm4
-    movu            [r8 + r3], xm5
-    movu            [r8 + r3 * 2], xm6
-    movu            [r8 + r6], xm7
-%endif
-%endmacro
-
-%macro FILTER_VER_CHROMA_S_AVX2_Nx16 2
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_%2x16, 4, 10, 10
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m9, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-    mov             r9d, %2 / 8
-.loopW:
-    PROCESS_CHROMA_S_AVX2_W8_16R %1
-%ifidn %1,sp
-    add             r2, 8
-%else
-    add             r2, 16
-%endif
-    add             r0, 16
-    dec             r9d
-    jnz             .loopW
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_Nx16 sp, 16
-    FILTER_VER_CHROMA_S_AVX2_Nx16 sp, 32
-    FILTER_VER_CHROMA_S_AVX2_Nx16 sp, 64
-    FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 16
-    FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 32
-    FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 64
-
-%macro FILTER_VER_CHROMA_S_AVX2_NxN 3
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%3_%1x%2, 4, 11, 10
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %3,sp
-    mova            m9, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-    mov             r9d, %2 / 16
-.loopH:
-    mov             r10d, %1 / 8
-.loopW:
-    PROCESS_CHROMA_S_AVX2_W8_16R %3
-%ifidn %3,sp
-    add             r2, 8
-%else
-    add             r2, 16
-%endif
-    add             r0, 16
-    dec             r10d
-    jnz             .loopW
-    lea             r0, [r7 - 2 * %1 + 16]
-%ifidn %3,sp
-    lea             r2, [r8 + r3 * 4 - %1 + 8]
-%else
-    lea             r2, [r8 + r3 * 4 - 2 * %1 + 16]
-%endif
-    dec             r9d
-    jnz             .loopH
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_NxN 16, 32, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 24, 32, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 32, 32, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 16, 32, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 24, 32, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 32, 32, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 16, 64, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 24, 64, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 32, 64, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 32, 48, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 32, 48, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 16, 64, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 24, 64, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 32, 64, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 64, 64, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 64, 32, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 64, 48, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 48, 64, sp
-    FILTER_VER_CHROMA_S_AVX2_NxN 64, 64, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 64, 32, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 64, 48, ss
-    FILTER_VER_CHROMA_S_AVX2_NxN 48, 64, ss
-
-%macro PROCESS_CHROMA_S_AVX2_W8_4R 1
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m4, [r5 + 1 * mmsize]
-    paddd           m2, m4
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm4, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm4, 1
-    pmaddwd         m5, [r5 + 1 * mmsize]
-    paddd           m3, m5
-%ifidn %1,sp
-    paddd           m0, m7
-    paddd           m1, m7
-    paddd           m2, m7
-    paddd           m3, m7
-    psrad           m0, 12
-    psrad           m1, 12
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m0, m1
-    packssdw        m2, m3
-%ifidn %1,sp
-    packuswb        m0, m2
-    mova            m3, [interp8_hps_shuf]
-    vpermd          m0, m3, m0
-    vextracti128    xm2, m0, 1
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm1, m0, 1
-    vextracti128    xm3, m2, 1
-%endif
-%endmacro
-
-%macro FILTER_VER_CHROMA_S_AVX2_8x4 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x4, 4, 6, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-
-    PROCESS_CHROMA_S_AVX2_W8_4R %1
-    lea             r4, [r3 * 3]
-%ifidn %1,sp
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r4], xm2
-%else
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r4], xm3
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_8x4 sp
-    FILTER_VER_CHROMA_S_AVX2_8x4 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_12x16 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_12x16, 4, 9, 10
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m9, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-    PROCESS_CHROMA_S_AVX2_W8_16R %1
-%ifidn %1,sp
-    add             r2, 8
-%else
-    add             r2, 16
-%endif
-    add             r0, 16
-    mova            m7, m9
-    PROCESS_CHROMA_AVX2_W4_16R %1
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_12x16 sp
-    FILTER_VER_CHROMA_S_AVX2_12x16 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_12x32 1
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_12x32, 4, 9, 10
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1, sp
-    mova            m9, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-%rep 2
-    PROCESS_CHROMA_S_AVX2_W8_16R %1
-%ifidn %1, sp
-    add             r2, 8
-%else
-    add             r2, 16
-%endif
-    add             r0, 16
-    mova            m7, m9
-    PROCESS_CHROMA_AVX2_W4_16R %1
-    sub             r0, 16
-%ifidn %1, sp
-    lea             r2, [r2 + r3 * 4 - 8]
-%else
-    lea             r2, [r2 + r3 * 4 - 16]
-%endif
-%endrep
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_12x32 sp
-    FILTER_VER_CHROMA_S_AVX2_12x32 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_16x12 1
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_16x12, 4, 9, 9
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m8, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-%rep 2
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r7, [r0 + r1 * 4]
-    movu            xm4, [r7]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-%ifidn %1,sp
-    paddd           m0, m8
-    paddd           m1, m8
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-
-    movu            xm5, [r7 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm1, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m1
-%ifidn %1,sp
-    paddd           m2, m8
-    paddd           m3, m8
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m2, m3
-%ifidn %1,sp
-    packuswb        m0, m2
-    mova            m3, [interp8_hps_shuf]
-    vpermd          m0, m3, m0
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    movu            [r2], xm0
-    vextracti128    xm0, m0, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3], xm0
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-%endif
-    lea             r8, [r2 + r3 * 4]
-
-    movu            xm1, [r7 + r4]                  ; m1 = row 7
-    punpckhwd       xm0, xm6, xm1
-    punpcklwd       xm6, xm1
-    vinserti128     m6, m6, xm0, 1
-    pmaddwd         m0, m6, [r5 + 1 * mmsize]
-    pmaddwd         m6, [r5]
-    paddd           m4, m0
-    lea             r7, [r7 + r1 * 4]
-    movu            xm0, [r7]                       ; m0 = row 8
-    punpckhwd       xm2, xm1, xm0
-    punpcklwd       xm1, xm0
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    pmaddwd         m1, [r5]
-    paddd           m5, m2
-%ifidn %1,sp
-    paddd           m4, m8
-    paddd           m5, m8
-    psrad           m4, 12
-    psrad           m5, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-%endif
-    packssdw        m4, m5
-
-    movu            xm2, [r7 + r1]                  ; m2 = row 9
-    punpckhwd       xm5, xm0, xm2
-    punpcklwd       xm0, xm2
-    vinserti128     m0, m0, xm5, 1
-    pmaddwd         m5, m0, [r5 + 1 * mmsize]
-    paddd           m6, m5
-    pmaddwd         m0, [r5]
-    movu            xm5, [r7 + r1 * 2]              ; m5 = row 10
-    punpckhwd       xm7, xm2, xm5
-    punpcklwd       xm2, xm5
-    vinserti128     m2, m2, xm7, 1
-    pmaddwd         m7, m2, [r5 + 1 * mmsize]
-    paddd           m1, m7
-    pmaddwd         m2, [r5]
-
-%ifidn %1,sp
-    paddd           m6, m8
-    paddd           m1, m8
-    psrad           m6, 12
-    psrad           m1, 12
-%else
-    psrad           m6, 6
-    psrad           m1, 6
-%endif
-    packssdw        m6, m1
-%ifidn %1,sp
-    packuswb        m4, m6
-    vpermd          m4, m3, m4
-    vextracti128    xm6, m4, 1
-    movq            [r8], xm4
-    movhps          [r8 + r3], xm4
-    movq            [r8 + r3 * 2], xm6
-    movhps          [r8 + r6], xm6
-%else
-    vpermq          m4, m4, 11011000b
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm7, m4, 1
-    vextracti128    xm1, m6, 1
-    movu            [r8], xm4
-    movu            [r8 + r3], xm7
-    movu            [r8 + r3 * 2], xm6
-    movu            [r8 + r6], xm1
-%endif
-    lea             r8, [r8 + r3 * 4]
-
-    movu            xm7, [r7 + r4]                  ; m7 = row 11
-    punpckhwd       xm1, xm5, xm7
-    punpcklwd       xm5, xm7
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    paddd           m0, m1
-    pmaddwd         m5, [r5]
-    lea             r7, [r7 + r1 * 4]
-    movu            xm1, [r7]                       ; m1 = row 12
-    punpckhwd       xm4, xm7, xm1
-    punpcklwd       xm7, xm1
-    vinserti128     m7, m7, xm4, 1
-    pmaddwd         m4, m7, [r5 + 1 * mmsize]
-    paddd           m2, m4
-    pmaddwd         m7, [r5]
-%ifidn %1,sp
-    paddd           m0, m8
-    paddd           m2, m8
-    psrad           m0, 12
-    psrad           m2, 12
-%else
-    psrad           m0, 6
-    psrad           m2, 6
-%endif
-    packssdw        m0, m2
-
-    movu            xm4, [r7 + r1]                  ; m4 = row 13
-    punpckhwd       xm2, xm1, xm4
-    punpcklwd       xm1, xm4
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m1, [r5 + 1 * mmsize]
-    paddd           m5, m1
-    movu            xm2, [r7 + r1 * 2]              ; m2 = row 14
-    punpckhwd       xm6, xm4, xm2
-    punpcklwd       xm4, xm2
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m4, [r5 + 1 * mmsize]
-    paddd           m7, m4
-%ifidn %1,sp
-    paddd           m5, m8
-    paddd           m7, m8
-    psrad           m5, 12
-    psrad           m7, 12
-%else
-    psrad           m5, 6
-    psrad           m7, 6
-%endif
-    packssdw        m5, m7
-%ifidn %1,sp
-    packuswb        m0, m5
-    vpermd          m0, m3, m0
-    vextracti128    xm5, m0, 1
-    movq            [r8], xm0
-    movhps          [r8 + r3], xm0
-    movq            [r8 + r3 * 2], xm5
-    movhps          [r8 + r6], xm5
-    add             r2, 8
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m5, m5, 11011000b
-    vextracti128    xm7, m0, 1
-    vextracti128    xm6, m5, 1
-    movu            [r8], xm0
-    movu            [r8 + r3], xm7
-    movu            [r8 + r3 * 2], xm5
-    movu            [r8 + r6], xm6
-    add             r2, 16
-%endif
-    add             r0, 16
-%endrep
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_16x12 sp
-    FILTER_VER_CHROMA_S_AVX2_16x12 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_8x12 1
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x12, 4, 7, 9
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m8, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-%ifidn %1,sp
-    paddd           m0, m8
-    paddd           m1, m8
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm1, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m1
-%ifidn %1,sp
-    paddd           m2, m8
-    paddd           m3, m8
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m2, m3
-%ifidn %1,sp
-    packuswb        m0, m2
-    mova            m3, [interp8_hps_shuf]
-    vpermd          m0, m3, m0
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    movu            [r2], xm0
-    vextracti128    xm0, m0, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3], xm0
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm1, [r0 + r4]                  ; m1 = row 7
-    punpckhwd       xm0, xm6, xm1
-    punpcklwd       xm6, xm1
-    vinserti128     m6, m6, xm0, 1
-    pmaddwd         m0, m6, [r5 + 1 * mmsize]
-    pmaddwd         m6, [r5]
-    paddd           m4, m0
-    lea             r0, [r0 + r1 * 4]
-    movu            xm0, [r0]                       ; m0 = row 8
-    punpckhwd       xm2, xm1, xm0
-    punpcklwd       xm1, xm0
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    pmaddwd         m1, [r5]
-    paddd           m5, m2
-%ifidn %1,sp
-    paddd           m4, m8
-    paddd           m5, m8
-    psrad           m4, 12
-    psrad           m5, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-%endif
-    packssdw        m4, m5
-
-    movu            xm2, [r0 + r1]                  ; m2 = row 9
-    punpckhwd       xm5, xm0, xm2
-    punpcklwd       xm0, xm2
-    vinserti128     m0, m0, xm5, 1
-    pmaddwd         m5, m0, [r5 + 1 * mmsize]
-    paddd           m6, m5
-    pmaddwd         m0, [r5]
-    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
-    punpckhwd       xm7, xm2, xm5
-    punpcklwd       xm2, xm5
-    vinserti128     m2, m2, xm7, 1
-    pmaddwd         m7, m2, [r5 + 1 * mmsize]
-    paddd           m1, m7
-    pmaddwd         m2, [r5]
-
-%ifidn %1,sp
-    paddd           m6, m8
-    paddd           m1, m8
-    psrad           m6, 12
-    psrad           m1, 12
-%else
-    psrad           m6, 6
-    psrad           m1, 6
-%endif
-    packssdw        m6, m1
-%ifidn %1,sp
-    packuswb        m4, m6
-    vpermd          m4, m3, m4
-    vextracti128    xm6, m4, 1
-    movq            [r2], xm4
-    movhps          [r2 + r3], xm4
-    movq            [r2 + r3 * 2], xm6
-    movhps          [r2 + r6], xm6
-%else
-    vpermq          m4, m4, 11011000b
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm7, m4, 1
-    vextracti128    xm1, m6, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm7
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r6], xm1
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm7, [r0 + r4]                  ; m7 = row 11
-    punpckhwd       xm1, xm5, xm7
-    punpcklwd       xm5, xm7
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    paddd           m0, m1
-    pmaddwd         m5, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm1, [r0]                       ; m1 = row 12
-    punpckhwd       xm4, xm7, xm1
-    punpcklwd       xm7, xm1
-    vinserti128     m7, m7, xm4, 1
-    pmaddwd         m4, m7, [r5 + 1 * mmsize]
-    paddd           m2, m4
-    pmaddwd         m7, [r5]
-%ifidn %1,sp
-    paddd           m0, m8
-    paddd           m2, m8
-    psrad           m0, 12
-    psrad           m2, 12
-%else
-    psrad           m0, 6
-    psrad           m2, 6
-%endif
-    packssdw        m0, m2
-
-    movu            xm4, [r0 + r1]                  ; m4 = row 13
-    punpckhwd       xm2, xm1, xm4
-    punpcklwd       xm1, xm4
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m1, [r5 + 1 * mmsize]
-    paddd           m5, m1
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 14
-    punpckhwd       xm6, xm4, xm2
-    punpcklwd       xm4, xm2
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m4, [r5 + 1 * mmsize]
-    paddd           m7, m4
-%ifidn %1,sp
-    paddd           m5, m8
-    paddd           m7, m8
-    psrad           m5, 12
-    psrad           m7, 12
-%else
-    psrad           m5, 6
-    psrad           m7, 6
-%endif
-    packssdw        m5, m7
-%ifidn %1,sp
-    packuswb        m0, m5
-    vpermd          m0, m3, m0
-    vextracti128    xm5, m0, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm5
-    movhps          [r2 + r6], xm5
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m5, m5, 11011000b
-    vextracti128    xm7, m0, 1
-    vextracti128    xm6, m5, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm7
-    movu            [r2 + r3 * 2], xm5
-    movu            [r2 + r6], xm6
-%endif
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_8x12 sp
-    FILTER_VER_CHROMA_S_AVX2_8x12 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_16x4 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_16x4, 4, 7, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-%rep 2
-    PROCESS_CHROMA_S_AVX2_W8_4R %1
-    lea             r6, [r3 * 3]
-%ifidn %1,sp
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-    add             r2, 8
-%else
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-    add             r2, 16
-%endif
-    lea             r6, [4 * r1 - 16]
-    sub             r0, r6
-%endrep
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_16x4 sp
-    FILTER_VER_CHROMA_S_AVX2_16x4 ss
-
-%macro PROCESS_CHROMA_S_AVX2_W8_8R 1
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r7, [r0 + r1 * 4]
-    movu            xm4, [r7]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-%ifidn %1,sp
-    paddd           m0, m7
-    paddd           m1, m7
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-
-    movu            xm5, [r7 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm1, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m1
-%ifidn %1,sp
-    paddd           m2, m7
-    paddd           m3, m7
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m2, m3
-%ifidn %1,sp
-    packuswb        m0, m2
-    mova            m3, [interp8_hps_shuf]
-    vpermd          m0, m3, m0
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    movu            [r2], xm0
-    vextracti128    xm0, m0, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3], xm0
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-%endif
-    lea             r8, [r2 + r3 * 4]
-
-    movu            xm1, [r7 + r4]                  ; m1 = row 7
-    punpckhwd       xm0, xm6, xm1
-    punpcklwd       xm6, xm1
-    vinserti128     m6, m6, xm0, 1
-    pmaddwd         m0, m6, [r5 + 1 * mmsize]
-    pmaddwd         m6, [r5]
-    paddd           m4, m0
-    lea             r7, [r7 + r1 * 4]
-    movu            xm0, [r7]                       ; m0 = row 8
-    punpckhwd       xm2, xm1, xm0
-    punpcklwd       xm1, xm0
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    pmaddwd         m1, [r5]
-    paddd           m5, m2
-%ifidn %1,sp
-    paddd           m4, m7
-    paddd           m5, m7
-    psrad           m4, 12
-    psrad           m5, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-%endif
-    packssdw        m4, m5
-
-    movu            xm2, [r7 + r1]                  ; m2 = row 9
-    punpckhwd       xm5, xm0, xm2
-    punpcklwd       xm0, xm2
-    vinserti128     m0, m0, xm5, 1
-    pmaddwd         m0, [r5 + 1 * mmsize]
-    paddd           m6, m0
-    movu            xm5, [r7 + r1 * 2]              ; m5 = row 10
-    punpckhwd       xm0, xm2, xm5
-    punpcklwd       xm2, xm5
-    vinserti128     m2, m2, xm0, 1
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m1, m2
-
-%ifidn %1,sp
-    paddd           m6, m7
-    paddd           m1, m7
-    psrad           m6, 12
-    psrad           m1, 12
-%else
-    psrad           m6, 6
-    psrad           m1, 6
-%endif
-    packssdw        m6, m1
-%ifidn %1,sp
-    packuswb        m4, m6
-    vpermd          m4, m3, m4
-    vextracti128    xm6, m4, 1
-    movq            [r8], xm4
-    movhps          [r8 + r3], xm4
-    movq            [r8 + r3 * 2], xm6
-    movhps          [r8 + r6], xm6
-%else
-    vpermq          m4, m4, 11011000b
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm7, m4, 1
-    vextracti128    xm1, m6, 1
-    movu            [r8], xm4
-    movu            [r8 + r3], xm7
-    movu            [r8 + r3 * 2], xm6
-    movu            [r8 + r6], xm1
-%endif
-%endmacro
-
-%macro FILTER_VER_CHROMA_S_AVX2_Nx8 2
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_%2x8, 4, 9, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-%rep %2 / 8
-    PROCESS_CHROMA_S_AVX2_W8_8R %1
-%ifidn %1,sp
-    add             r2, 8
-%else
-    add             r2, 16
-%endif
-    add             r0, 16
-%endrep
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_Nx8 sp, 32
-    FILTER_VER_CHROMA_S_AVX2_Nx8 sp, 16
-    FILTER_VER_CHROMA_S_AVX2_Nx8 ss, 32
-    FILTER_VER_CHROMA_S_AVX2_Nx8 ss, 16
-
-%macro FILTER_VER_CHROMA_S_AVX2_8x2 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x2, 4, 6, 6
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m5, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m0, m2
-    movu            xm4, [r0 + r1 * 4]              ; m4 = row 4
-    punpckhwd       xm2, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm2, 1
-    pmaddwd         m3, [r5 + 1 * mmsize]
-    paddd           m1, m3
-%ifidn %1,sp
-    paddd           m0, m5
-    paddd           m1, m5
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-%ifidn %1,sp
-    vextracti128    xm1, m0, 1
-    packuswb        xm0, xm1
-    pshufd          xm0, xm0, 11011000b
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-%else
-    vpermq          m0, m0, 11011000b
-    vextracti128    xm1, m0, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm1
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_8x2 sp
-    FILTER_VER_CHROMA_S_AVX2_8x2 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_8x6 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x6, 4, 6, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m4
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    pmaddwd         m3, [r5]
-    paddd           m1, m5
-%ifidn %1,sp
-    paddd           m0, m7
-    paddd           m1, m7
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm1, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m1
-%ifidn %1,sp
-    paddd           m2, m7
-    paddd           m3, m7
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m2, m3
-
-    movu            xm1, [r0 + r4]                  ; m1 = row 7
-    punpckhwd       xm3, xm6, xm1
-    punpcklwd       xm6, xm1
-    vinserti128     m6, m6, xm3, 1
-    pmaddwd         m6, [r5 + 1 * mmsize]
-    paddd           m4, m6
-    movu            xm6, [r0 + r1 * 4]              ; m6 = row 8
-    punpckhwd       xm3, xm1, xm6
-    punpcklwd       xm1, xm6
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5 + 1 * mmsize]
-    paddd           m5, m1
-%ifidn %1,sp
-    paddd           m4, m7
-    paddd           m5, m7
-    psrad           m4, 12
-    psrad           m5, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-%endif
-    packssdw        m4, m5
-    lea             r4, [r3 * 3]
-%ifidn %1,sp
-    packuswb        m0, m2
-    mova            m3, [interp8_hps_shuf]
-    vpermd          m0, m3, m0
-    vextracti128    xm2, m0, 1
-    vextracti128    xm5, m4, 1
-    packuswb        xm4, xm5
-    pshufd          xm4, xm4, 11011000b
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r4], xm2
-    lea             r2, [r2 + r3 * 4]
-    movq            [r2], xm4
-    movhps          [r2 + r3], xm4
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    vpermq          m4, m4, 11011000b
-    movu            [r2], xm0
-    vextracti128    xm0, m0, 1
-    vextracti128    xm3, m2, 1
-    vextracti128    xm5, m4, 1
-    movu            [r2 + r3], xm0
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r4], xm3
-    lea             r2, [r2 + r3 * 4]
-    movu            [r2], xm4
-    movu            [r2 + r3], xm5
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_8x6 sp
-    FILTER_VER_CHROMA_S_AVX2_8x6 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_8xN 2
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_8x%2, 4, 7, 9
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m8, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-%rep %2 / 16
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-%ifidn %1,sp
-    paddd           m0, m8
-    paddd           m1, m8
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm1, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m1
-%ifidn %1,sp
-    paddd           m2, m8
-    paddd           m3, m8
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m2, m3
-%ifidn %1,sp
-    packuswb        m0, m2
-    mova            m3, [interp8_hps_shuf]
-    vpermd          m0, m3, m0
-    vextracti128    xm2, m0, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m2, m2, 11011000b
-    movu            [r2], xm0
-    vextracti128    xm0, m0, 1
-    vextracti128    xm3, m2, 1
-    movu            [r2 + r3], xm0
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm3
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm1, [r0 + r4]                  ; m1 = row 7
-    punpckhwd       xm0, xm6, xm1
-    punpcklwd       xm6, xm1
-    vinserti128     m6, m6, xm0, 1
-    pmaddwd         m0, m6, [r5 + 1 * mmsize]
-    pmaddwd         m6, [r5]
-    paddd           m4, m0
-    lea             r0, [r0 + r1 * 4]
-    movu            xm0, [r0]                       ; m0 = row 8
-    punpckhwd       xm2, xm1, xm0
-    punpcklwd       xm1, xm0
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    pmaddwd         m1, [r5]
-    paddd           m5, m2
-%ifidn %1,sp
-    paddd           m4, m8
-    paddd           m5, m8
-    psrad           m4, 12
-    psrad           m5, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-%endif
-    packssdw        m4, m5
-
-    movu            xm2, [r0 + r1]                  ; m2 = row 9
-    punpckhwd       xm5, xm0, xm2
-    punpcklwd       xm0, xm2
-    vinserti128     m0, m0, xm5, 1
-    pmaddwd         m5, m0, [r5 + 1 * mmsize]
-    paddd           m6, m5
-    pmaddwd         m0, [r5]
-    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
-    punpckhwd       xm7, xm2, xm5
-    punpcklwd       xm2, xm5
-    vinserti128     m2, m2, xm7, 1
-    pmaddwd         m7, m2, [r5 + 1 * mmsize]
-    paddd           m1, m7
-    pmaddwd         m2, [r5]
-
-%ifidn %1,sp
-    paddd           m6, m8
-    paddd           m1, m8
-    psrad           m6, 12
-    psrad           m1, 12
-%else
-    psrad           m6, 6
-    psrad           m1, 6
-%endif
-    packssdw        m6, m1
-%ifidn %1,sp
-    packuswb        m4, m6
-    vpermd          m4, m3, m4
-    vextracti128    xm6, m4, 1
-    movq            [r2], xm4
-    movhps          [r2 + r3], xm4
-    movq            [r2 + r3 * 2], xm6
-    movhps          [r2 + r6], xm6
-%else
-    vpermq          m4, m4, 11011000b
-    vpermq          m6, m6, 11011000b
-    vextracti128    xm7, m4, 1
-    vextracti128    xm1, m6, 1
-    movu            [r2], xm4
-    movu            [r2 + r3], xm7
-    movu            [r2 + r3 * 2], xm6
-    movu            [r2 + r6], xm1
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm7, [r0 + r4]                  ; m7 = row 11
-    punpckhwd       xm1, xm5, xm7
-    punpcklwd       xm5, xm7
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    paddd           m0, m1
-    pmaddwd         m5, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm1, [r0]                       ; m1 = row 12
-    punpckhwd       xm4, xm7, xm1
-    punpcklwd       xm7, xm1
-    vinserti128     m7, m7, xm4, 1
-    pmaddwd         m4, m7, [r5 + 1 * mmsize]
-    paddd           m2, m4
-    pmaddwd         m7, [r5]
-%ifidn %1,sp
-    paddd           m0, m8
-    paddd           m2, m8
-    psrad           m0, 12
-    psrad           m2, 12
-%else
-    psrad           m0, 6
-    psrad           m2, 6
-%endif
-    packssdw        m0, m2
-
-    movu            xm4, [r0 + r1]                  ; m4 = row 13
-    punpckhwd       xm2, xm1, xm4
-    punpcklwd       xm1, xm4
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    paddd           m5, m2
-    pmaddwd         m1, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 14
-    punpckhwd       xm6, xm4, xm2
-    punpcklwd       xm4, xm2
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m7, m6
-    pmaddwd         m4, [r5]
-%ifidn %1,sp
-    paddd           m5, m8
-    paddd           m7, m8
-    psrad           m5, 12
-    psrad           m7, 12
-%else
-    psrad           m5, 6
-    psrad           m7, 6
-%endif
-    packssdw        m5, m7
-%ifidn %1,sp
-    packuswb        m0, m5
-    vpermd          m0, m3, m0
-    vextracti128    xm5, m0, 1
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm5
-    movhps          [r2 + r6], xm5
-%else
-    vpermq          m0, m0, 11011000b
-    vpermq          m5, m5, 11011000b
-    vextracti128    xm7, m0, 1
-    vextracti128    xm6, m5, 1
-    movu            [r2], xm0
-    movu            [r2 + r3], xm7
-    movu            [r2 + r3 * 2], xm5
-    movu            [r2 + r6], xm6
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm6, [r0 + r4]                  ; m6 = row 15
-    punpckhwd       xm5, xm2, xm6
-    punpcklwd       xm2, xm6
-    vinserti128     m2, m2, xm5, 1
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm0, [r0]                       ; m0 = row 16
-    punpckhwd       xm5, xm6, xm0
-    punpcklwd       xm6, xm0
-    vinserti128     m6, m6, xm5, 1
-    pmaddwd         m5, m6, [r5 + 1 * mmsize]
-    paddd           m4, m5
-    pmaddwd         m6, [r5]
-%ifidn %1,sp
-    paddd           m1, m8
-    paddd           m4, m8
-    psrad           m1, 12
-    psrad           m4, 12
-%else
-    psrad           m1, 6
-    psrad           m4, 6
-%endif
-    packssdw        m1, m4
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 17
-    punpckhwd       xm4, xm0, xm5
-    punpcklwd       xm0, xm5
-    vinserti128     m0, m0, xm4, 1
-    pmaddwd         m0, [r5 + 1 * mmsize]
-    paddd           m2, m0
-    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
-    punpckhwd       xm0, xm5, xm4
-    punpcklwd       xm5, xm4
-    vinserti128     m5, m5, xm0, 1
-    pmaddwd         m5, [r5 + 1 * mmsize]
-    paddd           m6, m5
-%ifidn %1,sp
-    paddd           m2, m8
-    paddd           m6, m8
-    psrad           m2, 12
-    psrad           m6, 12
-%else
-    psrad           m2, 6
-    psrad           m6, 6
-%endif
-    packssdw        m2, m6
-%ifidn %1,sp
-    packuswb        m1, m2
-    vpermd          m1, m3, m1
-    vextracti128    xm2, m1, 1
-    movq            [r2], xm1
-    movhps          [r2 + r3], xm1
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-%else
-    vpermq          m1, m1, 11011000b
-    vpermq          m2, m2, 11011000b
-    vextracti128    xm6, m1, 1
-    vextracti128    xm4, m2, 1
-    movu            [r2], xm1
-    movu            [r2 + r3], xm6
-    movu            [r2 + r3 * 2], xm2
-    movu            [r2 + r6], xm4
-%endif
-    lea             r2, [r2 + r3 * 4]
-%endrep
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_8xN sp, 16
-    FILTER_VER_CHROMA_S_AVX2_8xN sp, 32
-    FILTER_VER_CHROMA_S_AVX2_8xN sp, 64
-    FILTER_VER_CHROMA_S_AVX2_8xN ss, 16
-    FILTER_VER_CHROMA_S_AVX2_8xN ss, 32
-    FILTER_VER_CHROMA_S_AVX2_8xN ss, 64
-
-%macro FILTER_VER_CHROMA_S_AVX2_Nx24 2
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_%2x24, 4, 10, 10
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m9, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-    mov             r9d, %2 / 8
-.loopW:
-    PROCESS_CHROMA_S_AVX2_W8_16R %1
-%ifidn %1,sp
-    add             r2, 8
-%else
-    add             r2, 16
-%endif
-    add             r0, 16
-    dec             r9d
-    jnz             .loopW
-%ifidn %1,sp
-    lea             r2, [r8 + r3 * 4 - %2 + 8]
-%else
-    lea             r2, [r8 + r3 * 4 - 2 * %2 + 16]
-%endif
-    lea             r0, [r7 - 2 * %2 + 16]
-    mova            m7, m9
-    mov             r9d, %2 / 8
-.loop:
-    PROCESS_CHROMA_S_AVX2_W8_8R %1
-%ifidn %1,sp
-    add             r2, 8
-%else
-    add             r2, 16
-%endif
-    add             r0, 16
-    dec             r9d
-    jnz             .loop
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_Nx24 sp, 32
-    FILTER_VER_CHROMA_S_AVX2_Nx24 sp, 16
-    FILTER_VER_CHROMA_S_AVX2_Nx24 ss, 32
-    FILTER_VER_CHROMA_S_AVX2_Nx24 ss, 16
-
-%macro FILTER_VER_CHROMA_S_AVX2_2x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_2x8, 4, 6, 7
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-%ifidn %1,sp
-    mova            m6, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    movd            xm0, [r0]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm0, xm1
-    movd            xm2, [r0 + r1 * 2]
-    punpcklwd       xm1, xm2
-    punpcklqdq      xm0, xm1                        ; m0 = [2 1 1 0]
-    movd            xm3, [r0 + r4]
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movd            xm4, [r0]
-    punpcklwd       xm3, xm4
-    punpcklqdq      xm2, xm3                        ; m2 = [4 3 3 2]
-    vinserti128     m0, m0, xm2, 1                  ; m0 = [4 3 3 2 2 1 1 0]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm4, xm1
-    movd            xm3, [r0 + r1 * 2]
-    punpcklwd       xm1, xm3
-    punpcklqdq      xm4, xm1                        ; m4 = [6 5 5 4]
-    vinserti128     m2, m2, xm4, 1                  ; m2 = [6 5 5 4 4 3 3 2]
-    pmaddwd         m0, [r5]
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m0, m2
-    movd            xm1, [r0 + r4]
-    punpcklwd       xm3, xm1
-    lea             r0, [r0 + 4 * r1]
-    movd            xm2, [r0]
-    punpcklwd       xm1, xm2
-    punpcklqdq      xm3, xm1                        ; m3 = [8 7 7 6]
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [8 7 7 6 6 5 5 4]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm2, xm1
-    movd            xm5, [r0 + r1 * 2]
-    punpcklwd       xm1, xm5
-    punpcklqdq      xm2, xm1                        ; m2 = [10 9 9 8]
-    vinserti128     m3, m3, xm2, 1                  ; m3 = [10 9 9 8 8 7 7 6]
-    pmaddwd         m4, [r5]
-    pmaddwd         m3, [r5 + 1 * mmsize]
-    paddd           m4, m3
-%ifidn %1,sp
-    paddd           m0, m6
-    paddd           m4, m6
-    psrad           m0, 12
-    psrad           m4, 12
-%else
-    psrad           m0, 6
-    psrad           m4, 6
-%endif
-    packssdw        m0, m4
-    vextracti128    xm4, m0, 1
-    lea             r4, [r3 * 3]
-%ifidn %1,sp
-    packuswb        xm0, xm4
-    pextrw          [r2], xm0, 0
-    pextrw          [r2 + r3], xm0, 1
-    pextrw          [r2 + 2 * r3], xm0, 4
-    pextrw          [r2 + r4], xm0, 5
-    lea             r2, [r2 + r3 * 4]
-    pextrw          [r2], xm0, 2
-    pextrw          [r2 + r3], xm0, 3
-    pextrw          [r2 + 2 * r3], xm0, 6
-    pextrw          [r2 + r4], xm0, 7
-%else
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    movd            [r2 + 2 * r3], xm4
-    pextrd          [r2 + r4], xm4, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm0, 2
-    pextrd          [r2 + r3], xm0, 3
-    pextrd          [r2 + 2 * r3], xm4, 2
-    pextrd          [r2 + r4], xm4, 3
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_2x8 sp
-    FILTER_VER_CHROMA_S_AVX2_2x8 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_2x16 1
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_2x16, 4, 6, 9
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-    sub             r0, r1
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-%ifidn %1,sp
-    mova            m6, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    movd            xm0, [r0]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm0, xm1
-    movd            xm2, [r0 + r1 * 2]
-    punpcklwd       xm1, xm2
-    punpcklqdq      xm0, xm1                        ; m0 = [2 1 1 0]
-    movd            xm3, [r0 + r4]
-    punpcklwd       xm2, xm3
-    lea             r0, [r0 + 4 * r1]
-    movd            xm4, [r0]
-    punpcklwd       xm3, xm4
-    punpcklqdq      xm2, xm3                        ; m2 = [4 3 3 2]
-    vinserti128     m0, m0, xm2, 1                  ; m0 = [4 3 3 2 2 1 1 0]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm4, xm1
-    movd            xm3, [r0 + r1 * 2]
-    punpcklwd       xm1, xm3
-    punpcklqdq      xm4, xm1                        ; m4 = [6 5 5 4]
-    vinserti128     m2, m2, xm4, 1                  ; m2 = [6 5 5 4 4 3 3 2]
-    pmaddwd         m0, [r5]
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m0, m2
-    movd            xm1, [r0 + r4]
-    punpcklwd       xm3, xm1
-    lea             r0, [r0 + 4 * r1]
-    movd            xm2, [r0]
-    punpcklwd       xm1, xm2
-    punpcklqdq      xm3, xm1                        ; m3 = [8 7 7 6]
-    vinserti128     m4, m4, xm3, 1                  ; m4 = [8 7 7 6 6 5 5 4]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm2, xm1
-    movd            xm5, [r0 + r1 * 2]
-    punpcklwd       xm1, xm5
-    punpcklqdq      xm2, xm1                        ; m2 = [10 9 9 8]
-    vinserti128     m3, m3, xm2, 1                  ; m3 = [10 9 9 8 8 7 7 6]
-    pmaddwd         m4, [r5]
-    pmaddwd         m3, [r5 + 1 * mmsize]
-    paddd           m4, m3
-    movd            xm1, [r0 + r4]
-    punpcklwd       xm5, xm1
-    lea             r0, [r0 + 4 * r1]
-    movd            xm3, [r0]
-    punpcklwd       xm1, xm3
-    punpcklqdq      xm5, xm1                        ; m5 = [12 11 11 10]
-    vinserti128     m2, m2, xm5, 1                  ; m2 = [12 11 11 10 10 9 9 8]
-    movd            xm1, [r0 + r1]
-    punpcklwd       xm3, xm1
-    movd            xm7, [r0 + r1 * 2]
-    punpcklwd       xm1, xm7
-    punpcklqdq      xm3, xm1                        ; m3 = [14 13 13 12]
-    vinserti128     m5, m5, xm3, 1                  ; m5 = [14 13 13 12 12 11 11 10]
-    pmaddwd         m2, [r5]
-    pmaddwd         m5, [r5 + 1 * mmsize]
-    paddd           m2, m5
-    movd            xm5, [r0 + r4]
-    punpcklwd       xm7, xm5
-    lea             r0, [r0 + 4 * r1]
-    movd            xm1, [r0]
-    punpcklwd       xm5, xm1
-    punpcklqdq      xm7, xm5                        ; m7 = [16 15 15 14]
-    vinserti128     m3, m3, xm7, 1                  ; m3 = [16 15 15 14 14 13 13 12]
-    movd            xm5, [r0 + r1]
-    punpcklwd       xm1, xm5
-    movd            xm8, [r0 + r1 * 2]
-    punpcklwd       xm5, xm8
-    punpcklqdq      xm1, xm5                        ; m1 = [18 17 17 16]
-    vinserti128     m7, m7, xm1, 1                  ; m7 = [18 17 17 16 16 15 15 14]
-    pmaddwd         m3, [r5]
-    pmaddwd         m7, [r5 + 1 * mmsize]
-    paddd           m3, m7
-%ifidn %1,sp
-    paddd           m0, m6
-    paddd           m4, m6
-    paddd           m2, m6
-    paddd           m3, m6
-    psrad           m0, 12
-    psrad           m4, 12
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m0, 6
-    psrad           m4, 6
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m0, m4
-    packssdw        m2, m3
-    lea             r4, [r3 * 3]
-%ifidn %1,sp
-    packuswb        m0, m2
-    vextracti128    xm2, m0, 1
-    pextrw          [r2], xm0, 0
-    pextrw          [r2 + r3], xm0, 1
-    pextrw          [r2 + 2 * r3], xm2, 0
-    pextrw          [r2 + r4], xm2, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrw          [r2], xm0, 2
-    pextrw          [r2 + r3], xm0, 3
-    pextrw          [r2 + 2 * r3], xm2, 2
-    pextrw          [r2 + r4], xm2, 3
-    lea             r2, [r2 + r3 * 4]
-    pextrw          [r2], xm0, 4
-    pextrw          [r2 + r3], xm0, 5
-    pextrw          [r2 + 2 * r3], xm2, 4
-    pextrw          [r2 + r4], xm2, 5
-    lea             r2, [r2 + r3 * 4]
-    pextrw          [r2], xm0, 6
-    pextrw          [r2 + r3], xm0, 7
-    pextrw          [r2 + 2 * r3], xm2, 6
-    pextrw          [r2 + r4], xm2, 7
-%else
-    vextracti128    xm4, m0, 1
-    vextracti128    xm3, m2, 1
-    movd            [r2], xm0
-    pextrd          [r2 + r3], xm0, 1
-    movd            [r2 + 2 * r3], xm4
-    pextrd          [r2 + r4], xm4, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm0, 2
-    pextrd          [r2 + r3], xm0, 3
-    pextrd          [r2 + 2 * r3], xm4, 2
-    pextrd          [r2 + r4], xm4, 3
-    lea             r2, [r2 + r3 * 4]
-    movd            [r2], xm2
-    pextrd          [r2 + r3], xm2, 1
-    movd            [r2 + 2 * r3], xm3
-    pextrd          [r2 + r4], xm3, 1
-    lea             r2, [r2 + r3 * 4]
-    pextrd          [r2], xm2, 2
-    pextrd          [r2 + r3], xm2, 3
-    pextrd          [r2 + 2 * r3], xm3, 2
-    pextrd          [r2 + r4], xm3, 3
-%endif
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_2x16 sp
-    FILTER_VER_CHROMA_S_AVX2_2x16 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_6x8 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_6x8, 4, 6, 8
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m7, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    pmaddwd         m2, [r5]
-    paddd           m0, m4
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    pmaddwd         m3, [r5]
-    paddd           m1, m5
-%ifidn %1,sp
-    paddd           m0, m7
-    paddd           m1, m7
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm1, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m1
-%ifidn %1,sp
-    paddd           m2, m7
-    paddd           m3, m7
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m2, m3
-
-    movu            xm1, [r0 + r4]                  ; m1 = row 7
-    punpckhwd       xm3, xm6, xm1
-    punpcklwd       xm6, xm1
-    vinserti128     m6, m6, xm3, 1
-    pmaddwd         m3, m6, [r5 + 1 * mmsize]
-    pmaddwd         m6, [r5]
-    paddd           m4, m3
-
-    lea             r4, [r3 * 3]
-%ifidn %1,sp
-    packuswb        m0, m2
-    vextracti128    xm2, m0, 1
-    movd            [r2], xm0
-    pextrw          [r2 + 4], xm2, 0
-    pextrd          [r2 + r3], xm0, 1
-    pextrw          [r2 + r3 + 4], xm2, 2
-    pextrd          [r2 + r3 * 2], xm0, 2
-    pextrw          [r2 + r3 * 2 + 4], xm2, 4
-    pextrd          [r2 + r4], xm0, 3
-    pextrw          [r2 + r4 + 4], xm2, 6
-%else
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r4], xm2
-    vextracti128    xm0, m0, 1
-    vextracti128    xm3, m2, 1
-    movd            [r2 + 8], xm0
-    pextrd          [r2 + r3 + 8], xm0, 2
-    movd            [r2 + r3 * 2 + 8], xm3
-    pextrd          [r2 + r4 + 8], xm3, 2
-%endif
-    lea             r2, [r2 + r3 * 4]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm0, [r0]                       ; m0 = row 8
-    punpckhwd       xm2, xm1, xm0
-    punpcklwd       xm1, xm0
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    pmaddwd         m1, [r5]
-    paddd           m5, m2
-%ifidn %1,sp
-    paddd           m4, m7
-    paddd           m5, m7
-    psrad           m4, 12
-    psrad           m5, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-%endif
-    packssdw        m4, m5
-
-    movu            xm2, [r0 + r1]                  ; m2 = row 9
-    punpckhwd       xm5, xm0, xm2
-    punpcklwd       xm0, xm2
-    vinserti128     m0, m0, xm5, 1
-    pmaddwd         m0, [r5 + 1 * mmsize]
-    paddd           m6, m0
-    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
-    punpckhwd       xm0, xm2, xm5
-    punpcklwd       xm2, xm5
-    vinserti128     m2, m2, xm0, 1
-    pmaddwd         m2, [r5 + 1 * mmsize]
-    paddd           m1, m2
-
-%ifidn %1,sp
-    paddd           m6, m7
-    paddd           m1, m7
-    psrad           m6, 12
-    psrad           m1, 12
-%else
-    psrad           m6, 6
-    psrad           m1, 6
-%endif
-    packssdw        m6, m1
-%ifidn %1,sp
-    packuswb        m4, m6
-    vextracti128    xm6, m4, 1
-    movd            [r2], xm4
-    pextrw          [r2 + 4], xm6, 0
-    pextrd          [r2 + r3], xm4, 1
-    pextrw          [r2 + r3 + 4], xm6, 2
-    pextrd          [r2 + r3 * 2], xm4, 2
-    pextrw          [r2 + r3 * 2 + 4], xm6, 4
-    pextrd          [r2 + r4], xm4, 3
-    pextrw          [r2 + r4 + 4], xm6, 6
-%else
-    movq            [r2], xm4
-    movhps          [r2 + r3], xm4
-    movq            [r2 + r3 * 2], xm6
-    movhps          [r2 + r4], xm6
-    vextracti128    xm5, m4, 1
-    vextracti128    xm1, m6, 1
-    movd            [r2 + 8], xm5
-    pextrd          [r2 + r3 + 8], xm5, 2
-    movd            [r2 + r3 * 2 + 8], xm1
-    pextrd          [r2 + r4 + 8], xm1, 2
-%endif
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_6x8 sp
-    FILTER_VER_CHROMA_S_AVX2_6x8 ss
-
-%macro FILTER_VER_CHROMA_S_AVX2_6x16 1
-%if ARCH_X86_64 == 1
-INIT_YMM avx2
-cglobal interp_4tap_vert_%1_6x16, 4, 7, 9
-    mov             r4d, r4m
-    shl             r4d, 6
-    add             r1d, r1d
-
-%ifdef PIC
-    lea             r5, [pw_ChromaCoeffV]
-    add             r5, r4
-%else
-    lea             r5, [pw_ChromaCoeffV + r4]
-%endif
-
-    lea             r4, [r1 * 3]
-    sub             r0, r1
-%ifidn %1,sp
-    mova            m8, [pd_526336]
-%else
-    add             r3d, r3d
-%endif
-    lea             r6, [r3 * 3]
-    movu            xm0, [r0]                       ; m0 = row 0
-    movu            xm1, [r0 + r1]                  ; m1 = row 1
-    punpckhwd       xm2, xm0, xm1
-    punpcklwd       xm0, xm1
-    vinserti128     m0, m0, xm2, 1
-    pmaddwd         m0, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
-    punpckhwd       xm3, xm1, xm2
-    punpcklwd       xm1, xm2
-    vinserti128     m1, m1, xm3, 1
-    pmaddwd         m1, [r5]
-    movu            xm3, [r0 + r4]                  ; m3 = row 3
-    punpckhwd       xm4, xm2, xm3
-    punpcklwd       xm2, xm3
-    vinserti128     m2, m2, xm4, 1
-    pmaddwd         m4, m2, [r5 + 1 * mmsize]
-    paddd           m0, m4
-    pmaddwd         m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm4, [r0]                       ; m4 = row 4
-    punpckhwd       xm5, xm3, xm4
-    punpcklwd       xm3, xm4
-    vinserti128     m3, m3, xm5, 1
-    pmaddwd         m5, m3, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m3, [r5]
-%ifidn %1,sp
-    paddd           m0, m8
-    paddd           m1, m8
-    psrad           m0, 12
-    psrad           m1, 12
-%else
-    psrad           m0, 6
-    psrad           m1, 6
-%endif
-    packssdw        m0, m1
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 5
-    punpckhwd       xm6, xm4, xm5
-    punpcklwd       xm4, xm5
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m2, m6
-    pmaddwd         m4, [r5]
-    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
-    punpckhwd       xm1, xm5, xm6
-    punpcklwd       xm5, xm6
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    pmaddwd         m5, [r5]
-    paddd           m3, m1
-%ifidn %1,sp
-    paddd           m2, m8
-    paddd           m3, m8
-    psrad           m2, 12
-    psrad           m3, 12
-%else
-    psrad           m2, 6
-    psrad           m3, 6
-%endif
-    packssdw        m2, m3
-%ifidn %1,sp
-    packuswb        m0, m2
-    vextracti128    xm2, m0, 1
-    movd            [r2], xm0
-    pextrw          [r2 + 4], xm2, 0
-    pextrd          [r2 + r3], xm0, 1
-    pextrw          [r2 + r3 + 4], xm2, 2
-    pextrd          [r2 + r3 * 2], xm0, 2
-    pextrw          [r2 + r3 * 2 + 4], xm2, 4
-    pextrd          [r2 + r6], xm0, 3
-    pextrw          [r2 + r6 + 4], xm2, 6
-%else
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-    vextracti128    xm0, m0, 1
-    vextracti128    xm3, m2, 1
-    movd            [r2 + 8], xm0
-    pextrd          [r2 + r3 + 8], xm0, 2
-    movd            [r2 + r3 * 2 + 8], xm3
-    pextrd          [r2 + r6 + 8], xm3, 2
-%endif
-    lea             r2, [r2 + r3 * 4]
-    movu            xm1, [r0 + r4]                  ; m1 = row 7
-    punpckhwd       xm0, xm6, xm1
-    punpcklwd       xm6, xm1
-    vinserti128     m6, m6, xm0, 1
-    pmaddwd         m0, m6, [r5 + 1 * mmsize]
-    pmaddwd         m6, [r5]
-    paddd           m4, m0
-    lea             r0, [r0 + r1 * 4]
-    movu            xm0, [r0]                       ; m0 = row 8
-    punpckhwd       xm2, xm1, xm0
-    punpcklwd       xm1, xm0
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    pmaddwd         m1, [r5]
-    paddd           m5, m2
-%ifidn %1,sp
-    paddd           m4, m8
-    paddd           m5, m8
-    psrad           m4, 12
-    psrad           m5, 12
-%else
-    psrad           m4, 6
-    psrad           m5, 6
-%endif
-    packssdw        m4, m5
-
-    movu            xm2, [r0 + r1]                  ; m2 = row 9
-    punpckhwd       xm5, xm0, xm2
-    punpcklwd       xm0, xm2
-    vinserti128     m0, m0, xm5, 1
-    pmaddwd         m5, m0, [r5 + 1 * mmsize]
-    paddd           m6, m5
-    pmaddwd         m0, [r5]
-    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
-    punpckhwd       xm7, xm2, xm5
-    punpcklwd       xm2, xm5
-    vinserti128     m2, m2, xm7, 1
-    pmaddwd         m7, m2, [r5 + 1 * mmsize]
-    paddd           m1, m7
-    pmaddwd         m2, [r5]
-
-%ifidn %1,sp
-    paddd           m6, m8
-    paddd           m1, m8
-    psrad           m6, 12
-    psrad           m1, 12
-%else
-    psrad           m6, 6
-    psrad           m1, 6
-%endif
-    packssdw        m6, m1
-%ifidn %1,sp
-    packuswb        m4, m6
-    vextracti128    xm6, m4, 1
-    movd            [r2], xm4
-    pextrw          [r2 + 4], xm6, 0
-    pextrd          [r2 + r3], xm4, 1
-    pextrw          [r2 + r3 + 4], xm6, 2
-    pextrd          [r2 + r3 * 2], xm4, 2
-    pextrw          [r2 + r3 * 2 + 4], xm6, 4
-    pextrd          [r2 + r6], xm4, 3
-    pextrw          [r2 + r6 + 4], xm6, 6
-%else
-    movq            [r2], xm4
-    movhps          [r2 + r3], xm4
-    movq            [r2 + r3 * 2], xm6
-    movhps          [r2 + r6], xm6
-    vextracti128    xm4, m4, 1
-    vextracti128    xm1, m6, 1
-    movd            [r2 + 8], xm4
-    pextrd          [r2 + r3 + 8], xm4, 2
-    movd            [r2 + r3 * 2 + 8], xm1
-    pextrd          [r2 + r6 + 8], xm1, 2
-%endif
-    lea             r2, [r2 + r3 * 4]
-    movu            xm7, [r0 + r4]                  ; m7 = row 11
-    punpckhwd       xm1, xm5, xm7
-    punpcklwd       xm5, xm7
-    vinserti128     m5, m5, xm1, 1
-    pmaddwd         m1, m5, [r5 + 1 * mmsize]
-    paddd           m0, m1
-    pmaddwd         m5, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm1, [r0]                       ; m1 = row 12
-    punpckhwd       xm4, xm7, xm1
-    punpcklwd       xm7, xm1
-    vinserti128     m7, m7, xm4, 1
-    pmaddwd         m4, m7, [r5 + 1 * mmsize]
-    paddd           m2, m4
-    pmaddwd         m7, [r5]
-%ifidn %1,sp
-    paddd           m0, m8
-    paddd           m2, m8
-    psrad           m0, 12
-    psrad           m2, 12
-%else
-    psrad           m0, 6
-    psrad           m2, 6
-%endif
-    packssdw        m0, m2
-
-    movu            xm4, [r0 + r1]                  ; m4 = row 13
-    punpckhwd       xm2, xm1, xm4
-    punpcklwd       xm1, xm4
-    vinserti128     m1, m1, xm2, 1
-    pmaddwd         m2, m1, [r5 + 1 * mmsize]
-    paddd           m5, m2
-    pmaddwd         m1, [r5]
-    movu            xm2, [r0 + r1 * 2]              ; m2 = row 14
-    punpckhwd       xm6, xm4, xm2
-    punpcklwd       xm4, xm2
-    vinserti128     m4, m4, xm6, 1
-    pmaddwd         m6, m4, [r5 + 1 * mmsize]
-    paddd           m7, m6
-    pmaddwd         m4, [r5]
-%ifidn %1,sp
-    paddd           m5, m8
-    paddd           m7, m8
-    psrad           m5, 12
-    psrad           m7, 12
-%else
-    psrad           m5, 6
-    psrad           m7, 6
-%endif
-    packssdw        m5, m7
-%ifidn %1,sp
-    packuswb        m0, m5
-    vextracti128    xm5, m0, 1
-    movd            [r2], xm0
-    pextrw          [r2 + 4], xm5, 0
-    pextrd          [r2 + r3], xm0, 1
-    pextrw          [r2 + r3 + 4], xm5, 2
-    pextrd          [r2 + r3 * 2], xm0, 2
-    pextrw          [r2 + r3 * 2 + 4], xm5, 4
-    pextrd          [r2 + r6], xm0, 3
-    pextrw          [r2 + r6 + 4], xm5, 6
-%else
-    movq            [r2], xm0
-    movhps          [r2 + r3], xm0
-    movq            [r2 + r3 * 2], xm5
-    movhps          [r2 + r6], xm5
-    vextracti128    xm0, m0, 1
-    vextracti128    xm7, m5, 1
-    movd            [r2 + 8], xm0
-    pextrd          [r2 + r3 + 8], xm0, 2
-    movd            [r2 + r3 * 2 + 8], xm7
-    pextrd          [r2 + r6 + 8], xm7, 2
-%endif
-    lea             r2, [r2 + r3 * 4]
-
-    movu            xm6, [r0 + r4]                  ; m6 = row 15
-    punpckhwd       xm5, xm2, xm6
-    punpcklwd       xm2, xm6
-    vinserti128     m2, m2, xm5, 1
-    pmaddwd         m5, m2, [r5 + 1 * mmsize]
-    paddd           m1, m5
-    pmaddwd         m2, [r5]
-    lea             r0, [r0 + r1 * 4]
-    movu            xm0, [r0]                       ; m0 = row 16
-    punpckhwd       xm5, xm6, xm0
-    punpcklwd       xm6, xm0
-    vinserti128     m6, m6, xm5, 1
-    pmaddwd         m5, m6, [r5 + 1 * mmsize]
-    paddd           m4, m5
-    pmaddwd         m6, [r5]
-%ifidn %1,sp
-    paddd           m1, m8
-    paddd           m4, m8
-    psrad           m1, 12
-    psrad           m4, 12
-%else
-    psrad           m1, 6
-    psrad           m4, 6
-%endif
-    packssdw        m1, m4
-
-    movu            xm5, [r0 + r1]                  ; m5 = row 17
-    punpckhwd       xm4, xm0, xm5
-    punpcklwd       xm0, xm5
-    vinserti128     m0, m0, xm4, 1
-    pmaddwd         m0, [r5 + 1 * mmsize]
-    paddd           m2, m0
-    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
-    punpckhwd       xm0, xm5, xm4
-    punpcklwd       xm5, xm4
-    vinserti128     m5, m5, xm0, 1
-    pmaddwd         m5, [r5 + 1 * mmsize]
-    paddd           m6, m5
-%ifidn %1,sp
-    paddd           m2, m8
-    paddd           m6, m8
-    psrad           m2, 12
-    psrad           m6, 12
-%else
-    psrad           m2, 6
-    psrad           m6, 6
-%endif
-    packssdw        m2, m6
-%ifidn %1,sp
-    packuswb        m1, m2
-    vextracti128    xm2, m1, 1
-    movd            [r2], xm1
-    pextrw          [r2 + 4], xm2, 0
-    pextrd          [r2 + r3], xm1, 1
-    pextrw          [r2 + r3 + 4], xm2, 2
-    pextrd          [r2 + r3 * 2], xm1, 2
-    pextrw          [r2 + r3 * 2 + 4], xm2, 4
-    pextrd          [r2 + r6], xm1, 3
-    pextrw          [r2 + r6 + 4], xm2, 6
-%else
-    movq            [r2], xm1
-    movhps          [r2 + r3], xm1
-    movq            [r2 + r3 * 2], xm2
-    movhps          [r2 + r6], xm2
-    vextracti128    xm4, m1, 1
-    vextracti128    xm6, m2, 1
-    movd            [r2 + 8], xm4
-    pextrd          [r2 + r3 + 8], xm4, 2
-    movd            [r2 + r3 * 2 + 8], xm6
-    pextrd          [r2 + r6 + 8], xm6, 2
-%endif
-    RET
-%endif
-%endmacro
-
-    FILTER_VER_CHROMA_S_AVX2_6x16 sp
-    FILTER_VER_CHROMA_S_AVX2_6x16 ss
-
-;---------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vertical_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SS_W2_4R 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ss_%1x%2, 5, 6, 5
-
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 5
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       r4d, (%2/4)
-
-.loopH:
-    PROCESS_CHROMA_SP_W2_4R r5
-
-    psrad     m0, 6
-    psrad     m2, 6
-
-    packssdw  m0, m2
-
-    movd      [r2], m0
-    pextrd    [r2 + r3], m0, 1
-    lea       r2, [r2 + 2 * r3]
-    pextrd    [r2], m0, 2
-    pextrd    [r2 + r3], m0, 3
-
-    lea       r2, [r2 + 2 * r3]
-
-    dec       r4d
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SS_W2_4R 2, 4
-    FILTER_VER_CHROMA_SS_W2_4R 2, 8
-
-    FILTER_VER_CHROMA_SS_W2_4R 2, 16
-
-;---------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ss_4x2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;---------------------------------------------------------------------------------------------------------------
-INIT_XMM sse2
-cglobal interp_4tap_vert_ss_4x2, 5, 6, 4
-
-    add        r1d, r1d
-    add        r3d, r3d
-    sub        r0, r1
-    shl        r4d, 5
-
-%ifdef PIC
-    lea        r5, [tab_ChromaCoeffV]
-    lea        r5, [r5 + r4]
-%else
-    lea        r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    movq       m0, [r0]
-    movq       m1, [r0 + r1]
-    punpcklwd  m0, m1                          ;m0=[0 1]
-    pmaddwd    m0, [r5 + 0 *16]                ;m0=[0+1]  Row1
-
-    lea        r0, [r0 + 2 * r1]
-    movq       m2, [r0]
-    punpcklwd  m1, m2                          ;m1=[1 2]
-    pmaddwd    m1, [r5 + 0 *16]                ;m1=[1+2]  Row2
-
-    movq       m3, [r0 + r1]
-    punpcklwd  m2, m3                          ;m4=[2 3]
-    pmaddwd    m2, [r5 + 1 * 16]
-    paddd      m0, m2                          ;m0=[0+1+2+3]  Row1 done
-    psrad      m0, 6
-
-    movq       m2, [r0 + 2 * r1]
-    punpcklwd  m3, m2                          ;m5=[3 4]
-    pmaddwd    m3, [r5 + 1 * 16]
-    paddd      m1, m3                          ;m1=[1+2+3+4]  Row2 done
-    psrad      m1, 6
-
-    packssdw   m0, m1
-
-    movlps     [r2], m0
-    movhps     [r2 + r3], m0
-
-    RET
-
-;-------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vertical_ss_6x8(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;-------------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SS_W6_H4 2
-INIT_XMM sse4
-cglobal interp_4tap_vert_ss_6x%2, 5, 7, 6
-
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 5
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r6, [r5 + r4]
-%else
-    lea       r6, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       r4d, %2/4
-
-.loopH:
-    PROCESS_CHROMA_SP_W4_4R
-
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-
-    movlps    [r2], m0
-    movhps    [r2 + r3], m0
-    lea       r5, [r2 + 2 * r3]
-    movlps    [r5], m2
-    movhps    [r5 + r3], m2
-
-    lea       r5, [4 * r1 - 2 * 4]
-    sub       r0, r5
-    add       r2, 2 * 4
-
-    PROCESS_CHROMA_SP_W2_4R r6
-
-    psrad     m0, 6
-    psrad     m2, 6
-
-    packssdw  m0, m2
-
-    movd      [r2], m0
-    pextrd    [r2 + r3], m0, 1
-    lea       r2, [r2 + 2 * r3]
-    pextrd    [r2], m0, 2
-    pextrd    [r2 + r3], m0, 3
-
-    sub       r0, 2 * 4
-    lea       r2, [r2 + 2 * r3 - 2 * 4]
-
-    dec       r4d
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SS_W6_H4 6, 8
-
-    FILTER_VER_CHROMA_SS_W6_H4 6, 16
-
-
-;----------------------------------------------------------------------------------------------------------------
-; void interp_4tap_vert_ss_8x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
-;----------------------------------------------------------------------------------------------------------------
-%macro FILTER_VER_CHROMA_SS_W8_H2 2
-INIT_XMM sse2
-cglobal interp_4tap_vert_ss_%1x%2, 5, 6, 7
-
-    add       r1d, r1d
-    add       r3d, r3d
-    sub       r0, r1
-    shl       r4d, 5
-
-%ifdef PIC
-    lea       r5, [tab_ChromaCoeffV]
-    lea       r5, [r5 + r4]
-%else
-    lea       r5, [tab_ChromaCoeffV + r4]
-%endif
-
-    mov       r4d, %2/2
-.loopH:
-    PROCESS_CHROMA_SP_W8_2R
-
-    psrad     m0, 6
-    psrad     m1, 6
-    psrad     m2, 6
-    psrad     m3, 6
-
-    packssdw  m0, m1
-    packssdw  m2, m3
-
-    movu      [r2], m0
-    movu      [r2 + r3], m2
-
-    lea       r2, [r2 + 2 * r3]
-
-    dec       r4d
-    jnz       .loopH
-
-    RET
-%endmacro
-
-    FILTER_VER_CHROMA_SS_W8_H2 8, 2
-    FILTER_VER_CHROMA_SS_W8_H2 8, 4
-    FILTER_VER_CHROMA_SS_W8_H2 8, 6
-    FILTER_VER_CHROMA_SS_W8_H2 8, 8
-    FILTER_VER_CHROMA_SS_W8_H2 8, 16
-    FILTER_VER_CHROMA_SS_W8_H2 8, 32
-
-    FILTER_VER_CHROMA_SS_W8_H2 8, 12
-    FILTER_VER_CHROMA_SS_W8_H2 8, 64
-
 ;-----------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
 ;-----------------------------------------------------------------------------------------------------------------
@@ -25464,2816 +9353,3 @@
 
     FILTER_VER_LUMA_S_AVX2_32x24 sp
     FILTER_VER_LUMA_S_AVX2_32x24 ss
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_32x32(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------;
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_32x32, 4,6,8
-    mov             r4d, r4m
-    add             r3d, r3d
-    dec             r0
-
-    ; check isRowExt
-    cmp             r5m, byte 0
-
-    lea             r5, [tab_ChromaCoeff]
-    vpbroadcastw    m0, [r5 + r4 * 4 + 0]
-    vpbroadcastw    m1, [r5 + r4 * 4 + 2]
-    mova            m7, [pw_2000]
-
-    ; register map
-    ; m0 - interpolate coeff Low
-    ; m1 - interpolate coeff High
-    ; m7 - constant pw_2000
-    mov             r4d, 32
-    je             .loop
-    sub             r0, r1
-    add             r4d, 3
-
-.loop
-    ; Row 0
-    movu            m2, [r0]
-    movu            m3, [r0 + 1]
-    punpckhbw       m4, m2, m3
-    punpcklbw       m2, m3
-    pmaddubsw       m4, m0
-    pmaddubsw       m2, m0
-
-    movu            m3, [r0 + 2]
-    movu            m5, [r0 + 3]
-    punpckhbw       m6, m3, m5
-    punpcklbw       m3, m5
-    pmaddubsw       m6, m1
-    pmaddubsw       m3, m1
-
-    paddw           m4, m6
-    paddw           m2, m3
-    psubw           m4, m7
-    psubw           m2, m7
-    vperm2i128      m3, m2, m4, 0x20
-    vperm2i128      m5, m2, m4, 0x31
-    movu            [r2], m3
-    movu            [r2 + mmsize], m5
-
-    add             r2, r3
-    add             r0, r1
-    dec             r4d
-    jnz            .loop
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_16x16(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------;
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_16x16, 4,7,6
-    mov             r4d, r4m
-    mov             r5d, r5m
-    add             r3d, r3d
-
-%ifdef PIC
-    lea               r6,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,           [pw_1]
-    vbroadcasti128     m5,           [pw_2000]
-    mova               m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-    mov                r6d,         16
-    dec                r0
-    test                r5d,        r5d
-    je                 .loop
-    sub                r0 ,         r1
-    add                r6d ,        3
-
-.loop
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 8]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-    vpermq            m3,           m3,          11011000b
-    movu              [r2],         m3
-
-    add                r2,          r3
-    add                r0,          r1
-    dec                r6d
-    jnz                .loop
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_16xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_PS_16xN_AVX2 2
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_%1x%2, 4,7,6
-    mov                    r4d,        r4m
-    mov                    r5d,        r5m
-    add                    r3d,        r3d
-
-%ifdef PIC
-    lea                    r6,         [tab_ChromaCoeff]
-    vpbroadcastd           m0,         [r6 + r4 * 4]
-%else
-    vpbroadcastd           m0,         [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128         m2,         [pw_1]
-    vbroadcasti128         m5,         [pw_2000]
-    mova                   m1,         [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-    mov                    r6d,        %2
-    dec                    r0
-    test                   r5d,        r5d
-    je                     .loop
-    sub                    r0 ,        r1
-    add                    r6d ,       3
-
-.loop
-    ; Row 0
-    vbroadcasti128         m3,         [r0]
-    pshufb                 m3,         m1
-    pmaddubsw              m3,         m0
-    pmaddwd                m3,         m2
-    vbroadcasti128         m4,         [r0 + 8]
-    pshufb                 m4,         m1
-    pmaddubsw              m4,         m0
-    pmaddwd                m4,         m2
-
-    packssdw               m3,         m4
-    psubw                  m3,         m5
-
-    vpermq                 m3,         m3,          11011000b
-    movu                   [r2],       m3
-
-    add                    r2,         r3
-    add                    r0,         r1
-    dec                    r6d
-    jnz                    .loop
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 32
-    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 12
-    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 8
-    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 4
-    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 24
-    IPFILTER_CHROMA_PS_16xN_AVX2  16 , 64
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_32xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_PS_32xN_AVX2 2
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_%1x%2, 4,7,6
-    mov                r4d,          r4m
-    mov                r5d,          r5m
-    add                r3d,          r3d
-
-%ifdef PIC
-    lea                r6,           [tab_ChromaCoeff]
-    vpbroadcastd       m0,           [r6 + r4 * 4]
-%else
-    vpbroadcastd       m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,           [pw_1]
-    vbroadcasti128     m5,           [pw_2000]
-    mova               m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-    mov                r6d,          %2
-    dec                r0
-    test               r5d,          r5d
-    je                 .loop
-    sub                r0 ,          r1
-    add                r6d ,         3
-
-.loop
-    ; Row 0
-    vbroadcasti128     m3,           [r0]
-    pshufb             m3,           m1
-    pmaddubsw          m3,           m0
-    pmaddwd            m3,           m2
-    vbroadcasti128     m4,           [r0 + 8]
-    pshufb             m4,           m1
-    pmaddubsw          m4,           m0
-    pmaddwd            m4,           m2
-
-    packssdw           m3,           m4
-    psubw              m3,           m5
-
-    vpermq             m3,           m3,          11011000b
-    movu              [r2],          m3
-
-    vbroadcasti128     m3,           [r0 + 16]
-    pshufb             m3,           m1
-    pmaddubsw          m3,           m0
-    pmaddwd            m3,           m2
-    vbroadcasti128     m4,           [r0 + 24]
-    pshufb             m4,           m1
-    pmaddubsw          m4,           m0
-    pmaddwd            m4,           m2
-
-    packssdw           m3,           m4
-    psubw              m3,           m5
-
-    vpermq             m3,           m3,          11011000b
-    movu               [r2 + 32],    m3
-
-    add                r2,           r3
-    add                r0,           r1
-    dec                r6d
-    jnz                .loop
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 16
-    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 24
-    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 8
-    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 64
-    IPFILTER_CHROMA_PS_32xN_AVX2  32 , 48
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_4x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_4x4, 4,7,5
-    mov             r4d, r4m
-    mov             r5d, r5m
-    add             r3d, r3d
-
-%ifdef PIC
-    lea               r6,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,           [pw_1]
-    vbroadcasti128     m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec                r0
-    test                r5d,       r5d
-    je                 .label
-    sub                r0 , r1
-
-.label
-    ; Row 0-1
-    movu              xm3,           [r0]
-    vinserti128       m3,           m3,      [r0 + r1],     1
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 2-3
-    lea               r0,           [r0 + r1 * 2]
-    movu              xm4,           [r0]
-    vinserti128       m4,           m4,      [r0 + r1],     1
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           [pw_2000]
-    vextracti128      xm4,          m3,     1
-    movq              [r2],         xm3
-    movq              [r2+r3],      xm4
-    lea               r2,           [r2 + r3 * 2]
-    movhps            [r2],         xm3
-    movhps            [r2 + r3],    xm4
-
-    test                r5d,        r5d
-    jz                .end
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1 * 2]
-
-    ;Row 5-6
-    movu              xm3,          [r0]
-    vinserti128       m3,           m3,      [r0 + r1],     1
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 7
-    lea               r0,           [r0 + r1 * 2]
-    vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           [pw_2000]
-
-    vextracti128      xm4,          m3,     1
-    movq              [r2],         xm3
-    movq              [r2+r3],      xm4
-    lea               r2,           [r2 + r3 * 2]
-    movhps            [r2],         xm3
-.end
-    RET
-
-cglobal interp_4tap_horiz_ps_4x2, 4,7,5
-    mov             r4d, r4m
-    mov             r5d, r5m
-    add             r3d, r3d
-
-%ifdef PIC
-    lea               r6,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,           [pw_1]
-    vbroadcasti128     m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec                r0
-    test                r5d,       r5d
-    je                 .label
-    sub                r0 , r1
-
-.label
-    ; Row 0-1
-    movu              xm3,           [r0]
-    vinserti128       m3,           m3,      [r0 + r1],     1
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    packssdw          m3,           m3
-    psubw             m3,           [pw_2000]
-    vextracti128      xm4,          m3,     1
-    movq              [r2],         xm3
-    movq              [r2+r3],      xm4
-
-    test              r5d,          r5d
-    jz                .end
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1 * 2]
-
-    ;Row 2-3
-    movu              xm3,          [r0]
-    vinserti128       m3,           m3,      [r0 + r1],     1
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 5
-    lea               r0,           [r0 + r1 * 2]
-    vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           [pw_2000]
-
-    vextracti128      xm4,          m3,     1
-    movq              [r2],         xm3
-    movq              [r2+r3],      xm4
-    lea               r2,           [r2 + r3 * 2]
-    movhps            [r2],         xm3
-.end
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_4xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------;
-%macro IPFILTER_CHROMA_PS_4xN_AVX2 2
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_%1x%2, 4,7,5
-    mov             r4d, r4m
-    mov             r5d, r5m
-    add             r3d, r3d
-
-%ifdef PIC
-    lea               r6,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,           [pw_1]
-    vbroadcasti128     m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    mov              r4,                %2
-    dec              r0
-    test             r5d,       r5d
-    je               .loop
-    sub              r0 ,               r1
-
-
-.loop
-    sub              r4d,           4
-    ; Row 0-1
-    movu              xm3,          [r0]
-    vinserti128       m3,           m3,      [r0 + r1],     1
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 2-3
-    lea               r0,           [r0 + r1 * 2]
-    movu              xm4,          [r0]
-    vinserti128       m4,           m4,      [r0 + r1],     1
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           [pw_2000]
-    vextracti128      xm4,          m3,     1
-    movq              [r2],         xm3
-    movq              [r2+r3],      xm4
-    lea               r2,           [r2 + r3 * 2]
-    movhps            [r2],         xm3
-    movhps            [r2 + r3],    xm4
-
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1 * 2]
-
-    test              r4d,          r4d
-    jnz               .loop
-    test                r5d,        r5d
-    jz                .end
-
-    ;Row 5-6
-    movu              xm3,          [r0]
-    vinserti128       m3,           m3,      [r0 + r1],     1
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 7
-    lea               r0,           [r0 + r1 * 2]
-    vbroadcasti128    m4,           [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           [pw_2000]
-
-    vextracti128      xm4,          m3,     1
-    movq              [r2],         xm3
-    movq              [r2+r3],      xm4
-    lea               r2,           [r2 + r3 * 2]
-    movhps            [r2],         xm3
-.end
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PS_4xN_AVX2  4 , 8
-    IPFILTER_CHROMA_PS_4xN_AVX2  4 , 16
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_8x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------;
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_8x8, 4,7,6
-    mov             r4d, r4m
-    mov             r5d, r5m
-    add             r3d, r3d
-
-%ifdef PIC
-    lea               r6,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,           [pw_1]
-    vbroadcasti128     m5,           [pw_2000]
-    mova               m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    mov                r6d,      4
-    dec                r0
-    test                r5d,     r5d
-    je                 .loop
-    sub                r0 ,      r1
-    add                r6d ,     1
-
-.loop
-     dec               r6d
-    ; Row 0
-    vbroadcasti128    m3,           [r0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 1
-    vbroadcasti128    m4,           [r0 + r1]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-
-    vpermq            m3,           m3,          11011000b
-    vextracti128      xm4,          m3,     1
-    movu             [r2],         xm3
-    movu             [r2 + r3],    xm4
-
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1 * 2]
-    test               r6d,          r6d
-    jnz               .loop
-    test              r5d,         r5d
-    je                .end
-
-    ;Row 11
-    vbroadcasti128    m3,           [r0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    packssdw          m3,           m3
-    psubw             m3,           m5
-    vpermq            m3,           m3,          11011000b
-    movu             [r2],         xm3
-.end
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_4x2, 4,6,4
-    mov             r4d, r4m
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128    m1,           [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-
-    ; Row 0-1
-    movu              xm2,          [r0 - 1]
-    vinserti128       m2,           m2,      [r0 + r1 - 1],     1
-    pshufb            m2,           m1
-    pmaddubsw         m2,           m0
-    pmaddwd           m2,           [pw_1]
-
-    packssdw          m2,           m2
-    pmulhrsw          m2,           [pw_512]
-    vextracti128      xm3,          m2,     1
-    packuswb          xm2,          xm3
-
-    movd              [r2],         xm2
-    pextrd            [r2+r3],      xm2,     2
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_32xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_PP_32xN_AVX2 2
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_%1x%2, 4,6,7
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-    mova              m6,           [pw_512]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          %2
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    vbroadcasti128    m4,           [r0 + 16]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + 20]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-
-    movu              [r2],         m3
-    add               r2,           r3
-    add               r0,           r1
-    dec               r4d
-    jnz               .loop
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PP_32xN_AVX2 32, 16
-    IPFILTER_CHROMA_PP_32xN_AVX2 32, 24
-    IPFILTER_CHROMA_PP_32xN_AVX2 32, 8
-    IPFILTER_CHROMA_PP_32xN_AVX2 32, 64
-    IPFILTER_CHROMA_PP_32xN_AVX2 32, 48
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_8xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_PP_8xN_AVX2 2
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_%1x%2, 4,6,6
-    mov               r4d,    r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    movu              m1,           [tab_Tm]
-    vpbroadcastd      m2,           [pw_1]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    sub               r0,           1
-    mov               r4d,          %2
-
-.loop:
-    sub               r4d,          4
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 1
-    vbroadcasti128    m4,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           [pw_512]
-    lea               r0,           [r0 + r1 * 2]
-
-    ; Row 2
-    vbroadcasti128    m4,           [r0 ]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    ; Row 3
-    vbroadcasti128    m5,           [r0 + r1]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           [pw_512]
-
-    packuswb          m3,           m4
-    mova              m5,           [interp_4tap_8x8_horiz_shuf]
-    vpermd            m3,           m5,     m3
-    vextracti128      xm4,          m3,     1
-    movq              [r2],         xm3
-    movhps            [r2 + r3],    xm3
-    lea               r2,           [r2 + r3 * 2]
-    movq              [r2],         xm4
-    movhps            [r2 + r3],    xm4
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1*2]
-    test              r4d,          r4d
-    jnz               .loop
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 16
-    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 32
-    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 4
-    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 64
-    IPFILTER_CHROMA_PP_8xN_AVX2   8 , 12
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_4xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_PP_4xN_AVX2 2
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_%1x%2, 4,6,6
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vpbroadcastd      m2,           [pw_1]
-    vbroadcasti128    m1,           [tab_Tm]
-    mov               r4d,          %2
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec                r0
-
-.loop
-    sub               r4d,          4
-    ; Row 0-1
-    movu              xm3,          [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    vinserti128       m3,           m3,      [r0 + r1],     1
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-
-    ; Row 2-3
-    lea               r0,           [r0 + r1 * 2]
-    movu              xm4,          [r0]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    vinserti128       m4,           m4,      [r0 + r1],     1
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    pmulhrsw          m3,           [pw_512]
-    vextracti128      xm4,          m3,                     1
-    packuswb          xm3,          xm4
-
-    movd              [r2],         xm3
-    pextrd            [r2+r3],      xm3,                    2
-    lea               r2,           [r2 + r3 * 2]
-    pextrd            [r2],         xm3,                    1
-    pextrd            [r2+r3],      xm3,                    3
-
-    lea               r0,           [r0 + r1 * 2]
-    lea               r2,           [r2 + r3 * 2]
-    test              r4d,          r4d
-    jnz               .loop
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PP_4xN_AVX2  4 , 8
-    IPFILTER_CHROMA_PP_4xN_AVX2  4 , 16
-
-%macro IPFILTER_LUMA_PS_32xN_AVX2 2
-INIT_YMM avx2
-cglobal interp_8tap_horiz_ps_%1x%2, 4, 7, 8
-    mov                         r5d,               r5m
-    mov                         r4d,               r4m
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    vpbroadcastq                m0,                [r6 + r4 * 8]
-%else
-    vpbroadcastq                m0,                [tab_LumaCoeff + r4 * 8]
-%endif
-    mova                        m6,                [tab_Lm + 32]
-    mova                        m1,                [tab_Lm]
-    mov                         r4d,                %2                           ;height
-    add                         r3d,               r3d
-    vbroadcasti128              m2,                [pw_1]
-    mova                        m7,                [interp8_hps_shuf]
-
-    ; register map
-    ; m0      - interpolate coeff
-    ; m1 , m6 - shuffle order table
-    ; m2      - pw_1
-
-
-    sub                         r0,                3
-    test                        r5d,               r5d
-    jz                          .label
-    lea                         r6,                [r1 * 3]                     ; r8 = (N / 2 - 1) * srcStride
-    sub                         r0,                r6
-    add                         r4d,                7
-
-.label
-    lea                         r6,                 [pw_2000]
-.loop
-    ; Row 0
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
-    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-
-
-    vbroadcasti128              m4,                [r0 + 8]
-    pshufb                      m5,                m4,            m6            ;row 0 (col 12 to 15)
-    pshufb                      m4,                m1                           ;row 0 (col 8 to 11)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m2
-    pmaddwd                     m5,                m2
-    packssdw                    m4,                m5
-
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-    vpermd                      m3,                m7,               m3
-    psubw                       m3,                [r6]
-
-    movu                        [r2],              m3                          ;row 0
-
-    vbroadcasti128              m3,                [r0 + 16]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 20 to 23)
-    pshufb                      m3,                m1                           ; row 0 (col 16 to 19)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 24]
-    pshufb                      m5,                m4,            m6            ;row 0 (col 28 to 31)
-    pshufb                      m4,                m1                           ;row 0 (col 24 to 27)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m2
-    pmaddwd                     m5,                m2
-    packssdw                    m4,                m5
-
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-    vpermd                      m3,                m7,               m3
-    psubw                       m3,                [r6]
-
-    movu                        [r2 + 32],         m3                          ;row 0
-
-    add                         r0,                r1
-    add                         r2,                r3
-    dec                         r4d
-    jnz                         .loop
-    RET
-%endmacro
-
-    IPFILTER_LUMA_PS_32xN_AVX2 32 , 32
-    IPFILTER_LUMA_PS_32xN_AVX2 32 , 16
-    IPFILTER_LUMA_PS_32xN_AVX2 32 , 24
-    IPFILTER_LUMA_PS_32xN_AVX2 32 , 8
-    IPFILTER_LUMA_PS_32xN_AVX2 32 , 64
-
-INIT_YMM avx2
-cglobal interp_8tap_horiz_ps_48x64, 4, 7, 8
-    mov                         r5d,               r5m
-    mov                         r4d,               r4m
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    vpbroadcastq                m0,                [r6 + r4 * 8]
-%else
-    vpbroadcastq                m0,                [tab_LumaCoeff + r4 * 8]
-%endif
-    mova                        m6,                [tab_Lm + 32]
-    mova                        m1,                [tab_Lm]
-    mov                         r4d,               64                           ;height
-    add                         r3d,               r3d
-    vbroadcasti128              m2,                [pw_2000]
-    mova                        m7,                [pw_1]
-
-    ; register map
-    ; m0      - interpolate coeff
-    ; m1 , m6 - shuffle order table
-    ; m2      - pw_2000
-
-    sub                         r0,                3
-    test                        r5d,               r5d
-    jz                          .label
-    lea                         r6,                [r1 * 3]                     ; r6 = (N / 2 - 1) * srcStride
-    sub                         r0,                r6                           ; r0(src)-r6
-    add                         r4d,                7                            ; blkheight += N - 1  (7 - 1 = 6 ; since the last one row not in loop)
-
-.label
-    lea                         r6,                [interp8_hps_shuf]
-.loop
-    ; Row 0
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
-    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 8]
-    pshufb                      m5,                m4,             m6            ;row 0 (col 12 to 15)
-    pshufb                      m4,                m1                           ;row 0 (col 8 to 11)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m7
-    pmaddwd                     m5,                m7
-    packssdw                    m4,                m5
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-    mova                        m5,                [r6]
-    vpermd                      m3,                m5,             m3
-    psubw                       m3,                m2
-    movu                        [r2],              m3                          ;row 0
-
-    vbroadcasti128              m3,                [r0 + 16]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 20 to 23)
-    pshufb                      m3,                m1                           ; row 0 (col 16 to 19)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 24]
-    pshufb                      m5,                m4,             m6            ;row 0 (col 28 to 31)
-    pshufb                      m4,                m1                           ;row 0 (col 24 to 27)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m7
-    pmaddwd                     m5,                m7
-    packssdw                    m4,                m5
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-    mova                        m5,                [r6]
-    vpermd                      m3,                m5,               m3
-    psubw                       m3,                m2
-    movu                        [r2 + 32],         m3                          ;row 0
-
-    vbroadcasti128              m3,                [r0 + 32]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 36 to 39)
-    pshufb                      m3,                m1                           ; row 0 (col 32 to 35)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 40]
-    pshufb                      m5,                m4,            m6            ;row 0 (col 44 to 47)
-    pshufb                      m4,                m1                           ;row 0 (col 40 to 43)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m7
-    pmaddwd                     m5,                m7
-    packssdw                    m4,                m5
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-    mova                        m5,                [r6]
-    vpermd                      m3,                m5,               m3
-    psubw                       m3,                m2
-    movu                        [r2 + 64],         m3                          ;row 0
-
-    add                         r0,                r1
-    add                         r2,                r3
-    dec                         r4d
-    jnz                         .loop
-    RET
-
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_24x32, 4,6,8
-    sub               r0,         3
-    mov               r4d,        r4m
-%ifdef PIC
-    lea               r5,         [tab_LumaCoeff]
-    vpbroadcastd      m0,         [r5 + r4 * 8]
-    vpbroadcastd      m1,         [r5 + r4 * 8 + 4]
-%else
-    vpbroadcastd      m0,         [tab_LumaCoeff + r4 * 8]
-    vpbroadcastd      m1,         [tab_LumaCoeff + r4 * 8 + 4]
-%endif
-    movu              m3,         [tab_Tm + 16]
-    vpbroadcastd      m7,         [pw_1]
-    lea               r5,         [tab_Tm]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-    ; m2 , m2  shuffle order table
-    ; m7 - pw_1
-
-    mov               r4d,        32
-.loop:
-    ; Row 0
-    vbroadcasti128    m4,         [r0]                        ; [x E D C B A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,         m4,     m3
-    pshufb            m4,         [r5]
-    pmaddubsw         m4,         m0
-    pmaddubsw         m5,         m1
-    paddw             m4,         m5
-    pmaddwd           m4,         m7
-
-    vbroadcasti128    m5,         [r0 + 8]
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [r5]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
-    pmulhrsw          m4,         [pw_512]
-
-    vbroadcasti128    m2,         [r0 + 16]
-    pshufb            m5,         m2,     m3
-    pshufb            m2,         [r5]
-    pmaddubsw         m2,         m0
-    pmaddubsw         m5,         m1
-    paddw             m2,         m5
-    pmaddwd           m2,         m7
-
-    packssdw          m2,         m2
-    pmulhrsw          m2,         [pw_512]
-    packuswb          m4,         m2
-    vpermq            m4,         m4,     11011000b
-    vextracti128      xm5,        m4,     1
-    pshufd            xm4,        xm4,    11011000b
-    pshufd            xm5,        xm5,    11011000b
-
-    movu              [r2],       xm4
-    movq              [r2 + 16],  xm5
-    add               r0,         r1
-    add               r2,         r3
-    dec               r4d
-    jnz               .loop
-    RET
-
-INIT_YMM avx2
-cglobal interp_8tap_horiz_pp_12x16, 4,6,8
-    sub               r0,        3
-    mov               r4d,       r4m
-%ifdef PIC
-    lea               r5,        [tab_LumaCoeff]
-    vpbroadcastd      m0,        [r5 + r4 * 8]
-    vpbroadcastd      m1,        [r5 + r4 * 8 + 4]
-%else
-    vpbroadcastd      m0,         [tab_LumaCoeff + r4 * 8]
-    vpbroadcastd      m1,         [tab_LumaCoeff + r4 * 8 + 4]
-%endif
-    movu              m3,         [tab_Tm + 16]
-    vpbroadcastd      m7,         [pw_1]
-    lea               r5,         [tab_Tm]
-
-    ; register map
-    ; m0 , m1 interpolate coeff
-    ; m2 , m2  shuffle order table
-    ; m7 - pw_1
-
-    mov               r4d,        8
-.loop:
-    ; Row 0
-    vbroadcasti128    m4,         [r0]                        ;first 8 element
-    pshufb            m5,         m4,     m3
-    pshufb            m4,         [r5]
-    pmaddubsw         m4,         m0
-    pmaddubsw         m5,         m1
-    paddw             m4,         m5
-    pmaddwd           m4,         m7
-
-    vbroadcasti128    m5,         [r0 + 8]                    ; element 8 to 11
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [r5]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-
-    packssdw          m4,         m5                          ; [17 16 15 14 07 06 05 04 13 12 11 10 03 02 01 00]
-    pmulhrsw          m4,         [pw_512]
-
-    ;Row 1
-    vbroadcasti128    m2,         [r0 + r1]
-    pshufb            m5,         m2,     m3
-    pshufb            m2,         [r5]
-    pmaddubsw         m2,         m0
-    pmaddubsw         m5,         m1
-    paddw             m2,         m5
-    pmaddwd           m2,         m7
-
-    vbroadcasti128    m5,         [r0 + r1 + 8]
-    pshufb            m6,         m5,     m3
-    pshufb            m5,         [r5]
-    pmaddubsw         m5,         m0
-    pmaddubsw         m6,         m1
-    paddw             m5,         m6
-    pmaddwd           m5,         m7
-
-    packssdw          m2,         m5
-    pmulhrsw          m2,         [pw_512]
-    packuswb          m4,         m2
-    vpermq            m4,         m4,     11011000b
-    vextracti128      xm5,        m4,     1
-    pshufd            xm4,        xm4,    11011000b
-    pshufd            xm5,        xm5,    11011000b
-
-    movq              [r2],       xm4
-    pextrd            [r2+8],     xm4,    2
-    movq              [r2 + r3],  xm5
-    pextrd            [r2+r3+8],  xm5,    2
-    lea               r0,         [r0 + r1 * 2]
-    lea               r2,         [r2 + r3 * 2]
-    dec               r4d
-    jnz              .loop
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_16xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_PP_16xN_AVX2 2
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_%1x%2, 4, 6, 7
-    mov               r4d,          r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m6,           [pw_512]
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          %2/2
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    ; Row 1
-    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-
-    vextracti128      xm4,          m3,       1
-    movu              [r2],         xm3
-    movu              [r2 + r3],    xm4
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1 * 2]
-    dec               r4d
-    jnz               .loop
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 8
-    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 32
-    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 12
-    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 4
-    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 64
-    IPFILTER_CHROMA_PP_16xN_AVX2 16 , 24
-
-%macro IPFILTER_LUMA_PS_64xN_AVX2 1
-INIT_YMM avx2
-cglobal interp_8tap_horiz_ps_64x%1, 4, 7, 8
-    mov                         r5d,               r5m
-    mov                         r4d,               r4m
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    vpbroadcastq                m0,                [r6 + r4 * 8]
-%else
-    vpbroadcastq                m0,                [tab_LumaCoeff + r4 * 8]
-%endif
-    mova                        m6,                [tab_Lm + 32]
-    mova                        m1,                [tab_Lm]
-    mov                         r4d,               %1                           ;height
-    add                         r3d,               r3d
-    vbroadcasti128              m2,                [pw_1]
-    mova                        m7,                [interp8_hps_shuf]
-
-    ; register map
-    ; m0      - interpolate coeff
-    ; m1 , m6 - shuffle order table
-    ; m2      - pw_2000
-
-    sub                         r0,                3
-    test                        r5d,               r5d
-    jz                          .label
-    lea                         r6,                [r1 * 3]
-    sub                         r0,                r6                           ; r0(src)-r6
-    add                         r4d,               7                            ; blkheight += N - 1
-
-.label
-    lea                         r6,                [pw_2000]
-.loop
-    ; Row 0
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
-    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 8]
-    pshufb                      m5,                m4,            m6            ;row 0 (col 12 to 15)
-    pshufb                      m4,                m1                           ;row 0 (col 8 to 11)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m2
-    pmaddwd                     m5,                m2
-    packssdw                    m4,                m5
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-    vpermd                      m3,                m7,               m3
-    psubw                       m3,                [r6]
-    movu                        [r2],              m3                          ;row 0
-
-    vbroadcasti128              m3,                [r0 + 16]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 20 to 23)
-    pshufb                      m3,                m1                           ; row 0 (col 16 to 19)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 24]
-    pshufb                      m5,                m4,            m6            ;row 0 (col 28 to 31)
-    pshufb                      m4,                m1                           ;row 0 (col 24 to 27)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m2
-    pmaddwd                     m5,                m2
-    packssdw                    m4,                m5
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-    vpermd                      m3,                m7,               m3
-    psubw                       m3,                [r6]
-    movu                        [r2 + 32],         m3                          ;row 0
-
-    vbroadcasti128              m3,                [r0 + 32]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 36 to 39)
-    pshufb                      m3,                m1                           ; row 0 (col 32 to 35)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 40]
-    pshufb                      m5,                m4,            m6            ;row 0 (col 44 to 47)
-    pshufb                      m4,                m1                           ;row 0 (col 40 to 43)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m2
-    pmaddwd                     m5,                m2
-    packssdw                    m4,                m5
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-    vpermd                      m3,                m7,               m3
-    psubw                       m3,                [r6]
-    movu                        [r2 + 64],         m3                          ;row 0
-    vbroadcasti128              m3,                [r0 + 48]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 52 to 55)
-    pshufb                      m3,                m1                           ; row 0 (col 48 to 51)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 56]
-    pshufb                      m5,                m4,            m6            ;row 0 (col 60 to 63)
-    pshufb                      m4,                m1                           ;row 0 (col 56 to 59)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m2
-    pmaddwd                     m5,                m2
-    packssdw                    m4,                m5
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4
-    vpermd                      m3,                m7,               m3
-    psubw                       m3,                [r6]
-    movu                        [r2 + 96],         m3                          ;row 0
-
-    add                          r0,                r1
-    add                          r2,                r3
-    dec                          r4d
-    jnz                         .loop
-    RET
-%endmacro
-
-    IPFILTER_LUMA_PS_64xN_AVX2 64
-    IPFILTER_LUMA_PS_64xN_AVX2 48
-    IPFILTER_LUMA_PS_64xN_AVX2 32
-    IPFILTER_LUMA_PS_64xN_AVX2 16
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_8xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_PS_8xN_AVX2 1
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_8x%1, 4,7,6
-    mov                r4d,             r4m
-    mov                r5d,             r5m
-    add                r3d,             r3d
-
-%ifdef PIC
-    lea                r6,              [tab_ChromaCoeff]
-    vpbroadcastd       m0,              [r6 + r4 * 4]
-%else
-    vpbroadcastd       m0,              [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,              [pw_1]
-    vbroadcasti128     m5,              [pw_2000]
-    mova               m1,              [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    mov                r6d,             %1/2
-    dec                r0
-    test               r5d,             r5d
-    jz                 .loop
-    sub                r0 ,             r1
-    inc                r6d
-
-.loop
-    ; Row 0
-    vbroadcasti128     m3,              [r0]
-    pshufb             m3,              m1
-    pmaddubsw          m3,              m0
-    pmaddwd            m3,              m2
-
-    ; Row 1
-    vbroadcasti128     m4,              [r0 + r1]
-    pshufb             m4,              m1
-    pmaddubsw          m4,              m0
-    pmaddwd            m4,              m2
-    packssdw           m3,              m4
-    psubw              m3,              m5
-    vpermq             m3,              m3,          11011000b
-    vextracti128       xm4,             m3,          1
-    movu               [r2],            xm3
-    movu               [r2 + r3],       xm4
-
-    lea                r2,              [r2 + r3 * 2]
-    lea                r0,              [r0 + r1 * 2]
-    dec                r6d
-    jnz                .loop
-    test               r5d,             r5d
-    jz                 .end
-
-    ;Row 11
-    vbroadcasti128     m3,              [r0]
-    pshufb             m3,              m1
-    pmaddubsw          m3,              m0
-    pmaddwd            m3,              m2
-    packssdw           m3,              m3
-    psubw              m3,              m5
-    vpermq             m3,              m3,          11011000b
-    movu               [r2],            xm3
-.end
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PS_8xN_AVX2  2
-    IPFILTER_CHROMA_PS_8xN_AVX2  32
-    IPFILTER_CHROMA_PS_8xN_AVX2  16
-    IPFILTER_CHROMA_PS_8xN_AVX2  6
-    IPFILTER_CHROMA_PS_8xN_AVX2  4
-    IPFILTER_CHROMA_PS_8xN_AVX2  12
-    IPFILTER_CHROMA_PS_8xN_AVX2  64
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_2x4, 4, 7, 3
-    mov                r4d,            r4m
-    mov                r5d,            r5m
-    add                r3d,            r3d
-%ifdef PIC
-    lea                r6,             [tab_ChromaCoeff]
-    vpbroadcastd       m0,             [r6 + r4 * 4]
-%else
-    vpbroadcastd       m0,             [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova               xm3,            [pw_2000]
-    dec                r0
-    test               r5d,            r5d
-    jz                 .label
-    sub                r0,             r1
-
-.label
-    lea                r6,             [r1 * 3]
-    movq               xm1,            [r0]
-    movhps             xm1,            [r0 + r1]
-    movq               xm2,            [r0 + r1 * 2]
-    movhps             xm2,            [r0 + r6]
-
-    vinserti128        m1,             m1,          xm2,          1
-    pshufb             m1,             [interp4_hpp_shuf]
-    pmaddubsw          m1,             m0
-    pmaddwd            m1,             [pw_1]
-    vextracti128       xm2,            m1,          1
-    packssdw           xm1,            xm2
-    psubw              xm1,            xm3
-
-    lea                r4,             [r3 * 3]
-    movd               [r2],           xm1
-    pextrd             [r2 + r3],      xm1,         1
-    pextrd             [r2 + r3 * 2],  xm1,         2
-    pextrd             [r2 + r4],      xm1,         3
-
-    test               r5d,            r5d
-    jz                .end
-    lea                r2,             [r2 + r3 * 4]
-    lea                r0,             [r0 + r1 * 4]
-
-    movq               xm1,            [r0]
-    movhps             xm1,            [r0 + r1]
-    movq               xm2,            [r0 + r1 * 2]
-    vinserti128        m1,             m1,          xm2,          1
-    pshufb             m1,             [interp4_hpp_shuf]
-    pmaddubsw          m1,             m0
-    pmaddwd            m1,             [pw_1]
-    vextracti128       xm2,            m1,          1
-    packssdw           xm1,            xm2
-    psubw              xm1,            xm3
-
-    movd               [r2],           xm1
-    pextrd             [r2 + r3],      xm1,         1
-    pextrd             [r2 + r3 * 2],  xm1,         2
-.end
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_2x8, 4, 7, 7
-    mov               r4d,           r4m
-    mov               r5d,           r5m
-    add               r3d,           r3d
-
-%ifdef PIC
-    lea               r6,            [tab_ChromaCoeff]
-    vpbroadcastd      m0,            [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,            [tab_ChromaCoeff + r4 * 4]
-%endif
-    vbroadcasti128    m6,            [pw_2000]
-    test              r5d,            r5d
-    jz                .label
-    sub               r0,             r1
-
-.label
-    mova              m4,            [interp4_hpp_shuf]
-    mova              m5,            [pw_1]
-    dec               r0
-    lea               r4,            [r1 * 3]
-    movq              xm1,           [r0]                                   ;row 0
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m1,            m1,          xm2,          1
-    lea               r0,            [r0 + r1 * 4]
-    movq              xm3,           [r0]
-    movhps            xm3,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m3,            m3,          xm2,          1
-
-    pshufb            m1,            m4
-    pshufb            m3,            m4
-    pmaddubsw         m1,            m0
-    pmaddubsw         m3,            m0
-    pmaddwd           m1,            m5
-    pmaddwd           m3,            m5
-    packssdw          m1,            m3
-    psubw             m1,            m6
-
-    lea               r4,            [r3 * 3]
-    vextracti128      xm2,           m1,          1
-
-    movd              [r2],          xm1
-    pextrd            [r2 + r3],     xm1,         1
-    movd              [r2 + r3 * 2], xm2
-    pextrd            [r2 + r4],     xm2,         1
-    lea               r2,            [r2 + r3 * 4]
-    pextrd            [r2],          xm1,         2
-    pextrd            [r2 + r3],     xm1,         3
-    pextrd            [r2 + r3 * 2], xm2,         2
-    pextrd            [r2 + r4],     xm2,         3
-    test              r5d,            r5d
-    jz                .end
-
-    lea               r0,            [r0 + r1 * 4]
-    lea               r2,            [r2 + r3 * 4]
-    movq              xm1,           [r0]                                   ;row 0
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    vinserti128       m1,            m1,          xm2,          1
-    pshufb            m1,            m4
-    pmaddubsw         m1,            m0
-    pmaddwd           m1,            m5
-    packssdw          m1,            m1
-    psubw             m1,            m6
-    vextracti128      xm2,           m1,          1
-
-    movd              [r2],          xm1
-    pextrd            [r2 + r3],     xm1,         1
-    movd              [r2 + r3 * 2], xm2
-.end
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_12x16, 4, 6, 7
-    mov               r4d,          r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m6,           [pw_512]
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          8
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    ; Row 1
-    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-
-    vextracti128      xm4,          m3,       1
-    movq              [r2],         xm3
-    pextrd            [r2+8],       xm3,      2
-    movq              [r2 + r3],    xm4
-    pextrd            [r2 + r3 + 8],xm4,      2
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1 * 2]
-    dec               r4d
-    jnz               .loop
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_24x32, 4,6,7
-    mov              r4d,           r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-    mova              m6,           [pw_512]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          32
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    vbroadcasti128    m4,           [r0 + 16]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + 20]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-
-    vextracti128      xm4,          m3,       1
-    movu              [r2],         xm3
-    movq              [r2 + 16],    xm4
-    add               r2,           r3
-    add               r0,           r1
-    dec               r4d
-    jnz               .loop
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_6x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------;
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_6x8, 4,7,6
-    mov                r4d,            r4m
-    mov                r5d,            r5m
-    add                r3d,            r3d
-
-%ifdef PIC
-    lea                r6,             [tab_ChromaCoeff]
-    vpbroadcastd       m0,             [r6 + r4 * 4]
-%else
-    vpbroadcastd       m0,             [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,             [pw_1]
-    vbroadcasti128     m5,             [pw_2000]
-    mova               m1,             [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    mov               r6d,             8/2
-    dec               r0
-    test              r5d,             r5d
-    jz                .loop
-    sub               r0 ,             r1
-    inc               r6d
-
-.loop
-    ; Row 0
-    vbroadcasti128    m3,              [r0]
-    pshufb            m3,              m1
-    pmaddubsw         m3,              m0
-    pmaddwd           m3,              m2
-
-    ; Row 1
-    vbroadcasti128    m4,              [r0 + r1]
-    pshufb            m4,              m1
-    pmaddubsw         m4,              m0
-    pmaddwd           m4,              m2
-    packssdw          m3,              m4
-    psubw             m3,              m5
-    vpermq            m3,              m3,          11011000b
-    vextracti128      xm4,             m3,          1
-    movq              [r2],            xm3
-    pextrd            [r2 + 8],        xm3,         2
-    movq              [r2 + r3],       xm4
-    pextrd            [r2 + r3 + 8],   xm4,         2
-    lea               r2,              [r2 + r3 * 2]
-    lea               r0,              [r0 + r1 * 2]
-    dec               r6d
-    jnz              .loop
-    test              r5d,             r5d
-    jz               .end
-
-    ;Row 11
-    vbroadcasti128    m3,              [r0]
-    pshufb            m3,              m1
-    pmaddubsw         m3,              m0
-    pmaddwd           m3,              m2
-    packssdw          m3,              m3
-    psubw             m3,              m5
-    vextracti128      xm4,             m3,          1
-    movq              [r2],            xm3
-    movd              [r2+8],          xm4
-.end
-    RET
-
-INIT_YMM avx2
-cglobal interp_8tap_horiz_ps_12x16, 6, 7, 8
-    mov                         r5d,               r5m
-    mov                         r4d,               r4m
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    vpbroadcastq                m0,                [r6 + r4 * 8]
-%else
-    vpbroadcastq                m0,                [tab_LumaCoeff + r4 * 8]
-%endif
-    mova                        m6,                [tab_Lm + 32]
-    mova                        m1,                [tab_Lm]
-    add                         r3d,               r3d
-    vbroadcasti128              m2,                [pw_2000]
-    mov                         r4d,                16
-    vbroadcasti128              m7,                [pw_1]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - pw_2000
-
-    mova                        m5,                [interp8_hps_shuf]
-    sub                         r0,                3
-    test                        r5d,               r5d
-    jz                          .loop
-    lea                         r6,                [r1 * 3]                     ; r6 = (N / 2 - 1) * srcStride
-    sub                         r0,                r6                           ; r0(src)-r6
-    add                         r4d,                7
-.loop
-
-    ; Row 0
-
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m3,        m6
-    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 8]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m1
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m4,                m7
-    packssdw                    m4,                m4
-
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-
-    vpermd                      m3,                m5,               m3
-    psubw                       m3,                m2
-
-    vextracti128                xm4,               m3,               1
-    movu                        [r2],              xm3                          ;row 0
-    movq                        [r2 + 16],         xm4                          ;row 1
-
-    add                         r0,                r1
-    add                         r2,                r3
-    dec                         r4d
-    jnz                         .loop
-    RET
-
-INIT_YMM avx2
-cglobal interp_8tap_horiz_ps_24x32, 4, 7, 8
-    mov                         r5d,               r5m
-    mov                         r4d,               r4m
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    vpbroadcastq                m0,                [r6 + r4 * 8]
-%else
-    vpbroadcastq                m0,                [tab_LumaCoeff + r4 * 8]
-%endif
-    mova                        m6,                [tab_Lm + 32]
-    mova                        m1,                [tab_Lm]
-    mov                         r4d,               32                           ;height
-    add                         r3d,               r3d
-    vbroadcasti128              m2,                [pw_2000]
-    vbroadcasti128              m7,                [pw_1]
-
-    ; register map
-    ; m0      - interpolate coeff
-    ; m1 , m6 - shuffle order table
-    ; m2      - pw_2000
-
-    sub                         r0,                3
-    test                        r5d,               r5d
-    jz                          .label
-    lea                         r6,                [r1 * 3]                     ; r6 = (N / 2 - 1) * srcStride
-    sub                         r0,                r6                           ; r0(src)-r6
-    add                         r4d,               7                            ; blkheight += N - 1  (7 - 1 = 6 ; since the last one row not in loop)
-
-.label
-    lea                         r6,                [interp8_hps_shuf]
-.loop
-    ; Row 0
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
-    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-
-    vbroadcasti128              m4,                [r0 + 8]                     ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m5,                m4,            m6            ;row 1 (col 4 to 7)
-    pshufb                      m4,                m1                           ;row 1 (col 0 to 3)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m7
-    pmaddwd                     m5,                m7
-    packssdw                    m4,                m5
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-    mova                        m5,                [r6]
-    vpermd                      m3,                m5,               m3
-    psubw                       m3,                m2
-    movu                        [r2],              m3                          ;row 0
-
-    vbroadcasti128              m3,                [r0 + 16]
-    pshufb                      m4,                m3,          m6
-    pshufb                      m3,                m1
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-    pmaddwd                     m3,                m7
-    pmaddwd                     m4,                m7
-    packssdw                    m3,                m4
-    mova                        m4,                [r6]
-    vpermd                      m3,                m4,            m3
-    psubw                       m3,                m2
-    movu                        [r2 + 32],         xm3                          ;row 0
-
-    add                         r0,                r1
-    add                         r2,                r3
-    dec                         r4d
-    jnz                         .loop
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_24x32(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_24x32, 4,7,6
-    mov                r4d,            r4m
-    mov                r5d,            r5m
-    add                r3d,            r3d
-%ifdef PIC
-    lea                r6,             [tab_ChromaCoeff]
-    vpbroadcastd       m0,             [r6 + r4 * 4]
-%else
-    vpbroadcastd       m0,             [tab_ChromaCoeff + r4 * 4]
-%endif
-    vbroadcasti128     m2,             [pw_1]
-    vbroadcasti128     m5,             [pw_2000]
-    mova               m1,             [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-    mov                r6d,            32
-    dec                r0
-    test               r5d,            r5d
-    je                 .loop
-    sub                r0 ,            r1
-    add                r6d ,           3
-
-.loop
-    ; Row 0
-    vbroadcasti128     m3,             [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb             m3,             m1
-    pmaddubsw          m3,             m0
-    pmaddwd            m3,             m2
-    vbroadcasti128     m4,             [r0 + 8]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb             m4,             m1
-    pmaddubsw          m4,             m0
-    pmaddwd            m4,             m2
-    packssdw           m3,             m4
-    psubw              m3,             m5
-    vpermq             m3,             m3,          11011000b
-    movu               [r2],           m3
-
-    vbroadcasti128     m3,             [r0 + 16]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb             m3,             m1
-    pmaddubsw          m3,             m0
-    pmaddwd            m3,             m2
-    packssdw           m3,             m3
-    psubw              m3,             m5
-    vpermq             m3,             m3,          11011000b
-    movu               [r2 + 32],      xm3
-
-    add                r2,             r3
-    add                r0,             r1
-    dec                r6d
-    jnz                .loop
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------
-;macro FILTER_H8_W8_16N_AVX2
-;-----------------------------------------------------------------------------------------------------------------------
-%macro  FILTER_H8_W8_16N_AVX2 0
-    vbroadcasti128              m3,                [r0]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m4,                m3,             m6           ; row 0 (col 4 to 7)
-    pshufb                      m3,                m1                           ; shuffled based on the col order tab_Lm row 0 (col 0 to 3)
-    pmaddubsw                   m3,                m0
-    pmaddubsw                   m4,                m0
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4                         ; DWORD [R1D R1C R0D R0C R1B R1A R0B R0A]
-
-    vbroadcasti128              m4,                [r0 + 8]                         ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb                      m5,                m4,            m6            ;row 1 (col 4 to 7)
-    pshufb                      m4,                m1                           ;row 1 (col 0 to 3)
-    pmaddubsw                   m4,                m0
-    pmaddubsw                   m5,                m0
-    pmaddwd                     m4,                m2
-    pmaddwd                     m5,                m2
-    packssdw                    m4,                m5                         ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
-
-    pmaddwd                     m3,                m2
-    pmaddwd                     m4,                m2
-    packssdw                    m3,                m4                         ; all rows and col completed.
-
-    mova                        m5,                [interp8_hps_shuf]
-    vpermd                      m3,                m5,               m3
-    psubw                       m3,                m8
-
-    vextracti128                xm4,               m3,               1
-    mova                        [r4],              xm3
-    mova                        [r4 + 16],         xm4
-    %endmacro
-
-;-----------------------------------------------------------------------------
-; void interp_8tap_hv_pp_16x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int idxX, int idxY)
-;-----------------------------------------------------------------------------
-INIT_YMM avx2
-%if ARCH_X86_64 == 1
-cglobal interp_8tap_hv_pp_16x16, 4, 10, 15, 0-31*32
-%define stk_buf1    rsp
-    mov                         r4d,               r4m
-    mov                         r5d,               r5m
-%ifdef PIC
-    lea                         r6,                [tab_LumaCoeff]
-    vpbroadcastq                m0,                [r6 + r4 * 8]
-%else
-    vpbroadcastq                m0,                [tab_LumaCoeff + r4 * 8]
-%endif
-
-    xor                         r6,                 r6
-    mov                         r4,                 rsp
-    mova                        m6,                [tab_Lm + 32]
-    mova                        m1,                [tab_Lm]
-    mov                         r8,                16                           ;height
-    vbroadcasti128              m8,                [pw_2000]
-    vbroadcasti128              m2,                [pw_1]
-    sub                         r0,                3
-    lea                         r7,                [r1 * 3]                     ; r7 = (N / 2 - 1) * srcStride
-    sub                         r0,                r7                           ; r0(src)-r7
-    add                         r8,                7
-
-.loopH:
-    FILTER_H8_W8_16N_AVX2
-    add                         r0,                r1
-    add                         r4,                32
-    inc                         r6
-    cmp                         r6,                16+7
-    jnz                        .loopH
-
-; vertical phase
-    xor                         r6,                r6
-    xor                         r1,                r1
-.loopV:
-
-;load necessary variables
-    mov                         r4d,               r5d          ;coeff here for vertical is r5m
-    shl                         r4d,               7
-    mov                         r1d,               16
-    add                         r1d,               r1d
-
- ; load intermedia buffer
-    mov                         r0,                stk_buf1
-
-    ; register mapping
-    ; r0 - src
-    ; r5 - coeff
-    ; r6 - loop_i
-
-; load coeff table
-%ifdef PIC
-    lea                          r5,                [pw_LumaCoeffVer]
-    add                          r5,                r4
-%else
-    lea                          r5,                [pw_LumaCoeffVer + r4]
-%endif
-
-    lea                          r4,                [r1*3]
-    mova                         m14,               [pd_526336]
-    lea                          r6,                [r3 * 3]
-    mov                          r9d,               16 / 8
-
-.loopW:
-    PROCESS_LUMA_AVX2_W8_16R sp
-    add                          r2,                 8
-    add                          r0,                 16
-    dec                          r9d
-    jnz                          .loopW
-    RET
-%endif
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_12x32, 4, 6, 7
-    mov               r4d,          r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m6,           [pw_512]
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          16
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]                    ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    ; Row 1
-    vbroadcasti128    m4,           [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + r1 + 4]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-
-    vextracti128      xm4,          m3,       1
-    movq              [r2],         xm3
-    pextrd            [r2+8],       xm3,      2
-    movq              [r2 + r3],    xm4
-    pextrd            [r2 + r3 + 8],xm4,      2
-    lea               r2,           [r2 + r3 * 2]
-    lea               r0,           [r0 + r1 * 2]
-    dec               r4d
-    jnz               .loop
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_24x64, 4,6,7
-    mov              r4d,           r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-    mova              m6,           [pw_512]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          64
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    vbroadcasti128    m4,           [r0 + 16]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + 20]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-
-    vextracti128      xm4,          m3,       1
-    movu              [r2],         xm3
-    movq              [r2 + 16],    xm4
-    add               r2,           r3
-    add               r0,           r1
-    dec               r4d
-    jnz               .loop
-    RET
-
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_2x16, 4, 6, 6
-    mov               r4d,           r4m
-
-%ifdef PIC
-    lea               r5,            [tab_ChromaCoeff]
-    vpbroadcastd      m0,            [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,            [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m4,            [interp4_hpp_shuf]
-    mova              m5,            [pw_1]
-    dec               r0
-    lea               r4,            [r1 * 3]
-    movq              xm1,           [r0]
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m1,            m1,          xm2,          1
-    lea               r0,            [r0 + r1 * 4]
-    movq              xm3,           [r0]
-    movhps            xm3,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m3,            m3,          xm2,          1
-
-    pshufb            m1,            m4
-    pshufb            m3,            m4
-    pmaddubsw         m1,            m0
-    pmaddubsw         m3,            m0
-    pmaddwd           m1,            m5
-    pmaddwd           m3,            m5
-    packssdw          m1,            m3
-    pmulhrsw          m1,            [pw_512]
-    vextracti128      xm2,           m1,          1
-    packuswb          xm1,           xm2
-
-    lea               r4,            [r3 * 3]
-    pextrw            [r2],          xm1,         0
-    pextrw            [r2 + r3],     xm1,         1
-    pextrw            [r2 + r3 * 2], xm1,         4
-    pextrw            [r2 + r4],     xm1,         5
-    lea               r2,            [r2 + r3 * 4]
-    pextrw            [r2],          xm1,         2
-    pextrw            [r2 + r3],     xm1,         3
-    pextrw            [r2 + r3 * 2], xm1,         6
-    pextrw            [r2 + r4],     xm1,         7
-    lea               r2,            [r2 + r3 * 4]
-    lea               r0,            [r0 + r1 * 4]
-
-    lea               r4,            [r1 * 3]
-    movq              xm1,           [r0]
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m1,            m1,          xm2,          1
-    lea               r0,            [r0 + r1 * 4]
-    movq              xm3,           [r0]
-    movhps            xm3,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m3,            m3,          xm2,          1
-
-    pshufb            m1,            m4
-    pshufb            m3,            m4
-    pmaddubsw         m1,            m0
-    pmaddubsw         m3,            m0
-    pmaddwd           m1,            m5
-    pmaddwd           m3,            m5
-    packssdw          m1,            m3
-    pmulhrsw          m1,            [pw_512]
-    vextracti128      xm2,           m1,          1
-    packuswb          xm1,           xm2
-
-    lea               r4,            [r3 * 3]
-    pextrw            [r2],          xm1,         0
-    pextrw            [r2 + r3],     xm1,         1
-    pextrw            [r2 + r3 * 2], xm1,         4
-    pextrw            [r2 + r4],     xm1,         5
-    lea               r2,            [r2 + r3 * 4]
-    pextrw            [r2],          xm1,         2
-    pextrw            [r2 + r3],     xm1,         3
-    pextrw            [r2 + r3 * 2], xm1,         6
-    pextrw            [r2 + r4],     xm1,         7
-    RET
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_64xN(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-%macro IPFILTER_CHROMA_PP_64xN_AVX2 1
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_64x%1, 4,6,7
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea               r5,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,           [interp4_horiz_shuf1]
-    vpbroadcastd      m2,           [pw_1]
-    mova              m6,           [pw_512]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,          %1
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 4]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    vbroadcasti128    m4,           [r0 + 16]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + 20]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-    movu              [r2],         m3
-
-    vbroadcasti128    m3,           [r0 + 32]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 36]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    packssdw          m3,           m4
-    pmulhrsw          m3,           m6
-
-    vbroadcasti128    m4,           [r0 + 48]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-    vbroadcasti128    m5,           [r0 + 52]
-    pshufb            m5,           m1
-    pmaddubsw         m5,           m0
-    pmaddwd           m5,           m2
-    packssdw          m4,           m5
-    pmulhrsw          m4,           m6
-    packuswb          m3,           m4
-    vpermq            m3,           m3,      11011000b
-    movu              [r2 + 32],         m3
-
-    add               r2,           r3
-    add               r0,           r1
-    dec               r4d
-    jnz               .loop
-    RET
-%endmacro
-
-    IPFILTER_CHROMA_PP_64xN_AVX2  64
-    IPFILTER_CHROMA_PP_64xN_AVX2  32
-    IPFILTER_CHROMA_PP_64xN_AVX2  48
-    IPFILTER_CHROMA_PP_64xN_AVX2  16
-
-;-------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_pp_48x64(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx
-;-------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_48x64, 4,6,7
-    mov             r4d, r4m
-
-%ifdef PIC
-    lea               r5,            [tab_ChromaCoeff]
-    vpbroadcastd      m0,            [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,            [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,            [interp4_horiz_shuf1]
-    vpbroadcastd      m2,            [pw_1]
-    mova              m6,            [pw_512]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-    mov               r4d,           64
-
-.loop:
-    ; Row 0
-    vbroadcasti128    m3,            [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,            m1
-    pmaddubsw         m3,            m0
-    pmaddwd           m3,            m2
-    vbroadcasti128    m4,            [r0 + 4]
-    pshufb            m4,            m1
-    pmaddubsw         m4,            m0
-    pmaddwd           m4,            m2
-    packssdw          m3,            m4
-    pmulhrsw          m3,            m6
-
-    vbroadcasti128    m4,            [r0 + 16]
-    pshufb            m4,            m1
-    pmaddubsw         m4,            m0
-    pmaddwd           m4,            m2
-    vbroadcasti128    m5,            [r0 + 20]
-    pshufb            m5,            m1
-    pmaddubsw         m5,            m0
-    pmaddwd           m5,            m2
-    packssdw          m4,            m5
-    pmulhrsw          m4,            m6
-
-    packuswb          m3,            m4
-    vpermq            m3,            m3,      q3120
-
-    movu              [r2],          m3
-
-    vbroadcasti128    m3,            [r0 + mmsize]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,            m1
-    pmaddubsw         m3,            m0
-    pmaddwd           m3,            m2
-    vbroadcasti128    m4,            [r0 + mmsize + 4]
-    pshufb            m4,            m1
-    pmaddubsw         m4,            m0
-    pmaddwd           m4,            m2
-    packssdw          m3,            m4
-    pmulhrsw          m3,            m6
-
-    vbroadcasti128    m4,            [r0 + mmsize + 16]
-    pshufb            m4,            m1
-    pmaddubsw         m4,            m0
-    pmaddwd           m4,            m2
-    vbroadcasti128    m5,            [r0 + mmsize + 20]
-    pshufb            m5,            m1
-    pmaddubsw         m5,            m0
-    pmaddwd           m5,            m2
-    packssdw          m4,            m5
-    pmulhrsw          m4,            m6
-
-    packuswb          m3,            m4
-    vpermq            m3,            m3,      q3120
-    movu              [r2 + mmsize], xm3
-
-    add               r2,            r3
-    add               r0,            r1
-    dec               r4d
-    jnz               .loop
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_48x64(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------;
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_48x64, 4,7,6
-    mov             r4d, r4m
-    mov             r5d, r5m
-    add             r3d, r3d
-
-%ifdef PIC
-    lea               r6,           [tab_ChromaCoeff]
-    vpbroadcastd      m0,           [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    vbroadcasti128     m2,          [pw_1]
-    vbroadcasti128     m5,          [pw_2000]
-    mova               m1,          [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-    mov               r6d,          64
-    dec               r0
-    test              r5d,          r5d
-    je                .loop
-    sub               r0 ,          r1
-    add               r6d ,         3
-
-.loop
-    ; Row 0
-    vbroadcasti128    m3,           [r0]                           ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 8]                       ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-    vpermq            m3,           m3,          q3120
-    movu              [r2],         m3
-
-    vbroadcasti128    m3,           [r0 + 16]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 24]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-    vpermq            m3,           m3,          q3120
-    movu              [r2 + 32],    m3
-
-    vbroadcasti128    m3,           [r0 + 32]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,           m1
-    pmaddubsw         m3,           m0
-    pmaddwd           m3,           m2
-    vbroadcasti128    m4,           [r0 + 40]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,           m1
-    pmaddubsw         m4,           m0
-    pmaddwd           m4,           m2
-
-    packssdw          m3,           m4
-    psubw             m3,           m5
-    vpermq            m3,           m3,          q3120
-    movu              [r2 + 64],    m3
-
-    add               r2,          r3
-    add               r0,          r1
-    dec               r6d
-    jnz               .loop
-    RET
-
-;-----------------------------------------------------------------------------------------------------------------------------
-; void interp_4tap_horiz_ps_24x64(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
-;-----------------------------------------------------------------------------------------------------------------------------
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_24x64, 4,7,6
-    mov                r4d,            r4m
-    mov                r5d,            r5m
-    add                r3d,            r3d
-%ifdef PIC
-    lea                r6,             [tab_ChromaCoeff]
-    vpbroadcastd       m0,             [r6 + r4 * 4]
-%else
-    vpbroadcastd       m0,             [tab_ChromaCoeff + r4 * 4]
-%endif
-    vbroadcasti128     m2,             [pw_1]
-    vbroadcasti128     m5,             [pw_2000]
-    mova               m1,             [tab_Tm]
-
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-    mov                r6d,            64
-    dec                r0
-    test               r5d,            r5d
-    je                 .loop
-    sub                r0 ,            r1
-    add                r6d ,           3
-
-.loop
-    ; Row 0
-    vbroadcasti128     m3,             [r0]                          ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb             m3,             m1
-    pmaddubsw          m3,             m0
-    pmaddwd            m3,             m2
-    vbroadcasti128     m4,             [r0 + 8]                      ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb             m4,             m1
-    pmaddubsw          m4,             m0
-    pmaddwd            m4,             m2
-    packssdw           m3,             m4
-    psubw              m3,             m5
-    vpermq             m3,             m3,          q3120
-    movu               [r2],           m3
-
-    vbroadcasti128     m3,             [r0 + 16]                     ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb             m3,             m1
-    pmaddubsw          m3,             m0
-    pmaddwd            m3,             m2
-    packssdw           m3,             m3
-    psubw              m3,             m5
-    vpermq             m3,             m3,          q3120
-    movu               [r2 + 32],      xm3
-
-    add                r2,             r3
-    add                r0,             r1
-    dec                r6d
-    jnz                .loop
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_ps_2x16, 4, 7, 7
-    mov               r4d,           r4m
-    mov               r5d,           r5m
-    add               r3d,           r3d
-
-%ifdef PIC
-    lea               r6,            [tab_ChromaCoeff]
-    vpbroadcastd      m0,            [r6 + r4 * 4]
-%else
-    vpbroadcastd      m0,            [tab_ChromaCoeff + r4 * 4]
-%endif
-    vbroadcasti128    m6,            [pw_2000]
-    test              r5d,            r5d
-    jz                .label
-    sub               r0,             r1
-
-.label
-    mova              m4,            [interp4_hps_shuf]
-    mova              m5,            [pw_1]
-    dec               r0
-    lea               r4,            [r1 * 3]
-    movq              xm1,           [r0]                                   ;row 0
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m1,            m1,           xm2,          1
-    lea               r0,            [r0 + r1 * 4]
-    movq              xm3,           [r0]
-    movhps            xm3,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m3,            m3,           xm2,          1
-
-    pshufb            m1,            m4
-    pshufb            m3,            m4
-    pmaddubsw         m1,            m0
-    pmaddubsw         m3,            m0
-    pmaddwd           m1,            m5
-    pmaddwd           m3,            m5
-    packssdw          m1,            m3
-    psubw             m1,            m6
-
-    lea               r4,            [r3 * 3]
-    vextracti128      xm2,           m1,           1
-
-    movd              [r2],          xm1
-    pextrd            [r2 + r3],     xm1,          1
-    movd              [r2 + r3 * 2], xm2
-    pextrd            [r2 + r4],     xm2,          1
-    lea               r2,            [r2 + r3 * 4]
-    pextrd            [r2],          xm1,          2
-    pextrd            [r2 + r3],     xm1,          3
-    pextrd            [r2 + r3 * 2], xm2,          2
-    pextrd            [r2 + r4],     xm2,          3
-
-    lea               r0,            [r0 + r1 * 4]
-    lea               r2,            [r2 + r3 * 4]
-    lea               r4,            [r1 * 3]
-    movq              xm1,           [r0]
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m1,            m1,          xm2,           1
-    lea               r0,            [r0 + r1 * 4]
-    movq              xm3,           [r0]
-    movhps            xm3,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    movhps            xm2,           [r0 + r4]
-    vinserti128       m3,            m3,          xm2,           1
-
-    pshufb            m1,            m4
-    pshufb            m3,            m4
-    pmaddubsw         m1,            m0
-    pmaddubsw         m3,            m0
-    pmaddwd           m1,            m5
-    pmaddwd           m3,            m5
-    packssdw          m1,            m3
-    psubw             m1,            m6
-
-    lea               r4,            [r3 * 3]
-    vextracti128      xm2,           m1,           1
-
-    movd              [r2],          xm1
-    pextrd            [r2 + r3],     xm1,          1
-    movd              [r2 + r3 * 2], xm2
-    pextrd            [r2 + r4],     xm2,          1
-    lea               r2,            [r2 + r3 * 4]
-    pextrd            [r2],          xm1,          2
-    pextrd            [r2 + r3],     xm1,          3
-    pextrd            [r2 + r3 * 2], xm2,          2
-    pextrd            [r2 + r4],     xm2,          3
-
-    test              r5d,            r5d
-    jz                .end
-
-    lea               r0,            [r0 + r1 * 4]
-    lea               r2,            [r2 + r3 * 4]
-    movq              xm1,           [r0]
-    movhps            xm1,           [r0 + r1]
-    movq              xm2,           [r0 + r1 * 2]
-    vinserti128       m1,            m1,          xm2,           1
-    pshufb            m1,            m4
-    pmaddubsw         m1,            m0
-    pmaddwd           m1,            m5
-    packssdw          m1,            m1
-    psubw             m1,            m6
-    vextracti128      xm2,           m1,           1
-
-    movd              [r2],          xm1
-    pextrd            [r2 + r3],     xm1,          1
-    movd              [r2 + r3 * 2], xm2
-.end
-    RET
-
-INIT_YMM avx2
-cglobal interp_4tap_horiz_pp_6x16, 4, 6, 7
-    mov               r4d,               r4m
-
-%ifdef PIC
-    lea               r5,                [tab_ChromaCoeff]
-    vpbroadcastd      m0,                [r5 + r4 * 4]
-%else
-    vpbroadcastd      m0,                [tab_ChromaCoeff + r4 * 4]
-%endif
-
-    mova              m1,                [tab_Tm]
-    mova              m2,                [pw_1]
-    mova              m6,                [pw_512]
-    lea               r4,                [r1 * 3]
-    lea               r5,                [r3 * 3]
-    ; register map
-    ; m0 - interpolate coeff
-    ; m1 - shuffle order table
-    ; m2 - constant word 1
-
-    dec               r0
-%rep 4
-    ; Row 0
-    vbroadcasti128    m3,                [r0]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m3,                m1
-    pmaddubsw         m3,                m0
-    pmaddwd           m3,                m2
-
-    ; Row 1
-    vbroadcasti128    m4,                [r0 + r1]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,                m1
-    pmaddubsw         m4,                m0
-    pmaddwd           m4,                m2
-    packssdw          m3,                m4
-    pmulhrsw          m3,                m6
-
-    ; Row 2
-    vbroadcasti128    m4,                [r0 + r1 * 2]               ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m4,                m1
-    pmaddubsw         m4,                m0
-    pmaddwd           m4,                m2
-
-    ; Row 3
-    vbroadcasti128    m5,                [r0 + r4]                   ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
-    pshufb            m5,                m1
-    pmaddubsw         m5,                m0
-    pmaddwd           m5,                m2
-    packssdw          m4,                m5
-    pmulhrsw          m4,                m6
-
-    packuswb          m3,                m4
-    vextracti128      xm4,               m3,          1
-    movd              [r2],              xm3
-    pextrw            [r2 + 4],          xm4,         0
-    pextrd            [r2 + r3],         xm3,         1
-    pextrw            [r2 + r3 + 4],     xm4,         2
-    pextrd            [r2 + r3 * 2],     xm3,         2
-    pextrw            [r2 + r3 * 2 + 4], xm4,         4
-    pextrd            [r2 + r5],         xm3,         3
-    pextrw            [r2 + r5 + 4],     xm4,         6
-    lea               r2,                [r2 + r3 * 4]
-    lea               r0,                [r0 + r1 * 4]
-%endrep
-    RET

x265_2.6.tar.gz/source/common/x86/loopfilter.asm -> x265_2.7.tar.gz/source/common/x86/loopfilter.asm Changed

@@ -374,7 +374,7 @@
     pxor        m0, m0                      ; m0 = 0
     mova        m6, [pb_2]                  ; m6 = [2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2]
     shr         r4d, 4
-.loop
+.loop:
     movu        m7, [r0]
     movu        m5, [r0 + 16]
     movu        m3, [r0 + r3]
@@ -430,7 +430,7 @@
     mova        m6,    [pb_2]                  ; m6 = [2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2]
     mova        m7,    [pb_128]
     shr         r4d,   4
-.loop
+.loop:
     movu        m1,    [r0]                    ; m1 = pRec[x]
     movu        m2,    [r0 + r3]               ; m2 = pRec[x + iStride]
 
@@ -478,7 +478,7 @@
     mova        m4, [pb_2]
     shr         r4d, 4
     mova        m0, [pw_pixel_max]
-.loop
+.loop:
     movu        m5, [r0]
     movu        m3, [r0 + r3]
 
@@ -523,7 +523,7 @@
     mova          xm6,    [pb_2]                  ; xm6 = [2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2]
     mova          xm7,    [pb_128]
     shr           r4d,    4
-.loop
+.loop:
     movu          xm1,    [r0]                    ; xm1 = pRec[x]
     movu          xm2,    [r0 + r3]               ; xm2 = pRec[x + iStride]
 
@@ -572,7 +572,7 @@
     mov         r5d, r4d
     shr         r4d, 4
     mov         r6, r0
-.loop
+.loop:
     movu        m7, [r0]
     movu        m5, [r0 + 16]
     movu        m3, [r0 + r3]
@@ -674,7 +674,7 @@
     pxor        m0,         m0                      ; m0 = 0
     mova        m7,         [pb_128]
     shr         r4d,        4
-.loop
+.loop:
     movu        m1,         [r0]                    ; m1 = pRec[x]
     movu        m2,         [r0 + r3]               ; m2 = pRec[x + iStride]
 
@@ -748,7 +748,7 @@
     mova            m4, [pw_pixel_max]
     vbroadcasti128  m6, [r2]                ; m6 = m_iOffsetEo
     shr             r4d, 4
-.loop
+.loop:
     movu            m7, [r0]
     movu            m5, [r0 + r3]
     movu            m1, [r0 + r3 * 2]
@@ -804,7 +804,7 @@
     vbroadcasti128  m5,         [pb_128]
     vbroadcasti128  m6,         [r2]                         ; m6 = m_iOffsetEo
     shr             r4d,        4
-.loop
+.loop:
     movu            xm1,        [r0]                         ; m1 = pRec[x]
     movu            xm2,        [r0 + r3]                    ; m2 = pRec[x + iStride]
     vinserti128     m1,         m1,       xm2,            1
@@ -859,7 +859,7 @@
     movh        m6, [r0 + r4 * 2]
     movhps      m6, [r1 + r4]
 
-.loop
+.loop:
     movu        m7, [r0]
     movu        m5, [r0 + 16]
     movu        m3, [r0 + r5 + 2]
@@ -918,7 +918,7 @@
     movh        m5,    [r0 + r4]
     movhps      m5,    [r1 + r4]
 
-.loop
+.loop:
     movu        m1,    [r0]                    ; m1 = rec[x]
     movu        m2,    [r0 + r5 + 1]           ; m2 = rec[x + stride + 1]
     pxor        m3,    m1,    m7
@@ -970,7 +970,7 @@
     movhps          xm4, [r1 + r4]
     vbroadcasti128  m5, [r3]
     mova            m6, [pw_pixel_max]
-.loop
+.loop:
     movu            m1, [r0]
     movu            m3, [r0 + r5 + 2]
 
@@ -1061,7 +1061,7 @@
     movhps          xm4, [r1 + r4]
     vbroadcasti128  m5, [r3]
 
-.loop
+.loop:
     movu            m1, [r0]
     movu            m7, [r0 + 32]
     movu            m3, [r0 + r5 + 2]
@@ -1567,11 +1567,11 @@
     movu        m4, [r1 + 16]       ; offset[16-31]
     pxor        m7, m7
 
-.loopH
+.loopH:
     mov         r5d, r2d
     xor         r6,  r6
 
-.loopW
+.loopW:
     movu        m2, [r0 + r6]
     movu        m5, [r0 + r6 + 16]
     psrlw       m0, m2, (BIT_DEPTH - 5)
@@ -1617,11 +1617,11 @@
     movu        m3, [r1 + 0]      ; offset[0-15]
     movu        m4, [r1 + 16]     ; offset[16-31]
     pxor        m7, m7            ; m7 =[0]
-.loopH
+.loopH:
     mov         r5d, r2d
     xor         r6,  r6
 
-.loopW
+.loopW:
     movu        m2, [r0 + r6]     ; m0 = [rec]
     psrlw       m1, m2, 3
     pand        m1, [pb_31]       ; m1 = [index]
@@ -1670,9 +1670,9 @@
     mov             r6d, r3d
     shr             r3d, 1
 
-.loopH
+.loopH:
     mov             r5d, r2d
-.loopW
+.loopW:
     movu            m2, [r0]
     movu            m5, [r0 + r4]
     psrlw           m0, m2, (BIT_DEPTH - 5)
@@ -1751,9 +1751,9 @@
     shr             r2d,        4
     mov             r1d,        r3d
     shr             r3d,        1
-.loopH
+.loopH:
     mov             r5d,        r2d
-.loopW
+.loopW:
     movu            xm2,        [r0]                ; m2 = [rec]
     vinserti128     m2,         m2,  [r0 + r4],  1
     psrlw           m1,         m2,  3
@@ -1789,7 +1789,7 @@
     test            r1b,        1
     jz              .end
     mov             r5d,        r2d
-.loopW1
+.loopW1:
     movu            xm2,        [r0]                ; m2 = [rec]
     psrlw           xm1,        xm2, 3
     pand            xm1,        xm7                 ; m1 = [index]
@@ -1811,7 +1811,7 @@
     add             r0,         16
     dec             r5d
     jnz             .loopW1
-.end
+.end:
     RET
 %endif
 
@@ -1827,7 +1827,7 @@
     add             r3d, 1
     mov             r5, r0
     movu            m4, [r0 + r4]
-.loop
+.loop:
     movu            m1, [r1]        ; m2 = pRec[x]
     movu            m2, [r2]        ; m3 = pTmpU[x]
 
@@ -1921,7 +1921,7 @@
     mov             r5, r0
     movu            m4, [r0 + r4]
 
-.loop
+.loop:
     movu            m1, [r1]        ; m2 = pRec[x]
     movu            m2, [r2]        ; m3 = pTmpU[x]

x265_2.6.tar.gz/source/common/x86/mc-a.asm -> x265_2.7.tar.gz/source/common/x86/mc-a.asm Changed

@@ -4115,7 +4115,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 4
-.loop
+.loop:
     pixel_avg_W8
     dec     r9d
     jnz     .loop
@@ -4129,7 +4129,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 8
-.loop
+.loop:
     pixel_avg_W8
     dec     r9d
     jnz     .loop
@@ -4697,7 +4697,7 @@
     lea     r8, [r1 * 3]
     mov     r9d, 4
 
-.loop
+.loop:
     movu    m0, [r2]
     movu    m1, [r4]
     pavgw   m0, m1
@@ -4834,7 +4834,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 4
-.loop
+.loop:
     pixel_avg_H16
     dec r9d
     jnz .loop
@@ -4848,7 +4848,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 4
-.loop
+.loop:
     pixel_avg_H16
     pixel_avg_H16
     dec r9d
@@ -4863,7 +4863,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 4
-.loop
+.loop:
     pixel_avg_H16
     pixel_avg_H16
     pixel_avg_H16
@@ -4887,7 +4887,7 @@
     lea     r8, [r1 * 3]
     mov     r9d, 8
 
-.loop
+.loop:
     movu    m0, [r2]
     movu    m1, [r4]
     pavgw   m0, m1
@@ -4987,7 +4987,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 2
-.loop
+.loop:
     pixel_avg_W32
     dec     r9d
     jnz     .loop
@@ -5001,7 +5001,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 4
-.loop
+.loop:
     pixel_avg_W32
     dec     r9d
     jnz     .loop
@@ -5015,7 +5015,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 6
-.loop
+.loop:
     pixel_avg_W32
     dec     r9d
     jnz     .loop
@@ -5029,7 +5029,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 8
-.loop
+.loop:
     pixel_avg_W32
     dec     r9d
     jnz     .loop
@@ -5043,7 +5043,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 16
-.loop
+.loop:
     pixel_avg_W32
     dec     r9d
     jnz     .loop
@@ -5141,7 +5141,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 4
-.loop
+.loop:
     pixel_avg_W64
     dec     r9d
     jnz     .loop
@@ -5155,7 +5155,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 8
-.loop
+.loop:
     pixel_avg_W64
     dec     r9d
     jnz     .loop
@@ -5169,7 +5169,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 12
-.loop
+.loop:
     pixel_avg_W64
     dec     r9d
     jnz     .loop
@@ -5183,7 +5183,7 @@
     lea     r7, [r5 * 3]
     lea     r8, [r1 * 3]
     mov     r9d, 16
-.loop
+.loop:
     pixel_avg_W64
     dec     r9d
     jnz     .loop
@@ -5204,7 +5204,7 @@
     lea     r8, [r1 * 3]
     mov     r9d, 16
 
-.loop
+.loop:
     movu    m0, [r2]
     movu    m1, [r4]
     pavgw   m0, m1

x265_2.6.tar.gz/source/common/x86/pixel-util8.asm -> x265_2.7.tar.gz/source/common/x86/pixel-util8.asm Changed

x265_2.6.tar.gz/source/common/x86/sad-a.asm -> x265_2.7.tar.gz/source/common/x86/sad-a.asm Changed

@@ -5631,7 +5631,7 @@
     xorps           m5, m5
     mov             r4d, 4
 
-.loop
+.loop:
     movu           m1, [r0]               ; row 0 of pix0
     movu           m2, [r2]               ; row 0 of pix1
     movu           m3, [r0 + r1]          ; row 1 of pix0
@@ -5676,7 +5676,7 @@
     mov             r4d, 6
     lea             r5, [r1 * 3]
     lea             r6, [r3 * 3]
-.loop
+.loop:
     movu           m1, [r0]               ; row 0 of pix0
     movu           m2, [r2]               ; row 0 of pix1
     movu           m3, [r0 + r1]          ; row 1 of pix0
@@ -5718,7 +5718,7 @@
     lea             r5, [r1 * 3]
     lea             r6, [r3 * 3]
 
-.loop
+.loop:
     movu           m1, [r0]               ; row 0 of pix0
     movu           m2, [r2]               ; row 0 of pix1
     movu           m3, [r0 + r1]          ; row 1 of pix0
@@ -5759,7 +5759,7 @@
     lea             r5, [r1 * 3]
     lea             r6, [r3 * 3]
 
-.loop
+.loop:
     movu           m1, [r0]               ; row 0 of pix0
     movu           m2, [r2]               ; row 0 of pix1
     movu           m3, [r0 + r1]          ; row 1 of pix0
@@ -5822,7 +5822,7 @@
     mov             r4d, 64/4
     lea             r5, [r1 * 3]
     lea             r6, [r3 * 3]
-.loop
+.loop:
     movu           m1, [r0]               ; row 0 of pix0
     movu           m2, [r2]               ; row 0 of pix1
     movu           m3, [r0 + r1]          ; row 1 of pix0
@@ -5873,7 +5873,7 @@
     xorps           m0, m0
     xorps           m5, m5
     mov             r4d, 4
-.loop
+.loop:
     movu           m1, [r0]               ; first 32 of row 0 of pix0
     movu           m2, [r2]               ; first 32 of row 0 of pix1
     movu           m3, [r0 + 32]          ; second 32 of row 0 of pix0
@@ -5936,7 +5936,7 @@
     xorps           m0, m0
     xorps           m5, m5
     mov             r4d, 16
-.loop
+.loop:
     movu           m1, [r0]               ; first 32 of row 0 of pix0
     movu           m2, [r2]               ; first 32 of row 0 of pix1
     movu           m3, [r0 + 32]          ; second 32 of row 0 of pix0
@@ -5978,7 +5978,7 @@
     mov             r4d, 12
     lea             r5, [r1 * 3]
     lea             r6, [r3 * 3]
-.loop
+.loop:
     movu           m1, [r0]               ; first 32 of row 0 of pix0
     movu           m2, [r2]               ; first 32 of row 0 of pix1
     movu           m3, [r0 + 32]          ; second 32 of row 0 of pix0
@@ -6040,7 +6040,7 @@
     mov             r4d, 8
     lea             r5, [r1 * 3]
     lea             r6, [r3 * 3]
-.loop
+.loop:
     movu           m1, [r0]               ; first 32 of row 0 of pix0
     movu           m2, [r2]               ; first 32 of row 0 of pix1
     movu           m3, [r0 + 32]          ; second 32 of row 0 of pix0

x265_2.6.tar.gz/source/common/x86/seaintegral.asm -> x265_2.7.tar.gz/source/common/x86/seaintegral.asm Changed

@@ -36,7 +36,7 @@
     mov r2, r1
     shl r2, 4
 
-.loop
+.loop:
     movu    m0, [r0]
     movu    m1, [r0 + r2]
     psubd   m1, m0
@@ -54,7 +54,7 @@
     mov r2, r1
     shl r2, 5
 
-.loop
+.loop:
     movu    m0, [r0]
     movu    m1, [r0 + r2]
     psubd   m1, m0
@@ -75,7 +75,7 @@
     shl r3, 4
     add r2, r3
 
-.loop
+.loop:
     movu    m0, [r0]
     movu    m1, [r0 + r2]
     psubd   m1, m0
@@ -93,7 +93,7 @@
     mov r2, r1
     shl r2, 6
 
-.loop
+.loop:
     movu    m0, [r0]
     movu    m1, [r0 + r2]
     psubd   m1, m0
@@ -114,7 +114,7 @@
     shl r3, 5
     add r2, r3
 
-.loop
+.loop:
     movu    m0, [r0]
     movu    m1, [r0 + r2]
     psubd   m1, m0
@@ -132,7 +132,7 @@
     mov r2, r1
     shl r2, 7
 
-.loop
+.loop:
     movu    m0, [r0]
     movu    m1, [r0 + r2]
     psubd   m1, m0
@@ -264,7 +264,7 @@
     movu           [r0 + r3], xm0
     jmp            .end
 
-.end
+.end:
     RET
 %endif
 
@@ -379,7 +379,7 @@
     movu           [r0 + r3], m0
     jmp             .end
 
-.end
+.end:
     RET
 %endif
 
@@ -577,7 +577,7 @@
     movu           [r0 + r3], xm0
     jmp            .end
 
-.end
+.end:
     RET
 %endif
 
@@ -740,7 +740,7 @@
     movu           [r0 + r3], m0
     jmp             .end
 
-.end
+.end:
     RET
 %endif
 
@@ -883,7 +883,7 @@
     movu           [r0 + r3], m0
     jmp             .end
 
-.end
+.end:
     RET
 
 %macro INTEGRAL_THIRTYTWO_HORIZONTAL_16 0
@@ -1058,5 +1058,5 @@
     movu           [r0 + r3], m0
     jmp             .end
 
-.end
+.end:
     RET

x265_2.7.tar.gz/source/common/x86/v4-ipfilter16.asm Added

@@ -0,0 +1,3529 @@
+;*****************************************************************************
+;* Copyright (C) 2013-2017 MulticoreWare, Inc
+;*
+;* Authors: Nabajit Deka <nabajit@multicorewareinc.com>
+;*          Murugan Vairavel <murugan@multicorewareinc.com>
+;*          Min Chen <chenm003@163.com>
+;*
+;* This program is free software; you can redistribute it and/or modify
+;* it under the terms of the GNU General Public License as published by
+;* the Free Software Foundation; either version 2 of the License, or
+;* (at your option) any later version.
+;*
+;* This program is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+;* GNU General Public License for more details.
+;*
+;* You should have received a copy of the GNU General Public License
+;* along with this program; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
+;*
+;* This program is also available under a commercial proprietary license.
+;* For more information, contact us at license @ x265.com.
+;*****************************************************************************/
+
+%include "x86inc.asm"
+%include "x86util.asm"
+
+
+%define INTERP_OFFSET_PP        pd_32
+%define INTERP_SHIFT_PP         6
+
+%if BIT_DEPTH == 10
+    %define INTERP_SHIFT_PS         2
+    %define INTERP_OFFSET_PS        pd_n32768
+    %define INTERP_SHIFT_SP         10
+    %define INTERP_OFFSET_SP        v4_pd_524800
+%elif BIT_DEPTH == 12
+    %define INTERP_SHIFT_PS         4
+    %define INTERP_OFFSET_PS        pd_n131072
+    %define INTERP_SHIFT_SP         8
+    %define INTERP_OFFSET_SP        pd_524416
+%else
+    %error Unsupport bit depth!
+%endif
+
+
+SECTION_RODATA 32
+
+v4_pd_524800:        times 8 dd 524800
+tab_c_n8192:      times 8 dw -8192
+
+const tab_ChromaCoeffV,  times 8 dw 0, 64
+                         times 8 dw 0, 0
+
+                         times 8 dw -2, 58
+                         times 8 dw 10, -2
+
+                         times 8 dw -4, 54
+                         times 8 dw 16, -2
+
+                         times 8 dw -6, 46
+                         times 8 dw 28, -4
+
+                         times 8 dw -4, 36
+                         times 8 dw 36, -4
+
+                         times 8 dw -4, 28
+                         times 8 dw 46, -6
+
+                         times 8 dw -2, 16
+                         times 8 dw 54, -4
+
+                         times 8 dw -2, 10
+                         times 8 dw 58, -2
+                        
+tab_ChromaCoeffVer: times 8 dw 0, 64
+                    times 8 dw 0, 0
+
+                    times 8 dw -2, 58
+                    times 8 dw 10, -2
+
+                    times 8 dw -4, 54
+                    times 8 dw 16, -2
+
+                    times 8 dw -6, 46
+                    times 8 dw 28, -4
+
+                    times 8 dw -4, 36
+                    times 8 dw 36, -4
+
+                    times 8 dw -4, 28
+                    times 8 dw 46, -6
+
+                    times 8 dw -2, 16
+                    times 8 dw 54, -4
+
+                    times 8 dw -2, 10
+                    times 8 dw 58, -2
+                        
+SECTION .text
+cextern pd_8
+cextern pd_32
+cextern pw_pixel_max
+cextern pd_524416
+cextern pd_n32768
+cextern pd_n131072
+cextern pw_2000
+cextern idct8_shuf2
+
+%macro PROCESS_CHROMA_SP_W4_4R 0
+    movq       m0, [r0]
+    movq       m1, [r0 + r1]
+    punpcklwd  m0, m1                          ;m0=[0 1]
+    pmaddwd    m0, [r6 + 0 *32]                ;m0=[0+1]         Row1
+
+    lea        r0, [r0 + 2 * r1]
+    movq       m4, [r0]
+    punpcklwd  m1, m4                          ;m1=[1 2]
+    pmaddwd    m1, [r6 + 0 *32]                ;m1=[1+2]         Row2
+
+    movq       m5, [r0 + r1]
+    punpcklwd  m4, m5                          ;m4=[2 3]
+    pmaddwd    m2, m4, [r6 + 0 *32]            ;m2=[2+3]         Row3
+    pmaddwd    m4, [r6 + 1 * 32]
+    paddd      m0, m4                          ;m0=[0+1+2+3]     Row1 done
+
+    lea        r0, [r0 + 2 * r1]
+    movq       m4, [r0]
+    punpcklwd  m5, m4                          ;m5=[3 4]
+    pmaddwd    m3, m5, [r6 + 0 *32]            ;m3=[3+4]         Row4
+    pmaddwd    m5, [r6 + 1 * 32]
+    paddd      m1, m5                          ;m1 = [1+2+3+4]   Row2
+
+    movq       m5, [r0 + r1]
+    punpcklwd  m4, m5                          ;m4=[4 5]
+    pmaddwd    m4, [r6 + 1 * 32]
+    paddd      m2, m4                          ;m2=[2+3+4+5]     Row3
+
+    movq       m4, [r0 + 2 * r1]
+    punpcklwd  m5, m4                          ;m5=[5 6]
+    pmaddwd    m5, [r6 + 1 * 32]
+    paddd      m3, m5                          ;m3=[3+4+5+6]     Row4
+%endmacro
+
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_%3_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SS 4
+INIT_XMM sse2
+cglobal interp_4tap_vert_%3_%1x%2, 5, 7, %4 ,0-gprsize
+
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r6, [r5 + r4]
+%else
+    lea       r6, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       dword [rsp], %2/4
+
+%ifnidn %3, ss
+    %ifnidn %3, ps
+        mova      m7, [pw_pixel_max]
+        %ifidn %3, pp
+            mova      m6, [INTERP_OFFSET_PP]
+        %else
+            mova      m6, [INTERP_OFFSET_SP]
+        %endif
+    %else
+        mova      m6, [INTERP_OFFSET_PS]
+    %endif
+%endif
+
+.loopH:
+    mov       r4d, (%1/4)
+.loopW:
+    PROCESS_CHROMA_SP_W4_4R
+
+%ifidn %3, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%elifidn %3, ps
+    paddd     m0, m6
+    paddd     m1, m6
+    paddd     m2, m6
+    paddd     m3, m6
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%else
+    paddd     m0, m6
+    paddd     m1, m6
+    paddd     m2, m6
+    paddd     m3, m6
+    %ifidn %3, pp
+        psrad     m0, INTERP_SHIFT_PP
+        psrad     m1, INTERP_SHIFT_PP
+        psrad     m2, INTERP_SHIFT_PP
+        psrad     m3, INTERP_SHIFT_PP
+    %else
+        psrad     m0, INTERP_SHIFT_SP
+        psrad     m1, INTERP_SHIFT_SP
+        psrad     m2, INTERP_SHIFT_SP
+        psrad     m3, INTERP_SHIFT_SP
+    %endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    pxor      m5, m5
+    CLIPW2    m0, m2, m5, m7
+%endif
+
+    movh      [r2], m0
+    movhps    [r2 + r3], m0
+    lea       r5, [r2 + 2 * r3]
+    movh      [r5], m2
+    movhps    [r5 + r3], m2
+
+    lea       r5, [4 * r1 - 2 * 4]
+    sub       r0, r5
+    add       r2, 2 * 4
+
+    dec       r4d
+    jnz       .loopW
+
+    lea       r0, [r0 + 4 * r1 - 2 * %1]
+    lea       r2, [r2 + 4 * r3 - 2 * %1]
+
+    dec       dword [rsp]
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SS 4, 4, ss, 6
+    FILTER_VER_CHROMA_SS 4, 8, ss, 6
+    FILTER_VER_CHROMA_SS 16, 16, ss, 6
+    FILTER_VER_CHROMA_SS 16, 8, ss, 6
+    FILTER_VER_CHROMA_SS 16, 12, ss, 6
+    FILTER_VER_CHROMA_SS 12, 16, ss, 6
+    FILTER_VER_CHROMA_SS 16, 4, ss, 6
+    FILTER_VER_CHROMA_SS 4, 16, ss, 6
+    FILTER_VER_CHROMA_SS 32, 32, ss, 6
+    FILTER_VER_CHROMA_SS 32, 16, ss, 6
+    FILTER_VER_CHROMA_SS 16, 32, ss, 6
+    FILTER_VER_CHROMA_SS 32, 24, ss, 6
+    FILTER_VER_CHROMA_SS 24, 32, ss, 6
+    FILTER_VER_CHROMA_SS 32, 8, ss, 6
+
+    FILTER_VER_CHROMA_SS 4, 4, ps, 7
+    FILTER_VER_CHROMA_SS 4, 8, ps, 7
+    FILTER_VER_CHROMA_SS 16, 16, ps, 7
+    FILTER_VER_CHROMA_SS 16, 8, ps, 7
+    FILTER_VER_CHROMA_SS 16, 12, ps, 7
+    FILTER_VER_CHROMA_SS 12, 16, ps, 7
+    FILTER_VER_CHROMA_SS 16, 4, ps, 7
+    FILTER_VER_CHROMA_SS 4, 16, ps, 7
+    FILTER_VER_CHROMA_SS 32, 32, ps, 7
+    FILTER_VER_CHROMA_SS 32, 16, ps, 7
+    FILTER_VER_CHROMA_SS 16, 32, ps, 7
+    FILTER_VER_CHROMA_SS 32, 24, ps, 7
+    FILTER_VER_CHROMA_SS 24, 32, ps, 7
+    FILTER_VER_CHROMA_SS 32, 8, ps, 7
+
+    FILTER_VER_CHROMA_SS 4, 4, sp, 8
+    FILTER_VER_CHROMA_SS 4, 8, sp, 8
+    FILTER_VER_CHROMA_SS 16, 16, sp, 8
+    FILTER_VER_CHROMA_SS 16, 8, sp, 8
+    FILTER_VER_CHROMA_SS 16, 12, sp, 8
+    FILTER_VER_CHROMA_SS 12, 16, sp, 8
+    FILTER_VER_CHROMA_SS 16, 4, sp, 8
+    FILTER_VER_CHROMA_SS 4, 16, sp, 8
+    FILTER_VER_CHROMA_SS 32, 32, sp, 8
+    FILTER_VER_CHROMA_SS 32, 16, sp, 8
+    FILTER_VER_CHROMA_SS 16, 32, sp, 8
+    FILTER_VER_CHROMA_SS 32, 24, sp, 8
+    FILTER_VER_CHROMA_SS 24, 32, sp, 8
+    FILTER_VER_CHROMA_SS 32, 8, sp, 8
+
+    FILTER_VER_CHROMA_SS 4, 4, pp, 8
+    FILTER_VER_CHROMA_SS 4, 8, pp, 8
+    FILTER_VER_CHROMA_SS 16, 16, pp, 8
+    FILTER_VER_CHROMA_SS 16, 8, pp, 8
+    FILTER_VER_CHROMA_SS 16, 12, pp, 8
+    FILTER_VER_CHROMA_SS 12, 16, pp, 8
+    FILTER_VER_CHROMA_SS 16, 4, pp, 8
+    FILTER_VER_CHROMA_SS 4, 16, pp, 8
+    FILTER_VER_CHROMA_SS 32, 32, pp, 8
+    FILTER_VER_CHROMA_SS 32, 16, pp, 8
+    FILTER_VER_CHROMA_SS 16, 32, pp, 8
+    FILTER_VER_CHROMA_SS 32, 24, pp, 8
+    FILTER_VER_CHROMA_SS 24, 32, pp, 8
+    FILTER_VER_CHROMA_SS 32, 8, pp, 8
+
+
+    FILTER_VER_CHROMA_SS 16, 24, ss, 6
+    FILTER_VER_CHROMA_SS 12, 32, ss, 6
+    FILTER_VER_CHROMA_SS 4, 32, ss, 6
+    FILTER_VER_CHROMA_SS 32, 64, ss, 6
+    FILTER_VER_CHROMA_SS 16, 64, ss, 6
+    FILTER_VER_CHROMA_SS 32, 48, ss, 6
+    FILTER_VER_CHROMA_SS 24, 64, ss, 6
+
+    FILTER_VER_CHROMA_SS 16, 24, ps, 7
+    FILTER_VER_CHROMA_SS 12, 32, ps, 7
+    FILTER_VER_CHROMA_SS 4, 32, ps, 7
+    FILTER_VER_CHROMA_SS 32, 64, ps, 7
+    FILTER_VER_CHROMA_SS 16, 64, ps, 7
+    FILTER_VER_CHROMA_SS 32, 48, ps, 7
+    FILTER_VER_CHROMA_SS 24, 64, ps, 7
+
+    FILTER_VER_CHROMA_SS 16, 24, sp, 8
+    FILTER_VER_CHROMA_SS 12, 32, sp, 8
+    FILTER_VER_CHROMA_SS 4, 32, sp, 8
+    FILTER_VER_CHROMA_SS 32, 64, sp, 8
+    FILTER_VER_CHROMA_SS 16, 64, sp, 8
+    FILTER_VER_CHROMA_SS 32, 48, sp, 8
+    FILTER_VER_CHROMA_SS 24, 64, sp, 8
+
+    FILTER_VER_CHROMA_SS 16, 24, pp, 8
+    FILTER_VER_CHROMA_SS 12, 32, pp, 8
+    FILTER_VER_CHROMA_SS 4, 32, pp, 8
+    FILTER_VER_CHROMA_SS 32, 64, pp, 8
+    FILTER_VER_CHROMA_SS 16, 64, pp, 8
+    FILTER_VER_CHROMA_SS 32, 48, pp, 8
+    FILTER_VER_CHROMA_SS 24, 64, pp, 8
+
+
+    FILTER_VER_CHROMA_SS 48, 64, ss, 6
+    FILTER_VER_CHROMA_SS 64, 48, ss, 6
+    FILTER_VER_CHROMA_SS 64, 64, ss, 6
+    FILTER_VER_CHROMA_SS 64, 32, ss, 6
+    FILTER_VER_CHROMA_SS 64, 16, ss, 6
+
+    FILTER_VER_CHROMA_SS 48, 64, ps, 7
+    FILTER_VER_CHROMA_SS 64, 48, ps, 7
+    FILTER_VER_CHROMA_SS 64, 64, ps, 7
+    FILTER_VER_CHROMA_SS 64, 32, ps, 7
+    FILTER_VER_CHROMA_SS 64, 16, ps, 7
+
+    FILTER_VER_CHROMA_SS 48, 64, sp, 8
+    FILTER_VER_CHROMA_SS 64, 48, sp, 8
+    FILTER_VER_CHROMA_SS 64, 64, sp, 8
+    FILTER_VER_CHROMA_SS 64, 32, sp, 8
+    FILTER_VER_CHROMA_SS 64, 16, sp, 8
+
+    FILTER_VER_CHROMA_SS 48, 64, pp, 8
+    FILTER_VER_CHROMA_SS 64, 48, pp, 8
+    FILTER_VER_CHROMA_SS 64, 64, pp, 8
+    FILTER_VER_CHROMA_SS 64, 32, pp, 8
+    FILTER_VER_CHROMA_SS 64, 16, pp, 8
+
+
+%macro PROCESS_CHROMA_SP_W2_4R 1
+    movd       m0, [r0]
+    movd       m1, [r0 + r1]
+    punpcklwd  m0, m1                          ;m0=[0 1]
+
+    lea        r0, [r0 + 2 * r1]
+    movd       m2, [r0]
+    punpcklwd  m1, m2                          ;m1=[1 2]
+    punpcklqdq m0, m1                          ;m0=[0 1 1 2]
+    pmaddwd    m0, [%1 + 0 *32]                ;m0=[0+1 1+2] Row 1-2
+
+    movd       m1, [r0 + r1]
+    punpcklwd  m2, m1                          ;m2=[2 3]
+
+    lea        r0, [r0 + 2 * r1]
+    movd       m3, [r0]
+    punpcklwd  m1, m3                          ;m2=[3 4]
+    punpcklqdq m2, m1                          ;m2=[2 3 3 4]
+
+    pmaddwd    m4, m2, [%1 + 1 * 32]           ;m4=[2+3 3+4] Row 1-2
+    pmaddwd    m2, [%1 + 0 * 32]               ;m2=[2+3 3+4] Row 3-4
+    paddd      m0, m4                          ;m0=[0+1+2+3 1+2+3+4] Row 1-2
+
+    movd       m1, [r0 + r1]
+    punpcklwd  m3, m1                          ;m3=[4 5]
+
+    movd       m4, [r0 + 2 * r1]
+    punpcklwd  m1, m4                          ;m1=[5 6]
+    punpcklqdq m3, m1                          ;m2=[4 5 5 6]
+    pmaddwd    m3, [%1 + 1 * 32]               ;m3=[4+5 5+6] Row 3-4
+    paddd      m2, m3                          ;m2=[2+3+4+5 3+4+5+6] Row 3-4
+%endmacro
+;---------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vertical_%2_2x%1(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W2 3
+INIT_XMM sse4
+cglobal interp_4tap_vert_%2_2x%1, 5, 6, %3
+
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       r4d, (%1/4)
+%ifnidn %2, ss
+    %ifnidn %2, ps
+        pxor      m7, m7
+        mova      m6, [pw_pixel_max]
+        %ifidn %2, pp
+            mova      m5, [INTERP_OFFSET_PP]
+        %else
+            mova      m5, [INTERP_OFFSET_SP]
+        %endif
+    %else
+        mova      m5, [INTERP_OFFSET_PS]
+    %endif
+%endif
+
+.loopH:
+    PROCESS_CHROMA_SP_W2_4R r5
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m2, 6
+    packssdw  m0, m2
+%elifidn %2, ps
+    paddd     m0, m5
+    paddd     m2, m5
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    packssdw  m0, m2
+%else
+    paddd     m0, m5
+    paddd     m2, m5
+    %ifidn %2, pp
+        psrad     m0, INTERP_SHIFT_PP
+        psrad     m2, INTERP_SHIFT_PP
+    %else
+        psrad     m0, INTERP_SHIFT_SP
+        psrad     m2, INTERP_SHIFT_SP
+    %endif
+    packusdw  m0, m2
+    CLIPW     m0, m7,    m6
+%endif
+
+    movd      [r2], m0
+    pextrd    [r2 + r3], m0, 1
+    lea       r2, [r2 + 2 * r3]
+    pextrd    [r2], m0, 2
+    pextrd    [r2 + r3], m0, 3
+
+    lea       r2, [r2 + 2 * r3]
+
+    dec       r4d
+    jnz       .loopH
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_W2 4, ss, 5
+FILTER_VER_CHROMA_W2 8, ss, 5
+
+FILTER_VER_CHROMA_W2 4, pp, 8
+FILTER_VER_CHROMA_W2 8, pp, 8
+
+FILTER_VER_CHROMA_W2 4, ps, 6
+FILTER_VER_CHROMA_W2 8, ps, 6
+
+FILTER_VER_CHROMA_W2 4, sp, 8
+FILTER_VER_CHROMA_W2 8, sp, 8
+
+FILTER_VER_CHROMA_W2 16, ss, 5
+FILTER_VER_CHROMA_W2 16, pp, 8
+FILTER_VER_CHROMA_W2 16, ps, 6
+FILTER_VER_CHROMA_W2 16, sp, 8
+
+
+;---------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_%1_4x2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W4 3
+INIT_XMM sse4
+cglobal interp_4tap_vert_%2_4x%1, 5, 6, %3
+    add        r1d, r1d
+    add        r3d, r3d
+    sub        r0, r1
+    shl        r4d, 6
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeffV]
+    lea        r5, [r5 + r4]
+%else
+    lea        r5, [tab_ChromaCoeffV + r4]
+%endif
+
+%ifnidn %2, 2
+    mov        r4d, %1/2
+%endif
+
+%ifnidn %2, ss
+    %ifnidn %2, ps
+        pxor      m6, m6
+        mova      m5, [pw_pixel_max]
+        %ifidn %2, pp
+            mova      m4, [INTERP_OFFSET_PP]
+        %else
+            mova      m4, [INTERP_OFFSET_SP]
+        %endif
+    %else
+        mova      m4, [INTERP_OFFSET_PS]
+    %endif
+%endif
+
+%ifnidn %2, 2
+.loop:
+%endif
+
+    movh       m0, [r0]
+    movh       m1, [r0 + r1]
+    punpcklwd  m0, m1                          ;m0=[0 1]
+    pmaddwd    m0, [r5 + 0 *32]                ;m0=[0+1]  Row1
+
+    lea        r0, [r0 + 2 * r1]
+    movh       m2, [r0]
+    punpcklwd  m1, m2                          ;m1=[1 2]
+    pmaddwd    m1, [r5 + 0 *32]                ;m1=[1+2]  Row2
+
+    movh       m3, [r0 + r1]
+    punpcklwd  m2, m3                          ;m4=[2 3]
+    pmaddwd    m2, [r5 + 1 * 32]
+    paddd      m0, m2                          ;m0=[0+1+2+3]  Row1 done
+
+    movh       m2, [r0 + 2 * r1]
+    punpcklwd  m3, m2                          ;m5=[3 4]
+    pmaddwd    m3, [r5 + 1 * 32]
+    paddd      m1, m3                          ;m1=[1+2+3+4]  Row2 done
+
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    packssdw  m0, m1
+%elifidn %2, ps
+    paddd     m0, m4
+    paddd     m1, m4
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    packssdw  m0, m1
+%else
+    paddd     m0, m4
+    paddd     m1, m4
+    %ifidn %2, pp
+        psrad     m0, INTERP_SHIFT_PP
+        psrad     m1, INTERP_SHIFT_PP
+    %else
+        psrad     m0, INTERP_SHIFT_SP
+        psrad     m1, INTERP_SHIFT_SP
+    %endif
+    packusdw  m0, m1
+    CLIPW     m0, m6,    m5
+%endif
+
+    movh       [r2], m0
+    movhps     [r2 + r3], m0
+
+%ifnidn %2, 2
+    lea        r2, [r2 + r3 * 2]
+    dec        r4d
+    jnz        .loop
+%endif
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_W4 2, ss, 4
+FILTER_VER_CHROMA_W4 2, pp, 7
+FILTER_VER_CHROMA_W4 2, ps, 5
+FILTER_VER_CHROMA_W4 2, sp, 7
+
+FILTER_VER_CHROMA_W4 4, ss, 4
+FILTER_VER_CHROMA_W4 4, pp, 7
+FILTER_VER_CHROMA_W4 4, ps, 5
+FILTER_VER_CHROMA_W4 4, sp, 7
+
+;-------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vertical_%1_6x8(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-------------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W6 3
+INIT_XMM sse4
+cglobal interp_4tap_vert_%2_6x%1, 5, 7, %3
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r6, [r5 + r4]
+%else
+    lea       r6, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       r4d, %1/4
+
+%ifnidn %2, ss
+    %ifnidn %2, ps
+        mova      m7, [pw_pixel_max]
+        %ifidn %2, pp
+            mova      m6, [INTERP_OFFSET_PP]
+        %else
+            mova      m6, [INTERP_OFFSET_SP]
+        %endif
+    %else
+        mova      m6, [INTERP_OFFSET_PS]
+    %endif
+%endif
+
+.loopH:
+    PROCESS_CHROMA_SP_W4_4R
+
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%elifidn %2, ps
+    paddd     m0, m6
+    paddd     m1, m6
+    paddd     m2, m6
+    paddd     m3, m6
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%else
+    paddd     m0, m6
+    paddd     m1, m6
+    paddd     m2, m6
+    paddd     m3, m6
+    %ifidn %2, pp
+        psrad     m0, INTERP_SHIFT_PP
+        psrad     m1, INTERP_SHIFT_PP
+        psrad     m2, INTERP_SHIFT_PP
+        psrad     m3, INTERP_SHIFT_PP
+    %else
+        psrad     m0, INTERP_SHIFT_SP
+        psrad     m1, INTERP_SHIFT_SP
+        psrad     m2, INTERP_SHIFT_SP
+        psrad     m3, INTERP_SHIFT_SP
+    %endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    pxor      m5, m5
+    CLIPW2    m0, m2, m5, m7
+%endif
+
+    movh      [r2], m0
+    movhps    [r2 + r3], m0
+    lea       r5, [r2 + 2 * r3]
+    movh      [r5], m2
+    movhps    [r5 + r3], m2
+
+    lea       r5, [4 * r1 - 2 * 4]
+    sub       r0, r5
+    add       r2, 2 * 4
+
+    PROCESS_CHROMA_SP_W2_4R r6
+
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m2, 6
+    packssdw  m0, m2
+%elifidn %2, ps
+    paddd     m0, m6
+    paddd     m2, m6
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    packssdw  m0, m2
+%else
+    paddd     m0, m6
+    paddd     m2, m6
+    %ifidn %2, pp
+        psrad     m0, INTERP_SHIFT_PP
+        psrad     m2, INTERP_SHIFT_PP
+    %else
+        psrad     m0, INTERP_SHIFT_SP
+        psrad     m2, INTERP_SHIFT_SP
+    %endif
+    packusdw  m0, m2
+    CLIPW     m0, m5,    m7
+%endif
+
+    movd      [r2], m0
+    pextrd    [r2 + r3], m0, 1
+    lea       r2, [r2 + 2 * r3]
+    pextrd    [r2], m0, 2
+    pextrd    [r2 + r3], m0, 3
+
+    sub       r0, 2 * 4
+    lea       r2, [r2 + 2 * r3 - 2 * 4]
+
+    dec       r4d
+    jnz       .loopH
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_W6 8, ss, 6
+FILTER_VER_CHROMA_W6 8, ps, 7
+FILTER_VER_CHROMA_W6 8, sp, 8
+FILTER_VER_CHROMA_W6 8, pp, 8
+
+FILTER_VER_CHROMA_W6 16, ss, 6
+FILTER_VER_CHROMA_W6 16, ps, 7
+FILTER_VER_CHROMA_W6 16, sp, 8
+FILTER_VER_CHROMA_W6 16, pp, 8
+
+%macro PROCESS_CHROMA_SP_W8_2R 0
+    movu       m1, [r0]
+    movu       m3, [r0 + r1]
+    punpcklwd  m0, m1, m3
+    pmaddwd    m0, [r5 + 0 * 32]                ;m0 = [0l+1l]  Row1l
+    punpckhwd  m1, m3
+    pmaddwd    m1, [r5 + 0 * 32]                ;m1 = [0h+1h]  Row1h
+
+    movu       m4, [r0 + 2 * r1]
+    punpcklwd  m2, m3, m4
+    pmaddwd    m2, [r5 + 0 * 32]                ;m2 = [1l+2l]  Row2l
+    punpckhwd  m3, m4
+    pmaddwd    m3, [r5 + 0 * 32]                ;m3 = [1h+2h]  Row2h
+
+    lea        r0, [r0 + 2 * r1]
+    movu       m5, [r0 + r1]
+    punpcklwd  m6, m4, m5
+    pmaddwd    m6, [r5 + 1 * 32]                ;m6 = [2l+3l]  Row1l
+    paddd      m0, m6                           ;m0 = [0l+1l+2l+3l]  Row1l sum
+    punpckhwd  m4, m5
+    pmaddwd    m4, [r5 + 1 * 32]                ;m6 = [2h+3h]  Row1h
+    paddd      m1, m4                           ;m1 = [0h+1h+2h+3h]  Row1h sum
+
+    movu       m4, [r0 + 2 * r1]
+    punpcklwd  m6, m5, m4
+    pmaddwd    m6, [r5 + 1 * 32]                ;m6 = [3l+4l]  Row2l
+    paddd      m2, m6                           ;m2 = [1l+2l+3l+4l]  Row2l sum
+    punpckhwd  m5, m4
+    pmaddwd    m5, [r5 + 1 * 32]                ;m1 = [3h+4h]  Row2h
+    paddd      m3, m5                           ;m3 = [1h+2h+3h+4h]  Row2h sum
+%endmacro
+
+;----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_%3_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W8 4
+INIT_XMM sse2
+cglobal interp_4tap_vert_%3_%1x%2, 5, 6, %4
+
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       r4d, %2/2
+
+%ifidn %3, pp
+    mova      m7, [INTERP_OFFSET_PP]
+%elifidn %3, sp
+    mova      m7, [INTERP_OFFSET_SP]
+%elifidn %3, ps
+    mova      m7, [INTERP_OFFSET_PS]
+%endif
+
+.loopH:
+    PROCESS_CHROMA_SP_W8_2R
+
+%ifidn %3, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%elifidn %3, ps
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%else
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    %ifidn %3, pp
+        psrad     m0, INTERP_SHIFT_PP
+        psrad     m1, INTERP_SHIFT_PP
+        psrad     m2, INTERP_SHIFT_PP
+        psrad     m3, INTERP_SHIFT_PP
+    %else
+        psrad     m0, INTERP_SHIFT_SP
+        psrad     m1, INTERP_SHIFT_SP
+        psrad     m2, INTERP_SHIFT_SP
+        psrad     m3, INTERP_SHIFT_SP
+    %endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    pxor      m5, m5
+    mova      m6, [pw_pixel_max]
+    CLIPW2    m0, m2, m5, m6
+%endif
+
+    movu      [r2], m0
+    movu      [r2 + r3], m2
+
+    lea       r2, [r2 + 2 * r3]
+
+    dec       r4d
+    jnz       .loopH
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_W8 8, 2, ss, 7
+FILTER_VER_CHROMA_W8 8, 4, ss, 7
+FILTER_VER_CHROMA_W8 8, 6, ss, 7
+FILTER_VER_CHROMA_W8 8, 8, ss, 7
+FILTER_VER_CHROMA_W8 8, 16, ss, 7
+FILTER_VER_CHROMA_W8 8, 32, ss, 7
+
+FILTER_VER_CHROMA_W8 8, 2, sp, 8
+FILTER_VER_CHROMA_W8 8, 4, sp, 8
+FILTER_VER_CHROMA_W8 8, 6, sp, 8
+FILTER_VER_CHROMA_W8 8, 8, sp, 8
+FILTER_VER_CHROMA_W8 8, 16, sp, 8
+FILTER_VER_CHROMA_W8 8, 32, sp, 8
+
+FILTER_VER_CHROMA_W8 8, 2, ps, 8
+FILTER_VER_CHROMA_W8 8, 4, ps, 8
+FILTER_VER_CHROMA_W8 8, 6, ps, 8
+FILTER_VER_CHROMA_W8 8, 8, ps, 8
+FILTER_VER_CHROMA_W8 8, 16, ps, 8
+FILTER_VER_CHROMA_W8 8, 32, ps, 8
+
+FILTER_VER_CHROMA_W8 8, 2, pp, 8
+FILTER_VER_CHROMA_W8 8, 4, pp, 8
+FILTER_VER_CHROMA_W8 8, 6, pp, 8
+FILTER_VER_CHROMA_W8 8, 8, pp, 8
+FILTER_VER_CHROMA_W8 8, 16, pp, 8
+FILTER_VER_CHROMA_W8 8, 32, pp, 8
+
+FILTER_VER_CHROMA_W8 8, 12, ss, 7
+FILTER_VER_CHROMA_W8 8, 64, ss, 7
+FILTER_VER_CHROMA_W8 8, 12, sp, 8
+FILTER_VER_CHROMA_W8 8, 64, sp, 8
+FILTER_VER_CHROMA_W8 8, 12, ps, 8
+FILTER_VER_CHROMA_W8 8, 64, ps, 8
+FILTER_VER_CHROMA_W8 8, 12, pp, 8
+FILTER_VER_CHROMA_W8 8, 64, pp, 8
+
+%macro PROCESS_CHROMA_VERT_W16_2R 0
+    movu       m1, [r0]
+    movu       m3, [r0 + r1]
+    punpcklwd  m0, m1, m3
+    pmaddwd    m0, [r5 + 0 * 32]
+    punpckhwd  m1, m3
+    pmaddwd    m1, [r5 + 0 * 32]
+
+    movu       m4, [r0 + 2 * r1]
+    punpcklwd  m2, m3, m4
+    pmaddwd    m2, [r5 + 0 * 32]
+    punpckhwd  m3, m4
+    pmaddwd    m3, [r5 + 0 * 32]
+
+    lea        r0, [r0 + 2 * r1]
+    movu       m5, [r0 + r1]
+    punpcklwd  m6, m4, m5
+    pmaddwd    m6, [r5 + 1 * 32]
+    paddd      m0, m6
+    punpckhwd  m4, m5
+    pmaddwd    m4, [r5 + 1 * 32]
+    paddd      m1, m4
+
+    movu       m4, [r0 + 2 * r1]
+    punpcklwd  m6, m5, m4
+    pmaddwd    m6, [r5 + 1 * 32]
+    paddd      m2, m6
+    punpckhwd  m5, m4
+    pmaddwd    m5, [r5 + 1 * 32]
+    paddd      m3, m5
+%endmacro
+
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_AVX2_6xN 2
+INIT_YMM avx2
+%if ARCH_X86_64
+cglobal interp_4tap_vert_%2_6x%1, 4, 7, 10
+    mov             r4d, r4m
+    add             r1d, r1d
+    add             r3d, r3d
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    sub             r0, r1
+    mov             r6d, %1/4
+
+%ifidn %2,pp
+    vbroadcasti128  m8, [INTERP_OFFSET_PP]
+%elifidn %2, sp
+    vbroadcasti128  m8, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m8, [INTERP_OFFSET_PS]
+%endif
+
+.loopH:
+    movu            xm0, [r0]
+    movu            xm1, [r0 + r1]
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+
+    movu            xm2, [r0 + r1 * 2]
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+
+    lea             r4, [r1 * 3]
+    movu            xm3, [r0 + r4]
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m4
+
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    pmaddwd         m3, [r5]
+    paddd           m1, m5
+
+    movu            xm5, [r0 + r1]
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    pmaddwd         m4, [r5]
+    paddd           m2, m6
+
+    movu            xm6, [r0 + r1 * 2]
+    punpckhwd       xm7, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddwd         m7, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m7
+    lea             r4, [r3 * 3]
+%ifidn %2,ss
+    psrad           m0, 6
+    psrad           m1, 6
+    psrad           m2, 6
+    psrad           m3, 6
+%else
+    paddd           m0, m8
+    paddd           m1, m8
+    paddd           m2, m8
+    paddd           m3, m8
+%ifidn %2,pp
+    psrad           m0, INTERP_SHIFT_PP
+    psrad           m1, INTERP_SHIFT_PP
+    psrad           m2, INTERP_SHIFT_PP
+    psrad           m3, INTERP_SHIFT_PP
+%elifidn %2, sp
+    psrad           m0, INTERP_SHIFT_SP
+    psrad           m1, INTERP_SHIFT_SP
+    psrad           m2, INTERP_SHIFT_SP
+    psrad           m3, INTERP_SHIFT_SP
+%else
+    psrad           m0, INTERP_SHIFT_PS
+    psrad           m1, INTERP_SHIFT_PS
+    psrad           m2, INTERP_SHIFT_PS
+    psrad           m3, INTERP_SHIFT_PS
+%endif
+%endif
+
+    packssdw        m0, m1
+    packssdw        m2, m3
+    vpermq          m0, m0, q3120
+    vpermq          m2, m2, q3120
+    pxor            m5, m5
+    mova            m9, [pw_pixel_max]
+%ifidn %2,pp
+    CLIPW           m0, m5, m9
+    CLIPW           m2, m5, m9
+%elifidn %2, sp
+    CLIPW           m0, m5, m9
+    CLIPW           m2, m5, m9
+%endif
+
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+    movq            [r2], xm0
+    pextrd          [r2 + 8], xm0, 2
+    movq            [r2 + r3], xm1
+    pextrd          [r2 + r3 + 8], xm1, 2
+    movq            [r2 + r3 * 2], xm2
+    pextrd          [r2 + r3 * 2 + 8], xm2, 2
+    movq            [r2 + r4], xm3
+    pextrd          [r2 + r4 + 8], xm3, 2
+
+    lea             r2, [r2 + r3 * 4]
+    dec r6d
+    jnz .loopH
+    RET
+%endif
+%endmacro
+FILTER_VER_CHROMA_AVX2_6xN 8, pp
+FILTER_VER_CHROMA_AVX2_6xN 8, ps
+FILTER_VER_CHROMA_AVX2_6xN 8, ss
+FILTER_VER_CHROMA_AVX2_6xN 8, sp
+FILTER_VER_CHROMA_AVX2_6xN 16, pp
+FILTER_VER_CHROMA_AVX2_6xN 16, ps
+FILTER_VER_CHROMA_AVX2_6xN 16, ss
+FILTER_VER_CHROMA_AVX2_6xN 16, sp
+
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W16_16xN_avx2 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%2_16x%1, 5, 6, %3
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       r4d, %1/2
+
+%ifidn %2, pp
+    vbroadcasti128  m7, [INTERP_OFFSET_PP]
+%elifidn %2, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%elifidn %2, ps
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+
+.loopH:
+    PROCESS_CHROMA_VERT_W16_2R
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%elifidn %2, ps
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%else
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+ %ifidn %2, pp
+    psrad     m0, INTERP_SHIFT_PP
+    psrad     m1, INTERP_SHIFT_PP
+    psrad     m2, INTERP_SHIFT_PP
+    psrad     m3, INTERP_SHIFT_PP
+%else
+    psrad     m0, INTERP_SHIFT_SP
+    psrad     m1, INTERP_SHIFT_SP
+    psrad     m2, INTERP_SHIFT_SP
+    psrad     m3, INTERP_SHIFT_SP
+%endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    pxor      m5, m5
+    CLIPW2    m0, m2, m5, [pw_pixel_max]
+%endif
+
+    movu      [r2], m0
+    movu      [r2 + r3], m2
+    lea       r2, [r2 + 2 * r3]
+    dec       r4d
+    jnz       .loopH
+    RET
+%endmacro
+    FILTER_VER_CHROMA_W16_16xN_avx2 4, pp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 8, pp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 12, pp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 24, pp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 16, pp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 32, pp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 64, pp, 8
+
+    FILTER_VER_CHROMA_W16_16xN_avx2 4, ps, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 8, ps, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 12, ps, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 24, ps, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 16, ps, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 32, ps, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 64, ps, 8
+
+    FILTER_VER_CHROMA_W16_16xN_avx2 4, ss, 7
+    FILTER_VER_CHROMA_W16_16xN_avx2 8, ss, 7
+    FILTER_VER_CHROMA_W16_16xN_avx2 12, ss, 7
+    FILTER_VER_CHROMA_W16_16xN_avx2 24, ss, 7
+    FILTER_VER_CHROMA_W16_16xN_avx2 16, ss, 7
+    FILTER_VER_CHROMA_W16_16xN_avx2 32, ss, 7
+    FILTER_VER_CHROMA_W16_16xN_avx2 64, ss, 7
+
+    FILTER_VER_CHROMA_W16_16xN_avx2 4, sp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 8, sp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 12, sp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 24, sp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 16, sp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 32, sp, 8
+    FILTER_VER_CHROMA_W16_16xN_avx2 64, sp, 8
+
+%macro PROCESS_CHROMA_VERT_W32_2R 0
+    movu       m1, [r0]
+    movu       m3, [r0 + r1]
+    punpcklwd  m0, m1, m3
+    pmaddwd    m0, [r5 + 0 * mmsize]
+    punpckhwd  m1, m3
+    pmaddwd    m1, [r5 + 0 * mmsize]
+
+    movu       m9, [r0 + mmsize]
+    movu       m11, [r0 + r1 + mmsize]
+    punpcklwd  m8, m9, m11
+    pmaddwd    m8, [r5 + 0 * mmsize]
+    punpckhwd  m9, m11
+    pmaddwd    m9, [r5 + 0 * mmsize]
+
+    movu       m4, [r0 + 2 * r1]
+    punpcklwd  m2, m3, m4
+    pmaddwd    m2, [r5 + 0 * mmsize]
+    punpckhwd  m3, m4
+    pmaddwd    m3, [r5 + 0 * mmsize]
+
+    movu       m12, [r0 + 2 * r1 + mmsize]
+    punpcklwd  m10, m11, m12
+    pmaddwd    m10, [r5 + 0 * mmsize]
+    punpckhwd  m11, m12
+    pmaddwd    m11, [r5 + 0 * mmsize]
+
+    lea        r6, [r0 + 2 * r1]
+    movu       m5, [r6 + r1]
+    punpcklwd  m6, m4, m5
+    pmaddwd    m6, [r5 + 1 * mmsize]
+    paddd      m0, m6
+    punpckhwd  m4, m5
+    pmaddwd    m4, [r5 + 1 * mmsize]
+    paddd      m1, m4
+
+    movu       m13, [r6 + r1 + mmsize]
+    punpcklwd  m14, m12, m13
+    pmaddwd    m14, [r5 + 1 * mmsize]
+    paddd      m8, m14
+    punpckhwd  m12, m13
+    pmaddwd    m12, [r5 + 1 * mmsize]
+    paddd      m9, m12
+
+    movu       m4, [r6 + 2 * r1]
+    punpcklwd  m6, m5, m4
+    pmaddwd    m6, [r5 + 1 * mmsize]
+    paddd      m2, m6
+    punpckhwd  m5, m4
+    pmaddwd    m5, [r5 + 1 * mmsize]
+    paddd      m3, m5
+
+    movu       m12, [r6 + 2 * r1 + mmsize]
+    punpcklwd  m14, m13, m12
+    pmaddwd    m14, [r5 + 1 * mmsize]
+    paddd      m10, m14
+    punpckhwd  m13, m12
+    pmaddwd    m13, [r5 + 1 * mmsize]
+    paddd      m11, m13
+%endmacro
+
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W16_32xN_avx2 3
+INIT_YMM avx2
+%if ARCH_X86_64
+cglobal interp_4tap_vert_%2_32x%1, 5, 7, %3
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+    mov       r4d, %1/2
+
+%ifidn %2, pp
+    vbroadcasti128  m7, [INTERP_OFFSET_PP]
+%elifidn %2, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%elifidn %2, ps
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+
+.loopH:
+    PROCESS_CHROMA_VERT_W32_2R
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    psrad     m8, 6
+    psrad     m9, 6
+    psrad     m10, 6
+    psrad     m11, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+    packssdw  m8, m9
+    packssdw  m10, m11
+%elifidn %2, ps
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+    paddd     m8, m7
+    paddd     m9, m7
+    paddd     m10, m7
+    paddd     m11, m7
+    psrad     m8, INTERP_SHIFT_PS
+    psrad     m9, INTERP_SHIFT_PS
+    psrad     m10, INTERP_SHIFT_PS
+    psrad     m11, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+    packssdw  m8, m9
+    packssdw  m10, m11
+%else
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    paddd     m8, m7
+    paddd     m9, m7
+    paddd     m10, m7
+    paddd     m11, m7
+ %ifidn %2, pp
+    psrad     m0, INTERP_SHIFT_PP
+    psrad     m1, INTERP_SHIFT_PP
+    psrad     m2, INTERP_SHIFT_PP
+    psrad     m3, INTERP_SHIFT_PP
+    psrad     m8, INTERP_SHIFT_PP
+    psrad     m9, INTERP_SHIFT_PP
+    psrad     m10, INTERP_SHIFT_PP
+    psrad     m11, INTERP_SHIFT_PP
+%else
+    psrad     m0, INTERP_SHIFT_SP
+    psrad     m1, INTERP_SHIFT_SP
+    psrad     m2, INTERP_SHIFT_SP
+    psrad     m3, INTERP_SHIFT_SP
+    psrad     m8, INTERP_SHIFT_SP
+    psrad     m9, INTERP_SHIFT_SP
+    psrad     m10, INTERP_SHIFT_SP
+    psrad     m11, INTERP_SHIFT_SP
+%endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    packssdw  m8, m9
+    packssdw  m10, m11
+    pxor      m5, m5
+    CLIPW2    m0, m2, m5, [pw_pixel_max]
+    CLIPW2    m8, m10, m5, [pw_pixel_max]
+%endif
+
+    movu      [r2], m0
+    movu      [r2 + r3], m2
+    movu      [r2 + mmsize], m8
+    movu      [r2 + r3 + mmsize], m10
+    lea       r2, [r2 + 2 * r3]
+    lea       r0, [r0 + 2 * r1]
+    dec       r4d
+    jnz       .loopH
+    RET
+%endif
+%endmacro
+    FILTER_VER_CHROMA_W16_32xN_avx2 8, pp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 16, pp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 24, pp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 32, pp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 48, pp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 64, pp, 15
+
+    FILTER_VER_CHROMA_W16_32xN_avx2 8, ps, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 16, ps, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 24, ps, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 32, ps, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 48, ps, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 64, ps, 15
+
+    FILTER_VER_CHROMA_W16_32xN_avx2 8, ss, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 16, ss, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 24, ss, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 32, ss, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 48, ss, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 64, ss, 15
+
+    FILTER_VER_CHROMA_W16_32xN_avx2 8, sp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 16, sp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 24, sp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 32, sp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 48, sp, 15
+    FILTER_VER_CHROMA_W16_32xN_avx2 64, sp, 15
+
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W16_64xN_avx2 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%2_64x%1, 5, 7, %3
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+    mov       r4d, %1/2
+
+%ifidn %2, pp
+    vbroadcasti128  m7, [INTERP_OFFSET_PP]
+%elifidn %2, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%elifidn %2, ps
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+
+.loopH:
+%assign x 0
+%rep 4
+    movu       m1, [r0 + x]
+    movu       m3, [r0 + r1 + x]
+    movu       m5, [r5 + 0 * mmsize]
+    punpcklwd  m0, m1, m3
+    pmaddwd    m0, m5
+    punpckhwd  m1, m3
+    pmaddwd    m1, m5
+
+    movu       m4, [r0 + 2 * r1 + x]
+    punpcklwd  m2, m3, m4
+    pmaddwd    m2, m5
+    punpckhwd  m3, m4
+    pmaddwd    m3, m5
+
+    lea        r6, [r0 + 2 * r1]
+    movu       m5, [r6 + r1 + x]
+    punpcklwd  m6, m4, m5
+    pmaddwd    m6, [r5 + 1 * mmsize]
+    paddd      m0, m6
+    punpckhwd  m4, m5
+    pmaddwd    m4, [r5 + 1 * mmsize]
+    paddd      m1, m4
+
+    movu       m4, [r6 + 2 * r1 + x]
+    punpcklwd  m6, m5, m4
+    pmaddwd    m6, [r5 + 1 * mmsize]
+    paddd      m2, m6
+    punpckhwd  m5, m4
+    pmaddwd    m5, [r5 + 1 * mmsize]
+    paddd      m3, m5
+
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%elifidn %2, ps
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%else
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+%ifidn %2, pp
+    psrad     m0, INTERP_SHIFT_PP
+    psrad     m1, INTERP_SHIFT_PP
+    psrad     m2, INTERP_SHIFT_PP
+    psrad     m3, INTERP_SHIFT_PP
+%else
+    psrad     m0, INTERP_SHIFT_SP
+    psrad     m1, INTERP_SHIFT_SP
+    psrad     m2, INTERP_SHIFT_SP
+    psrad     m3, INTERP_SHIFT_SP
+%endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    pxor      m5, m5
+    CLIPW2    m0, m2, m5, [pw_pixel_max]
+%endif
+
+    movu      [r2 + x], m0
+    movu      [r2 + r3 + x], m2
+%assign x x+mmsize
+%endrep
+
+    lea       r2, [r2 + 2 * r3]
+    lea       r0, [r0 + 2 * r1]
+    dec       r4d
+    jnz       .loopH
+    RET
+%endmacro
+    FILTER_VER_CHROMA_W16_64xN_avx2 16, ss, 7
+    FILTER_VER_CHROMA_W16_64xN_avx2 32, ss, 7
+    FILTER_VER_CHROMA_W16_64xN_avx2 48, ss, 7
+    FILTER_VER_CHROMA_W16_64xN_avx2 64, ss, 7
+    FILTER_VER_CHROMA_W16_64xN_avx2 16, sp, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 32, sp, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 48, sp, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 64, sp, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 16, ps, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 32, ps, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 48, ps, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 64, ps, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 16, pp, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 32, pp, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 48, pp, 8
+    FILTER_VER_CHROMA_W16_64xN_avx2 64, pp, 8
+
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W16_12xN_avx2 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%2_12x%1, 5, 8, %3
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+    mov       r4d, %1/2
+
+%ifidn %2, pp
+    vbroadcasti128  m7, [INTERP_OFFSET_PP]
+%elifidn %2, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%elifidn %2, ps
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+
+.loopH:
+    PROCESS_CHROMA_VERT_W16_2R
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%elifidn %2, ps
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%else
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+ %ifidn %2, pp
+    psrad     m0, INTERP_SHIFT_PP
+    psrad     m1, INTERP_SHIFT_PP
+    psrad     m2, INTERP_SHIFT_PP
+    psrad     m3, INTERP_SHIFT_PP
+%else
+    psrad     m0, INTERP_SHIFT_SP
+    psrad     m1, INTERP_SHIFT_SP
+    psrad     m2, INTERP_SHIFT_SP
+    psrad     m3, INTERP_SHIFT_SP
+%endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    pxor      m5, m5
+    CLIPW2    m0, m2, m5, [pw_pixel_max]
+%endif
+
+    movu      [r2], xm0
+    movu      [r2 + r3], xm2
+    vextracti128 xm0, m0, 1
+    vextracti128 xm2, m2, 1
+    movq      [r2 + 16], xm0
+    movq      [r2 + r3 + 16], xm2
+    lea       r2, [r2 + 2 * r3]
+    dec       r4d
+    jnz       .loopH
+    RET
+%endmacro
+    FILTER_VER_CHROMA_W16_12xN_avx2 16, ss, 7
+    FILTER_VER_CHROMA_W16_12xN_avx2 16, sp, 8
+    FILTER_VER_CHROMA_W16_12xN_avx2 16, ps, 8
+    FILTER_VER_CHROMA_W16_12xN_avx2 16, pp, 8
+    FILTER_VER_CHROMA_W16_12xN_avx2 32, ss, 7
+    FILTER_VER_CHROMA_W16_12xN_avx2 32, sp, 8
+    FILTER_VER_CHROMA_W16_12xN_avx2 32, ps, 8
+    FILTER_VER_CHROMA_W16_12xN_avx2 32, pp, 8
+
+%macro PROCESS_CHROMA_VERT_W24_2R 0
+    movu       m1, [r0]
+    movu       m3, [r0 + r1]
+    punpcklwd  m0, m1, m3
+    pmaddwd    m0, [r5 + 0 * mmsize]
+    punpckhwd  m1, m3
+    pmaddwd    m1, [r5 + 0 * mmsize]
+
+    movu       xm9, [r0 + mmsize]
+    movu       xm11, [r0 + r1 + mmsize]
+    punpcklwd  xm8, xm9, xm11
+    pmaddwd    xm8, [r5 + 0 * mmsize]
+    punpckhwd  xm9, xm11
+    pmaddwd    xm9, [r5 + 0 * mmsize]
+
+    movu       m4, [r0 + 2 * r1]
+    punpcklwd  m2, m3, m4
+    pmaddwd    m2, [r5 + 0 * mmsize]
+    punpckhwd  m3, m4
+    pmaddwd    m3, [r5 + 0 * mmsize]
+
+    movu       xm12, [r0 + 2 * r1 + mmsize]
+    punpcklwd  xm10, xm11, xm12
+    pmaddwd    xm10, [r5 + 0 * mmsize]
+    punpckhwd  xm11, xm12
+    pmaddwd    xm11, [r5 + 0 * mmsize]
+
+    lea        r6, [r0 + 2 * r1]
+    movu       m5, [r6 + r1]
+    punpcklwd  m6, m4, m5
+    pmaddwd    m6, [r5 + 1 * mmsize]
+    paddd      m0, m6
+    punpckhwd  m4, m5
+    pmaddwd    m4, [r5 + 1 * mmsize]
+    paddd      m1, m4
+
+    movu       xm13, [r6 + r1 + mmsize]
+    punpcklwd  xm14, xm12, xm13
+    pmaddwd    xm14, [r5 + 1 * mmsize]
+    paddd      xm8, xm14
+    punpckhwd  xm12, xm13
+    pmaddwd    xm12, [r5 + 1 * mmsize]
+    paddd      xm9, xm12
+
+    movu       m4, [r6 + 2 * r1]
+    punpcklwd  m6, m5, m4
+    pmaddwd    m6, [r5 + 1 * mmsize]
+    paddd      m2, m6
+    punpckhwd  m5, m4
+    pmaddwd    m5, [r5 + 1 * mmsize]
+    paddd      m3, m5
+
+    movu       xm12, [r6 + 2 * r1 + mmsize]
+    punpcklwd  xm14, xm13, xm12
+    pmaddwd    xm14, [r5 + 1 * mmsize]
+    paddd      xm10, xm14
+    punpckhwd  xm13, xm12
+    pmaddwd    xm13, [r5 + 1 * mmsize]
+    paddd      xm11, xm13
+%endmacro
+
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W16_24xN_avx2 3
+INIT_YMM avx2
+%if ARCH_X86_64
+cglobal interp_4tap_vert_%2_24x%1, 5, 7, %3
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+    mov       r4d, %1/2
+
+%ifidn %2, pp
+    vbroadcasti128  m7, [INTERP_OFFSET_PP]
+%elifidn %2, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%elifidn %2, ps
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+
+.loopH:
+    PROCESS_CHROMA_VERT_W24_2R
+%ifidn %2, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    psrad     m8, 6
+    psrad     m9, 6
+    psrad     m10, 6
+    psrad     m11, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+    packssdw  m8, m9
+    packssdw  m10, m11
+%elifidn %2, ps
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+    paddd     m8, m7
+    paddd     m9, m7
+    paddd     m10, m7
+    paddd     m11, m7
+    psrad     m8, INTERP_SHIFT_PS
+    psrad     m9, INTERP_SHIFT_PS
+    psrad     m10, INTERP_SHIFT_PS
+    psrad     m11, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+    packssdw  m8, m9
+    packssdw  m10, m11
+%else
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    paddd     m8, m7
+    paddd     m9, m7
+    paddd     m10, m7
+    paddd     m11, m7
+ %ifidn %2, pp
+    psrad     m0, INTERP_SHIFT_PP
+    psrad     m1, INTERP_SHIFT_PP
+    psrad     m2, INTERP_SHIFT_PP
+    psrad     m3, INTERP_SHIFT_PP
+    psrad     m8, INTERP_SHIFT_PP
+    psrad     m9, INTERP_SHIFT_PP
+    psrad     m10, INTERP_SHIFT_PP
+    psrad     m11, INTERP_SHIFT_PP
+%else
+    psrad     m0, INTERP_SHIFT_SP
+    psrad     m1, INTERP_SHIFT_SP
+    psrad     m2, INTERP_SHIFT_SP
+    psrad     m3, INTERP_SHIFT_SP
+    psrad     m8, INTERP_SHIFT_SP
+    psrad     m9, INTERP_SHIFT_SP
+    psrad     m10, INTERP_SHIFT_SP
+    psrad     m11, INTERP_SHIFT_SP
+%endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    packssdw  m8, m9
+    packssdw  m10, m11
+    pxor      m5, m5
+    CLIPW2    m0, m2, m5, [pw_pixel_max]
+    CLIPW2    m8, m10, m5, [pw_pixel_max]
+%endif
+
+    movu      [r2], m0
+    movu      [r2 + r3], m2
+    movu      [r2 + mmsize], xm8
+    movu      [r2 + r3 + mmsize], xm10
+    lea       r2, [r2 + 2 * r3]
+    lea       r0, [r0 + 2 * r1]
+    dec       r4d
+    jnz       .loopH
+    RET
+%endif
+%endmacro
+    FILTER_VER_CHROMA_W16_24xN_avx2 32, ss, 15
+    FILTER_VER_CHROMA_W16_24xN_avx2 32, sp, 15
+    FILTER_VER_CHROMA_W16_24xN_avx2 32, ps, 15
+    FILTER_VER_CHROMA_W16_24xN_avx2 32, pp, 15
+    FILTER_VER_CHROMA_W16_24xN_avx2 64, ss, 15
+    FILTER_VER_CHROMA_W16_24xN_avx2 64, sp, 15
+    FILTER_VER_CHROMA_W16_24xN_avx2 64, ps, 15
+    FILTER_VER_CHROMA_W16_24xN_avx2 64, pp, 15
+
+   
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_W16_48x64_avx2 2
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_48x64, 5, 7, %2
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 6
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+    mov       r4d, 32
+
+%ifidn %1, pp
+    vbroadcasti128  m7, [INTERP_OFFSET_PP]
+%elifidn %1, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%elifidn %1, ps
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+
+.loopH:
+%assign x 0
+%rep 3
+    movu       m1, [r0 + x]
+    movu       m3, [r0 + r1 + x]
+    movu       m5, [r5 + 0 * mmsize]
+    punpcklwd  m0, m1, m3
+    pmaddwd    m0, m5
+    punpckhwd  m1, m3
+    pmaddwd    m1, m5
+
+    movu       m4, [r0 + 2 * r1 + x]
+    punpcklwd  m2, m3, m4
+    pmaddwd    m2, m5
+    punpckhwd  m3, m4
+    pmaddwd    m3, m5
+
+    lea        r6, [r0 + 2 * r1]
+    movu       m5, [r6 + r1 + x]
+    punpcklwd  m6, m4, m5
+    pmaddwd    m6, [r5 + 1 * mmsize]
+    paddd      m0, m6
+    punpckhwd  m4, m5
+    pmaddwd    m4, [r5 + 1 * mmsize]
+    paddd      m1, m4
+
+    movu       m4, [r6 + 2 * r1 + x]
+    punpcklwd  m6, m5, m4
+    pmaddwd    m6, [r5 + 1 * mmsize]
+    paddd      m2, m6
+    punpckhwd  m5, m4
+    pmaddwd    m5, [r5 + 1 * mmsize]
+    paddd      m3, m5
+
+%ifidn %1, ss
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%elifidn %1, ps
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+    psrad     m0, INTERP_SHIFT_PS
+    psrad     m1, INTERP_SHIFT_PS
+    psrad     m2, INTERP_SHIFT_PS
+    psrad     m3, INTERP_SHIFT_PS
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+%else
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+%ifidn %1, pp
+    psrad     m0, INTERP_SHIFT_PP
+    psrad     m1, INTERP_SHIFT_PP
+    psrad     m2, INTERP_SHIFT_PP
+    psrad     m3, INTERP_SHIFT_PP
+%else
+    psrad     m0, INTERP_SHIFT_SP
+    psrad     m1, INTERP_SHIFT_SP
+    psrad     m2, INTERP_SHIFT_SP
+    psrad     m3, INTERP_SHIFT_SP
+%endif
+    packssdw  m0, m1
+    packssdw  m2, m3
+    pxor      m5, m5
+    CLIPW2    m0, m2, m5, [pw_pixel_max]
+%endif
+
+    movu      [r2 + x], m0
+    movu      [r2 + r3 + x], m2
+%assign x x+mmsize
+%endrep
+
+    lea       r2, [r2 + 2 * r3]
+    lea       r0, [r0 + 2 * r1]
+    dec       r4d
+    jnz       .loopH
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_W16_48x64_avx2 pp, 8
+    FILTER_VER_CHROMA_W16_48x64_avx2 ps, 8
+    FILTER_VER_CHROMA_W16_48x64_avx2 ss, 7
+    FILTER_VER_CHROMA_W16_48x64_avx2 sp, 8
+
+INIT_XMM sse2
+cglobal chroma_p2s, 3, 7, 3
+    ; load width and height
+    mov         r3d, r3m
+    mov         r4d, r4m
+    add         r1, r1
+
+    ; load constant
+    mova        m2, [tab_c_n8192]
+
+.loopH:
+
+    xor         r5d, r5d
+.loopW:
+    lea         r6, [r0 + r5 * 2]
+
+    movu        m0, [r6]
+    psllw       m0, (14 - BIT_DEPTH)
+    paddw       m0, m2
+
+    movu        m1, [r6 + r1]
+    psllw       m1, (14 - BIT_DEPTH)
+    paddw       m1, m2
+
+    add         r5d, 8
+    cmp         r5d, r3d
+    lea         r6, [r2 + r5 * 2]
+    jg          .width4
+    movu        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
+    movu        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
+    je          .nextH
+    jmp         .loopW
+
+.width4:
+    test        r3d, 4
+    jz          .width2
+    test        r3d, 2
+    movh        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
+    movh        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
+    lea         r6, [r6 + 8]
+    pshufd      m0, m0, 2
+    pshufd      m1, m1, 2
+    jz          .nextH
+
+.width2:
+    movd        [r6 + FENC_STRIDE / 2 * 0 - 16], m0
+    movd        [r6 + FENC_STRIDE / 2 * 2 - 16], m1
+
+.nextH:
+    lea         r0, [r0 + r1 * 2]
+    add         r2, FENC_STRIDE / 2 * 4
+
+    sub         r4d, 2
+    jnz         .loopH
+    RET
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal filterPixelToShort_48x64, 3, 7, 4
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, 16
+
+    ; load constant
+    mova       m3, [pw_2000]
+
+.loop:
+    movu       m0, [r0]
+    movu       m1, [r0 + 32]
+    movu       m2, [r0 + 64]
+    psllw      m0, (14 - BIT_DEPTH)
+    psllw      m1, (14 - BIT_DEPTH)
+    psllw      m2, (14 - BIT_DEPTH)
+    psubw      m0, m3
+    psubw      m1, m3
+    psubw      m2, m3
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 0 + 32], m1
+    movu       [r2 + r3 * 0 + 64], m2
+
+    movu       m0, [r0 + r1]
+    movu       m1, [r0 + r1 + 32]
+    movu       m2, [r0 + r1 + 64]
+    psllw      m0, (14 - BIT_DEPTH)
+    psllw      m1, (14 - BIT_DEPTH)
+    psllw      m2, (14 - BIT_DEPTH)
+    psubw      m0, m3
+    psubw      m1, m3
+    psubw      m2, m3
+    movu       [r2 + r3 * 1], m0
+    movu       [r2 + r3 * 1 + 32], m1
+    movu       [r2 + r3 * 1 + 64], m2
+
+    movu       m0, [r0 + r1 * 2]
+    movu       m1, [r0 + r1 * 2 + 32]
+    movu       m2, [r0 + r1 * 2 + 64]
+    psllw      m0, (14 - BIT_DEPTH)
+    psllw      m1, (14 - BIT_DEPTH)
+    psllw      m2, (14 - BIT_DEPTH)
+    psubw      m0, m3
+    psubw      m1, m3
+    psubw      m2, m3
+    movu       [r2 + r3 * 2], m0
+    movu       [r2 + r3 * 2 + 32], m1
+    movu       [r2 + r3 * 2 + 64], m2
+
+    movu       m0, [r0 + r5]
+    movu       m1, [r0 + r5 + 32]
+    movu       m2, [r0 + r5 + 64]
+    psllw      m0, (14 - BIT_DEPTH)
+    psllw      m1, (14 - BIT_DEPTH)
+    psllw      m2, (14 - BIT_DEPTH)
+    psubw      m0, m3
+    psubw      m1, m3
+    psubw      m2, m3
+    movu       [r2 + r4], m0
+    movu       [r2 + r4 + 32], m1
+    movu       [r2 + r4 + 64], m2
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+
+   %macro FILTER_VER_CHROMA_AVX2_8xN 2
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_8x%2, 4, 9, 15
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    add             r3d, r3d
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    vbroadcasti128  m14, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m14, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m14, [INTERP_OFFSET_PS]
+%endif
+    lea             r6, [r3 * 3]
+    lea             r7, [r1 * 4]
+    mov             r8d, %2 / 16
+.loopH:
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm7, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddwd         m7, m5, [r5 + 1 * mmsize]
+    paddd           m3, m7
+    pmaddwd         m5, [r5]
+
+    movu            xm7, [r0 + r4]                  ; m7 = row 7
+    punpckhwd       xm8, xm6, xm7
+    punpcklwd       xm6, xm7
+    vinserti128     m6, m6, xm8, 1
+    pmaddwd         m8, m6, [r5 + 1 * mmsize]
+    paddd           m4, m8
+    pmaddwd         m6, [r5]
+
+    lea             r0, [r0 + r1 * 4]
+    movu            xm8, [r0]                       ; m8 = row 8
+    punpckhwd       xm9, xm7, xm8
+    punpcklwd       xm7, xm8
+    vinserti128     m7, m7, xm9, 1
+    pmaddwd         m9, m7, [r5 + 1 * mmsize]
+    paddd           m5, m9
+    pmaddwd         m7, [r5]
+
+
+    movu            xm9, [r0 + r1]                  ; m9 = row 9
+    punpckhwd       xm10, xm8, xm9
+    punpcklwd       xm8, xm9
+    vinserti128     m8, m8, xm10, 1
+    pmaddwd         m10, m8, [r5 + 1 * mmsize]
+    paddd           m6, m10
+    pmaddwd         m8, [r5]
+
+
+    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
+    punpckhwd       xm11, xm9, xm10
+    punpcklwd       xm9, xm10
+    vinserti128     m9, m9, xm11, 1
+    pmaddwd         m11, m9, [r5 + 1 * mmsize]
+    paddd           m7, m11
+    pmaddwd         m9, [r5]
+
+    movu            xm11, [r0 + r4]                 ; m11 = row 11
+    punpckhwd       xm12, xm10, xm11
+    punpcklwd       xm10, xm11
+    vinserti128     m10, m10, xm12, 1
+    pmaddwd         m12, m10, [r5 + 1 * mmsize]
+    paddd           m8, m12
+    pmaddwd         m10, [r5]
+
+    lea             r0, [r0 + r1 * 4]
+    movu            xm12, [r0]                      ; m12 = row 12
+    punpckhwd       xm13, xm11, xm12
+    punpcklwd       xm11, xm12
+    vinserti128     m11, m11, xm13, 1
+    pmaddwd         m13, m11, [r5 + 1 * mmsize]
+    paddd           m9, m13
+    pmaddwd         m11, [r5]
+
+%ifidn %1,ss
+    psrad           m0, 6
+    psrad           m1, 6
+    psrad           m2, 6
+    psrad           m3, 6
+    psrad           m4, 6
+    psrad           m5, 6
+%else
+    paddd           m0, m14
+    paddd           m1, m14
+    paddd           m2, m14
+    paddd           m3, m14
+    paddd           m4, m14
+    paddd           m5, m14
+%ifidn %1,pp
+    psrad           m0, 6
+    psrad           m1, 6
+    psrad           m2, 6
+    psrad           m3, 6
+    psrad           m4, 6
+    psrad           m5, 6
+%elifidn %1, sp
+    psrad           m0, INTERP_SHIFT_SP
+    psrad           m1, INTERP_SHIFT_SP
+    psrad           m2, INTERP_SHIFT_SP
+    psrad           m3, INTERP_SHIFT_SP
+    psrad           m4, INTERP_SHIFT_SP
+    psrad           m5, INTERP_SHIFT_SP
+%else
+    psrad           m0, INTERP_SHIFT_PS
+    psrad           m1, INTERP_SHIFT_PS
+    psrad           m2, INTERP_SHIFT_PS
+    psrad           m3, INTERP_SHIFT_PS
+    psrad           m4, INTERP_SHIFT_PS
+    psrad           m5, INTERP_SHIFT_PS
+%endif
+%endif
+
+    packssdw        m0, m1
+    packssdw        m2, m3
+    packssdw        m4, m5
+    vpermq          m0, m0, q3120
+    vpermq          m2, m2, q3120
+    vpermq          m4, m4, q3120
+    pxor            m5, m5
+    mova            m3, [pw_pixel_max]
+%ifidn %1,pp
+    CLIPW           m0, m5, m3
+    CLIPW           m2, m5, m3
+    CLIPW           m4, m5, m3
+%elifidn %1, sp
+    CLIPW           m0, m5, m3
+    CLIPW           m2, m5, m3
+    CLIPW           m4, m5, m3
+%endif
+
+    vextracti128    xm1, m0, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    vextracti128    xm1, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm1
+    lea             r2, [r2 + r3 * 4]
+    vextracti128    xm1, m4, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm1
+
+    movu            xm13, [r0 + r1]                 ; m13 = row 13
+    punpckhwd       xm0, xm12, xm13
+    punpcklwd       xm12, xm13
+    vinserti128     m12, m12, xm0, 1
+    pmaddwd         m0, m12, [r5 + 1 * mmsize]
+    paddd           m10, m0
+    pmaddwd         m12, [r5]
+
+    movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
+    punpckhwd       xm1, xm13, xm0
+    punpcklwd       xm13, xm0
+    vinserti128     m13, m13, xm1, 1
+    pmaddwd         m1, m13, [r5 + 1 * mmsize]
+    paddd           m11, m1
+    pmaddwd         m13, [r5]
+
+%ifidn %1,ss
+    psrad           m6, 6
+    psrad           m7, 6
+%else
+    paddd           m6, m14
+    paddd           m7, m14
+%ifidn %1,pp
+    psrad           m6, 6
+    psrad           m7, 6
+%elifidn %1, sp
+    psrad           m6, INTERP_SHIFT_SP
+    psrad           m7, INTERP_SHIFT_SP
+%else
+    psrad           m6, INTERP_SHIFT_PS
+    psrad           m7, INTERP_SHIFT_PS
+%endif
+%endif
+
+    packssdw        m6, m7
+    vpermq          m6, m6, q3120
+%ifidn %1,pp
+    CLIPW           m6, m5, m3
+%elifidn %1, sp
+    CLIPW           m6, m5, m3
+%endif
+    vextracti128    xm7, m6, 1
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r6], xm7
+
+    movu            xm1, [r0 + r4]                  ; m1 = row 15
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m2, m0, [r5 + 1 * mmsize]
+    paddd           m12, m2
+    pmaddwd         m0, [r5]
+
+    lea             r0, [r0 + r1 * 4]
+    movu            xm2, [r0]                       ; m2 = row 16
+    punpckhwd       xm6, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm6, 1
+    pmaddwd         m6, m1, [r5 + 1 * mmsize]
+    paddd           m13, m6
+    pmaddwd         m1, [r5]
+
+    movu            xm6, [r0 + r1]                  ; m6 = row 17
+    punpckhwd       xm4, xm2, xm6
+    punpcklwd       xm2, xm6
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+
+    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
+    punpckhwd       xm2, xm6, xm4
+    punpcklwd       xm6, xm4
+    vinserti128     m6, m6, xm2, 1
+    pmaddwd         m6, [r5 + 1 * mmsize]
+    paddd           m1, m6
+
+%ifidn %1,ss
+    psrad           m8, 6
+    psrad           m9, 6
+    psrad           m10, 6
+    psrad           m11, 6
+    psrad           m12, 6
+    psrad           m13, 6
+    psrad           m0, 6
+    psrad           m1, 6
+%else
+    paddd           m8, m14
+    paddd           m9, m14
+    paddd           m10, m14
+    paddd           m11, m14
+    paddd           m12, m14
+    paddd           m13, m14
+    paddd           m0, m14
+    paddd           m1, m14
+%ifidn %1,pp
+    psrad           m8, 6
+    psrad           m9, 6
+    psrad           m10, 6
+    psrad           m11, 6
+    psrad           m12, 6
+    psrad           m13, 6
+    psrad           m0, 6
+    psrad           m1, 6
+%elifidn %1, sp
+    psrad           m8, INTERP_SHIFT_SP
+    psrad           m9, INTERP_SHIFT_SP
+    psrad           m10, INTERP_SHIFT_SP
+    psrad           m11, INTERP_SHIFT_SP
+    psrad           m12, INTERP_SHIFT_SP
+    psrad           m13, INTERP_SHIFT_SP
+    psrad           m0, INTERP_SHIFT_SP
+    psrad           m1, INTERP_SHIFT_SP
+%else
+    psrad           m8, INTERP_SHIFT_PS
+    psrad           m9, INTERP_SHIFT_PS
+    psrad           m10, INTERP_SHIFT_PS
+    psrad           m11, INTERP_SHIFT_PS
+    psrad           m12, INTERP_SHIFT_PS
+    psrad           m13, INTERP_SHIFT_PS
+    psrad           m0, INTERP_SHIFT_PS
+    psrad           m1, INTERP_SHIFT_PS
+%endif
+%endif
+
+    packssdw        m8, m9
+    packssdw        m10, m11
+    packssdw        m12, m13
+    packssdw        m0, m1
+    vpermq          m8, m8, q3120
+    vpermq          m10, m10, q3120
+    vpermq          m12, m12, q3120
+    vpermq          m0, m0, q3120
+%ifidn %1,pp
+    CLIPW           m8, m5, m3
+    CLIPW           m10, m5, m3
+    CLIPW           m12, m5, m3
+    CLIPW           m0, m5, m3
+%elifidn %1, sp
+    CLIPW           m8, m5, m3
+    CLIPW           m10, m5, m3
+    CLIPW           m12, m5, m3
+    CLIPW           m0, m5, m3
+%endif
+    vextracti128    xm9, m8, 1
+    vextracti128    xm11, m10, 1
+    vextracti128    xm13, m12, 1
+    vextracti128    xm1, m0, 1
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm8
+    movu            [r2 + r3], xm9
+    movu            [r2 + r3 * 2], xm10
+    movu            [r2 + r6], xm11
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm12
+    movu            [r2 + r3], xm13
+    movu            [r2 + r3 * 2], xm0
+    movu            [r2 + r6], xm1
+    lea             r2, [r2 + r3 * 4]
+    dec             r8d
+    jnz             .loopH
+    RET
+%endif
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_8xN pp, 16
+FILTER_VER_CHROMA_AVX2_8xN ps, 16
+FILTER_VER_CHROMA_AVX2_8xN ss, 16
+FILTER_VER_CHROMA_AVX2_8xN sp, 16
+FILTER_VER_CHROMA_AVX2_8xN pp, 32
+FILTER_VER_CHROMA_AVX2_8xN ps, 32
+FILTER_VER_CHROMA_AVX2_8xN sp, 32
+FILTER_VER_CHROMA_AVX2_8xN ss, 32
+FILTER_VER_CHROMA_AVX2_8xN pp, 64
+FILTER_VER_CHROMA_AVX2_8xN ps, 64
+FILTER_VER_CHROMA_AVX2_8xN sp, 64
+FILTER_VER_CHROMA_AVX2_8xN ss, 64
+
+%macro PROCESS_CHROMA_AVX2_8x2 3
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m2, m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m3, m3, [r5 + 1 * mmsize]
+    paddd           m1, m3
+
+%ifnidn %1,ss
+    paddd           m0, m7
+    paddd           m1, m7
+%endif
+    psrad           m0, %3
+    psrad           m1, %3
+
+    packssdw        m0, m1
+    vpermq          m0, m0, q3120
+    pxor            m4, m4
+
+%if %2
+    CLIPW           m0, m4, [pw_pixel_max]
+%endif
+    vextracti128    xm1, m0, 1
+%endmacro
+
+
+%macro FILTER_VER_CHROMA_AVX2_8x2 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x2, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    add             r3d, r3d
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    vbroadcasti128  m7, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+
+    PROCESS_CHROMA_AVX2_8x2 %1, %2, %3
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_8x2 pp, 1, 6
+FILTER_VER_CHROMA_AVX2_8x2 ps, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_8x2 sp, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_8x2 ss, 0, 6
+
+%macro FILTER_VER_CHROMA_AVX2_4x2 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x2, 4, 6, 7
+    mov             r4d, r4m
+    add             r1d, r1d
+    add             r3d, r3d
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+
+%ifidn %1,pp
+    vbroadcasti128  m6, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m6, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m6, [INTERP_OFFSET_PS]
+%endif
+
+    movq            xm0, [r0]                       ; row 0
+    movq            xm1, [r0 + r1]                  ; row 1
+    punpcklwd       xm0, xm1
+
+    movq            xm2, [r0 + r1 * 2]              ; row 2
+    punpcklwd       xm1, xm2
+    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
+    pmaddwd         m0, [r5]
+
+    movq            xm3, [r0 + r4]                  ; row 3
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm4, [r0]                       ; row 4
+    punpcklwd       xm3, xm4
+    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    paddd           m0, m5
+
+%ifnidn %1, ss
+    paddd           m0, m6
+%endif
+    psrad           m0, %3
+    packssdw        m0, m0
+    pxor            m1, m1
+
+%if %2
+    CLIPW           m0, m1, [pw_pixel_max]
+%endif
+
+    vextracti128    xm2, m0, 1
+    lea             r4, [r3 * 3]
+    movq            [r2], xm0
+    movq            [r2 + r3], xm2
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_4x2 pp, 1, 6
+FILTER_VER_CHROMA_AVX2_4x2 ps, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_4x2 sp, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_4x2 ss, 0, 6
+
+%macro FILTER_VER_CHROMA_AVX2_4x4 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x4, 4, 6, 7
+    mov             r4d, r4m
+    add             r1d, r1d
+    add             r3d, r3d
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+
+%ifidn %1,pp
+   vbroadcasti128  m6, [pd_32]
+%elifidn %1, sp
+   vbroadcasti128  m6, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m6, [INTERP_OFFSET_PS]
+%endif
+    movq            xm0, [r0]                       ; row 0
+    movq            xm1, [r0 + r1]                  ; row 1
+    punpcklwd       xm0, xm1
+
+    movq            xm2, [r0 + r1 * 2]              ; row 2
+    punpcklwd       xm1, xm2
+    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
+    pmaddwd         m0, [r5]
+
+    movq            xm3, [r0 + r4]                  ; row 3
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm4, [r0]                       ; row 4
+    punpcklwd       xm3, xm4
+    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m5
+
+    movq            xm3, [r0 + r1]                  ; row 5
+    punpcklwd       xm4, xm3
+    movq            xm1, [r0 + r1 * 2]              ; row 6
+    punpcklwd       xm3, xm1
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
+    pmaddwd         m4, [r5 + 1 * mmsize]
+    paddd           m2, m4
+
+%ifnidn %1,ss
+    paddd           m0, m6
+    paddd           m2, m6
+%endif
+    psrad           m0, %3
+    psrad           m2, %3
+
+    packssdw        m0, m2
+    pxor            m1, m1
+%if %2
+    CLIPW           m0, m1, [pw_pixel_max]
+%endif
+
+    vextracti128    xm2, m0, 1
+    lea             r4, [r3 * 3]
+    movq            [r2], xm0
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r4], xm2
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_4x4 pp, 1, 6
+FILTER_VER_CHROMA_AVX2_4x4 ps, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_4x4 sp, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_4x4 ss, 0, 6
+
+
+%macro FILTER_VER_CHROMA_AVX2_4x8 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x8, 4, 7, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    add             r3d, r3d
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+
+%ifidn %1,pp
+    vbroadcasti128  m7, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+    lea             r6, [r3 * 3]
+
+    movq            xm0, [r0]                       ; row 0
+    movq            xm1, [r0 + r1]                  ; row 1
+    punpcklwd       xm0, xm1
+    movq            xm2, [r0 + r1 * 2]              ; row 2
+    punpcklwd       xm1, xm2
+    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
+    pmaddwd         m0, [r5]
+
+    movq            xm3, [r0 + r4]                  ; row 3
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm4, [r0]                       ; row 4
+    punpcklwd       xm3, xm4
+    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m5
+
+    movq            xm3, [r0 + r1]                  ; row 5
+    punpcklwd       xm4, xm3
+    movq            xm1, [r0 + r1 * 2]              ; row 6
+    punpcklwd       xm3, xm1
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
+    pmaddwd         m5, m4, [r5 + 1 * mmsize]
+    paddd           m2, m5
+    pmaddwd         m4, [r5]
+
+    movq            xm3, [r0 + r4]                  ; row 7
+    punpcklwd       xm1, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm6, [r0]                       ; row 8
+    punpcklwd       xm3, xm6
+    vinserti128     m1, m1, xm3, 1                  ; m1 = [8 7 7 6]
+    pmaddwd         m5, m1, [r5 + 1 * mmsize]
+    paddd           m4, m5
+    pmaddwd         m1, [r5]
+
+    movq            xm3, [r0 + r1]                  ; row 9
+    punpcklwd       xm6, xm3
+    movq            xm5, [r0 + 2 * r1]              ; row 10
+    punpcklwd       xm3, xm5
+    vinserti128     m6, m6, xm3, 1                  ; m6 = [A 9 9 8]
+    pmaddwd         m6, [r5 + 1 * mmsize]
+    paddd           m1, m6
+%ifnidn %1,ss
+    paddd           m0, m7
+    paddd           m2, m7
+%endif
+    psrad           m0, %3
+    psrad           m2, %3
+    packssdw        m0, m2
+    pxor            m6, m6
+    mova            m3, [pw_pixel_max]
+%if %2
+    CLIPW           m0, m6, m3
+%endif
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r6], xm2
+%ifnidn %1,ss
+    paddd           m4, m7
+    paddd           m1, m7
+%endif
+    psrad           m4, %3
+    psrad           m1, %3
+    packssdw        m4, m1
+%if %2
+    CLIPW           m4, m6, m3
+%endif
+    vextracti128    xm1, m4, 1
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm4
+    movq            [r2 + r3], xm1
+    movhps          [r2 + r3 * 2], xm4
+    movhps          [r2 + r6], xm1
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_4x8 pp, 1, 6
+FILTER_VER_CHROMA_AVX2_4x8 ps, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_4x8 sp, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_4x8 ss, 0 , 6
+
+%macro PROCESS_LUMA_AVX2_W4_16R_4TAP 3
+    movq            xm0, [r0]                       ; row 0
+    movq            xm1, [r0 + r1]                  ; row 1
+    punpcklwd       xm0, xm1
+    movq            xm2, [r0 + r1 * 2]              ; row 2
+    punpcklwd       xm1, xm2
+    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
+    pmaddwd         m0, [r5]
+    movq            xm3, [r0 + r4]                  ; row 3
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm4, [r0]                       ; row 4
+    punpcklwd       xm3, xm4
+    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m5
+    movq            xm3, [r0 + r1]                  ; row 5
+    punpcklwd       xm4, xm3
+    movq            xm1, [r0 + r1 * 2]              ; row 6
+    punpcklwd       xm3, xm1
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
+    pmaddwd         m5, m4, [r5 + 1 * mmsize]
+    paddd           m2, m5
+    pmaddwd         m4, [r5]
+    movq            xm3, [r0 + r4]                  ; row 7
+    punpcklwd       xm1, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm6, [r0]                       ; row 8
+    punpcklwd       xm3, xm6
+    vinserti128     m1, m1, xm3, 1                  ; m1 = [8 7 7 6]
+    pmaddwd         m5, m1, [r5 + 1 * mmsize]
+    paddd           m4, m5
+    pmaddwd         m1, [r5]
+    movq            xm3, [r0 + r1]                  ; row 9
+    punpcklwd       xm6, xm3
+    movq            xm5, [r0 + 2 * r1]              ; row 10
+    punpcklwd       xm3, xm5
+    vinserti128     m6, m6, xm3, 1                  ; m6 = [10 9 9 8]
+    pmaddwd         m3, m6, [r5 + 1 * mmsize]
+    paddd           m1, m3
+    pmaddwd         m6, [r5]
+%ifnidn %1,ss
+    paddd           m0, m7
+    paddd           m2, m7
+%endif
+    psrad           m0, %3
+    psrad           m2, %3
+    packssdw        m0, m2
+    pxor            m3, m3
+%if %2
+    CLIPW           m0, m3, [pw_pixel_max]
+%endif
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r6], xm2
+    movq            xm2, [r0 + r4]                  ;row 11
+    punpcklwd       xm5, xm2
+    lea             r0, [r0 + 4 * r1]
+    movq            xm0, [r0]                       ; row 12
+    punpcklwd       xm2, xm0
+    vinserti128     m5, m5, xm2, 1                  ; m5 = [12 11 11 10]
+    pmaddwd         m2, m5, [r5 + 1 * mmsize]
+    paddd           m6, m2
+    pmaddwd         m5, [r5]
+    movq            xm2, [r0 + r1]                  ; row 13
+    punpcklwd       xm0, xm2
+    movq            xm3, [r0 + 2 * r1]              ; row 14
+    punpcklwd       xm2, xm3
+    vinserti128     m0, m0, xm2, 1                  ; m0 = [14 13 13 12]
+    pmaddwd         m2, m0, [r5 + 1 * mmsize]
+    paddd           m5, m2
+    pmaddwd         m0, [r5]
+%ifnidn %1,ss
+    paddd           m4, m7
+    paddd           m1, m7
+%endif
+    psrad           m4, %3
+    psrad           m1, %3
+    packssdw        m4, m1
+    pxor            m2, m2
+%if %2
+    CLIPW           m4, m2, [pw_pixel_max]
+%endif
+
+    vextracti128    xm1, m4, 1
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm4
+    movq            [r2 + r3], xm1
+    movhps          [r2 + r3 * 2], xm4
+    movhps          [r2 + r6], xm1
+    movq            xm4, [r0 + r4]                  ; row 15
+    punpcklwd       xm3, xm4
+    lea             r0, [r0 + 4 * r1]
+    movq            xm1, [r0]                       ; row 16
+    punpcklwd       xm4, xm1
+    vinserti128     m3, m3, xm4, 1                  ; m3 = [16 15 15 14]
+    pmaddwd         m4, m3, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m3, [r5]
+    movq            xm4, [r0 + r1]                  ; row 17
+    punpcklwd       xm1, xm4
+    movq            xm2, [r0 + 2 * r1]              ; row 18
+    punpcklwd       xm4, xm2
+    vinserti128     m1, m1, xm4, 1                  ; m1 = [18 17 17 16]
+    pmaddwd         m1, [r5 + 1 * mmsize]
+    paddd           m3, m1
+
+%ifnidn %1,ss
+    paddd           m6, m7
+    paddd           m5, m7
+%endif
+    psrad           m6, %3
+    psrad           m5, %3
+    packssdw        m6, m5
+    pxor            m1, m1
+%if %2
+    CLIPW           m6, m1, [pw_pixel_max]
+%endif
+    vextracti128    xm5, m6, 1
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm6
+    movq            [r2 + r3], xm5
+    movhps          [r2 + r3 * 2], xm6
+    movhps          [r2 + r6], xm5
+%ifnidn %1,ss
+    paddd           m0, m7
+    paddd           m3, m7
+%endif
+    psrad           m0, %3
+    psrad           m3, %3
+    packssdw        m0, m3
+%if %2
+    CLIPW           m0, m1, [pw_pixel_max]
+%endif
+    vextracti128    xm3, m0, 1
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm0
+    movq            [r2 + r3], xm3
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r6], xm3
+%endmacro
+
+%macro FILTER_VER_CHROMA_AVX2_4xN 4
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x%2, 4, 8, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    add             r3d, r3d
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+    mov             r7d, %2 / 16
+%ifidn %1,pp
+    vbroadcasti128  m7, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+    lea             r6, [r3 * 3]
+.loopH:
+    PROCESS_LUMA_AVX2_W4_16R_4TAP %1, %3, %4
+    lea             r2, [r2 + r3 * 4]
+    dec             r7d
+    jnz             .loopH
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_4xN pp, 16, 1, 6
+FILTER_VER_CHROMA_AVX2_4xN ps, 16, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_4xN sp, 16, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_4xN ss, 16, 0, 6
+FILTER_VER_CHROMA_AVX2_4xN pp, 32, 1, 6
+FILTER_VER_CHROMA_AVX2_4xN ps, 32, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_4xN sp, 32, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_4xN ss, 32, 0, 6
+
+%macro FILTER_VER_CHROMA_AVX2_8x8 3
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_8x8, 4, 6, 12
+    mov             r4d, r4m
+    add             r1d, r1d
+    add             r3d, r3d
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+
+%ifidn %1,pp
+    vbroadcasti128  m11, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m11, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m11, [INTERP_OFFSET_PS]
+%endif
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m4                          ; res row0 done(0,1,2,3)
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    pmaddwd         m3, [r5]
+    paddd           m1, m5                          ;res row1 done(1, 2, 3, 4)
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    pmaddwd         m4, [r5]
+    paddd           m2, m6                          ;res row2 done(2,3,4,5)
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm7, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddwd         m7, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m7                          ;res row3 done(3,4,5,6)
+    movu            xm7, [r0 + r4]                  ; m7 = row 7
+    punpckhwd       xm8, xm6, xm7
+    punpcklwd       xm6, xm7
+    vinserti128     m6, m6, xm8, 1
+    pmaddwd         m8, m6, [r5 + 1 * mmsize]
+    pmaddwd         m6, [r5]
+    paddd           m4, m8                          ;res row4 done(4,5,6,7)
+    lea             r0, [r0 + r1 * 4]
+    movu            xm8, [r0]                       ; m8 = row 8
+    punpckhwd       xm9, xm7, xm8
+    punpcklwd       xm7, xm8
+    vinserti128     m7, m7, xm9, 1
+    pmaddwd         m9, m7, [r5 + 1 * mmsize]
+    pmaddwd         m7, [r5]
+    paddd           m5, m9                          ;res row5 done(5,6,7,8)
+    movu            xm9, [r0 + r1]                  ; m9 = row 9
+    punpckhwd       xm10, xm8, xm9
+    punpcklwd       xm8, xm9
+    vinserti128     m8, m8, xm10, 1
+    pmaddwd         m8, [r5 + 1 * mmsize]
+    paddd           m6, m8                          ;res row6 done(6,7,8,9)
+    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
+    punpckhwd       xm8, xm9, xm10
+    punpcklwd       xm9, xm10
+    vinserti128     m9, m9, xm8, 1
+    pmaddwd         m9, [r5 + 1 * mmsize]
+    paddd           m7, m9                          ;res row7 done 7,8,9,10
+    lea             r4, [r3 * 3]
+%ifnidn %1,ss
+    paddd           m0, m11
+    paddd           m1, m11
+    paddd           m2, m11
+    paddd           m3, m11
+%endif
+    psrad           m0, %3
+    psrad           m1, %3
+    psrad           m2, %3
+    psrad           m3, %3
+    packssdw        m0, m1
+    packssdw        m2, m3
+    vpermq          m0, m0, q3120
+    vpermq          m2, m2, q3120
+    pxor            m1, m1
+    mova            m3, [pw_pixel_max]
+%if %2
+    CLIPW           m0, m1, m3
+    CLIPW           m2, m1, m3
+%endif
+    vextracti128    xm9, m0, 1
+    vextracti128    xm8, m2, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm9
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r4], xm8
+%ifnidn %1,ss
+    paddd           m4, m11
+    paddd           m5, m11
+    paddd           m6, m11
+    paddd           m7, m11
+%endif
+    psrad           m4, %3
+    psrad           m5, %3
+    psrad           m6, %3
+    psrad           m7, %3
+    packssdw        m4, m5
+    packssdw        m6, m7
+    vpermq          m4, m4, q3120
+    vpermq          m6, m6, q3120
+%if %2
+    CLIPW           m4, m1, m3
+    CLIPW           m6, m1, m3
+%endif
+    vextracti128    xm5, m4, 1
+    vextracti128    xm7, m6, 1
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm4
+    movu            [r2 + r3], xm5
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r4], xm7
+    RET
+%endif
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_8x8 pp, 1, 6
+FILTER_VER_CHROMA_AVX2_8x8 ps, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_8x8 sp, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_8x8 ss, 0, 6
+
+%macro FILTER_VER_CHROMA_AVX2_8x6 3
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_8x6, 4, 6, 12
+    mov             r4d, r4m
+    add             r1d, r1d
+    add             r3d, r3d
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+
+%ifidn %1,pp
+    vbroadcasti128  m11, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m11, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m11, [INTERP_OFFSET_PS]
+%endif
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m4                          ; r0 done(0,1,2,3)
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    pmaddwd         m3, [r5]
+    paddd           m1, m5                          ;r1 done(1, 2, 3, 4)
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    pmaddwd         m4, [r5]
+    paddd           m2, m6                          ;r2 done(2,3,4,5)
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm7, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddwd         m7, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m7                          ;r3 done(3,4,5,6)
+    movu            xm7, [r0 + r4]                  ; m7 = row 7
+    punpckhwd       xm8, xm6, xm7
+    punpcklwd       xm6, xm7
+    vinserti128     m6, m6, xm8, 1
+    pmaddwd         m8, m6, [r5 + 1 * mmsize]
+    paddd           m4, m8                          ;r4 done(4,5,6,7)
+    lea             r0, [r0 + r1 * 4]
+    movu            xm8, [r0]                       ; m8 = row 8
+    punpckhwd       xm9, xm7, xm8
+    punpcklwd       xm7, xm8
+    vinserti128     m7, m7, xm9, 1
+    pmaddwd         m7, m7, [r5 + 1 * mmsize]
+    paddd           m5, m7                          ;r5 done(5,6,7,8)
+    lea             r4, [r3 * 3]
+%ifnidn %1,ss
+    paddd           m0, m11
+    paddd           m1, m11
+    paddd           m2, m11
+    paddd           m3, m11
+%endif
+    psrad           m0, %3
+    psrad           m1, %3
+    psrad           m2, %3
+    psrad           m3, %3
+    packssdw        m0, m1
+    packssdw        m2, m3
+    vpermq          m0, m0, q3120
+    vpermq          m2, m2, q3120
+    pxor            m10, m10
+    mova            m9, [pw_pixel_max]
+%if %2
+    CLIPW           m0, m10, m9
+    CLIPW           m2, m10, m9
+%endif
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r4], xm3
+%ifnidn %1,ss
+    paddd           m4, m11
+    paddd           m5, m11
+%endif
+    psrad           m4, %3
+    psrad           m5, %3
+    packssdw        m4, m5
+    vpermq          m4, m4, 11011000b
+%if %2
+    CLIPW           m4, m10, m9
+%endif
+    vextracti128    xm5, m4, 1
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm4
+    movu            [r2 + r3], xm5
+    RET
+%endif
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_8x6 pp, 1, 6
+FILTER_VER_CHROMA_AVX2_8x6 ps, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_8x6 sp, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_8x6 ss, 0, 6
+
+%macro PROCESS_CHROMA_AVX2 3
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m4, [r5 + 1 * mmsize]
+    paddd           m2, m4
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm4, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm4, 1
+    pmaddwd         m5, [r5 + 1 * mmsize]
+    paddd           m3, m5
+%ifnidn %1,ss
+    paddd           m0, m7
+    paddd           m1, m7
+    paddd           m2, m7
+    paddd           m3, m7
+%endif
+    psrad           m0, %3
+    psrad           m1, %3
+    psrad           m2, %3
+    psrad           m3, %3
+    packssdw        m0, m1
+    packssdw        m2, m3
+    vpermq          m0, m0, q3120
+    vpermq          m2, m2, q3120
+    pxor            m4, m4
+%if %2
+    CLIPW           m0, m4, [pw_pixel_max]
+    CLIPW           m2, m4, [pw_pixel_max]
+%endif
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+%endmacro
+
+
+%macro FILTER_VER_CHROMA_AVX2_8x4 3
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x4, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    add             r3d, r3d
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    vbroadcasti128  m7, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m7, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m7, [INTERP_OFFSET_PS]
+%endif
+    PROCESS_CHROMA_AVX2 %1, %2, %3
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    lea             r4, [r3 * 3]
+    movu            [r2 + r4], xm3
+    RET
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_8x4 pp, 1, 6
+FILTER_VER_CHROMA_AVX2_8x4 ps, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_8x4 sp, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_8x4 ss, 0, 6
+
+%macro FILTER_VER_CHROMA_AVX2_8x12 3
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_8x12, 4, 7, 15
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    add             r3d, r3d
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    vbroadcasti128  m14, [pd_32]
+%elifidn %1, sp
+    vbroadcasti128  m14, [INTERP_OFFSET_SP]
+%else
+    vbroadcasti128  m14, [INTERP_OFFSET_PS]
+%endif
+    lea             r6, [r3 * 3]
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm7, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddwd         m7, m5, [r5 + 1 * mmsize]
+    paddd           m3, m7
+    pmaddwd         m5, [r5]
+    movu            xm7, [r0 + r4]                  ; m7 = row 7
+    punpckhwd       xm8, xm6, xm7
+    punpcklwd       xm6, xm7
+    vinserti128     m6, m6, xm8, 1
+    pmaddwd         m8, m6, [r5 + 1 * mmsize]
+    paddd           m4, m8
+    pmaddwd         m6, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm8, [r0]                       ; m8 = row 8
+    punpckhwd       xm9, xm7, xm8
+    punpcklwd       xm7, xm8
+    vinserti128     m7, m7, xm9, 1
+    pmaddwd         m9, m7, [r5 + 1 * mmsize]
+    paddd           m5, m9
+    pmaddwd         m7, [r5]
+    movu            xm9, [r0 + r1]                  ; m9 = row 9
+    punpckhwd       xm10, xm8, xm9
+    punpcklwd       xm8, xm9
+    vinserti128     m8, m8, xm10, 1
+    pmaddwd         m10, m8, [r5 + 1 * mmsize]
+    paddd           m6, m10
+    pmaddwd         m8, [r5]
+    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
+    punpckhwd       xm11, xm9, xm10
+    punpcklwd       xm9, xm10
+    vinserti128     m9, m9, xm11, 1
+    pmaddwd         m11, m9, [r5 + 1 * mmsize]
+    paddd           m7, m11
+    pmaddwd         m9, [r5]
+    movu            xm11, [r0 + r4]                 ; m11 = row 11
+    punpckhwd       xm12, xm10, xm11
+    punpcklwd       xm10, xm11
+    vinserti128     m10, m10, xm12, 1
+    pmaddwd         m12, m10, [r5 + 1 * mmsize]
+    paddd           m8, m12
+    pmaddwd         m10, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm12, [r0]                      ; m12 = row 12
+    punpckhwd       xm13, xm11, xm12
+    punpcklwd       xm11, xm12
+    vinserti128     m11, m11, xm13, 1
+    pmaddwd         m13, m11, [r5 + 1 * mmsize]
+    paddd           m9, m13
+    pmaddwd         m11, [r5]
+%ifnidn %1,ss
+    paddd           m0, m14
+    paddd           m1, m14
+    paddd           m2, m14
+    paddd           m3, m14
+    paddd           m4, m14
+    paddd           m5, m14
+%endif
+    psrad           m0, %3
+    psrad           m1, %3
+    psrad           m2, %3
+    psrad           m3, %3
+    psrad           m4, %3
+    psrad           m5, %3
+    packssdw        m0, m1
+    packssdw        m2, m3
+    packssdw        m4, m5
+    vpermq          m0, m0, q3120
+    vpermq          m2, m2, q3120
+    vpermq          m4, m4, q3120
+    pxor            m5, m5
+    mova            m3, [pw_pixel_max]
+%if %2
+    CLIPW           m0, m5, m3
+    CLIPW           m2, m5, m3
+    CLIPW           m4, m5, m3
+%endif
+    vextracti128    xm1, m0, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    vextracti128    xm1, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm1
+    lea             r2, [r2 + r3 * 4]
+    vextracti128    xm1, m4, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm1
+    movu            xm13, [r0 + r1]                 ; m13 = row 13
+    punpckhwd       xm0, xm12, xm13
+    punpcklwd       xm12, xm13
+    vinserti128     m12, m12, xm0, 1
+    pmaddwd         m12, m12, [r5 + 1 * mmsize]
+    paddd           m10, m12
+    movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
+    punpckhwd       xm1, xm13, xm0
+    punpcklwd       xm13, xm0
+    vinserti128     m13, m13, xm1, 1
+    pmaddwd         m13, m13, [r5 + 1 * mmsize]
+    paddd           m11, m13
+%ifnidn %1,ss
+    paddd           m6, m14
+    paddd           m7, m14
+    paddd           m8, m14
+    paddd           m9, m14
+    paddd           m10, m14
+    paddd           m11, m14
+%endif
+    psrad           m6, %3
+    psrad           m7, %3
+    psrad           m8, %3
+    psrad           m9, %3
+    psrad           m10, %3
+    psrad           m11, %3
+    packssdw        m6, m7
+    packssdw        m8, m9
+    packssdw        m10, m11
+    vpermq          m6, m6, q3120
+    vpermq          m8, m8, q3120
+    vpermq          m10, m10, q3120
+%if %2
+    CLIPW           m6, m5, m3
+    CLIPW           m8, m5, m3
+    CLIPW           m10, m5, m3
+%endif
+    vextracti128    xm7, m6, 1
+    vextracti128    xm9, m8, 1
+    vextracti128    xm11, m10, 1
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r6], xm7
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm8
+    movu            [r2 + r3], xm9
+    movu            [r2 + r3 * 2], xm10
+    movu            [r2 + r6], xm11
+    RET
+%endif
+%endmacro
+
+FILTER_VER_CHROMA_AVX2_8x12 pp, 1, 6
+FILTER_VER_CHROMA_AVX2_8x12 ps, 0, INTERP_SHIFT_PS
+FILTER_VER_CHROMA_AVX2_8x12 sp, 1, INTERP_SHIFT_SP
+FILTER_VER_CHROMA_AVX2_8x12 ss, 0, 6
\ No newline at end of file

x265_2.7.tar.gz/source/common/x86/v4-ipfilter8.asm Added

@@ -0,0 +1,12799 @@
+;*****************************************************************************
+;* Copyright (C) 2013-2017 MulticoreWare, Inc
+;*
+;* Authors: Min Chen <chenm003@163.com>
+;*          Nabajit Deka <nabajit@multicorewareinc.com>
+;*          Praveen Kumar Tiwari <praveen@multicorewareinc.com>
+;*
+;* This program is free software; you can redistribute it and/or modify
+;* it under the terms of the GNU General Public License as published by
+;* the Free Software Foundation; either version 2 of the License, or
+;* (at your option) any later version.
+;*
+;* This program is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+;* GNU General Public License for more details.
+;*
+;* You should have received a copy of the GNU General Public License
+;* along with this program; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
+;*
+;* This program is also available under a commercial proprietary license.
+;* For more information, contact us at license @ x265.com.
+;*****************************************************************************/
+
+%include "x86inc.asm"
+%include "x86util.asm"
+
+SECTION_RODATA 32
+
+const v4_pd_526336, times 8 dd 8192*64+2048
+
+const tab_Vm,    db 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1
+                 db 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3, 2, 3
+
+const tab_Cm,    db 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3
+
+const interp_vert_shuf, times 2 db 0, 2, 1, 3, 2, 4, 3, 5, 4, 6, 5, 7, 6, 8, 7, 9
+                        times 2 db 4, 6, 5, 7, 6, 8, 7, 9, 8, 10, 9, 11, 10, 12, 11, 13
+
+const v4_interp4_vpp_shuf, times 2 db 0, 4, 1, 5, 2, 6, 3, 7, 8, 12, 9, 13, 10, 14, 11, 15
+
+const v4_interp4_vpp_shuf1, dd 0, 1, 1, 2, 2, 3, 3, 4
+                         dd 2, 3, 3, 4, 4, 5, 5, 6
+
+const tab_ChromaCoeff, db  0, 64,  0,  0
+                       db -2, 58, 10, -2
+                       db -4, 54, 16, -2
+                       db -6, 46, 28, -4
+                       db -4, 36, 36, -4
+                       db -4, 28, 46, -6
+                       db -2, 16, 54, -4
+                       db -2, 10, 58, -2
+
+const tabw_ChromaCoeff, dw  0, 64,  0,  0
+                        dw -2, 58, 10, -2
+                        dw -4, 54, 16, -2
+                        dw -6, 46, 28, -4
+                        dw -4, 36, 36, -4
+                        dw -4, 28, 46, -6
+                        dw -2, 16, 54, -4
+                        dw -2, 10, 58, -2
+
+const tab_ChromaCoeffV, times 4 dw 0, 64
+                        times 4 dw 0, 0
+
+                        times 4 dw -2, 58
+                        times 4 dw 10, -2
+
+                        times 4 dw -4, 54
+                        times 4 dw 16, -2
+
+                        times 4 dw -6, 46
+                        times 4 dw 28, -4
+
+                        times 4 dw -4, 36
+                        times 4 dw 36, -4
+
+                        times 4 dw -4, 28
+                        times 4 dw 46, -6
+
+                        times 4 dw -2, 16
+                        times 4 dw 54, -4
+
+                        times 4 dw -2, 10
+                        times 4 dw 58, -2
+
+const tab_ChromaCoeff_V, times 8 db 0, 64
+                         times 8 db 0,  0
+
+                         times 8 db -2, 58
+                         times 8 db 10, -2
+
+                         times 8 db -4, 54
+                         times 8 db 16, -2
+
+                         times 8 db -6, 46
+                         times 8 db 28, -4
+
+                         times 8 db -4, 36
+                         times 8 db 36, -4
+
+                         times 8 db -4, 28
+                         times 8 db 46, -6
+
+                         times 8 db -2, 16
+                         times 8 db 54, -4
+
+                         times 8 db -2, 10
+                         times 8 db 58, -2
+
+const tab_ChromaCoeffVer_32,    times 16 db 0, 64
+                                times 16 db 0, 0
+
+                                times 16 db -2, 58
+                                times 16 db 10, -2
+
+                                times 16 db -4, 54
+                                times 16 db 16, -2
+
+                                times 16 db -6, 46
+                                times 16 db 28, -4
+
+                                times 16 db -4, 36
+                                times 16 db 36, -4
+
+                                times 16 db -4, 28
+                                times 16 db 46, -6
+
+                                times 16 db -2, 16
+                                times 16 db 54, -4
+
+                                times 16 db -2, 10
+                                times 16 db 58, -2
+
+const pw_ChromaCoeffV,  times 8 dw 0, 64
+                        times 8 dw 0, 0
+
+                        times 8 dw -2, 58
+                        times 8 dw 10, -2
+
+                        times 8 dw -4, 54
+                        times 8 dw 16, -2
+
+                        times 8 dw -6, 46
+                        times 8 dw 28, -4
+
+                        times 8 dw -4, 36
+                        times 8 dw 36, -4
+
+                        times 8 dw -4, 28
+                        times 8 dw 46, -6
+
+                        times 8 dw -2, 16
+                        times 8 dw 54, -4
+
+                        times 8 dw -2, 10
+                        times 8 dw 58, -2
+
+const v4_interp8_hps_shuf,     dd 0, 4, 1, 5, 2, 6, 3, 7
+
+SECTION .text
+
+cextern pw_32
+cextern pw_512
+cextern pw_2000
+
+%macro  WORD_TO_DOUBLE 1
+%if ARCH_X86_64
+    punpcklbw   %1,     m8
+%else
+    punpcklbw   %1,     %1
+    psrlw       %1,     8
+%endif
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_2x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W2_H4_sse2 2
+INIT_XMM sse2
+%if ARCH_X86_64
+cglobal interp_4tap_vert_%1_2x%2, 4, 6, 9
+    pxor        m8,        m8
+%else
+cglobal interp_4tap_vert_%1_2x%2, 4, 6, 8
+%endif
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifidn %1,pp
+    mova        m1,        [pw_32]
+%elifidn %1,ps
+    mova        m1,        [pw_2000]
+    add         r3d,       r3d
+%endif
+
+%ifdef PIC
+    lea         r5,        [tabw_ChromaCoeff]
+    movh        m0,        [r5 + r4 * 8]
+%else
+    movh        m0,        [tabw_ChromaCoeff + r4 * 8]
+%endif
+
+    punpcklqdq  m0,        m0
+    lea         r5,        [3 * r1]
+
+%assign x 1
+%rep %2/4
+    movd        m2,        [r0]
+    movd        m3,        [r0 + r1]
+    movd        m4,        [r0 + 2 * r1]
+    movd        m5,        [r0 + r5]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklwd   m2,        m6
+
+    WORD_TO_DOUBLE         m2
+    pmaddwd     m2,        m0
+
+    lea         r0,        [r0 + 4 * r1]
+    movd        m6,        [r0]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklwd   m3,        m7
+
+    WORD_TO_DOUBLE         m3
+    pmaddwd     m3,        m0
+
+    packssdw    m2,        m3
+    pshuflw     m3,        m2,          q2301
+    pshufhw     m3,        m3,          q2301
+    paddw       m2,        m3
+
+    movd        m7,        [r0 + r1]
+
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklwd   m4,        m3
+
+    WORD_TO_DOUBLE         m4
+    pmaddwd     m4,        m0
+
+    movd        m3,        [r0 + 2 * r1]
+
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklwd   m5,        m7
+
+    WORD_TO_DOUBLE         m5
+    pmaddwd     m5,        m0
+
+    packssdw    m4,        m5
+    pshuflw     m5,        m4,          q2301
+    pshufhw     m5,        m5,          q2301
+    paddw       m4,        m5
+
+%ifidn %1,pp
+    psrld       m2,        16
+    psrld       m4,        16
+    packssdw    m2,        m4
+    paddw       m2,        m1
+    psraw       m2,        6
+    packuswb    m2,        m2
+
+%if ARCH_X86_64
+    movq        r4,        m2
+    mov         [r2],      r4w
+    shr         r4,        16
+    mov         [r2 + r3], r4w
+    lea         r2,        [r2 + 2 * r3]
+    shr         r4,        16
+    mov         [r2],      r4w
+    shr         r4,        16
+    mov         [r2 + r3], r4w
+%else
+    movd        r4,        m2
+    mov         [r2],      r4w
+    shr         r4,        16
+    mov         [r2 + r3], r4w
+    lea         r2,        [r2 + 2 * r3]
+    psrldq      m2,        4
+    movd        r4,        m2
+    mov         [r2],      r4w
+    shr         r4,        16
+    mov         [r2 + r3], r4w
+%endif
+%elifidn %1,ps
+    psrldq      m2,        2
+    psrldq      m4,        2
+    pshufd      m2,        m2, q3120
+    pshufd      m4,        m4, q3120
+    psubw       m4,        m1
+    psubw       m2,        m1
+
+    movd        [r2],      m2
+    psrldq      m2,        4
+    movd        [r2 + r3], m2
+    lea         r2,        [r2 + 2 * r3]
+    movd        [r2],      m4
+    psrldq      m4,        4
+    movd        [r2 + r3], m4
+%endif
+
+%if x < %2/4
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%assign x x+1
+%endrep
+    RET
+
+%endmacro
+
+    FILTER_V4_W2_H4_sse2 pp, 4
+    FILTER_V4_W2_H4_sse2 pp, 8
+    FILTER_V4_W2_H4_sse2 pp, 16
+
+    FILTER_V4_W2_H4_sse2 ps, 4
+    FILTER_V4_W2_H4_sse2 ps, 8
+    FILTER_V4_W2_H4_sse2 ps, 16
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro  FILTER_V2_W4_H4_sse2 1
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_4x2, 4, 6, 8
+    mov         r4d,       r4m
+    sub         r0,        r1
+    pxor        m7,        m7
+
+%ifdef PIC
+    lea         r5,        [tabw_ChromaCoeff]
+    movh        m0,        [r5 + r4 * 8]
+%else
+    movh        m0,        [tabw_ChromaCoeff + r4 * 8]
+%endif
+
+    lea         r5,        [r0 + 2 * r1]
+    punpcklqdq  m0,        m0
+    movd        m2,        [r0]
+    movd        m3,        [r0 + r1]
+    movd        m4,        [r5]
+    movd        m5,        [r5 + r1]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m1,        m4,        m5
+    punpcklwd   m2,        m1
+
+    movhlps     m6,        m2
+    punpcklbw   m2,        m7
+    punpcklbw   m6,        m7
+    pmaddwd     m2,        m0
+    pmaddwd     m6,        m0
+    packssdw    m2,        m6
+
+    movd        m1,        [r0 + 4 * r1]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m5,        m1
+    punpcklwd   m3,        m5
+
+    movhlps     m6,        m3
+    punpcklbw   m3,        m7
+    punpcklbw   m6,        m7
+    pmaddwd     m3,        m0
+    pmaddwd     m6,        m0
+    packssdw    m3,        m6
+
+    pshuflw     m4,        m2,        q2301
+    pshufhw     m4,        m4,        q2301
+    paddw       m2,        m4
+    pshuflw     m5,        m3,        q2301
+    pshufhw     m5,        m5,        q2301
+    paddw       m3,        m5
+
+%ifidn %1, pp
+    psrld       m2,        16
+    psrld       m3,        16
+    packssdw    m2,        m3
+
+    paddw       m2,        [pw_32]
+    psraw       m2,        6
+    packuswb    m2,        m2
+
+    movd        [r2],      m2
+    psrldq      m2,        4
+    movd        [r2 + r3], m2
+%elifidn %1, ps
+    psrldq      m2,        2
+    psrldq      m3,        2
+    pshufd      m2,        m2, q3120
+    pshufd      m3,        m3, q3120
+    punpcklqdq  m2, m3
+
+    add         r3d,       r3d
+    psubw       m2,        [pw_2000]
+    movh        [r2],      m2
+    movhps      [r2 + r3], m2
+%endif
+    RET
+
+%endmacro
+
+    FILTER_V2_W4_H4_sse2 pp
+    FILTER_V2_W4_H4_sse2 ps
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_4x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W4_H4_sse2 2
+INIT_XMM sse2
+%if ARCH_X86_64
+cglobal interp_4tap_vert_%1_4x%2, 4, 6, 9
+    pxor        m8,        m8
+%else
+cglobal interp_4tap_vert_%1_4x%2, 4, 6, 8
+%endif
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tabw_ChromaCoeff]
+    movh        m0,        [r5 + r4 * 8]
+%else
+    movh        m0,        [tabw_ChromaCoeff + r4 * 8]
+%endif
+
+%ifidn %1,pp
+    mova        m1,        [pw_32]
+%elifidn %1,ps
+    add         r3d,       r3d
+    mova        m1,        [pw_2000]
+%endif
+
+    lea         r5,        [3 * r1]
+    lea         r4,        [3 * r3]
+    punpcklqdq  m0,        m0
+
+%assign x 1
+%rep %2/4
+    movd        m2,        [r0]
+    movd        m3,        [r0 + r1]
+    movd        m4,        [r0 + 2 * r1]
+    movd        m5,        [r0 + r5]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklwd   m2,        m6
+
+    movhlps     m6,        m2
+    WORD_TO_DOUBLE         m2
+    WORD_TO_DOUBLE         m6
+    pmaddwd     m2,        m0
+    pmaddwd     m6,        m0
+    packssdw    m2,        m6
+
+    lea         r0,        [r0 + 4 * r1]
+    movd        m6,        [r0]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklwd   m3,        m7
+
+    movhlps     m7,        m3
+    WORD_TO_DOUBLE         m3
+    WORD_TO_DOUBLE         m7
+    pmaddwd     m3,        m0
+    pmaddwd     m7,        m0
+    packssdw    m3,        m7
+
+    pshuflw     m7,        m2,        q2301
+    pshufhw     m7,        m7,        q2301
+    paddw       m2,        m7
+    pshuflw     m7,        m3,        q2301
+    pshufhw     m7,        m7,        q2301
+    paddw       m3,        m7
+
+%ifidn %1,pp
+    psrld       m2,        16
+    psrld       m3,        16
+    packssdw    m2,        m3
+    paddw       m2,        m1
+    psraw       m2,        6
+%elifidn %1,ps
+    psrldq      m2,        2
+    psrldq      m3,        2
+    pshufd      m2,        m2, q3120
+    pshufd      m3,        m3, q3120
+    punpcklqdq  m2,        m3
+
+    psubw       m2,        m1
+    movh        [r2],      m2
+    movhps      [r2 + r3], m2
+%endif
+
+    movd        m7,        [r0 + r1]
+
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklwd   m4,        m3
+
+    movhlps     m3,        m4
+    WORD_TO_DOUBLE         m4
+    WORD_TO_DOUBLE         m3
+    pmaddwd     m4,        m0
+    pmaddwd     m3,        m0
+    packssdw    m4,        m3
+
+    movd        m3,        [r0 + 2 * r1]
+
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklwd   m5,        m7
+
+    movhlps     m3,        m5
+    WORD_TO_DOUBLE         m5
+    WORD_TO_DOUBLE         m3
+    pmaddwd     m5,        m0
+    pmaddwd     m3,        m0
+    packssdw    m5,        m3
+
+    pshuflw     m7,        m4,        q2301
+    pshufhw     m7,        m7,        q2301
+    paddw       m4,        m7
+    pshuflw     m7,        m5,        q2301
+    pshufhw     m7,        m7,        q2301
+    paddw       m5,        m7
+
+%ifidn %1,pp
+    psrld       m4,        16
+    psrld       m5,        16
+    packssdw    m4,        m5
+
+    paddw       m4,        m1
+    psraw       m4,        6
+    packuswb    m2,        m4
+
+    movd        [r2],      m2
+    psrldq      m2,        4
+    movd        [r2 + r3], m2
+    psrldq      m2,        4
+    movd        [r2 + 2 * r3],      m2
+    psrldq      m2,        4
+    movd        [r2 + r4], m2
+%elifidn %1,ps
+    psrldq      m4,        2
+    psrldq      m5,        2
+    pshufd      m4,        m4, q3120
+    pshufd      m5,        m5, q3120
+    punpcklqdq  m4,        m5
+    psubw       m4,        m1
+    movh        [r2 + 2 * r3],      m4
+    movhps      [r2 + r4], m4
+%endif
+
+%if x < %2/4
+    lea         r2,        [r2 + 4 * r3]
+%endif
+
+%assign x x+1
+%endrep
+    RET
+
+%endmacro
+
+    FILTER_V4_W4_H4_sse2 pp, 4
+    FILTER_V4_W4_H4_sse2 pp, 8
+    FILTER_V4_W4_H4_sse2 pp, 16
+    FILTER_V4_W4_H4_sse2 pp, 32
+
+    FILTER_V4_W4_H4_sse2 ps, 4
+    FILTER_V4_W4_H4_sse2 ps, 8
+    FILTER_V4_W4_H4_sse2 ps, 16
+    FILTER_V4_W4_H4_sse2 ps, 32
+
+;-----------------------------------------------------------------------------
+;void interp_4tap_vert_%1_6x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W6_H4_sse2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_6x%2, 4, 7, 10
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeffV]
+    mova        m6,        [r5 + r4]
+    mova        m5,        [r5 + r4 + 16]
+%else
+    mova        m6,        [tab_ChromaCoeffV + r4]
+    mova        m5,        [tab_ChromaCoeffV + r4 + 16]
+%endif
+
+%ifidn %1,pp
+    mova        m4,        [pw_32]
+%elifidn %1,ps
+    mova        m4,        [pw_2000]
+    add         r3d,       r3d
+%endif
+    lea         r5,        [3 * r1]
+
+%assign x 1
+%rep %2/4
+    movq        m0,        [r0]
+    movq        m1,        [r0 + r1]
+    movq        m2,        [r0 + 2 * r1]
+    movq        m3,        [r0 + r5]
+
+    punpcklbw   m0,        m1
+    punpcklbw   m1,        m2
+    punpcklbw   m2,        m3
+
+    movhlps     m7,        m0
+    punpcklbw   m0,        m9
+    punpcklbw   m7,        m9
+    pmaddwd     m0,        m6
+    pmaddwd     m7,        m6
+    packssdw    m0,        m7
+
+    movhlps     m8,        m2
+    movq        m7,        m2
+    punpcklbw   m8,        m9
+    punpcklbw   m7,        m9
+    pmaddwd     m8,        m5
+    pmaddwd     m7,        m5
+    packssdw    m7,        m8
+
+    paddw       m0,        m7
+
+%ifidn %1,pp
+    paddw       m0,        m4
+    psraw       m0,        6
+    packuswb    m0,        m0
+
+    movd        [r2],      m0
+    pextrw      r6d,       m0,        2
+    mov         [r2 + 4],  r6w
+%elifidn %1,ps
+    psubw       m0,        m4
+    movh        [r2],      m0
+    pshufd      m0,        m0,        2
+    movd        [r2 + 8],  m0
+%endif
+
+    lea         r0,        [r0 + 4 * r1]
+
+    movq        m0,        [r0]
+    punpcklbw   m3,        m0
+
+    movhlps     m8,        m1
+    punpcklbw   m1,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m1,        m6
+    pmaddwd     m8,        m6
+    packssdw    m1,        m8
+
+    movhlps     m8,        m3
+    movq        m7,        m3
+    punpcklbw   m8,        m9
+    punpcklbw   m7,        m9
+    pmaddwd     m8,        m5
+    pmaddwd     m7,        m5
+    packssdw    m7,        m8
+
+    paddw       m1,        m7
+
+%ifidn %1,pp
+    paddw       m1,        m4
+    psraw       m1,        6
+    packuswb    m1,        m1
+
+    movd        [r2 + r3], m1
+    pextrw      r6d,       m1,        2
+    mov         [r2 + r3 + 4], r6w
+%elifidn %1,ps
+    psubw       m1,        m4
+    movh        [r2 + r3], m1
+    pshufd      m1,        m1,        2
+    movd        [r2 + r3 + 8],  m1
+%endif
+
+    movq        m1,        [r0 + r1]
+    punpcklbw   m7,        m0,        m1
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m6
+    pmaddwd     m8,        m6
+    packssdw    m2,        m8
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m5
+    pmaddwd     m8,        m5
+    packssdw    m7,        m8
+
+    paddw       m2,        m7
+    lea         r2,        [r2 + 2 * r3]
+
+%ifidn %1,pp
+    paddw       m2,        m4
+    psraw       m2,        6
+    packuswb    m2,        m2
+    movd        [r2],      m2
+    pextrw      r6d,       m2,    2
+    mov         [r2 + 4],  r6w
+%elifidn %1,ps
+    psubw       m2,        m4
+    movh        [r2],      m2
+    pshufd      m2,        m2,        2
+    movd        [r2 + 8],  m2
+%endif
+
+    movq        m2,        [r0 + 2 * r1]
+    punpcklbw   m1,        m2
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m6
+    pmaddwd     m8,        m6
+    packssdw    m3,        m8
+
+    movhlps     m8,        m1
+    punpcklbw   m1,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m1,        m5
+    pmaddwd     m8,        m5
+    packssdw    m1,        m8
+
+    paddw       m3,        m1
+
+%ifidn %1,pp
+    paddw       m3,        m4
+    psraw       m3,        6
+    packuswb    m3,        m3
+
+    movd        [r2 + r3], m3
+    pextrw      r6d,       m3,    2
+    mov         [r2 + r3 + 4], r6w
+%elifidn %1,ps
+    psubw       m3,        m4
+    movh        [r2 + r3], m3
+    pshufd      m3,        m3,        2
+    movd        [r2 + r3 + 8],  m3
+%endif
+
+%if x < %2/4
+    lea         r2,        [r2 + 2 * r3]
+%endif
+
+%assign x x+1
+%endrep
+    RET
+
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W6_H4_sse2 pp, 8
+    FILTER_V4_W6_H4_sse2 pp, 16
+    FILTER_V4_W6_H4_sse2 ps, 8
+    FILTER_V4_W6_H4_sse2 ps, 16
+%endif
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W8_sse2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_8x%2, 4, 7, 12
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+
+%ifidn %1,pp
+    mova        m4,        [pw_32]
+%elifidn %1,ps
+    mova        m4,        [pw_2000]
+    add         r3d,       r3d
+%endif
+
+%ifdef PIC
+    lea         r6,        [tab_ChromaCoeffV]
+    mova        m6,        [r6 + r4]
+    mova        m5,        [r6 + r4 + 16]
+%else
+    mova        m6,        [tab_ChromaCoeffV + r4]
+    mova        m5,        [tab_ChromaCoeffV + r4 + 16]
+%endif
+
+    movq        m0,        [r0]
+    movq        m1,        [r0 + r1]
+    movq        m2,        [r0 + 2 * r1]
+    lea         r5,        [r0 + 2 * r1]
+    movq        m3,        [r5 + r1]
+
+    punpcklbw   m0,        m1
+    punpcklbw   m7,        m2,          m3
+
+    movhlps     m8,        m0
+    punpcklbw   m0,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m0,        m6
+    pmaddwd     m8,        m6
+    packssdw    m0,        m8
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m5
+    pmaddwd     m8,        m5
+    packssdw    m7,        m8
+
+    paddw       m0,        m7
+
+%ifidn %1,pp
+    paddw       m0,        m4
+    psraw       m0,        6
+%elifidn %1,ps
+    psubw       m0,        m4
+    movu        [r2],      m0
+%endif
+
+    movq        m11,        [r0 + 4 * r1]
+
+    punpcklbw   m1,        m2
+    punpcklbw   m7,        m3,        m11
+
+    movhlps     m8,        m1
+    punpcklbw   m1,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m1,        m6
+    pmaddwd     m8,        m6
+    packssdw    m1,        m8
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m5
+    pmaddwd     m8,        m5
+    packssdw    m7,        m8
+
+    paddw       m1,        m7
+
+%ifidn %1,pp
+    paddw       m1,        m4
+    psraw       m1,        6
+    packuswb    m1,        m0
+
+    movhps      [r2],      m1
+    movh        [r2 + r3], m1
+%elifidn %1,ps
+    psubw       m1,        m4
+    movu        [r2 + r3], m1
+%endif
+%if %2 == 2     ;end of 8x2
+    RET
+
+%else
+    lea         r6,        [r0 + 4 * r1]
+    movq        m1,        [r6 + r1]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m7,        m11,        m1
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m6
+    pmaddwd     m8,        m6
+    packssdw    m2,        m8
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m5
+    pmaddwd     m8,        m5
+    packssdw    m7,        m8
+
+    paddw       m2,        m7
+
+%ifidn %1,pp
+    paddw       m2,        m4
+    psraw       m2,        6
+%elifidn %1,ps
+    psubw       m2,        m4
+    movu        [r2 + 2 * r3], m2
+%endif
+
+    movq        m10,        [r6 + 2 * r1]
+
+    punpcklbw   m3,        m11
+    punpcklbw   m7,        m1,        m10
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m6
+    pmaddwd     m8,        m6
+    packssdw    m3,        m8
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m5
+    pmaddwd     m8,        m5
+    packssdw    m7,        m8
+
+    paddw       m3,        m7
+    lea         r5,        [r2 + 2 * r3]
+
+%ifidn %1,pp
+    paddw       m3,        m4
+    psraw       m3,        6
+    packuswb    m3,        m2
+
+    movhps      [r2 + 2 * r3], m3
+    movh        [r5 + r3], m3
+%elifidn %1,ps
+    psubw       m3,        m4
+    movu        [r5 + r3], m3
+%endif
+%if %2 == 4     ;end of 8x4
+    RET
+
+%else
+    lea         r6,        [r6 + 2 * r1]
+    movq        m3,        [r6 + r1]
+
+    punpcklbw   m11,        m1
+    punpcklbw   m7,        m10,        m3
+
+    movhlps     m8,        m11
+    punpcklbw   m11,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m11,        m6
+    pmaddwd     m8,        m6
+    packssdw    m11,        m8
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m5
+    pmaddwd     m8,        m5
+    packssdw    m7,        m8
+
+    paddw       m11,       m7
+
+%ifidn %1, pp
+    paddw       m11,       m4
+    psraw       m11,       6
+%elifidn %1,ps
+    psubw       m11,       m4
+    movu        [r2 + 4 * r3], m11
+%endif
+
+    movq        m7,        [r0 + 8 * r1]
+
+    punpcklbw   m1,        m10
+    punpcklbw   m3,        m7
+
+    movhlps     m8,        m1
+    punpcklbw   m1,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m1,        m6
+    pmaddwd     m8,        m6
+    packssdw    m1,        m8
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m5
+    pmaddwd     m8,        m5
+    packssdw    m3,        m8
+
+    paddw       m1,        m3
+    lea         r5,        [r2 + 4 * r3]
+
+%ifidn %1,pp
+    paddw       m1,        m4
+    psraw       m1,        6
+    packuswb    m1,        m11
+
+    movhps      [r2 + 4 * r3], m1
+    movh        [r5 + r3], m1
+%elifidn %1,ps
+    psubw       m1,        m4
+    movu        [r5 + r3], m1
+%endif
+%if %2 == 6
+    RET
+
+%else
+  %error INVALID macro argument, only 2, 4 or 6!
+%endif
+%endif
+%endif
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W8_sse2 pp, 2
+    FILTER_V4_W8_sse2 pp, 4
+    FILTER_V4_W8_sse2 pp, 6
+    FILTER_V4_W8_sse2 ps, 2
+    FILTER_V4_W8_sse2 ps, 4
+    FILTER_V4_W8_sse2 ps, 6
+%endif
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_8x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W8_H8_H16_H32_sse2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_8x%2, 4, 6, 11
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeffV]
+    mova        m6,        [r5 + r4]
+    mova        m5,        [r5 + r4 + 16]
+%else
+    mova        m6,        [tab_ChromaCoeff + r4]
+    mova        m5,        [tab_ChromaCoeff + r4 + 16]
+%endif
+
+%ifidn %1,pp
+    mova        m4,        [pw_32]
+%elifidn %1,ps
+    mova        m4,        [pw_2000]
+    add         r3d,       r3d
+%endif
+
+    lea         r5,        [r1 * 3]
+
+%assign x 1
+%rep %2/4
+    movq        m0,        [r0]
+    movq        m1,        [r0 + r1]
+    movq        m2,        [r0 + 2 * r1]
+    movq        m3,        [r0 + r5]
+
+    punpcklbw   m0,        m1
+    punpcklbw   m1,        m2
+    punpcklbw   m2,        m3
+
+    movhlps     m7,        m0
+    punpcklbw   m0,        m9
+    punpcklbw   m7,        m9
+    pmaddwd     m0,        m6
+    pmaddwd     m7,        m6
+    packssdw    m0,        m7
+
+    movhlps     m8,        m2
+    movq        m7,        m2
+    punpcklbw   m8,        m9
+    punpcklbw   m7,        m9
+    pmaddwd     m8,        m5
+    pmaddwd     m7,        m5
+    packssdw    m7,        m8
+
+    paddw       m0,        m7
+
+%ifidn %1,pp
+    paddw       m0,        m4
+    psraw       m0,        6
+%elifidn %1,ps
+    psubw       m0,        m4
+    movu        [r2],      m0
+%endif
+
+    lea         r0,        [r0 + 4 * r1]
+    movq        m10,       [r0]
+    punpcklbw   m3,        m10
+
+    movhlps     m8,        m1
+    punpcklbw   m1,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m1,        m6
+    pmaddwd     m8,        m6
+    packssdw    m1,        m8
+
+    movhlps     m8,        m3
+    movq        m7,        m3
+    punpcklbw   m8,        m9
+    punpcklbw   m7,        m9
+    pmaddwd     m8,        m5
+    pmaddwd     m7,        m5
+    packssdw    m7,        m8
+
+    paddw       m1,        m7
+
+%ifidn %1,pp
+    paddw       m1,        m4
+    psraw       m1,        6
+
+    packuswb    m0,        m1
+    movh        [r2],      m0
+    movhps      [r2 + r3], m0
+%elifidn %1,ps
+    psubw       m1,        m4
+    movu        [r2 + r3], m1
+%endif
+
+    movq        m1,        [r0 + r1]
+    punpcklbw   m10,       m1
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m6
+    pmaddwd     m8,        m6
+    packssdw    m2,        m8
+
+    movhlps     m8,        m10
+    punpcklbw   m10,       m9
+    punpcklbw   m8,        m9
+    pmaddwd     m10,       m5
+    pmaddwd     m8,        m5
+    packssdw    m10,       m8
+
+    paddw       m2,        m10
+    lea         r2,        [r2 + 2 * r3]
+
+%ifidn %1,pp
+    paddw       m2,        m4
+    psraw       m2,        6
+%elifidn %1,ps
+    psubw       m2,        m4
+    movu        [r2],      m2
+%endif
+
+    movq        m7,        [r0 + 2 * r1]
+    punpcklbw   m1,        m7
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m6
+    pmaddwd     m8,        m6
+    packssdw    m3,        m8
+
+    movhlps     m8,        m1
+    punpcklbw   m1,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m1,        m5
+    pmaddwd     m8,        m5
+    packssdw    m1,        m8
+
+    paddw       m3,        m1
+
+%ifidn %1,pp
+    paddw       m3,        m4
+    psraw       m3,        6
+
+    packuswb    m2,        m3
+    movh        [r2],      m2
+    movhps      [r2 + r3], m2
+%elifidn %1,ps
+    psubw       m3,        m4
+    movu        [r2 + r3], m3
+%endif
+
+%if x < %2/4
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%endrep
+    RET
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W8_H8_H16_H32_sse2 pp,  8
+    FILTER_V4_W8_H8_H16_H32_sse2 pp, 16
+    FILTER_V4_W8_H8_H16_H32_sse2 pp, 32
+
+    FILTER_V4_W8_H8_H16_H32_sse2 pp, 12
+    FILTER_V4_W8_H8_H16_H32_sse2 pp, 64
+
+    FILTER_V4_W8_H8_H16_H32_sse2 ps,  8
+    FILTER_V4_W8_H8_H16_H32_sse2 ps, 16
+    FILTER_V4_W8_H8_H16_H32_sse2 ps, 32
+
+    FILTER_V4_W8_H8_H16_H32_sse2 ps, 12
+    FILTER_V4_W8_H8_H16_H32_sse2 ps, 64
+%endif
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_12x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W12_H2_sse2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_12x%2, 4, 6, 11
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+
+%ifidn %1,pp
+    mova        m6,        [pw_32]
+%elifidn %1,ps
+    mova        m6,        [pw_2000]
+    add         r3d,       r3d
+%endif
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeffV]
+    mova        m1,        [r5 + r4]
+    mova        m0,        [r5 + r4 + 16]
+%else
+    mova        m1,        [tab_ChromaCoeffV + r4]
+    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
+%endif
+
+%assign x 1
+%rep %2/2
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m1
+    pmaddwd     m8,        m1
+    packssdw    m2,        m8
+
+    lea         r0,        [r0 + 2 * r1]
+    movu        m5,        [r0]
+    movu        m7,        [r0 + r1]
+
+    punpcklbw   m10,       m5,        m7
+    movhlps     m8,        m10
+    punpcklbw   m10,       m9
+    punpcklbw   m8,        m9
+    pmaddwd     m10,       m0
+    pmaddwd     m8,        m0
+    packssdw    m10,       m8
+
+    paddw       m4,        m10
+
+    punpckhbw   m10,       m5,        m7
+    movhlps     m8,        m10
+    punpcklbw   m10,       m9
+    punpcklbw   m8,        m9
+    pmaddwd     m10,       m0
+    pmaddwd     m8,        m0
+    packssdw    m10,       m8
+
+    paddw       m2,        m10
+
+%ifidn %1,pp
+    paddw       m4,        m6
+    psraw       m4,        6
+    paddw       m2,        m6
+    psraw       m2,        6
+
+    packuswb    m4,        m2
+    movh        [r2],      m4
+    psrldq      m4,        8
+    movd        [r2 + 8],  m4
+%elifidn %1,ps
+    psubw       m4,        m6
+    psubw       m2,        m6
+    movu        [r2],      m4
+    movh        [r2 + 16], m2
+%endif
+
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m4
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m1
+    pmaddwd     m8,        m1
+    packssdw    m3,        m8
+
+    movu        m5,        [r0 + 2 * r1]
+    punpcklbw   m2,        m7,        m5
+    punpckhbw   m7,        m5
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m0
+    pmaddwd     m8,        m0
+    packssdw    m2,        m8
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m0
+    pmaddwd     m8,        m0
+    packssdw    m7,        m8
+
+    paddw       m4,        m2
+    paddw       m3,        m7
+
+%ifidn %1,pp
+    paddw       m4,        m6
+    psraw       m4,        6
+    paddw       m3,        m6
+    psraw       m3,        6
+
+    packuswb    m4,        m3
+    movh        [r2 + r3], m4
+    psrldq      m4,        8
+    movd        [r2 + r3 + 8], m4
+%elifidn %1,ps
+    psubw       m4,        m6
+    psubw       m3,        m6
+    movu        [r2 + r3], m4
+    movh        [r2 + r3 + 16], m3
+%endif
+
+%if x < %2/2
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%assign x x+1
+%endrep
+    RET
+
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W12_H2_sse2 pp, 16
+    FILTER_V4_W12_H2_sse2 pp, 32
+    FILTER_V4_W12_H2_sse2 ps, 16
+    FILTER_V4_W12_H2_sse2 ps, 32
+%endif
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W16_H2_sse2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_16x%2, 4, 6, 11
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+
+%ifidn %1,pp
+    mova        m6,        [pw_32]
+%elifidn %1,ps
+    mova        m6,        [pw_2000]
+    add         r3d,       r3d
+%endif
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeffV]
+    mova        m1,        [r5 + r4]
+    mova        m0,        [r5 + r4 + 16]
+%else
+    mova        m1,        [tab_ChromaCoeffV + r4]
+    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
+%endif
+
+%assign x 1
+%rep %2/2
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m1
+    pmaddwd     m8,        m1
+    packssdw    m2,        m8
+
+    lea         r0,        [r0 + 2 * r1]
+    movu        m5,        [r0]
+    movu        m10,       [r0 + r1]
+
+    punpckhbw   m7,        m5,        m10
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m0
+    pmaddwd     m8,        m0
+    packssdw    m7,        m8
+    paddw       m2,        m7
+
+    punpcklbw   m7,        m5,        m10
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m0
+    pmaddwd     m8,        m0
+    packssdw    m7,        m8
+    paddw       m4,        m7
+
+%ifidn %1,pp
+    paddw       m4,        m6
+    psraw       m4,        6
+    paddw       m2,        m6
+    psraw       m2,        6
+
+    packuswb    m4,        m2
+    movu        [r2],      m4
+%elifidn %1,ps
+    psubw       m4,        m6
+    psubw       m2,        m6
+    movu        [r2],      m4
+    movu        [r2 + 16], m2
+%endif
+
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m1
+    pmaddwd     m8,        m1
+    packssdw    m3,        m8
+
+    movu        m5,        [r0 + 2 * r1]
+
+    punpcklbw   m2,        m10,       m5
+    punpckhbw   m10,       m5
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m0
+    pmaddwd     m8,        m0
+    packssdw    m2,        m8
+
+    movhlps     m8,        m10
+    punpcklbw   m10,       m9
+    punpcklbw   m8,        m9
+    pmaddwd     m10,       m0
+    pmaddwd     m8,        m0
+    packssdw    m10,       m8
+
+    paddw       m4,        m2
+    paddw       m3,        m10
+
+%ifidn %1,pp
+    paddw       m4,        m6
+    psraw       m4,        6
+    paddw       m3,        m6
+    psraw       m3,        6
+
+    packuswb    m4,        m3
+    movu        [r2 + r3], m4
+%elifidn %1,ps
+    psubw       m4,        m6
+    psubw       m3,        m6
+    movu        [r2 + r3], m4
+    movu        [r2 + r3 + 16], m3
+%endif
+
+%if x < %2/2
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%assign x x+1
+%endrep
+    RET
+
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W16_H2_sse2 pp, 4
+    FILTER_V4_W16_H2_sse2 pp, 8
+    FILTER_V4_W16_H2_sse2 pp, 12
+    FILTER_V4_W16_H2_sse2 pp, 16
+    FILTER_V4_W16_H2_sse2 pp, 32
+
+    FILTER_V4_W16_H2_sse2 pp, 24
+    FILTER_V4_W16_H2_sse2 pp, 64
+
+    FILTER_V4_W16_H2_sse2 ps, 4
+    FILTER_V4_W16_H2_sse2 ps, 8
+    FILTER_V4_W16_H2_sse2 ps, 12
+    FILTER_V4_W16_H2_sse2 ps, 16
+    FILTER_V4_W16_H2_sse2 ps, 32
+
+    FILTER_V4_W16_H2_sse2 ps, 24
+    FILTER_V4_W16_H2_sse2 ps, 64
+%endif
+
+;-----------------------------------------------------------------------------
+;void interp_4tap_vert_%1_24%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W24_sse2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_24x%2, 4, 6, 11
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+
+%ifidn %1,pp
+    mova        m6,        [pw_32]
+%elifidn %1,ps
+    mova        m6,        [pw_2000]
+    add         r3d,       r3d
+%endif
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeffV]
+    mova        m1,        [r5 + r4]
+    mova        m0,        [r5 + r4 + 16]
+%else
+    mova        m1,        [tab_ChromaCoeffV + r4]
+    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
+%endif
+
+%assign x 1
+%rep %2/2
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m1
+    pmaddwd     m8,        m1
+    packssdw    m2,        m8
+
+    lea         r5,        [r0 + 2 * r1]
+    movu        m5,        [r5]
+    movu        m10,       [r5 + r1]
+    punpcklbw   m7,        m5,        m10
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m0
+    pmaddwd     m8,        m0
+    packssdw    m7,        m8
+    paddw       m4,        m7
+
+    punpckhbw   m7,        m5,        m10
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m0
+    pmaddwd     m8,        m0
+    packssdw    m7,        m8
+
+    paddw       m2,        m7
+
+%ifidn %1,pp
+    paddw       m4,        m6
+    psraw       m4,        6
+    paddw       m2,        m6
+    psraw       m2,        6
+
+    packuswb    m4,        m2
+    movu        [r2],      m4
+%elifidn %1,ps
+    psubw       m4,        m6
+    psubw       m2,        m6
+    movu        [r2],      m4
+    movu        [r2 + 16], m2
+%endif
+
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m1
+    pmaddwd     m8,        m1
+    packssdw    m3,        m8
+
+    movu        m2,        [r5 + 2 * r1]
+
+    punpcklbw   m5,        m10,        m2
+    punpckhbw   m10,       m2
+
+    movhlps     m8,        m5
+    punpcklbw   m5,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m5,        m0
+    pmaddwd     m8,        m0
+    packssdw    m5,        m8
+
+    movhlps     m8,        m10
+    punpcklbw   m10,       m9
+    punpcklbw   m8,        m9
+    pmaddwd     m10,       m0
+    pmaddwd     m8,        m0
+    packssdw    m10,       m8
+
+    paddw       m4,        m5
+    paddw       m3,        m10
+
+%ifidn %1,pp
+    paddw       m4,        m6
+    psraw       m4,        6
+    paddw       m3,        m6
+    psraw       m3,        6
+
+    packuswb    m4,        m3
+    movu        [r2 + r3], m4
+%elifidn %1,ps
+    psubw       m4,        m6
+    psubw       m3,        m6
+    movu        [r2 + r3], m4
+    movu        [r2 + r3 + 16], m3
+%endif
+
+    movq        m2,        [r0 + 16]
+    movq        m3,        [r0 + r1 + 16]
+    movq        m4,        [r5 + 16]
+    movq        m5,        [r5 + r1 + 16]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m4,        m5
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m0
+    pmaddwd     m8,        m0
+    packssdw    m4,        m8
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m1
+    pmaddwd     m8,        m1
+    packssdw    m2,        m8
+
+    paddw       m2,        m4
+
+%ifidn %1,pp
+    paddw       m2,        m6
+    psraw       m2,        6
+%elifidn %1,ps
+    psubw       m2,        m6
+    movu        [r2 + 32], m2
+%endif
+
+    movq        m3,        [r0 + r1 + 16]
+    movq        m4,        [r5 + 16]
+    movq        m5,        [r5 + r1 + 16]
+    movq        m7,        [r5 + 2 * r1 + 16]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m5,        m7
+
+    movhlps     m8,        m5
+    punpcklbw   m5,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m5,        m0
+    pmaddwd     m8,        m0
+    packssdw    m5,        m8
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m1
+    pmaddwd     m8,        m1
+    packssdw    m3,        m8
+
+    paddw       m3,        m5
+
+%ifidn %1,pp
+    paddw       m3,        m6
+    psraw       m3,        6
+
+    packuswb    m2,        m3
+    movh        [r2 + 16], m2
+    movhps      [r2 + r3 + 16], m2
+%elifidn %1,ps
+    psubw       m3,        m6
+    movu        [r2 + r3 + 32], m3
+%endif
+
+%if x < %2/2
+    mov         r0,        r5
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%assign x x+1
+%endrep
+    RET
+
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W24_sse2 pp, 32
+    FILTER_V4_W24_sse2 pp, 64
+    FILTER_V4_W24_sse2 ps, 32
+    FILTER_V4_W24_sse2 ps, 64
+%endif
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W32_sse2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_32x%2, 4, 6, 10
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+
+%ifidn %1,pp
+    mova        m6,        [pw_32]
+%elifidn %1,ps
+    mova        m6,        [pw_2000]
+    add         r3d,       r3d
+%endif
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeffV]
+    mova        m1,        [r5 + r4]
+    mova        m0,        [r5 + r4 + 16]
+%else
+    mova        m1,        [tab_ChromaCoeffV + r4]
+    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
+%endif
+
+    mov         r4d,       %2
+
+.loop:
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m1
+    pmaddwd     m8,        m1
+    packssdw    m2,        m8
+
+    lea         r5,        [r0 + 2 * r1]
+    movu        m3,        [r5]
+    movu        m5,        [r5 + r1]
+
+    punpcklbw   m7,        m3,        m5
+    punpckhbw   m3,        m5
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m0
+    pmaddwd     m8,        m0
+    packssdw    m7,        m8
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m0
+    pmaddwd     m8,        m0
+    packssdw    m3,        m8
+
+    paddw       m4,        m7
+    paddw       m2,        m3
+
+%ifidn %1,pp
+    paddw       m4,        m6
+    psraw       m4,        6
+    paddw       m2,        m6
+    psraw       m2,        6
+
+    packuswb    m4,        m2
+    movu        [r2],      m4
+%elifidn %1,ps
+    psubw       m4,        m6
+    psubw       m2,        m6
+    movu        [r2],      m4
+    movu        [r2 + 16], m2
+%endif
+
+    movu        m2,        [r0 + 16]
+    movu        m3,        [r0 + r1 + 16]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m1
+    pmaddwd     m8,        m1
+    packssdw    m2,        m8
+
+    movu        m3,        [r5 + 16]
+    movu        m5,        [r5 + r1 + 16]
+
+    punpcklbw   m7,        m3,        m5
+    punpckhbw   m3,        m5
+
+    movhlps     m8,        m7
+    punpcklbw   m7,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m7,        m0
+    pmaddwd     m8,        m0
+    packssdw    m7,        m8
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m0
+    pmaddwd     m8,        m0
+    packssdw    m3,        m8
+
+    paddw       m4,        m7
+    paddw       m2,        m3
+
+%ifidn %1,pp
+    paddw       m4,        m6
+    psraw       m4,        6
+    paddw       m2,        m6
+    psraw       m2,        6
+
+    packuswb    m4,        m2
+    movu        [r2 + 16], m4
+%elifidn %1,ps
+    psubw       m4,        m6
+    psubw       m2,        m6
+    movu        [r2 + 32], m4
+    movu        [r2 + 48], m2
+%endif
+
+    lea         r0,        [r0 + r1]
+    lea         r2,        [r2 + r3]
+    dec         r4
+    jnz        .loop
+    RET
+
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W32_sse2 pp, 8
+    FILTER_V4_W32_sse2 pp, 16
+    FILTER_V4_W32_sse2 pp, 24
+    FILTER_V4_W32_sse2 pp, 32
+
+    FILTER_V4_W32_sse2 pp, 48
+    FILTER_V4_W32_sse2 pp, 64
+
+    FILTER_V4_W32_sse2 ps, 8
+    FILTER_V4_W32_sse2 ps, 16
+    FILTER_V4_W32_sse2 ps, 24
+    FILTER_V4_W32_sse2 ps, 32
+
+    FILTER_V4_W32_sse2 ps, 48
+    FILTER_V4_W32_sse2 ps, 64
+%endif
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_%1_%2x%3(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W16n_H2_sse2 3
+INIT_XMM sse2
+cglobal interp_4tap_vert_%1_%2x%3, 4, 7, 11
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+
+%ifidn %1,pp
+    mova        m7,        [pw_32]
+%elifidn %1,ps
+    mova        m7,        [pw_2000]
+    add         r3d,       r3d
+%endif
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeffV]
+    mova        m1,        [r5 + r4]
+    mova        m0,        [r5 + r4 + 16]
+%else
+    mova        m1,        [tab_ChromaCoeffV + r4]
+    mova        m0,        [tab_ChromaCoeffV + r4 + 16]
+%endif
+
+    mov         r4d,       %3/2
+
+.loop:
+
+    mov         r6d,       %2/16
+
+.loopW:
+
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m1
+    pmaddwd     m8,        m1
+    packssdw    m2,        m8
+
+    lea         r5,        [r0 + 2 * r1]
+    movu        m5,        [r5]
+    movu        m6,        [r5 + r1]
+
+    punpckhbw   m10,        m5,        m6
+    movhlps     m8,        m10
+    punpcklbw   m10,       m9
+    punpcklbw   m8,        m9
+    pmaddwd     m10,       m0
+    pmaddwd     m8,        m0
+    packssdw    m10,       m8
+    paddw       m2,        m10
+
+    punpcklbw   m10,        m5,        m6
+    movhlps     m8,        m10
+    punpcklbw   m10,       m9
+    punpcklbw   m8,        m9
+    pmaddwd     m10,       m0
+    pmaddwd     m8,        m0
+    packssdw    m10,       m8
+    paddw       m4,        m10
+
+%ifidn %1,pp
+    paddw       m4,        m7
+    psraw       m4,        6
+    paddw       m2,        m7
+    psraw       m2,        6
+
+    packuswb    m4,        m2
+    movu        [r2],      m4
+%elifidn %1,ps
+    psubw       m4,        m7
+    psubw       m2,        m7
+    movu        [r2],      m4
+    movu        [r2 + 16], m2
+%endif
+
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
+
+    movhlps     m8,        m4
+    punpcklbw   m4,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m4,        m1
+    pmaddwd     m8,        m1
+    packssdw    m4,        m8
+
+    movhlps     m8,        m3
+    punpcklbw   m3,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m3,        m1
+    pmaddwd     m8,        m1
+    packssdw    m3,        m8
+
+    movu        m5,        [r5 + 2 * r1]
+
+    punpcklbw   m2,        m6,        m5
+    punpckhbw   m6,        m5
+
+    movhlps     m8,        m2
+    punpcklbw   m2,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m2,        m0
+    pmaddwd     m8,        m0
+    packssdw    m2,        m8
+
+    movhlps     m8,        m6
+    punpcklbw   m6,        m9
+    punpcklbw   m8,        m9
+    pmaddwd     m6,        m0
+    pmaddwd     m8,        m0
+    packssdw    m6,        m8
+
+    paddw       m4,        m2
+    paddw       m3,        m6
+
+%ifidn %1,pp
+    paddw       m4,        m7
+    psraw       m4,        6
+    paddw       m3,        m7
+    psraw       m3,        6
+
+    packuswb    m4,        m3
+    movu        [r2 + r3], m4
+    add         r2,        16
+%elifidn %1,ps
+    psubw       m4,        m7
+    psubw       m3,        m7
+    movu        [r2 + r3], m4
+    movu        [r2 + r3 + 16], m3
+    add         r2,        32
+%endif
+
+    add         r0,        16
+    dec         r6d
+    jnz         .loopW
+
+    lea         r0,        [r0 + r1 * 2 - %2]
+
+%ifidn %1,pp
+    lea         r2,        [r2 + r3 * 2 - %2]
+%elifidn %1,ps
+    lea         r2,        [r2 + r3 * 2 - (%2 * 2)]
+%endif
+
+    dec         r4d
+    jnz        .loop
+    RET
+
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W16n_H2_sse2 pp, 64, 64
+    FILTER_V4_W16n_H2_sse2 pp, 64, 32
+    FILTER_V4_W16n_H2_sse2 pp, 64, 48
+    FILTER_V4_W16n_H2_sse2 pp, 48, 64
+    FILTER_V4_W16n_H2_sse2 pp, 64, 16
+    FILTER_V4_W16n_H2_sse2 ps, 64, 64
+    FILTER_V4_W16n_H2_sse2 ps, 64, 32
+    FILTER_V4_W16n_H2_sse2 ps, 64, 48
+    FILTER_V4_W16n_H2_sse2 ps, 48, 64
+    FILTER_V4_W16n_H2_sse2 ps, 64, 16
+%endif
+
+;-----------------------------------------------------------------------------
+;void interp_4tap_vert_pp_2x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_2x4, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+    lea         r4,        [r1 * 3]
+    lea         r5,        [r0 + 4 * r1]
+    pshufb      m0,        [tab_Cm]
+    mova        m1,        [pw_512]
+
+    movd        m2,        [r0]
+    movd        m3,        [r0 + r1]
+    movd        m4,        [r0 + 2 * r1]
+    movd        m5,        [r0 + r4]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklbw   m2,        m6
+
+    pmaddubsw   m2,        m0
+
+    movd        m6,        [r5]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklbw   m3,        m7
+
+    pmaddubsw   m3,        m0
+
+    phaddw      m2,        m3
+
+    pmulhrsw    m2,        m1
+
+    movd        m7,        [r5 + r1]
+
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklbw   m4,        m3
+
+    pmaddubsw   m4,        m0
+
+    movd        m3,        [r5 + 2 * r1]
+
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklbw   m5,        m7
+
+    pmaddubsw   m5,        m0
+
+    phaddw      m4,        m5
+
+    pmulhrsw    m4,        m1
+    packuswb    m2,        m4
+
+    pextrw      [r2],      m2, 0
+    pextrw      [r2 + r3], m2, 2
+    lea         r2,        [r2 + 2 * r3]
+    pextrw      [r2],      m2, 4
+    pextrw      [r2 + r3], m2, 6
+
+    RET
+
+%macro FILTER_VER_CHROMA_AVX2_2x4 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_2x4, 4, 6, 2
+    mov             r4d, r4m
+    shl             r4d, 5
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff_V]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeff_V + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+
+    pinsrw          xm1, [r0], 0
+    pinsrw          xm1, [r0 + r1], 1
+    pinsrw          xm1, [r0 + r1 * 2], 2
+    pinsrw          xm1, [r0 + r4], 3
+    lea             r0, [r0 + r1 * 4]
+    pinsrw          xm1, [r0], 4
+    pinsrw          xm1, [r0 + r1], 5
+    pinsrw          xm1, [r0 + r1 * 2], 6
+
+    pshufb          xm0, xm1, [interp_vert_shuf]
+    pshufb          xm1, [interp_vert_shuf + 32]
+    vinserti128     m0, m0, xm1, 1
+    pmaddubsw       m0, [r5]
+    vextracti128    xm1, m0, 1
+    paddw           xm0, xm1
+%ifidn %1,pp
+    pmulhrsw        xm0, [pw_512]
+    packuswb        xm0, xm0
+    lea             r4, [r3 * 3]
+    pextrw          [r2], xm0, 0
+    pextrw          [r2 + r3], xm0, 1
+    pextrw          [r2 + r3 * 2], xm0, 2
+    pextrw          [r2 + r4], xm0, 3
+%else
+    add             r3d, r3d
+    lea             r4, [r3 * 3]
+    psubw           xm0, [pw_2000]
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    pextrd          [r2 + r3 * 2], xm0, 2
+    pextrd          [r2 + r4], xm0, 3
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_2x4 pp
+    FILTER_VER_CHROMA_AVX2_2x4 ps
+
+%macro FILTER_VER_CHROMA_AVX2_2x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_2x8, 4, 6, 2
+    mov             r4d, r4m
+    shl             r4d, 6
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+
+    pinsrw          xm1, [r0], 0
+    pinsrw          xm1, [r0 + r1], 1
+    pinsrw          xm1, [r0 + r1 * 2], 2
+    pinsrw          xm1, [r0 + r4], 3
+    lea             r0, [r0 + r1 * 4]
+    pinsrw          xm1, [r0], 4
+    pinsrw          xm1, [r0 + r1], 5
+    pinsrw          xm1, [r0 + r1 * 2], 6
+    pinsrw          xm1, [r0 + r4], 7
+    movhlps         xm0, xm1
+    lea             r0, [r0 + r1 * 4]
+    pinsrw          xm0, [r0], 4
+    pinsrw          xm0, [r0 + r1], 5
+    pinsrw          xm0, [r0 + r1 * 2], 6
+    vinserti128     m1, m1, xm0, 1
+
+    pshufb          m0, m1, [interp_vert_shuf]
+    pshufb          m1, [interp_vert_shuf + 32]
+    pmaddubsw       m0, [r5]
+    pmaddubsw       m1, [r5 + 1 * mmsize]
+    paddw           m0, m1
+%ifidn %1,pp
+    pmulhrsw        m0, [pw_512]
+    vextracti128    xm1, m0, 1
+    packuswb        xm0, xm1
+    lea             r4, [r3 * 3]
+    pextrw          [r2], xm0, 0
+    pextrw          [r2 + r3], xm0, 1
+    pextrw          [r2 + r3 * 2], xm0, 2
+    pextrw          [r2 + r4], xm0, 3
+    lea             r2, [r2 + r3 * 4]
+    pextrw          [r2], xm0, 4
+    pextrw          [r2 + r3], xm0, 5
+    pextrw          [r2 + r3 * 2], xm0, 6
+    pextrw          [r2 + r4], xm0, 7
+%else
+    add             r3d, r3d
+    lea             r4, [r3 * 3]
+    psubw           m0, [pw_2000]
+    vextracti128    xm1, m0, 1
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    pextrd          [r2 + r3 * 2], xm0, 2
+    pextrd          [r2 + r4], xm0, 3
+    lea             r2, [r2 + r3 * 4]
+    movd            [r2], xm1
+    pextrd          [r2 + r3], xm1, 1
+    pextrd          [r2 + r3 * 2], xm1, 2
+    pextrd          [r2 + r4], xm1, 3
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_2x8 pp
+    FILTER_VER_CHROMA_AVX2_2x8 ps
+
+%macro FILTER_VER_CHROMA_AVX2_2x16 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_2x16, 4, 6, 3
+    mov             r4d, r4m
+    shl             r4d, 6
+    sub             r0,  r1
+
+%ifdef PIC
+    lea             r5,  [tab_ChromaCoeffVer_32]
+    add             r5,  r4
+%else
+    lea             r5,  [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4,  [r1 * 3]
+
+    movd            xm1, [r0]
+    pinsrw          xm1, [r0 + r1], 1
+    pinsrw          xm1, [r0 + r1 * 2], 2
+    pinsrw          xm1, [r0 + r4], 3
+    lea             r0,  [r0 + r1 * 4]
+    pinsrw          xm1, [r0], 4
+    pinsrw          xm1, [r0 + r1], 5
+    pinsrw          xm1, [r0 + r1 * 2], 6
+    pinsrw          xm1, [r0 + r4], 7
+    lea             r0,  [r0 + r1 * 4]
+    pinsrw          xm0, [r0], 4
+    pinsrw          xm0, [r0 + r1], 5
+    pinsrw          xm0, [r0 + r1 * 2], 6
+    pinsrw          xm0, [r0 + r4], 7
+    punpckhqdq      xm0, xm1, xm0
+    vinserti128     m1,  m1,  xm0,  1
+
+    pshufb          m2,  m1,  [interp_vert_shuf]
+    pshufb          m1,  [interp_vert_shuf + 32]
+    pmaddubsw       m2,  [r5]
+    pmaddubsw       m1,  [r5 + 1 * mmsize]
+    paddw           m2,  m1
+
+    lea             r0,  [r0 + r1 * 4]
+    pinsrw          xm1, [r0], 4
+    pinsrw          xm1, [r0 + r1], 5
+    pinsrw          xm1, [r0 + r1 * 2], 6
+    pinsrw          xm1, [r0 + r4], 7
+    punpckhqdq      xm1, xm0, xm1
+    lea             r0,  [r0 + r1 * 4]
+    pinsrw          xm0, [r0], 4
+    pinsrw          xm0, [r0 + r1], 5
+    pinsrw          xm0, [r0 + r1 * 2], 6
+    punpckhqdq      xm0, xm1, xm0
+    vinserti128     m1,  m1,  xm0,  1
+
+    pshufb          m0,  m1,  [interp_vert_shuf]
+    pshufb          m1,  [interp_vert_shuf + 32]
+    pmaddubsw       m0,  [r5]
+    pmaddubsw       m1,  [r5 + 1 * mmsize]
+    paddw           m0,  m1
+%ifidn %1,pp
+    mova            m1,  [pw_512]
+    pmulhrsw        m2,  m1
+    pmulhrsw        m0,  m1
+    packuswb        m2,  m0
+    lea             r4,  [r3 * 3]
+    pextrw          [r2], xm2, 0
+    pextrw          [r2 + r3], xm2, 1
+    pextrw          [r2 + r3 * 2], xm2, 2
+    pextrw          [r2 + r4], xm2, 3
+    vextracti128    xm0, m2, 1
+    lea             r2,  [r2 + r3 * 4]
+    pextrw          [r2], xm0, 0
+    pextrw          [r2 + r3], xm0, 1
+    pextrw          [r2 + r3 * 2], xm0, 2
+    pextrw          [r2 + r4], xm0, 3
+    lea             r2,  [r2 + r3 * 4]
+    pextrw          [r2], xm2, 4
+    pextrw          [r2 + r3], xm2, 5
+    pextrw          [r2 + r3 * 2], xm2, 6
+    pextrw          [r2 + r4], xm2, 7
+    lea             r2,  [r2 + r3 * 4]
+    pextrw          [r2], xm0, 4
+    pextrw          [r2 + r3], xm0, 5
+    pextrw          [r2 + r3 * 2], xm0, 6
+    pextrw          [r2 + r4], xm0, 7
+%else
+    add             r3d, r3d
+    lea             r4,  [r3 * 3]
+    vbroadcasti128  m1,  [pw_2000]
+    psubw           m2,  m1
+    psubw           m0,  m1
+    vextracti128    xm1, m2, 1
+    movd            [r2], xm2
+    pextrd          [r2 + r3], xm2, 1
+    pextrd          [r2 + r3 * 2], xm2, 2
+    pextrd          [r2 + r4], xm2, 3
+    lea             r2, [r2 + r3 * 4]
+    movd            [r2], xm1
+    pextrd          [r2 + r3], xm1, 1
+    pextrd          [r2 + r3 * 2], xm1, 2
+    pextrd          [r2 + r4], xm1, 3
+    vextracti128    xm1, m0, 1
+    lea             r2,  [r2 + r3 * 4]
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    pextrd          [r2 + r3 * 2], xm0, 2
+    pextrd          [r2 + r4], xm0, 3
+    lea             r2,  [r2 + r3 * 4]
+    movd            [r2], xm1
+    pextrd          [r2 + r3], xm1, 1
+    pextrd          [r2 + r3 * 2], xm1, 2
+    pextrd          [r2 + r4], xm1, 3
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_2x16 pp
+    FILTER_VER_CHROMA_AVX2_2x16 ps
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W2_H4 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_2x%2, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m0,        [tab_Cm]
+
+    mova        m1,        [pw_512]
+
+    mov         r4d,       %2
+    lea         r5,        [3 * r1]
+
+.loop:
+    movd        m2,        [r0]
+    movd        m3,        [r0 + r1]
+    movd        m4,        [r0 + 2 * r1]
+    movd        m5,        [r0 + r5]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklbw   m2,        m6
+
+    pmaddubsw   m2,        m0
+
+    lea         r0,        [r0 + 4 * r1]
+    movd        m6,        [r0]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklbw   m3,        m7
+
+    pmaddubsw   m3,        m0
+
+    phaddw      m2,        m3
+
+    pmulhrsw    m2,        m1
+
+    movd        m7,        [r0 + r1]
+
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklbw   m4,        m3
+
+    pmaddubsw   m4,        m0
+
+    movd        m3,        [r0 + 2 * r1]
+
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklbw   m5,        m7
+
+    pmaddubsw   m5,        m0
+
+    phaddw      m4,        m5
+
+    pmulhrsw    m4,        m1
+    packuswb    m2,        m4
+
+    pextrw      [r2],      m2, 0
+    pextrw      [r2 + r3], m2, 2
+    lea         r2,        [r2 + 2 * r3]
+    pextrw      [r2],      m2, 4
+    pextrw      [r2 + r3], m2, 6
+
+    lea         r2,        [r2 + 2 * r3]
+
+    sub         r4,        4
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W2_H4 2, 8
+    FILTER_V4_W2_H4 2, 16
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_4x2, 4, 6, 6
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m0,        [tab_Cm]
+    lea         r5,        [r0 + 2 * r1]
+
+    movd        m2,        [r0]
+    movd        m3,        [r0 + r1]
+    movd        m4,        [r5]
+    movd        m5,        [r5 + r1]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m1,        m4,        m5
+    punpcklbw   m2,        m1
+
+    pmaddubsw   m2,        m0
+
+    movd        m1,        [r0 + 4 * r1]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m5,        m1
+    punpcklbw   m3,        m5
+
+    pmaddubsw   m3,        m0
+
+    phaddw      m2,        m3
+
+    pmulhrsw    m2,        [pw_512]
+    packuswb    m2,        m2
+    movd        [r2],      m2
+    pextrd      [r2 + r3], m2,  1
+
+    RET
+
+%macro FILTER_VER_CHROMA_AVX2_4x2 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x2, 4, 6, 4
+    mov             r4d, r4m
+    shl             r4d, 5
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeff_V]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeff_V + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+
+    movd            xm1, [r0]
+    movd            xm2, [r0 + r1]
+    punpcklbw       xm1, xm2
+    movd            xm3, [r0 + r1 * 2]
+    punpcklbw       xm2, xm3
+    movlhps         xm1, xm2
+    movd            xm0, [r0 + r4]
+    punpcklbw       xm3, xm0
+    movd            xm2, [r0 + r1 * 4]
+    punpcklbw       xm0, xm2
+    movlhps         xm3, xm0
+    vinserti128     m1, m1, xm3, 1                          ; m1 = row[x x x 4 3 2 1 0]
+
+    pmaddubsw       m1, [r5]
+    vextracti128    xm3, m1, 1
+    paddw           xm1, xm3
+%ifidn %1,pp
+    pmulhrsw        xm1, [pw_512]
+    packuswb        xm1, xm1
+    movd            [r2], xm1
+    pextrd          [r2 + r3], xm1, 1
+%else
+    add             r3d, r3d
+    psubw           xm1, [pw_2000]
+    movq            [r2], xm1
+    movhps          [r2 + r3], xm1
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_4x2 pp
+    FILTER_VER_CHROMA_AVX2_4x2 ps
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_4x4, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m0,        [tab_Cm]
+    mova        m1,        [pw_512]
+    lea         r5,        [r0 + 4 * r1]
+    lea         r4,        [r1 * 3]
+
+    movd        m2,        [r0]
+    movd        m3,        [r0 + r1]
+    movd        m4,        [r0 + 2 * r1]
+    movd        m5,        [r0 + r4]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklbw   m2,        m6
+
+    pmaddubsw   m2,        m0
+
+    movd        m6,        [r5]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklbw   m3,        m7
+
+    pmaddubsw   m3,        m0
+
+    phaddw      m2,        m3
+
+    pmulhrsw    m2,        m1
+
+    movd        m7,        [r5 + r1]
+
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklbw   m4,        m3
+
+    pmaddubsw   m4,        m0
+
+    movd        m3,        [r5 + 2 * r1]
+
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklbw   m5,        m7
+
+    pmaddubsw   m5,        m0
+
+    phaddw      m4,        m5
+
+    pmulhrsw    m4,        m1
+
+    packuswb    m2,        m4
+    movd        [r2],      m2
+    pextrd      [r2 + r3], m2, 1
+    lea         r2,        [r2 + 2 * r3]
+    pextrd      [r2],      m2, 2
+    pextrd      [r2 + r3], m2, 3
+    RET
+
+%macro FILTER_VER_CHROMA_AVX2_4x4 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x4, 4, 6, 3
+    mov             r4d, r4m
+    shl             r4d, 6
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+
+    movd            xm1, [r0]
+    pinsrd          xm1, [r0 + r1], 1
+    pinsrd          xm1, [r0 + r1 * 2], 2
+    pinsrd          xm1, [r0 + r4], 3                       ; m1 = row[3 2 1 0]
+    lea             r0, [r0 + r1 * 4]
+    movd            xm2, [r0]
+    pinsrd          xm2, [r0 + r1], 1
+    pinsrd          xm2, [r0 + r1 * 2], 2                   ; m2 = row[x 6 5 4]
+    vinserti128     m1, m1, xm2, 1                          ; m1 = row[x 6 5 4 3 2 1 0]
+    mova            m2, [v4_interp4_vpp_shuf1]
+    vpermd          m0, m2, m1                              ; m0 = row[4 3 3 2 2 1 1 0]
+    mova            m2, [v4_interp4_vpp_shuf1 + mmsize]
+    vpermd          m1, m2, m1                              ; m1 = row[6 5 5 4 4 3 3 2]
+
+    mova            m2, [v4_interp4_vpp_shuf]
+    pshufb          m0, m0, m2
+    pshufb          m1, m1, m2
+    pmaddubsw       m0, [r5]
+    pmaddubsw       m1, [r5 + mmsize]
+    paddw           m0, m1                                  ; m0 = WORD ROW[3 2 1 0]
+%ifidn %1,pp
+    pmulhrsw        m0, [pw_512]
+    vextracti128    xm1, m0, 1
+    packuswb        xm0, xm1
+    lea             r5, [r3 * 3]
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    pextrd          [r2 + r3 * 2], xm0, 2
+    pextrd          [r2 + r5], xm0, 3
+%else
+    add             r3d, r3d
+    psubw           m0, [pw_2000]
+    vextracti128    xm1, m0, 1
+    lea             r5, [r3 * 3]
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm1
+    movhps          [r2 + r5], xm1
+%endif
+    RET
+%endmacro
+    FILTER_VER_CHROMA_AVX2_4x4 pp
+    FILTER_VER_CHROMA_AVX2_4x4 ps
+
+%macro FILTER_VER_CHROMA_AVX2_4x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x8, 4, 6, 5
+    mov             r4d, r4m
+    shl             r4d, 6
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+
+    movd            xm1, [r0]
+    pinsrd          xm1, [r0 + r1], 1
+    pinsrd          xm1, [r0 + r1 * 2], 2
+    pinsrd          xm1, [r0 + r4], 3                       ; m1 = row[3 2 1 0]
+    lea             r0, [r0 + r1 * 4]
+    movd            xm2, [r0]
+    pinsrd          xm2, [r0 + r1], 1
+    pinsrd          xm2, [r0 + r1 * 2], 2
+    pinsrd          xm2, [r0 + r4], 3                       ; m2 = row[7 6 5 4]
+    vinserti128     m1, m1, xm2, 1                          ; m1 = row[7 6 5 4 3 2 1 0]
+    lea             r0, [r0 + r1 * 4]
+    movd            xm3, [r0]
+    pinsrd          xm3, [r0 + r1], 1
+    pinsrd          xm3, [r0 + r1 * 2], 2                   ; m3 = row[x 10 9 8]
+    vinserti128     m2, m2, xm3, 1                          ; m2 = row[x 10 9 8 7 6 5 4]
+    mova            m3, [v4_interp4_vpp_shuf1]
+    vpermd          m0, m3, m1                              ; m0 = row[4 3 3 2 2 1 1 0]
+    vpermd          m4, m3, m2                              ; m4 = row[8 7 7 6 6 5 5 4]
+    mova            m3, [v4_interp4_vpp_shuf1 + mmsize]
+    vpermd          m1, m3, m1                              ; m1 = row[6 5 5 4 4 3 3 2]
+    vpermd          m2, m3, m2                              ; m2 = row[10 9 9 8 8 7 7 6]
+
+    mova            m3, [v4_interp4_vpp_shuf]
+    pshufb          m0, m0, m3
+    pshufb          m1, m1, m3
+    pshufb          m2, m2, m3
+    pshufb          m4, m4, m3
+    pmaddubsw       m0, [r5]
+    pmaddubsw       m4, [r5]
+    pmaddubsw       m1, [r5 + mmsize]
+    pmaddubsw       m2, [r5 + mmsize]
+    paddw           m0, m1                                  ; m0 = WORD ROW[3 2 1 0]
+    paddw           m4, m2                                  ; m4 = WORD ROW[7 6 5 4]
+%ifidn %1,pp
+    pmulhrsw        m0, [pw_512]
+    pmulhrsw        m4, [pw_512]
+    packuswb        m0, m4
+    vextracti128    xm1, m0, 1
+    lea             r5, [r3 * 3]
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    movd            [r2 + r3 * 2], xm1
+    pextrd          [r2 + r5], xm1, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm0, 2
+    pextrd          [r2 + r3], xm0, 3
+    pextrd          [r2 + r3 * 2], xm1, 2
+    pextrd          [r2 + r5], xm1, 3
+%else
+    add             r3d, r3d
+    psubw           m0, [pw_2000]
+    psubw           m4, [pw_2000]
+    vextracti128    xm1, m0, 1
+    vextracti128    xm2, m4, 1
+    lea             r5, [r3 * 3]
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm1
+    movhps          [r2 + r5], xm1
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm4
+    movhps          [r2 + r3], xm4
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r5], xm2
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_4x8 pp
+    FILTER_VER_CHROMA_AVX2_4x8 ps
+
+%macro FILTER_VER_CHROMA_AVX2_4xN 2
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x%2, 4, 6, 12
+    mov             r4d, r4m
+    shl             r4d, 6
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    mova            m10, [r5]
+    mova            m11, [r5 + mmsize]
+%ifidn %1,pp
+    mova            m9, [pw_512]
+%else
+    add             r3d, r3d
+    mova            m9, [pw_2000]
+%endif
+    lea             r5, [r3 * 3]
+%rep %2 / 16
+    movd            xm1, [r0]
+    pinsrd          xm1, [r0 + r1], 1
+    pinsrd          xm1, [r0 + r1 * 2], 2
+    pinsrd          xm1, [r0 + r4], 3                       ; m1 = row[3 2 1 0]
+    lea             r0, [r0 + r1 * 4]
+    movd            xm2, [r0]
+    pinsrd          xm2, [r0 + r1], 1
+    pinsrd          xm2, [r0 + r1 * 2], 2
+    pinsrd          xm2, [r0 + r4], 3                       ; m2 = row[7 6 5 4]
+    vinserti128     m1, m1, xm2, 1                          ; m1 = row[7 6 5 4 3 2 1 0]
+    lea             r0, [r0 + r1 * 4]
+    movd            xm3, [r0]
+    pinsrd          xm3, [r0 + r1], 1
+    pinsrd          xm3, [r0 + r1 * 2], 2
+    pinsrd          xm3, [r0 + r4], 3                       ; m3 = row[11 10 9 8]
+    vinserti128     m2, m2, xm3, 1                          ; m2 = row[11 10 9 8 7 6 5 4]
+    lea             r0, [r0 + r1 * 4]
+    movd            xm4, [r0]
+    pinsrd          xm4, [r0 + r1], 1
+    pinsrd          xm4, [r0 + r1 * 2], 2
+    pinsrd          xm4, [r0 + r4], 3                       ; m4 = row[15 14 13 12]
+    vinserti128     m3, m3, xm4, 1                          ; m3 = row[15 14 13 12 11 10 9 8]
+    lea             r0, [r0 + r1 * 4]
+    movd            xm5, [r0]
+    pinsrd          xm5, [r0 + r1], 1
+    pinsrd          xm5, [r0 + r1 * 2], 2                   ; m5 = row[x 18 17 16]
+    vinserti128     m4, m4, xm5, 1                          ; m4 = row[x 18 17 16 15 14 13 12]
+    mova            m5, [v4_interp4_vpp_shuf1]
+    vpermd          m0, m5, m1                              ; m0 = row[4 3 3 2 2 1 1 0]
+    vpermd          m6, m5, m2                              ; m6 = row[8 7 7 6 6 5 5 4]
+    vpermd          m7, m5, m3                              ; m7 = row[12 11 11 10 10 9 9 8]
+    vpermd          m8, m5, m4                              ; m8 = row[16 15 15 14 14 13 13 12]
+    mova            m5, [v4_interp4_vpp_shuf1 + mmsize]
+    vpermd          m1, m5, m1                              ; m1 = row[6 5 5 4 4 3 3 2]
+    vpermd          m2, m5, m2                              ; m2 = row[10 9 9 8 8 7 7 6]
+    vpermd          m3, m5, m3                              ; m3 = row[14 13 13 12 12 11 11 10]
+    vpermd          m4, m5, m4                              ; m4 = row[18 17 17 16 16 15 15 14]
+
+    mova            m5, [v4_interp4_vpp_shuf]
+    pshufb          m0, m0, m5
+    pshufb          m1, m1, m5
+    pshufb          m2, m2, m5
+    pshufb          m4, m4, m5
+    pshufb          m3, m3, m5
+    pshufb          m6, m6, m5
+    pshufb          m7, m7, m5
+    pshufb          m8, m8, m5
+    pmaddubsw       m0, m10
+    pmaddubsw       m6, m10
+    pmaddubsw       m7, m10
+    pmaddubsw       m8, m10
+    pmaddubsw       m1, m11
+    pmaddubsw       m2, m11
+    pmaddubsw       m3, m11
+    pmaddubsw       m4, m11
+    paddw           m0, m1                                  ; m0 = WORD ROW[3 2 1 0]
+    paddw           m6, m2                                  ; m6 = WORD ROW[7 6 5 4]
+    paddw           m7, m3                                  ; m7 = WORD ROW[11 10 9 8]
+    paddw           m8, m4                                  ; m8 = WORD ROW[15 14 13 12]
+%ifidn %1,pp
+    pmulhrsw        m0, m9
+    pmulhrsw        m6, m9
+    pmulhrsw        m7, m9
+    pmulhrsw        m8, m9
+    packuswb        m0, m6
+    packuswb        m7, m8
+    vextracti128    xm1, m0, 1
+    vextracti128    xm2, m7, 1
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    movd            [r2 + r3 * 2], xm1
+    pextrd          [r2 + r5], xm1, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm0, 2
+    pextrd          [r2 + r3], xm0, 3
+    pextrd          [r2 + r3 * 2], xm1, 2
+    pextrd          [r2 + r5], xm1, 3
+    lea             r2, [r2 + r3 * 4]
+    movd            [r2], xm7
+    pextrd          [r2 + r3], xm7, 1
+    movd            [r2 + r3 * 2], xm2
+    pextrd          [r2 + r5], xm2, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm7, 2
+    pextrd          [r2 + r3], xm7, 3
+    pextrd          [r2 + r3 * 2], xm2, 2
+    pextrd          [r2 + r5], xm2, 3
+%else
+    psubw           m0, m9
+    psubw           m6, m9
+    psubw           m7, m9
+    psubw           m8, m9
+    vextracti128    xm1, m0, 1
+    vextracti128    xm2, m6, 1
+    vextracti128    xm3, m7, 1
+    vextracti128    xm4, m8, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm1
+    movhps          [r2 + r5], xm1
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm6
+    movhps          [r2 + r3], xm6
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r5], xm2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm7
+    movhps          [r2 + r3], xm7
+    movq            [r2 + r3 * 2], xm3
+    movhps          [r2 + r5], xm3
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm8
+    movhps          [r2 + r3], xm8
+    movq            [r2 + r3 * 2], xm4
+    movhps          [r2 + r5], xm4
+%endif
+    lea             r2, [r2 + r3 * 4]
+%endrep
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_4xN pp, 16
+    FILTER_VER_CHROMA_AVX2_4xN ps, 16
+    FILTER_VER_CHROMA_AVX2_4xN pp, 32
+    FILTER_VER_CHROMA_AVX2_4xN ps, 32
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W4_H4 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m0,        [tab_Cm]
+
+    mova        m1,        [pw_512]
+
+    mov         r4d,       %2
+
+    lea         r5,        [3 * r1]
+
+.loop:
+    movd        m2,        [r0]
+    movd        m3,        [r0 + r1]
+    movd        m4,        [r0 + 2 * r1]
+    movd        m5,        [r0 + r5]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklbw   m2,        m6
+
+    pmaddubsw   m2,        m0
+
+    lea         r0,        [r0 + 4 * r1]
+    movd        m6,        [r0]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklbw   m3,        m7
+
+    pmaddubsw   m3,        m0
+
+    phaddw      m2,        m3
+
+    pmulhrsw    m2,        m1
+
+    movd        m7,        [r0 + r1]
+
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklbw   m4,        m3
+
+    pmaddubsw   m4,        m0
+
+    movd        m3,        [r0 + 2 * r1]
+
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklbw   m5,        m7
+
+    pmaddubsw   m5,        m0
+
+    phaddw      m4,        m5
+
+    pmulhrsw    m4,        m1
+    packuswb    m2,        m4
+    movd        [r2],      m2
+    pextrd      [r2 + r3], m2,  1
+    lea         r2,        [r2 + 2 * r3]
+    pextrd      [r2],      m2, 2
+    pextrd      [r2 + r3], m2, 3
+
+    lea         r2,        [r2 + 2 * r3]
+
+    sub         r4,        4
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W4_H4 4,  8
+    FILTER_V4_W4_H4 4, 16
+
+    FILTER_V4_W4_H4 4, 32
+
+%macro FILTER_V4_W8_H2 0
+    punpcklbw   m1,        m2
+    punpcklbw   m7,        m3,        m0
+
+    pmaddubsw   m1,        m6
+    pmaddubsw   m7,        m5
+
+    paddw       m1,        m7
+
+    pmulhrsw    m1,        m4
+    packuswb    m1,        m1
+%endmacro
+
+%macro FILTER_V4_W8_H3 0
+    punpcklbw   m2,        m3
+    punpcklbw   m7,        m0,        m1
+
+    pmaddubsw   m2,        m6
+    pmaddubsw   m7,        m5
+
+    paddw       m2,        m7
+
+    pmulhrsw    m2,        m4
+    packuswb    m2,        m2
+%endmacro
+
+%macro FILTER_V4_W8_H4 0
+    punpcklbw   m3,        m0
+    punpcklbw   m7,        m1,        m2
+
+    pmaddubsw   m3,        m6
+    pmaddubsw   m7,        m5
+
+    paddw       m3,        m7
+
+    pmulhrsw    m3,        m4
+    packuswb    m3,        m3
+%endmacro
+
+%macro FILTER_V4_W8_H5 0
+    punpcklbw   m0,        m1
+    punpcklbw   m7,        m2,        m3
+
+    pmaddubsw   m0,        m6
+    pmaddubsw   m7,        m5
+
+    paddw       m0,        m7
+
+    pmulhrsw    m0,        m4
+    packuswb    m0,        m0
+%endmacro
+
+%macro FILTER_V4_W8_8x2 2
+    FILTER_V4_W8 %1, %2
+    movq        m0,        [r0 + 4 * r1]
+
+    FILTER_V4_W8_H2
+
+    movh        [r2 + r3], m1
+%endmacro
+
+%macro FILTER_V4_W8_8x4 2
+    FILTER_V4_W8_8x2 %1, %2
+;8x3
+    lea         r6,        [r0 + 4 * r1]
+    movq        m1,        [r6 + r1]
+
+    FILTER_V4_W8_H3
+
+    movh        [r2 + 2 * r3], m2
+
+;8x4
+    movq        m2,        [r6 + 2 * r1]
+
+    FILTER_V4_W8_H4
+
+    lea         r5,        [r2 + 2 * r3]
+    movh        [r5 + r3], m3
+%endmacro
+
+%macro FILTER_V4_W8_8x6 2
+    FILTER_V4_W8_8x4 %1, %2
+;8x5
+    lea         r6,        [r6 + 2 * r1]
+    movq        m3,        [r6 + r1]
+
+    FILTER_V4_W8_H5
+
+    movh        [r2 + 4 * r3], m0
+
+;8x6
+    movq        m0,        [r0 + 8 * r1]
+
+    FILTER_V4_W8_H2
+
+    lea         r5,        [r2 + 4 * r3]
+    movh        [r5 + r3], m1
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W8 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_%1x%2, 4, 7, 8
+
+    mov         r4d,       r4m
+
+    sub         r0,        r1
+    movq        m0,        [r0]
+    movq        m1,        [r0 + r1]
+    movq        m2,        [r0 + 2 * r1]
+    lea         r5,        [r0 + 2 * r1]
+    movq        m3,        [r5 + r1]
+
+    punpcklbw   m0,        m1
+    punpcklbw   m4,        m2,          m3
+
+%ifdef PIC
+    lea         r6,        [tab_ChromaCoeff]
+    movd        m5,        [r6 + r4 * 4]
+%else
+    movd        m5,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m6,        m5,       [tab_Vm]
+    pmaddubsw   m0,        m6
+
+    pshufb      m5,        [tab_Vm + 16]
+    pmaddubsw   m4,        m5
+
+    paddw       m0,        m4
+
+    mova        m4,        [pw_512]
+
+    pmulhrsw    m0,        m4
+    packuswb    m0,        m0
+    movh        [r2],      m0
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_8x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+    FILTER_V4_W8_8x2 8, 2
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+    FILTER_V4_W8_8x4 8, 4
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_8x6(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+    FILTER_V4_W8_8x6 8, 6
+
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_4x2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-------------------------------------------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_4x2, 4, 6, 6
+
+    mov         r4d, r4m
+    sub         r0, r1
+    add         r3d, r3d
+
+%ifdef PIC
+    lea         r5, [tab_ChromaCoeff]
+    movd        m0, [r5 + r4 * 4]
+%else
+    movd        m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m0, [tab_Cm]
+
+    movd        m2, [r0]
+    movd        m3, [r0 + r1]
+    lea         r5, [r0 + 2 * r1]
+    movd        m4, [r5]
+    movd        m5, [r5 + r1]
+
+    punpcklbw   m2, m3
+    punpcklbw   m1, m4, m5
+    punpcklbw   m2, m1
+
+    pmaddubsw   m2, m0
+
+    movd        m1, [r0 + 4 * r1]
+
+    punpcklbw   m3, m4
+    punpcklbw   m5, m1
+    punpcklbw   m3, m5
+
+    pmaddubsw   m3, m0
+
+    phaddw      m2, m3
+
+    psubw       m2, [pw_2000]
+    movh        [r2], m2
+    movhps      [r2 + r3], m2
+
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_4x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-------------------------------------------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_4x4, 4, 6, 7
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m0, [r5 + r4 * 4]
+%else
+    movd       m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m0, [tab_Cm]
+
+    lea        r4, [r1 * 3]
+    lea        r5, [r0 + 4 * r1]
+
+    movd       m2, [r0]
+    movd       m3, [r0 + r1]
+    movd       m4, [r0 + 2 * r1]
+    movd       m5, [r0 + r4]
+
+    punpcklbw  m2, m3
+    punpcklbw  m6, m4, m5
+    punpcklbw  m2, m6
+
+    pmaddubsw  m2, m0
+
+    movd       m6, [r5]
+
+    punpcklbw  m3, m4
+    punpcklbw  m1, m5, m6
+    punpcklbw  m3, m1
+
+    pmaddubsw  m3, m0
+
+    phaddw     m2, m3
+
+    mova       m1, [pw_2000]
+
+    psubw      m2, m1
+    movh       [r2], m2
+    movhps     [r2 + r3], m2
+
+    movd       m2, [r5 + r1]
+
+    punpcklbw  m4, m5
+    punpcklbw  m3, m6, m2
+    punpcklbw  m4, m3
+
+    pmaddubsw  m4, m0
+
+    movd       m3, [r5 + 2 * r1]
+
+    punpcklbw  m5, m6
+    punpcklbw  m2, m3
+    punpcklbw  m5, m2
+
+    pmaddubsw  m5, m0
+
+    phaddw     m4, m5
+
+    psubw      m4, m1
+    lea        r2, [r2 + 2 * r3]
+    movh       [r2], m4
+    movhps     [r2 + r3], m4
+
+    RET
+
+;---------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W4_H4 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m0, [r5 + r4 * 4]
+%else
+    movd       m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m0, [tab_Cm]
+
+    mova       m1, [pw_2000]
+
+    mov        r4d, %2/4
+    lea        r5, [3 * r1]
+
+.loop:
+    movd       m2, [r0]
+    movd       m3, [r0 + r1]
+    movd       m4, [r0 + 2 * r1]
+    movd       m5, [r0 + r5]
+
+    punpcklbw  m2, m3
+    punpcklbw  m6, m4, m5
+    punpcklbw  m2, m6
+
+    pmaddubsw  m2, m0
+
+    lea        r0, [r0 + 4 * r1]
+    movd       m6, [r0]
+
+    punpcklbw  m3, m4
+    punpcklbw  m7, m5, m6
+    punpcklbw  m3, m7
+
+    pmaddubsw  m3, m0
+
+    phaddw     m2, m3
+
+    psubw      m2, m1
+    movh       [r2], m2
+    movhps     [r2 + r3], m2
+
+    movd       m2, [r0 + r1]
+
+    punpcklbw  m4, m5
+    punpcklbw  m3, m6, m2
+    punpcklbw  m4, m3
+
+    pmaddubsw  m4, m0
+
+    movd       m3, [r0 + 2 * r1]
+
+    punpcklbw  m5, m6
+    punpcklbw  m2, m3
+    punpcklbw  m5, m2
+
+    pmaddubsw  m5, m0
+
+    phaddw     m4, m5
+
+    psubw      m4, m1
+    lea        r2, [r2 + 2 * r3]
+    movh       [r2], m4
+    movhps     [r2 + r3], m4
+
+    lea        r2, [r2 + 2 * r3]
+
+    dec        r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V_PS_W4_H4 4, 8
+    FILTER_V_PS_W4_H4 4, 16
+
+    FILTER_V_PS_W4_H4 4, 32
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W8_H8_H16_H2 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 7
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m5, [r5 + r4 * 4]
+%else
+    movd       m5, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m6, m5, [tab_Vm]
+    pshufb     m5, [tab_Vm + 16]
+    mova       m4, [pw_2000]
+
+    mov        r4d, %2/2
+    lea        r5, [3 * r1]
+
+.loopH:
+    movq       m0, [r0]
+    movq       m1, [r0 + r1]
+    movq       m2, [r0 + 2 * r1]
+    movq       m3, [r0 + r5]
+
+    punpcklbw  m0, m1
+    punpcklbw  m1, m2
+    punpcklbw  m2, m3
+
+    pmaddubsw  m0, m6
+    pmaddubsw  m2, m5
+
+    paddw      m0, m2
+
+    psubw      m0, m4
+    movu       [r2], m0
+
+    movq       m0, [r0 + 4 * r1]
+
+    punpcklbw  m3, m0
+
+    pmaddubsw  m1, m6
+    pmaddubsw  m3, m5
+
+    paddw      m1, m3
+    psubw      m1, m4
+
+    movu       [r2 + r3], m1
+
+    lea        r0, [r0 + 2 * r1]
+    lea        r2, [r2 + 2 * r3]
+
+    dec        r4d
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_V_PS_W8_H8_H16_H2 8, 2
+    FILTER_V_PS_W8_H8_H16_H2 8, 4
+    FILTER_V_PS_W8_H8_H16_H2 8, 6
+
+    FILTER_V_PS_W8_H8_H16_H2 8, 12
+    FILTER_V_PS_W8_H8_H16_H2 8, 64
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_8x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W8_H8_H16_H32 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m5, [r5 + r4 * 4]
+%else
+    movd       m5, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m6, m5, [tab_Vm]
+    pshufb     m5, [tab_Vm + 16]
+    mova       m4, [pw_2000]
+
+    mov        r4d, %2/4
+    lea        r5, [3 * r1]
+
+.loop:
+    movq       m0, [r0]
+    movq       m1, [r0 + r1]
+    movq       m2, [r0 + 2 * r1]
+    movq       m3, [r0 + r5]
+
+    punpcklbw  m0, m1
+    punpcklbw  m1, m2
+    punpcklbw  m2, m3
+
+    pmaddubsw  m0, m6
+    pmaddubsw  m7, m2, m5
+
+    paddw      m0, m7
+
+    psubw       m0, m4
+    movu       [r2], m0
+
+    lea        r0, [r0 + 4 * r1]
+    movq       m0, [r0]
+
+    punpcklbw  m3, m0
+
+    pmaddubsw  m1, m6
+    pmaddubsw  m7, m3, m5
+
+    paddw      m1, m7
+
+    psubw      m1, m4
+    movu       [r2 + r3], m1
+
+    movq       m1, [r0 + r1]
+
+    punpcklbw  m0, m1
+
+    pmaddubsw  m2, m6
+    pmaddubsw  m0, m5
+
+    paddw      m2, m0
+
+    psubw      m2, m4
+    lea        r2, [r2 + 2 * r3]
+    movu       [r2], m2
+
+    movq       m2, [r0 + 2 * r1]
+
+    punpcklbw  m1, m2
+
+    pmaddubsw  m3, m6
+    pmaddubsw  m1, m5
+
+    paddw      m3, m1
+    psubw      m3, m4
+
+    movu       [r2 + r3], m3
+
+    lea        r2, [r2 + 2 * r3]
+
+    dec        r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V_PS_W8_H8_H16_H32 8,  8
+    FILTER_V_PS_W8_H8_H16_H32 8, 16
+    FILTER_V_PS_W8_H8_H16_H32 8, 32
+
+;------------------------------------------------------------------------------------------------------------
+;void interp_4tap_vert_ps_6x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W6 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_6x%2, 4, 6, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m5, [r5 + r4 * 4]
+%else
+    movd       m5, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m6, m5, [tab_Vm]
+    pshufb     m5, [tab_Vm + 16]
+    mova       m4, [pw_2000]
+    lea        r5, [3 * r1]
+    mov        r4d, %2/4
+
+.loop:
+    movq       m0, [r0]
+    movq       m1, [r0 + r1]
+    movq       m2, [r0 + 2 * r1]
+    movq       m3, [r0 + r5]
+
+    punpcklbw  m0, m1
+    punpcklbw  m1, m2
+    punpcklbw  m2, m3
+
+    pmaddubsw  m0, m6
+    pmaddubsw  m7, m2, m5
+
+    paddw      m0, m7
+    psubw      m0, m4
+
+    movh       [r2], m0
+    pshufd     m0, m0, 2
+    movd       [r2 + 8], m0
+
+    lea        r0, [r0 + 4 * r1]
+    movq       m0, [r0]
+    punpcklbw  m3, m0
+
+    pmaddubsw  m1, m6
+    pmaddubsw  m7, m3, m5
+
+    paddw      m1, m7
+    psubw      m1, m4
+
+    movh       [r2 + r3], m1
+    pshufd     m1, m1, 2
+    movd       [r2 + r3 + 8], m1
+
+    movq       m1, [r0 + r1]
+    punpcklbw  m0, m1
+
+    pmaddubsw  m2, m6
+    pmaddubsw  m0, m5
+
+    paddw      m2, m0
+    psubw      m2, m4
+
+    lea        r2,[r2 + 2 * r3]
+    movh       [r2], m2
+    pshufd     m2, m2, 2
+    movd       [r2 + 8], m2
+
+    movq       m2,[r0 + 2 * r1]
+    punpcklbw  m1, m2
+
+    pmaddubsw  m3, m6
+    pmaddubsw  m1, m5
+
+    paddw      m3, m1
+    psubw      m3, m4
+
+    movh       [r2 + r3], m3
+    pshufd     m3, m3, 2
+    movd       [r2 + r3 + 8], m3
+
+    lea        r2, [r2 + 2 * r3]
+
+    dec        r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V_PS_W6 6, 8
+    FILTER_V_PS_W6 6, 16
+
+;---------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_12x16(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W12 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_12x%2, 4, 6, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m0, [r5 + r4 * 4]
+%else
+    movd       m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m1, m0, [tab_Vm]
+    pshufb     m0, [tab_Vm + 16]
+
+    mov        r4d, %2/2
+
+.loop:
+    movu       m2, [r0]
+    movu       m3, [r0 + r1]
+
+    punpcklbw  m4, m2, m3
+    punpckhbw  m2, m3
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m2, m1
+
+    lea        r0, [r0 + 2 * r1]
+    movu       m5, [r0]
+    movu       m7, [r0 + r1]
+
+    punpcklbw  m6, m5, m7
+    pmaddubsw  m6, m0
+    paddw      m4, m6
+
+    punpckhbw  m6, m5, m7
+    pmaddubsw  m6, m0
+    paddw      m2, m6
+
+    mova       m6, [pw_2000]
+
+    psubw      m4, m6
+    psubw      m2, m6
+
+    movu       [r2], m4
+    movh       [r2 + 16], m2
+
+    punpcklbw  m4, m3, m5
+    punpckhbw  m3, m5
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m3, m1
+
+    movu       m2, [r0 + 2 * r1]
+
+    punpcklbw  m5, m7, m2
+    punpckhbw  m7, m2
+
+    pmaddubsw  m5, m0
+    pmaddubsw  m7, m0
+
+    paddw      m4, m5
+    paddw      m3, m7
+
+    psubw      m4, m6
+    psubw      m3, m6
+
+    movu       [r2 + r3], m4
+    movh       [r2 + r3 + 16], m3
+
+    lea        r2, [r2 + 2 * r3]
+
+    dec        r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V_PS_W12 12, 16
+    FILTER_V_PS_W12 12, 32
+
+;---------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_16x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W16 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m0, [r5 + r4 * 4]
+%else
+    movd       m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m1, m0, [tab_Vm]
+    pshufb     m0, [tab_Vm + 16]
+    mov        r4d, %2/2
+
+.loop:
+    movu       m2, [r0]
+    movu       m3, [r0 + r1]
+
+    punpcklbw  m4, m2, m3
+    punpckhbw  m2, m3
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m2, m1
+
+    lea        r0, [r0 + 2 * r1]
+    movu       m5, [r0]
+    movu       m7, [r0 + r1]
+
+    punpcklbw  m6, m5, m7
+    pmaddubsw  m6, m0
+    paddw      m4, m6
+
+    punpckhbw  m6, m5, m7
+    pmaddubsw  m6, m0
+    paddw      m2, m6
+
+    mova       m6, [pw_2000]
+
+    psubw      m4, m6
+    psubw      m2, m6
+
+    movu       [r2], m4
+    movu       [r2 + 16], m2
+
+    punpcklbw  m4, m3, m5
+    punpckhbw  m3, m5
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m3, m1
+
+    movu       m5, [r0 + 2 * r1]
+
+    punpcklbw  m2, m7, m5
+    punpckhbw  m7, m5
+
+    pmaddubsw  m2, m0
+    pmaddubsw  m7, m0
+
+    paddw      m4, m2
+    paddw      m3, m7
+
+    psubw      m4, m6
+    psubw      m3, m6
+
+    movu       [r2 + r3], m4
+    movu       [r2 + r3 + 16], m3
+
+    lea        r2, [r2 + 2 * r3]
+
+    dec        r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V_PS_W16 16,  4
+    FILTER_V_PS_W16 16,  8
+    FILTER_V_PS_W16 16, 12
+    FILTER_V_PS_W16 16, 16
+    FILTER_V_PS_W16 16, 32
+
+    FILTER_V_PS_W16 16, 24
+    FILTER_V_PS_W16 16, 64
+
+;--------------------------------------------------------------------------------------------------------------
+;void interp_4tap_vert_ps_24x32(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro FILTER_V4_PS_W24 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_24x%2, 4, 6, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m0, [r5 + r4 * 4]
+%else
+    movd       m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m1, m0, [tab_Vm]
+    pshufb     m0, [tab_Vm + 16]
+
+    mov        r4d, %2/2
+
+.loop:
+    movu       m2, [r0]
+    movu       m3, [r0 + r1]
+
+    punpcklbw  m4, m2, m3
+    punpckhbw  m2, m3
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m2, m1
+
+    lea        r5, [r0 + 2 * r1]
+
+    movu       m5, [r5]
+    movu       m7, [r5 + r1]
+
+    punpcklbw  m6, m5, m7
+    pmaddubsw  m6, m0
+    paddw      m4, m6
+
+    punpckhbw  m6, m5, m7
+    pmaddubsw  m6, m0
+    paddw      m2, m6
+
+    mova       m6, [pw_2000]
+
+    psubw      m4, m6
+    psubw      m2, m6
+
+    movu       [r2], m4
+    movu       [r2 + 16], m2
+
+    punpcklbw  m4, m3, m5
+    punpckhbw  m3, m5
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m3, m1
+
+    movu       m2, [r5 + 2 * r1]
+
+    punpcklbw  m5, m7, m2
+    punpckhbw  m7, m2
+
+    pmaddubsw  m5, m0
+    pmaddubsw  m7, m0
+
+    paddw      m4, m5
+    paddw      m3, m7
+
+    psubw      m4, m6
+    psubw      m3, m6
+
+    movu       [r2 + r3], m4
+    movu       [r2 + r3 + 16], m3
+
+    movq       m2, [r0 + 16]
+    movq       m3, [r0 + r1 + 16]
+    movq       m4, [r5 + 16]
+    movq       m5, [r5 + r1 + 16]
+
+    punpcklbw  m2, m3
+    punpcklbw  m7, m4, m5
+
+    pmaddubsw  m2, m1
+    pmaddubsw  m7, m0
+
+    paddw      m2, m7
+    psubw      m2, m6
+
+    movu       [r2 + 32], m2
+
+    movq       m2, [r5 + 2 * r1 + 16]
+
+    punpcklbw  m3, m4
+    punpcklbw  m5, m2
+
+    pmaddubsw  m3, m1
+    pmaddubsw  m5, m0
+
+    paddw      m3, m5
+    psubw      m3,  m6
+
+    movu       [r2 + r3 + 32], m3
+
+    mov        r0, r5
+    lea        r2, [r2 + 2 * r3]
+
+    dec        r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_PS_W24 24, 32
+
+    FILTER_V4_PS_W24 24, 64
+
+;---------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_32x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W32 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_%1x%2, 4, 6, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m0, [r5 + r4 * 4]
+%else
+    movd       m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m1, m0, [tab_Vm]
+    pshufb     m0, [tab_Vm + 16]
+
+    mova       m7, [pw_2000]
+
+    mov        r4d, %2
+
+.loop:
+    movu       m2, [r0]
+    movu       m3, [r0 + r1]
+
+    punpcklbw  m4, m2, m3
+    punpckhbw  m2, m3
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m2, m1
+
+    lea        r5, [r0 + 2 * r1]
+    movu       m3, [r5]
+    movu       m5, [r5 + r1]
+
+    punpcklbw  m6, m3, m5
+    punpckhbw  m3, m5
+
+    pmaddubsw  m6, m0
+    pmaddubsw  m3, m0
+
+    paddw      m4, m6
+    paddw      m2, m3
+
+    psubw      m4, m7
+    psubw      m2, m7
+
+    movu       [r2], m4
+    movu       [r2 + 16], m2
+
+    movu       m2, [r0 + 16]
+    movu       m3, [r0 + r1 + 16]
+
+    punpcklbw  m4, m2, m3
+    punpckhbw  m2, m3
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m2, m1
+
+    movu       m3, [r5 + 16]
+    movu       m5, [r5 + r1 + 16]
+
+    punpcklbw  m6, m3, m5
+    punpckhbw  m3, m5
+
+    pmaddubsw  m6, m0
+    pmaddubsw  m3, m0
+
+    paddw      m4, m6
+    paddw      m2, m3
+
+    psubw      m4, m7
+    psubw      m2, m7
+
+    movu       [r2 + 32], m4
+    movu       [r2 + 48], m2
+
+    lea        r0, [r0 + r1]
+    lea        r2, [r2 + r3]
+
+    dec        r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V_PS_W32 32,  8
+    FILTER_V_PS_W32 32, 16
+    FILTER_V_PS_W32 32, 24
+    FILTER_V_PS_W32 32, 32
+
+    FILTER_V_PS_W32 32, 48
+    FILTER_V_PS_W32 32, 64
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W8_H8_H16_H32 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m5,        [r5 + r4 * 4]
+%else
+    movd        m5,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m6,        m5,       [tab_Vm]
+    pshufb      m5,        [tab_Vm + 16]
+    mova        m4,        [pw_512]
+    lea         r5,        [r1 * 3]
+
+    mov         r4d,       %2
+
+.loop:
+    movq        m0,        [r0]
+    movq        m1,        [r0 + r1]
+    movq        m2,        [r0 + 2 * r1]
+    movq        m3,        [r0 + r5]
+
+    punpcklbw   m0,        m1
+    punpcklbw   m1,        m2
+    punpcklbw   m2,        m3
+
+    pmaddubsw   m0,        m6
+    pmaddubsw   m7,        m2, m5
+
+    paddw       m0,        m7
+
+    pmulhrsw    m0,        m4
+    packuswb    m0,        m0
+    movh        [r2],      m0
+
+    lea         r0,        [r0 + 4 * r1]
+    movq        m0,        [r0]
+
+    punpcklbw   m3,        m0
+
+    pmaddubsw   m1,        m6
+    pmaddubsw   m7,        m3, m5
+
+    paddw       m1,        m7
+
+    pmulhrsw    m1,        m4
+    packuswb    m1,        m1
+    movh        [r2 + r3], m1
+
+    movq        m1,        [r0 + r1]
+
+    punpcklbw   m0,        m1
+
+    pmaddubsw   m2,        m6
+    pmaddubsw   m0,        m5
+
+    paddw       m2,        m0
+
+    pmulhrsw    m2,        m4
+
+    movq        m7,        [r0 + 2 * r1]
+    punpcklbw   m1,        m7
+
+    pmaddubsw   m3,        m6
+    pmaddubsw   m1,        m5
+
+    paddw       m3,        m1
+
+    pmulhrsw    m3,        m4
+    packuswb    m2,        m3
+
+    lea         r2,        [r2 + 2 * r3]
+    movh        [r2],      m2
+    movhps      [r2 + r3], m2
+
+    lea         r2,        [r2 + 2 * r3]
+
+    sub         r4,         4
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W8_H8_H16_H32 8,  8
+    FILTER_V4_W8_H8_H16_H32 8, 16
+    FILTER_V4_W8_H8_H16_H32 8, 32
+
+    FILTER_V4_W8_H8_H16_H32 8, 12
+    FILTER_V4_W8_H8_H16_H32 8, 64
+
+%macro PROCESS_CHROMA_AVX2_W8_8R 0
+    movq            xm1, [r0]                       ; m1 = row 0
+    movq            xm2, [r0 + r1]                  ; m2 = row 1
+    punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
+    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
+    punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
+    vinserti128     m5, m1, xm2, 1                  ; m5 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
+    pmaddubsw       m5, [r5]
+    movq            xm4, [r0 + r4]                  ; m4 = row 3
+    punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
+    lea             r0, [r0 + r1 * 4]
+    movq            xm1, [r0]                       ; m1 = row 4
+    punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
+    vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
+    pmaddubsw       m0, m2, [r5 + 1 * mmsize]
+    paddw           m5, m0
+    pmaddubsw       m2, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 5
+    punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
+    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
+    punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
+    vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
+    pmaddubsw       m0, m1, [r5 + 1 * mmsize]
+    paddw           m2, m0
+    pmaddubsw       m1, [r5]
+    movq            xm3, [r0 + r4]                  ; m3 = row 7
+    punpcklbw       xm4, xm3                        ; m4 = [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
+    lea             r0, [r0 + r1 * 4]
+    movq            xm0, [r0]                       ; m0 = row 8
+    punpcklbw       xm3, xm0                        ; m3 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70]
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70] - [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
+    pmaddubsw       m3, m4, [r5 + 1 * mmsize]
+    paddw           m1, m3
+    pmaddubsw       m4, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 9
+    punpcklbw       xm0, xm3                        ; m0 = [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
+    movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
+    punpcklbw       xm3, xm6                        ; m3 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90]
+    vinserti128     m0, m0, xm3, 1                  ; m0 = [A7 97 A6 96 A5 95 A4 94 A3 93 A2 92 A1 91 A0 90] - [97 87 96 86 95 85 94 84 93 83 92 82 91 81 90 80]
+    pmaddubsw       m0, [r5 + 1 * mmsize]
+    paddw           m4, m0
+%endmacro
+
+%macro FILTER_VER_CHROMA_AVX2_8x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x8, 4, 6, 7
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+    PROCESS_CHROMA_AVX2_W8_8R
+%ifidn %1,pp
+    lea             r4, [r3 * 3]
+    mova            m3, [pw_512]
+    pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
+    pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
+    pmulhrsw        m1, m3                          ; m1 = word: row 4, row 5
+    pmulhrsw        m4, m3                          ; m4 = word: row 6, row 7
+    packuswb        m5, m2
+    packuswb        m1, m4
+    vextracti128    xm2, m5, 1
+    vextracti128    xm4, m1, 1
+    movq            [r2], xm5
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm5
+    movhps          [r2 + r4], xm2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm1
+    movq            [r2 + r3], xm4
+    movhps          [r2 + r3 * 2], xm1
+    movhps          [r2 + r4], xm4
+%else
+    add             r3d, r3d
+    vbroadcasti128  m3, [pw_2000]
+    lea             r4, [r3 * 3]
+    psubw           m5, m3                          ; m5 = word: row 0, row 1
+    psubw           m2, m3                          ; m2 = word: row 2, row 3
+    psubw           m1, m3                          ; m1 = word: row 4, row 5
+    psubw           m4, m3                          ; m4 = word: row 6, row 7
+    vextracti128    xm6, m5, 1
+    vextracti128    xm3, m2, 1
+    vextracti128    xm0, m1, 1
+    movu            [r2], xm5
+    movu            [r2 + r3], xm6
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r4], xm3
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm1
+    movu            [r2 + r3], xm0
+    movu            [r2 + r3 * 2], xm4
+    vextracti128    xm4, m4, 1
+    movu            [r2 + r4], xm4
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_8x8 pp
+    FILTER_VER_CHROMA_AVX2_8x8 ps
+
+%macro FILTER_VER_CHROMA_AVX2_8x6 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x6, 4, 6, 6
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+
+    movq            xm1, [r0]                       ; m1 = row 0
+    movq            xm2, [r0 + r1]                  ; m2 = row 1
+    punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
+    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
+    punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
+    vinserti128     m5, m1, xm2, 1                  ; m5 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
+    pmaddubsw       m5, [r5]
+    movq            xm4, [r0 + r4]                  ; m4 = row 3
+    punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
+    lea             r0, [r0 + r1 * 4]
+    movq            xm1, [r0]                       ; m1 = row 4
+    punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
+    vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
+    pmaddubsw       m0, m2, [r5 + 1 * mmsize]
+    paddw           m5, m0
+    pmaddubsw       m2, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 5
+    punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
+    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
+    punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
+    vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
+    pmaddubsw       m0, m1, [r5 + 1 * mmsize]
+    paddw           m2, m0
+    pmaddubsw       m1, [r5]
+    movq            xm3, [r0 + r4]                  ; m3 = row 7
+    punpcklbw       xm4, xm3                        ; m4 = [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
+    lea             r0, [r0 + r1 * 4]
+    movq            xm0, [r0]                       ; m0 = row 8
+    punpcklbw       xm3, xm0                        ; m3 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70]
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [87 77 86 76 85 75 84 74 83 73 82 72 81 71 80 70] - [77 67 76 66 75 65 74 64 73 63 72 62 71 61 70 60]
+    pmaddubsw       m4, [r5 + 1 * mmsize]
+    paddw           m1, m4
+%ifidn %1,pp
+    lea             r4, [r3 * 3]
+    mova            m3, [pw_512]
+    pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
+    pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
+    pmulhrsw        m1, m3                          ; m1 = word: row 4, row 5
+    packuswb        m5, m2
+    packuswb        m1, m1
+    vextracti128    xm2, m5, 1
+    vextracti128    xm4, m1, 1
+    movq            [r2], xm5
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm5
+    movhps          [r2 + r4], xm2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm1
+    movq            [r2 + r3], xm4
+%else
+    add             r3d, r3d
+    mova            m3, [pw_2000]
+    lea             r4, [r3 * 3]
+    psubw           m5, m3                          ; m5 = word: row 0, row 1
+    psubw           m2, m3                          ; m2 = word: row 2, row 3
+    psubw           m1, m3                          ; m1 = word: row 4, row 5
+    vextracti128    xm4, m5, 1
+    vextracti128    xm3, m2, 1
+    vextracti128    xm0, m1, 1
+    movu            [r2], xm5
+    movu            [r2 + r3], xm4
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r4], xm3
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm1
+    movu            [r2 + r3], xm0
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_8x6 pp
+    FILTER_VER_CHROMA_AVX2_8x6 ps
+
+%macro PROCESS_CHROMA_AVX2_W8_16R 1
+    movq            xm1, [r0]                       ; m1 = row 0
+    movq            xm2, [r0 + r1]                  ; m2 = row 1
+    punpcklbw       xm1, xm2
+    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
+    punpcklbw       xm2, xm3
+    vinserti128     m5, m1, xm2, 1
+    pmaddubsw       m5, [r5]
+    movq            xm4, [r0 + r4]                  ; m4 = row 3
+    punpcklbw       xm3, xm4
+    lea             r0, [r0 + r1 * 4]
+    movq            xm1, [r0]                       ; m1 = row 4
+    punpcklbw       xm4, xm1
+    vinserti128     m2, m3, xm4, 1
+    pmaddubsw       m0, m2, [r5 + 1 * mmsize]
+    paddw           m5, m0
+    pmaddubsw       m2, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 5
+    punpcklbw       xm1, xm3
+    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
+    punpcklbw       xm3, xm4
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m0, m1, [r5 + 1 * mmsize]
+    paddw           m2, m0
+    pmaddubsw       m1, [r5]
+    movq            xm3, [r0 + r4]                  ; m3 = row 7
+    punpcklbw       xm4, xm3
+    lea             r0, [r0 + r1 * 4]
+    movq            xm0, [r0]                       ; m0 = row 8
+    punpcklbw       xm3, xm0
+    vinserti128     m4, m4, xm3, 1
+    pmaddubsw       m3, m4, [r5 + 1 * mmsize]
+    paddw           m1, m3
+    pmaddubsw       m4, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 9
+    punpcklbw       xm0, xm3
+    movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
+    punpcklbw       xm3, xm6
+    vinserti128     m0, m0, xm3, 1
+    pmaddubsw       m3, m0, [r5 + 1 * mmsize]
+    paddw           m4, m3
+    pmaddubsw       m0, [r5]
+%ifidn %1,pp
+    pmulhrsw        m5, m7                          ; m5 = word: row 0, row 1
+    pmulhrsw        m2, m7                          ; m2 = word: row 2, row 3
+    pmulhrsw        m1, m7                          ; m1 = word: row 4, row 5
+    pmulhrsw        m4, m7                          ; m4 = word: row 6, row 7
+    packuswb        m5, m2
+    packuswb        m1, m4
+    vextracti128    xm2, m5, 1
+    vextracti128    xm4, m1, 1
+    movq            [r2], xm5
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm5
+    movhps          [r2 + r6], xm2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm1
+    movq            [r2 + r3], xm4
+    movhps          [r2 + r3 * 2], xm1
+    movhps          [r2 + r6], xm4
+%else
+    psubw           m5, m7                          ; m5 = word: row 0, row 1
+    psubw           m2, m7                          ; m2 = word: row 2, row 3
+    psubw           m1, m7                          ; m1 = word: row 4, row 5
+    psubw           m4, m7                          ; m4 = word: row 6, row 7
+    vextracti128    xm3, m5, 1
+    movu            [r2], xm5
+    movu            [r2 + r3], xm3
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+    lea             r2, [r2 + r3 * 4]
+    vextracti128    xm5, m1, 1
+    vextracti128    xm3, m4, 1
+    movu            [r2], xm1
+    movu            [r2 + r3], xm5
+    movu            [r2 + r3 * 2], xm4
+    movu            [r2 + r6], xm3
+%endif
+    movq            xm3, [r0 + r4]                  ; m3 = row 11
+    punpcklbw       xm6, xm3
+    lea             r0, [r0 + r1 * 4]
+    movq            xm5, [r0]                       ; m5 = row 12
+    punpcklbw       xm3, xm5
+    vinserti128     m6, m6, xm3, 1
+    pmaddubsw       m3, m6, [r5 + 1 * mmsize]
+    paddw           m0, m3
+    pmaddubsw       m6, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 13
+    punpcklbw       xm5, xm3
+    movq            xm2, [r0 + r1 * 2]              ; m2 = row 14
+    punpcklbw       xm3, xm2
+    vinserti128     m5, m5, xm3, 1
+    pmaddubsw       m3, m5, [r5 + 1 * mmsize]
+    paddw           m6, m3
+    pmaddubsw       m5, [r5]
+    movq            xm3, [r0 + r4]                  ; m3 = row 15
+    punpcklbw       xm2, xm3
+    lea             r0, [r0 + r1 * 4]
+    movq            xm1, [r0]                       ; m1 = row 16
+    punpcklbw       xm3, xm1
+    vinserti128     m2, m2, xm3, 1
+    pmaddubsw       m3, m2, [r5 + 1 * mmsize]
+    paddw           m5, m3
+    pmaddubsw       m2, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 17
+    punpcklbw       xm1, xm3
+    movq            xm4, [r0 + r1 * 2]              ; m4 = row 18
+    punpcklbw       xm3, xm4
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m1, [r5 + 1 * mmsize]
+    paddw           m2, m1
+    lea             r2, [r2 + r3 * 4]
+%ifidn %1,pp
+    pmulhrsw        m0, m7                          ; m0 = word: row 8, row 9
+    pmulhrsw        m6, m7                          ; m6 = word: row 10, row 11
+    pmulhrsw        m5, m7                          ; m5 = word: row 12, row 13
+    pmulhrsw        m2, m7                          ; m2 = word: row 14, row 15
+    packuswb        m0, m6
+    packuswb        m5, m2
+    vextracti128    xm6, m0, 1
+    vextracti128    xm2, m5, 1
+    movq            [r2], xm0
+    movq            [r2 + r3], xm6
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r6], xm6
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm5
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm5
+    movhps          [r2 + r6], xm2
+%else
+    psubw           m0, m7                          ; m0 = word: row 8, row 9
+    psubw           m6, m7                          ; m6 = word: row 10, row 11
+    psubw           m5, m7                          ; m5 = word: row 12, row 13
+    psubw           m2, m7                          ; m2 = word: row 14, row 15
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m6, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r6], xm3
+    lea             r2, [r2 + r3 * 4]
+    vextracti128    xm1, m5, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2], xm5
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+%endif
+%endmacro
+
+%macro FILTER_VER_CHROMA_AVX2_8x16 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x16, 4, 7, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m7, [pw_512]
+%else
+    add             r3d, r3d
+    mova            m7, [pw_2000]
+%endif
+    lea             r6, [r3 * 3]
+    PROCESS_CHROMA_AVX2_W8_16R %1
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_8x16 pp
+    FILTER_VER_CHROMA_AVX2_8x16 ps
+
+%macro FILTER_VER_CHROMA_AVX2_8x12 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x12, 4, 7, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1, pp
+    mova            m7, [pw_512]
+%else
+    add             r3d, r3d
+    mova            m7, [pw_2000]
+%endif
+    lea             r6, [r3 * 3]
+    movq            xm1, [r0]                       ; m1 = row 0
+    movq            xm2, [r0 + r1]                  ; m2 = row 1
+    punpcklbw       xm1, xm2
+    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
+    punpcklbw       xm2, xm3
+    vinserti128     m5, m1, xm2, 1
+    pmaddubsw       m5, [r5]
+    movq            xm4, [r0 + r4]                  ; m4 = row 3
+    punpcklbw       xm3, xm4
+    lea             r0, [r0 + r1 * 4]
+    movq            xm1, [r0]                       ; m1 = row 4
+    punpcklbw       xm4, xm1
+    vinserti128     m2, m3, xm4, 1
+    pmaddubsw       m0, m2, [r5 + 1 * mmsize]
+    paddw           m5, m0
+    pmaddubsw       m2, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 5
+    punpcklbw       xm1, xm3
+    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
+    punpcklbw       xm3, xm4
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m0, m1, [r5 + 1 * mmsize]
+    paddw           m2, m0
+    pmaddubsw       m1, [r5]
+    movq            xm3, [r0 + r4]                  ; m3 = row 7
+    punpcklbw       xm4, xm3
+    lea             r0, [r0 + r1 * 4]
+    movq            xm0, [r0]                       ; m0 = row 8
+    punpcklbw       xm3, xm0
+    vinserti128     m4, m4, xm3, 1
+    pmaddubsw       m3, m4, [r5 + 1 * mmsize]
+    paddw           m1, m3
+    pmaddubsw       m4, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 9
+    punpcklbw       xm0, xm3
+    movq            xm6, [r0 + r1 * 2]              ; m6 = row 10
+    punpcklbw       xm3, xm6
+    vinserti128     m0, m0, xm3, 1
+    pmaddubsw       m3, m0, [r5 + 1 * mmsize]
+    paddw           m4, m3
+    pmaddubsw       m0, [r5]
+%ifidn %1, pp
+    pmulhrsw        m5, m7                          ; m5 = word: row 0, row 1
+    pmulhrsw        m2, m7                          ; m2 = word: row 2, row 3
+    pmulhrsw        m1, m7                          ; m1 = word: row 4, row 5
+    pmulhrsw        m4, m7                          ; m4 = word: row 6, row 7
+    packuswb        m5, m2
+    packuswb        m1, m4
+    vextracti128    xm2, m5, 1
+    vextracti128    xm4, m1, 1
+    movq            [r2], xm5
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm5
+    movhps          [r2 + r6], xm2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm1
+    movq            [r2 + r3], xm4
+    movhps          [r2 + r3 * 2], xm1
+    movhps          [r2 + r6], xm4
+%else
+    psubw           m5, m7                          ; m5 = word: row 0, row 1
+    psubw           m2, m7                          ; m2 = word: row 2, row 3
+    psubw           m1, m7                          ; m1 = word: row 4, row 5
+    psubw           m4, m7                          ; m4 = word: row 6, row 7
+    vextracti128    xm3, m5, 1
+    movu            [r2], xm5
+    movu            [r2 + r3], xm3
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+    lea             r2, [r2 + r3 * 4]
+    vextracti128    xm5, m1, 1
+    vextracti128    xm3, m4, 1
+    movu            [r2], xm1
+    movu            [r2 + r3], xm5
+    movu            [r2 + r3 * 2], xm4
+    movu            [r2 + r6], xm3
+%endif
+    movq            xm3, [r0 + r4]                  ; m3 = row 11
+    punpcklbw       xm6, xm3
+    lea             r0, [r0 + r1 * 4]
+    movq            xm5, [r0]                       ; m5 = row 12
+    punpcklbw       xm3, xm5
+    vinserti128     m6, m6, xm3, 1
+    pmaddubsw       m3, m6, [r5 + 1 * mmsize]
+    paddw           m0, m3
+    pmaddubsw       m6, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 13
+    punpcklbw       xm5, xm3
+    movq            xm2, [r0 + r1 * 2]              ; m2 = row 14
+    punpcklbw       xm3, xm2
+    vinserti128     m5, m5, xm3, 1
+    pmaddubsw       m3, m5, [r5 + 1 * mmsize]
+    paddw           m6, m3
+    lea             r2, [r2 + r3 * 4]
+%ifidn %1, pp
+    pmulhrsw        m0, m7                          ; m0 = word: row 8, row 9
+    pmulhrsw        m6, m7                          ; m6 = word: row 10, row 11
+    packuswb        m0, m6
+    vextracti128    xm6, m0, 1
+    movq            [r2], xm0
+    movq            [r2 + r3], xm6
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r6], xm6
+%else
+    psubw           m0, m7                          ; m0 = word: row 8, row 9
+    psubw           m6, m7                          ; m6 = word: row 10, row 11
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m6, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r6], xm3
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_8x12 pp
+    FILTER_VER_CHROMA_AVX2_8x12 ps
+
+%macro FILTER_VER_CHROMA_AVX2_8xN 2
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x%2, 4, 7, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m7, [pw_512]
+%else
+    add             r3d, r3d
+    mova            m7, [pw_2000]
+%endif
+    lea             r6, [r3 * 3]
+%rep %2 / 16
+    PROCESS_CHROMA_AVX2_W8_16R %1
+    lea             r2, [r2 + r3 * 4]
+%endrep
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_8xN pp, 32
+    FILTER_VER_CHROMA_AVX2_8xN ps, 32
+    FILTER_VER_CHROMA_AVX2_8xN pp, 64
+    FILTER_VER_CHROMA_AVX2_8xN ps, 64
+
+%macro PROCESS_CHROMA_AVX2_W8_4R 0
+    movq            xm1, [r0]                       ; m1 = row 0
+    movq            xm2, [r0 + r1]                  ; m2 = row 1
+    punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
+    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
+    punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
+    vinserti128     m0, m1, xm2, 1                  ; m0 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
+    pmaddubsw       m0, [r5]
+    movq            xm4, [r0 + r4]                  ; m4 = row 3
+    punpcklbw       xm3, xm4                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
+    lea             r0, [r0 + r1 * 4]
+    movq            xm1, [r0]                       ; m1 = row 4
+    punpcklbw       xm4, xm1                        ; m4 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
+    vinserti128     m2, m3, xm4, 1                  ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
+    pmaddubsw       m4, m2, [r5 + 1 * mmsize]
+    paddw           m0, m4
+    pmaddubsw       m2, [r5]
+    movq            xm3, [r0 + r1]                  ; m3 = row 5
+    punpcklbw       xm1, xm3                        ; m1 = [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
+    movq            xm4, [r0 + r1 * 2]              ; m4 = row 6
+    punpcklbw       xm3, xm4                        ; m3 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50]
+    vinserti128     m1, m1, xm3, 1                  ; m1 = [67 57 66 56 65 55 64 54 63 53 62 52 61 51 60 50] - [57 47 56 46 55 45 54 44 53 43 52 42 51 41 50 40]
+    pmaddubsw       m1, [r5 + 1 * mmsize]
+    paddw           m2, m1
+%endmacro
+
+%macro FILTER_VER_CHROMA_AVX2_8x4 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x4, 4, 6, 5
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+    PROCESS_CHROMA_AVX2_W8_4R
+%ifidn %1,pp
+    lea             r4, [r3 * 3]
+    mova            m3, [pw_512]
+    pmulhrsw        m0, m3                          ; m0 = word: row 0, row 1
+    pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
+    packuswb        m0, m2
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r4], xm2
+%else
+    add             r3d, r3d
+    vbroadcasti128  m3, [pw_2000]
+    lea             r4, [r3 * 3]
+    psubw           m0, m3                          ; m0 = word: row 0, row 1
+    psubw           m2, m3                          ; m2 = word: row 2, row 3
+    vextracti128    xm1, m0, 1
+    vextracti128    xm4, m2, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r4], xm4
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_8x4 pp
+    FILTER_VER_CHROMA_AVX2_8x4 ps
+
+%macro FILTER_VER_CHROMA_AVX2_8x2 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x2, 4, 6, 4
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+
+    movq            xm1, [r0]                       ; m1 = row 0
+    movq            xm2, [r0 + r1]                  ; m2 = row 1
+    punpcklbw       xm1, xm2                        ; m1 = [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
+    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
+    punpcklbw       xm2, xm3                        ; m2 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10]
+    vinserti128     m1, m1, xm2, 1                  ; m1 = [27 17 26 16 25 15 24 14 23 13 22 12 21 11 20 10] - [17 07 16 06 15 05 14 04 13 03 12 02 11 01 10 00]
+    pmaddubsw       m1, [r5]
+    movq            xm2, [r0 + r4]                  ; m2 = row 3
+    punpcklbw       xm3, xm2                        ; m3 = [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
+    movq            xm0, [r0 + r1 * 4]              ; m0 = row 4
+    punpcklbw       xm2, xm0                        ; m2 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30]
+    vinserti128     m3, m3, xm2, 1                  ; m3 = [47 37 46 36 45 35 44 34 43 33 42 32 41 31 40 30] - [37 27 36 26 35 25 34 24 33 23 32 22 31 21 30 20]
+    pmaddubsw       m3, [r5 + 1 * mmsize]
+    paddw           m1, m3
+%ifidn %1,pp
+    pmulhrsw        m1, [pw_512]                    ; m1 = word: row 0, row 1
+    packuswb        m1, m1
+    vextracti128    xm0, m1, 1
+    movq            [r2], xm1
+    movq            [r2 + r3], xm0
+%else
+    add             r3d, r3d
+    psubw           m1, [pw_2000]                   ; m1 = word: row 0, row 1
+    vextracti128    xm0, m1, 1
+    movu            [r2], xm1
+    movu            [r2 + r3], xm0
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_8x2 pp
+    FILTER_VER_CHROMA_AVX2_8x2 ps
+
+%macro FILTER_VER_CHROMA_AVX2_6x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_6x8, 4, 6, 7
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+    PROCESS_CHROMA_AVX2_W8_8R
+%ifidn %1,pp
+    lea             r4, [r3 * 3]
+    mova            m3, [pw_512]
+    pmulhrsw        m5, m3                          ; m5 = word: row 0, row 1
+    pmulhrsw        m2, m3                          ; m2 = word: row 2, row 3
+    pmulhrsw        m1, m3                          ; m1 = word: row 4, row 5
+    pmulhrsw        m4, m3                          ; m4 = word: row 6, row 7
+    packuswb        m5, m2
+    packuswb        m1, m4
+    vextracti128    xm2, m5, 1
+    vextracti128    xm4, m1, 1
+    movd            [r2], xm5
+    pextrw          [r2 + 4], xm5, 2
+    movd            [r2 + r3], xm2
+    pextrw          [r2 + r3 + 4], xm2, 2
+    pextrd          [r2 + r3 * 2], xm5, 2
+    pextrw          [r2 + r3 * 2 + 4], xm5, 6
+    pextrd          [r2 + r4], xm2, 2
+    pextrw          [r2 + r4 + 4], xm2, 6
+    lea             r2, [r2 + r3 * 4]
+    movd            [r2], xm1
+    pextrw          [r2 + 4], xm1, 2
+    movd            [r2 + r3], xm4
+    pextrw          [r2 + r3 + 4], xm4, 2
+    pextrd          [r2 + r3 * 2], xm1, 2
+    pextrw          [r2 + r3 * 2 + 4], xm1, 6
+    pextrd          [r2 + r4], xm4, 2
+    pextrw          [r2 + r4 + 4], xm4, 6
+%else
+    add             r3d, r3d
+    vbroadcasti128  m3, [pw_2000]
+    lea             r4, [r3 * 3]
+    psubw           m5, m3                          ; m5 = word: row 0, row 1
+    psubw           m2, m3                          ; m2 = word: row 2, row 3
+    psubw           m1, m3                          ; m1 = word: row 4, row 5
+    psubw           m4, m3                          ; m4 = word: row 6, row 7
+    vextracti128    xm6, m5, 1
+    vextracti128    xm3, m2, 1
+    vextracti128    xm0, m1, 1
+    movq            [r2], xm5
+    pextrd          [r2 + 8], xm5, 2
+    movq            [r2 + r3], xm6
+    pextrd          [r2 + r3 + 8], xm6, 2
+    movq            [r2 + r3 * 2], xm2
+    pextrd          [r2 + r3 * 2 + 8], xm2, 2
+    movq            [r2 + r4], xm3
+    pextrd          [r2 + r4 + 8], xm3, 2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm1
+    pextrd          [r2 + 8], xm1, 2
+    movq            [r2 + r3], xm0
+    pextrd          [r2 + r3 + 8], xm0, 2
+    movq            [r2 + r3 * 2], xm4
+    pextrd          [r2 + r3 * 2 + 8], xm4, 2
+    vextracti128    xm4, m4, 1
+    movq            [r2 + r4], xm4
+    pextrd          [r2 + r4 + 8], xm4, 2
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_6x8 pp
+    FILTER_VER_CHROMA_AVX2_6x8 ps
+
+;-----------------------------------------------------------------------------
+;void interp_4tap_vert_pp_6x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W6_H4 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_6x%2, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m5,        [r5 + r4 * 4]
+%else
+    movd        m5,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m6,        m5,       [tab_Vm]
+    pshufb      m5,        [tab_Vm + 16]
+    mova        m4,        [pw_512]
+
+    mov         r4d,       %2
+    lea         r5,        [3 * r1]
+
+.loop:
+    movq        m0,        [r0]
+    movq        m1,        [r0 + r1]
+    movq        m2,        [r0 + 2 * r1]
+    movq        m3,        [r0 + r5]
+
+    punpcklbw   m0,        m1
+    punpcklbw   m1,        m2
+    punpcklbw   m2,        m3
+
+    pmaddubsw   m0,        m6
+    pmaddubsw   m7,        m2, m5
+
+    paddw       m0,        m7
+
+    pmulhrsw    m0,        m4
+    packuswb    m0,        m0
+    movd        [r2],      m0
+    pextrw      [r2 + 4],  m0,    2
+
+    lea         r0,        [r0 + 4 * r1]
+
+    movq        m0,        [r0]
+    punpcklbw   m3,        m0
+
+    pmaddubsw   m1,        m6
+    pmaddubsw   m7,        m3, m5
+
+    paddw       m1,        m7
+
+    pmulhrsw    m1,        m4
+    packuswb    m1,        m1
+    movd        [r2 + r3],      m1
+    pextrw      [r2 + r3 + 4],  m1,    2
+
+    movq        m1,        [r0 + r1]
+    punpcklbw   m7,        m0,        m1
+
+    pmaddubsw   m2,        m6
+    pmaddubsw   m7,        m5
+
+    paddw       m2,        m7
+
+    pmulhrsw    m2,        m4
+    packuswb    m2,        m2
+    lea         r2,        [r2 + 2 * r3]
+    movd        [r2],      m2
+    pextrw      [r2 + 4],  m2,    2
+
+    movq        m2,        [r0 + 2 * r1]
+    punpcklbw   m1,        m2
+
+    pmaddubsw   m3,        m6
+    pmaddubsw   m1,        m5
+
+    paddw       m3,        m1
+
+    pmulhrsw    m3,        m4
+    packuswb    m3,        m3
+
+    movd        [r2 + r3],        m3
+    pextrw      [r2 + r3 + 4],    m3,    2
+
+    lea         r2,        [r2 + 2 * r3]
+
+    sub         r4,         4
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W6_H4 6, 8
+
+    FILTER_V4_W6_H4 6, 16
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_12x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W12_H2 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_12x%2, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
+
+    mov         r4d,       %2
+
+.loop:
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
+
+    lea         r0,        [r0 + 2 * r1]
+    movu        m5,        [r0]
+    movu        m7,        [r0 + r1]
+
+    punpcklbw   m6,        m5,        m7
+    pmaddubsw   m6,        m0
+    paddw       m4,        m6
+
+    punpckhbw   m6,        m5,        m7
+    pmaddubsw   m6,        m0
+    paddw       m2,        m6
+
+    mova        m6,        [pw_512]
+
+    pmulhrsw    m4,        m6
+    pmulhrsw    m2,        m6
+
+    packuswb    m4,        m2
+
+    movh         [r2],     m4
+    pextrd       [r2 + 8], m4,  2
+
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m3,        m1
+
+    movu        m5,        [r0 + 2 * r1]
+
+    punpcklbw   m2,        m7,        m5
+    punpckhbw   m7,        m5
+
+    pmaddubsw   m2,        m0
+    pmaddubsw   m7,        m0
+
+    paddw       m4,        m2
+    paddw       m3,        m7
+
+    pmulhrsw    m4,        m6
+    pmulhrsw    m3,        m6
+
+    packuswb    m4,        m3
+
+    movh        [r2 + r3],      m4
+    pextrd      [r2 + r3 + 8],  m4,  2
+
+    lea         r2,        [r2 + 2 * r3]
+
+    sub         r4,        2
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W12_H2 12, 16
+
+    FILTER_V4_W12_H2 12, 32
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_16x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W16_H2 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_16x%2, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
+
+    mov         r4d,       %2/2
+
+.loop:
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
+
+    lea         r0,        [r0 + 2 * r1]
+    movu        m5,        [r0]
+    movu        m6,        [r0 + r1]
+
+    punpckhbw   m7,        m5,        m6
+    pmaddubsw   m7,        m0
+    paddw       m2,        m7
+
+    punpcklbw   m7,        m5,        m6
+    pmaddubsw   m7,        m0
+    paddw       m4,        m7
+
+    mova        m7,        [pw_512]
+
+    pmulhrsw    m4,        m7
+    pmulhrsw    m2,        m7
+
+    packuswb    m4,        m2
+
+    movu        [r2],      m4
+
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m3,        m1
+
+    movu        m5,        [r0 + 2 * r1]
+
+    punpcklbw   m2,        m6,        m5
+    punpckhbw   m6,        m5
+
+    pmaddubsw   m2,        m0
+    pmaddubsw   m6,        m0
+
+    paddw       m4,        m2
+    paddw       m3,        m6
+
+    pmulhrsw    m4,        m7
+    pmulhrsw    m3,        m7
+
+    packuswb    m4,        m3
+
+    movu        [r2 + r3],      m4
+
+    lea         r2,        [r2 + 2 * r3]
+
+    dec         r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W16_H2 16,  4
+    FILTER_V4_W16_H2 16,  8
+    FILTER_V4_W16_H2 16, 12
+    FILTER_V4_W16_H2 16, 16
+    FILTER_V4_W16_H2 16, 32
+
+    FILTER_V4_W16_H2 16, 24
+    FILTER_V4_W16_H2 16, 64
+
+%macro FILTER_VER_CHROMA_AVX2_16x16 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_16x16, 4, 6, 15
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    mova            m12, [r5]
+    mova            m13, [r5 + mmsize]
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m14, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m14, [pw_2000]
+%endif
+    lea             r5, [r3 * 3]
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhbw       xm2, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddubsw       m0, m12
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhbw       xm3, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m1, m12
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhbw       xm4, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddubsw       m4, m2, m13
+    paddw           m0, m4
+    pmaddubsw       m2, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhbw       xm5, xm3, xm4
+    punpcklbw       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddubsw       m5, m3, m13
+    paddw           m1, m5
+    pmaddubsw       m3, m12
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhbw       xm6, xm4, xm5
+    punpcklbw       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddubsw       m6, m4, m13
+    paddw           m2, m6
+    pmaddubsw       m4, m12
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhbw       xm7, xm5, xm6
+    punpcklbw       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddubsw       m7, m5, m13
+    paddw           m3, m7
+    pmaddubsw       m5, m12
+    movu            xm7, [r0 + r4]                  ; m7 = row 7
+    punpckhbw       xm8, xm6, xm7
+    punpcklbw       xm6, xm7
+    vinserti128     m6, m6, xm8, 1
+    pmaddubsw       m8, m6, m13
+    paddw           m4, m8
+    pmaddubsw       m6, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm8, [r0]                       ; m8 = row 8
+    punpckhbw       xm9, xm7, xm8
+    punpcklbw       xm7, xm8
+    vinserti128     m7, m7, xm9, 1
+    pmaddubsw       m9, m7, m13
+    paddw           m5, m9
+    pmaddubsw       m7, m12
+    movu            xm9, [r0 + r1]                  ; m9 = row 9
+    punpckhbw       xm10, xm8, xm9
+    punpcklbw       xm8, xm9
+    vinserti128     m8, m8, xm10, 1
+    pmaddubsw       m10, m8, m13
+    paddw           m6, m10
+    pmaddubsw       m8, m12
+    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
+    punpckhbw       xm11, xm9, xm10
+    punpcklbw       xm9, xm10
+    vinserti128     m9, m9, xm11, 1
+    pmaddubsw       m11, m9, m13
+    paddw           m7, m11
+    pmaddubsw       m9, m12
+
+%ifidn %1,pp
+    pmulhrsw        m0, m14                         ; m0 = word: row 0
+    pmulhrsw        m1, m14                         ; m1 = word: row 1
+    pmulhrsw        m2, m14                         ; m2 = word: row 2
+    pmulhrsw        m3, m14                         ; m3 = word: row 3
+    pmulhrsw        m4, m14                         ; m4 = word: row 4
+    pmulhrsw        m5, m14                         ; m5 = word: row 5
+    pmulhrsw        m6, m14                         ; m6 = word: row 6
+    pmulhrsw        m7, m14                         ; m7 = word: row 7
+    packuswb        m0, m1
+    packuswb        m2, m3
+    packuswb        m4, m5
+    packuswb        m6, m7
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+    vextracti128    xm5, m4, 1
+    vextracti128    xm7, m6, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r5], xm3
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm4
+    movu            [r2 + r3], xm5
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r5], xm7
+%else
+    psubw           m0, m14                         ; m0 = word: row 0
+    psubw           m1, m14                         ; m1 = word: row 1
+    psubw           m2, m14                         ; m2 = word: row 2
+    psubw           m3, m14                         ; m3 = word: row 3
+    psubw           m4, m14                         ; m4 = word: row 4
+    psubw           m5, m14                         ; m5 = word: row 5
+    psubw           m6, m14                         ; m6 = word: row 6
+    psubw           m7, m14                         ; m7 = word: row 7
+    movu            [r2], m0
+    movu            [r2 + r3], m1
+    movu            [r2 + r3 * 2], m2
+    movu            [r2 + r5], m3
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], m4
+    movu            [r2 + r3], m5
+    movu            [r2 + r3 * 2], m6
+    movu            [r2 + r5], m7
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm11, [r0 + r4]                 ; m11 = row 11
+    punpckhbw       xm6, xm10, xm11
+    punpcklbw       xm10, xm11
+    vinserti128     m10, m10, xm6, 1
+    pmaddubsw       m6, m10, m13
+    paddw           m8, m6
+    pmaddubsw       m10, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm6, [r0]                       ; m6 = row 12
+    punpckhbw       xm7, xm11, xm6
+    punpcklbw       xm11, xm6
+    vinserti128     m11, m11, xm7, 1
+    pmaddubsw       m7, m11, m13
+    paddw           m9, m7
+    pmaddubsw       m11, m12
+
+    movu            xm7, [r0 + r1]                  ; m7 = row 13
+    punpckhbw       xm0, xm6, xm7
+    punpcklbw       xm6, xm7
+    vinserti128     m6, m6, xm0, 1
+    pmaddubsw       m0, m6, m13
+    paddw           m10, m0
+    pmaddubsw       m6, m12
+    movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
+    punpckhbw       xm1, xm7, xm0
+    punpcklbw       xm7, xm0
+    vinserti128     m7, m7, xm1, 1
+    pmaddubsw       m1, m7, m13
+    paddw           m11, m1
+    pmaddubsw       m7, m12
+    movu            xm1, [r0 + r4]                  ; m1 = row 15
+    punpckhbw       xm2, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddubsw       m2, m0, m13
+    paddw           m6, m2
+    pmaddubsw       m0, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm2, [r0]                       ; m2 = row 16
+    punpckhbw       xm3, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m3, m1, m13
+    paddw           m7, m3
+    pmaddubsw       m1, m12
+    movu            xm3, [r0 + r1]                  ; m3 = row 17
+    punpckhbw       xm4, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddubsw       m2, m13
+    paddw           m0, m2
+    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
+    punpckhbw       xm5, xm3, xm4
+    punpcklbw       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddubsw       m3, m13
+    paddw           m1, m3
+
+%ifidn %1,pp
+    pmulhrsw        m8, m14                         ; m8 = word: row 8
+    pmulhrsw        m9, m14                         ; m9 = word: row 9
+    pmulhrsw        m10, m14                        ; m10 = word: row 10
+    pmulhrsw        m11, m14                        ; m11 = word: row 11
+    pmulhrsw        m6, m14                         ; m6 = word: row 12
+    pmulhrsw        m7, m14                         ; m7 = word: row 13
+    pmulhrsw        m0, m14                         ; m0 = word: row 14
+    pmulhrsw        m1, m14                         ; m1 = word: row 15
+    packuswb        m8, m9
+    packuswb        m10, m11
+    packuswb        m6, m7
+    packuswb        m0, m1
+    vpermq          m8, m8, 11011000b
+    vpermq          m10, m10, 11011000b
+    vpermq          m6, m6, 11011000b
+    vpermq          m0, m0, 11011000b
+    vextracti128    xm9, m8, 1
+    vextracti128    xm11, m10, 1
+    vextracti128    xm7, m6, 1
+    vextracti128    xm1, m0, 1
+    movu            [r2], xm8
+    movu            [r2 + r3], xm9
+    movu            [r2 + r3 * 2], xm10
+    movu            [r2 + r5], xm11
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm6
+    movu            [r2 + r3], xm7
+    movu            [r2 + r3 * 2], xm0
+    movu            [r2 + r5], xm1
+%else
+    psubw           m8, m14                         ; m8 = word: row 8
+    psubw           m9, m14                         ; m9 = word: row 9
+    psubw           m10, m14                        ; m10 = word: row 10
+    psubw           m11, m14                        ; m11 = word: row 11
+    psubw           m6, m14                         ; m6 = word: row 12
+    psubw           m7, m14                         ; m7 = word: row 13
+    psubw           m0, m14                         ; m0 = word: row 14
+    psubw           m1, m14                         ; m1 = word: row 15
+    movu            [r2], m8
+    movu            [r2 + r3], m9
+    movu            [r2 + r3 * 2], m10
+    movu            [r2 + r5], m11
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], m6
+    movu            [r2 + r3], m7
+    movu            [r2 + r3 * 2], m0
+    movu            [r2 + r5], m1
+%endif
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_16x16 pp
+    FILTER_VER_CHROMA_AVX2_16x16 ps
+%macro FILTER_VER_CHROMA_AVX2_16x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_16x8, 4, 7, 7
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m6, [pw_512]
+%else
+    add             r3d, r3d
+    mova            m6, [pw_2000]
+%endif
+    lea             r6, [r3 * 3]
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhbw       xm2, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddubsw       m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhbw       xm3, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhbw       xm4, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddubsw       m4, m2, [r5 + mmsize]
+    paddw           m0, m4
+    pmaddubsw       m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhbw       xm5, xm3, xm4
+    punpcklbw       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddubsw       m5, m3, [r5 + mmsize]
+    paddw           m1, m5
+    pmaddubsw       m3, [r5]
+%ifidn %1,pp
+    pmulhrsw        m0, m6                          ; m0 = word: row 0
+    pmulhrsw        m1, m6                          ; m1 = word: row 1
+    packuswb        m0, m1
+    vpermq          m0, m0, 11011000b
+    vextracti128    xm1, m0, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+%else
+    psubw           m0, m6                          ; m0 = word: row 0
+    psubw           m1, m6                          ; m1 = word: row 1
+    movu            [r2], m0
+    movu            [r2 + r3], m1
+%endif
+
+    movu            xm0, [r0 + r1]                  ; m0 = row 5
+    punpckhbw       xm1, xm4, xm0
+    punpcklbw       xm4, xm0
+    vinserti128     m4, m4, xm1, 1
+    pmaddubsw       m1, m4, [r5 + mmsize]
+    paddw           m2, m1
+    pmaddubsw       m4, [r5]
+    movu            xm1, [r0 + r1 * 2]              ; m1 = row 6
+    punpckhbw       xm5, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm5, 1
+    pmaddubsw       m5, m0, [r5 + mmsize]
+    paddw           m3, m5
+    pmaddubsw       m0, [r5]
+%ifidn %1,pp
+    pmulhrsw        m2, m6                          ; m2 = word: row 2
+    pmulhrsw        m3, m6                          ; m3 = word: row 3
+    packuswb        m2, m3
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+%else
+    psubw           m2, m6                          ; m2 = word: row 2
+    psubw           m3, m6                          ; m3 = word: row 3
+    movu            [r2 + r3 * 2], m2
+    movu            [r2 + r6], m3
+%endif
+
+    movu            xm2, [r0 + r4]                  ; m2 = row 7
+    punpckhbw       xm3, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m3, m1, [r5 + mmsize]
+    paddw           m4, m3
+    pmaddubsw       m1, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm3, [r0]                       ; m3 = row 8
+    punpckhbw       xm5, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm5, 1
+    pmaddubsw       m5, m2, [r5 + mmsize]
+    paddw           m0, m5
+    pmaddubsw       m2, [r5]
+    lea             r2, [r2 + r3 * 4]
+%ifidn %1,pp
+    pmulhrsw        m4, m6                          ; m4 = word: row 4
+    pmulhrsw        m0, m6                          ; m0 = word: row 5
+    packuswb        m4, m0
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm0, m4, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm0
+%else
+    psubw           m4, m6                          ; m4 = word: row 4
+    psubw           m0, m6                          ; m0 = word: row 5
+    movu            [r2], m4
+    movu            [r2 + r3], m0
+%endif
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 9
+    punpckhbw       xm4, xm3, xm5
+    punpcklbw       xm3, xm5
+    vinserti128     m3, m3, xm4, 1
+    pmaddubsw       m3, [r5 + mmsize]
+    paddw           m1, m3
+    movu            xm4, [r0 + r1 * 2]              ; m4 = row 10
+    punpckhbw       xm0, xm5, xm4
+    punpcklbw       xm5, xm4
+    vinserti128     m5, m5, xm0, 1
+    pmaddubsw       m5, [r5 + mmsize]
+    paddw           m2, m5
+%ifidn %1,pp
+    pmulhrsw        m1, m6                          ; m1 = word: row 6
+    pmulhrsw        m2, m6                          ; m2 = word: row 7
+    packuswb        m1, m2
+    vpermq          m1, m1, 11011000b
+    vextracti128    xm2, m1, 1
+    movu            [r2 + r3 * 2], xm1
+    movu            [r2 + r6], xm2
+%else
+    psubw           m1, m6                          ; m1 = word: row 6
+    psubw           m2, m6                          ; m2 = word: row 7
+    movu            [r2 + r3 * 2], m1
+    movu            [r2 + r6], m2
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_16x8 pp
+    FILTER_VER_CHROMA_AVX2_16x8 ps
+
+%macro FILTER_VER_CHROMA_AVX2_16x12 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_16x12, 4, 6, 10
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    mova            m8, [r5]
+    mova            m9, [r5 + mmsize]
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m7, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m7, [pw_2000]
+%endif
+    lea             r5, [r3 * 3]
+
+    movu            xm0, [r0]
+    vinserti128     m0, m0, [r0 + r1 * 2], 1
+    movu            xm1, [r0 + r1]
+    vinserti128     m1, m1, [r0 + r4], 1
+
+    punpcklbw       m2, m0, m1
+    punpckhbw       m3, m0, m1
+    vperm2i128      m4, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    pmaddubsw       m4, m8
+    pmaddubsw       m3, m2, m9
+    paddw           m4, m3
+    pmaddubsw       m2, m8
+
+    vextracti128    xm0, m0, 1
+    lea             r0, [r0 + r1 * 4]
+    vinserti128     m0, m0, [r0], 1
+
+    punpcklbw       m5, m1, m0
+    punpckhbw       m3, m1, m0
+    vperm2i128      m6, m5, m3, 0x20
+    vperm2i128      m5, m5, m3, 0x31
+    pmaddubsw       m6, m8
+    pmaddubsw       m3, m5, m9
+    paddw           m6, m3
+    pmaddubsw       m5, m8
+%ifidn %1,pp
+    pmulhrsw        m4, m7                         ; m4 = word: row 0
+    pmulhrsw        m6, m7                         ; m6 = word: row 1
+    packuswb        m4, m6
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm6, m4, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm6
+%else
+    psubw           m4, m7                         ; m4 = word: row 0
+    psubw           m6, m7                         ; m6 = word: row 1
+    movu            [r2], m4
+    movu            [r2 + r3], m6
+%endif
+
+    movu            xm4, [r0 + r1 * 2]
+    vinserti128     m4, m4, [r0 + r1], 1
+    vextracti128    xm1, m4, 1
+    vinserti128     m0, m0, xm1, 0
+
+    punpcklbw       m6, m0, m4
+    punpckhbw       m1, m0, m4
+    vperm2i128      m0, m6, m1, 0x20
+    vperm2i128      m6, m6, m1, 0x31
+    pmaddubsw       m1, m0, m9
+    paddw           m5, m1
+    pmaddubsw       m0, m8
+    pmaddubsw       m1, m6, m9
+    paddw           m2, m1
+    pmaddubsw       m6, m8
+
+%ifidn %1,pp
+    pmulhrsw        m2, m7                         ; m2 = word: row 2
+    pmulhrsw        m5, m7                         ; m5 = word: row 3
+    packuswb        m2, m5
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm5, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r5], xm5
+%else
+    psubw           m2, m7                         ; m2 = word: row 2
+    psubw           m5, m7                         ; m5 = word: row 3
+    movu            [r2 + r3 * 2], m2
+    movu            [r2 + r5], m5
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm1, [r0 + r4]
+    lea             r0, [r0 + r1 * 4]
+    vinserti128     m1, m1, [r0], 1
+    vinserti128     m4, m4, xm1, 1
+
+    punpcklbw       m2, m4, m1
+    punpckhbw       m5, m4, m1
+    vperm2i128      m3, m2, m5, 0x20
+    vperm2i128      m2, m2, m5, 0x31
+    pmaddubsw       m5, m3, m9
+    paddw           m6, m5
+    pmaddubsw       m3, m8
+    pmaddubsw       m5, m2, m9
+    paddw           m0, m5
+    pmaddubsw       m2, m8
+
+%ifidn %1,pp
+    pmulhrsw        m6, m7                         ; m6 = word: row 4
+    pmulhrsw        m0, m7                         ; m0 = word: row 5
+    packuswb        m6, m0
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm0, m6, 1
+    movu            [r2], xm6
+    movu            [r2 + r3], xm0
+%else
+    psubw           m6, m7                         ; m6 = word: row 4
+    psubw           m0, m7                         ; m0 = word: row 5
+    movu            [r2], m6
+    movu            [r2 + r3], m0
+%endif
+
+    movu            xm6, [r0 + r1 * 2]
+    vinserti128     m6, m6, [r0 + r1], 1
+    vextracti128    xm0, m6, 1
+    vinserti128     m1, m1, xm0, 0
+
+    punpcklbw       m4, m1, m6
+    punpckhbw       m5, m1, m6
+    vperm2i128      m0, m4, m5, 0x20
+    vperm2i128      m5, m4, m5, 0x31
+    pmaddubsw       m4, m0, m9
+    paddw           m2, m4
+    pmaddubsw       m0, m8
+    pmaddubsw       m4, m5, m9
+    paddw           m3, m4
+    pmaddubsw       m5, m8
+
+%ifidn %1,pp
+    pmulhrsw        m3, m7                         ; m3 = word: row 6
+    pmulhrsw        m2, m7                         ; m2 = word: row 7
+    packuswb        m3, m2
+    vpermq          m3, m3, 11011000b
+    vextracti128    xm2, m3, 1
+    movu            [r2 + r3 * 2], xm3
+    movu            [r2 + r5], xm2
+%else
+    psubw           m3, m7                         ; m3 = word: row 6
+    psubw           m2, m7                         ; m2 = word: row 7
+    movu            [r2 + r3 * 2], m3
+    movu            [r2 + r5], m2
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm3, [r0 + r4]
+    lea             r0, [r0 + r1 * 4]
+    vinserti128     m3, m3, [r0], 1
+    vinserti128     m6, m6, xm3, 1
+
+    punpcklbw       m2, m6, m3
+    punpckhbw       m1, m6, m3
+    vperm2i128      m4, m2, m1, 0x20
+    vperm2i128      m2, m2, m1, 0x31
+    pmaddubsw       m1, m4, m9
+    paddw           m5, m1
+    pmaddubsw       m4, m8
+    pmaddubsw       m1, m2, m9
+    paddw           m0, m1
+    pmaddubsw       m2, m8
+
+%ifidn %1,pp
+    pmulhrsw        m5, m7                         ; m5 = word: row 8
+    pmulhrsw        m0, m7                         ; m0 = word: row 9
+    packuswb        m5, m0
+    vpermq          m5, m5, 11011000b
+    vextracti128    xm0, m5, 1
+    movu            [r2], xm5
+    movu            [r2 + r3], xm0
+%else
+    psubw           m5, m7                         ; m5 = word: row 8
+    psubw           m0, m7                         ; m0 = word: row 9
+    movu            [r2], m5
+    movu            [r2 + r3], m0
+%endif
+
+    movu            xm5, [r0 + r1 * 2]
+    vinserti128     m5, m5, [r0 + r1], 1
+    vextracti128    xm0, m5, 1
+    vinserti128     m3, m3, xm0, 0
+
+    punpcklbw       m1, m3, m5
+    punpckhbw       m0, m3, m5
+    vperm2i128      m6, m1, m0, 0x20
+    vperm2i128      m0, m1, m0, 0x31
+    pmaddubsw       m1, m6, m9
+    paddw           m2, m1
+    pmaddubsw       m1, m0, m9
+    paddw           m4, m1
+
+%ifidn %1,pp
+    pmulhrsw        m4, m7                         ; m4 = word: row 10
+    pmulhrsw        m2, m7                         ; m2 = word: row 11
+    packuswb        m4, m2
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm2, m4, 1
+    movu            [r2 + r3 * 2], xm4
+    movu            [r2 + r5], xm2
+%else
+    psubw           m4, m7                         ; m4 = word: row 10
+    psubw           m2, m7                         ; m2 = word: row 11
+    movu            [r2 + r3 * 2], m4
+    movu            [r2 + r5], m2
+%endif
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_16x12 pp
+    FILTER_VER_CHROMA_AVX2_16x12 ps
+
+%macro FILTER_VER_CHROMA_AVX2_16xN 2
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_16x%2, 4, 8, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m7, [pw_512]
+%else
+    add             r3d, r3d
+    mova            m7, [pw_2000]
+%endif
+    lea             r6, [r3 * 3]
+    mov             r7d, %2 / 16
+.loopH:
+    movu            xm0, [r0]
+    vinserti128     m0, m0, [r0 + r1 * 2], 1
+    movu            xm1, [r0 + r1]
+    vinserti128     m1, m1, [r0 + r4], 1
+
+    punpcklbw       m2, m0, m1
+    punpckhbw       m3, m0, m1
+    vperm2i128      m4, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    pmaddubsw       m4, [r5]
+    pmaddubsw       m3, m2, [r5 + mmsize]
+    paddw           m4, m3
+    pmaddubsw       m2, [r5]
+
+    vextracti128    xm0, m0, 1
+    lea             r0, [r0 + r1 * 4]
+    vinserti128     m0, m0, [r0], 1
+
+    punpcklbw       m5, m1, m0
+    punpckhbw       m3, m1, m0
+    vperm2i128      m6, m5, m3, 0x20
+    vperm2i128      m5, m5, m3, 0x31
+    pmaddubsw       m6, [r5]
+    pmaddubsw       m3, m5, [r5 + mmsize]
+    paddw           m6, m3
+    pmaddubsw       m5, [r5]
+%ifidn %1,pp
+    pmulhrsw        m4, m7                         ; m4 = word: row 0
+    pmulhrsw        m6, m7                         ; m6 = word: row 1
+    packuswb        m4, m6
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm6, m4, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm6
+%else
+    psubw           m4, m7                         ; m4 = word: row 0
+    psubw           m6, m7                         ; m6 = word: row 1
+    movu            [r2], m4
+    movu            [r2 + r3], m6
+%endif
+
+    movu            xm4, [r0 + r1 * 2]
+    vinserti128     m4, m4, [r0 + r1], 1
+    vextracti128    xm1, m4, 1
+    vinserti128     m0, m0, xm1, 0
+
+    punpcklbw       m6, m0, m4
+    punpckhbw       m1, m0, m4
+    vperm2i128      m0, m6, m1, 0x20
+    vperm2i128      m6, m6, m1, 0x31
+    pmaddubsw       m1, m0, [r5 + mmsize]
+    paddw           m5, m1
+    pmaddubsw       m0, [r5]
+    pmaddubsw       m1, m6, [r5 + mmsize]
+    paddw           m2, m1
+    pmaddubsw       m6, [r5]
+
+%ifidn %1,pp
+    pmulhrsw        m2, m7                         ; m2 = word: row 2
+    pmulhrsw        m5, m7                         ; m5 = word: row 3
+    packuswb        m2, m5
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm5, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm5
+%else
+    psubw           m2, m7                         ; m2 = word: row 2
+    psubw           m5, m7                         ; m5 = word: row 3
+    movu            [r2 + r3 * 2], m2
+    movu            [r2 + r6], m5
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm1, [r0 + r4]
+    lea             r0, [r0 + r1 * 4]
+    vinserti128     m1, m1, [r0], 1
+    vinserti128     m4, m4, xm1, 1
+
+    punpcklbw       m2, m4, m1
+    punpckhbw       m5, m4, m1
+    vperm2i128      m3, m2, m5, 0x20
+    vperm2i128      m2, m2, m5, 0x31
+    pmaddubsw       m5, m3, [r5 + mmsize]
+    paddw           m6, m5
+    pmaddubsw       m3, [r5]
+    pmaddubsw       m5, m2, [r5 + mmsize]
+    paddw           m0, m5
+    pmaddubsw       m2, [r5]
+
+%ifidn %1,pp
+    pmulhrsw        m6, m7                         ; m6 = word: row 4
+    pmulhrsw        m0, m7                         ; m0 = word: row 5
+    packuswb        m6, m0
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm0, m6, 1
+    movu            [r2], xm6
+    movu            [r2 + r3], xm0
+%else
+    psubw           m6, m7                         ; m6 = word: row 4
+    psubw           m0, m7                         ; m0 = word: row 5
+    movu            [r2], m6
+    movu            [r2 + r3], m0
+%endif
+
+    movu            xm6, [r0 + r1 * 2]
+    vinserti128     m6, m6, [r0 + r1], 1
+    vextracti128    xm0, m6, 1
+    vinserti128     m1, m1, xm0, 0
+
+    punpcklbw       m4, m1, m6
+    punpckhbw       m5, m1, m6
+    vperm2i128      m0, m4, m5, 0x20
+    vperm2i128      m5, m4, m5, 0x31
+    pmaddubsw       m4, m0, [r5 + mmsize]
+    paddw           m2, m4
+    pmaddubsw       m0, [r5]
+    pmaddubsw       m4, m5, [r5 + mmsize]
+    paddw           m3, m4
+    pmaddubsw       m5, [r5]
+
+%ifidn %1,pp
+    pmulhrsw        m3, m7                         ; m3 = word: row 6
+    pmulhrsw        m2, m7                         ; m2 = word: row 7
+    packuswb        m3, m2
+    vpermq          m3, m3, 11011000b
+    vextracti128    xm2, m3, 1
+    movu            [r2 + r3 * 2], xm3
+    movu            [r2 + r6], xm2
+%else
+    psubw           m3, m7                         ; m3 = word: row 6
+    psubw           m2, m7                         ; m2 = word: row 7
+    movu            [r2 + r3 * 2], m3
+    movu            [r2 + r6], m2
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm3, [r0 + r4]
+    lea             r0, [r0 + r1 * 4]
+    vinserti128     m3, m3, [r0], 1
+    vinserti128     m6, m6, xm3, 1
+
+    punpcklbw       m2, m6, m3
+    punpckhbw       m1, m6, m3
+    vperm2i128      m4, m2, m1, 0x20
+    vperm2i128      m2, m2, m1, 0x31
+    pmaddubsw       m1, m4, [r5 + mmsize]
+    paddw           m5, m1
+    pmaddubsw       m4, [r5]
+    pmaddubsw       m1, m2, [r5 + mmsize]
+    paddw           m0, m1
+    pmaddubsw       m2, [r5]
+
+%ifidn %1,pp
+    pmulhrsw        m5, m7                         ; m5 = word: row 8
+    pmulhrsw        m0, m7                         ; m0 = word: row 9
+    packuswb        m5, m0
+    vpermq          m5, m5, 11011000b
+    vextracti128    xm0, m5, 1
+    movu            [r2], xm5
+    movu            [r2 + r3], xm0
+%else
+    psubw           m5, m7                         ; m5 = word: row 8
+    psubw           m0, m7                         ; m0 = word: row 9
+    movu            [r2], m5
+    movu            [r2 + r3], m0
+%endif
+
+    movu            xm5, [r0 + r1 * 2]
+    vinserti128     m5, m5, [r0 + r1], 1
+    vextracti128    xm0, m5, 1
+    vinserti128     m3, m3, xm0, 0
+
+    punpcklbw       m1, m3, m5
+    punpckhbw       m0, m3, m5
+    vperm2i128      m6, m1, m0, 0x20
+    vperm2i128      m0, m1, m0, 0x31
+    pmaddubsw       m1, m6, [r5 + mmsize]
+    paddw           m2, m1
+    pmaddubsw       m6, [r5]
+    pmaddubsw       m1, m0, [r5 + mmsize]
+    paddw           m4, m1
+    pmaddubsw       m0, [r5]
+
+%ifidn %1,pp
+    pmulhrsw        m4, m7                         ; m4 = word: row 10
+    pmulhrsw        m2, m7                         ; m2 = word: row 11
+    packuswb        m4, m2
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm2, m4, 1
+    movu            [r2 + r3 * 2], xm4
+    movu            [r2 + r6], xm2
+%else
+    psubw           m4, m7                         ; m4 = word: row 10
+    psubw           m2, m7                         ; m2 = word: row 11
+    movu            [r2 + r3 * 2], m4
+    movu            [r2 + r6], m2
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm3, [r0 + r4]
+    lea             r0, [r0 + r1 * 4]
+    vinserti128     m3, m3, [r0], 1
+    vinserti128     m5, m5, xm3, 1
+
+    punpcklbw       m2, m5, m3
+    punpckhbw       m1, m5, m3
+    vperm2i128      m4, m2, m1, 0x20
+    vperm2i128      m2, m2, m1, 0x31
+    pmaddubsw       m1, m4, [r5 + mmsize]
+    paddw           m0, m1
+    pmaddubsw       m4, [r5]
+    pmaddubsw       m1, m2, [r5 + mmsize]
+    paddw           m6, m1
+    pmaddubsw       m2, [r5]
+
+%ifidn %1,pp
+    pmulhrsw        m0, m7                         ; m0 = word: row 12
+    pmulhrsw        m6, m7                         ; m6 = word: row 13
+    packuswb        m0, m6
+    vpermq          m0, m0, 11011000b
+    vextracti128    xm6, m0, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm6
+%else
+    psubw           m0, m7                         ; m0 = word: row 12
+    psubw           m6, m7                         ; m6 = word: row 13
+    movu            [r2], m0
+    movu            [r2 + r3], m6
+%endif
+
+    movu            xm5, [r0 + r1 * 2]
+    vinserti128     m5, m5, [r0 + r1], 1
+    vextracti128    xm0, m5, 1
+    vinserti128     m3, m3, xm0, 0
+
+    punpcklbw       m1, m3, m5
+    punpckhbw       m0, m3, m5
+    vperm2i128      m6, m1, m0, 0x20
+    vperm2i128      m0, m1, m0, 0x31
+    pmaddubsw       m6, [r5 + mmsize]
+    paddw           m2, m6
+    pmaddubsw       m0, [r5 + mmsize]
+    paddw           m4, m0
+
+%ifidn %1,pp
+    pmulhrsw        m4, m7                         ; m4 = word: row 14
+    pmulhrsw        m2, m7                         ; m2 = word: row 15
+    packuswb        m4, m2
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm2, m4, 1
+    movu            [r2 + r3 * 2], xm4
+    movu            [r2 + r6], xm2
+%else
+    psubw           m4, m7                         ; m4 = word: row 14
+    psubw           m2, m7                         ; m2 = word: row 15
+    movu            [r2 + r3 * 2], m4
+    movu            [r2 + r6], m2
+%endif
+    lea             r2, [r2 + r3 * 4]
+    dec             r7d
+    jnz             .loopH
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_16xN pp, 32
+    FILTER_VER_CHROMA_AVX2_16xN ps, 32
+    FILTER_VER_CHROMA_AVX2_16xN pp, 64
+    FILTER_VER_CHROMA_AVX2_16xN ps, 64
+
+%macro FILTER_VER_CHROMA_AVX2_16x24 1
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_16x24, 4, 6, 15
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    mova            m12, [r5]
+    mova            m13, [r5 + mmsize]
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m14, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m14, [pw_2000]
+%endif
+    lea             r5, [r3 * 3]
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhbw       xm2, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddubsw       m0, m12
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhbw       xm3, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m1, m12
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhbw       xm4, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddubsw       m4, m2, m13
+    paddw           m0, m4
+    pmaddubsw       m2, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhbw       xm5, xm3, xm4
+    punpcklbw       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddubsw       m5, m3, m13
+    paddw           m1, m5
+    pmaddubsw       m3, m12
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhbw       xm6, xm4, xm5
+    punpcklbw       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddubsw       m6, m4, m13
+    paddw           m2, m6
+    pmaddubsw       m4, m12
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhbw       xm7, xm5, xm6
+    punpcklbw       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddubsw       m7, m5, m13
+    paddw           m3, m7
+    pmaddubsw       m5, m12
+    movu            xm7, [r0 + r4]                  ; m7 = row 7
+    punpckhbw       xm8, xm6, xm7
+    punpcklbw       xm6, xm7
+    vinserti128     m6, m6, xm8, 1
+    pmaddubsw       m8, m6, m13
+    paddw           m4, m8
+    pmaddubsw       m6, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm8, [r0]                       ; m8 = row 8
+    punpckhbw       xm9, xm7, xm8
+    punpcklbw       xm7, xm8
+    vinserti128     m7, m7, xm9, 1
+    pmaddubsw       m9, m7, m13
+    paddw           m5, m9
+    pmaddubsw       m7, m12
+    movu            xm9, [r0 + r1]                  ; m9 = row 9
+    punpckhbw       xm10, xm8, xm9
+    punpcklbw       xm8, xm9
+    vinserti128     m8, m8, xm10, 1
+    pmaddubsw       m10, m8, m13
+    paddw           m6, m10
+    pmaddubsw       m8, m12
+    movu            xm10, [r0 + r1 * 2]             ; m10 = row 10
+    punpckhbw       xm11, xm9, xm10
+    punpcklbw       xm9, xm10
+    vinserti128     m9, m9, xm11, 1
+    pmaddubsw       m11, m9, m13
+    paddw           m7, m11
+    pmaddubsw       m9, m12
+
+%ifidn %1,pp
+    pmulhrsw        m0, m14                         ; m0 = word: row 0
+    pmulhrsw        m1, m14                         ; m1 = word: row 1
+    pmulhrsw        m2, m14                         ; m2 = word: row 2
+    pmulhrsw        m3, m14                         ; m3 = word: row 3
+    pmulhrsw        m4, m14                         ; m4 = word: row 4
+    pmulhrsw        m5, m14                         ; m5 = word: row 5
+    pmulhrsw        m6, m14                         ; m6 = word: row 6
+    pmulhrsw        m7, m14                         ; m7 = word: row 7
+    packuswb        m0, m1
+    packuswb        m2, m3
+    packuswb        m4, m5
+    packuswb        m6, m7
+    vpermq          m0, m0, q3120
+    vpermq          m2, m2, q3120
+    vpermq          m4, m4, q3120
+    vpermq          m6, m6, q3120
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+    vextracti128    xm5, m4, 1
+    vextracti128    xm7, m6, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r5], xm3
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm4
+    movu            [r2 + r3], xm5
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r5], xm7
+%else
+    psubw           m0, m14                         ; m0 = word: row 0
+    psubw           m1, m14                         ; m1 = word: row 1
+    psubw           m2, m14                         ; m2 = word: row 2
+    psubw           m3, m14                         ; m3 = word: row 3
+    psubw           m4, m14                         ; m4 = word: row 4
+    psubw           m5, m14                         ; m5 = word: row 5
+    psubw           m6, m14                         ; m6 = word: row 6
+    psubw           m7, m14                         ; m7 = word: row 7
+    movu            [r2], m0
+    movu            [r2 + r3], m1
+    movu            [r2 + r3 * 2], m2
+    movu            [r2 + r5], m3
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], m4
+    movu            [r2 + r3], m5
+    movu            [r2 + r3 * 2], m6
+    movu            [r2 + r5], m7
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm11, [r0 + r4]                 ; m11 = row 11
+    punpckhbw       xm6, xm10, xm11
+    punpcklbw       xm10, xm11
+    vinserti128     m10, m10, xm6, 1
+    pmaddubsw       m6, m10, m13
+    paddw           m8, m6
+    pmaddubsw       m10, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm6, [r0]                       ; m6 = row 12
+    punpckhbw       xm7, xm11, xm6
+    punpcklbw       xm11, xm6
+    vinserti128     m11, m11, xm7, 1
+    pmaddubsw       m7, m11, m13
+    paddw           m9, m7
+    pmaddubsw       m11, m12
+
+    movu            xm7, [r0 + r1]                  ; m7 = row 13
+    punpckhbw       xm0, xm6, xm7
+    punpcklbw       xm6, xm7
+    vinserti128     m6, m6, xm0, 1
+    pmaddubsw       m0, m6, m13
+    paddw           m10, m0
+    pmaddubsw       m6, m12
+    movu            xm0, [r0 + r1 * 2]              ; m0 = row 14
+    punpckhbw       xm1, xm7, xm0
+    punpcklbw       xm7, xm0
+    vinserti128     m7, m7, xm1, 1
+    pmaddubsw       m1, m7, m13
+    paddw           m11, m1
+    pmaddubsw       m7, m12
+    movu            xm1, [r0 + r4]                  ; m1 = row 15
+    punpckhbw       xm2, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddubsw       m2, m0, m13
+    paddw           m6, m2
+    pmaddubsw       m0, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm2, [r0]                       ; m2 = row 16
+    punpckhbw       xm3, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m3, m1, m13
+    paddw           m7, m3
+    pmaddubsw       m1, m12
+    movu            xm3, [r0 + r1]                  ; m3 = row 17
+    punpckhbw       xm4, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddubsw       m4, m2, m13
+    paddw           m0, m4
+    pmaddubsw       m2, m12
+    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
+    punpckhbw       xm5, xm3, xm4
+    punpcklbw       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddubsw       m5, m3, m13
+    paddw           m1, m5
+    pmaddubsw       m3, m12
+
+%ifidn %1,pp
+    pmulhrsw        m8, m14                         ; m8 = word: row 8
+    pmulhrsw        m9, m14                         ; m9 = word: row 9
+    pmulhrsw        m10, m14                        ; m10 = word: row 10
+    pmulhrsw        m11, m14                        ; m11 = word: row 11
+    pmulhrsw        m6, m14                         ; m6 = word: row 12
+    pmulhrsw        m7, m14                         ; m7 = word: row 13
+    pmulhrsw        m0, m14                         ; m0 = word: row 14
+    pmulhrsw        m1, m14                         ; m1 = word: row 15
+    packuswb        m8, m9
+    packuswb        m10, m11
+    packuswb        m6, m7
+    packuswb        m0, m1
+    vpermq          m8, m8, q3120
+    vpermq          m10, m10, q3120
+    vpermq          m6, m6, q3120
+    vpermq          m0, m0, q3120
+    vextracti128    xm9, m8, 1
+    vextracti128    xm11, m10, 1
+    vextracti128    xm7, m6, 1
+    vextracti128    xm1, m0, 1
+    movu            [r2], xm8
+    movu            [r2 + r3], xm9
+    movu            [r2 + r3 * 2], xm10
+    movu            [r2 + r5], xm11
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm6
+    movu            [r2 + r3], xm7
+    movu            [r2 + r3 * 2], xm0
+    movu            [r2 + r5], xm1
+%else
+    psubw           m8, m14                         ; m8 = word: row 8
+    psubw           m9, m14                         ; m9 = word: row 9
+    psubw           m10, m14                        ; m10 = word: row 10
+    psubw           m11, m14                        ; m11 = word: row 11
+    psubw           m6, m14                         ; m6 = word: row 12
+    psubw           m7, m14                         ; m7 = word: row 13
+    psubw           m0, m14                         ; m0 = word: row 14
+    psubw           m1, m14                         ; m1 = word: row 15
+    movu            [r2], m8
+    movu            [r2 + r3], m9
+    movu            [r2 + r3 * 2], m10
+    movu            [r2 + r5], m11
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], m6
+    movu            [r2 + r3], m7
+    movu            [r2 + r3 * 2], m0
+    movu            [r2 + r5], m1
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm5, [r0 + r4]                  ; m5 = row 19
+    punpckhbw       xm6, xm4, xm5
+    punpcklbw       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddubsw       m6, m4, m13
+    paddw           m2, m6
+    pmaddubsw       m4, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm6, [r0]                       ; m6 = row 20
+    punpckhbw       xm7, xm5, xm6
+    punpcklbw       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddubsw       m7, m5, m13
+    paddw           m3, m7
+    pmaddubsw       m5, m12
+    movu            xm7, [r0 + r1]                  ; m7 = row 21
+    punpckhbw       xm0, xm6, xm7
+    punpcklbw       xm6, xm7
+    vinserti128     m6, m6, xm0, 1
+    pmaddubsw       m0, m6, m13
+    paddw           m4, m0
+    pmaddubsw       m6, m12
+    movu            xm0, [r0 + r1 * 2]              ; m0 = row 22
+    punpckhbw       xm1, xm7, xm0
+    punpcklbw       xm7, xm0
+    vinserti128     m7, m7, xm1, 1
+    pmaddubsw       m1, m7, m13
+    paddw           m5, m1
+    pmaddubsw       m7, m12
+    movu            xm1, [r0 + r4]                  ; m1 = row 23
+    punpckhbw       xm8, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm8, 1
+    pmaddubsw       m8, m0, m13
+    paddw           m6, m8
+    pmaddubsw       m0, m12
+    lea             r0, [r0 + r1 * 4]
+    movu            xm8, [r0]                       ; m8 = row 24
+    punpckhbw       xm9, xm1, xm8
+    punpcklbw       xm1, xm8
+    vinserti128     m1, m1, xm9, 1
+    pmaddubsw       m9, m1, m13
+    paddw           m7, m9
+    pmaddubsw       m1, m12
+    movu            xm9, [r0 + r1]                  ; m9 = row 25
+    punpckhbw       xm10, xm8, xm9
+    punpcklbw       xm8, xm9
+    vinserti128     m8, m8, xm10, 1
+    pmaddubsw       m8, m13
+    paddw           m0, m8
+    movu            xm10, [r0 + r1 * 2]             ; m10 = row 26
+    punpckhbw       xm11, xm9, xm10
+    punpcklbw       xm9, xm10
+    vinserti128     m9, m9, xm11, 1
+    pmaddubsw       m9, m13
+    paddw           m1, m9
+
+%ifidn %1,pp
+    pmulhrsw        m2, m14                         ; m2 = word: row 16
+    pmulhrsw        m3, m14                         ; m3 = word: row 17
+    pmulhrsw        m4, m14                         ; m4 = word: row 18
+    pmulhrsw        m5, m14                         ; m5 = word: row 19
+    pmulhrsw        m6, m14                         ; m6 = word: row 20
+    pmulhrsw        m7, m14                         ; m7 = word: row 21
+    pmulhrsw        m0, m14                         ; m0 = word: row 22
+    pmulhrsw        m1, m14                         ; m1 = word: row 23
+    packuswb        m2, m3
+    packuswb        m4, m5
+    packuswb        m6, m7
+    packuswb        m0, m1
+    vpermq          m2, m2, q3120
+    vpermq          m4, m4, q3120
+    vpermq          m6, m6, q3120
+    vpermq          m0, m0, q3120
+    vextracti128    xm3, m2, 1
+    vextracti128    xm5, m4, 1
+    vextracti128    xm7, m6, 1
+    vextracti128    xm1, m0, 1
+    movu            [r2], xm2
+    movu            [r2 + r3], xm3
+    movu            [r2 + r3 * 2], xm4
+    movu            [r2 + r5], xm5
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm6
+    movu            [r2 + r3], xm7
+    movu            [r2 + r3 * 2], xm0
+    movu            [r2 + r5], xm1
+%else
+    psubw           m2, m14                         ; m2 = word: row 16
+    psubw           m3, m14                         ; m3 = word: row 17
+    psubw           m4, m14                         ; m4 = word: row 18
+    psubw           m5, m14                         ; m5 = word: row 19
+    psubw           m6, m14                         ; m6 = word: row 20
+    psubw           m7, m14                         ; m7 = word: row 21
+    psubw           m0, m14                         ; m0 = word: row 22
+    psubw           m1, m14                         ; m1 = word: row 23
+    movu            [r2], m2
+    movu            [r2 + r3], m3
+    movu            [r2 + r3 * 2], m4
+    movu            [r2 + r5], m5
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], m6
+    movu            [r2 + r3], m7
+    movu            [r2 + r3 * 2], m0
+    movu            [r2 + r5], m1
+%endif
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_16x24 pp
+    FILTER_VER_CHROMA_AVX2_16x24 ps
+
+%macro FILTER_VER_CHROMA_AVX2_24x32 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_24x32, 4, 9, 10
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    mova            m8, [r5]
+    mova            m9, [r5 + mmsize]
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m7, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m7, [pw_2000]
+%endif
+    lea             r6, [r3 * 3]
+    mov             r5d, 2
+.loopH:
+    movu            xm0, [r0]
+    vinserti128     m0, m0, [r0 + r1 * 2], 1
+    movu            xm1, [r0 + r1]
+    vinserti128     m1, m1, [r0 + r4], 1
+
+    punpcklbw       m2, m0, m1
+    punpckhbw       m3, m0, m1
+    vperm2i128      m4, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    pmaddubsw       m4, m8
+    pmaddubsw       m3, m2, m9
+    paddw           m4, m3
+    pmaddubsw       m2, m8
+
+    vextracti128    xm0, m0, 1
+    lea             r7, [r0 + r1 * 4]
+    vinserti128     m0, m0, [r7], 1
+
+    punpcklbw       m5, m1, m0
+    punpckhbw       m3, m1, m0
+    vperm2i128      m6, m5, m3, 0x20
+    vperm2i128      m5, m5, m3, 0x31
+    pmaddubsw       m6, m8
+    pmaddubsw       m3, m5, m9
+    paddw           m6, m3
+    pmaddubsw       m5, m8
+%ifidn %1,pp
+    pmulhrsw        m4, m7                         ; m4 = word: row 0
+    pmulhrsw        m6, m7                         ; m6 = word: row 1
+    packuswb        m4, m6
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm6, m4, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm6
+%else
+    psubw           m4, m7                         ; m4 = word: row 0
+    psubw           m6, m7                         ; m6 = word: row 1
+    movu            [r2], m4
+    movu            [r2 + r3], m6
+%endif
+
+    movu            xm4, [r7 + r1 * 2]
+    vinserti128     m4, m4, [r7 + r1], 1
+    vextracti128    xm1, m4, 1
+    vinserti128     m0, m0, xm1, 0
+
+    punpcklbw       m6, m0, m4
+    punpckhbw       m1, m0, m4
+    vperm2i128      m0, m6, m1, 0x20
+    vperm2i128      m6, m6, m1, 0x31
+    pmaddubsw       m1, m0, m9
+    paddw           m5, m1
+    pmaddubsw       m0, m8
+    pmaddubsw       m1, m6, m9
+    paddw           m2, m1
+    pmaddubsw       m6, m8
+
+%ifidn %1,pp
+    pmulhrsw        m2, m7                         ; m2 = word: row 2
+    pmulhrsw        m5, m7                         ; m5 = word: row 3
+    packuswb        m2, m5
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm5, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm5
+%else
+    psubw           m2, m7                         ; m2 = word: row 2
+    psubw           m5, m7                         ; m5 = word: row 3
+    movu            [r2 + r3 * 2], m2
+    movu            [r2 + r6], m5
+%endif
+    lea             r8, [r2 + r3 * 4]
+
+    movu            xm1, [r7 + r4]
+    lea             r7, [r7 + r1 * 4]
+    vinserti128     m1, m1, [r7], 1
+    vinserti128     m4, m4, xm1, 1
+
+    punpcklbw       m2, m4, m1
+    punpckhbw       m5, m4, m1
+    vperm2i128      m3, m2, m5, 0x20
+    vperm2i128      m2, m2, m5, 0x31
+    pmaddubsw       m5, m3, m9
+    paddw           m6, m5
+    pmaddubsw       m3, m8
+    pmaddubsw       m5, m2, m9
+    paddw           m0, m5
+    pmaddubsw       m2, m8
+
+%ifidn %1,pp
+    pmulhrsw        m6, m7                         ; m6 = word: row 4
+    pmulhrsw        m0, m7                         ; m0 = word: row 5
+    packuswb        m6, m0
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm0, m6, 1
+    movu            [r8], xm6
+    movu            [r8 + r3], xm0
+%else
+    psubw           m6, m7                         ; m6 = word: row 4
+    psubw           m0, m7                         ; m0 = word: row 5
+    movu            [r8], m6
+    movu            [r8 + r3], m0
+%endif
+
+    movu            xm6, [r7 + r1 * 2]
+    vinserti128     m6, m6, [r7 + r1], 1
+    vextracti128    xm0, m6, 1
+    vinserti128     m1, m1, xm0, 0
+
+    punpcklbw       m4, m1, m6
+    punpckhbw       m5, m1, m6
+    vperm2i128      m0, m4, m5, 0x20
+    vperm2i128      m5, m4, m5, 0x31
+    pmaddubsw       m4, m0, m9
+    paddw           m2, m4
+    pmaddubsw       m0, m8
+    pmaddubsw       m4, m5, m9
+    paddw           m3, m4
+    pmaddubsw       m5, m8
+
+%ifidn %1,pp
+    pmulhrsw        m3, m7                         ; m3 = word: row 6
+    pmulhrsw        m2, m7                         ; m2 = word: row 7
+    packuswb        m3, m2
+    vpermq          m3, m3, 11011000b
+    vextracti128    xm2, m3, 1
+    movu            [r8 + r3 * 2], xm3
+    movu            [r8 + r6], xm2
+%else
+    psubw           m3, m7                         ; m3 = word: row 6
+    psubw           m2, m7                         ; m2 = word: row 7
+    movu            [r8 + r3 * 2], m3
+    movu            [r8 + r6], m2
+%endif
+    lea             r8, [r8 + r3 * 4]
+
+    movu            xm3, [r7 + r4]
+    lea             r7, [r7 + r1 * 4]
+    vinserti128     m3, m3, [r7], 1
+    vinserti128     m6, m6, xm3, 1
+
+    punpcklbw       m2, m6, m3
+    punpckhbw       m1, m6, m3
+    vperm2i128      m4, m2, m1, 0x20
+    vperm2i128      m2, m2, m1, 0x31
+    pmaddubsw       m1, m4, m9
+    paddw           m5, m1
+    pmaddubsw       m4, m8
+    pmaddubsw       m1, m2, m9
+    paddw           m0, m1
+    pmaddubsw       m2, m8
+
+%ifidn %1,pp
+    pmulhrsw        m5, m7                         ; m5 = word: row 8
+    pmulhrsw        m0, m7                         ; m0 = word: row 9
+    packuswb        m5, m0
+    vpermq          m5, m5, 11011000b
+    vextracti128    xm0, m5, 1
+    movu            [r8], xm5
+    movu            [r8 + r3], xm0
+%else
+    psubw           m5, m7                         ; m5 = word: row 8
+    psubw           m0, m7                         ; m0 = word: row 9
+    movu            [r8], m5
+    movu            [r8 + r3], m0
+%endif
+
+    movu            xm5, [r7 + r1 * 2]
+    vinserti128     m5, m5, [r7 + r1], 1
+    vextracti128    xm0, m5, 1
+    vinserti128     m3, m3, xm0, 0
+
+    punpcklbw       m1, m3, m5
+    punpckhbw       m0, m3, m5
+    vperm2i128      m6, m1, m0, 0x20
+    vperm2i128      m0, m1, m0, 0x31
+    pmaddubsw       m1, m6, m9
+    paddw           m2, m1
+    pmaddubsw       m6, m8
+    pmaddubsw       m1, m0, m9
+    paddw           m4, m1
+    pmaddubsw       m0, m8
+
+%ifidn %1,pp
+    pmulhrsw        m4, m7                         ; m4 = word: row 10
+    pmulhrsw        m2, m7                         ; m2 = word: row 11
+    packuswb        m4, m2
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm2, m4, 1
+    movu            [r8 + r3 * 2], xm4
+    movu            [r8 + r6], xm2
+%else
+    psubw           m4, m7                         ; m4 = word: row 10
+    psubw           m2, m7                         ; m2 = word: row 11
+    movu            [r8 + r3 * 2], m4
+    movu            [r8 + r6], m2
+%endif
+    lea             r8, [r8 + r3 * 4]
+
+    movu            xm3, [r7 + r4]
+    lea             r7, [r7 + r1 * 4]
+    vinserti128     m3, m3, [r7], 1
+    vinserti128     m5, m5, xm3, 1
+
+    punpcklbw       m2, m5, m3
+    punpckhbw       m1, m5, m3
+    vperm2i128      m4, m2, m1, 0x20
+    vperm2i128      m2, m2, m1, 0x31
+    pmaddubsw       m1, m4, m9
+    paddw           m0, m1
+    pmaddubsw       m4, m8
+    pmaddubsw       m1, m2, m9
+    paddw           m6, m1
+    pmaddubsw       m2, m8
+
+%ifidn %1,pp
+    pmulhrsw        m0, m7                         ; m0 = word: row 12
+    pmulhrsw        m6, m7                         ; m6 = word: row 13
+    packuswb        m0, m6
+    vpermq          m0, m0, 11011000b
+    vextracti128    xm6, m0, 1
+    movu            [r8], xm0
+    movu            [r8 + r3], xm6
+%else
+    psubw           m0, m7                         ; m0 = word: row 12
+    psubw           m6, m7                         ; m6 = word: row 13
+    movu            [r8], m0
+    movu            [r8 + r3], m6
+%endif
+
+    movu            xm5, [r7 + r1 * 2]
+    vinserti128     m5, m5, [r7 + r1], 1
+    vextracti128    xm0, m5, 1
+    vinserti128     m3, m3, xm0, 0
+
+    punpcklbw       m1, m3, m5
+    punpckhbw       m0, m3, m5
+    vperm2i128      m6, m1, m0, 0x20
+    vperm2i128      m0, m1, m0, 0x31
+    pmaddubsw       m6, m9
+    paddw           m2, m6
+    pmaddubsw       m0, m9
+    paddw           m4, m0
+
+%ifidn %1,pp
+    pmulhrsw        m4, m7                         ; m4 = word: row 14
+    pmulhrsw        m2, m7                         ; m2 = word: row 15
+    packuswb        m4, m2
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm2, m4, 1
+    movu            [r8 + r3 * 2], xm4
+    movu            [r8 + r6], xm2
+    add             r2, 16
+%else
+    psubw           m4, m7                         ; m4 = word: row 14
+    psubw           m2, m7                         ; m2 = word: row 15
+    movu            [r8 + r3 * 2], m4
+    movu            [r8 + r6], m2
+    add             r2, 32
+%endif
+    add             r0, 16
+    movq            xm1, [r0]                       ; m1 = row 0
+    movq            xm2, [r0 + r1]                  ; m2 = row 1
+    punpcklbw       xm1, xm2
+    movq            xm3, [r0 + r1 * 2]              ; m3 = row 2
+    punpcklbw       xm2, xm3
+    vinserti128     m5, m1, xm2, 1
+    pmaddubsw       m5, m8
+    movq            xm4, [r0 + r4]                  ; m4 = row 3
+    punpcklbw       xm3, xm4
+    lea             r7, [r0 + r1 * 4]
+    movq            xm1, [r7]                       ; m1 = row 4
+    punpcklbw       xm4, xm1
+    vinserti128     m2, m3, xm4, 1
+    pmaddubsw       m0, m2, m9
+    paddw           m5, m0
+    pmaddubsw       m2, m8
+    movq            xm3, [r7 + r1]                  ; m3 = row 5
+    punpcklbw       xm1, xm3
+    movq            xm4, [r7 + r1 * 2]              ; m4 = row 6
+    punpcklbw       xm3, xm4
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m0, m1, m9
+    paddw           m2, m0
+    pmaddubsw       m1, m8
+    movq            xm3, [r7 + r4]                  ; m3 = row 7
+    punpcklbw       xm4, xm3
+    lea             r7, [r7 + r1 * 4]
+    movq            xm0, [r7]                       ; m0 = row 8
+    punpcklbw       xm3, xm0
+    vinserti128     m4, m4, xm3, 1
+    pmaddubsw       m3, m4, m9
+    paddw           m1, m3
+    pmaddubsw       m4, m8
+    movq            xm3, [r7 + r1]                  ; m3 = row 9
+    punpcklbw       xm0, xm3
+    movq            xm6, [r7 + r1 * 2]              ; m6 = row 10
+    punpcklbw       xm3, xm6
+    vinserti128     m0, m0, xm3, 1
+    pmaddubsw       m3, m0, m9
+    paddw           m4, m3
+    pmaddubsw       m0, m8
+
+%ifidn %1,pp
+    pmulhrsw        m5, m7                          ; m5 = word: row 0, row 1
+    pmulhrsw        m2, m7                          ; m2 = word: row 2, row 3
+    pmulhrsw        m1, m7                          ; m1 = word: row 4, row 5
+    pmulhrsw        m4, m7                          ; m4 = word: row 6, row 7
+    packuswb        m5, m2
+    packuswb        m1, m4
+    vextracti128    xm2, m5, 1
+    vextracti128    xm4, m1, 1
+    movq            [r2], xm5
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm5
+    movhps          [r2 + r6], xm2
+    lea             r8, [r2 + r3 * 4]
+    movq            [r8], xm1
+    movq            [r8 + r3], xm4
+    movhps          [r8 + r3 * 2], xm1
+    movhps          [r8 + r6], xm4
+%else
+    psubw           m5, m7                          ; m5 = word: row 0, row 1
+    psubw           m2, m7                          ; m2 = word: row 2, row 3
+    psubw           m1, m7                          ; m1 = word: row 4, row 5
+    psubw           m4, m7                          ; m4 = word: row 6, row 7
+    vextracti128    xm3, m5, 1
+    movu            [r2], xm5
+    movu            [r2 + r3], xm3
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+    vextracti128    xm3, m1, 1
+    lea             r8, [r2 + r3 * 4]
+    movu            [r8], xm1
+    movu            [r8 + r3], xm3
+    vextracti128    xm3, m4, 1
+    movu            [r8 + r3 * 2], xm4
+    movu            [r8 + r6], xm3
+%endif
+    lea             r8, [r8 + r3 * 4]
+
+    movq            xm3, [r7 + r4]                  ; m3 = row 11
+    punpcklbw       xm6, xm3
+    lea             r7, [r7 + r1 * 4]
+    movq            xm5, [r7]                       ; m5 = row 12
+    punpcklbw       xm3, xm5
+    vinserti128     m6, m6, xm3, 1
+    pmaddubsw       m3, m6, m9
+    paddw           m0, m3
+    pmaddubsw       m6, m8
+    movq            xm3, [r7 + r1]                  ; m3 = row 13
+    punpcklbw       xm5, xm3
+    movq            xm2, [r7 + r1 * 2]              ; m2 = row 14
+    punpcklbw       xm3, xm2
+    vinserti128     m5, m5, xm3, 1
+    pmaddubsw       m3, m5, m9
+    paddw           m6, m3
+    pmaddubsw       m5, m8
+    movq            xm3, [r7 + r4]                  ; m3 = row 15
+    punpcklbw       xm2, xm3
+    lea             r7, [r7 + r1 * 4]
+    movq            xm1, [r7]                       ; m1 = row 16
+    punpcklbw       xm3, xm1
+    vinserti128     m2, m2, xm3, 1
+    pmaddubsw       m3, m2, m9
+    paddw           m5, m3
+    pmaddubsw       m2, m8
+    movq            xm3, [r7 + r1]                  ; m3 = row 17
+    punpcklbw       xm1, xm3
+    movq            xm4, [r7 + r1 * 2]              ; m4 = row 18
+    punpcklbw       xm3, xm4
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m3, m1, m9
+    paddw           m2, m3
+%ifidn %1,pp
+    pmulhrsw        m0, m7                          ; m0 = word: row 8, row 9
+    pmulhrsw        m6, m7                          ; m6 = word: row 10, row 11
+    pmulhrsw        m5, m7                          ; m5 = word: row 12, row 13
+    pmulhrsw        m2, m7                          ; m2 = word: row 14, row 15
+    packuswb        m0, m6
+    packuswb        m5, m2
+    vextracti128    xm6, m0, 1
+    vextracti128    xm2, m5, 1
+    movq            [r8], xm0
+    movq            [r8 + r3], xm6
+    movhps          [r8 + r3 * 2], xm0
+    movhps          [r8 + r6], xm6
+    lea             r8, [r8 + r3 * 4]
+    movq            [r8], xm5
+    movq            [r8 + r3], xm2
+    movhps          [r8 + r3 * 2], xm5
+    movhps          [r8 + r6], xm2
+    lea             r2, [r8 + r3 * 4 - 16]
+%else
+    psubw           m0, m7                          ; m0 = word: row 8, row 9
+    psubw           m6, m7                          ; m6 = word: row 10, row 11
+    psubw           m5, m7                          ; m5 = word: row 12, row 13
+    psubw           m2, m7                          ; m2 = word: row 14, row 15
+    vextracti128    xm3, m0, 1
+    movu            [r8], xm0
+    movu            [r8 + r3], xm3
+    vextracti128    xm3, m6, 1
+    movu            [r8 + r3 * 2], xm6
+    movu            [r8 + r6], xm3
+    vextracti128    xm3, m5, 1
+    lea             r8, [r8 + r3 * 4]
+    movu            [r8], xm5
+    movu            [r8 + r3], xm3
+    vextracti128    xm3, m2, 1
+    movu            [r8 + r3 * 2], xm2
+    movu            [r8 + r6], xm3
+    lea             r2, [r8 + r3 * 4 - 32]
+%endif
+    lea             r0, [r7 - 16]
+    dec             r5d
+    jnz             .loopH
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_24x32 pp
+    FILTER_VER_CHROMA_AVX2_24x32 ps
+
+%macro FILTER_VER_CHROMA_AVX2_24x64 1
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_24x64, 4, 7, 13
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    mova            m10, [r5]
+    mova            m11, [r5 + mmsize]
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m12, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m12, [pw_2000]
+%endif
+    lea             r5, [r3 * 3]
+    mov             r6d, 16
+.loopH:
+    movu            m0, [r0]                        ; m0 = row 0
+    movu            m1, [r0 + r1]                   ; m1 = row 1
+    punpcklbw       m2, m0, m1
+    punpckhbw       m3, m0, m1
+    pmaddubsw       m2, m10
+    pmaddubsw       m3, m10
+    movu            m0, [r0 + r1 * 2]               ; m0 = row 2
+    punpcklbw       m4, m1, m0
+    punpckhbw       m5, m1, m0
+    pmaddubsw       m4, m10
+    pmaddubsw       m5, m10
+    movu            m1, [r0 + r4]                   ; m1 = row 3
+    punpcklbw       m6, m0, m1
+    punpckhbw       m7, m0, m1
+    pmaddubsw       m8, m6, m11
+    pmaddubsw       m9, m7, m11
+    pmaddubsw       m6, m10
+    pmaddubsw       m7, m10
+    paddw           m2, m8
+    paddw           m3, m9
+%ifidn %1,pp
+    pmulhrsw        m2, m12
+    pmulhrsw        m3, m12
+    packuswb        m2, m3
+    movu            [r2], xm2
+    vextracti128    xm2, m2, 1
+    movq            [r2 + 16], xm2
+%else
+    psubw           m2, m12
+    psubw           m3, m12
+    vperm2i128      m0, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    movu            [r2], m0
+    movu            [r2 + mmsize], xm2
+%endif
+    lea             r0, [r0 + r1 * 4]
+    movu            m0, [r0]                        ; m0 = row 4
+    punpcklbw       m2, m1, m0
+    punpckhbw       m3, m1, m0
+    pmaddubsw       m8, m2, m11
+    pmaddubsw       m9, m3, m11
+    pmaddubsw       m2, m10
+    pmaddubsw       m3, m10
+    paddw           m4, m8
+    paddw           m5, m9
+%ifidn %1,pp
+    pmulhrsw        m4, m12
+    pmulhrsw        m5, m12
+    packuswb        m4, m5
+    movu            [r2 + r3], xm4
+    vextracti128    xm4, m4, 1
+    movq            [r2 + r3 + 16], xm4
+%else
+    psubw           m4, m12
+    psubw           m5, m12
+    vperm2i128      m1, m4, m5, 0x20
+    vperm2i128      m4, m4, m5, 0x31
+    movu            [r2 + r3], m1
+    movu            [r2 + r3 + mmsize], xm4
+%endif
+
+    movu            m1, [r0 + r1]                   ; m1 = row 5
+    punpcklbw       m4, m0, m1
+    punpckhbw       m5, m0, m1
+    pmaddubsw       m4, m11
+    pmaddubsw       m5, m11
+    paddw           m6, m4
+    paddw           m7, m5
+%ifidn %1,pp
+    pmulhrsw        m6, m12
+    pmulhrsw        m7, m12
+    packuswb        m6, m7
+    movu            [r2 + r3 * 2], xm6
+    vextracti128    xm6, m6, 1
+    movq            [r2 + r3 * 2 + 16], xm6
+%else
+    psubw           m6, m12
+    psubw           m7, m12
+    vperm2i128      m0, m6, m7, 0x20
+    vperm2i128      m6, m6, m7, 0x31
+    movu            [r2 + r3 * 2], m0
+    movu            [r2 + r3 * 2 + mmsize], xm6
+%endif
+
+    movu            m0, [r0 + r1 * 2]               ; m0 = row 6
+    punpcklbw       m6, m1, m0
+    punpckhbw       m7, m1, m0
+    pmaddubsw       m6, m11
+    pmaddubsw       m7, m11
+    paddw           m2, m6
+    paddw           m3, m7
+%ifidn %1,pp
+    pmulhrsw        m2, m12
+    pmulhrsw        m3, m12
+    packuswb        m2, m3
+    movu            [r2 + r5], xm2
+    vextracti128    xm2, m2, 1
+    movq            [r2 + r5 + 16], xm2
+%else
+    psubw           m2, m12
+    psubw           m3, m12
+    vperm2i128      m0, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    movu            [r2 + r5], m0
+    movu            [r2 + r5 + mmsize], xm2
+%endif
+    lea             r2, [r2 + r3 * 4]
+    dec             r6d
+    jnz             .loopH
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_24x64 pp
+    FILTER_VER_CHROMA_AVX2_24x64 ps
+
+%macro FILTER_VER_CHROMA_AVX2_16x4 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_16x4, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m7, [pw_512]
+%else
+    add             r3d, r3d
+    mova            m7, [pw_2000]
+%endif
+
+    movu            xm0, [r0]
+    vinserti128     m0, m0, [r0 + r1 * 2], 1
+    movu            xm1, [r0 + r1]
+    vinserti128     m1, m1, [r0 + r4], 1
+
+    punpcklbw       m2, m0, m1
+    punpckhbw       m3, m0, m1
+    vperm2i128      m4, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    pmaddubsw       m4, [r5]
+    pmaddubsw       m3, m2, [r5 + mmsize]
+    paddw           m4, m3
+    pmaddubsw       m2, [r5]
+
+    vextracti128    xm0, m0, 1
+    lea             r0, [r0 + r1 * 4]
+    vinserti128     m0, m0, [r0], 1
+
+    punpcklbw       m5, m1, m0
+    punpckhbw       m3, m1, m0
+    vperm2i128      m6, m5, m3, 0x20
+    vperm2i128      m5, m5, m3, 0x31
+    pmaddubsw       m6, [r5]
+    pmaddubsw       m3, m5, [r5 + mmsize]
+    paddw           m6, m3
+    pmaddubsw       m5, [r5]
+%ifidn %1,pp
+    pmulhrsw        m4, m7                          ; m4 = word: row 0
+    pmulhrsw        m6, m7                          ; m6 = word: row 1
+    packuswb        m4, m6
+    vpermq          m4, m4, 11011000b
+    vextracti128    xm6, m4, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm6
+%else
+    psubw           m4, m7                          ; m4 = word: row 0
+    psubw           m6, m7                          ; m6 = word: row 1
+    movu            [r2], m4
+    movu            [r2 + r3], m6
+%endif
+    lea             r2, [r2 + r3 * 2]
+
+    movu            xm4, [r0 + r1 * 2]
+    vinserti128     m4, m4, [r0 + r1], 1
+    vextracti128    xm1, m4, 1
+    vinserti128     m0, m0, xm1, 0
+
+    punpcklbw       m6, m0, m4
+    punpckhbw       m1, m0, m4
+    vperm2i128      m0, m6, m1, 0x20
+    vperm2i128      m6, m6, m1, 0x31
+    pmaddubsw       m0, [r5 + mmsize]
+    paddw           m5, m0
+    pmaddubsw       m6, [r5 + mmsize]
+    paddw           m2, m6
+
+%ifidn %1,pp
+    pmulhrsw        m2, m7                          ; m2 = word: row 2
+    pmulhrsw        m5, m7                          ; m5 = word: row 3
+    packuswb        m2, m5
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm5, m2, 1
+    movu            [r2], xm2
+    movu            [r2 + r3], xm5
+%else
+    psubw           m2, m7                          ; m2 = word: row 2
+    psubw           m5, m7                          ; m5 = word: row 3
+    movu            [r2], m2
+    movu            [r2 + r3], m5
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_16x4 pp
+    FILTER_VER_CHROMA_AVX2_16x4 ps
+
+%macro FILTER_VER_CHROMA_AVX2_12xN 2
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_12x%2, 4, 7, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m7, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m7, [pw_2000]
+%endif
+    lea             r6, [r3 * 3]
+%rep %2 / 16
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhbw       xm2, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddubsw       m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhbw       xm3, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhbw       xm4, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddubsw       m4, m2, [r5 + 1 * mmsize]
+    paddw           m0, m4
+    pmaddubsw       m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhbw       xm5, xm3, xm4
+    punpcklbw       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddubsw       m5, m3, [r5 + 1 * mmsize]
+    paddw           m1, m5
+    pmaddubsw       m3, [r5]
+%ifidn %1,pp
+    pmulhrsw        m0, m7                          ; m0 = word: row 0
+    pmulhrsw        m1, m7                          ; m1 = word: row 1
+    packuswb        m0, m1
+    vextracti128    xm1, m0, 1
+    movq            [r2], xm0
+    movd            [r2 + 8], xm1
+    movhps          [r2 + r3], xm0
+    pextrd          [r2 + r3 + 8], xm1, 2
+%else
+    psubw           m0, m7                          ; m0 = word: row 0
+    psubw           m1, m7                          ; m1 = word: row 1
+    movu            [r2], xm0
+    vextracti128    xm0, m0, 1
+    movq            [r2 + 16], xm0
+    movu            [r2 + r3], xm1
+    vextracti128    xm1, m1, 1
+    movq            [r2 + r3 + 16], xm1
+%endif
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhbw       xm6, xm4, xm5
+    punpcklbw       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddubsw       m6, m4, [r5 + 1 * mmsize]
+    paddw           m2, m6
+    pmaddubsw       m4, [r5]
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhbw       xm0, xm5, xm6
+    punpcklbw       xm5, xm6
+    vinserti128     m5, m5, xm0, 1
+    pmaddubsw       m0, m5, [r5 + 1 * mmsize]
+    paddw           m3, m0
+    pmaddubsw       m5, [r5]
+%ifidn %1,pp
+    pmulhrsw        m2, m7                          ; m2 = word: row 2
+    pmulhrsw        m3, m7                          ; m3 = word: row 3
+    packuswb        m2, m3
+    vextracti128    xm3, m2, 1
+    movq            [r2 + r3 * 2], xm2
+    movd            [r2 + r3 * 2 + 8], xm3
+    movhps          [r2 + r6], xm2
+    pextrd          [r2 + r6 + 8], xm3, 2
+%else
+    psubw           m2, m7                          ; m2 = word: row 2
+    psubw           m3, m7                          ; m3 = word: row 3
+    movu            [r2 + r3 * 2], xm2
+    vextracti128    xm2, m2, 1
+    movq            [r2 + r3 * 2 + 16], xm2
+    movu            [r2 + r6], xm3
+    vextracti128    xm3, m3, 1
+    movq            [r2 + r6 + 16], xm3
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm0, [r0 + r4]                  ; m0 = row 7
+    punpckhbw       xm3, xm6, xm0
+    punpcklbw       xm6, xm0
+    vinserti128     m6, m6, xm3, 1
+    pmaddubsw       m3, m6, [r5 + 1 * mmsize]
+    paddw           m4, m3
+    pmaddubsw       m6, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm3, [r0]                       ; m3 = row 8
+    punpckhbw       xm1, xm0, xm3
+    punpcklbw       xm0, xm3
+    vinserti128     m0, m0, xm1, 1
+    pmaddubsw       m1, m0, [r5 + 1 * mmsize]
+    paddw           m5, m1
+    pmaddubsw       m0, [r5]
+%ifidn %1,pp
+    pmulhrsw        m4, m7                          ; m4 = word: row 4
+    pmulhrsw        m5, m7                          ; m5 = word: row 5
+    packuswb        m4, m5
+    vextracti128    xm5, m4, 1
+    movq            [r2], xm4
+    movd            [r2 + 8], xm5
+    movhps          [r2 + r3], xm4
+    pextrd          [r2 + r3 + 8], xm5, 2
+%else
+    psubw           m4, m7                          ; m4 = word: row 4
+    psubw           m5, m7                          ; m5 = word: row 5
+    movu            [r2], xm4
+    vextracti128    xm4, m4, 1
+    movq            [r2 + 16], xm4
+    movu            [r2 + r3], xm5
+    vextracti128    xm5, m5, 1
+    movq            [r2 + r3 + 16], xm5
+%endif
+
+    movu            xm1, [r0 + r1]                  ; m1 = row 9
+    punpckhbw       xm2, xm3, xm1
+    punpcklbw       xm3, xm1
+    vinserti128     m3, m3, xm2, 1
+    pmaddubsw       m2, m3, [r5 + 1 * mmsize]
+    paddw           m6, m2
+    pmaddubsw       m3, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 10
+    punpckhbw       xm4, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm4, 1
+    pmaddubsw       m4, m1, [r5 + 1 * mmsize]
+    paddw           m0, m4
+    pmaddubsw       m1, [r5]
+
+%ifidn %1,pp
+    pmulhrsw        m6, m7                          ; m6 = word: row 6
+    pmulhrsw        m0, m7                          ; m0 = word: row 7
+    packuswb        m6, m0
+    vextracti128    xm0, m6, 1
+    movq            [r2 + r3 * 2], xm6
+    movd            [r2 + r3 * 2 + 8], xm0
+    movhps          [r2 + r6], xm6
+    pextrd          [r2 + r6 + 8], xm0, 2
+%else
+    psubw           m6, m7                          ; m6 = word: row 6
+    psubw           m0, m7                          ; m0 = word: row 7
+    movu            [r2 + r3 * 2], xm6
+    vextracti128    xm6, m6, 1
+    movq            [r2 + r3 * 2 + 16], xm6
+    movu            [r2 + r6], xm0
+    vextracti128    xm0, m0, 1
+    movq            [r2 + r6 + 16], xm0
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm4, [r0 + r4]                  ; m4 = row 11
+    punpckhbw       xm6, xm2, xm4
+    punpcklbw       xm2, xm4
+    vinserti128     m2, m2, xm6, 1
+    pmaddubsw       m6, m2, [r5 + 1 * mmsize]
+    paddw           m3, m6
+    pmaddubsw       m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm6, [r0]                       ; m6 = row 12
+    punpckhbw       xm0, xm4, xm6
+    punpcklbw       xm4, xm6
+    vinserti128     m4, m4, xm0, 1
+    pmaddubsw       m0, m4, [r5 + 1 * mmsize]
+    paddw           m1, m0
+    pmaddubsw       m4, [r5]
+%ifidn %1,pp
+    pmulhrsw        m3, m7                          ; m3 = word: row 8
+    pmulhrsw        m1, m7                          ; m1 = word: row 9
+    packuswb        m3, m1
+    vextracti128    xm1, m3, 1
+    movq            [r2], xm3
+    movd            [r2 + 8], xm1
+    movhps          [r2 + r3], xm3
+    pextrd          [r2 + r3 + 8], xm1, 2
+%else
+    psubw           m3, m7                          ; m3 = word: row 8
+    psubw           m1, m7                          ; m1 = word: row 9
+    movu            [r2], xm3
+    vextracti128    xm3, m3, 1
+    movq            [r2 + 16], xm3
+    movu            [r2 + r3], xm1
+    vextracti128    xm1, m1, 1
+    movq            [r2 + r3 + 16], xm1
+%endif
+
+    movu            xm0, [r0 + r1]                  ; m0 = row 13
+    punpckhbw       xm1, xm6, xm0
+    punpcklbw       xm6, xm0
+    vinserti128     m6, m6, xm1, 1
+    pmaddubsw       m1, m6, [r5 + 1 * mmsize]
+    paddw           m2, m1
+    pmaddubsw       m6, [r5]
+    movu            xm1, [r0 + r1 * 2]              ; m1 = row 14
+    punpckhbw       xm5, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm5, 1
+    pmaddubsw       m5, m0, [r5 + 1 * mmsize]
+    paddw           m4, m5
+    pmaddubsw       m0, [r5]
+%ifidn %1,pp
+    pmulhrsw        m2, m7                          ; m2 = word: row 10
+    pmulhrsw        m4, m7                          ; m4 = word: row 11
+    packuswb        m2, m4
+    vextracti128    xm4, m2, 1
+    movq            [r2 + r3 * 2], xm2
+    movd            [r2 + r3 * 2 + 8], xm4
+    movhps          [r2 + r6], xm2
+    pextrd          [r2 + r6 + 8], xm4, 2
+%else
+    psubw           m2, m7                          ; m2 = word: row 10
+    psubw           m4, m7                          ; m4 = word: row 11
+    movu            [r2 + r3 * 2], xm2
+    vextracti128    xm2, m2, 1
+    movq            [r2 + r3 * 2 + 16], xm2
+    movu            [r2 + r6], xm4
+    vextracti128    xm4, m4, 1
+    movq            [r2 + r6 + 16], xm4
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm5, [r0 + r4]                  ; m5 = row 15
+    punpckhbw       xm2, xm1, xm5
+    punpcklbw       xm1, xm5
+    vinserti128     m1, m1, xm2, 1
+    pmaddubsw       m2, m1, [r5 + 1 * mmsize]
+    paddw           m6, m2
+    pmaddubsw       m1, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm2, [r0]                       ; m2 = row 16
+    punpckhbw       xm3, xm5, xm2
+    punpcklbw       xm5, xm2
+    vinserti128     m5, m5, xm3, 1
+    pmaddubsw       m3, m5, [r5 + 1 * mmsize]
+    paddw           m0, m3
+    pmaddubsw       m5, [r5]
+    movu            xm3, [r0 + r1]                  ; m3 = row 17
+    punpckhbw       xm4, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddubsw       m2, [r5 + 1 * mmsize]
+    paddw           m1, m2
+    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
+    punpckhbw       xm2, xm3, xm4
+    punpcklbw       xm3, xm4
+    vinserti128     m3, m3, xm2, 1
+    pmaddubsw       m3, [r5 + 1 * mmsize]
+    paddw           m5, m3
+
+%ifidn %1,pp
+    pmulhrsw        m6, m7                          ; m6 = word: row 12
+    pmulhrsw        m0, m7                          ; m0 = word: row 13
+    pmulhrsw        m1, m7                          ; m1 = word: row 14
+    pmulhrsw        m5, m7                          ; m5 = word: row 15
+    packuswb        m6, m0
+    packuswb        m1, m5
+    vextracti128    xm0, m6, 1
+    vextracti128    xm5, m1, 1
+    movq            [r2], xm6
+    movd            [r2 + 8], xm0
+    movhps          [r2 + r3], xm6
+    pextrd          [r2 + r3 + 8], xm0, 2
+    movq            [r2 + r3 * 2], xm1
+    movd            [r2 + r3 * 2 + 8], xm5
+    movhps          [r2 + r6], xm1
+    pextrd          [r2 + r6 + 8], xm5, 2
+%else
+    psubw           m6, m7                          ; m6 = word: row 12
+    psubw           m0, m7                          ; m0 = word: row 13
+    psubw           m1, m7                          ; m1 = word: row 14
+    psubw           m5, m7                          ; m5 = word: row 15
+    movu            [r2], xm6
+    vextracti128    xm6, m6, 1
+    movq            [r2 + 16], xm6
+    movu            [r2 + r3], xm0
+    vextracti128    xm0, m0, 1
+    movq            [r2 + r3 + 16], xm0
+    movu            [r2 + r3 * 2], xm1
+    vextracti128    xm1, m1, 1
+    movq            [r2 + r3 * 2 + 16], xm1
+    movu            [r2 + r6], xm5
+    vextracti128    xm5, m5, 1
+    movq            [r2 + r6 + 16], xm5
+%endif
+    lea             r2, [r2 + r3 * 4]
+%endrep
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_12xN pp, 16
+    FILTER_VER_CHROMA_AVX2_12xN ps, 16
+    FILTER_VER_CHROMA_AVX2_12xN pp, 32
+    FILTER_VER_CHROMA_AVX2_12xN ps, 32
+
+;-----------------------------------------------------------------------------
+;void interp_4tap_vert_pp_24x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W24 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_24x%2, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
+
+    mov         r4d,       %2
+
+.loop:
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
+
+    lea         r5,        [r0 + 2 * r1]
+    movu        m5,        [r5]
+    movu        m7,        [r5 + r1]
+
+    punpcklbw   m6,        m5,        m7
+    pmaddubsw   m6,        m0
+    paddw       m4,        m6
+
+    punpckhbw   m6,        m5,        m7
+    pmaddubsw   m6,        m0
+    paddw       m2,        m6
+
+    mova        m6,        [pw_512]
+
+    pmulhrsw    m4,        m6
+    pmulhrsw    m2,        m6
+
+    packuswb    m4,        m2
+
+    movu        [r2],      m4
+
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m3,        m1
+
+    movu        m2,        [r5 + 2 * r1]
+
+    punpcklbw   m5,        m7,        m2
+    punpckhbw   m7,        m2
+
+    pmaddubsw   m5,        m0
+    pmaddubsw   m7,        m0
+
+    paddw       m4,        m5
+    paddw       m3,        m7
+
+    pmulhrsw    m4,        m6
+    pmulhrsw    m3,        m6
+
+    packuswb    m4,        m3
+
+    movu        [r2 + r3],      m4
+
+    movq        m2,        [r0 + 16]
+    movq        m3,        [r0 + r1 + 16]
+    movq        m4,        [r5 + 16]
+    movq        m5,        [r5 + r1 + 16]
+
+    punpcklbw   m2,        m3
+    punpcklbw   m4,        m5
+
+    pmaddubsw   m2,        m1
+    pmaddubsw   m4,        m0
+
+    paddw       m2,        m4
+
+    pmulhrsw    m2,        m6
+
+    movq        m3,        [r0 + r1 + 16]
+    movq        m4,        [r5 + 16]
+    movq        m5,        [r5 + r1 + 16]
+    movq        m7,        [r5 + 2 * r1 + 16]
+
+    punpcklbw   m3,        m4
+    punpcklbw   m5,        m7
+
+    pmaddubsw   m3,        m1
+    pmaddubsw   m5,        m0
+
+    paddw       m3,        m5
+
+    pmulhrsw    m3,        m6
+    packuswb    m2,        m3
+
+    movh        [r2 + 16], m2
+    movhps      [r2 + r3 + 16], m2
+
+    mov         r0,        r5
+    lea         r2,        [r2 + 2 * r3]
+
+    sub         r4,        2
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W24 24, 32
+
+    FILTER_V4_W24 24, 64
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W32 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
+
+    mova        m7,        [pw_512]
+
+    mov         r4d,       %2
+
+.loop:
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
+
+    lea         r5,        [r0 + 2 * r1]
+    movu        m3,        [r5]
+    movu        m5,        [r5 + r1]
+
+    punpcklbw   m6,        m3,        m5
+    punpckhbw   m3,        m5
+
+    pmaddubsw   m6,        m0
+    pmaddubsw   m3,        m0
+
+    paddw       m4,        m6
+    paddw       m2,        m3
+
+    pmulhrsw    m4,        m7
+    pmulhrsw    m2,        m7
+
+    packuswb    m4,        m2
+
+    movu        [r2],      m4
+
+    movu        m2,        [r0 + 16]
+    movu        m3,        [r0 + r1 + 16]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
+
+    movu        m3,        [r5 + 16]
+    movu        m5,        [r5 + r1 + 16]
+
+    punpcklbw   m6,        m3,        m5
+    punpckhbw   m3,        m5
+
+    pmaddubsw   m6,        m0
+    pmaddubsw   m3,        m0
+
+    paddw       m4,        m6
+    paddw       m2,        m3
+
+    pmulhrsw    m4,        m7
+    pmulhrsw    m2,        m7
+
+    packuswb    m4,        m2
+
+    movu        [r2 + 16], m4
+
+    lea         r0,        [r0 + r1]
+    lea         r2,        [r2 + r3]
+
+    dec         r4
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W32 32,  8
+    FILTER_V4_W32 32, 16
+    FILTER_V4_W32 32, 24
+    FILTER_V4_W32 32, 32
+
+    FILTER_V4_W32 32, 48
+    FILTER_V4_W32 32, 64
+
+%macro FILTER_VER_CHROMA_AVX2_32xN 2
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_32x%2, 4, 7, 13
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    mova            m10, [r5]
+    mova            m11, [r5 + mmsize]
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m12, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m12, [pw_2000]
+%endif
+    lea             r5, [r3 * 3]
+    mov             r6d, %2 / 4
+.loopW:
+    movu            m0, [r0]                        ; m0 = row 0
+    movu            m1, [r0 + r1]                   ; m1 = row 1
+    punpcklbw       m2, m0, m1
+    punpckhbw       m3, m0, m1
+    pmaddubsw       m2, m10
+    pmaddubsw       m3, m10
+    movu            m0, [r0 + r1 * 2]               ; m0 = row 2
+    punpcklbw       m4, m1, m0
+    punpckhbw       m5, m1, m0
+    pmaddubsw       m4, m10
+    pmaddubsw       m5, m10
+    movu            m1, [r0 + r4]                   ; m1 = row 3
+    punpcklbw       m6, m0, m1
+    punpckhbw       m7, m0, m1
+    pmaddubsw       m8, m6, m11
+    pmaddubsw       m9, m7, m11
+    pmaddubsw       m6, m10
+    pmaddubsw       m7, m10
+    paddw           m2, m8
+    paddw           m3, m9
+%ifidn %1,pp
+    pmulhrsw        m2, m12
+    pmulhrsw        m3, m12
+    packuswb        m2, m3
+    movu            [r2], m2
+%else
+    psubw           m2, m12
+    psubw           m3, m12
+    vperm2i128      m0, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    movu            [r2], m0
+    movu            [r2 + mmsize], m2
+%endif
+    lea             r0, [r0 + r1 * 4]
+    movu            m0, [r0]                        ; m0 = row 4
+    punpcklbw       m2, m1, m0
+    punpckhbw       m3, m1, m0
+    pmaddubsw       m8, m2, m11
+    pmaddubsw       m9, m3, m11
+    pmaddubsw       m2, m10
+    pmaddubsw       m3, m10
+    paddw           m4, m8
+    paddw           m5, m9
+%ifidn %1,pp
+    pmulhrsw        m4, m12
+    pmulhrsw        m5, m12
+    packuswb        m4, m5
+    movu            [r2 + r3], m4
+%else
+    psubw           m4, m12
+    psubw           m5, m12
+    vperm2i128      m1, m4, m5, 0x20
+    vperm2i128      m4, m4, m5, 0x31
+    movu            [r2 + r3], m1
+    movu            [r2 + r3 + mmsize], m4
+%endif
+
+    movu            m1, [r0 + r1]                   ; m1 = row 5
+    punpcklbw       m4, m0, m1
+    punpckhbw       m5, m0, m1
+    pmaddubsw       m4, m11
+    pmaddubsw       m5, m11
+    paddw           m6, m4
+    paddw           m7, m5
+%ifidn %1,pp
+    pmulhrsw        m6, m12
+    pmulhrsw        m7, m12
+    packuswb        m6, m7
+    movu            [r2 + r3 * 2], m6
+%else
+    psubw           m6, m12
+    psubw           m7, m12
+    vperm2i128      m0, m6, m7, 0x20
+    vperm2i128      m6, m6, m7, 0x31
+    movu            [r2 + r3 * 2], m0
+    movu            [r2 + r3 * 2 + mmsize], m6
+%endif
+
+    movu            m0, [r0 + r1 * 2]               ; m0 = row 6
+    punpcklbw       m6, m1, m0
+    punpckhbw       m7, m1, m0
+    pmaddubsw       m6, m11
+    pmaddubsw       m7, m11
+    paddw           m2, m6
+    paddw           m3, m7
+%ifidn %1,pp
+    pmulhrsw        m2, m12
+    pmulhrsw        m3, m12
+    packuswb        m2, m3
+    movu            [r2 + r5], m2
+%else
+    psubw           m2, m12
+    psubw           m3, m12
+    vperm2i128      m0, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    movu            [r2 + r5], m0
+    movu            [r2 + r5 + mmsize], m2
+%endif
+    lea             r2, [r2 + r3 * 4]
+    dec             r6d
+    jnz             .loopW
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_32xN pp, 64
+    FILTER_VER_CHROMA_AVX2_32xN pp, 48
+    FILTER_VER_CHROMA_AVX2_32xN pp, 32
+    FILTER_VER_CHROMA_AVX2_32xN pp, 24
+    FILTER_VER_CHROMA_AVX2_32xN pp, 16
+    FILTER_VER_CHROMA_AVX2_32xN pp, 8
+    FILTER_VER_CHROMA_AVX2_32xN ps, 64
+    FILTER_VER_CHROMA_AVX2_32xN ps, 48
+    FILTER_VER_CHROMA_AVX2_32xN ps, 32
+    FILTER_VER_CHROMA_AVX2_32xN ps, 24
+    FILTER_VER_CHROMA_AVX2_32xN ps, 16
+    FILTER_VER_CHROMA_AVX2_32xN ps, 8
+
+%macro FILTER_VER_CHROMA_AVX2_48x64 1
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_48x64, 4, 8, 13
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    mova            m10, [r5]
+    mova            m11, [r5 + mmsize]
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m12, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m12, [pw_2000]
+%endif
+    lea             r5, [r3 * 3]
+    lea             r7, [r1 * 4]
+    mov             r6d, 16
+.loopH:
+    movu            m0, [r0]                        ; m0 = row 0
+    movu            m1, [r0 + r1]                   ; m1 = row 1
+    punpcklbw       m2, m0, m1
+    punpckhbw       m3, m0, m1
+    pmaddubsw       m2, m10
+    pmaddubsw       m3, m10
+    movu            m0, [r0 + r1 * 2]               ; m0 = row 2
+    punpcklbw       m4, m1, m0
+    punpckhbw       m5, m1, m0
+    pmaddubsw       m4, m10
+    pmaddubsw       m5, m10
+    movu            m1, [r0 + r4]                   ; m1 = row 3
+    punpcklbw       m6, m0, m1
+    punpckhbw       m7, m0, m1
+    pmaddubsw       m8, m6, m11
+    pmaddubsw       m9, m7, m11
+    pmaddubsw       m6, m10
+    pmaddubsw       m7, m10
+    paddw           m2, m8
+    paddw           m3, m9
+%ifidn %1,pp
+    pmulhrsw        m2, m12
+    pmulhrsw        m3, m12
+    packuswb        m2, m3
+    movu            [r2], m2
+%else
+    psubw           m2, m12
+    psubw           m3, m12
+    vperm2i128      m0, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    movu            [r2], m0
+    movu            [r2 + mmsize], m2
+%endif
+    lea             r0, [r0 + r1 * 4]
+    movu            m0, [r0]                        ; m0 = row 4
+    punpcklbw       m2, m1, m0
+    punpckhbw       m3, m1, m0
+    pmaddubsw       m8, m2, m11
+    pmaddubsw       m9, m3, m11
+    pmaddubsw       m2, m10
+    pmaddubsw       m3, m10
+    paddw           m4, m8
+    paddw           m5, m9
+%ifidn %1,pp
+    pmulhrsw        m4, m12
+    pmulhrsw        m5, m12
+    packuswb        m4, m5
+    movu            [r2 + r3], m4
+%else
+    psubw           m4, m12
+    psubw           m5, m12
+    vperm2i128      m1, m4, m5, 0x20
+    vperm2i128      m4, m4, m5, 0x31
+    movu            [r2 + r3], m1
+    movu            [r2 + r3 + mmsize], m4
+%endif
+
+    movu            m1, [r0 + r1]                   ; m1 = row 5
+    punpcklbw       m4, m0, m1
+    punpckhbw       m5, m0, m1
+    pmaddubsw       m4, m11
+    pmaddubsw       m5, m11
+    paddw           m6, m4
+    paddw           m7, m5
+%ifidn %1,pp
+    pmulhrsw        m6, m12
+    pmulhrsw        m7, m12
+    packuswb        m6, m7
+    movu            [r2 + r3 * 2], m6
+%else
+    psubw           m6, m12
+    psubw           m7, m12
+    vperm2i128      m0, m6, m7, 0x20
+    vperm2i128      m6, m6, m7, 0x31
+    movu            [r2 + r3 * 2], m0
+    movu            [r2 + r3 * 2 + mmsize], m6
+%endif
+
+    movu            m0, [r0 + r1 * 2]               ; m0 = row 6
+    punpcklbw       m6, m1, m0
+    punpckhbw       m7, m1, m0
+    pmaddubsw       m6, m11
+    pmaddubsw       m7, m11
+    paddw           m2, m6
+    paddw           m3, m7
+%ifidn %1,pp
+    pmulhrsw        m2, m12
+    pmulhrsw        m3, m12
+    packuswb        m2, m3
+    movu            [r2 + r5], m2
+    add             r2, 32
+%else
+    psubw           m2, m12
+    psubw           m3, m12
+    vperm2i128      m0, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    movu            [r2 + r5], m0
+    movu            [r2 + r5 + mmsize], m2
+    add             r2, 64
+%endif
+    sub             r0, r7
+
+    movu            xm0, [r0 + 32]                  ; m0 = row 0
+    movu            xm1, [r0 + r1 + 32]             ; m1 = row 1
+    punpckhbw       xm2, xm0, xm1
+    punpcklbw       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddubsw       m0, m10
+    movu            xm2, [r0 + r1 * 2 + 32]         ; m2 = row 2
+    punpckhbw       xm3, xm1, xm2
+    punpcklbw       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddubsw       m1, m10
+    movu            xm3, [r0 + r4 + 32]             ; m3 = row 3
+    punpckhbw       xm4, xm2, xm3
+    punpcklbw       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddubsw       m4, m2, m11
+    paddw           m0, m4
+    pmaddubsw       m2, m10
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0 + 32]                  ; m4 = row 4
+    punpckhbw       xm5, xm3, xm4
+    punpcklbw       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddubsw       m5, m3, m11
+    paddw           m1, m5
+    pmaddubsw       m3, m10
+    movu            xm5, [r0 + r1 + 32]             ; m5 = row 5
+    punpckhbw       xm6, xm4, xm5
+    punpcklbw       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddubsw       m4, m11
+    paddw           m2, m4
+    movu            xm6, [r0 + r1 * 2 + 32]         ; m6 = row 6
+    punpckhbw       xm7, xm5, xm6
+    punpcklbw       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddubsw       m5, m11
+    paddw           m3, m5
+%ifidn %1,pp
+    pmulhrsw        m0, m12                         ; m0 = word: row 0
+    pmulhrsw        m1, m12                         ; m1 = word: row 1
+    pmulhrsw        m2, m12                         ; m2 = word: row 2
+    pmulhrsw        m3, m12                         ; m3 = word: row 3
+    packuswb        m0, m1
+    packuswb        m2, m3
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r5], xm3
+    lea             r2, [r2 + r3 * 4 - 32]
+%else
+    psubw           m0, m12                         ; m0 = word: row 0
+    psubw           m1, m12                         ; m1 = word: row 1
+    psubw           m2, m12                         ; m2 = word: row 2
+    psubw           m3, m12                         ; m3 = word: row 3
+    movu            [r2], m0
+    movu            [r2 + r3], m1
+    movu            [r2 + r3 * 2], m2
+    movu            [r2 + r5], m3
+    lea             r2, [r2 + r3 * 4 - 64]
+%endif
+    dec             r6d
+    jnz             .loopH
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_48x64 pp
+    FILTER_VER_CHROMA_AVX2_48x64 ps
+
+%macro FILTER_VER_CHROMA_AVX2_64xN 2
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_64x%2, 4, 8, 13
+    mov             r4d, r4m
+    shl             r4d, 6
+
+%ifdef PIC
+    lea             r5, [tab_ChromaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_ChromaCoeffVer_32 + r4]
+%endif
+
+    mova            m10, [r5]
+    mova            m11, [r5 + mmsize]
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,pp
+    mova            m12, [pw_512]
+%else
+    add             r3d, r3d
+    vbroadcasti128  m12, [pw_2000]
+%endif
+    lea             r5, [r3 * 3]
+    lea             r7, [r1 * 4]
+    mov             r6d, %2 / 4
+.loopH:
+%assign x 0
+%rep 2
+    movu            m0, [r0 + x]                    ; m0 = row 0
+    movu            m1, [r0 + r1 + x]               ; m1 = row 1
+    punpcklbw       m2, m0, m1
+    punpckhbw       m3, m0, m1
+    pmaddubsw       m2, m10
+    pmaddubsw       m3, m10
+    movu            m0, [r0 + r1 * 2 + x]           ; m0 = row 2
+    punpcklbw       m4, m1, m0
+    punpckhbw       m5, m1, m0
+    pmaddubsw       m4, m10
+    pmaddubsw       m5, m10
+    movu            m1, [r0 + r4 + x]               ; m1 = row 3
+    punpcklbw       m6, m0, m1
+    punpckhbw       m7, m0, m1
+    pmaddubsw       m8, m6, m11
+    pmaddubsw       m9, m7, m11
+    pmaddubsw       m6, m10
+    pmaddubsw       m7, m10
+    paddw           m2, m8
+    paddw           m3, m9
+%ifidn %1,pp
+    pmulhrsw        m2, m12
+    pmulhrsw        m3, m12
+    packuswb        m2, m3
+    movu            [r2], m2
+%else
+    psubw           m2, m12
+    psubw           m3, m12
+    vperm2i128      m0, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    movu            [r2], m0
+    movu            [r2 + mmsize], m2
+%endif
+    lea             r0, [r0 + r1 * 4]
+    movu            m0, [r0 + x]                    ; m0 = row 4
+    punpcklbw       m2, m1, m0
+    punpckhbw       m3, m1, m0
+    pmaddubsw       m8, m2, m11
+    pmaddubsw       m9, m3, m11
+    pmaddubsw       m2, m10
+    pmaddubsw       m3, m10
+    paddw           m4, m8
+    paddw           m5, m9
+%ifidn %1,pp
+    pmulhrsw        m4, m12
+    pmulhrsw        m5, m12
+    packuswb        m4, m5
+    movu            [r2 + r3], m4
+%else
+    psubw           m4, m12
+    psubw           m5, m12
+    vperm2i128      m1, m4, m5, 0x20
+    vperm2i128      m4, m4, m5, 0x31
+    movu            [r2 + r3], m1
+    movu            [r2 + r3 + mmsize], m4
+%endif
+
+    movu            m1, [r0 + r1 + x]               ; m1 = row 5
+    punpcklbw       m4, m0, m1
+    punpckhbw       m5, m0, m1
+    pmaddubsw       m4, m11
+    pmaddubsw       m5, m11
+    paddw           m6, m4
+    paddw           m7, m5
+%ifidn %1,pp
+    pmulhrsw        m6, m12
+    pmulhrsw        m7, m12
+    packuswb        m6, m7
+    movu            [r2 + r3 * 2], m6
+%else
+    psubw           m6, m12
+    psubw           m7, m12
+    vperm2i128      m0, m6, m7, 0x20
+    vperm2i128      m6, m6, m7, 0x31
+    movu            [r2 + r3 * 2], m0
+    movu            [r2 + r3 * 2 + mmsize], m6
+%endif
+
+    movu            m0, [r0 + r1 * 2 + x]           ; m0 = row 6
+    punpcklbw       m6, m1, m0
+    punpckhbw       m7, m1, m0
+    pmaddubsw       m6, m11
+    pmaddubsw       m7, m11
+    paddw           m2, m6
+    paddw           m3, m7
+%ifidn %1,pp
+    pmulhrsw        m2, m12
+    pmulhrsw        m3, m12
+    packuswb        m2, m3
+    movu            [r2 + r5], m2
+    add             r2, 32
+%else
+    psubw           m2, m12
+    psubw           m3, m12
+    vperm2i128      m0, m2, m3, 0x20
+    vperm2i128      m2, m2, m3, 0x31
+    movu            [r2 + r5], m0
+    movu            [r2 + r5 + mmsize], m2
+    add             r2, 64
+%endif
+    sub             r0, r7
+%assign x x+32
+%endrep
+%ifidn %1,pp
+    lea             r2, [r2 + r3 * 4 - 64]
+%else
+    lea             r2, [r2 + r3 * 4 - 128]
+%endif
+    add             r0, r7
+    dec             r6d
+    jnz             .loopH
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_AVX2_64xN pp, 64
+    FILTER_VER_CHROMA_AVX2_64xN pp, 48
+    FILTER_VER_CHROMA_AVX2_64xN pp, 32
+    FILTER_VER_CHROMA_AVX2_64xN pp, 16
+    FILTER_VER_CHROMA_AVX2_64xN ps, 64
+    FILTER_VER_CHROMA_AVX2_64xN ps, 48
+    FILTER_VER_CHROMA_AVX2_64xN ps, 32
+    FILTER_VER_CHROMA_AVX2_64xN ps, 16
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W16n_H2 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_pp_%1x%2, 4, 7, 8
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+
+%ifdef PIC
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
+%else
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
+
+    mov         r4d,       %2/2
+
+.loop:
+
+    mov         r6d,       %1/16
+
+.loopW:
+
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
+
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
+
+    lea         r5,        [r0 + 2 * r1]
+    movu        m5,        [r5]
+    movu        m6,        [r5 + r1]
+
+    punpckhbw   m7,        m5,        m6
+    pmaddubsw   m7,        m0
+    paddw       m2,        m7
+
+    punpcklbw   m7,        m5,        m6
+    pmaddubsw   m7,        m0
+    paddw       m4,        m7
+
+    mova        m7,        [pw_512]
+
+    pmulhrsw    m4,        m7
+    pmulhrsw    m2,        m7
+
+    packuswb    m4,        m2
+
+    movu        [r2],      m4
+
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
+
+    pmaddubsw   m4,        m1
+    pmaddubsw   m3,        m1
+
+    movu        m5,        [r5 + 2 * r1]
+
+    punpcklbw   m2,        m6,        m5
+    punpckhbw   m6,        m5
+
+    pmaddubsw   m2,        m0
+    pmaddubsw   m6,        m0
+
+    paddw       m4,        m2
+    paddw       m3,        m6
+
+    pmulhrsw    m4,        m7
+    pmulhrsw    m3,        m7
+
+    packuswb    m4,        m3
+
+    movu        [r2 + r3],      m4
+
+    add         r0,        16
+    add         r2,        16
+    dec         r6d
+    jnz         .loopW
+
+    lea         r0,        [r0 + r1 * 2 - %1]
+    lea         r2,        [r2 + r3 * 2 - %1]
+
+    dec         r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V4_W16n_H2 64, 64
+    FILTER_V4_W16n_H2 64, 32
+    FILTER_V4_W16n_H2 64, 48
+    FILTER_V4_W16n_H2 48, 64
+    FILTER_V4_W16n_H2 64, 16
+
+%macro PROCESS_CHROMA_SP_W4_4R 0
+    movq       m0, [r0]
+    movq       m1, [r0 + r1]
+    punpcklwd  m0, m1                          ;m0=[0 1]
+    pmaddwd    m0, [r6 + 0 *16]                ;m0=[0+1]         Row1
+
+    lea        r0, [r0 + 2 * r1]
+    movq       m4, [r0]
+    punpcklwd  m1, m4                          ;m1=[1 2]
+    pmaddwd    m1, [r6 + 0 *16]                ;m1=[1+2]         Row2
+
+    movq       m5, [r0 + r1]
+    punpcklwd  m4, m5                          ;m4=[2 3]
+    pmaddwd    m2, m4, [r6 + 0 *16]            ;m2=[2+3]         Row3
+    pmaddwd    m4, [r6 + 1 * 16]
+    paddd      m0, m4                          ;m0=[0+1+2+3]     Row1 done
+
+    lea        r0, [r0 + 2 * r1]
+    movq       m4, [r0]
+    punpcklwd  m5, m4                          ;m5=[3 4]
+    pmaddwd    m3, m5, [r6 + 0 *16]            ;m3=[3+4]         Row4
+    pmaddwd    m5, [r6 + 1 * 16]
+    paddd      m1, m5                          ;m1 = [1+2+3+4]   Row2
+
+    movq       m5, [r0 + r1]
+    punpcklwd  m4, m5                          ;m4=[4 5]
+    pmaddwd    m4, [r6 + 1 * 16]
+    paddd      m2, m4                          ;m2=[2+3+4+5]     Row3
+
+    movq       m4, [r0 + 2 * r1]
+    punpcklwd  m5, m4                          ;m5=[5 6]
+    pmaddwd    m5, [r6 + 1 * 16]
+    paddd      m3, m5                          ;m3=[3+4+5+6]     Row4
+%endmacro
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_sp_%1x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SP 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_sp_%1x%2, 5, 7, 7 ,0-gprsize
+
+    add       r1d, r1d
+    sub       r0, r1
+    shl       r4d, 5
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r6, [r5 + r4]
+%else
+    lea       r6, [tab_ChromaCoeffV + r4]
+%endif
+
+    mova      m6, [v4_pd_526336]
+
+    mov       dword [rsp], %2/4
+
+.loopH:
+    mov       r4d, (%1/4)
+.loopW:
+    PROCESS_CHROMA_SP_W4_4R
+
+    paddd     m0, m6
+    paddd     m1, m6
+    paddd     m2, m6
+    paddd     m3, m6
+
+    psrad     m0, 12
+    psrad     m1, 12
+    psrad     m2, 12
+    psrad     m3, 12
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+
+    packuswb  m0, m2
+
+    movd      [r2], m0
+    pextrd    [r2 + r3], m0, 1
+    lea       r5, [r2 + 2 * r3]
+    pextrd    [r5], m0, 2
+    pextrd    [r5 + r3], m0, 3
+
+    lea       r5, [4 * r1 - 2 * 4]
+    sub       r0, r5
+    add       r2, 4
+
+    dec       r4d
+    jnz       .loopW
+
+    lea       r0, [r0 + 4 * r1 - 2 * %1]
+    lea       r2, [r2 + 4 * r3 - %1]
+
+    dec       dword [rsp]
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SP 4, 4
+    FILTER_VER_CHROMA_SP 4, 8
+    FILTER_VER_CHROMA_SP 16, 16
+    FILTER_VER_CHROMA_SP 16, 8
+    FILTER_VER_CHROMA_SP 16, 12
+    FILTER_VER_CHROMA_SP 12, 16
+    FILTER_VER_CHROMA_SP 16, 4
+    FILTER_VER_CHROMA_SP 4, 16
+    FILTER_VER_CHROMA_SP 32, 32
+    FILTER_VER_CHROMA_SP 32, 16
+    FILTER_VER_CHROMA_SP 16, 32
+    FILTER_VER_CHROMA_SP 32, 24
+    FILTER_VER_CHROMA_SP 24, 32
+    FILTER_VER_CHROMA_SP 32, 8
+
+    FILTER_VER_CHROMA_SP 16, 24
+    FILTER_VER_CHROMA_SP 16, 64
+    FILTER_VER_CHROMA_SP 12, 32
+    FILTER_VER_CHROMA_SP 4, 32
+    FILTER_VER_CHROMA_SP 32, 64
+    FILTER_VER_CHROMA_SP 32, 48
+    FILTER_VER_CHROMA_SP 24, 64
+
+    FILTER_VER_CHROMA_SP 64, 64
+    FILTER_VER_CHROMA_SP 64, 32
+    FILTER_VER_CHROMA_SP 64, 48
+    FILTER_VER_CHROMA_SP 48, 64
+    FILTER_VER_CHROMA_SP 64, 16
+
+
+%macro PROCESS_CHROMA_SP_W2_4R 1
+    movd       m0, [r0]
+    movd       m1, [r0 + r1]
+    punpcklwd  m0, m1                          ;m0=[0 1]
+
+    lea        r0, [r0 + 2 * r1]
+    movd       m2, [r0]
+    punpcklwd  m1, m2                          ;m1=[1 2]
+    punpcklqdq m0, m1                          ;m0=[0 1 1 2]
+    pmaddwd    m0, [%1 + 0 *16]                ;m0=[0+1 1+2] Row 1-2
+
+    movd       m1, [r0 + r1]
+    punpcklwd  m2, m1                          ;m2=[2 3]
+
+    lea        r0, [r0 + 2 * r1]
+    movd       m3, [r0]
+    punpcklwd  m1, m3                          ;m2=[3 4]
+    punpcklqdq m2, m1                          ;m2=[2 3 3 4]
+
+    pmaddwd    m4, m2, [%1 + 1 * 16]           ;m4=[2+3 3+4] Row 1-2
+    pmaddwd    m2, [%1 + 0 * 16]               ;m2=[2+3 3+4] Row 3-4
+    paddd      m0, m4                          ;m0=[0+1+2+3 1+2+3+4] Row 1-2
+
+    movd       m1, [r0 + r1]
+    punpcklwd  m3, m1                          ;m3=[4 5]
+
+    movd       m4, [r0 + 2 * r1]
+    punpcklwd  m1, m4                          ;m1=[5 6]
+    punpcklqdq m3, m1                          ;m2=[4 5 5 6]
+    pmaddwd    m3, [%1 + 1 * 16]               ;m3=[4+5 5+6] Row 3-4
+    paddd      m2, m3                          ;m2=[2+3+4+5 3+4+5+6] Row 3-4
+%endmacro
+
+;-------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vertical_sp_%1x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-------------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SP_W2_4R 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_sp_%1x%2, 5, 6, 6
+
+    add       r1d, r1d
+    sub       r0, r1
+    shl       r4d, 5
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    mova      m5, [v4_pd_526336]
+
+    mov       r4d, (%2/4)
+
+.loopH:
+    PROCESS_CHROMA_SP_W2_4R r5
+
+    paddd     m0, m5
+    paddd     m2, m5
+
+    psrad     m0, 12
+    psrad     m2, 12
+
+    packssdw  m0, m2
+    packuswb  m0, m0
+
+    pextrw    [r2], m0, 0
+    pextrw    [r2 + r3], m0, 1
+    lea       r2, [r2 + 2 * r3]
+    pextrw    [r2], m0, 2
+    pextrw    [r2 + r3], m0, 3
+
+    lea       r2, [r2 + 2 * r3]
+
+    dec       r4d
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SP_W2_4R 2, 4
+    FILTER_VER_CHROMA_SP_W2_4R 2, 8
+
+    FILTER_VER_CHROMA_SP_W2_4R 2, 16
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_sp_4x2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_vert_sp_4x2, 5, 6, 5
+
+    add        r1d, r1d
+    sub        r0, r1
+    shl        r4d, 5
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeffV]
+    lea        r5, [r5 + r4]
+%else
+    lea        r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    mova       m4, [v4_pd_526336]
+
+    movq       m0, [r0]
+    movq       m1, [r0 + r1]
+    punpcklwd  m0, m1                          ;m0=[0 1]
+    pmaddwd    m0, [r5 + 0 *16]                ;m0=[0+1]  Row1
+
+    lea        r0, [r0 + 2 * r1]
+    movq       m2, [r0]
+    punpcklwd  m1, m2                          ;m1=[1 2]
+    pmaddwd    m1, [r5 + 0 *16]                ;m1=[1+2]  Row2
+
+    movq       m3, [r0 + r1]
+    punpcklwd  m2, m3                          ;m4=[2 3]
+    pmaddwd    m2, [r5 + 1 * 16]
+    paddd      m0, m2                          ;m0=[0+1+2+3]  Row1 done
+    paddd      m0, m4
+    psrad      m0, 12
+
+    movq       m2, [r0 + 2 * r1]
+    punpcklwd  m3, m2                          ;m5=[3 4]
+    pmaddwd    m3, [r5 + 1 * 16]
+    paddd      m1, m3                          ;m1 = [1+2+3+4]  Row2 done
+    paddd      m1, m4
+    psrad      m1, 12
+
+    packssdw   m0, m1
+    packuswb   m0, m0
+
+    movd       [r2], m0
+    pextrd     [r2 + r3], m0, 1
+
+    RET
+
+;-------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vertical_sp_6x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-------------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SP_W6_H4 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_sp_6x%2, 5, 7, 7
+
+    add       r1d, r1d
+    sub       r0, r1
+    shl       r4d, 5
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r6, [r5 + r4]
+%else
+    lea       r6, [tab_ChromaCoeffV + r4]
+%endif
+
+    mova      m6, [v4_pd_526336]
+
+    mov       r4d, %2/4
+
+.loopH:
+    PROCESS_CHROMA_SP_W4_4R
+
+    paddd     m0, m6
+    paddd     m1, m6
+    paddd     m2, m6
+    paddd     m3, m6
+
+    psrad     m0, 12
+    psrad     m1, 12
+    psrad     m2, 12
+    psrad     m3, 12
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+
+    packuswb  m0, m2
+
+    movd      [r2], m0
+    pextrd    [r2 + r3], m0, 1
+    lea       r5, [r2 + 2 * r3]
+    pextrd    [r5], m0, 2
+    pextrd    [r5 + r3], m0, 3
+
+    lea       r5, [4 * r1 - 2 * 4]
+    sub       r0, r5
+    add       r2, 4
+
+    PROCESS_CHROMA_SP_W2_4R r6
+
+    paddd     m0, m6
+    paddd     m2, m6
+
+    psrad     m0, 12
+    psrad     m2, 12
+
+    packssdw  m0, m2
+    packuswb  m0, m0
+
+    pextrw    [r2], m0, 0
+    pextrw    [r2 + r3], m0, 1
+    lea       r2, [r2 + 2 * r3]
+    pextrw    [r2], m0, 2
+    pextrw    [r2 + r3], m0, 3
+
+    sub       r0, 2 * 4
+    lea       r2, [r2 + 2 * r3 - 4]
+
+    dec       r4d
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SP_W6_H4 6, 8
+
+    FILTER_VER_CHROMA_SP_W6_H4 6, 16
+
+%macro PROCESS_CHROMA_SP_W8_2R 0
+    movu       m1, [r0]
+    movu       m3, [r0 + r1]
+    punpcklwd  m0, m1, m3
+    pmaddwd    m0, [r5 + 0 * 16]                ;m0 = [0l+1l]  Row1l
+    punpckhwd  m1, m3
+    pmaddwd    m1, [r5 + 0 * 16]                ;m1 = [0h+1h]  Row1h
+
+    movu       m4, [r0 + 2 * r1]
+    punpcklwd  m2, m3, m4
+    pmaddwd    m2, [r5 + 0 * 16]                ;m2 = [1l+2l]  Row2l
+    punpckhwd  m3, m4
+    pmaddwd    m3, [r5 + 0 * 16]                ;m3 = [1h+2h]  Row2h
+
+    lea        r0, [r0 + 2 * r1]
+    movu       m5, [r0 + r1]
+    punpcklwd  m6, m4, m5
+    pmaddwd    m6, [r5 + 1 * 16]                ;m6 = [2l+3l]  Row1l
+    paddd      m0, m6                           ;m0 = [0l+1l+2l+3l]  Row1l sum
+    punpckhwd  m4, m5
+    pmaddwd    m4, [r5 + 1 * 16]                ;m6 = [2h+3h]  Row1h
+    paddd      m1, m4                           ;m1 = [0h+1h+2h+3h]  Row1h sum
+
+    movu       m4, [r0 + 2 * r1]
+    punpcklwd  m6, m5, m4
+    pmaddwd    m6, [r5 + 1 * 16]                ;m6 = [3l+4l]  Row2l
+    paddd      m2, m6                           ;m2 = [1l+2l+3l+4l]  Row2l sum
+    punpckhwd  m5, m4
+    pmaddwd    m5, [r5 + 1 * 16]                ;m1 = [3h+4h]  Row2h
+    paddd      m3, m5                           ;m3 = [1h+2h+3h+4h]  Row2h sum
+%endmacro
+
+;--------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_sp_8x%2(int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;--------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SP_W8_H2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_sp_%1x%2, 5, 6, 8
+
+    add       r1d, r1d
+    sub       r0, r1
+    shl       r4d, 5
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    mova      m7, [v4_pd_526336]
+
+    mov       r4d, %2/2
+.loopH:
+    PROCESS_CHROMA_SP_W8_2R
+
+    paddd     m0, m7
+    paddd     m1, m7
+    paddd     m2, m7
+    paddd     m3, m7
+
+    psrad     m0, 12
+    psrad     m1, 12
+    psrad     m2, 12
+    psrad     m3, 12
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+
+    packuswb  m0, m2
+
+    movlps    [r2], m0
+    movhps    [r2 + r3], m0
+
+    lea       r2, [r2 + 2 * r3]
+
+    dec r4d
+    jnz .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SP_W8_H2 8, 2
+    FILTER_VER_CHROMA_SP_W8_H2 8, 4
+    FILTER_VER_CHROMA_SP_W8_H2 8, 6
+    FILTER_VER_CHROMA_SP_W8_H2 8, 8
+    FILTER_VER_CHROMA_SP_W8_H2 8, 16
+    FILTER_VER_CHROMA_SP_W8_H2 8, 32
+
+    FILTER_VER_CHROMA_SP_W8_H2 8, 12
+    FILTER_VER_CHROMA_SP_W8_H2 8, 64
+
+
+;---------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_%1x%2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W16n 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_%1x%2, 4, 7, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m0, [r5 + r4 * 4]
+%else
+    movd       m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m1, m0, [tab_Vm]
+    pshufb     m0, [tab_Vm + 16]
+    mov        r4d, %2/2
+
+.loop:
+
+    mov         r6d,       %1/16
+
+.loopW:
+
+    movu       m2, [r0]
+    movu       m3, [r0 + r1]
+
+    punpcklbw  m4, m2, m3
+    punpckhbw  m2, m3
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m2, m1
+
+    lea        r5, [r0 + 2 * r1]
+    movu       m5, [r5]
+    movu       m7, [r5 + r1]
+
+    punpcklbw  m6, m5, m7
+    pmaddubsw  m6, m0
+    paddw      m4, m6
+
+    punpckhbw  m6, m5, m7
+    pmaddubsw  m6, m0
+    paddw      m2, m6
+
+    mova       m6, [pw_2000]
+
+    psubw      m4, m6
+    psubw      m2, m6
+
+    movu       [r2], m4
+    movu       [r2 + 16], m2
+
+    punpcklbw  m4, m3, m5
+    punpckhbw  m3, m5
+
+    pmaddubsw  m4, m1
+    pmaddubsw  m3, m1
+
+    movu       m5, [r5 + 2 * r1]
+
+    punpcklbw  m2, m7, m5
+    punpckhbw  m7, m5
+
+    pmaddubsw  m2, m0
+    pmaddubsw  m7, m0
+
+    paddw      m4, m2
+    paddw      m3, m7
+
+    psubw      m4, m6
+    psubw      m3, m6
+
+    movu       [r2 + r3], m4
+    movu       [r2 + r3 + 16], m3
+
+    add         r0,        16
+    add         r2,        32
+    dec         r6d
+    jnz         .loopW
+
+    lea         r0,        [r0 + r1 * 2 - %1]
+    lea         r2,        [r2 + r3 * 2 - %1 * 2]
+
+    dec        r4d
+    jnz        .loop
+    RET
+%endmacro
+
+    FILTER_V_PS_W16n 64, 64
+    FILTER_V_PS_W16n 64, 32
+    FILTER_V_PS_W16n 64, 48
+    FILTER_V_PS_W16n 48, 64
+    FILTER_V_PS_W16n 64, 16
+
+
+;------------------------------------------------------------------------------------------------------------
+;void interp_4tap_vert_ps_2x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;------------------------------------------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_2x4, 4, 6, 7
+
+    mov         r4d, r4m
+    sub         r0, r1
+    add         r3d, r3d
+
+%ifdef PIC
+    lea         r5, [tab_ChromaCoeff]
+    movd        m0, [r5 + r4 * 4]
+%else
+    movd        m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb      m0, [tab_Cm]
+
+    lea         r5, [3 * r1]
+
+    movd        m2, [r0]
+    movd        m3, [r0 + r1]
+    movd        m4, [r0 + 2 * r1]
+    movd        m5, [r0 + r5]
+
+    punpcklbw   m2, m3
+    punpcklbw   m6, m4, m5
+    punpcklbw   m2, m6
+
+    pmaddubsw   m2, m0
+
+    lea         r0, [r0 + 4 * r1]
+    movd        m6, [r0]
+
+    punpcklbw   m3, m4
+    punpcklbw   m1, m5, m6
+    punpcklbw   m3, m1
+
+    pmaddubsw   m3, m0
+    phaddw      m2, m3
+
+    mova        m1, [pw_2000]
+
+    psubw       m2, m1
+
+    movd        [r2], m2
+    pextrd      [r2 + r3], m2, 2
+
+    movd        m2, [r0 + r1]
+
+    punpcklbw   m4, m5
+    punpcklbw   m3, m6, m2
+    punpcklbw   m4, m3
+
+    pmaddubsw   m4, m0
+
+    movd        m3, [r0 + 2 * r1]
+
+    punpcklbw   m5, m6
+    punpcklbw   m2, m3
+    punpcklbw   m5, m2
+
+    pmaddubsw   m5, m0
+    phaddw      m4, m5
+    psubw       m4, m1
+
+    lea         r2, [r2 + 2 * r3]
+    movd        [r2], m4
+    pextrd      [r2 + r3], m4, 2
+
+    RET
+
+;-------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ps_2x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-------------------------------------------------------------------------------------------------------------
+%macro FILTER_V_PS_W2 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ps_2x%2, 4, 6, 8
+
+    mov        r4d, r4m
+    sub        r0, r1
+    add        r3d, r3d
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeff]
+    movd       m0, [r5 + r4 * 4]
+%else
+    movd       m0, [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    pshufb     m0, [tab_Cm]
+
+    mova       m1, [pw_2000]
+    lea        r5, [3 * r1]
+    mov        r4d, %2/4
+.loop:
+    movd       m2, [r0]
+    movd       m3, [r0 + r1]
+    movd       m4, [r0 + 2 * r1]
+    movd       m5, [r0 + r5]
+
+    punpcklbw  m2, m3
+    punpcklbw  m6, m4, m5
+    punpcklbw  m2, m6
+
+    pmaddubsw  m2, m0
+
+    lea        r0, [r0 + 4 * r1]
+    movd       m6, [r0]
+
+    punpcklbw  m3, m4
+    punpcklbw  m7, m5, m6
+    punpcklbw  m3, m7
+
+    pmaddubsw  m3, m0
+
+    phaddw     m2, m3
+    psubw      m2, m1
+
+
+    movd       [r2], m2
+    pshufd     m2, m2, 2
+    movd       [r2 + r3], m2
+
+    movd       m2, [r0 + r1]
+
+    punpcklbw  m4, m5
+    punpcklbw  m3, m6, m2
+    punpcklbw  m4, m3
+
+    pmaddubsw  m4, m0
+
+    movd       m3, [r0 + 2 * r1]
+
+    punpcklbw  m5, m6
+    punpcklbw  m2, m3
+    punpcklbw  m5, m2
+
+    pmaddubsw  m5, m0
+
+    phaddw     m4, m5
+
+    psubw      m4, m1
+
+    lea        r2, [r2 + 2 * r3]
+    movd       [r2], m4
+    pshufd     m4 , m4 ,2
+    movd       [r2 + r3], m4
+
+    lea        r2, [r2 + 2 * r3]
+
+    dec        r4d
+    jnz        .loop
+
+    RET
+%endmacro
+
+    FILTER_V_PS_W2 2, 8
+
+    FILTER_V_PS_W2 2, 16
+
+;-----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SS 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_ss_%1x%2, 5, 7, 6 ,0-gprsize
+
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 5
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r6, [r5 + r4]
+%else
+    lea       r6, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       dword [rsp], %2/4
+
+.loopH:
+    mov       r4d, (%1/4)
+.loopW:
+    PROCESS_CHROMA_SP_W4_4R
+
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+
+    movlps    [r2], m0
+    movhps    [r2 + r3], m0
+    lea       r5, [r2 + 2 * r3]
+    movlps    [r5], m2
+    movhps    [r5 + r3], m2
+
+    lea       r5, [4 * r1 - 2 * 4]
+    sub       r0, r5
+    add       r2, 2 * 4
+
+    dec       r4d
+    jnz       .loopW
+
+    lea       r0, [r0 + 4 * r1 - 2 * %1]
+    lea       r2, [r2 + 4 * r3 - 2 * %1]
+
+    dec       dword [rsp]
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SS 4, 4
+    FILTER_VER_CHROMA_SS 4, 8
+    FILTER_VER_CHROMA_SS 16, 16
+    FILTER_VER_CHROMA_SS 16, 8
+    FILTER_VER_CHROMA_SS 16, 12
+    FILTER_VER_CHROMA_SS 12, 16
+    FILTER_VER_CHROMA_SS 16, 4
+    FILTER_VER_CHROMA_SS 4, 16
+    FILTER_VER_CHROMA_SS 32, 32
+    FILTER_VER_CHROMA_SS 32, 16
+    FILTER_VER_CHROMA_SS 16, 32
+    FILTER_VER_CHROMA_SS 32, 24
+    FILTER_VER_CHROMA_SS 24, 32
+    FILTER_VER_CHROMA_SS 32, 8
+
+    FILTER_VER_CHROMA_SS 16, 24
+    FILTER_VER_CHROMA_SS 12, 32
+    FILTER_VER_CHROMA_SS 4, 32
+    FILTER_VER_CHROMA_SS 32, 64
+    FILTER_VER_CHROMA_SS 16, 64
+    FILTER_VER_CHROMA_SS 32, 48
+    FILTER_VER_CHROMA_SS 24, 64
+
+    FILTER_VER_CHROMA_SS 64, 64
+    FILTER_VER_CHROMA_SS 64, 32
+    FILTER_VER_CHROMA_SS 64, 48
+    FILTER_VER_CHROMA_SS 48, 64
+    FILTER_VER_CHROMA_SS 64, 16
+
+%macro FILTER_VER_CHROMA_S_AVX2_4x4 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x4, 4, 6, 7
+    mov             r4d, r4m
+    add             r1d, r1d
+    shl             r4d, 6
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+%ifidn %1,sp
+    mova            m6, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+
+    movq            xm0, [r0]
+    movq            xm1, [r0 + r1]
+    punpcklwd       xm0, xm1
+    movq            xm2, [r0 + r1 * 2]
+    punpcklwd       xm1, xm2
+    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
+    pmaddwd         m0, [r5]
+    movq            xm3, [r0 + r4]
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm4, [r0]
+    punpcklwd       xm3, xm4
+    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m5
+    movq            xm3, [r0 + r1]
+    punpcklwd       xm4, xm3
+    movq            xm1, [r0 + r1 * 2]
+    punpcklwd       xm3, xm1
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
+    pmaddwd         m4, [r5 + 1 * mmsize]
+    paddd           m2, m4
+
+%ifidn %1,sp
+    paddd           m0, m6
+    paddd           m2, m6
+    psrad           m0, 12
+    psrad           m2, 12
+%else
+    psrad           m0, 6
+    psrad           m2, 6
+%endif
+    packssdw        m0, m2
+    vextracti128    xm2, m0, 1
+    lea             r4, [r3 * 3]
+
+%ifidn %1,sp
+    packuswb        xm0, xm2
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 2
+    pextrd          [r2 + r3 * 2], xm0, 1
+    pextrd          [r2 + r4], xm0, 3
+%else
+    movq            [r2], xm0
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r4], xm2
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_4x4 sp
+    FILTER_VER_CHROMA_S_AVX2_4x4 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_4x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x8, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+
+    movq            xm0, [r0]
+    movq            xm1, [r0 + r1]
+    punpcklwd       xm0, xm1
+    movq            xm2, [r0 + r1 * 2]
+    punpcklwd       xm1, xm2
+    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
+    pmaddwd         m0, [r5]
+    movq            xm3, [r0 + r4]
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm4, [r0]
+    punpcklwd       xm3, xm4
+    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m5
+    movq            xm3, [r0 + r1]
+    punpcklwd       xm4, xm3
+    movq            xm1, [r0 + r1 * 2]
+    punpcklwd       xm3, xm1
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
+    pmaddwd         m5, m4, [r5 + 1 * mmsize]
+    paddd           m2, m5
+    pmaddwd         m4, [r5]
+    movq            xm3, [r0 + r4]
+    punpcklwd       xm1, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm6, [r0]
+    punpcklwd       xm3, xm6
+    vinserti128     m1, m1, xm3, 1                  ; m1 = [8 7 7 6]
+    pmaddwd         m5, m1, [r5 + 1 * mmsize]
+    paddd           m4, m5
+    pmaddwd         m1, [r5]
+    movq            xm3, [r0 + r1]
+    punpcklwd       xm6, xm3
+    movq            xm5, [r0 + 2 * r1]
+    punpcklwd       xm3, xm5
+    vinserti128     m6, m6, xm3, 1                  ; m6 = [A 9 9 8]
+    pmaddwd         m6, [r5 + 1 * mmsize]
+    paddd           m1, m6
+    lea             r4, [r3 * 3]
+
+%ifidn %1,sp
+    paddd           m0, m7
+    paddd           m2, m7
+    paddd           m4, m7
+    paddd           m1, m7
+    psrad           m0, 12
+    psrad           m2, 12
+    psrad           m4, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m2, 6
+    psrad           m4, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m2
+    packssdw        m4, m1
+%ifidn %1,sp
+    packuswb        m0, m4
+    vextracti128    xm2, m0, 1
+    movd            [r2], xm0
+    movd            [r2 + r3], xm2
+    pextrd          [r2 + r3 * 2], xm0, 1
+    pextrd          [r2 + r4], xm2, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm0, 2
+    pextrd          [r2 + r3], xm2, 2
+    pextrd          [r2 + r3 * 2], xm0, 3
+    pextrd          [r2 + r4], xm2, 3
+%else
+    vextracti128    xm2, m0, 1
+    vextracti128    xm1, m4, 1
+    movq            [r2], xm0
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r4], xm2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm4
+    movq            [r2 + r3], xm1
+    movhps          [r2 + r3 * 2], xm4
+    movhps          [r2 + r4], xm1
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_4x8 sp
+    FILTER_VER_CHROMA_S_AVX2_4x8 ss
+
+%macro PROCESS_CHROMA_AVX2_W4_16R 1
+    movq            xm0, [r0]
+    movq            xm1, [r0 + r1]
+    punpcklwd       xm0, xm1
+    movq            xm2, [r0 + r1 * 2]
+    punpcklwd       xm1, xm2
+    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
+    pmaddwd         m0, [r5]
+    movq            xm3, [r0 + r4]
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm4, [r0]
+    punpcklwd       xm3, xm4
+    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m5
+    movq            xm3, [r0 + r1]
+    punpcklwd       xm4, xm3
+    movq            xm1, [r0 + r1 * 2]
+    punpcklwd       xm3, xm1
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [6 5 5 4]
+    pmaddwd         m5, m4, [r5 + 1 * mmsize]
+    paddd           m2, m5
+    pmaddwd         m4, [r5]
+    movq            xm3, [r0 + r4]
+    punpcklwd       xm1, xm3
+    lea             r0, [r0 + 4 * r1]
+    movq            xm6, [r0]
+    punpcklwd       xm3, xm6
+    vinserti128     m1, m1, xm3, 1                  ; m1 = [8 7 7 6]
+    pmaddwd         m5, m1, [r5 + 1 * mmsize]
+    paddd           m4, m5
+    pmaddwd         m1, [r5]
+    movq            xm3, [r0 + r1]
+    punpcklwd       xm6, xm3
+    movq            xm5, [r0 + 2 * r1]
+    punpcklwd       xm3, xm5
+    vinserti128     m6, m6, xm3, 1                  ; m6 = [10 9 9 8]
+    pmaddwd         m3, m6, [r5 + 1 * mmsize]
+    paddd           m1, m3
+    pmaddwd         m6, [r5]
+
+%ifidn %1,sp
+    paddd           m0, m7
+    paddd           m2, m7
+    paddd           m4, m7
+    paddd           m1, m7
+    psrad           m4, 12
+    psrad           m1, 12
+    psrad           m0, 12
+    psrad           m2, 12
+%else
+    psrad           m0, 6
+    psrad           m2, 6
+    psrad           m4, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m2
+    packssdw        m4, m1
+%ifidn %1,sp
+    packuswb        m0, m4
+    vextracti128    xm4, m0, 1
+    movd            [r2], xm0
+    movd            [r2 + r3], xm4
+    pextrd          [r2 + r3 * 2], xm0, 1
+    pextrd          [r2 + r6], xm4, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm0, 2
+    pextrd          [r2 + r3], xm4, 2
+    pextrd          [r2 + r3 * 2], xm0, 3
+    pextrd          [r2 + r6], xm4, 3
+%else
+    vextracti128    xm2, m0, 1
+    vextracti128    xm1, m4, 1
+    movq            [r2], xm0
+    movq            [r2 + r3], xm2
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r6], xm2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm4
+    movq            [r2 + r3], xm1
+    movhps          [r2 + r3 * 2], xm4
+    movhps          [r2 + r6], xm1
+%endif
+
+    movq            xm2, [r0 + r4]
+    punpcklwd       xm5, xm2
+    lea             r0, [r0 + 4 * r1]
+    movq            xm0, [r0]
+    punpcklwd       xm2, xm0
+    vinserti128     m5, m5, xm2, 1                  ; m5 = [12 11 11 10]
+    pmaddwd         m2, m5, [r5 + 1 * mmsize]
+    paddd           m6, m2
+    pmaddwd         m5, [r5]
+    movq            xm2, [r0 + r1]
+    punpcklwd       xm0, xm2
+    movq            xm3, [r0 + 2 * r1]
+    punpcklwd       xm2, xm3
+    vinserti128     m0, m0, xm2, 1                  ; m0 = [14 13 13 12]
+    pmaddwd         m2, m0, [r5 + 1 * mmsize]
+    paddd           m5, m2
+    pmaddwd         m0, [r5]
+    movq            xm4, [r0 + r4]
+    punpcklwd       xm3, xm4
+    lea             r0, [r0 + 4 * r1]
+    movq            xm1, [r0]
+    punpcklwd       xm4, xm1
+    vinserti128     m3, m3, xm4, 1                  ; m3 = [16 15 15 14]
+    pmaddwd         m4, m3, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m3, [r5]
+    movq            xm4, [r0 + r1]
+    punpcklwd       xm1, xm4
+    movq            xm2, [r0 + 2 * r1]
+    punpcklwd       xm4, xm2
+    vinserti128     m1, m1, xm4, 1                  ; m1 = [18 17 17 16]
+    pmaddwd         m1, [r5 + 1 * mmsize]
+    paddd           m3, m1
+
+%ifidn %1,sp
+    paddd           m6, m7
+    paddd           m5, m7
+    paddd           m0, m7
+    paddd           m3, m7
+    psrad           m6, 12
+    psrad           m5, 12
+    psrad           m0, 12
+    psrad           m3, 12
+%else
+    psrad           m6, 6
+    psrad           m5, 6
+    psrad           m0, 6
+    psrad           m3, 6
+%endif
+    packssdw        m6, m5
+    packssdw        m0, m3
+    lea             r2, [r2 + r3 * 4]
+
+%ifidn %1,sp
+    packuswb        m6, m0
+    vextracti128    xm0, m6, 1
+    movd            [r2], xm6
+    movd            [r2 + r3], xm0
+    pextrd          [r2 + r3 * 2], xm6, 1
+    pextrd          [r2 + r6], xm0, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm6, 2
+    pextrd          [r2 + r3], xm0, 2
+    pextrd          [r2 + r3 * 2], xm6, 3
+    pextrd          [r2 + r6], xm0, 3
+%else
+    vextracti128    xm5, m6, 1
+    vextracti128    xm3, m0, 1
+    movq            [r2], xm6
+    movq            [r2 + r3], xm5
+    movhps          [r2 + r3 * 2], xm6
+    movhps          [r2 + r6], xm5
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm0
+    movq            [r2 + r3], xm3
+    movhps          [r2 + r3 * 2], xm0
+    movhps          [r2 + r6], xm3
+%endif
+%endmacro
+
+%macro FILTER_VER_CHROMA_S_AVX2_4x16 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x16, 4, 7, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+    PROCESS_CHROMA_AVX2_W4_16R %1
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_4x16 sp
+    FILTER_VER_CHROMA_S_AVX2_4x16 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_4x32 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x32, 4, 7, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+%rep 2
+    PROCESS_CHROMA_AVX2_W4_16R %1
+    lea             r2, [r2 + r3 * 4]
+%endrep
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_4x32 sp
+    FILTER_VER_CHROMA_S_AVX2_4x32 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_4x2 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_4x2, 4, 6, 6
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+%ifidn %1,sp
+    mova            m5, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    movq            xm0, [r0]
+    movq            xm1, [r0 + r1]
+    punpcklwd       xm0, xm1
+    movq            xm2, [r0 + r1 * 2]
+    punpcklwd       xm1, xm2
+    vinserti128     m0, m0, xm1, 1                  ; m0 = [2 1 1 0]
+    pmaddwd         m0, [r5]
+    movq            xm3, [r0 + r4]
+    punpcklwd       xm2, xm3
+    movq            xm4, [r0 + 4 * r1]
+    punpcklwd       xm3, xm4
+    vinserti128     m2, m2, xm3, 1                  ; m2 = [4 3 3 2]
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+%ifidn %1,sp
+    paddd           m0, m5
+    psrad           m0, 12
+%else
+    psrad           m0, 6
+%endif
+    vextracti128    xm1, m0, 1
+    packssdw        xm0, xm1
+%ifidn %1,sp
+    packuswb        xm0, xm0
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+%else
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_4x2 sp
+    FILTER_VER_CHROMA_S_AVX2_4x2 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_2x4 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_2x4, 4, 6, 6
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+%ifidn %1,sp
+    mova            m5, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    movd            xm0, [r0]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm0, xm1
+    movd            xm2, [r0 + r1 * 2]
+    punpcklwd       xm1, xm2
+    punpcklqdq      xm0, xm1                        ; m0 = [2 1 1 0]
+    movd            xm3, [r0 + r4]
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movd            xm4, [r0]
+    punpcklwd       xm3, xm4
+    punpcklqdq      xm2, xm3                        ; m2 = [4 3 3 2]
+    vinserti128     m0, m0, xm2, 1                  ; m0 = [4 3 3 2 2 1 1 0]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm4, xm1
+    movd            xm3, [r0 + r1 * 2]
+    punpcklwd       xm1, xm3
+    punpcklqdq      xm4, xm1                        ; m4 = [6 5 5 4]
+    vinserti128     m2, m2, xm4, 1                  ; m2 = [6 5 5 4 4 3 3 2]
+    pmaddwd         m0, [r5]
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+%ifidn %1,sp
+    paddd           m0, m5
+    psrad           m0, 12
+%else
+    psrad           m0, 6
+%endif
+    vextracti128    xm1, m0, 1
+    packssdw        xm0, xm1
+    lea             r4, [r3 * 3]
+%ifidn %1,sp
+    packuswb        xm0, xm0
+    pextrw          [r2], xm0, 0
+    pextrw          [r2 + r3], xm0, 1
+    pextrw          [r2 + 2 * r3], xm0, 2
+    pextrw          [r2 + r4], xm0, 3
+%else
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    pextrd          [r2 + 2 * r3], xm0, 2
+    pextrd          [r2 + r4], xm0, 3
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_2x4 sp
+    FILTER_VER_CHROMA_S_AVX2_2x4 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_8x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x8, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m4
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    pmaddwd         m3, [r5]
+    paddd           m1, m5
+%ifidn %1,sp
+    paddd           m0, m7
+    paddd           m1, m7
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm1, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m1
+%ifidn %1,sp
+    paddd           m2, m7
+    paddd           m3, m7
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m2, m3
+
+    movu            xm1, [r0 + r4]                  ; m1 = row 7
+    punpckhwd       xm3, xm6, xm1
+    punpcklwd       xm6, xm1
+    vinserti128     m6, m6, xm3, 1
+    pmaddwd         m3, m6, [r5 + 1 * mmsize]
+    pmaddwd         m6, [r5]
+    paddd           m4, m3
+
+    lea             r4, [r3 * 3]
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m3, [v4_interp8_hps_shuf]
+    vpermd          m0, m3, m0
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r4], xm2
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    movu            [r2], xm0
+    vextracti128    xm0, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3], xm0
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r4], xm3
+%endif
+    lea             r2, [r2 + r3 * 4]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm0, [r0]                       ; m0 = row 8
+    punpckhwd       xm2, xm1, xm0
+    punpcklwd       xm1, xm0
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    pmaddwd         m1, [r5]
+    paddd           m5, m2
+%ifidn %1,sp
+    paddd           m4, m7
+    paddd           m5, m7
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m4, m5
+
+    movu            xm2, [r0 + r1]                  ; m2 = row 9
+    punpckhwd       xm5, xm0, xm2
+    punpcklwd       xm0, xm2
+    vinserti128     m0, m0, xm5, 1
+    pmaddwd         m0, [r5 + 1 * mmsize]
+    paddd           m6, m0
+    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
+    punpckhwd       xm0, xm2, xm5
+    punpcklwd       xm2, xm5
+    vinserti128     m2, m2, xm0, 1
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m1, m2
+
+%ifidn %1,sp
+    paddd           m6, m7
+    paddd           m1, m7
+    psrad           m6, 12
+    psrad           m1, 12
+%else
+    psrad           m6, 6
+    psrad           m1, 6
+%endif
+    packssdw        m6, m1
+%ifidn %1,sp
+    packuswb        m4, m6
+    vpermd          m4, m3, m4
+    vextracti128    xm6, m4, 1
+    movq            [r2], xm4
+    movhps          [r2 + r3], xm4
+    movq            [r2 + r3 * 2], xm6
+    movhps          [r2 + r4], xm6
+%else
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm5, m4, 1
+    vextracti128    xm1, m6, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm5
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r4], xm1
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_8x8 sp
+    FILTER_VER_CHROMA_S_AVX2_8x8 ss
+
+%macro PROCESS_CHROMA_S_AVX2_W8_16R 1
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r7, [r0 + r1 * 4]
+    movu            xm4, [r7]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+    movu            xm5, [r7 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm7, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddwd         m7, m5, [r5 + 1 * mmsize]
+    paddd           m3, m7
+    pmaddwd         m5, [r5]
+%ifidn %1,sp
+    paddd           m0, m9
+    paddd           m1, m9
+    paddd           m2, m9
+    paddd           m3, m9
+    psrad           m0, 12
+    psrad           m1, 12
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m0, m1
+    packssdw        m2, m3
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m3, [v4_interp8_hps_shuf]
+    vpermd          m0, m3, m0
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+%endif
+
+    movu            xm7, [r7 + r4]                  ; m7 = row 7
+    punpckhwd       xm8, xm6, xm7
+    punpcklwd       xm6, xm7
+    vinserti128     m6, m6, xm8, 1
+    pmaddwd         m8, m6, [r5 + 1 * mmsize]
+    paddd           m4, m8
+    pmaddwd         m6, [r5]
+    lea             r7, [r7 + r1 * 4]
+    movu            xm8, [r7]                       ; m8 = row 8
+    punpckhwd       xm0, xm7, xm8
+    punpcklwd       xm7, xm8
+    vinserti128     m7, m7, xm0, 1
+    pmaddwd         m0, m7, [r5 + 1 * mmsize]
+    paddd           m5, m0
+    pmaddwd         m7, [r5]
+    movu            xm0, [r7 + r1]                  ; m0 = row 9
+    punpckhwd       xm1, xm8, xm0
+    punpcklwd       xm8, xm0
+    vinserti128     m8, m8, xm1, 1
+    pmaddwd         m1, m8, [r5 + 1 * mmsize]
+    paddd           m6, m1
+    pmaddwd         m8, [r5]
+    movu            xm1, [r7 + r1 * 2]              ; m1 = row 10
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m2, m0, [r5 + 1 * mmsize]
+    paddd           m7, m2
+    pmaddwd         m0, [r5]
+%ifidn %1,sp
+    paddd           m4, m9
+    paddd           m5, m9
+    psrad           m4, 12
+    psrad           m5, 12
+    paddd           m6, m9
+    paddd           m7, m9
+    psrad           m6, 12
+    psrad           m7, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+    psrad           m6, 6
+    psrad           m7, 6
+%endif
+    packssdw        m4, m5
+    packssdw        m6, m7
+    lea             r8, [r2 + r3 * 4]
+%ifidn %1,sp
+    packuswb        m4, m6
+    vpermd          m4, m3, m4
+    vextracti128    xm6, m4, 1
+    movq            [r8], xm4
+    movhps          [r8 + r3], xm4
+    movq            [r8 + r3 * 2], xm6
+    movhps          [r8 + r6], xm6
+%else
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm5, m4, 1
+    vextracti128    xm7, m6, 1
+    movu            [r8], xm4
+    movu            [r8 + r3], xm5
+    movu            [r8 + r3 * 2], xm6
+    movu            [r8 + r6], xm7
+%endif
+
+    movu            xm2, [r7 + r4]                  ; m2 = row 11
+    punpckhwd       xm4, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm4, 1
+    pmaddwd         m4, m1, [r5 + 1 * mmsize]
+    paddd           m8, m4
+    pmaddwd         m1, [r5]
+    lea             r7, [r7 + r1 * 4]
+    movu            xm4, [r7]                       ; m4 = row 12
+    punpckhwd       xm5, xm2, xm4
+    punpcklwd       xm2, xm4
+    vinserti128     m2, m2, xm5, 1
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    paddd           m0, m5
+    pmaddwd         m2, [r5]
+    movu            xm5, [r7 + r1]                  ; m5 = row 13
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m1, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r7 + r1 * 2]              ; m6 = row 14
+    punpckhwd       xm7, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm7, 1
+    pmaddwd         m7, m5, [r5 + 1 * mmsize]
+    paddd           m2, m7
+    pmaddwd         m5, [r5]
+%ifidn %1,sp
+    paddd           m8, m9
+    paddd           m0, m9
+    paddd           m1, m9
+    paddd           m2, m9
+    psrad           m8, 12
+    psrad           m0, 12
+    psrad           m1, 12
+    psrad           m2, 12
+%else
+    psrad           m8, 6
+    psrad           m0, 6
+    psrad           m1, 6
+    psrad           m2, 6
+%endif
+    packssdw        m8, m0
+    packssdw        m1, m2
+    lea             r8, [r8 + r3 * 4]
+%ifidn %1,sp
+    packuswb        m8, m1
+    vpermd          m8, m3, m8
+    vextracti128    xm1, m8, 1
+    movq            [r8], xm8
+    movhps          [r8 + r3], xm8
+    movq            [r8 + r3 * 2], xm1
+    movhps          [r8 + r6], xm1
+%else
+    vpermq          m8, m8, 11011000b
+    vpermq          m1, m1, 11011000b
+    vextracti128    xm0, m8, 1
+    vextracti128    xm2, m1, 1
+    movu            [r8], xm8
+    movu            [r8 + r3], xm0
+    movu            [r8 + r3 * 2], xm1
+    movu            [r8 + r6], xm2
+%endif
+    lea             r8, [r8 + r3 * 4]
+
+    movu            xm7, [r7 + r4]                  ; m7 = row 15
+    punpckhwd       xm2, xm6, xm7
+    punpcklwd       xm6, xm7
+    vinserti128     m6, m6, xm2, 1
+    pmaddwd         m2, m6, [r5 + 1 * mmsize]
+    paddd           m4, m2
+    pmaddwd         m6, [r5]
+    lea             r7, [r7 + r1 * 4]
+    movu            xm2, [r7]                       ; m2 = row 16
+    punpckhwd       xm1, xm7, xm2
+    punpcklwd       xm7, xm2
+    vinserti128     m7, m7, xm1, 1
+    pmaddwd         m1, m7, [r5 + 1 * mmsize]
+    paddd           m5, m1
+    pmaddwd         m7, [r5]
+    movu            xm1, [r7 + r1]                  ; m1 = row 17
+    punpckhwd       xm0, xm2, xm1
+    punpcklwd       xm2, xm1
+    vinserti128     m2, m2, xm0, 1
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m6, m2
+    movu            xm0, [r7 + r1 * 2]              ; m0 = row 18
+    punpckhwd       xm2, xm1, xm0
+    punpcklwd       xm1, xm0
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m1, [r5 + 1 * mmsize]
+    paddd           m7, m1
+
+%ifidn %1,sp
+    paddd           m4, m9
+    paddd           m5, m9
+    paddd           m6, m9
+    paddd           m7, m9
+    psrad           m4, 12
+    psrad           m5, 12
+    psrad           m6, 12
+    psrad           m7, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+    psrad           m6, 6
+    psrad           m7, 6
+%endif
+    packssdw        m4, m5
+    packssdw        m6, m7
+%ifidn %1,sp
+    packuswb        m4, m6
+    vpermd          m4, m3, m4
+    vextracti128    xm6, m4, 1
+    movq            [r8], xm4
+    movhps          [r8 + r3], xm4
+    movq            [r8 + r3 * 2], xm6
+    movhps          [r8 + r6], xm6
+%else
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm5, m4, 1
+    vextracti128    xm7, m6, 1
+    movu            [r8], xm4
+    movu            [r8 + r3], xm5
+    movu            [r8 + r3 * 2], xm6
+    movu            [r8 + r6], xm7
+%endif
+%endmacro
+
+%macro FILTER_VER_CHROMA_S_AVX2_Nx16 2
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_%2x16, 4, 10, 10
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m9, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+    mov             r9d, %2 / 8
+.loopW:
+    PROCESS_CHROMA_S_AVX2_W8_16R %1
+%ifidn %1,sp
+    add             r2, 8
+%else
+    add             r2, 16
+%endif
+    add             r0, 16
+    dec             r9d
+    jnz             .loopW
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_Nx16 sp, 16
+    FILTER_VER_CHROMA_S_AVX2_Nx16 sp, 32
+    FILTER_VER_CHROMA_S_AVX2_Nx16 sp, 64
+    FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 16
+    FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 32
+    FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 64
+
+%macro FILTER_VER_CHROMA_S_AVX2_NxN 3
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%3_%1x%2, 4, 11, 10
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %3,sp
+    mova            m9, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+    mov             r9d, %2 / 16
+.loopH:
+    mov             r10d, %1 / 8
+.loopW:
+    PROCESS_CHROMA_S_AVX2_W8_16R %3
+%ifidn %3,sp
+    add             r2, 8
+%else
+    add             r2, 16
+%endif
+    add             r0, 16
+    dec             r10d
+    jnz             .loopW
+    lea             r0, [r7 - 2 * %1 + 16]
+%ifidn %3,sp
+    lea             r2, [r8 + r3 * 4 - %1 + 8]
+%else
+    lea             r2, [r8 + r3 * 4 - 2 * %1 + 16]
+%endif
+    dec             r9d
+    jnz             .loopH
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_NxN 16, 32, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 24, 32, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 32, 32, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 16, 32, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 24, 32, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 32, 32, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 16, 64, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 24, 64, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 32, 64, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 32, 48, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 32, 48, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 16, 64, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 24, 64, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 32, 64, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 64, 64, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 64, 32, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 64, 48, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 48, 64, sp
+    FILTER_VER_CHROMA_S_AVX2_NxN 64, 64, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 64, 32, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 64, 48, ss
+    FILTER_VER_CHROMA_S_AVX2_NxN 48, 64, ss
+
+%macro PROCESS_CHROMA_S_AVX2_W8_4R 1
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m4, [r5 + 1 * mmsize]
+    paddd           m2, m4
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm4, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm4, 1
+    pmaddwd         m5, [r5 + 1 * mmsize]
+    paddd           m3, m5
+%ifidn %1,sp
+    paddd           m0, m7
+    paddd           m1, m7
+    paddd           m2, m7
+    paddd           m3, m7
+    psrad           m0, 12
+    psrad           m1, 12
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m0, m1
+    packssdw        m2, m3
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m3, [v4_interp8_hps_shuf]
+    vpermd          m0, m3, m0
+    vextracti128    xm2, m0, 1
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm1, m0, 1
+    vextracti128    xm3, m2, 1
+%endif
+%endmacro
+
+%macro FILTER_VER_CHROMA_S_AVX2_8x4 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x4, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+
+    PROCESS_CHROMA_S_AVX2_W8_4R %1
+    lea             r4, [r3 * 3]
+%ifidn %1,sp
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r4], xm2
+%else
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r4], xm3
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_8x4 sp
+    FILTER_VER_CHROMA_S_AVX2_8x4 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_12x16 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_12x16, 4, 9, 10
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m9, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+    PROCESS_CHROMA_S_AVX2_W8_16R %1
+%ifidn %1,sp
+    add             r2, 8
+%else
+    add             r2, 16
+%endif
+    add             r0, 16
+    mova            m7, m9
+    PROCESS_CHROMA_AVX2_W4_16R %1
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_12x16 sp
+    FILTER_VER_CHROMA_S_AVX2_12x16 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_12x32 1
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_12x32, 4, 9, 10
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1, sp
+    mova            m9, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+%rep 2
+    PROCESS_CHROMA_S_AVX2_W8_16R %1
+%ifidn %1, sp
+    add             r2, 8
+%else
+    add             r2, 16
+%endif
+    add             r0, 16
+    mova            m7, m9
+    PROCESS_CHROMA_AVX2_W4_16R %1
+    sub             r0, 16
+%ifidn %1, sp
+    lea             r2, [r2 + r3 * 4 - 8]
+%else
+    lea             r2, [r2 + r3 * 4 - 16]
+%endif
+%endrep
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_12x32 sp
+    FILTER_VER_CHROMA_S_AVX2_12x32 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_16x12 1
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_16x12, 4, 9, 9
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m8, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+%rep 2
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r7, [r0 + r1 * 4]
+    movu            xm4, [r7]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+%ifidn %1,sp
+    paddd           m0, m8
+    paddd           m1, m8
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+
+    movu            xm5, [r7 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm1, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m1
+%ifidn %1,sp
+    paddd           m2, m8
+    paddd           m3, m8
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m2, m3
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m3, [v4_interp8_hps_shuf]
+    vpermd          m0, m3, m0
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    movu            [r2], xm0
+    vextracti128    xm0, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3], xm0
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+%endif
+    lea             r8, [r2 + r3 * 4]
+
+    movu            xm1, [r7 + r4]                  ; m1 = row 7
+    punpckhwd       xm0, xm6, xm1
+    punpcklwd       xm6, xm1
+    vinserti128     m6, m6, xm0, 1
+    pmaddwd         m0, m6, [r5 + 1 * mmsize]
+    pmaddwd         m6, [r5]
+    paddd           m4, m0
+    lea             r7, [r7 + r1 * 4]
+    movu            xm0, [r7]                       ; m0 = row 8
+    punpckhwd       xm2, xm1, xm0
+    punpcklwd       xm1, xm0
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    pmaddwd         m1, [r5]
+    paddd           m5, m2
+%ifidn %1,sp
+    paddd           m4, m8
+    paddd           m5, m8
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m4, m5
+
+    movu            xm2, [r7 + r1]                  ; m2 = row 9
+    punpckhwd       xm5, xm0, xm2
+    punpcklwd       xm0, xm2
+    vinserti128     m0, m0, xm5, 1
+    pmaddwd         m5, m0, [r5 + 1 * mmsize]
+    paddd           m6, m5
+    pmaddwd         m0, [r5]
+    movu            xm5, [r7 + r1 * 2]              ; m5 = row 10
+    punpckhwd       xm7, xm2, xm5
+    punpcklwd       xm2, xm5
+    vinserti128     m2, m2, xm7, 1
+    pmaddwd         m7, m2, [r5 + 1 * mmsize]
+    paddd           m1, m7
+    pmaddwd         m2, [r5]
+
+%ifidn %1,sp
+    paddd           m6, m8
+    paddd           m1, m8
+    psrad           m6, 12
+    psrad           m1, 12
+%else
+    psrad           m6, 6
+    psrad           m1, 6
+%endif
+    packssdw        m6, m1
+%ifidn %1,sp
+    packuswb        m4, m6
+    vpermd          m4, m3, m4
+    vextracti128    xm6, m4, 1
+    movq            [r8], xm4
+    movhps          [r8 + r3], xm4
+    movq            [r8 + r3 * 2], xm6
+    movhps          [r8 + r6], xm6
+%else
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm7, m4, 1
+    vextracti128    xm1, m6, 1
+    movu            [r8], xm4
+    movu            [r8 + r3], xm7
+    movu            [r8 + r3 * 2], xm6
+    movu            [r8 + r6], xm1
+%endif
+    lea             r8, [r8 + r3 * 4]
+
+    movu            xm7, [r7 + r4]                  ; m7 = row 11
+    punpckhwd       xm1, xm5, xm7
+    punpcklwd       xm5, xm7
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    paddd           m0, m1
+    pmaddwd         m5, [r5]
+    lea             r7, [r7 + r1 * 4]
+    movu            xm1, [r7]                       ; m1 = row 12
+    punpckhwd       xm4, xm7, xm1
+    punpcklwd       xm7, xm1
+    vinserti128     m7, m7, xm4, 1
+    pmaddwd         m4, m7, [r5 + 1 * mmsize]
+    paddd           m2, m4
+    pmaddwd         m7, [r5]
+%ifidn %1,sp
+    paddd           m0, m8
+    paddd           m2, m8
+    psrad           m0, 12
+    psrad           m2, 12
+%else
+    psrad           m0, 6
+    psrad           m2, 6
+%endif
+    packssdw        m0, m2
+
+    movu            xm4, [r7 + r1]                  ; m4 = row 13
+    punpckhwd       xm2, xm1, xm4
+    punpcklwd       xm1, xm4
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m1, [r5 + 1 * mmsize]
+    paddd           m5, m1
+    movu            xm2, [r7 + r1 * 2]              ; m2 = row 14
+    punpckhwd       xm6, xm4, xm2
+    punpcklwd       xm4, xm2
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m4, [r5 + 1 * mmsize]
+    paddd           m7, m4
+%ifidn %1,sp
+    paddd           m5, m8
+    paddd           m7, m8
+    psrad           m5, 12
+    psrad           m7, 12
+%else
+    psrad           m5, 6
+    psrad           m7, 6
+%endif
+    packssdw        m5, m7
+%ifidn %1,sp
+    packuswb        m0, m5
+    vpermd          m0, m3, m0
+    vextracti128    xm5, m0, 1
+    movq            [r8], xm0
+    movhps          [r8 + r3], xm0
+    movq            [r8 + r3 * 2], xm5
+    movhps          [r8 + r6], xm5
+    add             r2, 8
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m5, m5, 11011000b
+    vextracti128    xm7, m0, 1
+    vextracti128    xm6, m5, 1
+    movu            [r8], xm0
+    movu            [r8 + r3], xm7
+    movu            [r8 + r3 * 2], xm5
+    movu            [r8 + r6], xm6
+    add             r2, 16
+%endif
+    add             r0, 16
+%endrep
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_16x12 sp
+    FILTER_VER_CHROMA_S_AVX2_16x12 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_8x12 1
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x12, 4, 7, 9
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m8, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+%ifidn %1,sp
+    paddd           m0, m8
+    paddd           m1, m8
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm1, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m1
+%ifidn %1,sp
+    paddd           m2, m8
+    paddd           m3, m8
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m2, m3
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m3, [v4_interp8_hps_shuf]
+    vpermd          m0, m3, m0
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    movu            [r2], xm0
+    vextracti128    xm0, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3], xm0
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm1, [r0 + r4]                  ; m1 = row 7
+    punpckhwd       xm0, xm6, xm1
+    punpcklwd       xm6, xm1
+    vinserti128     m6, m6, xm0, 1
+    pmaddwd         m0, m6, [r5 + 1 * mmsize]
+    pmaddwd         m6, [r5]
+    paddd           m4, m0
+    lea             r0, [r0 + r1 * 4]
+    movu            xm0, [r0]                       ; m0 = row 8
+    punpckhwd       xm2, xm1, xm0
+    punpcklwd       xm1, xm0
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    pmaddwd         m1, [r5]
+    paddd           m5, m2
+%ifidn %1,sp
+    paddd           m4, m8
+    paddd           m5, m8
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m4, m5
+
+    movu            xm2, [r0 + r1]                  ; m2 = row 9
+    punpckhwd       xm5, xm0, xm2
+    punpcklwd       xm0, xm2
+    vinserti128     m0, m0, xm5, 1
+    pmaddwd         m5, m0, [r5 + 1 * mmsize]
+    paddd           m6, m5
+    pmaddwd         m0, [r5]
+    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
+    punpckhwd       xm7, xm2, xm5
+    punpcklwd       xm2, xm5
+    vinserti128     m2, m2, xm7, 1
+    pmaddwd         m7, m2, [r5 + 1 * mmsize]
+    paddd           m1, m7
+    pmaddwd         m2, [r5]
+
+%ifidn %1,sp
+    paddd           m6, m8
+    paddd           m1, m8
+    psrad           m6, 12
+    psrad           m1, 12
+%else
+    psrad           m6, 6
+    psrad           m1, 6
+%endif
+    packssdw        m6, m1
+%ifidn %1,sp
+    packuswb        m4, m6
+    vpermd          m4, m3, m4
+    vextracti128    xm6, m4, 1
+    movq            [r2], xm4
+    movhps          [r2 + r3], xm4
+    movq            [r2 + r3 * 2], xm6
+    movhps          [r2 + r6], xm6
+%else
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm7, m4, 1
+    vextracti128    xm1, m6, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm7
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r6], xm1
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm7, [r0 + r4]                  ; m7 = row 11
+    punpckhwd       xm1, xm5, xm7
+    punpcklwd       xm5, xm7
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    paddd           m0, m1
+    pmaddwd         m5, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm1, [r0]                       ; m1 = row 12
+    punpckhwd       xm4, xm7, xm1
+    punpcklwd       xm7, xm1
+    vinserti128     m7, m7, xm4, 1
+    pmaddwd         m4, m7, [r5 + 1 * mmsize]
+    paddd           m2, m4
+    pmaddwd         m7, [r5]
+%ifidn %1,sp
+    paddd           m0, m8
+    paddd           m2, m8
+    psrad           m0, 12
+    psrad           m2, 12
+%else
+    psrad           m0, 6
+    psrad           m2, 6
+%endif
+    packssdw        m0, m2
+
+    movu            xm4, [r0 + r1]                  ; m4 = row 13
+    punpckhwd       xm2, xm1, xm4
+    punpcklwd       xm1, xm4
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m1, [r5 + 1 * mmsize]
+    paddd           m5, m1
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 14
+    punpckhwd       xm6, xm4, xm2
+    punpcklwd       xm4, xm2
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m4, [r5 + 1 * mmsize]
+    paddd           m7, m4
+%ifidn %1,sp
+    paddd           m5, m8
+    paddd           m7, m8
+    psrad           m5, 12
+    psrad           m7, 12
+%else
+    psrad           m5, 6
+    psrad           m7, 6
+%endif
+    packssdw        m5, m7
+%ifidn %1,sp
+    packuswb        m0, m5
+    vpermd          m0, m3, m0
+    vextracti128    xm5, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm5
+    movhps          [r2 + r6], xm5
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m5, m5, 11011000b
+    vextracti128    xm7, m0, 1
+    vextracti128    xm6, m5, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm7
+    movu            [r2 + r3 * 2], xm5
+    movu            [r2 + r6], xm6
+%endif
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_8x12 sp
+    FILTER_VER_CHROMA_S_AVX2_8x12 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_16x4 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_16x4, 4, 7, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+%rep 2
+    PROCESS_CHROMA_S_AVX2_W8_4R %1
+    lea             r6, [r3 * 3]
+%ifidn %1,sp
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+    add             r2, 8
+%else
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+    add             r2, 16
+%endif
+    lea             r6, [4 * r1 - 16]
+    sub             r0, r6
+%endrep
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_16x4 sp
+    FILTER_VER_CHROMA_S_AVX2_16x4 ss
+
+%macro PROCESS_CHROMA_S_AVX2_W8_8R 1
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r7, [r0 + r1 * 4]
+    movu            xm4, [r7]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+%ifidn %1,sp
+    paddd           m0, m7
+    paddd           m1, m7
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+
+    movu            xm5, [r7 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r7 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm1, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m1
+%ifidn %1,sp
+    paddd           m2, m7
+    paddd           m3, m7
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m2, m3
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m3, [v4_interp8_hps_shuf]
+    vpermd          m0, m3, m0
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    movu            [r2], xm0
+    vextracti128    xm0, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3], xm0
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+%endif
+    lea             r8, [r2 + r3 * 4]
+
+    movu            xm1, [r7 + r4]                  ; m1 = row 7
+    punpckhwd       xm0, xm6, xm1
+    punpcklwd       xm6, xm1
+    vinserti128     m6, m6, xm0, 1
+    pmaddwd         m0, m6, [r5 + 1 * mmsize]
+    pmaddwd         m6, [r5]
+    paddd           m4, m0
+    lea             r7, [r7 + r1 * 4]
+    movu            xm0, [r7]                       ; m0 = row 8
+    punpckhwd       xm2, xm1, xm0
+    punpcklwd       xm1, xm0
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    pmaddwd         m1, [r5]
+    paddd           m5, m2
+%ifidn %1,sp
+    paddd           m4, m7
+    paddd           m5, m7
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m4, m5
+
+    movu            xm2, [r7 + r1]                  ; m2 = row 9
+    punpckhwd       xm5, xm0, xm2
+    punpcklwd       xm0, xm2
+    vinserti128     m0, m0, xm5, 1
+    pmaddwd         m0, [r5 + 1 * mmsize]
+    paddd           m6, m0
+    movu            xm5, [r7 + r1 * 2]              ; m5 = row 10
+    punpckhwd       xm0, xm2, xm5
+    punpcklwd       xm2, xm5
+    vinserti128     m2, m2, xm0, 1
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m1, m2
+
+%ifidn %1,sp
+    paddd           m6, m7
+    paddd           m1, m7
+    psrad           m6, 12
+    psrad           m1, 12
+%else
+    psrad           m6, 6
+    psrad           m1, 6
+%endif
+    packssdw        m6, m1
+%ifidn %1,sp
+    packuswb        m4, m6
+    vpermd          m4, m3, m4
+    vextracti128    xm6, m4, 1
+    movq            [r8], xm4
+    movhps          [r8 + r3], xm4
+    movq            [r8 + r3 * 2], xm6
+    movhps          [r8 + r6], xm6
+%else
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm7, m4, 1
+    vextracti128    xm1, m6, 1
+    movu            [r8], xm4
+    movu            [r8 + r3], xm7
+    movu            [r8 + r3 * 2], xm6
+    movu            [r8 + r6], xm1
+%endif
+%endmacro
+
+%macro FILTER_VER_CHROMA_S_AVX2_Nx8 2
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_%2x8, 4, 9, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+%rep %2 / 8
+    PROCESS_CHROMA_S_AVX2_W8_8R %1
+%ifidn %1,sp
+    add             r2, 8
+%else
+    add             r2, 16
+%endif
+    add             r0, 16
+%endrep
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_Nx8 sp, 32
+    FILTER_VER_CHROMA_S_AVX2_Nx8 sp, 16
+    FILTER_VER_CHROMA_S_AVX2_Nx8 ss, 32
+    FILTER_VER_CHROMA_S_AVX2_Nx8 ss, 16
+
+%macro FILTER_VER_CHROMA_S_AVX2_8x2 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x2, 4, 6, 6
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m5, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+    movu            xm4, [r0 + r1 * 4]              ; m4 = row 4
+    punpckhwd       xm2, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm2, 1
+    pmaddwd         m3, [r5 + 1 * mmsize]
+    paddd           m1, m3
+%ifidn %1,sp
+    paddd           m0, m5
+    paddd           m1, m5
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+%ifidn %1,sp
+    vextracti128    xm1, m0, 1
+    packuswb        xm0, xm1
+    pshufd          xm0, xm0, 11011000b
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+%else
+    vpermq          m0, m0, 11011000b
+    vextracti128    xm1, m0, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm1
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_8x2 sp
+    FILTER_VER_CHROMA_S_AVX2_8x2 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_8x6 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_8x6, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m4
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    pmaddwd         m3, [r5]
+    paddd           m1, m5
+%ifidn %1,sp
+    paddd           m0, m7
+    paddd           m1, m7
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm1, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m1
+%ifidn %1,sp
+    paddd           m2, m7
+    paddd           m3, m7
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m2, m3
+
+    movu            xm1, [r0 + r4]                  ; m1 = row 7
+    punpckhwd       xm3, xm6, xm1
+    punpcklwd       xm6, xm1
+    vinserti128     m6, m6, xm3, 1
+    pmaddwd         m6, [r5 + 1 * mmsize]
+    paddd           m4, m6
+    movu            xm6, [r0 + r1 * 4]              ; m6 = row 8
+    punpckhwd       xm3, xm1, xm6
+    punpcklwd       xm1, xm6
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5 + 1 * mmsize]
+    paddd           m5, m1
+%ifidn %1,sp
+    paddd           m4, m7
+    paddd           m5, m7
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m4, m5
+    lea             r4, [r3 * 3]
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m3, [v4_interp8_hps_shuf]
+    vpermd          m0, m3, m0
+    vextracti128    xm2, m0, 1
+    vextracti128    xm5, m4, 1
+    packuswb        xm4, xm5
+    pshufd          xm4, xm4, 11011000b
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r4], xm2
+    lea             r2, [r2 + r3 * 4]
+    movq            [r2], xm4
+    movhps          [r2 + r3], xm4
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    vpermq          m4, m4, 11011000b
+    movu            [r2], xm0
+    vextracti128    xm0, m0, 1
+    vextracti128    xm3, m2, 1
+    vextracti128    xm5, m4, 1
+    movu            [r2 + r3], xm0
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r4], xm3
+    lea             r2, [r2 + r3 * 4]
+    movu            [r2], xm4
+    movu            [r2 + r3], xm5
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_8x6 sp
+    FILTER_VER_CHROMA_S_AVX2_8x6 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_8xN 2
+INIT_YMM avx2
+%if ARCH_X86_64 == 1
+cglobal interp_4tap_vert_%1_8x%2, 4, 7, 9
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m8, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+%rep %2 / 16
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+%ifidn %1,sp
+    paddd           m0, m8
+    paddd           m1, m8
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm1, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m1
+%ifidn %1,sp
+    paddd           m2, m8
+    paddd           m3, m8
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m2, m3
+%ifidn %1,sp
+    packuswb        m0, m2
+    mova            m3, [v4_interp8_hps_shuf]
+    vpermd          m0, m3, m0
+    vextracti128    xm2, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m2, m2, 11011000b
+    movu            [r2], xm0
+    vextracti128    xm0, m0, 1
+    vextracti128    xm3, m2, 1
+    movu            [r2 + r3], xm0
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm3
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm1, [r0 + r4]                  ; m1 = row 7
+    punpckhwd       xm0, xm6, xm1
+    punpcklwd       xm6, xm1
+    vinserti128     m6, m6, xm0, 1
+    pmaddwd         m0, m6, [r5 + 1 * mmsize]
+    pmaddwd         m6, [r5]
+    paddd           m4, m0
+    lea             r0, [r0 + r1 * 4]
+    movu            xm0, [r0]                       ; m0 = row 8
+    punpckhwd       xm2, xm1, xm0
+    punpcklwd       xm1, xm0
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    pmaddwd         m1, [r5]
+    paddd           m5, m2
+%ifidn %1,sp
+    paddd           m4, m8
+    paddd           m5, m8
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m4, m5
+
+    movu            xm2, [r0 + r1]                  ; m2 = row 9
+    punpckhwd       xm5, xm0, xm2
+    punpcklwd       xm0, xm2
+    vinserti128     m0, m0, xm5, 1
+    pmaddwd         m5, m0, [r5 + 1 * mmsize]
+    paddd           m6, m5
+    pmaddwd         m0, [r5]
+    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
+    punpckhwd       xm7, xm2, xm5
+    punpcklwd       xm2, xm5
+    vinserti128     m2, m2, xm7, 1
+    pmaddwd         m7, m2, [r5 + 1 * mmsize]
+    paddd           m1, m7
+    pmaddwd         m2, [r5]
+
+%ifidn %1,sp
+    paddd           m6, m8
+    paddd           m1, m8
+    psrad           m6, 12
+    psrad           m1, 12
+%else
+    psrad           m6, 6
+    psrad           m1, 6
+%endif
+    packssdw        m6, m1
+%ifidn %1,sp
+    packuswb        m4, m6
+    vpermd          m4, m3, m4
+    vextracti128    xm6, m4, 1
+    movq            [r2], xm4
+    movhps          [r2 + r3], xm4
+    movq            [r2 + r3 * 2], xm6
+    movhps          [r2 + r6], xm6
+%else
+    vpermq          m4, m4, 11011000b
+    vpermq          m6, m6, 11011000b
+    vextracti128    xm7, m4, 1
+    vextracti128    xm1, m6, 1
+    movu            [r2], xm4
+    movu            [r2 + r3], xm7
+    movu            [r2 + r3 * 2], xm6
+    movu            [r2 + r6], xm1
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm7, [r0 + r4]                  ; m7 = row 11
+    punpckhwd       xm1, xm5, xm7
+    punpcklwd       xm5, xm7
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    paddd           m0, m1
+    pmaddwd         m5, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm1, [r0]                       ; m1 = row 12
+    punpckhwd       xm4, xm7, xm1
+    punpcklwd       xm7, xm1
+    vinserti128     m7, m7, xm4, 1
+    pmaddwd         m4, m7, [r5 + 1 * mmsize]
+    paddd           m2, m4
+    pmaddwd         m7, [r5]
+%ifidn %1,sp
+    paddd           m0, m8
+    paddd           m2, m8
+    psrad           m0, 12
+    psrad           m2, 12
+%else
+    psrad           m0, 6
+    psrad           m2, 6
+%endif
+    packssdw        m0, m2
+
+    movu            xm4, [r0 + r1]                  ; m4 = row 13
+    punpckhwd       xm2, xm1, xm4
+    punpcklwd       xm1, xm4
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    paddd           m5, m2
+    pmaddwd         m1, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 14
+    punpckhwd       xm6, xm4, xm2
+    punpcklwd       xm4, xm2
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m7, m6
+    pmaddwd         m4, [r5]
+%ifidn %1,sp
+    paddd           m5, m8
+    paddd           m7, m8
+    psrad           m5, 12
+    psrad           m7, 12
+%else
+    psrad           m5, 6
+    psrad           m7, 6
+%endif
+    packssdw        m5, m7
+%ifidn %1,sp
+    packuswb        m0, m5
+    vpermd          m0, m3, m0
+    vextracti128    xm5, m0, 1
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm5
+    movhps          [r2 + r6], xm5
+%else
+    vpermq          m0, m0, 11011000b
+    vpermq          m5, m5, 11011000b
+    vextracti128    xm7, m0, 1
+    vextracti128    xm6, m5, 1
+    movu            [r2], xm0
+    movu            [r2 + r3], xm7
+    movu            [r2 + r3 * 2], xm5
+    movu            [r2 + r6], xm6
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm6, [r0 + r4]                  ; m6 = row 15
+    punpckhwd       xm5, xm2, xm6
+    punpcklwd       xm2, xm6
+    vinserti128     m2, m2, xm5, 1
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm0, [r0]                       ; m0 = row 16
+    punpckhwd       xm5, xm6, xm0
+    punpcklwd       xm6, xm0
+    vinserti128     m6, m6, xm5, 1
+    pmaddwd         m5, m6, [r5 + 1 * mmsize]
+    paddd           m4, m5
+    pmaddwd         m6, [r5]
+%ifidn %1,sp
+    paddd           m1, m8
+    paddd           m4, m8
+    psrad           m1, 12
+    psrad           m4, 12
+%else
+    psrad           m1, 6
+    psrad           m4, 6
+%endif
+    packssdw        m1, m4
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 17
+    punpckhwd       xm4, xm0, xm5
+    punpcklwd       xm0, xm5
+    vinserti128     m0, m0, xm4, 1
+    pmaddwd         m0, [r5 + 1 * mmsize]
+    paddd           m2, m0
+    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
+    punpckhwd       xm0, xm5, xm4
+    punpcklwd       xm5, xm4
+    vinserti128     m5, m5, xm0, 1
+    pmaddwd         m5, [r5 + 1 * mmsize]
+    paddd           m6, m5
+%ifidn %1,sp
+    paddd           m2, m8
+    paddd           m6, m8
+    psrad           m2, 12
+    psrad           m6, 12
+%else
+    psrad           m2, 6
+    psrad           m6, 6
+%endif
+    packssdw        m2, m6
+%ifidn %1,sp
+    packuswb        m1, m2
+    vpermd          m1, m3, m1
+    vextracti128    xm2, m1, 1
+    movq            [r2], xm1
+    movhps          [r2 + r3], xm1
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+%else
+    vpermq          m1, m1, 11011000b
+    vpermq          m2, m2, 11011000b
+    vextracti128    xm6, m1, 1
+    vextracti128    xm4, m2, 1
+    movu            [r2], xm1
+    movu            [r2 + r3], xm6
+    movu            [r2 + r3 * 2], xm2
+    movu            [r2 + r6], xm4
+%endif
+    lea             r2, [r2 + r3 * 4]
+%endrep
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_8xN sp, 16
+    FILTER_VER_CHROMA_S_AVX2_8xN sp, 32
+    FILTER_VER_CHROMA_S_AVX2_8xN sp, 64
+    FILTER_VER_CHROMA_S_AVX2_8xN ss, 16
+    FILTER_VER_CHROMA_S_AVX2_8xN ss, 32
+    FILTER_VER_CHROMA_S_AVX2_8xN ss, 64
+
+%macro FILTER_VER_CHROMA_S_AVX2_Nx24 2
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_%2x24, 4, 10, 10
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m9, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+    mov             r9d, %2 / 8
+.loopW:
+    PROCESS_CHROMA_S_AVX2_W8_16R %1
+%ifidn %1,sp
+    add             r2, 8
+%else
+    add             r2, 16
+%endif
+    add             r0, 16
+    dec             r9d
+    jnz             .loopW
+%ifidn %1,sp
+    lea             r2, [r8 + r3 * 4 - %2 + 8]
+%else
+    lea             r2, [r8 + r3 * 4 - 2 * %2 + 16]
+%endif
+    lea             r0, [r7 - 2 * %2 + 16]
+    mova            m7, m9
+    mov             r9d, %2 / 8
+.loop:
+    PROCESS_CHROMA_S_AVX2_W8_8R %1
+%ifidn %1,sp
+    add             r2, 8
+%else
+    add             r2, 16
+%endif
+    add             r0, 16
+    dec             r9d
+    jnz             .loop
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_Nx24 sp, 32
+    FILTER_VER_CHROMA_S_AVX2_Nx24 sp, 16
+    FILTER_VER_CHROMA_S_AVX2_Nx24 ss, 32
+    FILTER_VER_CHROMA_S_AVX2_Nx24 ss, 16
+
+%macro FILTER_VER_CHROMA_S_AVX2_2x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_2x8, 4, 6, 7
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+%ifidn %1,sp
+    mova            m6, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    movd            xm0, [r0]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm0, xm1
+    movd            xm2, [r0 + r1 * 2]
+    punpcklwd       xm1, xm2
+    punpcklqdq      xm0, xm1                        ; m0 = [2 1 1 0]
+    movd            xm3, [r0 + r4]
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movd            xm4, [r0]
+    punpcklwd       xm3, xm4
+    punpcklqdq      xm2, xm3                        ; m2 = [4 3 3 2]
+    vinserti128     m0, m0, xm2, 1                  ; m0 = [4 3 3 2 2 1 1 0]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm4, xm1
+    movd            xm3, [r0 + r1 * 2]
+    punpcklwd       xm1, xm3
+    punpcklqdq      xm4, xm1                        ; m4 = [6 5 5 4]
+    vinserti128     m2, m2, xm4, 1                  ; m2 = [6 5 5 4 4 3 3 2]
+    pmaddwd         m0, [r5]
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+    movd            xm1, [r0 + r4]
+    punpcklwd       xm3, xm1
+    lea             r0, [r0 + 4 * r1]
+    movd            xm2, [r0]
+    punpcklwd       xm1, xm2
+    punpcklqdq      xm3, xm1                        ; m3 = [8 7 7 6]
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [8 7 7 6 6 5 5 4]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm2, xm1
+    movd            xm5, [r0 + r1 * 2]
+    punpcklwd       xm1, xm5
+    punpcklqdq      xm2, xm1                        ; m2 = [10 9 9 8]
+    vinserti128     m3, m3, xm2, 1                  ; m3 = [10 9 9 8 8 7 7 6]
+    pmaddwd         m4, [r5]
+    pmaddwd         m3, [r5 + 1 * mmsize]
+    paddd           m4, m3
+%ifidn %1,sp
+    paddd           m0, m6
+    paddd           m4, m6
+    psrad           m0, 12
+    psrad           m4, 12
+%else
+    psrad           m0, 6
+    psrad           m4, 6
+%endif
+    packssdw        m0, m4
+    vextracti128    xm4, m0, 1
+    lea             r4, [r3 * 3]
+%ifidn %1,sp
+    packuswb        xm0, xm4
+    pextrw          [r2], xm0, 0
+    pextrw          [r2 + r3], xm0, 1
+    pextrw          [r2 + 2 * r3], xm0, 4
+    pextrw          [r2 + r4], xm0, 5
+    lea             r2, [r2 + r3 * 4]
+    pextrw          [r2], xm0, 2
+    pextrw          [r2 + r3], xm0, 3
+    pextrw          [r2 + 2 * r3], xm0, 6
+    pextrw          [r2 + r4], xm0, 7
+%else
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    movd            [r2 + 2 * r3], xm4
+    pextrd          [r2 + r4], xm4, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm0, 2
+    pextrd          [r2 + r3], xm0, 3
+    pextrd          [r2 + 2 * r3], xm4, 2
+    pextrd          [r2 + r4], xm4, 3
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_2x8 sp
+    FILTER_VER_CHROMA_S_AVX2_2x8 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_2x16 1
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_2x16, 4, 6, 9
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+    sub             r0, r1
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+%ifidn %1,sp
+    mova            m6, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    movd            xm0, [r0]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm0, xm1
+    movd            xm2, [r0 + r1 * 2]
+    punpcklwd       xm1, xm2
+    punpcklqdq      xm0, xm1                        ; m0 = [2 1 1 0]
+    movd            xm3, [r0 + r4]
+    punpcklwd       xm2, xm3
+    lea             r0, [r0 + 4 * r1]
+    movd            xm4, [r0]
+    punpcklwd       xm3, xm4
+    punpcklqdq      xm2, xm3                        ; m2 = [4 3 3 2]
+    vinserti128     m0, m0, xm2, 1                  ; m0 = [4 3 3 2 2 1 1 0]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm4, xm1
+    movd            xm3, [r0 + r1 * 2]
+    punpcklwd       xm1, xm3
+    punpcklqdq      xm4, xm1                        ; m4 = [6 5 5 4]
+    vinserti128     m2, m2, xm4, 1                  ; m2 = [6 5 5 4 4 3 3 2]
+    pmaddwd         m0, [r5]
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m0, m2
+    movd            xm1, [r0 + r4]
+    punpcklwd       xm3, xm1
+    lea             r0, [r0 + 4 * r1]
+    movd            xm2, [r0]
+    punpcklwd       xm1, xm2
+    punpcklqdq      xm3, xm1                        ; m3 = [8 7 7 6]
+    vinserti128     m4, m4, xm3, 1                  ; m4 = [8 7 7 6 6 5 5 4]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm2, xm1
+    movd            xm5, [r0 + r1 * 2]
+    punpcklwd       xm1, xm5
+    punpcklqdq      xm2, xm1                        ; m2 = [10 9 9 8]
+    vinserti128     m3, m3, xm2, 1                  ; m3 = [10 9 9 8 8 7 7 6]
+    pmaddwd         m4, [r5]
+    pmaddwd         m3, [r5 + 1 * mmsize]
+    paddd           m4, m3
+    movd            xm1, [r0 + r4]
+    punpcklwd       xm5, xm1
+    lea             r0, [r0 + 4 * r1]
+    movd            xm3, [r0]
+    punpcklwd       xm1, xm3
+    punpcklqdq      xm5, xm1                        ; m5 = [12 11 11 10]
+    vinserti128     m2, m2, xm5, 1                  ; m2 = [12 11 11 10 10 9 9 8]
+    movd            xm1, [r0 + r1]
+    punpcklwd       xm3, xm1
+    movd            xm7, [r0 + r1 * 2]
+    punpcklwd       xm1, xm7
+    punpcklqdq      xm3, xm1                        ; m3 = [14 13 13 12]
+    vinserti128     m5, m5, xm3, 1                  ; m5 = [14 13 13 12 12 11 11 10]
+    pmaddwd         m2, [r5]
+    pmaddwd         m5, [r5 + 1 * mmsize]
+    paddd           m2, m5
+    movd            xm5, [r0 + r4]
+    punpcklwd       xm7, xm5
+    lea             r0, [r0 + 4 * r1]
+    movd            xm1, [r0]
+    punpcklwd       xm5, xm1
+    punpcklqdq      xm7, xm5                        ; m7 = [16 15 15 14]
+    vinserti128     m3, m3, xm7, 1                  ; m3 = [16 15 15 14 14 13 13 12]
+    movd            xm5, [r0 + r1]
+    punpcklwd       xm1, xm5
+    movd            xm8, [r0 + r1 * 2]
+    punpcklwd       xm5, xm8
+    punpcklqdq      xm1, xm5                        ; m1 = [18 17 17 16]
+    vinserti128     m7, m7, xm1, 1                  ; m7 = [18 17 17 16 16 15 15 14]
+    pmaddwd         m3, [r5]
+    pmaddwd         m7, [r5 + 1 * mmsize]
+    paddd           m3, m7
+%ifidn %1,sp
+    paddd           m0, m6
+    paddd           m4, m6
+    paddd           m2, m6
+    paddd           m3, m6
+    psrad           m0, 12
+    psrad           m4, 12
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m0, 6
+    psrad           m4, 6
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m0, m4
+    packssdw        m2, m3
+    lea             r4, [r3 * 3]
+%ifidn %1,sp
+    packuswb        m0, m2
+    vextracti128    xm2, m0, 1
+    pextrw          [r2], xm0, 0
+    pextrw          [r2 + r3], xm0, 1
+    pextrw          [r2 + 2 * r3], xm2, 0
+    pextrw          [r2 + r4], xm2, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrw          [r2], xm0, 2
+    pextrw          [r2 + r3], xm0, 3
+    pextrw          [r2 + 2 * r3], xm2, 2
+    pextrw          [r2 + r4], xm2, 3
+    lea             r2, [r2 + r3 * 4]
+    pextrw          [r2], xm0, 4
+    pextrw          [r2 + r3], xm0, 5
+    pextrw          [r2 + 2 * r3], xm2, 4
+    pextrw          [r2 + r4], xm2, 5
+    lea             r2, [r2 + r3 * 4]
+    pextrw          [r2], xm0, 6
+    pextrw          [r2 + r3], xm0, 7
+    pextrw          [r2 + 2 * r3], xm2, 6
+    pextrw          [r2 + r4], xm2, 7
+%else
+    vextracti128    xm4, m0, 1
+    vextracti128    xm3, m2, 1
+    movd            [r2], xm0
+    pextrd          [r2 + r3], xm0, 1
+    movd            [r2 + 2 * r3], xm4
+    pextrd          [r2 + r4], xm4, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm0, 2
+    pextrd          [r2 + r3], xm0, 3
+    pextrd          [r2 + 2 * r3], xm4, 2
+    pextrd          [r2 + r4], xm4, 3
+    lea             r2, [r2 + r3 * 4]
+    movd            [r2], xm2
+    pextrd          [r2 + r3], xm2, 1
+    movd            [r2 + 2 * r3], xm3
+    pextrd          [r2 + r4], xm3, 1
+    lea             r2, [r2 + r3 * 4]
+    pextrd          [r2], xm2, 2
+    pextrd          [r2 + r3], xm2, 3
+    pextrd          [r2 + 2 * r3], xm3, 2
+    pextrd          [r2 + r4], xm3, 3
+%endif
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_2x16 sp
+    FILTER_VER_CHROMA_S_AVX2_2x16 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_6x8 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_6x8, 4, 6, 8
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m7, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    pmaddwd         m2, [r5]
+    paddd           m0, m4
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    pmaddwd         m3, [r5]
+    paddd           m1, m5
+%ifidn %1,sp
+    paddd           m0, m7
+    paddd           m1, m7
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm1, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m1
+%ifidn %1,sp
+    paddd           m2, m7
+    paddd           m3, m7
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m2, m3
+
+    movu            xm1, [r0 + r4]                  ; m1 = row 7
+    punpckhwd       xm3, xm6, xm1
+    punpcklwd       xm6, xm1
+    vinserti128     m6, m6, xm3, 1
+    pmaddwd         m3, m6, [r5 + 1 * mmsize]
+    pmaddwd         m6, [r5]
+    paddd           m4, m3
+
+    lea             r4, [r3 * 3]
+%ifidn %1,sp
+    packuswb        m0, m2
+    vextracti128    xm2, m0, 1
+    movd            [r2], xm0
+    pextrw          [r2 + 4], xm2, 0
+    pextrd          [r2 + r3], xm0, 1
+    pextrw          [r2 + r3 + 4], xm2, 2
+    pextrd          [r2 + r3 * 2], xm0, 2
+    pextrw          [r2 + r3 * 2 + 4], xm2, 4
+    pextrd          [r2 + r4], xm0, 3
+    pextrw          [r2 + r4 + 4], xm2, 6
+%else
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r4], xm2
+    vextracti128    xm0, m0, 1
+    vextracti128    xm3, m2, 1
+    movd            [r2 + 8], xm0
+    pextrd          [r2 + r3 + 8], xm0, 2
+    movd            [r2 + r3 * 2 + 8], xm3
+    pextrd          [r2 + r4 + 8], xm3, 2
+%endif
+    lea             r2, [r2 + r3 * 4]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm0, [r0]                       ; m0 = row 8
+    punpckhwd       xm2, xm1, xm0
+    punpcklwd       xm1, xm0
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    pmaddwd         m1, [r5]
+    paddd           m5, m2
+%ifidn %1,sp
+    paddd           m4, m7
+    paddd           m5, m7
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m4, m5
+
+    movu            xm2, [r0 + r1]                  ; m2 = row 9
+    punpckhwd       xm5, xm0, xm2
+    punpcklwd       xm0, xm2
+    vinserti128     m0, m0, xm5, 1
+    pmaddwd         m0, [r5 + 1 * mmsize]
+    paddd           m6, m0
+    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
+    punpckhwd       xm0, xm2, xm5
+    punpcklwd       xm2, xm5
+    vinserti128     m2, m2, xm0, 1
+    pmaddwd         m2, [r5 + 1 * mmsize]
+    paddd           m1, m2
+
+%ifidn %1,sp
+    paddd           m6, m7
+    paddd           m1, m7
+    psrad           m6, 12
+    psrad           m1, 12
+%else
+    psrad           m6, 6
+    psrad           m1, 6
+%endif
+    packssdw        m6, m1
+%ifidn %1,sp
+    packuswb        m4, m6
+    vextracti128    xm6, m4, 1
+    movd            [r2], xm4
+    pextrw          [r2 + 4], xm6, 0
+    pextrd          [r2 + r3], xm4, 1
+    pextrw          [r2 + r3 + 4], xm6, 2
+    pextrd          [r2 + r3 * 2], xm4, 2
+    pextrw          [r2 + r3 * 2 + 4], xm6, 4
+    pextrd          [r2 + r4], xm4, 3
+    pextrw          [r2 + r4 + 4], xm6, 6
+%else
+    movq            [r2], xm4
+    movhps          [r2 + r3], xm4
+    movq            [r2 + r3 * 2], xm6
+    movhps          [r2 + r4], xm6
+    vextracti128    xm5, m4, 1
+    vextracti128    xm1, m6, 1
+    movd            [r2 + 8], xm5
+    pextrd          [r2 + r3 + 8], xm5, 2
+    movd            [r2 + r3 * 2 + 8], xm1
+    pextrd          [r2 + r4 + 8], xm1, 2
+%endif
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_6x8 sp
+    FILTER_VER_CHROMA_S_AVX2_6x8 ss
+
+%macro FILTER_VER_CHROMA_S_AVX2_6x16 1
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_6x16, 4, 7, 9
+    mov             r4d, r4m
+    shl             r4d, 6
+    add             r1d, r1d
+
+%ifdef PIC
+    lea             r5, [pw_ChromaCoeffV]
+    add             r5, r4
+%else
+    lea             r5, [pw_ChromaCoeffV + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r1
+%ifidn %1,sp
+    mova            m8, [v4_pd_526336]
+%else
+    add             r3d, r3d
+%endif
+    lea             r6, [r3 * 3]
+    movu            xm0, [r0]                       ; m0 = row 0
+    movu            xm1, [r0 + r1]                  ; m1 = row 1
+    punpckhwd       xm2, xm0, xm1
+    punpcklwd       xm0, xm1
+    vinserti128     m0, m0, xm2, 1
+    pmaddwd         m0, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 2
+    punpckhwd       xm3, xm1, xm2
+    punpcklwd       xm1, xm2
+    vinserti128     m1, m1, xm3, 1
+    pmaddwd         m1, [r5]
+    movu            xm3, [r0 + r4]                  ; m3 = row 3
+    punpckhwd       xm4, xm2, xm3
+    punpcklwd       xm2, xm3
+    vinserti128     m2, m2, xm4, 1
+    pmaddwd         m4, m2, [r5 + 1 * mmsize]
+    paddd           m0, m4
+    pmaddwd         m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm4, [r0]                       ; m4 = row 4
+    punpckhwd       xm5, xm3, xm4
+    punpcklwd       xm3, xm4
+    vinserti128     m3, m3, xm5, 1
+    pmaddwd         m5, m3, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m3, [r5]
+%ifidn %1,sp
+    paddd           m0, m8
+    paddd           m1, m8
+    psrad           m0, 12
+    psrad           m1, 12
+%else
+    psrad           m0, 6
+    psrad           m1, 6
+%endif
+    packssdw        m0, m1
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 5
+    punpckhwd       xm6, xm4, xm5
+    punpcklwd       xm4, xm5
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m2, m6
+    pmaddwd         m4, [r5]
+    movu            xm6, [r0 + r1 * 2]              ; m6 = row 6
+    punpckhwd       xm1, xm5, xm6
+    punpcklwd       xm5, xm6
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    pmaddwd         m5, [r5]
+    paddd           m3, m1
+%ifidn %1,sp
+    paddd           m2, m8
+    paddd           m3, m8
+    psrad           m2, 12
+    psrad           m3, 12
+%else
+    psrad           m2, 6
+    psrad           m3, 6
+%endif
+    packssdw        m2, m3
+%ifidn %1,sp
+    packuswb        m0, m2
+    vextracti128    xm2, m0, 1
+    movd            [r2], xm0
+    pextrw          [r2 + 4], xm2, 0
+    pextrd          [r2 + r3], xm0, 1
+    pextrw          [r2 + r3 + 4], xm2, 2
+    pextrd          [r2 + r3 * 2], xm0, 2
+    pextrw          [r2 + r3 * 2 + 4], xm2, 4
+    pextrd          [r2 + r6], xm0, 3
+    pextrw          [r2 + r6 + 4], xm2, 6
+%else
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+    vextracti128    xm0, m0, 1
+    vextracti128    xm3, m2, 1
+    movd            [r2 + 8], xm0
+    pextrd          [r2 + r3 + 8], xm0, 2
+    movd            [r2 + r3 * 2 + 8], xm3
+    pextrd          [r2 + r6 + 8], xm3, 2
+%endif
+    lea             r2, [r2 + r3 * 4]
+    movu            xm1, [r0 + r4]                  ; m1 = row 7
+    punpckhwd       xm0, xm6, xm1
+    punpcklwd       xm6, xm1
+    vinserti128     m6, m6, xm0, 1
+    pmaddwd         m0, m6, [r5 + 1 * mmsize]
+    pmaddwd         m6, [r5]
+    paddd           m4, m0
+    lea             r0, [r0 + r1 * 4]
+    movu            xm0, [r0]                       ; m0 = row 8
+    punpckhwd       xm2, xm1, xm0
+    punpcklwd       xm1, xm0
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    pmaddwd         m1, [r5]
+    paddd           m5, m2
+%ifidn %1,sp
+    paddd           m4, m8
+    paddd           m5, m8
+    psrad           m4, 12
+    psrad           m5, 12
+%else
+    psrad           m4, 6
+    psrad           m5, 6
+%endif
+    packssdw        m4, m5
+
+    movu            xm2, [r0 + r1]                  ; m2 = row 9
+    punpckhwd       xm5, xm0, xm2
+    punpcklwd       xm0, xm2
+    vinserti128     m0, m0, xm5, 1
+    pmaddwd         m5, m0, [r5 + 1 * mmsize]
+    paddd           m6, m5
+    pmaddwd         m0, [r5]
+    movu            xm5, [r0 + r1 * 2]              ; m5 = row 10
+    punpckhwd       xm7, xm2, xm5
+    punpcklwd       xm2, xm5
+    vinserti128     m2, m2, xm7, 1
+    pmaddwd         m7, m2, [r5 + 1 * mmsize]
+    paddd           m1, m7
+    pmaddwd         m2, [r5]
+
+%ifidn %1,sp
+    paddd           m6, m8
+    paddd           m1, m8
+    psrad           m6, 12
+    psrad           m1, 12
+%else
+    psrad           m6, 6
+    psrad           m1, 6
+%endif
+    packssdw        m6, m1
+%ifidn %1,sp
+    packuswb        m4, m6
+    vextracti128    xm6, m4, 1
+    movd            [r2], xm4
+    pextrw          [r2 + 4], xm6, 0
+    pextrd          [r2 + r3], xm4, 1
+    pextrw          [r2 + r3 + 4], xm6, 2
+    pextrd          [r2 + r3 * 2], xm4, 2
+    pextrw          [r2 + r3 * 2 + 4], xm6, 4
+    pextrd          [r2 + r6], xm4, 3
+    pextrw          [r2 + r6 + 4], xm6, 6
+%else
+    movq            [r2], xm4
+    movhps          [r2 + r3], xm4
+    movq            [r2 + r3 * 2], xm6
+    movhps          [r2 + r6], xm6
+    vextracti128    xm4, m4, 1
+    vextracti128    xm1, m6, 1
+    movd            [r2 + 8], xm4
+    pextrd          [r2 + r3 + 8], xm4, 2
+    movd            [r2 + r3 * 2 + 8], xm1
+    pextrd          [r2 + r6 + 8], xm1, 2
+%endif
+    lea             r2, [r2 + r3 * 4]
+    movu            xm7, [r0 + r4]                  ; m7 = row 11
+    punpckhwd       xm1, xm5, xm7
+    punpcklwd       xm5, xm7
+    vinserti128     m5, m5, xm1, 1
+    pmaddwd         m1, m5, [r5 + 1 * mmsize]
+    paddd           m0, m1
+    pmaddwd         m5, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm1, [r0]                       ; m1 = row 12
+    punpckhwd       xm4, xm7, xm1
+    punpcklwd       xm7, xm1
+    vinserti128     m7, m7, xm4, 1
+    pmaddwd         m4, m7, [r5 + 1 * mmsize]
+    paddd           m2, m4
+    pmaddwd         m7, [r5]
+%ifidn %1,sp
+    paddd           m0, m8
+    paddd           m2, m8
+    psrad           m0, 12
+    psrad           m2, 12
+%else
+    psrad           m0, 6
+    psrad           m2, 6
+%endif
+    packssdw        m0, m2
+
+    movu            xm4, [r0 + r1]                  ; m4 = row 13
+    punpckhwd       xm2, xm1, xm4
+    punpcklwd       xm1, xm4
+    vinserti128     m1, m1, xm2, 1
+    pmaddwd         m2, m1, [r5 + 1 * mmsize]
+    paddd           m5, m2
+    pmaddwd         m1, [r5]
+    movu            xm2, [r0 + r1 * 2]              ; m2 = row 14
+    punpckhwd       xm6, xm4, xm2
+    punpcklwd       xm4, xm2
+    vinserti128     m4, m4, xm6, 1
+    pmaddwd         m6, m4, [r5 + 1 * mmsize]
+    paddd           m7, m6
+    pmaddwd         m4, [r5]
+%ifidn %1,sp
+    paddd           m5, m8
+    paddd           m7, m8
+    psrad           m5, 12
+    psrad           m7, 12
+%else
+    psrad           m5, 6
+    psrad           m7, 6
+%endif
+    packssdw        m5, m7
+%ifidn %1,sp
+    packuswb        m0, m5
+    vextracti128    xm5, m0, 1
+    movd            [r2], xm0
+    pextrw          [r2 + 4], xm5, 0
+    pextrd          [r2 + r3], xm0, 1
+    pextrw          [r2 + r3 + 4], xm5, 2
+    pextrd          [r2 + r3 * 2], xm0, 2
+    pextrw          [r2 + r3 * 2 + 4], xm5, 4
+    pextrd          [r2 + r6], xm0, 3
+    pextrw          [r2 + r6 + 4], xm5, 6
+%else
+    movq            [r2], xm0
+    movhps          [r2 + r3], xm0
+    movq            [r2 + r3 * 2], xm5
+    movhps          [r2 + r6], xm5
+    vextracti128    xm0, m0, 1
+    vextracti128    xm7, m5, 1
+    movd            [r2 + 8], xm0
+    pextrd          [r2 + r3 + 8], xm0, 2
+    movd            [r2 + r3 * 2 + 8], xm7
+    pextrd          [r2 + r6 + 8], xm7, 2
+%endif
+    lea             r2, [r2 + r3 * 4]
+
+    movu            xm6, [r0 + r4]                  ; m6 = row 15
+    punpckhwd       xm5, xm2, xm6
+    punpcklwd       xm2, xm6
+    vinserti128     m2, m2, xm5, 1
+    pmaddwd         m5, m2, [r5 + 1 * mmsize]
+    paddd           m1, m5
+    pmaddwd         m2, [r5]
+    lea             r0, [r0 + r1 * 4]
+    movu            xm0, [r0]                       ; m0 = row 16
+    punpckhwd       xm5, xm6, xm0
+    punpcklwd       xm6, xm0
+    vinserti128     m6, m6, xm5, 1
+    pmaddwd         m5, m6, [r5 + 1 * mmsize]
+    paddd           m4, m5
+    pmaddwd         m6, [r5]
+%ifidn %1,sp
+    paddd           m1, m8
+    paddd           m4, m8
+    psrad           m1, 12
+    psrad           m4, 12
+%else
+    psrad           m1, 6
+    psrad           m4, 6
+%endif
+    packssdw        m1, m4
+
+    movu            xm5, [r0 + r1]                  ; m5 = row 17
+    punpckhwd       xm4, xm0, xm5
+    punpcklwd       xm0, xm5
+    vinserti128     m0, m0, xm4, 1
+    pmaddwd         m0, [r5 + 1 * mmsize]
+    paddd           m2, m0
+    movu            xm4, [r0 + r1 * 2]              ; m4 = row 18
+    punpckhwd       xm0, xm5, xm4
+    punpcklwd       xm5, xm4
+    vinserti128     m5, m5, xm0, 1
+    pmaddwd         m5, [r5 + 1 * mmsize]
+    paddd           m6, m5
+%ifidn %1,sp
+    paddd           m2, m8
+    paddd           m6, m8
+    psrad           m2, 12
+    psrad           m6, 12
+%else
+    psrad           m2, 6
+    psrad           m6, 6
+%endif
+    packssdw        m2, m6
+%ifidn %1,sp
+    packuswb        m1, m2
+    vextracti128    xm2, m1, 1
+    movd            [r2], xm1
+    pextrw          [r2 + 4], xm2, 0
+    pextrd          [r2 + r3], xm1, 1
+    pextrw          [r2 + r3 + 4], xm2, 2
+    pextrd          [r2 + r3 * 2], xm1, 2
+    pextrw          [r2 + r3 * 2 + 4], xm2, 4
+    pextrd          [r2 + r6], xm1, 3
+    pextrw          [r2 + r6 + 4], xm2, 6
+%else
+    movq            [r2], xm1
+    movhps          [r2 + r3], xm1
+    movq            [r2 + r3 * 2], xm2
+    movhps          [r2 + r6], xm2
+    vextracti128    xm4, m1, 1
+    vextracti128    xm6, m2, 1
+    movd            [r2 + 8], xm4
+    pextrd          [r2 + r3 + 8], xm4, 2
+    movd            [r2 + r3 * 2 + 8], xm6
+    pextrd          [r2 + r6 + 8], xm6, 2
+%endif
+    RET
+%endif
+%endmacro
+
+    FILTER_VER_CHROMA_S_AVX2_6x16 sp
+    FILTER_VER_CHROMA_S_AVX2_6x16 ss
+
+;---------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vertical_ss_%1x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SS_W2_4R 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ss_%1x%2, 5, 6, 5
+
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 5
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       r4d, (%2/4)
+
+.loopH:
+    PROCESS_CHROMA_SP_W2_4R r5
+
+    psrad     m0, 6
+    psrad     m2, 6
+
+    packssdw  m0, m2
+
+    movd      [r2], m0
+    pextrd    [r2 + r3], m0, 1
+    lea       r2, [r2 + 2 * r3]
+    pextrd    [r2], m0, 2
+    pextrd    [r2 + r3], m0, 3
+
+    lea       r2, [r2 + 2 * r3]
+
+    dec       r4d
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SS_W2_4R 2, 4
+    FILTER_VER_CHROMA_SS_W2_4R 2, 8
+
+    FILTER_VER_CHROMA_SS_W2_4R 2, 16
+
+;---------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ss_4x2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;---------------------------------------------------------------------------------------------------------------
+INIT_XMM sse2
+cglobal interp_4tap_vert_ss_4x2, 5, 6, 4
+
+    add        r1d, r1d
+    add        r3d, r3d
+    sub        r0, r1
+    shl        r4d, 5
+
+%ifdef PIC
+    lea        r5, [tab_ChromaCoeffV]
+    lea        r5, [r5 + r4]
+%else
+    lea        r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    movq       m0, [r0]
+    movq       m1, [r0 + r1]
+    punpcklwd  m0, m1                          ;m0=[0 1]
+    pmaddwd    m0, [r5 + 0 *16]                ;m0=[0+1]  Row1
+
+    lea        r0, [r0 + 2 * r1]
+    movq       m2, [r0]
+    punpcklwd  m1, m2                          ;m1=[1 2]
+    pmaddwd    m1, [r5 + 0 *16]                ;m1=[1+2]  Row2
+
+    movq       m3, [r0 + r1]
+    punpcklwd  m2, m3                          ;m4=[2 3]
+    pmaddwd    m2, [r5 + 1 * 16]
+    paddd      m0, m2                          ;m0=[0+1+2+3]  Row1 done
+    psrad      m0, 6
+
+    movq       m2, [r0 + 2 * r1]
+    punpcklwd  m3, m2                          ;m5=[3 4]
+    pmaddwd    m3, [r5 + 1 * 16]
+    paddd      m1, m3                          ;m1=[1+2+3+4]  Row2 done
+    psrad      m1, 6
+
+    packssdw   m0, m1
+
+    movlps     [r2], m0
+    movhps     [r2 + r3], m0
+
+    RET
+
+;-------------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vertical_ss_6x8(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;-------------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SS_W6_H4 2
+INIT_XMM sse4
+cglobal interp_4tap_vert_ss_6x%2, 5, 7, 6
+
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 5
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r6, [r5 + r4]
+%else
+    lea       r6, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       r4d, %2/4
+
+.loopH:
+    PROCESS_CHROMA_SP_W4_4R
+
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+
+    movlps    [r2], m0
+    movhps    [r2 + r3], m0
+    lea       r5, [r2 + 2 * r3]
+    movlps    [r5], m2
+    movhps    [r5 + r3], m2
+
+    lea       r5, [4 * r1 - 2 * 4]
+    sub       r0, r5
+    add       r2, 2 * 4
+
+    PROCESS_CHROMA_SP_W2_4R r6
+
+    psrad     m0, 6
+    psrad     m2, 6
+
+    packssdw  m0, m2
+
+    movd      [r2], m0
+    pextrd    [r2 + r3], m0, 1
+    lea       r2, [r2 + 2 * r3]
+    pextrd    [r2], m0, 2
+    pextrd    [r2 + r3], m0, 3
+
+    sub       r0, 2 * 4
+    lea       r2, [r2 + 2 * r3 - 2 * 4]
+
+    dec       r4d
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SS_W6_H4 6, 8
+
+    FILTER_VER_CHROMA_SS_W6_H4 6, 16
+
+
+;----------------------------------------------------------------------------------------------------------------
+; void interp_4tap_vert_ss_8x%2(int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
+;----------------------------------------------------------------------------------------------------------------
+%macro FILTER_VER_CHROMA_SS_W8_H2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_ss_%1x%2, 5, 6, 7
+
+    add       r1d, r1d
+    add       r3d, r3d
+    sub       r0, r1
+    shl       r4d, 5
+
+%ifdef PIC
+    lea       r5, [tab_ChromaCoeffV]
+    lea       r5, [r5 + r4]
+%else
+    lea       r5, [tab_ChromaCoeffV + r4]
+%endif
+
+    mov       r4d, %2/2
+.loopH:
+    PROCESS_CHROMA_SP_W8_2R
+
+    psrad     m0, 6
+    psrad     m1, 6
+    psrad     m2, 6
+    psrad     m3, 6
+
+    packssdw  m0, m1
+    packssdw  m2, m3
+
+    movu      [r2], m0
+    movu      [r2 + r3], m2
+
+    lea       r2, [r2 + 2 * r3]
+
+    dec       r4d
+    jnz       .loopH
+
+    RET
+%endmacro
+
+    FILTER_VER_CHROMA_SS_W8_H2 8, 2
+    FILTER_VER_CHROMA_SS_W8_H2 8, 4
+    FILTER_VER_CHROMA_SS_W8_H2 8, 6
+    FILTER_VER_CHROMA_SS_W8_H2 8, 8
+    FILTER_VER_CHROMA_SS_W8_H2 8, 16
+    FILTER_VER_CHROMA_SS_W8_H2 8, 32
+
+    FILTER_VER_CHROMA_SS_W8_H2 8, 12
+    FILTER_VER_CHROMA_SS_W8_H2 8, 64
+

x265_2.6.tar.gz/source/common/x86/x86inc.asm -> x265_2.7.tar.gz/source/common/x86/x86inc.asm Changed

@@ -66,6 +66,15 @@
     %endif
 %endif
 
+%define FORMAT_ELF 0
+%ifidn __OUTPUT_FORMAT__,elf
+    %define FORMAT_ELF 1
+%elifidn __OUTPUT_FORMAT__,elf32
+    %define FORMAT_ELF 1
+%elifidn __OUTPUT_FORMAT__,elf64
+    %define FORMAT_ELF 1
+%endif
+
 %ifdef PREFIX
     %define mangle(x) _ %+ x
 %else
@@ -88,6 +97,10 @@
     default rel
 %endif
 
+%ifdef __NASM_VER__
+    %use smartalign
+%endif
+
 ; Macros to eliminate most code duplication between x86_32 and x86_64:
 ; Currently this works only for leaf functions which load all their arguments
 ; into registers at the start, and make no other use of the stack. Luckily that
@@ -685,7 +698,7 @@
         CAT_XDEFINE cglobaled_, %2, 1
     %endif
     %xdefine current_function %2
-    %ifidn __OUTPUT_FORMAT__,elf
+    %if FORMAT_ELF
         global %2:function %%VISIBILITY
     %else
         global %2
@@ -711,14 +724,16 @@
 
 ; like cextern, but without the prefix
 %macro cextern_naked 1
-    %xdefine %1 mangle(%1)
+    %ifdef PREFIX
+        %xdefine %1 mangle(%1)
+    %endif
     CAT_XDEFINE cglobaled_, %1, 1
     extern %1
 %endmacro
 
 %macro const 1-2+
     %xdefine %1 mangle(private_prefix %+ _ %+ %1)
-    %ifidn __OUTPUT_FORMAT__,elf
+    %if FORMAT_ELF
         global %1:data hidden
     %else
         global %1
@@ -727,9 +742,8 @@
     %1: %2
 %endmacro
 
-; This is needed for ELF, otherwise the GNU linker assumes the stack is
-; executable by default.
-%ifidn __OUTPUT_FORMAT__,elf
+; This is needed for ELF, otherwise the GNU linker assumes the stack is executable by default.
+%if FORMAT_ELF
     [SECTION .note.GNU-stack noalloc noexec nowrite progbits]
 %endif
 
@@ -801,9 +815,17 @@
     %endif
 
     %if ARCH_X86_64 || cpuflag(sse2)
-        CPU amdnop
+        %ifdef __NASM_VER__
+            ALIGNMODE p6
+        %else
+            CPU amdnop
+        %endif
     %else
-        CPU basicnop
+        %ifdef __NASM_VER__
+            ALIGNMODE nop
+        %else
+            CPU basicnop
+        %endif
     %endif
 %endmacro
 
@@ -1467,7 +1489,7 @@
                 v%5%6 %1, %2, %3, %4
             %elifidn %1, %2
                 ; If %3 or %4 is a memory operand it needs to be encoded as the last operand.
-                %ifid %3
+                %ifnum sizeof%3
                     v%{5}213%6 %2, %3, %4
                 %else
                     v%{5}132%6 %2, %4, %3
@@ -1491,14 +1513,3 @@
 FMA4_INSTR fmsubadd, pd, ps
 FMA4_INSTR fnmadd,   pd, ps, sd, ss
 FMA4_INSTR fnmsub,   pd, ps, sd, ss
-
-; workaround: vpbroadcastq is broken in x86_32 due to a yasm bug (fixed in 1.3.0)
-%if __YASM_VERSION_ID__ < 0x01030000 && ARCH_X86_64 == 0
-    %macro vpbroadcastq 2
-        %if sizeof%1 == 16
-            movddup %1, %2
-        %else
-            vbroadcastsd %1, %2
-        %endif
-    %endmacro
-%endif

x265_2.6.tar.gz/source/dynamicHDR10/JsonHelper.cpp -> x265_2.7.tar.gz/source/dynamicHDR10/JsonHelper.cpp Changed

@@ -139,21 +139,13 @@
         return JsonObject();
     }
 
-    ifstream tfile;
-    string json_str;
-    string json_str2;
+    std::ifstream ifs(path);
+    const std::string json_str2((std::istreambuf_iterator<char>(ifs)),
+                                (std::istreambuf_iterator<char>()));
+
     string err = "";
-    tfile.open(path);
-    while(tfile)
-    {
-        std::getline(tfile, json_str);
-        json_str2.append(json_str);
-    }
 
-    tfile.close();
-    size_t beginning = json_str2.find_first_of("{");
-    int fixchar = json_str2[json_str2.size() - 2] == '}' ? 1 : 0;
-    return Json::parse(json_str2.substr(beginning,json_str2.size() - fixchar),err).object_items();
+    return Json::parse(json_str2,err, JsonParse::COMMENTS).object_items();
 }
 
 JsonArray JsonHelper::readJsonArray(const string &path)
@@ -174,28 +166,13 @@
         return JsonArray();
     }
 
-    ifstream tfile;
-    string json_str;
-    string json_str2;
-    string err = "";
-    tfile.open(path);
-    while(tfile)
-    {
-        std::getline(tfile, json_str);
-        json_str2.append(json_str);
-    }
+    std::ifstream ifs(path);
+    const std::string json_str2((std::istreambuf_iterator<char>(ifs)),
+                                (std::istreambuf_iterator<char>()));
 
-    tfile.close();
+    string err = "";
 
-    vector<Json> data;
-    if (json_str2.size() != 0)
-    {
-        size_t beginning = json_str2.find_first_of("[");
-        int fixchar = json_str2[json_str2.size() - 2] == ']' ? 1 : 0;
-        return Json::parse(json_str2.substr(beginning, json_str2.size() - fixchar), err).array_items();
-    }
-    else
-        return data;
+    return Json::parse(json_str2,err, JsonParse::COMMENTS).array_items();
 }
 
 bool JsonHelper::validatePathExtension(string &path)

x265_2.6.tar.gz/source/dynamicHDR10/SeiMetadataDictionary.cpp -> x265_2.7.tar.gz/source/dynamicHDR10/SeiMetadataDictionary.cpp Changed

x265_2.6.tar.gz/source/dynamicHDR10/SeiMetadataDictionary.h -> x265_2.7.tar.gz/source/dynamicHDR10/SeiMetadataDictionary.h Changed

x265_2.6.tar.gz/source/dynamicHDR10/metadataFromJson.cpp -> x265_2.7.tar.gz/source/dynamicHDR10/metadataFromJson.cpp Changed

@@ -372,7 +372,7 @@
     const uint16_t terminalProviderCode = 0x003C;
     const uint16_t terminalProviderOrientedCode = 0x0001;
     const uint8_t applicationIdentifier = 4;
-    const uint8_t applicationVersion = 0;
+    const uint8_t applicationVersion = 1;
 
     mPimpl->appendBits(metadata, countryCode, 8);
     mPimpl->appendBits(metadata, terminalProviderCode, 16);
@@ -384,9 +384,7 @@
     //Note: Validated only add up to two local selections, ignore the rest
     JsonArray jsonArray = fileData[frame][JsonDataKeys::LocalParameters].array_items();
     int ellipsesNum = static_cast<int>(jsonArray.size() > 2 ? 2 : jsonArray.size());
-
-    uint16_t numWindows = 1 + static_cast<uint16_t>(ellipsesNum);
-
+    uint16_t numWindows = (uint16_t)fileData[frame][JsonDataKeys::NumberOfWindows].int_value();
     mPimpl->appendBits(metadata, numWindows, 2);
     for (int i = 0; i < ellipsesNum; ++i)
     {
@@ -426,16 +424,15 @@
 
         mPimpl->appendBits(metadata, semimajorExternalAxis, 16);
         mPimpl->appendBits(metadata, semiminorExternalAxis, 16);
-        /*bool*/ uint8_t overlapProcessOption = static_cast<uint8_t>(ellipseJsonObject[EllipseNames::OverlapProcessOption].int_value()); //1; 
+        uint8_t overlapProcessOption = static_cast<uint8_t>(ellipseJsonObject[EllipseNames::OverlapProcessOption].int_value());
         //TODO: Uses Layering method, the value is "1"
         mPimpl->appendBits(metadata, overlapProcessOption, 1);
     }
     /* Targeted System Display Data */
-    uint32_t TEMPmonitorPeak = fileData[frame][JsonDataKeys::TargetDisplayLuminance].int_value();     //500;
-    mPimpl->appendBits(metadata, TEMPmonitorPeak, 27);
-
+    uint32_t monitorPeak = fileData[frame][JsonDataKeys::TargetDisplayLuminance].int_value();     //500;
+    mPimpl->appendBits(metadata, monitorPeak, 27);
     //NOTE: Set as false for now, as requested
-    /*bool*/uint8_t targetedSystemDisplayActualPeakLuminanceFlag = 0; /*false*/
+    uint8_t targetedSystemDisplayActualPeakLuminanceFlag = 0;
     mPimpl->appendBits(metadata, targetedSystemDisplayActualPeakLuminanceFlag, 1);
     if (targetedSystemDisplayActualPeakLuminanceFlag)
     {
@@ -463,7 +460,6 @@
         mPimpl->appendBits(metadata, static_cast<uint16_t>((int)luminanceData.maxGLuminance & 0xFFFF), 16);
         mPimpl->appendBits(metadata, static_cast<uint8_t>(((int)luminanceData.maxBLuminance & 0x10000) >> 16), 1);
         mPimpl->appendBits(metadata, static_cast<uint16_t>((int)luminanceData.maxBLuminance & 0xFFFF), 16);
-        /* changed from maxRGBLuminance to average luminance to match stms implementation */
         mPimpl->appendBits(metadata, static_cast<uint8_t>(((int)luminanceData.averageLuminance & 0x10000) >> 16), 1);
         mPimpl->appendBits(metadata, static_cast<uint16_t>((int)luminanceData.averageLuminance & 0xFFFF), 16);
 
@@ -478,7 +474,7 @@
             uint8_t distributionMaxrgbPercentage = static_cast<uint8_t>(percentilPercentages.at(i));
             mPimpl->appendBits(metadata, distributionMaxrgbPercentage, 7);
 
-            // 17bits: 1bit then 16
+            /* 17bits: 1bit then 16 */
             unsigned int ithPercentile = luminanceData.percentiles.at(i);
             uint8_t highValue = static_cast<uint8_t>((ithPercentile & 0x10000) >> 16);
             uint16_t lowValue = static_cast<uint16_t>(ithPercentile & 0xFFFF);
@@ -499,33 +495,32 @@
     {
         //TODO
     }
-    // BEZIER CURVE DATA
+    /* Bezier Curve Data */
     for (int w = 0; w < numWindows; ++w)
     {
-        //TODO: 
         uint8_t toneMappingFlag = 1;
-        mPimpl->appendBits(metadata, toneMappingFlag, 1);
-        if (toneMappingFlag)
+		/* Check if the window contains tone mapping bezier curve data and set toneMappingFlag appropriately */
+		//Json bezierData = fileData[frame][BezierCurveNames::TagName];
+        BezierCurveData curveData;
+		/* Select curve data based on global window */
+        if (w == 0)
         {
-            Json bezierData = fileData[frame][BezierCurveNames::TagName];
-            BezierCurveData curveData;
-
-            /* Select curve data based on global window or local window */
-            if (w == 0)
+            if (!mPimpl->bezierCurveFromJson(fileData[frame][BezierCurveNames::TagName], curveData))
             {
-                if (!mPimpl->bezierCurveFromJson(bezierData, curveData))
-                {
-                    std::cout << "error parsing bezierCurve frame: " << w << std::endl;
-                }
+				toneMappingFlag = 0;
             }
-            else
+        }
+	    /* Select curve data based on local window */
+        else
+        {
+            if (!mPimpl->bezierCurveFromJson(jsonArray[w - 1][BezierCurveNames::TagName], curveData))
             {
-                if (!mPimpl->bezierCurveFromJson(jsonArray[w - 1][BezierCurveNames::TagName], curveData))
-                {
-                    std::cout << "error parsing bezierCurve ellipse: " << w - 1 << std::endl;
-                }
+				toneMappingFlag = 0;
             }
-
+        }		
+        mPimpl->appendBits(metadata, toneMappingFlag, 1);
+        if (toneMappingFlag)
+        {
             uint16_t kneePointX = static_cast<uint16_t>(curveData.sPx);
             mPimpl->appendBits(metadata, kneePointX, 12);
             uint16_t kneePointY = static_cast<uint16_t>(curveData.sPy);
@@ -541,7 +536,7 @@
                 mPimpl->appendBits(metadata, anchor, 10);
             }
         }
-    }
+	}
     /* Set to false as requested */
     bool colorSaturationMappingFlag = 0;
     mPimpl->appendBits(metadata, colorSaturationMappingFlag, 1);

x265_2.6.tar.gz/source/encoder/analysis.cpp -> x265_2.7.tar.gz/source/encoder/analysis.cpp Changed

@@ -100,16 +100,17 @@
     for (uint32_t depth = 0; depth <= m_param->maxCUDepth; depth++, cuSize >>= 1)
     {
         ModeDepth &md = m_modeDepth[depth];
-
-        md.cuMemPool.create(depth, csp, MAX_PRED_TYPES, *m_param);
+        ok &= md.cuMemPool.create(depth, csp, MAX_PRED_TYPES, *m_param);
         ok &= md.fencYuv.create(cuSize, csp);
-
-        for (int j = 0; j < MAX_PRED_TYPES; j++)
+        if (ok)
         {
-            md.pred[j].cu.initialize(md.cuMemPool, depth, *m_param, j);
-            ok &= md.pred[j].predYuv.create(cuSize, csp);
-            ok &= md.pred[j].reconYuv.create(cuSize, csp);
-            md.pred[j].fencYuv = &md.fencYuv;
+            for (int j = 0; j < MAX_PRED_TYPES; j++)
+            {
+                md.pred[j].cu.initialize(md.cuMemPool, depth, *m_param, j);
+                ok &= md.pred[j].predYuv.create(cuSize, csp);
+                ok &= md.pred[j].reconYuv.create(cuSize, csp);
+                md.pred[j].fencYuv = &md.fencYuv;
+            }
         }
     }
     if (m_param->sourceHeight >= 1080)
@@ -158,38 +159,34 @@
     if (m_param->bCTUInfo && (*m_frame->m_ctuInfo + ctu.m_cuAddr))
     {
         x265_ctu_info_t* ctuTemp = *m_frame->m_ctuInfo + ctu.m_cuAddr;
-        if (ctuTemp->ctuPartitions)
-        {
-            int32_t depthIdx = 0;
-            uint32_t maxNum8x8Partitions = 64;
-            uint8_t* depthInfoPtr = m_frame->m_addOnDepth[ctu.m_cuAddr];
-            uint8_t* contentInfoPtr = m_frame->m_addOnCtuInfo[ctu.m_cuAddr];
-            int* prevCtuInfoChangePtr = m_frame->m_addOnPrevChange[ctu.m_cuAddr];
-            do
-            {
-                uint8_t depth = (uint8_t)ctuTemp->ctuPartitions[depthIdx];
-                uint8_t content = (uint8_t)(*((int32_t *)ctuTemp->ctuInfo + depthIdx));
-                int prevCtuInfoChange = m_frame->m_prevCtuInfoChange[ctu.m_cuAddr * maxNum8x8Partitions + depthIdx];
-                memset(depthInfoPtr, depth, sizeof(uint8_t) * numPartition >> 2 * depth);
-                memset(contentInfoPtr, content, sizeof(uint8_t) * numPartition >> 2 * depth);
-                memset(prevCtuInfoChangePtr, 0, sizeof(int) * numPartition >> 2 * depth);
-                for (uint32_t l = 0; l < numPartition >> 2 * depth; l++)
-                    prevCtuInfoChangePtr[l] = prevCtuInfoChange;
-                depthInfoPtr += ctu.m_numPartitions >> 2 * depth;
-                contentInfoPtr += ctu.m_numPartitions >> 2 * depth;
-                prevCtuInfoChangePtr += ctu.m_numPartitions >> 2 * depth;
-                depthIdx++;
-            } while (ctuTemp->ctuPartitions[depthIdx] != 0);
-
-            m_additionalCtuInfo = m_frame->m_addOnCtuInfo[ctu.m_cuAddr];
-            m_prevCtuInfoChange = m_frame->m_addOnPrevChange[ctu.m_cuAddr];
-            memcpy(ctu.m_cuDepth, m_frame->m_addOnDepth[ctu.m_cuAddr], sizeof(uint8_t) * numPartition);
-            //Calculate log2CUSize from depth
-            for (uint32_t i = 0; i < cuGeom.numPartitions; i++)
-                ctu.m_log2CUSize[i] = (uint8_t)m_param->maxLog2CUSize - ctu.m_cuDepth[i];
-        }
+        int32_t depthIdx = 0;
+        uint32_t maxNum8x8Partitions = 64;
+        uint8_t* depthInfoPtr = m_frame->m_addOnDepth[ctu.m_cuAddr];
+        uint8_t* contentInfoPtr = m_frame->m_addOnCtuInfo[ctu.m_cuAddr];
+        int* prevCtuInfoChangePtr = m_frame->m_addOnPrevChange[ctu.m_cuAddr];
+        do
+        {
+            uint8_t depth = (uint8_t)ctuTemp->ctuPartitions[depthIdx];
+            uint8_t content = (uint8_t)(*((int32_t *)ctuTemp->ctuInfo + depthIdx));
+            int prevCtuInfoChange = m_frame->m_prevCtuInfoChange[ctu.m_cuAddr * maxNum8x8Partitions + depthIdx];
+            memset(depthInfoPtr, depth, sizeof(uint8_t) * numPartition >> 2 * depth);
+            memset(contentInfoPtr, content, sizeof(uint8_t) * numPartition >> 2 * depth);
+            memset(prevCtuInfoChangePtr, 0, sizeof(int) * numPartition >> 2 * depth);
+            for (uint32_t l = 0; l < numPartition >> 2 * depth; l++)
+                prevCtuInfoChangePtr[l] = prevCtuInfoChange;
+            depthInfoPtr += ctu.m_numPartitions >> 2 * depth;
+            contentInfoPtr += ctu.m_numPartitions >> 2 * depth;
+            prevCtuInfoChangePtr += ctu.m_numPartitions >> 2 * depth;
+            depthIdx++;
+        } while (ctuTemp->ctuPartitions[depthIdx] != 0);
+
+        m_additionalCtuInfo = m_frame->m_addOnCtuInfo[ctu.m_cuAddr];
+        m_prevCtuInfoChange = m_frame->m_addOnPrevChange[ctu.m_cuAddr];
+        memcpy(ctu.m_cuDepth, m_frame->m_addOnDepth[ctu.m_cuAddr], sizeof(uint8_t) * numPartition);
+        //Calculate log2CUSize from depth
+        for (uint32_t i = 0; i < cuGeom.numPartitions; i++)
+            ctu.m_log2CUSize[i] = (uint8_t)m_param->maxLog2CUSize - ctu.m_cuDepth[i];
     }
-
     if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead)
     {
         m_multipassAnalysis = (analysis2PassFrameData*)m_frame->m_analysis2Pass.analysisFramedata;
@@ -207,11 +204,11 @@
         }
     }
 
-    if (m_param->analysisReuseMode && m_slice->m_sliceType != I_SLICE && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel < 10)
+    if ((m_param->analysisSave || m_param->analysisLoad) && m_slice->m_sliceType != I_SLICE && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel < 10)
     {
         int numPredDir = m_slice->isInterP() ? 1 : 2;
         m_reuseInterDataCTU = (analysis_inter_data*)m_frame->m_analysisData.interData;
-        m_reuseRef = &m_reuseInterDataCTU->ref[ctu.m_cuAddr * X265_MAX_PRED_MODE_PER_CTU * numPredDir];
+        m_reuseRef = &m_reuseInterDataCTU->ref [ctu.m_cuAddr * X265_MAX_PRED_MODE_PER_CTU * numPredDir];
         m_reuseDepth = &m_reuseInterDataCTU->depth[ctu.m_cuAddr * ctu.m_numPartitions];
         m_reuseModes = &m_reuseInterDataCTU->modes[ctu.m_cuAddr * ctu.m_numPartitions];
         if (m_param->analysisReuseLevel > 4)
@@ -219,7 +216,7 @@
             m_reusePartSize = &m_reuseInterDataCTU->partSize[ctu.m_cuAddr * ctu.m_numPartitions];
             m_reuseMergeFlag = &m_reuseInterDataCTU->mergeFlag[ctu.m_cuAddr * ctu.m_numPartitions];
         }
-        if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE)
+        if (m_param->analysisSave && !m_param->analysisLoad)
             for (int i = 0; i < X265_MAX_PRED_MODE_PER_CTU * numPredDir; i++)
                 m_reuseRef[i] = -1;
     }
@@ -228,7 +225,7 @@
     if (m_slice->m_sliceType == I_SLICE)
     {
         analysis_intra_data* intraDataCTU = (analysis_intra_data*)m_frame->m_analysisData.intraData;
-        if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1)
+        if (m_param->analysisLoad && m_param->analysisReuseLevel > 1)
         {
             memcpy(ctu.m_cuDepth, &intraDataCTU->depth[ctu.m_cuAddr * numPartition], sizeof(uint8_t) * numPartition);
             memcpy(ctu.m_lumaIntraDir, &intraDataCTU->modes[ctu.m_cuAddr * numPartition], sizeof(uint8_t) * numPartition);
@@ -239,7 +236,7 @@
     }
     else
     {
-        bool bCopyAnalysis = ((m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel == 10) || (m_param->bMVType && m_param->analysisReuseLevel >= 7 && ctu.m_numPartitions <= 16));
+        bool bCopyAnalysis = ((m_param->analysisLoad && m_param->analysisReuseLevel == 10) || (m_param->bMVType && m_param->analysisReuseLevel >= 7 && ctu.m_numPartitions <= 16));
         bool BCompressInterCUrd0_4 = (m_param->bMVType && m_param->analysisReuseLevel >= 7 && m_param->rdLevel <= 4);
         bool BCompressInterCUrd5_6 = (m_param->bMVType && m_param->analysisReuseLevel >= 7 && m_param->rdLevel >= 5 && m_param->rdLevel <= 6);
         bCopyAnalysis = bCopyAnalysis || BCompressInterCUrd0_4 || BCompressInterCUrd5_6;
@@ -280,7 +277,7 @@
             /* generate residual for entire CTU at once and copy to reconPic */
             encodeResidue(ctu, cuGeom);
         }
-        else if ((m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel == 10) || ((m_param->bMVType == AVC_INFO) && m_param->analysisReuseLevel >= 7))
+        else if ((m_param->analysisLoad && m_param->analysisReuseLevel == 10) || ((m_param->bMVType == AVC_INFO) && m_param->analysisReuseLevel >= 7 && ctu.m_numPartitions <= 16))
         {
             analysis_inter_data* interDataCTU = (analysis_inter_data*)m_frame->m_analysisData.interData;
             int posCTU = ctu.m_cuAddr * numPartition;
@@ -459,11 +456,9 @@
 
     int bestCUQP = qp;
     int lambdaQP = lqp;
-
     bool doQPRefine = (bDecidedDepth && depth <= m_slice->m_pps->maxCuDQPDepth) || (!bDecidedDepth && depth == m_slice->m_pps->maxCuDQPDepth);
-    if (m_param->analysisReuseLevel == 10)
+    if (m_param->analysisReuseLevel >= 7)
         doQPRefine = false;
-
     if (doQPRefine)
     {
         uint64_t bestCUCost, origCUCost, cuCost, cuPrevCost;
@@ -652,13 +647,12 @@
         cacheCost[cuIdx] = md.bestMode->rdCost;
     }
 
-    /* Save Intra CUs TU depth only when analysis mode is OFF */
-    if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4 && !m_param->analysisReuseMode)
+    if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)
     {
         CUData* ctu = md.bestMode->cu.m_encData->getPicCTU(parentCTU.m_cuAddr);
         int8_t maxTUDepth = -1;
         for (uint32_t i = 0; i < cuGeom.numPartitions; i++)
-            maxTUDepth = X265_MAX(maxTUDepth, md.pred[PRED_INTRA].cu.m_tuDepth[i]);
+            maxTUDepth = X265_MAX(maxTUDepth, md.bestMode->cu.m_tuDepth[i]);
         ctu->m_refTuDepth[cuGeom.geomRecurId] = maxTUDepth;
     }
 
@@ -1265,7 +1259,7 @@
                 mightSplit &= !bDecidedDepth;
             }
         }
-        if ((m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10))
+        if ((m_param->analysisLoad && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10))
         {
             if (mightNotSplit && depth == m_reuseDepth[cuGeom.absPartIdx])
             {
@@ -1305,9 +1299,8 @@
                 }
             }
         }
-
         /* Step 1. Evaluate Merge/Skip candidates for likely early-outs, if skip mode was not set above */
-        if ((mightNotSplit && depth >= minDepth && !md.bestMode && !bCtuInfoCheck) || (m_param->bMVType && (m_modeFlag[0] || m_modeFlag[1]))) /* TODO: Re-evaluate if analysis load/save still works */
+        if ((mightNotSplit && depth >= minDepth && !md.bestMode && !bCtuInfoCheck) || (m_param->bMVType && m_param->analysisReuseLevel == 7 && (m_modeFlag[0] || m_modeFlag[1]))) /* TODO: Re-evaluate if analysis load/save still works */
         {
             /* Compute Merge Cost */
             md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
@@ -1317,8 +1310,7 @@
                 skipModes = (m_param->bEnableEarlySkip || m_param->interRefine == 2)
                 && md.bestMode && md.bestMode->cu.isSkipped(0); // TODO: sa8d threshold per depth
         }
-
-        if (md.bestMode && m_param->bEnableRecursionSkip && !bCtuInfoCheck && !(m_param->bMVType && (m_modeFlag[0] || m_modeFlag[1])))
+        if (md.bestMode && m_param->bEnableRecursionSkip && !bCtuInfoCheck && !(m_param->bMVType && m_param->analysisReuseLevel == 7 && (m_modeFlag[0] || m_modeFlag[1])))
         {
             skipRecursion = md.bestMode->cu.isSkipped(0);
             if (mightSplit && depth >= minDepth && !skipRecursion)
@@ -1329,10 +1321,8 @@
                     skipRecursion = complexityCheckCU(*md.bestMode);
             }
         }
-
-        if (m_param->bMVType && md.bestMode && cuGeom.numPartitions <= 16)
+        if (m_param->bMVType && md.bestMode && cuGeom.numPartitions <= 16 && m_param->analysisReuseLevel == 7)
             skipRecursion = true;
-
         /* Step 2. Evaluate each of the 4 split sub-blocks in series */
         if (mightSplit && !skipRecursion)
         {
@@ -1387,11 +1377,20 @@
             else
                 splitPred->sa8dCost = m_rdCost.calcRdSADCost((uint32_t)splitPred->distortion, splitPred->sa8dBits);
         }
-
         /* If analysis mode is simple do not Evaluate other modes */
-        if ((m_param->bMVType && cuGeom.numPartitions <= 16) && (m_slice->m_sliceType == P_SLICE || m_slice->m_sliceType == B_SLICE))
-            mightNotSplit = !(m_checkMergeAndSkipOnly[0] || (m_checkMergeAndSkipOnly[0] && m_checkMergeAndSkipOnly[1]));
-
+        if (m_param->bMVType && m_param->analysisReuseLevel == 7)
+        {
+            if (m_slice->m_sliceType == P_SLICE)
+            {
+                if (m_checkMergeAndSkipOnly[0])
+                    skipModes = true;
+            }
+            else
+            {
+                if (m_checkMergeAndSkipOnly[0] && m_checkMergeAndSkipOnly[1])
+                    skipModes = true;
+            }
+        }
         /* Split CUs
          *   0  1
          *   2  3 */
@@ -1954,7 +1953,7 @@
                 mightSplit &= !bDecidedDepth;
             }
         }
-        if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
+        if (m_param->analysisLoad && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
         {
             if (mightNotSplit && depth == m_reuseDepth[cuGeom.absPartIdx])
             {
@@ -1998,10 +1997,9 @@
                 }
             }
         }
-
         /* Step 1. Evaluate Merge/Skip candidates for likely early-outs */
         if ((mightNotSplit && !md.bestMode && !bCtuInfoCheck) ||
-            (m_param->bMVType && (m_modeFlag[0] || m_modeFlag[1])))
+            (m_param->bMVType && m_param->analysisReuseLevel == 7 && (m_modeFlag[0] || m_modeFlag[1])))
         {
             md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
             md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
@@ -2016,10 +2014,8 @@
             if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)
                 skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);
         }
-
-        if (m_param->bMVType && md.bestMode && cuGeom.numPartitions <= 16)
+        if (m_param->bMVType && md.bestMode && cuGeom.numPartitions <= 16 && m_param->analysisReuseLevel == 7)
             skipRecursion = true;
-
         // estimate split cost
         /* Step 2. Evaluate each of the 4 split sub-blocks in series */
         if (mightSplit && !skipRecursion)
@@ -2071,11 +2067,20 @@
 
             checkDQPForSplitPred(*splitPred, cuGeom);
         }
-
         /* If analysis mode is simple do not Evaluate other modes */
-        if ((m_param->bMVType && cuGeom.numPartitions <= 16) && (m_slice->m_sliceType == P_SLICE || m_slice->m_sliceType == B_SLICE))
-            mightNotSplit = !(m_checkMergeAndSkipOnly[0] || (m_checkMergeAndSkipOnly[0] && m_checkMergeAndSkipOnly[1]));
-
+        if (m_param->bMVType && m_param->analysisReuseLevel == 7)
+        {
+            if (m_slice->m_sliceType == P_SLICE)
+            {
+                if (m_checkMergeAndSkipOnly[0])
+                    skipModes = true;
+            }
+            else
+            {
+                if (m_checkMergeAndSkipOnly[0] && m_checkMergeAndSkipOnly[1])
+                    skipModes = true;
+            }
+        }
         /* Split CUs
          *   0  1
          *   2  3 */
@@ -2881,7 +2886,7 @@
     interMode.cu.setPredModeSubParts(MODE_INTER);
     int numPredDir = m_slice->isInterP() ? 1 : 2;
 
-    if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
+    if (m_param->analysisLoad && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
     {
         int refOffset = cuGeom.geomRecurId * 16 * numPredDir + partSize * numPredDir * 2;
         int index = 0;
@@ -2923,7 +2928,7 @@
     }
     interMode.sa8dCost = m_rdCost.calcRdSADCost((uint32_t)interMode.distortion, interMode.sa8dBits);
 
-    if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1)
+    if (m_param->analysisSave && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1)
     {
         int refOffset = cuGeom.geomRecurId * 16 * numPredDir + partSize * numPredDir * 2;
         int index = 0;
@@ -2945,7 +2950,7 @@
     interMode.cu.setPredModeSubParts(MODE_INTER);
     int numPredDir = m_slice->isInterP() ? 1 : 2;
 
-    if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
+    if (m_param->analysisLoad && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
     {
         int refOffset = cuGeom.geomRecurId * 16 * numPredDir + partSize * numPredDir * 2;
         int index = 0;
@@ -2979,7 +2984,7 @@
     /* predInterSearch sets interMode.sa8dBits, but this is ignored */
     encodeResAndCalcRdInterCU(interMode, cuGeom);
 
-    if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1)
+    if (m_param->analysisSave && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1)
     {
         int refOffset = cuGeom.geomRecurId * 16 * numPredDir + partSize * numPredDir * 2;
         int index = 0;

x265_2.6.tar.gz/source/encoder/api.cpp -> x265_2.7.tar.gz/source/encoder/api.cpp Changed

@@ -67,9 +67,7 @@
     "Y PSNR, U PSNR, V PSNR, Global PSNR, SSIM, SSIM (dB), "
     "I count, I ave-QP, I kbps, I-PSNR Y, I-PSNR U, I-PSNR V, I-SSIM (dB), "
     "P count, P ave-QP, P kbps, P-PSNR Y, P-PSNR U, P-PSNR V, P-SSIM (dB), "
-    "B count, B ave-QP, B kbps, B-PSNR Y, B-PSNR U, B-PSNR V, B-SSIM (dB), "
-    "MaxCLL, MaxFALL, Version\n";
-
+    "B count, B ave-QP, B kbps, B-PSNR Y, B-PSNR U, B-PSNR V, B-SSIM (dB), ";
 x265_encoder *x265_encoder_open(x265_param *p)
 {
     if (!p)
@@ -192,9 +190,10 @@
 {
     if (!enc || !param_in)
         return -1;
-
     x265_param save;
     Encoder* encoder = static_cast<Encoder*>(enc);
+    if (encoder->m_param->csvfn == NULL && param_in->csvfpt != NULL)
+         encoder->m_param->csvfpt = param_in->csvfpt;
     if (encoder->m_latestParam->forceFlush != param_in->forceFlush)
         return encoder->reconfigureParam(encoder->m_latestParam, param_in);
     bool isReconfigureRc = encoder->isReconfigureRc(encoder->m_latestParam, param_in);
@@ -311,7 +310,9 @@
         Encoder *encoder = static_cast<Encoder*>(enc);
         x265_stats stats;
         encoder->fetchStats(&stats, sizeof(stats));
-        x265_csvlog_encode(enc, &stats, argc, argv);
+        int padx = encoder->m_sps.conformanceWindow.rightOffset;
+        int pady = encoder->m_sps.conformanceWindow.bottomOffset;
+        x265_csvlog_encode(encoder->m_param, &stats, padx, pady, argc, argv);
     }
 }
 
@@ -356,13 +357,13 @@
     return -1;
 }
 
-int x265_get_ref_frame_list(x265_encoder *enc, x265_picyuv** l0, x265_picyuv** l1, int sliceType, int poc)
+int x265_get_ref_frame_list(x265_encoder *enc, x265_picyuv** l0, x265_picyuv** l1, int sliceType, int poc, int* pocL0, int* pocL1)
 {
     if (!enc)
         return -1;
 
     Encoder *encoder = static_cast<Encoder*>(enc);
-    return encoder->getRefFrameList((PicYuv**)l0, (PicYuv**)l1, sliceType, poc);
+    return encoder->getRefFrameList((PicYuv**)l0, (PicYuv**)l1, sliceType, poc, pocL0, pocL1);
 }
 
 int x265_set_analysis_data(x265_encoder *enc, x265_analysis_data *analysis_data, int poc, uint32_t cuBytes)
@@ -398,7 +399,7 @@
     pic->userSEI.payloads = NULL;
     pic->userSEI.numPayloads = 0;
 
-    if (param->analysisReuseMode || (param->bMVType == AVC_INFO))
+    if ((param->analysisSave || param->analysisLoad) || (param->bMVType == AVC_INFO))
     {
         uint32_t widthInCU = (param->sourceWidth + param->maxCUSize - 1) >> param->maxLog2CUSize;
         uint32_t heightInCU = (param->sourceHeight + param->maxCUSize - 1) >> param->maxLog2CUSize;
@@ -754,7 +755,12 @@
                 fprintf(csvfp, "\n");
             }
             else
+            {
                 fputs(summaryCSVHeader, csvfp);
+                if (param->csvLogLevel >= 2 || param->maxCLL || param->maxFALL)
+                    fputs("MaxCLL, MaxFALL,", csvfp);
+                fputs(" Version\n", csvfp);
+            }
         }
         return csvfp;
     }
@@ -867,45 +873,40 @@
     fflush(stderr);
 }
 
-void x265_csvlog_encode(x265_encoder *enc, const x265_stats* stats, int argc, char** argv)
+void x265_csvlog_encode(const x265_param *p, const x265_stats *stats, int padx, int pady, int argc, char** argv)
 {
-    if (enc)
+    if (p && p->csvfpt)
     {
-        Encoder *encoder = static_cast<Encoder*>(enc);
-        int padx = encoder->m_sps.conformanceWindow.rightOffset;
-        int pady = encoder->m_sps.conformanceWindow.bottomOffset;
         const x265_api * api = x265_api_get(0);
 
-        if (!encoder->m_param->csvfpt)
-            return;
-
-        if (encoder->m_param->csvLogLevel)
+        if (p->csvLogLevel)
         {
             // adding summary to a per-frame csv log file, so it needs a summary header
-            fprintf(encoder->m_param->csvfpt, "\nSummary\n");
-            fputs(summaryCSVHeader, encoder->m_param->csvfpt);
+            fprintf(p->csvfpt, "\nSummary\n");
+            fputs(summaryCSVHeader, p->csvfpt);
+            if (p->csvLogLevel >= 2 || p->maxCLL || p->maxFALL)
+                fputs("MaxCLL, MaxFALL,", p->csvfpt);
+            fputs(" Version\n",p->csvfpt);
         }
-
         // CLI arguments or other
         if (argc)
         {
-            fputc('"', encoder->m_param->csvfpt);
+            fputc('"', p->csvfpt);
             for (int i = 1; i < argc; i++)
             {
-                fputc(' ', encoder->m_param->csvfpt);
-                fputs(argv[i], encoder->m_param->csvfpt);
+                fputc(' ', p->csvfpt);
+                fputs(argv[i], p->csvfpt);
             }
-            fputc('"', encoder->m_param->csvfpt);
+            fputc('"', p->csvfpt);
         }
         else
         {
-            const x265_param* paramTemp = encoder->m_param;
-            char *opts = x265_param2string((x265_param*)paramTemp, padx, pady);
+            char *opts = x265_param2string((x265_param*)p, padx, pady);
             if (opts)
             {
-                fputc('"', encoder->m_param->csvfpt);
-                fputs(opts, encoder->m_param->csvfpt);
-                fputc('"', encoder->m_param->csvfpt);
+                fputc('"', p->csvfpt);
+                fputs(opts, p->csvfpt);
+                fputc('"', p->csvfpt);
             }
         }
 
@@ -916,69 +917,70 @@
         timeinfo = localtime(&now);
         char buffer[200];
         strftime(buffer, 128, "%c", timeinfo);
-        fprintf(encoder->m_param->csvfpt, ", %s, ", buffer);
+        fprintf(p->csvfpt, ", %s, ", buffer);
 
         // elapsed time, fps, bitrate
-        fprintf(encoder->m_param->csvfpt, "%.2f, %.2f, %.2f,",
+        fprintf(p->csvfpt, "%.2f, %.2f, %.2f,",
             stats->elapsedEncodeTime, stats->encodedPictureCount / stats->elapsedEncodeTime, stats->bitrate);
 
-        if (encoder->m_param->bEnablePsnr)
-            fprintf(encoder->m_param->csvfpt, " %.3lf, %.3lf, %.3lf, %.3lf,",
+        if (p->bEnablePsnr)
+            fprintf(p->csvfpt, " %.3lf, %.3lf, %.3lf, %.3lf,",
             stats->globalPsnrY / stats->encodedPictureCount, stats->globalPsnrU / stats->encodedPictureCount,
             stats->globalPsnrV / stats->encodedPictureCount, stats->globalPsnr);
         else
-            fprintf(encoder->m_param->csvfpt, " -, -, -, -,");
-        if (encoder->m_param->bEnableSsim)
-            fprintf(encoder->m_param->csvfpt, " %.6f, %6.3f,", stats->globalSsim, x265_ssim2dB(stats->globalSsim));
+            fprintf(p->csvfpt, " -, -, -, -,");
+        if (p->bEnableSsim)
+            fprintf(p->csvfpt, " %.6f, %6.3f,", stats->globalSsim, x265_ssim2dB(stats->globalSsim));
         else
-            fprintf(encoder->m_param->csvfpt, " -, -,");
+            fprintf(p->csvfpt, " -, -,");
 
         if (stats->statsI.numPics)
         {
-            fprintf(encoder->m_param->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsI.numPics, stats->statsI.avgQp, stats->statsI.bitrate);
-            if (encoder->m_param->bEnablePsnr)
-                fprintf(encoder->m_param->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsI.psnrY, stats->statsI.psnrU, stats->statsI.psnrV);
+            fprintf(p->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsI.numPics, stats->statsI.avgQp, stats->statsI.bitrate);
+            if (p->bEnablePsnr)
+                fprintf(p->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsI.psnrY, stats->statsI.psnrU, stats->statsI.psnrV);
             else
-                fprintf(encoder->m_param->csvfpt, " -, -, -,");
-            if (encoder->m_param->bEnableSsim)
-                fprintf(encoder->m_param->csvfpt, " %.3lf,", stats->statsI.ssim);
+                fprintf(p->csvfpt, " -, -, -,");
+            if (p->bEnableSsim)
+                fprintf(p->csvfpt, " %.3lf,", stats->statsI.ssim);
             else
-                fprintf(encoder->m_param->csvfpt, " -,");
+                fprintf(p->csvfpt, " -,");
         }
         else
-            fprintf(encoder->m_param->csvfpt, " -, -, -, -, -, -, -,");
+            fprintf(p->csvfpt, " -, -, -, -, -, -, -,");
 
         if (stats->statsP.numPics)
         {
-            fprintf(encoder->m_param->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsP.numPics, stats->statsP.avgQp, stats->statsP.bitrate);
-            if (encoder->m_param->bEnablePsnr)
-                fprintf(encoder->m_param->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsP.psnrY, stats->statsP.psnrU, stats->statsP.psnrV);
+            fprintf(p->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsP.numPics, stats->statsP.avgQp, stats->statsP.bitrate);
+            if (p->bEnablePsnr)
+                fprintf(p->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsP.psnrY, stats->statsP.psnrU, stats->statsP.psnrV);
             else
-                fprintf(encoder->m_param->csvfpt, " -, -, -,");
-            if (encoder->m_param->bEnableSsim)
-                fprintf(encoder->m_param->csvfpt, " %.3lf,", stats->statsP.ssim);
+                fprintf(p->csvfpt, " -, -, -,");
+            if (p->bEnableSsim)
+                fprintf(p->csvfpt, " %.3lf,", stats->statsP.ssim);
             else
-                fprintf(encoder->m_param->csvfpt, " -,");
+                fprintf(p->csvfpt, " -,");
         }
         else
-            fprintf(encoder->m_param->csvfpt, " -, -, -, -, -, -, -,");
+            fprintf(p->csvfpt, " -, -, -, -, -, -, -,");
 
         if (stats->statsB.numPics)
         {
-            fprintf(encoder->m_param->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsB.numPics, stats->statsB.avgQp, stats->statsB.bitrate);
-            if (encoder->m_param->bEnablePsnr)
-                fprintf(encoder->m_param->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsB.psnrY, stats->statsB.psnrU, stats->statsB.psnrV);
+            fprintf(p->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsB.numPics, stats->statsB.avgQp, stats->statsB.bitrate);
+            if (p->bEnablePsnr)
+                fprintf(p->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsB.psnrY, stats->statsB.psnrU, stats->statsB.psnrV);
             else
-                fprintf(encoder->m_param->csvfpt, " -, -, -,");
-            if (encoder->m_param->bEnableSsim)
-                fprintf(encoder->m_param->csvfpt, " %.3lf,", stats->statsB.ssim);
+                fprintf(p->csvfpt, " -, -, -,");
+            if (p->bEnableSsim)
+                fprintf(p->csvfpt, " %.3lf,", stats->statsB.ssim);
             else
-                fprintf(encoder->m_param->csvfpt, " -,");
+                fprintf(p->csvfpt, " -,");
         }
         else
-            fprintf(encoder->m_param->csvfpt, " -, -, -, -, -, -, -,");
-
-        fprintf(encoder->m_param->csvfpt, " %-6u, %-6u, %s\n", stats->maxCLL, stats->maxFALL, api->version_str);
+            fprintf(p->csvfpt, " -, -, -, -, -, -, -,");
+        if (p->csvLogLevel >= 2 || p->maxCLL || p->maxFALL)
+            fprintf(p->csvfpt, " %-6u, %-6u,", stats->maxCLL, stats->maxFALL);
+        fprintf(p->csvfpt, " %s\n", api->version_str);
     }
 }

x265_2.6.tar.gz/source/encoder/dpb.cpp -> x265_2.7.tar.gz/source/encoder/dpb.cpp Changed

@@ -92,19 +92,14 @@
             m_freeList.pushBack(*curFrame);
             curFrame->m_encData->m_freeListNext = m_frameDataFreeList;
             m_frameDataFreeList = curFrame->m_encData;
-
-            if (curFrame->m_encData->m_meBuffer)
+            for (int i = 0; i < INTEGRAL_PLANE_NUM; i++)
             {
-                for (int i = 0; i < INTEGRAL_PLANE_NUM; i++)
+                if (curFrame->m_encData->m_meBuffer[i] != NULL)
                 {
-                    if (curFrame->m_encData->m_meBuffer[i] != NULL)
-                    {
-                        X265_FREE(curFrame->m_encData->m_meBuffer[i]);
-                        curFrame->m_encData->m_meBuffer[i] = NULL;
-                    }
+                    X265_FREE(curFrame->m_encData->m_meBuffer[i]);
+                    curFrame->m_encData->m_meBuffer[i] = NULL;
                 }
             }
-
             if (curFrame->m_ctuInfo != NULL)
             {
                 uint32_t widthInCU = (curFrame->m_param->sourceWidth + curFrame->m_param->maxCUSize - 1) >> curFrame->m_param->maxLog2CUSize;
@@ -181,7 +176,10 @@
     // Mark pictures in m_piclist as unreferenced if they are not included in RPS
     applyReferencePictureSet(&slice->m_rps, pocCurr);
 
-    slice->m_numRefIdx[0] = X265_MIN(newFrame->m_param->maxNumReferences, slice->m_rps.numberOfNegativePictures); // Ensuring L0 contains just the -ve POC
+    if (slice->m_sliceType != I_SLICE)
+        slice->m_numRefIdx[0] = x265_clip3(1, newFrame->m_param->maxNumReferences, slice->m_rps.numberOfNegativePictures);
+    else
+        slice->m_numRefIdx[0] = X265_MIN(newFrame->m_param->maxNumReferences, slice->m_rps.numberOfNegativePictures); // Ensuring L0 contains just the -ve POC
     slice->m_numRefIdx[1] = X265_MIN(newFrame->m_param->bBPyramid ? 2 : 1, slice->m_rps.numberOfPositivePictures);
     slice->setRefPicList(m_picList);
 
@@ -230,11 +228,14 @@
     {
         if ((iterPic->m_poc != curPoc) && iterPic->m_encData->m_bHasReferences)
         {
-            rps->poc[poci] = iterPic->m_poc;
-            rps->deltaPOC[poci] = rps->poc[poci] - curPoc;
-            (rps->deltaPOC[poci] < 0) ? numNeg++ : numPos++;
-            rps->bUsed[poci] = !isRAP;
-            poci++;
+            if ((m_lastIDR >= curPoc) || (m_lastIDR <= iterPic->m_poc))
+            {
+                    rps->poc[poci] = iterPic->m_poc;
+                    rps->deltaPOC[poci] = rps->poc[poci] - curPoc;
+                    (rps->deltaPOC[poci] < 0) ? numNeg++ : numPos++;
+                    rps->bUsed[poci] = !isRAP;
+                    poci++;
+            }
         }
         iterPic = iterPic->m_next;
     }

x265_2.6.tar.gz/source/encoder/encoder.cpp -> x265_2.7.tar.gz/source/encoder/encoder.cpp Changed

@@ -50,10 +50,8 @@
 
 /* Threshold for motion vection, based on expermental result.
  * TODO: come up an algorithm for adoptive threshold */
-
-#define MVTHRESHOLD 10
+#define MVTHRESHOLD (10*10)
 #define PU_2Nx2N 1
-
 static const char* defaultAnalysisFileName = "x265_analysis.dat";
 
 using namespace X265_NS;
@@ -79,7 +77,6 @@
     m_param = NULL;
     m_latestParam = NULL;
     m_threadPool = NULL;
-    m_analysisFile = NULL;
     m_analysisFileIn = NULL;
     m_analysisFileOut = NULL;
     m_offsetEmergency = NULL;
@@ -344,19 +341,29 @@
         m_aborted = true;
     if (!m_lookahead->create())
         m_aborted = true;
-
     initRefIdx();
-
-    if (m_param->analysisReuseMode)
+    if (m_param->analysisSave && m_param->bUseAnalysisFile)
     {
-        const char* name = m_param->analysisReuseFileName;
-        if (!name)
-            name = defaultAnalysisFileName;
-        const char* mode = m_param->analysisReuseMode == X265_ANALYSIS_LOAD ? "rb" : "wb";
-        m_analysisFile = x265_fopen(name, mode);
-        if (!m_analysisFile)
+        char* temp = strcatFilename(m_param->analysisSave, ".temp");
+        if (!temp)
+            m_aborted = true;
+        else
         {
-            x265_log_file(NULL, X265_LOG_ERROR, "Analysis load/save: failed to open file %s\n", name);
+            m_analysisFileOut = x265_fopen(temp, "wb");
+            X265_FREE(temp);
+        }
+        if (!m_analysisFileOut)
+        {
+            x265_log_file(NULL, X265_LOG_ERROR, "Analysis save: failed to open file %s.temp\n", m_param->analysisSave);
+            m_aborted = true;
+        }
+    }
+    if (m_param->analysisLoad && m_param->bUseAnalysisFile)
+    {
+        m_analysisFileIn = x265_fopen(m_param->analysisLoad, "rb");
+        if (!m_analysisFileIn)
+        {
+            x265_log_file(NULL, X265_LOG_ERROR, "Analysis load: failed to open file %s\n", m_param->analysisLoad);
             m_aborted = true;
         }
     }
@@ -450,7 +457,7 @@
     return 0;
 }
 
-int Encoder::getRefFrameList(PicYuv** l0, PicYuv** l1, int sliceType, int poc)
+int Encoder::getRefFrameList(PicYuv** l0, PicYuv** l1, int sliceType, int poc, int* pocL0, int* pocL1)
 {
     if (!(IS_X265_TYPE_I(sliceType)))
     {
@@ -462,9 +469,10 @@
                 if (framePtr->m_encData->m_slice->m_refFrameList[0][j] && framePtr->m_encData->m_slice->m_refFrameList[0][j]->m_reconPic != NULL)
                 {
                     int l0POC = framePtr->m_encData->m_slice->m_refFrameList[0][j]->m_poc;
+                    pocL0[j] = l0POC;
                     Frame* l0Fp = m_dpb->m_picList.getPOC(l0POC);
-                    if (l0Fp->m_reconPic->m_picOrg[0] == NULL)
-                        l0Fp->m_reconEncoded.wait(); /* If recon is not ready, current frame encoder need to wait. */
+                    while (l0Fp->m_reconRowFlag[l0Fp->m_numRows - 1].get() == 0)
+                        l0Fp->m_reconRowFlag[l0Fp->m_numRows - 1].waitForChange(0); /* If recon is not ready, current frame encoder has to wait. */
                     l0[j] = l0Fp->m_reconPic;
                 }
             }
@@ -473,15 +481,19 @@
                 if (framePtr->m_encData->m_slice->m_refFrameList[1][j] && framePtr->m_encData->m_slice->m_refFrameList[1][j]->m_reconPic != NULL)
                 {
                     int l1POC = framePtr->m_encData->m_slice->m_refFrameList[1][j]->m_poc;
+                    pocL1[j] = l1POC;
                     Frame* l1Fp = m_dpb->m_picList.getPOC(l1POC);
-                    if (l1Fp->m_reconPic->m_picOrg[0] == NULL)
-                        l1Fp->m_reconEncoded.wait(); /* If recon is not ready, current frame encoder need to wait. */
+                    while (l1Fp->m_reconRowFlag[l1Fp->m_numRows - 1].get() == 0)
+                        l1Fp->m_reconRowFlag[l1Fp->m_numRows - 1].waitForChange(0); /* If recon is not ready, current frame encoder has to wait. */
                     l1[j] = l1Fp->m_reconPic;
                 }
             }
         }
         else
-            x265_log(NULL, X265_LOG_WARNING, "Refrence List is not in piclist\n");
+        {
+            x265_log(NULL, X265_LOG_WARNING, "Current frame is not in DPB piclist.\n");
+            return 1;
+        }
     }
     else
     {
@@ -564,19 +576,19 @@
                     {
                         int cuOffset = cuI * bytes + pu;
                         (interData)->mergeFlag[cuPos + cuOffset] = (srcInterData)->mergeFlag[(mbIndex * 16) + cuOffset];
-
+                        (interData)->sadCost[cuPos + cuOffset] = (srcInterData)->sadCost[(mbIndex * 16) + cuOffset];
                         (interData)->interDir[cuPos + cuOffset] = (srcInterData)->interDir[(mbIndex * 16) + cuOffset];
                         for (uint32_t k = 0; k < numDir; k++)
                         {
                             (interData)->mvpIdx[k][cuPos + cuOffset] = (srcInterData)->mvpIdx[k][(mbIndex * 16) + cuOffset];
                             (interData)->refIdx[k][cuPos + cuOffset] = (srcInterData)->refIdx[k][(mbIndex * 16) + cuOffset];
                             memcpy(&(interData)->mv[k][cuPos + cuOffset], &(srcInterData)->mv[k][(mbIndex * 16) + cuOffset], sizeof(MV));
-                            if (m_param->analysisReuseLevel == 7)
+                            if (m_param->analysisReuseLevel == 7 && numPU == PU_2Nx2N &&
+                                ((interData)->depth[cuPos + cuOffset] == (m_param->maxCUSize >> 5)))
                             {
-                                int mv_x = ((analysis_inter_data *)curFrame->m_analysisData.interData)->mv[k][(mbIndex * 16) + cuOffset].x;
-                                int mv_y = ((analysis_inter_data *)curFrame->m_analysisData.interData)->mv[k][(mbIndex * 16) + cuOffset].y;
-                                double mv = sqrt(mv_x*mv_x + mv_y*mv_y);
-                                if (numPU == PU_2Nx2N && ((srcInterData)->depth[cuPos + cuOffset] == (m_param->maxCUSize >> 5)) && mv <= MVTHRESHOLD)
+                                int mv_x = (interData)->mv[k][cuPos + cuOffset].x;
+                                int mv_y = (interData)->mv[k][cuPos + cuOffset].y;
+                                if ((mv_x*mv_x + mv_y*mv_y) <= MVTHRESHOLD)
                                     memset(&curFrame->m_analysisData.modeFlag[k][cuPos + cuOffset], 1, bytes);
                             }
                         }
@@ -642,9 +654,10 @@
                     if (m_param->analysisReuseLevel > 4)
                     {
                         memset(&(currInterData)->partSize[count], (interData)->partSize[d], bytes);
-                        int numPU = nbPartsTable[(currInterData)->partSize[d]];
-                        for (int pu = 0; pu < numPU; pu++, d++)
+                        int numPU = nbPartsTable[(interData)->partSize[d]];
+                        for (int pu = 0; pu < numPU; pu++)
                         {
+                            if (pu) d++;
                             (currInterData)->mergeFlag[count + pu] = (interData)->mergeFlag[d];
                             if (m_param->analysisReuseLevel >= 7)
                             {
@@ -654,12 +667,11 @@
                                     (currInterData)->mvpIdx[i][count + pu] = (interData)->mvpIdx[i][d];
                                     (currInterData)->refIdx[i][count + pu] = (interData)->refIdx[i][d];
                                     memcpy(&(currInterData)->mv[i][count + pu], &(interData)->mv[i][d], sizeof(MV));
-                                    if (m_param->analysisReuseLevel == 7)
+                                    if (m_param->analysisReuseLevel == 7 && numPU == PU_2Nx2N && m_param->num4x4Partitions <= 16)
                                     {
-                                        int mv_x = ((analysis_inter_data *)curFrame->m_analysisData.interData)->mv[i][count + pu].x;
-                                        int mv_y = ((analysis_inter_data *)curFrame->m_analysisData.interData)->mv[i][count + pu].y;
-                                        double mv = sqrt(mv_x*mv_x + mv_y*mv_y);
-                                        if (numPU == PU_2Nx2N && m_param->num4x4Partitions <= 16 && mv <= MVTHRESHOLD)
+                                        int mv_x = (currInterData)->mv[i][count + pu].x;
+                                        int mv_y = (currInterData)->mv[i][count + pu].y;
+                                        if ((mv_x*mv_x + mv_y*mv_y) <= MVTHRESHOLD)
                                             memset(&curFrame->m_analysisData.modeFlag[i][count + pu], 1, bytes);
                                     }
                                 }
@@ -720,9 +732,6 @@
 
     X265_FREE(m_offsetEmergency);
 
-    if (m_analysisFile)
-        fclose(m_analysisFile);
-
     if (m_latestParam != NULL && m_latestParam != m_param)
     {
         if (m_latestParam->scalingLists != m_param->scalingLists)
@@ -737,7 +746,7 @@
     {
         int bError = 1;
         fclose(m_analysisFileOut);
-        const char* name = m_param->analysisReuseFileName;
+        const char* name = m_param->analysisSave ? m_param->analysisSave : m_param->analysisReuseFileName;
         if (!name)
             name = defaultAnalysisFileName;
         char* temp = strcatFilename(name, ".temp");
@@ -765,6 +774,8 @@
         free((char*)m_param->numaPools);
         free((char*)m_param->masteringDisplayColorVolume);
         free((char*)m_param->toneMapFile);
+        free((char*)m_param->analysisSave);
+        free((char*)m_param->analysisLoad);
         PARAM_NS::x265_param_free(m_param);
     }
 }
@@ -851,7 +862,7 @@
 
     if (m_exportedPic)
     {
-        if (!m_param->bUseAnalysisFile && m_param->analysisReuseMode == X265_ANALYSIS_SAVE)
+        if (!m_param->bUseAnalysisFile && m_param->analysisSave)
             freeAnalysis(&m_exportedPic->m_analysisData);
         ATOMIC_DEC(&m_exportedPic->m_countRefEncoders);
         m_exportedPic = NULL;
@@ -1036,7 +1047,7 @@
 
         /* In analysisSave mode, x265_analysis_data is allocated in pic_in and inFrame points to this */
         /* Load analysis data before lookahead->addPicture, since sliceType has been decided */
-        if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD)
+        if (m_param->analysisLoad)
         {
             /* readAnalysisFile reads analysis data for the frame and allocates memory based on slicetype */
             readAnalysisFile(&inFrame->m_analysisData, inFrame->m_poc, pic_in);
@@ -1049,11 +1060,14 @@
                 inFrame->m_lowres.sliceType = sliceType;
                 inFrame->m_lowres.bKeyframe = !!inFrame->m_analysisData.lookahead.keyframe;
                 inFrame->m_lowres.bLastMiniGopBFrame = !!inFrame->m_analysisData.lookahead.lastMiniGopBFrame;
-                int vbvCount = m_param->lookaheadDepth + m_param->bframes + 2;
-                for (int index = 0; index < vbvCount; index++)
+                if (m_rateControl->m_isVbv)
                 {
-                    inFrame->m_lowres.plannedSatd[index] = inFrame->m_analysisData.lookahead.plannedSatd[index];
-                    inFrame->m_lowres.plannedType[index] = inFrame->m_analysisData.lookahead.plannedType[index];
+                    int vbvCount = m_param->lookaheadDepth + m_param->bframes + 2;
+                    for (int index = 0; index < vbvCount; index++)
+                    {
+                        inFrame->m_lowres.plannedSatd[index] = inFrame->m_analysisData.lookahead.plannedSatd[index];
+                        inFrame->m_lowres.plannedType[index] = inFrame->m_analysisData.lookahead.plannedType[index];
+                    }
                 }
             }
         }
@@ -1118,7 +1132,7 @@
             x265_frame_stats* frameData = NULL;
 
             /* Free up pic_in->analysisData since it has already been used */
-            if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD || (m_param->bMVType && slice->m_sliceType != I_SLICE))
+            if ((m_param->analysisLoad && !m_param->analysisSave) || (m_param->bMVType && slice->m_sliceType != I_SLICE))
                 freeAnalysis(&outFrame->m_analysisData);
 
             if (pic_out)
@@ -1144,7 +1158,7 @@
                 }
 
                 /* Dump analysis data from pic_out to file in save mode and free */
-                if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE)
+                if (m_param->analysisSave)
                 {
                     pic_out->analysisData.poc = pic_out->poc;
                     pic_out->analysisData.sliceType = pic_out->sliceType;
@@ -1167,26 +1181,29 @@
                         pic_out->analysisData.satdCost *= factor;
                         pic_out->analysisData.lookahead.keyframe = outFrame->m_lowres.bKeyframe;
                         pic_out->analysisData.lookahead.lastMiniGopBFrame = outFrame->m_lowres.bLastMiniGopBFrame;
-                        int vbvCount = m_param->lookaheadDepth + m_param->bframes + 2;
-                        for (int index = 0; index < vbvCount; index++)
-                        {
-                            pic_out->analysisData.lookahead.plannedSatd[index] = outFrame->m_lowres.plannedSatd[index] * factor;
-                            pic_out->analysisData.lookahead.plannedType[index] = outFrame->m_lowres.plannedType[index];
-                        }
-                        for (uint32_t index = 0; index < pic_out->analysisData.numCuInHeight; index++)
+                        if (m_rateControl->m_isVbv)
                         {
-                            outFrame->m_analysisData.lookahead.intraSatdForVbv[index] = outFrame->m_encData->m_rowStat[index].intraSatdForVbv * factor;
-                            outFrame->m_analysisData.lookahead.satdForVbv[index] = outFrame->m_encData->m_rowStat[index].satdForVbv * factor;
-                        }
-                        pic_out->analysisData.lookahead.intraSatdForVbv = outFrame->m_analysisData.lookahead.intraSatdForVbv;
-                        pic_out->analysisData.lookahead.satdForVbv = outFrame->m_analysisData.lookahead.satdForVbv;
-                        for (uint32_t index = 0; index < pic_out->analysisData.numCUsInFrame; index++)
-                        {
-                            outFrame->m_analysisData.lookahead.intraVbvCost[index] = outFrame->m_encData->m_cuStat[index].intraVbvCost * factor;
-                            outFrame->m_analysisData.lookahead.vbvCost[index] = outFrame->m_encData->m_cuStat[index].vbvCost * factor;
+                            int vbvCount = m_param->lookaheadDepth + m_param->bframes + 2;
+                            for (int index = 0; index < vbvCount; index++)
+                            {
+                                pic_out->analysisData.lookahead.plannedSatd[index] = outFrame->m_lowres.plannedSatd[index] * factor;
+                                pic_out->analysisData.lookahead.plannedType[index] = outFrame->m_lowres.plannedType[index];
+                            }
+                            for (uint32_t index = 0; index < pic_out->analysisData.numCuInHeight; index++)
+                            {
+                                outFrame->m_analysisData.lookahead.intraSatdForVbv[index] = outFrame->m_encData->m_rowStat[index].intraSatdForVbv * factor;
+                                outFrame->m_analysisData.lookahead.satdForVbv[index] = outFrame->m_encData->m_rowStat[index].satdForVbv * factor;
+                            }
+                            pic_out->analysisData.lookahead.intraSatdForVbv = outFrame->m_analysisData.lookahead.intraSatdForVbv;
+                            pic_out->analysisData.lookahead.satdForVbv = outFrame->m_analysisData.lookahead.satdForVbv;
+                            for (uint32_t index = 0; index < pic_out->analysisData.numCUsInFrame; index++)
+                            {
+                                outFrame->m_analysisData.lookahead.intraVbvCost[index] = outFrame->m_encData->m_cuStat[index].intraVbvCost * factor;
+                                outFrame->m_analysisData.lookahead.vbvCost[index] = outFrame->m_encData->m_cuStat[index].vbvCost * factor;
+                            }
+                            pic_out->analysisData.lookahead.intraVbvCost = outFrame->m_analysisData.lookahead.intraVbvCost;
+                            pic_out->analysisData.lookahead.vbvCost = outFrame->m_analysisData.lookahead.vbvCost;
                         }
-                        pic_out->analysisData.lookahead.intraVbvCost = outFrame->m_analysisData.lookahead.intraVbvCost;
-                        pic_out->analysisData.lookahead.vbvCost = outFrame->m_analysisData.lookahead.vbvCost;
                     }
                     writeAnalysisFile(&pic_out->analysisData, *outFrame->m_encData);
                     if (m_param->bUseAnalysisFile)
@@ -1350,18 +1367,21 @@
                 slice->m_maxNumMergeCand = m_param->maxNumMergeCand;
                 slice->m_endCUAddr = slice->realEndAddress(m_sps.numCUsInFrame * m_param->num4x4Partitions);
             }
-            if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->bDisableLookahead)
+            if (m_param->analysisLoad && m_param->bDisableLookahead)
             {
                 frameEnc->m_dts = frameEnc->m_analysisData.lookahead.dts;
-                for (uint32_t index = 0; index < frameEnc->m_analysisData.numCuInHeight; index++)
-                {
-                    frameEnc->m_encData->m_rowStat[index].intraSatdForVbv = frameEnc->m_analysisData.lookahead.intraSatdForVbv[index];
-                    frameEnc->m_encData->m_rowStat[index].satdForVbv = frameEnc->m_analysisData.lookahead.satdForVbv[index];
-                }
-                for (uint32_t index = 0; index < frameEnc->m_analysisData.numCUsInFrame; index++)
+                if (m_rateControl->m_isVbv)
                 {
-                    frameEnc->m_encData->m_cuStat[index].intraVbvCost = frameEnc->m_analysisData.lookahead.intraVbvCost[index];
-                    frameEnc->m_encData->m_cuStat[index].vbvCost = frameEnc->m_analysisData.lookahead.vbvCost[index];
+                    for (uint32_t index = 0; index < frameEnc->m_analysisData.numCuInHeight; index++)
+                    {
+                        frameEnc->m_encData->m_rowStat[index].intraSatdForVbv = frameEnc->m_analysisData.lookahead.intraSatdForVbv[index];
+                        frameEnc->m_encData->m_rowStat[index].satdForVbv = frameEnc->m_analysisData.lookahead.satdForVbv[index];
+                    }
+                    for (uint32_t index = 0; index < frameEnc->m_analysisData.numCUsInFrame; index++)
+                    {
+                        frameEnc->m_encData->m_cuStat[index].intraVbvCost = frameEnc->m_analysisData.lookahead.intraVbvCost[index];
+                        frameEnc->m_encData->m_cuStat[index].vbvCost = frameEnc->m_analysisData.lookahead.vbvCost[index];
+                    }
                 }
             }
             if (m_param->searchMethod == X265_SEA && frameEnc->m_lowres.sliceType != X265_TYPE_B)
@@ -1416,7 +1436,7 @@
             frameEnc->m_encData->m_slice->m_iNumRPSInSPS = m_sps.spsrpsNum;
 
             curEncoder->m_rce.encodeOrder = frameEnc->m_encodeOrder = m_encodedFrameNum++;
-            if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || !m_param->bDisableLookahead)
+            if (!m_param->analysisLoad || !m_param->bDisableLookahead)
             {
                 if (m_bframeDelay)
                 {
@@ -1431,7 +1451,7 @@
             }
 
             /* Allocate analysis data before encode in save mode. This is allocated in frameEnc */
-            if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE)
+            if (m_param->analysisSave && !m_param->analysisLoad)
             {
                 x265_analysis_data* analysis = &frameEnc->m_analysisData;
                 analysis->poc = frameEnc->m_poc;
@@ -1963,11 +1983,12 @@
         stats->statsB.psnrU   = m_analyzeB.m_psnrSumU / (double)m_analyzeB.m_numPics;
         stats->statsB.psnrV   = m_analyzeB.m_psnrSumV / (double)m_analyzeB.m_numPics;
         stats->statsB.ssim    = x265_ssim2dB(m_analyzeB.m_globalSsim / (double)m_analyzeB.m_numPics);
-
-        stats->maxCLL         = m_analyzeAll.m_maxCLL;
-        stats->maxFALL        = (uint16_t)(m_analyzeAll.m_maxFALL / m_analyzeAll.m_numPics);
+        if (m_param->csvLogLevel >= 2 || m_param->maxCLL || m_param->maxFALL)
+        {
+            stats->maxCLL = m_analyzeAll.m_maxCLL;
+            stats->maxFALL = (uint16_t)(m_analyzeAll.m_maxFALL / m_analyzeAll.m_numPics);
+        }
     }
-
     /* If new statistics are added to x265_stats, we must check here whether the
      * structure provided by the user is the new structure or an older one (for
      * future safety) */
@@ -2039,10 +2060,11 @@
         if (m_param->bEnableSsim)
             m_analyzeB.addSsim(ssim);
     }
-
-    m_analyzeAll.m_maxFALL += curFrame->m_fencPic->m_avgLumaLevel;
-    m_analyzeAll.m_maxCLL = X265_MAX(m_analyzeAll.m_maxCLL, curFrame->m_fencPic->m_maxLumaLevel);
-
+    if (m_param->csvLogLevel >= 2 || m_param->maxCLL || m_param->maxFALL)
+    {
+        m_analyzeAll.m_maxFALL += curFrame->m_fencPic->m_avgLumaLevel;
+        m_analyzeAll.m_maxCLL = X265_MAX(m_analyzeAll.m_maxCLL, curFrame->m_fencPic->m_maxLumaLevel);
+    }
     char c = (slice->isIntra() ? (curFrame->m_lowres.sliceType == X265_TYPE_IDR ? 'I' : 'i') : slice->isInterP() ? 'P' : 'B');
     int poc = slice->m_poc;
     if (!IS_REFERENCED(curFrame))
@@ -2081,13 +2103,7 @@
                     frameStats->list1POC[ref] = ref < slice->m_numRefIdx[1] ? slice->m_refPOCList[1][ref] - slice->m_lastIDR : -1;
             }
         }
-
 #define ELAPSED_MSEC(start, end) (((double)(end) - (start)) / 1000)
-
-        frameStats->maxLumaLevel = curFrame->m_fencPic->m_maxLumaLevel;
-        frameStats->minLumaLevel = curFrame->m_fencPic->m_minLumaLevel;
-        frameStats->avgLumaLevel = curFrame->m_fencPic->m_avgLumaLevel;
-
         if (m_param->csvLogLevel >= 2)
         {
             frameStats->decideWaitTime = ELAPSED_MSEC(0, curEncoder->m_slicetypeWaitTime);
@@ -2107,6 +2123,9 @@
             frameStats->avgLumaDistortion = curFrame->m_encData->m_frameStats.avgLumaDistortion;
             frameStats->avgPsyEnergy = curFrame->m_encData->m_frameStats.avgPsyEnergy;
             frameStats->avgResEnergy = curFrame->m_encData->m_frameStats.avgResEnergy;
+            frameStats->maxLumaLevel = curFrame->m_fencPic->m_maxLumaLevel;
+            frameStats->minLumaLevel = curFrame->m_fencPic->m_minLumaLevel;
+            frameStats->avgLumaLevel = curFrame->m_fencPic->m_avgLumaLevel;
 
             frameStats->maxChromaULevel = curFrame->m_fencPic->m_maxChromaULevel;
             frameStats->minChromaULevel = curFrame->m_fencPic->m_minChromaULevel;
@@ -2285,7 +2304,7 @@
             if (buffer)
             {
                 sprintf(buffer, "x265 (build %d) - %s:%s - H.265/HEVC codec - "
-                        "Copyright 2013-2017 (c) Multicoreware, Inc - "
+                        "Copyright 2013-2018 (c) Multicoreware, Inc - "
                         "http://x265.org - options: %s",
                         X265_BUILD, PFX(version_str), PFX(build_info_str), opts);
                 
@@ -2451,6 +2470,18 @@
         this->m_externalFlush = true;
     else 
         this->m_externalFlush = false;
+
+    if (p->bMVType == AVC_INFO && (p->limitTU == 3 || p->limitTU == 4))
+    {
+        x265_log(p, X265_LOG_WARNING, "limit TU = 3 or 4 with MVType AVCINFO produces inconsistent output\n");
+    }
+
+    if (p->bMVType == AVC_INFO && p->minCUSize != 8)
+    {
+        p->minCUSize = 8;
+        x265_log(p, X265_LOG_WARNING, "Setting minCuSize = 8, AVCINFO expects 8x8 blocks\n");
+    }
+
     if (p->keyframeMax < 0)
     {
         /* A negative max GOP size indicates the user wants only one I frame at
@@ -2604,23 +2635,24 @@
         p->rc.rfConstantMin = 0;
     }
 
-    if (p->analysisReuseMode && (p->bDistributeModeAnalysis || p->bDistributeMotionEstimation))
+    if ((p->analysisLoad || p->analysisSave) && (p->bDistributeModeAnalysis || p->bDistributeMotionEstimation))
     {
         x265_log(p, X265_LOG_WARNING, "Analysis load/save options incompatible with pmode/pme, Disabling pmode/pme\n");
         p->bDistributeMotionEstimation = p->bDistributeModeAnalysis = 0;
     }
 
-    if (p->analysisReuseMode && p->rc.cuTree)
+    if ((p->analysisLoad || p->analysisSave) && p->rc.cuTree)
     {
         x265_log(p, X265_LOG_WARNING, "Analysis load/save options works only with cu-tree off, Disabling cu-tree\n");
         p->rc.cuTree = 0;
     }
 
-    if (p->analysisReuseMode && (p->analysisMultiPassRefine || p->analysisMultiPassDistortion))
+    if ((p->analysisLoad || p->analysisSave) && (p->analysisMultiPassRefine || p->analysisMultiPassDistortion))
     {
         x265_log(p, X265_LOG_WARNING, "Cannot use Analysis load/save option and multi-pass-opt-analysis/multi-pass-opt-distortion together,"
             "Disabling Analysis load/save and multi-pass-opt-analysis/multi-pass-opt-distortion\n");
-        p->analysisReuseMode = p->analysisMultiPassRefine = p->analysisMultiPassDistortion = 0;
+        p->analysisSave = p->analysisLoad = NULL;
+        p->analysisMultiPassRefine = p->analysisMultiPassDistortion = 0;
     }
     if (p->scaleFactor)
     {
@@ -2628,16 +2660,16 @@
         {
             p->scaleFactor = 0;
         }
-        else if (!p->analysisReuseMode || p->analysisReuseLevel < 10)
+        else if ((!p->analysisLoad && !p->analysisSave) || p->analysisReuseLevel < 10)
         {
-            x265_log(p, X265_LOG_WARNING, "Input scaling works with analysis-reuse-mode, analysis-reuse-level 10. Disabling scale-factor.\n");
+            x265_log(p, X265_LOG_WARNING, "Input scaling works with analysis load/save, analysis-reuse-level 10. Disabling scale-factor.\n");
             p->scaleFactor = 0;
         }
     }
 
     if (p->intraRefine)
     {
-        if (p->analysisReuseMode!= X265_ANALYSIS_LOAD || p->analysisReuseLevel < 10 || !p->scaleFactor)
+        if (!p->analysisLoad || p->analysisReuseLevel < 10 || !p->scaleFactor)
         {
             x265_log(p, X265_LOG_WARNING, "Intra refinement requires analysis load, analysis-reuse-level 10, scale factor. Disabling intra refine.\n");
             p->intraRefine = 0;
@@ -2646,7 +2678,7 @@
 
     if (p->interRefine)
     {
-        if (p->analysisReuseMode != X265_ANALYSIS_LOAD || p->analysisReuseLevel < 10 || !p->scaleFactor)
+        if (!p->analysisLoad || p->analysisReuseLevel < 10 || !p->scaleFactor)
         {
             x265_log(p, X265_LOG_WARNING, "Inter refinement requires analysis load, analysis-reuse-level 10, scale factor. Disabling inter refine.\n");
             p->interRefine = 0;
@@ -2661,7 +2693,7 @@
 
     if (p->mvRefine)
     {
-        if (p->analysisReuseMode != X265_ANALYSIS_LOAD || p->analysisReuseLevel < 10 || !p->scaleFactor)
+        if (!p->analysisLoad || p->analysisReuseLevel < 10 || !p->scaleFactor)
         {
             x265_log(p, X265_LOG_WARNING, "MV refinement requires analysis load, analysis-reuse-level 10, scale factor. Disabling MV refine.\n");
             p->mvRefine = 0;
@@ -2762,7 +2794,7 @@
     m_conformanceWindow.bottomOffset = 0;
     m_conformanceWindow.leftOffset = 0;
     /* set pad size if width is not multiple of the minimum CU size */
-    if (p->scaleFactor == 2 && ((p->sourceWidth / 2) & (p->minCUSize - 1)) && p->analysisReuseMode == X265_ANALYSIS_LOAD)
+    if (p->scaleFactor == 2 && ((p->sourceWidth / 2) & (p->minCUSize - 1)) && p->analysisLoad)
     {
         uint32_t rem = (p->sourceWidth / 2) & (p->minCUSize - 1);
         uint32_t padsize = p->minCUSize - rem;
@@ -2951,7 +2983,7 @@
         }
     }
     /* set pad size if height is not multiple of the minimum CU size */
-    if (p->scaleFactor == 2 && ((p->sourceHeight / 2) & (p->minCUSize - 1)) && p->analysisReuseMode == X265_ANALYSIS_LOAD)
+    if (p->scaleFactor == 2 && ((p->sourceHeight / 2) & (p->minCUSize - 1)) && p->analysisLoad)
     {
         uint32_t rem = (p->sourceHeight / 2) & (p->minCUSize - 1);
         uint32_t padsize = p->minCUSize - rem;
@@ -3016,13 +3048,19 @@
     p->maxCUDepth    = p->maxLog2CUSize - g_log2Size[p->minCUSize];
     p->unitSizeDepth = p->maxLog2CUSize - LOG2_UNIT_SIZE;
     p->num4x4Partitions = (1U << (p->unitSizeDepth << 1));
+
+    if (p->radl && (p->keyframeMax != p->keyframeMin))
+    {
+        p->radl = 0;
+        x265_log(p, X265_LOG_WARNING, "Radl requires fixed gop-length (keyint == min-keyint). Disabling radl.\n");
+    }
 }
 
 void Encoder::allocAnalysis(x265_analysis_data* analysis)
 {
     X265_CHECK(analysis->sliceType, "invalid slice type\n");
     analysis->interData = analysis->intraData = NULL;
-    if (m_param->bDisableLookahead)
+    if (m_param->bDisableLookahead && m_rateControl->m_isVbv)
     {
         CHECKED_MALLOC_ZERO(analysis->lookahead.intraSatdForVbv, uint32_t, analysis->numCuInHeight);
         CHECKED_MALLOC_ZERO(analysis->lookahead.satdForVbv, uint32_t, analysis->numCuInHeight);
@@ -3064,14 +3102,14 @@
         if (m_param->analysisReuseLevel >= 7)
         {
             CHECKED_MALLOC(interData->interDir, uint8_t, analysis->numPartitions * analysis->numCUsInFrame);
+            CHECKED_MALLOC(interData->sadCost, int64_t, analysis->numPartitions * analysis->numCUsInFrame);
             for (int dir = 0; dir < numDir; dir++)
             {
                 CHECKED_MALLOC(interData->mvpIdx[dir], uint8_t, analysis->numPartitions * analysis->numCUsInFrame);
                 CHECKED_MALLOC(interData->refIdx[dir], int8_t, analysis->numPartitions * analysis->numCUsInFrame);
                 CHECKED_MALLOC(interData->mv[dir], MV, analysis->numPartitions * analysis->numCUsInFrame);
-                CHECKED_MALLOC(analysis->modeFlag[dir], uint8_t, analysis->numPartitions * analysis->numCUsInFrame);
+                CHECKED_MALLOC_ZERO(analysis->modeFlag[dir], uint8_t, analysis->numPartitions * analysis->numCUsInFrame);
             }
-
             /* Allocate intra in inter */
             if (analysis->sliceType == X265_TYPE_P || m_param->bIntraInBFrames)
             {
@@ -3093,10 +3131,9 @@
     freeAnalysis(analysis);
     m_aborted = true;
 }
-
 void Encoder::freeAnalysis(x265_analysis_data* analysis)
 {
-    if (m_param->bDisableLookahead)
+    if (m_param->bDisableLookahead && m_rateControl->m_isVbv)
     {
         X265_FREE(analysis->lookahead.satdForVbv);
         X265_FREE(analysis->lookahead.intraSatdForVbv);
@@ -3249,31 +3286,31 @@
     static uint64_t consumedBytes = 0;
     static uint64_t totalConsumedBytes = 0;
     uint32_t depthBytes = 0;
-    fseeko(m_analysisFile, totalConsumedBytes, SEEK_SET);
-
+    if (m_param->bUseAnalysisFile)
+        fseeko(m_analysisFileIn, totalConsumedBytes, SEEK_SET);
     const x265_analysis_data *picData = &(picIn->analysisData);
     analysis_intra_data *intraPic = (analysis_intra_data *)picData->intraData;
     analysis_inter_data *interPic = (analysis_inter_data *)picData->interData;
 
     int poc; uint32_t frameRecordSize;
-    X265_FREAD(&frameRecordSize, sizeof(uint32_t), 1, m_analysisFile, &(picData->frameRecordSize));
-    X265_FREAD(&depthBytes, sizeof(uint32_t), 1, m_analysisFile, &(picData->depthBytes));
-    X265_FREAD(&poc, sizeof(int), 1, m_analysisFile, &(picData->poc));
+    X265_FREAD(&frameRecordSize, sizeof(uint32_t), 1, m_analysisFileIn, &(picData->frameRecordSize));
+    X265_FREAD(&depthBytes, sizeof(uint32_t), 1, m_analysisFileIn, &(picData->depthBytes));
+    X265_FREAD(&poc, sizeof(int), 1, m_analysisFileIn, &(picData->poc));
 
     if (m_param->bUseAnalysisFile)
     {
         uint64_t currentOffset = totalConsumedBytes;
 
         /* Seeking to the right frame Record */
-        while (poc != curPoc && !feof(m_analysisFile))
+        while (poc != curPoc && !feof(m_analysisFileIn))
         {
             currentOffset += frameRecordSize;
-            fseeko(m_analysisFile, currentOffset, SEEK_SET);
-            X265_FREAD(&frameRecordSize, sizeof(uint32_t), 1, m_analysisFile, &(picData->frameRecordSize));
-            X265_FREAD(&depthBytes, sizeof(uint32_t), 1, m_analysisFile, &(picData->depthBytes));
-            X265_FREAD(&poc, sizeof(int), 1, m_analysisFile, &(picData->poc));
+            fseeko(m_analysisFileIn, currentOffset, SEEK_SET);
+            X265_FREAD(&frameRecordSize, sizeof(uint32_t), 1, m_analysisFileIn, &(picData->frameRecordSize));
+            X265_FREAD(&depthBytes, sizeof(uint32_t), 1, m_analysisFileIn, &(picData->depthBytes));
+            X265_FREAD(&poc, sizeof(int), 1, m_analysisFileIn, &(picData->poc));
         }
-        if (poc != curPoc || feof(m_analysisFile))
+        if (poc != curPoc || feof(m_analysisFileIn))
         {
             x265_log(NULL, X265_LOG_WARNING, "Error reading analysis data: Cannot find POC %d\n", curPoc);
             freeAnalysis(analysis);
@@ -3284,30 +3321,29 @@
     /* Now arrived at the right frame, read the record */
     analysis->poc = poc;
     analysis->frameRecordSize = frameRecordSize;
-    X265_FREAD(&analysis->sliceType, sizeof(int), 1, m_analysisFile, &(picData->sliceType));
-    X265_FREAD(&analysis->bScenecut, sizeof(int), 1, m_analysisFile, &(picData->bScenecut));
-    X265_FREAD(&analysis->satdCost, sizeof(int64_t), 1, m_analysisFile, &(picData->satdCost));
-    X265_FREAD(&analysis->numCUsInFrame, sizeof(int), 1, m_analysisFile, &(picData->numCUsInFrame));
-    X265_FREAD(&analysis->numPartitions, sizeof(int), 1, m_analysisFile, &(picData->numPartitions));
+    X265_FREAD(&analysis->sliceType, sizeof(int), 1, m_analysisFileIn, &(picData->sliceType));
+    X265_FREAD(&analysis->bScenecut, sizeof(int), 1, m_analysisFileIn, &(picData->bScenecut));
+    X265_FREAD(&analysis->satdCost, sizeof(int64_t), 1, m_analysisFileIn, &(picData->satdCost));
+    X265_FREAD(&analysis->numCUsInFrame, sizeof(int), 1, m_analysisFileIn, &(picData->numCUsInFrame));
+    X265_FREAD(&analysis->numPartitions, sizeof(int), 1, m_analysisFileIn, &(picData->numPartitions));
     if (m_param->bDisableLookahead)
     {
-        X265_FREAD(&analysis->numCuInHeight, sizeof(uint32_t), 1, m_analysisFile, &(picData->numCuInHeight));
-        X265_FREAD(&analysis->lookahead, sizeof(x265_lookahead_data), 1, m_analysisFile, &(picData->lookahead));
+        X265_FREAD(&analysis->numCuInHeight, sizeof(uint32_t), 1, m_analysisFileIn, &(picData->numCuInHeight));
+        X265_FREAD(&analysis->lookahead, sizeof(x265_lookahead_data), 1, m_analysisFileIn, &(picData->lookahead));
     }
     int scaledNumPartition = analysis->numPartitions;
     int factor = 1 << m_param->scaleFactor;
 
     if (m_param->scaleFactor)
         analysis->numPartitions *= factor;
-
     /* Memory is allocated for inter and intra analysis data based on the slicetype */
     allocAnalysis(analysis);
-    if (m_param->bDisableLookahead)
+    if (m_param->bDisableLookahead && m_rateControl->m_isVbv)
     {
-        X265_FREAD(analysis->lookahead.intraVbvCost, sizeof(uint32_t), analysis->numCUsInFrame, m_analysisFile, picData->lookahead.intraVbvCost);
-        X265_FREAD(analysis->lookahead.vbvCost, sizeof(uint32_t), analysis->numCUsInFrame, m_analysisFile, picData->lookahead.vbvCost);
-        X265_FREAD(analysis->lookahead.satdForVbv, sizeof(uint32_t), analysis->numCuInHeight, m_analysisFile, picData->lookahead.satdForVbv);
-        X265_FREAD(analysis->lookahead.intraSatdForVbv, sizeof(uint32_t), analysis->numCuInHeight, m_analysisFile, picData->lookahead.intraSatdForVbv);
+        X265_FREAD(analysis->lookahead.intraVbvCost, sizeof(uint32_t), analysis->numCUsInFrame, m_analysisFileIn, picData->lookahead.intraVbvCost);
+        X265_FREAD(analysis->lookahead.vbvCost, sizeof(uint32_t), analysis->numCUsInFrame, m_analysisFileIn, picData->lookahead.vbvCost);
+        X265_FREAD(analysis->lookahead.satdForVbv, sizeof(uint32_t), analysis->numCuInHeight, m_analysisFileIn, picData->lookahead.satdForVbv);
+        X265_FREAD(analysis->lookahead.intraSatdForVbv, sizeof(uint32_t), analysis->numCuInHeight, m_analysisFileIn, picData->lookahead.intraSatdForVbv);
     }
     if (analysis->sliceType == X265_TYPE_IDR || analysis->sliceType == X265_TYPE_I)
     {
@@ -3321,9 +3357,9 @@
         modeBuf = tempBuf + depthBytes;
         partSizes = tempBuf + 2 * depthBytes;
 
-        X265_FREAD(depthBuf, sizeof(uint8_t), depthBytes, m_analysisFile, intraPic->depth);
-        X265_FREAD(modeBuf, sizeof(uint8_t), depthBytes, m_analysisFile, intraPic->chromaModes);
-        X265_FREAD(partSizes, sizeof(uint8_t), depthBytes, m_analysisFile, intraPic->partSizes);
+        X265_FREAD(depthBuf, sizeof(uint8_t), depthBytes, m_analysisFileIn, intraPic->depth);
+        X265_FREAD(modeBuf, sizeof(uint8_t), depthBytes, m_analysisFileIn, intraPic->chromaModes);
+        X265_FREAD(partSizes, sizeof(uint8_t), depthBytes, m_analysisFileIn, intraPic->partSizes);
 
         size_t count = 0;
         for (uint32_t d = 0; d < depthBytes; d++)
@@ -3344,12 +3380,12 @@
 
         if (!m_param->scaleFactor)
         {
-            X265_FREAD(((analysis_intra_data *)analysis->intraData)->modes, sizeof(uint8_t), analysis->numCUsInFrame * analysis->numPartitions, m_analysisFile, intraPic->modes);
+            X265_FREAD(((analysis_intra_data *)analysis->intraData)->modes, sizeof(uint8_t), analysis->numCUsInFrame * analysis->numPartitions, m_analysisFileIn, intraPic->modes);
         }
         else
         {
             uint8_t *tempLumaBuf = X265_MALLOC(uint8_t, analysis->numCUsInFrame * scaledNumPartition);
-            X265_FREAD(tempLumaBuf, sizeof(uint8_t), analysis->numCUsInFrame * scaledNumPartition, m_analysisFile, intraPic->modes);
+            X265_FREAD(tempLumaBuf, sizeof(uint8_t), analysis->numCUsInFrame * scaledNumPartition, m_analysisFileIn, intraPic->modes);
             for (uint32_t ctu32Idx = 0, cnt = 0; ctu32Idx < analysis->numCUsInFrame * scaledNumPartition; ctu32Idx++, cnt += factor)
                 memset(&((analysis_intra_data *)analysis->intraData)->modes[cnt], tempLumaBuf[ctu32Idx], factor);
             X265_FREE(tempLumaBuf);
@@ -3362,7 +3398,7 @@
     {
         uint32_t numDir = analysis->sliceType == X265_TYPE_P ? 1 : 2;
         uint32_t numPlanes = m_param->internalCsp == X265_CSP_I400 ? 1 : 3;
-        X265_FREAD((WeightParam*)analysis->wt, sizeof(WeightParam), numPlanes * numDir, m_analysisFile, (picIn->analysisData.wt));
+        X265_FREAD((WeightParam*)analysis->wt, sizeof(WeightParam), numPlanes * numDir, m_analysisFileIn, (picIn->analysisData.wt));
         if (m_param->analysisReuseLevel < 2)
             return;
 
@@ -3384,33 +3420,33 @@
         depthBuf = tempBuf;
         modeBuf = tempBuf + depthBytes;
 
-        X265_FREAD(depthBuf, sizeof(uint8_t), depthBytes, m_analysisFile, interPic->depth);
-        X265_FREAD(modeBuf, sizeof(uint8_t), depthBytes, m_analysisFile, interPic->modes);
+        X265_FREAD(depthBuf, sizeof(uint8_t), depthBytes, m_analysisFileIn, interPic->depth);
+        X265_FREAD(modeBuf, sizeof(uint8_t), depthBytes, m_analysisFileIn, interPic->modes);
 
         if (m_param->analysisReuseLevel > 4)
         {
             partSize = modeBuf + depthBytes;
             mergeFlag = partSize + depthBytes;
-            X265_FREAD(partSize, sizeof(uint8_t), depthBytes, m_analysisFile, interPic->partSize);
-            X265_FREAD(mergeFlag, sizeof(uint8_t), depthBytes, m_analysisFile, interPic->mergeFlag);
+            X265_FREAD(partSize, sizeof(uint8_t), depthBytes, m_analysisFileIn, interPic->partSize);
+            X265_FREAD(mergeFlag, sizeof(uint8_t), depthBytes, m_analysisFileIn, interPic->mergeFlag);
 
             if (m_param->analysisReuseLevel == 10)
             {
                 interDir = mergeFlag + depthBytes;
-                X265_FREAD(interDir, sizeof(uint8_t), depthBytes, m_analysisFile, interPic->interDir);
+                X265_FREAD(interDir, sizeof(uint8_t), depthBytes, m_analysisFileIn, interPic->interDir);
                 if (bIntraInInter)
                 {
                     chromaDir = interDir + depthBytes;
-                    X265_FREAD(chromaDir, sizeof(uint8_t), depthBytes, m_analysisFile, intraPic->chromaModes);
+                    X265_FREAD(chromaDir, sizeof(uint8_t), depthBytes, m_analysisFileIn, intraPic->chromaModes);
                 }
                 for (uint32_t i = 0; i < numDir; i++)
                 {
                     mvpIdx[i] = X265_MALLOC(uint8_t, depthBytes);
                     refIdx[i] = X265_MALLOC(int8_t, depthBytes);
                     mv[i] = X265_MALLOC(MV, depthBytes);
-                    X265_FREAD(mvpIdx[i], sizeof(uint8_t), depthBytes, m_analysisFile, interPic->mvpIdx[i]);
-                    X265_FREAD(refIdx[i], sizeof(int8_t), depthBytes, m_analysisFile, interPic->refIdx[i]);
-                    X265_FREAD(mv[i], sizeof(MV), depthBytes, m_analysisFile, interPic->mv[i]);
+                    X265_FREAD(mvpIdx[i], sizeof(uint8_t), depthBytes, m_analysisFileIn, interPic->mvpIdx[i]);
+                    X265_FREAD(refIdx[i], sizeof(int8_t), depthBytes, m_analysisFileIn, interPic->refIdx[i]);
+                    X265_FREAD(mv[i], sizeof(MV), depthBytes, m_analysisFileIn, interPic->mv[i]);
                 }
             }
         }
@@ -3469,12 +3505,12 @@
             {
                 if (!m_param->scaleFactor)
                 {
-                    X265_FREAD(((analysis_intra_data *)analysis->intraData)->modes, sizeof(uint8_t), analysis->numCUsInFrame * analysis->numPartitions, m_analysisFile, intraPic->modes);
+                    X265_FREAD(((analysis_intra_data *)analysis->intraData)->modes, sizeof(uint8_t), analysis->numCUsInFrame * analysis->numPartitions, m_analysisFileIn, intraPic->modes);
                 }
                 else
                 {
                     uint8_t *tempLumaBuf = X265_MALLOC(uint8_t, analysis->numCUsInFrame * scaledNumPartition);
-                    X265_FREAD(tempLumaBuf, sizeof(uint8_t), analysis->numCUsInFrame * scaledNumPartition, m_analysisFile, intraPic->modes);
+                    X265_FREAD(tempLumaBuf, sizeof(uint8_t), analysis->numCUsInFrame * scaledNumPartition, m_analysisFileIn, intraPic->modes);
                     for (uint32_t ctu32Idx = 0, cnt = 0; ctu32Idx < analysis->numCUsInFrame * scaledNumPartition; ctu32Idx++, cnt += factor)
                         memset(&((analysis_intra_data *)analysis->intraData)->modes[cnt], tempLumaBuf[ctu32Idx], factor);
                     X265_FREE(tempLumaBuf);
@@ -3482,7 +3518,7 @@
             }
         }
         else
-            X265_FREAD(((analysis_inter_data *)analysis->interData)->ref, sizeof(int32_t), analysis->numCUsInFrame * X265_MAX_PRED_MODE_PER_CTU * numDir, m_analysisFile, interPic->ref);
+            X265_FREAD(((analysis_inter_data *)analysis->interData)->ref, sizeof(int32_t), analysis->numCUsInFrame * X265_MAX_PRED_MODE_PER_CTU * numDir, m_analysisFileIn, interPic->ref);
 
         consumedBytes += frameRecordSize;
         if (numDir == 1)
@@ -3757,51 +3793,51 @@
     if (!m_param->bUseAnalysisFile)
         return;
 
-    X265_FWRITE(&analysis->frameRecordSize, sizeof(uint32_t), 1, m_analysisFile);
-    X265_FWRITE(&depthBytes, sizeof(uint32_t), 1, m_analysisFile);
-    X265_FWRITE(&analysis->poc, sizeof(int), 1, m_analysisFile);
-    X265_FWRITE(&analysis->sliceType, sizeof(int), 1, m_analysisFile);
-    X265_FWRITE(&analysis->bScenecut, sizeof(int), 1, m_analysisFile);
-    X265_FWRITE(&analysis->satdCost, sizeof(int64_t), 1, m_analysisFile);
-    X265_FWRITE(&analysis->numCUsInFrame, sizeof(int), 1, m_analysisFile);
-    X265_FWRITE(&analysis->numPartitions, sizeof(int), 1, m_analysisFile);
+    X265_FWRITE(&analysis->frameRecordSize, sizeof(uint32_t), 1, m_analysisFileOut);
+    X265_FWRITE(&depthBytes, sizeof(uint32_t), 1, m_analysisFileOut);
+    X265_FWRITE(&analysis->poc, sizeof(int), 1, m_analysisFileOut);
+    X265_FWRITE(&analysis->sliceType, sizeof(int), 1, m_analysisFileOut);
+    X265_FWRITE(&analysis->bScenecut, sizeof(int), 1, m_analysisFileOut);
+    X265_FWRITE(&analysis->satdCost, sizeof(int64_t), 1, m_analysisFileOut);
+    X265_FWRITE(&analysis->numCUsInFrame, sizeof(int), 1, m_analysisFileOut);
+    X265_FWRITE(&analysis->numPartitions, sizeof(int), 1, m_analysisFileOut);
     if (analysis->sliceType > X265_TYPE_I)
-        X265_FWRITE((WeightParam*)analysis->wt, sizeof(WeightParam), numPlanes * numDir, m_analysisFile);
+        X265_FWRITE((WeightParam*)analysis->wt, sizeof(WeightParam), numPlanes * numDir, m_analysisFileOut);
 
     if (m_param->analysisReuseLevel < 2)
         return;
 
     if (analysis->sliceType == X265_TYPE_IDR || analysis->sliceType == X265_TYPE_I)
     {
-        X265_FWRITE(((analysis_intra_data*)analysis->intraData)->depth, sizeof(uint8_t), depthBytes, m_analysisFile);
-        X265_FWRITE(((analysis_intra_data*)analysis->intraData)->chromaModes, sizeof(uint8_t), depthBytes, m_analysisFile);
-        X265_FWRITE(((analysis_intra_data*)analysis->intraData)->partSizes, sizeof(char), depthBytes, m_analysisFile);
-        X265_FWRITE(((analysis_intra_data*)analysis->intraData)->modes, sizeof(uint8_t), analysis->numCUsInFrame * analysis->numPartitions, m_analysisFile);
+        X265_FWRITE(((analysis_intra_data*)analysis->intraData)->depth, sizeof(uint8_t), depthBytes, m_analysisFileOut);
+        X265_FWRITE(((analysis_intra_data*)analysis->intraData)->chromaModes, sizeof(uint8_t), depthBytes, m_analysisFileOut);
+        X265_FWRITE(((analysis_intra_data*)analysis->intraData)->partSizes, sizeof(char), depthBytes, m_analysisFileOut);
+        X265_FWRITE(((analysis_intra_data*)analysis->intraData)->modes, sizeof(uint8_t), analysis->numCUsInFrame * analysis->numPartitions, m_analysisFileOut);
     }
     else
     {
-        X265_FWRITE(((analysis_inter_data*)analysis->interData)->depth, sizeof(uint8_t), depthBytes, m_analysisFile);
-        X265_FWRITE(((analysis_inter_data*)analysis->interData)->modes, sizeof(uint8_t), depthBytes, m_analysisFile);
+        X265_FWRITE(((analysis_inter_data*)analysis->interData)->depth, sizeof(uint8_t), depthBytes, m_analysisFileOut);
+        X265_FWRITE(((analysis_inter_data*)analysis->interData)->modes, sizeof(uint8_t), depthBytes, m_analysisFileOut);
         if (m_param->analysisReuseLevel > 4)
         {
-            X265_FWRITE(((analysis_inter_data*)analysis->interData)->partSize, sizeof(uint8_t), depthBytes, m_analysisFile);
-            X265_FWRITE(((analysis_inter_data*)analysis->interData)->mergeFlag, sizeof(uint8_t), depthBytes, m_analysisFile);
+            X265_FWRITE(((analysis_inter_data*)analysis->interData)->partSize, sizeof(uint8_t), depthBytes, m_analysisFileOut);
+            X265_FWRITE(((analysis_inter_data*)analysis->interData)->mergeFlag, sizeof(uint8_t), depthBytes, m_analysisFileOut);
             if (m_param->analysisReuseLevel == 10)
             {
-                X265_FWRITE(((analysis_inter_data*)analysis->interData)->interDir, sizeof(uint8_t), depthBytes, m_analysisFile);
-                if (bIntraInInter) X265_FWRITE(((analysis_intra_data*)analysis->intraData)->chromaModes, sizeof(uint8_t), depthBytes, m_analysisFile);
+                X265_FWRITE(((analysis_inter_data*)analysis->interData)->interDir, sizeof(uint8_t), depthBytes, m_analysisFileOut);
+                if (bIntraInInter) X265_FWRITE(((analysis_intra_data*)analysis->intraData)->chromaModes, sizeof(uint8_t), depthBytes, m_analysisFileOut);
                 for (uint32_t dir = 0; dir < numDir; dir++)
                 {
-                    X265_FWRITE(((analysis_inter_data*)analysis->interData)->mvpIdx[dir], sizeof(uint8_t), depthBytes, m_analysisFile);
-                    X265_FWRITE(((analysis_inter_data*)analysis->interData)->refIdx[dir], sizeof(int8_t), depthBytes, m_analysisFile);
-                    X265_FWRITE(((analysis_inter_data*)analysis->interData)->mv[dir], sizeof(MV), depthBytes, m_analysisFile);
+                    X265_FWRITE(((analysis_inter_data*)analysis->interData)->mvpIdx[dir], sizeof(uint8_t), depthBytes, m_analysisFileOut);
+                    X265_FWRITE(((analysis_inter_data*)analysis->interData)->refIdx[dir], sizeof(int8_t), depthBytes, m_analysisFileOut);
+                    X265_FWRITE(((analysis_inter_data*)analysis->interData)->mv[dir], sizeof(MV), depthBytes, m_analysisFileOut);
                 }
                 if (bIntraInInter)
-                    X265_FWRITE(((analysis_intra_data*)analysis->intraData)->modes, sizeof(uint8_t), analysis->numCUsInFrame * analysis->numPartitions, m_analysisFile);
+                    X265_FWRITE(((analysis_intra_data*)analysis->intraData)->modes, sizeof(uint8_t), analysis->numCUsInFrame * analysis->numPartitions, m_analysisFileOut);
             }
         }
         if (m_param->analysisReuseLevel != 10)
-            X265_FWRITE(((analysis_inter_data*)analysis->interData)->ref, sizeof(int32_t), analysis->numCUsInFrame * X265_MAX_PRED_MODE_PER_CTU * numDir, m_analysisFile);
+            X265_FWRITE(((analysis_inter_data*)analysis->interData)->ref, sizeof(int32_t), analysis->numCUsInFrame * X265_MAX_PRED_MODE_PER_CTU * numDir, m_analysisFileOut);
 
     }
 #undef X265_FWRITE

x265_2.6.tar.gz/source/encoder/encoder.h -> x265_2.7.tar.gz/source/encoder/encoder.h Changed

x265_2.6.tar.gz/source/encoder/frameencoder.cpp -> x265_2.7.tar.gz/source/encoder/frameencoder.cpp Changed

@@ -335,15 +335,13 @@
             while (!m_frame->m_ctuInfo)
                 m_frame->m_copied.wait();
         }
-        if ((m_param->bMVType == AVC_INFO) && !m_param->analysisReuseMode && !(IS_X265_TYPE_I(m_frame->m_lowres.sliceType)))
+        if ((m_param->bMVType == AVC_INFO) && !m_param->analysisSave && !m_param->analysisLoad && !(IS_X265_TYPE_I(m_frame->m_lowres.sliceType)))
         {
             while (((m_frame->m_analysisData.interData == NULL && m_frame->m_analysisData.intraData == NULL) || (uint32_t)m_frame->m_poc != m_frame->m_analysisData.poc))
                 m_frame->m_copyMVType.wait();
         }
         compressFrame();
         m_done.trigger(); /* FrameEncoder::getEncodedPicture() blocks for this event */
-        if (m_frame != NULL)
-            m_frame->m_reconEncoded.trigger();
         m_enable.wait();
     }
 }
@@ -432,7 +430,7 @@
     bool bUseWeightB = slice->m_sliceType == B_SLICE && slice->m_pps->bUseWeightedBiPred;
 
     WeightParam* reuseWP = NULL;
-    if (m_param->analysisReuseMode && (bUseWeightP || bUseWeightB))
+    if (m_param->analysisLoad && (bUseWeightP || bUseWeightB))
         reuseWP = (WeightParam*)m_frame->m_analysisData.wt;
 
     if (bUseWeightP || bUseWeightB)
@@ -441,7 +439,7 @@
         m_cuStats.countWeightAnalyze++;
         ScopedElapsedTime time(m_cuStats.weightAnalyzeTime);
 #endif
-        if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD)
+        if (m_param->analysisLoad)
         {
             for (int list = 0; list < slice->isInterB() + 1; list++) 
             {
@@ -468,6 +466,8 @@
     else
         slice->disableWeights();
 
+    if (m_param->analysisSave && (bUseWeightP || bUseWeightB))
+        reuseWP = (WeightParam*)m_frame->m_analysisData.wt;
     // Generate motion references
     int numPredDir = slice->isInterP() ? 1 : slice->isInterB() ? 2 : 0;
     for (int l = 0; l < numPredDir; l++)
@@ -480,7 +480,7 @@
             slice->m_refReconPicList[l][ref] = slice->m_refFrameList[l][ref]->m_reconPic;
             m_mref[l][ref].init(slice->m_refReconPicList[l][ref], w, *m_param);
         }
-        if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE && (bUseWeightP || bUseWeightB))
+        if (m_param->analysisSave && (bUseWeightP || bUseWeightB))
         {
             for (int i = 0; i < (m_param->internalCsp != X265_CSP_I400 ? 3 : 1); i++)
                 *(reuseWP++) = slice->m_weightPredTable[l][0][i];
@@ -1413,7 +1413,7 @@
             /* TODO: use defines from slicetype.h for lowres block size */
             uint32_t block_y = (ctu->m_cuPelY >> m_param->maxLog2CUSize) * noOfBlocks;
             uint32_t block_x = (ctu->m_cuPelX >> m_param->maxLog2CUSize) * noOfBlocks;
-            if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || !m_param->bDisableLookahead)
+            if (!m_param->analysisLoad || !m_param->bDisableLookahead)
             {
                 cuStat.vbvCost = 0;
                 cuStat.intraVbvCost = 0;
@@ -1748,8 +1748,8 @@
         if (rowInSlice == rowCount)
         {
             m_rowSliceTotalBits[sliceId] = 0;
-            if (bIsVbv)
-            {                
+            if (bIsVbv && !(m_param->rc.bEnableConstVbv && m_param->bEnableWavefront))
+            {          
                 for (uint32_t i = m_sliceBaseRow[sliceId]; i < rowCount + m_sliceBaseRow[sliceId]; i++)
                     m_rowSliceTotalBits[sliceId] += curEncData.m_rowStat[i].encodedBits;
             }

x265_2.6.tar.gz/source/encoder/framefilter.cpp -> x265_2.7.tar.gz/source/encoder/framefilter.cpp Changed

x265_2.6.tar.gz/source/encoder/ratecontrol.cpp -> x265_2.7.tar.gz/source/encoder/ratecontrol.cpp Changed

x265_2.6.tar.gz/source/encoder/ratecontrol.h -> x265_2.7.tar.gz/source/encoder/ratecontrol.h Changed

x265_2.6.tar.gz/source/encoder/sao.h -> x265_2.7.tar.gz/source/encoder/sao.h Changed

x265_2.6.tar.gz/source/encoder/search.cpp -> x265_2.7.tar.gz/source/encoder/search.cpp Changed

@@ -1947,7 +1947,7 @@
         /* poc difference is out of range for lookahead */
         return 0;
 
-    MV* mvs = m_frame->m_lowres.lowresMvs[list][diffPoc - 1];
+    MV* mvs = m_frame->m_lowres.lowresMvs[list][diffPoc];
     if (mvs[0].x == 0x7FFF)
         /* this motion search was not estimated by lookahead */
         return 0;
@@ -2073,7 +2073,7 @@
     int mvpIdx = selectMVP(interMode.cu, pu, amvp, list, ref);
     MV mvmin, mvmax, outmv, mvp = amvp[mvpIdx];
 
-    if (!m_param->analysisReuseMode) /* Prevents load/save outputs from diverging if lowresMV is not available */
+    if (!m_param->analysisSave && !m_param->analysisLoad) /* Prevents load/save outputs from diverging if lowresMV is not available */
     {
         MV lmv = getLowresMV(interMode.cu, pu, list, ref);
         if (lmv.notZero())
@@ -2161,7 +2161,7 @@
         cu.getNeighbourMV(puIdx, pu.puAbsPartIdx, interMode.interNeighbours);
 
         /* Uni-directional prediction */
-        if ((m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
+        if ((m_param->analysisLoad && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
             || (m_param->analysisMultiPassRefine && m_param->rc.bStatRead) || (m_param->bMVType == AVC_INFO))
         {
             for (int list = 0; list < numPredDir; list++)
@@ -2297,7 +2297,7 @@
                     int mvpIdx = selectMVP(cu, pu, amvp, list, ref);
                     MV mvmin, mvmax, outmv, mvp = amvp[mvpIdx];
 
-                    if (!m_param->analysisReuseMode) /* Prevents load/save outputs from diverging when lowresMV is not available */
+                    if (!m_param->analysisSave && !m_param->analysisLoad) /* Prevents load/save outputs from diverging when lowresMV is not available */
                     {
                         MV lmv = getLowresMV(cu, pu, list, ref);
                         if (lmv.notZero())

x265_2.6.tar.gz/source/encoder/slicetype.cpp -> x265_2.7.tar.gz/source/encoder/slicetype.cpp Changed

@@ -154,7 +154,7 @@
     int blockXY = 0;
     int blockX = 0, blockY = 0;
     double strength = 0.f;
-    if (param->rc.aqMode == X265_AQ_NONE || param->rc.aqStrength == 0)
+    if ((param->rc.aqMode == X265_AQ_NONE || param->rc.aqStrength == 0) || (param->rc.bStatRead && param->rc.cuTree && IS_REFERENCED(curFrame)))
     {
         /* Need to init it anyways for CU tree */
         int cuCount = blockCount;
@@ -589,7 +589,7 @@
     m_outputSignalRequired = false;
     m_isActive = true;
     m_inputCount = 0;
-
+    m_extendGopBoundary = false;
     m_8x8Height = ((m_param->sourceHeight / 2) + X265_LOWRES_CU_SIZE - 1) >> X265_LOWRES_CU_BITS;
     m_8x8Width = ((m_param->sourceWidth / 2) + X265_LOWRES_CU_SIZE - 1) >> X265_LOWRES_CU_BITS;
     m_cuCount = m_8x8Width * m_8x8Height;
@@ -646,7 +646,11 @@
         m_numRowsPerSlice = m_8x8Height;
         m_numCoopSlices = 1;
     }
-
+    if (param->gopLookahead && (param->gopLookahead > (param->lookaheadDepth - param->bframes - 2)))
+    {
+        param->gopLookahead = X265_MAX(0, param->lookaheadDepth - param->bframes - 2);
+        x265_log(param, X265_LOG_WARNING, "Gop-lookahead cannot be greater than (rc-lookahead - length of the mini-gop); Clipping gop-lookahead to %d\n", param->gopLookahead);
+    }
 #if DETAILED_CU_STATS
     m_slicetypeDecideElapsedTime = 0;
     m_preLookaheadElapsedTime = 0;
@@ -742,7 +746,7 @@
 /* Called by API thread */
 void Lookahead::addPicture(Frame& curFrame, int sliceType)
 {
-    if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->bDisableLookahead)
+    if (m_param->analysisLoad && m_param->bDisableLookahead)
     {
         if (!m_filled)
             m_filled = true;
@@ -843,7 +847,7 @@
             return out;
         }
 
-        if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->bDisableLookahead)
+        if (m_param->analysisLoad && m_param->bDisableLookahead)
             return NULL;
 
         findJob(-1); /* run slicetypeDecide() if necessary */
@@ -875,7 +879,7 @@
     Slice *slice = curFrame->m_encData->m_slice;
     int p0 = 0, p1, b;
     int poc = slice->m_poc;
-    int l0poc = slice->m_refPOCList[0][0];
+    int l0poc = slice->m_rps.numberOfNegativePictures ? slice->m_refPOCList[0][0] : -1;
     int l1poc = slice->m_refPOCList[1][0];
 
     switch (slice->m_sliceType)
@@ -892,23 +896,34 @@
         break;
 
     case B_SLICE:
-        b = poc - l0poc;
-        p1 = b + l1poc - poc;
-        frames[p0] = &slice->m_refFrameList[0][0]->m_lowres;
-        frames[b] = &curFrame->m_lowres;
-        frames[p1] = &slice->m_refFrameList[1][0]->m_lowres;
+        if (l0poc >= 0)
+        {
+            b = poc - l0poc;
+            p1 = b + l1poc - poc;
+            frames[p0] = &slice->m_refFrameList[0][0]->m_lowres;
+            frames[b] = &curFrame->m_lowres;
+            frames[p1] = &slice->m_refFrameList[1][0]->m_lowres;
+        }
+        else 
+        {
+            p0 = b = 0;
+            p1 = b + l1poc - poc;
+            frames[p0] = frames[b] = &curFrame->m_lowres;
+            frames[p1] = &slice->m_refFrameList[1][0]->m_lowres;
+        }
+        
         break;
 
     default:
         return;
     }
-    if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || !m_param->bDisableLookahead)
+    if (!m_param->analysisLoad || !m_param->bDisableLookahead)
     {
         X265_CHECK(curFrame->m_lowres.costEst[b - p0][p1 - b] > 0, "Slice cost not estimated\n")
         if (m_param->rc.cuTree && !m_param->rc.bStatRead)
             /* update row satds based on cutree offsets */
             curFrame->m_lowres.satdCost = frameCostRecalculate(frames, p0, p1, b);
-        else if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || m_param->scaleFactor)
+        else if (!m_param->analysisLoad || m_param->scaleFactor)
         {
             if (m_param->rc.aqMode)
                 curFrame->m_lowres.satdCost = curFrame->m_lowres.costEstAq[b - p0][p1 - b];
@@ -982,11 +997,8 @@
         ProfileLookaheadTime(m_lookahead.m_preLookaheadElapsedTime, m_lookahead.m_countPreLookahead);
         ProfileScopeEvent(prelookahead);
         m_lock.release();
-
         preFrame->m_lowres.init(preFrame->m_fencPic, preFrame->m_poc);
-        if (m_lookahead.m_param->rc.bStatRead && m_lookahead.m_param->rc.cuTree && IS_REFERENCED(preFrame))
-            /* cu-tree offsets were read from stats file */;
-        else if (m_lookahead.m_bAdaptiveQuant)
+        if (m_lookahead.m_bAdaptiveQuant)
             tld.calcAdaptiveQuantFrame(preFrame, m_lookahead.m_param);
         tld.lowresIntraEstimate(preFrame->m_lowres, m_lookahead.m_param->rc.qgSize);
         preFrame->m_lowresInit = true;
@@ -1052,7 +1064,7 @@
     {
         slicetypeAnalyse(frames, false);
         bool bIsVbv = m_param->rc.vbvBufferSize > 0 && m_param->rc.vbvMaxBitrate > 0;
-        if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->scaleFactor && bIsVbv)
+        if (m_param->analysisLoad && m_param->scaleFactor && bIsVbv)
         {
             int numFrames;
             for (numFrames = 0; numFrames < maxSearch; numFrames++)
@@ -1086,7 +1098,8 @@
             x265_log(m_param, X265_LOG_WARNING, "B-ref at frame %d incompatible with B-pyramid and %d reference frames\n",
                      frm.sliceType, m_param->maxNumReferences);
         }
-        if ((!m_param->bIntraRefresh || frm.frameNum == 0) && frm.frameNum - m_lastKeyframe >= m_param->keyframeMax)
+        if ((!m_param->bIntraRefresh || frm.frameNum == 0) && frm.frameNum - m_lastKeyframe >= m_param->keyframeMax &&
+            (!m_extendGopBoundary || frm.frameNum - m_lastKeyframe >= m_param->keyframeMax + m_param->gopLookahead))
         {
             if (frm.sliceType == X265_TYPE_AUTO || frm.sliceType == X265_TYPE_I)
                 frm.sliceType = m_param->bOpenGOP && m_lastKeyframe >= 0 ? X265_TYPE_I : X265_TYPE_IDR;
@@ -1115,12 +1128,20 @@
             /* Closed GOP */
             m_lastKeyframe = frm.frameNum;
             frm.bKeyframe = true;
-            if (bframes > 0)
+            if (bframes > 0 && !m_param->radl)
             {
                 list[bframes - 1]->m_lowres.sliceType = X265_TYPE_P;
                 bframes--;
             }
         }
+        if (m_param->radl && !m_param->bOpenGOP && list[bframes + 1])
+        {
+            if ((frm.frameNum - m_lastKeyframe) >  (m_param->keyframeMax - m_param->radl - 1) && (frm.frameNum - m_lastKeyframe) <  m_param->keyframeMax)
+                frm.sliceType = X265_TYPE_B;
+            if ((frm.frameNum - m_lastKeyframe) == (m_param->keyframeMax - m_param->radl - 1))
+                frm.sliceType = X265_TYPE_P;
+        }
+
         if (bframes == m_param->bframes || !list[bframes + 1])
         {
             if (IS_X265_TYPE_B(frm.sliceType))
@@ -1170,8 +1191,13 @@
         if (bframes)
         {
             p0 = 0; // last nonb
+            bool isp0available = frames[bframes + 1]->sliceType == X265_TYPE_IDR ? false : true;
+
             for (b = 1; b <= bframes; b++)
             {
+                if (!isp0available)
+                    p0 = b;
+
                 if (frames[b]->sliceType == X265_TYPE_B)
                     for (p1 = b; frames[p1]->sliceType == X265_TYPE_B; p1++)
                         ; // find new nonb or bref
@@ -1181,7 +1207,10 @@
                 estGroup.singleCost(p0, p1, b);
 
                 if (frames[b]->sliceType == X265_TYPE_BREF)
+                {
                     p0 = b;
+                    isp0available = true;
+                }
             }
         }
     }
@@ -1205,9 +1234,8 @@
     int idx = 0;
     list[bframes]->m_reorderedPts = pts[idx++];
     m_outputQueue.pushBack(*list[bframes]);
-
     /* Add B-ref frame next to P frame in output queue, the B-ref encode before non B-ref frame */
-    if (bframes > 1 && m_param->bBPyramid)
+    if (brefs)
     {
         for (int i = 0; i < bframes; i++)
         {
@@ -1247,7 +1275,7 @@
         frames[j + 1] = NULL;
         slicetypeAnalyse(frames, true);
         bool bIsVbv = m_param->rc.vbvBufferSize > 0 && m_param->rc.vbvMaxBitrate > 0;
-        if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->scaleFactor && bIsVbv)
+        if (m_param->analysisLoad && m_param->scaleFactor && bIsVbv)
         {
             int numFrames;
             for (numFrames = 0; numFrames < maxSearch; numFrames++)
@@ -1377,12 +1405,14 @@
             cuTree(frames, 0, bKeyframe);
         return;
     }
-
     frames[framecnt + 1] = NULL;
+    int keyFrameLimit = m_param->keyframeMax + m_lastKeyframe - frames[0]->frameNum - 1;
+    if (m_param->gopLookahead && keyFrameLimit <= m_param->bframes + 1)
+        keyintLimit = keyFrameLimit + m_param->gopLookahead;
+    else
+        keyintLimit = keyFrameLimit;
 
-    keyintLimit = m_param->keyframeMax - frames[0]->frameNum + m_lastKeyframe - 1;
     origNumFrames = numFrames = m_param->bIntraRefresh ? framecnt : X265_MIN(framecnt, keyintLimit);
-
     if (bIsVbvLookahead)
         numFrames = framecnt;
     else if (m_param->bOpenGOP && numFrames < framecnt)
@@ -1406,12 +1436,12 @@
                     continue;
 
                 /* Skip search if already done */
-                if (frames[b]->lowresMvs[0][i - 1][0].x != 0x7FFF)
+                if (frames[b]->lowresMvs[0][i][0].x != 0x7FFF)
                     continue;
 
                 /* perform search to p1 at same distance, if possible */
                 int p1 = b + i;
-                if (p1 >= numFrames || frames[b]->lowresMvs[1][i - 1][0].x != 0x7FFF)
+                if (p1 >= numFrames || frames[b]->lowresMvs[1][i][0].x != 0x7FFF)
                     p1 = b;
 
                 estGroup.add(p0, p1, b);
@@ -1433,7 +1463,7 @@
 
                     /* only measure frame cost in this pass if motion searches
                      * are already done */
-                    if (frames[b]->lowresMvs[0][i - 1][0].x == 0x7FFF)
+                    if (frames[b]->lowresMvs[0][i][0].x == 0x7FFF)
                         continue;
 
                     int p0 = b - i;
@@ -1445,7 +1475,7 @@
                             break;
 
                         /* ensure P1 search is done */
-                        if (j && frames[b]->lowresMvs[1][j - 1][0].x == 0x7FFF)
+                        if (j && frames[b]->lowresMvs[1][j][0].x == 0x7FFF)
                             continue;
 
                         /* ensure frame cost is not done */
@@ -1472,7 +1502,26 @@
         frames[1]->sliceType = X265_TYPE_I;
         return;
     }
-
+    if (m_param->gopLookahead && (keyFrameLimit >= 0) && (keyFrameLimit <= m_param->bframes + 1))
+    {
+        bool sceneTransition = m_isSceneTransition;
+        m_extendGopBoundary = false;
+        for (int i = m_param->bframes + 1; i < origNumFrames; i += m_param->bframes + 1)
+        {
+            scenecut(frames, i, i + 1, true, origNumFrames);
+            for (int j = i + 1; j <= X265_MIN(i + m_param->bframes + 1, origNumFrames); j++)
+            {
+                if (frames[j]->bScenecut && scenecutInternal(frames, j - 1, j, true) )
+                {
+                    m_extendGopBoundary = true;
+                    break;
+                }
+            }
+            if (m_extendGopBoundary)
+                break;
+        }
+        m_isSceneTransition = sceneTransition;
+    }
     if (m_param->bframes)
     {
         if (m_param->bFrameAdaptive == X265_B_ADAPT_TRELLIS)
@@ -1578,6 +1627,8 @@
 
     if (m_param->rc.cuTree)
         cuTree(frames, X265_MIN(numFrames, m_param->keyframeMax), bKeyframe);
+    if (m_param->gopLookahead && (keyFrameLimit >= 0) && (keyFrameLimit <= m_param->bframes + 1) && !m_extendGopBoundary)
+        keyintLimit = keyFrameLimit;
 
     if (!m_param->bIntraRefresh)
         for (int j = keyintLimit + 1; j <= numFrames; j += m_param->keyframeMax)
@@ -1588,8 +1639,8 @@
 
     if (bIsVbvLookahead)
         vbvLookahead(frames, numFrames, bKeyframe);
+    int maxp1 = X265_MIN(m_param->bframes + 1, origNumFrames);
 
-     int maxp1 = X265_MIN(m_param->bframes + 1, origNumFrames);
     /* Restore frame types for all frames that haven't actually been decided yet. */
     for (int j = resetStart; j <= numFrames; j++)
     {
@@ -1613,8 +1664,8 @@
         bool fluctuate = false;
         bool noScenecuts = false;
         int64_t avgSatdCost = 0;
-        if (frames[0]->costEst[1][0] > -1)
-            avgSatdCost = frames[0]->costEst[1][0];
+        if (frames[p0]->costEst[p1 - p0][0] > -1)
+            avgSatdCost = frames[p0]->costEst[p1 - p0][0];
         int cnt = 1;
         /* Where A and B are scenes: AAAAAABBBAAAAAA
          * If BBB is shorter than (maxp1-p0), it is detected as a flash
@@ -1700,12 +1751,10 @@
 
     CostEstimateGroup estGroup(*this, frames);
     estGroup.singleCost(p0, p1, p1);
-
     int64_t icost = frame->costEst[0][0];
     int64_t pcost = frame->costEst[p1 - p0][0];
-    int gopSize = frame->frameNum - m_lastKeyframe;
+    int gopSize = (frame->frameNum - m_lastKeyframe) % m_param->keyframeMax;
     float threshMax = (float)(m_param->scenecutThreshold / 100.0);
-
     /* magic numbers pulled out of thin air */
     float threshMin = (float)(threshMax * 0.25);
     double bias = m_param->scenecutBias;
@@ -1841,7 +1890,7 @@
 
 void Lookahead::calcMotionAdaptiveQuantFrame(Lowres **frames, int p0, int p1, int b)
 {
-    int listDist[2] = { b - p0 - 1, p1 - b - 1 };
+    int listDist[2] = { b - p0, p1 - b };
     int32_t strideInCU = m_8x8Width;
     double qp_adj = 0, avg_adj = 0, avg_adj_pow2 = 0, sd;
     for (uint16_t blocky = 0; blocky < m_8x8Height; blocky++)
@@ -2004,7 +2053,7 @@
     int32_t distScaleFactor = (((b - p0) << 8) + ((p1 - p0) >> 1)) / (p1 - p0);
     int32_t bipredWeight = m_param->bEnableWeightedBiPred ? 64 - (distScaleFactor >> 2) : 32;
     int32_t bipredWeights[2] = { bipredWeight, 64 - bipredWeight };
-    int listDist[2] = { b - p0 - 1, p1 - b - 1 };
+    int listDist[2] = { b - p0, p1 - b };
 
     memset(m_scratch, 0, m_8x8Width * sizeof(int));
 
@@ -2279,17 +2328,15 @@
         score = fenc->costEst[b - p0][p1 - b];
     else
     {
-        X265_CHECK(p0 != b, "I frame estimates should always be pre-calculated\n");
-
         bool bDoSearch[2];
-        bDoSearch[0] = p0 < b && fenc->lowresMvs[0][b - p0 - 1][0].x == 0x7FFF;
-        bDoSearch[1] = p1 > b && fenc->lowresMvs[1][p1 - b - 1][0].x == 0x7FFF;
+        bDoSearch[0] = fenc->lowresMvs[0][b - p0][0].x == 0x7FFF;
+        bDoSearch[1] = p1 > b && fenc->lowresMvs[1][p1 - b][0].x == 0x7FFF;
 
 #if CHECKED_BUILD
-        X265_CHECK(!(p0 < b && fenc->lowresMvs[0][b - p0 - 1][0].x == 0x7FFE), "motion search batch duplication L0\n");
-        X265_CHECK(!(p1 > b && fenc->lowresMvs[1][p1 - b - 1][0].x == 0x7FFE), "motion search batch duplication L1\n");
-        if (bDoSearch[0]) fenc->lowresMvs[0][b - p0 - 1][0].x = 0x7FFE;
-        if (bDoSearch[1]) fenc->lowresMvs[1][p1 - b - 1][0].x = 0x7FFE;
+        X265_CHECK(!(p0 < b && fenc->lowresMvs[0][b - p0][0].x == 0x7FFE), "motion search batch duplication L0\n");
+        X265_CHECK(!(p1 > b && fenc->lowresMvs[1][p1 - b][0].x == 0x7FFE), "motion search batch duplication L1\n");
+        if (bDoSearch[0]) fenc->lowresMvs[0][b - p0][0].x = 0x7FFE;
+        if (bDoSearch[1]) fenc->lowresMvs[1][p1 - b][0].x = 0x7FFE;
 #endif
 
         fenc->weightedRef[b - p0].isWeighted = false;
@@ -2380,7 +2427,7 @@
 
     /* A small, arbitrary bias to avoid VBV problems caused by zero-residual lookahead blocks. */
     int lowresPenalty = 4;
-    int listDist[2] = { b - p0 - 1, p1 - b - 1 };
+    int listDist[2] = { b - p0, p1 - b};
 
     MV mvmin, mvmax;
     int bcost = tld.me.COST_MAX;

x265_2.6.tar.gz/source/encoder/slicetype.h -> x265_2.7.tar.gz/source/encoder/slicetype.h Changed

x265_2.6.tar.gz/source/encoder/weightPrediction.cpp -> x265_2.7.tar.gz/source/encoder/weightPrediction.cpp Changed

x265_2.6.tar.gz/source/input/y4m.cpp -> x265_2.7.tar.gz/source/input/y4m.cpp Changed

@@ -20,7 +20,8 @@
  * This program is also available under a commercial proprietary license.
  * For more information, contact us at license @ x265.com.
  *****************************************************************************/
-
+#define _FILE_OFFSET_BITS 64
+#define _LARGEFILE_SOURCE
 #include "y4m.h"
 #include "common.h"
 
@@ -38,9 +39,7 @@
 
 using namespace X265_NS;
 using namespace std;
-
-static const char header[] = "FRAME";
-
+static const char header[] = {'F','R','A','M','E'};
 Y4MInput::Y4MInput(InputFileInfo& info)
 {
     for (int i = 0; i < QUEUE_SIZE; i++)
@@ -60,15 +59,14 @@
     ifs = NULL;
     if (!strcmp(info.filename, "-"))
     {
-        ifs = &cin;
+        ifs = stdin;
 #if _WIN32
         setmode(fileno(stdin), O_BINARY);
 #endif
     }
     else
-        ifs = new ifstream(info.filename, ios::binary | ios::in);
-
-    if (ifs && ifs->good() && parseHeader())
+        ifs = x265_fopen(info.filename, "rb");
+    if (ifs && !ferror(ifs) && parseHeader())
     {
         int pixelbytes = depth > 8 ? 2 : 1;
         for (int i = 0; i < x265_cli_csps[colorSpace].planes; i++)
@@ -91,8 +89,8 @@
     }
     if (!threadActive)
     {
-        if (ifs && ifs != &cin)
-            delete ifs;
+        if (ifs && ifs != stdin)
+            fclose(ifs);
         ifs = NULL;
         return;
     }
@@ -106,61 +104,34 @@
     info.csp = colorSpace;
     info.depth = depth;
     info.frameCount = -1;
-
-    size_t estFrameSize = framesize + strlen(header) + 1; /* assume basic FRAME\n headers */
-
+    size_t estFrameSize = framesize + sizeof(header) + 1; /* assume basic FRAME\n headers */
     /* try to estimate frame count, if this is not stdin */
-    if (ifs != &cin)
+    if (ifs != stdin)
     {
-        istream::pos_type cur = ifs->tellg();
-
-#if defined(_MSC_VER) && _MSC_VER < 1700
-        /* Older MSVC versions cannot handle 64bit file sizes properly, so go native */
-        HANDLE hFile = CreateFileA(info.filename, GENERIC_READ,
-                                   FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING,
-                                   FILE_ATTRIBUTE_NORMAL, NULL);
-        if (hFile != INVALID_HANDLE_VALUE)
-        {
-            LARGE_INTEGER size;
-            if (GetFileSizeEx(hFile, &size))
-                info.frameCount = (int)((size.QuadPart - (int64_t)cur) / estFrameSize);
-            CloseHandle(hFile);
-        }
-#else // if defined(_MSC_VER) && _MSC_VER < 1700
+        int64_t cur = ftello(ifs);
         if (cur >= 0)
         {
-            ifs->seekg(0, ios::end);
-            istream::pos_type size = ifs->tellg();
-            ifs->seekg(cur, ios::beg);
+            fseeko(ifs, 0, SEEK_END);
+            int64_t size = ftello(ifs);
+            fseeko(ifs, cur, SEEK_SET);
             if (size > 0)
                 info.frameCount = (int)((size - cur) / estFrameSize);
         }
-#endif // if defined(_MSC_VER) && _MSC_VER < 1700
     }
-
     if (info.skipFrames)
     {
-#if X86_64
-        if (ifs != &cin)
-            ifs->seekg((uint64_t)estFrameSize * info.skipFrames, ios::cur);
+        if (ifs != stdin)
+            fseeko(ifs, (int64_t)estFrameSize * info.skipFrames, SEEK_CUR);
         else
             for (int i = 0; i < info.skipFrames; i++)
-            {
-                ifs->read(buf[0], estFrameSize - framesize);
-                ifs->read(buf[0], framesize);
-            }
-#else
-        for (int i = 0; i < info.skipFrames; i++)
-            ifs->ignore(estFrameSize);
-#endif
+                if (fread(buf[0], estFrameSize - framesize, 1, ifs) + fread(buf[0], framesize, 1, ifs) != 2)
+                    break;
     }
 }
-
 Y4MInput::~Y4MInput()
 {
-    if (ifs && ifs != &cin)
-        delete ifs;
-
+    if (ifs && ifs != stdin)
+        fclose(ifs);
     for (int i = 0; i < QUEUE_SIZE; i++)
         X265_FREE(buf[i]);
 }
@@ -180,37 +151,31 @@
 
     int csp = 0;
     int d = 0;
-
-    while (ifs->good())
+    int c;
+    while ((c = fgetc(ifs)) != EOF)
     {
         // Skip Y4MPEG string
-        int c = ifs->get();
-        while (ifs->good() && (c != ' ') && (c != '\n'))
-            c = ifs->get();
-
-        while (c == ' ' && ifs->good())
+        while ((c != EOF) && (c != ' ') && (c != '\n'))
+            c = fgetc(ifs);
+        while (c == ' ')
         {
             // read parameter identifier
-            switch (ifs->get())
+            switch (fgetc(ifs))
             {
             case 'W':
                 width = 0;
-                while (ifs->good())
+                while ((c = fgetc(ifs)) != EOF)
                 {
-                    c = ifs->get();
-
                     if (c == ' ' || c == '\n')
                         break;
                     else
                         width = width * 10 + (c - '0');
                 }
                 break;
-
             case 'H':
                 height = 0;
-                while (ifs->good())
+                while ((c = fgetc(ifs)) != EOF)
                 {
-                    c = ifs->get();
                     if (c == ' ' || c == '\n')
                         break;
                     else
@@ -221,15 +186,13 @@
             case 'F':
                 rateNum = 0;
                 rateDenom = 0;
-                while (ifs->good())
+                while ((c = fgetc(ifs)) != EOF)
                 {
-                    c = ifs->get();
                     if (c == '.')
                     {
                         rateDenom = 1;
-                        while (ifs->good())
+                        while ((c = fgetc(ifs)) != EOF)
                         {
-                            c = ifs->get();
                             if (c == ' ' || c == '\n')
                                 break;
                             else
@@ -242,9 +205,8 @@
                     }
                     else if (c == ':')
                     {
-                        while (ifs->good())
+                        while ((c = fgetc(ifs)) != EOF)
                         {
-                            c = ifs->get();
                             if (c == ' ' || c == '\n')
                                 break;
                             else
@@ -260,14 +222,12 @@
             case 'A':
                 sarWidth = 0;
                 sarHeight = 0;
-                while (ifs->good())
+                while ((c = fgetc(ifs)) != EOF)
                 {
-                    c = ifs->get();
                     if (c == ':')
                     {
-                        while (ifs->good())
+                        while ((c = fgetc(ifs)) != EOF)
                         {
-                            c = ifs->get();
                             if (c == ' ' || c == '\n')
                                 break;
                             else
@@ -283,19 +243,15 @@
             case 'C':
                 csp = 0;
                 d = 0;
-                while (ifs->good())
+                while ((c = fgetc(ifs)) != EOF)
                 {
-                    c = ifs->get();
-
                     if (c <= 'o' && c >= '0')
                         csp = csp * 10 + (c - '0');
                     else if (c == 'p')
                     {
                         // example: C420p16
-                        while (ifs->good())
+                        while ((c = fgetc(ifs)) != EOF)
                         {
-                            c = ifs->get();
-
                             if (c <= '9' && c >= '0')
                                 d = d * 10 + (c - '0');
                             else
@@ -328,12 +284,10 @@
                 if (d >= 8 && d <= 16)
                     depth = d;
                 break;
-
             default:
-                while (ifs->good())
+                while ((c = fgetc(ifs)) != EOF)
                 {
                     // consume this unsupported configuration word
-                    c = ifs->get();
                     if (c == ' ' || c == '\n')
                         break;
                 }
@@ -375,30 +329,23 @@
     threadActive = false;
     writeCount.poke();
 }
-
 bool Y4MInput::populateFrameQueue()
 {
-    if (!ifs || ifs->fail())
+    if (!ifs || ferror(ifs))
         return false;
-
-    /* strip off the FRAME header */
-    char hbuf[sizeof(header)];
-
-    ifs->read(hbuf, strlen(header));
-    if (ifs->eof())
-        return false;
-
-    if (!ifs->good() || memcmp(hbuf, header, strlen(header)))
+    /* strip off the FRAME\n header */
+    char hbuf[sizeof(header) + 1];
+    if (fread(hbuf, sizeof(hbuf), 1, ifs) != 1 || memcmp(hbuf, header, sizeof(header)))
     {
-        x265_log(NULL, X265_LOG_ERROR, "y4m: frame header missing\n");
+        if (!feof(ifs))
+            x265_log(NULL, X265_LOG_ERROR, "y4m: frame header missing\n");
         return false;
     }
-
     /* consume bytes up to line feed */
-    int c = ifs->get();
-    while (c != '\n' && ifs->good())
-        c = ifs->get();
-
+    int c = hbuf[sizeof(header)];
+    while (c != '\n')
+        if ((c = fgetc(ifs)) == EOF)
+            break;
     /* wait for room in the ring buffer */
     int written = writeCount.get();
     int read = readCount.get();
@@ -408,10 +355,8 @@
         if (!threadActive)
             return false;
     }
-
     ProfileScopeEvent(frameRead);
-    ifs->read(buf[written % QUEUE_SIZE], framesize);
-    if (ifs->good())
+    if (fread(buf[written % QUEUE_SIZE], framesize, 1, ifs) == 1)
     {
         writeCount.incr();
         return true;

x265_2.6.tar.gz/source/input/y4m.h -> x265_2.7.tar.gz/source/input/y4m.h Changed

x265_2.6.tar.gz/source/input/yuv.cpp -> x265_2.7.tar.gz/source/input/yuv.cpp Changed

x265_2.6.tar.gz/source/input/yuv.h -> x265_2.7.tar.gz/source/input/yuv.h Changed

x265_2.6.tar.gz/source/output/raw.cpp -> x265_2.7.tar.gz/source/output/raw.cpp Changed

x265_2.6.tar.gz/source/test/CMakeLists.txt -> x265_2.7.tar.gz/source/test/CMakeLists.txt Changed

@@ -7,37 +7,37 @@
 
 # add X86 assembly files
 if(X86)
-enable_language(ASM_YASM)
+enable_language(ASM_NASM)
 
 if(MSVC_IDE)
-    set(YASM_SRC checkasm-a.obj)
+    set(NASM_SRC checkasm-a.obj)
     add_custom_command(
         OUTPUT checkasm-a.obj
-        COMMAND ${YASM_EXECUTABLE}
-        ARGS ${YASM_FLAGS} ${CMAKE_CURRENT_SOURCE_DIR}/checkasm-a.asm -o checkasm-a.obj
+        COMMAND ${NASM_EXECUTABLE}
+        ARGS ${NASM_FLAGS} ${CMAKE_CURRENT_SOURCE_DIR}/checkasm-a.asm -o checkasm-a.obj
         DEPENDS checkasm-a.asm)
 else()
-    set(YASM_SRC checkasm-a.asm)
+    set(NASM_SRC checkasm-a.asm)
 endif()
 endif(X86)
 
 # add ARM assembly files
 if(ARM OR CROSS_COMPILE_ARM)
     enable_language(ASM)
-    set(YASM_SRC checkasm-arm.S)
+    set(NASM_SRC checkasm-arm.S)
     add_custom_command(
         OUTPUT checkasm-arm.obj
         COMMAND ${CMAKE_CXX_COMPILER}
-        ARGS ${YASM_FLAGS} ${CMAKE_CURRENT_SOURCE_DIR}/checkasm-arm.S -o checkasm-arm.obj
+        ARGS ${NASM_FLAGS} ${CMAKE_CURRENT_SOURCE_DIR}/checkasm-arm.S -o checkasm-arm.obj
         DEPENDS checkasm-arm.S)
 endif(ARM OR CROSS_COMPILE_ARM)
 
 # add PowerPC assembly files
 if(POWER)
-    set(YASM_SRC)
+    set(NASM_SRC)
 endif(POWER)
 
-add_executable(TestBench ${YASM_SRC}
+add_executable(TestBench ${NASM_SRC}
     testbench.cpp testharness.h
     pixelharness.cpp pixelharness.h
     mbdstharness.cpp mbdstharness.h

x265_2.6.tar.gz/source/test/checkasm-a.asm -> x265_2.7.tar.gz/source/test/checkasm-a.asm Changed

@@ -26,7 +26,7 @@
 ;* For more information, contact us at license @ x265.com.
 ;*****************************************************************************
 
-%include "../common/x86/x86inc.asm"
+%include "x86inc.asm"
 
 SECTION_RODATA
 
@@ -35,24 +35,24 @@
 %if ARCH_X86_64
 ; just random numbers to reduce the chance of incidental match
 ALIGN 16
-x6:  ddq 0x79445c159ce790641a1b2550a612b48c
-x7:  ddq 0x86b2536fcd8cf6362eed899d5a28ddcd
-x8:  ddq 0x3f2bf84fc0fcca4eb0856806085e7943
-x9:  ddq 0xd229e1f5b281303facbd382dcf5b8de2
-x10: ddq 0xab63e2e11fa38ed971aeaff20b095fd9
-x11: ddq 0x77d410d5c42c882d89b0c0765892729a
-x12: ddq 0x24b3c1d2a024048bc45ea11a955d8dd5
-x13: ddq 0xdd7b8919edd427862e8ec680de14b47c
-x14: ddq 0x11e53e2b2ac655ef135ce6888fa02cbf
-x15: ddq 0x6de8f4c914c334d5011ff554472a7a10
-n7:   dq 0x21f86d66c8ca00ce
-n8:   dq 0x75b6ba21077c48ad
-n9:   dq 0xed56bb2dcb3c7736
-n10:  dq 0x8bda43d3fd1a7e06
-n11:  dq 0xb64a9c9e5d318408
-n12:  dq 0xdf9a54b303f1d3a3
-n13:  dq 0x4a75479abd64e097
-n14:  dq 0x249214109d5d1c88
+x6:  dq 0x1a1b2550a612b48c,0x79445c159ce79064
+x7:  dq 0x2eed899d5a28ddcd,0x86b2536fcd8cf636
+x8:  dq 0xb0856806085e7943,0x3f2bf84fc0fcca4e
+x9:  dq 0xacbd382dcf5b8de2,0xd229e1f5b281303f
+x10: dq 0x71aeaff20b095fd9,0xab63e2e11fa38ed9
+x11: dq 0x89b0c0765892729a,0x77d410d5c42c882d
+x12: dq 0xc45ea11a955d8dd5,0x24b3c1d2a024048b
+x13: dq 0x2e8ec680de14b47c,0xdd7b8919edd42786
+x14: dq 0x135ce6888fa02cbf,0x11e53e2b2ac655ef
+x15: dq 0x011ff554472a7a10,0x6de8f4c914c334d5
+n7:  dq 0x21f86d66c8ca00ce
+n8:  dq 0x75b6ba21077c48ad
+n9:  dq 0xed56bb2dcb3c7736
+n10: dq 0x8bda43d3fd1a7e06
+n11: dq 0xb64a9c9e5d318408
+n12: dq 0xdf9a54b303f1d3a3
+n13: dq 0x4a75479abd64e097
+n14: dq 0x249214109d5d1c88
 %endif
 
 SECTION .text
@@ -70,14 +70,14 @@
 ;-----------------------------------------------------------------------------
 cglobal checkasm_stack_clobber, 1,2
     ; Clobber the stack with junk below the stack pointer
-    %define size (max_args+6)*8
-    SUB  rsp, size
-    mov   r1, size-8
+    %define argsize (max_args+6)*8
+    SUB  rsp, argsize
+    mov   r1, argsize-8
 .loop:
     mov [rsp+r1], r0
     sub   r1, 8
     jge .loop
-    ADD  rsp, size
+    ADD  rsp, argsize
     RET
 
 %if WIN64
@@ -156,7 +156,11 @@
     mov  r9, rax
     mov r10, rdx
     lea  r0, [error_message]
+%if FORMAT_ELF
+    call puts wrt ..plt
+%else
     call puts
+%endif
     mov  r1, [rsp+max_args*8]
     mov  dword [r1], 0
     mov  rdx, r10

x265_2.6.tar.gz/source/test/regression-tests.txt -> x265_2.7.tar.gz/source/test/regression-tests.txt Changed

@@ -18,17 +18,17 @@
 BasketballDrive_1920x1080_50.y4m,--preset faster --aq-strength 2 --merange 190 --slices 3
 BasketballDrive_1920x1080_50.y4m,--preset medium --ctu 16 --max-tu-size 8 --subme 7 --qg-size 16 --cu-lossless --tu-inter-depth 3 --limit-tu 1
 BasketballDrive_1920x1080_50.y4m,--preset medium --keyint -1 --nr-inter 100 -F4 --no-sao
-BasketballDrive_1920x1080_50.y4m,--preset medium --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 2 --bitrate 7000 --limit-modes::--preset medium --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 2 --bitrate 7000 --limit-modes
+BasketballDrive_1920x1080_50.y4m,--preset medium --no-cutree --analysis-save x265_analysis.dat --analysis-reuse-level 2 --bitrate 7000 --limit-modes::--preset medium --no-cutree --analysis-load x265_analysis.dat --analysis-reuse-level 2 --bitrate 7000 --limit-modes
 BasketballDrive_1920x1080_50.y4m,--preset slow --nr-intra 100 -F4 --aq-strength 3 --qg-size 16 --limit-refs 1
 BasketballDrive_1920x1080_50.y4m,--preset slower --lossless --chromaloc 3 --subme 0 --limit-tu 4
-BasketballDrive_1920x1080_50.y4m,--preset slower --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 10 --bitrate 7000 --limit-tu 0::--preset slower --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 10 --bitrate 7000 --limit-tu 0
+BasketballDrive_1920x1080_50.y4m,--preset slower --no-cutree --analysis-save x265_analysis.dat --analysis-reuse-level 10 --bitrate 7000 --limit-tu 0::--preset slower --no-cutree --analysis-load x265_analysis.dat --analysis-reuse-level 10 --bitrate 7000 --limit-tu 0
 BasketballDrive_1920x1080_50.y4m,--preset veryslow --crf 4 --cu-lossless --pmode --limit-refs 1 --aq-mode 3 --limit-tu 3
-BasketballDrive_1920x1080_50.y4m,--preset veryslow --no-cutree --analysis-reuse-mode=save --bitrate 7000 --tskip-fast --limit-tu 4::--preset veryslow --no-cutree --analysis-reuse-mode=load --bitrate 7000  --tskip-fast --limit-tu 4
+BasketballDrive_1920x1080_50.y4m,--preset veryslow --no-cutree --analysis-save x265_analysis.dat --bitrate 7000 --tskip-fast --limit-tu 2::--preset veryslow --no-cutree --analysis-load x265_analysis.dat --bitrate 7000  --tskip-fast --limit-tu 2
 BasketballDrive_1920x1080_50.y4m,--preset veryslow --recon-y4m-exec "ffplay -i pipe:0 -autoexit"
 Coastguard-4k.y4m,--preset ultrafast --recon-y4m-exec "ffplay -i pipe:0 -autoexit"
 Coastguard-4k.y4m,--preset superfast --tune grain --overscan=crop
 Coastguard-4k.y4m,--preset superfast --tune grain --pme --aq-strength 2 --merange 190
-Coastguard-4k.y4m,--preset veryfast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 1 --bitrate 15000::--preset veryfast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 1 --bitrate 15000
+Coastguard-4k.y4m,--preset veryfast --no-cutree --analysis-save x265_analysis.dat --analysis-reuse-level 1 --bitrate 15000::--preset veryfast --no-cutree --analysis-load x265_analysis.dat --analysis-reuse-level 1 --bitrate 15000
 Coastguard-4k.y4m,--preset medium --rdoq-level 1 --tune ssim --no-signhide --me umh --slices 2
 Coastguard-4k.y4m,--preset slow --tune psnr --cbqpoffs -1 --crqpoffs 1 --limit-refs 1
 CrowdRun_1920x1080_50_10bit_422.yuv,--preset ultrafast --weightp --tune zerolatency --qg-size 16
@@ -52,7 +52,7 @@
 DucksAndLegs_1920x1080_60_10bit_444.yuv,--preset veryfast --weightp --nr-intra 1000 -F4
 DucksAndLegs_1920x1080_60_10bit_444.yuv,--preset medium --nr-inter 500 -F4 --no-psy-rdoq
 DucksAndLegs_1920x1080_60_10bit_444.yuv,--preset slower --no-weightp --rdoq-level 0 --limit-refs 3 --tu-inter-depth 4 --limit-tu 3
-DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset fast --no-cutree --analysis-reuse-mode=save --bitrate 3000 --early-skip --tu-inter-depth 3 --limit-tu 1::--preset fast --no-cutree --analysis-reuse-mode=load --bitrate 3000 --early-skip --tu-inter-depth 3 --limit-tu 1
+DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset fast --no-cutree --analysis-save x265_analysis.dat --bitrate 3000 --early-skip --tu-inter-depth 3 --limit-tu 1::--preset fast --no-cutree --analysis-load x265_analysis.dat --bitrate 3000 --early-skip --tu-inter-depth 3 --limit-tu 1
 FourPeople_1280x720_60.y4m,--preset superfast --no-wpp --lookahead-slices 2
 FourPeople_1280x720_60.y4m,--preset veryfast --aq-mode 2 --aq-strength 1.5 --qg-size 8
 FourPeople_1280x720_60.y4m,--preset medium --qp 38 --no-psy-rd
@@ -69,8 +69,8 @@
 KristenAndSara_1280x720_60.y4m,--preset slower --pmode --max-tu-size 8 --limit-refs 0 --limit-modes --limit-tu 1
 NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset superfast --tune psnr
 NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset medium --tune grain --limit-refs 2
-NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset slow --no-cutree --analysis-reuse-mode=save --rd 5 --analysis-reuse-level 10 --bitrate 9000::--preset slow --no-cutree --analysis-reuse-mode=load --rd 5 --analysis-reuse-level 10 --bitrate 9000
-News-4k.y4m,--preset ultrafast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 2 --bitrate 15000::--preset ultrafast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 2 --bitrate 15000
+NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset slow --no-cutree --analysis-save x265_analysis.dat --rd 5 --analysis-reuse-level 10 --bitrate 9000::--preset slow --no-cutree --analysis-load x265_analysis.dat --rd 5 --analysis-reuse-level 10 --bitrate 9000
+News-4k.y4m,--preset ultrafast --no-cutree --analysis-save x265_analysis.dat --analysis-reuse-level 2 --bitrate 15000::--preset ultrafast --no-cutree --analysis-load x265_analysis.dat --analysis-reuse-level 2 --bitrate 15000
 News-4k.y4m,--preset superfast --lookahead-slices 6 --aq-mode 0
 News-4k.y4m,--preset superfast --slices 4 --aq-mode 0 
 News-4k.y4m,--preset medium --tune ssim --no-sao --qg-size 16
@@ -125,7 +125,7 @@
 old_town_cross_444_720p50.y4m,--preset superfast --weightp --min-cu 16 --limit-modes
 old_town_cross_444_720p50.y4m,--preset veryfast --qp 1 --tune ssim
 old_town_cross_444_720p50.y4m,--preset faster --rd 1 --tune zero-latency
-old_town_cross_444_720p50.y4m,--preset fast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 1 --bitrate 3000 --early-skip::--preset fast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 1 --bitrate 3000 --early-skip
+old_town_cross_444_720p50.y4m,--preset fast --no-cutree --analysis-save pass1_analysis.dat --analysis-reuse-level 1 --bitrate 3000 --early-skip::--preset fast --no-cutree --analysis-load pass1_analysis.dat --analysis-save pass2_analysis.dat --analysis-reuse-level 1 --bitrate 3000 --early-skip::--preset fast --no-cutree --analysis-load pass2_analysis.dat --analysis-reuse-level 1 --bitrate 3000 --early-skip
 old_town_cross_444_720p50.y4m,--preset medium --keyint -1 --no-weightp --ref 6
 old_town_cross_444_720p50.y4m,--preset slow --rdoq-level 1 --early-skip --ref 7 --no-b-pyramid
 old_town_cross_444_720p50.y4m,--preset slower --crf 4 --cu-lossless
@@ -150,6 +150,8 @@
 Kimono1_1920x1080_24_400.yuv,--preset medium --rdoq-level 0 --limit-refs 3 --slices 2
 Kimono1_1920x1080_24_400.yuv,--preset veryslow --crf 4 --cu-lossless --slices 2 --limit-refs 3 --limit-modes
 Kimono1_1920x1080_24_400.yuv,--preset placebo --ctu 32 --max-tu-size 8 --limit-tu 2
+big_buck_bunny_360p24.y4m, --keyint 60 --min-keyint 40 --gop-lookahead 14
+BasketballDrive_1920x1080_50.y4m, --preset medium --no-open-gop --keyint 50 --min-keyint 50 --radl 2
 
 # Main12 intraCost overflow bug test
 720p50_parkrun_ter.y4m,--preset medium

x265_2.6.tar.gz/source/x265.cpp -> x265_2.7.tar.gz/source/x265.cpp Changed

@@ -301,9 +301,15 @@
                 if (!this->qpfile)
                     x265_log_file(param, X265_LOG_ERROR, "%s qpfile not found or error in opening qp file\n", optarg);
             }
+            OPT("fullhelp")
+            {
+                param->logLevel = X265_LOG_FULL;
+                printVersion(param, api);
+                showHelp(param);
+                break;
+            }
             else
                 bError |= !!api->param_parse(param, long_options[long_options_index].name, optarg);
-
             if (bError)
             {
                 const char *name = long_options_index > 0 ? long_options[long_options_index].name : argv[optind - 2];
@@ -579,9 +585,9 @@
 
     x265_picture pic_orig, pic_out;
     x265_picture *pic_in = &pic_orig;
-    /* Allocate recon picture if analysisReuseMode is enabled */
+    /* Allocate recon picture if analysis save/load is enabled */
     std::priority_queue<int64_t>* pts_queue = cliopt.output->needPTS() ? new std::priority_queue<int64_t>() : NULL;
-    x265_picture *pic_recon = (cliopt.recon || !!param->analysisReuseMode || pts_queue || reconPlay || param->csvLogLevel) ? &pic_out : NULL;
+    x265_picture *pic_recon = (cliopt.recon || param->analysisSave || param->analysisLoad || pts_queue || reconPlay || param->csvLogLevel) ? &pic_out : NULL;
     uint32_t inFrameCount = 0;
     uint32_t outFrameCount = 0;
     x265_nal *p_nal;

x265_2.6.tar.gz/source/x265.h -> x265_2.7.tar.gz/source/x265.h Changed

@@ -327,15 +327,15 @@
      * to allow the encoder to determine base QP */
     int     forceqp;
 
-    /* If param.analysisReuseMode is X265_ANALYSIS_OFF this field is ignored on input
-     * and output. Else the user must call x265_alloc_analysis_data() to
-     * allocate analysis buffers for every picture passed to the encoder.
+    /* If param.analysisLoad and param.analysisSave are disabled, this field is
+     * ignored on input and output. Else the user must call x265_alloc_analysis_data()
+     * to allocate analysis buffers for every picture passed to the encoder.
      *
-     * On input when param.analysisReuseMode is X265_ANALYSIS_LOAD and analysisData
+     * On input when param.analysisLoad is enabled and analysisData
      * member pointers are valid, the encoder will use the data stored here to
      * reduce encoder work.
      *
-     * On output when param.analysisReuseMode is X265_ANALYSIS_SAVE and analysisData
+     * On output when param.analysisSave is enabled and analysisData
      * member pointers are valid, the encoder will write output analysis into
      * this data structure */
     x265_analysis_data analysisData;
@@ -481,9 +481,7 @@
 #define X265_CSP_BGRA           7  /* packed bgr 32bits   */
 #define X265_CSP_RGB            8  /* packed rgb 24bits   */
 #define X265_CSP_MAX            9  /* end of list */
-
 #define X265_EXTENDED_SAR       255 /* aspect ratio explicitly specified as width:height */
-
 /* Analysis options */
 #define X265_ANALYSIS_OFF  0
 #define X265_ANALYSIS_SAVE 1
@@ -1129,13 +1127,13 @@
      * Default disabled */
     int       bEnableRdRefine;
 
-    /* If X265_ANALYSIS_SAVE, write per-frame analysis information into analysis
-     * buffers.  if X265_ANALYSIS_LOAD, read analysis information into analysis
-     * buffer and use this analysis information to reduce the amount of work
-     * the encoder must perform. Default X265_ANALYSIS_OFF */
+    /* If save, write per-frame analysis information into analysis buffers.
+     * If load, read analysis information into analysis buffer and use this
+     * analysis information to reduce the amount of work the encoder must perform.
+     * Default disabled. Now deprecated*/
     int       analysisReuseMode;
 
-    /* Filename for analysisReuseMode save/load. Default name is "x265_analysis.dat" */
+    /* Filename for multi-pass-opt-analysis/distortion. Default name is "x265_analysis.dat" */
     const char* analysisReuseFileName;
 
     /*== Rate Control ==*/
@@ -1273,6 +1271,7 @@
 
         /* internally enable if tune grain is set */
         int      bEnableConstVbv;
+
     } rc;
 
     /*== Video Usability Information ==*/
@@ -1455,7 +1454,7 @@
     int       bHDROpt;
 
     /* A value between 1 and 10 (both inclusive) determines the level of
-    * information stored/reused in save/load analysis-reuse-mode. Higher the refine
+    * information stored/reused in analysis save/load. Higher the refine
     * level higher the information stored/reused. Default is 5 */
     int       analysisReuseLevel;
 
@@ -1532,9 +1531,23 @@
 
     /* Reuse MV information obtained through API */
     int       bMVType;
-
     /* Allow the encoder to have a copy of the planes of x265_picture in Frame */
     int       bCopyPicToFrame;
+
+    /*Number of frames for GOP boundary decision lookahead.If a scenecut frame is found
+    * within this from the gop boundary set by keyint, the GOP will be extented until such a point,
+    * otherwise the GOP will be terminated as set by keyint*/
+    int       gopLookahead;
+
+    /*Write per-frame analysis information into analysis buffers. Default disabled. */
+    const char* analysisSave;
+
+    /* Read analysis information into analysis buffer and use this analysis information
+     * to reduce the amount of work the encoder must perform. Default disabled. */
+    const char* analysisLoad;
+
+    /*Number of RADL pictures allowed in front of IDR*/
+    int radl;
 } x265_param;
 
 /* x265_param_alloc:
@@ -1743,7 +1756,7 @@
 /* x265_get_ref_frame_list:
  *     returns negative on error, 0 when access unit were output.
  *     This API must be called after(poc >= lookaheadDepth + bframes + 2) condition check */
-int x265_get_ref_frame_list(x265_encoder *encoder, x265_picyuv**, x265_picyuv**, int, int);
+int x265_get_ref_frame_list(x265_encoder *encoder, x265_picyuv**, x265_picyuv**, int, int, int*, int*);
 
 /* x265_set_analysis_data:
  *     set the analysis data. The incoming analysis_data structure is assumed to be AVC-sized blocks.
@@ -1766,9 +1779,10 @@
 void x265_csvlog_frame(const x265_param *, const x265_picture *);
 
 /* Log final encode statistics to the CSV file handle. 'argc' and 'argv' are
- * intended to be command line arguments passed to the encoder. Encode
+ * intended to be command line arguments passed to the encoder. padx and pady are
+ * padding offsets for conformance and can be given from sps settings. Encode
  * statistics should be queried from the encoder just prior to closing it. */
-void x265_csvlog_encode(x265_encoder *encoder, const x265_stats *, int argc, char** argv);
+void x265_csvlog_encode(const x265_param*, const x265_stats *, int padx, int pady, int argc, char** argv);
 
 /* In-place downshift from a bit-depth greater than 8 to a bit-depth of 8, using
  * the residual bits to dither each row. */
@@ -1820,10 +1834,10 @@
     int           (*encoder_intra_refresh)(x265_encoder*);
     int           (*encoder_ctu_info)(x265_encoder*, int, x265_ctu_info_t**);
     int           (*get_slicetype_poc_and_scenecut)(x265_encoder*, int*, int*, int*);
-    int           (*get_ref_frame_list)(x265_encoder*, x265_picyuv**, x265_picyuv**, int, int);
+    int           (*get_ref_frame_list)(x265_encoder*, x265_picyuv**, x265_picyuv**, int, int, int*, int*);
     FILE*         (*csvlog_open)(const x265_param*);
     void          (*csvlog_frame)(const x265_param*, const x265_picture*);
-    void          (*csvlog_encode)(x265_encoder*, const x265_stats*, int, char**);
+    void          (*csvlog_encode)(const x265_param*, const x265_stats *, int, int, int, char**);
     void          (*dither_image)(x265_picture*, int, int, int16_t*, int);
     int           (*set_analysis_data)(x265_encoder *encoder, x265_analysis_data *analysis_data, int poc, uint32_t cuBytes);
     /* add new pointers to the end, or increment X265_MAJOR_VERSION */

x265_2.6.tar.gz/source/x265cli.h -> x265_2.7.tar.gz/source/x265cli.h Changed

@@ -38,6 +38,7 @@
 static const struct option long_options[] =
 {
     { "help",                 no_argument, NULL, 'h' },
+    { "fullhelp",             no_argument, NULL, 0 },
     { "version",              no_argument, NULL, 'V' },
     { "asm",            required_argument, NULL, 0 },
     { "no-asm",               no_argument, NULL, 0 },
@@ -119,9 +120,11 @@
     { "open-gop",             no_argument, NULL, 0 },
     { "keyint",         required_argument, NULL, 'I' },
     { "min-keyint",     required_argument, NULL, 'i' },
+    { "gop-lookahead",  required_argument, NULL, 0 },
     { "scenecut",       required_argument, NULL, 0 },
     { "no-scenecut",          no_argument, NULL, 0 },
     { "scenecut-bias",  required_argument, NULL, 0 },
+    { "radl",           required_argument, NULL, 0 },
     { "ctu-info",       required_argument, NULL, 0 },
     { "intra-refresh",        no_argument, NULL, 0 },
     { "rc-lookahead",   required_argument, NULL, 0 },
@@ -252,9 +255,11 @@
     { "no-slow-firstpass",    no_argument, NULL, 0 },
     { "multi-pass-opt-rps",   no_argument, NULL, 0 },
     { "no-multi-pass-opt-rps", no_argument, NULL, 0 },
-    { "analysis-reuse-mode",  required_argument, NULL, 0 },
-    { "analysis-reuse-file",  required_argument, NULL, 0 },
+    { "analysis-reuse-mode", required_argument, NULL, 0 }, /* DEPRECATED */
+    { "analysis-reuse-file", required_argument, NULL, 0 },
     { "analysis-reuse-level", required_argument, NULL, 0 },
+    { "analysis-save",  required_argument, NULL, 0 },
+    { "analysis-load",  required_argument, NULL, 0 },
     { "scale-factor",   required_argument, NULL, 0 },
     { "refine-intra",   required_argument, NULL, 0 },
     { "refine-inter",   required_argument, NULL, 0 },
@@ -314,6 +319,7 @@
     H0("    outfile is raw HEVC bitstream\n");
     H0("\nExecutable Options:\n");
     H0("-h/--help                        Show this help text and exit\n");
+    H0("   --fullhelp                    Show all options and exit\n");
     H0("-V/--version                     Show version info and exit\n");
     H0("\nOutput Options:\n");
     H0("-o/--output <filename>           Bitstream output file name\n");
@@ -418,9 +424,11 @@
     H0("   --[no-]open-gop               Enable open-GOP, allows I slices to be non-IDR. Default %s\n", OPT(param->bOpenGOP));
     H0("-I/--keyint <integer>            Max IDR period in frames. -1 for infinite-gop. Default %d\n", param->keyframeMax);
     H0("-i/--min-keyint <integer>        Scenecuts closer together than this are coded as I, not IDR. Default: auto\n");
+    H0("   --gop-lookahead <integer>     Extends gop boundary if a scenecut is found within this from keyint boundary. Default 0\n");
     H0("   --no-scenecut                 Disable adaptive I-frame decision\n");
     H0("   --scenecut <integer>          How aggressively to insert extra I-frames. Default %d\n", param->scenecutThreshold);
     H1("   --scenecut-bias <0..100.0>    Bias for scenecut detection. Default %.2f\n", param->scenecutBias);
+    H0("   --radl <integer>              Number of RADL pictures allowed in front of IDR. Default %d\n", param->radl);
     H0("   --intra-refresh               Use Periodic Intra Refresh instead of IDR frames\n");
     H0("   --rc-lookahead <integer>      Number of frames for frame-type lookahead (determines encoder latency) Default %d\n", param->lookaheadDepth);
     H1("   --lookahead-slices <0..16>    Number of slices to use per lookahead cost estimate. Default %d\n", param->lookaheadSlices);
@@ -461,18 +469,19 @@
     H0("   --[no-]analyze-src-pics       Motion estimation uses source frame planes. Default disable\n");
     H0("   --[no-]slow-firstpass         Enable a slow first pass in a multipass rate control mode. Default %s\n", OPT(param->rc.bEnableSlowFirstPass));
     H0("   --[no-]strict-cbr             Enable stricter conditions and tolerance for bitrate deviations in CBR mode. Default %s\n", OPT(param->rc.bStrictCbr));
-    H0("   --analysis-reuse-mode <string|int>  save - Dump analysis info into file, load - Load analysis buffers from the file. Default %d\n", param->analysisReuseMode);
+    H0("   --analysis-save <filename>    Dump analysis info into the specified file. Default Disabled\n");
+    H0("   --analysis-load <filename>    Load analysis buffers from the file specified. Default Disabled\n");
     H0("   --analysis-reuse-file <filename>    Specify file name used for either dumping or reading analysis data. Deault x265_analysis.dat\n");
     H0("   --analysis-reuse-level <1..10>      Level of analysis reuse indicates amount of info stored/reused in save/load mode, 1:least..10:most. Default %d\n", param->analysisReuseLevel);
     H0("   --refine-mv-type <string>     Reuse MV information received through API call. Supported option is avc. Default disabled - %d\n", param->bMVType);
     H0("   --scale-factor <int>          Specify factor by which input video is scaled down for analysis save mode. Default %d\n", param->scaleFactor);
-    H0("   --refine-intra <0..3>         Enable intra refinement for encode that uses analysis-reuse-mode=load.\n"
+    H0("   --refine-intra <0..3>         Enable intra refinement for encode that uses analysis-load.\n"
         "                                    - 0 : Forces both mode and depth from the save encode.\n"
         "                                    - 1 : Functionality of (0) + evaluate all intra modes at min-cu-size's depth when current depth is one smaller than min-cu-size's depth.\n"
         "                                    - 2 : Functionality of (1) + irrespective of size evaluate all angular modes when the save encode decides the best mode as angular.\n"
         "                                    - 3 : Functionality of (1) + irrespective of size evaluate all intra modes.\n"
         "                                Default:%d\n", param->intraRefine);
-    H0("   --refine-inter <0..3>         Enable inter refinement for encode that uses analysis-reuse-mode=load.\n"
+    H0("   --refine-inter <0..3>         Enable inter refinement for encode that uses analysis-load.\n"
         "                                    - 0 : Forces both mode and depth from the save encode.\n"
         "                                    - 1 : Functionality of (0) + evaluate all inter modes at min-cu-size's depth when current depth is one smaller than\n"
         "                                          min-cu-size's depth. When save encode decides the current block as skip(for all sizes) evaluate skip/merge.\n"
@@ -563,9 +572,8 @@
 #undef OPT
 #undef H0
 #undef H1
-
     if (level < X265_LOG_DEBUG)
-        printf("\nUse --log-level full --help for a full listing\n");
+        printf("\nUse --fullhelp for a full listing (or --log-level full --help)\n");
     printf("\n\nComplete documentation may be found at http://x265.readthedocs.org/en/default/cli.html\n");
     exit(1);
 }