Request 1963 (accepted)

Overview

Request 1963 (accepted)

Updated to 1.7

Created by Aloysius over 9 years ago
In state accepted

Submit x265

Submit package home:Aloysius:branches:Essentials / x265 to package Essentials / x265

x265.changes Changed

@@ -1,4 +1,45 @@
 -------------------------------------------------------------------
+Fri May 29 09:11:02 UTC 2015 - aloisio@gmx.com
+
+- soname bump to 59
+- Update to version 1.7
+  * large amount of assembly code optimizations
+  * some preliminary support for high dynamic range content
+  * improvements for multi-library support
+  * some new quality features
+    (full documentation at: http://x265.readthedocs.org/en/1.7)
+  * This release simplifies the multi-library support introduced
+    in version 1.6. Any libx265 can now forward API requests to
+    other installed libx265 libraries (by name) so applications
+    like ffmpeg and the x265 CLI can select between 8bit and 10bit
+    encodes at runtime without the need of a shim library or
+    library load path hacks. See --output-depth, and
+    http://x265.readthedocs.org/en/1.7/api.html#multi-library-interface
+  * For quality, x265 now allows you to configure the quantization
+    group size smaller than the CTU size (for finer grained AQ
+    adjustments). See --qg-size.
+  * x265 now supports limited mid-encode reconfigure via a new public
+    method: x265_encoder_reconfig()
+  * For HDR, x265 now supports signaling the SMPTE 2084 color transfer
+    function, the SMPTE 2086 mastering display color primaries, and the
+    content light levels. See --master-display, --max-cll
+  * x265 will no longer emit any non-conformant bitstreams unless
+    --allow-non-conformance is specified.
+  * The x265 CLI now supports a simple encode preview feature. See
+    --recon-y4m-exec.
+  * The AnnexB NAL headers can now be configured off, via x265_param.bAnnexB
+    This is not configurable via the CLI because it is a function of the
+    muxer being used, and the CLI only supports raw output files. See
+    --annexb
+  Misc:
+  * --lossless encodes are now signaled as level 8.5
+  * --profile now has a -P short option
+  * The regression scripts used by x265 are now public, and can be found at:
+    https://bitbucket.org/sborho/test-harness
+  * x265's cmake scripts now support PGO builds, the test-harness can be
+    used to drive the profile-guided build process.
+
+-------------------------------------------------------------------
 Tue Apr 28 20:08:06 UTC 2015 - aloisio@gmx.com
 
 - soname bumped to 51

x265.spec Changed

baselibs.conf Changed

x265_1.6.tar.gz/.hg_archival.txt -> x265_1.7.tar.gz/.hg_archival.txt Changed

x265_1.6.tar.gz/.hgtags -> x265_1.7.tar.gz/.hgtags Changed

x265_1.6.tar.gz/doc/reST/api.rst -> x265_1.7.tar.gz/doc/reST/api.rst Changed

@@ -171,8 +171,26 @@
 	 *      how x265_encoder_open has changed the parameters.
 	 *      note that the data accessible through pointers in the returned param struct
 	 *      (e.g. filenames) should not be modified by the calling application. */
-	void x265_encoder_parameters(x265_encoder *, x265_param *);                                                                      
-
+	void x265_encoder_parameters(x265_encoder *, x265_param *);
+
+**x265_encoder_reconfig()** may be used to reconfigure encoder parameters mid-encode::
+
+	/* x265_encoder_reconfig:
+	 *       used to modify encoder parameters.
+	 *      various parameters from x265_param are copied.
+	 *      this takes effect immediately, on whichever frame is encoded next;
+	 *      returns 0 on success, negative on parameter validation error.
+	 *
+	 *      not all parameters can be changed; see the actual function for a
+	 *      detailed breakdown.  since not all parameters can be changed, moving
+	 *      from preset to preset may not always fully copy all relevant parameters,
+	 *      but should still work usably in practice. however, more so than for
+	 *      other presets, many of the speed shortcuts used in ultrafast cannot be
+	 *      switched out of; using reconfig to switch between ultrafast and other
+	 *      presets is not recommended without a more fine-grained breakdown of
+	 *      parameters to take this into account. */
+	int x265_encoder_reconfig(x265_encoder *, x265_param *);
+	
 Pictures
 ========
 
@@ -352,7 +370,7 @@
 Multi-library Interface
 =======================
 
-If your application might want to make a runtime selection between among
+If your application might want to make a runtime selection between
 a number of libx265 libraries (perhaps 8bpp and 16bpp), then you will
 want to use the multi-library interface.
 
@@ -370,13 +388,34 @@
      *   libx265 */
     const x265_api* x265_api_get(int bitDepth);
 
-The general idea is to request the API for the bitDepth you would prefer
-the encoder to use (8 or 10), and if that returns NULL you request the
-API for bitDepth=0, which returns the system default libx265.
-
 Note that using this multi-library API in your application is only the
-first step. Next your application must dynamically link to libx265 and
-then you must build and install a multi-lib configuration of libx265,
-which includes 8bpp and 16bpp builds of libx265 and a shim library which
-forwards x265_api_get() calls to the appropriate library using dynamic
-loading and binding.
+first step.
+
+Your application must link to one build of libx265 (statically or 
+dynamically) and this linked version of libx265 will support one 
+bit-depth (8 or 10 bits). 
+
+Your application must now request the API for the bitDepth you would 
+prefer the encoder to use (8 or 10). If the requested bitdepth is zero, 
+or if it matches the bitdepth of the system default libx265 (the 
+currently linked library), then this library will be used for encode.
+If you request a different bit-depth, the linked libx265 will attempt 
+to dynamically bind a shared library with a name appropriate for the 
+requested bit-depth:
+
+    8-bit:  libx265_main.dll
+    10-bit: libx265_main10.dll
+
+    (the shared library extension is obviously platform specific. On
+    Linux it is .so while on Mac it is .dylib)
+
+For example on Windows, one could package together an x265.exe
+statically linked against the 8bpp libx265 together with a
+libx265_main10.dll in the same folder, and this executable would be able
+to encode main and main10 bitstreams.
+
+On Linux, x265 packagers could install 8bpp static and shared libraries
+under the name libx265 (so all applications link against 8bpp libx265)
+and then also install libx265_main10.so (symlinked to its numbered solib).
+Thus applications which use x265_api_get() will be able to generate main
+or main10 bitstreams.

x265_1.6.tar.gz/doc/reST/cli.rst -> x265_1.7.tar.gz/doc/reST/cli.rst Changed

@@ -159,6 +159,13 @@
 	handled implicitly.
 
 	One may also directly supply the CPU capability bitmap as an integer.
+	
+	Note that by specifying this option you are overriding x265's CPU
+	detection and it is possible to do this wrong. You can cause encoder
+	crashes by specifying SIMD architectures which are not supported on
+	your CPU.
+
+	Default: auto-detected SIMD architectures
 
 .. option:: --frame-threads, -F <integer>
 
@@ -171,7 +178,7 @@
 	Over-allocation of frame threads will not improve performance, it
 	will generally just increase memory use.
 
-	**Values:** any value between 8 and 16. Default is 0, auto-detect
+	**Values:** any value between 0 and 16. Default is 0, auto-detect
 
 .. option:: --pools <string>, --numa-pools <string>
 
@@ -201,11 +208,11 @@
 	their node, they will not be allowed to migrate between nodes, but they
 	will be allowed to move between CPU cores within their node.
 
-	If the three pool features: :option:`--wpp` :option:`--pmode` and
-	:option:`--pme` are all disabled, then :option:`--pools` is ignored
-	and no thread pools are created.
+	If the four pool features: :option:`--wpp`, :option:`--pmode`,
+	:option:`--pme` and :option:`--lookahead-slices` are all disabled,
+	then :option:`--pools` is ignored and no thread pools are created.
 
-	If "none" is specified, then all three of the thread pool features are
+	If "none" is specified, then all four of the thread pool features are
 	implicitly disabled.
 
 	Multiple thread pools will be allocated for any NUMA node with more than
@@ -217,9 +224,22 @@
 	:option:`--frame-threads`.  The pools are used for WPP and for
 	distributed analysis and motion search.
 
+	On Windows, the native APIs offer sufficient functionality to
+	discover the NUMA topology and enforce the thread affinity that
+	libx265 needs (so long as you have not chosen to target XP or
+	Vista), but on POSIX systems it relies on libnuma for this
+	functionality. If your target POSIX system is single socket, then
+	building without libnuma is a perfectly reasonable option, as it
+	will have no effect on the runtime behavior. On a multiple-socket
+	system, a POSIX build of libx265 without libnuma will be less work
+	efficient. See :ref:`thread pools <pools>` for more detail.
+
 	Default "", one thread is allocated per detected hardware thread
 	(logical CPU cores) and one thread pool per NUMA node.
 
+	Note that the string value will need to be escaped or quoted to
+	protect against shell expansion on many platforms
+
 .. option:: --wpp, --no-wpp
 
 	Enable Wavefront Parallel Processing. The encoder may begin encoding
@@ -399,10 +419,20 @@
 
 	**CLI ONLY**
 
+.. option:: --output-depth, -D 8|10
+
+	Bitdepth of output HEVC bitstream, which is also the internal bit
+	depth of the encoder. If the requested bit depth is not the bit
+	depth of the linked libx265, it will attempt to bind libx265_main
+	for an 8bit encoder, or libx265_main10 for a 10bit encoder, with the
+	same API version as the linked libx265.
+
+	**CLI ONLY**
+
 Profile, Level, Tier
 ====================
 
-.. option:: --profile <string>
+.. option:: --profile, -P <string>
 
 	Enforce the requirements of the specified profile, ensuring the
 	output stream will be decodable by a decoder which supports that
@@ -437,7 +467,7 @@
 	times 10, for example level **5.1** is specified as "5.1" or "51",
 	and level **5.0** is specified as "5.0" or "50".
 
-	Annex A levels: 1, 2, 2.1, 3, 3.1, 4, 4.1, 5, 5.1, 5.2, 6, 6.1, 6.2
+	Annex A levels: 1, 2, 2.1, 3, 3.1, 4, 4.1, 5, 5.1, 5.2, 6, 6.1, 6.2, 8.5
 
 .. option:: --high-tier, --no-high-tier
 
@@ -464,11 +494,22 @@
 	HEVC specification.  If x265 detects that the total reference count
 	is greater than 8, it will issue a warning that the resulting stream
 	is non-compliant and it signals the stream as profile NONE and level
-	NONE but still allows the encode to continue.  Compliant HEVC
+	NONE and will abort the encode unless
+	:option:`--allow-non-conformance` it specified.  Compliant HEVC
 	decoders may refuse to decode such streams.
 	
 	Default 3
 
+.. option:: --allow-non-conformance, --no-allow-non-conformance
+
+	Allow libx265 to generate a bitstream with profile and level NONE.
+	By default it will abort any encode which does not meet strict level
+	compliance. The two most likely causes for non-conformance are
+	:option:`--ctu` being too small, :option:`--ref` being too high,
+	or the bitrate or resolution being out of specification.
+
+	Default: disabled
+
 .. note::
 	:option:`--profile`, :option:`--level-idc`, and
 	:option:`--high-tier` are only intended for use when you are
@@ -476,7 +517,7 @@
 	limitations and must constrain the bitstream within those limits.
 	Specifying a profile or level may lower the encode quality
 	parameters to meet those requirements but it will never raise
-	them.
+	them. It may enable VBV constraints on a CRF encode.
 
 Mode decision / Analysis
 ========================
@@ -1111,6 +1152,14 @@
 
 	**Range of values:** 0.0 to 3.0
 
+.. option:: --qg-size <64|32|16>
+
+	Enable adaptive quantization for sub-CTUs. This parameter specifies 
+	the minimum CU size at which QP can be adjusted, ie. Quantization Group
+	size. Allowed range of values are 64, 32, 16 provided this falls within 
+	the inclusive range [maxCUSize, minCUSize]. Experimental.
+	Default: same as maxCUSize
+
 .. option:: --cutree, --no-cutree
 
 	Enable the use of lookahead's lowres motion vector fields to
@@ -1162,12 +1211,12 @@
 .. option:: --strict-cbr, --no-strict-cbr
 	
 	Enables stricter conditions to control bitrate deviance from the 
-	target bitrate in CBR mode. Bitrate adherence is prioritised
+	target bitrate in ABR mode. Bit rate adherence is prioritised
 	over quality. Rate tolerance is reduced to 50%. Default disabled.
 	
 	This option is for use-cases which require the final average bitrate 
-	to be within very strict limits of the target - preventing overshoots 
-	completely, and achieve bitrates within 5% of target bitrate, 
+	to be within very strict limits of the target; preventing overshoots, 
+	while keeping the bit rate within 5% of the target setting, 
 	especially in short segment encodes. Typically, the encoder stays 
 	conservative, waiting until there is enough feedback in terms of 
 	encoded frames to control QP. strict-cbr allows the encoder to be 
@@ -1209,7 +1258,7 @@
 	lookahead).  Default value is 0.6. Increasing it to 1 will
 	effectively generate CQP
 
-.. option:: --qstep <integer>
+.. option:: --qpstep <integer>
 
 	The maximum single adjustment in QP allowed to rate control. Default
 	4
@@ -1451,9 +1500,48 @@
 	specification for a description of these values. Default undefined
 	(not signaled)
 
+.. option:: --master-display <string>
+
+	SMPTE ST 2086 mastering display color volume SEI info, specified as
+	a string which is parsed when the stream header SEI are emitted. The
+	string format is "G(%hu,%hu)B(%hu,%hu)R(%hu,%hu)WP(%hu,%hu)L(%u,%u)"
+	where %hu are unsigned 16bit integers and %u are unsigned 32bit
+	integers. The SEI includes X,Y display primaries for RGB channels,
+	white point X,Y and max,min luminance values. (HDR)
+
+	Example for P65D3 1000-nits:
+
+		G(13200,34500)B(7500,3000)R(34000,16000)WP(15635,16450)L(10000000,1)
+
+	Note that this string value will need to be escaped or quoted to
+	protect against shell expansion on many platforms. No default.
+
+.. option:: --max-cll <string>
+
+	Maximum content light level and maximum frame average light level as
+	required by the Consumer Electronics Association 861.3 specification.
+
+	Specified as a string which is parsed when the stream header SEI are
+	emitted. The string format is "%hu,%hu" where %hu are unsigned 16bit
+	integers. The first value is the max content light level (or 0 if no
+	maximum is indicated), the second value is the maximum picture
+	average light level (or 0). (HDR)
+
+	Note that this string value will need to be escaped or quoted to
+	protect against shell expansion on many platforms. No default.
+
 Bitstream options
 =================
 
+.. option:: --annexb, --no-annexb
+
+	If enabled, x265 will produce Annex B bitstream format, which places
+	start codes before NAL. If disabled, x265 will produce file format,
+	which places length before NAL. x265 CLI will choose the right option
+	based on output format. Default enabled
+
+	**API ONLY**
+
 .. option:: --repeat-headers, --no-repeat-headers
 
 	If enabled, x265 will emit VPS, SPS, and PPS headers with every
@@ -1498,8 +1586,8 @@
 
 	Enable a temporal sub layer. All referenced I/P/B frames are in the
 	base layer and all unreferenced B frames are placed in a temporal
-	sublayer. A decoder may chose to drop the sublayer and only decode
-	and display the base layer slices.
+	enhancement layer. A decoder may chose to drop the enhancement layer 
+	and only decode and display the base layer slices.
 	
 	If used with a fixed GOP (:option:`b-adapt` 0) and :option:`bframes`
 	3 then the two layers evenly split the frame rate, with a cadence of
@@ -1525,4 +1613,20 @@
 
 	**CLI ONLY**
 
+.. option:: --recon-y4m-exec <string>
+
+	If you have an application which can play a Y4MPEG stream received
+	on stdin, the x265 CLI can feed it reconstructed pictures in display
+	order.  The pictures will have no timing info, obviously, so the
+	picture timing will be determined primarily by encoding elapsed time
+	and latencies, but it can be useful to preview the pictures being
+	output by the encoder to validate input settings and rate control
+	parameters.
+
+	Example command for ffplay (assuming it is in your PATH):
+
+	--recon-y4m-exec "ffplay -i pipe:0 -autoexit"
+
+	**CLI ONLY**
+
 .. vim: noet

x265_1.6.tar.gz/doc/reST/threading.rst -> x265_1.7.tar.gz/doc/reST/threading.rst Changed

@@ -2,6 +2,8 @@
 Threading
 *********
 
+.. _pools:
+
 Thread Pools
 ============
 
@@ -31,6 +33,18 @@
 expected to drop that job so the worker thread may go back to the pool
 and find more work.
 
+On Windows, the native APIs offer sufficient functionality to discover
+the NUMA topology and enforce the thread affinity that libx265 needs (so
+long as you have not chosen to target XP or Vista), but on POSIX systems
+it relies on libnuma for this functionality. If your target POSIX system
+is single socket, then building without libnuma is a perfectly
+reasonable option, as it will have no effect on the runtime behavior. On
+a multiple-socket system, a POSIX build of libx265 without libnuma will
+be less work efficient, but will still function correctly. You lose the
+work isolation effect that keeps each frame encoder from only using the
+threads of a single socket and so you incur a heavier context switching
+cost.
+
 Wavefront Parallel Processing
 =============================
 
@@ -225,6 +239,7 @@
 lowres cost analysis to worker threads. It will use bonded task groups
 to perform batches of frame cost estimates, and it may optionally use
 bonded task groups to measure single frame cost estimates using slices.
+(see :option:`--lookahead-slices`)
 
 The function slicetypeDecide() itself is also be performed by a worker
 thread if your encoder has a thread pool, else it runs within the

x265_1.6.tar.gz/readme.rst -> x265_1.7.tar.gz/readme.rst Changed

x265_1.6.tar.gz/source/CMakeLists.txt -> x265_1.7.tar.gz/source/CMakeLists.txt Changed

@@ -30,7 +30,7 @@
 mark_as_advanced(FPROFILE_USE FPROFILE_GENERATE NATIVE_BUILD)
 
 # X265_BUILD must be incremented each time the public API is changed
-set(X265_BUILD 51)
+set(X265_BUILD 59)
 configure_file("${PROJECT_SOURCE_DIR}/x265.def.in"
                "${PROJECT_BINARY_DIR}/x265.def")
 configure_file("${PROJECT_SOURCE_DIR}/x265_config.h.in"
@@ -65,15 +65,19 @@
     if(LIBRT)
         list(APPEND PLATFORM_LIBS rt)
     endif()
+    find_library(LIBDL dl)
+    if(LIBDL)
+        list(APPEND PLATFORM_LIBS dl)
+    endif()
     find_package(Numa)
     if(NUMA_FOUND)
-        list(APPEND CMAKE_REQUIRED_LIBRARIES ${NUMA_LIBRARY})
+        link_directories(${NUMA_LIBRARY_DIR})
+        list(APPEND CMAKE_REQUIRED_LIBRARIES numa)
         check_symbol_exists(numa_node_of_cpu numa.h NUMA_V2)
         if(NUMA_V2)
             add_definitions(-DHAVE_LIBNUMA)
             message(STATUS "libnuma found, building with support for NUMA nodes")
-            list(APPEND PLATFORM_LIBS ${NUMA_LIBRARY})
-            link_directories(${NUMA_LIBRARY_DIR})
+            list(APPEND PLATFORM_LIBS numa)
             include_directories(${NUMA_INCLUDE_DIR})
         endif()
     endif()
@@ -90,7 +94,7 @@
 if(CMAKE_GENERATOR STREQUAL "Xcode")
   set(XCODE 1)
 endif()
-if (APPLE)
+if(APPLE)
   add_definitions(-DMACOS)
 endif()
 
@@ -196,6 +200,7 @@
         add_definitions(-static)
         list(APPEND LINKER_OPTIONS "-static")
     endif(STATIC_LINK_CRT)
+    check_cxx_compiler_flag(-Wno-strict-overflow CC_HAS_NO_STRICT_OVERFLOW)
     check_cxx_compiler_flag(-Wno-narrowing CC_HAS_NO_NARROWING) 
     check_cxx_compiler_flag(-Wno-array-bounds CC_HAS_NO_ARRAY_BOUNDS) 
     if (CC_HAS_NO_ARRAY_BOUNDS)
@@ -291,7 +296,7 @@
     endif()
 endif(WARNINGS_AS_ERRORS)
 
-if (WIN32)
+if(WIN32)
     # Visual leak detector
     find_package(VLD QUIET)
     if(VLD_FOUND)
@@ -300,12 +305,15 @@
         list(APPEND PLATFORM_LIBS ${VLD_LIBRARIES})
         link_directories(${VLD_LIBRARY_DIRS})
     endif()
-    option(WINXP_SUPPORT "Make binaries compatible with Windows XP" OFF)
+    option(WINXP_SUPPORT "Make binaries compatible with Windows XP and Vista" OFF)
     if(WINXP_SUPPORT)
         # force use of workarounds for CONDITION_VARIABLE and atomic
         # intrinsics introduced after XP
-        add_definitions(-D_WIN32_WINNT=_WIN32_WINNT_WINXP)
-    endif()
+        add_definitions(-D_WIN32_WINNT=_WIN32_WINNT_WINXP -D_WIN32_WINNT_WIN7=0x0601)
+    else(WINXP_SUPPORT)
+        # default to targeting Windows 7 for the NUMA APIs
+        add_definitions(-D_WIN32_WINNT=_WIN32_WINNT_WIN7)
+    endif(WINXP_SUPPORT)
 endif()
 
 include(version) # determine X265_VERSION and X265_LATEST_TAG
@@ -462,8 +470,10 @@
 # Main CLI application
 option(ENABLE_CLI "Build standalone CLI application" ON)
 if(ENABLE_CLI)
-    file(GLOB InputFiles input/*.cpp input/*.h)
-    file(GLOB OutputFiles output/*.cpp output/*.h)
+    file(GLOB InputFiles input/input.cpp input/yuv.cpp input/y4m.cpp input/*.h)
+    file(GLOB OutputFiles output/output.cpp output/reconplay.cpp output/*.h
+                          output/yuv.cpp output/y4m.cpp # recon
+                          output/raw.cpp)               # muxers
     file(GLOB FilterFiles filters/*.cpp filters/*.h)
     source_group(input FILES ${InputFiles})
     source_group(output FILES ${OutputFiles})

x265_1.6.tar.gz/source/common/common.cpp -> x265_1.7.tar.gz/source/common/common.cpp Changed

x265_1.6.tar.gz/source/common/common.h -> x265_1.7.tar.gz/source/common/common.h Changed

x265_1.6.tar.gz/source/common/constants.cpp -> x265_1.7.tar.gz/source/common/constants.cpp Changed

x265_1.6.tar.gz/source/common/contexts.h -> x265_1.7.tar.gz/source/common/contexts.h Changed

x265_1.6.tar.gz/source/common/cudata.cpp -> x265_1.7.tar.gz/source/common/cudata.cpp Changed

@@ -298,7 +298,7 @@
 }
 
 // initialize Sub partition
-void CUData::initSubCU(const CUData& ctu, const CUGeom& cuGeom)
+void CUData::initSubCU(const CUData& ctu, const CUGeom& cuGeom, int qp)
 {
     m_absIdxInCTU   = cuGeom.absPartIdx;
     m_encData       = ctu.m_encData;
@@ -312,8 +312,8 @@
     m_cuAboveRight  = ctu.m_cuAboveRight;
     X265_CHECK(m_numPartitions == cuGeom.numPartitions, "initSubCU() size mismatch\n");
 
-    /* sequential memsets */
-    m_partSet((uint8_t*)m_qp, (uint8_t)ctu.m_qp[0]);
+    m_partSet((uint8_t*)m_qp, (uint8_t)qp);
+
     m_partSet(m_log2CUSize,   (uint8_t)cuGeom.log2CUSize);
     m_partSet(m_lumaIntraDir, (uint8_t)DC_IDX);
     m_partSet(m_tqBypass,     (uint8_t)m_encData->m_param->bLossless);
@@ -1830,6 +1830,10 @@
     }
 }
 
+/* Clip motion vector to within slightly padded boundary of picture (the
+ * MV may reference a block that is completely within the padded area).
+ * Note this function is unaware of how much of this picture is actually
+ * available for use (re: frame parallelism) */
 void CUData::clipMv(MV& outMV) const
 {
     const uint32_t mvshift = 2;
@@ -2027,6 +2031,7 @@
         uint32_t blockSize = 1 << log2CUSize;
         uint32_t sbWidth   = 1 << (g_log2Size[maxCUSize] - log2CUSize);
         int32_t lastLevelFlag = log2CUSize == g_log2Size[minCUSize];
+
         for (uint32_t sbY = 0; sbY < sbWidth; sbY++)
         {
             for (uint32_t sbX = 0; sbX < sbWidth; sbX++)

x265_1.6.tar.gz/source/common/cudata.h -> x265_1.7.tar.gz/source/common/cudata.h Changed

x265_1.6.tar.gz/source/common/dct.cpp -> x265_1.7.tar.gz/source/common/dct.cpp Changed

@@ -752,7 +752,7 @@
     }
 }
 
-int findPosLast_c(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig)
+int scanPosLast_c(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig, const uint16_t* /*scanCG4x4*/, const int /*trSize*/)
 {
     memset(coeffNum, 0, MLS_GRP_NUM * sizeof(*coeffNum));
     memset(coeffFlag, 0, MLS_GRP_NUM * sizeof(*coeffFlag));
@@ -785,6 +785,37 @@
     return scanPosLast - 1;
 }
 
+uint32_t findPosFirstLast_c(const int16_t *dstCoeff, const intptr_t trSize, const uint16_t scanTbl[16])
+{
+    int n;
+
+    for (n = SCAN_SET_SIZE - 1; n >= 0; --n)
+    {
+        const uint32_t idx = scanTbl[n];
+        const uint32_t idxY = idx / MLS_CG_SIZE;
+        const uint32_t idxX = idx % MLS_CG_SIZE;
+        if (dstCoeff[idxY * trSize + idxX])
+            break;
+    }
+
+    X265_CHECK(n >= 0, "non-zero coeff scan failuare!\n");
+
+    uint32_t lastNZPosInCG = (uint32_t)n;
+
+    for (n = 0;; n++)
+    {
+        const uint32_t idx = scanTbl[n];
+        const uint32_t idxY = idx / MLS_CG_SIZE;
+        const uint32_t idxX = idx % MLS_CG_SIZE;
+        if (dstCoeff[idxY * trSize + idxX])
+            break;
+    }
+
+    uint32_t firstNZPosInCG = (uint32_t)n;
+
+    return ((lastNZPosInCG << 16) | firstNZPosInCG);
+}
+
 }  // closing - anonymous file-static namespace
 
 namespace x265 {
@@ -817,6 +848,7 @@
     p.cu[BLOCK_16x16].copy_cnt = copy_count<16>;
     p.cu[BLOCK_32x32].copy_cnt = copy_count<32>;
 
-    p.findPosLast = findPosLast_c;
+    p.scanPosLast = scanPosLast_c;
+    p.findPosFirstLast = findPosFirstLast_c;
 }
 }

x265_1.6.tar.gz/source/common/frame.cpp -> x265_1.7.tar.gz/source/common/frame.cpp Changed

x265_1.6.tar.gz/source/common/frame.h -> x265_1.7.tar.gz/source/common/frame.h Changed

x265_1.6.tar.gz/source/common/framedata.h -> x265_1.7.tar.gz/source/common/framedata.h Changed

x265_1.6.tar.gz/source/common/ipfilter.cpp -> x265_1.7.tar.gz/source/common/ipfilter.cpp Changed

@@ -34,27 +34,8 @@
 #endif
 
 namespace {
-template<int dstStride, int width, int height>
-void pixelToShort_c(const pixel* src, intptr_t srcStride, int16_t* dst)
-{
-    int shift = IF_INTERNAL_PREC - X265_DEPTH;
-    int row, col;
-
-    for (row = 0; row < height; row++)
-    {
-        for (col = 0; col < width; col++)
-        {
-            int16_t val = src[col] << shift;
-            dst[col] = val - (int16_t)IF_INTERNAL_OFFS;
-        }
-
-        src += srcStride;
-        dst += dstStride;
-    }
-}
-
-template<int dstStride>
-void filterPixelToShort_c(const pixel* src, intptr_t srcStride, int16_t* dst, int width, int height)
+template<int width, int height>
+void filterPixelToShort_c(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride)
 {
     int shift = IF_INTERNAL_PREC - X265_DEPTH;
     int row, col;
@@ -398,7 +379,7 @@
     p.chroma[X265_CSP_I420].pu[CHROMA_420_ ## W ## x ## H].filter_vps = interp_vert_ps_c<4, W, H>;  \
     p.chroma[X265_CSP_I420].pu[CHROMA_420_ ## W ## x ## H].filter_vsp = interp_vert_sp_c<4, W, H>;  \
     p.chroma[X265_CSP_I420].pu[CHROMA_420_ ## W ## x ## H].filter_vss = interp_vert_ss_c<4, W, H>; \
-    p.chroma[X265_CSP_I420].pu[CHROMA_420_ ## W ## x ## H].chroma_p2s = pixelToShort_c<MAX_CU_SIZE / 2, W, H>; 
+    p.chroma[X265_CSP_I420].pu[CHROMA_420_ ## W ## x ## H].p2s = filterPixelToShort_c<W, H>;
 
 #define CHROMA_422(W, H) \
     p.chroma[X265_CSP_I422].pu[CHROMA_422_ ## W ## x ## H].filter_hpp = interp_horiz_pp_c<4, W, H>; \
@@ -407,7 +388,7 @@
     p.chroma[X265_CSP_I422].pu[CHROMA_422_ ## W ## x ## H].filter_vps = interp_vert_ps_c<4, W, H>;  \
     p.chroma[X265_CSP_I422].pu[CHROMA_422_ ## W ## x ## H].filter_vsp = interp_vert_sp_c<4, W, H>;  \
     p.chroma[X265_CSP_I422].pu[CHROMA_422_ ## W ## x ## H].filter_vss = interp_vert_ss_c<4, W, H>; \
-    p.chroma[X265_CSP_I422].pu[CHROMA_422_ ## W ## x ## H].chroma_p2s = pixelToShort_c<MAX_CU_SIZE / 2, W, H>; 
+    p.chroma[X265_CSP_I422].pu[CHROMA_422_ ## W ## x ## H].p2s = filterPixelToShort_c<W, H>;
 
 #define CHROMA_444(W, H) \
     p.chroma[X265_CSP_I444].pu[LUMA_ ## W ## x ## H].filter_hpp = interp_horiz_pp_c<4, W, H>; \
@@ -416,7 +397,7 @@
     p.chroma[X265_CSP_I444].pu[LUMA_ ## W ## x ## H].filter_vps = interp_vert_ps_c<4, W, H>;  \
     p.chroma[X265_CSP_I444].pu[LUMA_ ## W ## x ## H].filter_vsp = interp_vert_sp_c<4, W, H>;  \
     p.chroma[X265_CSP_I444].pu[LUMA_ ## W ## x ## H].filter_vss = interp_vert_ss_c<4, W, H>; \
-    p.chroma[X265_CSP_I444].pu[LUMA_ ## W ## x ## H].chroma_p2s = pixelToShort_c<MAX_CU_SIZE, W, H>; 
+    p.chroma[X265_CSP_I444].pu[LUMA_ ## W ## x ## H].p2s = filterPixelToShort_c<W, H>;
 
 #define LUMA(W, H) \
     p.pu[LUMA_ ## W ## x ## H].luma_hpp     = interp_horiz_pp_c<8, W, H>; \
@@ -426,7 +407,7 @@
     p.pu[LUMA_ ## W ## x ## H].luma_vsp     = interp_vert_sp_c<8, W, H>;  \
     p.pu[LUMA_ ## W ## x ## H].luma_vss     = interp_vert_ss_c<8, W, H>;  \
     p.pu[LUMA_ ## W ## x ## H].luma_hvpp    = interp_hv_pp_c<8, W, H>; \
-    p.pu[LUMA_ ## W ## x ## H].filter_p2s = pixelToShort_c<MAX_CU_SIZE, W, H>
+    p.pu[LUMA_ ## W ## x ## H].convert_p2s = filterPixelToShort_c<W, H>;
 
 void setupFilterPrimitives_c(EncoderPrimitives& p)
 {
@@ -482,6 +463,7 @@
 
     CHROMA_422(4, 8);
     CHROMA_422(4, 4);
+    CHROMA_422(2, 4);
     CHROMA_422(2, 8);
     CHROMA_422(8,  16);
     CHROMA_422(8,  8);
@@ -530,11 +512,6 @@
     CHROMA_444(48, 64);
     CHROMA_444(64, 16);
     CHROMA_444(16, 64);
-    p.luma_p2s = filterPixelToShort_c<MAX_CU_SIZE>;
-
-    p.chroma[X265_CSP_I444].p2s = filterPixelToShort_c<MAX_CU_SIZE>;
-    p.chroma[X265_CSP_I420].p2s = filterPixelToShort_c<MAX_CU_SIZE / 2>;
-    p.chroma[X265_CSP_I422].p2s = filterPixelToShort_c<MAX_CU_SIZE / 2>;
 
     p.extendRowBorder = extendCURowColBorder;
 }

x265_1.6.tar.gz/source/common/loopfilter.cpp -> x265_1.7.tar.gz/source/common/loopfilter.cpp Changed

@@ -42,18 +42,23 @@
         dst[x] = signOf(src1[x] - src2[x]);
 }
 
-void processSaoCUE0(pixel * rec, int8_t * offsetEo, int width, int8_t signLeft)
+void processSaoCUE0(pixel * rec, int8_t * offsetEo, int width, int8_t* signLeft, intptr_t stride)
 {
-    int x;
-    int8_t signRight;
+    int x, y;
+    int8_t signRight, signLeft0;
     int8_t edgeType;
 
-    for (x = 0; x < width; x++)
+    for (y = 0; y < 2; y++)
     {
-        signRight = ((rec[x] - rec[x + 1]) < 0) ? -1 : ((rec[x] - rec[x + 1]) > 0) ? 1 : 0;
-        edgeType = signRight + signLeft + 2;
-        signLeft  = -signRight;
-        rec[x] = x265_clip(rec[x] + offsetEo[edgeType]);
+        signLeft0 = signLeft[y];
+        for (x = 0; x < width; x++)
+        {
+            signRight = ((rec[x] - rec[x + 1]) < 0) ? -1 : ((rec[x] - rec[x + 1]) > 0) ? 1 : 0;
+            edgeType = signRight + signLeft0 + 2;
+            signLeft0 = -signRight;
+            rec[x] = x265_clip(rec[x] + offsetEo[edgeType]);
+        }
+        rec += stride;
     }
 }
 
@@ -72,6 +77,25 @@
     }
 }
 
+void processSaoCUE1_2Rows(pixel* rec, int8_t* upBuff1, int8_t* offsetEo, intptr_t stride, int width)
+{
+    int x, y;
+    int8_t signDown;
+    int edgeType;
+
+    for (y = 0; y < 2; y++)
+    {
+        for (x = 0; x < width; x++)
+        {
+            signDown = signOf(rec[x] - rec[x + stride]);
+            edgeType = signDown + upBuff1[x] + 2;
+            upBuff1[x] = -signDown;
+            rec[x] = x265_clip(rec[x] + offsetEo[edgeType]);
+        }
+        rec += stride;
+    }
+}
+
 void processSaoCUE2(pixel * rec, int8_t * bufft, int8_t * buff1, int8_t * offsetEo, int width, intptr_t stride)
 {
     int x;
@@ -119,8 +143,11 @@
 {
     p.saoCuOrgE0 = processSaoCUE0;
     p.saoCuOrgE1 = processSaoCUE1;
-    p.saoCuOrgE2 = processSaoCUE2;
-    p.saoCuOrgE3 = processSaoCUE3;
+    p.saoCuOrgE1_2Rows = processSaoCUE1_2Rows;
+    p.saoCuOrgE2[0] = processSaoCUE2;
+    p.saoCuOrgE2[1] = processSaoCUE2;
+    p.saoCuOrgE3[0] = processSaoCUE3;
+    p.saoCuOrgE3[1] = processSaoCUE3;
     p.saoCuOrgB0 = processSaoCUB0;
     p.sign = calSign;
 }

x265_1.6.tar.gz/source/common/param.cpp -> x265_1.7.tar.gz/source/common/param.cpp Changed

@@ -87,7 +87,7 @@
 extern "C"
 void x265_param_free(x265_param* p)
 {
-    return x265_free(p);
+    x265_free(p);
 }
 
 extern "C"
@@ -117,6 +117,7 @@
     param->levelIdc = 0;
     param->bHighTier = 0;
     param->interlaceMode = 0;
+    param->bAnnexB = 1;
     param->bRepeatHeaders = 0;
     param->bEnableAccessUnitDelimiters = 0;
     param->bEmitHRDSEI = 0;
@@ -209,6 +210,7 @@
     param->rc.zones = NULL;
     param->rc.bEnableSlowFirstPass = 0;
     param->rc.bStrictCbr = 0;
+    param->rc.qgSize = 64; /* Same as maxCUSize */
 
     /* Video Usability Information (VUI) */
     param->vui.aspectRatioIdc = 0;
@@ -263,6 +265,7 @@
             param->rc.aqStrength = 0.0;
             param->rc.aqMode = X265_AQ_NONE;
             param->rc.cuTree = 0;
+            param->rc.qgSize = 32;
             param->bEnableFastIntra = 1;
         }
         else if (!strcmp(preset, "superfast"))
@@ -279,6 +282,7 @@
             param->rc.aqStrength = 0.0;
             param->rc.aqMode = X265_AQ_NONE;
             param->rc.cuTree = 0;
+            param->rc.qgSize = 32;
             param->bEnableSAO = 0;
             param->bEnableFastIntra = 1;
         }
@@ -292,6 +296,7 @@
             param->rdLevel = 2;
             param->maxNumReferences = 1;
             param->rc.cuTree = 0;
+            param->rc.qgSize = 32;
             param->bEnableFastIntra = 1;
         }
         else if (!strcmp(preset, "faster"))
@@ -565,6 +570,7 @@
             p->levelIdc = atoi(value);
     }
     OPT("high-tier") p->bHighTier = atobool(value);
+    OPT("allow-non-conformance") p->bAllowNonConformance = atobool(value);
     OPT2("log-level", "log")
     {
         p->logLevel = atoi(value);
@@ -575,6 +581,7 @@
         }
     }
     OPT("cu-stats") p->bLogCuStats = atobool(value);
+    OPT("annexb") p->bAnnexB = atobool(value);
     OPT("repeat-headers") p->bRepeatHeaders = atobool(value);
     OPT("wpp") p->bEnableWavefront = atobool(value);
     OPT("ctu") p->maxCUSize = (uint32_t)atoi(value);
@@ -843,6 +850,9 @@
     OPT2("pools", "numa-pools") p->numaPools = strdup(value);
     OPT("lambda-file") p->rc.lambdaFileName = strdup(value);
     OPT("analysis-file") p->analysisFileName = strdup(value);
+    OPT("qg-size") p->rc.qgSize = atoi(value);
+    OPT("master-display") p->masteringDisplayColorVolume = strdup(value);
+    OPT("max-cll") p->contentLightLevelInfo = strdup(value);
     else
         return X265_PARAM_BAD_NAME;
 #undef OPT
@@ -1183,7 +1193,7 @@
     uint32_t maxLog2CUSize = (uint32_t)g_log2Size[param->maxCUSize];
     uint32_t minLog2CUSize = (uint32_t)g_log2Size[param->minCUSize];
 
-    if (g_ctuSizeConfigured || ATOMIC_INC(&g_ctuSizeConfigured) > 1)
+    if (ATOMIC_INC(&g_ctuSizeConfigured) > 1)
     {
         if (g_maxCUSize != param->maxCUSize)
         {
@@ -1264,22 +1274,20 @@
     x265_log(param, X265_LOG_INFO, "b-pyramid / weightp / weightb / refs: %d / %d / %d / %d\n",
              param->bBPyramid, param->bEnableWeightedPred, param->bEnableWeightedBiPred, param->maxNumReferences);
 
+    if (param->rc.aqMode)
+        x265_log(param, X265_LOG_INFO, "AQ: mode / str / qg-size / cu-tree  : %d / %0.1f / %d / %d\n", param->rc.aqMode,
+                 param->rc.aqStrength, param->rc.qgSize, param->rc.cuTree);
+
     if (param->bLossless)
         x265_log(param, X265_LOG_INFO, "Rate Control                        : Lossless\n");
     else switch (param->rc.rateControlMode)
     {
     case X265_RC_ABR:
-        x265_log(param, X265_LOG_INFO, "Rate Control / AQ-Strength / CUTree : ABR-%d kbps / %0.1f / %d\n", param->rc.bitrate,
-                 param->rc.aqStrength, param->rc.cuTree);
-        break;
+        x265_log(param, X265_LOG_INFO, "Rate Control / qCompress            : ABR-%d kbps / %0.2f\n", param->rc.bitrate, param->rc.qCompress); break;
     case X265_RC_CQP:
-        x265_log(param, X265_LOG_INFO, "Rate Control / AQ-Strength / CUTree : CQP-%d / %0.1f / %d\n", param->rc.qp, param->rc.aqStrength,
-                 param->rc.cuTree);
-        break;
+        x265_log(param, X265_LOG_INFO, "Rate Control                        : CQP-%d\n", param->rc.qp); break;
     case X265_RC_CRF:
-        x265_log(param, X265_LOG_INFO, "Rate Control / AQ-Strength / CUTree : CRF-%0.1f / %0.1f / %d\n", param->rc.rfConstant,
-                 param->rc.aqStrength, param->rc.cuTree);
-        break;
+        x265_log(param, X265_LOG_INFO, "Rate Control / qCompress            : CRF-%0.1f / %0.2f\n", param->rc.rfConstant, param->rc.qCompress); break;
     }
 
     if (param->rc.vbvBufferSize)
@@ -1327,6 +1335,43 @@
     fflush(stderr);
 }
 
+void x265_print_reconfigured_params(x265_param* param, x265_param* reconfiguredParam)
+{
+    if (!param || !reconfiguredParam)
+        return;
+
+    x265_log(param,X265_LOG_INFO, "Reconfigured param options :\n");
+
+    char buf[80] = { 0 };
+    char tmp[40];
+#define TOOLCMP(COND1, COND2, STR, VAL)  if (COND1 != COND2) { sprintf(tmp, STR, VAL); appendtool(param, buf, sizeof(buf), tmp); }
+    TOOLCMP(param->maxNumReferences, reconfiguredParam->maxNumReferences, "ref=%d", reconfiguredParam->maxNumReferences);
+    TOOLCMP(param->maxTUSize, reconfiguredParam->maxTUSize, "max-tu-size=%d", reconfiguredParam->maxTUSize);
+    TOOLCMP(param->searchRange, reconfiguredParam->searchRange, "merange=%d", reconfiguredParam->searchRange);
+    TOOLCMP(param->subpelRefine, reconfiguredParam->subpelRefine, "subme= %d", reconfiguredParam->subpelRefine);
+    TOOLCMP(param->rdLevel, reconfiguredParam->rdLevel, "rd=%d", reconfiguredParam->rdLevel);
+    TOOLCMP(param->psyRd, reconfiguredParam->psyRd, "psy-rd=%.2lf", reconfiguredParam->psyRd);
+    TOOLCMP(param->rdoqLevel, reconfiguredParam->rdoqLevel, "rdoq=%d", reconfiguredParam->rdoqLevel);
+    TOOLCMP(param->psyRdoq, reconfiguredParam->psyRdoq, "psy-rdoq=%.2lf", reconfiguredParam->psyRdoq);
+    TOOLCMP(param->noiseReductionIntra, reconfiguredParam->noiseReductionIntra, "nr-intra=%d", reconfiguredParam->noiseReductionIntra);
+    TOOLCMP(param->noiseReductionInter, reconfiguredParam->noiseReductionInter, "nr-inter=%d", reconfiguredParam->noiseReductionInter);
+    TOOLCMP(param->bEnableTSkipFast, reconfiguredParam->bEnableTSkipFast, "tskip-fast=%d", reconfiguredParam->bEnableTSkipFast);
+    TOOLCMP(param->bEnableSignHiding, reconfiguredParam->bEnableSignHiding, "signhide=%d", reconfiguredParam->bEnableSignHiding);
+    TOOLCMP(param->bEnableFastIntra, reconfiguredParam->bEnableFastIntra, "fast-intra=%d", reconfiguredParam->bEnableFastIntra);
+    if (param->bEnableLoopFilter && (param->deblockingFilterBetaOffset != reconfiguredParam->deblockingFilterBetaOffset 
+        || param->deblockingFilterTCOffset != reconfiguredParam->deblockingFilterTCOffset))
+    {
+        sprintf(tmp, "deblock(tC=%d:B=%d)", param->deblockingFilterTCOffset, param->deblockingFilterBetaOffset);
+        appendtool(param, buf, sizeof(buf), tmp);
+    }
+    else
+        TOOLCMP(param->bEnableLoopFilter,  reconfiguredParam->bEnableLoopFilter, "deblock=%d", reconfiguredParam->bEnableLoopFilter);
+
+    TOOLCMP(param->bEnableTemporalMvp, reconfiguredParam->bEnableTemporalMvp, "tmvp=%d", reconfiguredParam->bEnableTemporalMvp);
+    TOOLCMP(param->bEnableEarlySkip, reconfiguredParam->bEnableEarlySkip, "early-skip=%d", reconfiguredParam->bEnableEarlySkip);
+    x265_log(param, X265_LOG_INFO, "tools:%s\n", buf);
+}
+
 char *x265_param2string(x265_param* p)
 {
     char *buf, *s;

x265_1.6.tar.gz/source/common/param.h -> x265_1.7.tar.gz/source/common/param.h Changed

x265_1.6.tar.gz/source/common/picyuv.cpp -> x265_1.7.tar.gz/source/common/picyuv.cpp Changed

x265_1.6.tar.gz/source/common/pixel.cpp -> x265_1.7.tar.gz/source/common/pixel.cpp Changed

x265_1.6.tar.gz/source/common/predict.cpp -> x265_1.7.tar.gz/source/common/predict.cpp Changed

@@ -273,7 +273,7 @@
 void Predict::predInterLumaShort(const PredictionUnit& pu, ShortYuv& dstSYuv, const PicYuv& refPic, const MV& mv) const
 {
     int16_t* dst = dstSYuv.getLumaAddr(pu.puAbsPartIdx);
-    int dstStride = dstSYuv.m_size;
+    intptr_t dstStride = dstSYuv.m_size;
 
     intptr_t srcStride = refPic.m_stride;
     intptr_t srcOffset = (mv.x >> 2) + (mv.y >> 2) * srcStride;
@@ -288,7 +288,7 @@
     X265_CHECK(dstStride == MAX_CU_SIZE, "stride expected to be max cu size\n");
 
     if (!(yFrac | xFrac))
-        primitives.luma_p2s(src, srcStride, dst, pu.width, pu.height);
+        primitives.pu[partEnum].convert_p2s(src, srcStride, dst, dstStride);
     else if (!yFrac)
         primitives.pu[partEnum].luma_hps(src, srcStride, dst, dstStride, xFrac, 0);
     else if (!xFrac)
@@ -375,14 +375,13 @@
     int partEnum = partitionFromSizes(pu.width, pu.height);
     
     uint32_t cxWidth  = pu.width >> m_hChromaShift;
-    uint32_t cxHeight = pu.height >> m_vChromaShift;
 
-    X265_CHECK(((cxWidth | cxHeight) % 2) == 0, "chroma block size expected to be multiple of 2\n");
+    X265_CHECK(((cxWidth | (pu.height >> m_vChromaShift)) % 2) == 0, "chroma block size expected to be multiple of 2\n");
 
     if (!(yFrac | xFrac))
     {
-        primitives.chroma[m_csp].p2s(refCb, refStride, dstCb, cxWidth, cxHeight);
-        primitives.chroma[m_csp].p2s(refCr, refStride, dstCr, cxWidth, cxHeight);
+        primitives.chroma[m_csp].pu[partEnum].p2s(refCb, refStride, dstCb, dstStride);
+        primitives.chroma[m_csp].pu[partEnum].p2s(refCr, refStride, dstCr, dstStride);
     }
     else if (!yFrac)
     {
@@ -817,7 +816,9 @@
             const pixel refSample = *pAdiLineNext;
             // Pad unavailable samples with new value
             int nextOrTop = X265_MIN(next, leftUnits);
+
             // fill left column
+#if HIGH_BIT_DEPTH
             while (curr < nextOrTop)
             {
                 for (int i = 0; i < unitHeight; i++)
@@ -836,6 +837,24 @@
                 adi += unitWidth;
                 curr++;
             }
+#else
+            X265_CHECK(curr <= nextOrTop, "curr must be less than or equal to nextOrTop\n");
+            if (curr < nextOrTop)
+            {
+                const int fillSize = unitHeight * (nextOrTop - curr);
+                memset(adi, refSample, fillSize * sizeof(pixel));
+                curr = nextOrTop;
+                adi += fillSize;
+            }
+
+            if (curr < next)
+            {
+                const int fillSize = unitWidth * (next - curr);
+                memset(adi, refSample, fillSize * sizeof(pixel));
+                curr = next;
+                adi += fillSize;
+            }
+#endif
         }
 
         // pad all other reference samples.

x265_1.6.tar.gz/source/common/primitives.cpp -> x265_1.7.tar.gz/source/common/primitives.cpp Changed

x265_1.6.tar.gz/source/common/primitives.h -> x265_1.7.tar.gz/source/common/primitives.h Changed

@@ -140,7 +140,8 @@
 typedef int(*count_nonzero_t)(const int16_t* quantCoeff);
 typedef void (*weightp_pp_t)(const pixel* src, pixel* dst, intptr_t stride, int width, int height, int w0, int round, int shift, int offset);
 typedef void (*weightp_sp_t)(const int16_t* src, pixel* dst, intptr_t srcStride, intptr_t dstStride, int width, int height, int w0, int round, int shift, int offset);
-typedef void (*scale_t)(pixel* dst, const pixel* src, intptr_t stride);
+typedef void (*scale1D_t)(pixel* dst, const pixel* src);
+typedef void (*scale2D_t)(pixel* dst, const pixel* src, intptr_t stride);
 typedef void (*downscale_t)(const pixel* src0, pixel* dstf, pixel* dsth, pixel* dstv, pixel* dstc,
                             intptr_t src_stride, intptr_t dst_stride, int width, int height);
 typedef void (*extendCURowBorder_t)(pixel* txt, intptr_t stride, int width, int height, int marginX);
@@ -155,8 +156,7 @@
 typedef void (*filter_sp_t) (const int16_t* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int coeffIdx);
 typedef void (*filter_ss_t) (const int16_t* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx);
 typedef void (*filter_hv_pp_t) (const pixel* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int idxX, int idxY);
-typedef void (*filter_p2s_wxh_t)(const pixel* src, intptr_t srcStride, int16_t* dst, int width, int height);
-typedef void (*filter_p2s_t)(const pixel* src, intptr_t srcStride, int16_t* dst);
+typedef void (*filter_p2s_t)(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
 
 typedef void (*copy_pp_t)(pixel* dst, intptr_t dstStride, const pixel* src, intptr_t srcStride); // dst is aligned
 typedef void (*copy_sp_t)(pixel* dst, intptr_t dstStride, const int16_t* src, intptr_t srcStride);
@@ -168,7 +168,7 @@
 typedef void (*pixelavg_pp_t)(pixel* dst, intptr_t dstride, const pixel* src0, intptr_t sstride0, const pixel* src1, intptr_t sstride1, int weight);
 typedef void (*addAvg_t)(const int16_t* src0, const int16_t* src1, pixel* dst, intptr_t src0Stride, intptr_t src1Stride, intptr_t dstStride);
 
-typedef void (*saoCuOrgE0_t)(pixel* rec, int8_t* offsetEo, int width, int8_t signLeft);
+typedef void (*saoCuOrgE0_t)(pixel* rec, int8_t* offsetEo, int width, int8_t* signLeft, intptr_t stride);
 typedef void (*saoCuOrgE1_t)(pixel* rec, int8_t* upBuff1, int8_t* offsetEo, intptr_t stride, int width);
 typedef void (*saoCuOrgE2_t)(pixel* rec, int8_t* pBufft, int8_t* pBuff1, int8_t* offsetEo, int lcuWidth, intptr_t stride);
 typedef void (*saoCuOrgE3_t)(pixel* rec, int8_t* upBuff1, int8_t* m_offsetEo, intptr_t stride, int startX, int endX);
@@ -179,7 +179,8 @@
 
 typedef void (*cutree_propagate_cost) (int* dst, const uint16_t* propagateIn, const int32_t* intraCosts, const uint16_t* interCosts, const int32_t* invQscales, const double* fpsFactor, int len);
 
-typedef int (*findPosLast_t)(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig);
+typedef int (*scanPosLast_t)(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig, const uint16_t* scanCG4x4, const int trSize);
+typedef uint32_t (*findPosFirstLast_t)(const int16_t *dstCoeff, const intptr_t trSize, const uint16_t scanTbl[16]);
 
 /* Function pointers to optimized encoder primitives. Each pointer can reference
  * either an assembly routine, a SIMD intrinsic primitive, or a C function */
@@ -210,7 +211,7 @@
         addAvg_t       addAvg;      // bidir motion compensation, uses 16bit values
 
         copy_pp_t      copy_pp;
-        filter_p2s_t   filter_p2s;
+        filter_p2s_t   convert_p2s;
     }
     pu[NUM_PU_SIZES];
 
@@ -266,17 +267,26 @@
     dequant_scaling_t     dequant_scaling;
     dequant_normal_t      dequant_normal;
     denoiseDct_t          denoiseDct;
-    scale_t               scale1D_128to64;
-    scale_t               scale2D_64to32;
+    scale1D_t             scale1D_128to64;
+    scale2D_t             scale2D_64to32;
 
     ssim_4x4x2_core_t     ssim_4x4x2_core;
     ssim_end4_t           ssim_end_4;
 
     sign_t                sign;
     saoCuOrgE0_t          saoCuOrgE0;
-    saoCuOrgE1_t          saoCuOrgE1;
-    saoCuOrgE2_t          saoCuOrgE2;
-    saoCuOrgE3_t          saoCuOrgE3;
+
+    /* To avoid the overhead in avx2 optimization in handling width=16, SAO_E0_1 is split
+     * into two parts: saoCuOrgE1, saoCuOrgE1_2Rows */
+    saoCuOrgE1_t          saoCuOrgE1, saoCuOrgE1_2Rows;
+
+    // saoCuOrgE2[0] is used for width<=16 and saoCuOrgE2[1] is used for width > 16.
+    saoCuOrgE2_t          saoCuOrgE2[2];
+
+    /* In avx2 optimization, two rows cannot be handled simultaneously since it requires 
+     * a pixel from the previous row. So, saoCuOrgE3[0] is used for width<=16 and 
+     * saoCuOrgE3[1] is used for width > 16. */
+    saoCuOrgE3_t          saoCuOrgE3[2];
     saoCuOrgB0_t          saoCuOrgB0;
 
     downscale_t           frameInitLowres;
@@ -289,9 +299,9 @@
     weightp_sp_t          weight_sp;
     weightp_pp_t          weight_pp;
 
-    filter_p2s_wxh_t      luma_p2s;
 
-    findPosLast_t         findPosLast;
+    scanPosLast_t         scanPosLast;
+    findPosFirstLast_t    findPosFirstLast;
 
     /* There is one set of chroma primitives per color space. An encoder will
      * have just a single color space and thus it will only ever use one entry
@@ -316,7 +326,7 @@
             filter_hps_t filter_hps;
             addAvg_t     addAvg;
             copy_pp_t    copy_pp;
-            filter_p2s_t chroma_p2s;
+            filter_p2s_t p2s;
 
         }
         pu[NUM_PU_SIZES];
@@ -336,7 +346,6 @@
         }
         cu[NUM_CU_SIZES];
 
-        filter_p2s_wxh_t p2s; // takes width/height as arguments
     }
     chroma[X265_CSP_COUNT];
 };

x265_1.6.tar.gz/source/common/quant.cpp -> x265_1.7.tar.gz/source/common/quant.cpp Changed

@@ -198,7 +198,8 @@
 {
     m_entropyCoder = &entropy;
     m_rdoqLevel    = rdoqLevel;
-    m_psyRdoqScale = (int64_t)(psyScale * 256.0);
+    m_psyRdoqScale = (int32_t)(psyScale * 256.0);
+    X265_CHECK((psyScale * 256.0) < (double)MAX_INT, "psyScale value too large\n");
     m_scalingList  = &scalingList;
     m_resiDctCoeff = X265_MALLOC(int16_t, MAX_TR_SIZE * MAX_TR_SIZE * 2);
     m_fencDctCoeff = m_resiDctCoeff + (MAX_TR_SIZE * MAX_TR_SIZE);
@@ -225,16 +226,15 @@
     X265_FREE(m_fencShortBuf);
 }
 
-void Quant::setQPforQuant(const CUData& cu)
+void Quant::setQPforQuant(const CUData& ctu, int qp)
 {
-    m_tqBypass = !!cu.m_tqBypass[0];
+    m_tqBypass = !!ctu.m_tqBypass[0];
     if (m_tqBypass)
         return;
-    m_nr = m_frameNr ? &m_frameNr[cu.m_encData->m_frameEncoderID] : NULL;
-    int qpy = cu.m_qp[0];
-    m_qpParam[TEXT_LUMA].setQpParam(qpy + QP_BD_OFFSET);
-    setChromaQP(qpy + cu.m_slice->m_pps->chromaQpOffset[0], TEXT_CHROMA_U, cu.m_chromaFormat);
-    setChromaQP(qpy + cu.m_slice->m_pps->chromaQpOffset[1], TEXT_CHROMA_V, cu.m_chromaFormat);
+    m_nr = m_frameNr ? &m_frameNr[ctu.m_encData->m_frameEncoderID] : NULL;
+    m_qpParam[TEXT_LUMA].setQpParam(qp + QP_BD_OFFSET);
+    setChromaQP(qp + ctu.m_slice->m_pps->chromaQpOffset[0], TEXT_CHROMA_U, ctu.m_chromaFormat);
+    setChromaQP(qp + ctu.m_slice->m_pps->chromaQpOffset[1], TEXT_CHROMA_V, ctu.m_chromaFormat);
 }
 
 void Quant::setChromaQP(int qpin, TextType ttype, int chFmt)
@@ -515,6 +515,7 @@
 {
     int transformShift = MAX_TR_DYNAMIC_RANGE - X265_DEPTH - log2TrSize; /* Represents scaling through forward transform */
     int scalingListType = (cu.isIntra(absPartIdx) ? 0 : 3) + ttype;
+    const uint32_t usePsyMask = usePsy ? -1 : 0;
 
     X265_CHECK(scalingListType < 6, "scaling list type out of range\n");
 
@@ -529,9 +530,10 @@
     X265_CHECK((int)numSig == primitives.cu[log2TrSize - 2].count_nonzero(dstCoeff), "numSig differ\n");
     if (!numSig)
         return 0;
+
     uint32_t trSize = 1 << log2TrSize;
     int64_t lambda2 = m_qpParam[ttype].lambda2;
-    int64_t psyScale = (m_psyRdoqScale * m_qpParam[ttype].lambda);
+    const int64_t psyScale = ((int64_t)m_psyRdoqScale * m_qpParam[ttype].lambda);
 
     /* unquant constants for measuring distortion. Scaling list quant coefficients have a (1 << 4)
      * scale applied that must be removed during unquant. Note that in real dequant there is clipping
@@ -544,7 +546,7 @@
 #define UNQUANT(lvl)    (((lvl) * (unquantScale[blkPos] << per) + unquantRound) >> unquantShift)
 #define SIGCOST(bits)   ((lambda2 * (bits)) >> 8)
 #define RDCOST(d, bits) ((((int64_t)d * d) << scaleBits) + SIGCOST(bits))
-#define PSYVALUE(rec)   ((psyScale * (rec)) >> (16 - scaleBits))
+#define PSYVALUE(rec)   ((psyScale * (rec)) >> (2 * transformShift + 1))
 
     int64_t costCoeff[32 * 32];   /* d*d + lambda * bits */
     int64_t costUncoded[32 * 32]; /* d*d + lambda * 0    */
@@ -557,14 +559,6 @@
     int64_t costCoeffGroupSig[MLS_GRP_NUM]; /* lambda * bits of group coding cost */
     uint64_t sigCoeffGroupFlag64 = 0;
 
-    uint32_t ctxSet      = 0;
-    int    c1            = 1;
-    int    c2            = 0;
-    uint32_t goRiceParam = 0;
-    uint32_t c1Idx       = 0;
-    uint32_t c2Idx       = 0;
-    int cgLastScanPos    = -1;
-    int lastScanPos      = -1;
     const uint32_t cgSize = (1 << MLS_CG_SIZE); /* 4x4 num coef = 16 */
     bool bIsLuma = ttype == TEXT_LUMA;
 
@@ -579,30 +573,231 @@
     TUEntropyCodingParameters codeParams;
     cu.getTUEntropyCodingParameters(codeParams, absPartIdx, log2TrSize, bIsLuma);
     const uint32_t cgNum = 1 << (codeParams.log2TrSizeCG * 2);
+    const uint32_t cgStride = (trSize >> MLS_CG_LOG2_SIZE);
+
+    uint8_t coeffNum[MLS_GRP_NUM];      // value range[0, 16]
+    uint16_t coeffSign[MLS_GRP_NUM];    // bit mask map for non-zero coeff sign
+    uint16_t coeffFlag[MLS_GRP_NUM];    // bit mask map for non-zero coeff
+
+#if CHECKED_BUILD || _DEBUG
+    // clean output buffer, the asm version of scanPosLast Never output anything after latest non-zero coeff group
+    memset(coeffNum, 0, sizeof(coeffNum));
+    memset(coeffSign, 0, sizeof(coeffNum));
+    memset(coeffFlag, 0, sizeof(coeffNum));
+#endif
+    const int lastScanPos = primitives.scanPosLast(codeParams.scan, dstCoeff, coeffSign, coeffFlag, coeffNum, numSig, g_scan4x4[codeParams.scanType], trSize);
+    const int cgLastScanPos = (lastScanPos >> LOG2_SCAN_SET_SIZE);
+
 
     /* TODO: update bit estimates if dirty */
     EstBitsSbac& estBitsSbac = m_entropyCoder->m_estBitsSbac;
 
-    uint32_t scanPos;
-    coeffGroupRDStats cgRdStats;
+    uint32_t scanPos = 0;
+    uint32_t c1 = 1;
+
+    // process trail all zero Coeff Group
+
+    /* coefficients after lastNZ have no distortion signal cost */
+    const int zeroCG = cgNum - 1 - cgLastScanPos;
+    memset(&costCoeff[(cgLastScanPos + 1) << MLS_CG_SIZE], 0, zeroCG * MLS_CG_BLK_SIZE * sizeof(int64_t));
+    memset(&costSig[(cgLastScanPos + 1) << MLS_CG_SIZE], 0, zeroCG * MLS_CG_BLK_SIZE * sizeof(int64_t));
+
+    /* sum zero coeff (uncodec) cost */
+
+    // TODO: does we need these cost?
+    if (usePsyMask)
+    {
+        for (int cgScanPos = cgLastScanPos + 1; cgScanPos < (int)cgNum ; cgScanPos++)
+        {
+            X265_CHECK(coeffNum[cgScanPos] == 0, "count of coeff failure\n");
+
+            uint32_t scanPosBase = (cgScanPos << MLS_CG_SIZE);
+            uint32_t blkPos      = codeParams.scan[scanPosBase];
+
+            // TODO: we can't SIMD optimize because PSYVALUE need 64-bits multiplication, convert to Double can work faster by FMA
+            for (int y = 0; y < MLS_CG_SIZE; y++)
+            {
+                for (int x = 0; x < MLS_CG_SIZE; x++)
+                {
+                    int signCoef         = m_resiDctCoeff[blkPos + x];            /* pre-quantization DCT coeff */
+                    int predictedCoef    = m_fencDctCoeff[blkPos + x] - signCoef; /* predicted DCT = source DCT - residual DCT*/
+
+                    costUncoded[blkPos + x] = ((int64_t)signCoef * signCoef) << scaleBits;
+
+                    /* when no residual coefficient is coded, predicted coef == recon coef */
+                    costUncoded[blkPos + x] -= PSYVALUE(predictedCoef);
+
+                    totalUncodedCost += costUncoded[blkPos + x];
+                    totalRdCost += costUncoded[blkPos + x];
+                }
+                blkPos += trSize;
+            }
+        }
+    }
+    else
+    {
+        // non-psy path
+        for (int cgScanPos = cgLastScanPos + 1; cgScanPos < (int)cgNum ; cgScanPos++)
+        {
+            X265_CHECK(coeffNum[cgScanPos] == 0, "count of coeff failure\n");
+
+            uint32_t scanPosBase = (cgScanPos << MLS_CG_SIZE);
+            uint32_t blkPos      = codeParams.scan[scanPosBase];
+
+            for (int y = 0; y < MLS_CG_SIZE; y++)
+            {
+                for (int x = 0; x < MLS_CG_SIZE; x++)
+                {
+                    int signCoef = m_resiDctCoeff[blkPos + x];            /* pre-quantization DCT coeff */
+                    costUncoded[blkPos + x] = ((int64_t)signCoef * signCoef) << scaleBits;
+
+                    totalUncodedCost += costUncoded[blkPos + x];
+                    totalRdCost += costUncoded[blkPos + x];
+                }
+                blkPos += trSize;
+            }
+        }
+    }
+
+    static const uint8_t table_cnt[5][SCAN_SET_SIZE] =
+    {
+        // patternSigCtx = 0
+        {
+            2, 1, 1, 0,
+            1, 1, 0, 0,
+            1, 0, 0, 0,
+            0, 0, 0, 0,
+        },
+        // patternSigCtx = 1
+        {
+            2, 2, 2, 2,
+            1, 1, 1, 1,
+            0, 0, 0, 0,
+            0, 0, 0, 0,
+        },
+        // patternSigCtx = 2
+        {
+            2, 1, 0, 0,
+            2, 1, 0, 0,
+            2, 1, 0, 0,
+            2, 1, 0, 0,
+        },
+        // patternSigCtx = 3
+        {
+            2, 2, 2, 2,
+            2, 2, 2, 2,
+            2, 2, 2, 2,
+            2, 2, 2, 2,
+        },
+        // 4x4
+        {
+            0, 1, 4, 5,
+            2, 3, 4, 5,
+            6, 6, 8, 8,
+            7, 7, 8, 8
+        }
+    };
 
     /* iterate over coding groups in reverse scan order */
-    for (int cgScanPos = cgNum - 1; cgScanPos >= 0; cgScanPos--)
+    for (int cgScanPos = cgLastScanPos; cgScanPos >= 0; cgScanPos--)
     {
+        uint32_t ctxSet = (cgScanPos && bIsLuma) ? 2 : 0;
         const uint32_t cgBlkPos = codeParams.scanCG[cgScanPos];
         const uint32_t cgPosY   = cgBlkPos >> codeParams.log2TrSizeCG;
         const uint32_t cgPosX   = cgBlkPos - (cgPosY << codeParams.log2TrSizeCG);
         const uint64_t cgBlkPosMask = ((uint64_t)1 << cgBlkPos);
-        memset(&cgRdStats, 0, sizeof(coeffGroupRDStats));
+        const int patternSigCtx = calcPatternSigCtx(sigCoeffGroupFlag64, cgPosX, cgPosY, cgBlkPos, cgStride);
+        const int ctxSigOffset = codeParams.firstSignificanceMapContext + (cgScanPos && bIsLuma ? 3 : 0);
+
+        if (c1 == 0)
+            ctxSet++;
+        c1 = 1;
 
-        const int patternSigCtx = calcPatternSigCtx(sigCoeffGroupFlag64, cgPosX, cgPosY, codeParams.log2TrSizeCG);
+        if (cgScanPos && (coeffNum[cgScanPos] == 0))
+        {
+            // TODO: does we need zero-coeff cost?
+            const uint32_t scanPosBase = (cgScanPos << MLS_CG_SIZE);
+            uint32_t blkPos = codeParams.scan[scanPosBase];
 
+            if (usePsyMask)
+            {
+                // TODO: we can't SIMD optimize because PSYVALUE need 64-bits multiplication, convert to Double can work faster by FMA
+                for (int y = 0; y < MLS_CG_SIZE; y++)
+                {
+                    for (int x = 0; x < MLS_CG_SIZE; x++)
+                    {
+                        int signCoef         = m_resiDctCoeff[blkPos + x];            /* pre-quantization DCT coeff */
+                        int predictedCoef    = m_fencDctCoeff[blkPos + x] - signCoef; /* predicted DCT = source DCT - residual DCT*/
+
+                        costUncoded[blkPos + x] = ((int64_t)signCoef * signCoef) << scaleBits;
+
+                        /* when no residual coefficient is coded, predicted coef == recon coef */
+                        costUncoded[blkPos + x] -= PSYVALUE(predictedCoef);
+
+                        totalUncodedCost += costUncoded[blkPos + x];
+                        totalRdCost += costUncoded[blkPos + x];
+
+                        const uint32_t scanPosOffset =  y * MLS_CG_SIZE + x;
+                        const uint32_t ctxSig = table_cnt[patternSigCtx][g_scan4x4[codeParams.scanType][scanPosOffset]] + ctxSigOffset;
+                        X265_CHECK(trSize > 4, "trSize check failure\n");
+                        X265_CHECK(ctxSig == getSigCtxInc(patternSigCtx, log2TrSize, trSize, codeParams.scan[scanPosBase + scanPosOffset], bIsLuma, codeParams.firstSignificanceMapContext), "sigCtx check failure\n");
+
+                        costSig[scanPosBase + scanPosOffset] = SIGCOST(estBitsSbac.significantBits[0][ctxSig]);
+                        costCoeff[scanPosBase + scanPosOffset] = costUncoded[blkPos + x];
+                        sigRateDelta[blkPos + x] = estBitsSbac.significantBits[1][ctxSig] - estBitsSbac.significantBits[0][ctxSig];
+                    }
+                    blkPos += trSize;
+                }
+            }
+            else
+            {
+                // non-psy path
+                for (int y = 0; y < MLS_CG_SIZE; y++)
+                {
+                    for (int x = 0; x < MLS_CG_SIZE; x++)
+                    {
+                        int signCoef = m_resiDctCoeff[blkPos + x];            /* pre-quantization DCT coeff */
+                        costUncoded[blkPos + x] = ((int64_t)signCoef * signCoef) << scaleBits;
+
+                        totalUncodedCost += costUncoded[blkPos + x];
+                        totalRdCost += costUncoded[blkPos + x];
+
+                        const uint32_t scanPosOffset =  y * MLS_CG_SIZE + x;
+                        const uint32_t ctxSig = table_cnt[patternSigCtx][g_scan4x4[codeParams.scanType][scanPosOffset]] + ctxSigOffset;
+                        X265_CHECK(trSize > 4, "trSize check failure\n");
+                        X265_CHECK(ctxSig == getSigCtxInc(patternSigCtx, log2TrSize, trSize, codeParams.scan[scanPosBase + scanPosOffset], bIsLuma, codeParams.firstSignificanceMapContext), "sigCtx check failure\n");
+
+                        costSig[scanPosBase + scanPosOffset] = SIGCOST(estBitsSbac.significantBits[0][ctxSig]);
+                        costCoeff[scanPosBase + scanPosOffset] = costUncoded[blkPos + x];
+                        sigRateDelta[blkPos + x] = estBitsSbac.significantBits[1][ctxSig] - estBitsSbac.significantBits[0][ctxSig];
+                    }
+                    blkPos += trSize;
+                }
+            }
+
+            /* there were no coded coefficients in this coefficient group */
+            {
+                uint32_t ctxSig = getSigCoeffGroupCtxInc(sigCoeffGroupFlag64, cgPosX, cgPosY, cgBlkPos, cgStride);
+                costCoeffGroupSig[cgScanPos] = SIGCOST(estBitsSbac.significantCoeffGroupBits[ctxSig][0]);
+                totalRdCost += costCoeffGroupSig[cgScanPos];  /* add cost of 0 bit in significant CG bitmap */
+            }
+            continue;
+        }
+
+        coeffGroupRDStats cgRdStats;
+        memset(&cgRdStats, 0, sizeof(coeffGroupRDStats));
+
+        uint32_t subFlagMask = coeffFlag[cgScanPos];
+        int    c2            = 0;
+        uint32_t goRiceParam = 0;
+        uint32_t c1Idx       = 0;
+        uint32_t c2Idx       = 0;
         /* iterate over coefficients in each group in reverse scan order */
         for (int scanPosinCG = cgSize - 1; scanPosinCG >= 0; scanPosinCG--)
         {
             scanPos              = (cgScanPos << MLS_CG_SIZE) + scanPosinCG;
             uint32_t blkPos      = codeParams.scan[scanPos];
-            uint16_t maxAbsLevel = (int16_t)abs(dstCoeff[blkPos]);             /* abs(quantized coeff) */
+            uint32_t maxAbsLevel = abs(dstCoeff[blkPos]);             /* abs(quantized coeff) */
             int signCoef         = m_resiDctCoeff[blkPos];            /* pre-quantization DCT coeff */
             int predictedCoef    = m_fencDctCoeff[blkPos] - signCoef; /* predicted DCT = source DCT - residual DCT*/
 
@@ -611,22 +806,21 @@
              * FIX15 nature of the CABAC cost tables minus the forward transform scale */
 
             /* cost of not coding this coefficient (all distortion, no signal bits) */
-            costUncoded[scanPos] = (int64_t)(signCoef * signCoef) << scaleBits;
-            if (usePsy && blkPos)
+            costUncoded[blkPos] = ((int64_t)signCoef * signCoef) << scaleBits;
+            X265_CHECK((!!scanPos ^ !!blkPos) == 0, "failed on (blkPos=0 && scanPos!=0)\n");
+            if (usePsyMask & scanPos)
                 /* when no residual coefficient is coded, predicted coef == recon coef */
-                costUncoded[scanPos] -= PSYVALUE(predictedCoef);
+                costUncoded[blkPos] -= PSYVALUE(predictedCoef);
 
-            totalUncodedCost += costUncoded[scanPos];
+            totalUncodedCost += costUncoded[blkPos];
 
-            if (maxAbsLevel && lastScanPos < 0)
-            {
-                /* remember the first non-zero coef found in this reverse scan as the last pos */
-                lastScanPos   = scanPos;
-                ctxSet        = (scanPos < SCAN_SET_SIZE || !bIsLuma) ? 0 : 2;
-                cgLastScanPos = cgScanPos;
-            }
+            // coefficient level estimation
+            const int* greaterOneBits = estBitsSbac.greaterOneBits[4 * ctxSet + c1];
+            const uint32_t ctxSig = (blkPos == 0) ? 0 : table_cnt[(trSize == 4) ? 4 : patternSigCtx][g_scan4x4[codeParams.scanType][scanPosinCG]] + ctxSigOffset;
+            X265_CHECK(ctxSig == getSigCtxInc(patternSigCtx, log2TrSize, trSize, blkPos, bIsLuma, codeParams.firstSignificanceMapContext), "sigCtx check failure\n");
 
-            if (lastScanPos < 0)
+            // before find lastest non-zero coeff
+            if (scanPos > (uint32_t)lastScanPos)
             {
                 /* coefficients after lastNZ have no distortion signal cost */
                 costCoeff[scanPos] = 0;
@@ -635,10 +829,24 @@
                 /* No non-zero coefficient yet found, but this does not mean
                  * there is no uncoded-cost for this coefficient. Pre-
                  * quantization the coefficient may have been non-zero */
-                totalRdCost += costUncoded[scanPos];
+                totalRdCost += costUncoded[blkPos];
+            }
+            else if (!(subFlagMask & 1))
+            {
+                // fast zero coeff path
+                /* set default costs to uncoded costs */
+                costSig[scanPos] = SIGCOST(estBitsSbac.significantBits[0][ctxSig]);
+                costCoeff[scanPos] = costUncoded[blkPos] + costSig[scanPos];
+                sigRateDelta[blkPos] = estBitsSbac.significantBits[1][ctxSig] - estBitsSbac.significantBits[0][ctxSig];
+                totalRdCost += costCoeff[scanPos];
+                rateIncUp[blkPos] = greaterOneBits[0];
+
+                subFlagMask >>= 1;
             }
             else
             {
+                subFlagMask >>= 1;
+
                 const uint32_t c1c2Idx = ((c1Idx - 8) >> (sizeof(int) * CHAR_BIT - 1)) + (((-(int)c2Idx) >> (sizeof(int) * CHAR_BIT - 1)) + 1) * 2;
                 const uint32_t baseLevel = ((uint32_t)0xD9 >> (c1c2Idx * 2)) & 3;  // {1, 2, 1, 3}
 
@@ -647,12 +855,9 @@
                 X265_CHECK((int)baseLevel == ((c1Idx < C1FLAG_NUMBER) ? (2 + (c2Idx == 0)) : 1), "scan validation 3\n");
 
                 // coefficient level estimation
-                const uint32_t oneCtx = 4 * ctxSet + c1;
-                const uint32_t absCtx = ctxSet + c2;
-                const int* greaterOneBits = estBitsSbac.greaterOneBits[oneCtx];
-                const int* levelAbsBits = estBitsSbac.levelAbsBits[absCtx];
+                const int* levelAbsBits = estBitsSbac.levelAbsBits[ctxSet + c2];
 
-                uint16_t level = 0;
+                uint32_t level = 0;
                 uint32_t sigCoefBits = 0;
                 costCoeff[scanPos] = MAX_INT64;
 
@@ -660,48 +865,82 @@
                     sigRateDelta[blkPos] = 0;
                 else
                 {
-                    const uint32_t ctxSig = getSigCtxInc(patternSigCtx, log2TrSize, trSize, blkPos, bIsLuma, codeParams.firstSignificanceMapContext);
                     if (maxAbsLevel < 3)
                     {
                         /* set default costs to uncoded costs */
-                        costSig[scanPos] = SIGCOST(estBitsSbac.significantBits[ctxSig][0]);
-                        costCoeff[scanPos] = costUncoded[scanPos] + costSig[scanPos];
+                        costSig[scanPos] = SIGCOST(estBitsSbac.significantBits[0][ctxSig]);
+                        costCoeff[scanPos] = costUncoded[blkPos] + costSig[scanPos];
                     }
-                    sigRateDelta[blkPos] = estBitsSbac.significantBits[ctxSig][1] - estBitsSbac.significantBits[ctxSig][0];
-                    sigCoefBits = estBitsSbac.significantBits[ctxSig][1];
+                    sigRateDelta[blkPos] = estBitsSbac.significantBits[1][ctxSig] - estBitsSbac.significantBits[0][ctxSig];
+                    sigCoefBits = estBitsSbac.significantBits[1][ctxSig];
                 }
-                if (maxAbsLevel)
+
+                // NOTE: X265_MAX(maxAbsLevel - 1, 1) ==> (X>=2 -> X-1), (X<2 -> 1)  | (0 < X < 2 ==> X=1)
+                if (maxAbsLevel == 1)
                 {
-                    uint16_t minAbsLevel = X265_MAX(maxAbsLevel - 1, 1);
-                    for (uint16_t lvl = maxAbsLevel; lvl >= minAbsLevel; lvl--)
+                    uint32_t levelBits = (c1c2Idx & 1) ? greaterOneBits[0] + IEP_RATE : ((1 + goRiceParam) << 15) + IEP_RATE;
+                    X265_CHECK(levelBits == getICRateCost(1, 1 - baseLevel, greaterOneBits, levelAbsBits, goRiceParam, c1c2Idx) + IEP_RATE, "levelBits mistake\n");
+
+                    int unquantAbsLevel = UNQUANT(1);
+                    int d = abs(signCoef) - unquantAbsLevel;
+                    int64_t curCost = RDCOST(d, sigCoefBits + levelBits);
+
+                    /* Psy RDOQ: bias in favor of higher AC coefficients in the reconstructed frame */
+                    if (usePsyMask & scanPos)
                     {
-                        uint32_t levelBits = getICRateCost(lvl, lvl - baseLevel, greaterOneBits, levelAbsBits, goRiceParam, c1c2Idx) + IEP_RATE;
+                        int reconCoef = abs(unquantAbsLevel + SIGN(predictedCoef, signCoef));
+                        curCost -= PSYVALUE(reconCoef);
+                    }
 
-                        int unquantAbsLevel = UNQUANT(lvl);
-                        int d = abs(signCoef) - unquantAbsLevel;
-                        int64_t curCost = RDCOST(d, sigCoefBits + levelBits);
+                    if (curCost < costCoeff[scanPos])
+                    {
+                        level = 1;
+                        costCoeff[scanPos] = curCost;
+                        costSig[scanPos] = SIGCOST(sigCoefBits);
+                    }
+                }
+                else if (maxAbsLevel)
+                {
+                    uint32_t levelBits0 = getICRateCost(maxAbsLevel,     maxAbsLevel     - baseLevel, greaterOneBits, levelAbsBits, goRiceParam, c1c2Idx) + IEP_RATE;
+                    uint32_t levelBits1 = getICRateCost(maxAbsLevel - 1, maxAbsLevel - 1 - baseLevel, greaterOneBits, levelAbsBits, goRiceParam, c1c2Idx) + IEP_RATE;
 
-                        /* Psy RDOQ: bias in favor of higher AC coefficients in the reconstructed frame */
-                        if (usePsy && blkPos)
-                        {
-                            int reconCoef = abs(unquantAbsLevel + SIGN(predictedCoef, signCoef));
-                            curCost -= PSYVALUE(reconCoef);
-                        }
+                    int unquantAbsLevel0 = UNQUANT(maxAbsLevel);
+                    int d0 = abs(signCoef) - unquantAbsLevel0;
+                    int64_t curCost0 = RDCOST(d0, sigCoefBits + levelBits0);
 
-                        if (curCost < costCoeff[scanPos])
-                        {
-                            level = lvl;
-                            costCoeff[scanPos] = curCost;
-                            costSig[scanPos] = SIGCOST(sigCoefBits);
-                        }
+                    int unquantAbsLevel1 = UNQUANT(maxAbsLevel - 1);
+                    int d1 = abs(signCoef) - unquantAbsLevel1;
+                    int64_t curCost1 = RDCOST(d1, sigCoefBits + levelBits1);
+
+                    /* Psy RDOQ: bias in favor of higher AC coefficients in the reconstructed frame */
+                    if (usePsyMask & scanPos)
+                    {
+                        int reconCoef;
+                        reconCoef = abs(unquantAbsLevel0 + SIGN(predictedCoef, signCoef));
+                        curCost0 -= PSYVALUE(reconCoef);
+
+                        reconCoef = abs(unquantAbsLevel1 + SIGN(predictedCoef, signCoef));
+                        curCost1 -= PSYVALUE(reconCoef);
+                    }
+                    if (curCost0 < costCoeff[scanPos])
+                    {
+                        level = maxAbsLevel;
+                        costCoeff[scanPos] = curCost0;
+                        costSig[scanPos] = SIGCOST(sigCoefBits);
+                    }
+                    if (curCost1 < costCoeff[scanPos])
+                    {
+                        level = maxAbsLevel - 1;
+                        costCoeff[scanPos] = curCost1;
+                        costSig[scanPos] = SIGCOST(sigCoefBits);
                     }
                 }
 
-                dstCoeff[blkPos] = level;
+                dstCoeff[blkPos] = (int16_t)level;
                 totalRdCost += costCoeff[scanPos];
 
                 /* record costs for sign-hiding performed at the end */
-                if (level)
+                if ((cu.m_slice->m_pps->bSignHideEnabled ? ~0 : 0) & level)
                 {
                     const int32_t diff0 = level - 1 - baseLevel;
                     const int32_t diff2 = level + 1 - baseLevel;
@@ -763,41 +1002,27 @@
                 else if ((c1 < 3) && (c1 > 0) && level)
                     c1++;
 
-                /* context set update */
-                if (!(scanPos % SCAN_SET_SIZE) && scanPos)
+                if (dstCoeff[blkPos])
                 {
-                    c2 = 0;
-                    goRiceParam = 0;
-
-                    c1Idx = 0;
-                    c2Idx = 0;
-                    ctxSet = (scanPos == SCAN_SET_SIZE || !bIsLuma) ? 0 : 2;
-                    X265_CHECK(c1 >= 0, "c1 is negative\n");
-                    ctxSet -= ((int32_t)(c1 - 1) >> 31);
-                    c1 = 1;
+                    sigCoeffGroupFlag64 |= cgBlkPosMask;
+                    cgRdStats.codedLevelAndDist += costCoeff[scanPos] - costSig[scanPos];
+                    cgRdStats.uncodedDist += costUncoded[blkPos];
+                    cgRdStats.nnzBeforePos0 += scanPosinCG;
                 }
             }
 
             cgRdStats.sigCost += costSig[scanPos];
-            if (!scanPosinCG)
-                cgRdStats.sigCost0 = costSig[scanPos];
-
-            if (dstCoeff[blkPos])
-            {
-                sigCoeffGroupFlag64 |= cgBlkPosMask;
-                cgRdStats.codedLevelAndDist += costCoeff[scanPos] - costSig[scanPos];
-                cgRdStats.uncodedDist += costUncoded[scanPos];
-                cgRdStats.nnzBeforePos0 += scanPosinCG;
-            }
         } /* end for (scanPosinCG) */
 
+        X265_CHECK((cgScanPos << MLS_CG_SIZE) == (int)scanPos, "scanPos mistake\n");
+        cgRdStats.sigCost0 = costSig[scanPos];
+
         costCoeffGroupSig[cgScanPos] = 0;
 
-        if (cgLastScanPos < 0)
-        {
-            /* nothing to do at this point */
-        }
-        else if (!cgScanPos || cgScanPos == cgLastScanPos)
+        /* nothing to do at this case */
+        X265_CHECK(cgLastScanPos >= 0, "cgLastScanPos check failure\n");
+
+        if (!cgScanPos || cgScanPos == cgLastScanPos)
         {
             /* coeff group 0 is implied to be present, no signal cost */
             /* coeff group with last NZ is implied to be present, handled below */
@@ -815,7 +1040,7 @@
              * of the significant coefficient group flag and evaluate whether the RD cost of the
              * coded group is more than the RD cost of the uncoded group */
 
-            uint32_t sigCtx = getSigCoeffGroupCtxInc(sigCoeffGroupFlag64, cgPosX, cgPosY, codeParams.log2TrSizeCG);
+            uint32_t sigCtx = getSigCoeffGroupCtxInc(sigCoeffGroupFlag64, cgPosX, cgPosY, cgBlkPos, cgStride);
 
             int64_t costZeroCG = totalRdCost + SIGCOST(estBitsSbac.significantCoeffGroupBits[sigCtx][0]);
             costZeroCG += cgRdStats.uncodedDist;       /* add distortion for resetting non-zero levels to zero levels */
@@ -832,23 +1057,17 @@
                 costCoeffGroupSig[cgScanPos] = SIGCOST(estBitsSbac.significantCoeffGroupBits[sigCtx][0]);
 
                 /* reset all coeffs to 0. UNCODE THIS COEFF GROUP! */
-                for (int scanPosinCG = cgSize - 1; scanPosinCG >= 0; scanPosinCG--)
-                {
-                    scanPos = cgScanPos * cgSize + scanPosinCG;
-                    uint32_t blkPos = codeParams.scan[scanPos];
-                    if (dstCoeff[blkPos])
-                    {
-                        costCoeff[scanPos] = costUncoded[scanPos];
-                        costSig[scanPos] = 0;
-                    }
-                    dstCoeff[blkPos] = 0;
-                }
+                const uint32_t blkPos = codeParams.scan[cgScanPos * cgSize];
+                memset(&dstCoeff[blkPos + 0 * trSize], 0, 4 * sizeof(*dstCoeff));
+                memset(&dstCoeff[blkPos + 1 * trSize], 0, 4 * sizeof(*dstCoeff));
+                memset(&dstCoeff[blkPos + 2 * trSize], 0, 4 * sizeof(*dstCoeff));
+                memset(&dstCoeff[blkPos + 3 * trSize], 0, 4 * sizeof(*dstCoeff));
             }
         }
         else
         {
             /* there were no coded coefficients in this coefficient group */
-            uint32_t ctxSig = getSigCoeffGroupCtxInc(sigCoeffGroupFlag64, cgPosX, cgPosY, codeParams.log2TrSizeCG);
+            uint32_t ctxSig = getSigCoeffGroupCtxInc(sigCoeffGroupFlag64, cgPosX, cgPosY, cgBlkPos, cgStride);
             costCoeffGroupSig[cgScanPos] = SIGCOST(estBitsSbac.significantCoeffGroupBits[ctxSig][0]);
             totalRdCost += costCoeffGroupSig[cgScanPos];  /* add cost of 0 bit in significant CG bitmap */
             totalRdCost -= cgRdStats.sigCost;             /* remove cost of significant coefficient bitmap */
@@ -909,7 +1128,7 @@
              * cost of signaling it as not-significant */
             uint32_t blkPos = codeParams.scan[scanPos];
             if (dstCoeff[blkPos])
-            {                
+            {
                 // Calculates the cost of signaling the last significant coefficient in the block 
                 uint32_t pos[2] = { (blkPos & (trSize - 1)), (blkPos >> log2TrSize) };
                 if (codeParams.scanType == SCAN_VER)
@@ -940,7 +1159,7 @@
                 }
 
                 totalRdCost -= costCoeff[scanPos];
-                totalRdCost += costUncoded[scanPos];
+                totalRdCost += costUncoded[blkPos];
             }
             else
                 totalRdCost -= costSig[scanPos];
@@ -959,34 +1178,40 @@
         dstCoeff[blkPos] = (int16_t)((level ^ mask) - mask);
     }
 
+    // Average 49.62 pixels
     /* clean uncoded coefficients */
-    for (int pos = bestLastIdx; pos <= lastScanPos; pos++)
+    for (int pos = bestLastIdx; pos <= fastMin(lastScanPos, (bestLastIdx | (SCAN_SET_SIZE - 1))); pos++)
+    {
         dstCoeff[codeParams.scan[pos]] = 0;
+    }
+    for (int pos = (bestLastIdx & ~(SCAN_SET_SIZE - 1)) + SCAN_SET_SIZE; pos <= lastScanPos; pos += SCAN_SET_SIZE)
+    {
+        const uint32_t blkPos = codeParams.scan[pos];
+        memset(&dstCoeff[blkPos + 0 * trSize], 0, 4 * sizeof(*dstCoeff));
+        memset(&dstCoeff[blkPos + 1 * trSize], 0, 4 * sizeof(*dstCoeff));
+        memset(&dstCoeff[blkPos + 2 * trSize], 0, 4 * sizeof(*dstCoeff));
+        memset(&dstCoeff[blkPos + 3 * trSize], 0, 4 * sizeof(*dstCoeff));
+    }
 
     /* rate-distortion based sign-hiding */
     if (cu.m_slice->m_pps->bSignHideEnabled && numSig >= 2)
     {
+        const int realLastScanPos = (bestLastIdx - 1) >> LOG2_SCAN_SET_SIZE;
         int lastCG = true;
-        for (int subSet = cgLastScanPos; subSet >= 0; subSet--)
+        for (int subSet = realLastScanPos; subSet >= 0; subSet--)
         {
             int subPos = subSet << LOG2_SCAN_SET_SIZE;
             int n;
 
-            /* measure distance between first and last non-zero coef in this
-             * coding group */
-            for (n = SCAN_SET_SIZE - 1; n >= 0; --n)
-                if (dstCoeff[codeParams.scan[n + subPos]])
-                    break;
-            if (n < 0)
+            if (!(sigCoeffGroupFlag64 & (1ULL << codeParams.scanCG[subSet])))
                 continue;
 
-            int lastNZPosInCG = n;
-
-            for (n = 0;; n++)
-                if (dstCoeff[codeParams.scan[n + subPos]])
-                    break;
+            /* measure distance between first and last non-zero coef in this
+             * coding group */
+            const uint32_t posFirstLast = primitives.findPosFirstLast(&dstCoeff[codeParams.scan[subPos]], trSize, g_scan4x4[codeParams.scanType]);
+            int firstNZPosInCG = (uint16_t)posFirstLast;
+            int lastNZPosInCG = posFirstLast >> 16;
 
-            int firstNZPosInCG = n;
 
             if (lastNZPosInCG - firstNZPosInCG >= SBH_THRESHOLD)
             {
@@ -1092,22 +1317,6 @@
     return numSig;
 }
 
-/* Pattern decision for context derivation process of significant_coeff_flag */
-uint32_t Quant::calcPatternSigCtx(uint64_t sigCoeffGroupFlag64, uint32_t cgPosX, uint32_t cgPosY, uint32_t log2TrSizeCG)
-{
-    if (!log2TrSizeCG)
-        return 0;
-
-    const uint32_t trSizeCG = 1 << log2TrSizeCG;
-    X265_CHECK(trSizeCG <= 8, "transform CG is too large\n");
-    const uint32_t shift = (cgPosY << log2TrSizeCG) + cgPosX + 1;
-    const uint32_t sigPos = (uint32_t)(shift >= 64 ? 0 : sigCoeffGroupFlag64 >> shift);
-    const uint32_t sigRight = ((int32_t)(cgPosX - (trSizeCG - 1)) >> 31) & (sigPos & 1);
-    const uint32_t sigLower = ((int32_t)(cgPosY - (trSizeCG - 1)) >> 31) & (sigPos >> (trSizeCG - 2)) & 2;
-
-    return sigRight + sigLower;
-}
-
 /* Context derivation process of coeff_abs_significant_flag */
 uint32_t Quant::getSigCtxInc(uint32_t patternSigCtx, uint32_t log2TrSize, uint32_t trSize, uint32_t blkPos, bool bIsLuma,
                              uint32_t firstSignificanceMapContext)
@@ -1175,14 +1384,3 @@
     return (bIsLuma && (posX | posY) >= 4) ? 3 + offset : offset;
 }
 
-/* Context derivation process of coeff_abs_significant_flag */
-uint32_t Quant::getSigCoeffGroupCtxInc(uint64_t cgGroupMask, uint32_t cgPosX, uint32_t cgPosY, uint32_t log2TrSizeCG)
-{
-    const uint32_t trSizeCG = 1 << log2TrSizeCG;
-
-    const uint32_t sigPos = (uint32_t)(cgGroupMask >> (1 + (cgPosY << log2TrSizeCG) + cgPosX));
-    const uint32_t sigRight = ((int32_t)(cgPosX - (trSizeCG - 1)) >> 31) & sigPos;
-    const uint32_t sigLower = ((int32_t)(cgPosY - (trSizeCG - 1)) >> 31) & (sigPos >> (trSizeCG - 1));
-
-    return (sigRight | sigLower) & 1;
-}

x265_1.6.tar.gz/source/common/quant.h -> x265_1.7.tar.gz/source/common/quant.h Changed

@@ -41,7 +41,7 @@
     int per;
     int qp;
     int64_t lambda2; /* FIX8 */
-    int64_t lambda;  /* FIX8 */
+    int32_t lambda;  /* FIX8, dynamic range is 18-bits in 8bpp and 20-bits in 16bpp */
 
     QpParam() : qp(MAX_INT) {}
 
@@ -53,7 +53,8 @@
             per = qpScaled / 6;
             qp  = qpScaled;
             lambda2 = (int64_t)(x265_lambda2_tab[qp - QP_BD_OFFSET] * 256. + 0.5);
-            lambda  = (int64_t)(x265_lambda_tab[qp - QP_BD_OFFSET] * 256. + 0.5);
+            lambda  = (int32_t)(x265_lambda_tab[qp - QP_BD_OFFSET] * 256. + 0.5);
+            X265_CHECK((x265_lambda_tab[qp - QP_BD_OFFSET] * 256. + 0.5) < (double)MAX_INT, "x265_lambda_tab[] value too large\n");
         }
     }
 };
@@ -82,7 +83,7 @@
     QpParam            m_qpParam[3];
 
     int                m_rdoqLevel;
-    int64_t            m_psyRdoqScale;
+    int32_t            m_psyRdoqScale;  // dynamic range [0,50] * 256 = 14-bits
     int16_t*           m_resiDctCoeff;
     int16_t*           m_fencDctCoeff;
     int16_t*           m_fencShortBuf;
@@ -103,7 +104,7 @@
     bool allocNoiseReduction(const x265_param& param);
 
     /* CU setup */
-    void setQPforQuant(const CUData& cu);
+    void setQPforQuant(const CUData& ctu, int qp);
 
     uint32_t transformNxN(const CUData& cu, const pixel* fenc, uint32_t fencStride, const int16_t* residual, uint32_t resiStride, coeff_t* coeff,
                           uint32_t log2TrSize, TextType ttype, uint32_t absPartIdx, bool useTransformSkip);
@@ -111,10 +112,39 @@
     void invtransformNxN(int16_t* residual, uint32_t resiStride, const coeff_t* coeff,
                          uint32_t log2TrSize, TextType ttype, bool bIntra, bool useTransformSkip, uint32_t numSig);
 
+    /* Pattern decision for context derivation process of significant_coeff_flag */
+    static uint32_t calcPatternSigCtx(uint64_t sigCoeffGroupFlag64, uint32_t cgPosX, uint32_t cgPosY, uint32_t cgBlkPos, uint32_t trSizeCG)
+    {
+        if (trSizeCG == 1)
+            return 0;
+
+        X265_CHECK(trSizeCG <= 8, "transform CG is too large\n");
+        X265_CHECK(cgBlkPos < 64, "cgBlkPos is too large\n");
+        // NOTE: cgBlkPos+1 may more than 63, it is invalid for shift,
+        //       but in this case, both cgPosX and cgPosY equal to (trSizeCG - 1),
+        //       the sigRight and sigLower will clear value to zero, the final result will be correct
+        const uint32_t sigPos = (uint32_t)(sigCoeffGroupFlag64 >> (cgBlkPos + 1)); // just need lowest 7-bits valid
+
+        // TODO: instruction BT is faster, but _bittest64 still generate instruction 'BT m, r' in VS2012
+        const uint32_t sigRight = ((int32_t)(cgPosX - (trSizeCG - 1)) >> 31) & (sigPos & 1);
+        const uint32_t sigLower = ((int32_t)(cgPosY - (trSizeCG - 1)) >> 31) & (sigPos >> (trSizeCG - 2)) & 2;
+        return sigRight + sigLower;
+    }
+
+    /* Context derivation process of coeff_abs_significant_flag */
+    static uint32_t getSigCoeffGroupCtxInc(uint64_t cgGroupMask, uint32_t cgPosX, uint32_t cgPosY, uint32_t cgBlkPos, uint32_t trSizeCG)
+    {
+        X265_CHECK(cgBlkPos < 64, "cgBlkPos is too large\n");
+        // NOTE: unsafe shift operator, see NOTE in calcPatternSigCtx
+        const uint32_t sigPos = (uint32_t)(cgGroupMask >> (cgBlkPos + 1)); // just need lowest 8-bits valid
+        const uint32_t sigRight = ((int32_t)(cgPosX - (trSizeCG - 1)) >> 31) & sigPos;
+        const uint32_t sigLower = ((int32_t)(cgPosY - (trSizeCG - 1)) >> 31) & (sigPos >> (trSizeCG - 1));
+
+        return (sigRight | sigLower) & 1;
+    }
+
     /* static methods shared with entropy.cpp */
-    static uint32_t calcPatternSigCtx(uint64_t sigCoeffGroupFlag64, uint32_t cgPosX, uint32_t cgPosY, uint32_t log2TrSizeCG);
     static uint32_t getSigCtxInc(uint32_t patternSigCtx, uint32_t log2TrSize, uint32_t trSize, uint32_t blkPos, bool bIsLuma, uint32_t firstSignificanceMapContext);
-    static uint32_t getSigCoeffGroupCtxInc(uint64_t sigCoeffGroupFlag64, uint32_t cgPosX, uint32_t cgPosY, uint32_t log2TrSizeCG);
 
 protected:

x265_1.6.tar.gz/source/common/slice.h -> x265_1.7.tar.gz/source/common/slice.h Changed

x265_1.6.tar.gz/source/common/threading.h -> x265_1.7.tar.gz/source/common/threading.h Changed

x265_1.6.tar.gz/source/common/threadpool.cpp -> x265_1.7.tar.gz/source/common/threadpool.cpp Changed

@@ -232,7 +232,7 @@
     int cpuCount = getCpuCount();
     bool bNumaSupport = false;
 
-#if _WIN32_WINNT >= 0x0601
+#if defined(_WIN32_WINNT) && _WIN32_WINNT >= _WIN32_WINNT_WIN7 
     bNumaSupport = true;
 #elif HAVE_LIBNUMA
     bNumaSupport = numa_available() >= 0;
@@ -241,10 +241,10 @@
 
     for (int i = 0; i < cpuCount; i++)
     {
-#if _WIN32_WINNT >= 0x0601
+#if defined(_WIN32_WINNT) && _WIN32_WINNT >= _WIN32_WINNT_WIN7 
         UCHAR node;
         if (GetNumaProcessorNode((UCHAR)i, &node))
-            cpusPerNode[X265_MIN(node, MAX_NODE_NUM)]++;
+            cpusPerNode[X265_MIN(node, (UCHAR)MAX_NODE_NUM)]++;
         else
 #elif HAVE_LIBNUMA
         if (bNumaSupport >= 0)
@@ -261,7 +261,7 @@
     /* limit nodes based on param->numaPools */
     if (p->numaPools && *p->numaPools)
     {
-        char *nodeStr = p->numaPools;
+        const char *nodeStr = p->numaPools;
         for (int i = 0; i < numNumaNodes; i++)
         {
             if (!*nodeStr)
@@ -373,7 +373,7 @@
     return true;
 }
 
-void ThreadPool::stop()
+void ThreadPool::stopWorkers()
 {
     if (m_workers)
     {
@@ -408,7 +408,7 @@
 /* static */
 void ThreadPool::setThreadNodeAffinity(int numaNode)
 {
-#if _WIN32_WINNT >= 0x0601
+#if defined(_WIN32_WINNT) && _WIN32_WINNT >= _WIN32_WINNT_WIN7 
     GROUP_AFFINITY groupAffinity;
     if (GetNumaNodeProcessorMaskEx((USHORT)numaNode, &groupAffinity))
     {
@@ -433,7 +433,7 @@
 /* static */
 int ThreadPool::getNumaNodeCount()
 {
-#if _WIN32_WINNT >= 0x0601
+#if defined(_WIN32_WINNT) && _WIN32_WINNT >= _WIN32_WINNT_WIN7 
     ULONG num = 1;
     if (GetNumaHighestNodeNumber(&num))
         num++;

x265_1.6.tar.gz/source/common/threadpool.h -> x265_1.7.tar.gz/source/common/threadpool.h Changed

x265_1.6.tar.gz/source/common/x86/asm-primitives.cpp -> x265_1.7.tar.gz/source/common/x86/asm-primitives.cpp Changed

@@ -800,6 +800,10 @@
 #error "Unsupported build configuration (32bit x86 and HIGH_BIT_DEPTH), you must configure ENABLE_ASSEMBLY=OFF"
 #endif
 
+#if X86_64
+    p.scanPosLast = x265_scanPosLast_x64;
+#endif
+
     if (cpuMask & X265_CPU_SSE2)
     {
         /* We do not differentiate CPUs which support MMX and not SSE2. We only check
@@ -859,9 +863,6 @@
         PIXEL_AVG_W4(mmx2);
         LUMA_VAR(sse2);
 
-        p.luma_p2s = x265_luma_p2s_sse2;
-        p.chroma[X265_CSP_I420].p2s = x265_chroma_p2s_sse2;
-        p.chroma[X265_CSP_I422].p2s = x265_chroma_p2s_sse2;
 
         ALL_LUMA_TU(blockfill_s, blockfill_s, sse2);
         ALL_LUMA_TU_S(cpy1Dto2D_shr, cpy1Dto2D_shr_, sse2);
@@ -872,15 +873,41 @@
         ALL_LUMA_TU_S(calcresidual, getResidual, sse2);
         ALL_LUMA_TU_S(transpose, transpose, sse2);
 
-        p.cu[BLOCK_4x4].intra_pred[DC_IDX] = x265_intra_pred_dc4_sse2;
-        p.cu[BLOCK_8x8].intra_pred[DC_IDX] = x265_intra_pred_dc8_sse2;
-        p.cu[BLOCK_16x16].intra_pred[DC_IDX] = x265_intra_pred_dc16_sse2;
-        p.cu[BLOCK_32x32].intra_pred[DC_IDX] = x265_intra_pred_dc32_sse2;
-
-        p.cu[BLOCK_4x4].intra_pred[PLANAR_IDX] = x265_intra_pred_planar4_sse2;
-        p.cu[BLOCK_8x8].intra_pred[PLANAR_IDX] = x265_intra_pred_planar8_sse2;
-        p.cu[BLOCK_16x16].intra_pred[PLANAR_IDX] = x265_intra_pred_planar16_sse2;
-        p.cu[BLOCK_32x32].intra_pred[PLANAR_IDX] = x265_intra_pred_planar32_sse2;
+        ALL_LUMA_TU_S(intra_pred[PLANAR_IDX], intra_pred_planar, sse2);
+        ALL_LUMA_TU_S(intra_pred[DC_IDX], intra_pred_dc, sse2);
+
+        p.cu[BLOCK_4x4].intra_pred[2] = x265_intra_pred_ang4_2_sse2;
+        p.cu[BLOCK_4x4].intra_pred[3] = x265_intra_pred_ang4_3_sse2;
+        p.cu[BLOCK_4x4].intra_pred[4] = x265_intra_pred_ang4_4_sse2;
+        p.cu[BLOCK_4x4].intra_pred[5] = x265_intra_pred_ang4_5_sse2;
+        p.cu[BLOCK_4x4].intra_pred[6] = x265_intra_pred_ang4_6_sse2;
+        p.cu[BLOCK_4x4].intra_pred[7] = x265_intra_pred_ang4_7_sse2;
+        p.cu[BLOCK_4x4].intra_pred[8] = x265_intra_pred_ang4_8_sse2;
+        p.cu[BLOCK_4x4].intra_pred[9] = x265_intra_pred_ang4_9_sse2;
+        p.cu[BLOCK_4x4].intra_pred[10] = x265_intra_pred_ang4_10_sse2;
+        p.cu[BLOCK_4x4].intra_pred[11] = x265_intra_pred_ang4_11_sse2;
+        p.cu[BLOCK_4x4].intra_pred[12] = x265_intra_pred_ang4_12_sse2;
+        p.cu[BLOCK_4x4].intra_pred[13] = x265_intra_pred_ang4_13_sse2;
+        p.cu[BLOCK_4x4].intra_pred[14] = x265_intra_pred_ang4_14_sse2;
+        p.cu[BLOCK_4x4].intra_pred[15] = x265_intra_pred_ang4_15_sse2;
+        p.cu[BLOCK_4x4].intra_pred[16] = x265_intra_pred_ang4_16_sse2;
+        p.cu[BLOCK_4x4].intra_pred[17] = x265_intra_pred_ang4_17_sse2;
+        p.cu[BLOCK_4x4].intra_pred[18] = x265_intra_pred_ang4_18_sse2;
+        p.cu[BLOCK_4x4].intra_pred[19] = x265_intra_pred_ang4_17_sse2;
+        p.cu[BLOCK_4x4].intra_pred[20] = x265_intra_pred_ang4_16_sse2;
+        p.cu[BLOCK_4x4].intra_pred[21] = x265_intra_pred_ang4_15_sse2;
+        p.cu[BLOCK_4x4].intra_pred[22] = x265_intra_pred_ang4_14_sse2;
+        p.cu[BLOCK_4x4].intra_pred[23] = x265_intra_pred_ang4_13_sse2;
+        p.cu[BLOCK_4x4].intra_pred[24] = x265_intra_pred_ang4_12_sse2;
+        p.cu[BLOCK_4x4].intra_pred[25] = x265_intra_pred_ang4_11_sse2;
+        p.cu[BLOCK_4x4].intra_pred[26] = x265_intra_pred_ang4_26_sse2;
+        p.cu[BLOCK_4x4].intra_pred[27] = x265_intra_pred_ang4_9_sse2;
+        p.cu[BLOCK_4x4].intra_pred[28] = x265_intra_pred_ang4_8_sse2;
+        p.cu[BLOCK_4x4].intra_pred[29] = x265_intra_pred_ang4_7_sse2;
+        p.cu[BLOCK_4x4].intra_pred[30] = x265_intra_pred_ang4_6_sse2;
+        p.cu[BLOCK_4x4].intra_pred[31] = x265_intra_pred_ang4_5_sse2;
+        p.cu[BLOCK_4x4].intra_pred[32] = x265_intra_pred_ang4_4_sse2;
+        p.cu[BLOCK_4x4].intra_pred[33] = x265_intra_pred_ang4_3_sse2;
 
         p.cu[BLOCK_4x4].sse_ss = x265_pixel_ssd_ss_4x4_mmx2;
         ALL_LUMA_CU(sse_ss, pixel_ssd_ss, sse2);
@@ -918,6 +945,74 @@
         p.cu[BLOCK_16x16].count_nonzero = x265_count_nonzero_16x16_ssse3;
         p.cu[BLOCK_32x32].count_nonzero = x265_count_nonzero_32x32_ssse3;
         p.frameInitLowres = x265_frame_init_lowres_core_ssse3;
+
+        p.pu[LUMA_4x4].convert_p2s = x265_filterPixelToShort_4x4_ssse3;
+        p.pu[LUMA_4x8].convert_p2s = x265_filterPixelToShort_4x8_ssse3;
+        p.pu[LUMA_4x16].convert_p2s = x265_filterPixelToShort_4x16_ssse3;
+        p.pu[LUMA_8x4].convert_p2s = x265_filterPixelToShort_8x4_ssse3;
+        p.pu[LUMA_8x8].convert_p2s = x265_filterPixelToShort_8x8_ssse3;
+        p.pu[LUMA_8x16].convert_p2s = x265_filterPixelToShort_8x16_ssse3;
+        p.pu[LUMA_8x32].convert_p2s = x265_filterPixelToShort_8x32_ssse3;
+        p.pu[LUMA_16x4].convert_p2s = x265_filterPixelToShort_16x4_ssse3;
+        p.pu[LUMA_16x8].convert_p2s = x265_filterPixelToShort_16x8_ssse3;
+        p.pu[LUMA_16x12].convert_p2s = x265_filterPixelToShort_16x12_ssse3;
+        p.pu[LUMA_16x16].convert_p2s = x265_filterPixelToShort_16x16_ssse3;
+        p.pu[LUMA_16x32].convert_p2s = x265_filterPixelToShort_16x32_ssse3;
+        p.pu[LUMA_16x64].convert_p2s = x265_filterPixelToShort_16x64_ssse3;
+        p.pu[LUMA_32x8].convert_p2s = x265_filterPixelToShort_32x8_ssse3;
+        p.pu[LUMA_32x16].convert_p2s = x265_filterPixelToShort_32x16_ssse3;
+        p.pu[LUMA_32x24].convert_p2s = x265_filterPixelToShort_32x24_ssse3;
+        p.pu[LUMA_32x32].convert_p2s = x265_filterPixelToShort_32x32_ssse3;
+        p.pu[LUMA_32x64].convert_p2s = x265_filterPixelToShort_32x64_ssse3;
+        p.pu[LUMA_64x16].convert_p2s = x265_filterPixelToShort_64x16_ssse3;
+        p.pu[LUMA_64x32].convert_p2s = x265_filterPixelToShort_64x32_ssse3;
+        p.pu[LUMA_64x48].convert_p2s = x265_filterPixelToShort_64x48_ssse3;
+        p.pu[LUMA_64x64].convert_p2s = x265_filterPixelToShort_64x64_ssse3;
+        p.pu[LUMA_24x32].convert_p2s = x265_filterPixelToShort_24x32_ssse3;
+        p.pu[LUMA_12x16].convert_p2s = x265_filterPixelToShort_12x16_ssse3;
+        p.pu[LUMA_48x64].convert_p2s = x265_filterPixelToShort_48x64_ssse3;
+
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x4].p2s = x265_filterPixelToShort_4x4_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x8].p2s = x265_filterPixelToShort_4x8_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x16].p2s = x265_filterPixelToShort_4x16_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x4].p2s = x265_filterPixelToShort_8x4_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].p2s = x265_filterPixelToShort_8x8_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].p2s = x265_filterPixelToShort_8x16_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x32].p2s = x265_filterPixelToShort_8x32_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x4].p2s = x265_filterPixelToShort_16x4_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x8].p2s = x265_filterPixelToShort_16x8_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x12].p2s = x265_filterPixelToShort_16x12_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x16].p2s = x265_filterPixelToShort_16x16_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x32].p2s = x265_filterPixelToShort_16x32_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].p2s = x265_filterPixelToShort_32x8_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].p2s = x265_filterPixelToShort_32x16_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x24].p2s = x265_filterPixelToShort_32x24_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x32].p2s = x265_filterPixelToShort_32x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].p2s = x265_filterPixelToShort_4x4_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].p2s = x265_filterPixelToShort_4x8_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].p2s = x265_filterPixelToShort_4x16_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x32].p2s = x265_filterPixelToShort_4x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].p2s = x265_filterPixelToShort_8x4_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].p2s = x265_filterPixelToShort_8x8_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].p2s = x265_filterPixelToShort_8x12_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].p2s = x265_filterPixelToShort_8x16_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].p2s = x265_filterPixelToShort_8x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].p2s = x265_filterPixelToShort_8x64_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].p2s = x265_filterPixelToShort_12x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].p2s = x265_filterPixelToShort_16x8_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].p2s = x265_filterPixelToShort_16x16_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].p2s = x265_filterPixelToShort_16x24_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].p2s = x265_filterPixelToShort_16x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].p2s = x265_filterPixelToShort_16x64_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].p2s = x265_filterPixelToShort_24x64_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].p2s = x265_filterPixelToShort_32x16_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].p2s = x265_filterPixelToShort_32x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].p2s = x265_filterPixelToShort_32x48_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].p2s = x265_filterPixelToShort_32x64_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x2].p2s = x265_filterPixelToShort_4x2_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x2].p2s = x265_filterPixelToShort_8x2_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x6].p2s = x265_filterPixelToShort_8x6_ssse3;
+        p.findPosFirstLast = x265_findPosFirstLast_ssse3;
     }
     if (cpuMask & X265_CPU_SSE4)
     {
@@ -957,6 +1052,13 @@
         ALL_LUMA_TU_S(copy_cnt, copy_cnt_, sse4);
         ALL_LUMA_CU(psy_cost_pp, psyCost_pp, sse4);
         ALL_LUMA_CU(psy_cost_ss, psyCost_ss, sse4);
+
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_2x4].p2s = x265_filterPixelToShort_2x4_sse4;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_2x8].p2s = x265_filterPixelToShort_2x8_sse4;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_6x8].p2s = x265_filterPixelToShort_6x8_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].p2s = x265_filterPixelToShort_2x8_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].p2s = x265_filterPixelToShort_2x16_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_6x16].p2s = x265_filterPixelToShort_6x16_sse4;
     }
     if (cpuMask & X265_CPU_AVX)
     {
@@ -1079,6 +1181,26 @@
     }
     if (cpuMask & X265_CPU_AVX2)
     {
+        p.pu[LUMA_48x64].satd = x265_pixel_satd_48x64_avx2;
+
+        p.pu[LUMA_64x16].satd = x265_pixel_satd_64x16_avx2;
+        p.pu[LUMA_64x32].satd = x265_pixel_satd_64x32_avx2;
+        p.pu[LUMA_64x48].satd = x265_pixel_satd_64x48_avx2;
+        p.pu[LUMA_64x64].satd = x265_pixel_satd_64x64_avx2;
+
+        p.pu[LUMA_32x8].satd = x265_pixel_satd_32x8_avx2;
+        p.pu[LUMA_32x16].satd = x265_pixel_satd_32x16_avx2;
+        p.pu[LUMA_32x24].satd = x265_pixel_satd_32x24_avx2;
+        p.pu[LUMA_32x32].satd = x265_pixel_satd_32x32_avx2;
+        p.pu[LUMA_32x64].satd = x265_pixel_satd_32x64_avx2;
+
+        p.pu[LUMA_16x4].satd = x265_pixel_satd_16x4_avx2;
+        p.pu[LUMA_16x8].satd = x265_pixel_satd_16x8_avx2;
+        p.pu[LUMA_16x12].satd = x265_pixel_satd_16x12_avx2;
+        p.pu[LUMA_16x16].satd = x265_pixel_satd_16x16_avx2;
+        p.pu[LUMA_16x32].satd = x265_pixel_satd_16x32_avx2;
+        p.pu[LUMA_16x64].satd = x265_pixel_satd_16x64_avx2;
+
         p.cu[BLOCK_32x32].ssd_s = x265_pixel_ssd_s_32_avx2;
         p.cu[BLOCK_16x16].sse_ss = x265_pixel_ssd_ss_16x16_avx2;
 
@@ -1087,6 +1209,7 @@
         p.dequant_normal  = x265_dequant_normal_avx2;
 
         p.scale1D_128to64 = x265_scale1D_128to64_avx2;
+        p.scale2D_64to32 = x265_scale2D_64to32_avx2;
         // p.weight_pp = x265_weight_pp_avx2; fails tests
 
         p.cu[BLOCK_16x16].calcresidual = x265_getResidual16_avx2;
@@ -1119,12 +1242,84 @@
         ALL_LUMA_PU(luma_vps, interp_8tap_vert_ps, avx2);
         ALL_LUMA_PU(luma_vsp, interp_8tap_vert_sp, avx2);
         ALL_LUMA_PU(luma_vss, interp_8tap_vert_ss, avx2);
+
+        p.cu[BLOCK_16x16].add_ps = x265_pixel_add_ps_16x16_avx2;
+        p.cu[BLOCK_32x32].add_ps = x265_pixel_add_ps_32x32_avx2;
+        p.cu[BLOCK_64x64].add_ps = x265_pixel_add_ps_64x64_avx2;
+        p.chroma[X265_CSP_I420].cu[BLOCK_420_16x16].add_ps = x265_pixel_add_ps_16x16_avx2;
+        p.chroma[X265_CSP_I420].cu[BLOCK_420_32x32].add_ps = x265_pixel_add_ps_32x32_avx2;
+        p.chroma[X265_CSP_I422].cu[BLOCK_422_16x32].add_ps = x265_pixel_add_ps_16x32_avx2;
+        p.chroma[X265_CSP_I422].cu[BLOCK_422_32x64].add_ps = x265_pixel_add_ps_32x64_avx2;
+
+        p.cu[BLOCK_16x16].sub_ps = x265_pixel_sub_ps_16x16_avx2;
+        p.cu[BLOCK_32x32].sub_ps = x265_pixel_sub_ps_32x32_avx2;
+        p.cu[BLOCK_64x64].sub_ps = x265_pixel_sub_ps_64x64_avx2;
+        p.chroma[X265_CSP_I420].cu[BLOCK_420_16x16].sub_ps = x265_pixel_sub_ps_16x16_avx2;
+        p.chroma[X265_CSP_I420].cu[BLOCK_420_32x32].sub_ps = x265_pixel_sub_ps_32x32_avx2;
+        p.chroma[X265_CSP_I422].cu[BLOCK_422_16x32].sub_ps = x265_pixel_sub_ps_16x32_avx2;
+        p.chroma[X265_CSP_I422].cu[BLOCK_422_32x64].sub_ps = x265_pixel_sub_ps_32x64_avx2;
+
+        p.pu[LUMA_16x4].sad = x265_pixel_sad_16x4_avx2;
+        p.pu[LUMA_16x8].sad = x265_pixel_sad_16x8_avx2;
+        p.pu[LUMA_16x12].sad = x265_pixel_sad_16x12_avx2;
+        p.pu[LUMA_16x16].sad = x265_pixel_sad_16x16_avx2;
+        p.pu[LUMA_16x32].sad = x265_pixel_sad_16x32_avx2;
+
+        p.pu[LUMA_16x4].convert_p2s = x265_filterPixelToShort_16x4_avx2;
+        p.pu[LUMA_16x8].convert_p2s = x265_filterPixelToShort_16x8_avx2;
+        p.pu[LUMA_16x12].convert_p2s = x265_filterPixelToShort_16x12_avx2;
+        p.pu[LUMA_16x16].convert_p2s = x265_filterPixelToShort_16x16_avx2;
+        p.pu[LUMA_16x32].convert_p2s = x265_filterPixelToShort_16x32_avx2;
+        p.pu[LUMA_16x64].convert_p2s = x265_filterPixelToShort_16x64_avx2;
+        p.pu[LUMA_32x8].convert_p2s = x265_filterPixelToShort_32x8_avx2;
+        p.pu[LUMA_32x16].convert_p2s = x265_filterPixelToShort_32x16_avx2;
+        p.pu[LUMA_32x24].convert_p2s = x265_filterPixelToShort_32x24_avx2;
+        p.pu[LUMA_32x32].convert_p2s = x265_filterPixelToShort_32x32_avx2;
+        p.pu[LUMA_32x64].convert_p2s = x265_filterPixelToShort_32x64_avx2;
+        p.pu[LUMA_64x16].convert_p2s = x265_filterPixelToShort_64x16_avx2;
+        p.pu[LUMA_64x32].convert_p2s = x265_filterPixelToShort_64x32_avx2;
+        p.pu[LUMA_64x48].convert_p2s = x265_filterPixelToShort_64x48_avx2;
+        p.pu[LUMA_64x64].convert_p2s = x265_filterPixelToShort_64x64_avx2;
+        p.pu[LUMA_24x32].convert_p2s = x265_filterPixelToShort_24x32_avx2;
+        p.pu[LUMA_48x64].convert_p2s = x265_filterPixelToShort_48x64_avx2;
+
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x4].p2s = x265_filterPixelToShort_16x4_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x8].p2s = x265_filterPixelToShort_16x8_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x12].p2s = x265_filterPixelToShort_16x12_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x16].p2s = x265_filterPixelToShort_16x16_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x32].p2s = x265_filterPixelToShort_16x32_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_24x32].p2s = x265_filterPixelToShort_24x32_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].p2s = x265_filterPixelToShort_32x8_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].p2s = x265_filterPixelToShort_32x16_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x24].p2s = x265_filterPixelToShort_32x24_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x32].p2s = x265_filterPixelToShort_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].p2s = x265_filterPixelToShort_16x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].p2s = x265_filterPixelToShort_16x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].p2s = x265_filterPixelToShort_16x24_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].p2s = x265_filterPixelToShort_16x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].p2s = x265_filterPixelToShort_16x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].p2s = x265_filterPixelToShort_24x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].p2s = x265_filterPixelToShort_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].p2s = x265_filterPixelToShort_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].p2s = x265_filterPixelToShort_32x48_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].p2s = x265_filterPixelToShort_32x64_avx2;
+
+        p.pu[LUMA_4x4].luma_hps = x265_interp_8tap_horiz_ps_4x4_avx2;
+        p.pu[LUMA_4x8].luma_hps = x265_interp_8tap_horiz_ps_4x8_avx2;
+        p.pu[LUMA_4x16].luma_hps = x265_interp_8tap_horiz_ps_4x16_avx2;
+
+        if (cpuMask & X265_CPU_BMI2)
+            p.scanPosLast = x265_scanPosLast_avx2_bmi2;
     }
 }
 #else // if HIGH_BIT_DEPTH
 
 void setupAssemblyPrimitives(EncoderPrimitives &p, int cpuMask) // 8bpp
 {
+#if X86_64
+    p.scanPosLast = x265_scanPosLast_x64;
+#endif
+
     if (cpuMask & X265_CPU_SSE2)
     {
         /* We do not differentiate CPUs which support MMX and not SSE2. We only check
@@ -1175,6 +1370,47 @@
         CHROMA_420_VSP_FILTERS(_sse2);
         CHROMA_422_VSP_FILTERS(_sse2);
         CHROMA_444_VSP_FILTERS(_sse2);
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_2x4].filter_vpp = x265_interp_4tap_vert_pp_2x4_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_2x8].filter_vpp = x265_interp_4tap_vert_pp_2x8_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x2].filter_vpp = x265_interp_4tap_vert_pp_4x2_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x4].filter_vpp = x265_interp_4tap_vert_pp_4x4_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x8].filter_vpp = x265_interp_4tap_vert_pp_4x8_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x16].filter_vpp = x265_interp_4tap_vert_pp_4x16_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].filter_vpp = x265_interp_4tap_vert_pp_2x16_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].filter_vpp = x265_interp_4tap_vert_pp_4x4_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].filter_vpp = x265_interp_4tap_vert_pp_4x8_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].filter_vpp = x265_interp_4tap_vert_pp_4x16_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x32].filter_vpp = x265_interp_4tap_vert_pp_4x32_sse2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_vpp = x265_interp_4tap_vert_pp_4x4_sse2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x8].filter_vpp = x265_interp_4tap_vert_pp_4x8_sse2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x16].filter_vpp = x265_interp_4tap_vert_pp_4x16_sse2;
+#if X86_64
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_6x8].filter_vpp = x265_interp_4tap_vert_pp_6x8_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x2].filter_vpp = x265_interp_4tap_vert_pp_8x2_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x4].filter_vpp = x265_interp_4tap_vert_pp_8x4_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x6].filter_vpp = x265_interp_4tap_vert_pp_8x6_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].filter_vpp = x265_interp_4tap_vert_pp_8x8_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].filter_vpp = x265_interp_4tap_vert_pp_8x16_sse2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x32].filter_vpp = x265_interp_4tap_vert_pp_8x32_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_6x16].filter_vpp = x265_interp_4tap_vert_pp_6x16_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].filter_vpp = x265_interp_4tap_vert_pp_8x4_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].filter_vpp = x265_interp_4tap_vert_pp_8x4_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].filter_vpp = x265_interp_4tap_vert_pp_8x8_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].filter_vpp = x265_interp_4tap_vert_pp_8x12_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].filter_vpp = x265_interp_4tap_vert_pp_8x16_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].filter_vpp = x265_interp_4tap_vert_pp_8x32_sse2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].filter_vpp = x265_interp_4tap_vert_pp_8x64_sse2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x4].filter_vpp = x265_interp_4tap_vert_pp_8x4_sse2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x8].filter_vpp = x265_interp_4tap_vert_pp_8x8_sse2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x16].filter_vpp = x265_interp_4tap_vert_pp_8x16_sse2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x32].filter_vpp = x265_interp_4tap_vert_pp_8x32_sse2;
+#endif
+
+        ALL_LUMA_PU(luma_hpp, interp_8tap_horiz_pp, sse2);
+        p.pu[LUMA_4x4].luma_hpp = x265_interp_8tap_horiz_pp_4x4_sse2;
+        ALL_LUMA_PU(luma_hps, interp_8tap_horiz_ps, sse2);
+        p.pu[LUMA_4x4].luma_hps = x265_interp_8tap_horiz_ps_4x4_sse2;
+        p.pu[LUMA_8x8].luma_hvpp = x265_interp_8tap_hv_pp_8x8_sse3;
 
         //p.frameInitLowres = x265_frame_init_lowres_core_mmx2;
         p.frameInitLowres = x265_frame_init_lowres_core_sse2;
@@ -1186,15 +1422,8 @@
         ALL_LUMA_TU_S(cpy1Dto2D_shr, cpy1Dto2D_shr_, sse2);
         ALL_LUMA_TU_S(ssd_s, pixel_ssd_s_, sse2);
 
-        p.cu[BLOCK_4x4].intra_pred[DC_IDX] = x265_intra_pred_dc4_sse2;
-        p.cu[BLOCK_8x8].intra_pred[DC_IDX] = x265_intra_pred_dc8_sse2;
-        p.cu[BLOCK_16x16].intra_pred[DC_IDX] = x265_intra_pred_dc16_sse2;
-        p.cu[BLOCK_32x32].intra_pred[DC_IDX] = x265_intra_pred_dc32_sse2;
-
-        p.cu[BLOCK_4x4].intra_pred[PLANAR_IDX] = x265_intra_pred_planar4_sse2;
-        p.cu[BLOCK_8x8].intra_pred[PLANAR_IDX] = x265_intra_pred_planar8_sse2;
-        p.cu[BLOCK_16x16].intra_pred[PLANAR_IDX] = x265_intra_pred_planar16_sse2;
-        p.cu[BLOCK_32x32].intra_pred[PLANAR_IDX] = x265_intra_pred_planar32_sse2;
+        ALL_LUMA_TU_S(intra_pred[PLANAR_IDX], intra_pred_planar, sse2);
+        ALL_LUMA_TU_S(intra_pred[DC_IDX], intra_pred_dc, sse2);
 
         p.cu[BLOCK_4x4].intra_pred[2] = x265_intra_pred_ang4_2_sse2;
         p.cu[BLOCK_4x4].intra_pred[3] = x265_intra_pred_ang4_3_sse2;
@@ -1204,6 +1433,32 @@
         p.cu[BLOCK_4x4].intra_pred[7] = x265_intra_pred_ang4_7_sse2;
         p.cu[BLOCK_4x4].intra_pred[8] = x265_intra_pred_ang4_8_sse2;
         p.cu[BLOCK_4x4].intra_pred[9] = x265_intra_pred_ang4_9_sse2;
+        p.cu[BLOCK_4x4].intra_pred[10] = x265_intra_pred_ang4_10_sse2;
+        p.cu[BLOCK_4x4].intra_pred[11] = x265_intra_pred_ang4_11_sse2;
+        p.cu[BLOCK_4x4].intra_pred[12] = x265_intra_pred_ang4_12_sse2;
+        p.cu[BLOCK_4x4].intra_pred[13] = x265_intra_pred_ang4_13_sse2;
+        p.cu[BLOCK_4x4].intra_pred[14] = x265_intra_pred_ang4_14_sse2;
+        p.cu[BLOCK_4x4].intra_pred[15] = x265_intra_pred_ang4_15_sse2;
+        p.cu[BLOCK_4x4].intra_pred[16] = x265_intra_pred_ang4_16_sse2;
+        p.cu[BLOCK_4x4].intra_pred[17] = x265_intra_pred_ang4_17_sse2;
+        p.cu[BLOCK_4x4].intra_pred[18] = x265_intra_pred_ang4_18_sse2;
+        p.cu[BLOCK_4x4].intra_pred[19] = x265_intra_pred_ang4_17_sse2;
+        p.cu[BLOCK_4x4].intra_pred[20] = x265_intra_pred_ang4_16_sse2;
+        p.cu[BLOCK_4x4].intra_pred[21] = x265_intra_pred_ang4_15_sse2;
+        p.cu[BLOCK_4x4].intra_pred[22] = x265_intra_pred_ang4_14_sse2;
+        p.cu[BLOCK_4x4].intra_pred[23] = x265_intra_pred_ang4_13_sse2;
+        p.cu[BLOCK_4x4].intra_pred[24] = x265_intra_pred_ang4_12_sse2;
+        p.cu[BLOCK_4x4].intra_pred[25] = x265_intra_pred_ang4_11_sse2;
+        p.cu[BLOCK_4x4].intra_pred[26] = x265_intra_pred_ang4_26_sse2;
+        p.cu[BLOCK_4x4].intra_pred[27] = x265_intra_pred_ang4_9_sse2;
+        p.cu[BLOCK_4x4].intra_pred[28] = x265_intra_pred_ang4_8_sse2;
+        p.cu[BLOCK_4x4].intra_pred[29] = x265_intra_pred_ang4_7_sse2;
+        p.cu[BLOCK_4x4].intra_pred[30] = x265_intra_pred_ang4_6_sse2;
+        p.cu[BLOCK_4x4].intra_pred[31] = x265_intra_pred_ang4_5_sse2;
+        p.cu[BLOCK_4x4].intra_pred[32] = x265_intra_pred_ang4_4_sse2;
+        p.cu[BLOCK_4x4].intra_pred[33] = x265_intra_pred_ang4_3_sse2;
+
+        p.cu[BLOCK_4x4].intra_pred_allangs = x265_all_angs_pred_4x4_sse2;
 
         p.cu[BLOCK_4x4].calcresidual = x265_getResidual4_sse2;
         p.cu[BLOCK_8x8].calcresidual = x265_getResidual8_sse2;
@@ -1224,6 +1479,12 @@
 
         p.planecopy_sp = x265_downShift_16_sse2;
     }
+    if (cpuMask & X265_CPU_SSE3)
+    {
+        ALL_CHROMA_420_PU(filter_hpp, interp_4tap_horiz_pp, sse3);
+        ALL_CHROMA_422_PU(filter_hpp, interp_4tap_horiz_pp, sse3);
+        ALL_CHROMA_444_PU(filter_hpp, interp_4tap_horiz_pp, sse3);
+    }
     if (cpuMask & X265_CPU_SSSE3)
     {
         p.pu[LUMA_8x16].sad_x3 = x265_pixel_sad_x3_8x16_ssse3;
@@ -1249,48 +1510,86 @@
         ASSIGN_SSE_PP(ssse3);
         p.cu[BLOCK_4x4].sse_pp = x265_pixel_ssd_4x4_ssse3;
         p.chroma[X265_CSP_I422].cu[BLOCK_422_4x8].sse_pp = x265_pixel_ssd_4x8_ssse3;
-        p.pu[LUMA_4x4].filter_p2s = x265_pixelToShort_4x4_ssse3;
-        p.pu[LUMA_4x8].filter_p2s = x265_pixelToShort_4x8_ssse3;
-        p.pu[LUMA_4x16].filter_p2s = x265_pixelToShort_4x16_ssse3;
-        p.pu[LUMA_8x4].filter_p2s = x265_pixelToShort_8x4_ssse3;
-        p.pu[LUMA_8x8].filter_p2s = x265_pixelToShort_8x8_ssse3;
-        p.pu[LUMA_8x16].filter_p2s = x265_pixelToShort_8x16_ssse3;
-        p.pu[LUMA_8x32].filter_p2s = x265_pixelToShort_8x32_ssse3;
-        p.pu[LUMA_16x4].filter_p2s = x265_pixelToShort_16x4_ssse3;
-        p.pu[LUMA_16x8].filter_p2s = x265_pixelToShort_16x8_ssse3;
-        p.pu[LUMA_16x12].filter_p2s = x265_pixelToShort_16x12_ssse3;
-        p.pu[LUMA_16x16].filter_p2s = x265_pixelToShort_16x16_ssse3;
-        p.pu[LUMA_16x32].filter_p2s = x265_pixelToShort_16x32_ssse3;
-        p.pu[LUMA_16x64].filter_p2s = x265_pixelToShort_16x64_ssse3;
-        p.pu[LUMA_32x8].filter_p2s = x265_pixelToShort_32x8_ssse3;
-        p.pu[LUMA_32x16].filter_p2s = x265_pixelToShort_32x16_ssse3;
-        p.pu[LUMA_32x24].filter_p2s = x265_pixelToShort_32x24_ssse3;
-        p.pu[LUMA_32x32].filter_p2s = x265_pixelToShort_32x32_ssse3;
-        p.pu[LUMA_32x64].filter_p2s = x265_pixelToShort_32x64_ssse3;
-        p.pu[LUMA_64x16].filter_p2s = x265_pixelToShort_64x16_ssse3;
-        p.pu[LUMA_64x32].filter_p2s = x265_pixelToShort_64x32_ssse3;
-        p.pu[LUMA_64x48].filter_p2s = x265_pixelToShort_64x48_ssse3;
-        p.pu[LUMA_64x64].filter_p2s = x265_pixelToShort_64x64_ssse3;
-
-        p.chroma[X265_CSP_I420].p2s = x265_chroma_p2s_ssse3;
-        p.chroma[X265_CSP_I422].p2s = x265_chroma_p2s_ssse3;
 
         p.dst4x4 = x265_dst4_ssse3;
         p.cu[BLOCK_8x8].idct = x265_idct8_ssse3;
 
         ALL_LUMA_TU(count_nonzero, count_nonzero, ssse3);
 
+        // MUST be done after LUMA_FILTERS() to overwrite default version
+        p.pu[LUMA_8x8].luma_hvpp = x265_interp_8tap_hv_pp_8x8_ssse3;
+
         p.frameInitLowres = x265_frame_init_lowres_core_ssse3;
         p.scale1D_128to64 = x265_scale1D_128to64_ssse3;
         p.scale2D_64to32 = x265_scale2D_64to32_ssse3;
+
+        p.pu[LUMA_8x4].convert_p2s = x265_filterPixelToShort_8x4_ssse3;
+        p.pu[LUMA_8x8].convert_p2s = x265_filterPixelToShort_8x8_ssse3;
+        p.pu[LUMA_8x16].convert_p2s = x265_filterPixelToShort_8x16_ssse3;
+        p.pu[LUMA_8x32].convert_p2s = x265_filterPixelToShort_8x32_ssse3;
+        p.pu[LUMA_16x4].convert_p2s = x265_filterPixelToShort_16x4_ssse3;
+        p.pu[LUMA_16x8].convert_p2s = x265_filterPixelToShort_16x8_ssse3;
+        p.pu[LUMA_16x12].convert_p2s = x265_filterPixelToShort_16x12_ssse3;
+        p.pu[LUMA_16x16].convert_p2s = x265_filterPixelToShort_16x16_ssse3;
+        p.pu[LUMA_16x32].convert_p2s = x265_filterPixelToShort_16x32_ssse3;
+        p.pu[LUMA_16x64].convert_p2s = x265_filterPixelToShort_16x64_ssse3;
+        p.pu[LUMA_32x8].convert_p2s = x265_filterPixelToShort_32x8_ssse3;
+        p.pu[LUMA_32x16].convert_p2s = x265_filterPixelToShort_32x16_ssse3;
+        p.pu[LUMA_32x24].convert_p2s = x265_filterPixelToShort_32x24_ssse3;
+        p.pu[LUMA_32x32].convert_p2s = x265_filterPixelToShort_32x32_ssse3;
+        p.pu[LUMA_32x64].convert_p2s = x265_filterPixelToShort_32x64_ssse3;
+        p.pu[LUMA_64x16].convert_p2s = x265_filterPixelToShort_64x16_ssse3;
+        p.pu[LUMA_64x32].convert_p2s = x265_filterPixelToShort_64x32_ssse3;
+        p.pu[LUMA_64x48].convert_p2s = x265_filterPixelToShort_64x48_ssse3;
+        p.pu[LUMA_64x64].convert_p2s = x265_filterPixelToShort_64x64_ssse3;
+        p.pu[LUMA_12x16].convert_p2s = x265_filterPixelToShort_12x16_ssse3;
+        p.pu[LUMA_24x32].convert_p2s = x265_filterPixelToShort_24x32_ssse3;
+        p.pu[LUMA_48x64].convert_p2s = x265_filterPixelToShort_48x64_ssse3;
+
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x2].p2s = x265_filterPixelToShort_8x2_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x4].p2s = x265_filterPixelToShort_8x4_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x6].p2s = x265_filterPixelToShort_8x6_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].p2s = x265_filterPixelToShort_8x8_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].p2s = x265_filterPixelToShort_8x16_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x32].p2s = x265_filterPixelToShort_8x32_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x4].p2s = x265_filterPixelToShort_16x4_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x8].p2s = x265_filterPixelToShort_16x8_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x12].p2s = x265_filterPixelToShort_16x12_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x16].p2s = x265_filterPixelToShort_16x16_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_16x32].p2s = x265_filterPixelToShort_16x32_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].p2s = x265_filterPixelToShort_32x8_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].p2s = x265_filterPixelToShort_32x16_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x24].p2s = x265_filterPixelToShort_32x24_ssse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x32].p2s = x265_filterPixelToShort_32x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].p2s = x265_filterPixelToShort_8x4_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].p2s = x265_filterPixelToShort_8x8_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].p2s = x265_filterPixelToShort_8x12_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].p2s = x265_filterPixelToShort_8x16_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].p2s = x265_filterPixelToShort_8x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].p2s = x265_filterPixelToShort_8x64_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].p2s = x265_filterPixelToShort_12x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].p2s = x265_filterPixelToShort_16x8_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].p2s = x265_filterPixelToShort_16x16_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].p2s = x265_filterPixelToShort_16x24_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].p2s = x265_filterPixelToShort_16x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].p2s = x265_filterPixelToShort_16x64_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].p2s = x265_filterPixelToShort_24x64_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].p2s = x265_filterPixelToShort_32x16_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].p2s = x265_filterPixelToShort_32x32_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].p2s = x265_filterPixelToShort_32x48_ssse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].p2s = x265_filterPixelToShort_32x64_ssse3;
+        p.findPosFirstLast = x265_findPosFirstLast_ssse3;
     }
     if (cpuMask & X265_CPU_SSE4)
     {
         p.sign = x265_calSign_sse4;
         p.saoCuOrgE0 = x265_saoCuOrgE0_sse4;
         p.saoCuOrgE1 = x265_saoCuOrgE1_sse4;
-        p.saoCuOrgE2 = x265_saoCuOrgE2_sse4;
-        p.saoCuOrgE3 = x265_saoCuOrgE3_sse4;
+        p.saoCuOrgE1_2Rows = x265_saoCuOrgE1_2Rows_sse4;
+        p.saoCuOrgE2[0] = x265_saoCuOrgE2_sse4;
+        p.saoCuOrgE2[1] = x265_saoCuOrgE2_sse4;
+        p.saoCuOrgE3[0] = x265_saoCuOrgE3_sse4;
+        p.saoCuOrgE3[1] = x265_saoCuOrgE3_sse4;
         p.saoCuOrgB0 = x265_saoCuOrgB0_sse4;
 
         LUMA_ADDAVG(sse4);
@@ -1321,7 +1620,7 @@
         CHROMA_444_VSP_FILTERS_SSE4(_sse4);
 
         // MUST be done after LUMA_FILTERS() to overwrite default version
-        p.pu[LUMA_8x8].luma_hvpp = x265_interp_8tap_hv_pp_8x8_sse4;
+        p.pu[LUMA_8x8].luma_hvpp = x265_interp_8tap_hv_pp_8x8_ssse3;
 
         LUMA_CU_BLOCKCOPY(ps, sse4);
         CHROMA_420_CU_BLOCKCOPY(ps, sse4);
@@ -1348,6 +1647,25 @@
         p.cu[BLOCK_4x4].psy_cost_pp = x265_psyCost_pp_4x4_sse4;
         p.cu[BLOCK_4x4].psy_cost_ss = x265_psyCost_ss_4x4_sse4;
 
+        p.pu[LUMA_4x4].convert_p2s = x265_filterPixelToShort_4x4_sse4;
+        p.pu[LUMA_4x8].convert_p2s = x265_filterPixelToShort_4x8_sse4;
+        p.pu[LUMA_4x16].convert_p2s = x265_filterPixelToShort_4x16_sse4;
+
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_2x4].p2s = x265_filterPixelToShort_2x4_sse4;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_2x8].p2s = x265_filterPixelToShort_2x8_sse4;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x2].p2s = x265_filterPixelToShort_4x2_sse4;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x4].p2s = x265_filterPixelToShort_4x4_sse4;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x8].p2s = x265_filterPixelToShort_4x8_sse4;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x16].p2s = x265_filterPixelToShort_4x16_sse4;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_6x8].p2s = x265_filterPixelToShort_6x8_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].p2s = x265_filterPixelToShort_2x8_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].p2s = x265_filterPixelToShort_2x16_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].p2s = x265_filterPixelToShort_4x4_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].p2s = x265_filterPixelToShort_4x8_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].p2s = x265_filterPixelToShort_4x16_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x32].p2s = x265_filterPixelToShort_4x32_sse4;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_6x16].p2s = x265_filterPixelToShort_6x16_sse4;
+
 #if X86_64
         ALL_LUMA_CU(psy_cost_pp, psyCost_pp, sse4);
         ALL_LUMA_CU(psy_cost_ss, psyCost_ss, sse4);
@@ -1363,6 +1681,20 @@
         p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].satd = x265_pixel_satd_8x12_avx;
         p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].satd = x265_pixel_satd_12x32_avx;
         p.chroma[X265_CSP_I422].pu[CHROMA_422_4x32].satd = x265_pixel_satd_4x32_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].satd = x265_pixel_satd_16x32_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].satd = x265_pixel_satd_32x64_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].satd = x265_pixel_satd_16x16_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].satd = x265_pixel_satd_32x32_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].satd = x265_pixel_satd_16x64_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].satd = x265_pixel_satd_16x8_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].satd = x265_pixel_satd_32x16_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].satd = x265_pixel_satd_8x4_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].satd = x265_pixel_satd_8x16_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].satd = x265_pixel_satd_8x8_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].satd = x265_pixel_satd_8x32_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].satd = x265_pixel_satd_4x8_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].satd = x265_pixel_satd_4x16_avx;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].satd = x265_pixel_satd_4x4_avx;
         ALL_LUMA_PU(satd, pixel_satd, avx);
         p.chroma[X265_CSP_I420].pu[CHROMA_420_4x4].satd = x265_pixel_satd_4x4_avx;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].satd = x265_pixel_satd_8x8_avx;
@@ -1383,6 +1715,10 @@
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].satd = x265_pixel_satd_32x8_avx;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x32].satd = x265_pixel_satd_8x32_avx;
         ASSIGN_SA8D(avx);
+        p.chroma[X265_CSP_I420].cu[BLOCK_420_32x32].sa8d = x265_pixel_sa8d_32x32_avx;
+        p.chroma[X265_CSP_I420].cu[BLOCK_420_16x16].sa8d = x265_pixel_sa8d_16x16_avx;
+        p.chroma[X265_CSP_I420].cu[BLOCK_420_8x8].sa8d = x265_pixel_sa8d_8x8_avx;
+        p.chroma[X265_CSP_I420].cu[BLOCK_420_4x4].sa8d = x265_pixel_satd_4x4_avx;
         ASSIGN_SSE_PP(avx);
         p.chroma[X265_CSP_I420].cu[BLOCK_420_8x8].sse_pp = x265_pixel_ssd_8x8_avx;
         ASSIGN_SSE_SS(avx);
@@ -1405,6 +1741,7 @@
         p.chroma[X265_CSP_I420].cu[CHROMA_420_16x16].copy_ss = x265_blockcopy_ss_16x16_avx;
         p.chroma[X265_CSP_I420].cu[CHROMA_420_32x32].copy_ss = x265_blockcopy_ss_32x32_avx;
         p.chroma[X265_CSP_I422].cu[CHROMA_422_16x32].copy_ss = x265_blockcopy_ss_16x32_avx;
+        p.chroma[X265_CSP_I422].cu[CHROMA_422_32x64].copy_ss = x265_blockcopy_ss_32x64_avx;
 
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].copy_pp = x265_blockcopy_pp_32x8_avx;
         p.pu[LUMA_32x8].copy_pp = x265_blockcopy_pp_32x8_avx;
@@ -1447,6 +1784,26 @@
 #if X86_64
     if (cpuMask & X265_CPU_AVX2)
     {
+        p.planecopy_sp = x265_downShift_16_avx2;
+
+        p.cu[BLOCK_32x32].intra_pred[DC_IDX] = x265_intra_pred_dc32_avx2;
+
+        p.cu[BLOCK_16x16].intra_pred[PLANAR_IDX] = x265_intra_pred_planar16_avx2;
+        p.cu[BLOCK_32x32].intra_pred[PLANAR_IDX] = x265_intra_pred_planar32_avx2;
+
+        p.idst4x4 = x265_idst4_avx2;
+        p.dst4x4 = x265_dst4_avx2;
+        p.scale2D_64to32 = x265_scale2D_64to32_avx2;
+        p.saoCuOrgE0 = x265_saoCuOrgE0_avx2;
+        p.saoCuOrgE1 = x265_saoCuOrgE1_avx2;
+        p.saoCuOrgE1_2Rows = x265_saoCuOrgE1_2Rows_avx2;
+        p.saoCuOrgE2[0] = x265_saoCuOrgE2_avx2;
+        p.saoCuOrgE2[1] = x265_saoCuOrgE2_32_avx2;
+        p.saoCuOrgE3[0] = x265_saoCuOrgE3_avx2;
+        p.saoCuOrgE3[1] = x265_saoCuOrgE3_32_avx2;
+        p.saoCuOrgB0 = x265_saoCuOrgB0_avx2;
+        p.sign = x265_calSign_avx2;
+
         p.cu[BLOCK_4x4].psy_cost_ss = x265_psyCost_ss_4x4_avx2;
         p.cu[BLOCK_8x8].psy_cost_ss = x265_psyCost_ss_8x8_avx2;
         p.cu[BLOCK_16x16].psy_cost_ss = x265_psyCost_ss_16x16_avx2;
@@ -1494,31 +1851,50 @@
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].addAvg = x265_addAvg_8x8_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].addAvg = x265_addAvg_8x16_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x32].addAvg = x265_addAvg_8x32_avx2;
-
         p.chroma[X265_CSP_I420].pu[CHROMA_420_12x16].addAvg = x265_addAvg_12x16_avx2;
-
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x4].addAvg = x265_addAvg_16x4_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x8].addAvg = x265_addAvg_16x8_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x12].addAvg = x265_addAvg_16x12_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x16].addAvg = x265_addAvg_16x16_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x32].addAvg = x265_addAvg_16x32_avx2;
-
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].addAvg = x265_addAvg_32x8_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].addAvg = x265_addAvg_32x16_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x24].addAvg = x265_addAvg_32x24_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x32].addAvg = x265_addAvg_32x32_avx2;
 
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].addAvg = x265_addAvg_8x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].addAvg = x265_addAvg_8x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].addAvg = x265_addAvg_8x12_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].addAvg = x265_addAvg_8x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].addAvg = x265_addAvg_8x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].addAvg = x265_addAvg_8x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].addAvg = x265_addAvg_12x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].addAvg = x265_addAvg_16x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].addAvg = x265_addAvg_16x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].addAvg = x265_addAvg_16x24_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].addAvg = x265_addAvg_16x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].addAvg = x265_addAvg_16x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].addAvg = x265_addAvg_24x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].addAvg = x265_addAvg_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].addAvg = x265_addAvg_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].addAvg = x265_addAvg_32x48_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].addAvg = x265_addAvg_32x64_avx2;
+
         p.cu[BLOCK_16x16].add_ps = x265_pixel_add_ps_16x16_avx2;
         p.cu[BLOCK_32x32].add_ps = x265_pixel_add_ps_32x32_avx2;
         p.cu[BLOCK_64x64].add_ps = x265_pixel_add_ps_64x64_avx2;
         p.chroma[X265_CSP_I420].cu[BLOCK_420_16x16].add_ps = x265_pixel_add_ps_16x16_avx2;
         p.chroma[X265_CSP_I420].cu[BLOCK_420_32x32].add_ps = x265_pixel_add_ps_32x32_avx2;
+        p.chroma[X265_CSP_I422].cu[BLOCK_422_16x32].add_ps = x265_pixel_add_ps_16x32_avx2;
+        p.chroma[X265_CSP_I422].cu[BLOCK_422_32x64].add_ps = x265_pixel_add_ps_32x64_avx2;
 
         p.cu[BLOCK_16x16].sub_ps = x265_pixel_sub_ps_16x16_avx2;
         p.cu[BLOCK_32x32].sub_ps = x265_pixel_sub_ps_32x32_avx2;
         p.cu[BLOCK_64x64].sub_ps = x265_pixel_sub_ps_64x64_avx2;
         p.chroma[X265_CSP_I420].cu[BLOCK_420_16x16].sub_ps = x265_pixel_sub_ps_16x16_avx2;
         p.chroma[X265_CSP_I420].cu[BLOCK_420_32x32].sub_ps = x265_pixel_sub_ps_32x32_avx2;
+        p.chroma[X265_CSP_I422].cu[BLOCK_422_16x32].sub_ps = x265_pixel_sub_ps_16x32_avx2;
+        p.chroma[X265_CSP_I422].cu[BLOCK_422_32x64].sub_ps = x265_pixel_sub_ps_32x64_avx2;
 
         p.pu[LUMA_16x4].pixelavg_pp = x265_pixel_avg_16x4_avx2;
         p.pu[LUMA_16x8].pixelavg_pp = x265_pixel_avg_16x8_avx2;
@@ -1543,6 +1919,22 @@
         p.pu[LUMA_8x16].satd  = x265_pixel_satd_8x16_avx2;
         p.pu[LUMA_8x8].satd   = x265_pixel_satd_8x8_avx2;
 
+        p.pu[LUMA_16x4].satd  = x265_pixel_satd_16x4_avx2;
+        p.pu[LUMA_16x12].satd = x265_pixel_satd_16x12_avx2;
+        p.pu[LUMA_16x32].satd = x265_pixel_satd_16x32_avx2;
+        p.pu[LUMA_16x64].satd = x265_pixel_satd_16x64_avx2;
+
+        p.pu[LUMA_32x8].satd   = x265_pixel_satd_32x8_avx2;
+        p.pu[LUMA_32x16].satd   = x265_pixel_satd_32x16_avx2;
+        p.pu[LUMA_32x24].satd   = x265_pixel_satd_32x24_avx2;
+        p.pu[LUMA_32x32].satd   = x265_pixel_satd_32x32_avx2;
+        p.pu[LUMA_32x64].satd   = x265_pixel_satd_32x64_avx2;
+        p.pu[LUMA_48x64].satd   = x265_pixel_satd_48x64_avx2;
+        p.pu[LUMA_64x16].satd   = x265_pixel_satd_64x16_avx2;
+        p.pu[LUMA_64x32].satd   = x265_pixel_satd_64x32_avx2;
+        p.pu[LUMA_64x48].satd   = x265_pixel_satd_64x48_avx2;
+        p.pu[LUMA_64x64].satd   = x265_pixel_satd_64x64_avx2;
+
         p.pu[LUMA_32x8].sad = x265_pixel_sad_32x8_avx2;
         p.pu[LUMA_32x16].sad = x265_pixel_sad_32x16_avx2;
         p.pu[LUMA_32x24].sad = x265_pixel_sad_32x24_avx2;
@@ -1602,8 +1994,37 @@
 
         p.scale1D_128to64 = x265_scale1D_128to64_avx2;
         p.weight_pp = x265_weight_pp_avx2;
+        p.weight_sp = x265_weight_sp_avx2;
 
         // intra_pred functions
+        p.cu[BLOCK_4x4].intra_pred[3] = x265_intra_pred_ang4_3_avx2;
+        p.cu[BLOCK_4x4].intra_pred[4] = x265_intra_pred_ang4_4_avx2;
+        p.cu[BLOCK_4x4].intra_pred[5] = x265_intra_pred_ang4_5_avx2;
+        p.cu[BLOCK_4x4].intra_pred[6] = x265_intra_pred_ang4_6_avx2;
+        p.cu[BLOCK_4x4].intra_pred[7] = x265_intra_pred_ang4_7_avx2;
+        p.cu[BLOCK_4x4].intra_pred[8] = x265_intra_pred_ang4_8_avx2;
+        p.cu[BLOCK_4x4].intra_pred[9] = x265_intra_pred_ang4_9_avx2;
+        p.cu[BLOCK_4x4].intra_pred[11] = x265_intra_pred_ang4_11_avx2;
+        p.cu[BLOCK_4x4].intra_pred[12] = x265_intra_pred_ang4_12_avx2;
+        p.cu[BLOCK_4x4].intra_pred[13] = x265_intra_pred_ang4_13_avx2;
+        p.cu[BLOCK_4x4].intra_pred[14] = x265_intra_pred_ang4_14_avx2;
+        p.cu[BLOCK_4x4].intra_pred[15] = x265_intra_pred_ang4_15_avx2;
+        p.cu[BLOCK_4x4].intra_pred[16] = x265_intra_pred_ang4_16_avx2;
+        p.cu[BLOCK_4x4].intra_pred[17] = x265_intra_pred_ang4_17_avx2;
+        p.cu[BLOCK_4x4].intra_pred[19] = x265_intra_pred_ang4_19_avx2;
+        p.cu[BLOCK_4x4].intra_pred[20] = x265_intra_pred_ang4_20_avx2;
+        p.cu[BLOCK_4x4].intra_pred[21] = x265_intra_pred_ang4_21_avx2;
+        p.cu[BLOCK_4x4].intra_pred[22] = x265_intra_pred_ang4_22_avx2;
+        p.cu[BLOCK_4x4].intra_pred[23] = x265_intra_pred_ang4_23_avx2;
+        p.cu[BLOCK_4x4].intra_pred[24] = x265_intra_pred_ang4_24_avx2;
+        p.cu[BLOCK_4x4].intra_pred[25] = x265_intra_pred_ang4_25_avx2;
+        p.cu[BLOCK_4x4].intra_pred[27] = x265_intra_pred_ang4_27_avx2;
+        p.cu[BLOCK_4x4].intra_pred[28] = x265_intra_pred_ang4_28_avx2;
+        p.cu[BLOCK_4x4].intra_pred[29] = x265_intra_pred_ang4_29_avx2;
+        p.cu[BLOCK_4x4].intra_pred[30] = x265_intra_pred_ang4_30_avx2;
+        p.cu[BLOCK_4x4].intra_pred[31] = x265_intra_pred_ang4_31_avx2;
+        p.cu[BLOCK_4x4].intra_pred[32] = x265_intra_pred_ang4_32_avx2;
+        p.cu[BLOCK_4x4].intra_pred[33] = x265_intra_pred_ang4_33_avx2;
         p.cu[BLOCK_8x8].intra_pred[3] = x265_intra_pred_ang8_3_avx2;
         p.cu[BLOCK_8x8].intra_pred[33] = x265_intra_pred_ang8_33_avx2;
         p.cu[BLOCK_8x8].intra_pred[4] = x265_intra_pred_ang8_4_avx2;
@@ -1622,6 +2043,24 @@
         p.cu[BLOCK_8x8].intra_pred[12] = x265_intra_pred_ang8_12_avx2;
         p.cu[BLOCK_8x8].intra_pred[24] = x265_intra_pred_ang8_24_avx2;
         p.cu[BLOCK_8x8].intra_pred[11] = x265_intra_pred_ang8_11_avx2;
+        p.cu[BLOCK_8x8].intra_pred[13] = x265_intra_pred_ang8_13_avx2;
+        p.cu[BLOCK_8x8].intra_pred[20] = x265_intra_pred_ang8_20_avx2;
+        p.cu[BLOCK_8x8].intra_pred[21] = x265_intra_pred_ang8_21_avx2;
+        p.cu[BLOCK_8x8].intra_pred[22] = x265_intra_pred_ang8_22_avx2;
+        p.cu[BLOCK_8x8].intra_pred[23] = x265_intra_pred_ang8_23_avx2;
+        p.cu[BLOCK_8x8].intra_pred[14] = x265_intra_pred_ang8_14_avx2;
+        p.cu[BLOCK_8x8].intra_pred[15] = x265_intra_pred_ang8_15_avx2;
+        p.cu[BLOCK_8x8].intra_pred[16] = x265_intra_pred_ang8_16_avx2;
+        p.cu[BLOCK_16x16].intra_pred[3] = x265_intra_pred_ang16_3_avx2;
+        p.cu[BLOCK_16x16].intra_pred[4] = x265_intra_pred_ang16_4_avx2;
+        p.cu[BLOCK_16x16].intra_pred[5] = x265_intra_pred_ang16_5_avx2;
+        p.cu[BLOCK_16x16].intra_pred[6] = x265_intra_pred_ang16_6_avx2;
+        p.cu[BLOCK_16x16].intra_pred[7] = x265_intra_pred_ang16_7_avx2;
+        p.cu[BLOCK_16x16].intra_pred[8] = x265_intra_pred_ang16_8_avx2;
+        p.cu[BLOCK_16x16].intra_pred[9] = x265_intra_pred_ang16_9_avx2;
+        p.cu[BLOCK_16x16].intra_pred[12] = x265_intra_pred_ang16_12_avx2;
+        p.cu[BLOCK_16x16].intra_pred[11] = x265_intra_pred_ang16_11_avx2;
+        p.cu[BLOCK_16x16].intra_pred[13] = x265_intra_pred_ang16_13_avx2;
         p.cu[BLOCK_16x16].intra_pred[25] = x265_intra_pred_ang16_25_avx2;
         p.cu[BLOCK_16x16].intra_pred[28] = x265_intra_pred_ang16_28_avx2;
         p.cu[BLOCK_16x16].intra_pred[27] = x265_intra_pred_ang16_27_avx2;
@@ -1642,6 +2081,16 @@
         p.cu[BLOCK_32x32].intra_pred[30] = x265_intra_pred_ang32_30_avx2;
         p.cu[BLOCK_32x32].intra_pred[31] = x265_intra_pred_ang32_31_avx2;
         p.cu[BLOCK_32x32].intra_pred[32] = x265_intra_pred_ang32_32_avx2;
+        p.cu[BLOCK_32x32].intra_pred[33] = x265_intra_pred_ang32_33_avx2;
+        p.cu[BLOCK_32x32].intra_pred[25] = x265_intra_pred_ang32_25_avx2;
+        p.cu[BLOCK_32x32].intra_pred[24] = x265_intra_pred_ang32_24_avx2;
+        p.cu[BLOCK_32x32].intra_pred[23] = x265_intra_pred_ang32_23_avx2;
+        p.cu[BLOCK_32x32].intra_pred[22] = x265_intra_pred_ang32_22_avx2;
+        p.cu[BLOCK_32x32].intra_pred[21] = x265_intra_pred_ang32_21_avx2;
+        p.cu[BLOCK_32x32].intra_pred[18] = x265_intra_pred_ang32_18_avx2;
+
+        // all_angs primitives
+        p.cu[BLOCK_4x4].intra_pred_allangs = x265_all_angs_pred_4x4_avx2;
 
         // copy_sp primitives
         p.cu[BLOCK_16x16].copy_sp = x265_blockcopy_sp_16x16_avx2;
@@ -1725,6 +2174,8 @@
         p.pu[LUMA_64x48].luma_hps = x265_interp_8tap_horiz_ps_64x48_avx2;
         p.pu[LUMA_64x32].luma_hps = x265_interp_8tap_horiz_ps_64x32_avx2;
         p.pu[LUMA_64x16].luma_hps = x265_interp_8tap_horiz_ps_64x16_avx2;
+        p.pu[LUMA_12x16].luma_hps = x265_interp_8tap_horiz_ps_12x16_avx2;
+        p.pu[LUMA_24x32].luma_hps = x265_interp_8tap_horiz_ps_24x32_avx2;
 
         p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].filter_hpp = x265_interp_4tap_horiz_pp_8x8_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_4x4].filter_hpp = x265_interp_4tap_horiz_pp_4x4_avx2;
@@ -1744,6 +2195,7 @@
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x32].filter_hpp = x265_interp_4tap_horiz_pp_16x32_avx2;
 
         p.chroma[X265_CSP_I420].pu[CHROMA_420_6x8].filter_hpp = x265_interp_4tap_horiz_pp_6x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_6x16].filter_hpp = x265_interp_4tap_horiz_pp_6x16_avx2;
 
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].filter_hpp = x265_interp_4tap_horiz_pp_32x16_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x24].filter_hpp = x265_interp_4tap_horiz_pp_32x24_avx2;
@@ -1777,6 +2229,7 @@
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x8].filter_hps = x265_interp_4tap_horiz_ps_16x8_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_16x4].filter_hps = x265_interp_4tap_horiz_ps_16x4_avx2;
 
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_24x32].filter_hps = x265_interp_4tap_horiz_ps_24x32_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].filter_hps = x265_interp_4tap_horiz_ps_32x16_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x24].filter_hps = x265_interp_4tap_horiz_ps_32x24_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].filter_hps = x265_interp_4tap_horiz_ps_32x8_avx2;
@@ -1887,8 +2340,353 @@
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].filter_vss = x265_interp_4tap_vert_ss_32x16_avx2;
         p.chroma[X265_CSP_I420].pu[CHROMA_420_32x24].filter_vss = x265_interp_4tap_vert_ss_32x24_avx2;
 
-        if ((cpuMask & X265_CPU_BMI1) && (cpuMask & X265_CPU_BMI2))
-            p.findPosLast = x265_findPosLast_x64;
+        //i422 for chroma_vss
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].filter_vss = x265_interp_4tap_vert_ss_4x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].filter_vss = x265_interp_4tap_vert_ss_8x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].filter_vss = x265_interp_4tap_vert_ss_16x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].filter_vss = x265_interp_4tap_vert_ss_4x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].filter_vss = x265_interp_4tap_vert_ss_2x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].filter_vss = x265_interp_4tap_vert_ss_8x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].filter_vss = x265_interp_4tap_vert_ss_4x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].filter_vss = x265_interp_4tap_vert_ss_16x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].filter_vss = x265_interp_4tap_vert_ss_8x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].filter_vss = x265_interp_4tap_vert_ss_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].filter_vss = x265_interp_4tap_vert_ss_8x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].filter_vss = x265_interp_4tap_vert_ss_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].filter_vss = x265_interp_4tap_vert_ss_32x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].filter_vss = x265_interp_4tap_vert_ss_16x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].filter_vss = x265_interp_4tap_vert_ss_24x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].filter_vss = x265_interp_4tap_vert_ss_8x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].filter_vss = x265_interp_4tap_vert_ss_32x48_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].filter_vss = x265_interp_4tap_vert_ss_8x12_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_6x16].filter_vss = x265_interp_4tap_vert_ss_6x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].filter_vss = x265_interp_4tap_vert_ss_2x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].filter_vss = x265_interp_4tap_vert_ss_16x24_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].filter_vss = x265_interp_4tap_vert_ss_12x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x32].filter_vss = x265_interp_4tap_vert_ss_4x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x4].filter_vss = x265_interp_4tap_vert_ss_2x4_avx2;
+
+        //i444 for chroma_vss
+        p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_vss = x265_interp_4tap_vert_ss_4x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x8].filter_vss = x265_interp_4tap_vert_ss_8x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x16].filter_vss = x265_interp_4tap_vert_ss_16x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x32].filter_vss = x265_interp_4tap_vert_ss_32x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x64].filter_vss = x265_interp_4tap_vert_ss_64x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x4].filter_vss = x265_interp_4tap_vert_ss_8x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x8].filter_vss = x265_interp_4tap_vert_ss_4x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x8].filter_vss = x265_interp_4tap_vert_ss_16x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x16].filter_vss = x265_interp_4tap_vert_ss_8x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x16].filter_vss = x265_interp_4tap_vert_ss_32x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x32].filter_vss = x265_interp_4tap_vert_ss_16x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x12].filter_vss = x265_interp_4tap_vert_ss_16x12_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_12x16].filter_vss = x265_interp_4tap_vert_ss_12x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x4].filter_vss = x265_interp_4tap_vert_ss_16x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x16].filter_vss = x265_interp_4tap_vert_ss_4x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x24].filter_vss = x265_interp_4tap_vert_ss_32x24_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_24x32].filter_vss = x265_interp_4tap_vert_ss_24x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x8].filter_vss = x265_interp_4tap_vert_ss_32x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x32].filter_vss = x265_interp_4tap_vert_ss_8x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x32].filter_vss = x265_interp_4tap_vert_ss_64x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x64].filter_vss = x265_interp_4tap_vert_ss_32x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x48].filter_vss = x265_interp_4tap_vert_ss_64x48_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_48x64].filter_vss = x265_interp_4tap_vert_ss_48x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x16].filter_vss = x265_interp_4tap_vert_ss_64x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x64].filter_vss = x265_interp_4tap_vert_ss_16x64_avx2;
+
+        p.pu[LUMA_16x16].luma_hvpp = x265_interp_8tap_hv_pp_16x16_avx2;
+
+        p.pu[LUMA_32x8].convert_p2s = x265_filterPixelToShort_32x8_avx2;
+        p.pu[LUMA_32x16].convert_p2s = x265_filterPixelToShort_32x16_avx2;
+        p.pu[LUMA_32x24].convert_p2s = x265_filterPixelToShort_32x24_avx2;
+        p.pu[LUMA_32x32].convert_p2s = x265_filterPixelToShort_32x32_avx2;
+        p.pu[LUMA_32x64].convert_p2s = x265_filterPixelToShort_32x64_avx2;
+        p.pu[LUMA_64x16].convert_p2s = x265_filterPixelToShort_64x16_avx2;
+        p.pu[LUMA_64x32].convert_p2s = x265_filterPixelToShort_64x32_avx2;
+        p.pu[LUMA_64x48].convert_p2s = x265_filterPixelToShort_64x48_avx2;
+        p.pu[LUMA_64x64].convert_p2s = x265_filterPixelToShort_64x64_avx2;
+        p.pu[LUMA_48x64].convert_p2s = x265_filterPixelToShort_48x64_avx2;
+        p.pu[LUMA_24x32].convert_p2s = x265_filterPixelToShort_24x32_avx2;
+
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_24x32].p2s = x265_filterPixelToShort_24x32_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].p2s = x265_filterPixelToShort_32x8_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x16].p2s = x265_filterPixelToShort_32x16_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x24].p2s = x265_filterPixelToShort_32x24_avx2;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_32x32].p2s = x265_filterPixelToShort_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].p2s = x265_filterPixelToShort_24x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].p2s = x265_filterPixelToShort_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].p2s = x265_filterPixelToShort_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].p2s = x265_filterPixelToShort_32x48_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].p2s = x265_filterPixelToShort_32x64_avx2;
+
+        //i422 for chroma_hpp
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].filter_hpp = x265_interp_4tap_horiz_pp_12x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].filter_hpp = x265_interp_4tap_horiz_pp_24x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].filter_hpp = x265_interp_4tap_horiz_pp_2x16_avx2;
+
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].filter_hpp = x265_interp_4tap_horiz_pp_2x16_avx2;
+
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].filter_hpp = x265_interp_4tap_horiz_pp_4x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].filter_hpp = x265_interp_4tap_horiz_pp_4x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].filter_hpp = x265_interp_4tap_horiz_pp_4x16_avx2;
+        
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].filter_hpp = x265_interp_4tap_horiz_pp_8x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].filter_hpp = x265_interp_4tap_horiz_pp_8x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].filter_hpp = x265_interp_4tap_horiz_pp_8x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].filter_hpp = x265_interp_4tap_horiz_pp_8x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].filter_hpp = x265_interp_4tap_horiz_pp_8x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].filter_hpp = x265_interp_4tap_horiz_pp_8x12_avx2;
+        
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].filter_hpp = x265_interp_4tap_horiz_pp_16x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].filter_hpp = x265_interp_4tap_horiz_pp_16x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].filter_hpp = x265_interp_4tap_horiz_pp_16x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].filter_hpp = x265_interp_4tap_horiz_pp_16x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].filter_hpp = x265_interp_4tap_horiz_pp_16x24_avx2;
+        
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].filter_hpp = x265_interp_4tap_horiz_pp_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].filter_hpp = x265_interp_4tap_horiz_pp_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].filter_hpp = x265_interp_4tap_horiz_pp_32x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].filter_hpp = x265_interp_4tap_horiz_pp_32x48_avx2;
+        
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].filter_hpp = x265_interp_4tap_horiz_pp_2x8_avx2;
+
+        //i444 filters hpp
+
+        p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_hpp = x265_interp_4tap_horiz_pp_4x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x8].filter_hpp = x265_interp_4tap_horiz_pp_8x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x16].filter_hpp = x265_interp_4tap_horiz_pp_16x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x32].filter_hpp = x265_interp_4tap_horiz_pp_32x32_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_4x8].filter_hpp = x265_interp_4tap_horiz_pp_4x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x16].filter_hpp = x265_interp_4tap_horiz_pp_4x16_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_8x4].filter_hpp = x265_interp_4tap_horiz_pp_8x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x16].filter_hpp = x265_interp_4tap_horiz_pp_8x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x32].filter_hpp = x265_interp_4tap_horiz_pp_8x32_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_16x8].filter_hpp = x265_interp_4tap_horiz_pp_16x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x32].filter_hpp = x265_interp_4tap_horiz_pp_16x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x12].filter_hpp = x265_interp_4tap_horiz_pp_16x12_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x4].filter_hpp = x265_interp_4tap_horiz_pp_16x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x64].filter_hpp = x265_interp_4tap_horiz_pp_16x64_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_12x16].filter_hpp = x265_interp_4tap_horiz_pp_12x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_24x32].filter_hpp = x265_interp_4tap_horiz_pp_24x32_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_32x16].filter_hpp = x265_interp_4tap_horiz_pp_32x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x64].filter_hpp = x265_interp_4tap_horiz_pp_32x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x24].filter_hpp = x265_interp_4tap_horiz_pp_32x24_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x8].filter_hpp = x265_interp_4tap_horiz_pp_32x8_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_64x64].filter_hpp = x265_interp_4tap_horiz_pp_64x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x32].filter_hpp = x265_interp_4tap_horiz_pp_64x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x48].filter_hpp = x265_interp_4tap_horiz_pp_64x48_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x16].filter_hpp = x265_interp_4tap_horiz_pp_64x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_48x64].filter_hpp = x265_interp_4tap_horiz_pp_48x64_avx2;
+
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].filter_hps = x265_interp_4tap_horiz_ps_4x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].filter_hps = x265_interp_4tap_horiz_ps_4x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].filter_hps = x265_interp_4tap_horiz_ps_4x16_avx2;
+
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].filter_hps = x265_interp_4tap_horiz_ps_8x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].filter_hps = x265_interp_4tap_horiz_ps_8x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].filter_hps = x265_interp_4tap_horiz_ps_8x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].filter_hps = x265_interp_4tap_horiz_ps_8x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].filter_hps = x265_interp_4tap_horiz_ps_8x64_avx2; //adding macro call
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].filter_hps = x265_interp_4tap_horiz_ps_8x12_avx2; //adding macro call
+
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].filter_hps = x265_interp_4tap_horiz_ps_16x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].filter_hps = x265_interp_4tap_horiz_ps_16x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].filter_hps = x265_interp_4tap_horiz_ps_16x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].filter_hps = x265_interp_4tap_horiz_ps_16x64_avx2;//adding macro call
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].filter_hps = x265_interp_4tap_horiz_ps_16x24_avx2;//adding macro call
+
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].filter_hps = x265_interp_4tap_horiz_ps_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].filter_hps = x265_interp_4tap_horiz_ps_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].filter_hps = x265_interp_4tap_horiz_ps_32x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].filter_hps = x265_interp_4tap_horiz_ps_32x48_avx2;
+
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].filter_hps = x265_interp_4tap_horiz_ps_2x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].filter_hps = x265_interp_4tap_horiz_ps_24x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].filter_hps = x265_interp_4tap_horiz_ps_2x16_avx2;
+
+        //i444 chroma_hps
+        p.chroma[X265_CSP_I444].pu[LUMA_64x32].filter_hps = x265_interp_4tap_horiz_ps_64x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x48].filter_hps = x265_interp_4tap_horiz_ps_64x48_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x16].filter_hps = x265_interp_4tap_horiz_ps_64x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x64].filter_hps = x265_interp_4tap_horiz_ps_64x64_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_hps = x265_interp_4tap_horiz_ps_4x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x8].filter_hps = x265_interp_4tap_horiz_ps_8x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x16].filter_hps = x265_interp_4tap_horiz_ps_16x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x32].filter_hps = x265_interp_4tap_horiz_ps_32x32_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_4x8].filter_hps = x265_interp_4tap_horiz_ps_4x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x16].filter_hps = x265_interp_4tap_horiz_ps_4x16_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_8x4].filter_hps = x265_interp_4tap_horiz_ps_8x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x16].filter_hps = x265_interp_4tap_horiz_ps_8x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x32].filter_hps = x265_interp_4tap_horiz_ps_8x32_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_16x8].filter_hps = x265_interp_4tap_horiz_ps_16x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x32].filter_hps = x265_interp_4tap_horiz_ps_16x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x12].filter_hps = x265_interp_4tap_horiz_ps_16x12_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x4].filter_hps = x265_interp_4tap_horiz_ps_16x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x64].filter_hps = x265_interp_4tap_horiz_ps_16x64_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_24x32].filter_hps = x265_interp_4tap_horiz_ps_24x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_48x64].filter_hps = x265_interp_4tap_horiz_ps_48x64_avx2;
+
+        p.chroma[X265_CSP_I444].pu[LUMA_32x16].filter_hps = x265_interp_4tap_horiz_ps_32x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x64].filter_hps = x265_interp_4tap_horiz_ps_32x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x24].filter_hps = x265_interp_4tap_horiz_ps_32x24_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x8].filter_hps = x265_interp_4tap_horiz_ps_32x8_avx2;
+
+        //i422 for chroma_vsp
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].filter_vsp = x265_interp_4tap_vert_sp_4x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].filter_vsp = x265_interp_4tap_vert_sp_8x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].filter_vsp = x265_interp_4tap_vert_sp_16x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].filter_vsp = x265_interp_4tap_vert_sp_4x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].filter_vsp = x265_interp_4tap_vert_sp_2x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].filter_vsp = x265_interp_4tap_vert_sp_8x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].filter_vsp = x265_interp_4tap_vert_sp_4x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].filter_vsp = x265_interp_4tap_vert_sp_16x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].filter_vsp = x265_interp_4tap_vert_sp_8x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].filter_vsp = x265_interp_4tap_vert_sp_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].filter_vsp = x265_interp_4tap_vert_sp_8x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].filter_vsp = x265_interp_4tap_vert_sp_16x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].filter_vsp = x265_interp_4tap_vert_sp_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].filter_vsp = x265_interp_4tap_vert_sp_32x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].filter_vsp = x265_interp_4tap_vert_sp_16x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_24x64].filter_vsp = x265_interp_4tap_vert_sp_24x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].filter_vsp = x265_interp_4tap_vert_sp_8x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].filter_vsp = x265_interp_4tap_vert_sp_32x48_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].filter_vsp = x265_interp_4tap_vert_sp_8x12_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_6x16].filter_vsp = x265_interp_4tap_vert_sp_6x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x16].filter_vsp = x265_interp_4tap_vert_sp_2x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].filter_vsp = x265_interp_4tap_vert_sp_16x24_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].filter_vsp = x265_interp_4tap_vert_sp_12x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x32].filter_vsp = x265_interp_4tap_vert_sp_4x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x4].filter_vsp = x265_interp_4tap_vert_sp_2x4_avx2;
+
+        //i444 for chroma_vsp
+        p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_vsp = x265_interp_4tap_vert_sp_4x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x8].filter_vsp = x265_interp_4tap_vert_sp_8x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x16].filter_vsp = x265_interp_4tap_vert_sp_16x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x32].filter_vsp = x265_interp_4tap_vert_sp_32x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x64].filter_vsp = x265_interp_4tap_vert_sp_64x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x4].filter_vsp = x265_interp_4tap_vert_sp_8x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x8].filter_vsp = x265_interp_4tap_vert_sp_4x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x8].filter_vsp = x265_interp_4tap_vert_sp_16x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x16].filter_vsp = x265_interp_4tap_vert_sp_8x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x16].filter_vsp = x265_interp_4tap_vert_sp_32x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x32].filter_vsp = x265_interp_4tap_vert_sp_16x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x12].filter_vsp = x265_interp_4tap_vert_sp_16x12_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_12x16].filter_vsp = x265_interp_4tap_vert_sp_12x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x4].filter_vsp = x265_interp_4tap_vert_sp_16x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x16].filter_vsp = x265_interp_4tap_vert_sp_4x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x24].filter_vsp = x265_interp_4tap_vert_sp_32x24_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_24x32].filter_vsp = x265_interp_4tap_vert_sp_24x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x8].filter_vsp = x265_interp_4tap_vert_sp_32x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x32].filter_vsp = x265_interp_4tap_vert_sp_8x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x32].filter_vsp = x265_interp_4tap_vert_sp_64x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x64].filter_vsp = x265_interp_4tap_vert_sp_32x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x48].filter_vsp = x265_interp_4tap_vert_sp_64x48_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_48x64].filter_vsp = x265_interp_4tap_vert_sp_48x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_64x16].filter_vsp = x265_interp_4tap_vert_sp_64x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x64].filter_vsp = x265_interp_4tap_vert_sp_16x64_avx2;
+
+        //i422 for chroma_vps
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].filter_vps = x265_interp_4tap_vert_ps_4x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].filter_vps = x265_interp_4tap_vert_ps_8x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].filter_vps = x265_interp_4tap_vert_ps_16x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].filter_vps = x265_interp_4tap_vert_ps_4x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].filter_vps = x265_interp_4tap_vert_ps_2x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].filter_vps = x265_interp_4tap_vert_ps_8x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].filter_vps = x265_interp_4tap_vert_ps_4x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].filter_vps = x265_interp_4tap_vert_ps_16x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].filter_vps = x265_interp_4tap_vert_ps_8x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].filter_vps = x265_interp_4tap_vert_ps_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].filter_vps = x265_interp_4tap_vert_ps_8x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].filter_vps = x265_interp_4tap_vert_ps_16x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].filter_vps = x265_interp_4tap_vert_ps_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].filter_vps = x265_interp_4tap_vert_ps_16x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].filter_vps = x265_interp_4tap_vert_ps_8x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].filter_vps = x265_interp_4tap_vert_ps_32x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].filter_vps = x265_interp_4tap_vert_ps_32x48_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].filter_vps = x265_interp_4tap_vert_ps_12x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].filter_vps = x265_interp_4tap_vert_ps_8x12_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x4].filter_vps = x265_interp_4tap_vert_ps_2x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].filter_vps = x265_interp_4tap_vert_ps_16x24_avx2;
+
+        //i444 for chroma_vps
+        p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_vps = x265_interp_4tap_vert_ps_4x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x8].filter_vps = x265_interp_4tap_vert_ps_8x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x16].filter_vps = x265_interp_4tap_vert_ps_16x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x32].filter_vps = x265_interp_4tap_vert_ps_32x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x4].filter_vps = x265_interp_4tap_vert_ps_8x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x8].filter_vps = x265_interp_4tap_vert_ps_4x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x8].filter_vps = x265_interp_4tap_vert_ps_16x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x16].filter_vps = x265_interp_4tap_vert_ps_8x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x16].filter_vps = x265_interp_4tap_vert_ps_32x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x32].filter_vps = x265_interp_4tap_vert_ps_16x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x12].filter_vps = x265_interp_4tap_vert_ps_16x12_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_12x16].filter_vps = x265_interp_4tap_vert_ps_12x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x4].filter_vps = x265_interp_4tap_vert_ps_16x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x16].filter_vps = x265_interp_4tap_vert_ps_4x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x24].filter_vps = x265_interp_4tap_vert_ps_32x24_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_24x32].filter_vps = x265_interp_4tap_vert_ps_24x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x8].filter_vps = x265_interp_4tap_vert_ps_32x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x32].filter_vps = x265_interp_4tap_vert_ps_8x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x64].filter_vps = x265_interp_4tap_vert_ps_16x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x64].filter_vps = x265_interp_4tap_vert_ps_32x64_avx2;
+
+        //i422 for chroma_vpp
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].filter_vpp = x265_interp_4tap_vert_pp_4x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x16].filter_vpp = x265_interp_4tap_vert_pp_8x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x32].filter_vpp = x265_interp_4tap_vert_pp_16x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].filter_vpp = x265_interp_4tap_vert_pp_4x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].filter_vpp = x265_interp_4tap_vert_pp_2x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].filter_vpp = x265_interp_4tap_vert_pp_8x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].filter_vpp = x265_interp_4tap_vert_pp_4x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x16].filter_vpp = x265_interp_4tap_vert_pp_16x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x32].filter_vpp = x265_interp_4tap_vert_pp_8x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x32].filter_vpp = x265_interp_4tap_vert_pp_32x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].filter_vpp = x265_interp_4tap_vert_pp_8x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x8].filter_vpp = x265_interp_4tap_vert_pp_16x8_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x16].filter_vpp = x265_interp_4tap_vert_pp_32x16_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x64].filter_vpp = x265_interp_4tap_vert_pp_16x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x64].filter_vpp = x265_interp_4tap_vert_pp_8x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x64].filter_vpp = x265_interp_4tap_vert_pp_32x64_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_32x48].filter_vpp = x265_interp_4tap_vert_pp_32x48_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].filter_vpp = x265_interp_4tap_vert_pp_12x32_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x12].filter_vpp = x265_interp_4tap_vert_pp_8x12_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x4].filter_vpp = x265_interp_4tap_vert_pp_2x4_avx2;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_16x24].filter_vpp = x265_interp_4tap_vert_pp_16x24_avx2;
+
+        //i444 for chroma_vpp
+        p.chroma[X265_CSP_I444].pu[LUMA_4x4].filter_vpp = x265_interp_4tap_vert_pp_4x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x8].filter_vpp = x265_interp_4tap_vert_pp_8x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x16].filter_vpp = x265_interp_4tap_vert_pp_16x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x32].filter_vpp = x265_interp_4tap_vert_pp_32x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x4].filter_vpp = x265_interp_4tap_vert_pp_8x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x8].filter_vpp = x265_interp_4tap_vert_pp_4x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x8].filter_vpp = x265_interp_4tap_vert_pp_16x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x16].filter_vpp = x265_interp_4tap_vert_pp_8x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x16].filter_vpp = x265_interp_4tap_vert_pp_32x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x32].filter_vpp = x265_interp_4tap_vert_pp_16x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x12].filter_vpp = x265_interp_4tap_vert_pp_16x12_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_12x16].filter_vpp = x265_interp_4tap_vert_pp_12x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x4].filter_vpp = x265_interp_4tap_vert_pp_16x4_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_4x16].filter_vpp = x265_interp_4tap_vert_pp_4x16_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x24].filter_vpp = x265_interp_4tap_vert_pp_32x24_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_24x32].filter_vpp = x265_interp_4tap_vert_pp_24x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x8].filter_vpp = x265_interp_4tap_vert_pp_32x8_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_8x32].filter_vpp = x265_interp_4tap_vert_pp_8x32_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_16x64].filter_vpp = x265_interp_4tap_vert_pp_16x64_avx2;
+        p.chroma[X265_CSP_I444].pu[LUMA_32x64].filter_vpp = x265_interp_4tap_vert_pp_32x64_avx2;
+
+        if (cpuMask & X265_CPU_BMI2)
+            p.scanPosLast = x265_scanPosLast_avx2_bmi2;
     }
 #endif
 }

x265_1.6.tar.gz/source/common/x86/const-a.asm -> x265_1.7.tar.gz/source/common/x86/const-a.asm Changed

@@ -29,81 +29,100 @@
 
 SECTION_RODATA 32
 
-const pb_1,        times 32 db 1
+;; 8-bit constants
 
-const hsub_mul,    times 16 db 1, -1
-const pw_1,        times 16 dw 1
-const pw_16,       times 16 dw 16
-const pw_32,       times 16 dw 32
-const pw_128,      times 16 dw 128
-const pw_256,      times 16 dw 256
-const pw_257,      times 16 dw 257
-const pw_512,      times 16 dw 512
-const pw_1023,     times 8  dw 1023
-ALIGN 32
-const pw_1024,     times 16 dw 1024
-const pw_4096,     times 16 dw 4096
-const pw_00ff,     times 16 dw 0x00ff
-ALIGN 32
-const pw_pixel_max,times 16 dw ((1 << BIT_DEPTH)-1)
-const deinterleave_shufd, dd 0,4,1,5,2,6,3,7
-const pb_unpackbd1, times 2 db 0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3
-const pb_unpackbd2, times 2 db 4,4,4,4,5,5,5,5,6,6,6,6,7,7,7,7
-const pb_unpackwq1, db 0,1,0,1,0,1,0,1,2,3,2,3,2,3,2,3
-const pb_unpackwq2, db 4,5,4,5,4,5,4,5,6,7,6,7,6,7,6,7
-const pw_swap,      times 2 db 6,7,4,5,2,3,0,1
+const pb_0,                 times 16 db 0
+const pb_1,                 times 32 db 1
+const pb_2,                 times 32 db 2
+const pb_3,                 times 16 db 3
+const pb_4,                 times 32 db 4
+const pb_8,                 times 32 db 8
+const pb_15,                times 32 db 15
+const pb_16,                times 32 db 16
+const pb_32,                times 32 db 32
+const pb_64,                times 32 db 64
+const pb_128,               times 16 db 128
+const pb_a1,                times 16 db 0xa1
 
-const pb_2,        times 32 db 2
-const pb_4,        times 32 db 4
-const pb_16,       times 32 db 16
-const pb_64,       times 32 db 64
-const pb_01,       times  8 db 0,1
-const pb_0,        times 16 db 0
-const pb_a1,       times 16 db 0xa1
-const pb_3,        times 16 db 3
-const pb_8,        times 32 db 8
-const pb_32,       times 32 db 32
-const pb_128,      times 16 db 128
-const pb_shuf8x8c, db 0,0,0,0,2,2,2,2,4,4,4,4,6,6,6,6
+const pb_01,                times  8 db   0,   1
+const hsub_mul,             times 16 db   1,  -1
+const pw_swap,              times  2 db   6,   7,   4,   5,   2,   3,   0,   1
+const pb_unpackbd1,         times  2 db   0,   0,   0,   0,   1,   1,   1,   1,   2,   2,   2,   2,   3,   3,   3,   3
+const pb_unpackbd2,         times  2 db   4,   4,   4,   4,   5,   5,   5,   5,   6,   6,   6,   6,   7,   7,   7,   7
+const pb_unpackwq1,         times  1 db   0,   1,   0,   1,   0,   1,   0,   1,   2,   3,   2,   3,   2,   3,   2,   3
+const pb_unpackwq2,         times  1 db   4,   5,   4,   5,   4,   5,   4,   5,   6,   7,   6,   7,   6,   7,   6,   7
+const pb_shuf8x8c,          times  1 db   0,   0,   0,   0,   2,   2,   2,   2,   4,   4,   4,   4,   6,   6,   6,   6
+const pb_movemask,          times 16 db 0x00
+                            times 16 db 0xFF
+const pb_0000000000000F0F,  times  2 db 0xff, 0x00
+                            times 12 db 0x00
+const pb_000000000000000F,           db 0xff
+                            times 15 db 0x00
 
-const pw_0_15,     times 2 dw 0, 1, 2, 3, 4, 5, 6, 7
-const pw_2,        times 8 dw 2
-const pw_m2,       times 8 dw -2
-const pw_4,        times 8 dw 4
-const pw_8,        times 8 dw 8
-const pw_64,       times 8 dw 64
-const pw_256,      times 8 dw 256
-const pw_32_0,     times 4 dw 32,
-                   times 4 dw 0
-const pw_2000,     times 16 dw 0x2000
-const pw_8000,     times 8 dw 0x8000
-const pw_3fff,     times 8 dw 0x3fff
-const pw_ppppmmmm, dw 1,1,1,1,-1,-1,-1,-1
-const pw_ppmmppmm, dw 1,1,-1,-1,1,1,-1,-1
-const pw_pmpmpmpm, dw 1,-1,1,-1,1,-1,1,-1
-const pw_pmmpzzzz, dw 1,-1,-1,1,0,0,0,0
-const pd_1,        times 8 dd 1
-const pd_2,        times 8 dd 2
-const pd_4,        times 4 dd 4
-const pd_8,        times 4 dd 8
-const pd_16,       times 4 dd 16
-const pd_32,       times 4 dd 32
-const pd_64,       times 4 dd 64
-const pd_128,      times 4 dd 128
-const pd_256,      times 4 dd 256
-const pd_512,      times 4 dd 512
-const pd_1024,     times 4 dd 1024
-const pd_2048,     times 4 dd 2048
-const pd_ffff,     times 4 dd 0xffff
-const pd_32767,    times 4 dd 32767
-const pd_n32768,   times 4 dd 0xffff8000
-const pw_ff00,     times 8 dw 0xff00
+;; 16-bit constants
 
-const multi_2Row,  dw 1, 2, 3, 4, 1, 2, 3, 4
-const multiL,      dw 1, 2, 3, 4, 5, 6, 7, 8
-const multiH,      dw 9, 10, 11, 12, 13, 14, 15, 16
-const multiH2,     dw 17, 18, 19, 20, 21, 22, 23, 24
-const multiH3,     dw 25, 26, 27, 28, 29, 30, 31, 32
+const pw_1,                 times 16 dw 1
+const pw_2,                 times  8 dw 2
+const pw_m2,                times  8 dw -2
+const pw_4,                 times  8 dw 4
+const pw_8,                 times  8 dw 8
+const pw_16,                times 16 dw 16
+const pw_15,                times 16 dw 15
+const pw_31,                times 16 dw 31
+const pw_32,                times 16 dw 32
+const pw_64,                times  8 dw 64
+const pw_128,               times 16 dw 128
+const pw_256,               times 16 dw 256
+const pw_257,               times 16 dw 257
+const pw_512,               times 16 dw 512
+const pw_1023,              times  8 dw 1023
+const pw_1024,              times 16 dw 1024
+const pw_4096,              times 16 dw 4096
+const pw_00ff,              times 16 dw 0x00ff
+const pw_ff00,              times  8 dw 0xff00
+const pw_2000,              times 16 dw 0x2000
+const pw_8000,              times  8 dw 0x8000
+const pw_3fff,              times  8 dw 0x3fff
+const pw_32_0,              times  4 dw 32,
+                            times  4 dw 0
+const pw_pixel_max,         times 16 dw ((1 << BIT_DEPTH)-1)
+
+const pw_0_15,              times  2 dw   0,   1,   2,   3,   4,   5,   6,   7
+const pw_ppppmmmm,          times  1 dw   1,   1,   1,   1,  -1,  -1,  -1,  -1
+const pw_ppmmppmm,          times  1 dw   1,   1,  -1,  -1,   1,   1,  -1,  -1
+const pw_pmpmpmpm,          times  1 dw   1,  -1,   1,  -1,   1,  -1,   1,  -1
+const pw_pmmpzzzz,          times  1 dw   1,  -1,  -1,   1,   0,   0,   0,   0
+const multi_2Row,           times  1 dw   1,   2,   3,   4,   1,   2,   3,   4
+const multiH,               times  1 dw   9,  10,  11,  12,  13,  14,  15,  16
+const multiH3,              times  1 dw  25,  26,  27,  28,  29,  30,  31,  32
+const multiL,               times  1 dw   1,   2,   3,   4,   5,   6,   7,   8,   9,  10,  11,  12,  13,  14,  15,  16
+const multiH2,              times  1 dw  17,  18,  19,  20,  21,  22,  23,  24,  25,  26,  27,  28,  29,  30,  31,  32
+const pw_planar16_mul,      times  1 dw  15,  14,  13,  12,  11,  10,   9,   8,   7,   6,   5,   4,   3,   2,   1,   0
+const pw_planar32_mul,      times  1 dw  31,  30,  29,  28,  27,  26,  25,  24,  23,  22,  21,  20,  19,  18,  17,  16
+const pw_FFFFFFFFFFFFFFF0,           dw 0x00
+                            times 7  dw 0xff
+
+
+;; 32-bit constants
+
+const pd_1,                 times  8 dd 1
+const pd_2,                 times  8 dd 2
+const pd_4,                 times  4 dd 4
+const pd_8,                 times  4 dd 8
+const pd_16,                times  4 dd 16
+const pd_32,                times  4 dd 32
+const pd_64,                times  4 dd 64
+const pd_128,               times  4 dd 128
+const pd_256,               times  4 dd 256
+const pd_512,               times  4 dd 512
+const pd_1024,              times  4 dd 1024
+const pd_2048,              times  4 dd 2048
+const pd_ffff,              times  4 dd 0xffff
+const pd_32767,             times  4 dd 32767
+const pd_n32768,            times  4 dd 0xffff8000
+
+const trans8_shuf,          times  1 dd   0,   4,   1,   5,   2,   6,   3,   7
+const deinterleave_shufd,   times  1 dd   0,   4,   1,   5,   2,   6,   3,   7
 
 const popcnt_table
 %assign x 0

x265_1.6.tar.gz/source/common/x86/dct8.asm -> x265_1.7.tar.gz/source/common/x86/dct8.asm Changed

@@ -261,6 +261,11 @@
                 times 2 dw 84, -29, -74, 55
                 times 2 dw 55, -84, 74, -29
 
+pw_dst4_tab:    times 4 dw 29,  55,  74,  84
+                times 4 dw 74,  74,   0, -74
+                times 4 dw 84, -29, -74,  55
+                times 4 dw 55, -84,  74, -29
+
 tab_idst4:      times 4 dw 29, +84
                 times 4 dw +74, +55
                 times 4 dw 55, -29
@@ -270,6 +275,16 @@
                 times 4 dw 84, +55
                 times 4 dw -74, -29
 
+pw_idst4_tab:   times 4 dw  29,  84
+                times 4 dw  55, -29
+                times 4 dw  74,  55
+                times 4 dw  74, -84
+                times 4 dw  74, -74
+                times 4 dw  84,  55
+                times 4 dw  0,   74
+                times 4 dw -74, -29
+pb_idst4_shuf:  times 2 db 0, 1, 8, 9, 2, 3, 10, 11, 4, 5, 12, 13, 6, 7, 14, 15
+
 tab_dct8_1:     times 2 dw 89, 50, 75, 18
                 times 2 dw 75, -89, -18, -50
                 times 2 dw 50, 18, -89, 75
@@ -316,7 +331,7 @@
 cextern pd_1024
 cextern pd_2048
 cextern pw_ppppmmmm
-
+cextern trans8_shuf
 ;------------------------------------------------------
 ;void dct4(const int16_t* src, int16_t* dst, intptr_t srcStride)
 ;------------------------------------------------------
@@ -656,6 +671,59 @@
 
     RET
 
+;------------------------------------------------------------------
+;void dst4(const int16_t* src, int16_t* dst, intptr_t srcStride)
+;------------------------------------------------------------------
+INIT_YMM avx2
+cglobal dst4, 3, 4, 6
+%if BIT_DEPTH == 8
+  %define       DST_SHIFT 1
+  vpbroadcastd  m5, [pd_1]
+%elif BIT_DEPTH == 10
+  %define       DST_SHIFT 3
+  vpbroadcastd  m5, [pd_4]
+%endif
+    mova        m4, [trans8_shuf]
+    add         r2d, r2d
+    lea         r3, [pw_dst4_tab]
+
+    movq        xm0, [r0 + 0 * r2]
+    movhps      xm0, [r0 + 1 * r2]
+    lea         r0, [r0 + 2 * r2]
+    movq        xm1, [r0]
+    movhps      xm1, [r0 + r2]
+
+    vinserti128 m0, m0, xm1, 1          ; m0 = src[0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15]
+
+    pmaddwd     m2, m0, [r3 + 0 * 32]
+    pmaddwd     m1, m0, [r3 + 1 * 32]
+    phaddd      m2, m1
+    paddd       m2, m5
+    psrad       m2, DST_SHIFT
+    pmaddwd     m3, m0, [r3 + 2 * 32]
+    pmaddwd     m1, m0, [r3 + 3 * 32]
+    phaddd      m3, m1
+    paddd       m3, m5
+    psrad       m3, DST_SHIFT
+    packssdw    m2, m3
+    vpermd      m2, m4, m2
+
+    vpbroadcastd m5, [pd_128]
+    pmaddwd     m0, m2, [r3 + 0 * 32]
+    pmaddwd     m1, m2, [r3 + 1 * 32]
+    phaddd      m0, m1
+    paddd       m0, m5
+    psrad       m0, 8
+    pmaddwd     m3, m2, [r3 + 2 * 32]
+    pmaddwd     m2, m2, [r3 + 3 * 32]
+    phaddd      m3, m2
+    paddd       m3, m5
+    psrad       m3, 8
+    packssdw    m0, m3
+    vpermd      m0, m4, m0
+    movu        [r1], m0
+    RET
+
 ;-------------------------------------------------------
 ;void idst4(const int16_t* src, int16_t* dst, intptr_t dstStride)
 ;-------------------------------------------------------
@@ -748,6 +816,81 @@
     movhps      [r1 + r2], m1
     RET
 
+;-----------------------------------------------------------------
+;void idst4(const int16_t* src, int16_t* dst, intptr_t dstStride)
+;-----------------------------------------------------------------
+INIT_YMM avx2
+cglobal idst4, 3, 4, 6
+%if BIT_DEPTH == 8
+  vpbroadcastd  m4, [pd_2048]
+  %define       IDCT4_SHIFT 12
+%elif BIT_DEPTH == 10
+  vpbroadcastd  m4, [pd_512]
+  %define       IDCT4_SHIFT 10
+%else
+  %error Unsupported BIT_DEPTH!
+%endif
+    add         r2d, r2d
+    lea         r3, [pw_idst4_tab]
+
+    movu        xm0, [r0 + 0 * 16]
+    movu        xm1, [r0 + 1 * 16]
+
+    punpcklwd   m2, m0, m1
+    punpckhwd   m0, m1
+
+    vinserti128 m2, m2, xm2, 1
+    vinserti128 m0, m0, xm0, 1
+
+    vpbroadcastd m5, [pd_64]
+    pmaddwd     m1, m2, [r3 + 0 * 32]
+    pmaddwd     m3, m0, [r3 + 1 * 32]
+    paddd       m1, m3
+    paddd       m1, m5
+    psrad       m1, 7
+    pmaddwd     m3, m2, [r3 + 2 * 32]
+    pmaddwd     m0, [r3 + 3 * 32]
+    paddd       m3, m0
+    paddd       m3, m5
+    psrad       m3, 7
+
+    packssdw    m0, m1, m3
+    pshufb      m0, [pb_idst4_shuf]
+    vpermq      m1, m0, 11101110b
+
+    punpcklwd   m2, m0, m1
+    punpckhwd   m0, m1
+    punpcklwd   m1, m2, m0
+    punpckhwd   m2, m0
+
+    vpermq      m1, m1, 01000100b
+    vpermq      m2, m2, 01000100b
+
+    pmaddwd     m0, m1, [r3 + 0 * 32]
+    pmaddwd     m3, m2, [r3 + 1 * 32]
+    paddd       m0, m3
+    paddd       m0, m4
+    psrad       m0, IDCT4_SHIFT
+    pmaddwd     m3, m1, [r3 + 2 * 32]
+    pmaddwd     m2, m2, [r3 + 3 * 32]
+    paddd       m3, m2
+    paddd       m3, m4
+    psrad       m3, IDCT4_SHIFT
+
+    packssdw    m0, m3
+    pshufb      m1, m0, [pb_idst4_shuf]
+    vpermq      m0, m1, 11101110b
+
+    punpcklwd   m2, m1, m0
+    movq        [r1 + 0 * r2], xm2
+    movhps      [r1 + 1 * r2], xm2
+
+    punpckhwd   m1, m0
+    movq        [r1 + 2 * r2], xm1
+    lea         r1, [r1 + 2 * r2]
+    movhps      [r1 + r2], xm1
+    RET
+
 ;-------------------------------------------------------
 ; void dct8(const int16_t* src, int16_t* dst, intptr_t srcStride)
 ;-------------------------------------------------------

x265_1.6.tar.gz/source/common/x86/dct8.h -> x265_1.7.tar.gz/source/common/x86/dct8.h Changed

@@ -26,6 +26,7 @@
 void x265_dct4_sse2(const int16_t* src, int16_t* dst, intptr_t srcStride);
 void x265_dct8_sse2(const int16_t* src, int16_t* dst, intptr_t srcStride);
 void x265_dst4_ssse3(const int16_t* src, int16_t* dst, intptr_t srcStride);
+void x265_dst4_avx2(const int16_t* src, int16_t* dst, intptr_t srcStride);
 void x265_dct8_sse4(const int16_t* src, int16_t* dst, intptr_t srcStride);
 void x265_dct4_avx2(const int16_t* src, int16_t* dst, intptr_t srcStride);
 void x265_dct8_avx2(const int16_t* src, int16_t* dst, intptr_t srcStride);
@@ -33,6 +34,7 @@
 void x265_dct32_avx2(const int16_t* src, int16_t* dst, intptr_t srcStride);
 
 void x265_idst4_sse2(const int16_t* src, int16_t* dst, intptr_t dstStride);
+void x265_idst4_avx2(const int16_t* src, int16_t* dst, intptr_t dstStride);
 void x265_idct4_sse2(const int16_t* src, int16_t* dst, intptr_t dstStride);
 void x265_idct4_avx2(const int16_t* src, int16_t* dst, intptr_t dstStride);
 void x265_idct8_sse2(const int16_t* src, int16_t* dst, intptr_t dstStride);

x265_1.6.tar.gz/source/common/x86/intrapred.h -> x265_1.7.tar.gz/source/common/x86/intrapred.h Changed

@@ -34,6 +34,7 @@
 void x265_intra_pred_dc8_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int filter);
 void x265_intra_pred_dc16_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int filter);
 void x265_intra_pred_dc32_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int filter);
+void x265_intra_pred_dc32_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int filter);
 
 void x265_intra_pred_planar4_sse2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int);
 void x265_intra_pred_planar8_sse2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int);
@@ -43,6 +44,8 @@
 void x265_intra_pred_planar8_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int);
 void x265_intra_pred_planar16_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int);
 void x265_intra_pred_planar32_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int);
+void x265_intra_pred_planar16_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int);
+void x265_intra_pred_planar32_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int);
 
 #define DECL_ANG(bsize, mode, cpu) \
     void x265_intra_pred_ang ## bsize ## _ ## mode ## _ ## cpu(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
@@ -55,6 +58,16 @@
 DECL_ANG(4, 7, sse2);
 DECL_ANG(4, 8, sse2);
 DECL_ANG(4, 9, sse2);
+DECL_ANG(4, 10, sse2);
+DECL_ANG(4, 11, sse2);
+DECL_ANG(4, 12, sse2);
+DECL_ANG(4, 13, sse2);
+DECL_ANG(4, 14, sse2);
+DECL_ANG(4, 15, sse2);
+DECL_ANG(4, 16, sse2);
+DECL_ANG(4, 17, sse2);
+DECL_ANG(4, 18, sse2);
+DECL_ANG(4, 26, sse2);
 
 DECL_ANG(4, 2, ssse3);
 DECL_ANG(4, 3, sse4);
@@ -174,6 +187,34 @@
 DECL_ANG(32, 33, sse4);
 
 #undef DECL_ANG
+void x265_intra_pred_ang4_3_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_4_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_5_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_6_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_7_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_8_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_9_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_11_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_12_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_13_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_14_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_15_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_16_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_17_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_19_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_20_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_21_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_22_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_23_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_24_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_25_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_27_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_28_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_29_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_30_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_31_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_32_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang4_33_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_3_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_33_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_4_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
@@ -192,6 +233,24 @@
 void x265_intra_pred_ang8_12_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_24_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_11_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_13_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_14_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_15_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_16_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_20_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_21_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_22_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_23_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_3_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_4_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_5_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_6_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_7_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_8_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_9_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_12_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_11_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_13_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang16_25_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang16_28_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang16_27_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
@@ -212,8 +271,17 @@
 void x265_intra_pred_ang32_30_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang32_31_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang32_32_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang32_33_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang32_25_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang32_24_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang32_23_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang32_22_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang32_21_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang32_18_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_all_angs_pred_4x4_sse2(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
 void x265_all_angs_pred_4x4_sse4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
 void x265_all_angs_pred_8x8_sse4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
 void x265_all_angs_pred_16x16_sse4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
 void x265_all_angs_pred_32x32_sse4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
+void x265_all_angs_pred_4x4_avx2(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
 #endif // ifndef X265_INTRAPRED_H

x265_1.6.tar.gz/source/common/x86/intrapred16.asm -> x265_1.7.tar.gz/source/common/x86/intrapred16.asm Changed

@@ -690,6 +690,508 @@
 %endrep
     RET
 
+;-----------------------------------------------------------------------------------------
+; void intraPredAng4(pixel* dst, intptr_t dstStride, pixel* src, int dirMode, int bFilter)
+;-----------------------------------------------------------------------------------------
+INIT_XMM sse2
+cglobal intra_pred_ang4_2, 3,5,4
+    lea         r4,            [r2 + 4]
+    add         r2,            20
+    cmp         r3m,           byte 34
+    cmove       r2,            r4
+
+    add         r1,            r1
+    movu        m0,            [r2]
+    movh        [r0],          m0
+    psrldq      m0,            2
+    movh        [r0 + r1],     m0
+    psrldq      m0,            2
+    movh        [r0 + r1 * 2], m0
+    lea         r1,            [r1 * 3]
+    psrldq      m0,            2
+    movh        [r0 + r1],     m0
+    RET
+
+cglobal intra_pred_ang4_3, 3,5,8
+    mov         r4d, 2
+    cmp         r3m, byte 33
+    mov         r3d, 18
+    cmove       r3d, r4d
+
+    movu        m0, [r2 + r3]   ; [8 7 6 5 4 3 2 1]
+
+    mova        m2, m0
+    psrldq      m0, 2
+    punpcklwd   m2, m0      ; [5 4 4 3 3 2 2 1]
+    mova        m3, m0
+    psrldq      m0, 2
+    punpcklwd   m3, m0      ; [6 5 5 4 4 3 3 2]
+    mova        m4, m0
+    psrldq      m0, 2
+    punpcklwd   m4, m0      ; [7 6 6 5 5 4 4 3]
+    mova        m5, m0
+    psrldq      m0, 2
+    punpcklwd   m5, m0      ; [8 7 7 6 6 5 5 4]
+
+
+    lea         r3, [ang_table + 20 * 16]
+    mova        m0, [r3 + 6 * 16]   ; [26]
+    mova        m1, [r3]            ; [20]
+    mova        m6, [r3 - 6 * 16]   ; [14]
+    mova        m7, [r3 - 12 * 16]  ; [ 8]
+    jmp        .do_filter4x4
+
+
+ALIGN 16
+.do_filter4x4:
+    lea     r4, [pd_16]
+    pmaddwd m2, m0
+    paddd   m2, [r4]
+    psrld   m2, 5
+
+    pmaddwd m3, m1
+    paddd   m3, [r4]
+    psrld   m3, 5
+    packssdw m2, m3
+
+    pmaddwd m4, m6
+    paddd   m4, [r4]
+    psrld   m4, 5
+
+    pmaddwd m5, m7
+    paddd   m5, [r4]
+    psrld   m5, 5
+    packssdw m4, m5
+
+    jz         .store
+
+    ; transpose 4x4
+    punpckhwd    m0, m2, m4
+    punpcklwd    m2, m4
+    punpckhwd    m4, m2, m0
+    punpcklwd    m2, m0
+
+.store:
+    add         r1, r1
+    movh        [r0], m2
+    movhps      [r0 + r1], m2
+    movh        [r0 + r1 * 2], m4
+    lea         r1, [r1 * 3]
+    movhps      [r0 + r1], m4
+    RET
+
+cglobal intra_pred_ang4_4, 3,5,8
+    mov         r4d, 2
+    cmp         r3m, byte 32
+    mov         r3d, 18
+    cmove       r3d, r4d
+
+    movu        m0, [r2 + r3]   ; [8 7 6 5 4 3 2 1]
+    mova        m2, m0
+    psrldq      m0, 2
+    punpcklwd   m2, m0      ; [5 4 4 3 3 2 2 1]
+    mova        m3, m0
+    psrldq      m0, 2
+    punpcklwd   m3, m0      ; [6 5 5 4 4 3 3 2]
+    mova        m4, m3
+    mova        m5, m0
+    psrldq      m0, 2
+    punpcklwd   m5, m0      ; [7 6 6 5 5 4 4 3]
+
+    lea         r3, [ang_table + 18 * 16]
+    mova        m0, [r3 +  3 * 16]  ; [21]
+    mova        m1, [r3 -  8 * 16]  ; [10]
+    mova        m6, [r3 + 13 * 16]  ; [31]
+    mova        m7, [r3 +  2 * 16]  ; [20]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_5, 3,5,8
+    mov         r4d, 2
+    cmp         r3m, byte 31
+    mov         r3d, 18
+    cmove       r3d, r4d
+
+    movu        m0, [r2 + r3]   ; [8 7 6 5 4 3 2 1]
+    mova        m2, m0
+    psrldq      m0, 2
+    punpcklwd   m2, m0      ; [5 4 4 3 3 2 2 1]
+    mova        m3, m0
+    psrldq      m0, 2
+    punpcklwd   m3, m0      ; [6 5 5 4 4 3 3 2]
+    mova        m4, m3
+    mova        m5, m0
+    psrldq      m0, 2
+    punpcklwd   m5, m0      ; [7 6 6 5 5 4 4 3]
+
+    lea         r3, [ang_table + 10 * 16]
+    mova        m0, [r3 +  7 * 16]  ; [17]
+    mova        m1, [r3 -  8 * 16]  ; [ 2]
+    mova        m6, [r3 +  9 * 16]  ; [19]
+    mova        m7, [r3 -  6 * 16]  ; [ 4]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_6, 3,5,8
+    mov         r4d, 2
+    cmp         r3m, byte 30
+    mov         r3d, 18
+    cmove       r3d, r4d
+
+    movu        m0, [r2 + r3]   ; [8 7 6 5 4 3 2 1]
+    mova        m2, m0
+    psrldq      m0, 2
+    punpcklwd   m2, m0      ; [5 4 4 3 3 2 2 1]
+    mova        m3, m2
+    mova        m4, m0
+    psrldq      m0, 2
+    punpcklwd   m4, m0      ; [6 5 5 4 4 3 3 2]
+    mova        m5, m4
+
+    lea         r3, [ang_table + 19 * 16]
+    mova        m0, [r3 -  6 * 16]  ; [13]
+    mova        m1, [r3 +  7 * 16]  ; [26]
+    mova        m6, [r3 - 12 * 16]  ; [ 7]
+    mova        m7, [r3 +  1 * 16]  ; [20]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_7, 3,5,8
+    mov         r4d, 2
+    cmp         r3m, byte 29
+    mov         r3d, 18
+    cmove       r3d, r4d
+
+    movu        m0, [r2 + r3]   ; [8 7 6 5 4 3 2 1]
+    mova        m2, m0
+    psrldq      m0, 2
+    punpcklwd   m2, m0      ; [5 4 4 3 3 2 2 1]
+    mova        m3, m2
+    mova        m4, m2
+    mova        m5, m0
+    psrldq      m0, 2
+    punpcklwd   m5, m0      ; [6 5 5 4 4 3 3 2]
+
+    lea         r3, [ang_table + 20 * 16]
+    mova        m0, [r3 - 11 * 16]  ; [ 9]
+    mova        m1, [r3 -  2 * 16]  ; [18]
+    mova        m6, [r3 +  7 * 16]  ; [27]
+    mova        m7, [r3 - 16 * 16]  ; [ 4]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_8, 3,5,8
+    mov         r4d, 2
+    cmp         r3m, byte 28
+    mov         r3d, 18
+    cmove       r3d, r4d
+
+    movu        m0, [r2 + r3]   ; [8 7 6 5 4 3 2 1]
+    mova        m2, m0
+    psrldq      m0, 2
+    punpcklwd   m2, m0      ; [5 4 4 3 3 2 2 1]
+    mova        m3, m2
+    mova        m4, m2
+    mova        m5, m2
+
+    lea         r3, [ang_table + 13 * 16]
+    mova        m0, [r3 -  8 * 16]  ; [ 5]
+    mova        m1, [r3 -  3 * 16]  ; [10]
+    mova        m6, [r3 +  2 * 16]  ; [15]
+    mova        m7, [r3 +  7 * 16]  ; [20]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_9, 3,5,8
+    mov         r4d, 2
+    cmp         r3m, byte 27
+    mov         r3d, 18
+    cmove       r3d, r4d
+
+    movu        m0, [r2 + r3]   ; [8 7 6 5 4 3 2 1]
+    mova        m2, m0
+    psrldq      m0, 2
+    punpcklwd   m2, m0      ; [5 4 4 3 3 2 2 1]
+    mova        m3, m2
+    mova        m4, m2
+    mova        m5, m2
+
+    lea         r3, [ang_table + 4 * 16]
+    mova        m0, [r3 -  2 * 16]  ; [ 2]
+    mova        m1, [r3 -  0 * 16]  ; [ 4]
+    mova        m6, [r3 +  2 * 16]  ; [ 6]
+    mova        m7, [r3 +  4 * 16]  ; [ 8]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_10, 3,3,3
+    movh        m0,             [r2 + 18]           ; [4 3 2 1]
+
+    punpcklwd   m0,             m0              ;[4 4 3 3 2 2 1 1]
+    pshufd      m1,             m0, 0xFA
+    add         r1,             r1
+    pshufd      m0,             m0, 0x50
+    movhps      [r0 + r1],      m0
+    movh        [r0 + r1 * 2],  m1
+    lea         r1,             [r1 * 3]
+    movhps      [r0 + r1],      m1
+
+    cmp         r4m,            byte 0
+    jz         .quit
+
+    ; filter
+    movd        m2,             [r2]                ; [7 6 5 4 3 2 1 0]
+    pshuflw     m2,             m2, 0x00
+    movh        m1,             [r2 + 2]
+    psubw       m1,             m2
+    psraw       m1,             1
+    paddw       m0,             m1
+    pxor        m1,             m1
+    pmaxsw      m0,             m1
+    pminsw      m0,             [pw_1023]
+.quit:
+    movh        [r0],           m0
+    RET
+
+cglobal intra_pred_ang4_26, 3,3,3
+    movh        m0,             [r2 + 2]            ; [8 7 6 5 4 3 2 1]
+    add         r1d,            r1d
+    ; store
+    movh        [r0],           m0
+    movh        [r0 + r1],      m0
+    movh        [r0 + r1 * 2],  m0
+    lea         r3,             [r1 * 3]
+    movh        [r0 + r3],      m0
+
+    ; filter
+    cmp         r4m,            byte 0
+    jz         .quit
+
+    pshuflw     m0,             m0, 0x00
+    movd        m2,             [r2]
+    pshuflw     m2,             m2, 0x00
+    movh        m1,             [r2 + 18]
+    psubw       m1,             m2
+    psraw       m1,             1
+    paddw       m0,             m1
+    pxor        m1,             m1
+    pmaxsw      m0,             m1
+    pminsw      m0,             [pw_1023]
+
+    movh        r2,             m0
+    mov         [r0],           r2w
+    shr         r2,             16
+    mov         [r0 + r1],      r2w
+    shr         r2,             16
+    mov         [r0 + r1 * 2],  r2w
+    shr         r2,             16
+    mov         [r0 + r3],      r2w
+.quit:
+    RET
+
+cglobal intra_pred_ang4_11, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 25
+    mov         r3d, 16
+    cmove       r3d, r4d
+
+    movh        m1, [r2 + r3 + 2]   ; [x x x 4 3 2 1 0]
+    movh        m2, [r2 - 6]
+    punpcklqdq  m2, m1
+    psrldq      m2, 6
+    punpcklwd   m2, m1          ; [4 3 3 2 2 1 1 0]
+    mova        m3, m2
+    mova        m4, m2
+    mova        m5, m2
+
+    lea         r3, [ang_table + 24 * 16]
+    mova        m0, [r3 +  6 * 16]  ; [24]
+    mova        m1, [r3 +  4 * 16]  ; [26]
+    mova        m6, [r3 +  2 * 16]  ; [28]
+    mova        m7, [r3 +  0 * 16]  ; [30]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_12, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 24
+    mov         r3d, 16
+    cmove       r3d, r4d
+
+    movh        m1, [r2 + r3 + 2]
+    movh        m2, [r2 - 6]
+    punpcklqdq  m2, m1
+    psrldq      m2, 6
+    punpcklwd   m2, m1          ; [4 3 3 2 2 1 1 0]
+    mova        m3, m2
+    mova        m4, m2
+    mova        m5, m2
+
+    lea         r3, [ang_table + 20 * 16]
+    mova        m0, [r3 +  7 * 16]  ; [27]
+    mova        m1, [r3 +  2 * 16]  ; [22]
+    mova        m6, [r3 -  3 * 16]  ; [17]
+    mova        m7, [r3 -  8 * 16]  ; [12]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_13, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 23
+    mov         r3d, 16
+    jz          .next
+    xchg        r3d, r4d
+.next:
+    movd        m5, [r2 + r3 + 6]
+    movd        m2, [r2 - 2]
+    movh        m0, [r2 + r4 + 2]
+    punpcklwd   m5, m2
+    punpcklqdq  m5, m0
+    psrldq      m5, 4
+    mova        m2, m5
+    psrldq      m2, 2
+    punpcklwd   m5, m2          ; [3 2 2 1 1 0 0 x]
+    punpcklwd   m2, m0          ; [4 3 3 2 2 1 1 0]
+    mova        m3, m2
+    mova        m4, m2
+
+    lea         r3, [ang_table + 21 * 16]
+    mova        m0, [r3 +  2 * 16]  ; [23]
+    mova        m1, [r3 -  7 * 16]  ; [14]
+    mova        m6, [r3 - 16 * 16]  ; [ 5]
+    mova        m7, [r3 +  7 * 16]  ; [28]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_14, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 22
+    mov         r3d, 16
+    jz          .next
+    xchg        r3d, r4d
+.next:
+    movd        m5, [r2 + r3 + 2]
+    movd        m2, [r2 - 2]
+    movh        m0, [r2 + r4 + 2]
+    punpcklwd   m5, m2
+    punpcklqdq  m5, m0
+    psrldq      m5, 4
+    mova        m2, m5
+    psrldq      m2, 2
+    punpcklwd   m5, m2          ; [3 2 2 1 1 0 0 x]
+    punpcklwd   m2, m0          ; [4 3 3 2 2 1 1 0]
+    mova        m3, m2
+    mova        m4, m5
+
+    lea         r3, [ang_table + 19 * 16]
+    mova        m0, [r3 +  0 * 16]  ; [19]
+    mova        m1, [r3 - 13 * 16]  ; [ 6]
+    mova        m6, [r3 +  6 * 16]  ; [25]
+    mova        m7, [r3 -  7 * 16]  ; [12]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_15, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 21
+    mov         r3d, 16
+    jz          .next
+    xchg        r3d, r4d
+.next:
+    movd        m4, [r2]                ;[x x x A]
+    movh        m5, [r2 + r3 + 4]       ;[x C x B]
+    movh        m0, [r2 + r4 + 2]       ;[4 3 2 1]
+    pshuflw     m5, m5, 0x22            ;[B C B C]
+    punpcklqdq  m5, m4                  ;[x x x A B C B C]
+    psrldq      m5, 2                   ;[x x x x A B C B]
+    punpcklqdq  m5, m0
+    psrldq      m5, 2
+    mova        m2, m5
+    mova        m3, m5
+    psrldq      m2, 4
+    psrldq      m3, 2
+    punpcklwd   m5, m3          ; [2 1 1 0 0 x x y]
+    punpcklwd   m3, m2          ; [3 2 2 1 1 0 0 x]
+    punpcklwd   m2, m0          ; [4 3 3 2 2 1 1 0]
+    mova        m4, m3
+
+    lea         r3, [ang_table + 23 * 16]
+    mova        m0, [r3 -  8 * 16]  ; [15]
+    mova        m1, [r3 +  7 * 16]  ; [30]
+    mova        m6, [r3 - 10 * 16]  ; [13]
+    mova        m7, [r3 +  5 * 16]  ; [28]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_16, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 20
+    mov         r3d, 16
+    jz          .next
+    xchg        r3d, r4d
+.next:
+    movd        m4, [r2]                ;[x x x A]
+    movd        m5, [r2 + r3 + 4]       ;[x x C B]
+    movh        m0, [r2 + r4 + 2]       ;[4 3 2 1]
+    punpcklwd   m5, m4                  ;[x C A B]
+    pshuflw     m5, m5, 0x4A            ;[A B C C]
+    punpcklqdq  m5, m0                  ;[4 3 2 1 A B C C]
+    psrldq      m5, 2
+    mova        m2, m5
+    mova        m3, m5
+    psrldq      m2, 4
+    psrldq      m3, 2
+    punpcklwd   m5, m3          ; [2 1 1 0 0 x x y]
+    punpcklwd   m3, m2          ; [3 2 2 1 1 0 0 x]
+    punpcklwd   m2, m0          ; [4 3 3 2 2 1 1 0]
+    mova        m4, m3
+
+    lea         r3, [ang_table + 19 * 16]
+    mova        m0, [r3 -  8 * 16]  ; [11]
+    mova        m1, [r3 +  3 * 16]  ; [22]
+    mova        m6, [r3 - 18 * 16]  ; [ 1]
+    mova        m7, [r3 -  7 * 16]  ; [12]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_17, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 19
+    mov         r3d, 16
+    jz          .next
+    xchg        r3d, r4d
+.next:
+    movd        m4, [r2]
+    movh        m5, [r2 + r3 + 2]       ;[D x C B]
+    pshuflw     m5, m5, 0x1F            ;[B C D D]
+    punpcklqdq  m5, m4                  ;[x x x A B C D D]
+    psrldq      m5, 2                   ;[x x x x A B C D]
+    movhps      m5, [r2 + r4 + 2]
+
+    mova        m4, m5
+    psrldq      m4, 2
+    punpcklwd   m5, m4
+    mova        m3, m4
+    psrldq      m3, 2
+    punpcklwd   m4, m3
+    mova        m2, m3
+    psrldq      m2, 2
+    punpcklwd   m3, m2
+    mova        m1, m2
+    psrldq      m1, 2
+    punpcklwd   m2, m1
+
+    lea         r3, [ang_table + 14 * 16]
+    mova        m0, [r3 -  8 * 16]  ; [ 6]
+    mova        m1, [r3 -  2 * 16]  ; [12]
+    mova        m6, [r3 +  4 * 16]  ; [18]
+    mova        m7, [r3 + 10 * 16]  ; [24]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_18, 3,3,1
+    movh        m0, [r2 + 16]
+    pinsrw      m0, [r2], 0
+    pshuflw     m0, m0, q0123
+    movhps      m0, [r2 + 2]
+    add         r1, r1
+    lea         r2, [r1 * 3]
+    movh        [r0 + r2], m0
+    psrldq      m0, 2
+    movh        [r0 + r1 * 2], m0
+    psrldq      m0, 2
+    movh        [r0 + r1], m0
+    psrldq      m0, 2
+    movh        [r0], m0
+    RET
+
 ;-----------------------------------------------------------------------------------
 ; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel* above, int, int filter)
 ;-----------------------------------------------------------------------------------

x265_1.6.tar.gz/source/common/x86/intrapred8.asm -> x265_1.7.tar.gz/source/common/x86/intrapred8.asm Changed

@@ -28,6 +28,7 @@
 SECTION_RODATA 32
 
 intra_pred_shuff_0_8:    times 2 db 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8
+intra_pred_shuff_15_0:   times 2 db 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0
 
 pb_0_8        times 8 db  0,  8
 pb_unpackbw1  times 2 db  1,  8,  2,  8,  3,  8,  4,  8
@@ -58,7 +59,6 @@
 c_mode16_18:    db 0, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1
 
 ALIGN 32
-trans8_shuf:          dd 0, 4, 1, 5, 2, 6, 3, 7
 c_ang8_src1_9_2_10:   db 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9
 c_ang8_26_20:         db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
 c_ang8_src3_11_4_12:  db 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11
@@ -124,6 +124,37 @@
                       db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
                       db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
 
+ALIGN 32
+c_ang16_mode_11:      db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                      db 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                      db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                      db 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                      db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                      db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                      db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                      db 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+
+ALIGN 32
+c_ang16_mode_12:      db 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19
+                      db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                      db 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9
+                      db 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                      db 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31
+                      db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
+                      db 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21
+                      db  8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+
+
+ALIGN 32
+c_ang16_mode_13:      db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15
+                      db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                      db 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29
+                      db 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
+                      db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11
+                      db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                      db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25
+                      db 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
 
 ALIGN 32
 c_ang16_mode_28:      db 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
@@ -135,6 +166,15 @@
                       db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
                       db 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
 
+ALIGN 32
+c_ang16_mode_9:       db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                      db 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
+                      db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
+                      db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+                      db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
+                      db 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                      db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30
+                      db 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
 
 ALIGN 32
 c_ang16_mode_27:      db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
@@ -150,6 +190,15 @@
 ALIGN 32
 intra_pred_shuff_0_15: db 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14, 14, 15, 15, 15
 
+ALIGN 32
+c_ang16_mode_8:       db 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13
+                      db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                      db 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23
+                      db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                      db 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1
+                      db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                      db 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11
+                      db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
 
 ALIGN 32
 c_ang16_mode_29:     db 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9,  14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
@@ -162,6 +211,15 @@
                      db 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30
                      db 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
 
+ALIGN 32
+c_ang16_mode_7:      db 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17
+                     db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
+                     db 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3
+                     db 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                     db 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21
+                     db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30
+                     db 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7
+                     db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
 
 ALIGN 32
 c_ang16_mode_30:      db 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
@@ -175,6 +233,17 @@
                       db 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
 
 
+
+ALIGN 32
+c_ang16_mode_6:       db 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21
+                      db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                      db 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15
+                      db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                      db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9
+                      db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
+                      db 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3
+                      db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+
 ALIGN 32
 c_ang16_mode_31:      db 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17
                       db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19
@@ -186,6 +255,17 @@
                       db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31
                       db 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
 
+
+ALIGN 32
+c_ang16_mode_5:       db 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25
+                      db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                      db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27
+                      db 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                      db 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29
+                      db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                      db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31
+                      db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+
 ALIGN 32
 c_ang16_mode_32:      db 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21
                       db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31
@@ -200,6 +280,16 @@
                       db 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
 
 ALIGN 32
+c_ang16_mode_4:       db 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29
+                      db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                      db 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7
+                      db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                      db 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17
+                      db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                      db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27
+                      db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+
+ALIGN 32
 c_ang16_mode_33:     db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
                      db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
                      db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
@@ -216,6 +306,16 @@
                      db 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
 
 ALIGN 32
+c_ang16_mode_3:      db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                     db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                     db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30
+                     db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+                     db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                     db 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                     db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                     db 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+ALIGN 32
 c_ang16_mode_24:     db 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
                      db 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
                      db 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
@@ -376,6 +476,191 @@
                    db 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11
                    db 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
 
+
+ALIGN 32
+c_ang32_mode_33:   db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
+                   db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
+                   db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                   db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                   db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                   db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
+                   db 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                   db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                   db 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30
+                   db 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+                   db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                   db 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                   db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                   db 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
+                   db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
+                   db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                   db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                   db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                   db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
+                   db 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                   db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                   db 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30
+                   db 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+                   db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                   db 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                   db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                   db 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+
+
+ALIGN 32
+c_ang32_mode_25:   db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                   db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+                   db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
+                   db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                   db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                   db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                   db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                   db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+                   db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                   db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+                   db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
+                   db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                   db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                   db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                   db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                   db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+
+
+ALIGN 32
+c_ang32_mode_24:   db 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
+                   db 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                   db 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                   db 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+                   db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                   db 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                   db 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
+                   db 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                   db 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                   db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1
+                   db 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23
+                   db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13
+                   db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3
+                   db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25
+                   db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15
+                   db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5
+                   db 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+
+ALIGN 32
+c_ang32_mode_23:  db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                  db 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5
+                  db 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19
+                  db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1
+                  db 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15
+                  db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                  db 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
+                  db 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                  db 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                  db 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7
+                  db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21
+                  db 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3
+                  db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17
+                  db 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                  db 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
+                  db 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                  db 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                  db 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+
+ALIGN 32
+c_ang32_mode_22: db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                 db 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                 db 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                 db 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5
+                 db 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11
+                 db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17
+                 db 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                 db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                 db 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                 db 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3
+                 db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9
+                 db 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15
+                 db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                 db 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                 db 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                 db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1
+                 db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7
+                 db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13
+                 db 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+ALIGN 32
+c_ang32_mode_21: db 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15
+                 db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13
+                 db 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11
+                 db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9
+                 db 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7
+                 db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5
+                 db 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3
+                 db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1
+                 db 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                 db 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                 db 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                 db 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                 db 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                 db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                 db 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                 db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                 db 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+
+ALIGN 32
+intra_pred_shuff_0_4:    times 4 db 0, 1, 1, 2, 2, 3, 3, 4
+intra_pred4_shuff1:      db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5
+intra_pred4_shuff2:      db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5
+intra_pred4_shuff31:     db 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6
+intra_pred4_shuff33:     db 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6, 3, 4, 4, 5, 5, 6, 6, 7
+intra_pred4_shuff3:      db 8, 9, 9, 10, 10, 11, 11, 12, 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14, 11, 12, 12, 13, 13, 14, 14, 15
+intra_pred4_shuff4:      db 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14, 10, 11, 11, 12, 12, 13, 13, 14, 11, 12, 12, 13, 13, 14, 14, 15
+intra_pred4_shuff5:      db 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14, 10, 11, 11, 12, 12, 13, 13, 14, 11, 12, 12, 13, 13, 14, 14, 15
+intra_pred4_shuff6:      db 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14, 10, 11, 11, 12, 12, 13, 13, 14
+intra_pred4_shuff7:      db 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14
+intra_pred4_shuff9:      db 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13
+intra_pred4_shuff12:     db 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12,0, 9, 9, 10, 10, 11, 11, 12
+intra_pred4_shuff13:     db 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 4, 0, 0, 9, 9, 10, 10, 11
+intra_pred4_shuff14:     db 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 2, 0, 0, 9, 9, 10, 10, 11, 2, 0, 0, 9, 9, 10, 10, 11
+intra_pred4_shuff15:     db 0, 9, 9, 10, 10, 11, 11, 12, 2, 0, 0, 9, 9, 10, 10, 11, 2, 0, 0, 9, 9, 10, 10, 11, 4, 2, 2, 0, 0, 9, 9, 10
+intra_pred4_shuff16:     db 0, 9, 9, 10, 10, 11, 11, 12, 2, 0, 0, 9, 9, 10, 10, 11, 2, 0, 0, 9, 9, 10, 10, 11, 3, 2, 2, 0, 0, 9, 9, 10
+intra_pred4_shuff17:     db 0, 9, 9, 10, 10, 11, 11, 12, 1, 0, 0, 9, 9, 10, 10, 11, 2, 1, 1, 0, 0, 9, 9, 10, 4, 2, 2, 1, 1, 0, 0, 9
+intra_pred4_shuff19:     db 0, 1, 1, 2, 2, 3, 3, 4, 9, 0, 0, 1, 1, 2, 2, 3, 10, 9, 9, 0, 0, 1, 1, 2, 12, 10, 10, 9, 9, 0, 0, 1
+intra_pred4_shuff20:     db 0, 1, 1, 2, 2, 3, 3, 4, 10, 0, 0, 1, 1, 2, 2, 3, 10, 0, 0, 1, 1, 2, 2, 3, 11, 10, 10, 0, 0, 1, 1, 2
+intra_pred4_shuff21:     db 0, 1, 1, 2, 2, 3, 3, 4, 10, 0, 0, 1, 1, 2, 2, 3, 10, 0, 0, 1, 1, 2, 2, 3, 12, 10, 10, 0, 0, 1, 1, 2
+intra_pred4_shuff22:     db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 10, 0, 0, 1, 1, 2, 2, 3, 10, 0, 0, 1, 1, 2, 2, 3
+intra_pred4_shuff23:     db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 12, 0, 0, 1, 1, 2, 2, 3
+
+c_ang4_mode_27:          db 30, 2, 30, 2, 30, 2, 30, 2, 28, 4, 28, 4, 28, 4, 28, 4, 26, 6, 26, 6, 26, 6, 26, 6, 24, 8, 24, 8, 24, 8, 24, 8
+c_ang4_mode_28:          db 27, 5, 27, 5, 27, 5, 27, 5, 22, 10, 22, 10, 22, 10, 22, 10, 17, 15, 17, 15, 17, 15, 17, 15, 12, 20, 12, 20, 12, 20, 12, 20
+c_ang4_mode_29:          db 23, 9, 23, 9, 23, 9, 23, 9, 14, 18, 14, 18, 14, 18, 14, 18, 5, 27, 5, 27, 5, 27, 5, 27, 28, 4, 28, 4, 28, 4, 28, 4
+c_ang4_mode_30:          db 19, 13, 19, 13, 19, 13, 19, 13, 6, 26, 6, 26, 6, 26, 6, 26, 25, 7, 25, 7, 25, 7, 25, 7, 12, 20, 12, 20, 12, 20, 12, 20
+c_ang4_mode_31:          db 15, 17, 15, 17, 15, 17, 15, 17, 30, 2, 30, 2, 30, 2, 30, 2, 13, 19, 13, 19, 13, 19, 13, 19, 28, 4, 28, 4, 28, 4, 28, 4
+c_ang4_mode_32:          db 11, 21, 11, 21, 11, 21, 11, 21, 22, 10, 22, 10, 22, 10, 22, 10, 1, 31, 1, 31, 1, 31, 1, 31, 12, 20, 12, 20, 12, 20, 12, 20
+c_ang4_mode_33:          db 6, 26, 6, 26, 6, 26, 6, 26, 12, 20, 12, 20, 12, 20, 12, 20, 18, 14, 18, 14, 18, 14, 18, 14, 24, 8, 24, 8, 24, 8, 24, 8
+c_ang4_mode_5:           db 15, 17, 15, 17, 15, 17, 15, 17, 30, 2, 30, 2, 30, 2, 30, 2, 13, 19, 13, 19, 13, 19, 13, 19, 28, 4, 28, 4, 28, 4, 28, 4
+c_ang4_mode_6:           db 19, 13, 19, 13, 19, 13, 19, 13, 6, 26, 6, 26, 6, 26, 6, 26, 25, 7, 25, 7, 25, 7, 25, 7, 12, 20, 12, 20, 12, 20, 12, 20
+c_ang4_mode_7:           db 23, 9, 23, 9, 23, 9, 23, 9, 14, 18, 14, 18, 14, 18, 14, 18, 5, 27, 5, 27, 5, 27, 5, 27, 28, 4, 28, 4, 28, 4, 28, 4
+c_ang4_mode_8:           db 27, 5, 27, 5, 27, 5, 27, 5, 22, 10, 22, 10, 22, 10, 22, 10, 17, 15, 17, 15, 17, 15, 17, 15, 12, 20, 12, 20, 12, 20, 12, 20
+c_ang4_mode_9:           db 30, 2, 30, 2, 30, 2, 30, 2, 28, 4, 28, 4, 28, 4, 28, 4, 26, 6, 26, 6, 26, 6, 26, 6, 24, 8, 24, 8, 24, 8, 24, 8
+c_ang4_mode_11:          db 2, 30, 2, 30, 2, 30, 2, 30, 4, 28, 4, 28, 4, 28, 4, 28, 6, 26, 6, 26, 6, 26, 6, 26, 8, 24, 8, 24, 8, 24, 8, 24
+c_ang4_mode_12:          db 5, 27, 5, 27, 5, 27, 5, 27, 10, 22, 10, 22, 10, 22, 10, 22, 15, 17, 15, 17, 15, 17, 15, 17, 20, 12, 20, 12, 20, 12, 20, 12
+c_ang4_mode_13:          db 9, 23, 9, 23, 9, 23, 9, 23, 18, 14, 18, 14, 18, 14, 18, 14, 27, 5, 27, 5, 27, 5, 27, 5, 4, 28, 4, 28, 4, 28, 4, 28
+c_ang4_mode_14:          db 13, 19, 13, 19, 13, 19, 13, 19, 26, 6, 26, 6, 26, 6, 26, 6, 7, 25, 7, 25, 7, 25, 7, 25, 20, 12, 20, 12, 20, 12, 20, 12
+c_ang4_mode_15:          db 17, 15, 17, 15, 17, 15, 17, 15, 2, 30, 2, 30, 2, 30, 2, 30, 19, 13, 19, 13, 19, 13, 19, 13, 4, 28, 4, 28, 4, 28, 4, 28, 4
+c_ang4_mode_16:          db 21, 11, 21, 11, 21, 11, 21, 11, 10, 22, 10, 22, 10, 22, 10, 22, 31, 1, 31, 1, 31, 1, 31, 1, 20, 12, 20, 12, 20, 12, 20, 12
+c_ang4_mode_17:          db 26, 6, 26, 6, 26, 6, 26, 6, 20, 12, 20, 12, 20, 12, 20, 12, 14, 18, 14, 18, 14, 18, 14, 18, 8, 24, 8, 24, 8, 24, 8, 24
+c_ang4_mode_19:          db 26, 6, 26, 6, 26, 6, 26, 6, 20, 12, 20, 12, 20, 12, 20, 12, 14, 18, 14, 18, 14, 18, 14, 18, 8, 24, 8, 24, 8, 24, 8, 24
+c_ang4_mode_20:          db 21, 11, 21, 11, 21, 11, 21, 11, 10, 22, 10, 22, 10, 22, 10, 22, 31, 1, 31, 1, 31, 1, 31, 1, 20, 12, 20, 12, 20, 12, 20, 12
+c_ang4_mode_21:          db 17, 15, 17, 15, 17, 15, 17, 15, 2, 30, 2, 30, 2, 30, 2, 30, 19, 13, 19, 13, 19, 13, 19, 13, 4, 28, 4, 28, 4, 28, 4, 28
+c_ang4_mode_22:          db 13, 19, 13, 19, 13, 19, 13, 19, 26, 6, 26, 6, 26, 6, 26, 6, 7, 25, 7, 25, 7, 25, 7, 25, 20, 12, 20, 12, 20, 12, 20, 12
+c_ang4_mode_23:          db 9, 23, 9, 23, 9, 23, 9, 23, 18, 14, 18, 14, 18, 14, 18, 14, 27, 5, 27, 5, 27, 5, 27, 5, 4, 28, 4, 28, 4, 28, 4, 28
+c_ang4_mode_24:          db 5, 27, 5, 27, 5, 27, 5, 27, 10, 22, 10, 22, 10, 22, 10, 22, 15, 17, 15, 17, 15, 17, 15, 17, 20, 12, 20, 12, 20, 12, 20, 12
+c_ang4_mode_25:          db 2, 30, 2, 30, 2, 30, 2, 30, 4, 28, 4, 28, 4, 28, 4, 28, 6, 26, 6, 26, 6, 26, 6, 26, 8, 24, 8, 24, 8, 24, 8, 24
+
 ALIGN 32
 ;; (blkSize - 1 - x)
 pw_planar4_0:         dw 3,  2,  1,  0,  3,  2,  1,  0
@@ -388,6 +673,29 @@
 pw_planar32_L:        dw 31, 30, 29, 28, 27, 26, 25, 24
 pw_planar32_H:        dw 23, 22, 21, 20, 19, 18, 17, 16
 
+ALIGN 32
+c_ang8_mode_13:       db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                      db 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                      db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                      db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+
+ALIGN 32
+c_ang8_mode_14:       db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6
+                      db 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                      db 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18
+                      db 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+
+ALIGN 32
+c_ang8_mode_15:       db 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 17, 15, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30
+                      db 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                      db 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
+                      db 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+
+ALIGN 32
+c_ang8_mode_20:       db 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
+                      db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                      db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                      db 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
 
 const ang_table
 %assign x 0
@@ -409,8 +717,11 @@
 cextern pw_4
 cextern pw_8
 cextern pw_16
+cextern pw_15
+cextern pw_31
 cextern pw_32
 cextern pw_257
+cextern pw_512
 cextern pw_1024
 cextern pw_4096
 cextern pw_00ff
@@ -420,6 +731,9 @@
 cextern multiH2
 cextern multiH3
 cextern multi_2Row
+cextern trans8_shuf
+cextern pw_planar16_mul
+cextern pw_planar32_mul
 
 ;---------------------------------------------------------------------------------------------
 ; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int dirMode, int bFilter)
@@ -1249,7 +1563,7 @@
 ; void intraPredAng4(pixel* dst, intptr_t dstStride, pixel* src, int dirMode, int bFilter)
 ;-----------------------------------------------------------------------------------------
 INIT_XMM sse2
-cglobal intra_pred_ang4_2, 3,5,3
+cglobal intra_pred_ang4_2, 3,5,1
     lea         r4, [r2 + 2]
     add         r2, 10
     cmp         r3m, byte 34
@@ -1257,23 +1571,21 @@
 
     movh        m0, [r2]
     movd        [r0], m0
-    mova        m1, m0
-    psrldq      m1, 1
-    movd        [r0 + r1], m1
-    mova        m2, m0
-    psrldq      m2, 2
-    movd        [r0 + r1 * 2], m2
+    psrldq      m0, 1
+    movd        [r0 + r1], m0
+    psrldq      m0, 1
+    movd        [r0 + r1 * 2], m0
     lea         r1, [r1 * 3]
-    psrldq      m0, 3
+    psrldq      m0, 1
     movd        [r0 + r1], m0
     RET
 
 INIT_XMM sse2
 cglobal intra_pred_ang4_3, 3,5,8
-    mov         r4, 1
+    mov         r4d, 1
     cmp         r3m, byte 33
-    mov         r3, 9
-    cmove       r3, r4
+    mov         r3d, 9
+    cmove       r3d, r4d
 
     movh        m0, [r2 + r3]   ; [8 7 6 5 4 3 2 1]
     mova        m1, m0
@@ -1299,7 +1611,6 @@
 ALIGN 16
 .do_filter4x4:
     pxor        m1, m1
-    pxor        m3, m3
     punpckhbw   m3, m0
     psrlw       m3, 8
     pmaddwd     m3, m5
@@ -1308,7 +1619,6 @@
     packssdw    m0, m3
     paddw       m0, [pw_16]
     psraw       m0, 5
-    pxor        m3, m3
     punpckhbw   m3, m2
     psrlw       m3, 8
     pmaddwd     m3, m7
@@ -1335,32 +1645,31 @@
 .store:
     packuswb    m0, m2
     movd        [r0], m0
-    pshufd      m0, m0, 0x39
+    psrldq      m0, 4
     movd        [r0 + r1], m0
-    pshufd      m0, m0, 0x39
+    psrldq      m0, 4
     movd        [r0 + r1 * 2], m0
     lea         r1, [r1 * 3]
-    pshufd      m0, m0, 0x39
+    psrldq      m0, 4
     movd        [r0 + r1], m0
     RET
 
 cglobal intra_pred_ang4_4, 3,5,8
-    xor         r4, r4
-    inc         r4
+    xor         r4d, r4d
+    inc         r4d
     cmp         r3m, byte 32
-    mov         r3, 9
-    cmove       r3, r4
+    mov         r3d, 9
+    cmove       r3d, r4d
 
     movh        m0, [r2 + r3]    ; [8 7 6 5 4 3 2 1]
+    punpcklbw   m0, m0
+    psrldq      m0, 1
+    mova        m2, m0
+    psrldq      m2, 2           ; [x x x x x x x x 6 5 5 4 4 3 3 2]
     mova        m1, m0
-    psrldq      m1, 1           ; [x 8 7 6 5 4 3 2]
-    punpcklbw   m0, m1          ; [x 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1]
-    mova        m1, m0
-    psrldq      m1, 2           ; [x x x x x x x x 6 5 5 4 4 3 3 2]
-    mova        m3, m0
-    psrldq      m3, 4           ; [x x x x x x x x 7 6 6 5 5 4 4 3]
-    punpcklqdq  m0, m1
-    punpcklqdq  m2, m1, m3
+    psrldq      m1, 4           ; [x x x x x x x x 7 6 6 5 5 4 4 3]
+    punpcklqdq  m0, m2
+    punpcklqdq  m2, m1
 
     lea         r3, [pw_ang_table + 18 * 16]
     mova        m4, [r3 +  3 * 16]  ; [21]
@@ -1370,22 +1679,21 @@
     jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
 
 cglobal intra_pred_ang4_5, 3,5,8
-    xor         r4, r4
-    inc         r4
+    xor         r4d, r4d
+    inc         r4d
     cmp         r3m, byte 31
-    mov         r3, 9
-    cmove       r3, r4
+    mov         r3d, 9
+    cmove       r3d, r4d
 
     movh        m0, [r2 + r3]    ; [8 7 6 5 4 3 2 1]
-    mova        m1, m0
-    psrldq      m1, 1           ; [x 8 7 6 5 4 3 2]
-    punpcklbw   m0, m1          ; [x 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1]
-    mova        m1, m0
-    psrldq      m1, 2           ; [x x x x x x x x 6 5 5 4 4 3 3 2]
+    punpcklbw   m0, m0          ; [x 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1]
+    psrldq      m0, 1
+    mova        m2, m0
+    psrldq      m2, 2           ; [x x x x x x x x 6 5 5 4 4 3 3 2]
     mova        m3, m0
     psrldq      m3, 4           ; [x x x x x x x x 7 6 6 5 5 4 4 3]
-    punpcklqdq  m0, m1
-    punpcklqdq  m2, m1, m3
+    punpcklqdq  m0, m2
+    punpcklqdq  m2, m3
 
     lea         r3, [pw_ang_table + 10 * 16]
     mova        m4, [r3 +  7 * 16]  ; [17]
@@ -1395,18 +1703,17 @@
     jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
 
 cglobal intra_pred_ang4_6, 3,5,8
-    xor         r4, r4
-    inc         r4
+    xor         r4d, r4d
+    inc         r4d
     cmp         r3m, byte 30
-    mov         r3, 9
-    cmove       r3, r4
+    mov         r3d, 9
+    cmove       r3d, r4d
 
     movh        m0, [r2 + r3]    ; [8 7 6 5 4 3 2 1]
-    mova        m1, m0
-    psrldq      m1, 1           ; [x 8 7 6 5 4 3 2]
-    punpcklbw   m0, m1          ; [x 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1]
+    punpcklbw   m0, m0
+    psrldq      m0, 1           ; [x 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1]
     mova        m2, m0
-    psrldq      m2, 2           ; [x x x x x x x x 6 5 5 4 4 3 3 2]
+    psrldq      m2, 2           ; [x x x 8 8 7 7 6 6 5 5 4 4 3 3 2]
     punpcklqdq  m0, m0
     punpcklqdq  m2, m2
 
@@ -1418,20 +1725,20 @@
     jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
 
 cglobal intra_pred_ang4_7, 3,5,8
-    xor         r4, r4
-    inc         r4
+    xor         r4d, r4d
+    inc         r4d
     cmp         r3m, byte 29
-    mov         r3, 9
-    cmove       r3, r4
+    mov         r3d, 9
+    cmove       r3d, r4d
 
     movh        m0, [r2 + r3]    ; [8 7 6 5 4 3 2 1]
-    mova        m1, m0
-    psrldq      m1, 1           ; [x 8 7 6 5 4 3 2]
-    punpcklbw   m0, m1          ; [x 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1]
-    mova        m3, m0
-    psrldq      m3, 2           ; [x x x x x x x x 6 5 5 4 4 3 3 2]
-    punpcklqdq  m2, m0, m3
+    punpcklbw   m0, m0
+    psrldq      m0, 1
+    mova        m2, m0
+    psrldq      m2, 2           ; [x x x x x x x x 6 5 5 4 4 3 3 2]
     punpcklqdq  m0, m0
+    punpcklqdq  m2, m2
+    movhlps     m2, m0
 
     lea         r3, [pw_ang_table + 20 * 16]
     mova        m4, [r3 - 11 * 16]  ; [ 9]
@@ -1441,16 +1748,15 @@
     jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
 
 cglobal intra_pred_ang4_8, 3,5,8
-    xor         r4, r4
-    inc         r4
+    xor         r4d, r4d
+    inc         r4d
     cmp         r3m, byte 28
-    mov         r3, 9
-    cmove       r3, r4
+    mov         r3d, 9
+    cmove       r3d, r4d
 
     movh        m0, [r2 + r3]    ; [8 7 6 5 4 3 2 1]
-    mova        m1, m0
-    psrldq      m1, 1           ; [x 8 7 6 5 4 3 2]
-    punpcklbw   m0, m1          ; [x 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1]
+    punpcklbw   m0, m0
+    psrldq      m0, 1
     punpcklqdq  m0, m0
     mova        m2, m0
 
@@ -1462,16 +1768,15 @@
     jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
 
 cglobal intra_pred_ang4_9, 3,5,8
-    xor         r4, r4
-    inc         r4
+    xor         r4d, r4d
+    inc         r4d
     cmp         r3m, byte 27
-    mov         r3, 9
-    cmove       r3, r4
+    mov         r3d, 9
+    cmove       r3d, r4d
 
     movh        m0, [r2 + r3]    ; [8 7 6 5 4 3 2 1]
-    mova        m1, m0
-    psrldq      m1, 1           ; [x 8 7 6 5 4 3 2]
-    punpcklbw   m0, m1          ; [x 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1]
+    punpcklbw   m0, m0
+    psrldq      m0, 1           ; [x 8 7 6 5 4 3 2]
     punpcklqdq  m0, m0
     mova        m2, m0
 
@@ -1482,6 +1787,292 @@
     mova        m7, [r3 +  4 * 16]  ; [ 8]
     jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
 
+cglobal intra_pred_ang4_10, 3,5,4
+    movd        m0, [r2 + 9]            ; [8 7 6 5 4 3 2 1]
+    punpcklbw   m0, m0
+    punpcklwd   m0, m0
+    pshufd      m1, m0, 1
+    movhlps     m2, m0
+    pshufd      m3, m0, 3
+    movd        [r0 + r1], m1
+    movd        [r0 + r1 * 2], m2
+    lea         r1, [r1 * 3]
+    movd        [r0 + r1], m3
+    cmp         r4m, byte 0
+    jz          .quit
+
+    ; filter
+    pxor        m3, m3
+    punpcklbw   m0, m3
+    movh        m1, [r2]                ; [4 3 2 1 0]
+    punpcklbw   m1, m3
+    pshuflw     m2, m1, 0x00
+    psrldq      m1, 2
+    psubw       m1, m2
+    psraw       m1, 1
+    paddw       m0, m1
+    packuswb    m0, m0
+
+.quit:
+    movd        [r0], m0
+    RET
+
+cglobal intra_pred_ang4_26, 3,4,4
+    movd        m0, [r2 + 1]            ; [8 7 6 5 4 3 2 1]
+
+    ; store
+    movd        [r0], m0
+    movd        [r0 + r1], m0
+    movd        [r0 + r1 * 2], m0
+    lea         r3, [r1 * 3]
+    movd        [r0 + r3], m0
+
+    ; filter
+    cmp         r4m, byte 0
+    jz         .quit
+
+    pxor        m3, m3
+    punpcklbw   m0, m3
+    pshuflw     m0, m0, 0x00
+    movd        m2, [r2]
+    punpcklbw   m2, m3
+    pshuflw     m2, m2, 0x00
+    movd        m1, [r2 + 9]
+    punpcklbw   m1, m3
+    psubw       m1, m2
+    psraw       m1, 1
+    paddw       m0, m1
+    packuswb    m0, m0
+
+    movd        r2, m0
+    mov         [r0], r2b
+    shr         r2, 8
+    mov         [r0 + r1], r2b
+    shr         r2, 8
+    mov         [r0 + r1 * 2], r2b
+    shr         r2, 8
+    mov         [r0 + r3], r2b
+
+.quit:
+    RET
+
+cglobal intra_pred_ang4_11, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 25
+    mov         r3d, 8
+    cmove       r3d, r4d
+
+    movd        m1, [r2 + r3 + 1]       ;[4 3 2 1]
+    movh        m0, [r2 - 7]            ;[A x x x x x x x]
+    punpcklbw   m1, m1                  ;[4 4 3 3 2 2 1 1]
+    punpcklqdq  m0, m1                  ;[4 4 3 3 2 2 1 1 A x x x x x x x]]
+    psrldq      m0, 7                   ;[x x x x x x x x 4 3 3 2 2 1 1 A]
+    punpcklqdq  m0, m0
+    mova        m2, m0
+
+    lea         r3, [pw_ang_table + 24 * 16]
+
+    mova        m4, [r3 +  6 * 16]  ; [24]
+    mova        m5, [r3 +  4 * 16]  ; [26]
+    mova        m6, [r3 +  2 * 16]  ; [28]
+    mova        m7, [r3 +  0 * 16]  ; [30]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_12, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 24
+    mov         r3d, 8
+    cmove       r3d, r4d
+
+    movd        m1, [r2 + r3 + 1]
+    movh        m0, [r2 - 7]
+    punpcklbw   m1, m1
+    punpcklqdq  m0, m1
+    psrldq      m0, 7
+    punpcklqdq  m0, m0
+    mova        m2, m0
+
+    lea         r3, [pw_ang_table + 20 * 16]
+    mova        m4, [r3 +  7 * 16]  ; [27]
+    mova        m5, [r3 +  2 * 16]  ; [22]
+    mova        m6, [r3 -  3 * 16]  ; [17]
+    mova        m7, [r3 -  8 * 16]  ; [12]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_13, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 23
+    mov         r3d, 8
+    jz          .next
+    xchg        r3d, r4d
+
+.next:
+    movd        m1, [r2 - 1]            ;[x x A x]
+    movd        m2, [r2 + r4 + 1]       ;[4 3 2 1]
+    movd        m0, [r2 + r3 + 3]       ;[x x B x]
+    punpcklbw   m0, m1                  ;[x x x x A B x x]
+    punpckldq   m0, m2                  ;[4 3 2 1 A B x x]
+    psrldq      m0, 2                   ;[x x 4 3 2 1 A B]
+    punpcklbw   m0, m0                  ;[x x x x 4 4 3 3 2 2 1 1 A A B B]
+    mova        m1, m0
+    psrldq      m0, 3                   ;[x x x x x x x 4 4 3 3 2 2 1 1 A]
+    psrldq      m1, 1                   ;[x x x x x 4 4 3 3 2 2 1 1 A A B]
+    movh        m2, m0
+    punpcklqdq  m0, m0
+    punpcklqdq  m2, m1
+
+    lea         r3, [pw_ang_table + 21 * 16]
+    mova        m4, [r3 +  2 * 16]  ; [23]
+    mova        m5, [r3 -  7 * 16]  ; [14]
+    mova        m6, [r3 - 16 * 16]  ; [ 5]
+    mova        m7, [r3 +  7 * 16]  ; [28]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_14, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 22
+    mov         r3d, 8
+    jz          .next
+    xchg        r3d, r4d
+
+.next:
+    movd        m1, [r2 - 1]            ;[x x A x]
+    movd        m0, [r2 + r3 + 1]       ;[x x B x]
+    punpcklbw   m0, m1                  ;[A B x x]
+    movd        m1, [r2 + r4 + 1]       ;[4 3 2 1]
+    punpckldq   m0, m1                  ;[4 3 2 1 A B x x]
+    psrldq      m0, 2                   ;[x x 4 3 2 1 A B]
+    punpcklbw   m0, m0                  ;[x x x x 4 4 3 3 2 2 1 1 A A B B]
+    mova        m2, m0
+    psrldq      m0, 3                   ;[x x x x x x x 4 4 3 3 2 2 1 1 A]
+    psrldq      m2, 1                   ;[x x x x x 4 4 3 3 2 2 1 1 A A B]
+    punpcklqdq  m0, m0
+    punpcklqdq  m2, m2
+
+    lea         r3, [pw_ang_table + 19 * 16]
+    mova        m4, [r3 +  0 * 16]  ; [19]
+    mova        m5, [r3 - 13 * 16]  ; [ 6]
+    mova        m6, [r3 +  6 * 16]  ; [25]
+    mova        m7, [r3 -  7 * 16]  ; [12]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_15, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 21
+    mov         r3d, 8
+    jz          .next
+    xchg        r3d, r4d
+
+.next:
+    movd        m0, [r2]                ;[x x x A]
+    movd        m1, [r2 + r3 + 2]       ;[x x x B]
+    punpcklbw   m1, m0                  ;[x x A B]
+    movd        m0, [r2 + r3 + 3]       ;[x x C x]
+    punpcklwd   m0, m1                  ;[A B C x]
+    movd        m1, [r2 + r4 + 1]       ;[4 3 2 1]
+    punpckldq   m0, m1                  ;[4 3 2 1 A B C x]
+    psrldq      m0, 1                   ;[x 4 3 2 1 A B C]
+    punpcklbw   m0, m0                  ;[x x 4 4 3 3 2 2 1 1 A A B B C C]
+    psrldq      m0, 1
+    movh        m1, m0
+    psrldq      m0, 2
+    movh        m2, m0
+    psrldq      m0, 2
+    punpcklqdq  m0, m2
+    punpcklqdq  m2, m1
+
+    lea         r3, [pw_ang_table + 23 * 16]
+    mova        m4, [r3 -  8 * 16]  ; [15]
+    mova        m5, [r3 +  7 * 16]  ; [30]
+    mova        m6, [r3 - 10 * 16]  ; [13]
+    mova        m7, [r3 +  5 * 16]  ; [28]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_16, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 20
+    mov         r3d, 8
+    jz          .next
+    xchg        r3d, r4d
+
+.next:
+    movd        m2, [r2]                ;[x x x A]
+    movd        m1, [r2 + r3 + 2]       ;[x x x B]
+    punpcklbw   m1, m2                  ;[x x A B]
+    movh        m0, [r2 + r3 + 2]       ;[x x C x]
+    punpcklwd   m0, m1                  ;[A B C x]
+    movd        m1, [r2 + r4 + 1]       ;[4 3 2 1]
+    punpckldq   m0, m1                  ;[4 3 2 1 A B C x]
+    psrldq      m0, 1                   ;[x 4 3 2 1 A B C]
+    punpcklbw   m0, m0                  ;[x x 4 4 3 3 2 2 1 1 A A B B C C]
+    psrldq      m0, 1
+    movh        m1, m0
+    psrldq      m0, 2
+    movh        m2, m0
+    psrldq      m0, 2
+    punpcklqdq  m0, m2
+    punpcklqdq  m2, m1
+
+    lea         r3, [pw_ang_table + 19 * 16]
+    mova        m4, [r3 -  8 * 16]  ; [11]
+    mova        m5, [r3 +  3 * 16]  ; [22]
+    mova        m6, [r3 - 18 * 16]  ; [ 1]
+    mova        m7, [r3 -  7 * 16]  ; [12]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_17, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 19
+    mov         r3d, 8
+    jz          .next
+    xchg        r3d, r4d
+
+.next:
+    movd        m2, [r2]                ;[x x x A]
+    movd        m3, [r2 + r3 + 1]       ;[x x x B]
+    movd        m4, [r2 + r3 + 2]       ;[x x x C]
+    movd        m0, [r2 + r3 + 4]       ;[x x x D]
+    punpcklbw   m3, m2                  ;[x x A B]
+    punpcklbw   m0, m4                  ;[x x C D]
+    punpcklwd   m0, m3                  ;[A B C D]
+    movd        m1, [r2 + r4 + 1]       ;[4 3 2 1]
+    punpckldq   m0, m1                  ;[4 3 2 1 A B C D]
+    punpcklbw   m0, m0                  ;[4 4 3 3 2 2 1 1 A A B B C C D D]
+    psrldq      m0, 1
+    movh        m1, m0
+    psrldq      m0, 2
+    movh        m2, m0
+    punpcklqdq  m2, m1
+    psrldq      m0, 2
+    movh        m1, m0
+    psrldq      m0, 2
+    punpcklqdq  m0, m1
+
+    lea         r3, [pw_ang_table + 14 * 16]
+    mova        m4, [r3 -  8 * 16]  ; [ 6]
+    mova        m5, [r3 -  2 * 16]  ; [12]
+    mova        m6, [r3 +  4 * 16]  ; [18]
+    mova        m7, [r3 + 10 * 16]  ; [24]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
+cglobal intra_pred_ang4_18, 3,4,2
+    mov         r3d, [r2 + 8]
+    mov         r3b, byte [r2]
+    bswap       r3d
+    movd        m0, r3d
+
+    movd        m1, [r2 + 1]
+    punpckldq   m0, m1
+    lea         r3, [r1 * 3]
+    movd        [r0 + r3], m0
+    psrldq      m0, 1
+    movd        [r0 + r1 * 2], m0
+    psrldq      m0, 1
+    movd        [r0 + r1], m0
+    psrldq      m0, 1
+    movd        [r0], m0
+    RET
+
 ;---------------------------------------------------------------------------------------------
 ; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int dirMode, int bFilter)
 ;---------------------------------------------------------------------------------------------
@@ -1809,6 +2400,69 @@
 
     RET
 
+;---------------------------------------------------------------------------------------------
+; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int dirMode, int bFilter)
+;---------------------------------------------------------------------------------------------
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal intra_pred_dc32, 3, 4, 3
+    lea             r3, [r1 * 3]
+    pxor            m0, m0
+    movu            m1, [r2 + 1]
+    movu            m2, [r2 + 65]
+    psadbw          m1, m0
+    psadbw          m2, m0
+    paddw           m1, m2
+    vextracti128    xm2, m1, 1
+    paddw           m1, m2
+    pshufd          m2, m1, 2
+    paddw           m1, m2
+
+    pmulhrsw        m1, [pw_512]    ; sum = (sum + 32) / 64
+    vpbroadcastb    m1, xm1         ; m1 = byte [dc_val ...]
+
+    movu            [r0 + r1 * 0], m1
+    movu            [r0 + r1 * 1], m1
+    movu            [r0 + r1 * 2], m1
+    movu            [r0 + r3 * 1], m1
+    lea             r0, [r0 + 4 * r1]
+    movu            [r0 + r1 * 0], m1
+    movu            [r0 + r1 * 1], m1
+    movu            [r0 + r1 * 2], m1
+    movu            [r0 + r3 * 1], m1
+    lea             r0, [r0 + 4 * r1]
+    movu            [r0 + r1 * 0], m1
+    movu            [r0 + r1 * 1], m1
+    movu            [r0 + r1 * 2], m1
+    movu            [r0 + r3 * 1], m1
+    lea             r0, [r0 + 4 * r1]
+    movu            [r0 + r1 * 0], m1
+    movu            [r0 + r1 * 1], m1
+    movu            [r0 + r1 * 2], m1
+    movu            [r0 + r3 * 1], m1
+    lea             r0, [r0 + 4 * r1]
+    movu            [r0 + r1 * 0], m1
+    movu            [r0 + r1 * 1], m1
+    movu            [r0 + r1 * 2], m1
+    movu            [r0 + r3 * 1], m1
+    lea             r0, [r0 + 4 * r1]
+    movu            [r0 + r1 * 0], m1
+    movu            [r0 + r1 * 1], m1
+    movu            [r0 + r1 * 2], m1
+    movu            [r0 + r3 * 1], m1
+    lea             r0, [r0 + 4 * r1]
+    movu            [r0 + r1 * 0], m1
+    movu            [r0 + r1 * 1], m1
+    movu            [r0 + r1 * 2], m1
+    movu            [r0 + r3 * 1], m1
+    lea             r0, [r0 + 4 * r1]
+    movu            [r0 + r1 * 0], m1
+    movu            [r0 + r1 * 1], m1
+    movu            [r0 + r1 * 2], m1
+    movu            [r0 + r3 * 1], m1
+    RET
+%endif ;; ARCH_X86_64 == 1
+
 ;---------------------------------------------------------------------------------------
 ; void intra_pred_planar(pixel* dst, intptr_t dstStride, pixel*srcPix, int, int filter)
 ;---------------------------------------------------------------------------------------
@@ -2000,6 +2654,57 @@
 ;---------------------------------------------------------------------------------------
 ; void intra_pred_planar(pixel* dst, intptr_t dstStride, pixel*srcPix, int, int filter)
 ;---------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal intra_pred_planar16, 3,3,6
+    vpbroadcastw    m3, [r2 + 17]
+    mova            m5, [pw_00ff]
+    vpbroadcastw    m4, [r2 + 49]
+    mova            m0, [pw_planar16_mul]
+    pmovzxbw        m2, [r2 + 1]
+    pand            m3, m5                      ; v_topRight
+    pand            m4, m5                      ; v_bottomLeft
+
+    pmullw          m3, [multiL]                ; (x + 1) * topRight
+    pmullw          m1, m2, [pw_15]             ; (blkSize - 1 - y) * above[x]
+    paddw           m3, [pw_16]
+    paddw           m3, m4
+    paddw           m3, m1
+    psubw           m4, m2
+    add             r2, 33
+
+%macro INTRA_PRED_PLANAR16_AVX2 1
+    vpbroadcastw    m1, [r2 + %1]
+    vpsrlw          m2, m1, 8
+    pand            m1, m5
+
+    pmullw          m1, m0
+    pmullw          m2, m0
+    paddw           m1, m3
+    paddw           m3, m4
+    psraw           m1, 5
+    paddw           m2, m3
+    psraw           m2, 5
+    paddw           m3, m4
+    packuswb        m1, m2
+    vpermq          m1, m1, 11011000b
+    movu            [r0], xm1
+    vextracti128    [r0 + r1], m1, 1
+    lea             r0, [r0 + r1 * 2]
+%endmacro
+    INTRA_PRED_PLANAR16_AVX2 0
+    INTRA_PRED_PLANAR16_AVX2 2
+    INTRA_PRED_PLANAR16_AVX2 4
+    INTRA_PRED_PLANAR16_AVX2 6
+    INTRA_PRED_PLANAR16_AVX2 8
+    INTRA_PRED_PLANAR16_AVX2 10
+    INTRA_PRED_PLANAR16_AVX2 12
+    INTRA_PRED_PLANAR16_AVX2 14
+%undef INTRA_PRED_PLANAR16_AVX2
+    RET
+
+;---------------------------------------------------------------------------------------
+; void intra_pred_planar(pixel* dst, intptr_t dstStride, pixel*srcPix, int, int filter)
+;---------------------------------------------------------------------------------------
 INIT_XMM sse4
 %if ARCH_X86_64 == 1
 cglobal intra_pred_planar32, 3,4,12
@@ -2104,6 +2809,91 @@
     jnz             .loop
     RET
 
+;---------------------------------------------------------------------------------------
+; void intra_pred_planar(pixel* dst, intptr_t dstStride, pixel*srcPix, int, int filter)
+;---------------------------------------------------------------------------------------
+%if ARCH_X86_64 == 1
+INIT_YMM avx2
+cglobal intra_pred_planar32, 3,4,11
+    mova            m6, [pw_00ff]
+    vpbroadcastw    m3, [r2 + 33]               ; topRight   = above[32]
+    vpbroadcastw    m2, [r2 + 97]               ; bottomLeft = left[32]
+    pand            m3, m6
+    pand            m2, m6
+
+    pmullw          m0, m3, [multiL]            ; (x + 1) * topRight
+    pmullw          m3, [multiH2]               ; (x + 1) * topRight
+
+    paddw           m0, m2
+    paddw           m3, m2
+    paddw           m0, [pw_32]
+    paddw           m3, [pw_32]
+
+    pmovzxbw        m4, [r2 + 1]
+    pmovzxbw        m1, [r2 + 17]
+    pmullw          m5, m4, [pw_31]
+    paddw           m0, m5
+    psubw           m5, m2, m4
+    psubw           m2, m1
+    pmullw          m1, [pw_31]
+    paddw           m3, m1
+    mova            m1, m5
+
+    add             r2, 65                      ; (2 * blkSize + 1)
+    mova            m9, [pw_planar32_mul]
+    mova            m10, [pw_planar16_mul]
+
+%macro INTRA_PRED_PLANAR32_AVX2 0
+    vpbroadcastw    m4, [r2]
+    vpsrlw          m7, m4, 8
+    pand            m4, m6
+
+    pmullw          m5, m4, m9
+    pmullw          m4, m4, m10
+    paddw           m5, m0
+    paddw           m4, m3
+    paddw           m0, m1
+    paddw           m3, m2
+    psraw           m5, 6
+    psraw           m4, 6
+    packuswb        m5, m4
+    pmullw          m8, m7, m9
+    pmullw          m7, m7, m10
+    vpermq          m5, m5, 11011000b
+    paddw           m8, m0
+    paddw           m7, m3
+    paddw           m0, m1
+    paddw           m3, m2
+    psraw           m8, 6
+    psraw           m7, 6
+    packuswb        m8, m7
+    add             r2, 2
+    vpermq          m8, m8, 11011000b
+
+    movu            [r0], m5
+    movu            [r0 + r1], m8
+    lea             r0, [r0 + r1 * 2]
+%endmacro
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+    INTRA_PRED_PLANAR32_AVX2
+%undef INTRA_PRED_PLANAR32_AVX2
+    RET
+%endif ;; ARCH_X86_64 == 1
+
 ;-----------------------------------------------------------------------------------------
 ; void intraPredAng4(pixel* dst, intptr_t dstStride, pixel* src, int dirMode, int bFilter)
 ;-----------------------------------------------------------------------------------------
@@ -9577,6 +10367,99 @@
 
     RET
 
+INIT_YMM avx2
+cglobal intra_pred_ang32_18, 4, 4, 3
+    movu           m0, [r2]
+    movu           xm1, [r2 + 1 + 64]
+    pshufb         xm1, [intra_pred_shuff_15_0]
+    mova           xm2, xm0
+    vinserti128    m1, m1, xm2, 1
+
+    lea            r3, [r1 * 3]
+
+    movu           [r0], m0
+    palignr        m2, m0, m1, 15
+    movu           [r0 + r1], m2
+    palignr        m2, m0, m1, 14
+    movu           [r0 + r1 * 2], m2
+    palignr        m2, m0, m1, 13
+    movu           [r0 + r3], m2
+
+    lea            r0, [r0 + r1 * 4]
+    palignr        m2, m0, m1, 12
+    movu           [r0], m2
+    palignr        m2, m0, m1, 11
+    movu           [r0 + r1], m2
+    palignr        m2, m0, m1, 10
+    movu           [r0 + r1 * 2], m2
+    palignr        m2, m0, m1, 9
+    movu           [r0 + r3], m2
+
+    lea            r0, [r0 + r1 * 4]
+    palignr        m2, m0, m1, 8
+    movu           [r0], m2
+    palignr        m2, m0, m1, 7
+    movu           [r0 + r1], m2
+    palignr        m2, m0, m1, 6
+    movu           [r0 + r1 * 2], m2
+    palignr        m2, m0, m1, 5
+    movu           [r0 + r3], m2
+
+    lea            r0, [r0 + r1 * 4]
+    palignr        m2, m0, m1, 4
+    movu           [r0], m2
+    palignr        m2, m0, m1, 3
+    movu           [r0 + r1], m2
+    palignr        m2, m0, m1, 2
+    movu           [r0 + r1 * 2], m2
+    palignr        m2, m0, m1, 1
+    movu           [r0 + r3], m2
+
+    lea            r0, [r0 + r1 * 4]
+    movu           [r0], m1
+
+    movu           xm0, [r2 + 64 + 17]
+    pshufb         xm0, [intra_pred_shuff_15_0]
+    vinserti128    m0, m0, xm1, 1
+
+    palignr        m2, m1, m0, 15
+    movu           [r0 + r1], m2
+    palignr        m2, m1, m0, 14
+    movu           [r0 + r1 * 2], m2
+    palignr        m2, m1, m0, 13
+    movu           [r0 + r3], m2
+
+    lea            r0, [r0 + r1 * 4]
+    palignr        m2, m1, m0, 12
+    movu           [r0], m2
+    palignr        m2, m1, m0, 11
+    movu           [r0 + r1], m2
+    palignr        m2, m1, m0, 10
+    movu           [r0 + r1 * 2], m2
+    palignr        m2, m1, m0, 9
+    movu           [r0 + r3], m2
+
+    lea            r0, [r0 + r1 * 4]
+    palignr        m2, m1, m0, 8
+    movu           [r0], m2
+    palignr        m2, m1, m0, 7
+    movu           [r0 + r1], m2
+    palignr        m2, m1, m0,6
+    movu           [r0 + r1 * 2], m2
+    palignr        m2, m1, m0, 5
+    movu           [r0 + r3], m2
+
+    lea            r0, [r0 + r1 * 4]
+    palignr        m2, m1, m0, 4
+    movu           [r0], m2
+    palignr        m2, m1, m0, 3
+    movu           [r0 + r1], m2
+    palignr        m2, m1, m0,2
+    movu           [r0 + r1 * 2], m2
+    palignr        m2, m1, m0, 1
+    movu           [r0 + r3], m2
+    RET
+
 INIT_XMM sse4
 cglobal intra_pred_ang32_18, 4,5,5
     movu        m0, [r2]               ; [15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0]
@@ -11099,6 +11982,441 @@
     movhps            [r0 + r3], xm2
     RET
 
+INIT_YMM avx2
+cglobal intra_pred_ang8_15, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2 + 16]
+    pinsrb            xm5, [r2], 0
+    lea               r5, [intra_pred_shuff_0_8]
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 2], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_15]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 4], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 6], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 8], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    vperm2i128        m2, m1, m4, 00100000b
+    vperm2i128        m1, m1, m4, 00110001b
+    punpcklbw         m4, m2, m1
+    punpckhbw         m2, m1
+    punpcklwd         m1, m4, m2
+    punpckhwd         m4, m2
+    mova              m0, [trans8_shuf]
+    vpermd            m1, m0, m1
+    vpermd            m4, m0, m4
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    movhps            [r0 + r1], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    movhps            [r0 + r1], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_16, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2 + 16]
+    pinsrb            xm5, [r2], 0
+    lea               r5, [intra_pred_shuff_0_8]
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 2], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_20]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 3], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 5], 0
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 6], 0
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 8], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    vperm2i128        m2, m1, m4, 00100000b
+    vperm2i128        m1, m1, m4, 00110001b
+    punpcklbw         m4, m2, m1
+    punpckhbw         m2, m1
+    punpcklwd         m1, m4, m2
+    punpckhwd         m4, m2
+    mova              m0, [trans8_shuf]
+    vpermd            m1, m0, m1
+    vpermd            m4, m0, m4
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    movhps            [r0 + r1], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    movhps            [r0 + r1], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_20, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2]
+    lea               r5, [intra_pred_shuff_0_8]
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 2 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_20]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 3 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 5 + 16], 0
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 6 + 16], 0
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 8 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm1
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm4
+    movhps            [r0 + r3], xm2
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_21, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2]
+    lea               r5, [intra_pred_shuff_0_8]
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 2 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_15]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 4 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 6 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 8 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm1
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm4
+    movhps            [r0 + r3], xm2
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_22, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2]
+    lea               r5, [intra_pred_shuff_0_8]
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_14]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 2 + 16], 0
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 5 + 16], 0
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 7 + 16], 0
+    pshufb            xm5, [r5]
+    vinserti128       m0, m0, xm5, 1
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm1
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm4
+    movhps            [r0 + r3], xm2
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_14, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2 + 16]
+    pinsrb            xm5, [r2], 0
+    lea               r5, [intra_pred_shuff_0_8]
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_14]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 2], 0
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 5], 0
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 7], 0
+    pshufb            xm5, [r5]
+    vinserti128       m0, m0, xm5, 1
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    vperm2i128        m2, m1, m4, 00100000b
+    vperm2i128        m1, m1, m4, 00110001b
+    punpcklbw         m4, m2, m1
+    punpckhbw         m2, m1
+    punpcklwd         m1, m4, m2
+    punpckhwd         m4, m2
+    mova              m0, [trans8_shuf]
+    vpermd            m1, m0, m1
+    vpermd            m4, m0, m4
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    movhps            [r0 + r1], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    movhps            [r0 + r1], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_13, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2 + 16]
+    pinsrb            xm5, [r2], 0
+    lea               r5, [intra_pred_shuff_0_8]
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_13]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 4], 0
+    pshufb            xm4, xm5, [r5]
+    vinserti128       m0, m0, xm4, 1
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    vinserti128       m0, m0, xm4, 0
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 7], 0
+    pshufb            xm5, [r5]
+    vinserti128       m0, m0, xm5, 1
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    vperm2i128        m2, m1, m4, 00100000b
+    vperm2i128        m1, m1, m4, 00110001b
+    punpcklbw         m4, m2, m1
+    punpckhbw         m2, m1
+    punpcklwd         m1, m4, m2
+    punpckhwd         m4, m2
+    mova              m0, [trans8_shuf]
+    vpermd            m1, m0, m1
+    vpermd            m4, m0, m4
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    movhps            [r0 + r1], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    movhps            [r0 + r1], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    RET
+
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_23, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2]
+    lea               r5, [intra_pred_shuff_0_8]
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_13]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 4 + 16], 0
+    pshufb            xm4, xm5, [r5]
+    vinserti128       m0, m0, xm4, 1
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    vinserti128       m0, m0, xm4, 0
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 7 + 16], 0
+    pshufb            xm5, [r5]
+    vinserti128       m0, m0, xm5, 1
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm1
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm4
+    movhps            [r0 + r3], xm2
+    RET
 
 INIT_YMM avx2
 cglobal intra_pred_ang8_12, 3, 5, 5
@@ -11228,6 +12546,849 @@
     movu              [%2], xm3
 %endmacro
 
+%if ARCH_X86_64 == 1
+%macro INTRA_PRED_TRANS_STORE_16x16 0
+    punpcklbw    m8, m0, m1
+    punpckhbw    m0, m1
+
+    punpcklbw    m1, m2, m3
+    punpckhbw    m2, m3
+
+    punpcklbw    m3, m4, m5
+    punpckhbw    m4, m5
+
+    punpcklbw    m5, m6, m7
+    punpckhbw    m6, m7
+
+    punpcklwd    m7, m8, m1
+    punpckhwd    m8, m1
+
+    punpcklwd    m1, m3, m5
+    punpckhwd    m3, m5
+
+    punpcklwd    m5, m0, m2
+    punpckhwd    m0, m2
+
+    punpcklwd    m2, m4, m6
+    punpckhwd    m4, m6
+
+    punpckldq    m6, m7, m1
+    punpckhdq    m7, m1
+
+    punpckldq    m1, m8, m3
+    punpckhdq    m8, m3
+
+    punpckldq    m3, m5, m2
+    punpckhdq    m5, m2
+
+    punpckldq    m2, m0, m4
+    punpckhdq    m0, m4
+
+    vpermq       m6, m6, 0xD8
+    vpermq       m7, m7, 0xD8
+    vpermq       m1, m1, 0xD8
+    vpermq       m8, m8, 0xD8
+    vpermq       m3, m3, 0xD8
+    vpermq       m5, m5, 0xD8
+    vpermq       m2, m2, 0xD8
+    vpermq       m0, m0, 0xD8
+
+    movu            [r0], xm6
+    vextracti128    xm4, m6, 1
+    movu            [r0 + r1], xm4
+
+    movu            [r0 + 2 * r1], xm7
+    vextracti128    xm4, m7, 1
+    movu            [r0 + r3], xm4
+
+    lea             r0, [r0 + 4 * r1]
+
+    movu            [r0], xm1
+    vextracti128    xm4, m1, 1
+    movu            [r0 + r1], xm4
+
+    movu            [r0 + 2 * r1], xm8
+    vextracti128    xm4, m8, 1
+    movu            [r0 + r3], xm4
+
+    lea             r0, [r0 + 4 * r1]
+
+    movu            [r0], xm3
+    vextracti128    xm4, m3, 1
+    movu            [r0 + r1], xm4
+
+    movu            [r0 + 2 * r1], xm5
+    vextracti128    xm4, m5, 1
+    movu            [r0 + r3], xm4
+
+    lea             r0, [r0 + 4 * r1]
+
+    movu            [r0], xm2
+    vextracti128    xm4, m2, 1
+    movu            [r0 + r1], xm4
+
+    movu            [r0 + 2 * r1], xm0
+    vextracti128    xm4, m0, 1
+    movu            [r0 + r3], xm4
+%endmacro
+
+%macro INTRA_PRED_ANG16_CAL_ROW 3
+    pmaddubsw         %1, m9, [r4 + (%3 * mmsize)]
+    pmulhrsw          %1, m11
+    pmaddubsw         %2, m10, [r4 + (%3 * mmsize)]
+    pmulhrsw          %2, m11
+    packuswb          %1, %2
+%endmacro
+
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_12, 3, 6, 13
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    movu              xm9, [r2 + 32]
+    pinsrb            xm9, [r2], 0
+    pslldq            xm7, xm9, 1
+    pinsrb            xm7, [r2 + 6], 0
+    vinserti128       m9, m9, xm7, 1
+    pshufb            m9, [r5]
+
+    movu              xm12, [r2 + 6 + 32]
+
+    psrldq            xm10, xm12, 2
+    psrldq            xm8, xm12, 1
+    vinserti128       m10, m10, xm8, 1
+    pshufb            m10, [r5]
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_12]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    pslldq            xm7, 1
+    pinsrb            xm7, [r2 + 13], 0
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    mova              xm8, xm12
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+
+    movu              xm9, [r2 + 31]
+    pinsrb            xm9, [r2 + 6], 0
+    pinsrb            xm9, [r2 + 0], 1
+    pshufb            xm9, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    psrldq            xm10, xm12, 1
+    vinserti128       m10, m10, xm12, 1
+    pshufb            m10, [r5]
+
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_13, 3, 6, 14
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    movu              xm13, [r2 + 32]
+    pinsrb            xm13, [r2], 0
+    pslldq            xm7, xm13, 2
+    pinsrb            xm7, [r2 + 7], 0
+    pinsrb            xm7, [r2 + 4], 1
+    vinserti128       m9, m13, xm7, 1
+    pshufb            m9, [r5]
+
+    movu              xm12, [r2 + 4 + 32]
+
+    psrldq            xm10, xm12, 4
+    psrldq            xm8, xm12, 2
+    vinserti128       m10, m10, xm8, 1
+    pshufb            m10, [r5]
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_13]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+
+    pslldq            xm7, 1
+    pinsrb            xm7, [r2 + 11], 0
+    pshufb            xm2, xm7, [r5]
+    vinserti128       m9, m9, xm2, 1
+
+    psrldq            xm8, xm12, 1
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+
+    pslldq            xm13, 1
+    pinsrb            xm13, [r2 + 4], 0
+    pshufb            xm3, xm13, [r5]
+    vinserti128       m9, m9, xm3, 0
+
+    psrldq            xm8, xm12, 3
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+
+    pslldq            xm7, 1
+    pinsrb            xm7, [r2 + 14], 0
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    mova              xm8, xm12
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+
+    pslldq            xm13, 1
+    pinsrb            xm13, [r2 + 7], 0
+    pshufb            xm13, [r5]
+    vinserti128       m9, m9, xm13, 0
+
+    psrldq            xm12, 2
+    pshufb            xm12, [r5]
+    vinserti128       m10, m10, xm12, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_11, 3, 5, 12
+    mova              m11, [pw_1024]
+
+    movu              xm9, [r2 + 32]
+    pinsrb            xm9, [r2], 0
+    pshufb            xm9, [intra_pred_shuff_0_8]
+    vinserti128       m9, m9, xm9, 1
+
+    vbroadcasti128    m10, [r2 + 8 + 32]
+    pshufb            m10, [intra_pred_shuff_0_8]
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_11]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_3, 3, 6, 12
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    movu              xm9, [r2 + 1 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 9 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 8 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 16 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_3]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+
+    movu              xm9, [r2 + 2 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 10 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 9 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 17 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+
+    movu              xm7, [r2 + 3 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 0
+
+    movu              xm8, [r2 + 11 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+
+    movu              xm9, [r2 + 4 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 12 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 10 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 18 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    movu              xm9, [r2 + 5 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 13 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 11 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 19 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    add               r4, 4 * mmsize
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+
+    movu              xm7, [r2 + 12 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 20 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+
+    movu              xm9, [r2 + 6 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 14 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 13 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 21 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+
+    movu              xm9, [r2 + 7 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 15 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 14 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 22 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_4, 3, 6, 12
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    movu              xm9, [r2 + 1 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 9 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 6 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 14 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_4]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+
+    movu              xm9, [r2 + 2 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 10 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 7 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 15 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+
+    movu              xm7, [r2 + 8 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 16 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+
+    movu              xm7, [r2 + 3 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 0
+
+    movu              xm8, [r2 + 11 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    movu              xm9, [r2 + 4 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 12 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 9 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 17 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+
+    movu              xm7, [r2 + 10 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 18 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+
+    movu              xm7, [r2 + 5 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 0
+
+    movu              xm8, [r2 + 13 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+
+    movu              xm9, [r2 + 6 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 14 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 11 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 19 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_5, 3, 6, 12
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    movu              xm9, [r2 + 1 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 9 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 5 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 13 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_5]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+
+    movu              xm9, [r2 + 2 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 10 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 6 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 14 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+
+    movu              xm9, [r2 + 3 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 11 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 7 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 15 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+
+    movu              xm9, [r2 + 4 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 12 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 8 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 16 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+
+    movu              xm9, [r2 + 5 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 13 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 9 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 17 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_6, 3, 6, 12
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    movu              xm9, [r2 + 1 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 9 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 4 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 12 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_6]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+
+    movu              xm7, [r2 + 5 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 13 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+
+    movu              xm7, [r2 + 2 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 0
+
+    movu              xm8, [r2 + 10 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    movu              xm9, [r2 + 3 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 11 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 6 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 14 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+
+    movu              xm7, [r2 + 7 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 15 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+
+    movu              xm7, [r2 + 4 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 0
+
+    movu              xm8, [r2 + 12 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_7, 3, 6, 12
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    movu              xm9, [r2 + 1 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 9 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 3  + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 11 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_7]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+
+    movu              xm7, [r2 + 4  + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 12 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+
+    movu              xm7, [r2 + 2  + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 0
+
+    movu              xm8, [r2 + 10 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+
+    movu              xm7, [r2 + 5  + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 13 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+
+    movu              xm7, [r2 + 3  + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 0
+
+    movu              xm8, [r2 + 11 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_8, 3, 6, 12
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    movu              xm9, [r2 + 1 + 32]
+    pshufb            xm9, [r5]
+    movu              xm10, [r2 + 9 + 32]
+    pshufb            xm10, [r5]
+
+    movu              xm7, [r2 + 2  + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm8, [r2 + 10 + 32]
+    pshufb            xm8, [r5]
+    vinserti128       m10, m10, xm8, 1
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_8]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    movu              xm4, [r2 + 3  + 32]
+    pshufb            xm4, [r5]
+    vinserti128       m9, m9, xm4, 1
+
+    movu              xm5, [r2 + 11 + 32]
+    pshufb            xm5, [r5]
+    vinserti128       m10, m10, xm5, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+
+    vinserti128       m9, m9, xm7, 0
+    vinserti128       m10, m10, xm8, 0
+
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_9, 3, 6, 12
+    mova              m11, [pw_1024]
+    lea               r5, [intra_pred_shuff_0_8]
+
+    vbroadcasti128    m9, [r2 + 1 + 32]
+    pshufb            m9, [r5]
+    vbroadcasti128    m10, [r2 + 9 + 32]
+    pshufb            m10, [r5]
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_9]
+
+    INTRA_PRED_ANG16_CAL_ROW m0, m1, 0
+    INTRA_PRED_ANG16_CAL_ROW m1, m2, 1
+    INTRA_PRED_ANG16_CAL_ROW m2, m3, 2
+    INTRA_PRED_ANG16_CAL_ROW m3, m4, 3
+
+    add               r4, 4 * mmsize
+
+    INTRA_PRED_ANG16_CAL_ROW m4, m5, 0
+    INTRA_PRED_ANG16_CAL_ROW m5, m6, 1
+    INTRA_PRED_ANG16_CAL_ROW m6, m7, 2
+
+    movu              xm7, [r2 + 2 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m9, m9, xm7, 1
+
+    movu              xm7, [r2 + 10 + 32]
+    pshufb            xm7, [r5]
+    vinserti128       m10, m10, xm7, 1
+
+    INTRA_PRED_ANG16_CAL_ROW m7, m8, 3
+
+    ; transpose and store
+    INTRA_PRED_TRANS_STORE_16x16
+    RET
+%endif
+
 INIT_YMM avx2
 cglobal intra_pred_ang16_25, 3, 5, 5
     mova              m0, [pw_1024]
@@ -13514,5 +15675,2154 @@
     vpermq            m6, m6, 11011000b
     movu              [r0 + r3], m6
     RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang32_33, 3, 5, 11
+    mova              m0, [pw_1024]
+    mova              m1, [intra_pred_shuff_0_8]
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang32_mode_33]
+
+    ;row [0]
+    vbroadcasti128    m2, [r2 + 1]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 9]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 17]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 25]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 0 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 0 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0], m6
+
+    ;row [1]
+    vbroadcasti128    m2, [r2 + 2]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 10]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 18]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 26]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r1], m6
+
+    ;row [2]
+    vbroadcasti128    m2, [r2 + 3]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 11]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 19]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 27]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row [3]
+    vbroadcasti128    m2, [r2 + 4]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 12]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 20]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 28]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 3 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 3 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row [4, 5]
+    vbroadcasti128    m2, [r2 + 5]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 13]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 21]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 29]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row [6]
+    vbroadcasti128    m2, [r2 + 6]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 14]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 22]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 30]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row [7]
+    vbroadcasti128    m2, [r2 + 7]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 15]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 23]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 31]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row [8]
+    vbroadcasti128    m2, [r2 + 8]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 16]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 24]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 32]
+    pshufb            m5, m1
+
+    lea               r0, [r0 + 4 * r1]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 3 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 3 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0], m6
+
+    ;row [9, 10]
+    vbroadcasti128    m2, [r2 + 9]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 17]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 25]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 33]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row [11]
+    vbroadcasti128    m2, [r2 + 10]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 18]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 26]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 34]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row [12]
+    vbroadcasti128    m2, [r2 + 11]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 19]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 27]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 35]
+    pshufb            m5, m1
+
+    lea               r0, [r0 + 4 * r1]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0], m6
+
+    ;row [13]
+    vbroadcasti128    m2, [r2 + 12]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 20]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 28]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 36]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 3 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 3 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r1], m6
+
+    ;row [14]
+    vbroadcasti128    m2, [r2 + 13]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 21]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 29]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 37]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 0 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 0 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row [15, 16]
+    vbroadcasti128    m2, [r2 + 14]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 22]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 30]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 38]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row [17]
+    vbroadcasti128    m2, [r2 + 15]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 23]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 31]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 39]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r1], m6
+
+    ;row [18]
+    vbroadcasti128    m2, [r2 + 16]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 24]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 32]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 40]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 3 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 3 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row [19]
+    vbroadcasti128    m2, [r2 + 17]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 25]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 33]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 41]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 0 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 0 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row [20, 21]
+    vbroadcasti128    m2, [r2 + 18]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 26]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 34]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 42]
+    pshufb            m5, m1
+
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row [22]
+    vbroadcasti128    m2, [r2 + 19]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 27]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 35]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 43]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row [23]
+    vbroadcasti128    m2, [r2 + 20]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 28]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 36]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 44]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 3 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 3 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row [24]
+    vbroadcasti128    m2, [r2 + 21]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 29]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 37]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 45]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    lea               r0, [r0 + 4 * r1]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 0 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 0 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0], m6
+
+    ;row [25, 26]
+    vbroadcasti128    m2, [r2 + 22]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 30]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 38]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 46]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row [27]
+    vbroadcasti128    m2, [r2 + 23]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 31]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 39]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 47]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row [28]
+    vbroadcasti128    m2, [r2 + 24]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 32]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 40]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 48]
+    pshufb            m5, m1
+
+    lea               r0, [r0 + 4 * r1]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 3 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 3 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0], m6
+
+    ;row [29]
+    vbroadcasti128    m2, [r2 + 25]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 33]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 41]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 49]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 0 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 0 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r1], m6
+
+    ;row [30]
+    vbroadcasti128    m2, [r2 + 26]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 34]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 42]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 50]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row [31]
+    vbroadcasti128    m2, [r2 + 27]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 35]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 43]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 51]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang32_25, 3, 5, 11
+    mova              m0, [pw_1024]
+    mova              m1, [intra_pred_shuff_0_8]
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang32_mode_25]
+
+    ;row [0, 1]
+    vbroadcasti128    m2, [r2 + 0]
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 8]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 16]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 24]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[2, 3]
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[4, 5]
+    mova              m10, [r4 + 2 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[6, 7]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[8, 9]
+    add               r4, 4 * mmsize
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[10, 11]
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[12, 13]
+    mova              m10, [r4 + 2 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[14, 15]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[16, 17]
+    movu              xm2, [r2 - 1]
+    pinsrb            xm2, [r2 + 80], 0
+    vinserti128       m2, m2, xm2, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 7]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 15]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 23]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[18, 19]
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[20, 21]
+    mova              m10, [r4 + 2 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[22, 23]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[24, 25]
+    add               r4, 4 * mmsize
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[26, 27]
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[28, 29]
+    mova              m10, [r4 + 2 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[30, 31]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang32_24, 3, 5, 12
+    mova              m0, [pw_1024]
+    mova              m1, [intra_pred_shuff_0_8]
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang32_mode_24]
+
+    ;row[0, 1]
+    vbroadcasti128    m11, [r2 + 0]
+    pshufb            m2, m11, m1
+    vbroadcasti128    m3, [r2 + 8]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 16]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 24]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[2, 3]
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[4, 5]
+    mova              m10, [r4 + 2 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[6, 7]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 70], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 7]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 15]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 23]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[8, 9]
+    add               r4, 4 * mmsize
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[10, 11]
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[12, 13]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 77], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 6]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 14]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 22]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[14, 15]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[16, 17]
+    add               r4, 4 * mmsize
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[18]
+    mova              m10, [r4 + 1 * mmsize]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, m10
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, m10
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row[19, 20]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 83], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 5]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 13]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 21]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[21, 22]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[23, 24]
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[25, 26]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 90], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 4]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 12]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 20]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[27, 28]
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[29, 30]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;[row 31]
+    mova              m10, [r4 + 4 * mmsize]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, m10
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, m10
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang32_23, 3, 5, 12
+    mova              m0, [pw_1024]
+    mova              m1, [intra_pred_shuff_0_8]
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang32_mode_23]
+
+    ;row[0, 1]
+    vbroadcasti128    m11, [r2 + 0]
+    pshufb            m2, m11, m1
+    vbroadcasti128    m3, [r2 + 8]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 16]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 24]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[2]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row[3, 4]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 68], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 7]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 15]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 23]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[5, 6]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[7, 8]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 71], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 6]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 14]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 22]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[9]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r1], m6
+
+    ;row[10, 11]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 75], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 5]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 13]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 21]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[12, 13]
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[14, 15]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 78], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 4]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 12]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 20]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[16]
+    lea               r0, [r0 + 4 * r1]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0], m6
+
+    ;row[17, 18]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 82], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 3]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 11]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 19]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[19, 20]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[21, 22]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 85], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 2]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 10]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 18]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[23]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row[24, 25]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 89], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 1]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 9]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 17]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[26, 27]
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[28, 29]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 92], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 0]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 8]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 16]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[30, 31]
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang32_22, 3, 5, 13
+    mova              m0, [pw_1024]
+    mova              m1, [intra_pred_shuff_0_8]
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang32_mode_22]
+
+    ;row[0, 1]
+    vbroadcasti128    m11, [r2 + 0]
+    pshufb            m2, m11, m1
+    vbroadcasti128    m3, [r2 + 8]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 16]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 24]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[2, 3]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 66], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 7]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 15]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 23]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[4, 5]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 69], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 6]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 14]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 22]
+    pshufb            m5, m1
+
+    lea               r0, [r0 + 4 * r1]
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[6]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 3 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 3 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row[7, 8]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 71], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 5]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 13]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 21]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[9, 10]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 74], 0
+    vinserti128       m2, m11, xm11, 1
+    vinserti128       m2, m2, xm2, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 4]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 12]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 20]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[11]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row[12, 13]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 76], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 3]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 11]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 19]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 3 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[14, 15]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 79], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 2]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 10]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 18]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[16]
+    lea               r0, [r0 + 4 * r1]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 1 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 1 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0], m6
+
+    ;row[17, 18]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 81], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 1]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 9]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 17]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[19, 20]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 84], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m12, [r2 + 0]
+    pshufb            m3, m12, m1
+    vbroadcasti128    m4, [r2 + 8]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 16]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[21]
+    add               r4, 4 * mmsize
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 0 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 0 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r1], m6
+
+    ;row[22, 23]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 86], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 66], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 7]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 15]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[24, 25]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 89], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 69], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 6]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 14]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+    lea               r0, [r0 + 4 * r1]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[26]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 3 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 3 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + 2 * r1], m6
+
+    ;row[27, 28]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 91], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 71], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 5]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 13]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[29, 30]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 94], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 74], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 4]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 12]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[31]
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 2 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 2 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang32_21, 3, 5, 13
+    mova              m0, [pw_1024]
+    mova              m1, [intra_pred_shuff_0_8]
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang32_mode_21]
+
+    ;row[0]
+    vbroadcasti128    m11, [r2 + 0]
+    pshufb            m2, m11, m1
+    vbroadcasti128    m3, [r2 + 8]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 16]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 24]
+    pshufb            m5, m1
+
+    vperm2i128        m6, m2, m3, 00100000b
+    pmaddubsw         m6, [r4 + 0 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 0 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0], m6
+
+    ;row[1, 2]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 66], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 7]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 15]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 23]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[3, 4]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 68], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 6]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 14]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 22]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[5, 6]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 70], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 5]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 13]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 21]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[7, 8]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 72], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 4]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 12]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 20]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[9, 10]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 73], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 3]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 11]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 19]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[11, 12]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 75], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 2]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 10]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 18]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r3], m7
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m6
+
+    ;row[13, 14]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 77], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m3, [r2 + 1]
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 9]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 17]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + r1], m7
+    movu              [r0 + 2 * r1], m6
+
+    ;row[15]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 79], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    vbroadcasti128    m12, [r2 + 0]
+    pshufb            m3, m12, m1
+    vbroadcasti128    m4, [r2 + 8]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 16]
+    pshufb            m5, m1
+    vperm2i128        m6, m2, m3, 00100000b
+    add               r4, 4 * mmsize
+    pmaddubsw         m6, [r4 + 0 * mmsize]
+    pmulhrsw          m6, m0
+    vperm2i128        m7, m4, m5, 00100000b
+    pmaddubsw         m7, [r4 + 0 * mmsize]
+    pmulhrsw          m7, m0
+    packuswb          m6, m7
+    vpermq            m6, m6, 11011000b
+    movu              [r0 + r3], m6
+
+    ;row[16, 17]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 81], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 66], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 7]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 15]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[18, 19]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 83], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 68], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 6]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 14]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[20, 21]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 85], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 70], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 5]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 13]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[22, 23]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 87], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 72], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 4]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 12]
+    pshufb            m5, m1
+
+    add               r4, 4 * mmsize
+    mova              m10, [r4 + 0 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[24, 25]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 88], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 73], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 3]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 11]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 1 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[26, 27]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 90], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 75], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 2]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 10]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 2 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+
+    ;row[28, 29]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 92], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 77], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 1]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 9]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 3 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    lea               r0, [r0 + 4 * r1]
+    movu              [r0], m7
+    movu              [r0 + r1], m6
+
+    ;row[30, 31]
+    pslldq            xm11, 1
+    pinsrb            xm11, [r2 + 94], 0
+    vinserti128       m2, m11, xm11, 1
+    pshufb            m2, m1
+    pslldq            xm12, 1
+    pinsrb            xm12, [r2 + 79], 0
+    vinserti128       m3, m12, xm12, 1
+    pshufb            m3, m1
+    vbroadcasti128    m4, [r2 + 0]
+    pshufb            m4, m1
+    vbroadcasti128    m5, [r2 + 8]
+    pshufb            m5, m1
+
+    mova              m10, [r4 + 4 * mmsize]
+
+    INTRA_PRED_ANG32_CAL_ROW
+    movu              [r0 + 2 * r1], m7
+    movu              [r0 + r3], m6
+    RET
 %endif
 
+%macro INTRA_PRED_STORE_4x4 0
+    movd              [r0], xm0
+    pextrd            [r0 + r1], xm0, 1
+    vextracti128      xm0, m0, 1
+    lea               r0, [r0 + 2 * r1]
+    movd              [r0], xm0
+    pextrd            [r0 + r1], xm0, 1
+%endmacro
+
+%macro INTRA_PRED_TRANS_STORE_4x4 0
+    vpermq            m0, m0, 00001000b
+    pshufb            m0, [c_trans_4x4]
+
+    ;store
+    movd              [r0], xm0
+    pextrd            [r0 + r1], xm0, 1
+    lea               r0, [r0 + 2 * r1]
+    pextrd            [r0], xm0, 2
+    pextrd            [r0 + r1], xm0, 3
+%endmacro
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_27, 3, 3, 1
+    vbroadcasti128    m0, [r2 + 1]
+    pshufb            m0, [intra_pred_shuff_0_4]
+    pmaddubsw         m0, [c_ang4_mode_27]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_28, 3, 3, 1
+    vbroadcasti128    m0, [r2 + 1]
+    pshufb            m0, [intra_pred_shuff_0_4]
+    pmaddubsw         m0, [c_ang4_mode_28]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_29, 3, 3, 1
+    vbroadcasti128    m0, [r2 + 1]
+    pshufb            m0, [intra_pred4_shuff1]
+    pmaddubsw         m0, [c_ang4_mode_29]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_30, 3, 3, 1
+    vbroadcasti128    m0, [r2 + 1]
+    pshufb            m0, [intra_pred4_shuff2]
+    pmaddubsw         m0, [c_ang4_mode_30]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_31, 3, 3, 1
+    vbroadcasti128    m0, [r2 + 1]
+    pshufb            m0, [intra_pred4_shuff31]
+    pmaddubsw         m0, [c_ang4_mode_31]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_32, 3, 3, 1
+    vbroadcasti128    m0, [r2 + 1]
+    pshufb            m0, [intra_pred4_shuff31]
+    pmaddubsw         m0, [c_ang4_mode_32]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_33, 3, 3, 1
+    vbroadcasti128    m0, [r2 + 1]
+    pshufb            m0, [intra_pred4_shuff33]
+    pmaddubsw         m0, [c_ang4_mode_33]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_3, 3, 3, 1
+    vbroadcasti128    m0, [r2 + 1]
+    pshufb            m0, [intra_pred4_shuff3]
+    pmaddubsw         m0, [c_ang4_mode_33]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_4, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff5]
+    pmaddubsw         m0, [c_ang4_mode_32]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_5, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff5]
+    pmaddubsw         m0, [c_ang4_mode_5]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_6, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff6]
+    pmaddubsw         m0, [c_ang4_mode_6]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_7, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff7]
+    pmaddubsw         m0, [c_ang4_mode_7]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_8, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff9]
+    pmaddubsw         m0, [c_ang4_mode_8]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_9, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff9]
+    pmaddubsw         m0, [c_ang4_mode_9]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_11, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff12]
+    pmaddubsw         m0, [c_ang4_mode_11]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_12, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff12]
+    pmaddubsw         m0, [c_ang4_mode_12]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_13, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff13]
+    pmaddubsw         m0, [c_ang4_mode_13]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_14, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff14]
+    pmaddubsw         m0, [c_ang4_mode_14]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_15, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff15]
+    pmaddubsw         m0, [c_ang4_mode_15]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_16, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff16]
+    pmaddubsw         m0, [c_ang4_mode_16]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_17, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff17]
+    pmaddubsw         m0, [c_ang4_mode_17]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_19, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff19]
+    pmaddubsw         m0, [c_ang4_mode_19]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_20, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff20]
+    pmaddubsw         m0, [c_ang4_mode_20]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_21, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff21]
+    pmaddubsw         m0, [c_ang4_mode_21]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_22, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff22]
+    pmaddubsw         m0, [c_ang4_mode_22]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_23, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff23]
+    pmaddubsw         m0, [c_ang4_mode_23]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_24, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred_shuff_0_4]
+    pmaddubsw         m0, [c_ang4_mode_24]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang4_25, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred_shuff_0_4]
+    pmaddubsw         m0, [c_ang4_mode_25]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_STORE_4x4
+    RET

x265_1.6.tar.gz/source/common/x86/intrapred8_allangs.asm -> x265_1.7.tar.gz/source/common/x86/intrapred8_allangs.asm Changed

@@ -2,7 +2,7 @@
 ;* Copyright (C) 2013 x265 project
 ;*
 ;* Authors: Min Chen <chenm003@163.com> <min.chen@multicorewareinc.com>
-;*          Praveen Tiwari <praveen@multicorewareinc.com>
+;*          Praveen 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
@@ -27,6 +27,64 @@
 
 SECTION_RODATA 32
 
+all_ang4_shuff: db 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6, 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
+                db 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6, 3, 4, 4, 5, 5, 6, 6, 7
+                db 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6
+                db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5
+                db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5
+                db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4
+                db 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3
+                db 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12
+                db 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 4, 0, 0, 9, 9, 10, 10, 11
+                db 0, 9, 9, 10, 10, 11, 11, 12, 0, 9, 9, 10, 10, 11, 11, 12, 2, 0, 0, 9, 9, 10, 10, 11, 2, 0, 0, 9, 9, 10, 10, 11
+                db 0, 9, 9, 10, 10, 11, 11, 12, 2, 0, 0, 9, 9, 10, 10, 11, 2, 0, 0, 9, 9, 10, 10, 11, 4, 2, 2, 0, 0, 9, 9, 10
+                db 0, 9, 9, 10, 10, 11, 11, 12, 2, 0, 0, 9, 9, 10, 10, 11, 2, 0, 0, 9, 9, 10, 10, 11, 3, 2, 2, 0, 0, 9, 9, 10
+                db 0, 9, 9, 10, 10, 11, 11, 12, 1, 0, 0, 9, 9, 10, 10, 11, 2, 1, 1, 0, 0, 9, 9, 10, 4, 2, 2, 1, 1, 0, 0, 9
+                db 0, 1, 2, 3, 9, 0, 1, 2, 10, 9, 0, 1, 11, 10, 9, 0, 0, 1, 2, 3, 9, 0, 1, 2, 10, 9, 0, 1, 11, 10, 9, 0
+                db 0, 1, 1, 2, 2, 3, 3, 4, 9, 0, 0, 1, 1, 2, 2, 3, 10, 9, 9, 0, 0, 1, 1, 2, 12, 10, 10, 9, 9, 0, 0, 1
+                db 0, 1, 1, 2, 2, 3, 3, 4, 10, 0, 0, 1, 1, 2, 2, 3, 10, 0, 0, 1, 1, 2, 2, 3, 11, 10, 10, 0, 0, 1, 1, 2
+                db 0, 1, 1, 2, 2, 3, 3, 4, 10, 0, 0, 1, 1, 2, 2, 3, 10, 0, 0, 1, 1, 2, 2, 3, 12, 10, 10, 0, 0, 1, 1, 2
+                db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 10, 0, 0, 1, 1, 2, 2, 3, 10, 0, 0, 1, 1, 2, 2, 3
+                db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 12, 0, 0, 1, 1, 2, 2, 3
+                db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4
+                db 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4
+                db 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5
+                db 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6
+                db 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6, 2, 3, 3, 4, 4, 5, 5, 6
+                db 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6, 2, 3, 3, 4, 4, 5, 5, 6, 3, 4, 4, 5, 5, 6, 6, 7
+                db 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6, 3, 4, 4, 5, 5, 6, 6, 7, 4, 5, 5, 6, 6, 7, 7, 8
+                db 2, 3, 4, 5, 3, 4, 5, 6, 4, 5, 6, 7, 5, 6, 7, 8, 2, 3, 4, 5, 3, 4, 5, 6, 4, 5, 6, 7, 5, 6, 7, 8
+
+all_ang4: db 6, 26, 6, 26, 6, 26, 6, 26, 12, 20, 12, 20, 12, 20, 12, 20, 18, 14, 18, 14, 18, 14, 18, 14, 24, 8, 24, 8, 24, 8, 24, 8
+          db 11, 21, 11, 21, 11, 21, 11, 21, 22, 10, 22, 10, 22, 10, 22, 10, 1, 31, 1, 31, 1, 31, 1, 31, 12, 20, 12, 20, 12, 20, 12, 20
+          db 15, 17, 15, 17, 15, 17, 15, 17, 30, 2, 30, 2, 30, 2, 30, 2, 13, 19, 13, 19, 13, 19, 13, 19, 28, 4, 28, 4, 28, 4, 28, 4
+          db 19, 13, 19, 13, 19, 13, 19, 13, 6, 26, 6, 26, 6, 26, 6, 26, 25, 7, 25, 7, 25, 7, 25, 7, 12, 20, 12, 20, 12, 20, 12, 20
+          db 23, 9, 23, 9, 23, 9, 23, 9, 14, 18, 14, 18, 14, 18, 14, 18, 5, 27, 5, 27, 5, 27, 5, 27, 28, 4, 28, 4, 28, 4, 28, 4
+          db 27, 5, 27, 5, 27, 5, 27, 5, 22, 10, 22, 10, 22, 10, 22, 10, 17, 15, 17, 15, 17, 15, 17, 15, 12, 20, 12, 20, 12, 20, 12, 20
+          db 30, 2, 30, 2, 30, 2, 30, 2, 28, 4, 28, 4, 28, 4, 28, 4, 26, 6, 26, 6, 26, 6, 26, 6, 24, 8, 24, 8, 24, 8, 24, 8
+          db 2, 30, 2, 30, 2, 30, 2, 30, 4, 28, 4, 28, 4, 28, 4, 28, 6, 26, 6, 26, 6, 26, 6, 26, 8, 24, 8, 24, 8, 24, 8, 24
+          db 5, 27, 5, 27, 5, 27, 5, 27, 10, 22, 10, 22, 10, 22, 10, 22, 15, 17, 15, 17, 15, 17, 15, 17, 20, 12, 20, 12, 20, 12, 20, 12
+          db 9, 23, 9, 23, 9, 23, 9, 23, 18, 14, 18, 14, 18, 14, 18, 14, 27, 5, 27, 5, 27, 5, 27, 5, 4, 28, 4, 28, 4, 28, 4, 28
+          db 13, 19, 13, 19, 13, 19, 13, 19, 26, 6, 26, 6, 26, 6, 26, 6, 7, 25, 7, 25, 7, 25, 7, 25, 20, 12, 20, 12, 20, 12, 20, 12
+          db 17, 15, 17, 15, 17, 15, 17, 15, 2, 30, 2, 30, 2, 30, 2, 30, 19, 13, 19, 13, 19, 13, 19, 13, 4, 28, 4, 28, 4, 28, 4, 28
+          db 21, 11, 21, 11, 21, 11, 21, 11, 10, 22, 10, 22, 10, 22, 10, 22, 31, 1, 31, 1, 31, 1, 31, 1, 20, 12, 20, 12, 20, 12, 20, 12
+          db 26, 6, 26, 6, 26, 6, 26, 6, 20, 12, 20, 12, 20, 12, 20, 12, 14, 18, 14, 18, 14, 18, 14, 18, 8, 24, 8, 24, 8, 24, 8, 24
+          db 26, 6, 26, 6, 26, 6, 26, 6, 20, 12, 20, 12, 20, 12, 20, 12, 14, 18, 14, 18, 14, 18, 14, 18, 8, 24, 8, 24, 8, 24, 8, 24
+          db 21, 11, 21, 11, 21, 11, 21, 11, 10, 22, 10, 22, 10, 22, 10, 22, 31, 1, 31, 1, 31, 1, 31, 1, 20, 12, 20, 12, 20, 12, 20, 12
+          db 17, 15, 17, 15, 17, 15, 17, 15, 2, 30, 2, 30, 2, 30, 2, 30, 19, 13, 19, 13, 19, 13, 19, 13, 4, 28, 4, 28, 4, 28, 4, 28
+          db 13, 19, 13, 19, 13, 19, 13, 19, 26, 6, 26, 6, 26, 6, 26, 6, 7, 25, 7, 25, 7, 25, 7, 25, 20, 12, 20, 12, 20, 12, 20, 12
+          db 9, 23, 9, 23, 9, 23, 9, 23, 18, 14, 18, 14, 18, 14, 18, 14, 27, 5, 27, 5, 27, 5, 27, 5, 4, 28, 4, 28, 4, 28, 4, 28
+          db 5, 27, 5, 27, 5, 27, 5, 27, 10, 22, 10, 22, 10, 22, 10, 22, 15, 17, 15, 17, 15, 17, 15, 17, 20, 12, 20, 12, 20, 12, 20, 12
+          db 2, 30, 2, 30, 2, 30, 2, 30, 4, 28, 4, 28, 4, 28, 4, 28, 6, 26, 6, 26, 6, 26, 6, 26, 8, 24, 8, 24, 8, 24, 8, 24
+          db 30, 2, 30, 2, 30, 2, 30, 2, 28, 4, 28, 4, 28, 4, 28, 4, 26, 6, 26, 6, 26, 6, 26, 6, 24, 8, 24, 8, 24, 8, 24, 8
+          db 27, 5, 27, 5, 27, 5, 27, 5, 22, 10, 22, 10, 22, 10, 22, 10, 17, 15, 17, 15, 17, 15, 17, 15, 12, 20, 12, 20, 12, 20, 12, 20
+          db 23, 9, 23, 9, 23, 9, 23, 9, 14, 18, 14, 18, 14, 18, 14, 18, 5, 27, 5, 27, 5, 27, 5, 27, 28, 4, 28, 4, 28, 4, 28, 4
+          db 19, 13, 19, 13, 19, 13, 19, 13, 6, 26, 6, 26, 6, 26, 6, 26, 25, 7, 25, 7, 25, 7, 25, 7, 12, 20, 12, 20, 12, 20, 12, 20
+          db 15, 17, 15, 17, 15, 17, 15, 17, 30, 2, 30, 2, 30, 2, 30, 2, 13, 19, 13, 19, 13, 19, 13, 19, 28, 4, 28, 4, 28, 4, 28, 4
+          db 11, 21, 11, 21, 11, 21, 11, 21, 22, 10, 22, 10, 22, 10, 22, 10, 1, 31, 1, 31, 1, 31, 1, 31, 12, 20, 12, 20, 12, 20, 12, 20
+          db 6, 26, 6, 26, 6, 26, 6, 26, 12, 20, 12, 20, 12, 20, 12, 20, 18, 14, 18, 14, 18, 14, 18, 14, 24, 8, 24, 8, 24, 8, 24, 8
+
+
 SECTION .text
 
 ; global constant
@@ -34,9 +92,14 @@
 
 ; common constant with intrapred8.asm
 cextern ang_table
+cextern pw_ang_table
 cextern tab_S1
 cextern tab_S2
 cextern tab_Si
+cextern pw_16
+cextern pb_000000000000000F
+cextern pb_0000000000000F0F
+cextern pw_FFFFFFFFFFFFFFF0
 
 
 ;-----------------------------------------------------------------------------
@@ -23006,3 +23069,1098 @@
     palignr    m4,              m2,       m1,    14
     movu       [r0 + 2111 * 16],   m4
     RET
+
+
+;-----------------------------------------------------------------------------
+; void all_angs_pred_4x4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma)
+;-----------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal all_angs_pred_4x4, 4, 4, 6
+
+    mova           m5, [pw_1024]
+    lea            r2, [all_ang4]
+    lea            r3, [all_ang4_shuff]
+
+; mode 2
+
+    vbroadcasti128 m0, [r1 + 9]
+    mova           xm1, xm0
+    psrldq         xm1, 1
+    pshufb         xm1, [r3]
+    movu           [r0], xm1
+
+; mode 3
+
+    pshufb         m1, m0, [r3 + 1 * mmsize]
+    pmaddubsw      m1, [r2]
+    pmulhrsw       m1, m5
+
+; mode 4
+
+    pshufb         m2, m0, [r3 + 2 * mmsize]
+    pmaddubsw      m2, [r2 + 1 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (3 - 2) * 16], m1
+
+; mode 5
+
+    pshufb         m1, m0, [r3 + 2 * mmsize]
+    pmaddubsw      m1, [r2 + 2 * mmsize]
+    pmulhrsw       m1, m5
+
+; mode 6
+
+    pshufb         m2, m0, [r3 + 3 * mmsize]
+    pmaddubsw      m2, [r2 + 3 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (5 - 2) * 16], m1
+
+    add            r3, 4 * mmsize
+    add            r2, 4 * mmsize
+
+; mode 7
+
+    pshufb         m1, m0, [r3 + 0 * mmsize]
+    pmaddubsw      m1, [r2 + 0 * mmsize]
+    pmulhrsw       m1, m5
+
+; mode 8
+
+    pshufb         m2, m0, [r3 + 1 * mmsize]
+    pmaddubsw      m2, [r2 + 1 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (7 - 2) * 16], m1
+
+; mode 9
+
+    pshufb         m1, m0, [r3 + 1 * mmsize]
+    pmaddubsw      m1, [r2 + 2 * mmsize]
+    pmulhrsw       m1, m5
+    packuswb       m1, m1
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (9 - 2) * 16], xm1
+
+; mode 10
+
+    pshufb         xm1, xm0, [r3 + 2 * mmsize]
+    movu           [r0 + (10 - 2) * 16], xm1
+
+    pxor           xm1, xm1
+    movd           xm2, [r1 + 1]
+    pshufd         xm3, xm2, 0
+    punpcklbw      xm3, xm1
+    pinsrb         xm2, [r1], 0
+    pshufb         xm4, xm2, xm1
+    punpcklbw      xm4, xm1
+    psubw          xm3, xm4
+    psraw          xm3, 1
+    pshufb         xm4, xm0, xm1
+    punpcklbw      xm4, xm1
+    paddw          xm3, xm4
+    packuswb       xm3, xm1
+
+    pextrb         [r0 + 128], xm3, 0
+    pextrb         [r0 + 132], xm3, 1
+    pextrb         [r0 + 136], xm3, 2
+    pextrb         [r0 + 140], xm3, 3
+
+; mode 11
+
+    vbroadcasti128 m0, [r1]
+    pshufb         m1, m0, [r3 + 3 * mmsize]
+    pmaddubsw      m1, [r2 + 3 * mmsize]
+    pmulhrsw       m1, m5
+
+; mode 12
+
+    add            r2, 4 * mmsize
+
+    pshufb         m2, m0, [r3 + 3 * mmsize]
+    pmaddubsw      m2, [r2 + 0 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (11 - 2) * 16], m1
+
+; mode 13
+
+    add            r3, 4 * mmsize
+
+    pshufb         m1, m0, [r3 + 0 * mmsize]
+    pmaddubsw      m1, [r2 + 1 * mmsize]
+    pmulhrsw       m1, m5
+
+; mode 14
+
+    pshufb         m2, m0, [r3 + 1 * mmsize]
+    pmaddubsw      m2, [r2 + 2 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (13 - 2) * 16], m1
+
+; mode 15
+
+    pshufb         m1, m0, [r3 + 2 * mmsize]
+    pmaddubsw      m1, [r2 + 3 * mmsize]
+    pmulhrsw       m1, m5
+
+; mode 16
+
+    add            r2, 4 * mmsize
+
+    pshufb         m2, m0, [r3 + 3 * mmsize]
+    pmaddubsw      m2, [r2 + 0 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (15 - 2) * 16], m1
+
+; mode 17
+
+    add            r3, 4 * mmsize
+
+    pshufb         m1, m0, [r3 + 0 * mmsize]
+    pmaddubsw      m1, [r2 + 1 * mmsize]
+    pmulhrsw       m1, m5
+    packuswb       m1, m1
+    vpermq         m1, m1, 11011000b
+
+; mode 18
+
+    pshufb         m2, m0, [r3 + 1 * mmsize]
+    vinserti128    m1, m1, xm2, 1
+    movu           [r0 + (17 - 2) * 16], m1
+
+; mode 19
+
+    pshufb         m1, m0, [r3 + 2 * mmsize]
+    pmaddubsw      m1, [r2 + 2 * mmsize]
+    pmulhrsw       m1, m5
+
+; mode 20
+
+    pshufb         m2, m0, [r3 + 3 * mmsize]
+    pmaddubsw      m2, [r2 + 3 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (19 - 2) * 16], m1
+
+; mode 21
+
+    add            r2, 4 * mmsize
+    add            r3, 4 * mmsize
+
+    pshufb         m1, m0, [r3 + 0 * mmsize]
+    pmaddubsw      m1, [r2 + 0 * mmsize]
+    pmulhrsw       m1, m5
+
+; mode 22
+
+    pshufb         m2, m0, [r3 + 1 * mmsize]
+    pmaddubsw      m2, [r2 + 1 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (21 - 2) * 16], m1
+
+; mode 23
+
+    pshufb         m1, m0, [r3 + 2 * mmsize]
+    pmaddubsw      m1, [r2 + 2 * mmsize]
+    pmulhrsw       m1, m5
+
+; mode 24
+
+    pshufb         m2, m0, [r3 + 3 * mmsize]
+    pmaddubsw      m2, [r2 + 3 * mmsize]
+    pmulhrsw       m2, m5
+    packuswb       m1, m2
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (23 - 2) * 16], m1
+
+; mode 25
+
+    add            r2, 4 * mmsize
+
+    pshufb         m1, m0, [r3 + 3 * mmsize]
+    pmaddubsw      m1, [r2 + 0 * mmsize]
+    pmulhrsw       m1, m5
+    packuswb       m1, m1
+    vpermq         m1, m1, 11011000b
+    movu           [r0 + (25 - 2) * 16], xm1
+
+; mode 26
+
+    add            r3, 4 * mmsize
+
+    pshufb         xm1, xm0, [r3 + 0 * mmsize]
+    movu           [r0 + (26 - 2) * 16], xm1
+
+    pxor           xm1, xm1
+    movd           xm2, [r1 + 9]
+    pshufd         xm3, xm2, 0
+    punpcklbw      xm3, xm1
+    pinsrb         xm4, [r1 + 0], 0
+    pshufb         xm4, xm1
+    punpcklbw      xm4, xm1
+    psubw          xm3, xm4
+    psraw          xm3, 1
+    psrldq         xm2, xm0, 1
+    pshufb         xm2, xm1
+    punpcklbw      xm2, xm1
+    paddw          xm3, xm2
+    packuswb       xm3, xm1
+
+    pextrb       [r0 + 384], xm3, 0
+    pextrb       [r0 + 388], xm3, 1
+    pextrb       [r0 + 392], xm3, 2
+    pextrb       [r0 + 396], xm3, 3
+
+; mode 27
+
+    pshufb        m1, m0, [r3 + 1 * mmsize]
+    pmaddubsw     m1, [r2 + 1 * mmsize]
+    pmulhrsw      m1, m5
+
+; mode 28
+
+    pshufb        m2, m0, [r3 + 1 * mmsize]
+    pmaddubsw     m2, [r2 + 2 * mmsize]
+    pmulhrsw      m2, m5
+    packuswb      m1, m2
+    vpermq        m1, m1, 11011000b
+    movu          [r0 + (27 - 2) * 16], m1
+
+; mode 29
+
+    pshufb        m1, m0, [r3 + 2 * mmsize]
+    pmaddubsw     m1, [r2 + 3 * mmsize]
+    pmulhrsw      m1, m5
+
+; mode 30
+
+    add           r2, 4 * mmsize
+
+    pshufb        m2, m0, [r3 + 3 * mmsize]
+    pmaddubsw     m2, [r2 + 0 * mmsize]
+    pmulhrsw      m2, m5
+    packuswb      m1, m2
+    vpermq        m1, m1, 11011000b
+    movu          [r0 + (29 - 2) * 16], m1
+
+; mode 31
+
+    add           r3, 4 * mmsize
+
+    pshufb        m1, m0, [r3 + 0 * mmsize]
+    pmaddubsw     m1, [r2 + 1 * mmsize]
+    pmulhrsw      m1, m5
+
+; mode 32
+
+    pshufb        m2, m0, [r3 + 0 * mmsize]
+    pmaddubsw     m2, [r2 + 2 * mmsize]
+    pmulhrsw      m2, m5
+    packuswb      m1, m2
+    vpermq        m1, m1, 11011000b
+    movu          [r0 + (31 - 2) * 16], m1
+
+; mode 33
+
+    pshufb        m1, m0, [r3 + 1 * mmsize]
+    pmaddubsw     m1, [r2 + 3 * mmsize]
+    pmulhrsw      m1, m5
+    packuswb      m1, m2
+    vpermq        m1, m1, 11011000b
+
+; mode 34
+
+    pshufb        m0, [r3 + 2 * mmsize]
+    vinserti128   m1, m1, xm0, 1
+    movu          [r0 + (33 - 2) * 16], m1
+    RET
+
+;-----------------------------------------------------------------------------
+; void all_angs_pred_4x4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma)
+;-----------------------------------------------------------------------------
+INIT_XMM sse2
+cglobal all_angs_pred_4x4, 4, 4, 8
+
+; mode 2
+
+    movh        m6,             [r1 + 9]
+    mova        m2,             m6
+    psrldq      m2,             1
+    movd        [r0],           m2              ;byte[A, B, C, D]
+    psrldq      m2,             1
+    movd        [r0 + 4],       m2              ;byte[B, C, D, E]
+    psrldq      m2,             1
+    movd        [r0 + 8],       m2              ;byte[C, D, E, F]
+    psrldq      m2,             1
+    movd        [r0 + 12],      m2              ;byte[D, E, F, G]
+
+; mode 10/26
+
+    pxor        m7,             m7
+    pshufd      m5,             m6,        0
+    mova        [r0 + 128],     m5              ;mode 10 byte[9, A, B, C, 9, A, B, C, 9, A, B, C, 9, A, B, C]
+
+    movd        m4,             [r1 + 1]
+    pshufd      m4,             m4,        0
+    mova        [r0 + 384],     m4              ;mode 26 byte[1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4]
+
+    movd        m1,             [r1]
+    punpcklbw   m1,             m7
+    pshuflw     m1,             m1,     0x00
+    punpcklqdq  m1,             m1              ;m1 = byte[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+
+    punpckldq   m4,             m5
+    punpcklbw   m4,             m7              ;m4 = word[1, 2, 3, 4, 9, A, B, C]
+    pshuflw     m2,             m4,     0x00
+    pshufhw     m2,             m2,     0x00    ;m2 = word[1, 1, 1, 1, 9, 9, 9, 9]
+
+    psubw       m4,             m1
+    psraw       m4,             1
+
+    pshufd      m2,             m2,     q1032   ;m2 = word[9, 9, 9, 9, 1, 1, 1, 1]
+    paddw       m4,             m2
+    packuswb    m4,             m4
+
+%if ARCH_X86_64
+    movq        r2,             m4
+
+    mov         [r0 + 128],     r2b              ;mode 10
+    shr         r2,             8
+    mov         [r0 + 132],     r2b
+    shr         r2,             8
+    mov         [r0 + 136],     r2b
+    shr         r2,             8
+    mov         [r0 + 140],     r2b
+    shr         r2,             8
+    mov         [r0 + 384],     r2b              ;mode 26
+    shr         r2d,            8
+    mov         [r0 + 388],     r2b
+    shr         r2d,            8
+    mov         [r0 + 392],     r2b
+    shr         r2d,            8
+    mov         [r0 + 396],     r2b
+
+%else
+    movd        r2d,             m4
+
+    mov         [r0 + 128],     r2b              ;mode 10
+    shr         r2d,             8
+    mov         [r0 + 132],     r2b
+    shr         r2d,             8
+    mov         [r0 + 136],     r2b
+    shr         r2d,             8
+    mov         [r0 + 140],     r2b
+
+    psrldq      m4,             4
+    movd        r2d,            m4
+
+    mov         [r0 + 384],     r2b              ;mode 26
+    shr         r2d,            8
+    mov         [r0 + 388],     r2b
+    shr         r2d,            8
+    mov         [r0 + 392],     r2b
+    shr         r2d,            8
+    mov         [r0 + 396],     r2b
+%endif
+
+; mode 3
+
+    mova        m2,             [pw_16]
+    lea         r3,             [pw_ang_table + 7 * 16]
+    lea         r2,             [pw_ang_table + 23 * 16]
+    punpcklbw   m6,             m6
+    psrldq      m6,             1
+    movh        m1,             m6
+    psrldq      m6,             2
+    movh        m0,             m6
+    psrldq      m6,             2
+    movh        m3,             m6
+    psrldq      m6,             2
+    punpcklbw   m1,             m7              ;m1 = word[9, A, A, B, B, C, C, D]
+    punpcklbw   m0,             m7              ;m0 = word[A, B, B, C, C, D, D, E]
+    punpcklbw   m3,             m7              ;m3 = word[B, C, C, D, D, E, E, F]
+    punpcklbw   m6,             m7              ;m6 = word[C, D, D, E, E, F, F, G]
+
+    mova        m7,             [r2 - 3 * 16]
+
+    pmaddwd     m5,             m1,     [r2 + 3 * 16]
+    pmaddwd     m4,             m0,     m7
+
+    packssdw    m5,             m4
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m4,             m3,     [r3 + 7 * 16]
+    pmaddwd     m6,             [r3 + 1 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 16],      m5
+    movd        [r0 + 68],      m5              ;mode 6 row 1
+    psrldq      m5,             4
+    movd        [r0 + 76],      m5              ;mode 6 row 3
+
+; mode 4
+
+    pmaddwd     m4,             m0,     [r2 + 8 * 16]
+    pmaddwd     m6,             m3,     m7
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m5,             m1,     [r2 - 2 * 16]
+    pmaddwd     m6,             m0,     [r3 + 3 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 32],      m5
+
+; mode 5
+
+    pmaddwd     m5,             m1,     [r2 - 6 * 16]
+    pmaddwd     m6,             m0,     [r3 - 5 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m4,             m0,     [r2 - 4 * 16]
+    pmaddwd     m3,             [r3 - 3 * 16]
+
+    packssdw    m4,             m3
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 48],      m5
+
+; mode 6
+
+    pmaddwd     m5,             m1,     [r3 + 6 * 16]
+    pmaddwd     m6,             m0,     [r3 + 0 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    packuswb    m5,             m6
+    movd        [r0 + 64],      m5
+    psrldq      m5,             4
+    movd        [r0 + 72],      m5
+
+; mode 7
+
+    pmaddwd     m5,             m1,     [r3 + 2 * 16]
+    pmaddwd     m6,             m1,     [r2 - 5 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    mova        m3,             [r2 + 4 * 16]
+    pmaddwd     m4,             m1,     m3
+    pmaddwd     m0,             [r3 - 3 * 16]
+
+    packssdw    m4,             m0
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 80],      m5
+
+; mode 8
+
+    mova        m0,             [r3 - 2 * 16]
+    pmaddwd     m5,             m1,     m0
+    pmaddwd     m6,             m1,     [r3 + 3 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m4,             m1,     [r3 + 8 * 16]
+    pmaddwd     m7,             m1
+
+    packssdw    m4,             m7
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 96],      m5
+
+; mode 9
+
+    pmaddwd     m5,             m1,     [r3 - 5 * 16]
+    pmaddwd     m6,             m1,     [r3 - 3 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m4,             m1,     [r3 - 1 * 16]
+    pmaddwd     m6,             m1,     [r3 + 1 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 112],     m5
+
+; mode 11
+
+    movd        m5,             [r1]
+    punpcklwd   m5,             m1
+    pand        m5,             [pb_0000000000000F0F]
+    pslldq      m1,             4
+    por         m1,             m5              ;m1 = word[0, 9, 9, A, A, B, B, C]
+
+    pmaddwd     m5,             m1,     [r2 + 7 * 16]
+    pmaddwd     m6,             m1,     [r2 + 5 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m4,             m1,     [r2 + 3 * 16]
+    pmaddwd     m6,             m1,     [r2 + 1 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 144],     m5
+
+; mode 12
+
+    pmaddwd     m3,             m1
+    pmaddwd     m6,             m1,     [r2 - 1 * 16]
+
+    packssdw    m3,             m6
+    paddw       m3,             m2
+    psraw       m3,             5
+
+    pmaddwd     m4,             m1,     [r2 - 6 * 16]
+    pmaddwd     m6,             m1,     [r3 + 5 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m3,             m4
+    mova        [r0 + 160],     m3
+
+; mode 13
+
+    mova        m3,             m1
+    movd        m7,             [r1 + 4]
+    punpcklwd   m7,             m1
+    pand        m7,             [pb_0000000000000F0F]
+    pslldq      m3,             4
+    por         m3,             m7              ;m3 = word[4, 0, 0, 9, 9, A, A, B]
+
+    pmaddwd     m5,             m1,     [r2 + 0 * 16]
+    pmaddwd     m6,             m1,     [r3 + 7 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m4,             m1,     m0
+    pmaddwd     m6,             m3,     [r2 + 5 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 176],     m5
+
+; mode 14
+
+    pmaddwd     m5,             m1,     [r2 - 4 * 16]
+    pmaddwd     m6,             m1,     [r3 - 1 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    movd        m6,             [r1 + 2]
+    pand        m3,             [pw_FFFFFFFFFFFFFFF0]
+    pand        m6,             [pb_000000000000000F]
+    por         m3,             m6              ;m3 = word[2, 0, 0, 9, 9, A, A, B]
+
+    pmaddwd     m4,             m3,     [r2 + 2 * 16]
+    pmaddwd     m6,             m3,     [r3 + 5 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 192],     m5
+    psrldq      m5,             4
+    movd        [r0 + 240],     m5              ;mode 17 row 0
+
+; mode 15
+
+    pmaddwd     m5,             m1,     [r3 + 8 * 16]
+    pmaddwd     m6,             m3,     [r2 + 7 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m6,             m3,     [r3 + 6 * 16]
+
+    mova        m0,             m3
+    punpcklwd   m7,             m3
+    pslldq      m0,             4
+    pand        m7,             [pb_0000000000000F0F]
+    por         m0,             m7              ;m0 = word[4, 2, 2, 0, 0, 9, 9, A]
+
+    pmaddwd     m4,             m0,     [r2 + 5 * 16]
+
+    packssdw    m6,             m4
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m5,             m6
+    mova        [r0 + 208],     m5
+
+; mode 16
+
+    pmaddwd     m5,             m1,     [r3 + 4 * 16]
+    pmaddwd     m6,             m3,     [r2 - 1 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m3,             [r3 - 6 * 16]
+
+    movd        m6,             [r1 + 3]
+    pand        m0,             [pw_FFFFFFFFFFFFFFF0]
+    pand        m6,             [pb_000000000000000F]
+    por         m0,             m6              ;m0 = word[3, 2, 2, 0, 0, 9, 9, A]
+
+    pmaddwd     m0,             [r3 + 5 * 16]
+    packssdw    m3,             m0
+    paddw       m3,             m2
+    psraw       m3,             5
+
+    packuswb    m5,             m3
+    mova        [r0 + 224],     m5
+
+; mode 17
+
+    movd        m4,             [r1 + 1]
+    punpcklwd   m4,             m1
+    pand        m4,             [pb_0000000000000F0F]
+    pslldq      m1,             4
+    por         m1,             m4              ;m1 = word[1, 0, 0, 9, 9, A, A, B]
+
+    pmaddwd     m6,             m1,     [r3 + 5 * 16]
+
+    packssdw    m6,             m6
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    movd        m5,             [r1 + 2]
+    punpcklwd   m5,             m1
+    pand        m5,             [pb_0000000000000F0F]
+    pslldq      m1,             4
+    por         m1,             m5              ;m1 = word[2, 1, 1, 0, 0, 9, 9, A]
+
+    pmaddwd     m4,             m1,     [r2 - 5 * 16]
+
+    punpcklwd   m7,             m1
+    pand        m7,             [pb_0000000000000F0F]
+    pslldq      m1,             4
+    por         m1,             m7              ;m1 = word[4, 2, 2, 1, 1, 0, 0, 9]
+
+    pmaddwd     m1,             [r2 + 1 * 16]
+    packssdw    m4,             m1
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m6,             m4
+    movd        [r0 + 244],     m6
+    psrldq      m6,             8
+    movh        [r0 + 248],     m6
+
+; mode 18
+
+    movh        m1,             [r1]
+    movd        [r0 + 256],     m1              ;byte[0, 1, 2, 3]
+
+    movh        m3,             [r1 + 2]
+    punpcklqdq  m3,             m1
+    psrldq      m3,             7
+    movd        [r0 + 260],     m3              ;byte[2, 1, 0, 9]
+
+    movh        m4,             [r1 + 3]
+    punpcklqdq  m4,             m3
+    psrldq      m4,             7
+    movd        [r0 + 264],     m4              ;byte[1, 0, 9, A]
+
+    movh        m0,             [r1 + 4]
+    punpcklqdq  m0,             m4
+    psrldq      m0,             7
+    movd        [r0 + 268],     m0              ;byte[0, 9, A, B]
+
+; mode 19
+
+    pxor        m7,             m7
+    punpcklbw   m4,             m3
+    punpcklbw   m3,             m1
+    punpcklbw   m1,             m1
+    punpcklbw   m4,             m7              ;m4 = word[A, 9, 9, 0, 0, 1, 1, 2]
+    punpcklbw   m3,             m7              ;m3 = word[9, 0, 0, 1, 1, 2, 2, 3]
+    psrldq      m1,             1
+    punpcklbw   m1,             m7              ;m1 = word[0, 1, 1, 2, 2, 3, 3, 4]
+
+    pmaddwd     m6,             m1,     [r3 - 1 * 16]
+    pmaddwd     m7,             m3,     [r3 + 5 * 16]
+
+    packssdw    m6,             m7
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    pmaddwd     m5,             m4,     [r2 - 5 * 16]
+
+    movd        m7,             [r1 + 12]
+    punpcklwd   m7,             m4
+    pand        m7,             [pb_0000000000000F0F]
+    pslldq      m4,             4
+    por         m4,             m7              ;m4 = word[C, A, A, 9, 9, 0, 0, 1]
+
+    pmaddwd     m4,             [r2 + 1 * 16]
+    packssdw    m5,             m4
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    packuswb    m6,             m5
+    mova        [r0 + 272],     m6
+    movd        [r0 + 324],     m6              ;mode 22 row 1
+
+; mode 20
+
+    pmaddwd     m5,             m1,     [r3 + 4 * 16]
+
+    movd        m4,             [r1 + 10]
+    pand        m3,             [pw_FFFFFFFFFFFFFFF0]
+    pand        m4,             [pb_000000000000000F]
+    por         m3,             m4              ;m3 = word[A, 0, 0, 1, 1, 2, 2, 3]
+
+    pmaddwd     m6,             m3,     [r2 - 1 * 16]
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    pmaddwd     m4,             m3,     [r3 - 6 * 16]
+
+    punpcklwd   m0,             m3
+    pand        m0,             [pb_0000000000000F0F]
+    mova        m6,             m3
+    pslldq      m6,             4
+    por         m0,             m6              ;m0 = word[B, A, A, 0, 0, 1, 1, 2]
+
+    pmaddwd     m6,             m0,     [r3 + 5 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    packuswb    m5,             m4
+    mova        [r0 + 288],     m5
+
+; mode 21
+
+    pmaddwd     m4,             m1,     [r3 + 8 * 16]
+    pmaddwd     m6,             m3,     [r2 + 7 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m5,             m3,     [r3 + 6 * 16]
+
+    pand        m0,             [pw_FFFFFFFFFFFFFFF0]
+    pand        m7,             [pb_000000000000000F]
+    por         m0,             m7              ;m0 = word[C, A, A, 0, 0, 1, 1, 2]
+
+    pmaddwd     m0,             [r2 + 5 * 16]
+    packssdw    m5,             m0
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    packuswb    m4,             m5
+    mova        [r0 + 304],     m4
+
+; mode 22
+
+    pmaddwd     m4,             m1,     [r2 - 4 * 16]
+    packssdw    m4,             m4
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    mova        m0,             [r3 + 5 * 16]
+    pmaddwd     m5,             m3,     [r2 + 2 * 16]
+    pmaddwd     m6,             m3,     m0
+
+    packssdw    m5,             m6
+    paddw       m5,             m2
+    psraw       m5,             5
+
+    packuswb    m4,             m5
+    movd        [r0 + 320],     m4
+    psrldq      m4,             8
+    movh        [r0 + 328],     m4
+
+; mode 23
+
+    pmaddwd     m4,             m1,     [r2 + 0 * 16]
+    pmaddwd     m5,             m1,     [r3 + 7 * 16]
+
+    packssdw    m4,             m5
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m6,             m1,     [r3 - 2 * 16]
+
+    pand        m3,             [pw_FFFFFFFFFFFFFFF0]
+    por         m3,             m7              ;m3 = word[C, 0, 0, 1, 1, 2, 2, 3]
+
+    pmaddwd     m3,             [r2 + 5 * 16]
+    packssdw    m6,             m3
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m4,             m6
+    mova        [r0 + 336],     m4
+
+; mode 24
+
+    pmaddwd     m4,             m1,     [r2 + 4 * 16]
+    pmaddwd     m5,             m1,     [r2 - 1 * 16]
+
+    packssdw    m4,             m5
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m6,             m1,     [r2 - 6 * 16]
+    pmaddwd     m0,             m1
+
+    packssdw    m6,             m0
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m4,             m6
+    mova        [r0 + 352],     m4
+
+; mode 25
+
+    pmaddwd     m4,             m1,     [r2 + 7 * 16]
+    pmaddwd     m5,             m1,     [r2 + 5 * 16]
+
+    packssdw    m4,             m5
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m6,             m1,     [r2 + 3 * 16]
+    pmaddwd     m1,             [r2 + 1 * 16]
+
+    packssdw    m6,             m1
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m4,             m6
+    mova        [r0 + 368],     m4
+
+; mode 27
+
+    movh        m0,             [r1 + 1]
+    pxor        m7,             m7
+    punpcklbw   m0,             m0
+    psrldq      m0,             1
+    movh        m1,             m0
+    psrldq      m0,             2
+    movh        m3,             m0
+    psrldq      m0,             2
+    punpcklbw   m1,             m7              ;m1 = word[1, 2, 2, 3, 3, 4, 4, 5]
+    punpcklbw   m3,             m7              ;m3 = word[2, 3, 3, 4, 4, 5, 5, 6]
+    punpcklbw   m0,             m7              ;m0 = word[3, 4, 4, 5, 5, 6, 6, 7]
+
+    mova        m7,             [r3 - 3 * 16]
+
+    pmaddwd     m4,             m1,     [r3 - 5 * 16]
+    pmaddwd     m5,             m1,     m7
+
+    packssdw    m4,             m5
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m6,             m1,     [r3 - 1 * 16]
+    pmaddwd     m5,             m1,     [r3 + 1 * 16]
+
+    packssdw    m6,             m5
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m4,             m6
+    mova        [r0 + 400],     m4
+
+; mode 28
+
+    pmaddwd     m4,             m1,     [r3 - 2 * 16]
+    pmaddwd     m5,             m1,     [r3 + 3 * 16]
+
+    packssdw    m4,             m5
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m6,             m1,     [r3 + 8 * 16]
+    pmaddwd     m5,             m1,     [r2 - 3 * 16]
+
+    packssdw    m6,             m5
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m4,             m6
+    mova        [r0 + 416],     m4
+
+; mode 29
+
+    pmaddwd     m4,             m1,     [r3 + 2 * 16]
+    pmaddwd     m6,             m1,     [r2 - 5 * 16]
+
+    packssdw    m4,             m6
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m6,             m1,     [r2 + 4 * 16]
+    pmaddwd     m5,             m3,     m7
+
+    packssdw    m6,             m5
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m4,             m6
+    mova        [r0 + 432],     m4
+
+; mode 30
+
+    pmaddwd     m4,             m1,     [r3 + 6 * 16]
+    pmaddwd     m5,             m1,     [r2 + 3 * 16]
+
+    packssdw    m4,             m5
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m6,             m3,     [r3 + 0 * 16]
+    pmaddwd     m5,             m3,     [r2 - 3 * 16]
+
+    packssdw    m6,             m5
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m4,             m6
+    mova        [r0 + 448],     m4
+    psrldq      m4,             4
+    movh        [r0 + 496],     m4              ;mode 33 row 0
+    psrldq      m4,             8
+    movd        [r0 + 500],     m4              ;mode 33 row 1
+
+; mode 31
+
+    pmaddwd     m4,             m1,     [r2 - 6 * 16]
+    pmaddwd     m5,             m3,     [r3 - 5 * 16]
+
+    packssdw    m4,             m5
+    paddw       m4,             m2
+    psraw       m4,             5
+
+    pmaddwd     m6,             m3,     [r2 - 4 * 16]
+    pmaddwd     m7,             m0
+
+    packssdw    m6,             m7
+    paddw       m6,             m2
+    psraw       m6,             5
+
+    packuswb    m4,             m6
+    mova        [r0 + 464],     m4
+
+; mode 32
+
+    pmaddwd     m1,             [r2 - 2 * 16]
+    pmaddwd     m5,             m3,     [r3 + 3 * 16]
+
+    packssdw    m1,             m5
+    paddw       m1,             m2
+    psraw       m1,             5
+
+    pmaddwd     m3,             [r2 + 8 * 16]
+    pmaddwd     m5,             m0,     [r2 - 3 * 16]
+    packssdw    m3,             m5
+    paddw       m3,             m2
+    psraw       m3,             5
+
+    packuswb    m1,             m3
+    mova        [r0 + 480],     m1
+
+; mode 33
+
+    pmaddwd     m0,             [r3 + 7 * 16]
+    pxor        m7,             m7
+    movh        m4,             [r1 + 4]
+    punpcklbw   m4,             m4
+    psrldq      m4,             1
+    punpcklbw   m4,             m7
+
+    pmaddwd     m4,             [r3 + 1 * 16]
+
+    packssdw    m0,             m4
+    paddw       m0,             m2
+    psraw       m0,             5
+
+    packuswb    m0,             m0
+    movh        [r0 + 504],     m0
+
+; mode 34
+
+    movh        m7,             [r1 + 2]
+    movd        [r0 + 512],     m7              ;byte[2, 3, 4, 5]
+
+    psrldq      m7,             1
+    movd        [r0 + 516],     m7              ;byte[3, 4, 5, 6]
+
+    psrldq      m7,             1
+    movd        [r0 + 520],     m7              ;byte[4, 5, 6, 7]
+
+    psrldq      m7,             1
+    movd        [r0 + 524],     m7              ;byte[5, 6, 7, 8]
+
+RET

x265_1.6.tar.gz/source/common/x86/ipfilter16.asm -> x265_1.7.tar.gz/source/common/x86/ipfilter16.asm Changed

@@ -113,10 +113,13 @@
                   times 8 dw 58, -10
                   times 8 dw 4, -1
 
+const interp8_hps_shuf,     dd 0, 4, 1, 5, 2, 6, 3, 7
+
 SECTION .text
 cextern pd_32
 cextern pw_pixel_max
 cextern pd_n32768
+cextern pw_2000
 
 ;------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_horiz_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -5525,65 +5528,1409 @@
     FILTER_VER_LUMA_SS 64, 16
     FILTER_VER_LUMA_SS 16, 64
 
-;--------------------------------------------------------------------------------------------------
-; void filterConvertPelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int width, int height)
-;--------------------------------------------------------------------------------------------------
-INIT_XMM sse2
-cglobal luma_p2s, 3, 7, 5
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_2xN 1
+INIT_XMM sse4
+cglobal filterPixelToShort_2x%1, 3, 6, 2
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r1 * 3]
+    lea        r5, [r3 * 3]
 
-    add         r1, r1
+    ; load constant
+    mova       m1, [pw_2000]
 
-    ; load width and height
-    mov         r3d, r3m
-    mov         r4d, r4m
+%rep %1/4
+    movd       m0, [r0]
+    movhps     m0, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m1
+
+    movd       [r2 + r3 * 0], m0
+    pextrd     [r2 + r3 * 1], m0, 2
+
+    movd       m0, [r0 + r1 * 2]
+    movhps     m0, [r0 + r4]
+    psllw      m0, 4
+    psubw      m0, m1
+
+    movd       [r2 + r3 * 2], m0
+    pextrd     [r2 + r5], m0, 2
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+%endrep
+    RET
+%endmacro
+P2S_H_2xN 4
+P2S_H_2xN 8
+P2S_H_2xN 16
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_4xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_4x%1, 3, 6, 2
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
 
     ; load constant
-    mova        m4, [tab_c_n8192]
+    mova       m1, [pw_2000]
 
-.loopH:
+%rep %1/4
+    movh       m0, [r0]
+    movhps     m0, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m1
+    movh       [r2 + r3 * 0], m0
+    movhps     [r2 + r3 * 1], m0
+
+    movh       m0, [r0 + r1 * 2]
+    movhps     m0, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m1
+    movh       [r2 + r3 * 2], m0
+    movhps     [r2 + r4], m0
 
-    xor         r5d, r5d
-.loopW:
-    lea         r6, [r0 + r5 * 2]
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+%endrep
+    RET
+%endmacro
+P2S_H_4xN 4
+P2S_H_4xN 8
+P2S_H_4xN 16
+P2S_H_4xN 32
 
-    movu        m0, [r6]
-    psllw       m0, 4
-    paddw       m0, m4
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_XMM ssse3
+cglobal filterPixelToShort_4x2, 3, 4, 1
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
 
-    movu        m1, [r6 + r1]
-    psllw       m1, 4
-    paddw       m1, m4
+    movh       m0, [r0]
+    movhps     m0, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, [pw_2000]
+    movh       [r2 + r3 * 0], m0
+    movhps     [r2 + r3 * 1], m0
 
-    movu        m2, [r6 + r1 * 2]
-    psllw       m2, 4
-    paddw       m2, m4
-
-    lea         r6, [r6 + r1 * 2]
-    movu        m3, [r6 + r1]
-    psllw       m3, 4
-    paddw       m3, m4
+    RET
 
-    add         r5, 8
-    cmp         r5, r3
-    jg          .width4
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
-    je          .nextH
-    jmp         .loopW
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_6xN 1
+INIT_XMM sse4
+cglobal filterPixelToShort_6x%1, 3, 7, 3
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
 
-.width4:
-    movh        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
-    movh        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
-    movh        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
-    movh        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
+    ; load height
+    mov        r6d, %1/4
 
-.nextH:
-    lea         r0, [r0 + r1 * 4]
-    add         r2, FENC_STRIDE * 8
+    ; load constant
+    mova       m2, [pw_2000]
 
-    sub         r4d, 4
-    jnz         .loopH
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movh       [r2 + r3 * 0], m0
+    pextrd     [r2 + r3 * 0 + 8], m0, 2
+    movh       [r2 + r3 * 1], m1
+    pextrd     [r2 + r3 * 1 + 8], m1, 2
+
+    movu       m0, [r0 + r1 * 2]
+    movu       m1, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movh       [r2 + r3 * 2], m0
+    pextrd     [r2 + r3 * 2 + 8], m0, 2
+    movh       [r2 + r4], m1
+    pextrd     [r2 + r4 + 8], m1, 2
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_6xN 8
+P2S_H_6xN 16
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_8xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_8x%1, 3, 7, 2
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m1, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    psllw      m0, 4
+    psubw      m0, m1
+    movu       [r2 + r3 * 0], m0
+
+    movu       m0, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m1
+    movu       [r2 + r3 * 1], m0
+
+    movu       m0, [r0 + r1 * 2]
+    psllw      m0, 4
+    psubw      m0, m1
+    movu       [r2 + r3 * 2], m0
+
+    movu       m0, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m1
+    movu       [r2 + r4], m0
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_8xN 8
+P2S_H_8xN 4
+P2S_H_8xN 16
+P2S_H_8xN 32
+P2S_H_8xN 12
+P2S_H_8xN 64
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_XMM ssse3
+cglobal filterPixelToShort_8x2, 3, 4, 2
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+
+    psllw      m0, 4
+    psubw      m0, [pw_2000]
+    psllw      m1, 4
+    psubw      m1, [pw_2000]
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_XMM ssse3
+cglobal filterPixelToShort_8x6, 3, 7, 4
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r1 * 3]
+    lea        r5, [r1 * 5]
+    lea        r6, [r3 * 3]
+
+    ; load constant
+    mova       m3, [pw_2000]
+
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    movu       m2, [r0 + r1 * 2]
+
+    psllw      m0, 4
+    psubw      m0, m3
+    psllw      m1, 4
+    psubw      m1, m3
+    psllw      m2, 4
+    psubw      m2, m3
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+    movu       [r2 + r3 * 2], m2
+
+    movu       m0, [r0 + r4]
+    movu       m1, [r0 + r1 * 4]
+    movu       m2, [r0 + r5 ]
+
+    psllw      m0, 4
+    psubw      m0, m3
+    psllw      m1, 4
+    psubw      m1, m3
+    psllw      m2, 4
+    psubw      m2, m3
+
+    movu       [r2 + r6], m0
+    movu       [r2 + r3 * 4], m1
+    lea        r2, [r2 + r3 * 4]
+    movu       [r2 + r3], m2
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_16xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_16x%1, 3, 7, 3
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m2, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+
+    movu       m0, [r0 + r1 * 2]
+    movu       m1, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2], m0
+    movu       [r2 + r4], m1
+
+    movu       m0, [r0 + 16]
+    movu       m1, [r0 + r1 + 16]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0 + 16], m0
+    movu       [r2 + r3 * 1 + 16], m1
+
+    movu       m0, [r0 + r1 * 2 + 16]
+    movu       m1, [r0 + r5 + 16]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2 + 16], m0
+    movu       [r2 + r4 + 16], m1
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_16xN 16
+P2S_H_16xN 4
+P2S_H_16xN 8
+P2S_H_16xN 12
+P2S_H_16xN 32
+P2S_H_16xN 64
+P2S_H_16xN 24
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_16xN_avx2 1
+INIT_YMM avx2
+cglobal filterPixelToShort_16x%1, 3, 7, 3
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m2, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+
+    movu       m0, [r0 + r1 * 2]
+    movu       m1, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2], m0
+    movu       [r2 + r4], m1
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_16xN_avx2 16
+P2S_H_16xN_avx2 4
+P2S_H_16xN_avx2 8
+P2S_H_16xN_avx2 12
+P2S_H_16xN_avx2 32
+P2S_H_16xN_avx2 64
+P2S_H_16xN_avx2 24
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_32xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_32x%1, 3, 7, 5
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m4, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    movu       m2, [r0 + r1 * 2]
+    movu       m3, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+    movu       [r2 + r3 * 2], m2
+    movu       [r2 + r4], m3
+
+    movu       m0, [r0 + 16]
+    movu       m1, [r0 + r1 + 16]
+    movu       m2, [r0 + r1 * 2 + 16]
+    movu       m3, [r0 + r5 + 16]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 16], m0
+    movu       [r2 + r3 * 1 + 16], m1
+    movu       [r2 + r3 * 2 + 16], m2
+    movu       [r2 + r4 + 16], m3
+
+    movu       m0, [r0 + 32]
+    movu       m1, [r0 + r1 + 32]
+    movu       m2, [r0 + r1 * 2 + 32]
+    movu       m3, [r0 + r5 + 32]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 32], m0
+    movu       [r2 + r3 * 1 + 32], m1
+    movu       [r2 + r3 * 2 + 32], m2
+    movu       [r2 + r4 + 32], m3
+
+    movu       m0, [r0 + 48]
+    movu       m1, [r0 + r1 + 48]
+    movu       m2, [r0 + r1 * 2 + 48]
+    movu       m3, [r0 + r5 + 48]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 48], m0
+    movu       [r2 + r3 * 1 + 48], m1
+    movu       [r2 + r3 * 2 + 48], m2
+    movu       [r2 + r4 + 48], m3
 
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_32xN 32
+P2S_H_32xN 8
+P2S_H_32xN 16
+P2S_H_32xN 24
+P2S_H_32xN 64
+P2S_H_32xN 48
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_32xN_avx2 1
+INIT_YMM avx2
+cglobal filterPixelToShort_32x%1, 3, 7, 3
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m2, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+
+    movu       m0, [r0 + r1 * 2]
+    movu       m1, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2], m0
+    movu       [r2 + r4], m1
+
+    movu       m0, [r0 + 32]
+    movu       m1, [r0 + r1 + 32]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0 + 32], m0
+    movu       [r2 + r3 * 1 + 32], m1
+
+    movu       m0, [r0 + r1 * 2 + 32]
+    movu       m1, [r0 + r5 + 32]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2 + 32], m0
+    movu       [r2 + r4 + 32], m1
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_32xN_avx2 32
+P2S_H_32xN_avx2 8
+P2S_H_32xN_avx2 16
+P2S_H_32xN_avx2 24
+P2S_H_32xN_avx2 64
+P2S_H_32xN_avx2 48
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_64xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_64x%1, 3, 7, 5
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m4, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    movu       m2, [r0 + r1 * 2]
+    movu       m3, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+    movu       [r2 + r3 * 2], m2
+    movu       [r2 + r4], m3
+
+    movu       m0, [r0 + 16]
+    movu       m1, [r0 + r1 + 16]
+    movu       m2, [r0 + r1 * 2 + 16]
+    movu       m3, [r0 + r5 + 16]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 16], m0
+    movu       [r2 + r3 * 1 + 16], m1
+    movu       [r2 + r3 * 2 + 16], m2
+    movu       [r2 + r4 + 16], m3
+
+    movu       m0, [r0 + 32]
+    movu       m1, [r0 + r1 + 32]
+    movu       m2, [r0 + r1 * 2 + 32]
+    movu       m3, [r0 + r5 + 32]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 32], m0
+    movu       [r2 + r3 * 1 + 32], m1
+    movu       [r2 + r3 * 2 + 32], m2
+    movu       [r2 + r4 + 32], m3
+
+    movu       m0, [r0 + 48]
+    movu       m1, [r0 + r1 + 48]
+    movu       m2, [r0 + r1 * 2 + 48]
+    movu       m3, [r0 + r5 + 48]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 48], m0
+    movu       [r2 + r3 * 1 + 48], m1
+    movu       [r2 + r3 * 2 + 48], m2
+    movu       [r2 + r4 + 48], m3
+
+    movu       m0, [r0 + 64]
+    movu       m1, [r0 + r1 + 64]
+    movu       m2, [r0 + r1 * 2 + 64]
+    movu       m3, [r0 + r5 + 64]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 64], m0
+    movu       [r2 + r3 * 1 + 64], m1
+    movu       [r2 + r3 * 2 + 64], m2
+    movu       [r2 + r4 + 64], m3
+
+    movu       m0, [r0 + 80]
+    movu       m1, [r0 + r1 + 80]
+    movu       m2, [r0 + r1 * 2 + 80]
+    movu       m3, [r0 + r5 + 80]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 80], m0
+    movu       [r2 + r3 * 1 + 80], m1
+    movu       [r2 + r3 * 2 + 80], m2
+    movu       [r2 + r4 + 80], m3
+
+    movu       m0, [r0 + 96]
+    movu       m1, [r0 + r1 + 96]
+    movu       m2, [r0 + r1 * 2 + 96]
+    movu       m3, [r0 + r5 + 96]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 96], m0
+    movu       [r2 + r3 * 1 + 96], m1
+    movu       [r2 + r3 * 2 + 96], m2
+    movu       [r2 + r4 + 96], m3
+
+    movu       m0, [r0 + 112]
+    movu       m1, [r0 + r1 + 112]
+    movu       m2, [r0 + r1 * 2 + 112]
+    movu       m3, [r0 + r5 + 112]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 112], m0
+    movu       [r2 + r3 * 1 + 112], m1
+    movu       [r2 + r3 * 2 + 112], m2
+    movu       [r2 + r4 + 112], m3
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_64xN 64
+P2S_H_64xN 16
+P2S_H_64xN 32
+P2S_H_64xN 48
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_64xN_avx2 1
+INIT_YMM avx2
+cglobal filterPixelToShort_64x%1, 3, 7, 3
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m2, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+
+    movu       m0, [r0 + r1 * 2]
+    movu       m1, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2], m0
+    movu       [r2 + r4], m1
+
+    movu       m0, [r0 + 32]
+    movu       m1, [r0 + r1 + 32]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0 + 32], m0
+    movu       [r2 + r3 * 1 + 32], m1
+
+    movu       m0, [r0 + r1 * 2 + 32]
+    movu       m1, [r0 + r5 + 32]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2 + 32], m0
+    movu       [r2 + r4 + 32], m1
+
+    movu       m0, [r0 + 64]
+    movu       m1, [r0 + r1 + 64]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0 + 64], m0
+    movu       [r2 + r3 * 1 + 64], m1
+
+    movu       m0, [r0 + r1 * 2 + 64]
+    movu       m1, [r0 + r5 + 64]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2 + 64], m0
+    movu       [r2 + r4 + 64], m1
+
+    movu       m0, [r0 + 96]
+    movu       m1, [r0 + r1 + 96]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0 + 96], m0
+    movu       [r2 + r3 * 1 + 96], m1
+
+    movu       m0, [r0 + r1 * 2 + 96]
+    movu       m1, [r0 + r5 + 96]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2 + 96], m0
+    movu       [r2 + r4 + 96], m1
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_64xN_avx2 64
+P2S_H_64xN_avx2 16
+P2S_H_64xN_avx2 32
+P2S_H_64xN_avx2 48
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_24xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_24x%1, 3, 7, 5
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m4, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    movu       m2, [r0 + r1 * 2]
+    movu       m3, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+    movu       [r2 + r3 * 2], m2
+    movu       [r2 + r4], m3
+
+    movu       m0, [r0 + 16]
+    movu       m1, [r0 + r1 + 16]
+    movu       m2, [r0 + r1 * 2 + 16]
+    movu       m3, [r0 + r5 + 16]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 16], m0
+    movu       [r2 + r3 * 1 + 16], m1
+    movu       [r2 + r3 * 2 + 16], m2
+    movu       [r2 + r4 + 16], m3
+
+    movu       m0, [r0 + 32]
+    movu       m1, [r0 + r1 + 32]
+    movu       m2, [r0 + r1 * 2 + 32]
+    movu       m3, [r0 + r5 + 32]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 32], m0
+    movu       [r2 + r3 * 1 + 32], m1
+    movu       [r2 + r3 * 2 + 32], m2
+    movu       [r2 + r4 + 32], m3
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_24xN 32
+P2S_H_24xN 64
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_24xN_avx2 1
+INIT_YMM avx2
+cglobal filterPixelToShort_24x%1, 3, 7, 3
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m2, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + 32]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 0 + 32], xm1
+
+    movu       m0, [r0 + r1]
+    movu       m1, [r0 + r1 + 32]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+    movu       [r2 + r3 * 1], m0
+    movu       [r2 + r3 * 1 + 32], xm1
+
+    movu       m0, [r0 + r1 * 2]
+    movu       m1, [r0 + r1 * 2 + 32]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+    movu       [r2 + r3 * 2], m0
+    movu       [r2 + r3 * 2 + 32], xm1
+
+    movu       m0, [r0 + r5]
+    movu       m1, [r0 + r5 + 32]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+    movu       [r2 + r4], m0
+    movu       [r2 + r4 + 32], xm1
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_24xN_avx2 32
+P2S_H_24xN_avx2 64
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_12xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_12x%1, 3, 7, 3
+    add        r1d, r1d
+    mov        r3d, r3m
+    add        r3d, r3d
+    lea        r4, [r3 * 3]
+    lea        r5, [r1 * 3]
+
+    ; load height
+    mov        r6d, %1/4
+
+    ; load constant
+    mova       m2, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+
+    movu       m0, [r0 + r1 * 2]
+    movu       m1, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m2
+    psllw      m1, 4
+    psubw      m1, m2
+
+    movu       [r2 + r3 * 2], m0
+    movu       [r2 + r4], m1
+
+    movh       m0, [r0 + 16]
+    movhps     m0, [r0 + r1 + 16]
+    psllw      m0, 4
+    psubw      m0, m2
+
+    movh       [r2 + r3 * 0 + 16], m0
+    movhps     [r2 + r3 * 1 + 16], m0
+
+    movh       m0, [r0 + r1 * 2 + 16]
+    movhps     m0, [r0 + r5 + 16]
+    psllw      m0, 4
+    psubw      m0, m2
+
+    movh       [r2 + r3 * 2 + 16], m0
+    movhps     [r2 + r4 + 16], m0
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    jnz        .loop
+    RET
+%endmacro
+P2S_H_12xN 16
+P2S_H_12xN 32
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_XMM ssse3
+cglobal filterPixelToShort_48x64, 3, 7, 5
+    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       m4, [pw_2000]
+
+.loop
+    movu       m0, [r0]
+    movu       m1, [r0 + r1]
+    movu       m2, [r0 + r1 * 2]
+    movu       m3, [r0 + r5]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0], m0
+    movu       [r2 + r3 * 1], m1
+    movu       [r2 + r3 * 2], m2
+    movu       [r2 + r4], m3
+
+    movu       m0, [r0 + 16]
+    movu       m1, [r0 + r1 + 16]
+    movu       m2, [r0 + r1 * 2 + 16]
+    movu       m3, [r0 + r5 + 16]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 16], m0
+    movu       [r2 + r3 * 1 + 16], m1
+    movu       [r2 + r3 * 2 + 16], m2
+    movu       [r2 + r4 + 16], m3
+
+    movu       m0, [r0 + 32]
+    movu       m1, [r0 + r1 + 32]
+    movu       m2, [r0 + r1 * 2 + 32]
+    movu       m3, [r0 + r5 + 32]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 32], m0
+    movu       [r2 + r3 * 1 + 32], m1
+    movu       [r2 + r3 * 2 + 32], m2
+    movu       [r2 + r4 + 32], m3
+
+    movu       m0, [r0 + 48]
+    movu       m1, [r0 + r1 + 48]
+    movu       m2, [r0 + r1 * 2 + 48]
+    movu       m3, [r0 + r5 + 48]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 48], m0
+    movu       [r2 + r3 * 1 + 48], m1
+    movu       [r2 + r3 * 2 + 48], m2
+    movu       [r2 + r4 + 48], m3
+
+    movu       m0, [r0 + 64]
+    movu       m1, [r0 + r1 + 64]
+    movu       m2, [r0 + r1 * 2 + 64]
+    movu       m3, [r0 + r5 + 64]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 64], m0
+    movu       [r2 + r3 * 1 + 64], m1
+    movu       [r2 + r3 * 2 + 64], m2
+    movu       [r2 + r4 + 64], m3
+
+    movu       m0, [r0 + 80]
+    movu       m1, [r0 + r1 + 80]
+    movu       m2, [r0 + r1 * 2 + 80]
+    movu       m3, [r0 + r5 + 80]
+    psllw      m0, 4
+    psubw      m0, m4
+    psllw      m1, 4
+    psubw      m1, m4
+    psllw      m2, 4
+    psubw      m2, m4
+    psllw      m3, 4
+    psubw      m3, m4
+
+    movu       [r2 + r3 * 0 + 80], m0
+    movu       [r2 + r3 * 1 + 80], m1
+    movu       [r2 + r3 * 2 + 80], m2
+    movu       [r2 + r4 + 80], m3
+
+    lea        r0, [r0 + r1 * 4]
+    lea        r2, [r2 + r3 * 4]
+
+    dec        r6d
+    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, 4
+    psubw      m0, m3
+    psllw      m1, 4
+    psubw      m1, m3
+    psllw      m2, 4
+    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, 4
+    psubw      m0, m3
+    psllw      m1, 4
+    psubw      m1, m3
+    psllw      m2, 4
+    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, 4
+    psubw      m0, m3
+    psllw      m1, 4
+    psubw      m1, m3
+    psllw      m2, 4
+    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, 4
+    psubw      m0, m3
+    psllw      m1, 4
+    psubw      m1, m3
+    psllw      m2, 4
+    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,                [pd_n32768]
+
+    ; 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,               2
+    psrad                       xm4,               2
+    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,               2
+    psrad                       xm4,               2
+    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

x265_1.6.tar.gz/source/common/x86/ipfilter8.asm -> x265_1.7.tar.gz/source/common/x86/ipfilter8.asm Changed

@@ -27,269 +27,269 @@
 %include "x86util.asm"
 
 SECTION_RODATA 32
-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 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
 
-ALIGN 32
 const interp4_vpp_shuf, times 2 db 0, 4, 1, 5, 2, 6, 3, 7, 8, 12, 9, 13, 10, 14, 11, 15
 
-ALIGN 32
 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
 
-ALIGN 32
 const interp4_vpp_shuf1, dd 0, 1, 1, 2, 2, 3, 3, 4
                          dd 2, 3, 3, 4, 4, 5, 5, 6
 
-ALIGN 32
 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
-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
-
-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
-
-tab_Cm:    db 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3
-
-tab_c_526336:   times 4 dd 8192*64+2048
-
-pd_526336:      times 8 dd 8192*64+2048
-
-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
-ALIGN 32
-tab_ChromaCoeff_V: times 8 db 0, 64
-                   times 8 db 0,  0
+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
 
-                   times 8 db -2, 58
-                   times 8 db 10, -2
+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
 
-                   times 8 db -4, 54
-                   times 8 db 16, -2
+const tab_Cm,    db 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3, 0, 2, 1, 3
 
-                   times 8 db -6, 46
-                   times 8 db 28, -4
+const pd_526336, times 8 dd 8192*64+2048
 
-                   times 8 db -4, 36
-                   times 8 db 36, -4
+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
 
-                   times 8 db -4, 28
-                   times 8 db 46, -6
+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
 
-                   times 8 db -2, 16
-                   times 8 db 54, -4
+const tab_ChromaCoeff_V, times 8 db 0, 64
+                         times 8 db 0,  0
 
-                   times 8 db -2, 10
-                   times 8 db 58, -2
+                         times 8 db -2, 58
+                         times 8 db 10, -2
 
-tab_ChromaCoeffV: times 4 dw 0, 64
-                  times 4 dw 0, 0
+                         times 8 db -4, 54
+                         times 8 db 16, -2
 
-                  times 4 dw -2, 58
-                  times 4 dw 10, -2
+                         times 8 db -6, 46
+                         times 8 db 28, -4
 
-                  times 4 dw -4, 54
-                  times 4 dw 16, -2
+                         times 8 db -4, 36
+                         times 8 db 36, -4
 
-                  times 4 dw -6, 46 
-                  times 4 dw 28, -4
+                         times 8 db -4, 28
+                         times 8 db 46, -6
 
-                  times 4 dw -4, 36
-                  times 4 dw 36, -4
+                         times 8 db -2, 16
+                         times 8 db 54, -4
 
-                  times 4 dw -4, 28
-                  times 4 dw 46, -6
+                         times 8 db -2, 10
+                         times 8 db 58, -2
 
-                  times 4 dw -2, 16
-                  times 4 dw 54, -4
+const tab_ChromaCoeffV, times 4 dw 0, 64
+                        times 4 dw 0, 0
 
-                  times 4 dw -2, 10
-                  times 4 dw 58, -2
+                        times 4 dw -2, 58
+                        times 4 dw 10, -2
 
-ALIGN 32
-pw_ChromaCoeffV:  times 8 dw 0, 64
-                  times 8 dw 0, 0
+                        times 4 dw -4, 54
+                        times 4 dw 16, -2
 
-                  times 8 dw -2, 58
-                  times 8 dw 10, -2
+                        times 4 dw -6, 46
+                        times 4 dw 28, -4
 
-                  times 8 dw -4, 54
-                  times 8 dw 16, -2
+                        times 4 dw -4, 36
+                        times 4 dw 36, -4
 
-                  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:   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
-
-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
+                        times 4 dw -4, 28
+                        times 4 dw 46, -6
 
-ALIGN 32
-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
-
-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
-
-tab_LumaCoeffVer: times 8 db 0, 0
-                  times 8 db 0, 64
-                  times 8 db 0, 0
-                  times 8 db 0, 0
-
-                  times 8 db -1, 4
-                  times 8 db -10, 58
-                  times 8 db 17, -5
-                  times 8 db 1, 0
-
-                  times 8 db -1, 4
-                  times 8 db -11, 40
-                  times 8 db 40, -11
-                  times 8 db 4, -1
-
-                  times 8 db 0, 1
-                  times 8 db -5, 17
-                  times 8 db 58, -10
-                  times 8 db 4, -1
+                        times 4 dw -2, 16
+                        times 4 dw 54, -4
 
-ALIGN 32
-tab_LumaCoeffVer_32: 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
+                        times 4 dw -2, 10
+                        times 4 dw 58, -2
 
-ALIGN 32
-tab_ChromaCoeffVer_32: times 16 db 0, 64
-                       times 16 db 0, 0
+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
+                       db   0, 1, -5,  17,  58, -10, 4, -1
+
+const 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 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 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 16 db -2, 58
-                       times 16 db 10, -2
+                        times 8 dw -1, 4
+                        times 8 dw -11, 40
+                        times 8 dw 40, -11
+                        times 8 dw 4, -1
 
-                       times 16 db -4, 54
-                       times 16 db 16, -2
+                        times 8 dw 0, 1
+                        times 8 dw -5, 17
+                        times 8 dw 58, -10
+                        times 8 dw 4, -1
 
-                       times 16 db -6, 46
-                       times 16 db 28, -4
+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 -4, 36
-                       times 16 db 36, -4
+                        times 16 db -1, 4
+                        times 16 db -10, 58
+                        times 16 db 17, -5
+                        times 16 db 1, 0
 
-                       times 16 db -4, 28
-                       times 16 db 46, -6
+                        times 16 db -1, 4
+                        times 16 db -11, 40
+                        times 16 db 40, -11
+                        times 16 db 4, -1
 
-                       times 16 db -2, 16
-                       times 16 db 54, -4
+                        times 16 db 0, 1
+                        times 16 db -5, 17
+                        times 16 db 58, -10
+                        times 16 db 4, -1
 
-                       times 16 db -2, 10
-                       times 16 db 58, -2
+const tab_LumaCoeffVer, times 8 db 0, 0
+                        times 8 db 0, 64
+                        times 8 db 0, 0
+                        times 8 db 0, 0
 
-tab_c_64_n64:   times 8 db 64, -64
+                        times 8 db -1, 4
+                        times 8 db -10, 58
+                        times 8 db 17, -5
+                        times 8 db 1, 0
+
+                        times 8 db -1, 4
+                        times 8 db -11, 40
+                        times 8 db 40, -11
+                        times 8 db 4, -1
+
+                        times 8 db 0, 1
+                        times 8 db -5, 17
+                        times 8 db 58, -10
+                        times 8 db 4, -1
+
+const tab_LumaCoeffVer_32,  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_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
 
-ALIGN 32
-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_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
 
-ALIGN 32
-interp4_hpp_shuf: times 2 db 0, 1, 2, 3, 1, 2, 3, 4, 8, 9, 10, 11, 9, 10, 11, 12
+const interp4_hpp_shuf,     times 2 db 0, 1, 2, 3, 1, 2, 3, 4, 8, 9, 10, 11, 9, 10, 11, 12
 
-ALIGN 32
-interp8_hps_shuf: dd 0, 4, 1, 5, 2, 6, 3, 7
+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
@@ -298,9 +298,276 @@
 
 cextern pb_128
 cextern pw_1
+cextern pw_32
 cextern pw_512
 cextern pw_2000
 
+%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_2x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_2x4, 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
+
+    FILTER_H4_w2_2_sse2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+    FILTER_H4_w2_2_sse2
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_2x8, 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 4
+    FILTER_H4_w2_2_sse2
+%if x < 4
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endif
+%assign x x+1
+%endrep
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_2x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_2x16, 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 8
+    FILTER_H4_w2_2_sse2
+%if x < 8
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endif
+%assign x x+1
+%endrep
+
+    RET
+
+%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_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_4x2, 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
+
+    FILTER_H4_w4_2_sse2
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_4x4, 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
+
+    FILTER_H4_w4_2_sse2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+    FILTER_H4_w4_2_sse2
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_4x8, 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 4
+    FILTER_H4_w4_2_sse2
+%if x < 4
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endif
+%assign x x+1
+%endrep
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_4x16, 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 8
+    FILTER_H4_w4_2_sse2
+%if x < 8
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endif
+%assign x x+1
+%endrep
+
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_4x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_4x32, 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 16
+    FILTER_H4_w4_2_sse2
+%if x < 16
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+%endif
+%assign x x+1
+%endrep
+
+    RET
+
 %macro FILTER_H4_w2_2 3
     movh        %2, [srcq - 1]
     pshufb      %2, %2, Tm0
@@ -317,6 +584,1298 @@
     mov         [dstq + dststrideq], r4w
 %endmacro
 
+%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
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro IPFILTER_CHROMA_W_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_W_sse3 16,  4
+    IPFILTER_CHROMA_W_sse3 16,  8
+    IPFILTER_CHROMA_W_sse3 16, 12
+    IPFILTER_CHROMA_W_sse3 16, 16
+    IPFILTER_CHROMA_W_sse3 16, 32
+    IPFILTER_CHROMA_W_sse3 32,  8
+    IPFILTER_CHROMA_W_sse3 32, 16
+    IPFILTER_CHROMA_W_sse3 32, 24
+    IPFILTER_CHROMA_W_sse3 24, 32
+    IPFILTER_CHROMA_W_sse3 32, 32
+
+    IPFILTER_CHROMA_W_sse3 16, 24
+    IPFILTER_CHROMA_W_sse3 16, 64
+    IPFILTER_CHROMA_W_sse3 32, 48
+    IPFILTER_CHROMA_W_sse3 24, 64
+    IPFILTER_CHROMA_W_sse3 32, 64
+
+    IPFILTER_CHROMA_W_sse3 64, 64
+    IPFILTER_CHROMA_W_sse3 64, 32
+    IPFILTER_CHROMA_W_sse3 64, 48
+    IPFILTER_CHROMA_W_sse3 48, 64
+    IPFILTER_CHROMA_W_sse3 64, 16
+
+%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, [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  WORD_TO_DOUBLE 1
+%if ARCH_X86_64
+    punpcklbw   %1,     m8
+%else
+    punpcklbw   %1,     %1
+    psrlw       %1,     8
+%endif
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_2xn(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W2_H4_sse2 1
+INIT_XMM sse2
+%if ARCH_X86_64
+cglobal interp_4tap_vert_pp_2x%1, 4, 6, 9
+    pxor        m8,        m8
+%else
+cglobal interp_4tap_vert_pp_2x%1, 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
+
+    punpcklqdq  m0,        m0
+    mova        m1,        [pw_32]
+    lea         r5,        [3 * r1]
+
+%assign x 1
+%rep %1/4
+    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
+    psrld       m2,        16
+
+    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
+    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
+
+%if x < %1/4
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%assign x x+1
+%endrep
+    RET
+
+%endmacro
+
+    FILTER_V4_W2_H4_sse2 4
+    FILTER_V4_W2_H4_sse2 8
+    FILTER_V4_W2_H4_sse2 16
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse2
+cglobal interp_4tap_vert_pp_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
+    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
+    RET
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W4_H4_sse2 1
+INIT_XMM sse2
+%if ARCH_X86_64
+cglobal interp_4tap_vert_pp_4x%1, 4, 6, 9
+    pxor        m8,        m8
+%else
+cglobal interp_4tap_vert_pp_4x%1, 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
+
+    mova        m1,        [pw_32]
+    lea         r5,        [3 * r1]
+    punpcklqdq  m0,        m0
+
+%assign x 1
+%rep %1/4
+    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
+    psrld       m2,        16
+    psrld       m3,        16
+    packssdw    m2,        m3
+
+    paddw       m2,        m1
+    psraw       m2,        6
+
+    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
+    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
+    lea         r2,        [r2 + 2 * r3]
+    psrldq      m2,        4
+    movd        [r2],      m2
+    psrldq      m2,        4
+    movd        [r2 + r3], m2
+
+%if x < %1/4
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%assign x x+1
+%endrep
+    RET
+%endmacro
+
+    FILTER_V4_W4_H4_sse2 4
+    FILTER_V4_W4_H4_sse2 8
+    FILTER_V4_W4_H4_sse2 16
+    FILTER_V4_W4_H4_sse2 32
+
+;-----------------------------------------------------------------------------
+;void interp_4tap_vert_pp_6x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W6_H4_sse2 1
+INIT_XMM sse2
+cglobal interp_4tap_vert_pp_6x%1, 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
+
+    mova        m4,        [pw_32]
+    lea         r5,        [3 * r1]
+
+%assign x 1
+%rep %1/4
+    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
+
+    paddw       m0,        m4
+    psraw       m0,        6
+    packuswb    m0,        m0
+    movd        [r2],      m0
+    pextrw      r6d,       m0,        2
+    mov         [r2 + 4],  r6w
+
+    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
+
+    paddw       m1,        m4
+    psraw       m1,        6
+    packuswb    m1,        m1
+    movd        [r2 + r3], m1
+    pextrw      r6d,       m1,        2
+    mov         [r2 + r3 + 4], r6w
+    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
+
+    paddw       m2,        m4
+    psraw       m2,        6
+    packuswb    m2,        m2
+    lea         r2,        [r2 + 2 * r3]
+    movd        [r2],      m2
+    pextrw      r6d,       m2,    2
+    mov         [r2 + 4],  r6w
+
+    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
+
+    paddw       m3,        m4
+    psraw       m3,        6
+    packuswb    m3,        m3
+
+    movd        [r2 + r3], m3
+    pextrw      r6d,    m3,    2
+    mov         [r2 + r3 + 4], r6w
+
+%if x < %1/4
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%assign x x+1
+%endrep
+    RET
+
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W6_H4_sse2 8
+    FILTER_V4_W6_H4_sse2 16
+%endif
+
+;-----------------------------------------------------------------------------
+; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+%macro FILTER_V4_W8_sse2 1
+INIT_XMM sse2
+cglobal interp_4tap_vert_pp_8x%1, 4, 7, 12
+
+    mov         r4d,       r4m
+    sub         r0,        r1
+    shl         r4d,       5
+    pxor        m9,        m9
+    mova        m4,        [pw_32]
+
+%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
+
+    paddw       m0,        m4
+    psraw       m0,        6
+
+    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
+
+    paddw       m1,        m4
+    psraw       m1,        6
+    packuswb    m1,        m0
+
+    movhps      [r2],      m1
+    movh        [r2 + r3], m1
+%if %1 == 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
+
+    paddw       m2,        m4
+    psraw       m2,        6
+
+    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
+
+    paddw       m3,        m4
+    psraw       m3,        6
+    packuswb    m3,        m2
+
+    movhps      [r2 + 2 * r3], m3
+    lea         r5,        [r2 + 2 * r3]
+    movh        [r5 + r3], m3
+%if %1 == 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
+
+    paddw       m11,        m4
+    psraw       m11,        6
+
+    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
+
+    paddw       m1,        m4
+    psraw       m1,        6
+    packuswb    m1,        m11
+
+    movhps      [r2 + 4 * r3], m1
+    lea         r5,        [r2 + 4 * r3]
+    movh        [r5 + r3], m1
+%if %1 == 6
+    RET
+
+%else
+  %error INVALID macro argument, only 2, 4 or 6!
+%endif
+%endif
+%endif
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W8_sse2 2
+    FILTER_V4_W8_sse2 4
+    FILTER_V4_W8_sse2 6
+%endif
+
+;-----------------------------------------------------------------------------
+; 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_sse2 2
+INIT_XMM sse2
+cglobal interp_4tap_vert_pp_%1x%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
+
+    mova        m4,        [pw_32]
+    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
+    paddw       m0,        m4
+    psraw       m0,        6
+
+    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
+    paddw       m1,        m4
+    psraw       m1,        6
+
+    packuswb    m0,        m1
+    movh        [r2],      m0
+    movhps      [r2 + r3], m0
+
+    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
+    paddw       m2,        m4
+    psraw       m2,        6
+
+    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
+    paddw       m3,        m4
+    psraw       m3,        6
+
+    packuswb    m2,        m3
+    lea         r2,        [r2 + 2 * r3]
+    movh        [r2],      m2
+    movhps      [r2 + r3], m2
+%if x < %2/4
+    lea         r2,        [r2 + 2 * r3]
+%endif
+%endrep
+    RET
+%endmacro
+
+%if ARCH_X86_64
+    FILTER_V4_W8_H8_H16_H32_sse2 8,  8
+    FILTER_V4_W8_H8_H16_H32_sse2 8, 16
+    FILTER_V4_W8_H8_H16_H32_sse2 8, 32
+
+    FILTER_V4_W8_H8_H16_H32_sse2 8, 12
+    FILTER_V4_W8_H8_H16_H32_sse2 8, 64
+%endif
+
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_horiz_pp_2x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-----------------------------------------------------------------------------
@@ -328,26 +1887,26 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
+    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]
+    FILTER_H4_w2_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
 %endrep
 
-RET
+    RET
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_horiz_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -360,26 +1919,26 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
+    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]
+    FILTER_H4_w2_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
 %endrep
 
-RET
+    RET
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_horiz_pp_2x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -392,29 +1951,29 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [tab_Tm]
 
-mov         r5d,        16/2
+    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
+    FILTER_H4_w2_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+    dec         r5d
+    jnz         .loop
 
-RET
+    RET
 
 %macro FILTER_H4_w4_2 3
     movh        %2, [srcq - 1]
@@ -442,22 +2001,22 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [tab_Tm]
 
-FILTER_H4_w4_2   t0, t1, t2
+    FILTER_H4_w4_2   t0, t1, t2
 
-RET
+    RET
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_horiz_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -470,26 +2029,26 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
+    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]
+    FILTER_H4_w4_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
 %endrep
 
-RET
+    RET
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_horiz_pp_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -502,26 +2061,26 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
+    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]
+    FILTER_H4_w4_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
 %endrep
 
-RET
+    RET
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_horiz_pp_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -534,26 +2093,26 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
+    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]
+    FILTER_H4_w4_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
 %endrep
 
-RET
+    RET
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_horiz_pp_4x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -566,29 +2125,29 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
+    pshufd      coef2,       coef2,      0
+    mova        t2,          [pw_512]
+    mova        Tm0,         [tab_Tm]
 
-mov         r5d,        32/2
+    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
+    FILTER_H4_w4_2   t0, t1, t2
+    lea         srcq,       [srcq + srcstrideq * 2]
+    lea         dstq,       [dstq + dststrideq * 2]
+    dec         r5d
+    jnz         .loop
 
-RET
+    RET
 
 ALIGN 32
 const interp_4tap_8x8_horiz_shuf,   dd 0, 4, 1, 5, 2, 6, 3, 7
@@ -764,47 +2323,47 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,        r4m
+    mov         r4d,        r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-mov           r5d,       %2
+    mov           r5d,       %2
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
-mova        Tm1,         [tab_Tm + 16]
+    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
+    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
+    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)
@@ -820,55 +2379,55 @@
 %define t1          m1
 %define t0          m0
 
-mov         r4d,         r4m
+    mov         r4d,         r4m
 
 %ifdef PIC
-lea         r5,          [tab_ChromaCoeff]
-movd        coef2,       [r5 + r4 * 4]
+    lea         r5,          [tab_ChromaCoeff]
+    movd        coef2,       [r5 + r4 * 4]
 %else
-movd        coef2,       [tab_ChromaCoeff + r4 * 4]
+    movd        coef2,       [tab_ChromaCoeff + r4 * 4]
 %endif
 
-mov         r5d,          %2
+    mov         r5d,          %2
 
-pshufd      coef2,       coef2,      0
-mova        t2,          [pw_512]
-mova        Tm0,         [tab_Tm]
-mova        Tm1,         [tab_Tm + 16]
+    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
+    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
@@ -918,7 +2477,7 @@
 %endif
     punpcklqdq  m3, m3
 
-%ifidn %3, pp 
+%ifidn %3, pp
     mova      m2, [pw_512]
 %else
     mova      m2, [pw_2000]
@@ -937,7 +2496,7 @@
 .loopH:
     xor       r5, r5
 %rep %1 / 8
-  %ifidn %3, pp 
+  %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]
@@ -949,7 +2508,7 @@
 
 %rep (%1 % 8) / 4
     FILTER_H8_W4  m0, m1
-  %ifidn %3, pp 
+  %ifidn %3, pp
     pmulhrsw  m1, m2
     packuswb  m1, m1
     movd      [r2 + r5], m1
@@ -1120,8 +2679,8 @@
 %endif
 %endmacro
 
-FILTER_HORIZ_LUMA_AVX2_4xN 8
-FILTER_HORIZ_LUMA_AVX2_4xN 16
+    FILTER_HORIZ_LUMA_AVX2_4xN 8
+    FILTER_HORIZ_LUMA_AVX2_4xN 16
 
 INIT_YMM avx2
 cglobal interp_8tap_horiz_pp_8x4, 4, 6, 7
@@ -1271,9 +2830,9 @@
     RET
 %endmacro
 
-IPFILTER_LUMA_AVX2_8xN 8, 8
-IPFILTER_LUMA_AVX2_8xN 8, 16
-IPFILTER_LUMA_AVX2_8xN 8, 32
+    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
@@ -1306,7 +2865,7 @@
     pmaddubsw         m5,         m1
     paddw             m4,         m5
     pmaddwd           m4,         m7
-    vbroadcasti128    m5,         [r0 + 8]                    ; second 8 elements in Row0 
+    vbroadcasti128    m5,         [r0 + 8]                    ; second 8 elements in Row0
     pshufb            m6,         m5,     m3
     pshufb            m5,         [tab_Tm]
     pmaddubsw         m5,         m0
@@ -1322,7 +2881,7 @@
     pmaddubsw         m5,         m1
     paddw             m2,         m5
     pmaddwd           m2,         m7
-    vbroadcasti128    m5,         [r0 + r1 + 8]                    ; second 8 elements in Row0 
+    vbroadcasti128    m5,         [r0 + r1 + 8]                    ; second 8 elements in Row0
     pshufb            m6,         m5,     m3
     pshufb            m5,         [tab_Tm]
     pmaddubsw         m5,         m0
@@ -1617,7 +3176,7 @@
     jnz               .loop
     RET
 
-INIT_YMM avx2 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_pp_4x4, 4,6,6
     mov             r4d, r4m
 
@@ -1665,7 +3224,7 @@
     pextrd            [r2+r0],      xm3,     3
     RET
 
-INIT_YMM avx2 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_pp_2x4, 4, 6, 3
     mov               r4d,           r4m
 
@@ -1698,7 +3257,7 @@
     pextrw            [r2 + r4],     xm1,         3
     RET
 
-INIT_YMM avx2 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_pp_2x8, 4, 6, 6
     mov               r4d,           r4m
 
@@ -1941,7 +3500,7 @@
 
     IPFILTER_LUMA_AVX2 16, 4
     IPFILTER_LUMA_AVX2 16, 8
-    IPFILTER_LUMA_AVX2 16, 12 
+    IPFILTER_LUMA_AVX2 16, 12
     IPFILTER_LUMA_AVX2 16, 16
     IPFILTER_LUMA_AVX2 16, 32
     IPFILTER_LUMA_AVX2 16, 64
@@ -2144,6 +3703,108 @@
     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)
 ;-----------------------------------------------------------------------------------------------------------------------------
 
@@ -2230,7 +3891,7 @@
     pshufb                      m4,                m1
     pmaddubsw                   m4,                m0
     phaddw                      m4,                m4                           ; DWORD [R3D R3C R2D R2C R3B R3A R2B R2A]
-    phaddw                      m3,                m4 
+    phaddw                      m3,                m4
 
     vpermd                      m3,                m5,            m3            ; m5 don't broken in above
     psubw                       m3,                m2
@@ -2312,7 +3973,7 @@
     lea                         r2,         [r2 + r3 * 2]                   ; first loop dst ->5th row(i.e 4)
     sub                         r5d,        2
     jg                         .loop
-    jz                         .end             
+    jz                         .end
 
     ; last row
     movu                        xm1,        [r0]
@@ -2334,10 +3995,10 @@
 %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
+    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
@@ -2399,17 +4060,17 @@
     dec                         r9d
     jnz                         .label
 
-RET
+    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
+    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
 
 
 ;--------------------------------------------------------------------------------------------------------------
@@ -2460,27 +4121,27 @@
     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
+    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)
@@ -2547,10 +4208,10 @@
 
 ; Round and Saturate
 %macro FILTER_HV8_END 4 ; output in [1, 3]
-    paddd       %1, [tab_c_526336]
-    paddd       %2, [tab_c_526336]
-    paddd       %3, [tab_c_526336]
-    paddd       %4, [tab_c_526336]
+    paddd       %1, [pd_526336]
+    paddd       %2, [pd_526336]
+    paddd       %3, [pd_526336]
+    paddd       %4, [pd_526336]
     psrad       %1, 12
     psrad       %2, 12
     psrad       %3, 12
@@ -2565,7 +4226,7 @@
 ;-----------------------------------------------------------------------------
 ; void interp_8tap_hv_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int idxX, int idxY)
 ;-----------------------------------------------------------------------------
-INIT_XMM sse4
+INIT_XMM ssse3
 cglobal interp_8tap_hv_pp_8x8, 4, 7, 8, 0-15*16
 %define coef        m7
 %define stk_buf     rsp
@@ -2640,76 +4301,148 @@
     RET
 
 ;-----------------------------------------------------------------------------
+; void interp_8tap_hv_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int idxX, int idxY)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_8tap_hv_pp_8x8, 4, 7, 8, 0-15*16
+    mov         r4d,        r4m
+    mov         r5d,        r5m
+    add         r4d,        r4d
+    pxor        m6,         m6
+
+%ifdef PIC
+    lea         r6,         [tabw_LumaCoeff]
+    mova        m3,         [r6 + r4 * 8]
+%else
+    mova        m3,         [tabw_LumaCoeff + r4 * 8]
+%endif
+
+    ; move to row -3
+    lea         r6,         [r1 + r1 * 2]
+    sub         r0,         r6
+
+    mov         r4,         rsp
+
+%assign x 0     ;needed for FILTER_H8_W8_sse2 macro
+%assign y 1
+%rep 15
+    FILTER_H8_W8_sse2
+    psubw       m1,         [pw_2000]
+    mova        [r4],       m1
+
+%if y < 15
+    add         r0,         r1
+    add         r4,         16
+%endif
+%assign y y+1
+%endrep
+
+    ; ready to phase V
+    ; Here all of mN is free
+
+    ; load coeff table
+    shl         r5,         6
+    lea         r6,         [tab_LumaCoeffV]
+    lea         r5,         [r5 + r6]
+
+    ; load intermedia buffer
+    mov         r0,         rsp
+
+    ; register mapping
+    ; r0 - src
+    ; r5 - coeff
+
+    ; let's go
+%assign y 1
+%rep 4
+    FILTER_HV8_START    m1, m2, m3, m4, m0,             0, 0
+    FILTER_HV8_MID      m6, m2, m3, m4, m0, m1, m7, m5, 3, 1
+    FILTER_HV8_MID      m5, m6, m3, m4, m0, m1, m7, m2, 5, 2
+    FILTER_HV8_MID      m6, m5, m3, m4, m0, m1, m7, m2, 7, 3
+    FILTER_HV8_END      m3, m0, m4, m1
+
+    movh        [r2],       m3
+    movhps      [r2 + r3],  m3
+
+%if y < 4
+    lea         r0,         [r0 + 16 * 2]
+    lea         r2,         [r2 + r3 * 2]
+%endif
+%assign y y+1
+%endrep
+    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
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
-lea         r4,        [r1 * 3]
-lea         r5,        [r0 + 4 * r1]
-pshufb      m0,        [tab_Cm]
-mova        m1,        [pw_512]
+    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]
+    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
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklbw   m2,        m6
 
-pmaddubsw   m2,        m0
+    pmaddubsw   m2,        m0
 
-movd        m6,        [r5]
+    movd        m6,        [r5]
 
-punpcklbw   m3,        m4
-punpcklbw   m7,        m5,        m6
-punpcklbw   m3,        m7
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklbw   m3,        m7
 
-pmaddubsw   m3,        m0
+    pmaddubsw   m3,        m0
 
-phaddw      m2,        m3
+    phaddw      m2,        m3
 
-pmulhrsw    m2,        m1
+    pmulhrsw    m2,        m1
 
-movd        m7,        [r5 + r1]
+    movd        m7,        [r5 + r1]
 
-punpcklbw   m4,        m5
-punpcklbw   m3,        m6,        m7
-punpcklbw   m4,        m3
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklbw   m4,        m3
 
-pmaddubsw   m4,        m0
+    pmaddubsw   m4,        m0
 
-movd        m3,        [r5 + 2 * r1]
+    movd        m3,        [r5 + 2 * r1]
 
-punpcklbw   m5,        m6
-punpcklbw   m7,        m3
-punpcklbw   m5,        m7
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklbw   m5,        m7
 
-pmaddubsw   m5,        m0
+    pmaddubsw   m5,        m0
 
-phaddw      m4,        m5
+    phaddw      m4,        m5
 
-pmulhrsw    m4,        m1
-packuswb    m2,        m4
+    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
+    pextrw      [r2],      m2, 0
+    pextrw      [r2 + r3], m2, 2
+    lea         r2,        [r2 + 2 * r3]
+    pextrw      [r2],      m2, 4
+    pextrw      [r2 + r3], m2, 6
 
-RET
+    RET
 
 %macro FILTER_VER_CHROMA_AVX2_2x4 1
 INIT_YMM avx2
@@ -2762,8 +4495,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_2x4 pp
-FILTER_VER_CHROMA_AVX2_2x4 ps
+    FILTER_VER_CHROMA_AVX2_2x4 pp
+    FILTER_VER_CHROMA_AVX2_2x4 ps
 
 %macro FILTER_VER_CHROMA_AVX2_2x8 1
 INIT_YMM avx2
@@ -2834,8 +4567,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_2x8 pp
-FILTER_VER_CHROMA_AVX2_2x8 ps
+    FILTER_VER_CHROMA_AVX2_2x8 pp
+    FILTER_VER_CHROMA_AVX2_2x8 ps
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_vert_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -2844,85 +4577,85 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_2x%2, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m0,        [tab_Cm]
+    pshufb      m0,        [tab_Cm]
 
-mova        m1,        [pw_512]
+    mova        m1,        [pw_512]
 
-mov         r4d,       %2
-lea         r5,        [3 * r1]
+    mov         r4d,       %2
+    lea         r5,        [3 * r1]
 
 .loop:
-movd        m2,        [r0]
-movd        m3,        [r0 + r1]
-movd        m4,        [r0 + 2 * r1]
-movd        m5,        [r0 + r5]
+    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
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklbw   m2,        m6
 
-pmaddubsw   m2,        m0
+    pmaddubsw   m2,        m0
 
-lea         r0,        [r0 + 4 * r1]
-movd        m6,        [r0]
+    lea         r0,        [r0 + 4 * r1]
+    movd        m6,        [r0]
 
-punpcklbw   m3,        m4
-punpcklbw   m7,        m5,        m6
-punpcklbw   m3,        m7
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklbw   m3,        m7
 
-pmaddubsw   m3,        m0
+    pmaddubsw   m3,        m0
 
-phaddw      m2,        m3
+    phaddw      m2,        m3
 
-pmulhrsw    m2,        m1
+    pmulhrsw    m2,        m1
 
-movd        m7,        [r0 + r1]
+    movd        m7,        [r0 + r1]
 
-punpcklbw   m4,        m5
-punpcklbw   m3,        m6,        m7
-punpcklbw   m4,        m3
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklbw   m4,        m3
 
-pmaddubsw   m4,        m0
+    pmaddubsw   m4,        m0
 
-movd        m3,        [r0 + 2 * r1]
+    movd        m3,        [r0 + 2 * r1]
 
-punpcklbw   m5,        m6
-punpcklbw   m7,        m3
-punpcklbw   m5,        m7
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklbw   m5,        m7
 
-pmaddubsw   m5,        m0
+    pmaddubsw   m5,        m0
 
-phaddw      m4,        m5
+    phaddw      m4,        m5
 
-pmulhrsw    m4,        m1
-packuswb    m2,        m4
+    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
+    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]
+    lea         r2,        [r2 + 2 * r3]
 
-sub         r4,        4
-jnz        .loop
-RET
+    sub         r4,        4
+    jnz        .loop
+    RET
 %endmacro
 
-FILTER_V4_W2_H4 2, 8
+    FILTER_V4_W2_H4 2, 8
 
-FILTER_V4_W2_H4 2, 16
+    FILTER_V4_W2_H4 2, 16
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_vert_pp_4x2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -2930,46 +4663,46 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_4x2, 4, 6, 6
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m0,        [tab_Cm]
-lea         r5,        [r0 + 2 * r1]
+    pshufb      m0,        [tab_Cm]
+    lea         r5,        [r0 + 2 * r1]
 
-movd        m2,        [r0]
-movd        m3,        [r0 + r1]
-movd        m4,        [r5]
-movd        m5,        [r5 + 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
+    punpcklbw   m2,        m3
+    punpcklbw   m1,        m4,        m5
+    punpcklbw   m2,        m1
 
-pmaddubsw   m2,        m0
+    pmaddubsw   m2,        m0
 
-movd        m1,        [r0 + 4 * r1]
+    movd        m1,        [r0 + 4 * r1]
 
-punpcklbw   m3,        m4
-punpcklbw   m5,        m1
-punpcklbw   m3,        m5
+    punpcklbw   m3,        m4
+    punpcklbw   m5,        m1
+    punpcklbw   m3,        m5
 
-pmaddubsw   m3,        m0
+    pmaddubsw   m3,        m0
 
-phaddw      m2,        m3
+    phaddw      m2,        m3
 
-pmulhrsw    m2,        [pw_512]
-packuswb    m2,        m2
-movd        [r2],      m2
-pextrd      [r2 + r3], m2,  1
+    pmulhrsw    m2,        [pw_512]
+    packuswb    m2,        m2
+    movd        [r2],      m2
+    pextrd      [r2 + r3], m2,  1
 
-RET
+    RET
 
 %macro FILTER_VER_CHROMA_AVX2_4x2 1
 INIT_YMM avx2
@@ -3017,8 +4750,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_4x2 pp
-FILTER_VER_CHROMA_AVX2_4x2 ps
+    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)
@@ -3026,71 +4759,71 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_4x4, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m0,        [tab_Cm]
-mova        m1,        [pw_512]
-lea         r5,        [r0 + 4 * r1]
-lea         r4,        [r1 * 3]
+    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]
+    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
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklbw   m2,        m6
 
-pmaddubsw   m2,        m0
+    pmaddubsw   m2,        m0
 
-movd        m6,        [r5]
+    movd        m6,        [r5]
 
-punpcklbw   m3,        m4
-punpcklbw   m7,        m5,        m6
-punpcklbw   m3,        m7
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklbw   m3,        m7
 
-pmaddubsw   m3,        m0
+    pmaddubsw   m3,        m0
 
-phaddw      m2,        m3
+    phaddw      m2,        m3
 
-pmulhrsw    m2,        m1
+    pmulhrsw    m2,        m1
 
-movd        m7,        [r5 + r1]
+    movd        m7,        [r5 + r1]
 
-punpcklbw   m4,        m5
-punpcklbw   m3,        m6,        m7
-punpcklbw   m4,        m3
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklbw   m4,        m3
 
-pmaddubsw   m4,        m0
+    pmaddubsw   m4,        m0
 
-movd        m3,        [r5 + 2 * r1]
+    movd        m3,        [r5 + 2 * r1]
 
-punpcklbw   m5,        m6
-punpcklbw   m7,        m3
-punpcklbw   m5,        m7
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklbw   m5,        m7
 
-pmaddubsw   m5,        m0
+    pmaddubsw   m5,        m0
 
-phaddw      m4,        m5
+    phaddw      m4,        m5
 
-pmulhrsw    m4,        m1
+    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
+    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
@@ -3148,8 +4881,8 @@
 %endif
     RET
 %endmacro
-FILTER_VER_CHROMA_AVX2_4x4 pp
-FILTER_VER_CHROMA_AVX2_4x4 ps
+    FILTER_VER_CHROMA_AVX2_4x4 pp
+    FILTER_VER_CHROMA_AVX2_4x4 ps
 
 %macro FILTER_VER_CHROMA_AVX2_4x8 1
 INIT_YMM avx2
@@ -3235,8 +4968,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_4x8 pp
-FILTER_VER_CHROMA_AVX2_4x8 ps
+    FILTER_VER_CHROMA_AVX2_4x8 pp
+    FILTER_VER_CHROMA_AVX2_4x8 ps
 
 %macro FILTER_VER_CHROMA_AVX2_4x16 1
 INIT_YMM avx2
@@ -3380,8 +5113,8 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_4x16 pp
-FILTER_VER_CHROMA_AVX2_4x16 ps
+    FILTER_VER_CHROMA_AVX2_4x16 pp
+    FILTER_VER_CHROMA_AVX2_4x16 ps
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -3390,184 +5123,184 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m0,        [tab_Cm]
+    pshufb      m0,        [tab_Cm]
 
-mova        m1,        [pw_512]
+    mova        m1,        [pw_512]
 
-mov         r4d,       %2
+    mov         r4d,       %2
 
-lea         r5,        [3 * r1]
+    lea         r5,        [3 * r1]
 
 .loop:
-movd        m2,        [r0]
-movd        m3,        [r0 + r1]
-movd        m4,        [r0 + 2 * r1]
-movd        m5,        [r0 + r5]
+    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
+    punpcklbw   m2,        m3
+    punpcklbw   m6,        m4,        m5
+    punpcklbw   m2,        m6
 
-pmaddubsw   m2,        m0
+    pmaddubsw   m2,        m0
 
-lea         r0,        [r0 + 4 * r1]
-movd        m6,        [r0]
+    lea         r0,        [r0 + 4 * r1]
+    movd        m6,        [r0]
 
-punpcklbw   m3,        m4
-punpcklbw   m7,        m5,        m6
-punpcklbw   m3,        m7
+    punpcklbw   m3,        m4
+    punpcklbw   m7,        m5,        m6
+    punpcklbw   m3,        m7
 
-pmaddubsw   m3,        m0
+    pmaddubsw   m3,        m0
 
-phaddw      m2,        m3
+    phaddw      m2,        m3
 
-pmulhrsw    m2,        m1
+    pmulhrsw    m2,        m1
 
-movd        m7,        [r0 + r1]
+    movd        m7,        [r0 + r1]
 
-punpcklbw   m4,        m5
-punpcklbw   m3,        m6,        m7
-punpcklbw   m4,        m3
+    punpcklbw   m4,        m5
+    punpcklbw   m3,        m6,        m7
+    punpcklbw   m4,        m3
 
-pmaddubsw   m4,        m0
+    pmaddubsw   m4,        m0
 
-movd        m3,        [r0 + 2 * r1]
+    movd        m3,        [r0 + 2 * r1]
 
-punpcklbw   m5,        m6
-punpcklbw   m7,        m3
-punpcklbw   m5,        m7
+    punpcklbw   m5,        m6
+    punpcklbw   m7,        m3
+    punpcklbw   m5,        m7
 
-pmaddubsw   m5,        m0
+    pmaddubsw   m5,        m0
 
-phaddw      m4,        m5
+    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
+    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]
+    lea         r2,        [r2 + 2 * r3]
 
-sub         r4,        4
-jnz        .loop
-RET
+    sub         r4,        4
+    jnz        .loop
+    RET
 %endmacro
 
-FILTER_V4_W4_H4 4,  8
-FILTER_V4_W4_H4 4, 16
+    FILTER_V4_W4_H4 4,  8
+    FILTER_V4_W4_H4 4, 16
 
-FILTER_V4_W4_H4 4, 32
+    FILTER_V4_W4_H4 4, 32
 
 %macro FILTER_V4_W8_H2 0
-punpcklbw   m1,        m2
-punpcklbw   m7,        m3,        m0
+    punpcklbw   m1,        m2
+    punpcklbw   m7,        m3,        m0
 
-pmaddubsw   m1,        m6
-pmaddubsw   m7,        m5
+    pmaddubsw   m1,        m6
+    pmaddubsw   m7,        m5
 
-paddw       m1,        m7
+    paddw       m1,        m7
 
-pmulhrsw    m1,        m4
-packuswb    m1,        m1
+    pmulhrsw    m1,        m4
+    packuswb    m1,        m1
 %endmacro
 
 %macro FILTER_V4_W8_H3 0
-punpcklbw   m2,        m3
-punpcklbw   m7,        m0,        m1
+    punpcklbw   m2,        m3
+    punpcklbw   m7,        m0,        m1
 
-pmaddubsw   m2,        m6
-pmaddubsw   m7,        m5
+    pmaddubsw   m2,        m6
+    pmaddubsw   m7,        m5
 
-paddw       m2,        m7
+    paddw       m2,        m7
 
-pmulhrsw    m2,        m4
-packuswb    m2,        m2
+    pmulhrsw    m2,        m4
+    packuswb    m2,        m2
 %endmacro
 
 %macro FILTER_V4_W8_H4 0
-punpcklbw   m3,        m0
-punpcklbw   m7,        m1,        m2
+    punpcklbw   m3,        m0
+    punpcklbw   m7,        m1,        m2
 
-pmaddubsw   m3,        m6
-pmaddubsw   m7,        m5
+    pmaddubsw   m3,        m6
+    pmaddubsw   m7,        m5
 
-paddw       m3,        m7
+    paddw       m3,        m7
 
-pmulhrsw    m3,        m4
-packuswb    m3,        m3
+    pmulhrsw    m3,        m4
+    packuswb    m3,        m3
 %endmacro
 
 %macro FILTER_V4_W8_H5 0
-punpcklbw   m0,        m1
-punpcklbw   m7,        m2,        m3
+    punpcklbw   m0,        m1
+    punpcklbw   m7,        m2,        m3
 
-pmaddubsw   m0,        m6
-pmaddubsw   m7,        m5
+    pmaddubsw   m0,        m6
+    pmaddubsw   m7,        m5
 
-paddw       m0,        m7
+    paddw       m0,        m7
 
-pmulhrsw    m0,        m4
-packuswb    m0,        m0
+    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 %1, %2
+    movq        m0,        [r0 + 4 * r1]
 
-FILTER_V4_W8_H2
+    FILTER_V4_W8_H2
 
-movh        [r2 + r3], m1
+    movh        [r2 + r3], m1
 %endmacro
 
 %macro FILTER_V4_W8_8x4 2
-FILTER_V4_W8_8x2 %1, %2
+    FILTER_V4_W8_8x2 %1, %2
 ;8x3
-lea         r6,        [r0 + 4 * r1]
-movq        m1,        [r6 + r1]
+    lea         r6,        [r0 + 4 * r1]
+    movq        m1,        [r6 + r1]
 
-FILTER_V4_W8_H3
+    FILTER_V4_W8_H3
 
-movh        [r2 + 2 * r3], m2
+    movh        [r2 + 2 * r3], m2
 
 ;8x4
-movq        m2,        [r6 + 2 * r1]
+    movq        m2,        [r6 + 2 * r1]
 
-FILTER_V4_W8_H4
+    FILTER_V4_W8_H4
 
-lea         r5,        [r2 + 2 * r3]
-movh        [r5 + r3], m3
+    lea         r5,        [r2 + 2 * r3]
+    movh        [r5 + r3], m3
 %endmacro
 
 %macro FILTER_V4_W8_8x6 2
-FILTER_V4_W8_8x4 %1, %2
+    FILTER_V4_W8_8x4 %1, %2
 ;8x5
-lea         r6,        [r6 + 2 * r1]
-movq        m3,        [r6 + r1]
+    lea         r6,        [r6 + 2 * r1]
+    movq        m3,        [r6 + r1]
 
-FILTER_V4_W8_H5
+    FILTER_V4_W8_H5
 
-movh        [r2 + 4 * r3], m0
+    movh        [r2 + 4 * r3], m0
 
 ;8x6
-movq        m0,        [r0 + 8 * r1]
+    movq        m0,        [r0 + 8 * r1]
 
-FILTER_V4_W8_H2
+    FILTER_V4_W8_H2
 
-lea         r5,        [r2 + 4 * r3]
-movh        [r5 + r3], m1
+    lea         r5,        [r2 + 4 * r3]
+    movh        [r5 + r3], m1
 %endmacro
 
 ;-----------------------------------------------------------------------------
@@ -3577,60 +5310,60 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_%1x%2, 4, 7, 8
 
-mov         r4d,       r4m
+    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]
+    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
+    punpcklbw   m0,        m1
+    punpcklbw   m4,        m2,          m3
 
 %ifdef PIC
-lea         r6,        [tab_ChromaCoeff]
-movd        m5,        [r6 + r4 * 4]
+    lea         r6,        [tab_ChromaCoeff]
+    movd        m5,        [r6 + r4 * 4]
 %else
-movd        m5,        [tab_ChromaCoeff + r4 * 4]
+    movd        m5,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m6,        m5,       [tab_Vm]
-pmaddubsw   m0,        m6
+    pshufb      m6,        m5,       [tab_Vm]
+    pmaddubsw   m0,        m6
 
-pshufb      m5,        [tab_Vm + 16]
-pmaddubsw   m4,        m5
+    pshufb      m5,        [tab_Vm + 16]
+    pmaddubsw   m4,        m5
 
-paddw       m0,        m4
+    paddw       m0,        m4
 
-mova        m4,        [pw_512]
+    mova        m4,        [pw_512]
 
-pmulhrsw    m0,        m4
-packuswb    m0,        m0
-movh        [r2],      m0
+    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
+    FILTER_V4_W8_8x2 8, 2
 
-RET
+    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
+    FILTER_V4_W8_8x4 8, 4
 
-RET
+    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
+    FILTER_V4_W8_8x6 8, 6
 
-RET
+    RET
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_4tap_vert_ps_4x2(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
@@ -3638,46 +5371,46 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_ps_4x2, 4, 6, 6
 
-mov         r4d, r4m
-sub         r0, r1
-add         r3d, r3d
+    mov         r4d, r4m
+    sub         r0, r1
+    add         r3d, r3d
 
 %ifdef PIC
-lea         r5, [tab_ChromaCoeff]
-movd        m0, [r5 + r4 * 4]
+    lea         r5, [tab_ChromaCoeff]
+    movd        m0, [r5 + r4 * 4]
 %else
-movd        m0, [tab_ChromaCoeff + r4 * 4]
+    movd        m0, [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m0, [tab_Cm]
+    pshufb      m0, [tab_Cm]
 
-movd        m2, [r0]
-movd        m3, [r0 + r1]
-lea         r5, [r0 + 2 * r1]
-movd        m4, [r5]
-movd        m5, [r5 + r1]
+    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
+    punpcklbw   m2, m3
+    punpcklbw   m1, m4, m5
+    punpcklbw   m2, m1
 
-pmaddubsw   m2, m0
+    pmaddubsw   m2, m0
 
-movd        m1, [r0 + 4 * r1]
+    movd        m1, [r0 + 4 * r1]
 
-punpcklbw   m3, m4
-punpcklbw   m5, m1
-punpcklbw   m3, m5
+    punpcklbw   m3, m4
+    punpcklbw   m5, m1
+    punpcklbw   m3, m5
 
-pmaddubsw   m3, m0
+    pmaddubsw   m3, m0
 
-phaddw      m2, m3
+    phaddw      m2, m3
 
-psubw       m2, [pw_2000]
-movh        [r2], m2
-movhps      [r2 + r3], m2
+    psubw       m2, [pw_2000]
+    movh        [r2], m2
+    movhps      [r2 + r3], m2
 
-RET
+    RET
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_4tap_vert_ps_4x4(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx)
@@ -3835,10 +5568,10 @@
     RET
 %endmacro
 
-FILTER_V_PS_W4_H4 4, 8
-FILTER_V_PS_W4_H4 4, 16
+    FILTER_V_PS_W4_H4 4, 8
+    FILTER_V_PS_W4_H4 4, 16
 
-FILTER_V_PS_W4_H4 4, 32
+    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)
@@ -3904,12 +5637,12 @@
     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, 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
+    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)
@@ -3999,9 +5732,9 @@
     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
+    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)
@@ -4095,8 +5828,8 @@
     RET
 %endmacro
 
-FILTER_V_PS_W6 6, 8
-FILTER_V_PS_W6 6, 16
+    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)
@@ -4181,8 +5914,8 @@
     RET
 %endmacro
 
-FILTER_V_PS_W12 12, 16
-FILTER_V_PS_W12 12, 32
+    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)
@@ -4266,14 +5999,14 @@
     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,  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
+    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)
@@ -4389,9 +6122,9 @@
     RET
 %endmacro
 
-FILTER_V4_PS_W24 24, 32
+    FILTER_V4_PS_W24 24, 32
 
-FILTER_V4_PS_W24 24, 64
+    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)
@@ -4482,13 +6215,13 @@
     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,  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
+    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)
@@ -4497,95 +6230,95 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m5,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m5,        [r5 + r4 * 4]
 %else
-movd        m5,        [tab_ChromaCoeff + r4 * 4]
+    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]
+    pshufb      m6,        m5,       [tab_Vm]
+    pshufb      m5,        [tab_Vm + 16]
+    mova        m4,        [pw_512]
+    lea         r5,        [r1 * 3]
 
-mov         r4d,       %2
+    mov         r4d,       %2
 
 .loop:
-movq        m0,        [r0]
-movq        m1,        [r0 + r1]
-movq        m2,        [r0 + 2 * r1]
-movq        m3,        [r0 + r5]
+    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
+    punpcklbw   m0,        m1
+    punpcklbw   m1,        m2
+    punpcklbw   m2,        m3
 
-pmaddubsw   m0,        m6
-pmaddubsw   m7,        m2, m5
+    pmaddubsw   m0,        m6
+    pmaddubsw   m7,        m2, m5
 
-paddw       m0,        m7
+    paddw       m0,        m7
 
-pmulhrsw    m0,        m4
-packuswb    m0,        m0
-movh        [r2],      m0
+    pmulhrsw    m0,        m4
+    packuswb    m0,        m0
+    movh        [r2],      m0
 
-lea         r0,        [r0 + 4 * r1]
-movq        m0,        [r0]
+    lea         r0,        [r0 + 4 * r1]
+    movq        m0,        [r0]
 
-punpcklbw   m3,        m0
+    punpcklbw   m3,        m0
 
-pmaddubsw   m1,        m6
-pmaddubsw   m7,        m3, m5
+    pmaddubsw   m1,        m6
+    pmaddubsw   m7,        m3, m5
 
-paddw       m1,        m7
+    paddw       m1,        m7
 
-pmulhrsw    m1,        m4
-packuswb    m1,        m1
-movh        [r2 + r3], m1
+    pmulhrsw    m1,        m4
+    packuswb    m1,        m1
+    movh        [r2 + r3], m1
 
-movq        m1,        [r0 + r1]
+    movq        m1,        [r0 + r1]
 
-punpcklbw   m0,        m1
+    punpcklbw   m0,        m1
 
-pmaddubsw   m2,        m6
-pmaddubsw   m0,        m5
+    pmaddubsw   m2,        m6
+    pmaddubsw   m0,        m5
 
-paddw       m2,        m0
+    paddw       m2,        m0
 
-pmulhrsw    m2,        m4
+    pmulhrsw    m2,        m4
 
-movq        m7,        [r0 + 2 * r1]
-punpcklbw   m1,        m7
+    movq        m7,        [r0 + 2 * r1]
+    punpcklbw   m1,        m7
 
-pmaddubsw   m3,        m6
-pmaddubsw   m1,        m5
+    pmaddubsw   m3,        m6
+    pmaddubsw   m1,        m5
 
-paddw       m3,        m1
+    paddw       m3,        m1
 
-pmulhrsw    m3,        m4
-packuswb    m2,        m3
+    pmulhrsw    m3,        m4
+    packuswb    m2,        m3
 
-lea         r2,        [r2 + 2 * r3]
-movh        [r2],      m2
-movhps      [r2 + r3], m2
+    lea         r2,        [r2 + 2 * r3]
+    movh        [r2],      m2
+    movhps      [r2 + r3], m2
 
-lea         r2,        [r2 + 2 * r3]
+    lea         r2,        [r2 + 2 * r3]
 
-sub         r4,         4
-jnz        .loop
-RET
+    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,  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
+    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
@@ -4691,8 +6424,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_8x8 pp
-FILTER_VER_CHROMA_AVX2_8x8 ps
+    FILTER_VER_CHROMA_AVX2_8x8 pp
+    FILTER_VER_CHROMA_AVX2_8x8 ps
 
 %macro FILTER_VER_CHROMA_AVX2_8x6 1
 INIT_YMM avx2
@@ -4780,8 +6513,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_8x6 pp
-FILTER_VER_CHROMA_AVX2_8x6 ps
+    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
@@ -4961,12 +6694,154 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_8x16 pp
-FILTER_VER_CHROMA_AVX2_8x16 ps
+    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_8x32 1
+%macro FILTER_VER_CHROMA_AVX2_8xN 2
 INIT_YMM avx2
-cglobal interp_4tap_vert_%1_8x32, 4, 7, 8
+cglobal interp_4tap_vert_%1_8x%2, 4, 7, 8
     mov             r4d, r4m
     shl             r4d, 6
 
@@ -4986,15 +6861,17 @@
     mova            m7, [pw_2000]
 %endif
     lea             r6, [r3 * 3]
-%rep 2
+%rep %2 / 16
     PROCESS_CHROMA_AVX2_W8_16R %1
     lea             r2, [r2 + r3 * 4]
 %endrep
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_8x32 pp
-FILTER_VER_CHROMA_AVX2_8x32 ps
+    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
@@ -5065,8 +6942,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_8x4 pp
-FILTER_VER_CHROMA_AVX2_8x4 ps
+    FILTER_VER_CHROMA_AVX2_8x4 pp
+    FILTER_VER_CHROMA_AVX2_8x4 ps
 
 %macro FILTER_VER_CHROMA_AVX2_8x2 1
 INIT_YMM avx2
@@ -5114,8 +6991,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_8x2 pp
-FILTER_VER_CHROMA_AVX2_8x2 ps
+    FILTER_VER_CHROMA_AVX2_8x2 pp
+    FILTER_VER_CHROMA_AVX2_8x2 ps
 
 %macro FILTER_VER_CHROMA_AVX2_6x8 1
 INIT_YMM avx2
@@ -5194,8 +7071,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_6x8 pp
-FILTER_VER_CHROMA_AVX2_6x8 ps
+    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)
@@ -5204,96 +7081,96 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_6x%2, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m5,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m5,        [r5 + r4 * 4]
 %else
-movd        m5,        [tab_ChromaCoeff + r4 * 4]
+    movd        m5,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m6,        m5,       [tab_Vm]
-pshufb      m5,        [tab_Vm + 16]
-mova        m4,        [pw_512]
+    pshufb      m6,        m5,       [tab_Vm]
+    pshufb      m5,        [tab_Vm + 16]
+    mova        m4,        [pw_512]
 
-mov         r4d,       %2
-lea         r5,        [3 * r1]
+    mov         r4d,       %2
+    lea         r5,        [3 * r1]
 
 .loop:
-movq        m0,        [r0]
-movq        m1,        [r0 + r1]
-movq        m2,        [r0 + 2 * r1]
-movq        m3,        [r0 + r5]
+    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
+    punpcklbw   m0,        m1
+    punpcklbw   m1,        m2
+    punpcklbw   m2,        m3
 
-pmaddubsw   m0,        m6
-pmaddubsw   m7,        m2, m5
+    pmaddubsw   m0,        m6
+    pmaddubsw   m7,        m2, m5
 
-paddw       m0,        m7
+    paddw       m0,        m7
 
-pmulhrsw    m0,        m4
-packuswb    m0,        m0
-movd        [r2],      m0
-pextrw      [r2 + 4],  m0,    2
+    pmulhrsw    m0,        m4
+    packuswb    m0,        m0
+    movd        [r2],      m0
+    pextrw      [r2 + 4],  m0,    2
 
-lea         r0,        [r0 + 4 * r1]
+    lea         r0,        [r0 + 4 * r1]
 
-movq        m0,        [r0]
-punpcklbw   m3,        m0
+    movq        m0,        [r0]
+    punpcklbw   m3,        m0
 
-pmaddubsw   m1,        m6
-pmaddubsw   m7,        m3, m5
+    pmaddubsw   m1,        m6
+    pmaddubsw   m7,        m3, m5
 
-paddw       m1,        m7
+    paddw       m1,        m7
 
-pmulhrsw    m1,        m4
-packuswb    m1,        m1
-movd        [r2 + r3],      m1
-pextrw      [r2 + r3 + 4],  m1,    2
+    pmulhrsw    m1,        m4
+    packuswb    m1,        m1
+    movd        [r2 + r3],      m1
+    pextrw      [r2 + r3 + 4],  m1,    2
 
-movq        m1,        [r0 + r1]
-punpcklbw   m7,        m0,        m1
+    movq        m1,        [r0 + r1]
+    punpcklbw   m7,        m0,        m1
 
-pmaddubsw   m2,        m6
-pmaddubsw   m7,        m5
+    pmaddubsw   m2,        m6
+    pmaddubsw   m7,        m5
 
-paddw       m2,        m7
+    paddw       m2,        m7
 
-pmulhrsw    m2,        m4
-packuswb    m2,        m2
-lea         r2,        [r2 + 2 * r3]
-movd        [r2],      m2
-pextrw      [r2 + 4],  m2,    2
+    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
+    movq        m2,        [r0 + 2 * r1]
+    punpcklbw   m1,        m2
 
-pmaddubsw   m3,        m6
-pmaddubsw   m1,        m5
+    pmaddubsw   m3,        m6
+    pmaddubsw   m1,        m5
 
-paddw       m3,        m1
+    paddw       m3,        m1
 
-pmulhrsw    m3,        m4
-packuswb    m3,        m3
+    pmulhrsw    m3,        m4
+    packuswb    m3,        m3
 
-movd        [r2 + r3],        m3
-pextrw      [r2 + r3 + 4],    m3,    2
+    movd        [r2 + r3],        m3
+    pextrw      [r2 + r3 + 4],    m3,    2
 
-lea         r2,        [r2 + 2 * r3]
+    lea         r2,        [r2 + 2 * r3]
 
-sub         r4,         4
-jnz        .loop
-RET
+    sub         r4,         4
+    jnz        .loop
+    RET
 %endmacro
 
-FILTER_V4_W6_H4 6, 8
+    FILTER_V4_W6_H4 6, 8
 
-FILTER_V4_W6_H4 6, 16
+    FILTER_V4_W6_H4 6, 16
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_vert_pp_12x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -5302,88 +7179,88 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_12x%2, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m1,        m0,       [tab_Vm]
-pshufb      m0,        [tab_Vm + 16]
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
 
-mov         r4d,       %2
+    mov         r4d,       %2
 
 .loop:
-movu        m2,        [r0]
-movu        m3,        [r0 + r1]
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
 
-punpcklbw   m4,        m2,        m3
-punpckhbw   m2,        m3
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
 
-pmaddubsw   m4,        m1
-pmaddubsw   m2,        m1
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
 
-lea         r0,        [r0 + 2 * r1]
-movu        m5,        [r0]
-movu        m7,        [r0 + r1]
+    lea         r0,        [r0 + 2 * r1]
+    movu        m5,        [r0]
+    movu        m7,        [r0 + r1]
 
-punpcklbw   m6,        m5,        m7
-pmaddubsw   m6,        m0
-paddw       m4,        m6
+    punpcklbw   m6,        m5,        m7
+    pmaddubsw   m6,        m0
+    paddw       m4,        m6
 
-punpckhbw   m6,        m5,        m7
-pmaddubsw   m6,        m0
-paddw       m2,        m6
+    punpckhbw   m6,        m5,        m7
+    pmaddubsw   m6,        m0
+    paddw       m2,        m6
 
-mova        m6,        [pw_512]
+    mova        m6,        [pw_512]
 
-pmulhrsw    m4,        m6
-pmulhrsw    m2,        m6
+    pmulhrsw    m4,        m6
+    pmulhrsw    m2,        m6
 
-packuswb    m4,        m2
+    packuswb    m4,        m2
 
-movh         [r2],     m4
-pextrd       [r2 + 8], m4,  2
+    movh         [r2],     m4
+    pextrd       [r2 + 8], m4,  2
 
-punpcklbw   m4,        m3,        m5
-punpckhbw   m3,        m5
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
 
-pmaddubsw   m4,        m1
-pmaddubsw   m3,        m1
+    pmaddubsw   m4,        m1
+    pmaddubsw   m3,        m1
 
-movu        m5,        [r0 + 2 * r1]
+    movu        m5,        [r0 + 2 * r1]
 
-punpcklbw   m2,        m7,        m5
-punpckhbw   m7,        m5
+    punpcklbw   m2,        m7,        m5
+    punpckhbw   m7,        m5
 
-pmaddubsw   m2,        m0
-pmaddubsw   m7,        m0
+    pmaddubsw   m2,        m0
+    pmaddubsw   m7,        m0
 
-paddw       m4,        m2
-paddw       m3,        m7
+    paddw       m4,        m2
+    paddw       m3,        m7
 
-pmulhrsw    m4,        m6
-pmulhrsw    m3,        m6
+    pmulhrsw    m4,        m6
+    pmulhrsw    m3,        m6
 
-packuswb    m4,        m3
+    packuswb    m4,        m3
 
-movh        [r2 + r3],      m4
-pextrd      [r2 + r3 + 8],  m4,  2
+    movh        [r2 + r3],      m4
+    pextrd      [r2 + r3 + 8],  m4,  2
 
-lea         r2,        [r2 + 2 * r3]
+    lea         r2,        [r2 + 2 * r3]
 
-sub         r4,        2
-jnz        .loop
-RET
+    sub         r4,        2
+    jnz        .loop
+    RET
 %endmacro
 
-FILTER_V4_W12_H2 12, 16
+    FILTER_V4_W12_H2 12, 16
 
-FILTER_V4_W12_H2 12, 32
+    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)
@@ -5392,91 +7269,91 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_16x%2, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m1,        m0,       [tab_Vm]
-pshufb      m0,        [tab_Vm + 16]
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
 
-mov         r4d,       %2/2
+    mov         r4d,       %2/2
 
 .loop:
-movu        m2,        [r0]
-movu        m3,        [r0 + r1]
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
 
-punpcklbw   m4,        m2,        m3
-punpckhbw   m2,        m3
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
 
-pmaddubsw   m4,        m1
-pmaddubsw   m2,        m1
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
 
-lea         r0,        [r0 + 2 * r1]
-movu        m5,        [r0]
-movu        m6,        [r0 + r1]
+    lea         r0,        [r0 + 2 * r1]
+    movu        m5,        [r0]
+    movu        m6,        [r0 + r1]
 
-punpckhbw   m7,        m5,        m6
-pmaddubsw   m7,        m0
-paddw       m2,        m7
+    punpckhbw   m7,        m5,        m6
+    pmaddubsw   m7,        m0
+    paddw       m2,        m7
 
-punpcklbw   m7,        m5,        m6
-pmaddubsw   m7,        m0
-paddw       m4,        m7
+    punpcklbw   m7,        m5,        m6
+    pmaddubsw   m7,        m0
+    paddw       m4,        m7
 
-mova        m7,        [pw_512]
+    mova        m7,        [pw_512]
 
-pmulhrsw    m4,        m7
-pmulhrsw    m2,        m7
+    pmulhrsw    m4,        m7
+    pmulhrsw    m2,        m7
 
-packuswb    m4,        m2
+    packuswb    m4,        m2
 
-movu        [r2],      m4
+    movu        [r2],      m4
 
-punpcklbw   m4,        m3,        m5
-punpckhbw   m3,        m5
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
 
-pmaddubsw   m4,        m1
-pmaddubsw   m3,        m1
+    pmaddubsw   m4,        m1
+    pmaddubsw   m3,        m1
 
-movu        m5,        [r0 + 2 * r1]
+    movu        m5,        [r0 + 2 * r1]
 
-punpcklbw   m2,        m6,        m5
-punpckhbw   m6,        m5
+    punpcklbw   m2,        m6,        m5
+    punpckhbw   m6,        m5
 
-pmaddubsw   m2,        m0
-pmaddubsw   m6,        m0
+    pmaddubsw   m2,        m0
+    pmaddubsw   m6,        m0
 
-paddw       m4,        m2
-paddw       m3,        m6
+    paddw       m4,        m2
+    paddw       m3,        m6
 
-pmulhrsw    m4,        m7
-pmulhrsw    m3,        m7
+    pmulhrsw    m4,        m7
+    pmulhrsw    m3,        m7
 
-packuswb    m4,        m3
+    packuswb    m4,        m3
 
-movu        [r2 + r3],      m4
+    movu        [r2 + r3],      m4
 
-lea         r2,        [r2 + 2 * r3]
+    lea         r2,        [r2 + 2 * r3]
 
-dec         r4d
-jnz        .loop
-RET
+    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,  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
+    FILTER_V4_W16_H2 16, 24
+    FILTER_V4_W16_H2 16, 64
 
 %macro FILTER_VER_CHROMA_AVX2_16x16 1
 INIT_YMM avx2
@@ -5736,8 +7613,8 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_16x16 pp
-FILTER_VER_CHROMA_AVX2_16x16 ps
+    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
@@ -5891,8 +7768,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_16x8 pp
-FILTER_VER_CHROMA_AVX2_16x8 ps
+    FILTER_VER_CHROMA_AVX2_16x8 pp
+    FILTER_VER_CHROMA_AVX2_16x8 ps
 
 %macro FILTER_VER_CHROMA_AVX2_16x12 1
 INIT_YMM avx2
@@ -6119,13 +7996,13 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_16x12 pp
-FILTER_VER_CHROMA_AVX2_16x12 ps
+    FILTER_VER_CHROMA_AVX2_16x12 pp
+    FILTER_VER_CHROMA_AVX2_16x12 ps
 
-%macro FILTER_VER_CHROMA_AVX2_16x32 1
-INIT_YMM avx2
+%macro FILTER_VER_CHROMA_AVX2_16xN 2
 %if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_16x32, 4, 8, 8
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_16x%2, 4, 8, 8
     mov             r4d, r4m
     shl             r4d, 6
 
@@ -6145,7 +8022,7 @@
     mova            m7, [pw_2000]
 %endif
     lea             r6, [r3 * 3]
-    mov             r7d, 2
+    mov             r7d, %2 / 16
 .loopH:
     movu            xm0, [r0]
     vinserti128     m0, m0, [r0 + r1 * 2], 1
@@ -6412,8 +8289,381 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_16x32 pp
-FILTER_VER_CHROMA_AVX2_16x32 ps
+    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
@@ -6863,8 +9113,8 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_24x32 pp
-FILTER_VER_CHROMA_AVX2_24x32 ps
+    FILTER_VER_CHROMA_AVX2_24x32 pp
+    FILTER_VER_CHROMA_AVX2_24x32 ps
 
 %macro FILTER_VER_CHROMA_AVX2_16x4 1
 INIT_YMM avx2
@@ -6961,12 +9211,12 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_16x4 pp
-FILTER_VER_CHROMA_AVX2_16x4 ps
+    FILTER_VER_CHROMA_AVX2_16x4 pp
+    FILTER_VER_CHROMA_AVX2_16x4 ps
 
-%macro FILTER_VER_CHROMA_AVX2_12x16 1
+%macro FILTER_VER_CHROMA_AVX2_12xN 2
 INIT_YMM avx2
-cglobal interp_4tap_vert_%1_12x16, 4, 7, 8
+cglobal interp_4tap_vert_%1_12x%2, 4, 7, 8
     mov             r4d, r4m
     shl             r4d, 6
 
@@ -6986,7 +9236,7 @@
     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
@@ -7272,11 +9522,15 @@
     vextracti128    xm5, m5, 1
     movq            [r2 + r6 + 16], xm5
 %endif
+    lea             r2, [r2 + r3 * 4]
+%endrep
     RET
 %endmacro
 
-FILTER_VER_CHROMA_AVX2_12x16 pp
-FILTER_VER_CHROMA_AVX2_12x16 ps
+    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)
@@ -7285,121 +9539,121 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_24x%2, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m1,        m0,       [tab_Vm]
-pshufb      m0,        [tab_Vm + 16]
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
 
-mov         r4d,       %2
+    mov         r4d,       %2
 
 .loop:
-movu        m2,        [r0]
-movu        m3,        [r0 + r1]
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
 
-punpcklbw   m4,        m2,        m3
-punpckhbw   m2,        m3
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
 
-pmaddubsw   m4,        m1
-pmaddubsw   m2,        m1
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
 
-lea         r5,        [r0 + 2 * r1]
-movu        m5,        [r5]
-movu        m7,        [r5 + r1]
+    lea         r5,        [r0 + 2 * r1]
+    movu        m5,        [r5]
+    movu        m7,        [r5 + r1]
 
-punpcklbw   m6,        m5,        m7
-pmaddubsw   m6,        m0
-paddw       m4,        m6
+    punpcklbw   m6,        m5,        m7
+    pmaddubsw   m6,        m0
+    paddw       m4,        m6
 
-punpckhbw   m6,        m5,        m7
-pmaddubsw   m6,        m0
-paddw       m2,        m6
+    punpckhbw   m6,        m5,        m7
+    pmaddubsw   m6,        m0
+    paddw       m2,        m6
 
-mova        m6,        [pw_512]
+    mova        m6,        [pw_512]
 
-pmulhrsw    m4,        m6
-pmulhrsw    m2,        m6
+    pmulhrsw    m4,        m6
+    pmulhrsw    m2,        m6
 
-packuswb    m4,        m2
+    packuswb    m4,        m2
 
-movu        [r2],      m4
+    movu        [r2],      m4
 
-punpcklbw   m4,        m3,        m5
-punpckhbw   m3,        m5
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
 
-pmaddubsw   m4,        m1
-pmaddubsw   m3,        m1
+    pmaddubsw   m4,        m1
+    pmaddubsw   m3,        m1
 
-movu        m2,        [r5 + 2 * r1]
+    movu        m2,        [r5 + 2 * r1]
 
-punpcklbw   m5,        m7,        m2
-punpckhbw   m7,        m2
+    punpcklbw   m5,        m7,        m2
+    punpckhbw   m7,        m2
 
-pmaddubsw   m5,        m0
-pmaddubsw   m7,        m0
+    pmaddubsw   m5,        m0
+    pmaddubsw   m7,        m0
 
-paddw       m4,        m5
-paddw       m3,        m7
+    paddw       m4,        m5
+    paddw       m3,        m7
 
-pmulhrsw    m4,        m6
-pmulhrsw    m3,        m6
+    pmulhrsw    m4,        m6
+    pmulhrsw    m3,        m6
 
-packuswb    m4,        m3
+    packuswb    m4,        m3
 
-movu        [r2 + r3],      m4
+    movu        [r2 + r3],      m4
 
-movq        m2,        [r0 + 16]
-movq        m3,        [r0 + r1 + 16]
-movq        m4,        [r5 + 16]
-movq        m5,        [r5 + r1 + 16]
+    movq        m2,        [r0 + 16]
+    movq        m3,        [r0 + r1 + 16]
+    movq        m4,        [r5 + 16]
+    movq        m5,        [r5 + r1 + 16]
 
-punpcklbw   m2,        m3
-punpcklbw   m4,        m5
+    punpcklbw   m2,        m3
+    punpcklbw   m4,        m5
 
-pmaddubsw   m2,        m1
-pmaddubsw   m4,        m0
+    pmaddubsw   m2,        m1
+    pmaddubsw   m4,        m0
 
-paddw       m2,        m4
+    paddw       m2,        m4
 
-pmulhrsw    m2,        m6
+    pmulhrsw    m2,        m6
 
-movq        m3,        [r0 + r1 + 16]
-movq        m4,        [r5 + 16]
-movq        m5,        [r5 + r1 + 16]
-movq        m7,        [r5 + 2 * r1 + 16]
+    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
+    punpcklbw   m3,        m4
+    punpcklbw   m5,        m7
 
-pmaddubsw   m3,        m1
-pmaddubsw   m5,        m0
+    pmaddubsw   m3,        m1
+    pmaddubsw   m5,        m0
 
-paddw       m3,        m5
+    paddw       m3,        m5
 
-pmulhrsw    m3,        m6
-packuswb    m2,        m3
+    pmulhrsw    m3,        m6
+    packuswb    m2,        m3
 
-movh        [r2 + 16], m2
-movhps      [r2 + r3 + 16], m2
+    movh        [r2 + 16], m2
+    movhps      [r2 + r3 + 16], m2
 
-mov         r0,        r5
-lea         r2,        [r2 + 2 * r3]
+    mov         r0,        r5
+    lea         r2,        [r2 + 2 * r3]
 
-sub         r4,        2
-jnz        .loop
-RET
+    sub         r4,        2
+    jnz        .loop
+    RET
 %endmacro
 
-FILTER_V4_W24 24, 32
+    FILTER_V4_W24 24, 32
 
-FILTER_V4_W24 24, 64
+    FILTER_V4_W24 24, 64
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_vert_pp_32x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -7408,100 +9662,100 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_%1x%2, 4, 6, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m1,        m0,       [tab_Vm]
-pshufb      m0,        [tab_Vm + 16]
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
 
-mova        m7,        [pw_512]
+    mova        m7,        [pw_512]
 
-mov         r4d,       %2
+    mov         r4d,       %2
 
 .loop:
-movu        m2,        [r0]
-movu        m3,        [r0 + r1]
+    movu        m2,        [r0]
+    movu        m3,        [r0 + r1]
 
-punpcklbw   m4,        m2,        m3
-punpckhbw   m2,        m3
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
 
-pmaddubsw   m4,        m1
-pmaddubsw   m2,        m1
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
 
-lea         r5,        [r0 + 2 * r1]
-movu        m3,        [r5]
-movu        m5,        [r5 + r1]
+    lea         r5,        [r0 + 2 * r1]
+    movu        m3,        [r5]
+    movu        m5,        [r5 + r1]
 
-punpcklbw   m6,        m3,        m5
-punpckhbw   m3,        m5
+    punpcklbw   m6,        m3,        m5
+    punpckhbw   m3,        m5
 
-pmaddubsw   m6,        m0
-pmaddubsw   m3,        m0
+    pmaddubsw   m6,        m0
+    pmaddubsw   m3,        m0
 
-paddw       m4,        m6
-paddw       m2,        m3
+    paddw       m4,        m6
+    paddw       m2,        m3
 
-pmulhrsw    m4,        m7
-pmulhrsw    m2,        m7
+    pmulhrsw    m4,        m7
+    pmulhrsw    m2,        m7
 
-packuswb    m4,        m2
+    packuswb    m4,        m2
 
-movu        [r2],      m4
+    movu        [r2],      m4
 
-movu        m2,        [r0 + 16]
-movu        m3,        [r0 + r1 + 16]
+    movu        m2,        [r0 + 16]
+    movu        m3,        [r0 + r1 + 16]
 
-punpcklbw   m4,        m2,        m3
-punpckhbw   m2,        m3
+    punpcklbw   m4,        m2,        m3
+    punpckhbw   m2,        m3
 
-pmaddubsw   m4,        m1
-pmaddubsw   m2,        m1
+    pmaddubsw   m4,        m1
+    pmaddubsw   m2,        m1
 
-movu        m3,        [r5 + 16]
-movu        m5,        [r5 + r1 + 16]
+    movu        m3,        [r5 + 16]
+    movu        m5,        [r5 + r1 + 16]
 
-punpcklbw   m6,        m3,        m5
-punpckhbw   m3,        m5
+    punpcklbw   m6,        m3,        m5
+    punpckhbw   m3,        m5
 
-pmaddubsw   m6,        m0
-pmaddubsw   m3,        m0
+    pmaddubsw   m6,        m0
+    pmaddubsw   m3,        m0
 
-paddw       m4,        m6
-paddw       m2,        m3
+    paddw       m4,        m6
+    paddw       m2,        m3
 
-pmulhrsw    m4,        m7
-pmulhrsw    m2,        m7
+    pmulhrsw    m4,        m7
+    pmulhrsw    m2,        m7
 
-packuswb    m4,        m2
+    packuswb    m4,        m2
 
-movu        [r2 + 16], m4
+    movu        [r2 + 16], m4
 
-lea         r0,        [r0 + r1]
-lea         r2,        [r2 + r3]
+    lea         r0,        [r0 + r1]
+    lea         r2,        [r2 + r3]
 
-dec         r4
-jnz        .loop
-RET
+    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,  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
+    FILTER_V4_W32 32, 48
+    FILTER_V4_W32 32, 64
 
 %macro FILTER_VER_CHROMA_AVX2_32xN 2
-INIT_YMM avx2
 %if ARCH_X86_64 == 1
+INIT_YMM avx2
 cglobal interp_4tap_vert_%1_32x%2, 4, 7, 13
     mov             r4d, r4m
     shl             r4d, 6
@@ -7631,14 +9885,18 @@
 %endif
 %endmacro
 
-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, 32
-FILTER_VER_CHROMA_AVX2_32xN ps, 24
-FILTER_VER_CHROMA_AVX2_32xN ps, 16
-FILTER_VER_CHROMA_AVX2_32xN ps, 8
+    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
 
 ;-----------------------------------------------------------------------------
 ; void interp_4tap_vert_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -7647,413 +9905,1338 @@
 INIT_XMM sse4
 cglobal interp_4tap_vert_pp_%1x%2, 4, 7, 8
 
-mov         r4d,       r4m
-sub         r0,        r1
+    mov         r4d,       r4m
+    sub         r0,        r1
 
 %ifdef PIC
-lea         r5,        [tab_ChromaCoeff]
-movd        m0,        [r5 + r4 * 4]
+    lea         r5,        [tab_ChromaCoeff]
+    movd        m0,        [r5 + r4 * 4]
 %else
-movd        m0,        [tab_ChromaCoeff + r4 * 4]
+    movd        m0,        [tab_ChromaCoeff + r4 * 4]
 %endif
 
-pshufb      m1,        m0,       [tab_Vm]
-pshufb      m0,        [tab_Vm + 16]
+    pshufb      m1,        m0,       [tab_Vm]
+    pshufb      m0,        [tab_Vm + 16]
 
-mov         r4d,       %2/2
+    mov         r4d,       %2/2
 
 .loop:
 
-mov         r6d,       %1/16
+    mov         r6d,       %1/16
 
 .loopW:
 
-movu        m2,        [r0]
-movu        m3,        [r0 + r1]
+    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]
 
-punpcklbw   m4,        m2,        m3
-punpckhbw   m2,        m3
+    punpckhbw   m7,        m5,        m6
+    pmaddubsw   m7,        m0
+    paddw       m2,        m7
 
-pmaddubsw   m4,        m1
-pmaddubsw   m2,        m1
+    punpcklbw   m7,        m5,        m6
+    pmaddubsw   m7,        m0
+    paddw       m4,        m7
 
-lea         r5,        [r0 + 2 * r1]
-movu        m5,        [r5]
-movu        m6,        [r5 + r1]
+    mova        m7,        [pw_512]
 
-punpckhbw   m7,        m5,        m6
-pmaddubsw   m7,        m0
-paddw       m2,        m7
+    pmulhrsw    m4,        m7
+    pmulhrsw    m2,        m7
 
-punpcklbw   m7,        m5,        m6
-pmaddubsw   m7,        m0
-paddw       m4,        m7
+    packuswb    m4,        m2
 
-mova        m7,        [pw_512]
+    movu        [r2],      m4
 
-pmulhrsw    m4,        m7
-pmulhrsw    m2,        m7
+    punpcklbw   m4,        m3,        m5
+    punpckhbw   m3,        m5
 
-packuswb    m4,        m2
+    pmaddubsw   m4,        m1
+    pmaddubsw   m3,        m1
 
-movu        [r2],      m4
+    movu        m5,        [r5 + 2 * r1]
 
-punpcklbw   m4,        m3,        m5
-punpckhbw   m3,        m5
+    punpcklbw   m2,        m6,        m5
+    punpckhbw   m6,        m5
 
-pmaddubsw   m4,        m1
-pmaddubsw   m3,        m1
+    pmaddubsw   m2,        m0
+    pmaddubsw   m6,        m0
 
-movu        m5,        [r5 + 2 * r1]
+    paddw       m4,        m2
+    paddw       m3,        m6
 
-punpcklbw   m2,        m6,        m5
-punpckhbw   m6,        m5
+    pmulhrsw    m4,        m7
+    pmulhrsw    m3,        m7
 
-pmaddubsw   m2,        m0
-pmaddubsw   m6,        m0
+    packuswb    m4,        m3
 
-paddw       m4,        m2
-paddw       m3,        m6
+    movu        [r2 + r3],      m4
 
-pmulhrsw    m4,        m7
-pmulhrsw    m3,        m7
+    add         r0,        16
+    add         r2,        16
+    dec         r6d
+    jnz         .loopW
 
-packuswb    m4,        m3
+    lea         r0,        [r0 + r1 * 2 - %1]
+    lea         r2,        [r2 + r3 * 2 - %1]
 
-movu        [r2 + r3],      m4
+    dec         r4d
+    jnz        .loop
+    RET
+%endmacro
 
-add         r0,        16
-add         r2,        16
-dec         r6d
-jnz         .loopW
+    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
 
-lea         r0,        [r0 + r1 * 2 - %1]
-lea         r2,        [r2 + r3 * 2 - %1]
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_2xN 1
+INIT_XMM sse4
+cglobal filterPixelToShort_2x%1, 3, 4, 3
+    mov         r3d, r3m
+    add         r3d, r3d
 
-dec         r4d
-jnz        .loop
-RET
+    ; load constant
+    mova        m1, [pb_128]
+    mova        m2, [tab_c_64_n64]
+
+%rep %1/2
+    movd        m0, [r0]
+    pinsrd      m0, [r0 + r1], 1
+    punpcklbw   m0, m1
+    pmaddubsw   m0, m2
+
+    movd        [r2 + r3 * 0], m0
+    pextrd      [r2 + r3 * 1], m0, 2
+
+    lea         r0, [r0 + r1 * 2]
+    lea         r2, [r2 + r3 * 2]
+%endrep
+    RET
 %endmacro
+    P2S_H_2xN 4
+    P2S_H_2xN 8
+    P2S_H_2xN 16
 
-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 pixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int width, int height)
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
 ;-----------------------------------------------------------------------------
-%macro PIXEL_WH_4xN 2
-INIT_XMM ssse3
-cglobal pixelToShort_%1x%2, 3, 7, 6
+%macro P2S_H_4xN 1
+INIT_XMM sse4
+cglobal filterPixelToShort_4x%1, 3, 6, 4
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r3 * 3]
+    lea         r5, [r1 * 3]
+
+    ; load constant
+    mova        m2, [pb_128]
+    mova        m3, [tab_c_64_n64]
+
+%assign x 0
+%rep %1/4
+    movd        m0, [r0]
+    pinsrd      m0, [r0 + r1], 1
+    punpcklbw   m0, m2
+    pmaddubsw   m0, m3
+
+    movd        m1, [r0 + r1 * 2]
+    pinsrd      m1, [r0 + r5], 1
+    punpcklbw   m1, m2
+    pmaddubsw   m1, m3
+
+    movq        [r2 + r3 * 0], m0
+    movq        [r2 + r3 * 2], m1
+    movhps      [r2 + r3 * 1], m0
+    movhps      [r2 + r4], m1
+%assign x x+1
+%if (x != %1/4)
+    lea         r0, [r0 + r1 * 4]
+    lea         r2, [r2 + r3 * 4]
+%endif
+%endrep
+    RET
+%endmacro
+    P2S_H_4xN 4
+    P2S_H_4xN 8
+    P2S_H_4xN 16
+    P2S_H_4xN 32
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_6xN 1
+INIT_XMM sse4
+cglobal filterPixelToShort_6x%1, 3, 7, 6
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r1 * 3]
+    lea         r5, [r3 * 3]
+
+    ; load height
+    mov         r6d, %1/4
 
-    ; load width and height
-    mov         r3d, %1
-    mov         r4d, %2
     ; load constant
     mova        m4, [pb_128]
     mova        m5, [tab_c_64_n64]
-.loopH:
-    xor         r5d, r5d
 
-.loopW:
-    mov         r6, r0
-    movh        m0, [r6]
+.loop:
+    movh        m0, [r0]
     punpcklbw   m0, m4
     pmaddubsw   m0, m5
 
-    movh        m1, [r6 + r1]
+    movh        m1, [r0 + r1]
     punpcklbw   m1, m4
     pmaddubsw   m1, m5
 
-    movh        m2, [r6 + r1 * 2]
+    movh        m2, [r0 + r1 * 2]
     punpcklbw   m2, m4
     pmaddubsw   m2, m5
 
-    lea         r6, [r6 + r1 * 2]
-    movh        m3, [r6 + r1]
+    movh        m3, [r0 + r4]
     punpcklbw   m3, m4
     pmaddubsw   m3, m5
 
-    add         r5, 8
-    cmp         r5, r3
-    jg          .width4
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
-    je          .nextH
-    jmp         .loopW
-
-.width4:
-    movh        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
-    movh        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
-    movh        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
-    movh        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
+    movh        [r2 + r3 * 0], m0
+    pextrd      [r2 + r3 * 0 + 8], m0, 2
+    movh        [r2 + r3 * 1], m1
+    pextrd      [r2 + r3 * 1 + 8], m1, 2
+    movh        [r2 + r3 * 2], m2
+    pextrd      [r2 + r3 * 2 + 8], m2, 2
+    movh        [r2 + r5], m3
+    pextrd      [r2 + r5 + 8], m3, 2
 
-.nextH:
     lea         r0, [r0 + r1 * 4]
-    add         r2, FENC_STRIDE * 8
+    lea         r2, [r2 + r3 * 4]
 
-    sub         r4d, 4
-    jnz         .loopH
+    dec         r6d
+    jnz         .loop
     RET
 %endmacro
-PIXEL_WH_4xN 4, 4
-PIXEL_WH_4xN 4, 8
-PIXEL_WH_4xN 4, 16
+    P2S_H_6xN 8
+    P2S_H_6xN 16
 
 ;-----------------------------------------------------------------------------
-; void pixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int width, int height)
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
 ;-----------------------------------------------------------------------------
-%macro PIXEL_WH_8xN 2
+%macro P2S_H_8xN 1
 INIT_XMM ssse3
-cglobal pixelToShort_%1x%2, 3, 7, 6
+cglobal filterPixelToShort_8x%1, 3, 7, 6
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r5, [r1 * 3]
+    lea         r6, [r3 * 3]
 
-    ; load width and height
-    mov         r3d, %1
-    mov         r4d, %2
+    ; load height
+    mov         r4d, %1/4
 
     ; load constant
     mova        m4, [pb_128]
     mova        m5, [tab_c_64_n64]
 
-.loopH
-    xor         r5d, r5d
-.loopW
-    lea         r6, [r0 + r5]
-
-    movh        m0, [r6]
+.loop
+    movh        m0, [r0]
     punpcklbw   m0, m4
     pmaddubsw   m0, m5
 
-    movh        m1, [r6 + r1]
+    movh        m1, [r0 + r1]
     punpcklbw   m1, m4
     pmaddubsw   m1, m5
 
-    movh        m2, [r6 + r1 * 2]
+    movh        m2, [r0 + r1 * 2]
     punpcklbw   m2, m4
     pmaddubsw   m2, m5
 
-    lea         r6, [r6 + r1 * 2]
-    movh        m3, [r6 + r1]
+    movh        m3, [r0 + r5]
     punpcklbw   m3, m4
     pmaddubsw   m3, m5
 
-    add         r5, 8
-    cmp         r5, r3
-
-    movu        [r2 + FENC_STRIDE * 0], m0
-    movu        [r2 + FENC_STRIDE * 2], m1
-    movu        [r2 + FENC_STRIDE * 4], m2
-    movu        [r2 + FENC_STRIDE * 6], m3
-
-    je          .nextH
-    jmp         .loopW
-
+    movu        [r2 + r3 * 0], m0
+    movu        [r2 + r3 * 1], m1
+    movu        [r2 + r3 * 2], m2
+    movu        [r2 + r6 ], m3
 
-.nextH:
     lea         r0, [r0 + r1 * 4]
-    add         r2, FENC_STRIDE * 8
+    lea         r2, [r2 + r3 * 4]
 
-    sub         r4d, 4
-    jnz         .loopH
+    dec         r4d
+    jnz         .loop
     RET
 %endmacro
-PIXEL_WH_8xN 8, 8
-PIXEL_WH_8xN 8, 4
-PIXEL_WH_8xN 8, 16
-PIXEL_WH_8xN 8, 32
+    P2S_H_8xN 8
+    P2S_H_8xN 4
+    P2S_H_8xN 16
+    P2S_H_8xN 32
+    P2S_H_8xN 12
+    P2S_H_8xN 64
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_XMM ssse3
+cglobal filterPixelToShort_8x6, 3, 7, 5
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r1 * 3]
+    lea         r5, [r1 * 5]
+    lea         r6, [r3 * 3]
+
+    ; load constant
+    mova        m3, [pb_128]
+    mova        m4, [tab_c_64_n64]
+
+    movh        m0, [r0]
+    punpcklbw   m0, m3
+    pmaddubsw   m0, m4
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m3
+    pmaddubsw   m1, m4
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m3
+    pmaddubsw   m2, m4
 
+    movu        [r2 + r3 * 0], m0
+    movu        [r2 + r3 * 1], m1
+    movu        [r2 + r3 * 2], m2
+
+    movh        m0, [r0 + r4]
+    punpcklbw   m0, m3
+    pmaddubsw   m0, m4
+
+    movh        m1, [r0 + r1 * 4]
+    punpcklbw   m1, m3
+    pmaddubsw   m1, m4
+
+    movh        m2, [r0 + r5]
+    punpcklbw   m2, m3
+    pmaddubsw   m2, m4
+
+    movu        [r2 + r6 ], m0
+    movu        [r2 + r3 * 4], m1
+    lea         r2, [r2 + r3 * 4]
+    movu        [r2 + r3], m2
+
+    RET
 
 ;-----------------------------------------------------------------------------
-; void pixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int width, int height)
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
 ;-----------------------------------------------------------------------------
-%macro PIXEL_WH_16xN 2
+%macro P2S_H_16xN 1
 INIT_XMM ssse3
-cglobal pixelToShort_%1x%2, 3, 7, 6
+cglobal filterPixelToShort_16x%1, 3, 7, 6
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r3 * 3]
+    lea         r5, [r1 * 3]
 
-    ; load width and height
-    mov         r3d, %1
-    mov         r4d, %2
+   ; load height
+    mov         r6d, %1/4
 
     ; load constant
     mova        m4, [pb_128]
     mova        m5, [tab_c_64_n64]
 
-.loopH:
-    xor         r5d, r5d
-.loopW:
-    lea         r6, [r0 + r5]
-
-    movh        m0, [r6]
+.loop:
+    movh        m0, [r0]
     punpcklbw   m0, m4
     pmaddubsw   m0, m5
 
-    movh        m1, [r6 + r1]
+    movh        m1, [r0 + r1]
     punpcklbw   m1, m4
     pmaddubsw   m1, m5
 
-    movh        m2, [r6 + r1 * 2]
+    movh        m2, [r0 + r1 * 2]
     punpcklbw   m2, m4
     pmaddubsw   m2, m5
 
-    lea         r6, [r6 + r1 * 2]
-    movh        m3, [r6 + r1]
+    movh        m3, [r0 + r5]
     punpcklbw   m3, m4
     pmaddubsw   m3, m5
 
-    add         r5, 8
-    cmp         r5, r3
+    movu        [r2 + r3 * 0], m0
+    movu        [r2 + r3 * 1], m1
+    movu        [r2 + r3 * 2], m2
+    movu        [r2 + r4], m3
 
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
-    je          .nextH
-    jmp         .loopW
+    lea         r0, [r0 + 8]
 
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
 
-.nextH:
-    lea         r0, [r0 + r1 * 4]
-    add         r2, FENC_STRIDE * 8
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
 
-    sub         r4d, 4
-    jnz         .loopH
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0 + 16], m0
+    movu        [r2 + r3 * 1 + 16], m1
+    movu        [r2 + r3 * 2 + 16], m2
+    movu        [r2 + r4 + 16], m3
 
+    lea         r0, [r0 + r1 * 4 - 8]
+    lea         r2, [r2 + r3 * 4]
+
+    dec         r6d
+    jnz         .loop
     RET
 %endmacro
-PIXEL_WH_16xN 16, 16
-PIXEL_WH_16xN 16, 8
-PIXEL_WH_16xN 16, 4
-PIXEL_WH_16xN 16, 12
-PIXEL_WH_16xN 16, 32
-PIXEL_WH_16xN 16, 64
+    P2S_H_16xN 16
+    P2S_H_16xN 4
+    P2S_H_16xN 8
+    P2S_H_16xN 12
+    P2S_H_16xN 32
+    P2S_H_16xN 64
+    P2S_H_16xN 24
 
 ;-----------------------------------------------------------------------------
-; void pixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int width, int height)
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
 ;-----------------------------------------------------------------------------
-%macro PIXEL_WH_32xN 2
+%macro P2S_H_32xN 1
 INIT_XMM ssse3
-cglobal pixelToShort_%1x%2, 3, 7, 6
+cglobal filterPixelToShort_32x%1, 3, 7, 6
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r3 * 3]
+    lea         r5, [r1 * 3]
 
-    ; load width and height
-    mov         r3d, %1
-    mov         r4d, %2
+    ; load height
+    mov         r6d, %1/4
 
     ; load constant
     mova        m4, [pb_128]
     mova        m5, [tab_c_64_n64]
 
-.loopH:
-    xor         r5d, r5d
-.loopW:
-    lea         r6, [r0 + r5]
+.loop:
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0], m0
+    movu        [r2 + r3 * 1], m1
+    movu        [r2 + r3 * 2], m2
+    movu        [r2 + r4], m3
+
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
 
-    movh        m0, [r6]
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0 + 16], m0
+    movu        [r2 + r3 * 1 + 16], m1
+    movu        [r2 + r3 * 2 + 16], m2
+    movu        [r2 + r4 + 16], m3
+
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
     punpcklbw   m0, m4
     pmaddubsw   m0, m5
 
-    movh        m1, [r6 + r1]
+    movh        m1, [r0 + r1]
     punpcklbw   m1, m4
     pmaddubsw   m1, m5
 
-    movh        m2, [r6 + r1 * 2]
+    movh        m2, [r0 + r1 * 2]
     punpcklbw   m2, m4
     pmaddubsw   m2, m5
 
-    lea         r6, [r6 + r1 * 2]
-    movh        m3, [r6 + r1]
+    movh        m3, [r0 + r5]
     punpcklbw   m3, m4
     pmaddubsw   m3, m5
 
-    add         r5, 8
-    cmp         r5, r3
+    movu        [r2 + r3 * 0 + 32], m0
+    movu        [r2 + r3 * 1 + 32], m1
+    movu        [r2 + r3 * 2 + 32], m2
+    movu        [r2 + r4 + 32], m3
 
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
-    je          .nextH
-    jmp         .loopW
+    lea         r0, [r0 + 8]
 
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
 
-.nextH:
-    lea         r0, [r0 + r1 * 4]
-    add         r2, FENC_STRIDE * 8
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
 
-    sub         r4d, 4
-    jnz         .loopH
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0 + 48], m0
+    movu        [r2 + r3 * 1 + 48], m1
+    movu        [r2 + r3 * 2 + 48], m2
+    movu        [r2 + r4 + 48], m3
+
+    lea         r0, [r0 + r1 * 4 - 24]
+    lea         r2, [r2 + r3 * 4]
 
+    dec         r6d
+    jnz         .loop
     RET
 %endmacro
-PIXEL_WH_32xN 32, 32
-PIXEL_WH_32xN 32, 8
-PIXEL_WH_32xN 32, 16
-PIXEL_WH_32xN 32, 24
-PIXEL_WH_32xN 32, 64
+    P2S_H_32xN 32
+    P2S_H_32xN 8
+    P2S_H_32xN 16
+    P2S_H_32xN 24
+    P2S_H_32xN 64
+    P2S_H_32xN 48
 
 ;-----------------------------------------------------------------------------
-; void pixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int width, int height)
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
 ;-----------------------------------------------------------------------------
-%macro PIXEL_WH_64xN 2
+%macro P2S_H_32xN_avx2 1
+INIT_YMM avx2
+cglobal filterPixelToShort_32x%1, 3, 7, 3
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r5, [r1 * 3]
+    lea         r6, [r3 * 3]
+
+    ; load height
+    mov         r4d, %1/4
+
+    ; load constant
+    vpbroadcastd m2, [pw_2000]
+
+.loop:
+    pmovzxbw    m0, [r0 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + 1 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psubw       m0, m2
+    psubw       m1, m2
+    movu        [r2 + 0 * mmsize], m0
+    movu        [r2 + 1 * mmsize], m1
+
+    pmovzxbw    m0, [r0 + r1 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r1 + 1 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psubw       m0, m2
+    psubw       m1, m2
+    movu        [r2 + r3 + 0 * mmsize], m0
+    movu        [r2 + r3 + 1 * mmsize], m1
+
+    pmovzxbw    m0, [r0 + r1 * 2 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r1 * 2 + 1 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psubw       m0, m2
+    psubw       m1, m2
+    movu        [r2 + r3 * 2 + 0 * mmsize], m0
+    movu        [r2 + r3 * 2 + 1 * mmsize], m1
+
+    pmovzxbw    m0, [r0 + r5 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r5 + 1 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psubw       m0, m2
+    psubw       m1, m2
+    movu        [r2 + r6 + 0 * mmsize], m0
+    movu        [r2 + r6 + 1 * mmsize], m1
+
+    lea         r0, [r0 + r1 * 4]
+    lea         r2, [r2 + r3 * 4]
+
+    dec         r4d
+    jnz        .loop
+    RET
+%endmacro
+    P2S_H_32xN_avx2 32
+    P2S_H_32xN_avx2 8
+    P2S_H_32xN_avx2 16
+    P2S_H_32xN_avx2 24
+    P2S_H_32xN_avx2 64
+    P2S_H_32xN_avx2 48
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_64xN 1
 INIT_XMM ssse3
-cglobal pixelToShort_%1x%2, 3, 7, 6
+cglobal filterPixelToShort_64x%1, 3, 7, 6
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r3 * 3]
+    lea         r5, [r1 * 3]
 
-    ; load width and height
-    mov         r3d, %1
-    mov         r4d, %2
+    ; load height
+    mov         r6d, %1/4
 
     ; load constant
     mova        m4, [pb_128]
     mova        m5, [tab_c_64_n64]
 
-.loopH:
-    xor         r5d, r5d
-.loopW:
-    lea         r6, [r0 + r5]
+.loop:
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
 
-    movh        m0, [r6]
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0], m0
+    movu        [r2 + r3 * 1], m1
+    movu        [r2 + r3 * 2], m2
+    movu        [r2 + r4], m3
+
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
     punpcklbw   m0, m4
     pmaddubsw   m0, m5
 
-    movh        m1, [r6 + r1]
+    movh        m1, [r0 + r1]
     punpcklbw   m1, m4
     pmaddubsw   m1, m5
 
-    movh        m2, [r6 + r1 * 2]
+    movh        m2, [r0 + r1 * 2]
     punpcklbw   m2, m4
     pmaddubsw   m2, m5
 
-    lea         r6, [r6 + r1 * 2]
-    movh        m3, [r6 + r1]
+    movh        m3, [r0 + r5]
     punpcklbw   m3, m4
     pmaddubsw   m3, m5
 
-    add         r5, 8
-    cmp         r5, r3
+    movu        [r2 + r3 * 0 + 16], m0
+    movu        [r2 + r3 * 1 + 16], m1
+    movu        [r2 + r3 * 2 + 16], m2
+    movu        [r2 + r4 + 16], m3
+
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
 
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 0 - 16], m0
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 2 - 16], m1
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 4 - 16], m2
-    movu        [r2 + r5 * 2 + FENC_STRIDE * 6 - 16], m3
-    je          .nextH
-    jmp         .loopW
+    movu        [r2 + r3 * 0 + 32], m0
+    movu        [r2 + r3 * 1 + 32], m1
+    movu        [r2 + r3 * 2 + 32], m2
+    movu        [r2 + r4 + 32], m3
 
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0 + 48], m0
+    movu        [r2 + r3 * 1 + 48], m1
+    movu        [r2 + r3 * 2 + 48], m2
+    movu        [r2 + r4 + 48], m3
+
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0 + 64], m0
+    movu        [r2 + r3 * 1 + 64], m1
+    movu        [r2 + r3 * 2 + 64], m2
+    movu        [r2 + r4 + 64], m3
+
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0 + 80], m0
+    movu        [r2 + r3 * 1 + 80], m1
+    movu        [r2 + r3 * 2 + 80], m2
+    movu        [r2 + r4 + 80], m3
+
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0 + 96], m0
+    movu        [r2 + r3 * 1 + 96], m1
+    movu        [r2 + r3 * 2 + 96], m2
+    movu        [r2 + r4 + 96], m3
+
+    lea         r0, [r0 + 8]
+
+    movh        m0, [r0]
+    punpcklbw   m0, m4
+    pmaddubsw   m0, m5
+
+    movh        m1, [r0 + r1]
+    punpcklbw   m1, m4
+    pmaddubsw   m1, m5
+
+    movh        m2, [r0 + r1 * 2]
+    punpcklbw   m2, m4
+    pmaddubsw   m2, m5
+
+    movh        m3, [r0 + r5]
+    punpcklbw   m3, m4
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0 + 112], m0
+    movu        [r2 + r3 * 1 + 112], m1
+    movu        [r2 + r3 * 2 + 112], m2
+    movu        [r2 + r4 + 112], m3
+
+    lea         r0, [r0 + r1 * 4 - 56]
+    lea         r2, [r2 + r3 * 4]
+
+    dec         r6d
+    jnz         .loop
+    RET
+%endmacro
+    P2S_H_64xN 64
+    P2S_H_64xN 16
+    P2S_H_64xN 32
+    P2S_H_64xN 48
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_64xN_avx2 1
+INIT_YMM avx2
+cglobal filterPixelToShort_64x%1, 3, 7, 5
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r5, [r1 * 3]
+    lea         r6, [r3 * 3]
+
+    ; load height
+    mov         r4d, %1/4
+
+    ; load constant
+    vpbroadcastd m4, [pw_2000]
+
+.loop:
+    pmovzxbw    m0, [r0 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + 1 * mmsize/2]
+    pmovzxbw    m2, [r0 + 2 * mmsize/2]
+    pmovzxbw    m3, [r0 + 3 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psllw       m2, 6
+    psllw       m3, 6
+    psubw       m0, m4
+    psubw       m1, m4
+    psubw       m2, m4
+    psubw       m3, m4
+
+    movu        [r2 + 0 * mmsize], m0
+    movu        [r2 + 1 * mmsize], m1
+    movu        [r2 + 2 * mmsize], m2
+    movu        [r2 + 3 * mmsize], m3
+
+    pmovzxbw    m0, [r0 + r1 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r1 + 1 * mmsize/2]
+    pmovzxbw    m2, [r0 + r1 + 2 * mmsize/2]
+    pmovzxbw    m3, [r0 + r1 + 3 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psllw       m2, 6
+    psllw       m3, 6
+    psubw       m0, m4
+    psubw       m1, m4
+    psubw       m2, m4
+    psubw       m3, m4
+
+    movu        [r2 + r3 + 0 * mmsize], m0
+    movu        [r2 + r3 + 1 * mmsize], m1
+    movu        [r2 + r3 + 2 * mmsize], m2
+    movu        [r2 + r3 + 3 * mmsize], m3
+
+    pmovzxbw    m0, [r0 + r1 * 2 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r1 * 2 + 1 * mmsize/2]
+    pmovzxbw    m2, [r0 + r1 * 2 + 2 * mmsize/2]
+    pmovzxbw    m3, [r0 + r1 * 2 + 3 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psllw       m2, 6
+    psllw       m3, 6
+    psubw       m0, m4
+    psubw       m1, m4
+    psubw       m2, m4
+    psubw       m3, m4
+
+    movu        [r2 + r3 * 2 + 0 * mmsize], m0
+    movu        [r2 + r3 * 2 + 1 * mmsize], m1
+    movu        [r2 + r3 * 2 + 2 * mmsize], m2
+    movu        [r2 + r3 * 2 + 3 * mmsize], m3
+
+    pmovzxbw    m0, [r0 + r5 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r5 + 1 * mmsize/2]
+    pmovzxbw    m2, [r0 + r5 + 2 * mmsize/2]
+    pmovzxbw    m3, [r0 + r5 + 3 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psllw       m2, 6
+    psllw       m3, 6
+    psubw       m0, m4
+    psubw       m1, m4
+    psubw       m2, m4
+    psubw       m3, m4
+
+    movu        [r2 + r6 + 0 * mmsize], m0
+    movu        [r2 + r6 + 1 * mmsize], m1
+    movu        [r2 + r6 + 2 * mmsize], m2
+    movu        [r2 + r6 + 3 * mmsize], m3
 
-.nextH:
     lea         r0, [r0 + r1 * 4]
-    add         r2, FENC_STRIDE * 8
+    lea         r2, [r2 + r3 * 4]
 
-    sub         r4d, 4
-    jnz         .loopH
+    dec         r4d
+    jnz        .loop
+    RET
+%endmacro
+    P2S_H_64xN_avx2 64
+    P2S_H_64xN_avx2 16
+    P2S_H_64xN_avx2 32
+    P2S_H_64xN_avx2 48
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel src, intptr_t srcStride, int16_t dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_12xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_12x%1, 3, 7, 6
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r1 * 3]
+    lea         r6, [r3 * 3]
+    mov         r5d, %1/4
 
+    ; load constant
+    mova        m4, [pb_128]
+    mova        m5, [tab_c_64_n64]
+
+.loop:
+    movu        m0, [r0]
+    punpcklbw   m1, m0, m4
+    punpckhbw   m0, m4
+    pmaddubsw   m0, m5
+    pmaddubsw   m1, m5
+
+    movu        m2, [r0 + r1]
+    punpcklbw   m3, m2, m4
+    punpckhbw   m2, m4
+    pmaddubsw   m2, m5
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 0], m1
+    movu        [r2 + r3 * 1], m3
+
+    movh        [r2 + r3 * 0 + 16], m0
+    movh        [r2 + r3 * 1 + 16], m2
+
+    movu        m0, [r0 + r1 * 2]
+    punpcklbw   m1, m0, m4
+    punpckhbw   m0, m4
+    pmaddubsw   m0, m5
+    pmaddubsw   m1, m5
+
+    movu        m2, [r0 + r4]
+    punpcklbw   m3, m2, m4
+    punpckhbw   m2, m4
+    pmaddubsw   m2, m5
+    pmaddubsw   m3, m5
+
+    movu        [r2 + r3 * 2], m1
+    movu        [r2 + r6], m3
+
+    movh        [r2 + r3 * 2 + 16], m0
+    movh        [r2 + r6 + 16], m2
+
+    lea         r0, [r0 + r1 * 4]
+    lea         r2, [r2 + r3 * 4]
+
+    dec         r5d
+    jnz         .loop
+    RET
+%endmacro
+    P2S_H_12xN 16
+    P2S_H_12xN 32
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_24xN 1
+INIT_XMM ssse3
+cglobal filterPixelToShort_24x%1, 3, 7, 5
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r1 * 3]
+    lea         r5, [r3 * 3]
+    mov         r6d, %1/4
+
+    ; load constant
+    mova        m3, [pb_128]
+    mova        m4, [tab_c_64_n64]
+
+.loop:
+    movu        m0, [r0]
+    punpcklbw   m1, m0, m3
+    punpckhbw   m0, m3
+    pmaddubsw   m0, m4
+    pmaddubsw   m1, m4
+
+    movu        m2, [r0 + 16]
+    punpcklbw   m2, m3
+    pmaddubsw   m2, m4
+
+    movu        [r2 +  r3 * 0], m1
+    movu        [r2 +  r3 * 0 + 16], m0
+    movu        [r2 +  r3 * 0 + 32], m2
+
+    movu        m0, [r0 + r1]
+    punpcklbw   m1, m0, m3
+    punpckhbw   m0, m3
+    pmaddubsw   m0, m4
+    pmaddubsw   m1, m4
+
+    movu        m2, [r0 + r1 + 16]
+    punpcklbw   m2, m3
+    pmaddubsw   m2, m4
+
+    movu        [r2 +  r3 * 1], m1
+    movu        [r2 +  r3 * 1 + 16], m0
+    movu        [r2 +  r3 * 1 + 32], m2
+
+    movu        m0, [r0 + r1 * 2]
+    punpcklbw   m1, m0, m3
+    punpckhbw   m0, m3
+    pmaddubsw   m0, m4
+    pmaddubsw   m1, m4
+
+    movu        m2, [r0 + r1 * 2 + 16]
+    punpcklbw   m2, m3
+    pmaddubsw   m2, m4
+
+    movu        [r2 +  r3 * 2], m1
+    movu        [r2 +  r3 * 2 + 16], m0
+    movu        [r2 +  r3 * 2 + 32], m2
+
+    movu        m0, [r0 + r4]
+    punpcklbw   m1, m0, m3
+    punpckhbw   m0, m3
+    pmaddubsw   m0, m4
+    pmaddubsw   m1, m4
+
+    movu        m2, [r0 + r4 + 16]
+    punpcklbw   m2, m3
+    pmaddubsw   m2, m4
+    movu        [r2 +  r5], m1
+    movu        [r2 +  r5 + 16], m0
+    movu        [r2 +  r5 + 32], m2
+
+    lea         r0, [r0 + r1 * 4]
+    lea         r2, [r2 + r3 * 4]
+
+    dec         r6d
+    jnz         .loop
     RET
 %endmacro
-PIXEL_WH_64xN 64, 64
-PIXEL_WH_64xN 64, 16
-PIXEL_WH_64xN 64, 32
-PIXEL_WH_64xN 64, 48
+    P2S_H_24xN 32
+    P2S_H_24xN 64
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+%macro P2S_H_24xN_avx2 1
+INIT_YMM avx2
+cglobal filterPixelToShort_24x%1, 3, 7, 4
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r1 * 3]
+    lea         r5, [r3 * 3]
+    mov         r6d, %1/4
+
+    ; load constant
+    vpbroadcastd m1, [pw_2000]
+    vpbroadcastd m2, [pb_128]
+    vpbroadcastd m3, [tab_c_64_n64]
+
+.loop:
+    pmovzxbw    m0, [r0]
+    psllw       m0, 6
+    psubw       m0, m1
+    movu        [r2], m0
+
+    movu        m0, [r0 + mmsize/2]
+    punpcklbw   m0, m2
+    pmaddubsw   m0, m3
+    movu        [r2 +  r3 * 0 + mmsize], xm0
+
+    pmovzxbw    m0, [r0 + r1]
+    psllw       m0, 6
+    psubw       m0, m1
+    movu        [r2 + r3], m0
+
+    movu        m0, [r0 + r1 + mmsize/2]
+    punpcklbw   m0, m2
+    pmaddubsw   m0, m3
+    movu        [r2 +  r3 * 1 + mmsize], xm0
+
+    pmovzxbw    m0, [r0 + r1 * 2]
+    psllw       m0, 6
+    psubw       m0, m1
+    movu        [r2 + r3 * 2], m0
+
+    movu        m0, [r0 + r1 * 2 + mmsize/2]
+    punpcklbw   m0, m2
+    pmaddubsw   m0, m3
+    movu        [r2 +  r3 * 2 + mmsize], xm0
+
+    pmovzxbw    m0, [r0 + r4]
+    psllw       m0, 6
+    psubw       m0, m1
+    movu        [r2 + r5], m0
+
+    movu        m0, [r0 + r4 + mmsize/2]
+    punpcklbw   m0, m2
+    pmaddubsw   m0, m3
+    movu        [r2 + r5 + mmsize], xm0
+
+    lea         r0, [r0 + r1 * 4]
+    lea         r2, [r2 + r3 * 4]
+
+    dec         r6d
+    jnz         .loop
+    RET
+%endmacro
+    P2S_H_24xN_avx2 32
+    P2S_H_24xN_avx2 64
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_XMM ssse3
+cglobal filterPixelToShort_48x64, 3, 7, 4
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r4, [r1 * 3]
+    lea         r5, [r3 * 3]
+    mov         r6d, 16
+
+    ; load constant
+    mova        m2, [pb_128]
+    mova        m3, [tab_c_64_n64]
+
+.loop:
+    movu        m0, [r0]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 0], m1
+    movu        [r2 +  r3 * 0 + 16], m0
+
+    movu        m0, [r0 + 16]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 0 + 32], m1
+    movu        [r2 +  r3 * 0 + 48], m0
+
+    movu        m0, [r0 + 32]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 0 + 64], m1
+    movu        [r2 +  r3 * 0 + 80], m0
+
+    movu        m0, [r0 + r1]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 1], m1
+    movu        [r2 +  r3 * 1 + 16], m0
+
+    movu        m0, [r0 + r1 + 16]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 1 + 32], m1
+    movu        [r2 +  r3 * 1 + 48], m0
+
+    movu        m0, [r0 + r1 + 32]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 1 + 64], m1
+    movu        [r2 +  r3 * 1 + 80], m0
+
+    movu        m0, [r0 + r1 * 2]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 2], m1
+    movu        [r2 +  r3 * 2 + 16], m0
+
+    movu        m0, [r0 + r1 * 2 + 16]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 2 + 32], m1
+    movu        [r2 +  r3 * 2 + 48], m0
+
+    movu        m0, [r0 + r1 * 2 + 32]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r3 * 2 + 64], m1
+    movu        [r2 +  r3 * 2 + 80], m0
+
+    movu        m0, [r0 + r4]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r5], m1
+    movu        [r2 +  r5 + 16], m0
+
+    movu        m0, [r0 + r4 + 16]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r5 + 32], m1
+    movu        [r2 +  r5 + 48], m0
+
+    movu        m0, [r0 + r4 + 32]
+    punpcklbw   m1, m0, m2
+    punpckhbw   m0, m2
+    pmaddubsw   m0, m3
+    pmaddubsw   m1, m3
+
+    movu        [r2 +  r5 + 64], m1
+    movu        [r2 +  r5 + 80], m0
+
+    lea         r0, [r0 + r1 * 4]
+    lea         r2, [r2 + r3 * 4]
+
+    dec         r6d
+    jnz         .loop
+    RET
+
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal filterPixelToShort_48x64, 3,7,4
+    mov         r3d, r3m
+    add         r3d, r3d
+    lea         r5, [r1 * 3]
+    lea         r6, [r3 * 3]
+
+    ; load height
+    mov         r4d, 64/4
+
+    ; load constant
+    vpbroadcastd m3, [pw_2000]
+
+    ; just unroll(1) because it is best choice for 48x64
+.loop:
+    pmovzxbw    m0, [r0 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + 1 * mmsize/2]
+    pmovzxbw    m2, [r0 + 2 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psllw       m2, 6
+    psubw       m0, m3
+    psubw       m1, m3
+    psubw       m2, m3
+    movu        [r2 + 0 * mmsize], m0
+    movu        [r2 + 1 * mmsize], m1
+    movu        [r2 + 2 * mmsize], m2
+
+    pmovzxbw    m0, [r0 + r1 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r1 + 1 * mmsize/2]
+    pmovzxbw    m2, [r0 + r1 + 2 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psllw       m2, 6
+    psubw       m0, m3
+    psubw       m1, m3
+    psubw       m2, m3
+    movu        [r2 + r3 + 0 * mmsize], m0
+    movu        [r2 + r3 + 1 * mmsize], m1
+    movu        [r2 + r3 + 2 * mmsize], m2
+
+    pmovzxbw    m0, [r0 + r1 * 2 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r1 * 2 + 1 * mmsize/2]
+    pmovzxbw    m2, [r0 + r1 * 2 + 2 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psllw       m2, 6
+    psubw       m0, m3
+    psubw       m1, m3
+    psubw       m2, m3
+    movu        [r2 + r3 * 2 + 0 * mmsize], m0
+    movu        [r2 + r3 * 2 + 1 * mmsize], m1
+    movu        [r2 + r3 * 2 + 2 * mmsize], m2
+
+    pmovzxbw    m0, [r0 + r5 + 0 * mmsize/2]
+    pmovzxbw    m1, [r0 + r5 + 1 * mmsize/2]
+    pmovzxbw    m2, [r0 + r5 + 2 * mmsize/2]
+    psllw       m0, 6
+    psllw       m1, 6
+    psllw       m2, 6
+    psubw       m0, m3
+    psubw       m1, m3
+    psubw       m2, m3
+    movu        [r2 + r6 + 0 * mmsize], m0
+    movu        [r2 + r6 + 1 * mmsize], m1
+    movu        [r2 + r6 + 2 * mmsize], m2
+
+    lea         r0, [r0 + r1 * 4]
+    lea         r2, [r2 + r3 * 4]
+
+    dec         r4d
+    jnz        .loop
+    RET
+
 
 %macro PROCESS_LUMA_W4_4R 0
     movd        m0, [r0]
@@ -8495,36 +11678,36 @@
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_4xN 4, 4, pp
+    FILTER_VER_LUMA_4xN 4, 4, pp
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_4xN 4, 8, pp
-FILTER_VER_LUMA_AVX2_4xN 4, 8, pp
+    FILTER_VER_LUMA_4xN 4, 8, pp
+    FILTER_VER_LUMA_AVX2_4xN 4, 8, pp
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_4xN 4, 16, pp
-FILTER_VER_LUMA_AVX2_4xN 4, 16, pp
+    FILTER_VER_LUMA_4xN 4, 16, pp
+    FILTER_VER_LUMA_AVX2_4xN 4, 16, pp
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_4x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_4xN 4, 4, ps
+    FILTER_VER_LUMA_4xN 4, 4, ps
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_4x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_4xN 4, 8, ps
-FILTER_VER_LUMA_AVX2_4xN 4, 8, ps
+    FILTER_VER_LUMA_4xN 4, 8, ps
+    FILTER_VER_LUMA_AVX2_4xN 4, 8, ps
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_4x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_4xN 4, 16, ps
-FILTER_VER_LUMA_AVX2_4xN 4, 16, ps
+    FILTER_VER_LUMA_4xN 4, 16, ps
+    FILTER_VER_LUMA_AVX2_4xN 4, 16, ps
 
 %macro PROCESS_LUMA_AVX2_W8_8R 0
     movq            xm1, [r0]                       ; m1 = row 0
@@ -8895,50 +12078,50 @@
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_8xN 8, 4, pp
-FILTER_VER_LUMA_AVX2_8x4 pp
+    FILTER_VER_LUMA_8xN 8, 4, pp
+    FILTER_VER_LUMA_AVX2_8x4 pp
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_8x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_8xN 8, 8, pp
-FILTER_VER_LUMA_AVX2_8x8 pp
+    FILTER_VER_LUMA_8xN 8, 8, pp
+    FILTER_VER_LUMA_AVX2_8x8 pp
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_8x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_8xN 8, 16, pp
-FILTER_VER_LUMA_AVX2_8xN 8, 16, pp
+    FILTER_VER_LUMA_8xN 8, 16, pp
+    FILTER_VER_LUMA_AVX2_8xN 8, 16, pp
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_8x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_8xN 8, 32, pp
-FILTER_VER_LUMA_AVX2_8xN 8, 32, pp
+    FILTER_VER_LUMA_8xN 8, 32, pp
+    FILTER_VER_LUMA_AVX2_8xN 8, 32, pp
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_8xN 8, 4, ps
-FILTER_VER_LUMA_AVX2_8x4 ps
+    FILTER_VER_LUMA_8xN 8, 4, ps
+    FILTER_VER_LUMA_AVX2_8x4 ps
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_8x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_8xN 8, 8, ps
-FILTER_VER_LUMA_AVX2_8x8 ps
+    FILTER_VER_LUMA_8xN 8, 8, ps
+    FILTER_VER_LUMA_AVX2_8x8 ps
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_8x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_8xN 8, 16, ps
-FILTER_VER_LUMA_AVX2_8xN 8, 16, ps
+    FILTER_VER_LUMA_8xN 8, 16, ps
+    FILTER_VER_LUMA_AVX2_8xN 8, 16, ps
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_8x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_8xN 8, 32, ps
-FILTER_VER_LUMA_AVX2_8xN 8, 32, ps
+    FILTER_VER_LUMA_8xN 8, 32, ps
+    FILTER_VER_LUMA_AVX2_8xN 8, 32, ps
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_%3_12x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -9000,7 +12183,7 @@
 
     lea       r5, [8 * r1 - 8]
     sub       r0, r5
-%ifidn %3,pp 
+%ifidn %3,pp
     add       r2, 8
 %else
     add       r2, 16
@@ -9047,12 +12230,12 @@
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_12x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_12xN 12, 16, pp
+    FILTER_VER_LUMA_12xN 12, 16, pp
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_12x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
-FILTER_VER_LUMA_12xN 12, 16, ps
+    FILTER_VER_LUMA_12xN 12, 16, ps
 
 %macro FILTER_VER_LUMA_AVX2_12x16 1
 INIT_YMM avx2
@@ -9443,8 +12626,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_12x16 pp
-FILTER_VER_LUMA_AVX2_12x16 ps
+    FILTER_VER_LUMA_AVX2_12x16 pp
+    FILTER_VER_LUMA_AVX2_12x16 ps
 
 %macro FILTER_VER_LUMA_AVX2_16x16 1
 INIT_YMM avx2
@@ -9787,8 +12970,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_16x16 pp
-FILTER_VER_LUMA_AVX2_16x16 ps
+    FILTER_VER_LUMA_AVX2_16x16 pp
+    FILTER_VER_LUMA_AVX2_16x16 ps
 
 %macro FILTER_VER_LUMA_AVX2_16x12 1
 INIT_YMM avx2
@@ -10062,8 +13245,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_16x12 pp
-FILTER_VER_LUMA_AVX2_16x12 ps
+    FILTER_VER_LUMA_AVX2_16x12 pp
+    FILTER_VER_LUMA_AVX2_16x12 ps
 
 %macro FILTER_VER_LUMA_AVX2_16x8 1
 INIT_YMM avx2
@@ -10258,8 +13441,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_16x8 pp
-FILTER_VER_LUMA_AVX2_16x8 ps
+    FILTER_VER_LUMA_AVX2_16x8 pp
+    FILTER_VER_LUMA_AVX2_16x8 ps
 
 %macro FILTER_VER_LUMA_AVX2_16x4 1
 INIT_YMM avx2
@@ -10383,8 +13566,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_16x4 pp
-FILTER_VER_LUMA_AVX2_16x4 ps
+    FILTER_VER_LUMA_AVX2_16x4 pp
+    FILTER_VER_LUMA_AVX2_16x4 ps
 %macro FILTER_VER_LUMA_AVX2_16xN 3
 INIT_YMM avx2
 %if ARCH_X86_64 == 1
@@ -10735,10 +13918,10 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_16xN 16, 32, pp
-FILTER_VER_LUMA_AVX2_16xN 16, 64, pp
-FILTER_VER_LUMA_AVX2_16xN 16, 32, ps
-FILTER_VER_LUMA_AVX2_16xN 16, 64, ps
+    FILTER_VER_LUMA_AVX2_16xN 16, 32, pp
+    FILTER_VER_LUMA_AVX2_16xN 16, 64, pp
+    FILTER_VER_LUMA_AVX2_16xN 16, 32, ps
+    FILTER_VER_LUMA_AVX2_16xN 16, 64, ps
 
 %macro PROCESS_LUMA_AVX2_W16_16R 1
     movu            xm0, [r0]                       ; m0 = row 0
@@ -11466,8 +14649,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_24x32 pp
-FILTER_VER_LUMA_AVX2_24x32 ps
+    FILTER_VER_LUMA_AVX2_24x32 pp
+    FILTER_VER_LUMA_AVX2_24x32 ps
 
 %macro FILTER_VER_LUMA_AVX2_32xN 3
 INIT_YMM avx2
@@ -11517,10 +14700,10 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_32xN 32, 32, pp
-FILTER_VER_LUMA_AVX2_32xN 32, 64, pp
-FILTER_VER_LUMA_AVX2_32xN 32, 32, ps
-FILTER_VER_LUMA_AVX2_32xN 32, 64, ps
+    FILTER_VER_LUMA_AVX2_32xN 32, 32, pp
+    FILTER_VER_LUMA_AVX2_32xN 32, 64, pp
+    FILTER_VER_LUMA_AVX2_32xN 32, 32, ps
+    FILTER_VER_LUMA_AVX2_32xN 32, 64, ps
 
 %macro FILTER_VER_LUMA_AVX2_32x16 1
 INIT_YMM avx2
@@ -11560,9 +14743,9 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_32x16 pp
-FILTER_VER_LUMA_AVX2_32x16 ps
- 
+    FILTER_VER_LUMA_AVX2_32x16 pp
+    FILTER_VER_LUMA_AVX2_32x16 ps
+
 %macro FILTER_VER_LUMA_AVX2_32x24 1
 INIT_YMM avx2
 %if ARCH_X86_64 == 1
@@ -11620,8 +14803,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_32x24 pp
-FILTER_VER_LUMA_AVX2_32x24 ps
+    FILTER_VER_LUMA_AVX2_32x24 pp
+    FILTER_VER_LUMA_AVX2_32x24 ps
 
 %macro FILTER_VER_LUMA_AVX2_32x8 1
 INIT_YMM avx2
@@ -11663,8 +14846,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_32x8 pp
-FILTER_VER_LUMA_AVX2_32x8 ps
+    FILTER_VER_LUMA_AVX2_32x8 pp
+    FILTER_VER_LUMA_AVX2_32x8 ps
 
 %macro FILTER_VER_LUMA_AVX2_48x64 1
 INIT_YMM avx2
@@ -11722,8 +14905,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_48x64 pp
-FILTER_VER_LUMA_AVX2_48x64 ps
+    FILTER_VER_LUMA_AVX2_48x64 pp
+    FILTER_VER_LUMA_AVX2_48x64 ps
 
 %macro FILTER_VER_LUMA_AVX2_64xN 3
 INIT_YMM avx2
@@ -11781,12 +14964,12 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_64xN 64, 32, pp
-FILTER_VER_LUMA_AVX2_64xN 64, 48, pp
-FILTER_VER_LUMA_AVX2_64xN 64, 64, pp
-FILTER_VER_LUMA_AVX2_64xN 64, 32, ps
-FILTER_VER_LUMA_AVX2_64xN 64, 48, ps
-FILTER_VER_LUMA_AVX2_64xN 64, 64, ps
+    FILTER_VER_LUMA_AVX2_64xN 64, 32, pp
+    FILTER_VER_LUMA_AVX2_64xN 64, 48, pp
+    FILTER_VER_LUMA_AVX2_64xN 64, 64, pp
+    FILTER_VER_LUMA_AVX2_64xN 64, 32, ps
+    FILTER_VER_LUMA_AVX2_64xN 64, 48, ps
+    FILTER_VER_LUMA_AVX2_64xN 64, 64, ps
 
 %macro FILTER_VER_LUMA_AVX2_64x16 1
 INIT_YMM avx2
@@ -11832,8 +15015,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_64x16 pp
-FILTER_VER_LUMA_AVX2_64x16 ps
+    FILTER_VER_LUMA_AVX2_64x16 pp
+    FILTER_VER_LUMA_AVX2_64x16 ps
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_%3_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
@@ -11916,41 +15099,41 @@
     RET
 %endmacro
 
-FILTER_VER_LUMA 16, 4, pp
-FILTER_VER_LUMA 16, 8, pp
-FILTER_VER_LUMA 16, 12, pp
-FILTER_VER_LUMA 16, 16, pp
-FILTER_VER_LUMA 16, 32, pp
-FILTER_VER_LUMA 16, 64, pp
-FILTER_VER_LUMA 24, 32, pp
-FILTER_VER_LUMA 32, 8, pp
-FILTER_VER_LUMA 32, 16, pp
-FILTER_VER_LUMA 32, 24, pp
-FILTER_VER_LUMA 32, 32, pp
-FILTER_VER_LUMA 32, 64, pp
-FILTER_VER_LUMA 48, 64, pp
-FILTER_VER_LUMA 64, 16, pp
-FILTER_VER_LUMA 64, 32, pp
-FILTER_VER_LUMA 64, 48, pp
-FILTER_VER_LUMA 64, 64, pp
-
-FILTER_VER_LUMA 16, 4, ps
-FILTER_VER_LUMA 16, 8, ps
-FILTER_VER_LUMA 16, 12, ps
-FILTER_VER_LUMA 16, 16, ps
-FILTER_VER_LUMA 16, 32, ps
-FILTER_VER_LUMA 16, 64, ps
-FILTER_VER_LUMA 24, 32, ps
-FILTER_VER_LUMA 32, 8, ps
-FILTER_VER_LUMA 32, 16, ps
-FILTER_VER_LUMA 32, 24, ps
-FILTER_VER_LUMA 32, 32, ps
-FILTER_VER_LUMA 32, 64, ps
-FILTER_VER_LUMA 48, 64, ps
-FILTER_VER_LUMA 64, 16, ps
-FILTER_VER_LUMA 64, 32, ps
-FILTER_VER_LUMA 64, 48, ps
-FILTER_VER_LUMA 64, 64, ps
+    FILTER_VER_LUMA 16, 4, pp
+    FILTER_VER_LUMA 16, 8, pp
+    FILTER_VER_LUMA 16, 12, pp
+    FILTER_VER_LUMA 16, 16, pp
+    FILTER_VER_LUMA 16, 32, pp
+    FILTER_VER_LUMA 16, 64, pp
+    FILTER_VER_LUMA 24, 32, pp
+    FILTER_VER_LUMA 32, 8, pp
+    FILTER_VER_LUMA 32, 16, pp
+    FILTER_VER_LUMA 32, 24, pp
+    FILTER_VER_LUMA 32, 32, pp
+    FILTER_VER_LUMA 32, 64, pp
+    FILTER_VER_LUMA 48, 64, pp
+    FILTER_VER_LUMA 64, 16, pp
+    FILTER_VER_LUMA 64, 32, pp
+    FILTER_VER_LUMA 64, 48, pp
+    FILTER_VER_LUMA 64, 64, pp
+
+    FILTER_VER_LUMA 16, 4, ps
+    FILTER_VER_LUMA 16, 8, ps
+    FILTER_VER_LUMA 16, 12, ps
+    FILTER_VER_LUMA 16, 16, ps
+    FILTER_VER_LUMA 16, 32, ps
+    FILTER_VER_LUMA 16, 64, ps
+    FILTER_VER_LUMA 24, 32, ps
+    FILTER_VER_LUMA 32, 8, ps
+    FILTER_VER_LUMA 32, 16, ps
+    FILTER_VER_LUMA 32, 24, ps
+    FILTER_VER_LUMA 32, 32, ps
+    FILTER_VER_LUMA 32, 64, ps
+    FILTER_VER_LUMA 48, 64, ps
+    FILTER_VER_LUMA 64, 16, ps
+    FILTER_VER_LUMA 64, 32, ps
+    FILTER_VER_LUMA 64, 48, ps
+    FILTER_VER_LUMA 64, 64, ps
 
 %macro PROCESS_LUMA_SP_W4_4R 0
     movq       m0, [r0]
@@ -12036,7 +15219,7 @@
     lea       r6, [tab_LumaCoeffV + r4]
 %endif
 
-    mova      m7, [tab_c_526336]
+    mova      m7, [pd_526336]
 
     mov       dword [rsp], %2/4
 .loopH:
@@ -12110,63 +15293,49 @@
     FILTER_VER_LUMA_SP 64, 16
     FILTER_VER_LUMA_SP 16, 64
 
-; TODO: combin of U and V is more performance, but need more register
-; TODO: use two path for height alignment to 4 and otherwise may improvement 10% performance, but code is more complex, so I disable it
-INIT_XMM ssse3
-cglobal chroma_p2s, 3, 7, 4
-
-    ; load width and height
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_XMM sse4
+cglobal filterPixelToShort_4x2, 3, 4, 3
     mov         r3d, r3m
-    mov         r4d, r4m
+    add         r3d, r3d
 
     ; load constant
-    mova        m2, [pb_128]
-    mova        m3, [tab_c_64_n64]
+    mova        m1, [pb_128]
+    mova        m2, [tab_c_64_n64]
 
-.loopH:
-
-    xor         r5d, r5d
-.loopW:
-    lea         r6, [r0 + r5]
+    movd        m0, [r0]
+    pinsrd      m0, [r0 + r1], 1
+    punpcklbw   m0, m1
+    pmaddubsw   m0, m2
 
-    movh        m0, [r6]
-    punpcklbw   m0, m2
-    pmaddubsw   m0, m3
+    movq        [r2 + r3 * 0], m0
+    movhps      [r2 + r3 * 1], m0
 
-    movh        m1, [r6 + r1]
-    punpcklbw   m1, m2
-    pmaddubsw   m1, m3
+    RET
 
-    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
+;-----------------------------------------------------------------------------
+; void filterPixelToShort(pixel *src, intptr_t srcStride, int16_t *dst, int16_t dstStride)
+;-----------------------------------------------------------------------------
+INIT_XMM ssse3
+cglobal filterPixelToShort_8x2, 3, 4, 3
+    mov         r3d, r3m
+    add         r3d, r3d
 
-.nextH:
-    lea         r0, [r0 + r1 * 2]
-    add         r2, FENC_STRIDE / 2 * 4
+    ; load constant
+    mova        m1, [pb_128]
+    mova        m2, [tab_c_64_n64]
 
-    sub         r4d, 2
-    jnz         .loopH
+    movh        m0, [r0]
+    punpcklbw   m0, m1
+    pmaddubsw   m0, m2
+    movu        [r2 + r3 * 0], m0
+
+    movh        m0, [r0 + r1]
+    punpcklbw   m0, m1
+    pmaddubsw   m0, m2
+    movu        [r2 + r3 * 1], m0
 
     RET
 
@@ -12223,7 +15392,7 @@
     lea       r6, [tab_ChromaCoeffV + r4]
 %endif
 
-    mova      m6, [tab_c_526336]
+    mova      m6, [pd_526336]
 
     mov       dword [rsp], %2/4
 
@@ -12350,7 +15519,7 @@
     lea       r5, [tab_ChromaCoeffV + r4]
 %endif
 
-    mova      m5, [tab_c_526336]
+    mova      m5, [pd_526336]
 
     mov       r4d, (%2/4)
 
@@ -12380,10 +15549,10 @@
     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, 4
+    FILTER_VER_CHROMA_SP_W2_4R 2, 8
 
-FILTER_VER_CHROMA_SP_W2_4R 2, 16
+    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)
@@ -12402,7 +15571,7 @@
     lea        r5, [tab_ChromaCoeffV + r4]
 %endif
 
-    mova       m4, [tab_c_526336]
+    mova       m4, [pd_526336]
 
     movq       m0, [r0]
     movq       m1, [r0 + r1]
@@ -12454,7 +15623,7 @@
     lea       r6, [tab_ChromaCoeffV + r4]
 %endif
 
-    mova      m6, [tab_c_526336]
+    mova      m6, [pd_526336]
 
     mov       r4d, %2/4
 
@@ -12512,9 +15681,9 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_SP_W6_H4 6, 8
+    FILTER_VER_CHROMA_SP_W6_H4 6, 8
 
-FILTER_VER_CHROMA_SP_W6_H4 6, 16
+    FILTER_VER_CHROMA_SP_W6_H4 6, 16
 
 %macro PROCESS_CHROMA_SP_W8_2R 0
     movu       m1, [r0]
@@ -12566,7 +15735,7 @@
     lea       r5, [tab_ChromaCoeffV + r4]
 %endif
 
-    mova      m7, [tab_c_526336]
+    mova      m7, [pd_526336]
 
     mov       r4d, %2/2
 .loopH:
@@ -12598,15 +15767,15 @@
     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, 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
+    FILTER_VER_CHROMA_SP_W8_H2 8, 12
+    FILTER_VER_CHROMA_SP_W8_H2 8, 64
 
 
 ;-----------------------------------------------------------------------------------------------------------------------------
@@ -12658,10 +15827,10 @@
     RET
 %endmacro
 
-FILTER_HORIZ_CHROMA_2xN 2, 4
-FILTER_HORIZ_CHROMA_2xN 2, 8
+    FILTER_HORIZ_CHROMA_2xN 2, 4
+    FILTER_HORIZ_CHROMA_2xN 2, 8
 
-FILTER_HORIZ_CHROMA_2xN 2, 16
+    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)
@@ -12711,12 +15880,12 @@
     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, 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
+    FILTER_HORIZ_CHROMA_4xN 4, 32
 
 %macro PROCESS_CHROMA_W6 3
     movu       %1, [srcq]
@@ -12794,11 +15963,11 @@
     RET
 %endmacro
 
-FILTER_HORIZ_CHROMA 6, 8
-FILTER_HORIZ_CHROMA 12, 16
+    FILTER_HORIZ_CHROMA 6, 8
+    FILTER_HORIZ_CHROMA 12, 16
 
-FILTER_HORIZ_CHROMA 6, 16
-FILTER_HORIZ_CHROMA 12, 32
+    FILTER_HORIZ_CHROMA 6, 16
+    FILTER_HORIZ_CHROMA 12, 32
 
 %macro PROCESS_CHROMA_W8 3
     movu        %1, [srcq]
@@ -12857,15 +16026,15 @@
     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, 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
+    FILTER_HORIZ_CHROMA_8xN 8, 12
+    FILTER_HORIZ_CHROMA_8xN 8, 64
 
 %macro PROCESS_CHROMA_W16 4
     movu        %1, [srcq]
@@ -13027,28 +16196,28 @@
     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
+    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
 
 
 ;---------------------------------------------------------------------------------------------------------------
@@ -13144,11 +16313,11 @@
     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
+    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
 
 
 ;------------------------------------------------------------------------------------------------------------
@@ -13306,12 +16475,12 @@
     dec        r4d
     jnz        .loop
 
-RET
+    RET
 %endmacro
 
-FILTER_V_PS_W2 2, 8
+    FILTER_V_PS_W2 2, 8
 
-FILTER_V_PS_W2 2, 16
+    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)
@@ -13472,8 +16641,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_4x4 sp
-FILTER_VER_CHROMA_S_AVX2_4x4 ss
+    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
@@ -13584,8 +16753,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_4x8 sp
-FILTER_VER_CHROMA_S_AVX2_4x8 ss
+    FILTER_VER_CHROMA_S_AVX2_4x8 sp
+    FILTER_VER_CHROMA_S_AVX2_4x8 ss
 
 %macro PROCESS_CHROMA_AVX2_W4_16R 1
     movq            xm0, [r0]
@@ -13779,8 +16948,40 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_4x16 sp
-FILTER_VER_CHROMA_S_AVX2_4x16 ss
+    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
@@ -13836,8 +17037,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_4x2 sp
-FILTER_VER_CHROMA_S_AVX2_4x2 ss
+    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
@@ -13906,8 +17107,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_2x4 sp
-FILTER_VER_CHROMA_S_AVX2_2x4 ss
+    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
@@ -14085,8 +17286,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_8x8 sp
-FILTER_VER_CHROMA_S_AVX2_8x8 ss
+    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
@@ -14401,10 +17602,12 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_Nx16 sp, 16
-FILTER_VER_CHROMA_S_AVX2_Nx16 sp, 32
-FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 16
-FILTER_VER_CHROMA_S_AVX2_Nx16 ss, 32
+    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
@@ -14453,12 +17656,28 @@
 %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, 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
@@ -14567,8 +17786,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_8x4 sp
-FILTER_VER_CHROMA_S_AVX2_8x4 ss
+    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
@@ -14606,8 +17825,55 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_12x16 sp
-FILTER_VER_CHROMA_S_AVX2_12x16 ss
+    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
@@ -14860,8 +18126,257 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_16x12 sp
-FILTER_VER_CHROMA_S_AVX2_16x12 ss
+    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
@@ -14906,8 +18421,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_16x4 sp
-FILTER_VER_CHROMA_S_AVX2_16x4 ss
+    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
@@ -15097,10 +18612,10 @@
 %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
+    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
@@ -15172,8 +18687,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_8x2 sp
-FILTER_VER_CHROMA_S_AVX2_8x2 ss
+    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
@@ -15315,8 +18830,8 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_8x6 sp
-FILTER_VER_CHROMA_S_AVX2_8x6 ss
+    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
@@ -15637,15 +19152,17 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_8xN sp, 16
-FILTER_VER_CHROMA_S_AVX2_8xN sp, 32
-FILTER_VER_CHROMA_S_AVX2_8xN ss, 16
-FILTER_VER_CHROMA_S_AVX2_8xN ss, 32
+    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_32x24 1
-INIT_YMM avx2
+%macro FILTER_VER_CHROMA_S_AVX2_Nx24 2
 %if ARCH_X86_64 == 1
-cglobal interp_4tap_vert_%1_32x24, 4, 10, 10
+INIT_YMM avx2
+cglobal interp_4tap_vert_%1_%2x24, 4, 10, 10
     mov             r4d, r4m
     shl             r4d, 6
     add             r1d, r1d
@@ -15665,7 +19182,7 @@
     add             r3d, r3d
 %endif
     lea             r6, [r3 * 3]
-    mov             r9d, 4
+    mov             r9d, %2 / 8
 .loopW:
     PROCESS_CHROMA_S_AVX2_W8_16R %1
 %ifidn %1,sp
@@ -15677,13 +19194,13 @@
     dec             r9d
     jnz             .loopW
 %ifidn %1,sp
-    lea             r2, [r8 + r3 * 4 - 24]
+    lea             r2, [r8 + r3 * 4 - %2 + 8]
 %else
-    lea             r2, [r8 + r3 * 4 - 48]
+    lea             r2, [r8 + r3 * 4 - 2 * %2 + 16]
 %endif
-    lea             r0, [r7 - 48]
+    lea             r0, [r7 - 2 * %2 + 16]
     mova            m7, m9
-    mov             r9d, 4
+    mov             r9d, %2 / 8
 .loop:
     PROCESS_CHROMA_S_AVX2_W8_8R %1
 %ifidn %1,sp
@@ -15698,8 +19215,10 @@
 %endif
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_32x24 sp
-FILTER_VER_CHROMA_S_AVX2_32x24 ss
+    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
@@ -15797,8 +19316,170 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_2x8 sp
-FILTER_VER_CHROMA_S_AVX2_2x8 ss
+    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
@@ -15985,8 +19666,344 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_S_AVX2_6x8 sp
-FILTER_VER_CHROMA_S_AVX2_6x8 ss
+    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)
@@ -16031,10 +20048,10 @@
     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, 4
+    FILTER_VER_CHROMA_SS_W2_4R 2, 8
 
-FILTER_VER_CHROMA_SS_W2_4R 2, 16
+    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)
@@ -16147,9 +20164,9 @@
     RET
 %endmacro
 
-FILTER_VER_CHROMA_SS_W6_H4 6, 8
+    FILTER_VER_CHROMA_SS_W6_H4 6, 8
 
-FILTER_VER_CHROMA_SS_W6_H4 6, 16
+    FILTER_VER_CHROMA_SS_W6_H4 6, 16
 
 
 ;----------------------------------------------------------------------------------------------------------------
@@ -16194,15 +20211,15 @@
     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, 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
+    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)
@@ -16442,8 +20459,8 @@
     RET
 %endmacro
 
-FILTER_VER_LUMA_AVX2_4x4 sp
-FILTER_VER_LUMA_AVX2_4x4 ss
+    FILTER_VER_LUMA_AVX2_4x4 sp
+    FILTER_VER_LUMA_AVX2_4x4 ss
 
 %macro FILTER_VER_LUMA_AVX2_4x8 1
 INIT_YMM avx2
@@ -16588,8 +20605,8 @@
     RET
 %endmacro
 
-FILTER_VER_LUMA_AVX2_4x8 sp
-FILTER_VER_LUMA_AVX2_4x8 ss
+    FILTER_VER_LUMA_AVX2_4x8 sp
+    FILTER_VER_LUMA_AVX2_4x8 ss
 
 %macro PROCESS_LUMA_AVX2_W4_16R 1
     movq            xm0, [r0]
@@ -16833,8 +20850,8 @@
     RET
 %endmacro
 
-FILTER_VER_LUMA_AVX2_4x16 sp
-FILTER_VER_LUMA_AVX2_4x16 ss
+    FILTER_VER_LUMA_AVX2_4x16 sp
+    FILTER_VER_LUMA_AVX2_4x16 ss
 
 %macro FILTER_VER_LUMA_S_AVX2_8x8 1
 INIT_YMM avx2
@@ -17056,8 +21073,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_S_AVX2_8x8 sp
-FILTER_VER_LUMA_S_AVX2_8x8 ss
+    FILTER_VER_LUMA_S_AVX2_8x8 sp
+    FILTER_VER_LUMA_S_AVX2_8x8 ss
 
 %macro FILTER_VER_LUMA_S_AVX2_8xN 2
 INIT_YMM avx2
@@ -17446,10 +21463,10 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_S_AVX2_8xN sp, 16
-FILTER_VER_LUMA_S_AVX2_8xN sp, 32
-FILTER_VER_LUMA_S_AVX2_8xN ss, 16
-FILTER_VER_LUMA_S_AVX2_8xN ss, 32
+    FILTER_VER_LUMA_S_AVX2_8xN sp, 16
+    FILTER_VER_LUMA_S_AVX2_8xN sp, 32
+    FILTER_VER_LUMA_S_AVX2_8xN ss, 16
+    FILTER_VER_LUMA_S_AVX2_8xN ss, 32
 
 %macro PROCESS_LUMA_S_AVX2_W8_4R 1
     movu            xm0, [r0]                       ; m0 = row 0
@@ -17592,8 +21609,8 @@
     RET
 %endmacro
 
-FILTER_VER_LUMA_S_AVX2_8x4 sp
-FILTER_VER_LUMA_S_AVX2_8x4 ss
+    FILTER_VER_LUMA_S_AVX2_8x4 sp
+    FILTER_VER_LUMA_S_AVX2_8x4 ss
 
 %macro PROCESS_LUMA_AVX2_W8_16R 1
     movu            xm0, [r0]                       ; m0 = row 0
@@ -17988,12 +22005,12 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_Nx16 sp, 16
-FILTER_VER_LUMA_AVX2_Nx16 sp, 32
-FILTER_VER_LUMA_AVX2_Nx16 sp, 64
-FILTER_VER_LUMA_AVX2_Nx16 ss, 16
-FILTER_VER_LUMA_AVX2_Nx16 ss, 32
-FILTER_VER_LUMA_AVX2_Nx16 ss, 64
+    FILTER_VER_LUMA_AVX2_Nx16 sp, 16
+    FILTER_VER_LUMA_AVX2_Nx16 sp, 32
+    FILTER_VER_LUMA_AVX2_Nx16 sp, 64
+    FILTER_VER_LUMA_AVX2_Nx16 ss, 16
+    FILTER_VER_LUMA_AVX2_Nx16 ss, 32
+    FILTER_VER_LUMA_AVX2_Nx16 ss, 64
 
 %macro FILTER_VER_LUMA_AVX2_NxN 3
 INIT_YMM avx2
@@ -18047,24 +22064,24 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_NxN 16, 32, sp
-FILTER_VER_LUMA_AVX2_NxN 16, 64, sp
-FILTER_VER_LUMA_AVX2_NxN 24, 32, sp
-FILTER_VER_LUMA_AVX2_NxN 32, 32, sp
-FILTER_VER_LUMA_AVX2_NxN 32, 64, sp
-FILTER_VER_LUMA_AVX2_NxN 48, 64, sp
-FILTER_VER_LUMA_AVX2_NxN 64, 32, sp
-FILTER_VER_LUMA_AVX2_NxN 64, 48, sp
-FILTER_VER_LUMA_AVX2_NxN 64, 64, sp
-FILTER_VER_LUMA_AVX2_NxN 16, 32, ss
-FILTER_VER_LUMA_AVX2_NxN 16, 64, ss
-FILTER_VER_LUMA_AVX2_NxN 24, 32, ss
-FILTER_VER_LUMA_AVX2_NxN 32, 32, ss
-FILTER_VER_LUMA_AVX2_NxN 32, 64, ss
-FILTER_VER_LUMA_AVX2_NxN 48, 64, ss
-FILTER_VER_LUMA_AVX2_NxN 64, 32, ss
-FILTER_VER_LUMA_AVX2_NxN 64, 48, ss
-FILTER_VER_LUMA_AVX2_NxN 64, 64, ss
+    FILTER_VER_LUMA_AVX2_NxN 16, 32, sp
+    FILTER_VER_LUMA_AVX2_NxN 16, 64, sp
+    FILTER_VER_LUMA_AVX2_NxN 24, 32, sp
+    FILTER_VER_LUMA_AVX2_NxN 32, 32, sp
+    FILTER_VER_LUMA_AVX2_NxN 32, 64, sp
+    FILTER_VER_LUMA_AVX2_NxN 48, 64, sp
+    FILTER_VER_LUMA_AVX2_NxN 64, 32, sp
+    FILTER_VER_LUMA_AVX2_NxN 64, 48, sp
+    FILTER_VER_LUMA_AVX2_NxN 64, 64, sp
+    FILTER_VER_LUMA_AVX2_NxN 16, 32, ss
+    FILTER_VER_LUMA_AVX2_NxN 16, 64, ss
+    FILTER_VER_LUMA_AVX2_NxN 24, 32, ss
+    FILTER_VER_LUMA_AVX2_NxN 32, 32, ss
+    FILTER_VER_LUMA_AVX2_NxN 32, 64, ss
+    FILTER_VER_LUMA_AVX2_NxN 48, 64, ss
+    FILTER_VER_LUMA_AVX2_NxN 64, 32, ss
+    FILTER_VER_LUMA_AVX2_NxN 64, 48, ss
+    FILTER_VER_LUMA_AVX2_NxN 64, 64, ss
 
 %macro FILTER_VER_LUMA_S_AVX2_12x16 1
 INIT_YMM avx2
@@ -18102,8 +22119,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_S_AVX2_12x16 sp
-FILTER_VER_LUMA_S_AVX2_12x16 ss
+    FILTER_VER_LUMA_S_AVX2_12x16 sp
+    FILTER_VER_LUMA_S_AVX2_12x16 ss
 
 %macro FILTER_VER_LUMA_S_AVX2_16x12 1
 INIT_YMM avx2
@@ -18416,8 +22433,8 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_S_AVX2_16x12 sp
-FILTER_VER_LUMA_S_AVX2_16x12 ss
+    FILTER_VER_LUMA_S_AVX2_16x12 sp
+    FILTER_VER_LUMA_S_AVX2_16x12 ss
 
 %macro FILTER_VER_LUMA_S_AVX2_16x4 1
 INIT_YMM avx2
@@ -18464,8 +22481,8 @@
     RET
 %endmacro
 
-FILTER_VER_LUMA_S_AVX2_16x4 sp
-FILTER_VER_LUMA_S_AVX2_16x4 ss
+    FILTER_VER_LUMA_S_AVX2_16x4 sp
+    FILTER_VER_LUMA_S_AVX2_16x4 ss
 
 %macro PROCESS_LUMA_S_AVX2_W8_8R 1
     movu            xm0, [r0]                       ; m0 = row 0
@@ -18701,10 +22718,10 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_AVX2_Nx8 sp, 32
-FILTER_VER_LUMA_AVX2_Nx8 sp, 16
-FILTER_VER_LUMA_AVX2_Nx8 ss, 32
-FILTER_VER_LUMA_AVX2_Nx8 ss, 16
+    FILTER_VER_LUMA_AVX2_Nx8 sp, 32
+    FILTER_VER_LUMA_AVX2_Nx8 sp, 16
+    FILTER_VER_LUMA_AVX2_Nx8 ss, 32
+    FILTER_VER_LUMA_AVX2_Nx8 ss, 16
 
 %macro FILTER_VER_LUMA_S_AVX2_32x24 1
 INIT_YMM avx2
@@ -18764,13 +22781,13 @@
 %endif
 %endmacro
 
-FILTER_VER_LUMA_S_AVX2_32x24 sp
-FILTER_VER_LUMA_S_AVX2_32x24 ss
+    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 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_32x32, 4,7,6
     mov             r4d, r4m
     mov             r5d, r5m
@@ -18832,12 +22849,12 @@
     add                r0,           r1
     dec               r6d
     jnz                .loop
-   RET
+    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 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_16x16, 4,7,6
     mov             r4d, r4m
     mov             r5d, r5m
@@ -18885,13 +22902,13 @@
     add                r0,          r1
     dec                r6d
     jnz                .loop
-   RET
+    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 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_%1x%2, 4,7,6
     mov                    r4d,        r4m
     mov                    r5d,        r5m
@@ -18947,12 +22964,14 @@
     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 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_%1x%2, 4,7,6
     mov                r4d,          r4m
     mov                r5d,          r5m
@@ -19019,13 +23038,15 @@
     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 , 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 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_4x4, 4,7,5
     mov             r4d, r4m
     mov             r5d, r5m
@@ -19104,7 +23125,7 @@
     lea               r2,           [r2 + r3 * 2]
     movhps            [r2],         xm3
 .end
-   RET
+    RET
 
 cglobal interp_4tap_horiz_ps_4x2, 4,7,5
     mov             r4d, r4m
@@ -19173,13 +23194,13 @@
     lea               r2,           [r2 + r3 * 2]
     movhps            [r2],         xm3
 .end
-   RET
+    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 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_%1x%2, 4,7,5
     mov             r4d, r4m
     mov             r5d, r5m
@@ -19264,7 +23285,7 @@
     lea               r2,           [r2 + r3 * 2]
     movhps            [r2],         xm3
 .end
-RET
+    RET
 %endmacro
 
     IPFILTER_CHROMA_PS_4xN_AVX2  4 , 8
@@ -19272,7 +23293,7 @@
 ;-----------------------------------------------------------------------------------------------------------------------------
 ; void interp_4tap_horiz_ps_8x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
 ;-----------------------------------------------------------------------------------------------------------------------------;
-INIT_YMM avx2 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_8x8, 4,7,6
     mov             r4d, r4m
     mov             r5d, r5m
@@ -19341,9 +23362,9 @@
     vpermq            m3,           m3,          11011000b
     movu             [r2],         xm3
 .end
-   RET
+    RET
 
-INIT_YMM avx2 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_pp_4x2, 4,6,4
     mov             r4d, r4m
 %ifdef PIC
@@ -19436,9 +23457,11 @@
     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, 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
@@ -19512,15 +23535,17 @@
     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 , 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 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_pp_%1x%2, 4,6,6
     mov             r4d, r4m
 
@@ -19576,8 +23601,8 @@
     RET
 %endmacro
 
-IPFILTER_CHROMA_PP_4xN_AVX2  4 , 8
-IPFILTER_CHROMA_PP_4xN_AVX2  4 , 16
+    IPFILTER_CHROMA_PP_4xN_AVX2  4 , 8
+    IPFILTER_CHROMA_PP_4xN_AVX2  4 , 16
 
 %macro IPFILTER_LUMA_PS_32xN_AVX2 2
 INIT_YMM avx2
@@ -19674,11 +23699,11 @@
     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
+    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
@@ -20003,10 +24028,12 @@
     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 , 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
@@ -20144,16 +24171,16 @@
     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
+    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 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_8x%1, 4,7,6
     mov                r4d,             r4m
     mov                r5d,             r5m
@@ -20218,7 +24245,7 @@
     vpermq             m3,              m3,          11011000b
     movu               [r2],            xm3
 .end
-   RET
+    RET
 %endmacro
 
     IPFILTER_CHROMA_PS_8xN_AVX2  2
@@ -20226,6 +24253,8 @@
     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
@@ -20253,7 +24282,7 @@
     movhps             xm2,            [r0 + r6]
 
     vinserti128        m1,             m1,          xm2,          1
-    pshufb             m1,             [interp4_hps_shuf]
+    pshufb             m1,             [interp4_hpp_shuf]
     pmaddubsw          m1,             m0
     pmaddwd            m1,             [pw_1]
     vextracti128       xm2,            m1,          1
@@ -20275,7 +24304,7 @@
     movhps             xm1,            [r0 + r1]
     movq               xm2,            [r0 + r1 * 2]
     vinserti128        m1,             m1,          xm2,          1
-    pshufb             m1,             [interp4_hps_shuf]
+    pshufb             m1,             [interp4_hpp_shuf]
     pmaddubsw          m1,             m0
     pmaddwd            m1,             [pw_1]
     vextracti128       xm2,            m1,          1
@@ -20306,7 +24335,7 @@
     sub               r0,             r1
 
 .label
-    mova              m4,            [interp4_hps_shuf]
+    mova              m4,            [interp4_hpp_shuf]
     mova              m5,            [pw_1]
     dec               r0
     lea               r4,            [r1 * 3]
@@ -20488,7 +24517,7 @@
 ;-----------------------------------------------------------------------------------------------------------------------------
 ; void interp_4tap_horiz_ps_6x8(pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt)
 ;-----------------------------------------------------------------------------------------------------------------------------;
-INIT_YMM avx2 
+INIT_YMM avx2
 cglobal interp_4tap_horiz_ps_6x8, 4,7,6
     mov                r4d,            r4m
     mov                r5d,            r5m
@@ -20556,3 +24585,1024 @@
     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_1.6.tar.gz/source/common/x86/ipfilter8.h -> x265_1.7.tar.gz/source/common/x86/ipfilter8.h Changed

@@ -289,16 +289,114 @@
     SETUP_CHROMA_420_HORIZ_FUNC_DEF(64, 16, cpu); \
     SETUP_CHROMA_420_HORIZ_FUNC_DEF(16, 64, cpu)
 
-void x265_chroma_p2s_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, int width, int height);
-void x265_luma_p2s_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, int width, int height);
+void x265_filterPixelToShort_4x4_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_4x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_4x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x4_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x4_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x12_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x24_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x48_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_24x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_12x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_48x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x4_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x8_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x12_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x16_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x32_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x64_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x8_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x16_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x24_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x32_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x64_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x16_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x32_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x48_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x64_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_24x32_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_48x64_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+
+#define SETUP_CHROMA_P2S_FUNC_DEF(W, H, cpu) \
+    void x265_filterPixelToShort_ ## W ## x ## H ## cpu(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+
+#define CHROMA_420_P2S_FILTERS_SSSE3(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(4, 2, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(8, 2, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(8, 6, cpu);
+
+#define CHROMA_420_P2S_FILTERS_SSE4(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(2, 4, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(2, 8, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(6, 8, cpu);
+
+#define CHROMA_422_P2S_FILTERS_SSSE3(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(4, 32, cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(8, 12, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(8, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(12, 32, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 24, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(24, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 48, cpu);
+
+#define CHROMA_422_P2S_FILTERS_SSE4(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(2, 8, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(2, 16, cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(6, 16, cpu);
+
+#define CHROMA_420_P2S_FILTERS_AVX2(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 4, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 8, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 12, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 16, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 32, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(24, 32, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 8, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 16, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 24, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 32, cpu);
+
+#define CHROMA_422_P2S_FILTERS_AVX2(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 8, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 16, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 24, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 32, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(24, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 16, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 32, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 48, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 64, cpu);
 
 CHROMA_420_VERT_FILTERS(_sse2);
 CHROMA_420_HORIZ_FILTERS(_sse4);
 CHROMA_420_VERT_FILTERS_SSE4(_sse4);
+CHROMA_420_P2S_FILTERS_SSSE3(_ssse3);
+CHROMA_420_P2S_FILTERS_SSE4(_sse4);
+CHROMA_420_P2S_FILTERS_AVX2(_avx2);
 
 CHROMA_422_VERT_FILTERS(_sse2);
 CHROMA_422_HORIZ_FILTERS(_sse4);
 CHROMA_422_VERT_FILTERS_SSE4(_sse4);
+CHROMA_422_P2S_FILTERS_SSE4(_sse4);
+CHROMA_422_P2S_FILTERS_SSSE3(_ssse3);
+CHROMA_422_P2S_FILTERS_AVX2(_avx2);
 
 CHROMA_444_VERT_FILTERS(_sse2);
 CHROMA_444_HORIZ_FILTERS(_sse4);
@@ -572,6 +670,48 @@
     SETUP_CHROMA_SS_FUNC_DEF(64, 16, cpu); \
     SETUP_CHROMA_SS_FUNC_DEF(16, 64, cpu);
 
+#define SETUP_CHROMA_P2S_FUNC_DEF(W, H, cpu) \
+    void x265_filterPixelToShort_ ## W ## x ## H ## cpu(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+
+#define CHROMA_420_P2S_FILTERS_SSE4(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(2, 4, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(2, 8, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(4, 2, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(6, 8, cpu); 
+
+#define CHROMA_420_P2S_FILTERS_SSSE3(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(8, 2, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(8, 6, cpu);
+
+#define CHROMA_422_P2S_FILTERS_SSE4(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(2, 8, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(2, 16, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(6, 16, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(4, 32, cpu);
+
+#define CHROMA_422_P2S_FILTERS_SSSE3(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(8, 12, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(8, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(12, 32, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 24, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(16, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(24, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 48, cpu);
+
+#define CHROMA_420_P2S_FILTERS_AVX2(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(24, 32, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 8, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 16, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 24, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 32, cpu);
+
+#define CHROMA_422_P2S_FILTERS_AVX2(cpu) \
+    SETUP_CHROMA_P2S_FUNC_DEF(24, 64, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 16, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 32, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 48, cpu); \
+    SETUP_CHROMA_P2S_FUNC_DEF(32, 64, cpu);
+
 CHROMA_420_FILTERS(_sse4);
 CHROMA_420_FILTERS(_avx2);
 CHROMA_420_SP_FILTERS(_sse2);
@@ -582,19 +722,32 @@
 CHROMA_420_SS_FILTERS_SSE4(_sse4);
 CHROMA_420_SS_FILTERS(_avx2);
 CHROMA_420_SS_FILTERS_SSE4(_avx2);
+CHROMA_420_P2S_FILTERS_SSE4(_sse4);
+CHROMA_420_P2S_FILTERS_SSSE3(_ssse3);
+CHROMA_420_P2S_FILTERS_AVX2(_avx2);
 
 CHROMA_422_FILTERS(_sse4);
 CHROMA_422_FILTERS(_avx2);
 CHROMA_422_SP_FILTERS(_sse2);
+CHROMA_422_SP_FILTERS(_avx2);
 CHROMA_422_SP_FILTERS_SSE4(_sse4);
+CHROMA_422_SP_FILTERS_SSE4(_avx2);
 CHROMA_422_SS_FILTERS(_sse2);
+CHROMA_422_SS_FILTERS(_avx2);
 CHROMA_422_SS_FILTERS_SSE4(_sse4);
+CHROMA_422_SS_FILTERS_SSE4(_avx2);
+CHROMA_422_P2S_FILTERS_SSE4(_sse4);
+CHROMA_422_P2S_FILTERS_SSSE3(_ssse3);
+CHROMA_422_P2S_FILTERS_AVX2(_avx2);
+void x265_interp_4tap_vert_ss_2x4_avx2(const int16_t* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_sp_2x4_avx2(const int16_t* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int coeffIdx);
 
 CHROMA_444_FILTERS(_sse4);
 CHROMA_444_SP_FILTERS(_sse4);
 CHROMA_444_SS_FILTERS(_sse2);
-
-void x265_chroma_p2s_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, int width, int height);
+CHROMA_444_FILTERS(_avx2);
+CHROMA_444_SP_FILTERS(_avx2);
+CHROMA_444_SS_FILTERS(_avx2);
 
 #undef SETUP_CHROMA_FUNC_DEF
 #undef SETUP_CHROMA_SP_FUNC_DEF
@@ -623,29 +776,155 @@
 LUMA_FILTERS(_avx2);
 LUMA_SP_FILTERS(_avx2);
 LUMA_SS_FILTERS(_avx2);
-void x265_interp_8tap_hv_pp_8x8_sse4(const pixel* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int idxX, int idxY);
-void x265_pixelToShort_4x4_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_4x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_4x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_8x4_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_8x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_8x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_8x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_16x4_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_16x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_16x12_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_16x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_16x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_16x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_32x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_32x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_32x24_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_32x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_32x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_64x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_64x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_64x48_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
-void x265_pixelToShort_64x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst);
+void x265_interp_8tap_hv_pp_8x8_ssse3(const pixel* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int idxX, int idxY);
+void x265_interp_8tap_hv_pp_16x16_avx2(const pixel* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int idxX, int idxY);
+void x265_filterPixelToShort_4x4_sse4(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_4x8_sse4(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_4x16_sse4(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x4_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_8x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x4_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x12_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_16x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x8_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x24_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x48_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_12x16_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_24x32_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_48x64_ssse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x8_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x16_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x24_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x32_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_32x64_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x16_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x32_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x48_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_64x64_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_48x64_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_filterPixelToShort_24x32_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
+void x265_interp_4tap_horiz_pp_2x4_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_2x8_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_2x16_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_4x2_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_4x4_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_4x8_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_4x16_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_4x32_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_6x8_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_6x16_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_8x2_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_8x4_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_8x6_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_8x8_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_8x12_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_8x16_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_8x32_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_8x64_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_12x16_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_12x32_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_16x4_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_16x8_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_16x12_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_16x16_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_16x24_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_16x32_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_16x64_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_24x32_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_24x64_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_32x8_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_32x16_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_32x24_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_32x32_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_32x48_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_32x64_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_48x64_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_64x16_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_64x32_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_64x48_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_64x64_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_4x4_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_4x8_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_4x16_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_8x4_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_8x8_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_8x16_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_8x32_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_12x16_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_16x4_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_16x8_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_16x12_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_16x16_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_16x32_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_16x64_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_24x32_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_32x8_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_32x16_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_32x24_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_32x32_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_32x64_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_48x64_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_64x16_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_64x32_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_64x48_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_pp_64x64_sse2(const pixel* src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_8tap_horiz_ps_4x4_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_4x8_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_4x16_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_8x4_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_8x8_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_8x16_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_8x32_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_12x16_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_16x4_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_16x8_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_16x12_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_16x16_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_16x32_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_16x64_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_24x32_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_32x8_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_32x16_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_32x24_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_32x32_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_32x64_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_48x64_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_64x16_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_64x32_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_64x48_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_horiz_ps_64x64_sse2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
+void x265_interp_8tap_hv_pp_8x8_sse3(const pixel* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int idxX, int idxY);
+void x265_interp_4tap_vert_pp_2x4_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_2x8_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_2x16_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_4x2_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_4x4_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_4x8_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_4x16_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_4x32_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+#ifdef X86_64
+void x265_interp_4tap_vert_pp_6x8_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_6x16_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_8x2_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_8x4_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_8x6_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_8x8_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_8x12_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_8x16_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_8x32_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_vert_pp_8x64_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+#endif
 #undef LUMA_FILTERS
 #undef LUMA_SP_FILTERS
 #undef LUMA_SS_FILTERS

x265_1.6.tar.gz/source/common/x86/loopfilter.asm -> x265_1.7.tar.gz/source/common/x86/loopfilter.asm Changed

@@ -28,31 +28,39 @@
 %include "x86inc.asm"
 
 SECTION_RODATA 32
-pb_31:      times 16 db 31
-pb_15:      times 16 db 15
+pb_31:      times 32 db 31
+pb_15:      times 32 db 15
+pb_movemask_32:  times 32 db 0x00
+                 times 32 db 0xFF
 
 SECTION .text
 cextern pb_1
 cextern pb_128
 cextern pb_2
 cextern pw_2
+cextern pb_movemask
 
 
 ;============================================================================================================
-; void saoCuOrgE0(pixel * rec, int8_t * offsetEo, int lcuWidth, int8_t signLeft)
+; void saoCuOrgE0(pixel * rec, int8_t * offsetEo, int lcuWidth, int8_t* signLeft, intptr_t stride)
 ;============================================================================================================
 INIT_XMM sse4
-cglobal saoCuOrgE0, 4, 4, 8, rec, offsetEo, lcuWidth, signLeft
+cglobal saoCuOrgE0, 5, 5, 8, rec, offsetEo, lcuWidth, signLeft, stride
 
-    neg         r3                          ; r3 = -signLeft
-    movzx       r3d, r3b
-    movd        m0, r3d
-    mova        m4, [pb_128]                ; m4 = [80]
-    pxor        m5, m5                      ; m5 = 0
-    movu        m6, [r1]                    ; m6 = offsetEo
+    mov         r4d, r4m
+    mova        m4,  [pb_128]                ; m4 = [80]
+    pxor        m5,  m5                      ; m5 = 0
+    movu        m6,  [r1]                    ; m6 = offsetEo
+
+    movzx       r1d, byte [r3]
+    inc         r3
+    neg         r1b
+    movd        m0, r1d
+    lea         r1, [r0 + r4]
+    mov         r4d, r2d
 
 .loop:
-    movu        m7, [r0]                    ; m1 = rec[x]
+    movu        m7, [r0]                    ; m7 = rec[x]
     movu        m2, [r0 + 1]                ; m2 = rec[x+1]
 
     pxor        m1, m7, m4
@@ -69,7 +77,7 @@
     pxor        m0, m0
     palignr     m0, m2, 15
     paddb       m2, m3
-    paddb       m2, [pb_2]                  ; m1 = uiEdgeType
+    paddb       m2, [pb_2]                  ; m2 = uiEdgeType
     pshufb      m3, m6, m2
     pmovzxbw    m2, m7                      ; rec
     punpckhbw   m7, m5
@@ -84,6 +92,97 @@
     add         r0q, 16
     sub         r2d, 16
     jnz        .loop
+
+    movzx       r3d, byte [r3]
+    neg         r3b
+    movd        m0, r3d
+.loopH:
+    movu        m7, [r1]                    ; m7 = rec[x]
+    movu        m2, [r1 + 1]                ; m2 = rec[x+1]
+
+    pxor        m1, m7, m4
+    pxor        m3, m2, m4
+    pcmpgtb     m2, m1, m3
+    pcmpgtb     m3, m1
+    pand        m2, [pb_1]
+    por         m2, m3
+
+    pslldq      m3, m2, 1
+    por         m3, m0
+
+    psignb      m3, m4                      ; m3 = signLeft
+    pxor        m0, m0
+    palignr     m0, m2, 15
+    paddb       m2, m3
+    paddb       m2, [pb_2]                  ; m2 = uiEdgeType
+    pshufb      m3, m6, m2
+    pmovzxbw    m2, m7                      ; rec
+    punpckhbw   m7, m5
+    pmovsxbw    m1, m3                      ; offsetEo
+    punpckhbw   m3, m3
+    psraw       m3, 8
+    paddw       m2, m1
+    paddw       m7, m3
+    packuswb    m2, m7
+    movu        [r1], m2
+
+    add         r1q, 16
+    sub         r4d, 16
+    jnz        .loopH
+    RET
+
+INIT_YMM avx2
+cglobal saoCuOrgE0, 5, 5, 7, rec, offsetEo, lcuWidth, signLeft, stride
+
+    mov                 r4d,        r4m
+    vbroadcasti128      m4,         [pb_128]                   ; m4 = [80]
+    vbroadcasti128      m6,         [r1]                       ; m6 = offsetEo
+    movzx               r1d,        byte [r3]
+    neg                 r1b
+    movd                xm0,        r1d
+    movzx               r1d,        byte [r3 + 1]
+    neg                 r1b
+    movd                xm1,        r1d
+    vinserti128         m0,         m0,        xm1,           1
+
+.loop:
+    movu                xm5,        [r0]                       ; xm5 = rec[x]
+    movu                xm2,        [r0 + 1]                   ; xm2 = rec[x + 1]
+    vinserti128         m5,         m5,        [r0 + r4],     1
+    vinserti128         m2,         m2,        [r0 + r4 + 1], 1
+
+    pxor                m1,         m5,        m4
+    pxor                m3,         m2,        m4
+    pcmpgtb             m2,         m1,        m3
+    pcmpgtb             m3,         m1
+    pand                m2,         [pb_1]
+    por                 m2,         m3
+
+    pslldq              m3,         m2,        1
+    por                 m3,         m0
+
+    psignb              m3,         m4                         ; m3 = signLeft
+    pxor                m0,         m0
+    palignr             m0,         m2,        15
+    paddb               m2,         m3
+    paddb               m2,         [pb_2]                     ; m2 = uiEdgeType
+    pshufb              m3,         m6,        m2
+    pmovzxbw            m2,         xm5                        ; rec
+    vextracti128        xm5,        m5,        1
+    pmovzxbw            m5,         xm5
+    pmovsxbw            m1,         xm3                        ; offsetEo
+    vextracti128        xm3,        m3,        1
+    pmovsxbw            m3,         xm3
+    paddw               m2,         m1
+    paddw               m5,         m3
+    packuswb            m2,         m5
+    vpermq              m2,         m2,        11011000b
+    movu                [r0],       xm2
+    vextracti128        [r0 + r4],  m2,        1
+
+    add                 r0q,        16
+    sub                 r2d,        16
+    jnz                 .loop
     RET
 
 ;==================================================================================================
@@ -94,117 +193,382 @@
     mov         r3d, r3m
     mov         r4d, r4m
     pxor        m0,    m0                      ; m0 = 0
-    movu        m6,    [pb_2]                  ; m6 = [2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2]
+    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
-         movu        m1,    [r0]                    ; m1 = pRec[x]
-         movu        m2,    [r0 + r3]               ; m2 = pRec[x + iStride]
-
-         pxor        m3,    m1,    m7
-         pxor        m4,    m2,    m7
-         pcmpgtb     m2,    m3,    m4
-         pcmpgtb     m4,    m3
-         pand        m2,    [pb_1]
-         por         m2,    m4
-
-         movu        m3,    [r1]                    ; m3 = m_iUpBuff1
-
-         paddb       m3,    m2
-         paddb       m3,    m6
-
-         movu        m4,    [r2]                    ; m4 = m_iOffsetEo
-         pshufb      m5,    m4,    m3
-
-         psubb       m3,    m0,    m2
-         movu        [r1],  m3
-
-         pmovzxbw    m2,    m1
-         punpckhbw   m1,    m0
-         pmovsxbw    m3,    m5
-         punpckhbw   m5,    m5
-         psraw       m5,    8
-
-         paddw       m2,    m3
-         paddw       m1,    m5
-         packuswb    m2,    m1
-         movu        [r0],  m2
-
-         add         r0,    16
-         add         r1,    16
-         dec         r4d
-         jnz         .loop
+.loop
+    movu        m1,    [r0]                    ; m1 = pRec[x]
+    movu        m2,    [r0 + r3]               ; m2 = pRec[x + iStride]
+
+    pxor        m3,    m1,    m7
+    pxor        m4,    m2,    m7
+    pcmpgtb     m2,    m3,    m4
+    pcmpgtb     m4,    m3
+    pand        m2,    [pb_1]
+    por         m2,    m4
+
+    movu        m3,    [r1]                    ; m3 = m_iUpBuff1
+
+    paddb       m3,    m2
+    paddb       m3,    m6
+
+    movu        m4,    [r2]                    ; m4 = m_iOffsetEo
+    pshufb      m5,    m4,    m3
+
+    psubb       m3,    m0,    m2
+    movu        [r1],  m3
+
+    pmovzxbw    m2,    m1
+    punpckhbw   m1,    m0
+    pmovsxbw    m3,    m5
+    punpckhbw   m5,    m5
+    psraw       m5,    8
+
+    paddw       m2,    m3
+    paddw       m1,    m5
+    packuswb    m2,    m1
+    movu        [r0],  m2
+
+    add         r0,    16
+    add         r1,    16
+    dec         r4d
+    jnz         .loop
+    RET
+
+INIT_YMM avx2
+cglobal saoCuOrgE1, 3, 5, 8, pRec, m_iUpBuff1, m_iOffsetEo, iStride, iLcuWidth
+    mov           r3d,    r3m
+    mov           r4d,    r4m
+    movu          xm0,    [r2]                    ; xm0 = m_iOffsetEo
+    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
+    movu          xm1,    [r0]                    ; xm1 = pRec[x]
+    movu          xm2,    [r0 + r3]               ; xm2 = pRec[x + iStride]
+
+    pxor          xm3,    xm1,    xm7
+    pxor          xm4,    xm2,    xm7
+    pcmpgtb       xm2,    xm3,    xm4
+    pcmpgtb       xm4,    xm3
+    pand          xm2,    [pb_1]
+    por           xm2,    xm4
+
+    movu          xm3,    [r1]                    ; xm3 = m_iUpBuff1
+
+    paddb         xm3,    xm2
+    paddb         xm3,    xm6
+
+    pshufb        xm5,    xm0,    xm3
+    pxor          xm4,    xm4
+    psubb         xm3,    xm4,    xm2
+    movu          [r1],   xm3
+
+    pmovzxbw      m2,     xm1
+    pmovsxbw      m3,     xm5
+
+    paddw         m2,     m3
+    vextracti128  xm3,    m2,     1
+    packuswb      xm2,    xm3
+    movu          [r0],   xm2
+
+    add           r0,     16
+    add           r1,     16
+    dec           r4d
+    jnz           .loop
+    RET
+
+;========================================================================================================
+; void saoCuOrgE1_2Rows(pixel *pRec, int8_t *m_iUpBuff1, int8_t *m_iOffsetEo, Int iStride, Int iLcuWidth)
+;========================================================================================================
+INIT_XMM sse4
+cglobal saoCuOrgE1_2Rows, 3, 5, 8, pRec, m_iUpBuff1, m_iOffsetEo, iStride, iLcuWidth
+    mov         r3d,        r3m
+    mov         r4d,        r4m
+    pxor        m0,         m0                      ; m0 = 0
+    mova        m7,         [pb_128]
+    shr         r4d,        4
+.loop
+    movu        m1,         [r0]                    ; m1 = pRec[x]
+    movu        m2,         [r0 + r3]               ; m2 = pRec[x + iStride]
+
+    pxor        m3,         m1,         m7
+    pxor        m4,         m2,         m7
+    pcmpgtb     m6,         m3,         m4
+    pcmpgtb     m5,         m4,         m3
+    pand        m6,         [pb_1]
+    por         m6,         m5
+
+    movu        m5,         [r0 + r3 * 2]
+    pxor        m3,         m5,         m7
+    pcmpgtb     m5,         m4,         m3
+    pcmpgtb     m3,         m4
+    pand        m5,         [pb_1]
+    por         m5,         m3
+
+    movu        m3,         [r1]                    ; m3 = m_iUpBuff1
+    paddb       m3,         m6
+    paddb       m3,         [pb_2]
+
+    movu        m4,         [r2]                    ; m4 = m_iOffsetEo
+    pshufb      m4,         m3
+
+    psubb       m3,         m0,         m6
+    movu        [r1],       m3
+
+    pmovzxbw    m6,         m1
+    punpckhbw   m1,         m0
+    pmovsxbw    m3,         m4
+    punpckhbw   m4,         m4
+    psraw       m4,         8
+
+    paddw       m6,         m3
+    paddw       m1,         m4
+    packuswb    m6,         m1
+    movu        [r0],       m6
+
+    movu        m3,         [r1]                    ; m3 = m_iUpBuff1
+    paddb       m3,         m5
+    paddb       m3,         [pb_2]
+
+    movu        m4,         [r2]                    ; m4 = m_iOffsetEo
+    pshufb      m4,         m3
+    psubb       m3,         m0,         m5
+    movu        [r1],       m3
+
+    pmovzxbw    m5,         m2
+    punpckhbw   m2,         m0
+    pmovsxbw    m3,         m4
+    punpckhbw   m4,         m4
+    psraw       m4,         8
+
+    paddw       m5,         m3
+    paddw       m2,         m4
+    packuswb    m5,         m2
+    movu        [r0 + r3],  m5
+
+    add         r0,         16
+    add         r1,         16
+    dec         r4d
+    jnz         .loop
+    RET
+
+INIT_YMM avx2
+cglobal saoCuOrgE1_2Rows, 3, 5, 7, pRec, m_iUpBuff1, m_iOffsetEo, iStride, iLcuWidth
+    mov             r3d,        r3m
+    mov             r4d,        r4m
+    pxor            m0,         m0                           ; m0 = 0
+    vbroadcasti128  m5,         [pb_128]
+    vbroadcasti128  m6,         [r2]                         ; m6 = m_iOffsetEo
+    shr             r4d,        4
+.loop
+    movu            xm1,        [r0]                         ; m1 = pRec[x]
+    movu            xm2,        [r0 + r3]                    ; m2 = pRec[x + iStride]
+    vinserti128     m1,         m1,       xm2,            1
+    vinserti128     m2,         m2,       [r0 + r3 * 2],  1
+
+    pxor            m3,         m1,       m5
+    pxor            m4,         m2,       m5
+    pcmpgtb         m2,         m3,       m4
+    pcmpgtb         m4,         m3
+    pand            m2,         [pb_1]
+    por             m2,         m4
+
+    movu            xm3,        [r1]                         ; xm3 = m_iUpBuff
+    psubb           m4,         m0,       m2
+    vinserti128     m3,         m3,       xm4,            1
+    paddb           m3,         m2
+    paddb           m3,         [pb_2]
+    pshufb          m2,         m6,       m3
+    vextracti128    [r1],       m4,       1
+
+    pmovzxbw        m4,         xm1
+    vextracti128    xm3,        m1,       1
+    pmovzxbw        m3,         xm3
+    pmovsxbw        m1,         xm2
+    vextracti128    xm2,        m2,       1
+    pmovsxbw        m2,         xm2
+
+    paddw           m4,         m1
+    paddw           m3,         m2
+    packuswb        m4,         m3
+    vpermq          m4,         m4,       11011000b
+    movu            [r0],       xm4
+    vextracti128    [r0 + r3],  m4,       1
+
+    add             r0,         16
+    add             r1,         16
+    dec             r4d
+    jnz             .loop
     RET
 
 ;======================================================================================================================================================
 ; void saoCuOrgE2(pixel * rec, int8_t * bufft, int8_t * buff1, int8_t * offsetEo, int lcuWidth, intptr_t stride)
 ;======================================================================================================================================================
 INIT_XMM sse4
-cglobal saoCuOrgE2, 5, 7, 8, rec, bufft, buff1, offsetEo, lcuWidth
-
-    mov         r6,    16
+cglobal saoCuOrgE2, 5, 6, 8, rec, bufft, buff1, offsetEo, lcuWidth
+    mov         r4d,   r4m
     mov         r5d,   r5m
     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]
     mova        m7,    [pb_128]
-    shr         r4d,   4
-    inc         r1q
-
-    .loop
-         movu        m1,    [r0]                    ; m1 = rec[x]
-         movu        m2,    [r0 + r5 + 1]           ; m2 = rec[x + stride + 1]
-         pxor        m3,    m1,    m7
-         pxor        m4,    m2,    m7
-         pcmpgtb     m2,    m3,    m4
-         pcmpgtb     m4,    m3
-         pand        m2,    [pb_1]
-         por         m2,    m4
-         movu        m3,    [r2]                    ; m3 = buff1
-
-         paddb       m3,    m2
-         paddb       m3,    m6                      ; m3 = edgeType
-
-         movu        m4,    [r3]                    ; m4 = offsetEo
-         pshufb      m4,    m3
-
-         psubb       m3,    m0,    m2
-         movu        [r1],  m3
-
-         pmovzxbw    m2,    m1
-         punpckhbw   m1,    m0
-         pmovsxbw    m3,    m4
-         punpckhbw   m4,    m4
-         psraw       m4,    8
-
-         paddw       m2,    m3
-         paddw       m1,    m4
-         packuswb    m2,    m1
-         movu        [r0],  m2
-
-         add         r0,    r6
-         add         r1,    r6
-         add         r2,    r6
-         dec         r4d
-         jnz         .loop
+    inc         r1
+    movh        m5,    [r0 + r4]
+    movhps      m5,    [r1 + r4]
+
+.loop
+    movu        m1,    [r0]                    ; m1 = rec[x]
+    movu        m2,    [r0 + r5 + 1]           ; m2 = rec[x + stride + 1]
+    pxor        m3,    m1,    m7
+    pxor        m4,    m2,    m7
+    pcmpgtb     m2,    m3,    m4
+    pcmpgtb     m4,    m3
+    pand        m2,    [pb_1]
+    por         m2,    m4
+    movu        m3,    [r2]                    ; m3 = buff1
+
+    paddb       m3,    m2
+    paddb       m3,    m6                      ; m3 = edgeType
+
+    movu        m4,    [r3]                    ; m4 = offsetEo
+    pshufb      m4,    m3
+
+    psubb       m3,    m0,    m2
+    movu        [r1],  m3
+
+    pmovzxbw    m2,    m1
+    punpckhbw   m1,    m0
+    pmovsxbw    m3,    m4
+    punpckhbw   m4,    m4
+    psraw       m4,    8
+
+    paddw       m2,    m3
+    paddw       m1,    m4
+    packuswb    m2,    m1
+    movu        [r0],  m2
+
+    add         r0,    16
+    add         r1,    16
+    add         r2,    16
+    sub         r4,    16
+    jg          .loop
+
+    movh        [r0 + r4], m5
+    movhps      [r1 + r4], m5
+    RET
+
+INIT_YMM avx2
+cglobal saoCuOrgE2, 5, 6, 7, rec, bufft, buff1, offsetEo, lcuWidth
+    mov            r4d,   r4m
+    mov            r5d,   r5m
+    pxor           xm0,   xm0                     ; xm0 = 0
+    mova           xm5,   [pb_128]
+    inc            r1
+    movq           xm6,   [r0 + r4]
+    movhps         xm6,   [r1 + r4]
+
+    movu           xm1,   [r0]                    ; xm1 = rec[x]
+    movu           xm2,   [r0 + r5 + 1]           ; xm2 = rec[x + stride + 1]
+    pxor           xm3,   xm1,   xm5
+    pxor           xm4,   xm2,   xm5
+    pcmpgtb        xm2,   xm3,   xm4
+    pcmpgtb        xm4,   xm3
+    pand           xm2,   [pb_1]
+    por            xm2,   xm4
+    movu           xm3,   [r2]                    ; xm3 = buff1
+
+    paddb          xm3,   xm2
+    paddb          xm3,   [pb_2]                  ; xm3 = edgeType
+
+    movu           xm4,   [r3]                    ; xm4 = offsetEo
+    pshufb         xm4,   xm3
+
+    psubb          xm3,   xm0,   xm2
+    movu           [r1],  xm3
+
+    pmovzxbw       m2,    xm1
+    pmovsxbw       m3,    xm4
+
+    paddw          m2,    m3
+    vextracti128   xm3,   m2,    1
+    packuswb       xm2,   xm3
+    movu           [r0],  xm2
+
+    movq           [r0 + r4], xm6
+    movhps         [r1 + r4], xm6
+    RET
+
+INIT_YMM avx2
+cglobal saoCuOrgE2_32, 5, 6, 8, rec, bufft, buff1, offsetEo, lcuWidth
+    mov             r4d,   r4m
+    mov             r5d,   r5m
+    pxor            m0,    m0                      ; m0 = 0
+    vbroadcasti128  m7,    [pb_128]
+    vbroadcasti128  m5,    [r3]                    ; m5 = offsetEo
+    inc             r1
+    movq            xm6,   [r0 + r4]
+    movhps          xm6,   [r1 + r4]
+
+.loop:
+    movu            m1,    [r0]                    ; m1 = rec[x]
+    movu            m2,    [r0 + r5 + 1]           ; m2 = rec[x + stride + 1]
+    pxor            m3,    m1,    m7
+    pxor            m4,    m2,    m7
+    pcmpgtb         m2,    m3,    m4
+    pcmpgtb         m4,    m3
+    pand            m2,    [pb_1]
+    por             m2,    m4
+    movu            m3,    [r2]                    ; m3 = buff1
+
+    paddb           m3,    m2
+    paddb           m3,    [pb_2]                  ; m3 = edgeType
+
+    pshufb          m4,    m5,    m3
+
+    psubb           m3,    m0,    m2
+    movu            [r1],  m3
+
+    pmovzxbw        m2,    xm1
+    vextracti128    xm1,   m1,    1
+    pmovzxbw        m1,    xm1
+    pmovsxbw        m3,    xm4
+    vextracti128    xm4,   m4,    1
+    pmovsxbw        m4,    xm4
+
+    paddw           m2,    m3
+    paddw           m1,    m4
+    packuswb        m2,    m1
+    vpermq          m2,    m2,    11011000b
+    movu            [r0],  m2
+
+    add             r0,    32
+    add             r1,    32
+    add             r2,    32
+    sub             r4,    32
+    jg              .loop
+
+    movq            [r0 + r4], xm6
+    movhps          [r1 + r4], xm6
     RET
 
 ;=======================================================================================================
 ;void saoCuOrgE3(pixel *rec, int8_t *upBuff1, int8_t *m_offsetEo, intptr_t stride, int startX, int endX)
 ;=======================================================================================================
 INIT_XMM sse4
-cglobal saoCuOrgE3, 3, 7, 8
+cglobal saoCuOrgE3, 3,6,8
     mov             r3d, r3m
     mov             r4d, r4m
     mov             r5d, r5m
 
-    mov             r6d, r5d
-    sub             r6d, r4d
+    ; save latest 2 pixels for case startX=1 or left_endX=15
+    movh            m7, [r0 + r5]
+    movhps          m7, [r1 + r5 - 1]
 
+    ; move to startX+1
     inc             r4d
     add             r0, r4
     add             r1, r4
-    movh            m7, [r0 + r6 - 1]
-    mov             r6, [r1 + r6 - 2]
+    sub             r5d, r4d
     pxor            m0, m0                      ; m0 = 0
     movu            m6, [pb_2]                  ; m6 = [2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2]
 
@@ -244,30 +608,143 @@
     packuswb        m2, m1
     movu            [r0], m2
 
-    sub             r5d, 16
-    jle             .end
+    add             r0, 16
+    add             r1, 16
 
-    lea             r0, [r0 + 16]
-    lea             r1, [r1 + 16]
+    sub             r5, 16
+    jg             .loop
 
-    jnz             .loop
+    ; restore last pixels (up to 2)
+    movh            [r0 + r5], m7
+    movhps          [r1 + r5 - 1], m7
+    RET
 
-.end:
-    js              .skip
-    sub             r0, r4
-    sub             r1, r4
-    movh            [r0 + 16], m7
-    mov             [r1 + 15], r6
-    jmp             .quit
+INIT_YMM avx2
+cglobal saoCuOrgE3, 3, 6, 8
+    mov             r3d,  r3m
+    mov             r4d,  r4m
+    mov             r5d,  r5m
+
+    ; save latest 2 pixels for case startX=1 or left_endX=15
+    movq            xm7,  [r0 + r5]
+    movhps          xm7,  [r1 + r5 - 1]
+
+    ; move to startX+1
+    inc             r4d
+    add             r0,   r4
+    add             r1,   r4
+    sub             r5d,  r4d
+    pxor            xm0,  xm0                     ; xm0 = 0
+    mova            xm6,  [pb_2]                  ; xm6 = [2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2]
+    movu            xm5,  [r2]                    ; xm5 = m_iOffsetEo
+
+.loop:
+    movu            xm1,  [r0]                    ; xm1 = pRec[x]
+    movu            xm2,  [r0 + r3]               ; xm2 = pRec[x + iStride]
+
+    psubusb         xm3,  xm2,  xm1
+    psubusb         xm4,  xm1,  xm2
+    pcmpeqb         xm3,  xm0
+    pcmpeqb         xm4,  xm0
+    pcmpeqb         xm2,  xm1
+
+    pabsb           xm3,  xm3
+    por             xm4,  xm3
+    pandn           xm2,  xm4                     ; xm2 = iSignDown
+
+    movu            xm3,  [r1]                    ; xm3 = m_iUpBuff1
+
+    paddb           xm3,  xm2
+    paddb           xm3,  xm6                     ; xm3 = uiEdgeType
+
+    pshufb          xm4,  xm5,  xm3
+
+    psubb           xm3,  xm0,  xm2
+    movu            [r1 - 1],   xm3
+
+    pmovzxbw        m2,   xm1
+    pmovsxbw        m3,   xm4
+
+    paddw           m2,   m3
+    vextracti128    xm3,  m2,   1
+    packuswb        xm2,  xm3
+    movu            [r0], xm2
+
+    add             r0,   16
+    add             r1,   16
+
+    sub             r5,   16
+    jg             .loop
+
+    ; restore last pixels (up to 2)
+    movq            [r0 + r5],     xm7
+    movhps          [r1 + r5 - 1], xm7
+    RET
+
+INIT_YMM avx2
+cglobal saoCuOrgE3_32, 3, 6, 8
+    mov             r3d,  r3m
+    mov             r4d,  r4m
+    mov             r5d,  r5m
+
+    ; save latest 2 pixels for case startX=1 or left_endX=15
+    movq            xm7,  [r0 + r5]
+    movhps          xm7,  [r1 + r5 - 1]
+
+    ; move to startX+1
+    inc             r4d
+    add             r0,   r4
+    add             r1,   r4
+    sub             r5d,  r4d
+    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]
+    vbroadcasti128  m5,   [r2]                    ; m5 = m_iOffsetEo
+
+.loop:
+    movu            m1,   [r0]                    ; m1 = pRec[x]
+    movu            m2,   [r0 + r3]               ; m2 = pRec[x + iStride]
+
+    psubusb         m3,   m2,   m1
+    psubusb         m4,   m1,   m2
+    pcmpeqb         m3,   m0
+    pcmpeqb         m4,   m0
+    pcmpeqb         m2,   m1
+
+    pabsb           m3,   m3
+    por             m4,   m3
+    pandn           m2,   m4                      ; m2 = iSignDown
+
+    movu            m3,   [r1]                    ; m3 = m_iUpBuff1
+
+    paddb           m3,   m2
+    paddb           m3,   m6                      ; m3 = uiEdgeType
+
+    pshufb          m4,   m5,   m3
+
+    psubb           m3,   m0,   m2
+    movu            [r1 - 1],   m3
+
+    pmovzxbw        m2,   xm1
+    vextracti128    xm1,  m1,   1
+    pmovzxbw        m1,   xm1
+    pmovsxbw        m3,   xm4
+    vextracti128    xm4,  m4,   1
+    pmovsxbw        m4,   xm4
 
-.skip:
-    sub             r0, r4
-    sub             r1, r4
-    movh            [r0 + 15], m7
-    mov             [r1 + 14], r6
+    paddw           m2,   m3
+    paddw           m1,   m4
+    packuswb        m2,   m1
+    vpermq          m2,   m2,   11011000b
+    movu            [r0], m2
 
-.quit:
+    add             r0,   32
+    add             r1,   32
+    sub             r5,   32
+    jg             .loop
 
+    ; restore last pixels (up to 2)
+    movq            [r0 + r5],     xm7
+    movhps          [r1 + r5 - 1], xm7
     RET
 
 ;=====================================================================================
@@ -320,32 +797,181 @@
     jnz         .loopH
     RET
 
+INIT_YMM avx2
+cglobal saoCuOrgB0, 4, 7, 8
+
+    mov             r3d,        r3m
+    mov             r4d,        r4m
+    mova            m7,         [pb_31]
+    vbroadcasti128  m3,         [r1 + 0]            ; offset[0-15]
+    vbroadcasti128  m4,         [r1 + 16]           ; offset[16-31]
+    lea             r6,         [r4 * 2]
+    sub             r6d,        r2d
+    shr             r2d,        4
+    mov             r1d,        r3d
+    shr             r3d,        1
+.loopH
+    mov             r5d,        r2d
+.loopW
+    movu            xm2,        [r0]                ; m2 = [rec]
+    vinserti128     m2,         m2,  [r0 + r4],  1
+    psrlw           m1,         m2,  3
+    pand            m1,         m7                  ; m1 = [index]
+    pcmpgtb         m0,         m1,  [pb_15]        ; m0 = [mask]
+
+    pshufb          m6,         m3,  m1
+    pshufb          m5,         m4,  m1
+
+    pblendvb        m6,         m6,  m5,  m0        ; NOTE: don't use 3 parameters style, x264 macro have some bug!
+
+    pmovzxbw        m1,         xm2                 ; rec
+    vextracti128    xm2,        m2,  1
+    pmovzxbw        m2,         xm2
+    pmovsxbw        m0,         xm6                 ; offset
+    vextracti128    xm6,        m6,  1
+    pmovsxbw        m6,         xm6
+
+    paddw           m1,         m0
+    paddw           m2,         m6
+    packuswb        m1,         m2
+    vpermq          m1,         m1,  11011000b
+
+    movu            [r0],       xm1
+    vextracti128    [r0 + r4],  m1,  1
+    add             r0,         16
+    dec             r5d
+    jnz             .loopW
+
+    add             r0,         r6
+    dec             r3d
+    jnz             .loopH
+    test            r1b,        1
+    jz              .end
+    mov             r5d,        r2d
+.loopW1
+    movu            xm2,        [r0]                ; m2 = [rec]
+    psrlw           xm1,        xm2, 3
+    pand            xm1,        xm7                 ; m1 = [index]
+    pcmpgtb         xm0,        xm1, [pb_15]        ; m0 = [mask]
+
+    pshufb          xm6,        xm3, xm1
+    pshufb          xm5,        xm4, xm1
+
+    pblendvb        xm6,        xm6, xm5, xm0       ; NOTE: don't use 3 parameters style, x264 macro have some bug!
+
+    pmovzxbw        m1,         xm2                 ; rec
+    pmovsxbw        m0,         xm6                 ; offset
+
+    paddw           m1,         m0
+    vextracti128    xm0,        m1,  1
+    packuswb        xm1,        xm0
+
+    movu            [r0],       xm1
+    add             r0,         16
+    dec             r5d
+    jnz             .loopW1
+.end
+    RET
+
 ;============================================================================================================
-; void calSign(int8_t *dst, const Pixel *src1, const Pixel *src2, const int endX)
+; void calSign(int8_t *dst, const Pixel *src1, const Pixel *src2, const int width)
 ;============================================================================================================
 INIT_XMM sse4
-cglobal calSign, 4, 5, 7
+cglobal calSign, 4,5,6
+    mova        m0,     [pb_128]
+    mova        m1,     [pb_1]
 
-    mov         r4,    16
-    mova        m1,    [pb_128]
-    mova        m0,    [pb_1]
-    shr         r3d,   4
-.loop
-    movu        m2,    [r1]        ; m2 = pRec[x]
-    movu        m3,    [r2]        ; m3 = pTmpU[x]
+    sub         r1,     r0
+    sub         r2,     r0
 
-    pxor        m4,    m2,    m1
-    pxor        m5,    m3,    m1
-    pcmpgtb     m6,    m4,    m5
-    pcmpgtb     m5,    m4
-    pand        m6,    m0
-    por         m6,    m5
+    mov         r4d,    r3d
+    shr         r3d,    4
+    jz         .next
+.loop:
+    movu        m2,     [r0 + r1]            ; m2 = pRec[x]
+    movu        m3,     [r0 + r2]            ; m3 = pTmpU[x]
+    pxor        m4,     m2,     m0
+    pxor        m3,     m0
+    pcmpgtb     m5,     m4,     m3
+    pcmpgtb     m3,     m4
+    pand        m5,     m1
+    por         m5,     m3
+    movu        [r0],   m5
+
+    add         r0,     16
+    dec         r3d
+    jnz        .loop
 
-    movu        [r0],  m6
+    ; process partial
+.next:
+    and         r4d, 15
+    jz         .end
+
+    movu        m2,     [r0 + r1]            ; m2 = pRec[x]
+    movu        m3,     [r0 + r2]            ; m3 = pTmpU[x]
+    pxor        m4,     m2,     m0
+    pxor        m3,     m0
+    pcmpgtb     m5,     m4,     m3
+    pcmpgtb     m3,     m4
+    pand        m5,     m1
+    por         m5,     m3
+
+    lea         r3,     [pb_movemask + 16]
+    sub         r3,     r4
+    movu        xmm0,   [r3]
+    movu        m3,     [r0]
+    pblendvb    m5,     m5,     m3,     xmm0
+    movu        [r0],   m5
 
-    add         r0,    r4
-    add         r1,    r4
-    add         r2,    r4
-    dec         r3d
-    jnz         .loop
+.end:
+    RET
+
+INIT_YMM avx2
+cglobal calSign, 4, 5, 6
+    vbroadcasti128  m0,     [pb_128]
+    mova            m1,     [pb_1]
+
+    sub             r1,     r0
+    sub             r2,     r0
+
+    mov             r4d,    r3d
+    shr             r3d,    5
+    jz              .next
+.loop:
+    movu            m2,     [r0 + r1]            ; m2 = pRec[x]
+    movu            m3,     [r0 + r2]            ; m3 = pTmpU[x]
+    pxor            m4,     m2,     m0
+    pxor            m3,     m0
+    pcmpgtb         m5,     m4,     m3
+    pcmpgtb         m3,     m4
+    pand            m5,     m1
+    por             m5,     m3
+    movu            [r0],   m5
+
+    add             r0,     mmsize
+    dec             r3d
+    jnz             .loop
+
+    ; process partial
+.next:
+    and             r4d,    31
+    jz              .end
+
+    movu            m2,     [r0 + r1]            ; m2 = pRec[x]
+    movu            m3,     [r0 + r2]            ; m3 = pTmpU[x]
+    pxor            m4,     m2,     m0
+    pxor            m3,     m0
+    pcmpgtb         m5,     m4,     m3
+    pcmpgtb         m3,     m4
+    pand            m5,     m1
+    por             m5,     m3
+
+    lea             r3,     [pb_movemask_32 + 32]
+    sub             r3,     r4
+    movu            m0,     [r3]
+    movu            m3,     [r0]
+    pblendvb        m5,     m5,     m3,     m0
+    movu            [r0],   m5
+
+.end:
     RET

x265_1.6.tar.gz/source/common/x86/loopfilter.h -> x265_1.7.tar.gz/source/common/x86/loopfilter.h Changed

@@ -25,11 +25,21 @@
 #ifndef X265_LOOPFILTER_H
 #define X265_LOOPFILTER_H
 
-void x265_saoCuOrgE0_sse4(pixel * rec, int8_t * offsetEo, int endX, int8_t signLeft);
+void x265_saoCuOrgE0_sse4(pixel * rec, int8_t * offsetEo, int endX, int8_t* signLeft, intptr_t stride);
+void x265_saoCuOrgE0_avx2(pixel * rec, int8_t * offsetEo, int endX, int8_t* signLeft, intptr_t stride);
 void x265_saoCuOrgE1_sse4(pixel* rec, int8_t* upBuff1, int8_t* offsetEo, intptr_t stride, int width);
+void x265_saoCuOrgE1_avx2(pixel* rec, int8_t* upBuff1, int8_t* offsetEo, intptr_t stride, int width);
+void x265_saoCuOrgE1_2Rows_sse4(pixel* rec, int8_t* upBuff1, int8_t* offsetEo, intptr_t stride, int width);
+void x265_saoCuOrgE1_2Rows_avx2(pixel* rec, int8_t* upBuff1, int8_t* offsetEo, intptr_t stride, int width);
 void x265_saoCuOrgE2_sse4(pixel* rec, int8_t* pBufft, int8_t* pBuff1, int8_t* offsetEo, int lcuWidth, intptr_t stride);
+void x265_saoCuOrgE2_avx2(pixel* rec, int8_t* pBufft, int8_t* pBuff1, int8_t* offsetEo, int lcuWidth, intptr_t stride);
+void x265_saoCuOrgE2_32_avx2(pixel* rec, int8_t* pBufft, int8_t* pBuff1, int8_t* offsetEo, int lcuWidth, intptr_t stride);
 void x265_saoCuOrgE3_sse4(pixel *rec, int8_t *upBuff1, int8_t *m_offsetEo, intptr_t stride, int startX, int endX);
+void x265_saoCuOrgE3_avx2(pixel *rec, int8_t *upBuff1, int8_t *m_offsetEo, intptr_t stride, int startX, int endX);
+void x265_saoCuOrgE3_32_avx2(pixel *rec, int8_t *upBuff1, int8_t *m_offsetEo, intptr_t stride, int startX, int endX);
 void x265_saoCuOrgB0_sse4(pixel* rec, const int8_t* offsetBo, int ctuWidth, int ctuHeight, intptr_t stride);
+void x265_saoCuOrgB0_avx2(pixel* rec, const int8_t* offsetBo, int ctuWidth, int ctuHeight, intptr_t stride);
 void x265_calSign_sse4(int8_t *dst, const pixel *src1, const pixel *src2, const int endX);
+void x265_calSign_avx2(int8_t *dst, const pixel *src1, const pixel *src2, const int endX);
 
 #endif // ifndef X265_LOOPFILTER_H

x265_1.6.tar.gz/source/common/x86/mc-a.asm -> x265_1.7.tar.gz/source/common/x86/mc-a.asm Changed

x265_1.6.tar.gz/source/common/x86/pixel-a.asm -> x265_1.7.tar.gz/source/common/x86/pixel-a.asm Changed

@@ -7078,6 +7078,117 @@
 .end:
     RET
 
+; Input 16bpp, Output 8bpp
+;-------------------------------------------------------------------------------------------------------------------------------------
+;void planecopy_sp(uint16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int width, int height, int shift, uint16_t mask)
+;-------------------------------------------------------------------------------------------------------------------------------------
+INIT_YMM avx2
+cglobal downShift_16, 6,7,3
+    movd        xm0, r6m        ; m0 = shift
+    add         r1d, r1d
+    dec         r5d
+.loopH:
+    xor         r6, r6
+.loopW:
+    movu        m1, [r0 + r6 * 2 +  0]
+    movu        m2, [r0 + r6 * 2 + 32]
+    vpsrlw      m1, xm0
+    vpsrlw      m2, xm0
+    packuswb    m1, m2
+    vpermq      m1, m1, 11011000b
+    movu        [r2 + r6], m1
+
+    add         r6d, mmsize
+    cmp         r6d, r4d
+    jl          .loopW
+
+    ; move to next row
+    add         r0, r1
+    add         r2, r3
+    dec         r5d
+    jnz         .loopH
+
+; processing last row of every frame [To handle width which not a multiple of 32]
+    mov         r6d, r4d
+    and         r4d, 31
+    shr         r6d, 5
+
+.loop32:
+    movu        m1, [r0]
+    movu        m2, [r0 + 32]
+    psrlw       m1, xm0
+    psrlw       m2, xm0
+    packuswb    m1, m2
+    vpermq      m1, m1, 11011000b
+    movu        [r2], m1
+
+    add         r0, 2*mmsize
+    add         r2, mmsize
+    dec         r6d
+    jnz         .loop32
+
+    cmp         r4d, 16
+    jl          .process8
+    movu        m1, [r0]
+    psrlw       m1, xm0
+    packuswb    m1, m1
+    vpermq      m1, m1, 10001000b
+    movu        [r2], xm1
+
+    add         r0, mmsize
+    add         r2, 16
+    sub         r4d, 16
+    jz          .end
+
+.process8:
+    cmp         r4d, 8
+    jl          .process4
+    movu        m1, [r0]
+    psrlw       m1, xm0
+    packuswb    m1, m1
+    movq        [r2], xm1
+
+    add         r0, 16
+    add         r2, 8
+    sub         r4d, 8
+    jz          .end
+
+.process4:
+    cmp         r4d, 4
+    jl          .process2
+    movq        xm1,[r0]
+    psrlw       m1, xm0
+    packuswb    m1, m1
+    movd        [r2], xm1
+
+    add         r0, 8
+    add         r2, 4
+    sub         r4d, 4
+    jz          .end
+
+.process2:
+    cmp         r4d, 2
+    jl          .process1
+    movd        xm1, [r0]
+    psrlw       m1, xm0
+    packuswb    m1, m1
+    movd        r6d, xm1
+    mov         [r2], r6w
+
+    add         r0, 4
+    add         r2, 2
+    sub         r4d, 2
+    jz          .end
+
+.process1:
+    movd        xm1, [r0]
+    psrlw       m1, xm0
+    packuswb    m1, m1
+    movd        r3d, xm1
+    mov         [r2], r3b
+.end:
+    RET
+
 ; Input 8bpp, Output 16bpp
 ;---------------------------------------------------------------------------------------------------------------------
 ;void planecopy_cp(uint8_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int width, int height, int shift)
@@ -10395,3 +10506,1372 @@
     mov             rsp, r5
     RET
 %endif
+
+;;---------------------------------------------------------------
+;; SATD AVX2
+;; int pixel_satd(const pixel*, intptr_t, const pixel*, intptr_t)
+;;---------------------------------------------------------------
+;; r0   - pix0
+;; r1   - pix0Stride
+;; r2   - pix1
+;; r3   - pix1Stride
+
+%if ARCH_X86_64 == 1 && HIGH_BIT_DEPTH == 0
+INIT_YMM avx2
+cglobal calc_satd_16x8    ; function to compute satd cost for 16 columns, 8 rows
+    pxor                m6, m6
+    vbroadcasti128      m0, [r0]
+    vbroadcasti128      m4, [r2]
+    vbroadcasti128      m1, [r0 + r1]
+    vbroadcasti128      m5, [r2 + r3]
+    pmaddubsw           m4, m7
+    pmaddubsw           m0, m7
+    pmaddubsw           m5, m7
+    pmaddubsw           m1, m7
+    psubw               m0, m4
+    psubw               m1, m5
+    vbroadcasti128      m2, [r0 + r1 * 2]
+    vbroadcasti128      m4, [r2 + r3 * 2]
+    vbroadcasti128      m3, [r0 + r4]
+    vbroadcasti128      m5, [r2 + r5]
+    pmaddubsw           m4, m7
+    pmaddubsw           m2, m7
+    pmaddubsw           m5, m7
+    pmaddubsw           m3, m7
+    psubw               m2, m4
+    psubw               m3, m5
+    lea                 r0, [r0 + r1 * 4]
+    lea                 r2, [r2 + r3 * 4]
+    paddw               m4, m0, m1
+    psubw               m1, m1, m0
+    paddw               m0, m2, m3
+    psubw               m3, m2
+    paddw               m2, m4, m0
+    psubw               m0, m4
+    paddw               m4, m1, m3
+    psubw               m3, m1
+    pabsw               m2, m2
+    pabsw               m0, m0
+    pabsw               m4, m4
+    pabsw               m3, m3
+    pblendw             m1, m2, m0, 10101010b
+    pslld               m0, 16
+    psrld               m2, 16
+    por                 m0, m2
+    pmaxsw              m1, m0
+    paddw               m6, m1
+    pblendw             m2, m4, m3, 10101010b
+    pslld               m3, 16
+    psrld               m4, 16
+    por                 m3, m4
+    pmaxsw              m2, m3
+    paddw               m6, m2
+    vbroadcasti128      m1, [r0]
+    vbroadcasti128      m4, [r2]
+    vbroadcasti128      m2, [r0 + r1]
+    vbroadcasti128      m5, [r2 + r3]
+    pmaddubsw           m4, m7
+    pmaddubsw           m1, m7
+    pmaddubsw           m5, m7
+    pmaddubsw           m2, m7
+    psubw               m1, m4
+    psubw               m2, m5
+    vbroadcasti128      m0, [r0 + r1 * 2]
+    vbroadcasti128      m4, [r2 + r3 * 2]
+    vbroadcasti128      m3, [r0 + r4]
+    vbroadcasti128      m5, [r2 + r5]
+    lea                 r0, [r0 + r1 * 4]
+    lea                 r2, [r2 + r3 * 4]
+    pmaddubsw           m4, m7
+    pmaddubsw           m0, m7
+    pmaddubsw           m5, m7
+    pmaddubsw           m3, m7
+    psubw               m0, m4
+    psubw               m3, m5
+    paddw               m4, m1, m2
+    psubw               m2, m1
+    paddw               m1, m0, m3
+    psubw               m3, m0
+    paddw               m0, m4, m1
+    psubw               m1, m4
+    paddw               m4, m2, m3
+    psubw               m3, m2
+    pabsw               m0, m0
+    pabsw               m1, m1
+    pabsw               m4, m4
+    pabsw               m3, m3
+    pblendw             m2, m0, m1, 10101010b
+    pslld               m1, 16
+    psrld               m0, 16
+    por                 m1, m0
+    pmaxsw              m2, m1
+    paddw               m6, m2
+    pblendw             m0, m4, m3, 10101010b
+    pslld               m3, 16
+    psrld               m4, 16
+    por                 m3, m4
+    pmaxsw              m0, m3
+    paddw               m6, m0
+    vextracti128        xm0, m6, 1
+    pmovzxwd            m6, xm6
+    pmovzxwd            m0, xm0
+    paddd               m8, m6
+    paddd               m9, m0
+    ret
+
+cglobal calc_satd_16x4    ; function to compute satd cost for 16 columns, 4 rows
+    pxor                m6, m6
+    vbroadcasti128      m0, [r0]
+    vbroadcasti128      m4, [r2]
+    vbroadcasti128      m1, [r0 + r1]
+    vbroadcasti128      m5, [r2 + r3]
+    pmaddubsw           m4, m7
+    pmaddubsw           m0, m7
+    pmaddubsw           m5, m7
+    pmaddubsw           m1, m7
+    psubw               m0, m4
+    psubw               m1, m5
+    vbroadcasti128      m2, [r0 + r1 * 2]
+    vbroadcasti128      m4, [r2 + r3 * 2]
+    vbroadcasti128      m3, [r0 + r4]
+    vbroadcasti128      m5, [r2 + r5]
+    pmaddubsw           m4, m7
+    pmaddubsw           m2, m7
+    pmaddubsw           m5, m7
+    pmaddubsw           m3, m7
+    psubw               m2, m4
+    psubw               m3, m5
+    paddw               m4, m0, m1
+    psubw               m1, m1, m0
+    paddw               m0, m2, m3
+    psubw               m3, m2
+    paddw               m2, m4, m0
+    psubw               m0, m4
+    paddw               m4, m1, m3
+    psubw               m3, m1
+    pabsw               m2, m2
+    pabsw               m0, m0
+    pabsw               m4, m4
+    pabsw               m3, m3
+    pblendw             m1, m2, m0, 10101010b
+    pslld               m0, 16
+    psrld               m2, 16
+    por                 m0, m2
+    pmaxsw              m1, m0
+    paddw               m6, m1
+    pblendw             m2, m4, m3, 10101010b
+    pslld               m3, 16
+    psrld               m4, 16
+    por                 m3, m4
+    pmaxsw              m2, m3
+    paddw               m6, m2
+    vextracti128        xm0, m6, 1
+    pmovzxwd            m6, xm6
+    pmovzxwd            m0, xm0
+    paddd               m8, m6
+    paddd               m9, m0
+    ret
+
+cglobal pixel_satd_16x4, 4,6,10         ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+
+    call            calc_satd_16x4
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_16x12, 4,6,10        ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+
+    call            calc_satd_16x8
+    call            calc_satd_16x4
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_16x32, 4,6,10        ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_16x64, 4,6,10        ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_32x8, 4,8,10          ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_32x16, 4,8,10         ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_32x24, 4,8,10         ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_32x32, 4,8,10         ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_32x64, 4,8,10         ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_48x64, 4,8,10        ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_64x16, 4,8,10         ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 48]
+    lea             r2, [r7 + 48]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_64x32, 4,8,10         ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 48]
+    lea             r2, [r7 + 48]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_64x48, 4,8,10        ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 48]
+    lea             r2, [r7 + 48]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+
+cglobal pixel_satd_64x64, 4,8,10        ; if WIN64 && cpuflag(avx2)
+    mova            m7, [hmul_16p]
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m8, m8
+    pxor            m9, m9
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 16]
+    lea             r2, [r7 + 16]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    lea             r0, [r6 + 48]
+    lea             r2, [r7 + 48]
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    paddd           m8, m9
+    vextracti128    xm0, m8, 1
+    paddd           xm0, xm8
+    movhlps         xm1, xm0
+    paddd           xm0, xm1
+    pshuflw         xm1, xm0, q0032
+    paddd           xm0, xm1
+    movd            eax, xm0
+    RET
+%endif ; ARCH_X86_64 == 1 && HIGH_BIT_DEPTH == 0
+
+%if ARCH_X86_64 == 1 && HIGH_BIT_DEPTH == 1
+INIT_YMM avx2
+cglobal calc_satd_16x8    ; function to compute satd cost for 16 columns, 8 rows
+    ; rows 0-3
+    movu            m0, [r0]
+    movu            m4, [r2]
+    psubw           m0, m4
+    movu            m1, [r0 + r1]
+    movu            m5, [r2 + r3]
+    psubw           m1, m5
+    movu            m2, [r0 + r1 * 2]
+    movu            m4, [r2 + r3 * 2]
+    psubw           m2, m4
+    movu            m3, [r0 + r4]
+    movu            m5, [r2 + r5]
+    psubw           m3, m5
+    lea             r0, [r0 + r1 * 4]
+    lea             r2, [r2 + r3 * 4]
+    paddw           m4, m0, m1
+    psubw           m1, m0
+    paddw           m0, m2, m3
+    psubw           m3, m2
+    punpckhwd       m2, m4, m1
+    punpcklwd       m4, m1
+    punpckhwd       m1, m0, m3
+    punpcklwd       m0, m3
+    paddw           m3, m4, m0
+    psubw           m0, m4
+    paddw           m4, m2, m1
+    psubw           m1, m2
+    punpckhdq       m2, m3, m0
+    punpckldq       m3, m0
+    paddw           m0, m3, m2
+    psubw           m2, m3
+    punpckhdq       m3, m4, m1
+    punpckldq       m4, m1
+    paddw           m1, m4, m3
+    psubw           m3, m4
+    punpckhqdq      m4, m0, m1
+    punpcklqdq      m0, m1
+    pabsw           m0, m0
+    pabsw           m4, m4
+    pmaxsw          m0, m0, m4
+    punpckhqdq      m1, m2, m3
+    punpcklqdq      m2, m3
+    pabsw           m2, m2
+    pabsw           m1, m1
+    pmaxsw          m2, m1
+    pxor            m7, m7
+    mova            m1, m0
+    punpcklwd       m1, m7
+    paddd           m6, m1
+    mova            m1, m0
+    punpckhwd       m1, m7
+    paddd           m6, m1
+    pxor            m7, m7
+    mova            m1, m2
+    punpcklwd       m1, m7
+    paddd           m6, m1
+    mova            m1, m2
+    punpckhwd       m1, m7
+    paddd           m6, m1
+    ; rows 4-7
+    movu            m0, [r0]
+    movu            m4, [r2]
+    psubw           m0, m4
+    movu            m1, [r0 + r1]
+    movu            m5, [r2 + r3]
+    psubw           m1, m5
+    movu            m2, [r0 + r1 * 2]
+    movu            m4, [r2 + r3 * 2]
+    psubw           m2, m4
+    movu            m3, [r0 + r4]
+    movu            m5, [r2 + r5]
+    psubw           m3, m5
+    lea             r0, [r0 + r1 * 4]
+    lea             r2, [r2 + r3 * 4]
+    paddw           m4, m0, m1
+    psubw           m1, m0
+    paddw           m0, m2, m3
+    psubw           m3, m2
+    punpckhwd       m2, m4, m1
+    punpcklwd       m4, m1
+    punpckhwd       m1, m0, m3
+    punpcklwd       m0, m3
+    paddw           m3, m4, m0
+    psubw           m0, m4
+    paddw           m4, m2, m1
+    psubw           m1, m2
+    punpckhdq       m2, m3, m0
+    punpckldq       m3, m0
+    paddw           m0, m3, m2
+    psubw           m2, m3
+    punpckhdq       m3, m4, m1
+    punpckldq       m4, m1
+    paddw           m1, m4, m3
+    psubw           m3, m4
+    punpckhqdq      m4, m0, m1
+    punpcklqdq      m0, m1
+    pabsw           m0, m0
+    pabsw           m4, m4
+    pmaxsw          m0, m0, m4
+    punpckhqdq      m1, m2, m3
+    punpcklqdq      m2, m3
+    pabsw           m2, m2
+    pabsw           m1, m1
+    pmaxsw          m2, m1
+    pxor            m7, m7
+    mova            m1, m0
+    punpcklwd       m1, m7
+    paddd           m6, m1
+    mova            m1, m0
+    punpckhwd       m1, m7
+    paddd           m6, m1
+    pxor            m7, m7
+    mova            m1, m2
+    punpcklwd       m1, m7
+    paddd           m6, m1
+    mova            m1, m2
+    punpckhwd       m1, m7
+    paddd           m6, m1
+    ret
+
+cglobal calc_satd_16x4    ; function to compute satd cost for 16 columns, 4 rows
+    ; rows 0-3
+    movu            m0, [r0]
+    movu            m4, [r2]
+    psubw           m0, m4
+    movu            m1, [r0 + r1]
+    movu            m5, [r2 + r3]
+    psubw           m1, m5
+    movu            m2, [r0 + r1 * 2]
+    movu            m4, [r2 + r3 * 2]
+    psubw           m2, m4
+    movu            m3, [r0 + r4]
+    movu            m5, [r2 + r5]
+    psubw           m3, m5
+    lea             r0, [r0 + r1 * 4]
+    lea             r2, [r2 + r3 * 4]
+    paddw           m4, m0, m1
+    psubw           m1, m0
+    paddw           m0, m2, m3
+    psubw           m3, m2
+    punpckhwd       m2, m4, m1
+    punpcklwd       m4, m1
+    punpckhwd       m1, m0, m3
+    punpcklwd       m0, m3
+    paddw           m3, m4, m0
+    psubw           m0, m4
+    paddw           m4, m2, m1
+    psubw           m1, m2
+    punpckhdq       m2, m3, m0
+    punpckldq       m3, m0
+    paddw           m0, m3, m2
+    psubw           m2, m3
+    punpckhdq       m3, m4, m1
+    punpckldq       m4, m1
+    paddw           m1, m4, m3
+    psubw           m3, m4
+    punpckhqdq      m4, m0, m1
+    punpcklqdq      m0, m1
+    pabsw           m0, m0
+    pabsw           m4, m4
+    pmaxsw          m0, m0, m4
+    punpckhqdq      m1, m2, m3
+    punpcklqdq      m2, m3
+    pabsw           m2, m2
+    pabsw           m1, m1
+    pmaxsw          m2, m1
+    pxor            m7, m7
+    mova            m1, m0
+    punpcklwd       m1, m7
+    paddd           m6, m1
+    mova            m1, m0
+    punpckhwd       m1, m7
+    paddd           m6, m1
+    pxor            m7, m7
+    mova            m1, m2
+    punpcklwd       m1, m7
+    paddd           m6, m1
+    mova            m1, m2
+    punpckhwd       m1, m7
+    paddd           m6, m1
+    ret
+
+cglobal pixel_satd_16x4, 4,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+
+    call            calc_satd_16x4
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_16x8, 4,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_16x12, 4,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+
+    call            calc_satd_16x8
+    call            calc_satd_16x4
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_16x16, 4,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_16x32, 4,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_16x64, 4,6,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_32x8, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_32x16, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_32x24, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_32x32, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_32x64, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_48x64, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 64]
+    lea             r2, [r7 + 64]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_64x16, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 64]
+    lea             r2, [r7 + 64]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 96]
+    lea             r2, [r7 + 96]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_64x32, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 64]
+    lea             r2, [r7 + 64]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 96]
+    lea             r2, [r7 + 96]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_64x48, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 64]
+    lea             r2, [r7 + 64]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 96]
+    lea             r2, [r7 + 96]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
+cglobal pixel_satd_64x64, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 64]
+    lea             r2, [r7 + 64]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 96]
+    lea             r2, [r7 + 96]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+%endif ; ARCH_X86_64 == 1 && HIGH_BIT_DEPTH == 1

x265_1.6.tar.gz/source/common/x86/pixel-util.h -> x265_1.7.tar.gz/source/common/x86/pixel-util.h Changed

@@ -73,15 +73,18 @@
 float x265_pixel_ssim_end4_sse2(int sum0[5][4], int sum1[5][4], int width);
 float x265_pixel_ssim_end4_avx(int sum0[5][4], int sum1[5][4], int width);
 
-void x265_scale1D_128to64_ssse3(pixel*, const pixel*, intptr_t);
-void x265_scale1D_128to64_avx2(pixel*, const pixel*, intptr_t);
+void x265_scale1D_128to64_ssse3(pixel*, const pixel*);
+void x265_scale1D_128to64_avx2(pixel*, const pixel*);
 void x265_scale2D_64to32_ssse3(pixel*, const pixel*, intptr_t);
+void x265_scale2D_64to32_avx2(pixel*, const pixel*, intptr_t);
 
-int x265_findPosLast_x64(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig);
+int x265_scanPosLast_x64(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig, const uint16_t* scanCG4x4, const int trSize);
+int x265_scanPosLast_avx2_bmi2(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig, const uint16_t* scanCG4x4, const int trSize);
+uint32_t x265_findPosFirstLast_ssse3(const int16_t *dstCoeff, const intptr_t trSize, const uint16_t scanTbl[16]);
 
 #define SETUP_CHROMA_PIXELSUB_PS_FUNC(W, H, cpu) \
     void x265_pixel_sub_ps_ ## W ## x ## H ## cpu(int16_t*  dest, intptr_t destride, const pixel* src0, const pixel* src1, intptr_t srcstride0, intptr_t srcstride1); \
-    void x265_pixel_add_ps_ ## W ## x ## H ## cpu(pixel* dest, intptr_t destride, const pixel* src0, const int16_t*  scr1, intptr_t srcStride0, intptr_t srcStride1);
+    void x265_pixel_add_ps_ ## W ## x ## H ## cpu(pixel* dest, intptr_t destride, const pixel* src0, const int16_t*  src1, intptr_t srcStride0, intptr_t srcStride1);
 
 #define CHROMA_420_PIXELSUB_DEF(cpu) \
     SETUP_CHROMA_PIXELSUB_PS_FUNC(4, 4, cpu); \
@@ -97,7 +100,7 @@
 
 #define SETUP_LUMA_PIXELSUB_PS_FUNC(W, H, cpu) \
     void x265_pixel_sub_ps_ ## W ## x ## H ## cpu(int16_t*  dest, intptr_t destride, const pixel* src0, const pixel* src1, intptr_t srcstride0, intptr_t srcstride1); \
-    void x265_pixel_add_ps_ ## W ## x ## H ## cpu(pixel* dest, intptr_t destride, const pixel* src0, const int16_t*  scr1, intptr_t srcStride0, intptr_t srcStride1);
+    void x265_pixel_add_ps_ ## W ## x ## H ## cpu(pixel* dest, intptr_t destride, const pixel* src0, const int16_t*  src1, intptr_t srcStride0, intptr_t srcStride1);
 
 #define LUMA_PIXELSUB_DEF(cpu) \
     SETUP_LUMA_PIXELSUB_PS_FUNC(8,   8, cpu); \

x265_1.6.tar.gz/source/common/x86/pixel-util8.asm -> x265_1.7.tar.gz/source/common/x86/pixel-util8.asm Changed

@@ -40,16 +40,17 @@
 ssim_c1:   times 4 dd 416          ; .01*.01*255*255*64
 ssim_c2:   times 4 dd 235963       ; .03*.03*255*255*64*63
 %endif
-mask_ff:   times 16 db 0xff
-           times 16 db 0
-deinterleave_shuf: db 0, 2, 4, 6, 8, 10, 12, 14, 1, 3, 5, 7, 9, 11, 13, 15
-deinterleave_word_shuf: db 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15
-hmul_16p:  times 16 db 1
-           times 8 db 1, -1
-hmulw_16p:  times 8 dw 1
-            times 4 dw 1, -1
 
-trans8_shuf: dd 0, 4, 1, 5, 2, 6, 3, 7
+mask_ff:                times 16 db 0xff
+                        times 16 db 0
+deinterleave_shuf:      times  2 db 0, 2, 4, 6, 8, 10, 12, 14, 1, 3, 5, 7, 9, 11, 13, 15
+deinterleave_word_shuf: times  2 db 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15
+hmul_16p:               times 16 db 1
+                        times  8 db 1, -1
+hmulw_16p:              times  8 dw 1
+                        times  4 dw 1, -1
+
+trans8_shuf:            dd 0, 4, 1, 5, 2, 6, 3, 7
 
 SECTION .text
 
@@ -67,6 +68,7 @@
 cextern pb_2
 cextern pb_4
 cextern pb_8
+cextern pb_15
 cextern pb_16
 cextern pb_32
 cextern pb_64
@@ -616,7 +618,7 @@
 
 %if ARCH_X86_64 == 1
 INIT_YMM avx2
-cglobal quant, 5,5,10
+cglobal quant, 5,6,9
     ; fill qbits
     movd            xm4, r4d            ; m4 = qbits
 
@@ -627,7 +629,7 @@
     ; fill offset
     vpbroadcastd    m5, r5m             ; m5 = add
 
-    vpbroadcastw    m9, [pw_1]          ; m9 = word [1]
+    lea             r5, [pw_1]
 
     mov             r4d, r6m
     shr             r4d, 4
@@ -665,7 +667,7 @@
 
     ; count non-zero coeff
     ; TODO: popcnt is faster, but some CPU can't support
-    pminuw          m2, m9
+    pminuw          m2, [r5]
     paddw           m7, m2
 
     add             r0, mmsize
@@ -1285,9 +1287,8 @@
     mov          r6d, r6m
     shl          r6d, 16
     or           r6d, r5d          ; assuming both (w0<<6) and round are using maximum of 16 bits each.
-    movd         xm0, r6d
-    pshufd       xm0, xm0, 0       ; m0 = [w0<<6, round]
-    vinserti128  m0, m0, xm0, 1    ; document says (pshufd + vinserti128) can be replaced with vpbroadcastd m0, xm0, but having build problem, need to investigate
+
+    vpbroadcastd m0, r6d
 
     movd         xm1, r7m
     vpbroadcastd m2, r8m
@@ -1492,6 +1493,84 @@
     dec         r5d
     jnz         .loopH
     RET
+
+%if ARCH_X86_64
+INIT_YMM avx2
+cglobal weight_sp, 6, 9, 7
+    mov             r7d, r7m
+    shl             r7d, 16
+    or              r7d, r6m
+    vpbroadcastd    m0, r7d            ; m0 = times 8 dw w0, round
+    movd            xm1, r8m            ; m1 = [shift]
+    vpbroadcastd    m2, r9m            ; m2 = times 16 dw offset
+    vpbroadcastw    m3, [pw_1]
+    vpbroadcastw    m4, [pw_2000]
+
+    add             r2d, r2d            ; 2 * srcstride
+
+    mov             r7, r0
+    mov             r8, r1
+.loopH:
+    mov             r6d, r4d            ; width
+
+    ; save old src and dst
+    mov             r0, r7              ; src
+    mov             r1, r8              ; dst
+.loopW:
+    movu            m5, [r0]
+    paddw           m5, m4
+
+    punpcklwd       m6,m5, m3
+    pmaddwd         m6, m0
+    psrad           m6, xm1
+    paddd           m6, m2
+
+    punpckhwd       m5, m3
+    pmaddwd         m5, m0
+    psrad           m5, xm1
+    paddd           m5, m2
+
+    packssdw        m6, m5
+    packuswb        m6, m6
+    vpermq          m6, m6, 10001000b
+
+    sub             r6d, 16
+    jl              .width8
+    movu            [r1], xm6
+    je              .nextH
+    add             r0, 32
+    add             r1, 16
+    jmp             .loopW
+
+.width8:
+    add             r6d, 16
+    cmp             r6d, 8
+    jl              .width4
+    movq            [r1], xm6
+    je              .nextH
+    psrldq          m6, 8
+    sub             r6d, 8
+    add             r1, 8
+
+.width4:
+    cmp             r6d, 4
+    jl              .width2
+    movd            [r1], xm6
+    je              .nextH
+    add             r1, 4
+    pshufd          m6, m6, 1
+
+.width2:
+    pextrw          [r1], xm6, 0
+
+.nextH:
+    lea             r7, [r7 + r2]
+    lea             r8, [r8 + r3]
+
+    dec             r5d
+    jnz             .loopH
+    RET
+%endif
 %endif  ; end of (HIGH_BIT_DEPTH == 0)
     
 
@@ -3944,6 +4023,150 @@
     RET
 %endif
 
+;-----------------------------------------------------------------
+; void scale2D_64to32(pixel *dst, pixel *src, intptr_t stride)
+;-----------------------------------------------------------------
+%if HIGH_BIT_DEPTH
+INIT_YMM avx2
+cglobal scale2D_64to32, 3, 4, 5, dest, src, stride
+    mov         r3d,     32
+    add         r2d,     r2d
+    mova        m4,      [pw_2000]
+
+.loop:
+    movu        m0,      [r1]
+    movu        m1,      [r1 + 1 * mmsize]
+    movu        m2,      [r1 + r2]
+    movu        m3,      [r1 + r2 + 1 * mmsize]
+
+    paddw       m0,      m2
+    paddw       m1,      m3
+    phaddw      m0,      m1
+
+    pmulhrsw    m0,      m4
+    vpermq      m0,      m0, q3120
+    movu        [r0],    m0
+
+    movu        m0,      [r1 + 2 * mmsize]
+    movu        m1,      [r1 + 3 * mmsize]
+    movu        m2,      [r1 + r2 + 2 * mmsize]
+    movu        m3,      [r1 + r2 + 3 * mmsize]
+
+    paddw       m0,      m2
+    paddw       m1,      m3
+    phaddw      m0,      m1
+
+    pmulhrsw    m0,      m4
+    vpermq      m0,      m0, q3120
+    movu        [r0 + mmsize], m0
+
+    add         r0,      64
+    lea         r1,      [r1 + 2 * r2]
+    dec         r3d
+    jnz         .loop
+    RET
+%else
+
+INIT_YMM avx2
+cglobal scale2D_64to32, 3, 5, 8, dest, src, stride
+    mov         r3d,     16
+    mova        m7,      [deinterleave_shuf]
+.loop:
+    movu        m0,      [r1]                  ; i
+    lea         r4,      [r1 + r2 * 2]
+    psrlw       m1,      m0, 8                 ; j
+    movu        m2,      [r1 + r2]             ; k
+    psrlw       m3,      m2, 8                 ; l
+
+    pxor        m4,      m0, m1                ; i^j
+    pxor        m5,      m2, m3                ; k^l
+    por         m4,      m5                    ; ij|kl
+
+    pavgb       m0,      m1                    ; s
+    pavgb       m2,      m3                    ; t
+    mova        m5,      m0
+    pavgb       m0,      m2                    ; (s+t+1)/2
+    pxor        m5,      m2                    ; s^t
+    pand        m4,      m5                    ; (ij|kl)&st
+    pand        m4,      [pb_1]
+    psubb       m0,      m4                    ; Result
+
+    movu        m1,      [r1 + 32]             ; i
+    psrlw       m2,      m1, 8                 ; j
+    movu        m3,      [r1 + r2 + 32]        ; k
+    psrlw       m4,      m3, 8                 ; l
+
+    pxor        m5,      m1, m2                ; i^j
+    pxor        m6,      m3, m4                ; k^l
+    por         m5,      m6                    ; ij|kl
+
+    pavgb       m1,      m2                    ; s
+    pavgb       m3,      m4                    ; t
+    mova        m6,      m1
+    pavgb       m1,      m3                    ; (s+t+1)/2
+    pxor        m6,      m3                    ; s^t
+    pand        m5,      m6                    ; (ij|kl)&st
+    pand        m5,      [pb_1]
+    psubb       m1,      m5                    ; Result
+
+    pshufb      m0,      m0, m7
+    pshufb      m1,      m1, m7
+
+    punpcklqdq  m0,      m1
+    vpermq      m0,      m0, 11011000b
+    movu        [r0],    m0
+
+    add         r0,      32
+
+    movu        m0,      [r4]                  ; i
+    psrlw       m1,      m0, 8                 ; j
+    movu        m2,      [r4 + r2]             ; k
+    psrlw       m3,      m2, 8                 ; l
+
+    pxor        m4,      m0, m1                ; i^j
+    pxor        m5,      m2, m3                ; k^l
+    por         m4,      m5                    ; ij|kl
+
+    pavgb       m0,      m1                    ; s
+    pavgb       m2,      m3                    ; t
+    mova        m5,      m0
+    pavgb       m0,      m2                    ; (s+t+1)/2
+    pxor        m5,      m2                    ; s^t
+    pand        m4,      m5                    ; (ij|kl)&st
+    pand        m4,      [pb_1]
+    psubb       m0,      m4                    ; Result
+
+    movu        m1,      [r4 + 32]             ; i
+    psrlw       m2,      m1, 8                 ; j
+    movu        m3,      [r4 + r2 + 32]        ; k
+    psrlw       m4,      m3, 8                 ; l
+
+    pxor        m5,      m1, m2                ; i^j
+    pxor        m6,      m3, m4                ; k^l
+    por         m5,      m6                    ; ij|kl
+
+    pavgb       m1,      m2                    ; s
+    pavgb       m3,      m4                    ; t
+    mova        m6,      m1
+    pavgb       m1,      m3                    ; (s+t+1)/2
+    pxor        m6,      m3                    ; s^t
+    pand        m5,      m6                    ; (ij|kl)&st
+    pand        m5,      [pb_1]
+    psubb       m1,      m5                    ; Result
+
+    pshufb      m0,      m0, m7
+    pshufb      m1,      m1, m7
+
+    punpcklqdq  m0,      m1
+    vpermq      m0,      m0, 11011000b
+    movu        [r0],    m0
+
+    lea         r1,      [r1 + 4 * r2]
+    add         r0,      32
+    dec         r3d
+    jnz         .loop
+    RET
+%endif
 
 ;-----------------------------------------------------------------------------
 ; void pixel_sub_ps_4x4(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
@@ -4337,18 +4560,70 @@
 ;-----------------------------------------------------------------------------
 ; void pixel_sub_ps_16x16(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
 ;-----------------------------------------------------------------------------
+%if HIGH_BIT_DEPTH
+%macro PIXELSUB_PS_W16_H4_avx2 1
+%if ARCH_X86_64
+INIT_YMM avx2
+cglobal pixel_sub_ps_16x%1, 6, 9, 4, dest, deststride, src0, src1, srcstride0, srcstride1
+    add         r1d,    r1d
+    add         r4d,    r4d
+    add         r5d,    r5d
+    lea         r6,     [r1 * 3]
+    lea         r7,     [r4 * 3]
+    lea         r8,     [r5 * 3]
+
+%rep %1/4
+    movu        m0,     [r2]
+    movu        m1,     [r3]
+    movu        m2,     [r2 + r4]
+    movu        m3,     [r3 + r5]
+
+    psubw       m0,     m1
+    psubw       m2,     m3
+
+    movu        [r0],            m0
+    movu        [r0 + r1],       m2
+
+    movu        m0,     [r2 + r4 * 2]
+    movu        m1,     [r3 + r5 * 2]
+    movu        m2,     [r2 + r7]
+    movu        m3,     [r3 + r8]
+
+    psubw       m0,     m1
+    psubw       m2,     m3
+
+    movu        [r0 + r1 * 2],   m0
+    movu        [r0 + r6],       m2
+
+    lea         r0,     [r0 + r1 * 4]
+    lea         r2,     [r2 + r4 * 4]
+    lea         r3,     [r3 + r5 * 4]
+%endrep
+    RET
+%endif
+%endmacro
+PIXELSUB_PS_W16_H4_avx2 16
+PIXELSUB_PS_W16_H4_avx2 32
+%else
+;-----------------------------------------------------------------------------
+; void pixel_sub_ps_16x16(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
+;-----------------------------------------------------------------------------
+%macro PIXELSUB_PS_W16_H8_avx2 2
+%if ARCH_X86_64
 INIT_YMM avx2
-cglobal pixel_sub_ps_16x16, 6, 7, 4, dest, deststride, src0, src1, srcstride0, srcstride1
+cglobal pixel_sub_ps_16x%2, 6, 10, 4, dest, deststride, src0, src1, srcstride0, srcstride1
     add         r1,     r1
     lea         r6,     [r1 * 3]
+    mov         r7d,    %2/8
 
-%rep 4
+    lea         r9,     [r4 * 3]
+    lea         r8,     [r5 * 3]
+
+.loop
     pmovzxbw    m0,     [r2]
     pmovzxbw    m1,     [r3]
     pmovzxbw    m2,     [r2 + r4]
     pmovzxbw    m3,     [r3 + r5]
-    lea         r2,     [r2 + r4 * 2]
-    lea         r3,     [r3 + r5 * 2]
 
     psubw       m0,     m1
     psubw       m2,     m3
@@ -4356,6 +4631,21 @@
     movu        [r0],            m0
     movu        [r0 + r1],       m2
 
+    pmovzxbw    m0,     [r2 + 2 * r4]
+    pmovzxbw    m1,     [r3 + 2 * r5]
+    pmovzxbw    m2,     [r2 + r9]
+    pmovzxbw    m3,     [r3 + r8]
+
+    psubw       m0,     m1
+    psubw       m2,     m3
+
+    movu        [r0 + r1 * 2],   m0
+    movu        [r0 + r6],       m2
+
+    lea         r0,     [r0 + r1 * 4]
+    lea         r2,     [r2 + r4 * 4]
+    lea         r3,     [r3 + r5 * 4]
+
     pmovzxbw    m0,     [r2]
     pmovzxbw    m1,     [r3]
     pmovzxbw    m2,     [r2 + r4]
@@ -4364,14 +4654,34 @@
     psubw       m0,     m1
     psubw       m2,     m3
 
+    movu        [r0],            m0
+    movu        [r0 + r1],       m2
+
+    pmovzxbw    m0,     [r2 + 2 * r4]
+    pmovzxbw    m1,     [r3 + 2 * r5]
+    pmovzxbw    m2,     [r2 + r9]
+    pmovzxbw    m3,     [r3 + r8]
+
+    psubw       m0,     m1
+    psubw       m2,     m3
+
     movu        [r0 + r1 * 2],   m0
     movu        [r0 + r6],       m2
 
     lea         r0,     [r0 + r1 * 4]
-    lea         r2,     [r2 + r4 * 2]
-    lea         r3,     [r3 + r5 * 2]
-%endrep
+    lea         r2,     [r2 + r4 * 4]
+    lea         r3,     [r3 + r5 * 4]
+
+    dec         r7d
+    jnz         .loop
     RET
+%endif
+%endmacro
+
+PIXELSUB_PS_W16_H8_avx2 16, 16
+PIXELSUB_PS_W16_H8_avx2 16, 32
+%endif
+
 ;-----------------------------------------------------------------------------
 ; void pixel_sub_ps_32x%2(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
 ;-----------------------------------------------------------------------------
@@ -4509,10 +4819,83 @@
 ;-----------------------------------------------------------------------------
 ; void pixel_sub_ps_32x32(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
 ;-----------------------------------------------------------------------------
+%if HIGH_BIT_DEPTH
+%macro PIXELSUB_PS_W32_H4_avx2 1
+%if ARCH_X86_64
 INIT_YMM avx2
-cglobal pixel_sub_ps_32x32, 6, 7, 4, dest, deststride, src0, src1, srcstride0, srcstride1
-     mov        r6d,    4
-     add        r1,     r1
+cglobal pixel_sub_ps_32x%1, 6, 10, 4, dest, deststride, src0, src1, srcstride0, srcstride1
+    add         r1d,    r1d
+    add         r4d,    r4d
+    add         r5d,    r5d
+    mov         r9d,    %1/4
+    lea         r6,     [r1 * 3]
+    lea         r7,     [r4 * 3]
+    lea         r8,     [r5 * 3]
+
+.loop
+    movu        m0,     [r2]
+    movu        m1,     [r2 + 32]
+    movu        m2,     [r3]
+    movu        m3,     [r3 + 32]
+    psubw       m0,     m2
+    psubw       m1,     m3
+
+    movu        [r0],                 m0
+    movu        [r0 + 32],            m1
+
+    movu        m0,     [r2 + r4]
+    movu        m1,     [r2 + r4 + 32]
+    movu        m2,     [r3 + r5]
+    movu        m3,     [r3 + r5 + 32]
+    psubw       m0,     m2
+    psubw       m1,     m3
+
+    movu        [r0 + r1],            m0
+    movu        [r0 + r1 + 32],       m1
+
+    movu        m0,     [r2 + r4 * 2]
+    movu        m1,     [r2 + r4 * 2 + 32]
+    movu        m2,     [r3 + r5 * 2]
+    movu        m3,     [r3 + r5 * 2 + 32]
+    psubw       m0,     m2
+    psubw       m1,     m3
+
+    movu        [r0 + r1 * 2],        m0
+    movu        [r0 + r1 * 2 + 32],   m1
+
+    movu        m0,     [r2 + r7]
+    movu        m1,     [r2 + r7 + 32]
+    movu        m2,     [r3 + r8]
+    movu        m3,     [r3 + r8 + 32]
+    psubw       m0,     m2
+    psubw       m1,     m3
+
+    movu        [r0 + r6],            m0
+    movu        [r0 + r6 + 32],       m1
+
+    lea         r0,     [r0 + r1 * 4]
+    lea         r2,     [r2 + r4 * 4]
+    lea         r3,     [r3 + r5 * 4]
+    dec         r9d
+    jnz         .loop
+    RET
+%endif
+%endmacro
+PIXELSUB_PS_W32_H4_avx2 32
+PIXELSUB_PS_W32_H4_avx2 64
+%else
+;-----------------------------------------------------------------------------
+; void pixel_sub_ps_32x32(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
+;-----------------------------------------------------------------------------
+%macro PIXELSUB_PS_W32_H8_avx2 2
+%if ARCH_X86_64
+INIT_YMM avx2
+cglobal pixel_sub_ps_32x%2, 6, 10, 4, dest, deststride, src0, src1, srcstride0, srcstride1
+    mov        r6d,    %2/8
+    add        r1,     r1
+    lea         r7,         [r4 * 3]
+    lea         r8,         [r5 * 3]
+    lea         r9,         [r1 * 3]
 
 .loop:
     pmovzxbw    m0,     [r2]
@@ -4537,55 +4920,44 @@
     movu        [r0 + r1],       m0
     movu        [r0 + r1 + 32],  m1
 
-    add         r2,     r4
-    add         r3,     r5
-
-    pmovzxbw    m0,     [r2 + r4]
-    pmovzxbw    m1,     [r2 + r4 + 16]
-    pmovzxbw    m2,     [r3 + r5]
-    pmovzxbw    m3,     [r3 + r5 + 16]
+    pmovzxbw    m0,     [r2 + 2 * r4]
+    pmovzxbw    m1,     [r2 + 2 * r4 + 16]
+    pmovzxbw    m2,     [r3 + 2 * r5]
+    pmovzxbw    m3,     [r3 + 2 * r5 + 16]
 
     psubw       m0,     m2
     psubw       m1,     m3
-    lea         r0,     [r0 + r1 * 2]
 
-    movu        [r0 ],           m0
-    movu        [r0 + 32],       m1
-
-    add         r2,     r4
-    add         r3,     r5
+    movu        [r0 + r1 * 2 ],           m0
+    movu        [r0 + r1 * 2 + 32],       m1
 
-    pmovzxbw    m0,     [r2 + r4]
-    pmovzxbw    m1,     [r2 + r4 + 16]
-    pmovzxbw    m2,     [r3 + r5]
-    pmovzxbw    m3,     [r3 + r5 + 16]
+    pmovzxbw    m0,     [r2 + r7]
+    pmovzxbw    m1,     [r2 + r7 + 16]
+    pmovzxbw    m2,     [r3 + r8]
+    pmovzxbw    m3,     [r3 + r8 + 16]
 
 
     psubw       m0,     m2
     psubw       m1,     m3
-    add         r0,     r1
 
-    movu        [r0 ],           m0
-    movu        [r0 + 32],       m1
+    movu        [r0 + r9],           m0
+    movu        [r0 + r9 +32],       m1
 
-    add         r2,     r4
-    add         r3,     r5
+    lea         r2,     [r2 + r4 * 4]
+    lea         r3,     [r3 + r5 * 4]
+    lea         r0,     [r0 + r1 * 4]
 
-    pmovzxbw    m0,     [r2 + r4]
-    pmovzxbw    m1,     [r2 + r4 + 16]
-    pmovzxbw    m2,     [r3 + r5]
-    pmovzxbw    m3,     [r3 + r5 + 16]
+    pmovzxbw    m0,     [r2]
+    pmovzxbw    m1,     [r2 + 16]
+    pmovzxbw    m2,     [r3]
+    pmovzxbw    m3,     [r3 + 16]
 
     psubw       m0,     m2
     psubw       m1,     m3
-    add         r0,     r1
 
     movu        [r0 ],           m0
     movu        [r0 + 32],       m1
 
-    add         r2,     r4
-    add         r3,     r5
-
     pmovzxbw    m0,     [r2 + r4]
     pmovzxbw    m1,     [r2 + r4 + 16]
     pmovzxbw    m2,     [r3 + r5]
@@ -4593,48 +4965,45 @@
 
     psubw       m0,     m2
     psubw       m1,     m3
-    add         r0,     r1
 
-    movu        [r0 ],           m0
-    movu        [r0 + 32],       m1
+    movu        [r0 + r1],           m0
+    movu        [r0 + r1 + 32],       m1
 
-    add         r2,     r4
-    add         r3,     r5
-
-    pmovzxbw    m0,     [r2 + r4]
-    pmovzxbw    m1,     [r2 + r4 + 16]
-    pmovzxbw    m2,     [r3 + r5]
-    pmovzxbw    m3,     [r3 + r5 + 16]
+    pmovzxbw    m0,     [r2 + 2 * r4]
+    pmovzxbw    m1,     [r2 + 2 * r4 + 16]
+    pmovzxbw    m2,     [r3 + 2 * r5]
+    pmovzxbw    m3,     [r3 + 2 * r5 + 16]
 
     psubw       m0,     m2
     psubw       m1,     m3
-    add         r0,     r1
 
-    movu        [r0 ],           m0
-    movu        [r0 + 32],       m1
+    movu        [r0 + r1 * 2],           m0
+    movu        [r0 + r1 * 2 + 32],       m1
 
-    add         r2,     r4
-    add         r3,     r5
-
-    pmovzxbw    m0,     [r2 + r4]
-    pmovzxbw    m1,     [r2 + r4 + 16]
-    pmovzxbw    m2,     [r3 + r5]
-    pmovzxbw    m3,     [r3 + r5 + 16]
+    pmovzxbw    m0,     [r2 + r7]
+    pmovzxbw    m1,     [r2 + r7 + 16]
+    pmovzxbw    m2,     [r3 + r8]
+    pmovzxbw    m3,     [r3 + r8 + 16]
 
     psubw       m0,     m2
     psubw       m1,     m3
-    add         r0,     r1
 
-    movu        [r0 ],           m0
-    movu        [r0 + 32],       m1
+    movu        [r0 + r9],           m0
+    movu        [r0 + r9 + 32],       m1
 
-    lea         r0,     [r0 + r1]
-    lea         r2,     [r2 + r4 * 2]
-    lea         r3,     [r3 + r5 * 2]
+    lea         r0,     [r0 + r1 * 4]
+    lea         r2,     [r2 + r4 * 4]
+    lea         r3,     [r3 + r5 * 4]
 
     dec         r6d
     jnz         .loop
     RET
+%endif
+%endmacro
+
+PIXELSUB_PS_W32_H8_avx2 32, 32
+PIXELSUB_PS_W32_H8_avx2 32, 64
+%endif
 
 ;-----------------------------------------------------------------------------
 ; void pixel_sub_ps_64x%2(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
@@ -4858,6 +5227,102 @@
 ;-----------------------------------------------------------------------------
 ; void pixel_sub_ps_64x64(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
 ;-----------------------------------------------------------------------------
+%if HIGH_BIT_DEPTH
+%if ARCH_X86_64
+INIT_YMM avx2
+cglobal pixel_sub_ps_64x64, 6, 10, 8, dest, deststride, src0, src1, srcstride0, srcstride1
+    add         r1d,    r1d
+    add         r4d,    r4d
+    add         r5d,    r5d
+    mov         r9d,    16
+    lea         r6,     [r1 * 3]
+    lea         r7,     [r4 * 3]
+    lea         r8,     [r5 * 3]
+
+.loop
+    movu        m0,     [r2]
+    movu        m1,     [r2 + 32]
+    movu        m2,     [r2 + 64]
+    movu        m3,     [r2 + 96]
+    movu        m4,     [r3]
+    movu        m5,     [r3 + 32]
+    movu        m6,     [r3 + 64]
+    movu        m7,     [r3 + 96]
+    psubw       m0,     m4
+    psubw       m1,     m5
+    psubw       m2,     m6
+    psubw       m3,     m7
+
+    movu        [r0],                 m0
+    movu        [r0 + 32],            m1
+    movu        [r0 + 64],            m2
+    movu        [r0 + 96],            m3
+
+    movu        m0,     [r2 + r4]
+    movu        m1,     [r2 + r4 + 32]
+    movu        m2,     [r2 + r4 + 64]
+    movu        m3,     [r2 + r4 + 96]
+    movu        m4,     [r3 + r5]
+    movu        m5,     [r3 + r5 + 32]
+    movu        m6,     [r3 + r5 + 64]
+    movu        m7,     [r3 + r5 + 96]
+    psubw       m0,     m4
+    psubw       m1,     m5
+    psubw       m2,     m6
+    psubw       m3,     m7
+
+    movu        [r0 + r1],            m0
+    movu        [r0 + r1 + 32],       m1
+    movu        [r0 + r1 + 64],       m2
+    movu        [r0 + r1 + 96],       m3
+
+    movu        m0,     [r2 + r4 * 2]
+    movu        m1,     [r2 + r4 * 2 + 32]
+    movu        m2,     [r2 + r4 * 2 + 64]
+    movu        m3,     [r2 + r4 * 2 + 96]
+    movu        m4,     [r3 + r5 * 2]
+    movu        m5,     [r3 + r5 * 2 + 32]
+    movu        m6,     [r3 + r5 * 2 + 64]
+    movu        m7,     [r3 + r5 * 2 + 96]
+    psubw       m0,     m4
+    psubw       m1,     m5
+    psubw       m2,     m6
+    psubw       m3,     m7
+
+    movu        [r0 + r1 * 2],        m0
+    movu        [r0 + r1 * 2 + 32],   m1
+    movu        [r0 + r1 * 2 + 64],   m2
+    movu        [r0 + r1 * 2 + 96],   m3
+
+    movu        m0,     [r2 + r7]
+    movu        m1,     [r2 + r7 + 32]
+    movu        m2,     [r2 + r7 + 64]
+    movu        m3,     [r2 + r7 + 96]
+    movu        m4,     [r3 + r8]
+    movu        m5,     [r3 + r8 + 32]
+    movu        m6,     [r3 + r8 + 64]
+    movu        m7,     [r3 + r8 + 96]
+    psubw       m0,     m4
+    psubw       m1,     m5
+    psubw       m2,     m6
+    psubw       m3,     m7
+
+    movu        [r0 + r6],            m0
+    movu        [r0 + r6 + 32],       m1
+    movu        [r0 + r6 + 64],       m2
+    movu        [r0 + r6 + 96],       m3
+
+    lea         r0,     [r0 + r1 * 4]
+    lea         r2,     [r2 + r4 * 4]
+    lea         r3,     [r3 + r5 * 4]
+    dec         r9d
+    jnz         .loop
+    RET
+%endif
+%else
+;-----------------------------------------------------------------------------
+; void pixel_sub_ps_64x64(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
+;-----------------------------------------------------------------------------
 INIT_YMM avx2
 cglobal pixel_sub_ps_64x64, 6, 7, 8, dest, deststride, src0, src1, srcstride0, srcstride1
      mov        r6d,    16
@@ -4963,7 +5428,7 @@
     dec         r6d
     jnz         .loop
     RET
-
+%endif
 ;=============================================================================
 ; variance
 ;=============================================================================
@@ -5387,7 +5852,7 @@
     RET
 %endmacro
 
-;int x265_test_func(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig)
+;int scanPosLast(const uint16_t *scan, const coeff_t *coeff, uint16_t *coeffSign, uint16_t *coeffFlag, uint8_t *coeffNum, int numSig, const uint16_t* scanCG4x4, const int trSize)
 ;{
 ;    int scanPosLast = 0;
 ;    do
@@ -5409,8 +5874,104 @@
 ;}
 
 %if ARCH_X86_64 == 1
+INIT_XMM avx2,bmi2
+cglobal scanPosLast, 7,11,6
+    ; convert unit of Stride(trSize) to int16_t
+    mov         r7d, r7m
+    add         r7d, r7d
+
+    ; loading scan table and convert to Byte
+    mova        m0, [r6]
+    packuswb    m0, [r6 + mmsize]
+    pxor        m1, m0, [pb_15]
+
+    ; clear CG count
+    xor         r9d, r9d
+
+    ; m0 - Zigzag scan table
+    ; m1 - revert order scan table
+    ; m4 - zero
+    ; m5 - ones
+
+    pxor        m4, m4
+    pcmpeqb     m5, m5
+    lea         r8d, [r7d * 3]
+
+.loop:
+    ; position of current CG
+    movzx       r6d, word [r0]
+    lea         r6, [r6 * 2 + r1]
+    add         r0, 16 * 2
+
+    ; loading current CG
+    movh        m2, [r6]
+    movhps      m2, [r6 + r7]
+    movh        m3, [r6 + r7 * 2]
+    movhps      m3, [r6 + r8]
+    packsswb    m2, m3
+
+    ; Zigzag
+    pshufb      m3, m2, m0
+    pshufb      m2, m1
+
+    ; get sign
+    pmovmskb    r6d, m3
+    pcmpeqb     m3, m4
+    pmovmskb    r10d, m3
+    not         r10d
+    pext        r6d, r6d, r10d
+    mov         [r2 + r9 * 2], r6w
+
+    ; get non-zero flag
+    ; TODO: reuse above result with reorder
+    pcmpeqb     m2, m4
+    pxor        m2, m5
+    pmovmskb    r6d, m2
+    mov         [r3 + r9 * 2], r6w
+
+    ; get non-zero number, POPCNT is faster
+    pabsb       m2, m2
+    psadbw      m2, m4
+    movhlps     m3, m2
+    paddd       m2, m3
+    movd        r6d, m2
+    mov         [r4 + r9], r6b
+
+    inc         r9d
+    sub         r5d, r6d
+    jg         .loop
+
+    ; fixup last CG non-zero flag
+    dec         r9d
+    movzx       r0d, word [r3 + r9 * 2]
+;%if cpuflag(bmi1)  ; 2uops?
+;    tzcnt       r1d, r0d
+;%else
+    bsf         r1d, r0d
+;%endif
+    shrx        r0d, r0d, r1d
+    mov         [r3 + r9 * 2], r0w
+
+    ; get last pos
+    mov         eax, r9d
+    shl         eax, 4
+    xor         r1d, 15
+    add         eax, r1d
+    RET
+
+
+; t3 must be ecx, since it's used for shift.
+%if WIN64
+    DECLARE_REG_TMP 3,1,2,0
+%elif ARCH_X86_64
+    DECLARE_REG_TMP 0,1,2,3
+%else ; X86_32
+    %error Unsupport platform X86_32
+%endif
 INIT_CPUFLAGS
-cglobal findPosLast_x64, 5,12
+cglobal scanPosLast_x64, 5,12
+    mov         r10, r3mp
+    movifnidn   t0, r0mp
     mov         r5d, r5m
     xor         r11d, r11d                  ; cgIdx
     xor         r7d, r7d                    ; tmp for non-zero flag
@@ -5418,40 +5979,78 @@
 .loop:
     xor         r8d, r8d                    ; coeffSign[]
     xor         r9d, r9d                    ; coeffFlag[]
-    xor         r10d, r10d                  ; coeffNum[]
+    xor         t3d, t3d                    ; coeffNum[]
 
 %assign x 0
 %rep 16
-    movzx       r6d, word [r0 + x * 2]
-    movsx       r6d, word [r1 + r6 * 2]
+    movzx       r6d, word [t0 + x * 2]
+    movsx       r6d, word [t1 + r6 * 2]
     test        r6d, r6d
     setnz       r7b
     shr         r6d, 31
-    shlx        r6d, r6d, r10d
+    shl         r6d, t3b
     or          r8d, r6d
     lea         r9, [r9 * 2 + r7]
-    add         r10d, r7d
+    add         t3d, r7d
 %assign x x+1
 %endrep
 
     ; store latest group data
-    mov         [r2 + r11 * 2], r8w
-    mov         [r3 + r11 * 2], r9w
-    mov         [r4 + r11], r10b
+    mov         [t2 + r11 * 2], r8w
+    mov         [r10 + r11 * 2], r9w
+    mov         [r4 + r11], t3b
     inc         r11d
 
-    add         r0, 16 * 2
-    sub         r5d, r10d
+    add         t0, 16 * 2
+    sub         r5d, t3d
     jnz        .loop
 
     ; store group data
-    tzcnt       r6d, r9d
-    shrx        r9d, r9d, r6d
-    mov         [r3 + (r11 - 1) * 2], r9w
+    bsf         t3d, r9d
+    shr         r9d, t3b
+    mov         [r10 + (r11 - 1) * 2], r9w
 
     ; get posLast
     shl         r11d, 4
-    sub         r11d, r6d
+    sub         r11d, t3d
     lea         eax, [r11d - 1]
     RET
 %endif
+
+
+;-----------------------------------------------------------------------------
+; uint32_t[last first] findPosFirstAndLast(const int16_t *dstCoeff, const intptr_t trSize, const uint16_t scanTbl[16])
+;-----------------------------------------------------------------------------
+INIT_XMM ssse3
+cglobal findPosFirstLast, 3,3,3
+    ; convert stride to int16_t
+    add         r1d, r1d
+
+    ; loading scan table and convert to Byte
+    mova        m0, [r2]
+    packuswb    m0, [r2 + mmsize]
+
+    ; loading 16 of coeff
+    movh        m1, [r0]
+    movhps      m1, [r0 + r1]
+    movh        m2, [r0 + r1 * 2]
+    lea         r1, [r1 * 3]
+    movhps      m2, [r0 + r1]
+    packsswb    m1, m2
+
+    ; get non-zero mask
+    pxor        m2, m2
+    pcmpeqb     m1, m2
+
+    ; reorder by Zigzag scan
+    pshufb      m1, m0
+
+    ; get First and Last pos
+    xor         eax, eax
+    pmovmskb    r0d, m1
+    not         r0w
+    bsr         r1w, r0w
+    bsf          ax, r0w
+    shl         r1d, 16
+    or          eax, r1d
+    RET

x265_1.6.tar.gz/source/common/x86/pixel.h -> x265_1.7.tar.gz/source/common/x86/pixel.h Changed

@@ -226,6 +226,7 @@
 ADDAVG(addAvg_32x48)
 
 void x265_downShift_16_sse2(const uint16_t* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int width, int height, int shift, uint16_t mask);
+void x265_downShift_16_avx2(const uint16_t* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int width, int height, int shift, uint16_t mask);
 void x265_upShift_8_sse4(const uint8_t* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int width, int height, int shift);
 int x265_psyCost_pp_4x4_sse4(const pixel* source, intptr_t sstride, const pixel* recon, intptr_t rstride);
 int x265_psyCost_pp_8x8_sse4(const pixel* source, intptr_t sstride, const pixel* recon, intptr_t rstride);
@@ -256,10 +257,14 @@
 void x265_pixel_add_ps_16x16_avx2(pixel* a, intptr_t dstride, const pixel* b0, const int16_t* b1, intptr_t sstride0, intptr_t sstride1);
 void x265_pixel_add_ps_32x32_avx2(pixel* a, intptr_t dstride, const pixel* b0, const int16_t* b1, intptr_t sstride0, intptr_t sstride1);
 void x265_pixel_add_ps_64x64_avx2(pixel* a, intptr_t dstride, const pixel* b0, const int16_t* b1, intptr_t sstride0, intptr_t sstride1);
+void x265_pixel_add_ps_16x32_avx2(pixel* a, intptr_t dstride, const pixel* b0, const int16_t* b1, intptr_t sstride0, intptr_t sstride1);
+void x265_pixel_add_ps_32x64_avx2(pixel* a, intptr_t dstride, const pixel* b0, const int16_t* b1, intptr_t sstride0, intptr_t sstride1);
 
 void x265_pixel_sub_ps_16x16_avx2(int16_t* a, intptr_t dstride, const pixel* b0, const pixel* b1, intptr_t sstride0, intptr_t sstride1);
 void x265_pixel_sub_ps_32x32_avx2(int16_t* a, intptr_t dstride, const pixel* b0, const pixel* b1, intptr_t sstride0, intptr_t sstride1);
 void x265_pixel_sub_ps_64x64_avx2(int16_t* a, intptr_t dstride, const pixel* b0, const pixel* b1, intptr_t sstride0, intptr_t sstride1);
+void x265_pixel_sub_ps_16x32_avx2(int16_t* a, intptr_t dstride, const pixel* b0, const pixel* b1, intptr_t sstride0, intptr_t sstride1);
+void x265_pixel_sub_ps_32x64_avx2(int16_t* a, intptr_t dstride, const pixel* b0, const pixel* b1, intptr_t sstride0, intptr_t sstride1);
 
 int x265_psyCost_pp_4x4_avx2(const pixel* source, intptr_t sstride, const pixel* recon, intptr_t rstride);
 int x265_psyCost_pp_8x8_avx2(const pixel* source, intptr_t sstride, const pixel* recon, intptr_t rstride);
@@ -272,6 +277,7 @@
 int x265_psyCost_ss_16x16_avx2(const int16_t* source, intptr_t sstride, const int16_t* recon, intptr_t rstride);
 int x265_psyCost_ss_32x32_avx2(const int16_t* source, intptr_t sstride, const int16_t* recon, intptr_t rstride);
 int x265_psyCost_ss_64x64_avx2(const int16_t* source, intptr_t sstride, const int16_t* recon, intptr_t rstride);
+void x265_weight_sp_avx2(const int16_t* src, pixel* dst, intptr_t srcStride, intptr_t dstStride, int width, int height, int w0, int round, int shift, int offset);
 
 #undef DECL_PIXELS
 #undef DECL_HEVC_SSD

x265_1.6.tar.gz/source/common/x86/pixeladd8.asm -> x265_1.7.tar.gz/source/common/x86/pixeladd8.asm Changed

@@ -398,10 +398,65 @@
 
     jnz         .loop
     RET
+%endif
+%endmacro
+PIXEL_ADD_PS_W16_H4 16, 16
+PIXEL_ADD_PS_W16_H4 16, 32
 
+;-----------------------------------------------------------------------------
+; void pixel_add_ps_16x16(pixel *dest, intptr_t destride, pixel *src0, int16_t *scr1, intptr_t srcStride0, intptr_t srcStride1)
+;-----------------------------------------------------------------------------
+%macro PIXEL_ADD_PS_W16_H4_avx2 1
+%if HIGH_BIT_DEPTH
+%if ARCH_X86_64
 INIT_YMM avx2
-cglobal pixel_add_ps_16x%2, 6, 7, 8, dest, destride, src0, scr1, srcStride0, srcStride1
-    mov         r6d,        %2/4
+cglobal pixel_add_ps_16x%1, 6, 10, 4, dest, destride, src0, scr1, srcStride0, srcStride1
+    mova    m3,     [pw_pixel_max]
+    pxor    m2,     m2
+    mov     r6d,    %1/4
+    add     r4d,    r4d
+    add     r5d,    r5d
+    add     r1d,    r1d
+    lea     r7,     [r4 * 3]
+    lea     r8,     [r5 * 3]
+    lea     r9,     [r1 * 3]
+
+.loop:
+    movu    m0,     [r2]
+    movu    m1,     [r3]
+    paddw   m0,     m1
+    CLIPW   m0, m2, m3
+    movu    [r0],              m0
+
+    movu    m0,     [r2 + r4]
+    movu    m1,     [r3 + r5]
+    paddw   m0,     m1
+    CLIPW   m0, m2, m3
+    movu    [r0 + r1],         m0
+
+    movu    m0,     [r2 + r4 * 2]
+    movu    m1,     [r3 + r5 * 2]
+    paddw   m0,     m1
+    CLIPW   m0, m2, m3
+    movu    [r0 + r1 * 2],     m0
+
+    movu    m0,     [r2 + r7]
+    movu    m1,     [r3 + r8]
+    paddw   m0,     m1
+    CLIPW   m0, m2, m3
+    movu    [r0 + r9],         m0
+
+    dec     r6d
+    lea     r0,     [r0 + r1 * 4]
+    lea     r2,     [r2 + r4 * 4]
+    lea     r3,     [r3 + r5 * 4]
+    jnz     .loop
+    RET
+%endif
+%else
+INIT_YMM avx2
+cglobal pixel_add_ps_16x%1, 6, 7, 8, dest, destride, src0, scr1, srcStride0, srcStride1
+    mov         r6d,        %1/4
     add         r5,         r5
 .loop:
 
@@ -447,8 +502,8 @@
 %endif
 %endmacro
 
-PIXEL_ADD_PS_W16_H4 16, 16
-PIXEL_ADD_PS_W16_H4 16, 32
+PIXEL_ADD_PS_W16_H4_avx2 16
+PIXEL_ADD_PS_W16_H4_avx2 32
 
 
 ;-----------------------------------------------------------------------------
@@ -569,11 +624,90 @@
 
     jnz         .loop
     RET
+%endif
+%endmacro
+PIXEL_ADD_PS_W32_H2 32, 32
+PIXEL_ADD_PS_W32_H2 32, 64
 
+;-----------------------------------------------------------------------------
+; void pixel_add_ps_32x32(pixel *dest, intptr_t destride, pixel *src0, int16_t *scr1, intptr_t srcStride0, intptr_t srcStride1)
+;-----------------------------------------------------------------------------
+%macro PIXEL_ADD_PS_W32_H4_avx2 1
+%if HIGH_BIT_DEPTH
+%if ARCH_X86_64
 INIT_YMM avx2
-cglobal pixel_add_ps_32x%2, 6, 7, 8, dest, destride, src0, scr1, srcStride0, srcStride1
-    mov         r6d,        %2/4
+cglobal pixel_add_ps_32x%1, 6, 10, 6, dest, destride, src0, scr1, srcStride0, srcStride1
+    mova    m5,     [pw_pixel_max]
+    pxor    m4,     m4
+    mov     r6d,    %1/4
+    add     r4d,    r4d
+    add     r5d,    r5d
+    add     r1d,    r1d
+    lea     r7,     [r4 * 3]
+    lea     r8,     [r5 * 3]
+    lea     r9,     [r1 * 3]
+
+.loop:
+    movu    m0,     [r2]
+    movu    m2,     [r2 + 32]
+    movu    m1,     [r3]
+    movu    m3,     [r3 + 32]
+    paddw   m0,     m1
+    paddw   m2,     m3
+    CLIPW2  m0, m2, m4, m5
+
+    movu    [r0],               m0
+    movu    [r0 + 32],          m2
+
+    movu    m0,     [r2 + r4]
+    movu    m2,     [r2 + r4 + 32]
+    movu    m1,     [r3 + r5]
+    movu    m3,     [r3 + r5 + 32]
+    paddw   m0,     m1
+    paddw   m2,     m3
+    CLIPW2  m0, m2, m4, m5
+
+    movu    [r0 + r1],          m0
+    movu    [r0 + r1 + 32],     m2
+
+    movu    m0,     [r2 + r4 * 2]
+    movu    m2,     [r2 + r4 * 2 + 32]
+    movu    m1,     [r3 + r5 * 2]
+    movu    m3,     [r3 + r5 * 2 + 32]
+    paddw   m0,     m1
+    paddw   m2,     m3
+    CLIPW2  m0, m2, m4, m5
+
+    movu    [r0 + r1 * 2],      m0
+    movu    [r0 + r1 * 2 + 32], m2
+
+    movu    m0,     [r2 + r7]
+    movu    m2,     [r2 + r7 + 32]
+    movu    m1,     [r3 + r8]
+    movu    m3,     [r3 + r8 + 32]
+    paddw   m0,     m1
+    paddw   m2,     m3
+    CLIPW2  m0, m2, m4, m5
+
+    movu    [r0 + r9],          m0
+    movu    [r0 + r9 + 32],     m2
+
+    dec     r6d
+    lea     r0,     [r0 + r1 * 4]
+    lea     r2,     [r2 + r4 * 4]
+    lea     r3,     [r3 + r5 * 4]
+    jnz     .loop
+    RET
+%endif
+%else
+%if ARCH_X86_64
+INIT_YMM avx2
+cglobal pixel_add_ps_32x%1, 6, 10, 8, dest, destride, src0, scr1, srcStride0, srcStride1
+    mov         r6d,        %1/4
     add         r5,         r5
+    lea         r7,         [r4 * 3]
+    lea         r8,         [r5 * 3]
+    lea         r9,         [r1 * 3]
 .loop:
     pmovzxbw    m0,         [r2]                ; first half of row 0 of src0
     pmovzxbw    m1,         [r2 + 16]           ; second half of row 0 of src0
@@ -597,44 +731,41 @@
     vpermq      m0, m0, 11011000b
     movu        [r0 + r1],      m0              ; row 1 of dst
 
-    lea         r2,         [r2 + r4 * 2]
-    lea         r3,         [r3 + r5 * 2]
-    lea         r0,         [r0 + r1 * 2]
-
-    pmovzxbw    m0,         [r2]                ; first half of row 2 of src0
-    pmovzxbw    m1,         [r2 + 16]           ; second half of row 2 of src0
-    movu        m2,         [r3]                ; first half of row 2 of src1
-    movu        m3,         [r3 + 32]           ; second half of row 2 of src1
+    pmovzxbw    m0,         [r2 + r4 * 2]       ; first half of row 2 of src0
+    pmovzxbw    m1,         [r2 + r4 * 2 + 16]  ; second half of row 2 of src0
+    movu        m2,         [r3 + r5 * 2]       ; first half of row 2 of src1
+    movu        m3,         [r3 + + r5 * 2 + 32]; second half of row 2 of src1
 
     paddw       m0,         m2
     paddw       m1,         m3
     packuswb    m0,         m1
     vpermq      m0, m0, 11011000b
-    movu        [r0],      m0                   ; row 2 of dst
+    movu        [r0 + r1 * 2],      m0          ; row 2 of dst
 
-    pmovzxbw    m0,         [r2 + r4]           ; first half of row 3 of src0
-    pmovzxbw    m1,         [r2 + r4 + 16]      ; second half of row 3 of src0
-    movu        m2,         [r3 + r5]           ; first half of row 3 of src1
-    movu        m3,         [r3 + r5 + 32]      ; second half of row 3 of src1
+    pmovzxbw    m0,         [r2 + r7]           ; first half of row 3 of src0
+    pmovzxbw    m1,         [r2 + r7 + 16]      ; second half of row 3 of src0
+    movu        m2,         [r3 + r8]           ; first half of row 3 of src1
+    movu        m3,         [r3 + r8 + 32]      ; second half of row 3 of src1
 
     paddw       m0,         m2
     paddw       m1,         m3
     packuswb    m0,         m1
     vpermq      m0, m0, 11011000b
-    movu        [r0 + r1],      m0              ; row 3 of dst
+    movu        [r0 + r9],      m0              ; row 3 of dst
 
-    lea         r2,         [r2 + r4 * 2]
-    lea         r3,         [r3 + r5 * 2]
-    lea         r0,         [r0 + r1 * 2]
+    lea         r2,         [r2 + r4 * 4]
+    lea         r3,         [r3 + r5 * 4]
+    lea         r0,         [r0 + r1 * 4]
 
     dec         r6d
     jnz         .loop
     RET
 %endif
+%endif
 %endmacro
 
-PIXEL_ADD_PS_W32_H2 32, 32
-PIXEL_ADD_PS_W32_H2 32, 64
+PIXEL_ADD_PS_W32_H4_avx2 32
+PIXEL_ADD_PS_W32_H4_avx2 64
 
 
 ;-----------------------------------------------------------------------------
@@ -841,10 +972,127 @@
 
     jnz         .loop
     RET
+%endif
+%endmacro
+PIXEL_ADD_PS_W64_H2 64, 64
 
+;-----------------------------------------------------------------------------
+; void pixel_add_ps_64x64(pixel *dest, intptr_t destride, pixel *src0, int16_t *scr1, intptr_t srcStride0, intptr_t srcStride1)
+;-----------------------------------------------------------------------------
+%if HIGH_BIT_DEPTH
+%if ARCH_X86_64
 INIT_YMM avx2
-cglobal pixel_add_ps_64x%2, 6, 7, 8, dest, destride, src0, scr1, srcStride0, srcStride1
-    mov         r6d,        %2/2
+cglobal pixel_add_ps_64x64, 6, 10, 6, dest, destride, src0, scr1, srcStride0, srcStride1
+    mova    m5,     [pw_pixel_max]
+    pxor    m4,     m4
+    mov     r6d,    16
+    add     r4d,    r4d
+    add     r5d,    r5d
+    add     r1d,    r1d
+    lea     r7,     [r4 * 3]
+    lea     r8,     [r5 * 3]
+    lea     r9,     [r1 * 3]
+
+.loop:
+    movu    m0,     [r2]
+    movu    m1,     [r2 + 32]
+    movu    m2,     [r3]
+    movu    m3,     [r3 + 32]
+    paddw   m0,     m2
+    paddw   m1,     m3
+
+    CLIPW2  m0, m1, m4, m5
+    movu    [r0],                m0
+    movu    [r0 + 32],           m1
+
+    movu    m0,     [r2 + 64]
+    movu    m1,     [r2 + 96]
+    movu    m2,     [r3 + 64]
+    movu    m3,     [r3 + 96]
+    paddw   m0,     m2
+    paddw   m1,     m3
+
+    CLIPW2  m0, m1, m4, m5
+    movu    [r0 + 64],           m0
+    movu    [r0 + 96],           m1
+
+    movu    m0,     [r2 + r4]
+    movu    m1,     [r2 + r4 + 32]
+    movu    m2,     [r3 + r5]
+    movu    m3,     [r3 + r5 + 32]
+    paddw   m0,     m2
+    paddw   m1,     m3
+
+    CLIPW2  m0, m1, m4, m5
+    movu    [r0 + r1],           m0
+    movu    [r0 + r1 + 32],      m1
+
+    movu    m0,     [r2 + r4 + 64]
+    movu    m1,     [r2 + r4 + 96]
+    movu    m2,     [r3 + r5 + 64]
+    movu    m3,     [r3 + r5 + 96]
+    paddw   m0,     m2
+    paddw   m1,     m3
+
+    CLIPW2  m0, m1, m4, m5
+    movu    [r0 + r1 + 64],      m0
+    movu    [r0 + r1 + 96],      m1
+
+    movu    m0,     [r2 + r4 * 2]
+    movu    m1,     [r2 + r4 * 2 + 32]
+    movu    m2,     [r3 + r5 * 2]
+    movu    m3,     [r3 + r5 * 2+ 32]
+    paddw   m0,     m2
+    paddw   m1,     m3
+
+    CLIPW2  m0, m1, m4, m5
+    movu    [r0 + r1 * 2],       m0
+    movu    [r0 + r1 * 2 + 32],  m1
+
+    movu    m0,     [r2 + r4 * 2 + 64]
+    movu    m1,     [r2 + r4 * 2 + 96]
+    movu    m2,     [r3 + r5 * 2 + 64]
+    movu    m3,     [r3 + r5 * 2 + 96]
+    paddw   m0,     m2
+    paddw   m1,     m3
+
+    CLIPW2  m0, m1, m4, m5
+    movu    [r0 + r1 * 2 + 64],  m0
+    movu    [r0 + r1 * 2 + 96],  m1
+
+    movu    m0,     [r2 + r7]
+    movu    m1,     [r2 + r7 + 32]
+    movu    m2,     [r3 + r8]
+    movu    m3,     [r3 + r8 + 32]
+    paddw   m0,     m2
+    paddw   m1,     m3
+
+    CLIPW2  m0, m1, m4, m5
+    movu    [r0 + r9],           m0
+    movu    [r0 + r9 + 32],      m1
+
+    movu    m0,     [r2 + r7 + 64]
+    movu    m1,     [r2 + r7 + 96]
+    movu    m2,     [r3 + r8 + 64]
+    movu    m3,     [r3 + r8 + 96]
+    paddw   m0,     m2
+    paddw   m1,     m3
+
+    CLIPW2  m0, m1, m4, m5
+    movu    [r0 + r9 + 64],      m0
+    movu    [r0 + r9 + 96],      m1
+
+    dec     r6d
+    lea     r0,     [r0 + r1 * 4]
+    lea     r2,     [r2 + r4 * 4]
+    lea     r3,     [r3 + r5 * 4]
+    jnz     .loop
+    RET
+%endif
+%else
+INIT_YMM avx2
+cglobal pixel_add_ps_64x64, 6, 7, 8, dest, destride, src0, scr1, srcStride0, srcStride1
+    mov         r6d,        32
     add         r5,         r5
 .loop:
     pmovzxbw    m0,         [r2]                ; first 16 of row 0 of src0
@@ -896,6 +1144,3 @@
     RET
 
 %endif
-%endmacro
-
-PIXEL_ADD_PS_W64_H2 64, 64

x265_1.6.tar.gz/source/common/x86/sad-a.asm -> x265_1.7.tar.gz/source/common/x86/sad-a.asm Changed

@@ -4004,10 +4004,12 @@
     RET
 
 INIT_YMM avx2
-cglobal pixel_sad_32x24, 4,5,6
+cglobal pixel_sad_32x24, 4,7,6
     xorps           m0, m0
     xorps           m5, m5
     mov             r4d, 6
+    lea             r5, [r1 * 3]
+    lea             r6, [r3 * 3]
 .loop
     movu           m1, [r0]               ; row 0 of pix0
     movu           m2, [r2]               ; row 0 of pix1
@@ -4019,21 +4021,18 @@
     paddd          m0, m1
     paddd          m5, m3
 
-    lea     r2,     [r2 + 2 * r3]
-    lea     r0,     [r0 + 2 * r1]
-
-    movu           m1, [r0]               ; row 2 of pix0
-    movu           m2, [r2]               ; row 2 of pix1
-    movu           m3, [r0 + r1]          ; row 3 of pix0
-    movu           m4, [r2 + r3]          ; row 3 of pix1
+    movu           m1, [r0 + 2 * r1]      ; row 2 of pix0
+    movu           m2, [r2 + 2 * r3]      ; row 2 of pix1
+    movu           m3, [r0 + r5]          ; row 3 of pix0
+    movu           m4, [r2 + r6]          ; row 3 of pix1
 
     psadbw         m1, m2
     psadbw         m3, m4
     paddd          m0, m1
     paddd          m5, m3
 
-    lea     r2,     [r2 + 2 * r3]
-    lea     r0,     [r0 + 2 * r1]
+    lea     r2,     [r2 + 4 * r3]
+    lea     r0,     [r0 + 4 * r1]
 
     dec         r4d
     jnz         .loop
@@ -4307,10 +4306,12 @@
     RET
 
 INIT_YMM avx2
-cglobal pixel_sad_64x48, 4,5,6
+cglobal pixel_sad_64x48, 4,7,6
     xorps           m0, m0
     xorps           m5, m5
-    mov             r4d, 24
+    mov             r4d, 12
+    lea             r5, [r1 * 3]
+    lea             r6, [r3 * 3]
 .loop
     movu           m1, [r0]               ; first 32 of row 0 of pix0
     movu           m2, [r2]               ; first 32 of row 0 of pix1
@@ -4332,8 +4333,28 @@
     paddd          m0, m1
     paddd          m5, m3
 
-    lea     r2,     [r2 + 2 * r3]
-    lea     r0,     [r0 + 2 * r1]
+    movu           m1, [r0 + 2 * r1]      ; first 32 of row 0 of pix0
+    movu           m2, [r2 + 2 * r3]      ; first 32 of row 0 of pix1
+    movu           m3, [r0 + 2 * r1 + 32] ; second 32 of row 0 of pix0
+    movu           m4, [r2 + 2 * r3 + 32] ; second 32 of row 0 of pix1
+
+    psadbw         m1, m2
+    psadbw         m3, m4
+    paddd          m0, m1
+    paddd          m5, m3
+
+    movu           m1, [r0 + r5]          ; first 32 of row 1 of pix0
+    movu           m2, [r2 + r6]          ; first 32 of row 1 of pix1
+    movu           m3, [r0 + 32 + r5]     ; second 32 of row 1 of pix0
+    movu           m4, [r2 + 32 + r6]     ; second 32 of row 1 of pix1
+
+    psadbw         m1, m2
+    psadbw         m3, m4
+    paddd          m0, m1
+    paddd          m5, m3
+
+    lea     r2,     [r2 + 4 * r3]
+    lea     r0,     [r0 + 4 * r1]
 
     dec         r4d
     jnz         .loop
@@ -4347,10 +4368,12 @@
     RET
 
 INIT_YMM avx2
-cglobal pixel_sad_64x64, 4,5,6
+cglobal pixel_sad_64x64, 4,7,6
     xorps           m0, m0
     xorps           m5, m5
     mov             r4d, 8
+    lea             r5, [r1 * 3]
+    lea             r6, [r3 * 3]
 .loop
     movu           m1, [r0]               ; first 32 of row 0 of pix0
     movu           m2, [r2]               ; first 32 of row 0 of pix1
@@ -4372,31 +4395,28 @@
     paddd          m0, m1
     paddd          m5, m3
 
-    lea     r2,     [r2 + 2 * r3]
-    lea     r0,     [r0 + 2 * r1]
-
-    movu           m1, [r0]               ; first 32 of row 2 of pix0
-    movu           m2, [r2]               ; first 32 of row 2 of pix1
-    movu           m3, [r0 + 32]          ; second 32 of row 2 of pix0
-    movu           m4, [r2 + 32]          ; second 32 of row 2 of pix1
+    movu           m1, [r0 + 2 * r1]      ; first 32 of row 2 of pix0
+    movu           m2, [r2 + 2 * r3]      ; first 32 of row 2 of pix1
+    movu           m3, [r0 + 2 * r1 + 32] ; second 32 of row 2 of pix0
+    movu           m4, [r2 + 2 * r3 + 32] ; second 32 of row 2 of pix1
 
     psadbw         m1, m2
     psadbw         m3, m4
     paddd          m0, m1
     paddd          m5, m3
 
-    movu           m1, [r0 + r1]          ; first 32 of row 3 of pix0
-    movu           m2, [r2 + r3]          ; first 32 of row 3 of pix1
-    movu           m3, [r0 + 32 + r1]     ; second 32 of row 3 of pix0
-    movu           m4, [r2 + 32 + r3]     ; second 32 of row 3 of pix1
+    movu           m1, [r0 + r5]          ; first 32 of row 3 of pix0
+    movu           m2, [r2 + r6]          ; first 32 of row 3 of pix1
+    movu           m3, [r0 + 32 + r5]     ; second 32 of row 3 of pix0
+    movu           m4, [r2 + 32 + r6]     ; second 32 of row 3 of pix1
 
     psadbw         m1, m2
     psadbw         m3, m4
     paddd          m0, m1
     paddd          m5, m3
 
-    lea     r2,     [r2 + 2 * r3]
-    lea     r0,     [r0 + 2 * r1]
+    lea     r2,     [r2 + 4 * r3]
+    lea     r0,     [r0 + 4 * r1]
 
     movu           m1, [r0]               ; first 32 of row 4 of pix0
     movu           m2, [r2]               ; first 32 of row 4 of pix1
@@ -4418,31 +4438,28 @@
     paddd          m0, m1
     paddd          m5, m3
 
-    lea     r2,     [r2 + 2 * r3]
-    lea     r0,     [r0 + 2 * r1]
-
-    movu           m1, [r0]               ; first 32 of row 6 of pix0
-    movu           m2, [r2]               ; first 32 of row 6 of pix1
-    movu           m3, [r0 + 32]          ; second 32 of row 6 of pix0
-    movu           m4, [r2 + 32]          ; second 32 of row 6 of pix1
+    movu           m1, [r0 + 2 * r1]      ; first 32 of row 6 of pix0
+    movu           m2, [r2 + 2 * r3]      ; first 32 of row 6 of pix1
+    movu           m3, [r0 + 2 * r1 + 32] ; second 32 of row 6 of pix0
+    movu           m4, [r2 + 2 * r3 + 32] ; second 32 of row 6 of pix1
 
     psadbw         m1, m2
     psadbw         m3, m4
     paddd          m0, m1
     paddd          m5, m3
 
-    movu           m1, [r0 + r1]          ; first 32 of row 7 of pix0
-    movu           m2, [r2 + r3]          ; first 32 of row 7 of pix1
-    movu           m3, [r0 + 32 + r1]     ; second 32 of row 7 of pix0
-    movu           m4, [r2 + 32 + r3]     ; second 32 of row 7 of pix1
+    movu           m1, [r0 + r5]          ; first 32 of row 7 of pix0
+    movu           m2, [r2 + r6]          ; first 32 of row 7 of pix1
+    movu           m3, [r0 + 32 + r5]     ; second 32 of row 7 of pix0
+    movu           m4, [r2 + 32 + r6]     ; second 32 of row 7 of pix1
 
     psadbw         m1, m2
     psadbw         m3, m4
     paddd          m0, m1
     paddd          m5, m3
 
-    lea     r2,     [r2 + 2 * r3]
-    lea     r0,     [r0 + 2 * r1]
+    lea     r2,     [r2 + 4 * r3]
+    lea     r0,     [r0 + 4 * r1]
 
     dec         r4d
     jnz         .loop

x265_1.6.tar.gz/source/common/x86/sad16-a.asm -> x265_1.7.tar.gz/source/common/x86/sad16-a.asm Changed

@@ -276,9 +276,8 @@
     ABSW2   m3, m4, m3, m4, m7, m5
     paddw   m1, m2
     paddw   m3, m4
-    paddw   m3, m1
-    pmaddwd m3, [pw_1]
-    paddd   m0, m3
+    paddw   m0, m1
+    paddw   m0, m3
 %else
     movu    m1, [r2]
     movu    m2, [r2+2*r3]
@@ -287,15 +286,45 @@
     ABSW2   m1, m2, m1, m2, m3, m4
     lea     r0, [r0+4*r1]
     lea     r2, [r2+4*r3]
-    paddw   m2, m1
-    pmaddwd m2, [pw_1]
-    paddd   m0, m2
+    paddw   m0, m1
+    paddw   m0, m2
 %endif
 %endmacro
 
-;-----------------------------------------------------------------------------
-; int pixel_sad_NxM( uint16_t *, intptr_t, uint16_t *, intptr_t )
-;-----------------------------------------------------------------------------
+%macro SAD_INC_2ROW_Nx64 1
+%if 2*%1 > mmsize
+    movu    m1, [r2 + 0]
+    movu    m2, [r2 + 16]
+    movu    m3, [r2 + 2 * r3 + 0]
+    movu    m4, [r2 + 2 * r3 + 16]
+    psubw   m1, [r0 + 0]
+    psubw   m2, [r0 + 16]
+    psubw   m3, [r0 + 2 * r1 + 0]
+    psubw   m4, [r0 + 2 * r1 + 16]
+    ABSW2   m1, m2, m1, m2, m5, m6
+    lea     r0, [r0 + 4 * r1]
+    lea     r2, [r2 + 4 * r3]
+    ABSW2   m3, m4, m3, m4, m7, m5
+    paddw   m1, m2
+    paddw   m3, m4
+    paddw   m0, m1
+    paddw   m8, m3
+%else
+    movu    m1, [r2]
+    movu    m2, [r2 + 2 * r3]
+    psubw   m1, [r0]
+    psubw   m2, [r0 + 2 * r1]
+    ABSW2   m1, m2, m1, m2, m3, m4
+    lea     r0, [r0 + 4 * r1]
+    lea     r2, [r2 + 4 * r3]
+    paddw   m0, m1
+    paddw   m8, m2
+%endif
+%endmacro
+
+; ---------------------------------------------------------------------------- -
+; int pixel_sad_NxM(uint16_t *, intptr_t, uint16_t *, intptr_t)
+; ---------------------------------------------------------------------------- -
 %macro SAD 2
 cglobal pixel_sad_%1x%2, 4,5-(%2&4/4),8*(%1/mmsize)
     pxor    m0, m0
@@ -309,8 +338,35 @@
     dec    r4d
     jg .loop
 %endif
+%if %2 == 32
+    HADDUWD m0, m1
+    HADDD   m0, m1
+%else
+    HADDW   m0, m1
+%endif
+    movd    eax, xm0
+    RET
+%endmacro
 
+; ---------------------------------------------------------------------------- -
+; int pixel_sad_Nx64(uint16_t *, intptr_t, uint16_t *, intptr_t)
+; ---------------------------------------------------------------------------- -
+%macro SAD_Nx64 1
+cglobal pixel_sad_%1x64, 4,5-(64&4/4), 9
+    pxor    m0, m0
+    pxor    m8, m8
+    mov     r4d, 64 / 2
+.loop:
+    SAD_INC_2ROW_Nx64 %1
+    dec    r4d
+    jg .loop
+
+    HADDUWD m0, m1
+    HADDUWD m8, m1
     HADDD   m0, m1
+    HADDD   m8, m1
+    paddd   m0, m8
+
     movd    eax, xm0
     RET
 %endmacro
@@ -321,7 +377,7 @@
 SAD  16, 12
 SAD  16, 16
 SAD  16, 32
-SAD  16, 64
+SAD_Nx64  16
 
 INIT_XMM sse2
 SAD  8,  4
@@ -329,6 +385,13 @@
 SAD  8, 16
 SAD  8, 32
 
+INIT_YMM avx2
+SAD  16,  4
+SAD  16,  8
+SAD  16, 12
+SAD  16, 16
+SAD  16, 32
+
 ;------------------------------------------------------------------
 ; int pixel_sad_32xN( uint16_t *, intptr_t, uint16_t *, intptr_t )
 ;------------------------------------------------------------------
@@ -716,7 +779,6 @@
 %endif
     movd     eax, xm0
     RET
-
 ;-----------------------------------------------------------------------------
 ; void pixel_sad_xN_WxH( uint16_t *fenc, uint16_t *pix0, uint16_t *pix1,
 ;                        uint16_t *pix2, intptr_t i_stride, int scores[3] )

x265_1.6.tar.gz/source/common/x86/x86inc.asm -> x265_1.7.tar.gz/source/common/x86/x86inc.asm Changed

x265_1.6.tar.gz/source/encoder/CMakeLists.txt -> x265_1.7.tar.gz/source/encoder/CMakeLists.txt Changed

x265_1.6.tar.gz/source/encoder/analysis.cpp -> x265_1.7.tar.gz/source/encoder/analysis.cpp Changed

@@ -130,9 +130,12 @@
     for (uint32_t i = 0; i <= g_maxCUDepth; i++)
         for (uint32_t j = 0; j < MAX_PRED_TYPES; j++)
             m_modeDepth[i].pred[j].invalidate();
-#endif
     invalidateContexts(0);
-    m_quant.setQPforQuant(ctu);
+#endif
+
+    int qp = setLambdaFromQP(ctu, m_slice->m_pps->bUseDQP ? calculateQpforCuSize(ctu, cuGeom) : m_slice->m_sliceQp);
+    ctu.setQPSubParts((int8_t)qp, 0, 0);
+
     m_rqt[0].cur.load(initialContext);
     m_modeDepth[0].fencYuv.copyFromPicYuv(*m_frame->m_fencPic, ctu.m_cuAddr, 0);
 
@@ -140,11 +143,11 @@
     if (m_param->analysisMode)
     {
         if (m_slice->m_sliceType == I_SLICE)
-            m_reuseIntraDataCTU = (analysis_intra_data *)m_frame->m_analysisData.intraData;
+            m_reuseIntraDataCTU = (analysis_intra_data*)m_frame->m_analysisData.intraData;
         else
         {
             int numPredDir = m_slice->isInterP() ? 1 : 2;
-            m_reuseInterDataCTU = (analysis_inter_data *)m_frame->m_analysisData.interData;
+            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_reuseBestMergeCand = &m_reuseInterDataCTU->bestMergeCand[ctu.m_cuAddr * CUGeom::MAX_GEOMS];
         }
@@ -155,10 +158,10 @@
     uint32_t zOrder = 0;
     if (m_slice->m_sliceType == I_SLICE)
     {
-        compressIntraCU(ctu, cuGeom, zOrder);
+        compressIntraCU(ctu, cuGeom, zOrder, qp);
         if (m_param->analysisMode == X265_ANALYSIS_SAVE && m_frame->m_analysisData.intraData)
         {
-            CUData *bestCU = &m_modeDepth[0].bestMode->cu;
+            CUData* bestCU = &m_modeDepth[0].bestMode->cu;
             memcpy(&m_reuseIntraDataCTU->depth[ctu.m_cuAddr * numPartition], bestCU->m_cuDepth, sizeof(uint8_t) * numPartition);
             memcpy(&m_reuseIntraDataCTU->modes[ctu.m_cuAddr * numPartition], bestCU->m_lumaIntraDir, sizeof(uint8_t) * numPartition);
             memcpy(&m_reuseIntraDataCTU->partSizes[ctu.m_cuAddr * numPartition], bestCU->m_partSize, sizeof(uint8_t) * numPartition);
@@ -173,21 +176,21 @@
             * they are available for intra predictions */
             m_modeDepth[0].fencYuv.copyToPicYuv(*m_frame->m_reconPic, ctu.m_cuAddr, 0);
 
-            compressInterCU_rd0_4(ctu, cuGeom);
+            compressInterCU_rd0_4(ctu, cuGeom, qp);
 
             /* generate residual for entire CTU at once and copy to reconPic */
             encodeResidue(ctu, cuGeom);
         }
         else if (m_param->bDistributeModeAnalysis && m_param->rdLevel >= 2)
-            compressInterCU_dist(ctu, cuGeom);
+            compressInterCU_dist(ctu, cuGeom, qp);
         else if (m_param->rdLevel <= 4)
-            compressInterCU_rd0_4(ctu, cuGeom);
+            compressInterCU_rd0_4(ctu, cuGeom, qp);
         else
         {
-            compressInterCU_rd5_6(ctu, cuGeom, zOrder);
+            compressInterCU_rd5_6(ctu, cuGeom, zOrder, qp);
             if (m_param->analysisMode == X265_ANALYSIS_SAVE && m_frame->m_analysisData.interData)
             {
-                CUData *bestCU = &m_modeDepth[0].bestMode->cu;
+                CUData* bestCU = &m_modeDepth[0].bestMode->cu;
                 memcpy(&m_reuseInterDataCTU->depth[ctu.m_cuAddr * numPartition], bestCU->m_cuDepth, sizeof(uint8_t) * numPartition);
                 memcpy(&m_reuseInterDataCTU->modes[ctu.m_cuAddr * numPartition], bestCU->m_predMode, sizeof(uint8_t) * numPartition);
             }
@@ -206,24 +209,28 @@
         return;
     else if (md.bestMode->cu.isIntra(0))
     {
+        m_quant.m_tqBypass = true;
         md.pred[PRED_LOSSLESS].initCosts();
         md.pred[PRED_LOSSLESS].cu.initLosslessCU(md.bestMode->cu, cuGeom);
         PartSize size = (PartSize)md.pred[PRED_LOSSLESS].cu.m_partSize[0];
         uint8_t* modes = md.pred[PRED_LOSSLESS].cu.m_lumaIntraDir;
         checkIntra(md.pred[PRED_LOSSLESS], cuGeom, size, modes, NULL);
         checkBestMode(md.pred[PRED_LOSSLESS], cuGeom.depth);
+        m_quant.m_tqBypass = false;
     }
     else
     {
+        m_quant.m_tqBypass = true;
         md.pred[PRED_LOSSLESS].initCosts();
         md.pred[PRED_LOSSLESS].cu.initLosslessCU(md.bestMode->cu, cuGeom);
         md.pred[PRED_LOSSLESS].predYuv.copyFromYuv(md.bestMode->predYuv);
         encodeResAndCalcRdInterCU(md.pred[PRED_LOSSLESS], cuGeom);
         checkBestMode(md.pred[PRED_LOSSLESS], cuGeom.depth);
+        m_quant.m_tqBypass = false;
     }
 }
 
-void Analysis::compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, uint32_t& zOrder)
+void Analysis::compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, uint32_t& zOrder, int32_t qp)
 {
     uint32_t depth = cuGeom.depth;
     ModeDepth& md = m_modeDepth[depth];
@@ -241,11 +248,9 @@
 
         if (mightNotSplit && depth == reuseDepth[zOrder] && zOrder == cuGeom.absPartIdx)
         {
-            m_quant.setQPforQuant(parentCTU);
-
             PartSize size = (PartSize)reusePartSizes[zOrder];
             Mode& mode = size == SIZE_2Nx2N ? md.pred[PRED_INTRA] : md.pred[PRED_INTRA_NxN];
-            mode.cu.initSubCU(parentCTU, cuGeom);
+            mode.cu.initSubCU(parentCTU, cuGeom, qp);
             checkIntra(mode, cuGeom, size, &reuseModes[zOrder], &reuseChromaModes[zOrder]);
             checkBestMode(mode, depth);
 
@@ -262,15 +267,13 @@
     }
     else if (mightNotSplit)
     {
-        m_quant.setQPforQuant(parentCTU);
-
-        md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom);
+        md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
         checkIntra(md.pred[PRED_INTRA], cuGeom, SIZE_2Nx2N, NULL, NULL);
         checkBestMode(md.pred[PRED_INTRA], depth);
 
         if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3)
         {
-            md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom);
+            md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom, qp);
             checkIntra(md.pred[PRED_INTRA_NxN], cuGeom, SIZE_NxN, NULL, NULL);
             checkBestMode(md.pred[PRED_INTRA_NxN], depth);
         }
@@ -287,12 +290,13 @@
         Mode* splitPred = &md.pred[PRED_SPLIT];
         splitPred->initCosts();
         CUData* splitCU = &splitPred->cu;
-        splitCU->initSubCU(parentCTU, cuGeom);
+        splitCU->initSubCU(parentCTU, cuGeom, qp);
 
         uint32_t nextDepth = depth + 1;
         ModeDepth& nd = m_modeDepth[nextDepth];
         invalidateContexts(nextDepth);
         Entropy* nextContext = &m_rqt[depth].cur;
+        int32_t nextQP = qp;
 
         for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
         {
@@ -301,7 +305,11 @@
             {
                 m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                 m_rqt[nextDepth].cur.load(*nextContext);
-                compressIntraCU(parentCTU, childGeom, zOrder);
+
+                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
+                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));
+
+                compressIntraCU(parentCTU, childGeom, zOrder, nextQP);
 
                 // Save best CU and pred data for this sub CU
                 splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
@@ -322,7 +330,7 @@
         else
             updateModeCost(*splitPred);
 
-        checkDQPForSplitPred(splitPred->cu, cuGeom);
+        checkDQPForSplitPred(*splitPred, cuGeom);
         checkBestMode(*splitPred, depth);
     }
 
@@ -362,24 +370,18 @@
     }
 
     ModeDepth& md = m_modeDepth[pmode.cuGeom.depth];
-    bool bMergeOnly = pmode.cuGeom.log2CUSize == 6;
 
     /* setup slave Analysis */
     if (&slave != this)
     {
         slave.m_slice = m_slice;
         slave.m_frame = m_frame;
-        slave.setQP(*m_slice, m_rdCost.m_qp);
+        slave.m_param = m_param;
+        slave.setLambdaFromQP(md.pred[PRED_2Nx2N].cu, m_rdCost.m_qp);
         slave.invalidateContexts(0);
-
-        if (m_param->rdLevel >= 5)
-        {
-            slave.m_rqt[pmode.cuGeom.depth].cur.load(m_rqt[pmode.cuGeom.depth].cur);
-            slave.m_quant.setQPforQuant(md.pred[PRED_2Nx2N].cu);
-        }
+        slave.m_rqt[pmode.cuGeom.depth].cur.load(m_rqt[pmode.cuGeom.depth].cur);
     }
 
-
     /* perform Mode task, repeat until no more work is available */
     do
     {
@@ -388,8 +390,6 @@
             switch (pmode.modes[task])
             {
             case PRED_INTRA:
-                if (&slave != this)
-                    slave.m_rqt[pmode.cuGeom.depth].cur.load(m_rqt[pmode.cuGeom.depth].cur);
                 slave.checkIntraInInter(md.pred[PRED_INTRA], pmode.cuGeom);
                 if (m_param->rdLevel > 2)
                     slave.encodeIntraInInter(md.pred[PRED_INTRA], pmode.cuGeom);
@@ -441,7 +441,7 @@
                 break;
 
             case PRED_2Nx2N:
-                slave.checkInter_rd5_6(md.pred[PRED_2Nx2N], pmode.cuGeom, SIZE_2Nx2N, false);
+                slave.checkInter_rd5_6(md.pred[PRED_2Nx2N], pmode.cuGeom, SIZE_2Nx2N);
                 md.pred[PRED_BIDIR].rdCost = MAX_INT64;
                 if (m_slice->m_sliceType == B_SLICE)
                 {
@@ -452,27 +452,27 @@
                 break;
 
             case PRED_Nx2N:
-                slave.checkInter_rd5_6(md.pred[PRED_Nx2N], pmode.cuGeom, SIZE_Nx2N, false);
+                slave.checkInter_rd5_6(md.pred[PRED_Nx2N], pmode.cuGeom, SIZE_Nx2N);
                 break;
 
             case PRED_2NxN:
-                slave.checkInter_rd5_6(md.pred[PRED_2NxN], pmode.cuGeom, SIZE_2NxN, false);
+                slave.checkInter_rd5_6(md.pred[PRED_2NxN], pmode.cuGeom, SIZE_2NxN);
                 break;
 
             case PRED_2NxnU:
-                slave.checkInter_rd5_6(md.pred[PRED_2NxnU], pmode.cuGeom, SIZE_2NxnU, bMergeOnly);
+                slave.checkInter_rd5_6(md.pred[PRED_2NxnU], pmode.cuGeom, SIZE_2NxnU);
                 break;
 
             case PRED_2NxnD:
-                slave.checkInter_rd5_6(md.pred[PRED_2NxnD], pmode.cuGeom, SIZE_2NxnD, bMergeOnly);
+                slave.checkInter_rd5_6(md.pred[PRED_2NxnD], pmode.cuGeom, SIZE_2NxnD);
                 break;
 
             case PRED_nLx2N:
-                slave.checkInter_rd5_6(md.pred[PRED_nLx2N], pmode.cuGeom, SIZE_nLx2N, bMergeOnly);
+                slave.checkInter_rd5_6(md.pred[PRED_nLx2N], pmode.cuGeom, SIZE_nLx2N);
                 break;
 
             case PRED_nRx2N:
-                slave.checkInter_rd5_6(md.pred[PRED_nRx2N], pmode.cuGeom, SIZE_nRx2N, bMergeOnly);
+                slave.checkInter_rd5_6(md.pred[PRED_nRx2N], pmode.cuGeom, SIZE_nRx2N);
                 break;
 
             default:
@@ -490,7 +490,7 @@
     while (task >= 0);
 }
 
-void Analysis::compressInterCU_dist(const CUData& parentCTU, const CUGeom& cuGeom)
+void Analysis::compressInterCU_dist(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp)
 {
     uint32_t depth = cuGeom.depth;
     uint32_t cuAddr = parentCTU.m_cuAddr;
@@ -505,34 +505,34 @@
 
     if (mightNotSplit && depth >= minDepth)
     {
-        int bTryAmp = m_slice->m_sps->maxAMPDepth > depth && (cuGeom.log2CUSize < 6 || m_param->rdLevel > 4);
+        int bTryAmp = m_slice->m_sps->maxAMPDepth > depth;
         int bTryIntra = m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames;
 
         PMODE pmode(*this, cuGeom);
 
         /* Initialize all prediction CUs based on parentCTU */
-        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom);
-        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom);
+        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
+        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
         if (bTryIntra)
         {
-            md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom);
+            md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
             if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3 && m_param->rdLevel >= 5)
-                md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom);
+                md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom, qp);
             pmode.modes[pmode.m_jobTotal++] = PRED_INTRA;
         }
-        md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom); pmode.modes[pmode.m_jobTotal++] = PRED_2Nx2N;
-        md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom);
+        md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_2Nx2N;
+        md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);
         if (m_param->bEnableRectInter)
         {
-            md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom); pmode.modes[pmode.m_jobTotal++] = PRED_2NxN;
-            md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom); pmode.modes[pmode.m_jobTotal++] = PRED_Nx2N;
+            md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_2NxN;
+            md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_Nx2N;
         }
         if (bTryAmp)
         {
-            md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom); pmode.modes[pmode.m_jobTotal++] = PRED_2NxnU;
-            md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom); pmode.modes[pmode.m_jobTotal++] = PRED_2NxnD;
-            md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom); pmode.modes[pmode.m_jobTotal++] = PRED_nLx2N;
-            md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom); pmode.modes[pmode.m_jobTotal++] = PRED_nRx2N;
+            md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_2NxnU;
+            md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_2NxnD;
+            md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_nLx2N;
+            md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_nRx2N;
         }
 
         pmode.tryBondPeers(*m_frame->m_encData->m_jobProvider, pmode.m_jobTotal);
@@ -662,7 +662,7 @@
 
         if (md.bestMode->rdCost == MAX_INT64 && !bTryIntra)
         {
-            md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom);
+            md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
             checkIntraInInter(md.pred[PRED_INTRA], cuGeom);
             encodeIntraInInter(md.pred[PRED_INTRA], cuGeom);
             checkBestMode(md.pred[PRED_INTRA], depth);
@@ -688,12 +688,13 @@
         Mode* splitPred = &md.pred[PRED_SPLIT];
         splitPred->initCosts();
         CUData* splitCU = &splitPred->cu;
-        splitCU->initSubCU(parentCTU, cuGeom);
+        splitCU->initSubCU(parentCTU, cuGeom, qp);
 
         uint32_t nextDepth = depth + 1;
         ModeDepth& nd = m_modeDepth[nextDepth];
         invalidateContexts(nextDepth);
         Entropy* nextContext = &m_rqt[depth].cur;
+        int nextQP = qp;
 
         for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
         {
@@ -702,7 +703,11 @@
             {
                 m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                 m_rqt[nextDepth].cur.load(*nextContext);
-                compressInterCU_dist(parentCTU, childGeom);
+
+                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
+                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));
+
+                compressInterCU_dist(parentCTU, childGeom, nextQP);
 
                 // Save best CU and pred data for this sub CU
                 splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
@@ -721,7 +726,7 @@
         else
             updateModeCost(*splitPred);
 
-        checkDQPForSplitPred(splitPred->cu, cuGeom);
+        checkDQPForSplitPred(*splitPred, cuGeom);
         checkBestMode(*splitPred, depth);
     }
 
@@ -741,7 +746,7 @@
         md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, cuAddr, cuGeom.absPartIdx);
 }
 
-void Analysis::compressInterCU_rd0_4(const CUData& parentCTU, const CUGeom& cuGeom)
+void Analysis::compressInterCU_rd0_4(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp)
 {
     uint32_t depth = cuGeom.depth;
     uint32_t cuAddr = parentCTU.m_cuAddr;
@@ -757,8 +762,8 @@
         bool bTryIntra = m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames;
 
         /* Compute Merge Cost */
-        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom);
-        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom);
+        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
+        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
         checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
 
         bool earlyskip = false;
@@ -767,30 +772,30 @@
 
         if (!earlyskip)
         {
-            md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom);
+            md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
             checkInter_rd0_4(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N);
 
             if (m_slice->m_sliceType == B_SLICE)
             {
-                md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom);
+                md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);
                 checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], cuGeom);
             }
 
             Mode *bestInter = &md.pred[PRED_2Nx2N];
             if (m_param->bEnableRectInter)
             {
-                md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom);
+                md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                 checkInter_rd0_4(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N);
                 if (md.pred[PRED_Nx2N].sa8dCost < bestInter->sa8dCost)
                     bestInter = &md.pred[PRED_Nx2N];
 
-                md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom);
+                md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);
                 checkInter_rd0_4(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN);
                 if (md.pred[PRED_2NxN].sa8dCost < bestInter->sa8dCost)
                     bestInter = &md.pred[PRED_2NxN];
             }
 
-            if (m_slice->m_sps->maxAMPDepth > depth && cuGeom.log2CUSize < 6)
+            if (m_slice->m_sps->maxAMPDepth > depth)
             {
                 bool bHor = false, bVer = false;
                 if (bestInter->cu.m_partSize[0] == SIZE_2NxN)
@@ -806,24 +811,24 @@
 
                 if (bHor)
                 {
-                    md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom);
+                    md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp);
                     checkInter_rd0_4(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU);
                     if (md.pred[PRED_2NxnU].sa8dCost < bestInter->sa8dCost)
                         bestInter = &md.pred[PRED_2NxnU];
 
-                    md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom);
+                    md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);
                     checkInter_rd0_4(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD);
                     if (md.pred[PRED_2NxnD].sa8dCost < bestInter->sa8dCost)
                         bestInter = &md.pred[PRED_2NxnD];
                 }
                 if (bVer)
                 {
-                    md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom);
+                    md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                     checkInter_rd0_4(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N);
                     if (md.pred[PRED_nLx2N].sa8dCost < bestInter->sa8dCost)
                         bestInter = &md.pred[PRED_nLx2N];
 
-                    md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom);
+                    md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                     checkInter_rd0_4(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N);
                     if (md.pred[PRED_nRx2N].sa8dCost < bestInter->sa8dCost)
                         bestInter = &md.pred[PRED_nRx2N];
@@ -855,7 +860,7 @@
                 if ((bTryIntra && md.bestMode->cu.getQtRootCbf(0)) ||
                     md.bestMode->sa8dCost == MAX_INT64)
                 {
-                    md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom);
+                    md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
                     checkIntraInInter(md.pred[PRED_INTRA], cuGeom);
                     encodeIntraInInter(md.pred[PRED_INTRA], cuGeom);
                     checkBestMode(md.pred[PRED_INTRA], depth);
@@ -873,7 +878,7 @@
 
                 if (bTryIntra || md.bestMode->sa8dCost == MAX_INT64)
                 {
-                    md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom);
+                    md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
                     checkIntraInInter(md.pred[PRED_INTRA], cuGeom);
                     if (md.pred[PRED_INTRA].sa8dCost < md.bestMode->sa8dCost)
                         md.bestMode = &md.pred[PRED_INTRA];
@@ -901,7 +906,6 @@
                     {
                         /* generate recon pixels with no rate distortion considerations */
                         CUData& cu = md.bestMode->cu;
-                        m_quant.setQPforQuant(cu);
 
                         uint32_t tuDepthRange[2];
                         cu.getInterTUQtDepthRange(tuDepthRange, 0);
@@ -926,7 +930,6 @@
                     {
                         /* generate recon pixels with no rate distortion considerations */
                         CUData& cu = md.bestMode->cu;
-                        m_quant.setQPforQuant(cu);
 
                         uint32_t tuDepthRange[2];
                         cu.getIntraTUQtDepthRange(tuDepthRange, 0);
@@ -960,12 +963,13 @@
         Mode* splitPred = &md.pred[PRED_SPLIT];
         splitPred->initCosts();
         CUData* splitCU = &splitPred->cu;
-        splitCU->initSubCU(parentCTU, cuGeom);
+        splitCU->initSubCU(parentCTU, cuGeom, qp);
 
         uint32_t nextDepth = depth + 1;
         ModeDepth& nd = m_modeDepth[nextDepth];
         invalidateContexts(nextDepth);
         Entropy* nextContext = &m_rqt[depth].cur;
+        int nextQP = qp;
 
         for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
         {
@@ -974,7 +978,11 @@
             {
                 m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                 m_rqt[nextDepth].cur.load(*nextContext);
-                compressInterCU_rd0_4(parentCTU, childGeom);
+
+                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
+                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));
+
+                compressInterCU_rd0_4(parentCTU, childGeom, nextQP);
 
                 // Save best CU and pred data for this sub CU
                 splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
@@ -1006,7 +1014,7 @@
         else if (splitPred->sa8dCost < md.bestMode->sa8dCost)
             md.bestMode = splitPred;
 
-        checkDQPForSplitPred(md.bestMode->cu, cuGeom);
+        checkDQPForSplitPred(*md.bestMode, cuGeom);
     }
     if (mightNotSplit)
     {
@@ -1025,7 +1033,7 @@
         md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, cuAddr, cuGeom.absPartIdx);
 }
 
-void Analysis::compressInterCU_rd5_6(const CUData& parentCTU, const CUGeom& cuGeom, uint32_t &zOrder)
+void Analysis::compressInterCU_rd5_6(const CUData& parentCTU, const CUGeom& cuGeom, uint32_t &zOrder, int32_t qp)
 {
     uint32_t depth = cuGeom.depth;
     ModeDepth& md = m_modeDepth[depth];
@@ -1040,8 +1048,8 @@
         uint8_t* reuseModes  = &m_reuseInterDataCTU->modes[parentCTU.m_cuAddr * parentCTU.m_numPartitions];
         if (mightNotSplit && depth == reuseDepth[zOrder] && zOrder == cuGeom.absPartIdx && reuseModes[zOrder] == MODE_SKIP)
         {
-            md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom);
-            md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom);
+            md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
+            md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
             checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom, true);
 
             if (m_bTryLossless)
@@ -1060,20 +1068,20 @@
 
     if (mightNotSplit)
     {
-        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom);
-        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom);
+        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
+        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
         checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom, false);
         bool earlySkip = m_param->bEnableEarlySkip && md.bestMode && !md.bestMode->cu.getQtRootCbf(0);
 
         if (!earlySkip)
         {
-            md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom);
-            checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, false);
+            md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
+            checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N);
             checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);
 
             if (m_slice->m_sliceType == B_SLICE)
             {
-                md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom);
+                md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);
                 checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], cuGeom);
                 if (md.pred[PRED_BIDIR].sa8dCost < MAX_INT64)
                 {
@@ -1084,20 +1092,18 @@
 
             if (m_param->bEnableRectInter)
             {
-                md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom);
-                checkInter_rd5_6(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N, false);
+                md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
+                checkInter_rd5_6(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N);
                 checkBestMode(md.pred[PRED_Nx2N], cuGeom.depth);
 
-                md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom);
-                checkInter_rd5_6(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, false);
+                md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);
+                checkInter_rd5_6(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN);
                 checkBestMode(md.pred[PRED_2NxN], cuGeom.depth);
             }
 
             // Try AMP (SIZE_2NxnU, SIZE_2NxnD, SIZE_nLx2N, SIZE_nRx2N)
             if (m_slice->m_sps->maxAMPDepth > depth)
             {
-                bool bMergeOnly = cuGeom.log2CUSize == 6;
-
                 bool bHor = false, bVer = false;
                 if (md.bestMode->cu.m_partSize[0] == SIZE_2NxN)
                     bHor = true;
@@ -1111,35 +1117,35 @@
 
                 if (bHor)
                 {
-                    md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom);
-                    checkInter_rd5_6(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU, bMergeOnly);
+                    md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp);
+                    checkInter_rd5_6(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU);
                     checkBestMode(md.pred[PRED_2NxnU], cuGeom.depth);
 
-                    md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom);
-                    checkInter_rd5_6(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, bMergeOnly);
+                    md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);
+                    checkInter_rd5_6(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD);
                     checkBestMode(md.pred[PRED_2NxnD], cuGeom.depth);
                 }
                 if (bVer)
                 {
-                    md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom);
-                    checkInter_rd5_6(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N, bMergeOnly);
+                    md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp);
+                    checkInter_rd5_6(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N);
                     checkBestMode(md.pred[PRED_nLx2N], cuGeom.depth);
 
-                    md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom);
-                    checkInter_rd5_6(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, bMergeOnly);
+                    md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);
+                    checkInter_rd5_6(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N);
                     checkBestMode(md.pred[PRED_nRx2N], cuGeom.depth);
                 }
             }
 
             if (m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames)
             {
-                md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom);
+                md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
                 checkIntra(md.pred[PRED_INTRA], cuGeom, SIZE_2Nx2N, NULL, NULL);
                 checkBestMode(md.pred[PRED_INTRA], depth);
 
                 if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3)
                 {
-                    md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom);
+                    md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom, qp);
                     checkIntra(md.pred[PRED_INTRA_NxN], cuGeom, SIZE_NxN, NULL, NULL);
                     checkBestMode(md.pred[PRED_INTRA_NxN], depth);
                 }
@@ -1159,12 +1165,13 @@
         Mode* splitPred = &md.pred[PRED_SPLIT];
         splitPred->initCosts();
         CUData* splitCU = &splitPred->cu;
-        splitCU->initSubCU(parentCTU, cuGeom);
+        splitCU->initSubCU(parentCTU, cuGeom, qp);
 
         uint32_t nextDepth = depth + 1;
         ModeDepth& nd = m_modeDepth[nextDepth];
         invalidateContexts(nextDepth);
         Entropy* nextContext = &m_rqt[depth].cur;
+        int nextQP = qp;
 
         for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
         {
@@ -1173,7 +1180,11 @@
             {
                 m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                 m_rqt[nextDepth].cur.load(*nextContext);
-                compressInterCU_rd5_6(parentCTU, childGeom, zOrder);
+
+                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
+                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));
+
+                compressInterCU_rd5_6(parentCTU, childGeom, zOrder, nextQP);
 
                 // Save best CU and pred data for this sub CU
                 splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
@@ -1193,7 +1204,7 @@
         else
             updateModeCost(*splitPred);
 
-        checkDQPForSplitPred(splitPred->cu, cuGeom);
+        checkDQPForSplitPred(*splitPred, cuGeom);
         checkBestMode(*splitPred, depth);
     }
 
@@ -1308,7 +1319,7 @@
     md.bestMode->cu.setPUMv(1, candMvField[bestSadCand][1].mv, 0, 0);
     md.bestMode->cu.setPURefIdx(0, (int8_t)candMvField[bestSadCand][0].refIdx, 0, 0);
     md.bestMode->cu.setPURefIdx(1, (int8_t)candMvField[bestSadCand][1].refIdx, 0, 0);
-    checkDQP(md.bestMode->cu, cuGeom);
+    checkDQP(*md.bestMode, cuGeom);
     X265_CHECK(md.bestMode->ok(), "Merge mode not ok\n");
 }
 
@@ -1440,7 +1451,7 @@
         bestPred->cu.setPUMv(1, candMvField[bestCand][1].mv, 0, 0);
         bestPred->cu.setPURefIdx(0, (int8_t)candMvField[bestCand][0].refIdx, 0, 0);
         bestPred->cu.setPURefIdx(1, (int8_t)candMvField[bestCand][1].refIdx, 0, 0);
-        checkDQP(bestPred->cu, cuGeom);
+        checkDQP(*bestPred, cuGeom);
         X265_CHECK(bestPred->ok(), "merge mode is not ok");
     }
 
@@ -1472,7 +1483,7 @@
         }
     }
 
-    predInterSearch(interMode, cuGeom, false, m_bChromaSa8d);
+    predInterSearch(interMode, cuGeom, m_bChromaSa8d);
 
     /* predInterSearch sets interMode.sa8dBits */
     const Yuv& fencYuv = *interMode.fencYuv;
@@ -1500,7 +1511,7 @@
     }
 }
 
-void Analysis::checkInter_rd5_6(Mode& interMode, const CUGeom& cuGeom, PartSize partSize, bool bMergeOnly)
+void Analysis::checkInter_rd5_6(Mode& interMode, const CUGeom& cuGeom, PartSize partSize)
 {
     interMode.initCosts();
     interMode.cu.setPartSizeSubParts(partSize);
@@ -1520,7 +1531,7 @@
         }
     }
 
-    predInterSearch(interMode, cuGeom, bMergeOnly, true);
+    predInterSearch(interMode, cuGeom, true);
 
     /* predInterSearch sets interMode.sa8dBits, but this is ignored */
     encodeResAndCalcRdInterCU(interMode, cuGeom);
@@ -1642,8 +1653,8 @@
         uint32_t zcost = zsa8d + m_rdCost.getCost(bits0) + m_rdCost.getCost(bits1);
 
         /* refine MVP selection for zero mv, updates: mvp, mvpidx, bits, cost */
-        checkBestMVP(inter2Nx2N.amvpCand[0][ref0], mvzero, mvp0, mvpIdx0, bits0, zcost);
-        checkBestMVP(inter2Nx2N.amvpCand[1][ref1], mvzero, mvp1, mvpIdx1, bits1, zcost);
+        mvp0 = checkBestMVP(inter2Nx2N.amvpCand[0][ref0], mvzero, mvpIdx0, bits0, zcost);
+        mvp1 = checkBestMVP(inter2Nx2N.amvpCand[1][ref1], mvzero, mvpIdx1, bits1, zcost);
 
         uint32_t zbits = bits0 + bits1 + m_listSelBits[2] - (m_listSelBits[0] + m_listSelBits[1]);
         zcost = zsa8d + m_rdCost.getCost(zbits);
@@ -1697,7 +1708,6 @@
     CUData& cu = bestMode->cu;
 
     cu.copyFromPic(ctu, cuGeom);
-    m_quant.setQPforQuant(cu);
 
     Yuv& fencYuv = m_modeDepth[cuGeom.depth].fencYuv;
     if (cuGeom.depth)
@@ -1913,37 +1923,39 @@
     return false;
 }
 
-int Analysis::calculateQpforCuSize(CUData& ctu, const CUGeom& cuGeom)
+int Analysis::calculateQpforCuSize(const CUData& ctu, const CUGeom& cuGeom)
 {
-    uint32_t ctuAddr = ctu.m_cuAddr;
     FrameData& curEncData = *m_frame->m_encData;
-    double qp = curEncData.m_cuStat[ctuAddr].baseQp;
-
-    uint32_t width = m_frame->m_fencPic->m_picWidth;
-    uint32_t height = m_frame->m_fencPic->m_picHeight;
-    uint32_t block_x = ctu.m_cuPelX + g_zscanToPelX[cuGeom.absPartIdx];
-    uint32_t block_y = ctu.m_cuPelY + g_zscanToPelY[cuGeom.absPartIdx];
-    uint32_t maxCols = (m_frame->m_fencPic->m_picWidth + (16 - 1)) / 16;
-    uint32_t blockSize = g_maxCUSize >> cuGeom.depth;
-    double qp_offset = 0;
-    uint32_t cnt = 0;
-    uint32_t idx;
+    double qp = curEncData.m_cuStat[ctu.m_cuAddr].baseQp;
 
     /* Use cuTree offsets if cuTree enabled and frame is referenced, else use AQ offsets */
     bool isReferenced = IS_REFERENCED(m_frame);
     double *qpoffs = (isReferenced && m_param->rc.cuTree) ? m_frame->m_lowres.qpCuTreeOffset : m_frame->m_lowres.qpAqOffset;
-
-    for (uint32_t block_yy = block_y; block_yy < block_y + blockSize && block_yy < height; block_yy += 16)
+    if (qpoffs)
     {
-        for (uint32_t block_xx = block_x; block_xx < block_x + blockSize && block_xx < width; block_xx += 16)
+        uint32_t width = m_frame->m_fencPic->m_picWidth;
+        uint32_t height = m_frame->m_fencPic->m_picHeight;
+        uint32_t block_x = ctu.m_cuPelX + g_zscanToPelX[cuGeom.absPartIdx];
+        uint32_t block_y = ctu.m_cuPelY + g_zscanToPelY[cuGeom.absPartIdx];
+        uint32_t maxCols = (m_frame->m_fencPic->m_picWidth + (16 - 1)) / 16;
+        uint32_t blockSize = g_maxCUSize >> cuGeom.depth;
+        double qp_offset = 0;
+        uint32_t cnt = 0;
+        uint32_t idx;
+
+        for (uint32_t block_yy = block_y; block_yy < block_y + blockSize && block_yy < height; block_yy += 16)
         {
-            idx = ((block_yy / 16) * (maxCols)) + (block_xx / 16);
-            qp_offset += qpoffs[idx];
-            cnt++;
+            for (uint32_t block_xx = block_x; block_xx < block_x + blockSize && block_xx < width; block_xx += 16)
+            {
+                idx = ((block_yy / 16) * (maxCols)) + (block_xx / 16);
+                qp_offset += qpoffs[idx];
+                cnt++;
+            }
         }
+
+        qp_offset /= cnt;
+        qp += qp_offset;
     }
 
-    qp_offset /= cnt;
-    qp += qp_offset;
     return x265_clip3(QP_MIN, QP_MAX_MAX, (int)(qp + 0.5));
 }

x265_1.6.tar.gz/source/encoder/analysis.h -> x265_1.7.tar.gz/source/encoder/analysis.h Changed

@@ -109,12 +109,12 @@
     uint32_t*            m_reuseBestMergeCand;
 
     /* full analysis for an I-slice CU */
-    void compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, uint32_t &zOrder);
+    void compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, uint32_t &zOrder, int32_t qp);
 
     /* full analysis for a P or B slice CU */
-    void compressInterCU_dist(const CUData& parentCTU, const CUGeom& cuGeom);
-    void compressInterCU_rd0_4(const CUData& parentCTU, const CUGeom& cuGeom);
-    void compressInterCU_rd5_6(const CUData& parentCTU, const CUGeom& cuGeom, uint32_t &zOrder);
+    void compressInterCU_dist(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp);
+    void compressInterCU_rd0_4(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp);
+    void compressInterCU_rd5_6(const CUData& parentCTU, const CUGeom& cuGeom, uint32_t &zOrder, int32_t qp);
 
     /* measure merge and skip */
     void checkMerge2Nx2N_rd0_4(Mode& skip, Mode& merge, const CUGeom& cuGeom);
@@ -122,7 +122,7 @@
 
     /* measure inter options */
     void checkInter_rd0_4(Mode& interMode, const CUGeom& cuGeom, PartSize partSize);
-    void checkInter_rd5_6(Mode& interMode, const CUGeom& cuGeom, PartSize partSize, bool bMergeOnly);
+    void checkInter_rd5_6(Mode& interMode, const CUGeom& cuGeom, PartSize partSize);
 
     void checkBidir2Nx2N(Mode& inter2Nx2N, Mode& bidir2Nx2N, const CUGeom& cuGeom);
 
@@ -139,7 +139,7 @@
     /* generate residual and recon pixels for an entire CTU recursively (RD0) */
     void encodeResidue(const CUData& parentCTU, const CUGeom& cuGeom);
 
-    int calculateQpforCuSize(CUData& ctu, const CUGeom& cuGeom);
+    int calculateQpforCuSize(const CUData& ctu, const CUGeom& cuGeom);
 
     /* check whether current mode is the new best */
     inline void checkBestMode(Mode& mode, uint32_t depth)

x265_1.6.tar.gz/source/encoder/api.cpp -> x265_1.7.tar.gz/source/encoder/api.cpp Changed

@@ -39,9 +39,11 @@
     if (!p)
         return NULL;
 
-    x265_param *param = X265_MALLOC(x265_param, 1);
-    if (!param)
-        return NULL;
+    Encoder* encoder = NULL;
+    x265_param* param = x265_param_alloc();
+    x265_param* latestParam = x265_param_alloc();
+    if (!param || !latestParam)
+        goto fail;
 
     memcpy(param, p, sizeof(x265_param));
     x265_log(param, X265_LOG_INFO, "HEVC encoder version %s\n", x265_version_str);
@@ -50,38 +52,44 @@
     x265_setup_primitives(param, param->cpuid);
 
     if (x265_check_params(param))
-        return NULL;
+        goto fail;
 
     if (x265_set_globals(param))
-        return NULL;
+        goto fail;
 
-    Encoder *encoder = new Encoder;
+    encoder = new Encoder;
     if (!param->rc.bEnableSlowFirstPass)
         x265_param_apply_fastfirstpass(param);
 
     // may change params for auto-detect, etc
     encoder->configure(param);
-    
     // may change rate control and CPB params
     if (!enforceLevel(*param, encoder->m_vps))
-    {
-        delete encoder;
-        return NULL;
-    }
+        goto fail;
 
     // will detect and set profile/tier/level in VPS
     determineLevel(*param, encoder->m_vps);
 
-    encoder->create();
-    if (encoder->m_aborted)
+    if (!param->bAllowNonConformance && encoder->m_vps.ptl.profileIdc == Profile::NONE)
     {
-        delete encoder;
-        return NULL;
+        x265_log(param, X265_LOG_INFO, "non-conformant bitstreams not allowed (--allow-non-conformance)\n");
+        goto fail;
     }
 
-    x265_print_params(param);
+    encoder->create();
+    encoder->m_latestParam = latestParam;
+    memcpy(latestParam, param, sizeof(x265_param));
+    if (encoder->m_aborted)
+        goto fail;
 
+    x265_print_params(param);
     return encoder;
+
+fail:
+    delete encoder;
+    x265_param_free(param);
+    x265_param_free(latestParam);
+    return NULL;
 }
 
 extern "C"
@@ -112,6 +120,27 @@
 }
 
 extern "C"
+int x265_encoder_reconfig(x265_encoder* enc, x265_param* param_in)
+{
+    if (!enc || !param_in)
+        return -1;
+
+    x265_param save;
+    Encoder* encoder = static_cast<Encoder*>(enc);
+    memcpy(&save, encoder->m_latestParam, sizeof(x265_param));
+    int ret = encoder->reconfigureParam(encoder->m_latestParam, param_in);
+    if (ret)
+        /* reconfigure failed, recover saved param set */
+        memcpy(encoder->m_latestParam, &save, sizeof(x265_param));
+    else
+    {
+        encoder->m_reconfigured = true;
+        x265_print_reconfigured_params(&save, encoder->m_latestParam);
+    }
+    return ret;
+}
+
+extern "C"
 int x265_encoder_encode(x265_encoder *enc, x265_nal **pp_nal, uint32_t *pi_nal, x265_picture *pic_in, x265_picture *pic_out)
 {
     if (!enc)
@@ -173,19 +202,22 @@
     {
         Encoder *encoder = static_cast<Encoder*>(enc);
 
-        encoder->stop();
+        encoder->stopJobs();
         encoder->printSummary();
         encoder->destroy();
         delete encoder;
+        ATOMIC_DEC(&g_ctuSizeConfigured);
     }
 }
 
 extern "C"
 void x265_cleanup(void)
 {
-    BitCost::destroy();
-    CUData::s_partSet[0] = NULL; /* allow CUData to adjust to new CTU size */
-    g_ctuSizeConfigured = 0;
+    if (!g_ctuSizeConfigured)
+    {
+        BitCost::destroy();
+        CUData::s_partSet[0] = NULL; /* allow CUData to adjust to new CTU size */
+    }
 }
 
 extern "C"
@@ -232,6 +264,7 @@
     &x265_picture_init,
     &x265_encoder_open,
     &x265_encoder_parameters,
+    &x265_encoder_reconfig,
     &x265_encoder_headers,
     &x265_encoder_encode,
     &x265_encoder_get_stats,
@@ -243,11 +276,66 @@
     x265_max_bit_depth,
 };
 
+typedef const x265_api* (*api_get_func)(int bitDepth);
+
+#define xstr(s) str(s)
+#define str(s) #s
+
+#if _WIN32
+#define ext ".dll"
+#elif MACOS
+#include <dlfcn.h>
+#define ext ".dylib"
+#else
+#include <dlfcn.h>
+#define ext ".so"
+#endif
+
 extern "C"
 const x265_api* x265_api_get(int bitDepth)
 {
     if (bitDepth && bitDepth != X265_DEPTH)
-        return NULL;
+    {
+        const char* libname = NULL;
+        const char* method = "x265_api_get_" xstr(X265_BUILD);
+
+        if (bitDepth == 12)
+            libname = "libx265_main12" ext;
+        else if (bitDepth == 10)
+            libname = "libx265_main10" ext;
+        else if (bitDepth == 8)
+            libname = "libx265_main" ext;
+        else
+            return NULL;
+
+        const x265_api* api = NULL;
+
+#if _WIN32
+        HMODULE h = LoadLibraryA(libname);
+        if (h)
+        {
+            api_get_func get = (api_get_func)GetProcAddress(h, method);
+            if (get)
+                api = get(0);
+        }
+#else
+        void* h = dlopen(libname, RTLD_LAZY | RTLD_LOCAL);
+        if (h)
+        {
+            api_get_func get = (api_get_func)dlsym(h, method);
+            if (get)
+                api = get(0);
+        }
+#endif
+
+        if (api && bitDepth != api->max_bit_depth)
+        {
+            x265_log(NULL, X265_LOG_WARNING, "%s does not support requested bitDepth %d\n", libname, bitDepth);
+            return NULL;
+        }
+
+        return api;
+    }
 
     return &libapi;
 }

x265_1.6.tar.gz/source/encoder/encoder.cpp -> x265_1.7.tar.gz/source/encoder/encoder.cpp Changed

@@ -58,6 +58,7 @@
 Encoder::Encoder()
 {
     m_aborted = false;
+    m_reconfigured = false;
     m_encodedFrameNum = 0;
     m_pocLast = -1;
     m_curEncoder = 0;
@@ -73,6 +74,7 @@
     m_outputCount = 0;
     m_csvfpt = NULL;
     m_param = NULL;
+    m_latestParam = NULL;
     m_cuOffsetY = NULL;
     m_cuOffsetC = NULL;
     m_buOffsetY = NULL;
@@ -106,7 +108,7 @@
     bool allowPools = !p->numaPools || strcmp(p->numaPools, "none");
 
     // Trim the thread pool if --wpp, --pme, and --pmode are disabled
-    if (!p->bEnableWavefront && !p->bDistributeModeAnalysis && !p->bDistributeMotionEstimation)
+    if (!p->bEnableWavefront && !p->bDistributeModeAnalysis && !p->bDistributeMotionEstimation && !p->lookaheadSlices)
         allowPools = false;
 
     if (!p->frameNumThreads)
@@ -140,9 +142,11 @@
             x265_log(p, X265_LOG_WARNING, "No thread pool allocated, --pme disabled\n");
         if (p->bDistributeModeAnalysis)
             x265_log(p, X265_LOG_WARNING, "No thread pool allocated, --pmode disabled\n");
+        if (p->lookaheadSlices)
+            x265_log(p, X265_LOG_WARNING, "No thread pool allocated, --lookahead-slices disabled\n");
 
         // disable all pool features if the thread pool is disabled or unusable.
-        p->bEnableWavefront = p->bDistributeModeAnalysis = p->bDistributeMotionEstimation = 0;
+        p->bEnableWavefront = p->bDistributeModeAnalysis = p->bDistributeMotionEstimation = p->lookaheadSlices = 0;
     }
 
     char buf[128];
@@ -159,7 +163,10 @@
     x265_log(p, X265_LOG_INFO, "frame threads / pool features       : %d / %s\n", p->frameNumThreads, buf);
 
     for (int i = 0; i < m_param->frameNumThreads; i++)
+    {
         m_frameEncoder[i] = new FrameEncoder;
+        m_frameEncoder[i]->m_nalList.m_annexB = !!m_param->bAnnexB;
+    }
 
     if (m_numPools)
     {
@@ -287,15 +294,17 @@
     m_aborted |= parseLambdaFile(m_param);
 
     m_encodeStartTime = x265_mdate();
+
+    m_nalList.m_annexB = !!m_param->bAnnexB;
 }
 
-void Encoder::stop()
+void Encoder::stopJobs()
 {
     if (m_rateControl)
         m_rateControl->terminate(); // unblock all blocked RC calls
 
     if (m_lookahead)
-        m_lookahead->stop();
+        m_lookahead->stopJobs();
     
     for (int i = 0; i < m_param->frameNumThreads; i++)
     {
@@ -309,7 +318,7 @@
     }
 
     if (m_threadPool)
-        m_threadPool->stop();
+        m_threadPool->stopWorkers();
 }
 
 void Encoder::destroy()
@@ -358,15 +367,20 @@
 
     if (m_param)
     {
-        free((void*)m_param->rc.lambdaFileName); // allocs by strdup
-        free(m_param->rc.statFileName);
-        free(m_param->analysisFileName);
-        free((void*)m_param->scalingLists);
-        free(m_param->csvfn);
-        free(m_param->numaPools);
+        /* release string arguments that were strdup'd */
+        free((char*)m_param->rc.lambdaFileName);
+        free((char*)m_param->rc.statFileName);
+        free((char*)m_param->analysisFileName);
+        free((char*)m_param->scalingLists);
+        free((char*)m_param->csvfn);
+        free((char*)m_param->numaPools);
+        free((char*)m_param->masteringDisplayColorVolume);
+        free((char*)m_param->contentLightLevelInfo);
 
-        X265_FREE(m_param);
+        x265_param_free(m_param);
     }
+
+    x265_param_free(m_latestParam);
 }
 
 void Encoder::updateVbvPlan(RateControl* rc)
@@ -436,7 +450,8 @@
         if (m_dpb->m_freeList.empty())
         {
             inFrame = new Frame;
-            if (inFrame->create(m_param))
+            x265_param* p = m_reconfigured? m_latestParam : m_param;
+            if (inFrame->create(p))
             {
                 /* the first PicYuv created is asked to generate the CU and block unit offset
                  * arrays which are then shared with all subsequent PicYuv (orig and recon) 
@@ -477,7 +492,10 @@
             }
         }
         else
+        {
             inFrame = m_dpb->m_freeList.popBack();
+            inFrame->m_lowresInit = false;
+        }
 
         /* Copy input picture into a Frame and PicYuv, send to lookahead */
         inFrame->m_fencPic->copyFromPicture(*pic_in, m_sps.conformanceWindow.rightOffset, m_sps.conformanceWindow.bottomOffset);
@@ -486,6 +504,7 @@
         inFrame->m_userData  = pic_in->userData;
         inFrame->m_pts       = pic_in->pts;
         inFrame->m_forceqp   = pic_in->forceqp;
+        inFrame->m_param     = m_reconfigured ? m_latestParam : m_param;
 
         if (m_pocLast == 0)
             m_firstPts = inFrame->m_pts;
@@ -717,6 +736,34 @@
     return ret;
 }
 
+int Encoder::reconfigureParam(x265_param* encParam, x265_param* param)
+{
+    encParam->maxNumReferences = param->maxNumReferences; // never uses more refs than specified in stream headers
+    encParam->bEnableLoopFilter = param->bEnableLoopFilter;
+    encParam->deblockingFilterTCOffset = param->deblockingFilterTCOffset;
+    encParam->deblockingFilterBetaOffset = param->deblockingFilterBetaOffset; 
+    encParam->bEnableFastIntra = param->bEnableFastIntra;
+    encParam->bEnableEarlySkip = param->bEnableEarlySkip;
+    encParam->bEnableTemporalMvp = param->bEnableTemporalMvp;
+    /* Scratch buffer prevents me_range from being increased for esa/tesa
+    if (param->searchMethod < X265_FULL_SEARCH || param->searchMethod < encParam->searchRange)
+        encParam->searchRange = param->searchRange; */
+    encParam->noiseReductionInter = param->noiseReductionInter;
+    encParam->noiseReductionIntra = param->noiseReductionIntra;
+    /* We can't switch out of subme=0 during encoding. */
+    if (encParam->subpelRefine)
+        encParam->subpelRefine = param->subpelRefine;
+    encParam->rdoqLevel = param->rdoqLevel;
+    encParam->rdLevel = param->rdLevel;
+    encParam->bEnableTSkipFast = param->bEnableTSkipFast;
+    encParam->psyRd = param->psyRd;
+    encParam->psyRdoq = param->psyRdoq;
+    encParam->bEnableSignHiding = param->bEnableSignHiding;
+    encParam->bEnableFastIntra = param->bEnableFastIntra;
+    encParam->maxTUSize = param->maxTUSize;
+    return x265_check_params(encParam);
+}
+
 void EncStats::addPsnr(double psnrY, double psnrU, double psnrV)
 {
     m_psnrSumY += psnrY;
@@ -1430,6 +1477,34 @@
     bs.writeByteAlignment();
     list.serialize(NAL_UNIT_PPS, bs);
 
+    if (m_param->masteringDisplayColorVolume)
+    {
+        SEIMasteringDisplayColorVolume mdsei;
+        if (mdsei.parse(m_param->masteringDisplayColorVolume))
+        {
+            bs.resetBits();
+            mdsei.write(bs, m_sps);
+            bs.writeByteAlignment();
+            list.serialize(NAL_UNIT_PREFIX_SEI, bs);
+        }
+        else
+            x265_log(m_param, X265_LOG_WARNING, "unable to parse mastering display color volume info\n");
+    }
+
+    if (m_param->contentLightLevelInfo)
+    {
+        SEIContentLightLevel cllsei;
+        if (cllsei.parse(m_param->contentLightLevelInfo))
+        {
+            bs.resetBits();
+            cllsei.write(bs, m_sps);
+            bs.writeByteAlignment();
+            list.serialize(NAL_UNIT_PREFIX_SEI, bs);
+        }
+        else
+            x265_log(m_param, X265_LOG_WARNING, "unable to parse content light level info\n");
+    }
+
     if (m_param->bEmitInfoSEI)
     {
         char *opts = x265_param2string(m_param);
@@ -1559,7 +1634,8 @@
     if (!m_param->bLossless && (m_param->rc.aqMode || bIsVbv))
     {
         pps->bUseDQP = true;
-        pps->maxCuDQPDepth = 0; /* TODO: make configurable? */
+        pps->maxCuDQPDepth = g_log2Size[m_param->maxCUSize] - g_log2Size[m_param->rc.qgSize];
+        X265_CHECK(pps->maxCuDQPDepth <= 2, "max CU DQP depth cannot be greater than 2\n");
     }
     else
     {
@@ -1788,6 +1864,23 @@
         p->analysisMode = X265_ANALYSIS_OFF;
         x265_log(p, X265_LOG_WARNING, "Analysis save and load mode not supported for distributed mode analysis\n");
     }
+
+    bool bIsVbv = m_param->rc.vbvBufferSize > 0 && m_param->rc.vbvMaxBitrate > 0;
+    if (!m_param->bLossless && (m_param->rc.aqMode || bIsVbv))
+    {
+        if (p->rc.qgSize < X265_MAX(16, p->minCUSize))
+        {
+            p->rc.qgSize = X265_MAX(16, p->minCUSize);
+            x265_log(p, X265_LOG_WARNING, "QGSize should be greater than or equal to 16 and minCUSize, setting QGSize = %d\n", p->rc.qgSize);
+        }
+        if (p->rc.qgSize > p->maxCUSize)
+        {
+            p->rc.qgSize = p->maxCUSize;
+            x265_log(p, X265_LOG_WARNING, "QGSize should be less than or equal to maxCUSize, setting QGSize = %d\n", p->rc.qgSize);
+        }
+    }
+    else
+        m_param->rc.qgSize = p->maxCUSize;
 }
 
 void Encoder::allocAnalysis(x265_analysis_data* analysis)

x265_1.6.tar.gz/source/encoder/encoder.h -> x265_1.7.tar.gz/source/encoder/encoder.h Changed

x265_1.6.tar.gz/source/encoder/entropy.cpp -> x265_1.7.tar.gz/source/encoder/entropy.cpp Changed

@@ -585,7 +585,7 @@
         if (ctu.isSkipped(absPartIdx))
         {
             codeMergeIndex(ctu, absPartIdx);
-            finishCU(ctu, absPartIdx, depth);
+            finishCU(ctu, absPartIdx, depth, bEncodeDQP);
             return;
         }
         codePredMode(ctu.m_predMode[absPartIdx]);
@@ -606,7 +606,7 @@
     codeCoeff(ctu, absPartIdx, bEncodeDQP, tuDepthRange);
 
     // --- write terminating bit ---
-    finishCU(ctu, absPartIdx, depth);
+    finishCU(ctu, absPartIdx, depth, bEncodeDQP);
 }
 
 /* Return bit count of signaling inter mode */
@@ -658,7 +658,7 @@
 }
 
 /* finish encoding a cu and handle end-of-slice conditions */
-void Entropy::finishCU(const CUData& ctu, uint32_t absPartIdx, uint32_t depth)
+void Entropy::finishCU(const CUData& ctu, uint32_t absPartIdx, uint32_t depth, bool bCodeDQP)
 {
     const Slice* slice = ctu.m_slice;
     uint32_t realEndAddress = slice->m_endCUAddr;
@@ -672,6 +672,9 @@
     bool granularityBoundary = (((rpelx & granularityMask) == 0 || (rpelx == slice->m_sps->picWidthInLumaSamples )) &&
                                 ((bpely & granularityMask) == 0 || (bpely == slice->m_sps->picHeightInLumaSamples)));
 
+    if (slice->m_pps->bUseDQP)
+        const_cast<CUData&>(ctu).setQPSubParts(bCodeDQP ? ctu.getRefQP(absPartIdx) : ctu.m_qp[absPartIdx], absPartIdx, depth);
+
     if (granularityBoundary)
     {
         // Encode slice finish
@@ -1141,11 +1144,11 @@
     {
         length = 0;
         codeNumber = (codeNumber >> absGoRice) - COEF_REMAIN_BIN_REDUCTION;
-        if (codeNumber != 0)
         {
             unsigned long idx;
             CLZ(idx, codeNumber + 1);
             length = idx;
+            X265_CHECK((codeNumber != 0) || (length == 0), "length check failure\n");
             codeNumber -= (1 << idx) - 1;
         }
         codeNumber = (codeNumber << absGoRice) + codeRemain;
@@ -1461,7 +1464,7 @@
     //const uint32_t maskPosXY = ((uint32_t)~0 >> (31 - log2TrSize + MLS_CG_LOG2_SIZE)) >> 1;
     X265_CHECK((uint32_t)((1 << (log2TrSize - MLS_CG_LOG2_SIZE)) - 1) == (((uint32_t)~0 >> (31 - log2TrSize + MLS_CG_LOG2_SIZE)) >> 1), "maskPosXY fault\n");
 
-    scanPosLast = primitives.findPosLast(codingParameters.scan, coeff, coeffSign, coeffFlag, coeffNum, numSig);
+    scanPosLast = primitives.scanPosLast(codingParameters.scan, coeff, coeffSign, coeffFlag, coeffNum, numSig, g_scan4x4[codingParameters.scanType], trSize);
     posLast = codingParameters.scan[scanPosLast];
 
     const int lastScanSet = scanPosLast >> MLS_CG_SIZE;
@@ -1515,7 +1518,6 @@
     uint8_t * const baseCoeffGroupCtx = &m_contextState[OFF_SIG_CG_FLAG_CTX + (bIsLuma ? 0 : NUM_SIG_CG_FLAG_CTX)];
     uint8_t * const baseCtx = bIsLuma ? &m_contextState[OFF_SIG_FLAG_CTX] : &m_contextState[OFF_SIG_FLAG_CTX + NUM_SIG_FLAG_CTX_LUMA];
     uint32_t c1 = 1;
-    uint32_t goRiceParam = 0;
     int scanPosSigOff = scanPosLast - (lastScanSet << MLS_CG_SIZE) - 1;
     int absCoeff[1 << MLS_CG_SIZE];
     int numNonZero = 1;
@@ -1529,7 +1531,6 @@
         const uint32_t subCoeffFlag = coeffFlag[subSet];
         uint32_t scanFlagMask = subCoeffFlag;
         int subPosBase = subSet << MLS_CG_SIZE;
-        goRiceParam    = 0;
         
         if (subSet == lastScanSet)
         {
@@ -1548,7 +1549,7 @@
         else
         {
             uint32_t sigCoeffGroup = ((sigCoeffGroupFlag64 & cgBlkPosMask) != 0);
-            uint32_t ctxSig = Quant::getSigCoeffGroupCtxInc(sigCoeffGroupFlag64, cgPosX, cgPosY, codingParameters.log2TrSizeCG);
+            uint32_t ctxSig = Quant::getSigCoeffGroupCtxInc(sigCoeffGroupFlag64, cgPosX, cgPosY, cgBlkPos, (trSize >> MLS_CG_LOG2_SIZE));
             encodeBin(sigCoeffGroup, baseCoeffGroupCtx[ctxSig]);
         }
 
@@ -1556,7 +1557,8 @@
         if (sigCoeffGroupFlag64 & cgBlkPosMask)
         {
             X265_CHECK((log2TrSize != 2) || (log2TrSize == 2 && subSet == 0), "log2TrSize and subSet mistake!\n");
-            const int patternSigCtx = Quant::calcPatternSigCtx(sigCoeffGroupFlag64, cgPosX, cgPosY, codingParameters.log2TrSizeCG);
+            const int patternSigCtx = Quant::calcPatternSigCtx(sigCoeffGroupFlag64, cgPosX, cgPosY, cgBlkPos, (trSize >> MLS_CG_LOG2_SIZE));
+            const uint32_t posOffset = (bIsLuma && subSet) ? 3 : 0;
 
             static const uint8_t ctxIndMap4x4[16] =
             {
@@ -1566,37 +1568,50 @@
                 7, 7, 8, 8
             };
             // NOTE: [patternSigCtx][posXinSubset][posYinSubset]
-            static const uint8_t table_cnt[4][4][4] =
+            static const uint8_t table_cnt[4][SCAN_SET_SIZE] =
             {
                 // patternSigCtx = 0
                 {
-                    { 2, 1, 1, 0 },
-                    { 1, 1, 0, 0 },
-                    { 1, 0, 0, 0 },
-                    { 0, 0, 0, 0 },
+                    2, 1, 1, 0,
+                    1, 1, 0, 0,
+                    1, 0, 0, 0,
+                    0, 0, 0, 0,
                 },
                 // patternSigCtx = 1
                 {
-                    { 2, 1, 0, 0 },
-                    { 2, 1, 0, 0 },
-                    { 2, 1, 0, 0 },
-                    { 2, 1, 0, 0 },
+                    2, 2, 2, 2,
+                    1, 1, 1, 1,
+                    0, 0, 0, 0,
+                    0, 0, 0, 0,
                 },
                 // patternSigCtx = 2
                 {
-                    { 2, 2, 2, 2 },
-                    { 1, 1, 1, 1 },
-                    { 0, 0, 0, 0 },
-                    { 0, 0, 0, 0 },
+                    2, 1, 0, 0,
+                    2, 1, 0, 0,
+                    2, 1, 0, 0,
+                    2, 1, 0, 0,
                 },
                 // patternSigCtx = 3
                 {
-                    { 2, 2, 2, 2 },
-                    { 2, 2, 2, 2 },
-                    { 2, 2, 2, 2 },
-                    { 2, 2, 2, 2 },
+                    2, 2, 2, 2,
+                    2, 2, 2, 2,
+                    2, 2, 2, 2,
+                    2, 2, 2, 2,
                 }
             };
+
+            const int offset = codingParameters.firstSignificanceMapContext;
+            ALIGN_VAR_32(uint16_t, tmpCoeff[SCAN_SET_SIZE]);
+            // TODO: accelerate by PABSW
+            const uint32_t blkPosBase  = codingParameters.scan[subPosBase];
+            for (int i = 0; i < MLS_CG_SIZE; i++)
+            {
+                tmpCoeff[i * MLS_CG_SIZE + 0] = (uint16_t)abs(coeff[blkPosBase + i * trSize + 0]);
+                tmpCoeff[i * MLS_CG_SIZE + 1] = (uint16_t)abs(coeff[blkPosBase + i * trSize + 1]);
+                tmpCoeff[i * MLS_CG_SIZE + 2] = (uint16_t)abs(coeff[blkPosBase + i * trSize + 2]);
+                tmpCoeff[i * MLS_CG_SIZE + 3] = (uint16_t)abs(coeff[blkPosBase + i * trSize + 3]);
+            }
+
             if (m_bitIf)
             {
                 if (log2TrSize == 2)
@@ -1604,16 +1619,16 @@
                     uint32_t blkPos, sig, ctxSig;
                     for (; scanPosSigOff >= 0; scanPosSigOff--)
                     {
-                        blkPos  = codingParameters.scan[subPosBase + scanPosSigOff];
+                        blkPos = g_scan4x4[codingParameters.scanType][scanPosSigOff];
                         sig     = scanFlagMask & 1;
                         scanFlagMask >>= 1;
-                        X265_CHECK((uint32_t)(coeff[blkPos] != 0) == sig, "sign bit mistake\n");
+                        X265_CHECK((uint32_t)(tmpCoeff[blkPos] != 0) == sig, "sign bit mistake\n");
                         {
                             ctxSig = ctxIndMap4x4[blkPos];
                             X265_CHECK(ctxSig == Quant::getSigCtxInc(patternSigCtx, log2TrSize, trSize, blkPos, bIsLuma, codingParameters.firstSignificanceMapContext), "sigCtx mistake!\n");;
                             encodeBin(sig, baseCtx[ctxSig]);
                         }
-                        absCoeff[numNonZero] = int(abs(coeff[blkPos]));
+                        absCoeff[numNonZero] = tmpCoeff[blkPos];
                         numNonZero += sig;
                     }
                 }
@@ -1621,35 +1636,25 @@
                 {
                     X265_CHECK((log2TrSize > 2), "log2TrSize must be more than 2 in this path!\n");
 
-                    const uint8_t (*tabSigCtx)[4] = table_cnt[(uint32_t)patternSigCtx];
-                    const int offset = codingParameters.firstSignificanceMapContext;
-                    const uint32_t lumaMask = bIsLuma ? ~0 : 0;
-                    static const uint32_t posXY4Mask[] = {0x024, 0x0CC, 0x39C};
-                    const uint32_t posGT4Mask = posXY4Mask[log2TrSize - 3] & lumaMask;
+                    const uint8_t *tabSigCtx = table_cnt[(uint32_t)patternSigCtx];
 
                     uint32_t blkPos, sig, ctxSig;
                     for (; scanPosSigOff >= 0; scanPosSigOff--)
                     {
-                        blkPos  = codingParameters.scan[subPosBase + scanPosSigOff];
-                        X265_CHECK(blkPos || (subPosBase + scanPosSigOff == 0), "blkPos==0 must be at scan[0]\n");
+                        blkPos = g_scan4x4[codingParameters.scanType][scanPosSigOff];
                         const uint32_t posZeroMask = (subPosBase + scanPosSigOff) ? ~0 : 0;
                         sig     = scanFlagMask & 1;
                         scanFlagMask >>= 1;
-                        X265_CHECK((uint32_t)(coeff[blkPos] != 0) == sig, "sign bit mistake\n");
+                        X265_CHECK((uint32_t)(tmpCoeff[blkPos] != 0) == sig, "sign bit mistake\n");
                         if (scanPosSigOff != 0 || subSet == 0 || numNonZero)
                         {
-                            const uint32_t posY = blkPos >> log2TrSize;
-                            const uint32_t posOffset = (blkPos & posGT4Mask) ? 3 : 0;
-
-                            const uint32_t posXinSubset = blkPos & 3;
-                            const uint32_t posYinSubset = posY & 3;
-                            const uint32_t cnt = tabSigCtx[posXinSubset][posYinSubset] + offset;
+                            const uint32_t cnt = tabSigCtx[blkPos] + offset;
                             ctxSig = (cnt + posOffset) & posZeroMask;
 
-                            X265_CHECK(ctxSig == Quant::getSigCtxInc(patternSigCtx, log2TrSize, trSize, blkPos, bIsLuma, codingParameters.firstSignificanceMapContext), "sigCtx mistake!\n");;
+                            X265_CHECK(ctxSig == Quant::getSigCtxInc(patternSigCtx, log2TrSize, trSize, codingParameters.scan[subPosBase + scanPosSigOff], bIsLuma, codingParameters.firstSignificanceMapContext), "sigCtx mistake!\n");;
                             encodeBin(sig, baseCtx[ctxSig]);
                         }
-                        absCoeff[numNonZero] = int(abs(coeff[blkPos]));
+                        absCoeff[numNonZero] = tmpCoeff[blkPos];
                         numNonZero += sig;
                     }
                 }
@@ -1663,19 +1668,26 @@
                     uint32_t blkPos, sig, ctxSig;
                     for (; scanPosSigOff >= 0; scanPosSigOff--)
                     {
-                        blkPos  = codingParameters.scan[subPosBase + scanPosSigOff];
+                        blkPos = g_scan4x4[codingParameters.scanType][scanPosSigOff];
                         sig     = scanFlagMask & 1;
                         scanFlagMask >>= 1;
-                        X265_CHECK((uint32_t)(coeff[blkPos] != 0) == sig, "sign bit mistake\n");
+                        X265_CHECK((uint32_t)(tmpCoeff[blkPos] != 0) == sig, "sign bit mistake\n");
                         {
                             ctxSig = ctxIndMap4x4[blkPos];
-                            X265_CHECK(ctxSig == Quant::getSigCtxInc(patternSigCtx, log2TrSize, trSize, blkPos, bIsLuma, codingParameters.firstSignificanceMapContext), "sigCtx mistake!\n");;
+                            X265_CHECK(ctxSig == Quant::getSigCtxInc(patternSigCtx, log2TrSize, trSize, codingParameters.scan[subPosBase + scanPosSigOff], bIsLuma, codingParameters.firstSignificanceMapContext), "sigCtx mistake!\n");;
                             //encodeBin(sig, baseCtx[ctxSig]);
                             const uint32_t mstate = baseCtx[ctxSig];
-                            baseCtx[ctxSig] = sbacNext(mstate, sig);
-                            sum += sbacGetEntropyBits(mstate, sig);
+                            const uint32_t mps = mstate & 1;
+                            const uint32_t stateBits = g_entropyStateBits[mstate ^ sig];
+                            uint32_t nextState = (stateBits >> 23) + mps;
+                            if ((mstate ^ sig) == 1)
+                                nextState = sig;
+                            X265_CHECK(sbacNext(mstate, sig) == nextState, "nextState check failure\n");
+                            X265_CHECK(sbacGetEntropyBits(mstate, sig) == (stateBits & 0xFFFFFF), "entropyBits check failure\n");
+                            baseCtx[ctxSig] = (uint8_t)nextState;
+                            sum += stateBits;
                         }
-                        absCoeff[numNonZero] = int(abs(coeff[blkPos]));
+                        absCoeff[numNonZero] = tmpCoeff[blkPos];
                         numNonZero += sig;
                     }
                 } // end of 4x4
@@ -1683,41 +1695,39 @@
                 {
                     X265_CHECK((log2TrSize > 2), "log2TrSize must be more than 2 in this path!\n");
 
-                    const uint8_t (*tabSigCtx)[4] = table_cnt[(uint32_t)patternSigCtx];
-                    const int offset = codingParameters.firstSignificanceMapContext;
-                    const uint32_t lumaMask = bIsLuma ? ~0 : 0;
-                    static const uint32_t posXY4Mask[] = {0x024, 0x0CC, 0x39C};
-                    const uint32_t posGT4Mask = posXY4Mask[log2TrSize - 3] & lumaMask;
+                    const uint8_t *tabSigCtx = table_cnt[(uint32_t)patternSigCtx];
 
                     uint32_t blkPos, sig, ctxSig;
                     for (; scanPosSigOff >= 0; scanPosSigOff--)
                     {
-                        blkPos  = codingParameters.scan[subPosBase + scanPosSigOff];
-                        X265_CHECK(blkPos || (subPosBase + scanPosSigOff == 0), "blkPos==0 must be at scan[0]\n");
+                        blkPos = g_scan4x4[codingParameters.scanType][scanPosSigOff];
                         const uint32_t posZeroMask = (subPosBase + scanPosSigOff) ? ~0 : 0;
                         sig     = scanFlagMask & 1;
                         scanFlagMask >>= 1;
-                        X265_CHECK((uint32_t)(coeff[blkPos] != 0) == sig, "sign bit mistake\n");
+                        X265_CHECK((uint32_t)(tmpCoeff[blkPos] != 0) == sig, "sign bit mistake\n");
                         if (scanPosSigOff != 0 || subSet == 0 || numNonZero)
                         {
-                            const uint32_t posY = blkPos >> log2TrSize;
-                            const uint32_t posOffset = (blkPos & posGT4Mask) ? 3 : 0;
-
-                            const uint32_t posXinSubset = blkPos & 3;
-                            const uint32_t posYinSubset = posY & 3;
-                            const uint32_t cnt = tabSigCtx[posXinSubset][posYinSubset] + offset;
+                            const uint32_t cnt = tabSigCtx[blkPos] + offset;
                             ctxSig = (cnt + posOffset) & posZeroMask;
 
-                            X265_CHECK(ctxSig == Quant::getSigCtxInc(patternSigCtx, log2TrSize, trSize, blkPos, bIsLuma, codingParameters.firstSignificanceMapContext), "sigCtx mistake!\n");;
+                            X265_CHECK(ctxSig == Quant::getSigCtxInc(patternSigCtx, log2TrSize, trSize, codingParameters.scan[subPosBase + scanPosSigOff], bIsLuma, codingParameters.firstSignificanceMapContext), "sigCtx mistake!\n");;
                             //encodeBin(sig, baseCtx[ctxSig]);
                             const uint32_t mstate = baseCtx[ctxSig];
-                            baseCtx[ctxSig] = sbacNext(mstate, sig);
-                            sum += sbacGetEntropyBits(mstate, sig);
+                            const uint32_t mps = mstate & 1;
+                            const uint32_t stateBits = g_entropyStateBits[mstate ^ sig];
+                            uint32_t nextState = (stateBits >> 23) + mps;
+                            if ((mstate ^ sig) == 1)
+                                nextState = sig;
+                            X265_CHECK(sbacNext(mstate, sig) == nextState, "nextState check failure\n");
+                            X265_CHECK(sbacGetEntropyBits(mstate, sig) == (stateBits & 0xFFFFFF), "entropyBits check failure\n");
+                            baseCtx[ctxSig] = (uint8_t)nextState;
+                            sum += stateBits;
                         }
-                        absCoeff[numNonZero] = int(abs(coeff[blkPos]));
+                        absCoeff[numNonZero] = tmpCoeff[blkPos];
                         numNonZero += sig;
                     }
                 } // end of non 4x4 path
+                sum &= 0xFFFFFF;
 
                 // update RD cost
                 m_fracBits += sum;
@@ -1762,31 +1772,77 @@
             if (!c1)
             {
                 baseCtxMod = bIsLuma ? &m_contextState[OFF_ABS_FLAG_CTX + ctxSet] : &m_contextState[OFF_ABS_FLAG_CTX + NUM_ABS_FLAG_CTX_LUMA + ctxSet];
-                if (firstC2FlagIdx != -1)
-                {
-                    uint32_t symbol = absCoeff[firstC2FlagIdx] > 2;
-                    encodeBin(symbol, baseCtxMod[0]);
-                }
+
+                X265_CHECK((firstC2FlagIdx != -1), "firstC2FlagIdx check failure\n");
+                uint32_t symbol = absCoeff[firstC2FlagIdx] > 2;
+                encodeBin(symbol, baseCtxMod[0]);
             }
 
             const int hiddenShift = (bHideFirstSign && signHidden) ? 1 : 0;
             encodeBinsEP((coeffSigns >> hiddenShift), numNonZero - hiddenShift);
 
-            int firstCoeff2 = 1;
             if (!c1 || numNonZero > C1FLAG_NUMBER)
             {
-                for (int idx = 0; idx < numNonZero; idx++)
+                uint32_t goRiceParam = 0;
+                int firstCoeff2 = 1;
+                uint32_t baseLevelN = 0x5555AAAA; // 2-bits encode format baseLevel
+
+                if (!m_bitIf)
                 {
-                    int baseLevel = (idx < C1FLAG_NUMBER) ? (2 + firstCoeff2) : 1;
+                    // FastRd path
+                    for (int idx = 0; idx < numNonZero; idx++)
+                    {
+                        int baseLevel = (baseLevelN & 3) | firstCoeff2;
+                        X265_CHECK(baseLevel == ((idx < C1FLAG_NUMBER) ? (2 + firstCoeff2) : 1), "baseLevel check failurr\n");
+                        baseLevelN >>= 2;
+                        int codeNumber = absCoeff[idx] - baseLevel;
 
-                    if (absCoeff[idx] >= baseLevel)
+                        if (codeNumber >= 0)
+                        {
+                            //writeCoefRemainExGolomb(absCoeff[idx] - baseLevel, goRiceParam);
+                            uint32_t length = 0;
+
+                            codeNumber = ((uint32_t)codeNumber >> goRiceParam) - COEF_REMAIN_BIN_REDUCTION;
+                            if (codeNumber >= 0)
+                            {
+                                {
+                                    unsigned long cidx;
+                                    CLZ(cidx, codeNumber + 1);
+                                    length = cidx;
+                                }
+                                X265_CHECK((codeNumber != 0) || (length == 0), "length check failure\n");
+
+                                codeNumber = (length + length);
+                            }
+                            m_fracBits += (COEF_REMAIN_BIN_REDUCTION + 1 + goRiceParam + codeNumber) << 15;
+
+                            if (absCoeff[idx] > (COEF_REMAIN_BIN_REDUCTION << goRiceParam))
+                                goRiceParam = (goRiceParam + 1) - (goRiceParam >> 2);
+                            X265_CHECK(goRiceParam <= 4, "goRiceParam check failure\n");
+                        }
+                        if (absCoeff[idx] >= 2)
+                            firstCoeff2 = 0;
+                    }
+                }
+                else
+                {
+                    // Standard path
+                    for (int idx = 0; idx < numNonZero; idx++)
                     {
-                        writeCoefRemainExGolomb(absCoeff[idx] - baseLevel, goRiceParam);
-                        if (absCoeff[idx] > 3 * (1 << goRiceParam))
-                            goRiceParam = std::min<uint32_t>(goRiceParam + 1, 4);
+                        int baseLevel = (baseLevelN & 3) | firstCoeff2;
+                        X265_CHECK(baseLevel == ((idx < C1FLAG_NUMBER) ? (2 + firstCoeff2) : 1), "baseLevel check failurr\n");
+                        baseLevelN >>= 2;
+
+                        if (absCoeff[idx] >= baseLevel)
+                        {
+                            writeCoefRemainExGolomb(absCoeff[idx] - baseLevel, goRiceParam);
+                            if (absCoeff[idx] > (COEF_REMAIN_BIN_REDUCTION << goRiceParam))
+                                goRiceParam = (goRiceParam + 1) - (goRiceParam >> 2);
+                            X265_CHECK(goRiceParam <= 4, "goRiceParam check failure\n");
+                        }
+                        if (absCoeff[idx] >= 2)
+                            firstCoeff2 = 0;
                     }
-                    if (absCoeff[idx] >= 2)
-                        firstCoeff2 = 0;
                 }
             }
         }
@@ -1874,20 +1930,20 @@
     if (bIsLuma)
     {
         for (uint32_t bin = 0; bin < 2; bin++)
-            estBitsSbac.significantBits[0][bin] = sbacGetEntropyBits(m_contextState[OFF_SIG_FLAG_CTX], bin);
+            estBitsSbac.significantBits[bin][0] = sbacGetEntropyBits(m_contextState[OFF_SIG_FLAG_CTX], bin);
 
         for (int ctxIdx = firstCtx; ctxIdx < firstCtx + numCtx; ctxIdx++)
             for (uint32_t bin = 0; bin < 2; bin++)
-                estBitsSbac.significantBits[ctxIdx][bin] = sbacGetEntropyBits(m_contextState[OFF_SIG_FLAG_CTX + ctxIdx], bin);
+                estBitsSbac.significantBits[bin][ctxIdx] = sbacGetEntropyBits(m_contextState[OFF_SIG_FLAG_CTX + ctxIdx], bin);
     }
     else
     {
         for (uint32_t bin = 0; bin < 2; bin++)
-            estBitsSbac.significantBits[0][bin] = sbacGetEntropyBits(m_contextState[OFF_SIG_FLAG_CTX + (NUM_SIG_FLAG_CTX_LUMA + 0)], bin);
+            estBitsSbac.significantBits[bin][0] = sbacGetEntropyBits(m_contextState[OFF_SIG_FLAG_CTX + (NUM_SIG_FLAG_CTX_LUMA + 0)], bin);
 
         for (int ctxIdx = firstCtx; ctxIdx < firstCtx + numCtx; ctxIdx++)
             for (uint32_t bin = 0; bin < 2; bin++)
-                estBitsSbac.significantBits[ctxIdx][bin] = sbacGetEntropyBits(m_contextState[OFF_SIG_FLAG_CTX + (NUM_SIG_FLAG_CTX_LUMA + ctxIdx)], bin);
+                estBitsSbac.significantBits[bin][ctxIdx] = sbacGetEntropyBits(m_contextState[OFF_SIG_FLAG_CTX + (NUM_SIG_FLAG_CTX_LUMA + ctxIdx)], bin);
     }
 
     int blkSizeOffset = bIsLuma ? ((log2TrSize - 2) * 3 + ((log2TrSize - 1) >> 2)) : NUM_CTX_LAST_FLAG_XY_LUMA;
@@ -2187,6 +2243,28 @@
     0x0050c, 0x29bab, 0x004c1, 0x2a674, 0x004a7, 0x2aa5e, 0x0046f, 0x2b32f, 0x0041f, 0x2c0ad, 0x003e7, 0x2ca8d, 0x003ba, 0x2d323, 0x0010c, 0x3bfbb
 };
 
+// [8 24] --> [stateMPS BitCost], [stateLPS BitCost]
+const uint32_t g_entropyStateBits[128] =
+{
+    // Corrected table, most notably for last state
+    0x01007b23, 0x000085f9, 0x020074a0, 0x00008cbc, 0x03006ee4, 0x01009354, 0x040067f4, 0x02009c1b,
+    0x050060b0, 0x0200a62a, 0x06005a9c, 0x0400af5b, 0x0700548d, 0x0400b955, 0x08004f56, 0x0500c2a9,
+    0x09004a87, 0x0600cbf7, 0x0a0045d6, 0x0700d5c3, 0x0b004144, 0x0800e01b, 0x0c003d88, 0x0900e937,
+    0x0d0039e0, 0x0900f2cd, 0x0e003663, 0x0b00fc9e, 0x0f003347, 0x0b010600, 0x10003050, 0x0c010f95,
+    0x11002d4d, 0x0d011a02, 0x12002ad3, 0x0d012333, 0x1300286e, 0x0f012cad, 0x14002604, 0x0f0136df,
+    0x15002425, 0x10013f48, 0x160021f4, 0x100149c4, 0x1700203e, 0x1201527b, 0x18001e4d, 0x12015d00,
+    0x19001c99, 0x130166de, 0x1a001b18, 0x13017017, 0x1b0019a5, 0x15017988, 0x1c001841, 0x15018327,
+    0x1d0016df, 0x16018d50, 0x1e0015d9, 0x16019547, 0x1f00147c, 0x1701a083, 0x2000138e, 0x1801a8a3,
+    0x21001251, 0x1801b418, 0x22001166, 0x1901bd27, 0x23001068, 0x1a01c77b, 0x24000f7f, 0x1a01d18e,
+    0x25000eda, 0x1b01d91a, 0x26000e19, 0x1b01e254, 0x27000d4f, 0x1c01ec9a, 0x28000c90, 0x1d01f6e0,
+    0x29000c01, 0x1d01fef8, 0x2a000b5f, 0x1e0208b1, 0x2b000ab6, 0x1e021362, 0x2c000a15, 0x1e021e46,
+    0x2d000988, 0x1f02285d, 0x2e000934, 0x20022ea8, 0x2f0008a8, 0x200239b2, 0x3000081d, 0x21024577,
+    0x310007c9, 0x21024ce6, 0x32000763, 0x21025663, 0x33000710, 0x22025e8f, 0x340006a0, 0x22026a26,
+    0x35000672, 0x23026f23, 0x360005e8, 0x23027ef8, 0x370005ba, 0x230284b5, 0x3800055e, 0x24029057,
+    0x3900050c, 0x24029bab, 0x3a0004c1, 0x2402a674, 0x3b0004a7, 0x2502aa5e, 0x3c00046f, 0x2502b32f,
+    0x3d00041f, 0x2502c0ad, 0x3e0003e7, 0x2602ca8d, 0x3e0003ba, 0x2602d323, 0x3f00010c, 0x3f03bfbb,
+};
+
 const uint8_t g_nextState[128][2] =
 {
     { 2, 1 }, { 0, 3 }, { 4, 0 }, { 1, 5 }, { 6, 2 }, { 3, 7 }, { 8, 4 }, { 5, 9 },

x265_1.6.tar.gz/source/encoder/entropy.h -> x265_1.7.tar.gz/source/encoder/entropy.h Changed

@@ -87,7 +87,7 @@
 struct EstBitsSbac
 {
     int significantCoeffGroupBits[NUM_SIG_CG_FLAG_CTX][2];
-    int significantBits[NUM_SIG_FLAG_CTX][2];
+    int significantBits[2][NUM_SIG_FLAG_CTX];
     int lastBits[2][10];
     int greaterOneBits[NUM_ONE_FLAG_CTX][2];
     int levelAbsBits[NUM_ABS_FLAG_CTX][2];
@@ -179,7 +179,7 @@
     inline void codeQtCbfChroma(uint32_t cbf, uint32_t tuDepth)           { encodeBin(cbf, m_contextState[OFF_QT_CBF_CTX + 2 + tuDepth]); }
     inline void codeQtRootCbf(uint32_t cbf)                               { encodeBin(cbf, m_contextState[OFF_QT_ROOT_CBF_CTX]); }
     inline void codeTransformSkipFlags(uint32_t transformSkip, TextType ttype) { encodeBin(transformSkip, m_contextState[OFF_TRANSFORMSKIP_FLAG_CTX + (ttype ? NUM_TRANSFORMSKIP_FLAG_CTX : 0)]); }
-
+    void codeDeltaQP(const CUData& cu, uint32_t absPartIdx);
     void codeSaoOffset(const SaoCtuParam& ctuParam, int plane);
 
     /* RDO functions */
@@ -221,7 +221,7 @@
     }
 
     void encodeCU(const CUData& ctu, const CUGeom &cuGeom, uint32_t absPartIdx, uint32_t depth, bool& bEncodeDQP);
-    void finishCU(const CUData& ctu, uint32_t absPartIdx, uint32_t depth);
+    void finishCU(const CUData& ctu, uint32_t absPartIdx, uint32_t depth, bool bEncodeDQP);
 
     void writeOut();
 
@@ -242,7 +242,6 @@
 
     void codeSaoMaxUvlc(uint32_t code, uint32_t maxSymbol);
 
-    void codeDeltaQP(const CUData& cu, uint32_t absPartIdx);
     void codeLastSignificantXY(uint32_t posx, uint32_t posy, uint32_t log2TrSize, bool bIsLuma, uint32_t scanIdx);
 
     void encodeTransform(const CUData& cu, uint32_t absPartIdx, uint32_t tuDepth, uint32_t log2TrSize,

x265_1.6.tar.gz/source/encoder/frameencoder.cpp -> x265_1.7.tar.gz/source/encoder/frameencoder.cpp Changed

@@ -213,6 +213,7 @@
 {
     m_slicetypeWaitTime = x265_mdate() - m_prevOutputTime;
     m_frame = curFrame;
+    m_param = curFrame->m_param;
     m_sliceType = curFrame->m_lowres.sliceType;
     curFrame->m_encData->m_frameEncoderID = m_jpId;
     curFrame->m_encData->m_jobProvider = this;
@@ -794,6 +795,7 @@
     uint32_t row = (uint32_t)intRow;
     CTURow& curRow = m_rows[row];
 
+    tld.analysis.m_param = m_param;
     if (m_param->bEnableWavefront)
     {
         ScopedLock self(curRow.lock);
@@ -824,6 +826,13 @@
     const uint32_t lineStartCUAddr = row * numCols;
     bool bIsVbv = m_param->rc.vbvBufferSize > 0 && m_param->rc.vbvMaxBitrate > 0;
 
+    /* These store the count of inter, intra and skip cus within quad tree structure of each CTU */
+    uint32_t qTreeInterCnt[NUM_CU_DEPTH];
+    uint32_t qTreeIntraCnt[NUM_CU_DEPTH];
+    uint32_t qTreeSkipCnt[NUM_CU_DEPTH];
+    for (uint32_t depth = 0; depth <= g_maxCUDepth; depth++)
+        qTreeIntraCnt[depth] = qTreeInterCnt[depth] = qTreeSkipCnt[depth] = 0;
+
     while (curRow.completed < numCols)
     {
         ProfileScopeEvent(encodeCTU);
@@ -841,24 +850,34 @@
                 curEncData.m_rowStat[row].diagQpScale = x265_qp2qScale(curEncData.m_avgQpRc);
             }
 
+            FrameData::RCStatCU& cuStat = curEncData.m_cuStat[cuAddr];
             if (row >= col && row && m_vbvResetTriggerRow != intRow)
-                curEncData.m_cuStat[cuAddr].baseQp = curEncData.m_cuStat[cuAddr - numCols + 1].baseQp;
+                cuStat.baseQp = curEncData.m_cuStat[cuAddr - numCols + 1].baseQp;
             else
-                curEncData.m_cuStat[cuAddr].baseQp = curEncData.m_rowStat[row].diagQp;
-        }
-        else
-            curEncData.m_cuStat[cuAddr].baseQp = curEncData.m_avgQpRc;
+                cuStat.baseQp = curEncData.m_rowStat[row].diagQp;
+
+            /* TODO: use defines from slicetype.h for lowres block size */
+            uint32_t maxBlockCols = (m_frame->m_fencPic->m_picWidth + (16 - 1)) / 16;
+            uint32_t maxBlockRows = (m_frame->m_fencPic->m_picHeight + (16 - 1)) / 16;
+            uint32_t noOfBlocks = g_maxCUSize / 16;
+            uint32_t block_y = (cuAddr / curEncData.m_slice->m_sps->numCuInWidth) * noOfBlocks;
+            uint32_t block_x = (cuAddr * noOfBlocks) - block_y * curEncData.m_slice->m_sps->numCuInWidth;
+            
+            cuStat.vbvCost = 0;
+            cuStat.intraVbvCost = 0;
+            for (uint32_t h = 0; h < noOfBlocks && block_y < maxBlockRows; h++, block_y++)
+            {
+                uint32_t idx = block_x + (block_y * maxBlockCols);
 
-        if (m_param->rc.aqMode || bIsVbv)
-        {
-            int qp = calcQpForCu(cuAddr, curEncData.m_cuStat[cuAddr].baseQp);
-            tld.analysis.setQP(*slice, qp);
-            qp = x265_clip3(QP_MIN, QP_MAX_SPEC, qp);
-            ctu->setQPSubParts((int8_t)qp, 0, 0);
-            curEncData.m_rowStat[row].sumQpAq += qp;
+                for (uint32_t w = 0; w < noOfBlocks && (block_x + w) < maxBlockCols; w++, idx++)
+                {
+                    cuStat.vbvCost += m_frame->m_lowres.lowresCostForRc[idx] & LOWRES_COST_MASK;
+                    cuStat.intraVbvCost += m_frame->m_lowres.intraCost[idx];
+                }
+            }
         }
         else
-            tld.analysis.setQP(*slice, slice->m_sliceQp);
+            curEncData.m_cuStat[cuAddr].baseQp = curEncData.m_avgQpRc;
 
         if (m_param->bEnableWavefront && !col && row)
         {
@@ -886,7 +905,9 @@
         curRow.completed++;
 
         if (m_param->bLogCuStats || m_param->rc.bStatWrite)
-            collectCTUStatistics(*ctu);
+            curEncData.m_rowStat[row].sumQpAq += collectCTUStatistics(*ctu, qTreeInterCnt, qTreeIntraCnt, qTreeSkipCnt);
+        else if (m_param->rc.aqMode)
+            curEncData.m_rowStat[row].sumQpAq += calcCTUQP(*ctu);
 
         // copy no. of intra, inter Cu cnt per row into frame stats for 2 pass
         if (m_param->rc.bStatWrite)
@@ -894,18 +915,17 @@
             curRow.rowStats.mvBits += best.mvBits;
             curRow.rowStats.coeffBits += best.coeffBits;
             curRow.rowStats.miscBits += best.totalBits - (best.mvBits + best.coeffBits);
-            StatisticLog* log = &m_sliceTypeLog[slice->m_sliceType];
 
             for (uint32_t depth = 0; depth <= g_maxCUDepth; depth++)
             {
                 /* 1 << shift == number of 8x8 blocks at current depth */
                 int shift = 2 * (g_maxCUDepth - depth);
-                curRow.rowStats.iCuCnt += log->qTreeIntraCnt[depth] << shift;
-                curRow.rowStats.pCuCnt += log->qTreeInterCnt[depth] << shift;
-                curRow.rowStats.skipCuCnt += log->qTreeSkipCnt[depth] << shift;
+                curRow.rowStats.iCuCnt += qTreeIntraCnt[depth] << shift;
+                curRow.rowStats.pCuCnt += qTreeInterCnt[depth] << shift;
+                curRow.rowStats.skipCuCnt += qTreeSkipCnt[depth] << shift;
 
                 // clear the row cu data from thread local object
-                log->qTreeIntraCnt[depth] = log->qTreeInterCnt[depth] = log->qTreeSkipCnt[depth] = 0;
+                qTreeIntraCnt[depth] = qTreeInterCnt[depth] = qTreeSkipCnt[depth] = 0;
             }
         }
 
@@ -1075,15 +1095,18 @@
         }
     }
 
+    tld.analysis.m_param = NULL;
     curRow.busy = false;
 
     if (ATOMIC_INC(&m_completionCount) == 2 * (int)m_numRows)
         m_completionEvent.trigger();
 }
 
-void FrameEncoder::collectCTUStatistics(CUData& ctu)
+/* collect statistics about CU coding decisions, return total QP */
+int FrameEncoder::collectCTUStatistics(const CUData& ctu, uint32_t* qtreeInterCnt, uint32_t* qtreeIntraCnt, uint32_t* qtreeSkipCnt)
 {
     StatisticLog* log = &m_sliceTypeLog[ctu.m_slice->m_sliceType];
+    int totQP = 0;
 
     if (ctu.m_slice->m_sliceType == I_SLICE)
     {
@@ -1094,13 +1117,14 @@
 
             log->totalCu++;
             log->cntIntra[depth]++;
-            log->qTreeIntraCnt[depth]++;
+            qtreeIntraCnt[depth]++;
+            totQP += ctu.m_qp[absPartIdx] * (ctu.m_numPartitions >> (depth * 2));
 
             if (ctu.m_predMode[absPartIdx] == MODE_NONE)
             {
                 log->totalCu--;
                 log->cntIntra[depth]--;
-                log->qTreeIntraCnt[depth]--;
+                qtreeIntraCnt[depth]--;
             }
             else if (ctu.m_partSize[absPartIdx] != SIZE_2Nx2N)
             {
@@ -1124,6 +1148,7 @@
 
             log->totalCu++;
             log->cntTotalCu[depth]++;
+            totQP += ctu.m_qp[absPartIdx] * (ctu.m_numPartitions >> (depth * 2));
 
             if (ctu.m_predMode[absPartIdx] == MODE_NONE)
             {
@@ -1134,12 +1159,12 @@
             {
                 log->totalCu--;
                 log->cntSkipCu[depth]++;
-                log->qTreeSkipCnt[depth]++;
+                qtreeSkipCnt[depth]++;
             }
             else if (ctu.isInter(absPartIdx))
             {
                 log->cntInter[depth]++;
-                log->qTreeInterCnt[depth]++;
+                qtreeInterCnt[depth]++;
 
                 if (ctu.m_partSize[absPartIdx] < AMP_ID)
                     log->cuInterDistribution[depth][ctu.m_partSize[absPartIdx]]++;
@@ -1149,12 +1174,13 @@
             else if (ctu.isIntra(absPartIdx))
             {
                 log->cntIntra[depth]++;
-                log->qTreeIntraCnt[depth]++;
+                qtreeIntraCnt[depth]++;
 
                 if (ctu.m_partSize[absPartIdx] != SIZE_2Nx2N)
                 {
                     X265_CHECK(ctu.m_log2CUSize[absPartIdx] == 3 && ctu.m_slice->m_sps->quadtreeTULog2MinSize < 3, "Intra NxN found at improbable depth\n");
                     log->cntIntraNxN++;
+                    log->cntIntra[depth]--;
                     /* TODO: log intra modes at absPartIdx +0 to +3 */
                 }
                 else if (ctu.m_lumaIntraDir[absPartIdx] > 1)
@@ -1164,6 +1190,23 @@
             }
         }
     }
+
+    return totQP;
+}
+
+/* iterate over coded CUs and determine total QP */
+int FrameEncoder::calcCTUQP(const CUData& ctu)
+{
+    int totQP = 0;
+    uint32_t depth = 0, numParts = ctu.m_numPartitions;
+
+    for (uint32_t absPartIdx = 0; absPartIdx < ctu.m_numPartitions; absPartIdx += numParts)
+    {
+        depth = ctu.m_cuDepth[absPartIdx];
+        numParts = ctu.m_numPartitions >> (depth * 2);
+        totQP += ctu.m_qp[absPartIdx] * numParts;
+    }
+    return totQP;
 }
 
 /* DCT-domain noise reduction / adaptive deadzone from libavcodec */
@@ -1198,55 +1241,6 @@
     }
 }
 
-int FrameEncoder::calcQpForCu(uint32_t ctuAddr, double baseQp)
-{
-    x265_emms();
-    double qp = baseQp;
-
-    FrameData& curEncData = *m_frame->m_encData;
-    /* clear cuCostsForVbv from when vbv row reset was triggered */
-    bool bIsVbv = m_param->rc.vbvBufferSize > 0 && m_param->rc.vbvMaxBitrate > 0;
-    if (bIsVbv)
-    {
-        curEncData.m_cuStat[ctuAddr].vbvCost = 0;
-        curEncData.m_cuStat[ctuAddr].intraVbvCost = 0;
-    }
-
-    /* Derive qpOffet for each CU by averaging offsets for all 16x16 blocks in the cu. */
-    double qp_offset = 0;
-    uint32_t maxBlockCols = (m_frame->m_fencPic->m_picWidth + (16 - 1)) / 16;
-    uint32_t maxBlockRows = (m_frame->m_fencPic->m_picHeight + (16 - 1)) / 16;
-    uint32_t noOfBlocks = g_maxCUSize / 16;
-    uint32_t block_y = (ctuAddr / curEncData.m_slice->m_sps->numCuInWidth) * noOfBlocks;
-    uint32_t block_x = (ctuAddr * noOfBlocks) - block_y * curEncData.m_slice->m_sps->numCuInWidth;
-
-    /* Use cuTree offsets if cuTree enabled and frame is referenced, else use AQ offsets */
-    bool isReferenced = IS_REFERENCED(m_frame);
-    double *qpoffs = (isReferenced && m_param->rc.cuTree) ? m_frame->m_lowres.qpCuTreeOffset : m_frame->m_lowres.qpAqOffset;
-
-    uint32_t cnt = 0, idx = 0;
-    for (uint32_t h = 0; h < noOfBlocks && block_y < maxBlockRows; h++, block_y++)
-    {
-        for (uint32_t w = 0; w < noOfBlocks && (block_x + w) < maxBlockCols; w++)
-        {
-            idx = block_x + w + (block_y * maxBlockCols);
-            if (m_param->rc.aqMode)
-                qp_offset += qpoffs[idx];
-            if (bIsVbv)
-            {
-                curEncData.m_cuStat[ctuAddr].vbvCost += m_frame->m_lowres.lowresCostForRc[idx] & LOWRES_COST_MASK;
-                curEncData.m_cuStat[ctuAddr].intraVbvCost += m_frame->m_lowres.intraCost[idx];
-            }
-            cnt++;
-        }
-    }
-
-    qp_offset /= cnt;
-    qp += qp_offset;
-
-    return x265_clip3(QP_MIN, QP_MAX_MAX, (int)(qp + 0.5));
-}
-
 Frame *FrameEncoder::getEncodedPicture(NALList& output)
 {
     if (m_frame)

x265_1.6.tar.gz/source/encoder/frameencoder.h -> x265_1.7.tar.gz/source/encoder/frameencoder.h Changed

x265_1.6.tar.gz/source/encoder/level.cpp -> x265_1.7.tar.gz/source/encoder/level.cpp Changed

@@ -55,15 +55,14 @@
     { 35651584, 1069547520, 60000,    240000,   60000,  240000,   8, Level::LEVEL6,   "6",   60 },
     { 35651584, 2139095040, 120000,   480000,   120000, 480000,   8, Level::LEVEL6_1, "6.1", 61 },
     { 35651584, 4278190080U, 240000,  800000,   240000, 800000,   6, Level::LEVEL6_2, "6.2", 62 },
+    { MAX_UINT, MAX_UINT, MAX_UINT, MAX_UINT, MAX_UINT, MAX_UINT, 1, Level::LEVEL8_5, "8.5", 85 },
 };
 
 /* determine minimum decoder level required to decode the described video */
 void determineLevel(const x265_param &param, VPS& vps)
 {
     vps.maxTempSubLayers = param.bEnableTemporalSubLayers ? 2 : 1;
-    if (param.bLossless)
-        vps.ptl.profileIdc = Profile::NONE;
-    else if (param.internalCsp == X265_CSP_I420)
+    if (param.internalCsp == X265_CSP_I420)
     {
         if (param.internalBitDepth == 8)
         {
@@ -104,7 +103,15 @@
 
     const size_t NumLevels = sizeof(levels) / sizeof(levels[0]);
     uint32_t i;
-    for (i = 0; i < NumLevels; i++)
+    if (param.bLossless)
+    {
+        i = 13;
+        vps.ptl.minCrForLevel = 1;
+        vps.ptl.maxLumaSrForLevel = MAX_UINT;
+        vps.ptl.levelIdc = Level::LEVEL8_5;
+        vps.ptl.tierFlag = Level::MAIN;
+    }
+    else for (i = 0; i < NumLevels; i++)
     {
         if (lumaSamples > levels[i].maxLumaSamples)
             continue;
@@ -337,31 +344,40 @@
 extern "C"
 int x265_param_apply_profile(x265_param *param, const char *profile)
 {
-    if (!profile)
+    if (!param || !profile)
         return 0;
-    if (!strcmp(profile, "main"))
-    {
-        /* SPSs shall have chroma_format_idc equal to 1 only */
-        param->internalCsp = X265_CSP_I420;
 
 #if HIGH_BIT_DEPTH
-        /* SPSs shall have bit_depth_luma_minus8 equal to 0 only */
-        x265_log(param, X265_LOG_ERROR, "Main profile not supported, compiled for Main10.\n");
+    if (!strcmp(profile, "main") || !strcmp(profile, "mainstillpicture") || !strcmp(profile, "msp") || !strcmp(profile, "main444-8"))
+    {
+        x265_log(param, X265_LOG_ERROR, "%s profile not supported, compiled for Main10.\n", profile);
         return -1;
-#endif
     }
-    else if (!strcmp(profile, "main10"))
+#else
+    if (!strcmp(profile, "main10") || !strcmp(profile, "main422-10") || !strcmp(profile, "main444-10"))
     {
-        /* SPSs shall have chroma_format_idc equal to 1 only */
-        param->internalCsp = X265_CSP_I420;
-
-        /* SPSs shall have bit_depth_luma_minus8 in the range of 0 to 2, inclusive 
-         * this covers all builds of x265, currently */
+        x265_log(param, X265_LOG_ERROR, "%s profile not supported, compiled for Main.\n", profile);
+        return -1;
+    }
+#endif
+    
+    if (!strcmp(profile, "main"))
+    {
+        if (!(param->internalCsp & X265_CSP_I420))
+        {
+            x265_log(param, X265_LOG_ERROR, "%s profile not compatible with %s input color space.\n",
+                     profile, x265_source_csp_names[param->internalCsp]);
+            return -1;
+        }
     }
     else if (!strcmp(profile, "mainstillpicture") || !strcmp(profile, "msp"))
     {
-        /* SPSs shall have chroma_format_idc equal to 1 only */
-        param->internalCsp = X265_CSP_I420;
+        if (!(param->internalCsp & X265_CSP_I420))
+        {
+            x265_log(param, X265_LOG_ERROR, "%s profile not compatible with %s input color space.\n",
+                     profile, x265_source_csp_names[param->internalCsp]);
+            return -1;
+        }
 
         /* SPSs shall have sps_max_dec_pic_buffering_minus1[ sps_max_sub_layers_minus1 ] equal to 0 only */
         param->maxNumReferences = 1;
@@ -378,25 +394,29 @@
         param->rc.cuTree = 0;
         param->bEnableWeightedPred = 0;
         param->bEnableWeightedBiPred = 0;
-
-#if HIGH_BIT_DEPTH
-        /* SPSs shall have bit_depth_luma_minus8 equal to 0 only */
-        x265_log(param, X265_LOG_ERROR, "Mainstillpicture profile not supported, compiled for Main10.\n");
-        return -1;
-#endif
+    }
+    else if (!strcmp(profile, "main10"))
+    {
+        if (!(param->internalCsp & X265_CSP_I420))
+        {
+            x265_log(param, X265_LOG_ERROR, "%s profile not compatible with %s input color space.\n",
+                     profile, x265_source_csp_names[param->internalCsp]);
+            return -1;
+        }
     }
     else if (!strcmp(profile, "main422-10"))
-        param->internalCsp = X265_CSP_I422;
-    else if (!strcmp(profile, "main444-8"))
     {
-        param->internalCsp = X265_CSP_I444;
-#if HIGH_BIT_DEPTH
-        x265_log(param, X265_LOG_ERROR, "Main 4:4:4 8 profile not supported, compiled for Main10.\n");
-        return -1;
-#endif
+        if (!(param->internalCsp & (X265_CSP_I420 | X265_CSP_I422)))
+        {
+            x265_log(param, X265_LOG_ERROR, "%s profile not compatible with %s input color space.\n",
+                     profile, x265_source_csp_names[param->internalCsp]);
+            return -1;
+        }
+    }
+    else if (!strcmp(profile, "main444-8") || !strcmp(profile, "main444-10"))
+    {
+        /* any color space allowed */
     }
-    else if (!strcmp(profile, "main444-10"))
-        param->internalCsp = X265_CSP_I444;
     else
     {
         x265_log(param, X265_LOG_ERROR, "unknown profile <%s>\n", profile);

x265_1.6.tar.gz/source/encoder/motion.cpp -> x265_1.7.tar.gz/source/encoder/motion.cpp Changed

@@ -234,9 +234,14 @@
                pix_base + (m1x) + (m1y) * stride, \
                pix_base + (m2x) + (m2y) * stride, \
                stride, costs); \
-        (costs)[0] += mvcost((bmv + MV(m0x, m0y)) << 2); \
-        (costs)[1] += mvcost((bmv + MV(m1x, m1y)) << 2); \
-        (costs)[2] += mvcost((bmv + MV(m2x, m2y)) << 2); \
+        const uint16_t *base_mvx = &m_cost_mvx[(bmv.x + (m0x)) << 2]; \
+        const uint16_t *base_mvy = &m_cost_mvy[(bmv.y + (m0y)) << 2]; \
+        X265_CHECK(mvcost((bmv + MV(m0x, m0y)) << 2) == (base_mvx[((m0x) - (m0x)) << 2] + base_mvy[((m0y) - (m0y)) << 2]), "mvcost() check failure\n"); \
+        X265_CHECK(mvcost((bmv + MV(m1x, m1y)) << 2) == (base_mvx[((m1x) - (m0x)) << 2] + base_mvy[((m1y) - (m0y)) << 2]), "mvcost() check failure\n"); \
+        X265_CHECK(mvcost((bmv + MV(m2x, m2y)) << 2) == (base_mvx[((m2x) - (m0x)) << 2] + base_mvy[((m2y) - (m0y)) << 2]), "mvcost() check failure\n"); \
+        (costs)[0] += (base_mvx[((m0x) - (m0x)) << 2] + base_mvy[((m0y) - (m0y)) << 2]); \
+        (costs)[1] += (base_mvx[((m1x) - (m0x)) << 2] + base_mvy[((m1y) - (m0y)) << 2]); \
+        (costs)[2] += (base_mvx[((m2x) - (m0x)) << 2] + base_mvy[((m2y) - (m0y)) << 2]); \
     }
 
 #define COST_MV_PT_DIST_X4(m0x, m0y, p0, d0, m1x, m1y, p1, d1, m2x, m2y, p2, d2, m3x, m3y, p3, d3) \
@@ -247,10 +252,10 @@
                fref + (m2x) + (m2y) * stride, \
                fref + (m3x) + (m3y) * stride, \
                stride, costs); \
-        costs[0] += mvcost(MV(m0x, m0y) << 2); \
-        costs[1] += mvcost(MV(m1x, m1y) << 2); \
-        costs[2] += mvcost(MV(m2x, m2y) << 2); \
-        costs[3] += mvcost(MV(m3x, m3y) << 2); \
+        (costs)[0] += mvcost(MV(m0x, m0y) << 2); \
+        (costs)[1] += mvcost(MV(m1x, m1y) << 2); \
+        (costs)[2] += mvcost(MV(m2x, m2y) << 2); \
+        (costs)[3] += mvcost(MV(m3x, m3y) << 2); \
         COPY4_IF_LT(bcost, costs[0], bmv, MV(m0x, m0y), bPointNr, p0, bDistance, d0); \
         COPY4_IF_LT(bcost, costs[1], bmv, MV(m1x, m1y), bPointNr, p1, bDistance, d1); \
         COPY4_IF_LT(bcost, costs[2], bmv, MV(m2x, m2y), bPointNr, p2, bDistance, d2); \
@@ -266,10 +271,16 @@
                pix_base + (m2x) + (m2y) * stride, \
                pix_base + (m3x) + (m3y) * stride, \
                stride, costs); \
-        costs[0] += mvcost((omv + MV(m0x, m0y)) << 2); \
-        costs[1] += mvcost((omv + MV(m1x, m1y)) << 2); \
-        costs[2] += mvcost((omv + MV(m2x, m2y)) << 2); \
-        costs[3] += mvcost((omv + MV(m3x, m3y)) << 2); \
+        const uint16_t *base_mvx = &m_cost_mvx[(omv.x << 2)]; \
+        const uint16_t *base_mvy = &m_cost_mvy[(omv.y << 2)]; \
+        X265_CHECK(mvcost((omv + MV(m0x, m0y)) << 2) == (base_mvx[(m0x) << 2] + base_mvy[(m0y) << 2]), "mvcost() check failure\n"); \
+        X265_CHECK(mvcost((omv + MV(m1x, m1y)) << 2) == (base_mvx[(m1x) << 2] + base_mvy[(m1y) << 2]), "mvcost() check failure\n"); \
+        X265_CHECK(mvcost((omv + MV(m2x, m2y)) << 2) == (base_mvx[(m2x) << 2] + base_mvy[(m2y) << 2]), "mvcost() check failure\n"); \
+        X265_CHECK(mvcost((omv + MV(m3x, m3y)) << 2) == (base_mvx[(m3x) << 2] + base_mvy[(m3y) << 2]), "mvcost() check failure\n"); \
+        costs[0] += (base_mvx[(m0x) << 2] + base_mvy[(m0y) << 2]); \
+        costs[1] += (base_mvx[(m1x) << 2] + base_mvy[(m1y) << 2]); \
+        costs[2] += (base_mvx[(m2x) << 2] + base_mvy[(m2y) << 2]); \
+        costs[3] += (base_mvx[(m3x) << 2] + base_mvy[(m3y) << 2]); \
         COPY2_IF_LT(bcost, costs[0], bmv, omv + MV(m0x, m0y)); \
         COPY2_IF_LT(bcost, costs[1], bmv, omv + MV(m1x, m1y)); \
         COPY2_IF_LT(bcost, costs[2], bmv, omv + MV(m2x, m2y)); \
@@ -285,10 +296,17 @@
                pix_base + (m2x) + (m2y) * stride, \
                pix_base + (m3x) + (m3y) * stride, \
                stride, costs); \
-        (costs)[0] += mvcost((bmv + MV(m0x, m0y)) << 2); \
-        (costs)[1] += mvcost((bmv + MV(m1x, m1y)) << 2); \
-        (costs)[2] += mvcost((bmv + MV(m2x, m2y)) << 2); \
-        (costs)[3] += mvcost((bmv + MV(m3x, m3y)) << 2); \
+        /* TODO: use restrict keyword in ICL */ \
+        const uint16_t *base_mvx = &m_cost_mvx[(bmv.x << 2)]; \
+        const uint16_t *base_mvy = &m_cost_mvy[(bmv.y << 2)]; \
+        X265_CHECK(mvcost((bmv + MV(m0x, m0y)) << 2) == (base_mvx[(m0x) << 2] + base_mvy[(m0y) << 2]), "mvcost() check failure\n"); \
+        X265_CHECK(mvcost((bmv + MV(m1x, m1y)) << 2) == (base_mvx[(m1x) << 2] + base_mvy[(m1y) << 2]), "mvcost() check failure\n"); \
+        X265_CHECK(mvcost((bmv + MV(m2x, m2y)) << 2) == (base_mvx[(m2x) << 2] + base_mvy[(m2y) << 2]), "mvcost() check failure\n"); \
+        X265_CHECK(mvcost((bmv + MV(m3x, m3y)) << 2) == (base_mvx[(m3x) << 2] + base_mvy[(m3y) << 2]), "mvcost() check failure\n"); \
+        (costs)[0] += (base_mvx[(m0x) << 2] + base_mvy[(m0y) << 2]); \
+        (costs)[1] += (base_mvx[(m1x) << 2] + base_mvy[(m1y) << 2]); \
+        (costs)[2] += (base_mvx[(m2x) << 2] + base_mvy[(m2y) << 2]); \
+        (costs)[3] += (base_mvx[(m3x) << 2] + base_mvy[(m3y) << 2]); \
     }
 
 #define DIA1_ITER(mx, my) \

x265_1.6.tar.gz/source/encoder/nal.cpp -> x265_1.7.tar.gz/source/encoder/nal.cpp Changed

@@ -35,6 +35,7 @@
     , m_extraBuffer(NULL)
     , m_extraOccupancy(0)
     , m_extraAllocSize(0)
+    , m_annexB(true)
 {}
 
 void NALList::takeContents(NALList& other)
@@ -90,7 +91,12 @@
     uint8_t *out = m_buffer + m_occupancy;
     uint32_t bytes = 0;
 
-    if (!m_numNal || nalUnitType == NAL_UNIT_VPS || nalUnitType == NAL_UNIT_SPS || nalUnitType == NAL_UNIT_PPS)
+    if (!m_annexB)
+    {
+        /* Will write size later */
+        bytes += 4;
+    }
+    else if (!m_numNal || nalUnitType == NAL_UNIT_VPS || nalUnitType == NAL_UNIT_SPS || nalUnitType == NAL_UNIT_PPS)
     {
         memcpy(out, startCodePrefix, 4);
         bytes += 4;
@@ -144,6 +150,16 @@
      * to 0x03 is appended to the end of the data.  */
     if (!out[bytes - 1])
         out[bytes++] = 0x03;
+
+    if (!m_annexB)
+    {
+        uint32_t dataSize = bytes - 4;
+        out[0] = (uint8_t)(dataSize >> 24);
+        out[1] = (uint8_t)(dataSize >> 16);
+        out[2] = (uint8_t)(dataSize >> 8);
+        out[3] = (uint8_t)dataSize;
+    }
+
     m_occupancy += bytes;
 
     X265_CHECK(m_numNal < (uint32_t)MAX_NAL_UNITS, "NAL count overflow\n");

x265_1.6.tar.gz/source/encoder/nal.h -> x265_1.7.tar.gz/source/encoder/nal.h Changed

x265_1.6.tar.gz/source/encoder/ratecontrol.cpp -> x265_1.7.tar.gz/source/encoder/ratecontrol.cpp Changed

@@ -300,7 +300,7 @@
         }
     }
 
-    /* qstep - value set as encoder specific */
+    /* qpstep - value set as encoder specific */
     m_lstep = pow(2, m_param->rc.qpStep / 6.0);
 
     for (int i = 0; i < 2; i++)
@@ -370,14 +370,19 @@
     m_accumPQp = (m_param->rc.rateControlMode == X265_RC_CRF ? CRF_INIT_QP : ABR_INIT_QP_MIN) * m_accumPNorm;
 
     /* Frame Predictors and Row predictors used in vbv */
-    for (int i = 0; i < 5; i++)
+    for (int i = 0; i < 4; i++)
     {
-        m_pred[i].coeff = 1.5;
+        m_pred[i].coeff = 1.0;
         m_pred[i].count = 1.0;
         m_pred[i].decay = 0.5;
         m_pred[i].offset = 0.0;
     }
-    m_pred[0].coeff = 1.0;
+    m_pred[0].coeff = m_pred[3].coeff = 0.75;
+    if (m_param->rc.qCompress >= 0.8) // when tuned for grain 
+    {
+        m_pred[1].coeff = 0.75;
+        m_pred[0].coeff = m_pred[3].coeff = 0.50;
+    }
     if (!m_statFileOut && (m_param->rc.bStatWrite || m_param->rc.bStatRead))
     {
         /* If the user hasn't defined the stat filename, use the default value */
@@ -945,6 +950,9 @@
     m_curSlice = curEncData.m_slice;
     m_sliceType = m_curSlice->m_sliceType;
     rce->sliceType = m_sliceType;
+    if (!m_2pass)
+        rce->keptAsRef = IS_REFERENCED(curFrame);
+    m_predType = getPredictorType(curFrame->m_lowres.sliceType, m_sliceType);
     rce->poc = m_curSlice->m_poc;
     if (m_param->rc.bStatRead)
     {
@@ -1074,7 +1082,7 @@
             m_lastQScaleFor[m_sliceType] = x265_qp2qScale(rce->qpaRc);
             if (rce->poc == 0)
                  m_lastQScaleFor[P_SLICE] = m_lastQScaleFor[m_sliceType] * fabs(m_param->rc.ipFactor);
-            rce->frameSizePlanned = predictSize(&m_pred[m_sliceType], m_qp, (double)m_currentSatd);
+            rce->frameSizePlanned = predictSize(&m_pred[m_predType], m_qp, (double)m_currentSatd);
         }
     }
     m_framesDone++;
@@ -1105,6 +1113,14 @@
         m_accumPQp += m_qp;
 }
 
+int RateControl::getPredictorType(int lowresSliceType, int sliceType)
+{
+    /* Use a different predictor for B Ref and B frames for vbv frame size predictions */
+    if (lowresSliceType == X265_TYPE_BREF)
+        return 3;
+    return sliceType;
+}
+
 double RateControl::getDiffLimitedQScale(RateControlEntry *rce, double q)
 {
     // force I/B quants as a function of P quants
@@ -1379,6 +1395,7 @@
             q += m_pbOffset;
 
         double qScale = x265_qp2qScale(q);
+        rce->qpNoVbv = q;
         double lmin = 0, lmax = 0;
         if (m_isVbv)
         {
@@ -1391,16 +1408,15 @@
                     qScale = x265_clip3(lmin, lmax, qScale);
                 q = x265_qScale2qp(qScale);
             }
-            rce->qpNoVbv = q;
             if (!m_2pass)
             {
                 qScale = clipQscale(curFrame, rce, qScale);
                 /* clip qp to permissible range after vbv-lookahead estimation to avoid possible 
                  * mispredictions by initial frame size predictors */
-                if (m_pred[m_sliceType].count == 1)
+                if (m_pred[m_predType].count == 1)
                     qScale = x265_clip3(lmin, lmax, qScale);
                 m_lastQScaleFor[m_sliceType] = qScale;
-                rce->frameSizePlanned = predictSize(&m_pred[m_sliceType], qScale, (double)m_currentSatd);
+                rce->frameSizePlanned = predictSize(&m_pred[m_predType], qScale, (double)m_currentSatd);
             }
             else
                 rce->frameSizePlanned = qScale2bits(rce, qScale);
@@ -1544,7 +1560,7 @@
             q = clipQscale(curFrame, rce, q);
             /*  clip qp to permissible range after vbv-lookahead estimation to avoid possible
              * mispredictions by initial frame size predictors */
-            if (!m_2pass && m_isVbv && m_pred[m_sliceType].count == 1)
+            if (!m_2pass && m_isVbv && m_pred[m_predType].count == 1)
                 q = x265_clip3(lqmin, lqmax, q);
         }
         m_lastQScaleFor[m_sliceType] = q;
@@ -1554,7 +1570,7 @@
         if (m_2pass && m_isVbv)
             rce->frameSizePlanned = qScale2bits(rce, q);
         else
-            rce->frameSizePlanned = predictSize(&m_pred[m_sliceType], q, (double)m_currentSatd);
+            rce->frameSizePlanned = predictSize(&m_pred[m_predType], q, (double)m_currentSatd);
 
         /* Always use up the whole VBV in this case. */
         if (m_singleFrameVbv)
@@ -1707,7 +1723,7 @@
             {
                 double frameQ[3];
                 double curBits;
-                curBits = predictSize(&m_pred[m_sliceType], q, (double)m_currentSatd);
+                curBits = predictSize(&m_pred[m_predType], q, (double)m_currentSatd);
                 double bufferFillCur = m_bufferFill - curBits;
                 double targetFill;
                 double totalDuration = m_frameDuration;
@@ -1726,7 +1742,8 @@
                         bufferFillCur += wantedFrameSize;
                     int64_t satd = curFrame->m_lowres.plannedSatd[j] >> (X265_DEPTH - 8);
                     type = IS_X265_TYPE_I(type) ? I_SLICE : IS_X265_TYPE_B(type) ? B_SLICE : P_SLICE;
-                    curBits = predictSize(&m_pred[type], frameQ[type], (double)satd);
+                    int predType = getPredictorType(curFrame->m_lowres.plannedType[j], type);
+                    curBits = predictSize(&m_pred[predType], frameQ[type], (double)satd);
                     bufferFillCur -= curBits;
                 }
 
@@ -1766,7 +1783,7 @@
             }
             // Now a hard threshold to make sure the frame fits in VBV.
             // This one is mostly for I-frames.
-            double bits = predictSize(&m_pred[m_sliceType], q, (double)m_currentSatd);
+            double bits = predictSize(&m_pred[m_predType], q, (double)m_currentSatd);
 
             // For small VBVs, allow the frame to use up the entire VBV.
             double maxFillFactor;
@@ -1783,18 +1800,21 @@
                 bits *= qf;
                 if (bits < m_bufferRate / minFillFactor)
                     q *= bits * minFillFactor / m_bufferRate;
-                bits = predictSize(&m_pred[m_sliceType], q, (double)m_currentSatd);
+                bits = predictSize(&m_pred[m_predType], q, (double)m_currentSatd);
             }
 
             q = X265_MAX(q0, q);
         }
 
         /* Apply MinCR restrictions */
-        double pbits = predictSize(&m_pred[m_sliceType], q, (double)m_currentSatd);
+        double pbits = predictSize(&m_pred[m_predType], q, (double)m_currentSatd);
         if (pbits > rce->frameSizeMaximum)
             q *= pbits / rce->frameSizeMaximum;
-
-        if (!m_isCbr || (m_isAbr && m_currentSatd >= rce->movingAvgSum && q <= q0 / 2))
+        /* To detect frames that are more complex in SATD costs compared to prev window, yet 
+         * lookahead vbv reduces its qscale by half its value. Be on safer side and avoid drastic 
+         * qscale reductions for frames high in complexity */
+        bool mispredCheck = rce->movingAvgSum && m_currentSatd >= rce->movingAvgSum && q <= q0 / 2;
+        if (!m_isCbr || (m_isAbr && mispredCheck))
             q = X265_MAX(q0, q);
 
         if (m_rateFactorMaxIncrement)
@@ -1838,18 +1858,26 @@
         if (satdCostForPendingCus  > 0)
         {
             double pred_s = predictSize(rce->rowPred[0], qScale, satdCostForPendingCus);
-            uint32_t refRowSatdCost = 0, refRowBits = 0, intraCost = 0;
+            uint32_t refRowSatdCost = 0, refRowBits = 0, intraCostForPendingCus = 0;
             double refQScale = 0;
 
             if (picType != I_SLICE)
             {
                 FrameData& refEncData = *refFrame->m_encData;
                 uint32_t endCuAddr = maxCols * (row + 1);
-                for (uint32_t cuAddr = curEncData.m_rowStat[row].numEncodedCUs + 1; cuAddr < endCuAddr; cuAddr++)
+                uint32_t startCuAddr = curEncData.m_rowStat[row].numEncodedCUs;
+                if (startCuAddr)
                 {
-                    refRowSatdCost += refEncData.m_cuStat[cuAddr].vbvCost;
-                    refRowBits += refEncData.m_cuStat[cuAddr].totalBits;
-                    intraCost += curEncData.m_cuStat[cuAddr].intraVbvCost;
+                    for (uint32_t cuAddr = startCuAddr + 1 ; cuAddr < endCuAddr; cuAddr++)
+                    {
+                        refRowSatdCost += refEncData.m_cuStat[cuAddr].vbvCost;
+                        refRowBits += refEncData.m_cuStat[cuAddr].totalBits;
+                    }
+                }
+                else
+                {
+                    refRowBits = refEncData.m_rowStat[row].encodedBits;
+                    refRowSatdCost = refEncData.m_rowStat[row].satdForVbv;
                 }
 
                 refRowSatdCost >>= X265_DEPTH - 8;
@@ -1859,7 +1887,7 @@
             if (picType == I_SLICE || qScale >= refQScale)
             {
                 if (picType == P_SLICE 
-                    && !refFrame 
+                    && refFrame 
                     && refFrame->m_encData->m_slice->m_sliceType == picType
                     && refQScale > 0
                     && refRowSatdCost > 0)
@@ -1875,8 +1903,9 @@
             }
             else if (picType == P_SLICE)
             {
+                intraCostForPendingCus = curEncData.m_rowStat[row].intraSatdForVbv - curEncData.m_rowStat[row].diagIntraSatd;
                 /* Our QP is lower than the reference! */
-                double pred_intra = predictSize(rce->rowPred[1], qScale, intraCost);
+                double pred_intra = predictSize(rce->rowPred[1], qScale, intraCostForPendingCus);
                 /* Sum: better to overestimate than underestimate by using only one of the two predictors. */
                 totalSatdBits += (int32_t)(pred_intra + pred_s);
             }
@@ -2099,8 +2128,10 @@
 
 void RateControl::updateVbv(int64_t bits, RateControlEntry* rce)
 {
+    int predType = rce->sliceType;
+    predType = rce->sliceType == B_SLICE && rce->keptAsRef ? 3 : predType;
     if (rce->lastSatd >= m_ncu)
-        updatePredictor(&m_pred[rce->sliceType], x265_qp2qScale(rce->qpaRc), (double)rce->lastSatd, (double)bits);
+        updatePredictor(&m_pred[predType], x265_qp2qScale(rce->qpaRc), (double)rce->lastSatd, (double)bits);
     if (!m_isVbv)
         return;
 
@@ -2156,23 +2187,24 @@
     {
         if (m_isVbv)
         {
+            /* determine avg QP decided by VBV rate control */
             for (uint32_t i = 0; i < slice->m_sps->numCuInHeight; i++)
                 curEncData.m_avgQpRc += curEncData.m_rowStat[i].sumQpRc;
 
             curEncData.m_avgQpRc /= slice->m_sps->numCUsInFrame;
             rce->qpaRc = curEncData.m_avgQpRc;
-
-            // copy avg RC qp to m_avgQpAq. To print out the correct qp when aq/cutree is disabled.
-            curEncData.m_avgQpAq = curEncData.m_avgQpRc;
         }
 
         if (m_param->rc.aqMode)
         {
+            /* determine actual avg encoded QP, after AQ/cutree adjustments */
             for (uint32_t i = 0; i < slice->m_sps->numCuInHeight; i++)
                 curEncData.m_avgQpAq += curEncData.m_rowStat[i].sumQpAq;
 
-            curEncData.m_avgQpAq /= slice->m_sps->numCUsInFrame;
+            curEncData.m_avgQpAq /= (slice->m_sps->numCUsInFrame * NUM_4x4_PARTITIONS);
         }
+        else
+            curEncData.m_avgQpAq = curEncData.m_avgQpRc;
     }
 
     // Write frame stats into the stats file if 2 pass is enabled.
@@ -2301,7 +2333,7 @@
 {
     m_finalFrameCount = count;
     /* unblock waiting threads */
-    m_startEndOrder.set(m_startEndOrder.get());
+    m_startEndOrder.poke();
 }
 
 /* called when the encoder is closing, and no more frames will be output.
@@ -2311,7 +2343,7 @@
 {
     m_bTerminated = true;
     /* unblock waiting threads */
-    m_startEndOrder.set(m_startEndOrder.get());
+    m_startEndOrder.poke();
 }
 
 void RateControl::destroy()

x265_1.6.tar.gz/source/encoder/ratecontrol.h -> x265_1.7.tar.gz/source/encoder/ratecontrol.h Changed

x265_1.6.tar.gz/source/encoder/rdcost.h -> x265_1.7.tar.gz/source/encoder/rdcost.h Changed

@@ -40,13 +40,15 @@
     uint32_t  m_chromaDistWeight[2];
     uint32_t  m_psyRdBase;
     uint32_t  m_psyRd;
-    int       m_qp;
+    int       m_qp; /* QP used to configure lambda, may be higher than QP_MAX_SPEC but <= QP_MAX_MAX */
 
     void setPsyRdScale(double scale)                { m_psyRdBase = (uint32_t)floor(65536.0 * scale * 0.33); }
 
     void setQP(const Slice& slice, int qp)
     {
+        x265_emms(); /* TODO: if the lambda tables were ints, this would not be necessary */
         m_qp = qp;
+        setLambda(x265_lambda2_tab[qp], x265_lambda_tab[qp]);
 
         /* Scale PSY RD factor by a slice type factor */
         static const uint32_t psyScaleFix8[3] = { 300, 256, 96 }; /* B, P, I */
@@ -60,19 +62,21 @@
         }
 
         int qpCb, qpCr;
-        setLambda(x265_lambda2_tab[qp], x265_lambda_tab[qp]);
         if (slice.m_sps->chromaFormatIdc == X265_CSP_I420)
-            qpCb = x265_clip3(QP_MIN, QP_MAX_MAX, (int)g_chromaScale[qp + slice.m_pps->chromaQpOffset[0]]);
+        {
+            qpCb = (int)g_chromaScale[x265_clip3(QP_MIN, QP_MAX_MAX, qp + slice.m_pps->chromaQpOffset[0])];
+            qpCr = (int)g_chromaScale[x265_clip3(QP_MIN, QP_MAX_MAX, qp + slice.m_pps->chromaQpOffset[1])];
+        }
         else
-            qpCb = X265_MIN(qp + slice.m_pps->chromaQpOffset[0], QP_MAX_SPEC);
+        {
+            qpCb = x265_clip3(QP_MIN, QP_MAX_SPEC, qp + slice.m_pps->chromaQpOffset[0]);
+            qpCr = x265_clip3(QP_MIN, QP_MAX_SPEC, qp + slice.m_pps->chromaQpOffset[1]);
+        }
+
         int chroma_offset_idx = X265_MIN(qp - qpCb + 12, MAX_CHROMA_LAMBDA_OFFSET);
         uint16_t lambdaOffset = m_psyRd ? x265_chroma_lambda2_offset_tab[chroma_offset_idx] : 256;
         m_chromaDistWeight[0] = lambdaOffset;
 
-        if (slice.m_sps->chromaFormatIdc == X265_CSP_I420)
-            qpCr = x265_clip3(QP_MIN, QP_MAX_MAX, (int)g_chromaScale[qp + slice.m_pps->chromaQpOffset[0]]);
-        else
-            qpCr = X265_MIN(qp + slice.m_pps->chromaQpOffset[0], QP_MAX_SPEC);
         chroma_offset_idx = X265_MIN(qp - qpCr + 12, MAX_CHROMA_LAMBDA_OFFSET);
         lambdaOffset = m_psyRd ? x265_chroma_lambda2_offset_tab[chroma_offset_idx] : 256;
         m_chromaDistWeight[1] = lambdaOffset;

x265_1.6.tar.gz/source/encoder/sao.cpp -> x265_1.7.tar.gz/source/encoder/sao.cpp Changed

@@ -258,7 +258,7 @@
     pixel* tmpL;
     pixel* tmpU;
 
-    int8_t _upBuff1[MAX_CU_SIZE + 2], *upBuff1 = _upBuff1 + 1;
+    int8_t _upBuff1[MAX_CU_SIZE + 2], *upBuff1 = _upBuff1 + 1, signLeft1[2];
     int8_t _upBufft[MAX_CU_SIZE + 2], *upBufft = _upBufft + 1;
 
     memset(_upBuff1 + MAX_CU_SIZE, 0, 2 * sizeof(int8_t)); /* avoid valgrind uninit warnings */
@@ -279,7 +279,7 @@
     {
     case SAO_EO_0: // dir: -
     {
-        pixel firstPxl = 0, lastPxl = 0;
+        pixel firstPxl = 0, lastPxl = 0, row1FirstPxl = 0, row1LastPxl = 0;
         startX = !lpelx;
         endX   = (rpelx == picWidth) ? ctuWidth - 1 : ctuWidth;
         if (ctuWidth & 15)
@@ -301,25 +301,38 @@
         }
         else
         {
-            for (y = 0; y < ctuHeight; y++)
+            for (y = 0; y < ctuHeight; y += 2)
             {
-                int signLeft = signOf(rec[startX] - tmpL[y]);
+                signLeft1[0] = signOf(rec[startX] - tmpL[y]);
+                signLeft1[1] = signOf(rec[stride + startX] - tmpL[y + 1]);
 
                 if (!lpelx)
+                {
                     firstPxl = rec[0];
+                    row1FirstPxl = rec[stride];
+                }
 
                 if (rpelx == picWidth)
+                {
                     lastPxl = rec[ctuWidth - 1];
+                    row1LastPxl = rec[stride + ctuWidth - 1];
+                }
 
-                primitives.saoCuOrgE0(rec, m_offsetEo, ctuWidth, (int8_t)signLeft);
+                primitives.saoCuOrgE0(rec, m_offsetEo, ctuWidth, signLeft1, stride);
 
                 if (!lpelx)
+                {
                     rec[0] = firstPxl;
+                    rec[stride] = row1FirstPxl;
+                }
 
                 if (rpelx == picWidth)
+                {
                     rec[ctuWidth - 1] = lastPxl;
+                    rec[stride + ctuWidth - 1] = row1LastPxl;
+                }
 
-                rec += stride;
+                rec += 2 * stride;
             }
         }
         break;
@@ -354,11 +367,14 @@
         {
             primitives.sign(upBuff1, rec, tmpU, ctuWidth);
 
-            for (y = startY; y < endY; y++)
+            int diff = (endY - startY) % 2;
+            for (y = startY; y < endY - diff; y += 2)
             {
-                primitives.saoCuOrgE1(rec, upBuff1, m_offsetEo, stride, ctuWidth);
-                rec += stride;
+                primitives.saoCuOrgE1_2Rows(rec, upBuff1, m_offsetEo, stride, ctuWidth);
+                rec += 2 * stride;
             }
+            if (diff & 1)
+                primitives.saoCuOrgE1(rec, upBuff1, m_offsetEo, stride, ctuWidth);
         }
 
         break;
@@ -421,23 +437,8 @@
             for (y = startY; y < endY; y++)
             {
                 int8_t iSignDown2 = signOf(rec[stride + startX] - tmpL[y]);
-                pixel firstPxl = rec[0];  // copy first Pxl
-                pixel lastPxl = rec[ctuWidth - 1];
-                int8_t one = upBufft[1];
-                int8_t two = upBufft[endX + 1];
 
-                primitives.saoCuOrgE2(rec, upBufft, upBuff1, m_offsetEo, ctuWidth, stride);
-                if (!lpelx)
-                {
-                    rec[0] = firstPxl;
-                    upBufft[1] = one;
-                }
-
-                if (rpelx == picWidth)
-                {
-                    rec[ctuWidth - 1] = lastPxl;
-                    upBufft[endX + 1] = two;
-                }
+                primitives.saoCuOrgE2[endX > 16](rec + startX, upBufft + startX, upBuff1 + startX, m_offsetEo, endX - startX, stride);
 
                 upBufft[startX] = iSignDown2;
 
@@ -508,7 +509,7 @@
                 upBuff1[x - 1] = -signDown;
                 rec[x] = m_clipTable[rec[x] + m_offsetEo[edgeType]];
 
-                primitives.saoCuOrgE3(rec, upBuff1, m_offsetEo, stride - 1, startX, endX);
+                primitives.saoCuOrgE3[endX > 16](rec, upBuff1, m_offsetEo, stride - 1, startX, endX);
 
                 upBuff1[endX - 1] = signOf(rec[endX - 1 + stride] - rec[endX]);
 
@@ -783,13 +784,7 @@
                 rec += stride;
             }
 
-            if (!(ctuWidth & 15))
-                primitives.sign(upBuff1, rec, &rec[- stride], ctuWidth);
-            else
-            {
-                for (x = 0; x < ctuWidth; x++)
-                    upBuff1[x] = signOf(rec[x] - rec[x - stride]);
-            }
+            primitives.sign(upBuff1, rec, &rec[- stride], ctuWidth);
 
             for (y = startY; y < endY; y++)
             {
@@ -832,8 +827,7 @@
                 rec += stride;
             }
 
-            for (x = startX; x < endX; x++)
-                upBuff1[x] = signOf(rec[x] - rec[x - stride - 1]);
+            primitives.sign(&upBuff1[startX], &rec[startX], &rec[startX - stride - 1], (endX - startX));
 
             for (y = startY; y < endY; y++)
             {
@@ -879,8 +873,7 @@
                 rec += stride;
             }
 
-            for (x = startX - 1; x < endX; x++)
-                upBuff1[x] = signOf(rec[x] - rec[x - stride + 1]);
+            primitives.sign(&upBuff1[startX - 1], &rec[startX - 1], &rec[startX - 1 - stride + 1], (endX - startX + 1));
 
             for (y = startY; y < endY; y++)
             {

x265_1.6.tar.gz/source/encoder/search.cpp -> x265_1.7.tar.gz/source/encoder/search.cpp Changed

@@ -163,11 +163,16 @@
     X265_FREE(m_tsRecon);
 }
 
-void Search::setQP(const Slice& slice, int qp)
+int Search::setLambdaFromQP(const CUData& ctu, int qp)
 {
-    x265_emms(); /* TODO: if the lambda tables were ints, this would not be necessary */
+    X265_CHECK(qp >= QP_MIN && qp <= QP_MAX_MAX, "QP used for lambda is out of range\n");
+
     m_me.setQP(qp);
-    m_rdCost.setQP(slice, qp);
+    m_rdCost.setQP(*m_slice, qp);
+
+    int quantQP = x265_clip3(QP_MIN, QP_MAX_SPEC, qp);
+    m_quant.setQPforQuant(ctu, quantQP);
+    return quantQP;
 }
 
 #if CHECKED_BUILD || _DEBUG
@@ -1185,7 +1190,7 @@
         intraMode.psyEnergy = m_rdCost.psyCost(cuGeom.log2CUSize - 2, fencYuv->m_buf[0], fencYuv->m_size, intraMode.reconYuv.m_buf[0], intraMode.reconYuv.m_size);
     }
     updateModeCost(intraMode);
-    checkDQP(cu, cuGeom);
+    checkDQP(intraMode, cuGeom);
 }
 
 /* Note that this function does not save the best intra prediction, it must
@@ -1231,16 +1236,11 @@
 
         pixel nScale[129];
         intraNeighbourBuf[1][0] = intraNeighbourBuf[0][0];
-        primitives.scale1D_128to64(nScale + 1, intraNeighbourBuf[0] + 1, 0);
+        primitives.scale1D_128to64(nScale + 1, intraNeighbourBuf[0] + 1);
 
         // we do not estimate filtering for downscaled samples
-        for (int x = 1; x < 65; x++)
-        {
-            intraNeighbourBuf[0][x] = nScale[x];           // Top pixel
-            intraNeighbourBuf[0][x + 64] = nScale[x + 64]; // Left pixel
-            intraNeighbourBuf[1][x] = nScale[x];           // Top pixel
-            intraNeighbourBuf[1][x + 64] = nScale[x + 64]; // Left pixel
-        }
+        memcpy(&intraNeighbourBuf[0][1], &nScale[1], 2 * 64 * sizeof(pixel));   // Top & Left pixels
+        memcpy(&intraNeighbourBuf[1][1], &nScale[1], 2 * 64 * sizeof(pixel));
 
         scaleTuSize = 32;
         scaleStride = 32;
@@ -1369,8 +1369,6 @@
     X265_CHECK(cu.m_partSize[0] == SIZE_2Nx2N, "encodeIntraInInter does not expect NxN intra\n");
     X265_CHECK(!m_slice->isIntra(), "encodeIntraInInter does not expect to be used in I slices\n");
 
-    m_quant.setQPforQuant(cu);
-
     uint32_t tuDepthRange[2];
     cu.getIntraTUQtDepthRange(tuDepthRange, 0);
 
@@ -1405,7 +1403,7 @@
 
     m_entropyCoder.store(intraMode.contexts);
     updateModeCost(intraMode);
-    checkDQP(intraMode.cu, cuGeom);
+    checkDQP(intraMode, cuGeom);
 }
 
 uint32_t Search::estIntraPredQT(Mode &intraMode, const CUGeom& cuGeom, const uint32_t depthRange[2], uint8_t* sharedModes)
@@ -1465,16 +1463,10 @@
 
                     pixel nScale[129];
                     intraNeighbourBuf[1][0] = intraNeighbourBuf[0][0];
-                    primitives.scale1D_128to64(nScale + 1, intraNeighbourBuf[0] + 1, 0);
+                    primitives.scale1D_128to64(nScale + 1, intraNeighbourBuf[0] + 1);
 
-                    // TO DO: primitive
-                    for (int x = 1; x < 65; x++)
-                    {
-                        intraNeighbourBuf[0][x] = nScale[x];           // Top pixel
-                        intraNeighbourBuf[0][x + 64] = nScale[x + 64]; // Left pixel
-                        intraNeighbourBuf[1][x] = nScale[x];           // Top pixel
-                        intraNeighbourBuf[1][x + 64] = nScale[x + 64]; // Left pixel
-                    }
+                    memcpy(&intraNeighbourBuf[0][1], &nScale[1], 2 * 64 * sizeof(pixel));
+                    memcpy(&intraNeighbourBuf[1][1], &nScale[1], 2 * 64 * sizeof(pixel));
 
                     scaleTuSize = 32;
                     scaleStride = 32;
@@ -1869,6 +1861,34 @@
     return outCost;
 }
 
+/* Pick between the two AMVP candidates which is the best one to use as
+ * MVP for the motion search, based on SAD cost */
+int Search::selectMVP(const CUData& cu, const PredictionUnit& pu, const MV amvp[AMVP_NUM_CANDS], int list, int ref)
+{
+    if (amvp[0] == amvp[1])
+        return 0;
+
+    Yuv& tmpPredYuv = m_rqt[cu.m_cuDepth[0]].tmpPredYuv;
+    uint32_t costs[AMVP_NUM_CANDS];
+
+    for (int i = 0; i < AMVP_NUM_CANDS; i++)
+    {
+        MV mvCand = amvp[i];
+
+        // NOTE: skip mvCand if Y is > merange and -FN>1
+        if (m_bFrameParallel && (mvCand.y >= (m_param->searchRange + 1) * 4))
+            costs[i] = m_me.COST_MAX;
+        else
+        {
+            cu.clipMv(mvCand);
+            predInterLumaPixel(pu, tmpPredYuv, *m_slice->m_refPicList[list][ref]->m_reconPic, mvCand);
+            costs[i] = m_me.bufSAD(tmpPredYuv.getLumaAddr(pu.puAbsPartIdx), tmpPredYuv.m_size);
+        }
+    }
+
+    return costs[0] <= costs[1] ? 0 : 1;
+}
+
 void Search::PME::processTasks(int workerThreadId)
 {
 #if DETAILED_CU_STATS
@@ -1899,10 +1919,10 @@
     /* Setup slave Search instance for ME for master's CU */
     if (&slave != this)
     {
-        slave.setQP(*m_slice, m_rdCost.m_qp);
         slave.m_slice = m_slice;
         slave.m_frame = m_frame;
-
+        slave.m_param = m_param;
+        slave.setLambdaFromQP(pme.mode.cu, m_rdCost.m_qp);
         slave.m_me.setSourcePU(*pme.mode.fencYuv, pme.pu.ctuAddr, pme.pu.cuAbsPartIdx, pme.pu.puAbsPartIdx, pme.pu.width, pme.pu.height);
     }
 
@@ -1910,9 +1930,9 @@
     do
     {
         if (meId < m_slice->m_numRefIdx[0])
-            slave.singleMotionEstimation(*this, pme.mode, pme.cuGeom, pme.pu, pme.puIdx, 0, meId);
+            slave.singleMotionEstimation(*this, pme.mode, pme.pu, pme.puIdx, 0, meId);
         else
-            slave.singleMotionEstimation(*this, pme.mode, pme.cuGeom, pme.pu, pme.puIdx, 1, meId - m_slice->m_numRefIdx[0]);
+            slave.singleMotionEstimation(*this, pme.mode, pme.pu, pme.puIdx, 1, meId - m_slice->m_numRefIdx[0]);
 
         meId = -1;
         pme.m_lock.acquire();
@@ -1923,55 +1943,30 @@
     while (meId >= 0);
 }
 
-void Search::singleMotionEstimation(Search& master, Mode& interMode, const CUGeom& cuGeom, const PredictionUnit& pu,
-                                    int part, int list, int ref)
+void Search::singleMotionEstimation(Search& master, Mode& interMode, const PredictionUnit& pu, int part, int list, int ref)
 {
     uint32_t bits = master.m_listSelBits[list] + MVP_IDX_BITS;
     bits += getTUBits(ref, m_slice->m_numRefIdx[list]);
 
-    MV mvc[(MD_ABOVE_LEFT + 1) * 2 + 1];
-    int numMvc = interMode.cu.getPMV(interMode.interNeighbours, list, ref, interMode.amvpCand[list][ref], mvc);
-
-    int mvpIdx = 0;
-    int merange = m_param->searchRange;
     MotionData* bestME = interMode.bestME[part];
 
-    if (interMode.amvpCand[list][ref][0] != interMode.amvpCand[list][ref][1])
-    {
-        uint32_t bestCost = MAX_INT;
-        for (int i = 0; i < AMVP_NUM_CANDS; i++)
-        {
-            MV mvCand = interMode.amvpCand[list][ref][i];
-
-            // NOTE: skip mvCand if Y is > merange and -FN>1
-            if (m_bFrameParallel && (mvCand.y >= (merange + 1) * 4))
-                continue;
-
-            interMode.cu.clipMv(mvCand);
-
-            Yuv& tmpPredYuv = m_rqt[cuGeom.depth].tmpPredYuv;
-            predInterLumaPixel(pu, tmpPredYuv, *m_slice->m_refPicList[list][ref]->m_reconPic, mvCand);
-            uint32_t cost = m_me.bufSAD(tmpPredYuv.getLumaAddr(pu.puAbsPartIdx), tmpPredYuv.m_size);
+    MV  mvc[(MD_ABOVE_LEFT + 1) * 2 + 1];
+    int numMvc = interMode.cu.getPMV(interMode.interNeighbours, list, ref, interMode.amvpCand[list][ref], mvc);
 
-            if (bestCost > cost)
-            {
-                bestCost = cost;
-                mvpIdx = i;
-            }
-        }
-    }
+    const MV* amvp = interMode.amvpCand[list][ref];
+    int mvpIdx = selectMVP(interMode.cu, pu, amvp, list, ref);
+    MV mvmin, mvmax, outmv, mvp = amvp[mvpIdx];
 
-    MV mvmin, mvmax, outmv, mvp = interMode.amvpCand[list][ref][mvpIdx];
-    setSearchRange(interMode.cu, mvp, merange, mvmin, mvmax);
+    setSearchRange(interMode.cu, mvp, m_param->searchRange, mvmin, mvmax);
 
-    int satdCost = m_me.motionEstimate(&m_slice->m_mref[list][ref], mvmin, mvmax, mvp, numMvc, mvc, merange, outmv);
+    int satdCost = m_me.motionEstimate(&m_slice->m_mref[list][ref], mvmin, mvmax, mvp, numMvc, mvc, m_param->searchRange, outmv);
 
     /* Get total cost of partition, but only include MV bit cost once */
     bits += m_me.bitcost(outmv);
     uint32_t cost = (satdCost - m_me.mvcost(outmv)) + m_rdCost.getCost(bits);
 
-    /* Refine MVP selection, updates: mvp, mvpIdx, bits, cost */
-    checkBestMVP(interMode.amvpCand[list][ref], outmv, mvp, mvpIdx, bits, cost);
+    /* Refine MVP selection, updates: mvpIdx, bits, cost */
+    mvp = checkBestMVP(amvp, outmv, mvpIdx, bits, cost);
 
     /* tie goes to the smallest ref ID, just like --no-pme */
     ScopedLock _lock(master.m_meLock);
@@ -1988,7 +1983,7 @@
 }
 
 /* find the best inter prediction for each PU of specified mode */
-void Search::predInterSearch(Mode& interMode, const CUGeom& cuGeom, bool bMergeOnly, bool bChromaSA8D)
+void Search::predInterSearch(Mode& interMode, const CUGeom& cuGeom, bool bChromaMC)
 {
     ProfileCUScope(interMode.cu, motionEstimationElapsedTime, countMotionEstimate);
 
@@ -2009,7 +2004,6 @@
     Yuv&     tmpPredYuv = m_rqt[cuGeom.depth].tmpPredYuv;
 
     MergeData merge;
-    uint32_t mrgCost;
     memset(&merge, 0, sizeof(merge));
 
     for (int puIdx = 0; puIdx < numPart; puIdx++)
@@ -2020,27 +2014,7 @@
         m_me.setSourcePU(*interMode.fencYuv, pu.ctuAddr, pu.cuAbsPartIdx, pu.puAbsPartIdx, pu.width, pu.height);
 
         /* find best cost merge candidate. note: 2Nx2N merge and bidir are handled as separate modes */
-        if (cu.m_partSize[0] != SIZE_2Nx2N)
-        {
-            mrgCost = mergeEstimation(cu, cuGeom, pu, puIdx, merge);
-
-            if (bMergeOnly && mrgCost != MAX_UINT)
-            {
-                cu.m_mergeFlag[pu.puAbsPartIdx] = true;
-                cu.m_mvpIdx[0][pu.puAbsPartIdx] = merge.index; // merge candidate ID is stored in L0 MVP idx
-                cu.setPUInterDir(merge.dir, pu.puAbsPartIdx, puIdx);
-                cu.setPUMv(0, merge.mvField[0].mv, pu.puAbsPartIdx, puIdx);
-                cu.setPURefIdx(0, merge.mvField[0].refIdx, pu.puAbsPartIdx, puIdx);
-                cu.setPUMv(1, merge.mvField[1].mv, pu.puAbsPartIdx, puIdx);
-                cu.setPURefIdx(1, merge.mvField[1].refIdx, pu.puAbsPartIdx, puIdx);
-                totalmebits += merge.bits;
-
-                motionCompensation(cu, pu, *predYuv, true, bChromaSA8D);
-                continue;
-            }
-        }
-        else
-            mrgCost = MAX_UINT;
+        uint32_t mrgCost = numPart == 1 ? MAX_UINT : mergeEstimation(cu, cuGeom, pu, puIdx, merge);
 
         bestME[0].cost = MAX_UINT;
         bestME[1].cost = MAX_UINT;
@@ -2061,45 +2035,19 @@
 
                 int numMvc = cu.getPMV(interMode.interNeighbours, list, ref, interMode.amvpCand[list][ref], mvc);
 
-                // Pick the best possible MVP from AMVP candidates based on least residual
-                int mvpIdx = 0;
-                int merange = m_param->searchRange;
+                const MV* amvp = interMode.amvpCand[list][ref];
+                int mvpIdx = selectMVP(cu, pu, amvp, list, ref);
+                MV mvmin, mvmax, outmv, mvp = amvp[mvpIdx];
 
-                if (interMode.amvpCand[list][ref][0] != interMode.amvpCand[list][ref][1])
-                {
-                    uint32_t bestCost = MAX_INT;
-                    for (int i = 0; i < AMVP_NUM_CANDS; i++)
-                    {
-                        MV mvCand = interMode.amvpCand[list][ref][i];
-
-                        // NOTE: skip mvCand if Y is > merange and -FN>1
-                        if (m_bFrameParallel && (mvCand.y >= (merange + 1) * 4))
-                            continue;
-
-                        cu.clipMv(mvCand);
-                        predInterLumaPixel(pu, tmpPredYuv, *slice->m_refPicList[list][ref]->m_reconPic, mvCand);
-                        uint32_t cost = m_me.bufSAD(tmpPredYuv.getLumaAddr(pu.puAbsPartIdx), tmpPredYuv.m_size);
-
-                        if (bestCost > cost)
-                        {
-                            bestCost = cost;
-                            mvpIdx = i;
-                        }
-                    }
-                }
-
-                MV mvmin, mvmax, outmv, mvp = interMode.amvpCand[list][ref][mvpIdx];
-
-                int satdCost;
-                setSearchRange(cu, mvp, merange, mvmin, mvmax);
-                satdCost = m_me.motionEstimate(&slice->m_mref[list][ref], mvmin, mvmax, mvp, numMvc, mvc, merange, outmv);
+                setSearchRange(cu, mvp, m_param->searchRange, mvmin, mvmax);
+                int satdCost = m_me.motionEstimate(&slice->m_mref[list][ref], mvmin, mvmax, mvp, numMvc, mvc, m_param->searchRange, outmv);
 
                 /* Get total cost of partition, but only include MV bit cost once */
                 bits += m_me.bitcost(outmv);
                 uint32_t cost = (satdCost - m_me.mvcost(outmv)) + m_rdCost.getCost(bits);
 
-                /* Refine MVP selection, updates: mvp, mvpIdx, bits, cost */
-                checkBestMVP(interMode.amvpCand[list][ref], outmv, mvp, mvpIdx, bits, cost);
+                /* Refine MVP selection, updates: mvpIdx, bits, cost */
+                mvp = checkBestMVP(amvp, outmv, mvpIdx, bits, cost);
 
                 if (cost < bestME[list].cost)
                 {
@@ -2122,7 +2070,7 @@
             {
                 processPME(pme, *this);
 
-                singleMotionEstimation(*this, interMode, cuGeom, pu, puIdx, 0, 0); /* L0-0 */
+                singleMotionEstimation(*this, interMode, pu, puIdx, 0, 0); /* L0-0 */
 
                 bDoUnidir = false;
 
@@ -2144,44 +2092,19 @@
 
                     int numMvc = cu.getPMV(interMode.interNeighbours, list, ref, interMode.amvpCand[list][ref], mvc);
 
-                    // Pick the best possible MVP from AMVP candidates based on least residual
-                    int mvpIdx = 0;
-                    int merange = m_param->searchRange;
-
-                    if (interMode.amvpCand[list][ref][0] != interMode.amvpCand[list][ref][1])
-                    {
-                        uint32_t bestCost = MAX_INT;
-                        for (int i = 0; i < AMVP_NUM_CANDS; i++)
-                        {
-                            MV mvCand = interMode.amvpCand[list][ref][i];
-
-                            // NOTE: skip mvCand if Y is > merange and -FN>1
-                            if (m_bFrameParallel && (mvCand.y >= (merange + 1) * 4))
-                                continue;
+                    const MV* amvp = interMode.amvpCand[list][ref];
+                    int mvpIdx = selectMVP(cu, pu, amvp, list, ref);
+                    MV mvmin, mvmax, outmv, mvp = amvp[mvpIdx];
 
-                            cu.clipMv(mvCand);
-                            predInterLumaPixel(pu, tmpPredYuv, *slice->m_refPicList[list][ref]->m_reconPic, mvCand);
-                            uint32_t cost = m_me.bufSAD(tmpPredYuv.getLumaAddr(pu.puAbsPartIdx), tmpPredYuv.m_size);
-
-                            if (bestCost > cost)
-                            {
-                                bestCost = cost;
-                                mvpIdx = i;
-                            }
-                        }
-                    }
-
-                    MV mvmin, mvmax, outmv, mvp = interMode.amvpCand[list][ref][mvpIdx];
-
-                    setSearchRange(cu, mvp, merange, mvmin, mvmax);
-                    int satdCost = m_me.motionEstimate(&slice->m_mref[list][ref], mvmin, mvmax, mvp, numMvc, mvc, merange, outmv);
+                    setSearchRange(cu, mvp, m_param->searchRange, mvmin, mvmax);
+                    int satdCost = m_me.motionEstimate(&slice->m_mref[list][ref], mvmin, mvmax, mvp, numMvc, mvc, m_param->searchRange, outmv);
 
                     /* Get total cost of partition, but only include MV bit cost once */
                     bits += m_me.bitcost(outmv);
                     uint32_t cost = (satdCost - m_me.mvcost(outmv)) + m_rdCost.getCost(bits);
 
-                    /* Refine MVP selection, updates: mvp, mvpIdx, bits, cost */
-                    checkBestMVP(interMode.amvpCand[list][ref], outmv, mvp, mvpIdx, bits, cost);
+                    /* Refine MVP selection, updates: mvpIdx, bits, cost */
+                    mvp = checkBestMVP(amvp, outmv, mvpIdx, bits, cost);
 
                     if (cost < bestME[list].cost)
                     {
@@ -2289,8 +2212,8 @@
                 uint32_t cost = satdCost + m_rdCost.getCost(bits0) + m_rdCost.getCost(bits1);
 
                 /* refine MVP selection for zero mv, updates: mvp, mvpidx, bits, cost */
-                checkBestMVP(interMode.amvpCand[0][bestME[0].ref], mvzero, mvp0, mvpIdx0, bits0, cost);
-                checkBestMVP(interMode.amvpCand[1][bestME[1].ref], mvzero, mvp1, mvpIdx1, bits1, cost);
+                mvp0 = checkBestMVP(interMode.amvpCand[0][bestME[0].ref], mvzero, mvpIdx0, bits0, cost);
+                mvp1 = checkBestMVP(interMode.amvpCand[1][bestME[1].ref], mvzero, mvpIdx1, bits1, cost);
 
                 if (cost < bidirCost)
                 {
@@ -2370,7 +2293,7 @@
             totalmebits += bestME[1].bits;
         }
 
-        motionCompensation(cu, pu, *predYuv, true, bChromaSA8D);
+        motionCompensation(cu, pu, *predYuv, true, bChromaMC);
     }
     X265_CHECK(interMode.ok(), "inter mode is not ok");
     interMode.sa8dBits += totalmebits;
@@ -2429,27 +2352,21 @@
 }
 
 /* Check if using an alternative MVP would result in a smaller MVD + signal bits */
-void Search::checkBestMVP(MV* amvpCand, MV mv, MV& mvPred, int& outMvpIdx, uint32_t& outBits, uint32_t& outCost) const
+const MV& Search::checkBestMVP(const MV* amvpCand, const MV& mv, int& mvpIdx, uint32_t& outBits, uint32_t& outCost) const
 {
-    X265_CHECK(amvpCand[outMvpIdx] == mvPred, "checkBestMVP: unexpected mvPred\n");
-
-    int mvpIdx = !outMvpIdx;
-    MV mvp = amvpCand[mvpIdx];
-    int diffBits = m_me.bitcost(mv, mvp) - m_me.bitcost(mv, mvPred);
+    int diffBits = m_me.bitcost(mv, amvpCand[!mvpIdx]) - m_me.bitcost(mv, amvpCand[mvpIdx]);
     if (diffBits < 0)
     {
-        outMvpIdx = mvpIdx;
-        mvPred = mvp;
+        mvpIdx = !mvpIdx;
         uint32_t origOutBits = outBits;
         outBits = origOutBits + diffBits;
         outCost = (outCost - m_rdCost.getCost(origOutBits)) + m_rdCost.getCost(outBits);
     }
+    return amvpCand[mvpIdx];
 }
 
-void Search::setSearchRange(const CUData& cu, MV mvp, int merange, MV& mvmin, MV& mvmax) const
+void Search::setSearchRange(const CUData& cu, const MV& mvp, int merange, MV& mvmin, MV& mvmax) const
 {
-    cu.clipMv(mvp);
-
     MV dist((int16_t)merange << 2, (int16_t)merange << 2);
     mvmin = mvp - dist;
     mvmax = mvp + dist;
@@ -2534,9 +2451,6 @@
     uint32_t log2CUSize = cuGeom.log2CUSize;
     int sizeIdx = log2CUSize - 2;
 
-    uint32_t tqBypass = cu.m_tqBypass[0];
-    m_quant.setQPforQuant(interMode.cu);
-
     resiYuv->subtract(*fencYuv, *predYuv, log2CUSize);
 
     uint32_t tuDepthRange[2];
@@ -2547,6 +2461,7 @@
     Cost costs;
     estimateResidualQT(interMode, cuGeom, 0, 0, *resiYuv, costs, tuDepthRange);
 
+    uint32_t tqBypass = cu.m_tqBypass[0];
     if (!tqBypass)
     {
         uint32_t cbf0Dist = primitives.cu[sizeIdx].sse_pp(fencYuv->m_buf[0], fencYuv->m_size, predYuv->m_buf[0], predYuv->m_size);
@@ -2631,7 +2546,7 @@
     interMode.coeffBits = coeffBits;
     interMode.mvBits = bits - coeffBits;
     updateModeCost(interMode);
-    checkDQP(interMode.cu, cuGeom);
+    checkDQP(interMode, cuGeom);
 }
 
 void Search::residualTransformQuantInter(Mode& mode, const CUGeom& cuGeom, uint32_t absPartIdx, uint32_t tuDepth, const uint32_t depthRange[2])
@@ -3448,22 +3363,43 @@
     }
 }
 
-void Search::checkDQP(CUData& cu, const CUGeom& cuGeom)
+void Search::checkDQP(Mode& mode, const CUGeom& cuGeom)
 {
+    CUData& cu = mode.cu;
     if (cu.m_slice->m_pps->bUseDQP && cuGeom.depth <= cu.m_slice->m_pps->maxCuDQPDepth)
     {
         if (cu.getQtRootCbf(0))
         {
-            /* When analysing RDO with DQP bits, the entropy encoder should add the cost of DQP bits here
-             * i.e Encode QP */
+            if (m_param->rdLevel >= 3)
+            {
+                mode.contexts.resetBits();
+                mode.contexts.codeDeltaQP(cu, 0);
+                uint32_t bits = mode.contexts.getNumberOfWrittenBits();
+                mode.mvBits += bits;
+                mode.totalBits += bits;
+                updateModeCost(mode);
+            }
+            else if (m_param->rdLevel <= 1)
+            {
+                mode.sa8dBits++;
+                mode.sa8dCost = m_rdCost.calcRdSADCost(mode.distortion, mode.sa8dBits);
+            }
+            else
+            {
+                mode.mvBits++;
+                mode.totalBits++;
+                updateModeCost(mode);
+            }
         }
         else
             cu.setQPSubParts(cu.getRefQP(0), 0, cuGeom.depth);
     }
 }
 
-void Search::checkDQPForSplitPred(CUData& cu, const CUGeom& cuGeom)
+void Search::checkDQPForSplitPred(Mode& mode, const CUGeom& cuGeom)
 {
+    CUData& cu = mode.cu;
+
     if ((cuGeom.depth == cu.m_slice->m_pps->maxCuDQPDepth) && cu.m_slice->m_pps->bUseDQP)
     {
         bool hasResidual = false;
@@ -3478,10 +3414,31 @@
             }
         }
         if (hasResidual)
-            /* TODO: Encode QP, and recalculate RD cost of splitPred */
+        {
+            if (m_param->rdLevel >= 3)
+            {
+                mode.contexts.resetBits();
+                mode.contexts.codeDeltaQP(cu, 0);
+                uint32_t bits = mode.contexts.getNumberOfWrittenBits();
+                mode.mvBits += bits;
+                mode.totalBits += bits;
+                updateModeCost(mode);
+            }
+            else if (m_param->rdLevel <= 1)
+            {
+                mode.sa8dBits++;
+                mode.sa8dCost = m_rdCost.calcRdSADCost(mode.distortion, mode.sa8dBits);
+            }
+            else
+            {
+                mode.mvBits++;
+                mode.totalBits++;
+                updateModeCost(mode);
+            }
             /* For all zero CBF sub-CUs, reset QP to RefQP (so that deltaQP is not signalled).
             When the non-zero CBF sub-CU is found, stop */
             cu.setQPSubCUs(cu.getRefQP(0), 0, cuGeom.depth);
+        }
         else
             /* No residual within this CU or subCU, so reset QP to RefQP */
             cu.setQPSubParts(cu.getRefQP(0), 0, cuGeom.depth);

x265_1.6.tar.gz/source/encoder/search.h -> x265_1.7.tar.gz/source/encoder/search.h Changed

@@ -287,7 +287,7 @@
     ~Search();
 
     bool     initSearch(const x265_param& param, ScalingList& scalingList);
-    void     setQP(const Slice& slice, int qp);
+    int      setLambdaFromQP(const CUData& ctu, int qp); /* returns real quant QP in valid spec range */
 
     // mark temp RD entropy contexts as uninitialized; useful for finding loads without stores
     void     invalidateContexts(int fromDepth);
@@ -301,7 +301,7 @@
     void     encodeIntraInInter(Mode& intraMode, const CUGeom& cuGeom);
 
     // estimation inter prediction (non-skip)
-    void     predInterSearch(Mode& interMode, const CUGeom& cuGeom, bool bMergeOnly, bool bChroma);
+    void     predInterSearch(Mode& interMode, const CUGeom& cuGeom, bool bChromaMC);
 
     // encode residual and compute rd-cost for inter mode
     void     encodeResAndCalcRdInterCU(Mode& interMode, const CUGeom& cuGeom);
@@ -316,8 +316,8 @@
     void     getBestIntraModeChroma(Mode& intraMode, const CUGeom& cuGeom);
 
     /* update CBF flags and QP values to be internally consistent */
-    void checkDQP(CUData& cu, const CUGeom& cuGeom);
-    void checkDQPForSplitPred(CUData& cu, const CUGeom& cuGeom);
+    void checkDQP(Mode& mode, const CUGeom& cuGeom);
+    void checkDQPForSplitPred(Mode& mode, const CUGeom& cuGeom);
 
     class PME : public BondedTaskGroup
     {
@@ -339,7 +339,7 @@
     };
 
     void     processPME(PME& pme, Search& slave);
-    void     singleMotionEstimation(Search& master, Mode& interMode, const CUGeom& cuGeom, const PredictionUnit& pu, int part, int list, int ref);
+    void     singleMotionEstimation(Search& master, Mode& interMode, const PredictionUnit& pu, int part, int list, int ref);
 
 protected:
 
@@ -396,8 +396,9 @@
     };
 
     /* inter/ME helper functions */
-    void     checkBestMVP(MV* amvpCand, MV cMv, MV& mvPred, int& mvpIdx, uint32_t& outBits, uint32_t& outCost) const;
-    void     setSearchRange(const CUData& cu, MV mvp, int merange, MV& mvmin, MV& mvmax) const;
+    int       selectMVP(const CUData& cu, const PredictionUnit& pu, const MV amvp[AMVP_NUM_CANDS], int list, int ref);
+    const MV& checkBestMVP(const MV amvpCand[2], const MV& mv, int& mvpIdx, uint32_t& outBits, uint32_t& outCost) const;
+    void     setSearchRange(const CUData& cu, const MV& mvp, int merange, MV& mvmin, MV& mvmax) const;
     uint32_t mergeEstimation(CUData& cu, const CUGeom& cuGeom, const PredictionUnit& pu, int puIdx, MergeData& m);
     static void getBlkBits(PartSize cuMode, bool bPSlice, int puIdx, uint32_t lastMode, uint32_t blockBit[3]);

x265_1.6.tar.gz/source/encoder/sei.h -> x265_1.7.tar.gz/source/encoder/sei.h Changed

@@ -71,6 +71,8 @@
         DECODED_PICTURE_HASH                 = 132,
         SCALABLE_NESTING                     = 133,
         REGION_REFRESH_INFO                  = 134,
+        MASTERING_DISPLAY_INFO               = 137,
+        CONTENT_LIGHT_LEVEL_INFO             = 144,
     };
 
     virtual PayloadType payloadType() const = 0;
@@ -111,6 +113,73 @@
     }
 };
 
+class SEIMasteringDisplayColorVolume : public SEI
+{
+public:
+
+    uint16_t displayPrimaryX[3];
+    uint16_t displayPrimaryY[3];
+    uint16_t whitePointX, whitePointY;
+    uint32_t maxDisplayMasteringLuminance;
+    uint32_t minDisplayMasteringLuminance;
+
+    PayloadType payloadType() const { return MASTERING_DISPLAY_INFO; }
+
+    bool parse(const char* value)
+    {
+        return sscanf(value, "G(%hu,%hu)B(%hu,%hu)R(%hu,%hu)WP(%hu,%hu)L(%u,%u)",
+                      &displayPrimaryX[0], &displayPrimaryY[0],
+                      &displayPrimaryX[1], &displayPrimaryY[1],
+                      &displayPrimaryX[2], &displayPrimaryY[2],
+                      &whitePointX, &whitePointY,
+                      &maxDisplayMasteringLuminance, &minDisplayMasteringLuminance) == 10;
+    }
+
+    void write(Bitstream& bs, const SPS&)
+    {
+        m_bitIf = &bs;
+
+        WRITE_CODE(MASTERING_DISPLAY_INFO, 8, "payload_type");
+        WRITE_CODE(8 * 2 + 2 * 4, 8, "payload_size");
+
+        for (uint32_t i = 0; i < 3; i++)
+        {
+            WRITE_CODE(displayPrimaryX[i], 16, "display_primaries_x[ c ]");
+            WRITE_CODE(displayPrimaryY[i], 16, "display_primaries_y[ c ]");
+        }
+        WRITE_CODE(whitePointX, 16, "white_point_x");
+        WRITE_CODE(whitePointY, 16, "white_point_y");
+        WRITE_CODE(maxDisplayMasteringLuminance, 32, "max_display_mastering_luminance");
+        WRITE_CODE(minDisplayMasteringLuminance, 32, "min_display_mastering_luminance");
+    }
+};
+
+class SEIContentLightLevel : public SEI
+{
+public:
+
+    uint16_t max_content_light_level;
+    uint16_t max_pic_average_light_level;
+
+    PayloadType payloadType() const { return CONTENT_LIGHT_LEVEL_INFO; }
+
+    bool parse(const char* value)
+    {
+        return sscanf(value, "%hu,%hu",
+                      &max_content_light_level, &max_pic_average_light_level) == 2;
+    }
+
+    void write(Bitstream& bs, const SPS&)
+    {
+        m_bitIf = &bs;
+
+        WRITE_CODE(CONTENT_LIGHT_LEVEL_INFO, 8, "payload_type");
+        WRITE_CODE(4, 8, "payload_size");
+        WRITE_CODE(max_content_light_level,     16, "max_content_light_level");
+        WRITE_CODE(max_pic_average_light_level, 16, "max_pic_average_light_level");
+    }
+};
+
 class SEIDecodedPictureHash : public SEI
 {
 public:

x265_1.6.tar.gz/source/encoder/slicetype.cpp -> x265_1.7.tar.gz/source/encoder/slicetype.cpp Changed

@@ -44,23 +44,6 @@
 
 namespace {
 
-inline int16_t median(int16_t a, int16_t b, int16_t c)
-{
-    int16_t t = (a - b) & ((a - b) >> 31);
-
-    a -= t;
-    b += t;
-    b -= (b - c) & ((b - c) >> 31);
-    b += (a - b) & ((a - b) >> 31);
-    return b;
-}
-
-inline void median_mv(MV &dst, MV a, MV b, MV c)
-{
-    dst.x = median(a.x, b.x, c.x);
-    dst.y = median(a.y, b.y, c.y);
-}
-
 /* Compute variance to derive AC energy of each block */
 inline uint32_t acEnergyVar(Frame *curFrame, uint64_t sum_ssd, int shift, int plane)
 {
@@ -492,8 +475,6 @@
     m_8x8Blocks = m_8x8Width > 2 && m_8x8Height > 2 ? (m_8x8Width - 2) * (m_8x8Height - 2) : m_8x8Width * m_8x8Height;
 
     m_lastKeyframe = -m_param->keyframeMax;
-    memset(m_preframes, 0, sizeof(m_preframes));
-    m_preTotal = m_preAcquired = m_preCompleted = 0;
     m_sliceTypeBusy = false;
     m_fullQueueSize = X265_MAX(1, m_param->lookaheadDepth);
     m_bAdaptiveQuant = m_param->rc.aqMode || m_param->bEnableWeightedPred || m_param->bEnableWeightedBiPred;
@@ -568,14 +549,14 @@
     return m_tld && m_scratch;
 }
 
-void Lookahead::stop()
+void Lookahead::stopJobs()
 {
     if (m_pool && !m_inputQueue.empty())
     {
-        m_preLookaheadLock.acquire();
+        m_inputLock.acquire();
         m_isActive = false;
         bool wait = m_outputSignalRequired = m_sliceTypeBusy;
-        m_preLookaheadLock.release();
+        m_inputLock.release();
 
         if (wait)
             m_outputSignal.wait();
@@ -634,19 +615,11 @@
             m_filled = true; /* full capacity plus mini-gop lag */
     }
 
-    m_preLookaheadLock.acquire();
-
     m_inputLock.acquire();
     m_inputQueue.pushBack(curFrame);
-    m_inputLock.release();
-
-    m_preframes[m_preTotal++] = &curFrame;
-    X265_CHECK(m_preTotal <= X265_LOOKAHEAD_MAX, "prelookahead overflow\n");
-    
-    m_preLookaheadLock.release();
-
-    if (m_pool)
+    if (m_pool && m_inputQueue.size() >= m_fullQueueSize)
         tryWakeOne();
+    m_inputLock.release();
 }
 
 /* Called by API thread */
@@ -657,74 +630,33 @@
     m_filled = true;
 }
 
-void Lookahead::findJob(int workerThreadID)
+void Lookahead::findJob(int /*workerThreadID*/)
 {
-    Frame* preFrame;
-    bool   doDecide;
-
-    if (!m_isActive)
-        return;
-
-    int tld = workerThreadID;
-    if (workerThreadID < 0)
-        tld = m_pool ? m_pool->m_numWorkers : 0;
+    bool doDecide;
 
-    m_preLookaheadLock.acquire();
-    do
-    {
-        preFrame = NULL;
-        doDecide = false;
+    m_inputLock.acquire();
+    if (m_inputQueue.size() >= m_fullQueueSize && !m_sliceTypeBusy && m_isActive)
+        doDecide = m_sliceTypeBusy = true;
+    else
+        doDecide = m_helpWanted = false;
+    m_inputLock.release();
 
-        if (m_preTotal > m_preAcquired)
-            preFrame = m_preframes[m_preAcquired++];
-        else
-        {
-            if (m_preTotal == m_preCompleted)
-                m_preAcquired = m_preTotal = m_preCompleted = 0;
-
-            /* the worker thread that performs the last pre-lookahead will generally get to run
-             * slicetypeDecide() */
-            m_inputLock.acquire();
-            if (!m_sliceTypeBusy && !m_preTotal && m_inputQueue.size() >= m_fullQueueSize && m_isActive)
-                doDecide = m_sliceTypeBusy = true;
-            else
-                m_helpWanted = false;
-            m_inputLock.release();
-        }
-        m_preLookaheadLock.release();
+    if (!doDecide)
+        return;
 
-        if (preFrame)
-        {
-            ProfileLookaheadTime(m_preLookaheadElapsedTime, m_countPreLookahead);
-            ProfileScopeEvent(prelookahead);
-
-            preFrame->m_lowres.init(preFrame->m_fencPic, preFrame->m_poc);
-            if (m_param->rc.bStatRead && m_param->rc.cuTree && IS_REFERENCED(preFrame))
-                /* cu-tree offsets were read from stats file */;
-            else if (m_bAdaptiveQuant)
-                m_tld[tld].calcAdaptiveQuantFrame(preFrame, m_param);
-            m_tld[tld].lowresIntraEstimate(preFrame->m_lowres);
-
-            m_preLookaheadLock.acquire(); /* re-acquire for next pass */
-            m_preCompleted++;
-        }
-        else if (doDecide)
-        {
-            ProfileLookaheadTime(m_slicetypeDecideElapsedTime, m_countSlicetypeDecide);
-            ProfileScopeEvent(slicetypeDecideEV);
+    ProfileLookaheadTime(m_slicetypeDecideElapsedTime, m_countSlicetypeDecide);
+    ProfileScopeEvent(slicetypeDecideEV);
 
-            slicetypeDecide();
+    slicetypeDecide();
 
-            m_preLookaheadLock.acquire(); /* re-acquire for next pass */
-            if (m_outputSignalRequired)
-            {
-                m_outputSignal.trigger();
-                m_outputSignalRequired = false;
-            }
-            m_sliceTypeBusy = false;
-        }
+    m_inputLock.acquire();
+    if (m_outputSignalRequired)
+    {
+        m_outputSignal.trigger();
+        m_outputSignalRequired = false;
     }
-    while (preFrame || doDecide);
+    m_sliceTypeBusy = false;
+    m_inputLock.release();
 }
 
 /* Called by API thread */
@@ -739,13 +671,11 @@
         if (out)
             return out;
 
-        /* process all pending pre-lookahead frames and run slicetypeDecide() if
-         * necessary */
-        findJob(-1);
+        findJob(-1); /* run slicetypeDecide() if necessary */
 
-        m_preLookaheadLock.acquire();
-        bool wait = m_outputSignalRequired = m_sliceTypeBusy || m_preTotal;
-        m_preLookaheadLock.release();
+        m_inputLock.acquire();
+        bool wait = m_outputSignalRequired = m_sliceTypeBusy;
+        m_inputLock.release();
 
         if (wait)
             m_outputSignal.wait();
@@ -809,7 +739,7 @@
     {
         /* aggregate lowres row satds to CTU resolution */
         curFrame->m_lowres.lowresCostForRc = curFrame->m_lowres.lowresCosts[b - p0][p1 - b];
-        uint32_t lowresRow = 0, lowresCol = 0, lowresCuIdx = 0, sum = 0;
+        uint32_t lowresRow = 0, lowresCol = 0, lowresCuIdx = 0, sum = 0, intraSum = 0;
         uint32_t scale = m_param->maxCUSize / (2 * X265_LOWRES_CU_SIZE);
         uint32_t numCuInHeight = (m_param->sourceHeight + g_maxCUSize - 1) / g_maxCUSize;
         uint32_t widthInLowresCu = (uint32_t)m_8x8Width, heightInLowresCu = (uint32_t)m_8x8Height;
@@ -823,7 +753,7 @@
             lowresRow = row * scale;
             for (uint32_t cnt = 0; cnt < scale && lowresRow < heightInLowresCu; lowresRow++, cnt++)
             {
-                sum = 0;
+                sum = 0; intraSum = 0;
                 lowresCuIdx = lowresRow * widthInLowresCu;
                 for (lowresCol = 0; lowresCol < widthInLowresCu; lowresCol++, lowresCuIdx++)
                 {
@@ -836,24 +766,57 @@
                     }
                     curFrame->m_lowres.lowresCostForRc[lowresCuIdx] = lowresCuCost;
                     sum += lowresCuCost;
+                    intraSum += curFrame->m_lowres.intraCost[lowresCuIdx];
                 }
                 curFrame->m_encData->m_rowStat[row].satdForVbv += sum;
+                curFrame->m_encData->m_rowStat[row].intraSatdForVbv += intraSum;
             }
         }
     }
 }
 
+void PreLookaheadGroup::processTasks(int workerThreadID)
+{
+    if (workerThreadID < 0)
+        workerThreadID = m_lookahead.m_pool ? m_lookahead.m_pool->m_numWorkers : 0;
+    LookaheadTLD& tld = m_lookahead.m_tld[workerThreadID];
+
+    m_lock.acquire();
+    while (m_jobAcquired < m_jobTotal)
+    {
+        Frame* preFrame = m_preframes[m_jobAcquired++];
+        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)
+            tld.calcAdaptiveQuantFrame(preFrame, m_lookahead.m_param);
+        tld.lowresIntraEstimate(preFrame->m_lowres);
+        preFrame->m_lowresInit = true;
+
+        m_lock.acquire();
+    }
+    m_lock.release();
+}
+
 /* called by API thread or worker thread with inputQueueLock acquired */
 void Lookahead::slicetypeDecide()
 {
-    Lowres *frames[X265_LOOKAHEAD_MAX];
-    Frame *list[X265_LOOKAHEAD_MAX];
-    int maxSearch = X265_MIN(m_param->lookaheadDepth, X265_LOOKAHEAD_MAX);
+    PreLookaheadGroup pre(*this);
 
+    Lowres* frames[X265_LOOKAHEAD_MAX + X265_BFRAME_MAX + 4];
+    Frame*  list[X265_BFRAME_MAX + 4];
     memset(frames, 0, sizeof(frames));
     memset(list, 0, sizeof(list));
+    int maxSearch = X265_MIN(m_param->lookaheadDepth, X265_LOOKAHEAD_MAX);
+    maxSearch = X265_MAX(1, maxSearch);
+
     {
         ScopedLock lock(m_inputLock);
+
         Frame *curFrame = m_inputQueue.first();
         int j;
         for (j = 0; j < m_param->bframes + 2; j++)
@@ -869,13 +832,25 @@
         {
             if (!curFrame) break;
             frames[j + 1] = &curFrame->m_lowres;
-            X265_CHECK(curFrame->m_lowres.costEst[0][0] > 0, "prelookahead not completed for input picture\n");
+
+            if (!curFrame->m_lowresInit)
+                pre.m_preframes[pre.m_jobTotal++] = curFrame;
+
             curFrame = curFrame->m_next;
         }
 
         maxSearch = j;
     }
 
+    /* perform pre-analysis on frames which need it, using a bonded task group */
+    if (pre.m_jobTotal)
+    {
+        if (m_pool)
+            pre.tryBondPeers(*m_pool, pre.m_jobTotal);
+        pre.processTasks(-1);
+        pre.waitForExit();
+    }
+
     if (m_lastNonB && !m_param->rc.bStatRead &&
         ((m_param->bFrameAdaptive && m_param->bframes) ||
          m_param->rc.cuTree || m_param->scenecutThreshold ||
@@ -2038,12 +2013,10 @@
 
         int numc = 0;
         MV mvc[4], mvp;
-
         MV* fencMV = &fenc->lowresMvs[i][listDist[i]][cuXY];
+        ReferencePlanes* fref = i ? fref1 : wfref0;
 
         /* Reverse-order MV prediction */
-        mvc[0] = 0;
-        mvc[2] = 0;
 #define MVC(mv) mvc[numc++] = mv;
         if (cuX < widthInCU - 1)
             MVC(fencMV[1]);
@@ -2056,12 +2029,29 @@
                 MVC(fencMV[widthInCU + 1]);
         }
 #undef MVC
-        if (numc <= 1)
-            mvp = mvc[0];
+
+        if (!numc)
+            mvp = 0;
         else
-            median_mv(mvp, mvc[0], mvc[1], mvc[2]);
+        {
+            ALIGN_VAR_32(pixel, subpelbuf[X265_LOWRES_CU_SIZE * X265_LOWRES_CU_SIZE]);
+            int mvpcost = MotionEstimate::COST_MAX;
+
+            /* measure SATD cost of each neighbor MV (estimating merge analysis)
+             * and use the lowest cost MV as MVP (estimating AMVP). Since all
+             * mvc[] candidates are measured here, none are passed to motionEstimate */
+            for (int idx = 0; idx < numc; idx++)
+            {
+                intptr_t stride = X265_LOWRES_CU_SIZE;
+                pixel *src = fref->lowresMC(pelOffset, mvc[idx], subpelbuf, stride);
+                int cost = tld.me.bufSATD(src, stride);
+                COPY2_IF_LT(mvpcost, cost, mvp, mvc[idx]);
+            }
+        }
 
-        fencCost = tld.me.motionEstimate(i ? fref1 : wfref0, mvmin, mvmax, mvp, numc, mvc, s_merange, *fencMV);
+        /* ME will never return a cost larger than the cost @MVP, so we do not
+         * have to check that ME cost is more than the estimated merge cost */
+        fencCost = tld.me.motionEstimate(fref, mvmin, mvmax, mvp, 0, NULL, s_merange, *fencMV);
         COPY2_IF_LT(bcost, fencCost, listused, i + 1);
     }

x265_1.6.tar.gz/source/encoder/slicetype.h -> x265_1.7.tar.gz/source/encoder/slicetype.h Changed

@@ -105,8 +105,6 @@
     Lock          m_outputLock;
 
     /* pre-lookahead */
-    Frame*        m_preframes[X265_LOOKAHEAD_MAX];
-    int           m_preTotal, m_preAcquired, m_preCompleted;
     int           m_fullQueueSize;
     bool          m_isActive;
     bool          m_sliceTypeBusy;
@@ -114,7 +112,6 @@
     bool          m_outputSignalRequired;
     bool          m_bBatchMotionSearch;
     bool          m_bBatchFrameCosts;
-    Lock          m_preLookaheadLock;
     Event         m_outputSignal;
 
     LookaheadTLD* m_tld;
@@ -143,7 +140,7 @@
 
     bool    create();
     void    destroy();
-    void    stop();
+    void    stopJobs();
 
     void    addPicture(Frame&, int sliceType);
     void    flush();
@@ -176,6 +173,22 @@
     int64_t frameCostRecalculate(Lowres **frames, int p0, int p1, int b);
 };
 
+class PreLookaheadGroup : public BondedTaskGroup
+{
+public:
+
+    Frame* m_preframes[X265_LOOKAHEAD_MAX];
+    Lookahead& m_lookahead;
+
+    PreLookaheadGroup(Lookahead& l) : m_lookahead(l) {}
+
+    void processTasks(int workerThreadID);
+
+protected:
+
+    PreLookaheadGroup& operator=(const PreLookaheadGroup&);
+};
+
 class CostEstimateGroup : public BondedTaskGroup
 {
 public:

x265_1.6.tar.gz/source/input/input.cpp -> x265_1.7.tar.gz/source/input/input.cpp Changed

x265_1.6.tar.gz/source/input/input.h -> x265_1.7.tar.gz/source/input/input.h Changed

x265_1.6.tar.gz/source/input/y4m.cpp -> x265_1.7.tar.gz/source/input/y4m.cpp Changed

x265_1.6.tar.gz/source/input/y4m.h -> x265_1.7.tar.gz/source/input/y4m.h Changed

x265_1.6.tar.gz/source/input/yuv.cpp -> x265_1.7.tar.gz/source/input/yuv.cpp Changed

x265_1.6.tar.gz/source/input/yuv.h -> x265_1.7.tar.gz/source/input/yuv.h Changed

x265_1.6.tar.gz/source/output/output.cpp -> x265_1.7.tar.gz/source/output/output.cpp Changed

x265_1.6.tar.gz/source/output/output.h -> x265_1.7.tar.gz/source/output/output.h Changed

@@ -1,7 +1,8 @@
 /*****************************************************************************
- * Copyright (C) 2013 x265 project
+ * Copyright (C) 2013-2015 x265 project
  *
  * Authors: Steve Borho <steve@borho.org>
+ *          Xinyue Lu <i@7086.in>
  *
  * 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
@@ -25,22 +26,23 @@
 #define X265_OUTPUT_H
 
 #include "x265.h"
+#include "input/input.h"
 
 namespace x265 {
 // private x265 namespace
 
-class Output
+class ReconFile
 {
 protected:
 
-    virtual ~Output()  {}
+    virtual ~ReconFile()  {}
 
 public:
 
-    Output()           {}
+    ReconFile()           {}
 
-    static Output* open(const char *fname, int width, int height, uint32_t bitdepth,
-                        uint32_t fpsNum, uint32_t fpsDenom, int csp);
+    static ReconFile* open(const char *fname, int width, int height, uint32_t bitdepth,
+                           uint32_t fpsNum, uint32_t fpsDenom, int csp);
 
     virtual bool isFail() const = 0;
 
@@ -50,6 +52,35 @@
 
     virtual const char *getName() const = 0;
 };
+
+class OutputFile
+{
+protected:
+
+    virtual ~OutputFile() {}
+
+public:
+
+    OutputFile() {}
+
+    static OutputFile* open(const char* fname, InputFileInfo& inputInfo);
+
+    virtual bool isFail() const = 0;
+
+    virtual bool needPTS() const = 0;
+
+    virtual void release() = 0;
+
+    virtual const char* getName() const = 0;
+
+    virtual void setParam(x265_param* param) = 0;
+
+    virtual int writeHeaders(const x265_nal* nal, uint32_t nalcount) = 0;
+
+    virtual int writeFrame(const x265_nal* nal, uint32_t nalcount, x265_picture& pic) = 0;
+
+    virtual void closeFile(int64_t largest_pts, int64_t second_largest_pts) = 0;
+};
 }
 
 #endif // ifndef X265_OUTPUT_H

x265_1.7.tar.gz/source/output/raw.cpp Added

@@ -0,0 +1,80 @@
+/*****************************************************************************
+ * Copyright (C) 2013-2015 x265 project
+ *
+ * Authors: Steve Borho <steve@borho.org>
+ *          Xinyue Lu <i@7086.in>
+ *
+ * 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 "raw.h"
+
+using namespace x265;
+using namespace std;
+
+RAWOutput::RAWOutput(const char* fname, InputFileInfo&)
+{
+    b_fail = false;
+    if (!strcmp(fname, "-"))
+    {
+        ofs = &cout;
+        return;
+    }
+    ofs = new ofstream(fname, ios::binary | ios::out);
+    if (ofs->fail())
+        b_fail = true;
+}
+
+void RAWOutput::setParam(x265_param* param)
+{
+    param->bAnnexB = true;
+}
+
+int RAWOutput::writeHeaders(const x265_nal* nal, uint32_t nalcount)
+{
+    uint32_t bytes = 0;
+
+    for (uint32_t i = 0; i < nalcount; i++)
+    {
+        ofs->write((const char*)nal->payload, nal->sizeBytes);
+        bytes += nal->sizeBytes;
+        nal++;
+    }
+
+    return bytes;
+}
+
+int RAWOutput::writeFrame(const x265_nal* nal, uint32_t nalcount, x265_picture&)
+{
+    uint32_t bytes = 0;
+
+    for (uint32_t i = 0; i < nalcount; i++)
+    {
+        ofs->write((const char*)nal->payload, nal->sizeBytes);
+        bytes += nal->sizeBytes;
+        nal++;
+    }
+
+    return bytes;
+}
+
+void RAWOutput::closeFile(int64_t, int64_t)
+{
+    if (ofs != &cout)
+        delete ofs;
+}

x265_1.7.tar.gz/source/output/raw.h Added

@@ -0,0 +1,64 @@
+/*****************************************************************************
+ * Copyright (C) 2013-2015 x265 project
+ *
+ * Authors: Steve Borho <steve@borho.org>
+ *          Xinyue Lu <i@7086.in>
+ *
+ * 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.
+ *****************************************************************************/
+
+#ifndef X265_HEVC_RAW_H
+#define X265_HEVC_RAW_H
+
+#include "output.h"
+#include "common.h"
+#include <fstream>
+#include <iostream>
+
+namespace x265 {
+class RAWOutput : public OutputFile
+{
+protected:
+
+    std::ostream* ofs;
+
+    bool b_fail;
+
+public:
+
+    RAWOutput(const char* fname, InputFileInfo&);
+
+    bool isFail() const { return b_fail; }
+
+    bool needPTS() const { return false; }
+
+    void release() { delete this; }
+
+    const char* getName() const { return "raw"; }
+
+    void setParam(x265_param* param);
+
+    int writeHeaders(const x265_nal* nal, uint32_t nalcount);
+
+    int writeFrame(const x265_nal* nal, uint32_t nalcount, x265_picture&);
+
+    void closeFile(int64_t largest_pts, int64_t second_largest_pts);
+};
+}
+
+#endif // ifndef X265_HEVC_RAW_H

x265_1.7.tar.gz/source/output/reconplay.cpp Added

@@ -0,0 +1,197 @@
+/*****************************************************************************
+ * Copyright (C) 2013 x265 project
+ *
+ * Authors: Peixuan Zhang <zhangpeixuancn@gmail.com>
+ *          Chunli Zhang <chunli@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 "common.h"
+#include "reconplay.h"
+
+#include <signal.h>
+
+using namespace x265;
+
+#if _WIN32
+#define popen  _popen
+#define pclose _pclose
+#define pipemode "wb"
+#else
+#define pipemode "w"
+#endif
+
+bool ReconPlay::pipeValid;
+
+#ifndef _WIN32
+static void sigpipe_handler(int)
+{
+    if (ReconPlay::pipeValid)
+        general_log(NULL, "exec", X265_LOG_ERROR, "pipe closed\n");
+    ReconPlay::pipeValid = false;
+}
+#endif
+
+ReconPlay::ReconPlay(const char* commandLine, x265_param& param)
+{
+#ifndef _WIN32
+    if (signal(SIGPIPE, sigpipe_handler) == SIG_ERR)
+        general_log(&param, "exec", X265_LOG_ERROR, "Unable to register SIGPIPE handler: %s\n", strerror(errno));
+#endif
+
+    width = param.sourceWidth;
+    height = param.sourceHeight;
+    colorSpace = param.internalCsp;
+
+    frameSize = 0;
+    for (int i = 0; i < x265_cli_csps[colorSpace].planes; i++)
+        frameSize += (uint32_t)((width >> x265_cli_csps[colorSpace].width[i]) * (height >> x265_cli_csps[colorSpace].height[i]));
+
+    for (int i = 0; i < RECON_BUF_SIZE; i++)
+    {
+        poc[i] = -1;
+        CHECKED_MALLOC(frameData[i], pixel, frameSize);
+    }
+
+    outputPipe = popen(commandLine, pipemode);
+    if (outputPipe)
+    {
+        const char* csp = (colorSpace >= X265_CSP_I444) ? "444" : (colorSpace >= X265_CSP_I422) ? "422" : "420";
+        const char* depth = (param.internalBitDepth == 10) ? "p10" : "";
+
+        fprintf(outputPipe, "YUV4MPEG2 W%d H%d F%d:%d Ip C%s%s\n", width, height, param.fpsNum, param.fpsDenom, csp, depth);
+
+        pipeValid = true;
+        threadActive = true;
+        start();
+        return;
+    }
+    else
+        general_log(&param, "exec", X265_LOG_ERROR, "popen(%s) failed\n", commandLine);
+
+fail:
+    threadActive = false;
+}
+
+ReconPlay::~ReconPlay()
+{
+    if (threadActive)
+    {
+        threadActive = false;
+        writeCount.poke();
+        stop();
+    }
+
+    if (outputPipe) 
+        pclose(outputPipe);
+
+    for (int i = 0; i < RECON_BUF_SIZE; i++)
+        X265_FREE(frameData[i]);
+}
+
+bool ReconPlay::writePicture(const x265_picture& pic)
+{
+    if (!threadActive || !pipeValid)
+        return false;
+
+    int written = writeCount.get();
+    int read = readCount.get();
+    int currentCursor = pic.poc % RECON_BUF_SIZE;
+
+    /* TODO: it's probably better to drop recon pictures when the ring buffer is
+     * backed up on the display app */
+    while (written - read > RECON_BUF_SIZE - 2 || poc[currentCursor] != -1)
+    {
+        read = readCount.waitForChange(read);
+        if (!threadActive)
+            return false;
+    }
+
+    X265_CHECK(pic.colorSpace == colorSpace, "invalid color space\n");
+    X265_CHECK(pic.bitDepth == X265_DEPTH,   "invalid bit depth\n");
+
+    pixel* buf = frameData[currentCursor];
+    for (int i = 0; i < x265_cli_csps[colorSpace].planes; i++)
+    {
+        char* src = (char*)pic.planes[i];
+        int pwidth = width >> x265_cli_csps[colorSpace].width[i];
+
+        for (int h = 0; h < height >> x265_cli_csps[colorSpace].height[i]; h++)
+        {
+            memcpy(buf, src, pwidth * sizeof(pixel));
+            src += pic.stride[i];
+            buf += pwidth;
+        }
+    }
+
+    poc[currentCursor] = pic.poc;
+    writeCount.incr();
+
+    return true;
+}
+
+void ReconPlay::threadMain()
+{
+    THREAD_NAME("ReconPlayOutput", 0);
+
+    do
+    {
+        /* extract the next output picture in display order and write to pipe */
+        if (!outputFrame())
+            break;
+    }
+    while (threadActive);
+
+    threadActive = false;
+    readCount.poke();
+}
+
+bool ReconPlay::outputFrame()
+{
+    int written = writeCount.get();
+    int read = readCount.get();
+    int currentCursor = read % RECON_BUF_SIZE;
+
+    while (poc[currentCursor] != read)
+    {
+        written = writeCount.waitForChange(written);
+        if (!threadActive)
+            return false;
+    }
+
+    char* buf = (char*)frameData[currentCursor];
+    intptr_t remainSize = frameSize * sizeof(pixel);
+
+    fprintf(outputPipe, "FRAME\n");
+    while (remainSize > 0)
+    {
+        intptr_t retCount = (intptr_t)fwrite(buf, sizeof(char), remainSize, outputPipe);
+
+        if (retCount < 0 || !pipeValid)
+            /* pipe failure, stop writing and start dropping recon pictures */
+            return false;
+    
+        buf += retCount;
+        remainSize -= retCount;
+    }
+
+    poc[currentCursor] = -1;
+    readCount.incr();
+    return true;
+}

x265_1.7.tar.gz/source/output/reconplay.h Added

@@ -0,0 +1,74 @@
+/*****************************************************************************
+ * Copyright (C) 2013 x265 project
+ *
+ * Authors: Peixuan Zhang <zhangpeixuancn@gmail.com>
+ *          Chunli Zhang <chunli@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.
+ *****************************************************************************/
+
+#ifndef X265_RECONPLAY_H
+#define X265_RECONPLAY_H
+
+#include "x265.h"
+#include "threading.h"
+#include <cstdio>
+
+namespace x265 {
+// private x265 namespace
+
+class ReconPlay : public Thread
+{
+public:
+
+    ReconPlay(const char* commandLine, x265_param& param);
+
+    virtual ~ReconPlay();
+
+    bool writePicture(const x265_picture& pic);
+
+    static bool pipeValid;
+
+protected:
+
+    enum { RECON_BUF_SIZE = 40 };
+
+    FILE*  outputPipe;     /* The output pipe for player */
+    size_t frameSize;      /* size of one frame in pixels */
+    bool   threadActive;   /* worker thread is active */
+    int    width;          /* width of frame */
+    int    height;         /* height of frame */
+    int    colorSpace;     /* color space of frame */
+
+    int    poc[RECON_BUF_SIZE];
+    pixel* frameData[RECON_BUF_SIZE];
+
+    /* Note that the class uses read and write counters to signal that reads and
+     * writes have occurred in the ring buffer, but writes into the buffer
+     * happen in decode order and the reader must check that the POC it next
+     * needs to send to the pipe is in fact present.  The counters are used to
+     * prevent the writer from getting too far ahead of the reader */
+    ThreadSafeInteger readCount;
+    ThreadSafeInteger writeCount;
+
+    void threadMain();
+    bool outputFrame();
+};
+}
+
+#endif // ifndef X265_RECONPLAY_H

x265_1.6.tar.gz/source/output/y4m.h -> x265_1.7.tar.gz/source/output/y4m.h Changed

x265_1.6.tar.gz/source/output/yuv.h -> x265_1.7.tar.gz/source/output/yuv.h Changed

x265_1.6.tar.gz/source/test/ipfilterharness.cpp -> x265_1.7.tar.gz/source/test/ipfilterharness.cpp Changed

@@ -61,55 +61,6 @@
     }
 }
 
-bool IPFilterHarness::check_IPFilter_primitive(filter_p2s_wxh_t ref, filter_p2s_wxh_t opt, int isChroma, int csp)
-{
-    intptr_t rand_srcStride;
-    int min_size = isChroma ? 2 : 4;
-    int max_size = isChroma ? (MAX_CU_SIZE >> 1) : MAX_CU_SIZE;
-
-    if (isChroma && (csp == X265_CSP_I444))
-    {
-        min_size = 4;
-        max_size = MAX_CU_SIZE;
-    }
-
-    for (int i = 0; i < ITERS; i++)
-    {
-        int index = i % TEST_CASES;
-        int rand_height = (int16_t)rand() % 100;
-        int rand_width = (int16_t)rand() % 100;
-
-        rand_srcStride = rand_width + rand() % 100;
-        if (rand_srcStride < rand_width)
-            rand_srcStride = rand_width;
-
-        rand_width &= ~(min_size - 1);
-        rand_width = x265_clip3(min_size, max_size, rand_width);
-
-        rand_height &= ~(min_size - 1);
-        rand_height = x265_clip3(min_size, max_size, rand_height);
-
-        ref(pixel_test_buff[index],
-            rand_srcStride,
-            IPF_C_output_s,
-            rand_width,
-            rand_height);
-
-        checked(opt, pixel_test_buff[index],
-                rand_srcStride,
-                IPF_vec_output_s,
-                rand_width,
-                rand_height);
-
-        if (memcmp(IPF_vec_output_s, IPF_C_output_s, TEST_BUF_SIZE * sizeof(int16_t)))
-            return false;
-
-        reportfail();
-    }
-
-    return true;
-}
-
 bool IPFilterHarness::check_IPFilterChroma_primitive(filter_pp_t ref, filter_pp_t opt)
 {
     intptr_t rand_srcStride, rand_dstStride;
@@ -518,12 +469,13 @@
     {
         intptr_t rand_srcStride = rand() % 100;
         int index = i % TEST_CASES;
+        intptr_t dstStride = rand() % 100 + 64;
 
-        ref(pixel_test_buff[index] + i, rand_srcStride, IPF_C_output_s);
+        ref(pixel_test_buff[index] + i, rand_srcStride, IPF_C_output_s, dstStride);
 
-        checked(opt, pixel_test_buff[index] + i, rand_srcStride, IPF_vec_output_s);
+        checked(opt, pixel_test_buff[index] + i, rand_srcStride, IPF_vec_output_s, dstStride);
 
-        if (memcmp(IPF_vec_output_s, IPF_C_output_s, TEST_BUF_SIZE * sizeof(pixel)))
+        if (memcmp(IPF_vec_output_s, IPF_C_output_s, TEST_BUF_SIZE * sizeof(int16_t)))
             return false;
 
         reportfail();
@@ -538,12 +490,13 @@
     {
         intptr_t rand_srcStride = rand() % 100;
         int index = i % TEST_CASES;
+        intptr_t dstStride = rand() % 100 + 64;
 
-        ref(pixel_test_buff[index] + i, rand_srcStride, IPF_C_output_s);
+        ref(pixel_test_buff[index] + i, rand_srcStride, IPF_C_output_s, dstStride);
 
-        checked(opt, pixel_test_buff[index] + i, rand_srcStride, IPF_vec_output_s);
+        checked(opt, pixel_test_buff[index] + i, rand_srcStride, IPF_vec_output_s, dstStride);
 
-        if (memcmp(IPF_vec_output_s, IPF_C_output_s, TEST_BUF_SIZE * sizeof(pixel)))
+        if (memcmp(IPF_vec_output_s, IPF_C_output_s, TEST_BUF_SIZE * sizeof(int16_t)))
             return false;
 
         reportfail();
@@ -554,15 +507,6 @@
 
 bool IPFilterHarness::testCorrectness(const EncoderPrimitives& ref, const EncoderPrimitives& opt)
 {
-    if (opt.luma_p2s)
-    {
-        // last parameter does not matter in case of luma
-        if (!check_IPFilter_primitive(ref.luma_p2s, opt.luma_p2s, 0, 1))
-        {
-            printf("luma_p2s failed\n");
-            return false;
-        }
-    }
 
     for (int value = 0; value < NUM_PU_SIZES; value++)
     {
@@ -622,11 +566,11 @@
                 return false;
             }
         }
-        if (opt.pu[value].filter_p2s)
+        if (opt.pu[value].convert_p2s)
         {
-            if (!check_IPFilterLumaP2S_primitive(ref.pu[value].filter_p2s, opt.pu[value].filter_p2s))
+            if (!check_IPFilterLumaP2S_primitive(ref.pu[value].convert_p2s, opt.pu[value].convert_p2s))
             {
-                printf("filter_p2s[%s]", lumaPartStr[value]);
+                printf("convert_p2s[%s]", lumaPartStr[value]);
                 return false;
             }
         }
@@ -634,14 +578,6 @@
 
     for (int csp = X265_CSP_I420; csp < X265_CSP_COUNT; csp++)
     {
-        if (opt.chroma[csp].p2s)
-        {
-            if (!check_IPFilter_primitive(ref.chroma[csp].p2s, opt.chroma[csp].p2s, 1, csp))
-            {
-                printf("chroma_p2s[%s]", x265_source_csp_names[csp]);
-                return false;
-            }
-        }
         for (int value = 0; value < NUM_PU_SIZES; value++)
         {
             if (opt.chroma[csp].pu[value].filter_hpp)
@@ -692,9 +628,9 @@
                     return false;
                 }
             }
-            if (opt.chroma[csp].pu[value].chroma_p2s)
+            if (opt.chroma[csp].pu[value].p2s)
             {
-                if (!check_IPFilterChromaP2S_primitive(ref.chroma[csp].pu[value].chroma_p2s, opt.chroma[csp].pu[value].chroma_p2s))
+                if (!check_IPFilterChromaP2S_primitive(ref.chroma[csp].pu[value].p2s, opt.chroma[csp].pu[value].p2s))
                 {
                     printf("chroma_p2s[%s]", chromaPartStr[csp][value]);
                     return false;
@@ -708,19 +644,10 @@
 
 void IPFilterHarness::measureSpeed(const EncoderPrimitives& ref, const EncoderPrimitives& opt)
 {
-    int height = 64;
-    int width = 64;
     int16_t srcStride = 96;
     int16_t dstStride = 96;
     int maxVerticalfilterHalfDistance = 3;
 
-    if (opt.luma_p2s)
-    {
-        printf("luma_p2s\t");
-        REPORT_SPEEDUP(opt.luma_p2s, ref.luma_p2s,
-                       pixel_buff, srcStride, IPF_vec_output_s, width, height);
-    }
-
     for (int value = 0; value < NUM_PU_SIZES; value++)
     {
         if (opt.pu[value].luma_hpp)
@@ -777,23 +704,18 @@
                            pixel_buff + 3 * srcStride, srcStride, IPF_vec_output_p, srcStride, 1, 3);
         }
 
-        if (opt.pu[value].filter_p2s)
+        if (opt.pu[value].convert_p2s)
         {
-            printf("filter_p2s [%s]\t", lumaPartStr[value]);
-            REPORT_SPEEDUP(opt.pu[value].filter_p2s, ref.pu[value].filter_p2s,
-                           pixel_buff, srcStride, IPF_vec_output_s);
+            printf("convert_p2s[%s]\t", lumaPartStr[value]);
+            REPORT_SPEEDUP(opt.pu[value].convert_p2s, ref.pu[value].convert_p2s,
+                               pixel_buff, srcStride,
+                               IPF_vec_output_s, dstStride);
         }
     }
 
     for (int csp = X265_CSP_I420; csp < X265_CSP_COUNT; csp++)
     {
         printf("= Color Space %s =\n", x265_source_csp_names[csp]);
-        if (opt.chroma[csp].p2s)
-        {
-            printf("chroma_p2s\t");
-            REPORT_SPEEDUP(opt.chroma[csp].p2s, ref.chroma[csp].p2s,
-                           pixel_buff, srcStride, IPF_vec_output_s, width, height);
-        }
         for (int value = 0; value < NUM_PU_SIZES; value++)
         {
             if (opt.chroma[csp].pu[value].filter_hpp)
@@ -836,13 +758,11 @@
                                short_buff + maxVerticalfilterHalfDistance * srcStride, srcStride,
                                IPF_vec_output_s, dstStride, 1);
             }
-
-            if (opt.chroma[csp].pu[value].chroma_p2s)
+            if (opt.chroma[csp].pu[value].p2s)
             {
                 printf("chroma_p2s[%s]\t", chromaPartStr[csp][value]);
-                REPORT_SPEEDUP(opt.chroma[csp].pu[value].chroma_p2s, ref.chroma[csp].pu[value].chroma_p2s,
-                               pixel_buff, srcStride,
-                               IPF_vec_output_s);
+                REPORT_SPEEDUP(opt.chroma[csp].pu[value].p2s, ref.chroma[csp].pu[value].p2s,
+                               pixel_buff, srcStride, IPF_vec_output_s, dstStride);
             }
         }
     }

x265_1.6.tar.gz/source/test/ipfilterharness.h -> x265_1.7.tar.gz/source/test/ipfilterharness.h Changed

x265_1.6.tar.gz/source/test/pixelharness.cpp -> x265_1.7.tar.gz/source/test/pixelharness.cpp Changed

@@ -666,7 +666,32 @@
     return true;
 }
 
-bool PixelHarness::check_scale_pp(scale_t ref, scale_t opt)
+bool PixelHarness::check_scale1D_pp(scale1D_t ref, scale1D_t opt)
+{
+    ALIGN_VAR_16(pixel, ref_dest[64 * 64]);
+    ALIGN_VAR_16(pixel, opt_dest[64 * 64]);
+
+    memset(ref_dest, 0, sizeof(ref_dest));
+    memset(opt_dest, 0, sizeof(opt_dest));
+
+    int j = 0;
+    for (int i = 0; i < ITERS; i++)
+    {
+        int index = i % TEST_CASES;
+        checked(opt, opt_dest, pixel_test_buff[index] + j);
+        ref(ref_dest, pixel_test_buff[index] + j);
+
+        if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel)))
+            return false;
+
+        reportfail();
+        j += INCR;
+    }
+
+    return true;
+}
+
+bool PixelHarness::check_scale2D_pp(scale2D_t ref, scale2D_t opt)
 {
     ALIGN_VAR_16(pixel, ref_dest[64 * 64]);
     ALIGN_VAR_16(pixel, opt_dest[64 * 64]);
@@ -845,8 +870,8 @@
 
 bool PixelHarness::check_calSign(sign_t ref, sign_t opt)
 {
-    ALIGN_VAR_16(int8_t, ref_dest[64 * 64]);
-    ALIGN_VAR_16(int8_t, opt_dest[64 * 64]);
+    ALIGN_VAR_16(int8_t, ref_dest[64 * 2]);
+    ALIGN_VAR_16(int8_t, opt_dest[64 * 2]);
 
     memset(ref_dest, 0xCD, sizeof(ref_dest));
     memset(opt_dest, 0xCD, sizeof(opt_dest));
@@ -855,12 +880,12 @@
 
     for (int i = 0; i < ITERS; i++)
     {
-        int width = 16 * (rand() % 4 + 1);
+        int width = (rand() % 64) + 1;
 
         ref(ref_dest, pbuf2 + j, pbuf3 + j, width);
         checked(opt, opt_dest, pbuf2 + j, pbuf3 + j, width);
 
-        if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(int8_t)))
+        if (memcmp(ref_dest, opt_dest, sizeof(ref_dest)))
             return false;
 
         reportfail();
@@ -883,12 +908,10 @@
     for (int i = 0; i < ITERS; i++)
     {
         int width = 16 * (rand() % 4 + 1);
-        int8_t sign = rand() % 3;
-        if (sign == 2)
-            sign = -1;
+        int stride = width + 1;
 
-        ref(ref_dest, psbuf1 + j, width, sign);
-        checked(opt, opt_dest, psbuf1 + j, width, sign);
+        ref(ref_dest, psbuf1 + j, width, psbuf2 + j, stride);
+        checked(opt, opt_dest, psbuf1 + j, width, psbuf5 + j, stride);
 
         if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel)))
             return false;
@@ -928,7 +951,43 @@
     return true;
 }
 
-bool PixelHarness::check_saoCuOrgE2_t(saoCuOrgE2_t ref, saoCuOrgE2_t opt)
+bool PixelHarness::check_saoCuOrgE2_t(saoCuOrgE2_t ref[2], saoCuOrgE2_t opt[2])
+{
+    ALIGN_VAR_16(pixel, ref_dest[64 * 64]);
+    ALIGN_VAR_16(pixel, opt_dest[64 * 64]);
+
+    memset(ref_dest, 0xCD, sizeof(ref_dest));
+    memset(opt_dest, 0xCD, sizeof(opt_dest));
+
+    for (int id = 0; id < 2; id++)
+    {
+        int j = 0;
+        if (opt[id])
+        {
+            for (int i = 0; i < ITERS; i++)
+            {
+                int width = 16 * (1 << (id * (rand() % 2 + 1))) - (rand() % 2);
+                int stride = width + 1;
+
+                ref[width > 16](ref_dest, psbuf1 + j, psbuf2 + j, psbuf3 + j, width, stride);
+                checked(opt[width > 16], opt_dest, psbuf4 + j, psbuf2 + j, psbuf3 + j, width, stride);
+
+                if (memcmp(psbuf1 + j, psbuf4 + j, width * sizeof(int8_t)))
+                    return false;
+
+                if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel)))
+                    return false;
+
+                reportfail();
+                j += INCR;
+            }
+        }
+    }
+
+    return true;
+}
+
+bool PixelHarness::check_saoCuOrgE3_t(saoCuOrgE3_t ref, saoCuOrgE3_t opt)
 {
     ALIGN_VAR_16(pixel, ref_dest[64 * 64]);
     ALIGN_VAR_16(pixel, opt_dest[64 * 64]);
@@ -940,16 +999,14 @@
 
     for (int i = 0; i < ITERS; i++)
     {
-        int width = 16 * (rand() % 4 + 1);
-        int stride = width + 1;
-
-        ref(ref_dest, psbuf1 + j, psbuf2 + j, psbuf3 + j, width, stride);
-        checked(opt, opt_dest, psbuf4 + j, psbuf2 + j, psbuf3 + j, width, stride);
+        int stride = 16 * (rand() % 4 + 1);
+        int start = rand() % 2;
+        int end = 16 - rand() % 2;
 
-        if (memcmp(psbuf1 + j, psbuf4 + j, width * sizeof(int8_t)))
-            return false;
+        ref(ref_dest, psbuf2 + j, psbuf1 + j, stride, start, end);
+        checked(opt, opt_dest, psbuf5 + j, psbuf1 + j, stride, start, end);
 
-        if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel)))
+        if (memcmp(ref_dest, opt_dest, 64 * 64 * sizeof(pixel)) || memcmp(psbuf2, psbuf5, BUFFSIZE))
             return false;
 
         reportfail();
@@ -959,7 +1016,7 @@
     return true;
 }
 
-bool PixelHarness::check_saoCuOrgE3_t(saoCuOrgE3_t ref, saoCuOrgE3_t opt)
+bool PixelHarness::check_saoCuOrgE3_32_t(saoCuOrgE3_t ref, saoCuOrgE3_t opt)
 {
     ALIGN_VAR_16(pixel, ref_dest[64 * 64]);
     ALIGN_VAR_16(pixel, opt_dest[64 * 64]);
@@ -971,9 +1028,9 @@
 
     for (int i = 0; i < ITERS; i++)
     {
-        int stride = 16 * (rand() % 4 + 1);
+        int stride = 32 * (rand() % 2 + 1);
         int start = rand() % 2;
-        int end = (16 * (rand() % 4 + 1)) - rand() % 2;
+        int end = (32 * (rand() % 2 + 1)) - rand() % 2;
 
         ref(ref_dest, psbuf2 + j, psbuf1 + j, stride, start, end);
         checked(opt, opt_dest, psbuf5 + j, psbuf1 + j, stride, start, end);
@@ -995,9 +1052,8 @@
 
     memset(ref_dest, 0xCD, sizeof(ref_dest));
     memset(opt_dest, 0xCD, sizeof(opt_dest));
-
-    int width = 16 + rand() % 48;
-    int height = 16 + rand() % 48;
+    int width = 32 + rand() % 32;
+    int height = 32 + rand() % 32;
     intptr_t srcStride = 64;
     intptr_t dstStride = width;
     int j = 0;
@@ -1133,8 +1189,8 @@
     for (int i = 0; i < ITERS; i++)
     {
         int width = 16 * (rand() % 4 + 1);
-        int height = rand() % 64 +1;
-        int stride = rand() % 65;
+        int height = rand() % 63 + 2;
+        int stride = width;
 
         ref(ref_dest, psbuf1 + j, width, height, stride);
         checked(opt, opt_dest, psbuf1 + j, width, height, stride);
@@ -1149,7 +1205,7 @@
     return true;
 }
 
-bool PixelHarness::check_findPosLast(findPosLast_t ref, findPosLast_t opt)
+bool PixelHarness::check_scanPosLast(scanPosLast_t ref, scanPosLast_t opt)
 {
     ALIGN_VAR_16(coeff_t, ref_src[32 * 32 + ITERS * 2]);
     uint8_t ref_coeffNum[MLS_GRP_NUM], opt_coeffNum[MLS_GRP_NUM];      // value range[0, 16]
@@ -1160,6 +1216,14 @@
     for (int i = 0; i < 32 * 32; i++)
     {
         ref_src[i] = rand() & SHORT_MAX;
+
+        // more zero coeff
+        if (ref_src[i] < SHORT_MAX * 2 / 3)
+            ref_src[i] = 0;
+
+        // more negtive
+        if ((rand() % 10) < 8)
+            ref_src[i] *= -1;
         totalCoeffs += (ref_src[i] != 0);
     }
 
@@ -1187,10 +1251,19 @@
         for (int j = 0; j < 1 << (2 * (rand_scan_size + 2)); j++)
             rand_numCoeff += (ref_src[i + j] != 0);
 
+        // at least one coeff in transform block
+        if (rand_numCoeff == 0)
+        {
+            ref_src[i + (1 << (2 * (rand_scan_size + 2))) - 1] = -1;
+            rand_numCoeff = 1;
+        }
+
+        const int trSize = (1 << (rand_scan_size + 2));
         const uint16_t* const scanTbl = g_scanOrder[rand_scan_type][rand_scan_size];
+        const uint16_t* const scanTblCG4x4 = g_scan4x4[rand_scan_size <= (MDCS_LOG2_MAX_SIZE - 2) ? rand_scan_type : SCAN_DIAG];
 
-        int ref_scanPos = ref(scanTbl, ref_src + i, ref_coeffSign, ref_coeffFlag, ref_coeffNum, rand_numCoeff);
-        int opt_scanPos = (int)checked(opt, scanTbl, ref_src + i, opt_coeffSign, opt_coeffFlag, opt_coeffNum, rand_numCoeff);
+        int ref_scanPos = ref(scanTbl, ref_src + i, ref_coeffSign, ref_coeffFlag, ref_coeffNum, rand_numCoeff, scanTblCG4x4, trSize);
+        int opt_scanPos = (int)checked(opt, scanTbl, ref_src + i, opt_coeffSign, opt_coeffFlag, opt_coeffNum, rand_numCoeff, scanTblCG4x4, trSize);
 
         if (ref_scanPos != opt_scanPos)
             return false;
@@ -1209,6 +1282,56 @@
             rand_numCoeff -= ref_coeffNum[j];
         }
 
+        if (rand_numCoeff != 0)
+            return false;
+
+        reportfail();
+    }
+
+    return true;
+}
+
+bool PixelHarness::check_findPosFirstLast(findPosFirstLast_t ref, findPosFirstLast_t opt)
+{
+    ALIGN_VAR_16(coeff_t, ref_src[32 * 32 + ITERS * 2]);
+
+    for (int i = 0; i < 32 * 32; i++)
+    {
+        ref_src[i] = rand() & SHORT_MAX;
+    }
+
+    // extra test area all of 0x1234
+    for (int i = 0; i < ITERS * 2; i++)
+    {
+        ref_src[32 * 32 + i] = 0x1234;
+    }
+
+    for (int i = 0; i < ITERS; i++)
+    {
+        int rand_scan_type = rand() % NUM_SCAN_TYPE;
+        int rand_scan_size = (rand() % NUM_SCAN_SIZE) + 2;
+        coeff_t *rand_src = ref_src + i;
+
+        const uint16_t* const scanTbl = g_scan4x4[rand_scan_type];
+
+        int j;
+        for (j = 0; j < SCAN_SET_SIZE; j++)
+        {
+            const uint32_t idxY = j / MLS_CG_SIZE;
+            const uint32_t idxX = j % MLS_CG_SIZE;
+            if (rand_src[idxY * rand_scan_size + idxX]) break;
+        }
+
+        // fill one coeff when all coeff group are zero
+        if (j >= SCAN_SET_SIZE)
+            rand_src[0] = 0x0BAD;
+
+        uint32_t ref_scanPos = ref(rand_src, (1 << rand_scan_size), scanTbl);
+        uint32_t opt_scanPos = (int)checked(opt, rand_src, (1 << rand_scan_size), scanTbl);
+
+        if (ref_scanPos != opt_scanPos)
+            return false;
+
         reportfail();
     }
 
@@ -1414,6 +1537,14 @@
                     return false;
                 }
             }
+            if (opt.chroma[i].cu[part].sa8d)
+            {
+                if (!check_pixelcmp(ref.chroma[i].cu[part].sa8d, opt.chroma[i].cu[part].sa8d))
+                {
+                    printf("chroma_sa8d[%s][%s] failed\n", x265_source_csp_names[i], chromaPartStr[i][part]);
+                    return false;
+                }
+            }
         }
     }
 
@@ -1603,7 +1734,7 @@
 
     if (opt.scale1D_128to64)
     {
-        if (!check_scale_pp(ref.scale1D_128to64, opt.scale1D_128to64))
+        if (!check_scale1D_pp(ref.scale1D_128to64, opt.scale1D_128to64))
         {
             printf("scale1D_128to64 failed!\n");
             return false;
@@ -1612,7 +1743,7 @@
 
     if (opt.scale2D_64to32)
     {
-        if (!check_scale_pp(ref.scale2D_64to32, opt.scale2D_64to32))
+        if (!check_scale2D_pp(ref.scale2D_64to32, opt.scale2D_64to32))
         {
             printf("scale2D_64to32 failed!\n");
             return false;
@@ -1664,20 +1795,41 @@
         }
     }
 
-    if (opt.saoCuOrgE2)
+    if (opt.saoCuOrgE1_2Rows)
+    {
+        if (!check_saoCuOrgE1_t(ref.saoCuOrgE1_2Rows, opt.saoCuOrgE1_2Rows))
+        {
+            printf("SAO_EO_1_2Rows failed\n");
+            return false;
+        }
+    }
+
+    if (opt.saoCuOrgE2[0] || opt.saoCuOrgE2[1])
+    {
+        saoCuOrgE2_t ref1[] = { ref.saoCuOrgE2[0], ref.saoCuOrgE2[1] };
+        saoCuOrgE2_t opt1[] = { opt.saoCuOrgE2[0], opt.saoCuOrgE2[1] };
+
+        if (!check_saoCuOrgE2_t(ref1, opt1))
+        {
+            printf("SAO_EO_2[0] && SAO_EO_2[1] failed\n");
+            return false;
+        }
+    }
+
+    if (opt.saoCuOrgE3[0])
     {
-        if (!check_saoCuOrgE2_t(ref.saoCuOrgE2, opt.saoCuOrgE2))
+        if (!check_saoCuOrgE3_t(ref.saoCuOrgE3[0], opt.saoCuOrgE3[0]))
         {
-            printf("SAO_EO_2 failed\n");
+            printf("SAO_EO_3[0] failed\n");
             return false;
         }
     }
 
-    if (opt.saoCuOrgE3)
+    if (opt.saoCuOrgE3[1])
     {
-        if (!check_saoCuOrgE3_t(ref.saoCuOrgE3, opt.saoCuOrgE3))
+        if (!check_saoCuOrgE3_32_t(ref.saoCuOrgE3[1], opt.saoCuOrgE3[1]))
         {
-            printf("SAO_EO_3 failed\n");
+            printf("SAO_EO_3[1] failed\n");
             return false;
         }
     }
@@ -1718,11 +1870,20 @@
         }
     }
 
-    if (opt.findPosLast)
+    if (opt.scanPosLast)
     {
-        if (!check_findPosLast(ref.findPosLast, opt.findPosLast))
+        if (!check_scanPosLast(ref.scanPosLast, opt.scanPosLast))
         {
-            printf("findPosLast failed!\n");
+            printf("scanPosLast failed!\n");
+            return false;
+        }
+    }
+
+    if (opt.findPosFirstLast)
+    {
+        if (!check_findPosFirstLast(ref.findPosFirstLast, opt.findPosFirstLast))
+        {
+            printf("findPosFirstLast failed!\n");
             return false;
         }
     }
@@ -1863,6 +2024,11 @@
                 HEADER("[%s]  add_ps[%s]", x265_source_csp_names[i], chromaPartStr[i][part]);
                 REPORT_SPEEDUP(opt.chroma[i].cu[part].add_ps, ref.chroma[i].cu[part].add_ps, pbuf1, FENC_STRIDE, pbuf2, sbuf1, STRIDE, STRIDE);
             }
+            if (opt.chroma[i].cu[part].sa8d)
+            {
+                HEADER("[%s] sa8d[%s]", x265_source_csp_names[i], chromaPartStr[i][part]);
+                REPORT_SPEEDUP(opt.chroma[i].cu[part].sa8d, ref.chroma[i].cu[part].sa8d, pbuf1, STRIDE, pbuf2, STRIDE);
+            }
         }
     }
 
@@ -2003,7 +2169,7 @@
     if (opt.scale1D_128to64)
     {
         HEADER0("scale1D_128to64");
-        REPORT_SPEEDUP(opt.scale1D_128to64, ref.scale1D_128to64, pbuf2, pbuf1, 64);
+        REPORT_SPEEDUP(opt.scale1D_128to64, ref.scale1D_128to64, pbuf2, pbuf1);
     }
 
     if (opt.scale2D_64to32)
@@ -2033,7 +2199,7 @@
     if (opt.saoCuOrgE0)
     {
         HEADER0("SAO_EO_0");
-        REPORT_SPEEDUP(opt.saoCuOrgE0, ref.saoCuOrgE0, pbuf1, psbuf1, 64, 1);
+        REPORT_SPEEDUP(opt.saoCuOrgE0, ref.saoCuOrgE0, pbuf1, psbuf1, 64, psbuf2, 64);
     }
 
     if (opt.saoCuOrgE1)
@@ -2042,16 +2208,34 @@
         REPORT_SPEEDUP(opt.saoCuOrgE1, ref.saoCuOrgE1, pbuf1, psbuf2, psbuf1, 64, 64);
     }
 
-    if (opt.saoCuOrgE2)
+    if (opt.saoCuOrgE1_2Rows)
     {
-        HEADER0("SAO_EO_2");
-        REPORT_SPEEDUP(opt.saoCuOrgE2, ref.saoCuOrgE2, pbuf1, psbuf1, psbuf2, psbuf3, 64, 64);
+        HEADER0("SAO_EO_1_2Rows");
+        REPORT_SPEEDUP(opt.saoCuOrgE1_2Rows, ref.saoCuOrgE1_2Rows, pbuf1, psbuf2, psbuf1, 64, 64);
     }
 
-    if (opt.saoCuOrgE3)
+    if (opt.saoCuOrgE2[0])
     {
-        HEADER0("SAO_EO_3");
-        REPORT_SPEEDUP(opt.saoCuOrgE3, ref.saoCuOrgE3, pbuf1, psbuf2, psbuf1, 64, 0, 64);
+        HEADER0("SAO_EO_2[0]");
+        REPORT_SPEEDUP(opt.saoCuOrgE2[0], ref.saoCuOrgE2[0], pbuf1, psbuf1, psbuf2, psbuf3, 16, 64);
+    }
+
+    if (opt.saoCuOrgE2[1])
+    {
+        HEADER0("SAO_EO_2[1]");
+        REPORT_SPEEDUP(opt.saoCuOrgE2[1], ref.saoCuOrgE2[1], pbuf1, psbuf1, psbuf2, psbuf3, 64, 64);
+    }
+
+    if (opt.saoCuOrgE3[0])
+    {
+        HEADER0("SAO_EO_3[0]");
+        REPORT_SPEEDUP(opt.saoCuOrgE3[0], ref.saoCuOrgE3[0], pbuf1, psbuf2, psbuf1, 64, 0, 16);
+    }
+
+    if (opt.saoCuOrgE3[1])
+    {
+        HEADER0("SAO_EO_3[1]");
+        REPORT_SPEEDUP(opt.saoCuOrgE3[1], ref.saoCuOrgE3[1], pbuf1, psbuf2, psbuf1, 64, 0, 64);
     }
 
     if (opt.saoCuOrgB0)
@@ -2078,12 +2262,25 @@
         REPORT_SPEEDUP(opt.propagateCost, ref.propagateCost, ibuf1, ushort_test_buff[0], int_test_buff[0], ushort_test_buff[0], int_test_buff[0], double_test_buff[0], 80);
     }
 
-    if (opt.findPosLast)
+    if (opt.scanPosLast)
     {
-        HEADER0("findPosLast");
+        HEADER0("scanPosLast");
         coeff_t coefBuf[32 * 32];
         memset(coefBuf, 0, sizeof(coefBuf));
         memset(coefBuf + 32 * 31, 1, 32 * sizeof(coeff_t));
-        REPORT_SPEEDUP(opt.findPosLast, ref.findPosLast, g_scanOrder[SCAN_DIAG][NUM_SCAN_SIZE - 1], coefBuf, (uint16_t*)sbuf1, (uint16_t*)sbuf2, (uint8_t*)psbuf1, 32);
+        REPORT_SPEEDUP(opt.scanPosLast, ref.scanPosLast, g_scanOrder[SCAN_DIAG][NUM_SCAN_SIZE - 1], coefBuf, (uint16_t*)sbuf1, (uint16_t*)sbuf2, (uint8_t*)psbuf1, 32, g_scan4x4[SCAN_DIAG], 32);
+    }
+
+    if (opt.findPosFirstLast)
+    {
+        HEADER0("findPosFirstLast");
+        coeff_t coefBuf[32 * MLS_CG_SIZE];
+        memset(coefBuf, 0, sizeof(coefBuf));
+        // every CG can't be all zeros!
+        coefBuf[3 + 0 * 32] = 0x0BAD;
+        coefBuf[3 + 1 * 32] = 0x0BAD;
+        coefBuf[3 + 2 * 32] = 0x0BAD;
+        coefBuf[3 + 3 * 32] = 0x0BAD;
+        REPORT_SPEEDUP(opt.findPosFirstLast, ref.findPosFirstLast, coefBuf, 32, g_scan4x4[SCAN_DIAG]);
     }
 }

x265_1.6.tar.gz/source/test/pixelharness.h -> x265_1.7.tar.gz/source/test/pixelharness.h Changed

@@ -76,7 +76,8 @@
     bool check_pixelavg_pp(pixelavg_pp_t ref, pixelavg_pp_t opt);
     bool check_pixel_sub_ps(pixel_sub_ps_t ref, pixel_sub_ps_t opt);
     bool check_pixel_add_ps(pixel_add_ps_t ref, pixel_add_ps_t opt);
-    bool check_scale_pp(scale_t ref, scale_t opt);
+    bool check_scale1D_pp(scale1D_t ref, scale1D_t opt);
+    bool check_scale2D_pp(scale2D_t ref, scale2D_t opt);
     bool check_ssd_s(pixel_ssd_s_t ref, pixel_ssd_s_t opt);
     bool check_blockfill_s(blockfill_s_t ref, blockfill_s_t opt);
     bool check_calresidual(calcresidual_t ref, calcresidual_t opt);
@@ -95,8 +96,9 @@
     bool check_addAvg(addAvg_t, addAvg_t);
     bool check_saoCuOrgE0_t(saoCuOrgE0_t ref, saoCuOrgE0_t opt);
     bool check_saoCuOrgE1_t(saoCuOrgE1_t ref, saoCuOrgE1_t opt);
-    bool check_saoCuOrgE2_t(saoCuOrgE2_t ref, saoCuOrgE2_t opt);
+    bool check_saoCuOrgE2_t(saoCuOrgE2_t ref[], saoCuOrgE2_t opt[]);
     bool check_saoCuOrgE3_t(saoCuOrgE3_t ref, saoCuOrgE3_t opt);
+    bool check_saoCuOrgE3_32_t(saoCuOrgE3_t ref, saoCuOrgE3_t opt);
     bool check_saoCuOrgB0_t(saoCuOrgB0_t ref, saoCuOrgB0_t opt);
     bool check_planecopy_sp(planecopy_sp_t ref, planecopy_sp_t opt);
     bool check_planecopy_cp(planecopy_cp_t ref, planecopy_cp_t opt);
@@ -104,7 +106,8 @@
     bool check_psyCost_pp(pixelcmp_t ref, pixelcmp_t opt);
     bool check_psyCost_ss(pixelcmp_ss_t ref, pixelcmp_ss_t opt);
     bool check_calSign(sign_t ref, sign_t opt);
-    bool check_findPosLast(findPosLast_t ref, findPosLast_t opt);
+    bool check_scanPosLast(scanPosLast_t ref, scanPosLast_t opt);
+    bool check_findPosFirstLast(findPosFirstLast_t ref, findPosFirstLast_t opt);
 
 public:

x265_1.6.tar.gz/source/test/rate-control-tests.txt -> x265_1.7.tar.gz/source/test/rate-control-tests.txt Changed

@@ -1,34 +1,36 @@
-# List of command lines to be run by rate control regression tests, see https://bitbucket.org/sborho/test-harness
-
-# This test is listed first since it currently reproduces bugs
-big_buck_bunny_360p24.y4m,--preset medium --bitrate 1000 --pass 1 -F4,--preset medium --bitrate 1000 --pass 2 -F4
-
-# VBV tests, non-deterministic so testing for correctness and bitrate
-# fluctuations - up to 1% bitrate fluctuation is allowed between runs
-RaceHorses_416x240_30_10bit.yuv,--preset medium --bitrate 700 --vbv-bufsize 900 --vbv-maxrate 700
-RaceHorses_416x240_30_10bit.yuv,--preset superfast --bitrate 600 --vbv-bufsize 600 --vbv-maxrate 600
-RaceHorses_416x240_30_10bit.yuv,--preset veryslow --bitrate 1100 --vbv-bufsize 1100 --vbv-maxrate 1200
-112_1920x1080_25.yuv,--preset medium --bitrate 1000 --vbv-maxrate 1500 --vbv-bufsize 1500 --aud
-112_1920x1080_25.yuv,--preset medium --bitrate 10000 --vbv-maxrate 10000 --vbv-bufsize 15000 --hrd
-112_1920x1080_25.yuv,--preset medium --bitrate 4000 --vbv-maxrate 12000 --vbv-bufsize 12000 --repeat-headers
-112_1920x1080_25.yuv,--preset superfast --bitrate 1000 --vbv-maxrate 1000 --vbv-bufsize 1500 --hrd --strict-cbr
-112_1920x1080_25.yuv,--preset superfast --bitrate 30000 --vbv-maxrate 30000 --vbv-bufsize 30000 --repeat-headers
-112_1920x1080_25.yuv,--preset superfast --bitrate 4000 --vbv-maxrate 6000 --vbv-bufsize 6000 --aud
-112_1920x1080_25.yuv,--preset veryslow --bitrate 1000 --vbv-maxrate 3000 --vbv-bufsize 3000 --repeat-headers
-big_buck_bunny_360p24.y4m,--preset medium --bitrate 1000 --vbv-bufsize 3000 --vbv-maxrate 3000 --repeat-headers
-big_buck_bunny_360p24.y4m,--preset medium --bitrate 3000 --vbv-bufsize 3000 --vbv-maxrate 3000 --hrd
-big_buck_bunny_360p24.y4m,--preset medium --bitrate 400 --vbv-bufsize 600 --vbv-maxrate 600 --aud
-big_buck_bunny_360p24.y4m,--preset medium --crf 1 --vbv-bufsize 3000 --vbv-maxrate 3000 --hrd
-big_buck_bunny_360p24.y4m,--preset superfast --bitrate 1000 --vbv-bufsize 1000 --vbv-maxrate 1000 --aud --strict-cbr
-big_buck_bunny_360p24.y4m,--preset superfast --bitrate 3000 --vbv-bufsize 9000 --vbv-maxrate 9000 --repeat-headers
-big_buck_bunny_360p24.y4m,--preset superfast --bitrate 400 --vbv-bufsize 600 --vbv-maxrate 400 --hrd
-big_buck_bunny_360p24.y4m,--preset superfast --crf 6 --vbv-bufsize 1000 --vbv-maxrate 1000 --aud
-
-# multi-pass rate control tests
-big_buck_bunny_360p24.y4m,--preset slow --crf 40 --pass 1,--preset slow --bitrate 200 --pass 2
-big_buck_bunny_360p24.y4m,--preset medium --bitrate 700 --pass 1 -F4 --slow-firstpass,--preset medium --bitrate 700 --vbv-bufsize 900 --vbv-maxrate 700 --pass 2 -F4
-112_1920x1080_25.yuv,--preset slow --bitrate 1000 --pass 1 -F4,--preset slow --bitrate 1000 --pass 2 -F4
-112_1920x1080_25.yuv,--preset superfast --crf 12 --pass 1,--preset superfast --bitrate 4000 --pass 2 -F4
-RaceHorses_416x240_30_10bit.yuv,--preset veryslow --crf 40 --pass 1, --preset veryslow --bitrate 200 --pass 2 -F4
-RaceHorses_416x240_30_10bit.yuv,--preset superfast --bitrate 600 --pass 1 -F4 --slow-firstpass,--preset superfast --bitrate 600 --pass 2 -F4
-RaceHorses_416x240_30_10bit.yuv,--preset medium --crf 26 --pass 1,--preset medium --bitrate 500 --pass 3 -F4,--preset medium --bitrate 500 --pass 2 -F4
+# List of command lines to be run by rate control regression tests, see https://bitbucket.org/sborho/test-harness
+
+#These tests should yeild deterministic results
+# This test is listed first since it currently reproduces bugs
+big_buck_bunny_360p24.y4m,--preset medium --bitrate 1000 --pass 1 -F4,--preset medium --bitrate 1000 --pass 2 -F4
+fire_1920x1080_30.yuv, --preset slow --bitrate 2000 --tune zero-latency 
+
+
+# VBV tests, non-deterministic so testing for correctness and bitrate
+# fluctuations - up to 1% bitrate fluctuation is allowed between runs
+night_cars_1920x1080_30.yuv,--preset medium --crf 25 --vbv-bufsize 5000 --vbv-maxrate 5000 -F6 --crf-max 34 --crf-min 22
+ducks_take_off_420_720p50.y4m,--preset slow --bitrate 1600 --vbv-bufsize 1600 --vbv-maxrate 1600 --strict-cbr --aq-mode 2 --aq-strength 0.5
+CrowdRun_1920x1080_50_10bit_422.yuv,--preset veryslow --bitrate 4000 --vbv-bufsize 3000 --vbv-maxrate 4000 --tune grain
+fire_1920x1080_30.yuv,--preset medium --bitrate 1000 --vbv-maxrate 1500 --vbv-bufsize 1500 --aud --pmode --tune ssim
+112_1920x1080_25.yuv,--preset ultrafast --bitrate 10000 --vbv-maxrate 10000 --vbv-bufsize 15000 --hrd --strict-cbr
+Traffic_4096x2048_30.yuv,--preset superfast --bitrate 20000 --vbv-maxrate 20000 --vbv-bufsize 20000 --repeat-headers --strict-cbr
+Traffic_4096x2048_30.yuv,--preset faster --bitrate 8000 --vbv-maxrate 8000 --vbv-bufsize 6000 --aud --repeat-headers --no-open-gop --hrd --pmode --pme
+News-4k.y4m,--preset veryfast --bitrate 3000 --vbv-maxrate 5000 --vbv-bufsize 5000 --repeat-headers --temporal-layers
+NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset medium --bitrate 18000 --vbv-bufsize 20000 --vbv-maxrate 18000 --strict-cbr
+NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset medium --bitrate 8000 --vbv-bufsize 12000 --vbv-maxrate 10000  --tune grain
+big_buck_bunny_360p24.y4m,--preset medium --bitrate 400 --vbv-bufsize 600 --vbv-maxrate 600 --aud --hrd --tune fast-decode
+sita_1920x1080_30.yuv,--preset superfast --crf 25 --vbv-bufsize 3000 --vbv-maxrate 4000 --vbv-bufsize 5000 --hrd  --crf-max 30
+sita_1920x1080_30.yuv,--preset superfast --bitrate 3000 --vbv-bufsize 3000 --vbv-maxrate 3000 --aud --strict-cbr
+
+
+
+# multi-pass rate control tests
+big_buck_bunny_360p24.y4m,--preset slow --crf 40 --pass 1 -f 5000,--preset slow --bitrate 200 --pass 2 -f 5000
+big_buck_bunny_360p24.y4m,--preset medium --bitrate 700 --pass 1 -F4 --slow-firstpass -f 5000 ,--preset medium --bitrate 700 --vbv-bufsize 900 --vbv-maxrate 700 --pass 2 -F4 -f 5000
+112_1920x1080_25.yuv,--preset fast --bitrate 1000 --vbv-maxrate 1000 --vbv-bufsize 1000 --strict-cbr --pass 1 -F4,--preset fast --bitrate 1000 --vbv-maxrate 3000 --vbv-bufsize 3000 --pass 2 -F4
+pine_tree_1920x1080_30.yuv,--preset veryfast --crf 12 --pass 1 -F4,--preset faster --bitrate 4000 --pass 2 -F4
+SteamLocomotiveTrain_2560x1600_60_10bit_crop.yuv, --tune grain --preset ultrafast --bitrate 5000 --vbv-maxrate 5000 --vbv-bufsize 8000 --strict-cbr -F4 --pass 1, --tune grain --preset ultrafast --bitrate 8000 --vbv-maxrate 8000 --vbv-bufsize 8000 -F4 --pass 2
+RaceHorses_416x240_30_10bit.yuv,--preset medium --crf 40 --pass 1, --preset faster --bitrate 200 --pass 2 -F4
+CrowdRun_1920x1080_50_10bit_422.yuv,--preset superfast --bitrate 2500 --pass 1 -F4 --slow-firstpass,--preset superfast --bitrate 2500 --pass 2 -F4
+RaceHorses_416x240_30_10bit.yuv,--preset medium --crf 26 --vbv-maxrate 1000 --vbv-bufsize 1000 --pass 1,--preset fast --bitrate 1000  --vbv-maxrate 1000 --vbv-bufsize 700 --pass 3 -F4,--preset slow --bitrate 500 --vbv-maxrate 500  --vbv-bufsize 700 --pass 2 -F4
+

x265_1.6.tar.gz/source/test/regression-tests.txt -> x265_1.7.tar.gz/source/test/regression-tests.txt Changed

@@ -12,9 +12,9 @@
 # not auto-detected.
 
 BasketballDrive_1920x1080_50.y4m,--preset faster --aq-strength 2 --merange 190
-BasketballDrive_1920x1080_50.y4m,--preset medium --ctu 16 --max-tu-size 8 --subme 7
+BasketballDrive_1920x1080_50.y4m,--preset medium --ctu 16 --max-tu-size 8 --subme 7 --qg-size 32
 BasketballDrive_1920x1080_50.y4m,--preset medium --keyint -1 --nr-inter 100 -F4 --no-sao
-BasketballDrive_1920x1080_50.y4m,--preset slow --nr-intra 100 -F4 --aq-strength 3
+BasketballDrive_1920x1080_50.y4m,--preset slow --nr-intra 100 -F4 --aq-strength 3 --qg-size 16
 BasketballDrive_1920x1080_50.y4m,--preset slower --lossless --chromaloc 3 --subme 0
 BasketballDrive_1920x1080_50.y4m,--preset superfast --psy-rd 1 --ctu 16 --no-wpp
 BasketballDrive_1920x1080_50.y4m,--preset ultrafast --signhide --colormatrix bt709
@@ -29,7 +29,7 @@
 CrowdRun_1920x1080_50_10bit_422.yuv,--preset slow --no-wpp --tune ssim --transfer smpte240m
 CrowdRun_1920x1080_50_10bit_422.yuv,--preset slower --tune ssim --tune fastdecode
 CrowdRun_1920x1080_50_10bit_422.yuv,--preset superfast --weightp --no-wpp --sao
-CrowdRun_1920x1080_50_10bit_422.yuv,--preset ultrafast --weightp --tune zerolatency
+CrowdRun_1920x1080_50_10bit_422.yuv,--preset ultrafast --weightp --tune zerolatency --qg-size 16
 CrowdRun_1920x1080_50_10bit_422.yuv,--preset veryfast --temporal-layers --tune grain
 CrowdRun_1920x1080_50_10bit_444.yuv,--preset medium --dither --keyint -1 --rdoq-level 1
 CrowdRun_1920x1080_50_10bit_444.yuv,--preset superfast --weightp --dither --no-psy-rd
@@ -37,8 +37,8 @@
 CrowdRun_1920x1080_50_10bit_444.yuv,--preset veryfast --temporal-layers --repeat-headers
 CrowdRun_1920x1080_50_10bit_444.yuv,--preset veryslow --tskip --tskip-fast --no-scenecut
 DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset medium --tune psnr --bframes 16
-DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset slow --temporal-layers --no-psy-rd
-DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset superfast --weightp
+DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset slow --temporal-layers --no-psy-rd --qg-size 32
+DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset superfast --weightp --qg-size 16
 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
 DucksAndLegs_1920x1080_60_10bit_444.yuv,--preset veryfast --weightp --nr-intra 1000 -F4
@@ -51,11 +51,11 @@
 Kimono1_1920x1080_24_10bit_444.yuv,--preset superfast --weightb
 KristenAndSara_1280x720_60.y4m,--preset medium --no-cutree --max-tu-size 16
 KristenAndSara_1280x720_60.y4m,--preset slower --pmode --max-tu-size 8
-KristenAndSara_1280x720_60.y4m,--preset superfast --min-cu-size 16
+KristenAndSara_1280x720_60.y4m,--preset superfast --min-cu-size 16 --qg-size 16
 KristenAndSara_1280x720_60.y4m,--preset ultrafast --strong-intra-smoothing
 NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset medium --tune grain
 NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset superfast --tune psnr
-News-4k.y4m,--preset medium --tune ssim --no-sao
+News-4k.y4m,--preset medium --tune ssim --no-sao --qg-size 32
 News-4k.y4m,--preset superfast --lookahead-slices 6 --aq-mode 0
 OldTownCross_1920x1080_50_10bit_422.yuv,--preset medium --no-weightp
 OldTownCross_1920x1080_50_10bit_422.yuv,--preset slower --tune fastdecode
@@ -108,13 +108,13 @@
 parkrun_ter_720p50.y4m,--preset slower --fast-intra --no-rect --tune grain
 silent_cif_420.y4m,--preset medium --me full --rect --amp
 silent_cif_420.y4m,--preset superfast --weightp --rect
-silent_cif_420.y4m,--preset placebo --ctu 32 --no-sao
+silent_cif_420.y4m,--preset placebo --ctu 32 --no-sao --qg-size 16
 vtc1nw_422_ntsc.y4m,--preset medium --scaling-list default --ctu 16 --ref 5
-vtc1nw_422_ntsc.y4m,--preset slower --nr-inter 1000 -F4 --tune fast-decode
+vtc1nw_422_ntsc.y4m,--preset slower --nr-inter 1000 -F4 --tune fast-decode --qg-size 16
 vtc1nw_422_ntsc.y4m,--preset superfast --weightp --nr-intra 100 -F4
 washdc_422_ntsc.y4m,--preset faster --rdoq-level 1 --max-merge 5
 washdc_422_ntsc.y4m,--preset medium --no-weightp --max-tu-size 4
-washdc_422_ntsc.y4m,--preset slower --psy-rdoq 2.0 --rdoq-level 2
+washdc_422_ntsc.y4m,--preset slower --psy-rdoq 2.0 --rdoq-level 2 --qg-size 32
 washdc_422_ntsc.y4m,--preset superfast --psy-rd 1 --tune zerolatency
 washdc_422_ntsc.y4m,--preset ultrafast --weightp --tu-intra-depth 4
 washdc_422_ntsc.y4m,--preset veryfast --tu-inter-depth 4

x265_1.6.tar.gz/source/test/smoke-tests.txt -> x265_1.7.tar.gz/source/test/smoke-tests.txt Changed

@@ -1,14 +1,18 @@
 # List of command lines to be run by smoke tests, see https://bitbucket.org/sborho/test-harness
 
+# consider VBV tests a failure if new bitrate is more than 5% different
+# from the old bitrate
+# vbv-tolerance = 0.05
+
 big_buck_bunny_360p24.y4m,--preset=superfast --bitrate 400 --vbv-bufsize 600 --vbv-maxrate 400 --hrd --aud --repeat-headers
 big_buck_bunny_360p24.y4m,--preset=medium --bitrate 1000 -F4 --cu-lossless --scaling-list default
-big_buck_bunny_360p24.y4m,--preset=slower --no-weightp --cu-stats --pme
-washdc_422_ntsc.y4m,--preset=faster --no-strong-intra-smoothing --keyint 1
+big_buck_bunny_360p24.y4m,--preset=slower --no-weightp --cu-stats --pme --qg-size 16
+washdc_422_ntsc.y4m,--preset=faster --no-strong-intra-smoothing --keyint 1 --qg-size 16
 washdc_422_ntsc.y4m,--preset=medium --qp 40 --nr-inter 400 -F4
 washdc_422_ntsc.y4m,--preset=veryslow --pmode --tskip --rdoq-level 0
 old_town_cross_444_720p50.y4m,--preset=ultrafast --weightp --keyint -1
 old_town_cross_444_720p50.y4m,--preset=fast --keyint 20 --min-cu-size 16
-old_town_cross_444_720p50.y4m,--preset=slow --sao-non-deblock --pmode
+old_town_cross_444_720p50.y4m,--preset=slow --sao-non-deblock --pmode --qg-size 32
 RaceHorses_416x240_30_10bit.yuv,--preset=veryfast --cu-stats --max-tu-size 8
 RaceHorses_416x240_30_10bit.yuv,--preset=slower --bitrate 500 -F4 --rdoq-level 1
 CrowdRun_1920x1080_50_10bit_444.yuv,--preset=ultrafast --constrained-intra --min-keyint 5 --keyint 10

x265_1.6.tar.gz/source/test/testbench.cpp -> x265_1.7.tar.gz/source/test/testbench.cpp Changed

x265_1.6.tar.gz/source/x265.cpp -> x265_1.7.tar.gz/source/x265.cpp Changed

@@ -27,6 +27,7 @@
 
 #include "input/input.h"
 #include "output/output.h"
+#include "output/reconplay.h"
 #include "filters/filters.h"
 #include "common.h"
 #include "param.h"
@@ -46,12 +47,16 @@
 #include <string>
 #include <ostream>
 #include <fstream>
+#include <queue>
 
+#define CONSOLE_TITLE_SIZE 200
 #ifdef _WIN32
 #include <windows.h>
+static char orgConsoleTitle[CONSOLE_TITLE_SIZE] = "";
 #else
 #define GetConsoleTitle(t, n)
 #define SetConsoleTitle(t)
+#define SetThreadExecutionState(es)
 #endif
 
 using namespace x265;
@@ -65,33 +70,34 @@
 
 struct CLIOptions
 {
-    Input*  input;
-    Output* recon;
-    std::fstream bitstreamFile;
+    InputFile* input;
+    ReconFile* recon;
+    OutputFile* output;
+    FILE*       qpfile;
+    const char* reconPlayCmd;
+    const x265_api* api;
+    x265_param* param;
     bool bProgress;
     bool bForceY4m;
     bool bDither;
-
     uint32_t seek;              // number of frames to skip from the beginning
     uint32_t framesToBeEncoded; // number of frames to encode
     uint64_t totalbytes;
-    size_t   analysisRecordSize; // number of bytes read from or dumped into file
-    int      analysisHeaderSize;
-
     int64_t startTime;
     int64_t prevUpdateTime;
-    float   frameRate;
-    FILE*   qpfile;
-    FILE*   analysisFile;
 
     /* in microseconds */
     static const int UPDATE_INTERVAL = 250000;
 
     CLIOptions()
     {
-        frameRate = 0.f;
         input = NULL;
         recon = NULL;
+        output = NULL;
+        qpfile = NULL;
+        reconPlayCmd = NULL;
+        api = NULL;
+        param = NULL;
         framesToBeEncoded = seek = 0;
         totalbytes = 0;
         bProgress = true;
@@ -99,18 +105,12 @@
         startTime = x265_mdate();
         prevUpdateTime = 0;
         bDither = false;
-        qpfile = NULL;
-        analysisFile = NULL;
-        analysisRecordSize = 0;
-        analysisHeaderSize = 0;
     }
 
     void destroy();
-    void writeNALs(const x265_nal* nal, uint32_t nalcount);
-    void printStatus(uint32_t frameNum, x265_param *param);
-    bool parse(int argc, char **argv, x265_param* param);
+    void printStatus(uint32_t frameNum);
+    bool parse(int argc, char **argv);
     bool parseQPFile(x265_picture &pic_org);
-    bool validateFanout(x265_param*);
 };
 
 void CLIOptions::destroy()
@@ -124,23 +124,12 @@
     if (qpfile)
         fclose(qpfile);
     qpfile = NULL;
-    if (analysisFile)
-        fclose(analysisFile);
-    analysisFile = NULL;
+    if (output)
+        output->release();
+    output = NULL;
 }
 
-void CLIOptions::writeNALs(const x265_nal* nal, uint32_t nalcount)
-{
-    ProfileScopeEvent(bitstreamWrite);
-    for (uint32_t i = 0; i < nalcount; i++)
-    {
-        bitstreamFile.write((const char*)nal->payload, nal->sizeBytes);
-        totalbytes += nal->sizeBytes;
-        nal++;
-    }
-}
-
-void CLIOptions::printStatus(uint32_t frameNum, x265_param *param)
+void CLIOptions::printStatus(uint32_t frameNum)
 {
     char buf[200];
     int64_t time = x265_mdate();
@@ -167,15 +156,16 @@
     prevUpdateTime = time;
 }
 
-bool CLIOptions::parse(int argc, char **argv, x265_param* param)
+bool CLIOptions::parse(int argc, char **argv)
 {
     bool bError = 0;
     int help = 0;
     int inputBitDepth = 8;
+    int outputBitDepth = 0;
     int reconFileBitDepth = 0;
     const char *inputfn = NULL;
     const char *reconfn = NULL;
-    const char *bitstreamfn = NULL;
+    const char *outputfn = NULL;
     const char *preset = NULL;
     const char *tune = NULL;
     const char *profile = NULL;
@@ -192,15 +182,31 @@
         int c = getopt_long(argc, argv, short_options, long_options, NULL);
         if (c == -1)
             break;
-        if (c == 'p')
+        else if (c == 'p')
             preset = optarg;
-        if (c == 't')
+        else if (c == 't')
             tune = optarg;
+        else if (c == 'D')
+            outputBitDepth = atoi(optarg);
         else if (c == '?')
             showHelp(param);
     }
 
-    if (x265_param_default_preset(param, preset, tune) < 0)
+    api = x265_api_get(outputBitDepth);
+    if (!api)
+    {
+        x265_log(NULL, X265_LOG_WARNING, "falling back to default bit-depth\n");
+        api = x265_api_get(0);
+    }
+
+    param = api->param_alloc();
+    if (!param)
+    {
+        x265_log(NULL, X265_LOG_ERROR, "param alloc failed\n");
+        return true;
+    }
+
+    if (api->param_default_preset(param, preset, tune) < 0)
     {
         x265_log(NULL, X265_LOG_ERROR, "preset or tune unrecognized\n");
         return true;
@@ -211,9 +217,7 @@
         int long_options_index = -1;
         int c = getopt_long(argc, argv, short_options, long_options, &long_options_index);
         if (c == -1)
-        {
             break;
-        }
 
         switch (c)
         {
@@ -261,7 +265,7 @@
             OPT2("frame-skip", "seek") this->seek = (uint32_t)x265_atoi(optarg, bError);
             OPT("frames") this->framesToBeEncoded = (uint32_t)x265_atoi(optarg, bError);
             OPT("no-progress") this->bProgress = false;
-            OPT("output") bitstreamfn = optarg;
+            OPT("output") outputfn = optarg;
             OPT("input") inputfn = optarg;
             OPT("recon") reconfn = optarg;
             OPT("input-depth") inputBitDepth = (uint32_t)x265_atoi(optarg, bError);
@@ -271,17 +275,19 @@
             OPT("profile") profile = optarg; /* handled last */
             OPT("preset") /* handled above */;
             OPT("tune")   /* handled above */;
+            OPT("output-depth")   /* handled above */;
+            OPT("recon-y4m-exec") reconPlayCmd = optarg;
             OPT("qpfile")
             {
                 this->qpfile = fopen(optarg, "rb");
                 if (!this->qpfile)
                 {
-                    x265_log(param, X265_LOG_ERROR, "%s qpfile not found or error in opening qp file \n", optarg);
+                    x265_log(param, X265_LOG_ERROR, "%s qpfile not found or error in opening qp file\n", optarg);
                     return false;
                 }
             }
             else
-                bError |= !!x265_param_parse(param, long_options[long_options_index].name, optarg);
+                bError |= !!api->param_parse(param, long_options[long_options_index].name, optarg);
 
             if (bError)
             {
@@ -295,8 +301,8 @@
 
     if (optind < argc && !inputfn)
         inputfn = argv[optind++];
-    if (optind < argc && !bitstreamfn)
-        bitstreamfn = argv[optind++];
+    if (optind < argc && !outputfn)
+        outputfn = argv[optind++];
     if (optind < argc)
     {
         x265_log(param, X265_LOG_WARNING, "extra unused command arguments given <%s>\n", argv[optind]);
@@ -306,15 +312,15 @@
     if (argc <= 1 || help)
         showHelp(param);
 
-    if (inputfn == NULL || bitstreamfn == NULL)
+    if (inputfn == NULL || outputfn == NULL)
     {
         x265_log(param, X265_LOG_ERROR, "input or output file not specified, try -V for help\n");
         return true;
     }
 
-    if (param->internalBitDepth != x265_max_bit_depth)
+    if (param->internalBitDepth != api->max_bit_depth)
     {
-        x265_log(param, X265_LOG_ERROR, "Only bit depths of %d are supported in this build\n", x265_max_bit_depth);
+        x265_log(param, X265_LOG_ERROR, "Only bit depths of %d are supported in this build\n", api->max_bit_depth);
         return true;
     }
 
@@ -332,7 +338,7 @@
     info.frameCount = 0;
     getParamAspectRatio(param, info.sarWidth, info.sarHeight);
 
-    this->input = Input::open(info, this->bForceY4m);
+    this->input = InputFile::open(info, this->bForceY4m);
     if (!this->input || this->input->isFail())
     {
         x265_log(param, X265_LOG_ERROR, "unable to open input file <%s>\n", inputfn);
@@ -362,7 +368,11 @@
         this->framesToBeEncoded = info.frameCount - seek;
     param->totalFrames = this->framesToBeEncoded;
 
-    if (x265_param_apply_profile(param, profile))
+    /* Force CFR until we have support for VFR */
+    info.timebaseNum = param->fpsDenom;
+    info.timebaseDenom = param->fpsNum;
+
+    if (api->param_apply_profile(param, profile))
         return true;
 
     if (param->logLevel >= X265_LOG_INFO)
@@ -381,7 +391,7 @@
         else
             sprintf(buf + p, " frames %u - %d of %d", this->seek, this->seek + this->framesToBeEncoded - 1, info.frameCount);
 
-        fprintf(stderr, "%s  [info]: %s\n", input->getName(), buf);
+        general_log(param, input->getName(), X265_LOG_INFO, "%s\n", buf);
     }
 
     this->input->startReader();
@@ -390,26 +400,28 @@
     {
         if (reconFileBitDepth == 0)
             reconFileBitDepth = param->internalBitDepth;
-        this->recon = Output::open(reconfn, param->sourceWidth, param->sourceHeight, reconFileBitDepth,
-                                   param->fpsNum, param->fpsDenom, param->internalCsp);
+        this->recon = ReconFile::open(reconfn, param->sourceWidth, param->sourceHeight, reconFileBitDepth,
+                                      param->fpsNum, param->fpsDenom, param->internalCsp);
         if (this->recon->isFail())
         {
-            x265_log(param, X265_LOG_WARNING, "unable to write reconstruction file\n");
+            x265_log(param, X265_LOG_WARNING, "unable to write reconstructed outputs file\n");
             this->recon->release();
             this->recon = 0;
         }
         else
-            fprintf(stderr, "%s  [info]: reconstructed images %dx%d fps %d/%d %s\n", this->recon->getName(),
+            general_log(param, this->recon->getName(), X265_LOG_INFO,
+                    "reconstructed images %dx%d fps %d/%d %s\n",
                     param->sourceWidth, param->sourceHeight, param->fpsNum, param->fpsDenom,
                     x265_source_csp_names[param->internalCsp]);
     }
 
-    this->bitstreamFile.open(bitstreamfn, std::fstream::binary | std::fstream::out);
-    if (!this->bitstreamFile)
+    this->output = OutputFile::open(outputfn, info);
+    if (this->output->isFail())
     {
-        x265_log(NULL, X265_LOG_ERROR, "failed to open bitstream file <%s> for writing\n", bitstreamfn);
+        x265_log(param, X265_LOG_ERROR, "failed to open output file <%s> for writing\n", outputfn);
         return true;
     }
+    general_log(param, this->output->getName(), X265_LOG_INFO, "output file: %s\n", outputfn);
     return false;
 }
 
@@ -464,28 +476,45 @@
     PROFILE_INIT();
     THREAD_NAME("API", 0);
 
-    x265_param *param = x265_param_alloc();
+    GetConsoleTitle(orgConsoleTitle, CONSOLE_TITLE_SIZE);
+    SetThreadExecutionState(ES_CONTINUOUS | ES_SYSTEM_REQUIRED | ES_AWAYMODE_REQUIRED);
+
+    ReconPlay* reconPlay = NULL;
     CLIOptions cliopt;
 
-    if (cliopt.parse(argc, argv, param))
+    if (cliopt.parse(argc, argv))
     {
         cliopt.destroy();
-        x265_param_free(param);
+        if (cliopt.api)
+            cliopt.api->param_free(cliopt.param);
         exit(1);
     }
 
-    x265_encoder *encoder = x265_encoder_open(param);
+    x265_param* param = cliopt.param;
+    const x265_api* api = cliopt.api;
+
+    /* This allows muxers to modify bitstream format */
+    cliopt.output->setParam(param);
+
+    if (cliopt.reconPlayCmd)
+        reconPlay = new ReconPlay(cliopt.reconPlayCmd, *param);
+
+    /* note: we could try to acquire a different libx265 API here based on
+     * the profile found during option parsing, but it must be done before
+     * opening an encoder */
+
+    x265_encoder *encoder = api->encoder_open(param);
     if (!encoder)
     {
         x265_log(param, X265_LOG_ERROR, "failed to open encoder\n");
         cliopt.destroy();
-        x265_param_free(param);
-        x265_cleanup();
+        api->param_free(param);
+        api->cleanup();
         exit(2);
     }
 
     /* get the encoder parameters post-initialization */
-    x265_encoder_parameters(encoder, param);
+    api->encoder_parameters(encoder, param);
 
     /* Control-C handler */
     if (signal(SIGINT, sigint_handler) == SIG_ERR)
@@ -494,7 +523,8 @@
     x265_picture pic_orig, pic_out;
     x265_picture *pic_in = &pic_orig;
     /* Allocate recon picture if analysisMode is enabled */
-    x265_picture *pic_recon = (cliopt.recon || !!param->analysisMode) ? &pic_out : NULL;
+    std::priority_queue<int64_t>* pts_queue = cliopt.output->needPTS() ? new std::priority_queue<int64_t>() : NULL;
+    x265_picture *pic_recon = (cliopt.recon || !!param->analysisMode || pts_queue || reconPlay) ? &pic_out : NULL;
     uint32_t inFrameCount = 0;
     uint32_t outFrameCount = 0;
     x265_nal *p_nal;
@@ -505,17 +535,17 @@
 
     if (!param->bRepeatHeaders)
     {
-        if (x265_encoder_headers(encoder, &p_nal, &nal) < 0)
+        if (api->encoder_headers(encoder, &p_nal, &nal) < 0)
         {
             x265_log(param, X265_LOG_ERROR, "Failure generating stream headers\n");
             ret = 3;
             goto fail;
         }
         else
-            cliopt.writeNALs(p_nal, nal);
+            cliopt.totalbytes += cliopt.output->writeHeaders(p_nal, nal);
     }
 
-    x265_picture_init(param, pic_in);
+    api->picture_init(param, pic_in);
 
     if (cliopt.bDither)
     {
@@ -549,46 +579,72 @@
 
         if (pic_in)
         {
-            if (pic_in->bitDepth > X265_DEPTH && cliopt.bDither)
+            if (pic_in->bitDepth > param->internalBitDepth && cliopt.bDither)
             {
-                ditherImage(*pic_in, param->sourceWidth, param->sourceHeight, errorBuf, X265_DEPTH);
-                pic_in->bitDepth = X265_DEPTH;
+                ditherImage(*pic_in, param->sourceWidth, param->sourceHeight, errorBuf, param->internalBitDepth);
+                pic_in->bitDepth = param->internalBitDepth;
             }
+            /* Overwrite PTS */
+            pic_in->pts = pic_in->poc;
         }
 
-        int numEncoded = x265_encoder_encode(encoder, &p_nal, &nal, pic_in, pic_recon);
+        int numEncoded = api->encoder_encode(encoder, &p_nal, &nal, pic_in, pic_recon);
         if (numEncoded < 0)
         {
             b_ctrl_c = 1;
             ret = 4;
             break;
         }
+
+        if (reconPlay && numEncoded)
+            reconPlay->writePicture(*pic_recon);
+
         outFrameCount += numEncoded;
 
         if (numEncoded && pic_recon && cliopt.recon)
             cliopt.recon->writePicture(pic_out);
         if (nal)
-            cliopt.writeNALs(p_nal, nal);
+        {
+            cliopt.totalbytes += cliopt.output->writeFrame(p_nal, nal, pic_out);
+            if (pts_queue)
+            {
+                pts_queue->push(-pic_out.pts);
+                if (pts_queue->size() > 2)
+                    pts_queue->pop();
+            }
+        }
 
-        cliopt.printStatus(outFrameCount, param);
+        cliopt.printStatus(outFrameCount);
     }
 
     /* Flush the encoder */
     while (!b_ctrl_c)
     {
-        int numEncoded = x265_encoder_encode(encoder, &p_nal, &nal, NULL, pic_recon);
+        int numEncoded = api->encoder_encode(encoder, &p_nal, &nal, NULL, pic_recon);
         if (numEncoded < 0)
         {
             ret = 4;
             break;
         }
+
+        if (reconPlay && numEncoded)
+            reconPlay->writePicture(*pic_recon);
+
         outFrameCount += numEncoded;
         if (numEncoded && pic_recon && cliopt.recon)
             cliopt.recon->writePicture(pic_out);
         if (nal)
-            cliopt.writeNALs(p_nal, nal);
+        {
+            cliopt.totalbytes += cliopt.output->writeFrame(p_nal, nal, pic_out);
+            if (pts_queue)
+            {
+                pts_queue->push(-pic_out.pts);
+                if (pts_queue->size() > 2)
+                    pts_queue->pop();
+            }
+        }
 
-        cliopt.printStatus(outFrameCount, param);
+        cliopt.printStatus(outFrameCount);
 
         if (!numEncoded)
             break;
@@ -599,42 +655,62 @@
         fprintf(stderr, "%*s\r", 80, " ");
 
 fail:
-    x265_encoder_get_stats(encoder, &stats, sizeof(stats));
+
+    delete reconPlay;
+
+    api->encoder_get_stats(encoder, &stats, sizeof(stats));
     if (param->csvfn && !b_ctrl_c)
-        x265_encoder_log(encoder, argc, argv);
-    x265_encoder_close(encoder);
-    cliopt.bitstreamFile.close();
+        api->encoder_log(encoder, argc, argv);
+    api->encoder_close(encoder);
+
+    int64_t second_largest_pts = 0;
+    int64_t largest_pts = 0;
+    if (pts_queue && pts_queue->size() >= 2)
+    {
+        second_largest_pts = -pts_queue->top();
+        pts_queue->pop();
+        largest_pts = -pts_queue->top();
+        pts_queue->pop();
+        delete pts_queue;
+        pts_queue = NULL;
+    }
+    cliopt.output->closeFile(largest_pts, second_largest_pts);
 
     if (b_ctrl_c)
-        fprintf(stderr, "aborted at input frame %d, output frame %d\n",
-                cliopt.seek + inFrameCount, stats.encodedPictureCount);
+        general_log(param, NULL, X265_LOG_INFO, "aborted at input frame %d, output frame %d\n",
+                    cliopt.seek + inFrameCount, stats.encodedPictureCount);
 
     if (stats.encodedPictureCount)
     {
-        printf("\nencoded %d frames in %.2fs (%.2f fps), %.2f kb/s", stats.encodedPictureCount,
-               stats.elapsedEncodeTime, stats.encodedPictureCount / stats.elapsedEncodeTime, stats.bitrate);
+        char buffer[4096];
+        int p = sprintf(buffer, "\nencoded %d frames in %.2fs (%.2f fps), %.2f kb/s", stats.encodedPictureCount,
+                        stats.elapsedEncodeTime, stats.encodedPictureCount / stats.elapsedEncodeTime, stats.bitrate);
 
         if (param->bEnablePsnr)
-            printf(", Global PSNR: %.3f", stats.globalPsnr);
+            p += sprintf(buffer + p, ", Global PSNR: %.3f", stats.globalPsnr);
 
         if (param->bEnableSsim)
-            printf(", SSIM Mean Y: %.7f (%6.3f dB)", stats.globalSsim, x265_ssim2dB(stats.globalSsim));
+            p += sprintf(buffer + p, ", SSIM Mean Y: %.7f (%6.3f dB)", stats.globalSsim, x265_ssim2dB(stats.globalSsim));
 
-        printf("\n");
+        sprintf(buffer + p, "\n");
+        general_log(param, NULL, X265_LOG_INFO, buffer);
     }
     else
     {
-        printf("\nencoded 0 frames\n");
+        general_log(param, NULL, X265_LOG_INFO, "\nencoded 0 frames\n");
     }
 
-    x265_cleanup(); /* Free library singletons */
+    api->cleanup(); /* Free library singletons */
 
     cliopt.destroy();
 
-    x265_param_free(param);
+    api->param_free(param);
 
     X265_FREE(errorBuf);
 
+    SetConsoleTitle(orgConsoleTitle);
+    SetThreadExecutionState(ES_CONTINUOUS);
+
 #if HAVE_VLD
     assert(VLDReportLeaks() == 0);
 #endif

x265_1.6.tar.gz/source/x265.def.in -> x265_1.7.tar.gz/source/x265.def.in Changed

x265_1.6.tar.gz/source/x265.h -> x265_1.7.tar.gz/source/x265.h Changed

@@ -416,7 +416,7 @@
      *
      * Frame encoders are distributed between the available thread pools, and
      * the encoder will never generate more thread pools than frameNumThreads */
-    char*     numaPools;
+    const char* numaPools;
 
     /* Enable wavefront parallel processing, greatly increases parallelism for
      * less than 1% compression efficiency loss. Requires a thread pool, enabled
@@ -458,7 +458,7 @@
      * order. Otherwise the encoder will emit per-stream statistics into the log
      * file when x265_encoder_log is called (presumably at the end of the
      * encode) */
-    char*     csvfn;
+    const char* csvfn;
 
     /*== Internal Picture Specification ==*/
 
@@ -522,12 +522,21 @@
      * performance. Value must be between 1 and 16, default is 3 */
     int       maxNumReferences;
 
+    /* Allow libx265 to emit HEVC bitstreams which do not meet strict level
+     * requirements. Defaults to false */
+    int       bAllowNonConformance;
+
     /*== Bitstream Options ==*/
 
     /* Flag indicating whether VPS, SPS and PPS headers should be output with
      * each keyframe. Default false */
     int       bRepeatHeaders;
 
+    /* Flag indicating whether the encoder should generate start codes (Annex B
+     * format) or length (file format) before NAL units. Default true, Annex B.
+     * Muxers should set this to the correct value */
+    int       bAnnexB;
+
     /* Flag indicating whether the encoder should emit an Access Unit Delimiter
      * NAL at the start of every access unit. Default false */
     int       bEnableAccessUnitDelimiters;
@@ -869,7 +878,7 @@
     int       analysisMode;
 
     /* Filename for analysisMode save/load. Default name is "x265_analysis.dat" */
-    char*     analysisFileName;
+    const char* analysisFileName;
 
     /*== Rate Control ==*/
 
@@ -962,7 +971,7 @@
 
         /* Filename of the 2pass output/input stats file, if unspecified the
          * encoder will default to using x265_2pass.log */
-        char*     statFileName;
+        const char* statFileName;
 
         /* temporally blur quants */
         double    qblur;
@@ -988,6 +997,12 @@
         /* Enable stricter conditions to check bitrate deviations in CBR mode. May compromise 
          * quality to maintain bitrate adherence */
         int bStrictCbr;
+
+        /* Enable adaptive quantization at CU granularity. This parameter specifies 
+         * the minimum CU size at which QP can be adjusted, i.e. Quantization Group 
+         * (QG) size. Allowed values are 64, 32, 16 provided it falls within the 
+         * inclusuve range [maxCUSize, minCUSize]. Experimental, default: maxCUSize*/
+        uint32_t qgSize;
     } rc;
 
     /*== Video Usability Information ==*/
@@ -1084,6 +1099,22 @@
          * conformance cropping window to further crop the displayed window */
         int defDispWinBottomOffset;
     } vui;
+
+    /* SMPTE ST 2086 mastering display color volume SEI info, specified as a
+     * string which is parsed when the stream header SEI are emitted. The string
+     * format is "G(%hu,%hu)B(%hu,%hu)R(%hu,%hu)WP(%hu,%hu)L(%u,%u)" where %hu
+     * are unsigned 16bit integers and %u are unsigned 32bit integers. The SEI
+     * includes X,Y display primaries for RGB channels, white point X,Y and
+     * max,min luminance values. */
+    const char* masteringDisplayColorVolume;
+
+    /* Content light level info SEI, specified as a string which is parsed when
+     * the stream header SEI are emitted. The string format is "%hu,%hu" where
+     * %hu are unsigned 16bit integers. The first value is the max content light
+     * level (or 0 if no maximum is indicated), the second value is the maximum
+     * picture average light level (or 0). */
+    const char* contentLightLevelInfo;
+
 } x265_param;
 
 /* x265_param_alloc:
@@ -1162,12 +1193,10 @@
 void x265_picture_init(x265_param *param, x265_picture *pic);
 
 /* x265_max_bit_depth:
- *      Specifies the maximum number of bits per pixel that x265 can input. This
- *      is also the max bit depth that x265 encodes in.  When x265_max_bit_depth
- *      is 8, the internal and input bit depths must be 8.  When
- *      x265_max_bit_depth is 12, the internal and input bit depths can be
- *      either 8, 10, or 12. Note that the internal bit depth must be the same
- *      for all encoders allocated in the same process. */
+ *      Specifies the numer of bits per pixel that x265 uses internally to
+ *      represent a pixel, and the bit depth of the output bitstream.
+ *      param->internalBitDepth must be set to this value. x265_max_bit_depth
+ *      will be 8 for default builds, 10 for HIGH_BIT_DEPTH builds. */
 X265_API extern const int x265_max_bit_depth;
 
 /* x265_version_str:
@@ -1214,6 +1243,21 @@
  *      Once flushing has begun, all subsequent calls must pass pic_in as NULL. */
 int x265_encoder_encode(x265_encoder *encoder, x265_nal **pp_nal, uint32_t *pi_nal, x265_picture *pic_in, x265_picture *pic_out);
 
+/* x265_encoder_reconfig:
+ *      various parameters from x265_param are copied.
+ *      this takes effect immediately, on whichever frame is encoded next;
+ *      returns 0 on success, negative on parameter validation error.
+ *
+ *      not all parameters can be changed; see the actual function for a
+ *      detailed breakdown.  since not all parameters can be changed, moving
+ *      from preset to preset may not always fully copy all relevant parameters,
+ *      but should still work usably in practice. however, more so than for
+ *      other presets, many of the speed shortcuts used in ultrafast cannot be
+ *      switched out of; using reconfig to switch between ultrafast and other
+ *      presets is not recommended without a more fine-grained breakdown of
+ *      parameters to take this into account. */
+int x265_encoder_reconfig(x265_encoder *, x265_param *);
+
 /* x265_encoder_get_stats:
  *       returns encoder statistics */
 void x265_encoder_get_stats(x265_encoder *encoder, x265_stats *, uint32_t statsSizeBytes);
@@ -1253,6 +1297,7 @@
     void          (*picture_init)(x265_param*, x265_picture*);
     x265_encoder* (*encoder_open)(x265_param*);
     void          (*encoder_parameters)(x265_encoder*, x265_param*);
+    int           (*encoder_reconfig)(x265_encoder*, x265_param*);
     int           (*encoder_headers)(x265_encoder*, x265_nal**, uint32_t*);
     int           (*encoder_encode)(x265_encoder*, x265_nal**, uint32_t*, x265_picture*, x265_picture*);
     void          (*encoder_get_stats)(x265_encoder*, x265_stats*, uint32_t);
@@ -1275,8 +1320,14 @@
  *   Retrieve the programming interface for a linked x265 library.
  *   May return NULL if no library is available that supports the
  *   requested bit depth. If bitDepth is 0 the function is guarunteed
- *   to return a non-NULL x265_api pointer, from the system default
- *   libx265 */
+ *   to return a non-NULL x265_api pointer, from the linked libx265.
+ *
+ *   If the requested bitDepth is not supported by the linked libx265,
+ *   it will attempt to dynamically bind x265_api_get() from a shared
+ *   library with an appropriate name:
+ *     8bit:  libx265_main.so
+ *     10bit: libx265_main10.so
+ *   Obviously the shared library file extension is platform specific */
 const x265_api* x265_api_get(int bitDepth);
 
 #ifdef __cplusplus

x265_1.6.tar.gz/source/x265cli.h -> x265_1.7.tar.gz/source/x265cli.h Changed

@@ -30,7 +30,7 @@
 namespace x265 {
 #endif
 
-static const char short_options[] = "o:p:f:F:r:I:i:b:s:t:q:m:hwV?";
+static const char short_options[] = "o:D:P:p:f:F:r:I:i:b:s:t:q:m:hwV?";
 static const struct option long_options[] =
 {
     { "help",                 no_argument, NULL, 'h' },
@@ -47,16 +47,19 @@
     { "no-pme",               no_argument, NULL, 0 },
     { "pme",                  no_argument, NULL, 0 },
     { "log-level",      required_argument, NULL, 0 },
-    { "profile",        required_argument, NULL, 0 },
+    { "profile",        required_argument, NULL, 'P' },
     { "level-idc",      required_argument, NULL, 0 },
     { "high-tier",            no_argument, NULL, 0 },
     { "no-high-tier",         no_argument, NULL, 0 },
+    { "allow-non-conformance",no_argument, NULL, 0 },
+    { "no-allow-non-conformance",no_argument, NULL, 0 },
     { "csv",            required_argument, NULL, 0 },
     { "no-cu-stats",          no_argument, NULL, 0 },
     { "cu-stats",             no_argument, NULL, 0 },
     { "y4m",                  no_argument, NULL, 0 },
     { "no-progress",          no_argument, NULL, 0 },
     { "output",         required_argument, NULL, 'o' },
+    { "output-depth",   required_argument, NULL, 'D' },
     { "input",          required_argument, NULL, 0 },
     { "input-depth",    required_argument, NULL, 0 },
     { "input-res",      required_argument, NULL, 0 },
@@ -181,6 +184,8 @@
     { "colormatrix",    required_argument, NULL, 0 },
     { "chromaloc",      required_argument, NULL, 0 },
     { "crop-rect",      required_argument, NULL, 0 },
+    { "master-display", required_argument, NULL, 0 },
+    { "max-cll",        required_argument, NULL, 0 },
     { "no-dither",            no_argument, NULL, 0 },
     { "dither",               no_argument, NULL, 0 },
     { "no-repeat-headers",    no_argument, NULL, 0 },
@@ -205,6 +210,8 @@
     { "strict-cbr",           no_argument, NULL, 0 },
     { "temporal-layers",      no_argument, NULL, 0 },
     { "no-temporal-layers",   no_argument, NULL, 0 },
+    { "qg-size",        required_argument, NULL, 0 },
+    { "recon-y4m-exec", required_argument, NULL, 0 },
     { 0, 0, 0, 0 },
     { 0, 0, 0, 0 },
     { 0, 0, 0, 0 },
@@ -236,6 +243,7 @@
     H0("-V/--version                     Show version info and exit\n");
     H0("\nOutput Options:\n");
     H0("-o/--output <filename>           Bitstream output file name\n");
+    H0("-D/--output-depth 8|10           Output bit depth (also internal bit depth). Default %d\n", param->internalBitDepth);
     H0("   --log-level <string>          Logging level: none error warning info debug full. Default %s\n", x265::logLevelNames[param->logLevel + 1]);
     H0("   --no-progress                 Disable CLI progress reports\n");
     H0("   --[no-]cu-stats               Enable logging stats about distribution of cu across all modes. Default %s\n",OPT(param->bLogCuStats));
@@ -255,9 +263,10 @@
     H0("   --[no-]ssim                   Enable reporting SSIM metric scores. Default %s\n", OPT(param->bEnableSsim));
     H0("   --[no-]psnr                   Enable reporting PSNR metric scores. Default %s\n", OPT(param->bEnablePsnr));
     H0("\nProfile, Level, Tier:\n");
-    H0("   --profile <string>            Enforce an encode profile: main, main10, mainstillpicture\n");
+    H0("-P/--profile <string>            Enforce an encode profile: main, main10, mainstillpicture\n");
     H0("   --level-idc <integer|float>   Force a minimum required decoder level (as '5.0' or '50')\n");
     H0("   --[no-]high-tier              If a decoder level is specified, this modifier selects High tier of that level\n");
+    H0("   --[no-]allow-non-conformance  Allow the encoder to generate profile NONE bitstreams. Default %s\n", OPT(param->bAllowNonConformance));
     H0("\nThreading, performance:\n");
     H0("   --pools <integer,...>         Comma separated thread count per thread pool (pool per NUMA node)\n");
     H0("                                 '-' implies no threads on node, '+' implies one thread per core on node\n");
@@ -352,12 +361,14 @@
     H0("   --analysis-file <filename>    Specify file name used for either dumping or reading analysis data.\n");
     H0("   --aq-mode <integer>           Mode for Adaptive Quantization - 0:none 1:uniform AQ 2:auto variance. Default %d\n", param->rc.aqMode);
     H0("   --aq-strength <float>         Reduces blocking and blurring in flat and textured areas (0 to 3.0). Default %.2f\n", param->rc.aqStrength);
+    H0("   --qg-size <int>               Specifies the size of the quantization group (64, 32, 16). Default %d\n", param->rc.qgSize);
     H0("   --[no-]cutree                 Enable cutree for Adaptive Quantization. Default %s\n", OPT(param->rc.cuTree));
     H1("   --ipratio <float>             QP factor between I and P. Default %.2f\n", param->rc.ipFactor);
     H1("   --pbratio <float>             QP factor between P and B. Default %.2f\n", param->rc.pbFactor);
     H1("   --qcomp <float>               Weight given to predicted complexity. Default %.2f\n", param->rc.qCompress);
-    H1("   --cbqpoffs <integer>          Chroma Cb QP Offset. Default %d\n", param->cbQpOffset);
-    H1("   --crqpoffs <integer>          Chroma Cr QP Offset. Default %d\n", param->crQpOffset);
+    H1("   --qpstep <integer>            The maximum single adjustment in QP allowed to rate control. Default %d\n", param->rc.qpStep);
+    H1("   --cbqpoffs <integer>          Chroma Cb QP Offset [-12..12]. Default %d\n", param->cbQpOffset);
+    H1("   --crqpoffs <integer>          Chroma Cr QP Offset [-12..12]. Default %d\n", param->crQpOffset);
     H1("   --scaling-list <string>       Specify a file containing HM style quant scaling lists or 'default' or 'off'. Default: off\n");
     H1("   --lambda-file <string>        Specify a file containing replacement values for the lambda tables\n");
     H1("                                 MAX_MAX_QP+1 floats for lambda table, then again for lambda2 table\n");
@@ -384,6 +395,9 @@
     H1("   --colormatrix <string>        Specify color matrix setting from undef, bt709, fcc, bt470bg, smpte170m,\n");
     H1("                                 smpte240m, GBR, YCgCo, bt2020nc, bt2020c. Default undef\n");
     H1("   --chromaloc <integer>         Specify chroma sample location (0 to 5). Default of %d\n", param->vui.chromaSampleLocTypeTopField);
+    H0("   --master-display <string>     SMPTE ST 2086 master display color volume info SEI (HDR)\n");
+    H0("                                    format: G(x,y)B(x,y)R(x,y)WP(x,y)L(max,min)\n");
+    H0("   --max-cll <string>            Emit content light level info SEI as \"cll,fall\" (HDR)\n");
     H0("\nBitstream options:\n");
     H0("   --[no-]repeat-headers         Emit SPS and PPS headers at each keyframe. Default %s\n", OPT(param->bRepeatHeaders));
     H0("   --[no-]info                   Emit SEI identifying encoder and parameters. Default %s\n", OPT(param->bEmitInfoSEI));
@@ -394,6 +408,7 @@
     H1("\nReconstructed video options (debugging):\n");
     H1("-r/--recon <filename>            Reconstructed raw image YUV or Y4M output file name\n");
     H1("   --recon-depth <integer>       Bit-depth of reconstructed raw image file. Defaults to input bit depth, or 8 if Y4M\n");
+    H1("   --recon-y4m-exec <string>     pipe reconstructed frames to Y4M viewer, ex:\"ffplay -i pipe:0 -autoexit\"\n");
     H1("\nExecutable return codes:\n");
     H1("    0 - encode successful\n");
     H1("    1 - unable to parse command line\n");

Build Results
RPM Lint

Refresh

Comments for request 1963 0

Request History

Aloysius created request over 9 years ago

Updated to 1.7

scarabeus accepted request over 9 years ago

Thanks for the bump