Changes of Revision 88
obs-studio.changes
Changed
x
1
2
-------------------------------------------------------------------3
+Wed Jan 06 18:27:38 UTC 2021 - jimmy@boombatower.com4
+5
+- Update to version 26.1.1:6
+ * win-dshow: Fix dshowcapture not linking audio of certain devices7
+ * linux-jack: fix deadlock when closing the client8
+ * linux-jack: mark ports as JackPortIsTerminal9
+ * linux-jack: fix timestamp calculation10
+ * obs-browser: Initialize CEF early to fix macOS crash11
+ * libobs: Update version to 26.1.112
+ * rtmp-services: Add Loola.tv service13
+ * rtmp-services: Fix json formatting14
+ * libobs: Avoid unnecessary mallocs in audio processing15
+ * UI: Fix padding on Acri context bar buttons16
+ * image-source: Fix slideshow transition bug when randomized17
+ * docs/sphinx: Add missing obs_frontend_open_projector18
+ * libobs: Update to SIMDe 0.7.119
+ * libobs: Set lock state when duplicating scene item20
+ * libobs: Add definitions in ARCH_SIMD_DEFINES21
+ * cmake: Add ARCH_SIMD_DEFINES variable22
+ * coreaudio-encoder: Fix cmake for mingw23
+ * Revert "UI: Only apply new scaling behavior on newer installs"24
+ * UI: Only apply new scaling behavior on newer installs25
+ * UI: Support fractional scaling for Canvas/Base size26
+ * mac-virtualcam: Remove unnecessary logging27
+ * mac-virtualcam: Mark parameters as unused28
+ * image-source: Add .webp to "All formats" option29
+ * image-source: Add webp to file filter30
+ * CI: Remove jack, speex and fdk-aac from default builds for macOS31
+ * libobs, obs-ffmpeg: Use correct value for EINVAL error check32
+ * UI/updater: Increase number of download workers33
+ * UI/updater: Enable HTTP2 and TLS 1.334
+ * UI: Fix name of kab-KAB locale35
+ * decklink: Fix automatic pixel format detection36
+ * CI: Fix macOS 10.13 crashes due to unsupported library symbols37
+ * UI/installer: Add additional VS2019 DLL check38
+ * mac-virtualcam: Fix file mode39
+ * CI: Run make with -j$(nproc)40
+ * CI: Remove obsolete and unused files41
+ * libobs: Add texture sharing support for macOS/OpenGL42
+ * CI: Add necessary changes for CEF 418343
+ * UI/updater: Move in-use files away before writing44
+ * UI/updater: Always clean up temporary files45
+ * UI: Remove Smashcast from AutoConfig46
+ * rtmp-services: Remove Smashcast47
+48
+-------------------------------------------------------------------49
Tue Dec 15 23:25:38 UTC 2020 - Jimmy Berry <jimmy@boombatower.com>50
51
- Add modinfo-use-full-path.patch for new v4l2lookback support.52
obs-studio.spec
Changed
8
1
2
Name: obs-studio3
-Version: 26.1.04
+Version: 26.1.15
Release: 06
Summary: A recording/broadcasting program7
Group: Productivity/Multimedia/Video/Editors and Convertors8
_service
Changed
10
1
2
<services>3
<service name="tar_scm" mode="disabled">4
<param name="versionformat">@PARENT_TAG@</param>5
- <param name="revision">refs/tags/26.1.0</param>6
+ <param name="revision">refs/tags/26.1.1</param>7
<param name="url">git://github.com/jp9000/obs-studio.git</param>8
<param name="scm">git</param>9
<param name="changesgenerate">enable</param>10
_servicedata
Changed
9
1
2
<servicedata>3
<service name="tar_scm">4
<param name="url">git://github.com/jp9000/obs-studio.git</param>5
- <param name="changesrevision">38ad3ba18fc27846e122bd56f589ccb34c4578e2</param>6
+ <param name="changesrevision">dffa8221124106bc2a4c92e5f5d0fa21128a61f6</param>7
</service>8
</servicedata>9
obs-studio-26.1.0.tar.xz/CI/before-deploy-osx.sh
Deleted
116
1
2
-hr() {3
- echo "───────────────────────────────────────────────────"4
- echo $15
- echo "───────────────────────────────────────────────────"6
-}7
-8
-# Exit if something fails9
-set -e10
-11
-# Generate file name variables12
-export GIT_TAG=$(git describe --abbrev=0)13
-export GIT_HASH=$(git rev-parse --short HEAD)14
-export FILE_DATE=$(date +%Y-%m-%d.%H-%M-%S)15
-export FILENAME=$FILE_DATE-$GIT_HASH-$TRAVIS_BRANCH-osx.dmg16
-17
-echo "git tag: $GIT_TAG"18
-19
-cd ./build20
-21
-# Move obslua22
-hr "Moving OBS LUA"23
-mv ./rundir/RelWithDebInfo/data/obs-scripting/obslua.so ./rundir/RelWithDebInfo/bin/24
-25
-# Move obspython26
-hr "Moving OBS Python"27
-# mv ./rundir/RelWithDebInfo/data/obs-scripting/_obspython.so ./rundir/RelWithDebInfo/bin/28
-# mv ./rundir/RelWithDebInfo/data/obs-scripting/obspython.py ./rundir/RelWithDebInfo/bin/29
-30
-# Package everything into a nice .app31
-hr "Packaging .app"32
-STABLE=false33
-if [ -n "${TRAVIS_TAG}" ]; then34
- STABLE=true35
-fi36
-37
-#sudo python ../CI/install/osx/build_app.py --public-key ../CI/install/osx/OBSPublicDSAKey.pem --sparkle-framework ../../sparkle/Sparkle.framework --stable=$STABLE38
-39
-../CI/install/osx/packageApp.sh40
-41
-# fix obs outputs plugin it doesn't play nicely with dylibBundler at the moment42
-if [ -f /usr/local/opt/mbedtls/lib/libmbedtls.12.dylib ]; then43
- cp /usr/local/opt/mbedtls/lib/libmbedtls.12.dylib ./OBS.app/Contents/Frameworks/44
- cp /usr/local/opt/mbedtls/lib/libmbedcrypto.3.dylib ./OBS.app/Contents/Frameworks/45
- cp /usr/local/opt/mbedtls/lib/libmbedx509.0.dylib ./OBS.app/Contents/Frameworks/46
- chmod +w ./OBS.app/Contents/Frameworks/*.dylib47
- install_name_tool -id @executable_path/../Frameworks/libmbedtls.12.dylib ./OBS.app/Contents/Frameworks/libmbedtls.12.dylib48
- install_name_tool -id @executable_path/../Frameworks/libmbedcrypto.3.dylib ./OBS.app/Contents/Frameworks/libmbedcrypto.3.dylib49
- install_name_tool -id @executable_path/../Frameworks/libmbedx509.0.dylib ./OBS.app/Contents/Frameworks/libmbedx509.0.dylib50
- install_name_tool -change libmbedtls.12.dylib @executable_path/../Frameworks/libmbedtls.12.dylib ./OBS.app/Contents/Plugins/obs-outputs.so51
- install_name_tool -change libmbedcrypto.3.dylib @executable_path/../Frameworks/libmbedcrypto.3.dylib ./OBS.app/Contents/Plugins/obs-outputs.so52
- install_name_tool -change libmbedx509.0.dylib @executable_path/../Frameworks/libmbedx509.0.dylib ./OBS.app/Contents/Plugins/obs-outputs.so53
-elif [ -f /usr/local/opt/mbedtls/lib/libmbedtls.13.dylib ]; then54
- cp /usr/local/opt/mbedtls/lib/libmbedtls.13.dylib ./OBS.app/Contents/Frameworks/55
- cp /usr/local/opt/mbedtls/lib/libmbedcrypto.5.dylib ./OBS.app/Contents/Frameworks/56
- cp /usr/local/opt/mbedtls/lib/libmbedx509.1.dylib ./OBS.app/Contents/Frameworks/57
- chmod +w ./OBS.app/Contents/Frameworks/*.dylib58
- install_name_tool -id @executable_path/../Frameworks/libmbedtls.13.dylib ./OBS.app/Contents/Frameworks/libmbedtls.13.dylib59
- install_name_tool -id @executable_path/../Frameworks/libmbedcrypto.5.dylib ./OBS.app/Contents/Frameworks/libmbedcrypto.5.dylib60
- install_name_tool -id @executable_path/../Frameworks/libmbedx509.1.dylib ./OBS.app/Contents/Frameworks/libmbedx509.1.dylib61
- install_name_tool -change libmbedtls.13.dylib @executable_path/../Frameworks/libmbedtls.13.dylib ./OBS.app/Contents/Plugins/obs-outputs.so62
- install_name_tool -change libmbedcrypto.5.dylib @executable_path/../Frameworks/libmbedcrypto.5.dylib ./OBS.app/Contents/Plugins/obs-outputs.so63
- install_name_tool -change libmbedx509.1.dylib @executable_path/../Frameworks/libmbedx509.1.dylib ./OBS.app/Contents/Plugins/obs-outputs.so64
-fi65
-66
-install_name_tool -change /usr/local/opt/curl/lib/libcurl.4.dylib @executable_path/../Frameworks/libcurl.4.dylib ./OBS.app/Contents/Plugins/obs-outputs.so67
-install_name_tool -change @rpath/libobs.0.dylib @executable_path/../Frameworks/libobs.0.dylib ./OBS.app/Contents/Plugins/obs-outputs.so68
-install_name_tool -change /tmp/obsdeps/bin/libjansson.4.dylib @executable_path/../Frameworks/libjansson.4.dylib ./OBS.app/Contents/Plugins/obs-outputs.so69
-70
-# copy sparkle into the app71
-hr "Copying Sparkle.framework"72
-cp -R ../../sparkle/Sparkle.framework ./OBS.app/Contents/Frameworks/73
-install_name_tool -change @rpath/Sparkle.framework/Versions/A/Sparkle @executable_path/../Frameworks/Sparkle.framework/Versions/A/Sparkle ./OBS.app/Contents/MacOS/obs74
-75
-# Copy Chromium embedded framework to app Frameworks directory76
-hr "Copying Chromium Embedded Framework.framework"77
-sudo mkdir -p OBS.app/Contents/Frameworks78
-sudo cp -R ../../cef_binary_${CEF_BUILD_VERSION}_macosx64/Release/Chromium\ Embedded\ Framework.framework OBS.app/Contents/Frameworks/79
-80
-install_name_tool -change /usr/local/opt/qt/lib/QtGui.framework/Versions/5/QtGui @executable_path/../Frameworks/QtGui.framework/Versions/5/QtGui ./OBS.app/Contents/Plugins/obs-browser.so81
-install_name_tool -change /usr/local/opt/qt/lib/QtCore.framework/Versions/5/QtCore @executable_path/../Frameworks/QtCore.framework/Versions/5/QtCore ./OBS.app/Contents/Plugins/obs-browser.so82
-install_name_tool -change /usr/local/opt/qt/lib/QtWidgets.framework/Versions/5/QtWidgets @executable_path/../Frameworks/QtWidgets.framework/Versions/5/QtWidgets ./OBS.app/Contents/Plugins/obs-browser.so83
-84
-cp ../CI/install/osx/OBSPublicDSAKey.pem OBS.app/Contents/Resources85
-86
-# edit plist87
-plutil -insert CFBundleVersion -string $GIT_TAG ./OBS.app/Contents/Info.plist88
-plutil -insert CFBundleShortVersionString -string $GIT_TAG ./OBS.app/Contents/Info.plist89
-plutil -insert OBSFeedsURL -string https://obsproject.com/osx_update/feeds.xml ./OBS.app/Contents/Info.plist90
-plutil -insert SUFeedURL -string https://obsproject.com/osx_update/stable/updates.xml ./OBS.app/Contents/Info.plist91
-plutil -insert SUPublicDSAKeyFile -string OBSPublicDSAKey.pem ./OBS.app/Contents/Info.plist92
-93
-dmgbuild -s ../CI/install/osx/settings.json "OBS" obs.dmg94
-95
-if [ -v "$TRAVIS" ]; then96
- # Signing stuff97
- hr "Decrypting Cert"98
- openssl aes-256-cbc -K $encrypted_dd3c7f5e9db9_key -iv $encrypted_dd3c7f5e9db9_iv -in ../CI/osxcert/Certificates.p12.enc -out Certificates.p12 -d99
- hr "Creating Keychain"100
- security create-keychain -p mysecretpassword build.keychain101
- security default-keychain -s build.keychain102
- security unlock-keychain -p mysecretpassword build.keychain103
- security set-keychain-settings -t 3600 -u build.keychain104
- hr "Importing certs into keychain"105
- security import ./Certificates.p12 -k build.keychain -T /usr/bin/productsign -P ""106
- # macOS 10.12+107
- security set-key-partition-list -S apple-tool:,apple: -s -k mysecretpassword build.keychain108
-fi109
-110
-cp ./OBS.dmg ./$FILENAME111
-112
-# Move to the folder that travis uses to upload artifacts from113
-hr "Moving package to nightly folder for distribution"114
-mkdir ../nightly115
-sudo mv ./$FILENAME ../nightly116
obs-studio-26.1.0.tar.xz/CI/before-script-osx.sh
Deleted
19
1
2
-# Make sure ccache is found3
-export PATH=/usr/local/opt/ccache/libexec:$PATH4
-5
-git fetch --tags6
-7
-mkdir build8
-cd build9
-cmake -DENABLE_SPARKLE_UPDATER=ON \10
--DCMAKE_OSX_DEPLOYMENT_TARGET=10.13 \11
--DDISABLE_PYTHON=ON \12
--DQTDIR=/usr/local/Cellar/qt/5.14.1 \13
--DDepsPath=/tmp/obsdeps \14
--DVLCPath=$PWD/../../vlc-3.0.8 \15
--DBUILD_BROWSER=ON \16
--DBROWSER_DEPLOY=ON \17
--DWITH_RTMPS=ON \18
--DCEF_ROOT_DIR=$PWD/../../cef_binary_${CEF_BUILD_VERSION}_macosx64 ..19
obs-studio-26.1.0.tar.xz/CI/install
Deleted
2
1
-(directory)2
obs-studio-26.1.0.tar.xz/CI/install-dependencies-osx.sh
Deleted
83
1
2
-hr() {3
- echo "───────────────────────────────────────────────────"4
- echo $15
- echo "───────────────────────────────────────────────────"6
-}7
-8
-# Exit if something fails9
-set -e10
-11
-# Echo all commands before executing12
-set -v13
-14
-if [[ $TRAVIS ]]; then15
- git fetch --unshallow16
-fi17
-18
-git fetch origin --tags19
-20
-# Leave obs-studio folder21
-cd ../22
-23
-# Install Packages app so we can build a package later24
-# http://s.sudre.free.fr/Software/Packages/about.html25
-hr "Downloading Packages app"26
-wget --quiet --retry-connrefused --waitretry=1 https://s3-us-west-2.amazonaws.com/obs-nightly/Packages.pkg27
-sudo installer -pkg ./Packages.pkg -target /28
-29
-brew update30
-31
-#Base OBS Deps and ccache32
-for DEPENDENCY in jack speexdsp ccache mbedtls freetype fdk-aac cmocka; do33
- if [ ! -d "$(brew --cellar)/${DEPENDENCY}" ]; then34
- brew install $DEPENDENCY35
- else36
- brew upgrade $DEPENDENCY37
- fi38
-done39
-40
-brew install https://gist.githubusercontent.com/DDRBoxman/9c7a2b08933166f4b61ed9a44b242609/raw/ef4de6c587c6bd7f50210eccd5bd51ff08e6de13/qt.rb41
-if [ -d "$(brew --cellar)/swig" ]; then42
- brew unlink swig43
-fi44
-brew install https://gist.githubusercontent.com/DDRBoxman/4cada55c51803a2f963fa40ce55c9d3e/raw/572c67e908bfbc1bcb8c476ea77ea3935133f5b5/swig.rb45
-46
-pip install dmgbuild47
-48
-export PATH=/usr/local/opt/ccache/libexec:$PATH49
-ccache -s || echo "CCache is not available."50
-51
-# Fetch and untar prebuilt OBS deps that are compatible with older versions of OSX52
-hr "Downloading OBS deps"53
-wget --quiet --retry-connrefused --waitretry=1 https://github.com/obsproject/obs-deps/releases/download/2020-04-24/osx-deps-2020-04-24.tar.gz54
-tar -xf ./osx-deps-2020-04-24.tar.gz -C /tmp55
-56
-# Fetch vlc codebase57
-hr "Downloading VLC repo"58
-wget --quiet --retry-connrefused --waitretry=1 https://downloads.videolan.org/vlc/3.0.8/vlc-3.0.8.tar.xz59
-tar -xf vlc-3.0.8.tar.xz60
-61
-# Get sparkle62
-hr "Downloading Sparkle framework"63
-wget --quiet --retry-connrefused --waitretry=1 -O sparkle.tar.bz2 https://github.com/sparkle-project/Sparkle/releases/download/1.23.0/Sparkle-1.23.0.tar.bz264
-mkdir ./sparkle65
-tar -xf ./sparkle.tar.bz2 -C ./sparkle66
-sudo cp -R ./sparkle/Sparkle.framework /Library/Frameworks/Sparkle.framework67
-68
-# CEF Stuff69
-hr "Downloading CEF"70
-wget --quiet --retry-connrefused --waitretry=1 https://obs-nightly.s3-us-west-2.amazonaws.com/cef_binary_${CEF_BUILD_VERSION}_macosx64.tar.bz271
-tar -xf ./cef_binary_${CEF_BUILD_VERSION}_macosx64.tar.bz272
-cd ./cef_binary_${CEF_BUILD_VERSION}_macosx6473
-# remove a broken test74
-sed -i '.orig' '/add_subdirectory(tests\/ceftests)/d' ./CMakeLists.txt75
-# target 10.1176
-sed -i '.orig' s/\"10.9\"/\"10.11\"/ ./cmake/cef_variables.cmake77
-mkdir build78
-cd ./build79
-cmake -DCMAKE_CXX_FLAGS="-std=c++11 -stdlib=libc++" -DCMAKE_EXE_LINKER_FLAGS="-std=c++11 -stdlib=libc++" -DCMAKE_OSX_DEPLOYMENT_TARGET=10.11 ..80
-make -j481
-mkdir libcef_dll82
-cd ../../83
obs-studio-26.1.0.tar.xz/CI/install/osx
Deleted
2
1
-(directory)2
obs-studio-26.1.0.tar.xz/CI/install/osx/CMakeLists.pkgproj
Deleted
1028
1
2
-<?xml version="1.0" encoding="UTF-8"?>3
-<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">4
-<plist version="1.0">5
-<dict>6
- <key>PACKAGES</key>7
- <array>8
- <dict>9
- <key>PACKAGE_FILES</key>10
- <dict>11
- <key>DEFAULT_INSTALL_LOCATION</key>12
- <string>/</string>13
- <key>HIERARCHY</key>14
- <dict>15
- <key>CHILDREN</key>16
- <array>17
- <dict>18
- <key>CHILDREN</key>19
- <array>20
- <dict>21
- <key>CHILDREN</key>22
- <array/>23
- <key>GID</key>24
- <integer>80</integer>25
- <key>PATH</key>26
- <string>../../../build/OBS.app</string>27
- <key>PATH_TYPE</key>28
- <integer>3</integer>29
- <key>PERMISSIONS</key>30
- <integer>493</integer>31
- <key>TYPE</key>32
- <integer>3</integer>33
- <key>UID</key>34
- <integer>0</integer>35
- </dict>36
- <dict>37
- <key>CHILDREN</key>38
- <array/>39
- <key>GID</key>40
- <integer>80</integer>41
- <key>PATH</key>42
- <string>Utilities</string>43
- <key>PATH_TYPE</key>44
- <integer>0</integer>45
- <key>PERMISSIONS</key>46
- <integer>493</integer>47
- <key>TYPE</key>48
- <integer>1</integer>49
- <key>UID</key>50
- <integer>0</integer>51
- </dict>52
- </array>53
- <key>GID</key>54
- <integer>80</integer>55
- <key>PATH</key>56
- <string>Applications</string>57
- <key>PATH_TYPE</key>58
- <integer>0</integer>59
- <key>PERMISSIONS</key>60
- <integer>509</integer>61
- <key>TYPE</key>62
- <integer>1</integer>63
- <key>UID</key>64
- <integer>0</integer>65
- </dict>66
- <dict>67
- <key>CHILDREN</key>68
- <array>69
- <dict>70
- <key>CHILDREN</key>71
- <array>72
- <dict>73
- <key>CHILDREN</key>74
- <array>75
- <dict>76
- <key>CHILDREN</key>77
- <array>78
- <dict>79
- <key>CHILDREN</key>80
- <array>81
- <dict>82
- <key>CHILDREN</key>83
- <array>84
- <dict>85
- <key>CHILDREN</key>86
- <array/>87
- <key>GID</key>88
- <integer>80</integer>89
- <key>PATH</key>90
- <string>../../../build/plugins/obs-browser/obs-browser-page</string>91
- <key>PATH_TYPE</key>92
- <integer>3</integer>93
- <key>PERMISSIONS</key>94
- <integer>493</integer>95
- <key>TYPE</key>96
- <integer>3</integer>97
- <key>UID</key>98
- <integer>0</integer>99
- </dict>100
- <dict>101
- <key>CHILDREN</key>102
- <array/>103
- <key>GID</key>104
- <integer>80</integer>105
- <key>PATH</key>106
- <string>../../../build/plugins/obs-browser/obs-browser.so</string>107
- <key>PATH_TYPE</key>108
- <integer>3</integer>109
- <key>PERMISSIONS</key>110
- <integer>493</integer>111
- <key>TYPE</key>112
- <integer>3</integer>113
- <key>UID</key>114
- <integer>0</integer>115
- </dict>116
- </array>117
- <key>GID</key>118
- <integer>80</integer>119
- <key>PATH</key>120
- <string>bin</string>121
- <key>PATH_TYPE</key>122
- <integer>0</integer>123
- <key>PERMISSIONS</key>124
- <integer>493</integer>125
- <key>TYPE</key>126
- <integer>2</integer>127
- <key>UID</key>128
- <integer>0</integer>129
- </dict>130
- </array>131
- <key>GID</key>132
- <integer>80</integer>133
- <key>PATH</key>134
- <string>obs-browser</string>135
- <key>PATH_TYPE</key>136
- <integer>0</integer>137
- <key>PERMISSIONS</key>138
- <integer>493</integer>139
- <key>TYPE</key>140
- <integer>2</integer>141
- <key>UID</key>142
- <integer>0</integer>143
- </dict>144
- </array>145
- <key>GID</key>146
- <integer>80</integer>147
- <key>PATH</key>148
- <string>plugins</string>149
- <key>PATH_TYPE</key>150
- <integer>0</integer>151
- <key>PERMISSIONS</key>152
- <integer>493</integer>153
- <key>TYPE</key>154
- <integer>2</integer>155
- <key>UID</key>156
- <integer>0</integer>157
- </dict>158
- </array>159
- <key>GID</key>160
- <integer>80</integer>161
- <key>PATH</key>162
- <string>obs-studio</string>163
- <key>PATH_TYPE</key>164
- <integer>0</integer>165
- <key>PERMISSIONS</key>166
- <integer>493</integer>167
- <key>TYPE</key>168
- <integer>2</integer>169
- <key>UID</key>170
- <integer>0</integer>171
- </dict>172
- </array>173
- <key>GID</key>174
- <integer>80</integer>175
- <key>PATH</key>176
- <string>Application Support</string>177
- <key>PATH_TYPE</key>178
- <integer>0</integer>179
- <key>PERMISSIONS</key>180
- <integer>493</integer>181
- <key>TYPE</key>182
- <integer>1</integer>183
- <key>UID</key>184
- <integer>0</integer>185
- </dict>186
- <dict>187
- <key>CHILDREN</key>188
- <array/>189
- <key>GID</key>190
- <integer>0</integer>191
- <key>PATH</key>192
- <string>Documentation</string>193
- <key>PATH_TYPE</key>194
- <integer>0</integer>195
- <key>PERMISSIONS</key>196
- <integer>493</integer>197
- <key>TYPE</key>198
- <integer>1</integer>199
- <key>UID</key>200
- <integer>0</integer>201
- </dict>202
- <dict>203
- <key>CHILDREN</key>204
- <array/>205
- <key>GID</key>206
- <integer>0</integer>207
- <key>PATH</key>208
- <string>Filesystems</string>209
- <key>PATH_TYPE</key>210
- <integer>0</integer>211
- <key>PERMISSIONS</key>212
- <integer>493</integer>213
- <key>TYPE</key>214
- <integer>1</integer>215
- <key>UID</key>216
- <integer>0</integer>217
- </dict>218
- <dict>219
- <key>CHILDREN</key>220
- <array/>221
- <key>GID</key>222
- <integer>0</integer>223
- <key>PATH</key>224
- <string>Frameworks</string>225
- <key>PATH_TYPE</key>226
- <integer>0</integer>227
- <key>PERMISSIONS</key>228
- <integer>493</integer>229
- <key>TYPE</key>230
- <integer>1</integer>231
- <key>UID</key>232
- <integer>0</integer>233
- </dict>234
- <dict>235
- <key>CHILDREN</key>236
- <array/>237
- <key>GID</key>238
- <integer>0</integer>239
- <key>PATH</key>240
- <string>Input Methods</string>241
- <key>PATH_TYPE</key>242
- <integer>0</integer>243
- <key>PERMISSIONS</key>244
- <integer>493</integer>245
- <key>TYPE</key>246
- <integer>1</integer>247
- <key>UID</key>248
- <integer>0</integer>249
- </dict>250
- <dict>251
- <key>CHILDREN</key>252
- <array/>253
- <key>GID</key>254
- <integer>0</integer>255
- <key>PATH</key>256
- <string>Internet Plug-Ins</string>257
- <key>PATH_TYPE</key>258
- <integer>0</integer>259
- <key>PERMISSIONS</key>260
- <integer>493</integer>261
- <key>TYPE</key>262
- <integer>1</integer>263
- <key>UID</key>264
- <integer>0</integer>265
- </dict>266
- <dict>267
- <key>CHILDREN</key>268
- <array/>269
- <key>GID</key>270
- <integer>0</integer>271
- <key>PATH</key>272
- <string>LaunchAgents</string>273
- <key>PATH_TYPE</key>274
- <integer>0</integer>275
- <key>PERMISSIONS</key>276
- <integer>493</integer>277
- <key>TYPE</key>278
- <integer>1</integer>279
- <key>UID</key>280
- <integer>0</integer>281
- </dict>282
- <dict>283
- <key>CHILDREN</key>284
- <array/>285
- <key>GID</key>286
- <integer>0</integer>287
- <key>PATH</key>288
- <string>LaunchDaemons</string>289
- <key>PATH_TYPE</key>290
- <integer>0</integer>291
- <key>PERMISSIONS</key>292
- <integer>493</integer>293
- <key>TYPE</key>294
- <integer>1</integer>295
- <key>UID</key>296
- <integer>0</integer>297
- </dict>298
- <dict>299
- <key>CHILDREN</key>300
- <array/>301
- <key>GID</key>302
- <integer>0</integer>303
- <key>PATH</key>304
- <string>PreferencePanes</string>305
- <key>PATH_TYPE</key>306
- <integer>0</integer>307
- <key>PERMISSIONS</key>308
- <integer>493</integer>309
- <key>TYPE</key>310
- <integer>1</integer>311
- <key>UID</key>312
- <integer>0</integer>313
- </dict>314
- <dict>315
- <key>CHILDREN</key>316
- <array/>317
- <key>GID</key>318
- <integer>0</integer>319
- <key>PATH</key>320
- <string>Preferences</string>321
- <key>PATH_TYPE</key>322
- <integer>0</integer>323
- <key>PERMISSIONS</key>324
- <integer>493</integer>325
- <key>TYPE</key>326
- <integer>1</integer>327
- <key>UID</key>328
- <integer>0</integer>329
- </dict>330
- <dict>331
- <key>CHILDREN</key>332
- <array/>333
- <key>GID</key>334
- <integer>80</integer>335
- <key>PATH</key>336
- <string>Printers</string>337
- <key>PATH_TYPE</key>338
- <integer>0</integer>339
- <key>PERMISSIONS</key>340
- <integer>493</integer>341
- <key>TYPE</key>342
- <integer>1</integer>343
- <key>UID</key>344
- <integer>0</integer>345
- </dict>346
- <dict>347
- <key>CHILDREN</key>348
- <array/>349
- <key>GID</key>350
- <integer>0</integer>351
- <key>PATH</key>352
- <string>PrivilegedHelperTools</string>353
- <key>PATH_TYPE</key>354
- <integer>0</integer>355
- <key>PERMISSIONS</key>356
- <integer>493</integer>357
- <key>TYPE</key>358
- <integer>1</integer>359
- <key>UID</key>360
- <integer>0</integer>361
- </dict>362
- <dict>363
- <key>CHILDREN</key>364
- <array/>365
- <key>GID</key>366
- <integer>0</integer>367
- <key>PATH</key>368
- <string>QuickLook</string>369
- <key>PATH_TYPE</key>370
- <integer>0</integer>371
- <key>PERMISSIONS</key>372
- <integer>493</integer>373
- <key>TYPE</key>374
- <integer>1</integer>375
- <key>UID</key>376
- <integer>0</integer>377
- </dict>378
- <dict>379
- <key>CHILDREN</key>380
- <array/>381
- <key>GID</key>382
- <integer>0</integer>383
- <key>PATH</key>384
- <string>QuickTime</string>385
- <key>PATH_TYPE</key>386
- <integer>0</integer>387
- <key>PERMISSIONS</key>388
- <integer>493</integer>389
- <key>TYPE</key>390
- <integer>1</integer>391
- <key>UID</key>392
- <integer>0</integer>393
- </dict>394
- <dict>395
- <key>CHILDREN</key>396
- <array/>397
- <key>GID</key>398
- <integer>0</integer>399
- <key>PATH</key>400
- <string>Screen Savers</string>401
- <key>PATH_TYPE</key>402
- <integer>0</integer>403
- <key>PERMISSIONS</key>404
- <integer>493</integer>405
- <key>TYPE</key>406
- <integer>1</integer>407
- <key>UID</key>408
- <integer>0</integer>409
- </dict>410
- <dict>411
- <key>CHILDREN</key>412
- <array/>413
- <key>GID</key>414
- <integer>0</integer>415
- <key>PATH</key>416
- <string>Scripts</string>417
- <key>PATH_TYPE</key>418
- <integer>0</integer>419
- <key>PERMISSIONS</key>420
- <integer>493</integer>421
- <key>TYPE</key>422
- <integer>1</integer>423
- <key>UID</key>424
- <integer>0</integer>425
- </dict>426
- <dict>427
- <key>CHILDREN</key>428
- <array/>429
- <key>GID</key>430
- <integer>0</integer>431
- <key>PATH</key>432
- <string>Services</string>433
- <key>PATH_TYPE</key>434
- <integer>0</integer>435
- <key>PERMISSIONS</key>436
- <integer>493</integer>437
- <key>TYPE</key>438
- <integer>1</integer>439
- <key>UID</key>440
- <integer>0</integer>441
- </dict>442
- <dict>443
- <key>CHILDREN</key>444
- <array/>445
- <key>GID</key>446
- <integer>0</integer>447
- <key>PATH</key>448
- <string>Widgets</string>449
- <key>PATH_TYPE</key>450
- <integer>0</integer>451
- <key>PERMISSIONS</key>452
- <integer>493</integer>453
- <key>TYPE</key>454
- <integer>1</integer>455
- <key>UID</key>456
- <integer>0</integer>457
- </dict>458
- </array>459
- <key>GID</key>460
- <integer>0</integer>461
- <key>PATH</key>462
- <string>Library</string>463
- <key>PATH_TYPE</key>464
- <integer>0</integer>465
- <key>PERMISSIONS</key>466
- <integer>493</integer>467
- <key>TYPE</key>468
- <integer>1</integer>469
- <key>UID</key>470
- <integer>0</integer>471
- </dict>472
- <dict>473
- <key>CHILDREN</key>474
- <array>475
- <dict>476
- <key>CHILDREN</key>477
- <array>478
- <dict>479
- <key>CHILDREN</key>480
- <array/>481
- <key>GID</key>482
- <integer>0</integer>483
- <key>PATH</key>484
- <string>Extensions</string>485
- <key>PATH_TYPE</key>486
- <integer>0</integer>487
- <key>PERMISSIONS</key>488
- <integer>493</integer>489
- <key>TYPE</key>490
- <integer>1</integer>491
- <key>UID</key>492
- <integer>0</integer>493
- </dict>494
- </array>495
- <key>GID</key>496
- <integer>0</integer>497
- <key>PATH</key>498
- <string>Library</string>499
- <key>PATH_TYPE</key>500
- <integer>0</integer>501
- <key>PERMISSIONS</key>502
- <integer>493</integer>503
- <key>TYPE</key>504
- <integer>1</integer>505
- <key>UID</key>506
- <integer>0</integer>507
- </dict>508
- </array>509
- <key>GID</key>510
- <integer>0</integer>511
- <key>PATH</key>512
- <string>System</string>513
- <key>PATH_TYPE</key>514
- <integer>0</integer>515
- <key>PERMISSIONS</key>516
- <integer>493</integer>517
- <key>TYPE</key>518
- <integer>1</integer>519
- <key>UID</key>520
- <integer>0</integer>521
- </dict>522
- <dict>523
- <key>CHILDREN</key>524
- <array>525
- <dict>526
- <key>CHILDREN</key>527
- <array/>528
- <key>GID</key>529
- <integer>0</integer>530
- <key>PATH</key>531
- <string>Shared</string>532
- <key>PATH_TYPE</key>533
- <integer>0</integer>534
- <key>PERMISSIONS</key>535
- <integer>1023</integer>536
- <key>TYPE</key>537
- <integer>1</integer>538
- <key>UID</key>539
- <integer>0</integer>540
- </dict>541
- </array>542
- <key>GID</key>543
- <integer>80</integer>544
- <key>PATH</key>545
- <string>Users</string>546
- <key>PATH_TYPE</key>547
- <integer>0</integer>548
- <key>PERMISSIONS</key>549
- <integer>493</integer>550
- <key>TYPE</key>551
- <integer>1</integer>552
- <key>UID</key>553
- <integer>0</integer>554
- </dict>555
- </array>556
- <key>GID</key>557
- <integer>0</integer>558
- <key>PATH</key>559
- <string>/</string>560
- <key>PATH_TYPE</key>561
- <integer>0</integer>562
- <key>PERMISSIONS</key>563
- <integer>493</integer>564
- <key>TYPE</key>565
- <integer>1</integer>566
- <key>UID</key>567
- <integer>0</integer>568
- </dict>569
- <key>PAYLOAD_TYPE</key>570
- <integer>0</integer>571
- <key>VERSION</key>572
- <integer>2</integer>573
- </dict>574
- <key>PACKAGE_SCRIPTS</key>575
- <dict>576
- <key>POSTINSTALL_PATH</key>577
- <dict>578
- <key>PATH</key>579
- <string>post-install.sh</string>580
- <key>PATH_TYPE</key>581
- <integer>3</integer>582
- </dict>583
- <key>PREINSTALL_PATH</key>584
- <dict/>585
- <key>RESOURCES</key>586
- <array/>587
- </dict>588
- <key>PACKAGE_SETTINGS</key>589
- <dict>590
- <key>AUTHENTICATION</key>591
- <integer>1</integer>592
- <key>CONCLUSION_ACTION</key>593
- <integer>0</integer>594
- <key>IDENTIFIER</key>595
- <string>org.obsproject.pkg.obs-studio</string>596
- <key>NAME</key>597
- <string>OBS</string>598
- <key>OVERWRITE_PERMISSIONS</key>599
- <false/>600
- <key>VERSION</key>601
- <string>1.0</string>602
- </dict>603
- <key>UUID</key>604
- <string>19CCE3F2-8911-4364-B673-8B5BC3ABD4DA</string>605
- </dict>606
- <dict>607
- <key>PACKAGE_SETTINGS</key>608
- <dict>609
- <key>LOCATION</key>610
- <integer>0</integer>611
- <key>NAME</key>612
- <string>SyphonInject</string>613
- </dict>614
- <key>PATH</key>615
- <dict>616
- <key>PATH</key>617
- <string>SyphonInject.pkg</string>618
- <key>PATH_TYPE</key>619
- <integer>1</integer>620
- </dict>621
- <key>TYPE</key>622
- <integer>1</integer>623
- <key>UUID</key>624
- <string>0CC9C67E-4D14-4794-9930-019925513B1C</string>625
- </dict>626
- </array>627
- <key>PROJECT</key>628
- <dict>629
- <key>PROJECT_COMMENTS</key>630
- <dict>631
- <key>NOTES</key>632
- <data>633
- PCFET0NUWVBFIGh0bWwgUFVCTElDICItLy9XM0MvL0RURCBIVE1M634
- IDQuMDEvL0VOIiAiaHR0cDovL3d3dy53My5vcmcvVFIvaHRtbDQv635
- c3RyaWN0LmR0ZCI+CjxodG1sPgo8aGVhZD4KPG1ldGEgaHR0cC1l636
- cXVpdj0iQ29udGVudC1UeXBlIiBjb250ZW50PSJ0ZXh0L2h0bWw7637
- IGNoYXJzZXQ9VVRGLTgiPgo8bWV0YSBodHRwLWVxdWl2PSJDb250638
- ZW50LVN0eWxlLVR5cGUiIGNvbnRlbnQ9InRleHQvY3NzIj4KPHRp639
- dGxlPjwvdGl0bGU+CjxtZXRhIG5hbWU9IkdlbmVyYXRvciIgY29u640
- dGVudD0iQ29jb2EgSFRNTCBXcml0ZXIiPgo8bWV0YSBuYW1lPSJD641
- b2NvYVZlcnNpb24iIGNvbnRlbnQ9IjE1MDQuODEiPgo8c3R5bGUg642
- dHlwZT0idGV4dC9jc3MiPgo8L3N0eWxlPgo8L2hlYWQ+Cjxib2R5643
- Pgo8L2JvZHk+CjwvaHRtbD4K644
- </data>645
- </dict>646
- <key>PROJECT_PRESENTATION</key>647
- <dict>648
- <key>BACKGROUND</key>649
- <dict>650
- <key>ALIGNMENT</key>651
- <integer>4</integer>652
- <key>BACKGROUND_PATH</key>653
- <dict>654
- <key>PATH</key>655
- <string>obs.png</string>656
- <key>PATH_TYPE</key>657
- <integer>1</integer>658
- </dict>659
- <key>CUSTOM</key>660
- <integer>1</integer>661
- <key>SCALING</key>662
- <integer>0</integer>663
- </dict>664
- <key>INSTALLATION TYPE</key>665
- <dict>666
- <key>HIERARCHIES</key>667
- <dict>668
- <key>INSTALLER</key>669
- <dict>670
- <key>LIST</key>671
- <array>672
- <dict>673
- <key>DESCRIPTION</key>674
- <array/>675
- <key>OPTIONS</key>676
- <dict>677
- <key>HIDDEN</key>678
- <false/>679
- <key>STATE</key>680
- <integer>0</integer>681
- </dict>682
- <key>PACKAGE_UUID</key>683
- <string>19CCE3F2-8911-4364-B673-8B5BC3ABD4DA</string>684
- <key>REQUIREMENTS</key>685
- <array/>686
- <key>TITLE</key>687
- <array/>688
- <key>TOOLTIP</key>689
- <array/>690
- <key>TYPE</key>691
- <integer>0</integer>692
- <key>UUID</key>693
- <string>7C540711-59F4-479C-9CFD-8C4D6594992E</string>694
- </dict>695
- <dict>696
- <key>DESCRIPTION</key>697
- <array/>698
- <key>OPTIONS</key>699
- <dict>700
- <key>HIDDEN</key>701
- <false/>702
- <key>STATE</key>703
- <integer>1</integer>704
- </dict>705
- <key>PACKAGE_UUID</key>706
- <string>0CC9C67E-4D14-4794-9930-019925513B1C</string>707
- <key>REQUIREMENTS</key>708
- <array/>709
- <key>TITLE</key>710
- <array/>711
- <key>TOOLTIP</key>712
- <array/>713
- <key>TYPE</key>714
- <integer>0</integer>715
- <key>UUID</key>716
- <string>BBDE08F6-D7EE-47CB-881F-7F208B3A604B</string>717
- </dict>718
- </array>719
- <key>REMOVED</key>720
- <dict/>721
- </dict>722
- </dict>723
- <key>INSTALLATION TYPE</key>724
- <integer>0</integer>725
- <key>MODE</key>726
- <integer>0</integer>727
- </dict>728
- <key>INSTALLATION_STEPS</key>729
- <array>730
- <dict>731
- <key>ICPRESENTATION_CHAPTER_VIEW_CONTROLLER_CLASS</key>732
- <string>ICPresentationViewIntroductionController</string>733
- <key>INSTALLER_PLUGIN</key>734
- <string>Introduction</string>735
- <key>LIST_TITLE_KEY</key>736
- <string>InstallerSectionTitle</string>737
- </dict>738
- <dict>739
- <key>ICPRESENTATION_CHAPTER_VIEW_CONTROLLER_CLASS</key>740
- <string>ICPresentationViewReadMeController</string>741
- <key>INSTALLER_PLUGIN</key>742
- <string>ReadMe</string>743
- <key>LIST_TITLE_KEY</key>744
- <string>InstallerSectionTitle</string>745
- </dict>746
- <dict>747
- <key>ICPRESENTATION_CHAPTER_VIEW_CONTROLLER_CLASS</key>748
- <string>ICPresentationViewLicenseController</string>749
- <key>INSTALLER_PLUGIN</key>750
- <string>License</string>751
- <key>LIST_TITLE_KEY</key>752
- <string>InstallerSectionTitle</string>753
- </dict>754
- <dict>755
- <key>ICPRESENTATION_CHAPTER_VIEW_CONTROLLER_CLASS</key>756
- <string>ICPresentationViewDestinationSelectController</string>757
- <key>INSTALLER_PLUGIN</key>758
- <string>TargetSelect</string>759
- <key>LIST_TITLE_KEY</key>760
- <string>InstallerSectionTitle</string>761
- </dict>762
- <dict>763
- <key>ICPRESENTATION_CHAPTER_VIEW_CONTROLLER_CLASS</key>764
- <string>ICPresentationViewInstallationTypeController</string>765
- <key>INSTALLER_PLUGIN</key>766
- <string>PackageSelection</string>767
- <key>LIST_TITLE_KEY</key>768
- <string>InstallerSectionTitle</string>769
- </dict>770
- <dict>771
- <key>ICPRESENTATION_CHAPTER_VIEW_CONTROLLER_CLASS</key>772
- <string>ICPresentationViewInstallationController</string>773
- <key>INSTALLER_PLUGIN</key>774
- <string>Install</string>775
- <key>LIST_TITLE_KEY</key>776
- <string>InstallerSectionTitle</string>777
- </dict>778
- <dict>779
- <key>ICPRESENTATION_CHAPTER_VIEW_CONTROLLER_CLASS</key>780
- <string>ICPresentationViewSummaryController</string>781
- <key>INSTALLER_PLUGIN</key>782
- <string>Summary</string>783
- <key>LIST_TITLE_KEY</key>784
- <string>InstallerSectionTitle</string>785
- </dict>786
- </array>787
- <key>INTRODUCTION</key>788
- <dict>789
- <key>LOCALIZATIONS</key>790
- <array/>791
- </dict>792
- <key>LICENSE</key>793
- <dict>794
- <key>KEYWORDS</key>795
- <dict/>796
- <key>LOCALIZATIONS</key>797
- <array/>798
- <key>MODE</key>799
- <integer>0</integer>800
- </dict>801
- <key>README</key>802
- <dict>803
- <key>LOCALIZATIONS</key>804
- <array/>805
- </dict>806
- <key>SUMMARY</key>807
- <dict>808
- <key>LOCALIZATIONS</key>809
- <array/>810
- </dict>811
- <key>TITLE</key>812
- <dict>813
- <key>LOCALIZATIONS</key>814
- <array>815
- <dict>816
- <key>LANGUAGE</key>817
- <string>English</string>818
- <key>VALUE</key>819
- <string>OBS</string>820
- </dict>821
- </array>822
- </dict>823
- </dict>824
- <key>PROJECT_REQUIREMENTS</key>825
- <dict>826
- <key>LIST</key>827
- <array/>828
- <key>POSTINSTALL_PATH</key>829
- <dict/>830
- <key>PREINSTALL_PATH</key>831
- <dict/>832
- <key>RESOURCES</key>833
- <array/>834
- <key>ROOT_VOLUME_ONLY</key>835
- <false/>836
- </dict>837
- <key>PROJECT_SETTINGS</key>838
- <dict>839
- <key>ADVANCED_OPTIONS</key>840
- <dict/>841
- <key>BUILD_FORMAT</key>842
- <integer>0</integer>843
- <key>BUILD_PATH</key>844
- <dict>845
- <key>PATH</key>846
- <string>../../../build</string>847
- <key>PATH_TYPE</key>848
- <integer>3</integer>849
- </dict>850
- <key>EXCLUDED_FILES</key>851
- <array>852
- <dict>853
- <key>PATTERNS_ARRAY</key>854
- <array>855
- <dict>856
- <key>REGULAR_EXPRESSION</key>857
- <false/>858
- <key>STRING</key>859
- <string>.DS_Store</string>860
- <key>TYPE</key>861
- <integer>0</integer>862
- </dict>863
- </array>864
- <key>PROTECTED</key>865
- <true/>866
- <key>PROXY_NAME</key>867
- <string>Remove .DS_Store files</string>868
- <key>PROXY_TOOLTIP</key>869
- <string>Remove ".DS_Store" files created by the Finder.</string>870
- <key>STATE</key>871
- <true/>872
- </dict>873
- <dict>874
- <key>PATTERNS_ARRAY</key>875
- <array>876
- <dict>877
- <key>REGULAR_EXPRESSION</key>878
- <false/>879
- <key>STRING</key>880
- <string>.pbdevelopment</string>881
- <key>TYPE</key>882
- <integer>0</integer>883
- </dict>884
- </array>885
- <key>PROTECTED</key>886
- <true/>887
- <key>PROXY_NAME</key>888
- <string>Remove .pbdevelopment files</string>889
- <key>PROXY_TOOLTIP</key>890
- <string>Remove ".pbdevelopment" files created by ProjectBuilder or Xcode.</string>891
- <key>STATE</key>892
- <true/>893
- </dict>894
- <dict>895
- <key>PATTERNS_ARRAY</key>896
- <array>897
- <dict>898
- <key>REGULAR_EXPRESSION</key>899
- <false/>900
- <key>STRING</key>901
- <string>CVS</string>902
- <key>TYPE</key>903
- <integer>1</integer>904
- </dict>905
- <dict>906
- <key>REGULAR_EXPRESSION</key>907
- <false/>908
- <key>STRING</key>909
- <string>.cvsignore</string>910
- <key>TYPE</key>911
- <integer>0</integer>912
- </dict>913
- <dict>914
- <key>REGULAR_EXPRESSION</key>915
- <false/>916
- <key>STRING</key>917
- <string>.cvspass</string>918
- <key>TYPE</key>919
- <integer>0</integer>920
- </dict>921
- <dict>922
- <key>REGULAR_EXPRESSION</key>923
- <false/>924
- <key>STRING</key>925
- <string>.svn</string>926
- <key>TYPE</key>927
- <integer>1</integer>928
- </dict>929
- <dict>930
- <key>REGULAR_EXPRESSION</key>931
- <false/>932
- <key>STRING</key>933
- <string>.git</string>934
- <key>TYPE</key>935
- <integer>1</integer>936
- </dict>937
- <dict>938
- <key>REGULAR_EXPRESSION</key>939
- <false/>940
- <key>STRING</key>941
- <string>.gitignore</string>942
- <key>TYPE</key>943
- <integer>0</integer>944
- </dict>945
- </array>946
- <key>PROTECTED</key>947
- <true/>948
- <key>PROXY_NAME</key>949
- <string>Remove SCM metadata</string>950
- <key>PROXY_TOOLTIP</key>951
- <string>Remove helper files and folders used by the CVS, SVN or Git Source Code Management systems.</string>952
- <key>STATE</key>953
- <true/>954
- </dict>955
- <dict>956
- <key>PATTERNS_ARRAY</key>957
- <array>958
- <dict>959
- <key>REGULAR_EXPRESSION</key>960
- <false/>961
- <key>STRING</key>962
- <string>classes.nib</string>963
- <key>TYPE</key>964
- <integer>0</integer>965
- </dict>966
- <dict>967
- <key>REGULAR_EXPRESSION</key>968
- <false/>969
- <key>STRING</key>970
- <string>designable.db</string>971
- <key>TYPE</key>972
- <integer>0</integer>973
- </dict>974
- <dict>975
- <key>REGULAR_EXPRESSION</key>976
- <false/>977
- <key>STRING</key>978
- <string>info.nib</string>979
- <key>TYPE</key>980
- <integer>0</integer>981
- </dict>982
- </array>983
- <key>PROTECTED</key>984
- <true/>985
- <key>PROXY_NAME</key>986
- <string>Optimize nib files</string>987
- <key>PROXY_TOOLTIP</key>988
- <string>Remove "classes.nib", "info.nib" and "designable.nib" files within .nib bundles.</string>989
- <key>STATE</key>990
- <true/>991
- </dict>992
- <dict>993
- <key>PATTERNS_ARRAY</key>994
- <array>995
- <dict>996
- <key>REGULAR_EXPRESSION</key>997
- <false/>998
- <key>STRING</key>999
- <string>Resources Disabled</string>1000
- <key>TYPE</key>1001
- <integer>1</integer>1002
- </dict>1003
- </array>1004
- <key>PROTECTED</key>1005
- <true/>1006
- <key>PROXY_NAME</key>1007
- <string>Remove Resources Disabled folders</string>1008
- <key>PROXY_TOOLTIP</key>1009
- <string>Remove "Resources Disabled" folders.</string>1010
- <key>STATE</key>1011
- <true/>1012
- </dict>1013
- <dict>1014
- <key>SEPARATOR</key>1015
- <true/>1016
- </dict>1017
- </array>1018
- <key>NAME</key>1019
- <string>OBS</string>1020
- </dict>1021
- </dict>1022
- <key>TYPE</key>1023
- <integer>0</integer>1024
- <key>VERSION</key>1025
- <integer>2</integer>1026
-</dict>1027
-</plist>1028
obs-studio-26.1.0.tar.xz/CI/install/osx/Info.plist
Deleted
30
1
2
-<?xml version="1.0" encoding="UTF-8"?>3
-<!DOCTYPE plist PUBLIC "-//Apple Computer//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">4
-<plist version="1.0">5
-<dict>6
- <key>CFBundleIconFile</key>7
- <string>obs.icns</string>8
- <key>CFBundleName</key>9
- <string>OBS</string>10
- <key>CFBundleGetInfoString</key>11
- <string>OBS - Free and Open Source Streaming/Recording Software</string>12
- <key>CFBundleExecutable</key>13
- <string>obs</string>14
- <key>CFBundleIdentifier</key>15
- <string>com.obsproject.obs-studio</string>16
- <key>CFBundlePackageType</key>17
- <string>APPL</string>18
- <key>LSMinimumSystemVersion</key>19
- <string>10.8.5</string>20
- <key>NSHighResolutionCapable</key>21
- <true/>22
- <key>LSAppNapIsDisabled</key>23
- <true/>24
- <key>NSCameraUsageDescription</key>25
- <string>OBS needs to access the camera to enable camera sources to work.</string>26
- <key>NSMicrophoneUsageDescription</key>27
- <string>OBS needs to access the microphone to enable audio input.</string>28
-</dict>29
-</plist>30
obs-studio-26.1.0.tar.xz/CI/install/osx/OBSPublicDSAKey.pem
Deleted
38
1
2
------BEGIN PUBLIC KEY-----3
-MIIGPDCCBC4GByqGSM44BAEwggQhAoICAQCZZZ2y7H2GJmMfP4KQihJTJOoiGNUw4
-mue6sqMbH+utRykRnSKBZux6R665eRFMpNgrgFO1TLLGbdD2U31KiGtCvFJOmOl35
-+QP055BuXjEG36NU7AWEFLAlbDlr/2D3oumq3Ib3iMnnr9RrVztJ2VFOvVio1eWr6
-ZxboVwKPK8D6BqsWiv15vbYlJnTC4Fls6ySmdjVBxwoPlTaMu1ysi5DfbIZ93s5u7
-aQt1FvXuWtPBWjyVUORcNbcWf49E5R2pV0OSBK95Hw2/wXz4vmj+w92dTePGnVaW8
-Me4CoF5PIeZILwp6DCLStX4eW2WG1NChJTC8zeQ/3bMMoGyKM/MadyvrDqMywsKY9
-caxkIwHrDKOEdXXGo80dIwZMMLipPA8DKhx5ojphfkeXjIhKSx+49knXT3ED5okE10
-Wai7tGUXj/8D8sGh+7b+AVsdujvr4v8WQaZiKUOZ2IIHOg3VLz9T9v0zet1Yt98711
-KNymFcp2CHeJ6KnDP/ZGQ6Nl0HsPxUgscsXV+R2FEc8Q1j0Ukkuxnopa0E4/huUu12
-gjyRzpXD734qFMDf7LcXca6qNjBor6gVj5sRyRKCpZ+KQfMUlr8jp506ztYSyeJu13
-dxJV30tQgztwkbrs02CqOt4Z3Peo6sdht7hWKSPVwmja3tq8/TfUSSoo6wKYN9/w14
-Mf3dVeRF8hCzJQIVAJnzuzmzQhCKPiQnl3jh5qGII2XfAoICAQCCVATAff89ceHj15
-ROHEbHTQFpVxJ/kRZPfxnU46DSw79Tih7tthV68oakPSOTP3cx/Tga0GwogarZ9N16
-F2VVan5w9OQSSewXsr5UDT5bnmJF+h+JB7TMy+sXZBYobUqjlUd5VtKc8RsN86P417
-s7xbK0mA+hfe+27r18JT81/eH3xUfh7UOUGSdMN2Ch9f7RFSMZIgUAZUzu2K3ODp18
-hPgtc2QJ8QVAp7GLvQgw8ZUME/ChZslyBIyJvYgUIxfxlgRWYro5pQT7/ngkgdXo19
-wlghHKkldwMuY3zaFdhPnFNuEUEtc18ILsbz0+AnagCUd6n+3safskCRqLIHMOY620
-iLBSZPX9hJQhVCqSqz1VNDDww8FNa/fojJ1Lr/TI0I+0Ib2pCiY2LChXUqGY5SLZ21
-2KNs5qFsyZP+I0L8YsGwqvUYyFwk7Ok224n0NtaOwqpLCrtXd/i6DaDNiaoJuwJC22
-1ELCfaZivorgkC5rhBt2H7qWUAR+EtrFE/gb0k/G5EIhjYql7onGbX+G2re38vQA23
-fg1pzguhig2dafP/BxMLZrn1Gg61xzmEYPuS9gclktaf675srv8GVb46VkOxXL+D24
-YvTmpJPP7UUOVlmAMCo4j4y09MW3jq9TDp42VTLeZVubyjslGnavlnq1O+ZyXUye25
-1FMeby65sIbSHHHwoFnRv3hLSEXI5gOCAgYAAoICAQCUkYnZkPfHfOJZI403xUYP26
-CE/bLpkza074Xo6EXElsWRnpQgNTx+JFOvItgj3v0OkIqDin9UredKOwfkiftslV27
-jxUVKA6I5kwnGvCpvTpQMLyLjq+VQr+J2D6eId6tV/iajhdu5r4JThU8KllT7Ywb28
-NAur34ftLNCVAMRUaDNeEoHfePgderW384e+lbvpmtifmBluammGSxxRtUsdjvJZ29
-BFkhaJu86CKxcU7D1lbPVOtV/jaxz6d16VdGcfBdi2LzXZzZtYpT9XGPX3NF+xii30
-spAURWsoe11LTRXF+eJhgCm5iIDN3kh1HEQKYKAVpmrcM0aFzk/NpS+tFyU72vaq31
-IRSSJw/aa1oELOAakG5oPldc4RcYWl32sbnVwXHO7TZvgTrBSC10o65MAC5CHP/s32
-b07heDYAIt7re7szvOYq+c/9zAMAlu3pcO8MqaXYMmybdHBXHQ2b+DdJWHmIUWcX33
-CbUzr09vzGkJAvqsXqbmJPr8aixrO75DhT0iDTILLWe/GWK51nf+Tg0pNxVgGyAl34
-BqvRqqo7SSDu9FMkwQesFFHhuoHLyEHwVPJ+sMQTNwQcm9c6YuW8EYDRSkeKLWYk35
-3fkjG+Pe9uVE8a1taDg3FjSY0UqjUT6XMw+i0Lajyus2L6wFBwrrGM6E4xa6x1CC36
-MGjmuSOlPA1umQsToIcO4g==37
------END PUBLIC KEY-----38
obs-studio-26.1.0.tar.xz/CI/install/osx/SyphonInject.pkg
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background.png
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd
Deleted
2
1
-(directory)2
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd/QuickLook
Deleted
2
1
-(directory)2
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd/QuickLook/Icon.tiff
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd/QuickLook/Preview.tiff
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd/QuickLook/Thumbnail.tiff
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd/data
Deleted
2
1
-(directory)2
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd/data/556CF265-5721-4F18-BE83-8CF39483B4C2
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd/data/8CA689C3-ED2A-459E-952C-E08026CFCD07
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background.pxd/metadata.info
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background.tiff
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/background@2x.png
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/buildDMG
Deleted
3
1
2
-dmgbuild -s ./settings.json "OBS" obs.dmg3
obs-studio-26.1.0.tar.xz/CI/install/osx/build_app.py
Deleted
240
1
2
-#!/usr/bin/env python3
-4
-candidate_paths = "bin obs-plugins data".split()5
-6
-plist_path = "../cmake/osxbundle/Info.plist"7
-icon_path = "../cmake/osxbundle/obs.icns"8
-run_path = "../cmake/osxbundle/obslaunch.sh"9
-10
-#not copied11
-blacklist = """/usr /System""".split()12
-13
-#copied14
-whitelist = """/usr/local""".split()15
-16
-#17
-#18
-#19
-20
-21
-from sys import argv22
-from glob import glob23
-from subprocess import check_output, call24
-from collections import namedtuple25
-from shutil import copy, copytree, rmtree26
-from os import makedirs, rename, walk, path as ospath27
-import plistlib28
-29
-import argparse30
-31
-def _str_to_bool(s):32
- """Convert string to bool (in argparse context)."""33
- if s.lower() not in ['true', 'false']:34
- raise ValueError('Need bool; got %r' % s)35
- return {'true': True, 'false': False}[s.lower()]36
-37
-def add_boolean_argument(parser, name, default=False):38
- """Add a boolean argument to an ArgumentParser instance."""39
- group = parser.add_mutually_exclusive_group()40
- group.add_argument(41
- '--' + name, nargs='?', default=default, const=True, type=_str_to_bool)42
- group.add_argument('--no' + name, dest=name, action='store_false')43
-44
-parser = argparse.ArgumentParser(description='obs-studio package util')45
-parser.add_argument('-d', '--base-dir', dest='dir', default='rundir/RelWithDebInfo')46
-parser.add_argument('-n', '--build-number', dest='build_number', default='0')47
-parser.add_argument('-k', '--public-key', dest='public_key', default='OBSPublicDSAKey.pem')48
-parser.add_argument('-f', '--sparkle-framework', dest='sparkle', default=None)49
-parser.add_argument('-b', '--base-url', dest='base_url', default='https://obsproject.com/osx_update')50
-parser.add_argument('-u', '--user', dest='user', default='jp9000')51
-parser.add_argument('-c', '--channel', dest='channel', default='master')52
-add_boolean_argument(parser, 'stable', default=False)53
-parser.add_argument('-p', '--prefix', dest='prefix', default='')54
-args = parser.parse_args()55
-56
-def cmd(cmd):57
- import subprocess58
- import shlex59
- return subprocess.check_output(shlex.split(cmd)).rstrip('\r\n')60
-61
-LibTarget = namedtuple("LibTarget", ("path", "external", "copy_as"))62
-63
-inspect = list()64
-65
-inspected = set()66
-67
-build_path = args.dir68
-build_path = build_path.replace("\\ ", " ")69
-70
-def add(name, external=False, copy_as=None):71
- if external and copy_as is None:72
- copy_as = name.split("/")[-1]73
- if name[0] != "/":74
- name = build_path+"/"+name75
- t = LibTarget(name, external, copy_as)76
- if t in inspected:77
- return78
- inspect.append(t)79
- inspected.add(t)80
-81
-82
-for i in candidate_paths:83
- print("Checking " + i)84
- for root, dirs, files in walk(build_path+"/"+i):85
- for file_ in files:86
- if ".ini" in file_:87
- continue88
- if ".png" in file_:89
- continue90
- if ".effect" in file_:91
- continue92
- if ".py" in file_:93
- continue94
- if ".json" in file_:95
- continue96
- path = root + "/" + file_97
- try:98
- out = check_output("{0}otool -L '{1}'".format(args.prefix, path), shell=True,99
- universal_newlines=True)100
- if "is not an object file" in out:101
- continue102
- except:103
- continue104
- rel_path = path[len(build_path)+1:]105
- print(repr(path), repr(rel_path))106
- add(rel_path)107
-108
-def add_plugins(path, replace):109
- for img in glob(path.replace(110
- "lib/QtCore.framework/Versions/5/QtCore",111
- "plugins/%s/*"%replace).replace(112
- "Library/Frameworks/QtCore.framework/Versions/5/QtCore",113
- "share/qt5/plugins/%s/*"%replace)):114
- if "_debug" in img:115
- continue116
- add(img, True, img.split("plugins/")[-1])117
-118
-actual_sparkle_path = '@loader_path/Frameworks/Sparkle.framework/Versions/A/Sparkle'119
-120
-while inspect:121
- target = inspect.pop()122
- print("inspecting", repr(target))123
- path = target.path124
- if path[0] == "@":125
- continue126
- out = check_output("{0}otool -L '{1}'".format(args.prefix, path), shell=True,127
- universal_newlines=True)128
-129
- if "QtCore" in path:130
- add_plugins(path, "platforms")131
- add_plugins(path, "imageformats")132
- add_plugins(path, "accessible")133
- add_plugins(path, "styles")134
-135
-136
- for line in out.split("\n")[1:]:137
- new = line.strip().split(" (")[0]138
- if '@' in new and "sparkle.framework" in new.lower():139
- actual_sparkle_path = new140
- print "Using sparkle path:", repr(actual_sparkle_path)141
- if not new or new[0] == "@" or new.endswith(path.split("/")[-1]):142
- continue143
- whitelisted = False144
- for i in whitelist:145
- if new.startswith(i):146
- whitelisted = True147
- if not whitelisted:148
- blacklisted = False149
- for i in blacklist:150
- if new.startswith(i):151
- blacklisted = True152
- break153
- if blacklisted:154
- continue155
- add(new, True)156
-157
-changes = list()158
-for path, external, copy_as in inspected:159
- if not external:160
- continue #built with install_rpath hopefully161
- changes.append("-change '%s' '@rpath/%s'"%(path, copy_as))162
-changes = " ".join(changes)163
-164
-info = plistlib.readPlist(plist_path)165
-166
-latest_tag = cmd('git describe --tags --abbrev=0')167
-log = cmd('git log --pretty=oneline {0}...HEAD'.format(latest_tag))168
-169
-from os import path170
-# set version171
-if args.stable:172
- info["CFBundleVersion"] = latest_tag173
- info["CFBundleShortVersionString"] = latest_tag174
- info["SUFeedURL"] = '{0}/stable/updates.xml'.format(args.base_url)175
-else:176
- info["CFBundleVersion"] = args.build_number177
- info["CFBundleShortVersionString"] = '{0}.{1}'.format(latest_tag, args.build_number)178
- info["SUFeedURL"] = '{0}/{1}/{2}/updates.xml'.format(args.base_url, args.user, args.channel)179
-180
-info["SUPublicDSAKeyFile"] = path.basename(args.public_key)181
-info["OBSFeedsURL"] = '{0}/feeds.xml'.format(args.base_url)182
-183
-app_name = info["CFBundleName"]+".app"184
-icon_file = "tmp/Contents/Resources/%s"%info["CFBundleIconFile"]185
-186
-copytree(build_path, "tmp/Contents/Resources/", symlinks=True)187
-copy(icon_path, icon_file)188
-plistlib.writePlist(info, "tmp/Contents/Info.plist")189
-makedirs("tmp/Contents/MacOS")190
-copy(run_path, "tmp/Contents/MacOS/%s"%info["CFBundleExecutable"])191
-try:192
- copy(args.public_key, "tmp/Contents/Resources")193
-except:194
- pass195
-196
-if args.sparkle is not None:197
- copytree(args.sparkle, "tmp/Contents/Frameworks/Sparkle.framework", symlinks=True)198
-199
-prefix = "tmp/Contents/Resources/"200
-sparkle_path = '@loader_path/{0}/Frameworks/Sparkle.framework/Versions/A/Sparkle'201
-202
-cmd('{0}install_name_tool -change {1} {2} {3}/bin/obs'.format(203
- args.prefix, actual_sparkle_path, sparkle_path.format('../..'), prefix))204
-205
-206
-207
-for path, external, copy_as in inspected:208
- id_ = ""209
- filename = path210
- rpath = ""211
- if external:212
- if copy_as == "Python":213
- continue214
- id_ = "-id '@rpath/%s'"%copy_as215
- filename = prefix + "bin/" +copy_as216
- rpath = "-add_rpath @loader_path/ -add_rpath @executable_path/"217
- if "/" in copy_as:218
- try:219
- dirs = copy_as.rsplit("/", 1)[0]220
- makedirs(prefix + "bin/" + dirs)221
- except:222
- pass223
- copy(path, filename)224
- else:225
- filename = path[len(build_path)+1:]226
- id_ = "-id '@rpath/../%s'"%filename227
- if not filename.startswith("bin"):228
- print(filename)229
- rpath = "-add_rpath '@loader_path/{}/'".format(ospath.relpath("bin/", ospath.dirname(filename)))230
- filename = prefix + filename231
-232
- cmd = "{0}install_name_tool {1} {2} {3} '{4}'".format(args.prefix, changes, id_, rpath, filename)233
- call(cmd, shell=True)234
-235
-try:236
- rename("tmp", app_name)237
-except:238
- print("App already exists")239
- rmtree("tmp")240
obs-studio-26.1.0.tar.xz/CI/install/osx/dylibBundler
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/makeRetinaBG
Deleted
3
1
2
-tiffutil -cathidpicheck background.png background@2x.png -out background.tiff3
obs-studio-26.1.0.tar.xz/CI/install/osx/obs.icns
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/obs.png
Deleted
obs-studio-26.1.0.tar.xz/CI/install/osx/packageApp.sh
Deleted
79
1
2
-# Exit if something fails3
-set -e4
-5
-rm -rf ./OBS.app6
-7
-mkdir OBS.app8
-mkdir OBS.app/Contents9
-mkdir OBS.app/Contents/MacOS10
-mkdir OBS.app/Contents/PlugIns11
-mkdir OBS.app/Contents/Resources12
-13
-cp -R rundir/RelWithDebInfo/bin/ ./OBS.app/Contents/MacOS14
-cp -R rundir/RelWithDebInfo/data ./OBS.app/Contents/Resources15
-cp ../CI/install/osx/obs.icns ./OBS.app/Contents/Resources16
-cp -R rundir/RelWithDebInfo/obs-plugins/ ./OBS.app/Contents/PlugIns17
-cp ../CI/install/osx/Info.plist ./OBS.app/Contents18
-19
-../CI/install/osx/dylibBundler -b -cd -d ./OBS.app/Contents/Frameworks -p @executable_path/../Frameworks/ \20
--s ./OBS.app/Contents/MacOS \21
--s /usr/local/opt/mbedtls/lib/ \22
--x ./OBS.app/Contents/PlugIns/coreaudio-encoder.so \23
--x ./OBS.app/Contents/PlugIns/decklink-ouput-ui.so \24
--x ./OBS.app/Contents/PlugIns/frontend-tools.so \25
--x ./OBS.app/Contents/PlugIns/image-source.so \26
--x ./OBS.app/Contents/PlugIns/linux-jack.so \27
--x ./OBS.app/Contents/PlugIns/mac-avcapture.so \28
--x ./OBS.app/Contents/PlugIns/mac-capture.so \29
--x ./OBS.app/Contents/PlugIns/mac-decklink.so \30
--x ./OBS.app/Contents/PlugIns/mac-syphon.so \31
--x ./OBS.app/Contents/PlugIns/mac-vth264.so \32
--x ./OBS.app/Contents/PlugIns/obs-browser.so \33
--x ./OBS.app/Contents/PlugIns/obs-browser-page \34
--x ./OBS.app/Contents/PlugIns/obs-ffmpeg.so \35
--x ./OBS.app/Contents/PlugIns/obs-filters.so \36
--x ./OBS.app/Contents/PlugIns/obs-transitions.so \37
--x ./OBS.app/Contents/PlugIns/obs-vst.so \38
--x ./OBS.app/Contents/PlugIns/rtmp-services.so \39
--x ./OBS.app/Contents/MacOS/obs \40
--x ./OBS.app/Contents/MacOS/obs-ffmpeg-mux \41
--x ./OBS.app/Contents/MacOS/obslua.so \42
--x ./OBS.app/Contents/PlugIns/obs-x264.so \43
--x ./OBS.app/Contents/PlugIns/text-freetype2.so \44
--x ./OBS.app/Contents/PlugIns/obs-libfdk.so45
-# -x ./OBS.app/Contents/MacOS/_obspython.so \46
-# -x ./OBS.app/Contents/PlugIns/obs-outputs.so \47
-48
-/usr/local/Cellar/qt/5.14.1/bin/macdeployqt ./OBS.app49
-50
-mv ./OBS.app/Contents/MacOS/libobs-opengl.so ./OBS.app/Contents/Frameworks51
-52
-rm -f -r ./OBS.app/Contents/Frameworks/QtNetwork.framework53
-54
-# put qt network in here becasuse streamdeck uses it55
-cp -R /usr/local/opt/qt/lib/QtNetwork.framework ./OBS.app/Contents/Frameworks56
-chmod -R +w ./OBS.app/Contents/Frameworks/QtNetwork.framework57
-rm -r ./OBS.app/Contents/Frameworks/QtNetwork.framework/Headers58
-rm -r ./OBS.app/Contents/Frameworks/QtNetwork.framework/Versions/5/Headers/59
-chmod 644 ./OBS.app/Contents/Frameworks/QtNetwork.framework/Versions/5/Resources/Info.plist60
-install_name_tool -id @executable_path/../Frameworks/QtNetwork.framework/Versions/5/QtNetwork ./OBS.app/Contents/Frameworks/QtNetwork.framework/Versions/5/QtNetwork61
-install_name_tool -change /usr/local/Cellar/qt/5.14.1/lib/QtCore.framework/Versions/5/QtCore @executable_path/../Frameworks/QtCore.framework/Versions/5/QtCore ./OBS.app/Contents/Frameworks/QtNetwork.framework/Versions/5/QtNetwork62
-63
-64
-# decklink ui qt65
-install_name_tool -change /usr/local/opt/qt/lib/QtGui.framework/Versions/5/QtGui @executable_path/../Frameworks/QtGui.framework/Versions/5/QtGui ./OBS.app/Contents/PlugIns/decklink-ouput-ui.so66
-install_name_tool -change /usr/local/opt/qt/lib/QtCore.framework/Versions/5/QtCore @executable_path/../Frameworks/QtCore.framework/Versions/5/QtCore ./OBS.app/Contents/PlugIns/decklink-ouput-ui.so67
-install_name_tool -change /usr/local/opt/qt/lib/QtWidgets.framework/Versions/5/QtWidgets @executable_path/../Frameworks/QtWidgets.framework/Versions/5/QtWidgets ./OBS.app/Contents/PlugIns/decklink-ouput-ui.so68
-69
-# frontend tools qt70
-install_name_tool -change /usr/local/opt/qt/lib/QtGui.framework/Versions/5/QtGui @executable_path/../Frameworks/QtGui.framework/Versions/5/QtGui ./OBS.app/Contents/PlugIns/frontend-tools.so71
-install_name_tool -change /usr/local/opt/qt/lib/QtCore.framework/Versions/5/QtCore @executable_path/../Frameworks/QtCore.framework/Versions/5/QtCore ./OBS.app/Contents/PlugIns/frontend-tools.so72
-install_name_tool -change /usr/local/opt/qt/lib/QtWidgets.framework/Versions/5/QtWidgets @executable_path/../Frameworks/QtWidgets.framework/Versions/5/QtWidgets ./OBS.app/Contents/PlugIns/frontend-tools.so73
-74
-# vst qt75
-install_name_tool -change /usr/local/opt/qt/lib/QtGui.framework/Versions/5/QtGui @executable_path/../Frameworks/QtGui.framework/Versions/5/QtGui ./OBS.app/Contents/PlugIns/obs-vst.so76
-install_name_tool -change /usr/local/opt/qt/lib/QtCore.framework/Versions/5/QtCore @executable_path/../Frameworks/QtCore.framework/Versions/5/QtCore ./OBS.app/Contents/PlugIns/obs-vst.so77
-install_name_tool -change /usr/local/opt/qt/lib/QtWidgets.framework/Versions/5/QtWidgets @executable_path/../Frameworks/QtWidgets.framework/Versions/5/QtWidgets ./OBS.app/Contents/PlugIns/obs-vst.so78
-install_name_tool -change /usr/local/opt/qt/lib/QtMacExtras.framework/Versions/5/QtMacExtras @executable_path/../Frameworks/QtMacExtras.framework/Versions/5/QtMacExtras ./OBS.app/Contents/PlugIns/obs-vst.so79
obs-studio-26.1.0.tar.xz/CI/install/osx/package_util.py
Deleted
96
1
2
-def cmd(cmd):3
- import subprocess4
- import shlex5
- return subprocess.check_output(shlex.split(cmd)).rstrip('\r\n')6
-7
-def get_tag_info(tag):8
- rev = cmd('git rev-parse {0}'.format(latest_tag))9
- anno = cmd('git cat-file -p {0}'.format(rev))10
- tag_info = []11
- for i, v in enumerate(anno.splitlines()):12
- if i <= 4:13
- continue14
- tag_info.append(v.lstrip())15
-16
- return tag_info17
-18
-def gen_html(github_user, latest_tag):19
-20
- url = 'https://github.com/{0}/obs-studio/commit/%H'.format(github_user)21
-22
- with open('readme.html', 'w') as f:23
- f.write("<html><body>")24
- log_cmd = """git log {0}...HEAD --pretty=format:'<li>• <a href="{1}">(view)</a> %s</li>'"""25
- log_res = cmd(log_cmd.format(latest_tag, url))26
- if len(log_res.splitlines()):27
- f.write('<p>Changes since {0}: (Newest to oldest)</p>'.format(latest_tag))28
- f.write(log_res)29
-30
- ul = False31
- f.write('<p>')32
- import re33
-34
- for l in get_tag_info(latest_tag):35
- if not len(l):36
- continue37
- if l.startswith('*'):38
- ul = True39
- if not ul:40
- f.write('<ul>')41
- f.write('<li>• {0}</li>'.format(re.sub(r'^(\s*)?[*](\s*)?', '', l)))42
- else:43
- if ul:44
- f.write('</ul><p/>')45
- ul = False46
- f.write('<p>{0}</p>'.format(l))47
- if ul:48
- f.write('</ul>')49
- f.write('</p></body></html>')50
-51
- cmd('textutil -convert rtf readme.html -output readme.rtf')52
- cmd("""sed -i '' 's/Times-Roman/Verdana/g' readme.rtf""")53
-54
-def save_manifest(latest_tag, user, jenkins_build, branch, stable):55
- log = cmd('git log --pretty=oneline {0}...HEAD'.format(latest_tag))56
- manifest = {}57
- manifest['commits'] = []58
- for v in log.splitlines():59
- manifest['commits'].append(v)60
- manifest['tag'] = {61
- 'name': latest_tag,62
- 'description': get_tag_info(latest_tag)63
- }64
-65
- manifest['version'] = cmd('git rev-list HEAD --count')66
- manifest['sha1'] = cmd('git rev-parse HEAD')67
- manifest['jenkins_build'] = jenkins_build68
-69
- manifest['user'] = user70
- manifest['branch'] = branch71
- manifest['stable'] = stable72
-73
- import cPickle74
- with open('manifest', 'w') as f:75
- cPickle.dump(manifest, f)76
-77
-def prepare_pkg(project, package_id):78
- cmd('packagesutil --file "{0}" set package-1 identifier {1}'.format(project, package_id))79
- cmd('packagesutil --file "{0}" set package-1 version {1}'.format(project, '1.0'))80
-81
-82
-import argparse83
-parser = argparse.ArgumentParser(description='obs-studio package util')84
-parser.add_argument('-u', '--user', dest='user', default='jp9000')85
-parser.add_argument('-p', '--package-id', dest='package_id', default='org.obsproject.pkg.obs-studio')86
-parser.add_argument('-f', '--project-file', dest='project', default='OBS.pkgproj')87
-parser.add_argument('-j', '--jenkins-build', dest='jenkins_build', default='0')88
-parser.add_argument('-b', '--branch', dest='branch', default='master')89
-parser.add_argument('-s', '--stable', dest='stable', required=False, action='store_true', default=False)90
-args = parser.parse_args()91
-92
-latest_tag = cmd('git describe --tags --abbrev=0')93
-gen_html(args.user, latest_tag)94
-prepare_pkg(args.project, args.package_id)95
-save_manifest(latest_tag, args.user, args.jenkins_build, args.branch, args.stable)96
obs-studio-26.1.0.tar.xz/CI/install/osx/post-install.sh
Deleted
3
1
2
-#!/usr/bin/env bash3
obs-studio-26.1.0.tar.xz/CI/install/osx/settings.json
Deleted
15
1
2
-{3
- "title": "OBS",4
- "background": "../CI/install/osx/background.tiff",5
- "format": "UDZO",6
- "compression-level": 9,7
- "window": { "position": { "x": 100, "y": 100 },8
- "size": { "width": 540, "height": 380 } },9
- "contents": [10
- { "x": 120, "y": 180, "type": "file",11
- "path": "./OBS.app" },12
- { "x": 420, "y": 180, "type": "link", "path": "/Applications" }13
- ]14
-}15
obs-studio-26.1.0.tar.xz/CI/osxcert
Deleted
2
1
-(directory)2
obs-studio-26.1.0.tar.xz/CI/osxcert/Certificates.p12.enc
Deleted
obs-studio-26.1.0.tar.xz/CI/util
Deleted
2
1
-(directory)2
obs-studio-26.1.0.tar.xz/CI/util/build-package-deps-osx.sh
Deleted
163
1
2
-#!/usr/bin/env bash3
-4
-set -e5
-6
-# This script builds a tar file that contains a bunch of deps that OBS needs for7
-# advanced functionality on OSX. Currently this tar file is pulled down off of s38
-# and used in the CI build process on travis.9
-# Mostly this sets build flags to compile with older SDKS and make sure that 10
-# the libs are portable.11
-12
-exists()13
-{14
- command -v "$1" >/dev/null 2>&115
-}16
-17
-if ! exists nasm; then18
- echo "nasm not found. Try brew install nasm"19
- exit20
-fi21
-22
-CURDIR=$(pwd)23
-24
-# the temp directory25
-WORK_DIR=`mktemp -d`26
-27
-# deletes the temp directory28
-function cleanup {29
- #rm -rf "$WORK_DIR"30
- echo "Deleted temp working directory $WORK_DIR"31
-}32
-33
-# register the cleanup function to be called on the EXIT signal34
-trap cleanup EXIT35
-36
-cd $WORK_DIR37
-38
-DEPS_DEST=$WORK_DIR/obsdeps39
-40
-# make dest dirs41
-mkdir $DEPS_DEST42
-mkdir $DEPS_DEST/bin43
-mkdir $DEPS_DEST/include44
-mkdir $DEPS_DEST/lib45
-46
-# OSX COMPAT47
-export MACOSX_DEPLOYMENT_TARGET=10.1148
-49
-# If you need an olders SDK and Xcode won't give it to you50
-# https://github.com/phracker/MacOSX-SDKs51
-52
-# libopus53
-curl -L -O https://ftp.osuosl.org/pub/xiph/releases/opus/opus-1.2.1.tar.gz54
-tar -xf opus-1.2.1.tar.gz55
-cd ./opus-1.2.156
-mkdir build57
-cd ./build58
-../configure --disable-shared --enable-static --prefix="/tmp/obsdeps"59
-make -j 1260
-make install61
-62
-cd $WORK_DIR63
-64
-# libogg65
-curl -L -O https://ftp.osuosl.org/pub/xiph/releases/ogg/libogg-1.3.3.tar.gz66
-tar -xf libogg-1.3.3.tar.gz67
-cd ./libogg-1.3.368
-mkdir build69
-cd ./build70
-../configure --disable-shared --enable-static --prefix="/tmp/obsdeps"71
-make -j 1272
-make install73
-74
-cd $WORK_DIR75
-76
-# libvorbis77
-curl -L -O https://ftp.osuosl.org/pub/xiph/releases/vorbis/libvorbis-1.3.6.tar.gz78
-tar -xf libvorbis-1.3.6.tar.gz79
-cd ./libvorbis-1.3.680
-mkdir build81
-cd ./build82
-../configure --disable-shared --enable-static --prefix="/tmp/obsdeps"83
-make -j 1284
-make install85
-86
-cd $WORK_DIR87
-88
-# libvpx89
-curl -L -O https://chromium.googlesource.com/webm/libvpx/+archive/v1.7.0.tar.gz90
-mkdir -p ./libvpx-v1.7.091
-tar -xf v1.7.0.tar.gz -C $PWD/libvpx-v1.7.092
-cd ./libvpx-v1.7.093
-mkdir -p build94
-cd ./build95
-../configure --disable-shared --prefix="/tmp/obsdeps" --libdir="/tmp/obsdeps/lib"96
-make -j 1297
-make install98
-99
-cd $WORK_DIR100
-101
-# x264102
-git clone git://git.videolan.org/x264.git103
-cd ./x264104
-git checkout origin/stable105
-mkdir build106
-cd ./build107
-../configure --extra-ldflags="-mmacosx-version-min=10.11" --enable-static --prefix="/tmp/obsdeps"108
-make -j 12109
-make install110
-../configure --extra-ldflags="-mmacosx-version-min=10.11" --enable-shared --libdir="/tmp/obsdeps/bin" --prefix="/tmp/obsdeps"111
-make -j 12112
-ln -f -s libx264.*.dylib libx264.dylib113
-find . -name \*.dylib -exec cp \{\} $DEPS_DEST/bin/ \;114
-rsync -avh --include="*/" --include="*.h" --exclude="*" ../* $DEPS_DEST/include/115
-rsync -avh --include="*/" --include="*.h" --exclude="*" ./* $DEPS_DEST/include/116
-117
-cd $WORK_DIR118
-119
-# janson120
-curl -L -O http://www.digip.org/jansson/releases/jansson-2.11.tar.gz121
-tar -xf jansson-2.11.tar.gz122
-cd jansson-2.11123
-mkdir build124
-cd ./build125
-../configure --libdir="/tmp/obsdeps/bin" --enable-shared --disable-static126
-make -j 12127
-find . -name \*.dylib -exec cp \{\} $DEPS_DEST/bin/ \;128
-rsync -avh --include="*/" --include="*.h" --exclude="*" ../* $DEPS_DEST/include/129
-rsync -avh --include="*/" --include="*.h" --exclude="*" ./* $DEPS_DEST/include/130
-131
-cd $WORK_DIR132
-133
-export LDFLAGS="-L/tmp/obsdeps/lib"134
-export CFLAGS="-I/tmp/obsdeps/include"135
-136
-# FFMPEG137
-curl -L -O https://github.com/FFmpeg/FFmpeg/archive/n4.0.2.zip138
-unzip ./n4.0.2.zip139
-cd ./FFmpeg-n4.0.2140
-mkdir build141
-cd ./build142
-../configure --pkg-config-flags="--static" --extra-ldflags="-mmacosx-version-min=10.11" --enable-shared --disable-static --shlibdir="/tmp/obsdeps/bin" --enable-gpl --disable-doc --enable-libx264 --enable-libopus --enable-libvorbis --enable-libvpx --disable-outdev=sdl143
-make -j 12144
-find . -name \*.dylib -exec cp \{\} $DEPS_DEST/bin/ \;145
-rsync -avh --include="*/" --include="*.h" --exclude="*" ../* $DEPS_DEST/include/146
-rsync -avh --include="*/" --include="*.h" --exclude="*" ./* $DEPS_DEST/include/147
-148
-#luajit149
-curl -L -O https://luajit.org/download/LuaJIT-2.0.5.tar.gz150
-tar -xf LuaJIT-2.0.5.tar.gz151
-cd LuaJIT-2.0.5152
-make PREFIX=/tmp/obsdeps153
-make PREFIX=/tmp/obsdeps install154
-find /tmp/obsdeps/lib -name libluajit\*.dylib -exec cp \{\} $DEPS_DEST/lib/ \;155
-rsync -avh --include="*/" --include="*.h" --exclude="*" src/* $DEPS_DEST/include/156
-make PREFIX=/tmp/obsdeps uninstall157
-158
-cd $WORK_DIR159
-160
-tar -czf osx-deps.tar.gz obsdeps161
-162
-cp ./osx-deps.tar.gz $CURDIR163
obs-studio-26.1.0.tar.xz/CI/util/win32.sh
Deleted
71
1
2
-#/bin/bash3
-4
-cd x2645
-make clean6
-LDFLAGS="-static-libgcc" ./configure --enable-shared --enable-win32thread --disable-avs --disable-ffms --disable-gpac --disable-interlaced --disable-lavf --cross-prefix=i686-w64-mingw32- --host=i686-pc-mingw32 --prefix="/home/jim/packages/win32"7
-make -j6 fprofiled VIDS="CITY_704x576_60_orig_01.yuv"8
-make install9
-i686-w64-mingw32-dlltool -z /home/jim/packages/win32/bin/x264.orig.def --export-all-symbols /home/jim/packages/win32/bin/libx264-148.dll10
-grep "EXPORTS\|x264" /home/jim/packages/win32/bin/x264.orig.def > /home/jim/packages/win32/bin/x264.def11
-rm -f /home/jim/packages/win32/bin/x264.org.def12
-sed -i -e "/\\t.*DATA/d" -e "/\\t\".*/d" -e "s/\s@.*//" /home/jim/packages/win32/bin/x264.def13
-i686-w64-mingw32-dlltool -m i386 -d /home/jim/packages/win32/bin/x264.def -l /home/jim/packages/win32/bin/x264.lib -D /home/jim/win32/packages/bin/libx264-148.dll14
-cd ..15
-16
-cd opus17
-make clean18
-LDFLAGS="-static-libgcc" ./configure -host=i686-w64-mingw32 --prefix="/home/jim/packages/win32" --enable-shared19
-make -j620
-make install21
-cd ..22
-23
-cd zlib/build3224
-make clean25
-cmake .. -DCMAKE_SYSTEM_NAME=Windows -DCMAKE_C_COMPILER=i686-w64-mingw32-gcc -DCMAKE_INSTALL_PREFIX=/home/jim/packages/win32 -DINSTALL_PKGCONFIG_DIR=/home/jim/packages/win32/lib/pkgconfig -DCMAKE_RC_COMPILER=i686-w64-mingw32-windres -DCMAKE_SHARED_LINKER_FLAGS="-static-libgcc"26
-make -j627
-make install28
-mv ../../win32/lib/libzlib.dll.a ../../win32/lib/libz.dll.a29
-mv ../../win32/lib/libzlibstatic.a ../../win32/lib/libz.a30
-cp ../win32/zlib.def /home/jim/packages/win32/bin31
-i686-w64-mingw32-dlltool -m i386 -d ../win32/zlib.def -l /home/jim/packages/win32/bin/zlib.lib -D /home/jim/win32/packages/bin/zlib.dll32
-cd ../..33
-34
-cd libpng35
-make clean36
-PKG_CONFIG_PATH="/home/jim/packages/win32/lib/pkgconfig" LDFLAGS="-L/home/jim/packages/win32/lib -static-libgcc" CPPFLAGS="-I/home/jim/packages/win32/include" ./configure -host=i686-w64-mingw32 --prefix="/home/jim/packages/win32" --enable-shared37
-make -j638
-make install39
-cd ..40
-41
-cd libogg42
-make clean43
-PKG_CONFIG_PATH="/home/jim/packages/win32/lib/pkgconfig" LDFLAGS="-L/home/jim/packages/win32/lib -static-libgcc" CPPFLAGS="-I/home/jim/packages/win32/include" ./configure -host=i686-w64-mingw32 --prefix="/home/jim/packages/win32" --enable-shared44
-make -j645
-make install46
-cd ..47
-48
-cd libvorbis49
-make clean50
-PKG_CONFIG_PATH="/home/jim/packages/win32/lib/pkgconfig" LDFLAGS="-L/home/jim/packages/win32/lib -static-libgcc" CPPFLAGS="-I/home/jim/packages/win32/include" ./configure -host=i686-w64-mingw32 --prefix="/home/jim/packages/win32" --enable-shared --with-ogg="/home/jim/packages/win32"51
-make -j652
-make install53
-cd ..54
-55
-cd libvpxbuild56
-make clean57
-PKG_CONFIG_PATH="/home/jim/packages/win32/lib/pkgconfig" CROSS=i686-w64-mingw32- LDFLAGS="-static-libgcc" ../libvpx/configure --prefix=/home/jim/packages/win32 --enable-vp8 --enable-vp9 --disable-docs --disable-examples --enable-shared --disable-static --enable-runtime-cpu-detect --enable-realtime-only --disable-install-bins --disable-install-docs --disable-unit-tests --target=x86-win32-gcc58
-make -j659
-make install60
-i686-w64-mingw32-dlltool -m i386 -d libvpx.def -l /home/jim/packages/win32/bin/vpx.lib -D /home/jim/win32/packages/bin/libvpx-1.dll61
-cd ..62
-63
-cd ffmpeg64
-make clean65
-cp /media/sf_linux/nvEncodeAPI.h /home/jim/packages/win32/include66
-PKG_CONFIG_PATH="/home/jim/packages/win32/lib/pkgconfig" LDFLAGS="-L/home/jim/packages/win32/lib -static-libgcc" CFLAGS="-I/home/jim/packages/win32/include" ./configure --enable-memalign-hack --enable-gpl --disable-programs --disable-doc --arch=x86 --enable-shared --enable-nvenc --enable-libx264 --enable-libopus --enable-libvorbis --enable-libvpx --disable-debug --cross-prefix=i686-w64-mingw32- --target-os=mingw32 --pkg-config=pkg-config --prefix="/home/jim/packages/win32" --disable-postproc67
-read -n1 -r -p "Press any key to continue building FFmpeg..." key68
-make -j669
-make install70
-cd ..71
obs-studio-26.1.0.tar.xz/CI/util/win64.sh
Deleted
71
1
2
-#/bin/bash3
-4
-cd x2645
-make clean6
-LDFLAGS="-static-libgcc" ./configure --enable-shared --enable-win32thread --disable-avs --disable-ffms --disable-gpac --disable-interlaced --disable-lavf --cross-prefix=x86_64-w64-mingw32- --host=x86_64-pc-mingw32 --prefix="/home/jim/packages/win64"7
-make -j6 fprofiled VIDS="CITY_704x576_60_orig_01.yuv"8
-make install9
-x86_64-w64-mingw32-dlltool -z /home/jim/packages/win64/bin/x264.orig.def --export-all-symbols /home/jim/packages/win64/bin/libx264-148.dll10
-grep "EXPORTS\|x264" /home/jim/packages/win64/bin/x264.orig.def > /home/jim/packages/win64/bin/x264.def11
-rm -f /home/jim/packages/win64/bin/x264.org.def12
-sed -i -e "/\\t.*DATA/d" -e "/\\t\".*/d" -e "s/\s@.*//" /home/jim/packages/win64/bin/x264.def13
-x86_64-w64-mingw32-dlltool -m i386:x86-64 -d /home/jim/packages/win64/bin/x264.def -l /home/jim/packages/win64/bin/x264.lib -D /home/jim/win64/packages/bin/libx264-148.dll14
-cd ..15
-16
-cd opus17
-make clean18
-LDFLAGS="-static-libgcc" ./configure -host=x86_64-w64-mingw32 --prefix="/home/jim/packages/win64" --enable-shared19
-make -j620
-make install21
-cd ..22
-23
-cd zlib/build6424
-make clean25
-cmake .. -DCMAKE_SYSTEM_NAME=Windows -DCMAKE_C_COMPILER=x86_64-w64-mingw32-gcc -DCMAKE_INSTALL_PREFIX=/home/jim/packages/win64 -DCMAKE_RC_COMPILER=x86_64-w64-mingw32-windres -DCMAKE_SHARED_LINKER_FLAGS="-static-libgcc"26
-make -j627
-make install28
-mv ../../win64/lib/libzlib.dll.a ../../win64/lib/libz.dll.a29
-mv ../../win64/lib/libzlibstatic.a ../../win64/lib/libz.a30
-cp ../win64/zlib.def /home/jim/packages/win64/bin31
-x86_64-w64-mingw32-dlltool -m i386:x86-64 -d ../win32/zlib.def -l /home/jim/packages/win64/bin/zlib.lib -D /home/jim/win64/packages/bin/zlib.dll32
-cd ../..33
-34
-cd libpng35
-make clean36
-PKG_CONFIG_PATH="/home/jim/packages/win64/lib/pkgconfig" LDFLAGS="-L/home/jim/packages/win64/lib" CPPFLAGS="-I/home/jim/packages/win64/include" ./configure -host=x86_64-w64-mingw32 --prefix="/home/jim/packages/win64" --enable-shared37
-make -j638
-make install39
-cd ..40
-41
-cd libogg42
-make clean43
-PKG_CONFIG_PATH="/home/jim/packages/win64/lib/pkgconfig" LDFLAGS="-L/home/jim/packages/win64/lib -static-libgcc" CPPFLAGS="-I/home/jim/packages/win64/include" ./configure -host=x86_64-w64-mingw32 --prefix="/home/jim/packages/win64" --enable-shared44
-make -j645
-make install46
-cd ..47
-48
-cd libvorbis49
-make clean50
-PKG_CONFIG_PATH="/home/jim/packages/win64/lib/pkgconfig" LDFLAGS="-L/home/jim/packages/win64/lib -static-libgcc" CPPFLAGS="-I/home/jim/packages/win64/include" ./configure -host=x86_64-w64-mingw32 --prefix="/home/jim/packages/win64" --enable-shared --with-ogg="/home/jim/packages/win64"51
-make -j652
-make install53
-cd ..54
-55
-cd libvpxbuild56
-make clean57
-PKG_CONFIG_PATH="/home/jim/packages/win64/lib/pkgconfig" CROSS=x86_64-w64-mingw32- LDFLAGS="-static-libgcc" ../libvpx/configure --prefix=/home/jim/packages/win64 --enable-vp8 --enable-vp9 --disable-docs --disable-examples --enable-shared --disable-static --enable-runtime-cpu-detect --enable-realtime-only --disable-install-bins --disable-install-docs --disable-unit-tests --target=x86_64-win64-gcc58
-make -j659
-make install60
-x86_64-w64-mingw32-dlltool -m i386:x86-64 -d libvpx.def -l /home/jim/packages/win64/bin/vpx.lib -D /home/jim/win64/packages/bin/libvpx-1.dll61
-cd ..62
-63
-cd ffmpeg64
-make clean65
-cp /media/sf_linux/nvEncodeAPI.h /home/jim/packages/win64/include66
-PKG_CONFIG_PATH="/home/jim/packages/win64/lib/pkgconfig" LDFLAGS="-L/home/jim/packages/win64/lib" CPPFLAGS="-I/home/jim/packages/win64/include" ./configure --enable-memalign-hack --enable-gpl --disable-doc --arch=x86_64 --enable-shared --enable-nvenc --enable-libx264 --enable-libopus --enable-libvorbis --enable-libvpx --disable-debug --cross-prefix=x86_64-w64-mingw32- --target-os=mingw32 --pkg-config=pkg-config --prefix="/home/jim/packages/win64" --disable-postproc67
-read -n1 -r -p "Press any key to continue building FFmpeg..." key68
-make -j669
-make install70
-cd ..71
obs-studio-26.1.0.tar.xz/libobs/util/simde/mmx.h
Deleted
2272
1
2
-/* SPDX-License-Identifier: MIT3
- *4
- * Permission is hereby granted, free of charge, to any person5
- * obtaining a copy of this software and associated documentation6
- * files (the "Software"), to deal in the Software without7
- * restriction, including without limitation the rights to use, copy,8
- * modify, merge, publish, distribute, sublicense, and/or sell copies9
- * of the Software, and to permit persons to whom the Software is10
- * furnished to do so, subject to the following conditions:11
- *12
- * The above copyright notice and this permission notice shall be13
- * included in all copies or substantial portions of the Software.14
- *15
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,16
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF17
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND18
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS19
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN20
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN21
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE22
- * SOFTWARE.23
- *24
- * Copyright:25
- * 2017-2020 Evan Nemerson <evan@nemerson.com>26
- */27
-28
-#if !defined(SIMDE_X86_MMX_H)29
-#define SIMDE_X86_MMX_H30
-31
-#include "simde-common.h"32
-33
-#if !defined(SIMDE_X86_MMX_NATIVE) && defined(SIMDE_ENABLE_NATIVE_ALIASES)34
-#define SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES35
-#endif36
-37
-HEDLEY_DIAGNOSTIC_PUSH38
-SIMDE_DISABLE_UNWANTED_DIAGNOSTICS39
-40
-#if defined(SIMDE_X86_MMX_NATIVE)41
-#define SIMDE_X86_MMX_USE_NATIVE_TYPE42
-#elif defined(SIMDE_X86_SSE_NATIVE)43
-#define SIMDE_X86_MMX_USE_NATIVE_TYPE44
-#endif45
-46
-#if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)47
-#include <mmintrin.h>48
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)49
-#include <arm_neon.h>50
-#endif51
-52
-#include <stdint.h>53
-#include <limits.h>54
-55
-SIMDE_BEGIN_DECLS_56
-57
-typedef union {58
-#if defined(SIMDE_VECTOR_SUBSCRIPT)59
- SIMDE_ALIGN(8) int8_t i8 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;60
- SIMDE_ALIGN(8) int16_t i16 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;61
- SIMDE_ALIGN(8) int32_t i32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;62
- SIMDE_ALIGN(8) int64_t i64 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;63
- SIMDE_ALIGN(8) uint8_t u8 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;64
- SIMDE_ALIGN(8) uint16_t u16 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;65
- SIMDE_ALIGN(8) uint32_t u32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;66
- SIMDE_ALIGN(8) uint64_t u64 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;67
- SIMDE_ALIGN(8) simde_float32 f32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;68
- SIMDE_ALIGN(8) int_fast32_t i32f SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;69
- SIMDE_ALIGN(8) uint_fast32_t u32f SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;70
-#else71
- SIMDE_ALIGN(8) int8_t i8[8];72
- SIMDE_ALIGN(8) int16_t i16[4];73
- SIMDE_ALIGN(8) int32_t i32[2];74
- SIMDE_ALIGN(8) int64_t i64[1];75
- SIMDE_ALIGN(8) uint8_t u8[8];76
- SIMDE_ALIGN(8) uint16_t u16[4];77
- SIMDE_ALIGN(8) uint32_t u32[2];78
- SIMDE_ALIGN(8) uint64_t u64[1];79
- SIMDE_ALIGN(8) simde_float32 f32[2];80
- SIMDE_ALIGN(8) int_fast32_t i32f[8 / sizeof(int_fast32_t)];81
- SIMDE_ALIGN(8) uint_fast32_t u32f[8 / sizeof(uint_fast32_t)];82
-#endif83
-84
-#if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)85
- __m64 n;86
-#endif87
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)88
- int8x8_t neon_i8;89
- int16x4_t neon_i16;90
- int32x2_t neon_i32;91
- int64x1_t neon_i64;92
- uint8x8_t neon_u8;93
- uint16x4_t neon_u16;94
- uint32x2_t neon_u32;95
- uint64x1_t neon_u64;96
- float32x2_t neon_f32;97
-#endif98
-} simde__m64_private;99
-100
-#if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)101
-typedef __m64 simde__m64;102
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)103
-typedef int32x2_t simde__m64;104
-#elif defined(SIMDE_VECTOR_SUBSCRIPT)105
-typedef int32_t simde__m64 SIMDE_ALIGN(8) SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;106
-#else107
-typedef simde__m64_private simde__m64;108
-#endif109
-110
-#if !defined(SIMDE_X86_MMX_USE_NATIVE_TYPE) && \111
- defined(SIMDE_ENABLE_NATIVE_ALIASES)112
-#define SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES113
-typedef simde__m64 __m64;114
-#endif115
-116
-HEDLEY_STATIC_ASSERT(8 == sizeof(simde__m64), "__m64 size incorrect");117
-HEDLEY_STATIC_ASSERT(8 == sizeof(simde__m64_private), "__m64 size incorrect");118
-#if defined(SIMDE_CHECK_ALIGNMENT) && defined(SIMDE_ALIGN_OF)119
-HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m64) == 8,120
- "simde__m64 is not 8-byte aligned");121
-HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m64_private) == 8,122
- "simde__m64_private is not 8-byte aligned");123
-#endif124
-125
-SIMDE_FUNCTION_ATTRIBUTES126
-simde__m64 simde__m64_from_private(simde__m64_private v)127
-{128
- simde__m64 r;129
- simde_memcpy(&r, &v, sizeof(r));130
- return r;131
-}132
-133
-SIMDE_FUNCTION_ATTRIBUTES134
-simde__m64_private simde__m64_to_private(simde__m64 v)135
-{136
- simde__m64_private r;137
- simde_memcpy(&r, &v, sizeof(r));138
- return r;139
-}140
-141
-#define SIMDE_X86_GENERATE_CONVERSION_FUNCTION(simde_type, source_type, isax, \142
- fragment) \143
- SIMDE_FUNCTION_ATTRIBUTES \144
- simde__##simde_type simde__##simde_type##_from_##isax##_##fragment( \145
- source_type value) \146
- { \147
- simde__##simde_type##_private r_; \148
- r_.isax##_##fragment = value; \149
- return simde__##simde_type##_from_private(r_); \150
- } \151
- \152
- SIMDE_FUNCTION_ATTRIBUTES \153
- source_type simde__##simde_type##_to_##isax##_##fragment( \154
- simde__##simde_type value) \155
- { \156
- simde__##simde_type##_private r_ = \157
- simde__##simde_type##_to_private(value); \158
- return r_.isax##_##fragment; \159
- }160
-161
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)162
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int8x8_t, neon, i8)163
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int16x4_t, neon, i16)164
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int32x2_t, neon, i32)165
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int64x1_t, neon, i64)166
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint8x8_t, neon, u8)167
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint16x4_t, neon, u16)168
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint32x2_t, neon, u32)169
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint64x1_t, neon, u64)170
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, float32x2_t, neon, f32)171
-#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */172
-173
-SIMDE_FUNCTION_ATTRIBUTES174
-simde__m64 simde_mm_add_pi8(simde__m64 a, simde__m64 b)175
-{176
-#if defined(SIMDE_X86_MMX_NATIVE)177
- return _mm_add_pi8(a, b);178
-#else179
- simde__m64_private r_;180
- simde__m64_private a_ = simde__m64_to_private(a);181
- simde__m64_private b_ = simde__m64_to_private(b);182
-183
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)184
- r_.neon_i8 = vadd_s8(a_.neon_i8, b_.neon_i8);185
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)186
- r_.i8 = a_.i8 + b_.i8;187
-#else188
- SIMDE_VECTORIZE189
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {190
- r_.i8[i] = a_.i8[i] + b_.i8[i];191
- }192
-#endif193
-194
- return simde__m64_from_private(r_);195
-#endif196
-}197
-#define simde_m_paddb(a, b) simde_mm_add_pi8(a, b)198
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)199
-#define _mm_add_pi8(a, b) simde_mm_add_pi8(a, b)200
-#define _m_paddb(a, b) simde_m_paddb(a, b)201
-#endif202
-203
-SIMDE_FUNCTION_ATTRIBUTES204
-simde__m64 simde_mm_add_pi16(simde__m64 a, simde__m64 b)205
-{206
-#if defined(SIMDE_X86_MMX_NATIVE)207
- return _mm_add_pi16(a, b);208
-#else209
- simde__m64_private r_;210
- simde__m64_private a_ = simde__m64_to_private(a);211
- simde__m64_private b_ = simde__m64_to_private(b);212
-213
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)214
- r_.neon_i16 = vadd_s16(a_.neon_i16, b_.neon_i16);215
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)216
- r_.i16 = a_.i16 + b_.i16;217
-#else218
- SIMDE_VECTORIZE219
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {220
- r_.i16[i] = a_.i16[i] + b_.i16[i];221
- }222
-#endif223
-224
- return simde__m64_from_private(r_);225
-#endif226
-}227
-#define simde_m_paddw(a, b) simde_mm_add_pi16(a, b)228
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)229
-#define _mm_add_pi16(a, b) simde_mm_add_pi16(a, b)230
-#define _m_add_paddw(a, b) simde_mm_add_pi16(a, b)231
-#endif232
-233
-SIMDE_FUNCTION_ATTRIBUTES234
-simde__m64 simde_mm_add_pi32(simde__m64 a, simde__m64 b)235
-{236
-#if defined(SIMDE_X86_MMX_NATIVE)237
- return _mm_add_pi32(a, b);238
-#else239
- simde__m64_private r_;240
- simde__m64_private a_ = simde__m64_to_private(a);241
- simde__m64_private b_ = simde__m64_to_private(b);242
-243
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)244
- r_.neon_i32 = vadd_s32(a_.neon_i32, b_.neon_i32);245
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)246
- r_.i32 = a_.i32 + b_.i32;247
-#else248
- SIMDE_VECTORIZE249
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {250
- r_.i32[i] = a_.i32[i] + b_.i32[i];251
- }252
-#endif253
-254
- return simde__m64_from_private(r_);255
-#endif256
-}257
-#define simde_m_paddd(a, b) simde_mm_add_pi32(a, b)258
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)259
-#define _mm_add_pi32(a, b) simde_mm_add_pi32(a, b)260
-#define _m_add_paddd(a, b) simde_mm_add_pi32(a, b)261
-#endif262
-263
-SIMDE_FUNCTION_ATTRIBUTES264
-simde__m64 simde_mm_adds_pi8(simde__m64 a, simde__m64 b)265
-{266
-#if defined(SIMDE_X86_MMX_NATIVE)267
- return _mm_adds_pi8(a, b);268
-#else269
- simde__m64_private r_, a_ = simde__m64_to_private(a),270
- b_ = simde__m64_to_private(b);271
-272
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)273
- r_.neon_i8 = vqadd_s8(a_.neon_i8, b_.neon_i8);274
-#else275
- SIMDE_VECTORIZE276
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {277
- if ((((b_.i8[i]) > 0) &&278
- ((a_.i8[i]) > (INT8_MAX - (b_.i8[i]))))) {279
- r_.i8[i] = INT8_MAX;280
- } else if ((((b_.i8[i]) < 0) &&281
- ((a_.i8[i]) < (INT8_MIN - (b_.i8[i]))))) {282
- r_.i8[i] = INT8_MIN;283
- } else {284
- r_.i8[i] = (a_.i8[i]) + (b_.i8[i]);285
- }286
- }287
-#endif288
-289
- return simde__m64_from_private(r_);290
-#endif291
-}292
-#define simde_m_paddsb(a, b) simde_mm_adds_pi8(a, b)293
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)294
-#define _mm_adds_pi8(a, b) simde_mm_adds_pi8(a, b)295
-#define _m_add_paddsb(a, b) simde_mm_adds_pi8(a, b)296
-#endif297
-298
-SIMDE_FUNCTION_ATTRIBUTES299
-simde__m64 simde_mm_adds_pu8(simde__m64 a, simde__m64 b)300
-{301
-#if defined(SIMDE_X86_MMX_NATIVE)302
- return _mm_adds_pu8(a, b);303
-#else304
- simde__m64_private r_;305
- simde__m64_private a_ = simde__m64_to_private(a);306
- simde__m64_private b_ = simde__m64_to_private(b);307
-308
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)309
- r_.neon_u8 = vqadd_u8(a_.neon_u8, b_.neon_u8);310
-#else311
- SIMDE_VECTORIZE312
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {313
- const uint_fast16_t x =314
- HEDLEY_STATIC_CAST(uint_fast16_t, a_.u8[i]) +315
- HEDLEY_STATIC_CAST(uint_fast16_t, b_.u8[i]);316
- if (x > UINT8_MAX)317
- r_.u8[i] = UINT8_MAX;318
- else319
- r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, x);320
- }321
-#endif322
-323
- return simde__m64_from_private(r_);324
-#endif325
-}326
-#define simde_m_paddusb(a, b) simde_mm_adds_pu8(a, b)327
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)328
-#define _mm_adds_pu8(a, b) simde_mm_adds_pu8(a, b)329
-#define _m_paddusb(a, b) simde_mm_adds_pu8(a, b)330
-#endif331
-332
-SIMDE_FUNCTION_ATTRIBUTES333
-simde__m64 simde_mm_adds_pi16(simde__m64 a, simde__m64 b)334
-{335
-#if defined(SIMDE_X86_MMX_NATIVE)336
- return _mm_adds_pi16(a, b);337
-#else338
- simde__m64_private r_;339
- simde__m64_private a_ = simde__m64_to_private(a);340
- simde__m64_private b_ = simde__m64_to_private(b);341
-342
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)343
- r_.neon_i16 = vqadd_s16(a_.neon_i16, b_.neon_i16);344
-#else345
- SIMDE_VECTORIZE346
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {347
- if ((((b_.i16[i]) > 0) &&348
- ((a_.i16[i]) > (INT16_MAX - (b_.i16[i]))))) {349
- r_.i16[i] = INT16_MAX;350
- } else if ((((b_.i16[i]) < 0) &&351
- ((a_.i16[i]) < (SHRT_MIN - (b_.i16[i]))))) {352
- r_.i16[i] = SHRT_MIN;353
- } else {354
- r_.i16[i] = (a_.i16[i]) + (b_.i16[i]);355
- }356
- }357
-#endif358
-359
- return simde__m64_from_private(r_);360
-#endif361
-}362
-#define simde_m_paddsw(a, b) simde_mm_adds_pi16(a, b)363
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)364
-#define _mm_adds_pi16(a, b) simde_mm_adds_pi16(a, b)365
-#define _m_paddsw(a, b) simde_mm_adds_pi16(a, b)366
-#endif367
-368
-SIMDE_FUNCTION_ATTRIBUTES369
-simde__m64 simde_mm_adds_pu16(simde__m64 a, simde__m64 b)370
-{371
-#if defined(SIMDE_X86_MMX_NATIVE)372
- return _mm_adds_pu16(a, b);373
-#else374
- simde__m64_private r_;375
- simde__m64_private a_ = simde__m64_to_private(a);376
- simde__m64_private b_ = simde__m64_to_private(b);377
-378
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)379
- r_.neon_u16 = vqadd_u16(a_.neon_u16, b_.neon_u16);380
-#else381
- SIMDE_VECTORIZE382
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {383
- const uint32_t x = a_.u16[i] + b_.u16[i];384
- if (x > UINT16_MAX)385
- r_.u16[i] = UINT16_MAX;386
- else387
- r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, x);388
- }389
-#endif390
-391
- return simde__m64_from_private(r_);392
-#endif393
-}394
-#define simde_m_paddusw(a, b) simde_mm_adds_pu16(a, b)395
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)396
-#define _mm_adds_pu16(a, b) simde_mm_adds_pu16(a, b)397
-#define _m_paddusw(a, b) simde_mm_adds_pu16(a, b)398
-#endif399
-400
-SIMDE_FUNCTION_ATTRIBUTES401
-simde__m64 simde_mm_and_si64(simde__m64 a, simde__m64 b)402
-{403
-#if defined(SIMDE_X86_MMX_NATIVE)404
- return _mm_and_si64(a, b);405
-#else406
- simde__m64_private r_;407
- simde__m64_private a_ = simde__m64_to_private(a);408
- simde__m64_private b_ = simde__m64_to_private(b);409
-410
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)411
- r_.neon_i32 = vand_s32(a_.neon_i32, b_.neon_i32);412
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)413
- r_.i64 = a_.i64 & b_.i64;414
-#else415
- r_.i64[0] = a_.i64[0] & b_.i64[0];416
-#endif417
-418
- return simde__m64_from_private(r_);419
-#endif420
-}421
-#define simde_m_pand(a, b) simde_mm_and_si64(a, b)422
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)423
-#define _mm_and_si64(a, b) simde_mm_and_si64(a, b)424
-#define _m_pand(a, b) simde_mm_and_si64(a, b)425
-#endif426
-427
-SIMDE_FUNCTION_ATTRIBUTES428
-simde__m64 simde_mm_andnot_si64(simde__m64 a, simde__m64 b)429
-{430
-#if defined(SIMDE_X86_MMX_NATIVE)431
- return _mm_andnot_si64(a, b);432
-#else433
- simde__m64_private r_;434
- simde__m64_private a_ = simde__m64_to_private(a);435
- simde__m64_private b_ = simde__m64_to_private(b);436
-437
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)438
- r_.neon_i32 = vbic_s32(b_.neon_i32, a_.neon_i32);439
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)440
- r_.i32f = ~a_.i32f & b_.i32f;441
-#else442
- r_.u64[0] = (~(a_.u64[0])) & (b_.u64[0]);443
-#endif444
-445
- return simde__m64_from_private(r_);446
-#endif447
-}448
-#define simde_m_pandn(a, b) simde_mm_andnot_si64(a, b)449
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)450
-#define _mm_andnot_si64(a, b) simde_mm_andnot_si64(a, b)451
-#define _m_pandn(a, b) simde_mm_andnot_si64(a, b)452
-#endif453
-454
-SIMDE_FUNCTION_ATTRIBUTES455
-simde__m64 simde_mm_cmpeq_pi8(simde__m64 a, simde__m64 b)456
-{457
-#if defined(SIMDE_X86_MMX_NATIVE)458
- return _mm_cmpeq_pi8(a, b);459
-#else460
- simde__m64_private r_;461
- simde__m64_private a_ = simde__m64_to_private(a);462
- simde__m64_private b_ = simde__m64_to_private(b);463
-464
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)465
- r_.neon_i8 = vreinterpret_s8_u8(vceq_s8(a_.neon_i8, b_.neon_i8));466
-#else467
- SIMDE_VECTORIZE468
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {469
- r_.i8[i] = (a_.i8[i] == b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);470
- }471
-#endif472
-473
- return simde__m64_from_private(r_);474
-#endif475
-}476
-#define simde_m_pcmpeqb(a, b) simde_mm_cmpeq_pi8(a, b)477
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)478
-#define _mm_cmpeq_pi8(a, b) simde_mm_cmpeq_pi8(a, b)479
-#define _m_pcmpeqb(a, b) simde_mm_cmpeq_pi8(a, b)480
-#endif481
-482
-SIMDE_FUNCTION_ATTRIBUTES483
-simde__m64 simde_mm_cmpeq_pi16(simde__m64 a, simde__m64 b)484
-{485
-#if defined(SIMDE_X86_MMX_NATIVE)486
- return _mm_cmpeq_pi16(a, b);487
-#else488
- simde__m64_private r_;489
- simde__m64_private a_ = simde__m64_to_private(a);490
- simde__m64_private b_ = simde__m64_to_private(b);491
-492
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)493
- r_.neon_i16 = vreinterpret_s16_u16(vceq_s16(a_.neon_i16, b_.neon_i16));494
-#else495
- SIMDE_VECTORIZE496
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {497
- r_.i16[i] = (a_.i16[i] == b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);498
- }499
-#endif500
-501
- return simde__m64_from_private(r_);502
-#endif503
-}504
-#define simde_m_pcmpeqw(a, b) simde_mm_cmpeq_pi16(a, b)505
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)506
-#define _mm_cmpeq_pi16(a, b) simde_mm_cmpeq_pi16(a, b)507
-#define _m_pcmpeqw(a, b) simde_mm_cmpeq_pi16(a, b)508
-#endif509
-510
-SIMDE_FUNCTION_ATTRIBUTES511
-simde__m64 simde_mm_cmpeq_pi32(simde__m64 a, simde__m64 b)512
-{513
-#if defined(SIMDE_X86_MMX_NATIVE)514
- return _mm_cmpeq_pi32(a, b);515
-#else516
- simde__m64_private r_;517
- simde__m64_private a_ = simde__m64_to_private(a);518
- simde__m64_private b_ = simde__m64_to_private(b);519
-520
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)521
- r_.neon_i32 = vreinterpret_s32_u32(vceq_s32(a_.neon_i32, b_.neon_i32));522
-#else523
- SIMDE_VECTORIZE524
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {525
- r_.i32[i] = (a_.i32[i] == b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);526
- }527
-#endif528
-529
- return simde__m64_from_private(r_);530
-#endif531
-}532
-#define simde_m_pcmpeqd(a, b) simde_mm_cmpeq_pi32(a, b)533
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)534
-#define _mm_cmpeq_pi32(a, b) simde_mm_cmpeq_pi32(a, b)535
-#define _m_pcmpeqd(a, b) simde_mm_cmpeq_pi32(a, b)536
-#endif537
-538
-SIMDE_FUNCTION_ATTRIBUTES539
-simde__m64 simde_mm_cmpgt_pi8(simde__m64 a, simde__m64 b)540
-{541
-#if defined(SIMDE_X86_MMX_NATIVE)542
- return _mm_cmpgt_pi8(a, b);543
-#else544
- simde__m64_private r_;545
- simde__m64_private a_ = simde__m64_to_private(a);546
- simde__m64_private b_ = simde__m64_to_private(b);547
-548
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)549
- r_.neon_i8 = vreinterpret_s8_u8(vcgt_s8(a_.neon_i8, b_.neon_i8));550
-#else551
- SIMDE_VECTORIZE552
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {553
- r_.i8[i] = (a_.i8[i] > b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);554
- }555
-#endif556
-557
- return simde__m64_from_private(r_);558
-#endif559
-}560
-#define simde_m_pcmpgtb(a, b) simde_mm_cmpgt_pi8(a, b)561
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)562
-#define _mm_cmpgt_pi8(a, b) simde_mm_cmpgt_pi8(a, b)563
-#define _m_pcmpgtb(a, b) simde_mm_cmpgt_pi8(a, b)564
-#endif565
-566
-SIMDE_FUNCTION_ATTRIBUTES567
-simde__m64 simde_mm_cmpgt_pi16(simde__m64 a, simde__m64 b)568
-{569
-#if defined(SIMDE_X86_MMX_NATIVE)570
- return _mm_cmpgt_pi16(a, b);571
-#else572
- simde__m64_private r_;573
- simde__m64_private a_ = simde__m64_to_private(a);574
- simde__m64_private b_ = simde__m64_to_private(b);575
-576
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)577
- r_.neon_i16 = vreinterpret_s16_u16(vcgt_s16(a_.neon_i16, b_.neon_i16));578
-#else579
- SIMDE_VECTORIZE580
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {581
- r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);582
- }583
-#endif584
-585
- return simde__m64_from_private(r_);586
-#endif587
-}588
-#define simde_m_pcmpgtw(a, b) simde_mm_cmpgt_pi16(a, b)589
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)590
-#define _mm_cmpgt_pi16(a, b) simde_mm_cmpgt_pi16(a, b)591
-#define _m_pcmpgtw(a, b) simde_mm_cmpgt_pi16(a, b)592
-#endif593
-594
-SIMDE_FUNCTION_ATTRIBUTES595
-simde__m64 simde_mm_cmpgt_pi32(simde__m64 a, simde__m64 b)596
-{597
-#if defined(SIMDE_X86_MMX_NATIVE)598
- return _mm_cmpgt_pi32(a, b);599
-#else600
- simde__m64_private r_;601
- simde__m64_private a_ = simde__m64_to_private(a);602
- simde__m64_private b_ = simde__m64_to_private(b);603
-604
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)605
- r_.neon_i32 = vreinterpret_s32_u32(vcgt_s32(a_.neon_i32, b_.neon_i32));606
-#else607
- SIMDE_VECTORIZE608
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {609
- r_.i32[i] = (a_.i32[i] > b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);610
- }611
-#endif612
-613
- return simde__m64_from_private(r_);614
-#endif615
-}616
-#define simde_m_pcmpgtd(a, b) simde_mm_cmpgt_pi32(a, b)617
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)618
-#define _mm_cmpgt_pi32(a, b) simde_mm_cmpgt_pi32(a, b)619
-#define _m_pcmpgtd(a, b) simde_mm_cmpgt_pi32(a, b)620
-#endif621
-622
-SIMDE_FUNCTION_ATTRIBUTES623
-int64_t simde_mm_cvtm64_si64(simde__m64 a)624
-{625
-#if defined(SIMDE_X86_MMX_NATIVE) && defined(SIMDE_ARCH_AMD64) && \626
- !defined(__PGI)627
- return _mm_cvtm64_si64(a);628
-#else629
- simde__m64_private a_ = simde__m64_to_private(a);630
-631
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)632
- return vget_lane_s64(a_.neon_i64, 0);633
-#else634
- return a_.i64[0];635
-#endif636
-#endif637
-}638
-#define simde_m_to_int64(a) simde_mm_cvtm64_si64(a)639
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)640
-#define _mm_cvtm64_si64(a) simde_mm_cvtm64_si64(a)641
-#define _m_to_int64(a) simde_mm_cvtm64_si64(a)642
-#endif643
-644
-SIMDE_FUNCTION_ATTRIBUTES645
-simde__m64 simde_mm_cvtsi32_si64(int32_t a)646
-{647
-#if defined(SIMDE_X86_MMX_NATIVE)648
- return _mm_cvtsi32_si64(a);649
-#else650
- simde__m64_private r_;651
-652
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)653
- const int32_t av[sizeof(r_.neon_i32) / sizeof(r_.neon_i32[0])] = {a, 0};654
- r_.neon_i32 = vld1_s32(av);655
-#else656
- r_.i32[0] = a;657
- r_.i32[1] = 0;658
-#endif659
-660
- return simde__m64_from_private(r_);661
-#endif662
-}663
-#define simde_m_from_int(a) simde_mm_cvtsi32_si64(a)664
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)665
-#define _mm_cvtsi32_si64(a) simde_mm_cvtsi32_si64(a)666
-#define _m_from_int(a) simde_mm_cvtsi32_si64(a)667
-#endif668
-669
-SIMDE_FUNCTION_ATTRIBUTES670
-simde__m64 simde_mm_cvtsi64_m64(int64_t a)671
-{672
-#if defined(SIMDE_X86_MMX_NATIVE) && defined(SIMDE_ARCH_AMD64) && \673
- !defined(__PGI)674
- return _mm_cvtsi64_m64(a);675
-#else676
- simde__m64_private r_;677
-678
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)679
- r_.neon_i64 = vld1_s64(&a);680
-#else681
- r_.i64[0] = a;682
-#endif683
-684
- return simde__m64_from_private(r_);685
-#endif686
-}687
-#define simde_m_from_int64(a) simde_mm_cvtsi64_m64(a)688
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)689
-#define _mm_cvtsi64_m64(a) simde_mm_cvtsi64_m64(a)690
-#define _m_from_int64(a) simde_mm_cvtsi64_m64(a)691
-#endif692
-693
-SIMDE_FUNCTION_ATTRIBUTES694
-int32_t simde_mm_cvtsi64_si32(simde__m64 a)695
-{696
-#if defined(SIMDE_X86_MMX_NATIVE)697
- return _mm_cvtsi64_si32(a);698
-#else699
- simde__m64_private a_ = simde__m64_to_private(a);700
-701
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)702
- return vget_lane_s32(a_.neon_i32, 0);703
-#else704
- return a_.i32[0];705
-#endif706
-#endif707
-}708
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)709
-#define _mm_cvtsi64_si32(a) simde_mm_cvtsi64_si32(a)710
-#endif711
-712
-SIMDE_FUNCTION_ATTRIBUTES713
-void simde_mm_empty(void)714
-{715
-#if defined(SIMDE_X86_MMX_NATIVE)716
- _mm_empty();717
-#else718
-#endif719
-}720
-#define simde_m_empty() simde_mm_empty()721
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)722
-#define _mm_empty() simde_mm_empty()723
-#define _m_empty() simde_mm_empty()724
-#endif725
-726
-SIMDE_FUNCTION_ATTRIBUTES727
-simde__m64 simde_mm_madd_pi16(simde__m64 a, simde__m64 b)728
-{729
-#if defined(SIMDE_X86_MMX_NATIVE)730
- return _mm_madd_pi16(a, b);731
-#else732
- simde__m64_private r_;733
- simde__m64_private a_ = simde__m64_to_private(a);734
- simde__m64_private b_ = simde__m64_to_private(b);735
-736
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)737
- int32x4_t i1 = vmull_s16(a_.neon_i16, b_.neon_i16);738
- r_.neon_i32 = vpadd_s32(vget_low_s32(i1), vget_high_s32(i1));739
-#else740
- SIMDE_VECTORIZE741
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i += 2) {742
- r_.i32[i / 2] = (a_.i16[i] * b_.i16[i]) +743
- (a_.i16[i + 1] * b_.i16[i + 1]);744
- }745
-#endif746
-747
- return simde__m64_from_private(r_);748
-#endif749
-}750
-#define simde_m_pmaddwd(a, b) simde_mm_madd_pi16(a, b)751
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)752
-#define _mm_madd_pi16(a, b) simde_mm_madd_pi16(a, b)753
-#define _m_pmaddwd(a, b) simde_mm_madd_pi16(a, b)754
-#endif755
-756
-SIMDE_FUNCTION_ATTRIBUTES757
-simde__m64 simde_mm_mulhi_pi16(simde__m64 a, simde__m64 b)758
-{759
-#if defined(SIMDE_X86_MMX_NATIVE)760
- return _mm_mulhi_pi16(a, b);761
-#else762
- simde__m64_private r_;763
- simde__m64_private a_ = simde__m64_to_private(a);764
- simde__m64_private b_ = simde__m64_to_private(b);765
-766
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)767
- const int32x4_t t1 = vmull_s16(a_.neon_i16, b_.neon_i16);768
- const uint32x4_t t2 = vshrq_n_u32(vreinterpretq_u32_s32(t1), 16);769
- const uint16x4_t t3 = vmovn_u32(t2);770
- r_.neon_i16 = vreinterpret_s16_u16(t3);771
-#else772
- SIMDE_VECTORIZE773
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {774
- r_.i16[i] = HEDLEY_STATIC_CAST(int16_t,775
- ((a_.i16[i] * b_.i16[i]) >> 16));776
- }777
-#endif778
-779
- return simde__m64_from_private(r_);780
-#endif781
-}782
-#define simde_m_pmulhw(a, b) simde_mm_mulhi_pi16(a, b)783
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)784
-#define _mm_mulhi_pi16(a, b) simde_mm_mulhi_pi16(a, b)785
-#define _m_pmulhw(a, b) simde_mm_mulhi_pi16(a, b)786
-#endif787
-788
-SIMDE_FUNCTION_ATTRIBUTES789
-simde__m64 simde_mm_mullo_pi16(simde__m64 a, simde__m64 b)790
-{791
-#if defined(SIMDE_X86_MMX_NATIVE)792
- return _mm_mullo_pi16(a, b);793
-#else794
- simde__m64_private r_;795
- simde__m64_private a_ = simde__m64_to_private(a);796
- simde__m64_private b_ = simde__m64_to_private(b);797
-798
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)799
- const int32x4_t t1 = vmull_s16(a_.neon_i16, b_.neon_i16);800
- const uint16x4_t t2 = vmovn_u32(vreinterpretq_u32_s32(t1));801
- r_.neon_i16 = vreinterpret_s16_u16(t2);802
-#else803
- SIMDE_VECTORIZE804
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {805
- r_.i16[i] = HEDLEY_STATIC_CAST(806
- int16_t, ((a_.i16[i] * b_.i16[i]) & 0xffff));807
- }808
-#endif809
-810
- return simde__m64_from_private(r_);811
-#endif812
-}813
-#define simde_m_pmullw(a, b) simde_mm_mullo_pi16(a, b)814
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)815
-#define _mm_mullo_pi16(a, b) simde_mm_mullo_pi16(a, b)816
-#define _m_pmullw(a, b) simde_mm_mullo_pi16(a, b)817
-#endif818
-819
-SIMDE_FUNCTION_ATTRIBUTES820
-simde__m64 simde_mm_or_si64(simde__m64 a, simde__m64 b)821
-{822
-#if defined(SIMDE_X86_MMX_NATIVE)823
- return _mm_or_si64(a, b);824
-#else825
- simde__m64_private r_;826
- simde__m64_private a_ = simde__m64_to_private(a);827
- simde__m64_private b_ = simde__m64_to_private(b);828
-829
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)830
- r_.neon_i32 = vorr_s32(a_.neon_i32, b_.neon_i32);831
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)832
- r_.i64 = a_.i64 | b_.i64;833
-#else834
- r_.i64[0] = a_.i64[0] | b_.i64[0];835
-#endif836
-837
- return simde__m64_from_private(r_);838
-#endif839
-}840
-#define simde_m_por(a, b) simde_mm_or_si64(a, b)841
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)842
-#define _mm_or_si64(a, b) simde_mm_or_si64(a, b)843
-#define _m_por(a, b) simde_mm_or_si64(a, b)844
-#endif845
-846
-SIMDE_FUNCTION_ATTRIBUTES847
-simde__m64 simde_mm_packs_pi16(simde__m64 a, simde__m64 b)848
-{849
-#if defined(SIMDE_X86_MMX_NATIVE)850
- return _mm_packs_pi16(a, b);851
-#else852
- simde__m64_private r_;853
- simde__m64_private a_ = simde__m64_to_private(a);854
- simde__m64_private b_ = simde__m64_to_private(b);855
-856
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)857
- r_.neon_i8 = vqmovn_s16(vcombine_s16(a_.neon_i16, b_.neon_i16));858
-#else859
- SIMDE_VECTORIZE860
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {861
- if (a_.i16[i] < INT8_MIN) {862
- r_.i8[i] = INT8_MIN;863
- } else if (a_.i16[i] > INT8_MAX) {864
- r_.i8[i] = INT8_MAX;865
- } else {866
- r_.i8[i] = HEDLEY_STATIC_CAST(int8_t, a_.i16[i]);867
- }868
- }869
-870
- SIMDE_VECTORIZE871
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {872
- if (b_.i16[i] < INT8_MIN) {873
- r_.i8[i + 4] = INT8_MIN;874
- } else if (b_.i16[i] > INT8_MAX) {875
- r_.i8[i + 4] = INT8_MAX;876
- } else {877
- r_.i8[i + 4] = HEDLEY_STATIC_CAST(int8_t, b_.i16[i]);878
- }879
- }880
-#endif881
-882
- return simde__m64_from_private(r_);883
-#endif884
-}885
-#define simde_m_packsswb(a, b) simde_mm_packs_pi16(a, b)886
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)887
-#define _mm_packs_pi16(a, b) simde_mm_packs_pi16(a, b)888
-#define _m_packsswb(a, b) mm_packs_pi16(a, b)889
-#endif890
-891
-SIMDE_FUNCTION_ATTRIBUTES892
-simde__m64 simde_mm_packs_pi32(simde__m64 a, simde__m64 b)893
-{894
-#if defined(SIMDE_X86_MMX_NATIVE)895
- return _mm_packs_pi32(a, b);896
-#else897
- simde__m64_private r_;898
- simde__m64_private a_ = simde__m64_to_private(a);899
- simde__m64_private b_ = simde__m64_to_private(b);900
-901
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)902
- r_.neon_i16 = vqmovn_s32(vcombine_s32(a_.neon_i32, b_.neon_i32));903
-#else904
- SIMDE_VECTORIZE905
- for (size_t i = 0; i < (8 / sizeof(a_.i32[0])); i++) {906
- if (a_.i32[i] < SHRT_MIN) {907
- r_.i16[i] = SHRT_MIN;908
- } else if (a_.i32[i] > INT16_MAX) {909
- r_.i16[i] = INT16_MAX;910
- } else {911
- r_.i16[i] = HEDLEY_STATIC_CAST(int16_t, a_.i32[i]);912
- }913
- }914
-915
- SIMDE_VECTORIZE916
- for (size_t i = 0; i < (8 / sizeof(b_.i32[0])); i++) {917
- if (b_.i32[i] < SHRT_MIN) {918
- r_.i16[i + 2] = SHRT_MIN;919
- } else if (b_.i32[i] > INT16_MAX) {920
- r_.i16[i + 2] = INT16_MAX;921
- } else {922
- r_.i16[i + 2] = HEDLEY_STATIC_CAST(int16_t, b_.i32[i]);923
- }924
- }925
-#endif926
-927
- return simde__m64_from_private(r_);928
-#endif929
-}930
-#define simde_m_packssdw(a, b) simde_mm_packs_pi32(a, b)931
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)932
-#define _mm_packs_pi32(a, b) simde_mm_packs_pi32(a, b)933
-#define _m_packssdw(a, b) simde_mm_packs_pi32(a, b)934
-#endif935
-936
-SIMDE_FUNCTION_ATTRIBUTES937
-simde__m64 simde_mm_packs_pu16(simde__m64 a, simde__m64 b)938
-{939
-#if defined(SIMDE_X86_MMX_NATIVE)940
- return _mm_packs_pu16(a, b);941
-#else942
- simde__m64_private r_;943
- simde__m64_private a_ = simde__m64_to_private(a);944
- simde__m64_private b_ = simde__m64_to_private(b);945
-946
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)947
- const int16x8_t t1 = vcombine_s16(a_.neon_i16, b_.neon_i16);948
-949
- /* Set elements which are < 0 to 0 */950
- const int16x8_t t2 =951
- vandq_s16(t1, vreinterpretq_s16_u16(vcgezq_s16(t1)));952
-953
- /* Vector with all s16 elements set to UINT8_MAX */954
- const int16x8_t vmax = vmovq_n_s16((int16_t)UINT8_MAX);955
-956
- /* Elements which are within the acceptable range */957
- const int16x8_t le_max =958
- vandq_s16(t2, vreinterpretq_s16_u16(vcleq_s16(t2, vmax)));959
- const int16x8_t gt_max =960
- vandq_s16(vmax, vreinterpretq_s16_u16(vcgtq_s16(t2, vmax)));961
-962
- /* Final values as 16-bit integers */963
- const int16x8_t values = vorrq_s16(le_max, gt_max);964
-965
- r_.neon_u8 = vmovn_u16(vreinterpretq_u16_s16(values));966
-#else967
- SIMDE_VECTORIZE968
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {969
- if (a_.i16[i] > UINT8_MAX) {970
- r_.u8[i] = UINT8_MAX;971
- } else if (a_.i16[i] < 0) {972
- r_.u8[i] = 0;973
- } else {974
- r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, a_.i16[i]);975
- }976
- }977
-978
- SIMDE_VECTORIZE979
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {980
- if (b_.i16[i] > UINT8_MAX) {981
- r_.u8[i + 4] = UINT8_MAX;982
- } else if (b_.i16[i] < 0) {983
- r_.u8[i + 4] = 0;984
- } else {985
- r_.u8[i + 4] = HEDLEY_STATIC_CAST(uint8_t, b_.i16[i]);986
- }987
- }988
-#endif989
-990
- return simde__m64_from_private(r_);991
-#endif992
-}993
-#define simde_m_packuswb(a, b) simde_mm_packs_pu16(a, b)994
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)995
-#define _mm_packs_pu16(a, b) simde_mm_packs_pu16(a, b)996
-#define _m_packuswb(a, b) simde_mm_packs_pu16(a, b)997
-#endif998
-999
-SIMDE_FUNCTION_ATTRIBUTES1000
-simde__m64 simde_mm_set_pi8(int8_t e7, int8_t e6, int8_t e5, int8_t e4,1001
- int8_t e3, int8_t e2, int8_t e1, int8_t e0)1002
-{1003
-#if defined(SIMDE_X86_MMX_NATIVE)1004
- return _mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0);1005
-#else1006
- simde__m64_private r_;1007
-1008
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1009
- const int8_t v[sizeof(r_.i8) / sizeof(r_.i8[0])] = {e0, e1, e2, e3,1010
- e4, e5, e6, e7};1011
- r_.neon_i8 = vld1_s8(v);1012
-#else1013
- r_.i8[0] = e0;1014
- r_.i8[1] = e1;1015
- r_.i8[2] = e2;1016
- r_.i8[3] = e3;1017
- r_.i8[4] = e4;1018
- r_.i8[5] = e5;1019
- r_.i8[6] = e6;1020
- r_.i8[7] = e7;1021
-#endif1022
-1023
- return simde__m64_from_private(r_);1024
-#endif1025
-}1026
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1027
-#define _mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0) \1028
- simde_mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0)1029
-#endif1030
-1031
-SIMDE_FUNCTION_ATTRIBUTES1032
-simde__m64 simde_x_mm_set_pu8(uint8_t e7, uint8_t e6, uint8_t e5, uint8_t e4,1033
- uint8_t e3, uint8_t e2, uint8_t e1, uint8_t e0)1034
-{1035
- simde__m64_private r_;1036
-1037
-#if defined(SIMDE_X86_MMX_NATIVE)1038
- r_.n = _mm_set_pi8(1039
- HEDLEY_STATIC_CAST(int8_t, e7), HEDLEY_STATIC_CAST(int8_t, e6),1040
- HEDLEY_STATIC_CAST(int8_t, e5), HEDLEY_STATIC_CAST(int8_t, e4),1041
- HEDLEY_STATIC_CAST(int8_t, e3), HEDLEY_STATIC_CAST(int8_t, e2),1042
- HEDLEY_STATIC_CAST(int8_t, e1), HEDLEY_STATIC_CAST(int8_t, e0));1043
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1044
- const uint8_t v[sizeof(r_.u8) / sizeof(r_.u8[0])] = {e0, e1, e2, e3,1045
- e4, e5, e6, e7};1046
- r_.neon_u8 = vld1_u8(v);1047
-#else1048
- r_.u8[0] = e0;1049
- r_.u8[1] = e1;1050
- r_.u8[2] = e2;1051
- r_.u8[3] = e3;1052
- r_.u8[4] = e4;1053
- r_.u8[5] = e5;1054
- r_.u8[6] = e6;1055
- r_.u8[7] = e7;1056
-#endif1057
-1058
- return simde__m64_from_private(r_);1059
-}1060
-1061
-SIMDE_FUNCTION_ATTRIBUTES1062
-simde__m64 simde_mm_set_pi16(int16_t e3, int16_t e2, int16_t e1, int16_t e0)1063
-{1064
-#if defined(SIMDE_X86_MMX_NATIVE)1065
- return _mm_set_pi16(e3, e2, e1, e0);1066
-#else1067
- simde__m64_private r_;1068
-1069
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1070
- const int16_t v[sizeof(r_.i16) / sizeof(r_.i16[0])] = {e0, e1, e2, e3};1071
- r_.neon_i16 = vld1_s16(v);1072
-#else1073
- r_.i16[0] = e0;1074
- r_.i16[1] = e1;1075
- r_.i16[2] = e2;1076
- r_.i16[3] = e3;1077
-#endif1078
- return simde__m64_from_private(r_);1079
-#endif1080
-}1081
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1082
-#define _mm_set_pi16(e3, e2, e1, e0) simde_mm_set_pi16(e3, e2, e1, e0)1083
-#endif1084
-1085
-SIMDE_FUNCTION_ATTRIBUTES1086
-simde__m64 simde_x_mm_set_pu16(uint16_t e3, uint16_t e2, uint16_t e1,1087
- uint16_t e0)1088
-{1089
- simde__m64_private r_;1090
-1091
-#if defined(SIMDE_X86_MMX_NATIVE)1092
- r_.n = _mm_set_pi16(HEDLEY_STATIC_CAST(int16_t, e3),1093
- HEDLEY_STATIC_CAST(int16_t, e2),1094
- HEDLEY_STATIC_CAST(int16_t, e1),1095
- HEDLEY_STATIC_CAST(int16_t, e0));1096
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1097
- const uint16_t v[sizeof(r_.u16) / sizeof(r_.u16[0])] = {e0, e1, e2, e3};1098
- r_.neon_u16 = vld1_u16(v);1099
-#else1100
- r_.u16[0] = e0;1101
- r_.u16[1] = e1;1102
- r_.u16[2] = e2;1103
- r_.u16[3] = e3;1104
-#endif1105
-1106
- return simde__m64_from_private(r_);1107
-}1108
-1109
-SIMDE_FUNCTION_ATTRIBUTES1110
-simde__m64 simde_x_mm_set_pu32(uint32_t e1, uint32_t e0)1111
-{1112
- simde__m64_private r_;1113
-1114
-#if defined(SIMDE_X86_MMX_NATIVE)1115
- r_.n = _mm_set_pi32(HEDLEY_STATIC_CAST(int32_t, e1),1116
- HEDLEY_STATIC_CAST(int32_t, e0));1117
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1118
- const uint32_t v[sizeof(r_.u32) / sizeof(r_.u32[0])] = {e0, e1};1119
- r_.neon_u32 = vld1_u32(v);1120
-#else1121
- r_.u32[0] = e0;1122
- r_.u32[1] = e1;1123
-#endif1124
-1125
- return simde__m64_from_private(r_);1126
-}1127
-1128
-SIMDE_FUNCTION_ATTRIBUTES1129
-simde__m64 simde_mm_set_pi32(int32_t e1, int32_t e0)1130
-{1131
- simde__m64_private r_;1132
-1133
-#if defined(SIMDE_X86_MMX_NATIVE)1134
- r_.n = _mm_set_pi32(e1, e0);1135
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1136
- const int32_t v[sizeof(r_.i32) / sizeof(r_.i32[0])] = {e0, e1};1137
- r_.neon_i32 = vld1_s32(v);1138
-#else1139
- r_.i32[0] = e0;1140
- r_.i32[1] = e1;1141
-#endif1142
-1143
- return simde__m64_from_private(r_);1144
-}1145
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1146
-#define _mm_set_pi32(e1, e0) simde_mm_set_pi32(e1, e0)1147
-#endif1148
-1149
-SIMDE_FUNCTION_ATTRIBUTES1150
-simde__m64 simde_x_mm_set_pi64(int64_t e0)1151
-{1152
- simde__m64_private r_;1153
-1154
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1155
- const int64_t v[sizeof(r_.i64) / sizeof(r_.i64[0])] = {e0};1156
- r_.neon_i64 = vld1_s64(v);1157
-#else1158
- r_.i64[0] = e0;1159
-#endif1160
-1161
- return simde__m64_from_private(r_);1162
-}1163
-1164
-SIMDE_FUNCTION_ATTRIBUTES1165
-simde__m64 simde_x_mm_set_f32x2(simde_float32 e1, simde_float32 e0)1166
-{1167
- simde__m64_private r_;1168
-1169
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1170
- const simde_float32 v[sizeof(r_.f32) / sizeof(r_.f32[0])] = {e0, e1};1171
- r_.neon_f32 = vld1_f32(v);1172
-#else1173
- r_.f32[0] = e0;1174
- r_.f32[1] = e1;1175
-#endif1176
-1177
- return simde__m64_from_private(r_);1178
-}1179
-1180
-SIMDE_FUNCTION_ATTRIBUTES1181
-simde__m64 simde_mm_set1_pi8(int8_t a)1182
-{1183
-#if defined(SIMDE_X86_MMX_NATIVE)1184
- return _mm_set1_pi8(a);1185
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1186
- simde__m64_private r_;1187
- r_.neon_i8 = vmov_n_s8(a);1188
- return simde__m64_from_private(r_);1189
-#else1190
- return simde_mm_set_pi8(a, a, a, a, a, a, a, a);1191
-#endif1192
-}1193
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1194
-#define _mm_set1_pi8(a) simde_mm_set1_pi8(a)1195
-#endif1196
-1197
-SIMDE_FUNCTION_ATTRIBUTES1198
-simde__m64 simde_mm_set1_pi16(int16_t a)1199
-{1200
-#if defined(SIMDE_X86_MMX_NATIVE)1201
- return _mm_set1_pi16(a);1202
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1203
- simde__m64_private r_;1204
- r_.neon_i16 = vmov_n_s16(a);1205
- return simde__m64_from_private(r_);1206
-#else1207
- return simde_mm_set_pi16(a, a, a, a);1208
-#endif1209
-}1210
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1211
-#define _mm_set1_pi16(a) simde_mm_set1_pi16(a)1212
-#endif1213
-1214
-SIMDE_FUNCTION_ATTRIBUTES1215
-simde__m64 simde_mm_set1_pi32(int32_t a)1216
-{1217
-#if defined(SIMDE_X86_MMX_NATIVE)1218
- return _mm_set1_pi32(a);1219
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1220
- simde__m64_private r_;1221
- r_.neon_i32 = vmov_n_s32(a);1222
- return simde__m64_from_private(r_);1223
-#else1224
- return simde_mm_set_pi32(a, a);1225
-#endif1226
-}1227
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1228
-#define _mm_set1_pi32(a) simde_mm_set1_pi32(a)1229
-#endif1230
-1231
-SIMDE_FUNCTION_ATTRIBUTES1232
-simde__m64 simde_mm_setr_pi8(int8_t e7, int8_t e6, int8_t e5, int8_t e4,1233
- int8_t e3, int8_t e2, int8_t e1, int8_t e0)1234
-{1235
-#if defined(SIMDE_X86_MMX_NATIVE)1236
- return _mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0);1237
-#else1238
- return simde_mm_set_pi8(e0, e1, e2, e3, e4, e5, e6, e7);1239
-#endif1240
-}1241
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1242
-#define _mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0) \1243
- simde_mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0)1244
-#endif1245
-1246
-SIMDE_FUNCTION_ATTRIBUTES1247
-simde__m64 simde_mm_setr_pi16(int16_t e3, int16_t e2, int16_t e1, int16_t e0)1248
-{1249
-#if defined(SIMDE_X86_MMX_NATIVE)1250
- return _mm_setr_pi16(e3, e2, e1, e0);1251
-#else1252
- return simde_mm_set_pi16(e0, e1, e2, e3);1253
-#endif1254
-}1255
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1256
-#define _mm_setr_pi16(e3, e2, e1, e0) simde_mm_setr_pi16(e3, e2, e1, e0)1257
-#endif1258
-1259
-SIMDE_FUNCTION_ATTRIBUTES1260
-simde__m64 simde_mm_setr_pi32(int32_t e1, int32_t e0)1261
-{1262
-#if defined(SIMDE_X86_MMX_NATIVE)1263
- return _mm_setr_pi32(e1, e0);1264
-#else1265
- return simde_mm_set_pi32(e0, e1);1266
-#endif1267
-}1268
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1269
-#define _mm_setr_pi32(e1, e0) simde_mm_setr_pi32(e1, e0)1270
-#endif1271
-1272
-SIMDE_FUNCTION_ATTRIBUTES1273
-simde__m64 simde_mm_setzero_si64(void)1274
-{1275
-#if defined(SIMDE_X86_MMX_NATIVE)1276
- return _mm_setzero_si64();1277
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1278
- simde__m64_private r_;1279
- r_.neon_u32 = vmov_n_u32(0);1280
- return simde__m64_from_private(r_);1281
-#else1282
- return simde_mm_set_pi32(0, 0);1283
-#endif1284
-}1285
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1286
-#define _mm_setzero_si64() simde_mm_setzero_si64()1287
-#endif1288
-1289
-SIMDE_FUNCTION_ATTRIBUTES1290
-simde__m64 simde_x_mm_setone_si64(void)1291
-{1292
- return simde_mm_set1_pi32(~INT32_C(0));1293
-}1294
-1295
-SIMDE_FUNCTION_ATTRIBUTES1296
-simde__m64 simde_mm_sll_pi16(simde__m64 a, simde__m64 count)1297
-{1298
-#if defined(SIMDE_X86_MMX_NATIVE)1299
- return _mm_sll_pi16(a, count);1300
-#else1301
- simde__m64_private r_;1302
- simde__m64_private a_ = simde__m64_to_private(a);1303
- simde__m64_private count_ = simde__m64_to_private(count);1304
-1305
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1306
- r_.neon_i16 = vshl_s16(a_.neon_i16, vmov_n_s16((int16_t)vget_lane_u64(1307
- count_.neon_u64, 0)));1308
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1309
- r_.i16 = a_.i16 << count_.u64[0];1310
-#else1311
- if (HEDLEY_UNLIKELY(count_.u64[0] > 15)) {1312
- simde_memset(&r_, 0, sizeof(r_));1313
- return simde__m64_from_private(r_);1314
- }1315
-1316
- SIMDE_VECTORIZE1317
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {1318
- r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t,1319
- a_.u16[i] << count_.u64[0]);1320
- }1321
-#endif1322
-1323
- return simde__m64_from_private(r_);1324
-#endif1325
-}1326
-#define simde_m_psllw(a, count) simde_mm_sll_pi16(a, count)1327
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1328
-#define _mm_sll_pi16(a, count) simde_mm_sll_pi16(a, count)1329
-#define _m_psllw(a, count) simde_mm_sll_pi16(a, count)1330
-#endif1331
-1332
-SIMDE_FUNCTION_ATTRIBUTES1333
-simde__m64 simde_mm_sll_pi32(simde__m64 a, simde__m64 count)1334
-{1335
-#if defined(SIMDE_X86_MMX_NATIVE)1336
- return _mm_sll_pi32(a, count);1337
-#else1338
- simde__m64_private r_;1339
- simde__m64_private a_ = simde__m64_to_private(a);1340
- simde__m64_private count_ = simde__m64_to_private(count);1341
-1342
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1343
- r_.neon_i32 = vshl_s32(a_.neon_i32, vmov_n_s32((int32_t)vget_lane_u64(1344
- count_.neon_u64, 0)));1345
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1346
- r_.i32 = a_.i32 << count_.u64[0];1347
-#else1348
- if (HEDLEY_UNLIKELY(count_.u64[0] > 31)) {1349
- simde_memset(&r_, 0, sizeof(r_));1350
- return simde__m64_from_private(r_);1351
- }1352
-1353
- SIMDE_VECTORIZE1354
- for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {1355
- r_.u32[i] = a_.u32[i] << count_.u64[0];1356
- }1357
-#endif1358
-1359
- return simde__m64_from_private(r_);1360
-#endif1361
-}1362
-#define simde_m_pslld(a, count) simde_mm_sll_pi32(a, count)1363
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1364
-#define _mm_sll_pi32(a, count) simde_mm_sll_pi32(a, count)1365
-#define _m_pslld(a, count) simde_mm_sll_pi32(a, count)1366
-#endif1367
-1368
-SIMDE_FUNCTION_ATTRIBUTES1369
-simde__m64 simde_mm_slli_pi16(simde__m64 a, int count)1370
-{1371
-#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1372
- return _mm_slli_pi16(a, count);1373
-#else1374
- simde__m64_private r_;1375
- simde__m64_private a_ = simde__m64_to_private(a);1376
-1377
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1378
- r_.i16 = a_.i16 << count;1379
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1380
- r_.neon_i16 = vshl_s16(a_.neon_i16, vmov_n_s16((int16_t)count));1381
-#else1382
- SIMDE_VECTORIZE1383
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {1384
- r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, a_.u16[i] << count);1385
- }1386
-#endif1387
-1388
- return simde__m64_from_private(r_);1389
-#endif1390
-}1391
-#define simde_m_psllwi(a, count) simde_mm_slli_pi16(a, count)1392
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1393
-#define _mm_slli_pi16(a, count) simde_mm_slli_pi16(a, count)1394
-#define _m_psllwi(a, count) simde_mm_slli_pi16(a, count)1395
-#endif1396
-1397
-SIMDE_FUNCTION_ATTRIBUTES1398
-simde__m64 simde_mm_slli_pi32(simde__m64 a, int count)1399
-{1400
-#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1401
- return _mm_slli_pi32(a, count);1402
-#else1403
- simde__m64_private r_;1404
- simde__m64_private a_ = simde__m64_to_private(a);1405
-1406
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1407
- r_.i32 = a_.i32 << count;1408
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1409
- r_.neon_i32 = vshl_s32(a_.neon_i32, vmov_n_s32((int32_t)count));1410
-#else1411
- SIMDE_VECTORIZE1412
- for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {1413
- r_.u32[i] = a_.u32[i] << count;1414
- }1415
-#endif1416
-1417
- return simde__m64_from_private(r_);1418
-#endif1419
-}1420
-#define simde_m_pslldi(a, b) simde_mm_slli_pi32(a, b)1421
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1422
-#define _mm_slli_pi32(a, count) simde_mm_slli_pi32(a, count)1423
-#define _m_pslldi(a, count) simde_mm_slli_pi32(a, count)1424
-#endif1425
-1426
-SIMDE_FUNCTION_ATTRIBUTES1427
-simde__m64 simde_mm_slli_si64(simde__m64 a, int count)1428
-{1429
-#if defined(SIMDE_X86_MMX_NATIVE)1430
- return _mm_slli_si64(a, count);1431
-#else1432
- simde__m64_private r_;1433
- simde__m64_private a_ = simde__m64_to_private(a);1434
-1435
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1436
- r_.i64 = a_.i64 << count;1437
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1438
- r_.neon_i64 = vshl_s64(a_.neon_i64, vmov_n_s64((int64_t)count));1439
-#else1440
- r_.u64[0] = a_.u64[0] << count;1441
-#endif1442
-1443
- return simde__m64_from_private(r_);1444
-#endif1445
-}1446
-#define simde_m_psllqi(a, count) simde_mm_slli_si64(a, count)1447
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1448
-#define _mm_slli_si64(a, count) simde_mm_slli_si64(a, count)1449
-#define _m_psllqi(a, count) simde_mm_slli_si64(a, count)1450
-#endif1451
-1452
-SIMDE_FUNCTION_ATTRIBUTES1453
-simde__m64 simde_mm_sll_si64(simde__m64 a, simde__m64 count)1454
-{1455
-#if defined(SIMDE_X86_MMX_NATIVE)1456
- return _mm_sll_si64(a, count);1457
-#else1458
- simde__m64_private r_;1459
- simde__m64_private a_ = simde__m64_to_private(a);1460
- simde__m64_private count_ = simde__m64_to_private(count);1461
-1462
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1463
- r_.neon_i64 = vshl_s64(a_.neon_i64, count_.neon_i64);1464
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1465
- r_.i64 = a_.i64 << count_.i64;1466
-#else1467
- if (HEDLEY_UNLIKELY(count_.u64[0] > 63)) {1468
- simde_memset(&r_, 0, sizeof(r_));1469
- return simde__m64_from_private(r_);1470
- }1471
-1472
- r_.u64[0] = a_.u64[0] << count_.u64[0];1473
-#endif1474
-1475
- return simde__m64_from_private(r_);1476
-#endif1477
-}1478
-#define simde_m_psllq(a, count) simde_mm_sll_si64(a, count)1479
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1480
-#define _mm_sll_si64(a, count) simde_mm_sll_si64(a, count)1481
-#define _m_psllq(a, count) simde_mm_sll_si64(a, count)1482
-#endif1483
-1484
-SIMDE_FUNCTION_ATTRIBUTES1485
-simde__m64 simde_mm_srl_pi16(simde__m64 a, simde__m64 count)1486
-{1487
-#if defined(SIMDE_X86_MMX_NATIVE)1488
- return _mm_srl_pi16(a, count);1489
-#else1490
- simde__m64_private r_;1491
- simde__m64_private a_ = simde__m64_to_private(a);1492
- simde__m64_private count_ = simde__m64_to_private(count);1493
-1494
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1495
- r_.u16 = a_.u16 >> count_.u64[0];1496
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1497
- r_.neon_u16 = vshl_u16(1498
- a_.neon_u16,1499
- vmov_n_s16(-((int16_t)vget_lane_u64(count_.neon_u64, 0))));1500
-#else1501
- if (HEDLEY_UNLIKELY(count_.u64[0] > 15)) {1502
- simde_memset(&r_, 0, sizeof(r_));1503
- return simde__m64_from_private(r_);1504
- }1505
-1506
- SIMDE_VECTORIZE1507
- for (size_t i = 0; i < sizeof(r_.u16) / sizeof(r_.u16[0]); i++) {1508
- r_.u16[i] = a_.u16[i] >> count_.u64[0];1509
- }1510
-#endif1511
-1512
- return simde__m64_from_private(r_);1513
-#endif1514
-}1515
-#define simde_m_psrlw(a, count) simde_mm_srl_pi16(a, count)1516
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1517
-#define _mm_srl_pi16(a, count) simde_mm_srl_pi16(a, count)1518
-#define _m_psrlw(a, count) simde_mm_srl_pi16(a, count)1519
-#endif1520
-1521
-SIMDE_FUNCTION_ATTRIBUTES1522
-simde__m64 simde_mm_srl_pi32(simde__m64 a, simde__m64 count)1523
-{1524
-#if defined(SIMDE_X86_MMX_NATIVE)1525
- return _mm_srl_pi32(a, count);1526
-#else1527
- simde__m64_private r_;1528
- simde__m64_private a_ = simde__m64_to_private(a);1529
- simde__m64_private count_ = simde__m64_to_private(count);1530
-1531
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1532
- r_.u32 = a_.u32 >> count_.u64[0];1533
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1534
- r_.neon_u32 = vshl_u32(1535
- a_.neon_u32,1536
- vmov_n_s32(-((int32_t)vget_lane_u64(count_.neon_u64, 0))));1537
-#else1538
- if (HEDLEY_UNLIKELY(count_.u64[0] > 31)) {1539
- simde_memset(&r_, 0, sizeof(r_));1540
- return simde__m64_from_private(r_);1541
- }1542
-1543
- SIMDE_VECTORIZE1544
- for (size_t i = 0; i < sizeof(r_.u32) / sizeof(r_.u32[0]); i++) {1545
- r_.u32[i] = a_.u32[i] >> count_.u64[0];1546
- }1547
-#endif1548
-1549
- return simde__m64_from_private(r_);1550
-#endif1551
-}1552
-#define simde_m_psrld(a, count) simde_mm_srl_pi32(a, count)1553
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1554
-#define _mm_srl_pi32(a, count) simde_mm_srl_pi32(a, count)1555
-#define _m_psrld(a, count) simde_mm_srl_pi32(a, count)1556
-#endif1557
-1558
-SIMDE_FUNCTION_ATTRIBUTES1559
-simde__m64 simde_mm_srli_pi16(simde__m64 a, int count)1560
-{1561
-#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1562
- return _mm_srli_pi16(a, count);1563
-#else1564
- simde__m64_private r_;1565
- simde__m64_private a_ = simde__m64_to_private(a);1566
-1567
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1568
- r_.u16 = a_.u16 >> count;1569
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1570
- r_.neon_u16 = vshl_u16(a_.neon_u16, vmov_n_s16(-((int16_t)count)));1571
-#else1572
- SIMDE_VECTORIZE1573
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {1574
- r_.u16[i] = a_.u16[i] >> count;1575
- }1576
-#endif1577
-1578
- return simde__m64_from_private(r_);1579
-#endif1580
-}1581
-#define simde_m_psrlwi(a, count) simde_mm_srli_pi16(a, count)1582
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1583
-#define _mm_srli_pi16(a, count) simde_mm_srli_pi16(a, count)1584
-#define _m_psrlwi(a, count) simde_mm_srli_pi16(a, count)1585
-#endif1586
-1587
-SIMDE_FUNCTION_ATTRIBUTES1588
-simde__m64 simde_mm_srli_pi32(simde__m64 a, int count)1589
-{1590
-#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1591
- return _mm_srli_pi32(a, count);1592
-#else1593
- simde__m64_private r_;1594
- simde__m64_private a_ = simde__m64_to_private(a);1595
-1596
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1597
- r_.u32 = a_.u32 >> count;1598
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1599
- r_.neon_u32 = vshl_u32(a_.neon_u32, vmov_n_s32(-((int32_t)count)));1600
-#else1601
- SIMDE_VECTORIZE1602
- for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {1603
- r_.u32[i] = a_.u32[i] >> count;1604
- }1605
-#endif1606
-1607
- return simde__m64_from_private(r_);1608
-#endif1609
-}1610
-#define simde_m_psrldi(a, count) simde_mm_srli_pi32(a, count)1611
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1612
-#define _mm_srli_pi32(a, count) simde_mm_srli_pi32(a, count)1613
-#define _m_psrldi(a, count) simde_mm_srli_pi32(a, count)1614
-#endif1615
-1616
-SIMDE_FUNCTION_ATTRIBUTES1617
-simde__m64 simde_mm_srli_si64(simde__m64 a, int count)1618
-{1619
-#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1620
- return _mm_srli_si64(a, count);1621
-#else1622
- simde__m64_private r_;1623
- simde__m64_private a_ = simde__m64_to_private(a);1624
-1625
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1626
- r_.neon_u64 = vshl_u64(a_.neon_u64, vmov_n_s64(-count));1627
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1628
- r_.u64 = a_.u64 >> count;1629
-#else1630
- r_.u64[0] = a_.u64[0] >> count;1631
-#endif1632
-1633
- return simde__m64_from_private(r_);1634
-#endif1635
-}1636
-#define simde_m_psrlqi(a, count) simde_mm_srli_si64(a, count)1637
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1638
-#define _mm_srli_si64(a, count) simde_mm_srli_si64(a, count)1639
-#define _m_psrlqi(a, count) simde_mm_srli_si64(a, count)1640
-#endif1641
-1642
-SIMDE_FUNCTION_ATTRIBUTES1643
-simde__m64 simde_mm_srl_si64(simde__m64 a, simde__m64 count)1644
-{1645
-#if defined(SIMDE_X86_MMX_NATIVE)1646
- return _mm_srl_si64(a, count);1647
-#else1648
- simde__m64_private r_;1649
- simde__m64_private a_ = simde__m64_to_private(a);1650
- simde__m64_private count_ = simde__m64_to_private(count);1651
-1652
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1653
- r_.neon_u64 = vshl_u64(a_.neon_u64, vneg_s64(count_.neon_i64));1654
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1655
- r_.u64 = a_.u64 >> count_.u64;1656
-#else1657
- if (HEDLEY_UNLIKELY(count_.u64[0] > 63)) {1658
- simde_memset(&r_, 0, sizeof(r_));1659
- return simde__m64_from_private(r_);1660
- }1661
-1662
- r_.u64[0] = a_.u64[0] >> count_.u64[0];1663
-#endif1664
-1665
- return simde__m64_from_private(r_);1666
-#endif1667
-}1668
-#define simde_m_psrlq(a, count) simde_mm_srl_si64(a, count)1669
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1670
-#define _mm_srl_si64(a, count) simde_mm_srl_si64(a, count)1671
-#define _m_psrlq(a, count) simde_mm_srl_si64(a, count)1672
-#endif1673
-1674
-SIMDE_FUNCTION_ATTRIBUTES1675
-simde__m64 simde_mm_srai_pi16(simde__m64 a, int count)1676
-{1677
-#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1678
- return _mm_srai_pi16(a, count);1679
-#else1680
- simde__m64_private r_;1681
- simde__m64_private a_ = simde__m64_to_private(a);1682
-1683
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1684
- r_.i16 = a_.i16 >> (count & 0xff);1685
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1686
- r_.neon_i16 = vshl_s16(a_.neon_i16, vmov_n_s16(-HEDLEY_STATIC_CAST(int16_t, count));1687
-#else1688
- SIMDE_VECTORIZE1689
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1690
- r_.i16[i] = a_.i16[i] >> (count & 0xff);1691
- }1692
-#endif1693
-1694
- return simde__m64_from_private(r_);1695
-#endif1696
-}1697
-#define simde_m_psrawi(a, count) simde_mm_srai_pi16(a, count)1698
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1699
-#define _mm_srai_pi16(a, count) simde_mm_srai_pi16(a, count)1700
-#define _m_psrawi(a, count) simde_mm_srai_pi16(a, count)1701
-#endif1702
-1703
-SIMDE_FUNCTION_ATTRIBUTES1704
-simde__m64 simde_mm_srai_pi32(simde__m64 a, int count)1705
-{1706
-#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1707
- return _mm_srai_pi32(a, count);1708
-#else1709
- simde__m64_private r_;1710
- simde__m64_private a_ = simde__m64_to_private(a);1711
-1712
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1713
- r_.i32 = a_.i32 >> (count & 0xff);1714
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1715
- r_.neon_i32 = vshl_s32(a_.neon_i32,1716
- vmov_n_s32(-HEDLEY_STATIC_CAST(int32_t, count)));1717
-#else1718
- SIMDE_VECTORIZE1719
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1720
- r_.i32[i] = a_.i32[i] >> (count & 0xff);1721
- }1722
-#endif1723
-1724
- return simde__m64_from_private(r_);1725
-#endif1726
-}1727
-#define simde_m_psradi(a, count) simde_mm_srai_pi32(a, count)1728
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1729
-#define _mm_srai_pi32(a, count) simde_mm_srai_pi32(a, count)1730
-#define _m_srai_pi32(a, count) simde_mm_srai_pi32(a, count)1731
-#endif1732
-1733
-SIMDE_FUNCTION_ATTRIBUTES1734
-simde__m64 simde_mm_sra_pi16(simde__m64 a, simde__m64 count)1735
-{1736
-#if defined(SIMDE_X86_MMX_NATIVE)1737
- return _mm_sra_pi16(a, count);1738
-#else1739
- simde__m64_private r_;1740
- simde__m64_private a_ = simde__m64_to_private(a);1741
- simde__m64_private count_ = simde__m64_to_private(count);1742
- const int cnt = HEDLEY_STATIC_CAST(1743
- int, (count_.i64[0] > 15 ? 15 : count_.i64[0]));1744
-1745
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1746
- r_.i16 = a_.i16 >> cnt;1747
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1748
- r_.neon_i16 =1749
- vshl_s16(a_.neon_i16,1750
- vmov_n_s16(-HEDLEY_STATIC_CAST(1751
- int16_t, vget_lane_u64(count_.neon_u64, 0))));1752
-#else1753
- SIMDE_VECTORIZE1754
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1755
- r_.i16[i] = a_.i16[i] >> cnt;1756
- }1757
-#endif1758
-1759
- return simde__m64_from_private(r_);1760
-#endif1761
-}1762
-#define simde_m_psraw(a, count) simde_mm_sra_pi16(a, count)1763
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1764
-#define _mm_sra_pi16(a, count) simde_mm_sra_pi16(a, count)1765
-#define _m_psraw(a, count) simde_mm_sra_pi16(a, count)1766
-#endif1767
-1768
-SIMDE_FUNCTION_ATTRIBUTES1769
-simde__m64 simde_mm_sra_pi32(simde__m64 a, simde__m64 count)1770
-{1771
-#if defined(SIMDE_X86_MMX_NATIVE)1772
- return _mm_sra_pi32(a, count);1773
-#else1774
- simde__m64_private r_;1775
- simde__m64_private a_ = simde__m64_to_private(a);1776
- simde__m64_private count_ = simde__m64_to_private(count);1777
- const int32_t cnt =1778
- (count_.u64[0] > 31)1779
- ? 311780
- : HEDLEY_STATIC_CAST(int32_t, count_.u64[0]);1781
-1782
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1783
- r_.i32 = a_.i32 >> cnt;1784
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1785
- r_.neon_i32 =1786
- vshl_s32(a_.neon_i32,1787
- vmov_n_s32(-HEDLEY_STATIC_CAST(1788
- int32_t, vget_lane_u64(count_.neon_u64, 0))));1789
-#else1790
- SIMDE_VECTORIZE1791
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1792
- r_.i32[i] = a_.i32[i] >> cnt;1793
- }1794
-#endif1795
-1796
- return simde__m64_from_private(r_);1797
-#endif1798
-}1799
-#define simde_m_psrad(a, b) simde_mm_sra_pi32(a, b)1800
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1801
-#define _mm_sra_pi32(a, count) simde_mm_sra_pi32(a, count)1802
-#define _m_psrad(a, count) simde_mm_sra_pi32(a, count)1803
-#endif1804
-1805
-SIMDE_FUNCTION_ATTRIBUTES1806
-simde__m64 simde_mm_sub_pi8(simde__m64 a, simde__m64 b)1807
-{1808
-#if defined(SIMDE_X86_MMX_NATIVE)1809
- return _mm_sub_pi8(a, b);1810
-#else1811
- simde__m64_private r_;1812
- simde__m64_private a_ = simde__m64_to_private(a);1813
- simde__m64_private b_ = simde__m64_to_private(b);1814
-1815
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1816
- r_.neon_i8 = vsub_s8(a_.neon_i8, b_.neon_i8);1817
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1818
- r_.i8 = a_.i8 - b_.i8;1819
-#else1820
- SIMDE_VECTORIZE1821
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1822
- r_.i8[i] = a_.i8[i] - b_.i8[i];1823
- }1824
-#endif1825
-1826
- return simde__m64_from_private(r_);1827
-#endif1828
-}1829
-#define simde_m_psubb(a, b) simde_mm_sub_pi8(a, b)1830
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1831
-#define _mm_sub_pi8(a, b) simde_mm_sub_pi8(a, b)1832
-#define _m_psubb(a, b) simde_mm_sub_pi8(a, b)1833
-#endif1834
-1835
-SIMDE_FUNCTION_ATTRIBUTES1836
-simde__m64 simde_mm_sub_pi16(simde__m64 a, simde__m64 b)1837
-{1838
-#if defined(SIMDE_X86_MMX_NATIVE)1839
- return _mm_sub_pi16(a, b);1840
-#else1841
- simde__m64_private r_;1842
- simde__m64_private a_ = simde__m64_to_private(a);1843
- simde__m64_private b_ = simde__m64_to_private(b);1844
-1845
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1846
- r_.neon_i16 = vsub_s16(a_.neon_i16, b_.neon_i16);1847
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1848
- r_.i16 = a_.i16 - b_.i16;1849
-#else1850
- SIMDE_VECTORIZE1851
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1852
- r_.i16[i] = a_.i16[i] - b_.i16[i];1853
- }1854
-#endif1855
-1856
- return simde__m64_from_private(r_);1857
-#endif1858
-}1859
-#define simde_m_psubw(a, b) simde_mm_sub_pi16(a, b)1860
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1861
-#define _mm_sub_pi16(a, b) simde_mm_sub_pi16(a, b)1862
-#define _m_psubw(a, b) simde_mm_sub_pi16(a, b)1863
-#endif1864
-1865
-SIMDE_FUNCTION_ATTRIBUTES1866
-simde__m64 simde_mm_sub_pi32(simde__m64 a, simde__m64 b)1867
-{1868
-#if defined(SIMDE_X86_MMX_NATIVE)1869
- return _mm_sub_pi32(a, b);1870
-#else1871
- simde__m64_private r_;1872
- simde__m64_private a_ = simde__m64_to_private(a);1873
- simde__m64_private b_ = simde__m64_to_private(b);1874
-1875
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1876
- r_.neon_i32 = vsub_s32(a_.neon_i32, b_.neon_i32);1877
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1878
- r_.i32 = a_.i32 - b_.i32;1879
-#else1880
- SIMDE_VECTORIZE1881
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1882
- r_.i32[i] = a_.i32[i] - b_.i32[i];1883
- }1884
-#endif1885
-1886
- return simde__m64_from_private(r_);1887
-#endif1888
-}1889
-#define simde_m_psubd(a, b) simde_mm_sub_pi32(a, b)1890
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1891
-#define _mm_sub_pi32(a, b) simde_mm_sub_pi32(a, b)1892
-#define _m_psubd(a, b) simde_mm_sub_pi32(a, b)1893
-#endif1894
-1895
-SIMDE_FUNCTION_ATTRIBUTES1896
-simde__m64 simde_mm_subs_pi8(simde__m64 a, simde__m64 b)1897
-{1898
-#if defined(SIMDE_X86_MMX_NATIVE)1899
- return _mm_subs_pi8(a, b);1900
-#else1901
- simde__m64_private r_;1902
- simde__m64_private a_ = simde__m64_to_private(a);1903
- simde__m64_private b_ = simde__m64_to_private(b);1904
-1905
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1906
- r_.neon_i8 = vqsub_s8(a_.neon_i8, b_.neon_i8);1907
-#else1908
- SIMDE_VECTORIZE1909
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1910
- if (((b_.i8[i]) > 0 && (a_.i8[i]) < INT8_MIN + (b_.i8[i]))) {1911
- r_.i8[i] = INT8_MIN;1912
- } else if ((b_.i8[i]) < 0 &&1913
- (a_.i8[i]) > INT8_MAX + (b_.i8[i])) {1914
- r_.i8[i] = INT8_MAX;1915
- } else {1916
- r_.i8[i] = (a_.i8[i]) - (b_.i8[i]);1917
- }1918
- }1919
-#endif1920
-1921
- return simde__m64_from_private(r_);1922
-#endif1923
-}1924
-#define simde_m_psubsb(a, b) simde_mm_subs_pi8(a, b)1925
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1926
-#define _mm_subs_pi8(a, b) simde_mm_subs_pi8(a, b)1927
-#define _m_psubsb(a, b) simde_mm_subs_pi8(a, b)1928
-#endif1929
-1930
-SIMDE_FUNCTION_ATTRIBUTES1931
-simde__m64 simde_mm_subs_pu8(simde__m64 a, simde__m64 b)1932
-{1933
-#if defined(SIMDE_X86_MMX_NATIVE)1934
- return _mm_subs_pu8(a, b);1935
-#else1936
- simde__m64_private r_;1937
- simde__m64_private a_ = simde__m64_to_private(a);1938
- simde__m64_private b_ = simde__m64_to_private(b);1939
-1940
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1941
- r_.neon_u8 = vqsub_u8(a_.neon_u8, b_.neon_u8);1942
-#else1943
- SIMDE_VECTORIZE1944
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {1945
- const int32_t x = a_.u8[i] - b_.u8[i];1946
- if (x < 0) {1947
- r_.u8[i] = 0;1948
- } else if (x > UINT8_MAX) {1949
- r_.u8[i] = UINT8_MAX;1950
- } else {1951
- r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, x);1952
- }1953
- }1954
-#endif1955
-1956
- return simde__m64_from_private(r_);1957
-#endif1958
-}1959
-#define simde_m_psubusb(a, b) simde_mm_subs_pu8(a, b)1960
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1961
-#define _mm_subs_pu8(a, b) simde_mm_subs_pu8(a, b)1962
-#define _m_psubusb(a, b) simde_mm_subs_pu8(a, b)1963
-#endif1964
-1965
-SIMDE_FUNCTION_ATTRIBUTES1966
-simde__m64 simde_mm_subs_pi16(simde__m64 a, simde__m64 b)1967
-{1968
-#if defined(SIMDE_X86_MMX_NATIVE)1969
- return _mm_subs_pi16(a, b);1970
-#else1971
- simde__m64_private r_;1972
- simde__m64_private a_ = simde__m64_to_private(a);1973
- simde__m64_private b_ = simde__m64_to_private(b);1974
-1975
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1976
- r_.neon_i16 = vqsub_s16(a_.neon_i16, b_.neon_i16);1977
-#else1978
- SIMDE_VECTORIZE1979
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1980
- if (((b_.i16[i]) > 0 && (a_.i16[i]) < SHRT_MIN + (b_.i16[i]))) {1981
- r_.i16[i] = SHRT_MIN;1982
- } else if ((b_.i16[i]) < 0 &&1983
- (a_.i16[i]) > INT16_MAX + (b_.i16[i])) {1984
- r_.i16[i] = INT16_MAX;1985
- } else {1986
- r_.i16[i] = (a_.i16[i]) - (b_.i16[i]);1987
- }1988
- }1989
-#endif1990
-1991
- return simde__m64_from_private(r_);1992
-#endif1993
-}1994
-#define simde_m_psubsw(a, b) simde_mm_subs_pi16(a, b)1995
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1996
-#define _mm_subs_pi16(a, b) simde_mm_subs_pi16(a, b)1997
-#define _m_psubsw(a, b) simde_mm_subs_pi16(a, b)1998
-#endif1999
-2000
-SIMDE_FUNCTION_ATTRIBUTES2001
-simde__m64 simde_mm_subs_pu16(simde__m64 a, simde__m64 b)2002
-{2003
-#if defined(SIMDE_X86_MMX_NATIVE)2004
- return _mm_subs_pu16(a, b);2005
-#else2006
- simde__m64_private r_;2007
- simde__m64_private a_ = simde__m64_to_private(a);2008
- simde__m64_private b_ = simde__m64_to_private(b);2009
-2010
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2011
- r_.neon_u16 = vqsub_u16(a_.neon_u16, b_.neon_u16);2012
-#else2013
- SIMDE_VECTORIZE2014
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {2015
- const int x = a_.u16[i] - b_.u16[i];2016
- if (x < 0) {2017
- r_.u16[i] = 0;2018
- } else if (x > UINT16_MAX) {2019
- r_.u16[i] = UINT16_MAX;2020
- } else {2021
- r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, x);2022
- }2023
- }2024
-#endif2025
-2026
- return simde__m64_from_private(r_);2027
-#endif2028
-}2029
-#define simde_m_psubusw(a, b) simde_mm_subs_pu16(a, b)2030
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2031
-#define _mm_subs_pu16(a, b) simde_mm_subs_pu16(a, b)2032
-#define _m_psubusw(a, b) simde_mm_subs_pu16(a, b)2033
-#endif2034
-2035
-SIMDE_FUNCTION_ATTRIBUTES2036
-simde__m64 simde_mm_unpackhi_pi8(simde__m64 a, simde__m64 b)2037
-{2038
-#if defined(SIMDE_X86_MMX_NATIVE)2039
- return _mm_unpackhi_pi8(a, b);2040
-#else2041
- simde__m64_private r_;2042
- simde__m64_private a_ = simde__m64_to_private(a);2043
- simde__m64_private b_ = simde__m64_to_private(b);2044
-2045
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2046
- r_.neon_i8 = vzip2_s8(a_.neon_i8, b_.neon_i8);2047
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2048
- r_.i8 = SIMDE_SHUFFLE_VECTOR_(8, 8, a_.i8, b_.i8, 4, 12, 5, 13, 6, 14,2049
- 7, 15);2050
-#else2051
- r_.i8[0] = a_.i8[4];2052
- r_.i8[1] = b_.i8[4];2053
- r_.i8[2] = a_.i8[5];2054
- r_.i8[3] = b_.i8[5];2055
- r_.i8[4] = a_.i8[6];2056
- r_.i8[5] = b_.i8[6];2057
- r_.i8[6] = a_.i8[7];2058
- r_.i8[7] = b_.i8[7];2059
-#endif2060
-2061
- return simde__m64_from_private(r_);2062
-#endif2063
-}2064
-#define simde_m_punpckhbw(a, b) simde_mm_unpackhi_pi8(a, b)2065
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2066
-#define _mm_unpackhi_pi8(a, b) simde_mm_unpackhi_pi8(a, b)2067
-#define _m_punpckhbw(a, b) simde_mm_unpackhi_pi8(a, b)2068
-#endif2069
-2070
-SIMDE_FUNCTION_ATTRIBUTES2071
-simde__m64 simde_mm_unpackhi_pi16(simde__m64 a, simde__m64 b)2072
-{2073
-#if defined(SIMDE_X86_MMX_NATIVE)2074
- return _mm_unpackhi_pi16(a, b);2075
-#else2076
- simde__m64_private r_;2077
- simde__m64_private a_ = simde__m64_to_private(a);2078
- simde__m64_private b_ = simde__m64_to_private(b);2079
-2080
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2081
- r_.neon_i16 = vzip2_s16(a_.neon_i16, b_.neon_i16);2082
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2083
- r_.i16 = SIMDE_SHUFFLE_VECTOR_(16, 8, a_.i16, b_.i16, 2, 6, 3, 7);2084
-#else2085
- r_.i16[0] = a_.i16[2];2086
- r_.i16[1] = b_.i16[2];2087
- r_.i16[2] = a_.i16[3];2088
- r_.i16[3] = b_.i16[3];2089
-#endif2090
-2091
- return simde__m64_from_private(r_);2092
-#endif2093
-}2094
-#define simde_m_punpckhwd(a, b) simde_mm_unpackhi_pi16(a, b)2095
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2096
-#define _mm_unpackhi_pi16(a, b) simde_mm_unpackhi_pi16(a, b)2097
-#define _m_punpckhwd(a, b) simde_mm_unpackhi_pi16(a, b)2098
-#endif2099
-2100
-SIMDE_FUNCTION_ATTRIBUTES2101
-simde__m64 simde_mm_unpackhi_pi32(simde__m64 a, simde__m64 b)2102
-{2103
-#if defined(SIMDE_X86_MMX_NATIVE)2104
- return _mm_unpackhi_pi32(a, b);2105
-#else2106
- simde__m64_private r_;2107
- simde__m64_private a_ = simde__m64_to_private(a);2108
- simde__m64_private b_ = simde__m64_to_private(b);2109
-2110
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2111
- r_.neon_i32 = vzip2_s32(a_.neon_i32, b_.neon_i32);2112
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2113
- r_.i32 = SIMDE_SHUFFLE_VECTOR_(32, 8, a_.i32, b_.i32, 1, 3);2114
-#else2115
- r_.i32[0] = a_.i32[1];2116
- r_.i32[1] = b_.i32[1];2117
-#endif2118
-2119
- return simde__m64_from_private(r_);2120
-#endif2121
-}2122
-#define simde_m_punpckhdq(a, b) simde_mm_unpackhi_pi32(a, b)2123
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2124
-#define _mm_unpackhi_pi32(a, b) simde_mm_unpackhi_pi32(a, b)2125
-#define _m_punpckhdq(a, b) simde_mm_unpackhi_pi32(a, b)2126
-#endif2127
-2128
-SIMDE_FUNCTION_ATTRIBUTES2129
-simde__m64 simde_mm_unpacklo_pi8(simde__m64 a, simde__m64 b)2130
-{2131
-#if defined(SIMDE_X86_MMX_NATIVE)2132
- return _mm_unpacklo_pi8(a, b);2133
-#else2134
- simde__m64_private r_;2135
- simde__m64_private a_ = simde__m64_to_private(a);2136
- simde__m64_private b_ = simde__m64_to_private(b);2137
-2138
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2139
- r_.neon_i8 = vzip1_s8(a_.neon_i8, b_.neon_i8);2140
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2141
- r_.i8 = SIMDE_SHUFFLE_VECTOR_(8, 8, a_.i8, b_.i8, 0, 8, 1, 9, 2, 10, 3,2142
- 11);2143
-#else2144
- r_.i8[0] = a_.i8[0];2145
- r_.i8[1] = b_.i8[0];2146
- r_.i8[2] = a_.i8[1];2147
- r_.i8[3] = b_.i8[1];2148
- r_.i8[4] = a_.i8[2];2149
- r_.i8[5] = b_.i8[2];2150
- r_.i8[6] = a_.i8[3];2151
- r_.i8[7] = b_.i8[3];2152
-#endif2153
-2154
- return simde__m64_from_private(r_);2155
-#endif2156
-}2157
-#define simde_m_punpcklbw(a, b) simde_mm_unpacklo_pi8(a, b)2158
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2159
-#define _mm_unpacklo_pi8(a, b) simde_mm_unpacklo_pi8(a, b)2160
-#define _m_punpcklbw(a, b) simde_mm_unpacklo_pi8(a, b)2161
-#endif2162
-2163
-SIMDE_FUNCTION_ATTRIBUTES2164
-simde__m64 simde_mm_unpacklo_pi16(simde__m64 a, simde__m64 b)2165
-{2166
-#if defined(SIMDE_X86_MMX_NATIVE)2167
- return _mm_unpacklo_pi16(a, b);2168
-#else2169
- simde__m64_private r_;2170
- simde__m64_private a_ = simde__m64_to_private(a);2171
- simde__m64_private b_ = simde__m64_to_private(b);2172
-2173
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2174
- r_.neon_i16 = vzip1_s16(a_.neon_i16, b_.neon_i16);2175
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2176
- r_.i16 = SIMDE_SHUFFLE_VECTOR_(16, 8, a_.i16, b_.i16, 0, 4, 1, 5);2177
-#else2178
- r_.i16[0] = a_.i16[0];2179
- r_.i16[1] = b_.i16[0];2180
- r_.i16[2] = a_.i16[1];2181
- r_.i16[3] = b_.i16[1];2182
-#endif2183
-2184
- return simde__m64_from_private(r_);2185
-#endif2186
-}2187
-#define simde_m_punpcklwd(a, b) simde_mm_unpacklo_pi16(a, b)2188
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2189
-#define _mm_unpacklo_pi16(a, b) simde_mm_unpacklo_pi16(a, b)2190
-#define _m_punpcklwd(a, b) simde_mm_unpacklo_pi16(a, b)2191
-#endif2192
-2193
-SIMDE_FUNCTION_ATTRIBUTES2194
-simde__m64 simde_mm_unpacklo_pi32(simde__m64 a, simde__m64 b)2195
-{2196
-#if defined(SIMDE_X86_MMX_NATIVE)2197
- return _mm_unpacklo_pi32(a, b);2198
-#else2199
- simde__m64_private r_;2200
- simde__m64_private a_ = simde__m64_to_private(a);2201
- simde__m64_private b_ = simde__m64_to_private(b);2202
-2203
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2204
- r_.neon_i32 = vzip1_s32(a_.neon_i32, b_.neon_i32);2205
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2206
- r_.i32 = SIMDE_SHUFFLE_VECTOR_(32, 8, a_.i32, b_.i32, 0, 2);2207
-#else2208
- r_.i32[0] = a_.i32[0];2209
- r_.i32[1] = b_.i32[0];2210
-#endif2211
-2212
- return simde__m64_from_private(r_);2213
-#endif2214
-}2215
-#define simde_m_punpckldq(a, b) simde_mm_unpacklo_pi32(a, b)2216
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2217
-#define _mm_unpacklo_pi32(a, b) simde_mm_unpacklo_pi32(a, b)2218
-#define _m_punpckldq(a, b) simde_mm_unpacklo_pi32(a, b)2219
-#endif2220
-2221
-SIMDE_FUNCTION_ATTRIBUTES2222
-simde__m64 simde_mm_xor_si64(simde__m64 a, simde__m64 b)2223
-{2224
-#if defined(SIMDE_X86_MMX_NATIVE)2225
- return _mm_xor_si64(a, b);2226
-#else2227
- simde__m64_private r_;2228
- simde__m64_private a_ = simde__m64_to_private(a);2229
- simde__m64_private b_ = simde__m64_to_private(b);2230
-2231
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2232
- r_.neon_i32 = veor_s32(a_.neon_i32, b_.neon_i32);2233
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2234
- r_.i32f = a_.i32f ^ b_.i32f;2235
-#else2236
- r_.u64[0] = a_.u64[0] ^ b_.u64[0];2237
-#endif2238
-2239
- return simde__m64_from_private(r_);2240
-#endif2241
-}2242
-#define simde_m_pxor(a, b) simde_mm_xor_si64(a, b)2243
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2244
-#define _mm_xor_si64(a, b) simde_mm_xor_si64(a, b)2245
-#define _m_pxor(a, b) simde_mm_xor_si64(a, b)2246
-#endif2247
-2248
-SIMDE_FUNCTION_ATTRIBUTES2249
-int32_t simde_m_to_int(simde__m64 a)2250
-{2251
-#if defined(SIMDE_X86_MMX_NATIVE)2252
- return _m_to_int(a);2253
-#else2254
- simde__m64_private a_ = simde__m64_to_private(a);2255
-2256
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2257
- return vget_lane_s32(a_.neon_i32, 0);2258
-#else2259
- return a_.i32[0];2260
-#endif2261
-#endif2262
-}2263
-#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2264
-#define _m_to_int(a) simde_m_to_int(a)2265
-#endif2266
-2267
-SIMDE_END_DECLS_2268
-2269
-HEDLEY_DIAGNOSTIC_POP2270
-2271
-#endif /* !defined(SIMDE_X86_MMX_H) */2272
obs-studio-26.1.0.tar.xz/libobs/util/simde/sse.h
Deleted
3720
1
2
-/* SPDX-License-Identifier: MIT3
- *4
- * Permission is hereby granted, free of charge, to any person5
- * obtaining a copy of this software and associated documentation6
- * files (the "Software"), to deal in the Software without7
- * restriction, including without limitation the rights to use, copy,8
- * modify, merge, publish, distribute, sublicense, and/or sell copies9
- * of the Software, and to permit persons to whom the Software is10
- * furnished to do so, subject to the following conditions:11
- *12
- * The above copyright notice and this permission notice shall be13
- * included in all copies or substantial portions of the Software.14
- *15
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,16
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF17
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND18
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS19
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN20
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN21
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE22
- * SOFTWARE.23
- *24
- * Copyright:25
- * 2017-2020 Evan Nemerson <evan@nemerson.com>26
- * 2015-2017 John W. Ratcliff <jratcliffscarab@gmail.com>27
- * 2015 Brandon Rowlett <browlett@nvidia.com>28
- * 2015 Ken Fast <kfast@gdeb.com>29
- */30
-31
-#if !defined(SIMDE_X86_SSE_H)32
-#define SIMDE_X86_SSE_H33
-34
-#include "mmx.h"35
-36
-#if !defined(SIMDE_X86_AVX512F_NATIVE) && defined(SIMDE_ENABLE_NATIVE_ALIASES)37
-#define SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES38
-#endif39
-40
-#if defined(_WIN32)41
-#include <windows.h>42
-#endif43
-44
-HEDLEY_DIAGNOSTIC_PUSH45
-SIMDE_DISABLE_UNWANTED_DIAGNOSTICS46
-SIMDE_BEGIN_DECLS_47
-48
-typedef union {49
-#if defined(SIMDE_VECTOR_SUBSCRIPT)50
- SIMDE_ALIGN(16) int8_t i8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;51
- SIMDE_ALIGN(16) int16_t i16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;52
- SIMDE_ALIGN(16) int32_t i32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;53
- SIMDE_ALIGN(16) int64_t i64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;54
- SIMDE_ALIGN(16) uint8_t u8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;55
- SIMDE_ALIGN(16) uint16_t u16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;56
- SIMDE_ALIGN(16) uint32_t u32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;57
- SIMDE_ALIGN(16) uint64_t u64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;58
-#if defined(SIMDE_HAVE_INT128_)59
- SIMDE_ALIGN(16) simde_int128 i128 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;60
- SIMDE_ALIGN(16) simde_uint128 u128 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;61
-#endif62
- SIMDE_ALIGN(16) simde_float32 f32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;63
- SIMDE_ALIGN(16) int_fast32_t i32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;64
- SIMDE_ALIGN(16) uint_fast32_t u32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;65
-#else66
- SIMDE_ALIGN(16) int8_t i8[16];67
- SIMDE_ALIGN(16) int16_t i16[8];68
- SIMDE_ALIGN(16) int32_t i32[4];69
- SIMDE_ALIGN(16) int64_t i64[2];70
- SIMDE_ALIGN(16) uint8_t u8[16];71
- SIMDE_ALIGN(16) uint16_t u16[8];72
- SIMDE_ALIGN(16) uint32_t u32[4];73
- SIMDE_ALIGN(16) uint64_t u64[2];74
-#if defined(SIMDE_HAVE_INT128_)75
- SIMDE_ALIGN(16) simde_int128 i128[1];76
- SIMDE_ALIGN(16) simde_uint128 u128[1];77
-#endif78
- SIMDE_ALIGN(16) simde_float32 f32[4];79
- SIMDE_ALIGN(16) int_fast32_t i32f[16 / sizeof(int_fast32_t)];80
- SIMDE_ALIGN(16) uint_fast32_t u32f[16 / sizeof(uint_fast32_t)];81
-#endif82
-83
- SIMDE_ALIGN(16) simde__m64_private m64_private[2];84
- SIMDE_ALIGN(16) simde__m64 m64[2];85
-86
-#if defined(SIMDE_X86_SSE_NATIVE)87
- SIMDE_ALIGN(16) __m128 n;88
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)89
- SIMDE_ALIGN(16) int8x16_t neon_i8;90
- SIMDE_ALIGN(16) int16x8_t neon_i16;91
- SIMDE_ALIGN(16) int32x4_t neon_i32;92
- SIMDE_ALIGN(16) int64x2_t neon_i64;93
- SIMDE_ALIGN(16) uint8x16_t neon_u8;94
- SIMDE_ALIGN(16) uint16x8_t neon_u16;95
- SIMDE_ALIGN(16) uint32x4_t neon_u32;96
- SIMDE_ALIGN(16) uint64x2_t neon_u64;97
- SIMDE_ALIGN(16) float32x4_t neon_f32;98
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)99
- SIMDE_ALIGN(16) float64x2_t neon_f64;100
-#endif101
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)102
- SIMDE_ALIGN(16) v128_t wasm_v128;103
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)104
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned char) altivec_u8;105
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned short) altivec_u16;106
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32;107
- SIMDE_ALIGN(16)108
- SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long) altivec_u64;109
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed char) altivec_i8;110
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed short) altivec_i16;111
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32;112
- SIMDE_ALIGN(16)113
- SIMDE_POWER_ALTIVEC_VECTOR(signed long long) altivec_i64;114
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(float) altivec_f32;115
-#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)116
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(double) altivec_f64;117
-#endif118
-#endif119
-} simde__m128_private;120
-121
-#if defined(SIMDE_X86_SSE_NATIVE)122
-typedef __m128 simde__m128;123
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)124
-typedef float32x4_t simde__m128;125
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)126
-typedef v128_t simde__m128;127
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)128
-typedef SIMDE_POWER_ALTIVEC_VECTOR(float) simde__m128;129
-#elif defined(SIMDE_VECTOR_SUBSCRIPT)130
-typedef simde_float32 simde__m128 SIMDE_ALIGN(16)131
- SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;132
-#else133
-typedef simde__m128_private simde__m128;134
-#endif135
-136
-#if !defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_ENABLE_NATIVE_ALIASES)137
-#define SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES138
-typedef simde__m128 __m128;139
-#endif140
-141
-HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128), "simde__m128 size incorrect");142
-HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128_private),143
- "simde__m128_private size incorrect");144
-#if defined(SIMDE_CHECK_ALIGNMENT) && defined(SIMDE_ALIGN_OF)145
-HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128) == 16,146
- "simde__m128 is not 16-byte aligned");147
-HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128_private) == 16,148
- "simde__m128_private is not 16-byte aligned");149
-#endif150
-151
-SIMDE_FUNCTION_ATTRIBUTES152
-simde__m128 simde__m128_from_private(simde__m128_private v)153
-{154
- simde__m128 r;155
- simde_memcpy(&r, &v, sizeof(r));156
- return r;157
-}158
-159
-SIMDE_FUNCTION_ATTRIBUTES160
-simde__m128_private simde__m128_to_private(simde__m128 v)161
-{162
- simde__m128_private r;163
- simde_memcpy(&r, &v, sizeof(r));164
- return r;165
-}166
-167
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)168
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, int8x16_t, neon, i8)169
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, int16x8_t, neon, i16)170
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, int32x4_t, neon, i32)171
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, int64x2_t, neon, i64)172
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, uint8x16_t, neon, u8)173
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, uint16x8_t, neon, u16)174
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, uint32x4_t, neon, u32)175
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, uint64x2_t, neon, u64)176
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, float32x4_t, neon, f32)177
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)178
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, float64x2_t, neon, f64)179
-#endif180
-#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */181
-182
-#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)183
-HEDLEY_DIAGNOSTIC_POP184
-#endif185
-186
-SIMDE_FUNCTION_ATTRIBUTES187
-simde__m128 simde_mm_set_ps(simde_float32 e3, simde_float32 e2,188
- simde_float32 e1, simde_float32 e0)189
-{190
-#if defined(SIMDE_X86_SSE_NATIVE)191
- return _mm_set_ps(e3, e2, e1, e0);192
-#else193
- simde__m128_private r_;194
-195
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)196
- SIMDE_ALIGN(16) simde_float32 data[4] = {e0, e1, e2, e3};197
- r_.neon_f32 = vld1q_f32(data);198
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)199
- r_.wasm_v128 = wasm_f32x4_make(e0, e1, e2, e3);200
-#else201
- r_.f32[0] = e0;202
- r_.f32[1] = e1;203
- r_.f32[2] = e2;204
- r_.f32[3] = e3;205
-#endif206
-207
- return simde__m128_from_private(r_);208
-#endif209
-}210
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)211
-#define _mm_set_ps(e3, e2, e1, e0) simde_mm_set_ps(e3, e2, e1, e0)212
-#endif213
-214
-SIMDE_FUNCTION_ATTRIBUTES215
-simde__m128 simde_mm_set_ps1(simde_float32 a)216
-{217
-#if defined(SIMDE_X86_SSE_NATIVE)218
- return _mm_set_ps1(a);219
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)220
- return vdupq_n_f32(a);221
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)222
- (void)a;223
- return vec_splats(a);224
-#else225
- return simde_mm_set_ps(a, a, a, a);226
-#endif227
-}228
-#define simde_mm_set1_ps(a) simde_mm_set_ps1(a)229
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)230
-#define _mm_set_ps1(a) simde_mm_set_ps1(a)231
-#define _mm_set1_ps(a) simde_mm_set1_ps(a)232
-#endif233
-234
-SIMDE_FUNCTION_ATTRIBUTES235
-simde__m128 simde_mm_move_ss(simde__m128 a, simde__m128 b)236
-{237
-#if defined(SIMDE_X86_SSE_NATIVE)238
- return _mm_move_ss(a, b);239
-#else240
- simde__m128_private r_, a_ = simde__m128_to_private(a),241
- b_ = simde__m128_to_private(b);242
-243
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)244
- r_.neon_f32 =245
- vsetq_lane_f32(vgetq_lane_f32(b_.neon_f32, 0), a_.neon_f32, 0);246
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)247
- SIMDE_POWER_ALTIVEC_VECTOR(unsigned char)248
- m = {16, 17, 18, 19, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};249
- r_.altivec_f32 = vec_perm(a_.altivec_f32, b_.altivec_f32, m);250
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)251
- r_.wasm_v128 = wasm_v8x16_shuffle(b_.wasm_v128, a_.wasm_v128, 0, 1, 2,252
- 3, 20, 21, 22, 23, 24, 25, 26, 27, 28,253
- 29, 30, 31);254
-#elif defined(SIMDE_SHUFFLE_VECTOR_)255
- r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 4, 1, 2, 3);256
-#else257
- r_.f32[0] = b_.f32[0];258
- r_.f32[1] = a_.f32[1];259
- r_.f32[2] = a_.f32[2];260
- r_.f32[3] = a_.f32[3];261
-#endif262
-263
- return simde__m128_from_private(r_);264
-#endif265
-}266
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)267
-#define _mm_move_ss(a, b) simde_mm_move_ss((a), (b))268
-#endif269
-270
-SIMDE_FUNCTION_ATTRIBUTES271
-simde__m128 simde_mm_add_ps(simde__m128 a, simde__m128 b)272
-{273
-#if defined(SIMDE_X86_SSE_NATIVE)274
- return _mm_add_ps(a, b);275
-#else276
- simde__m128_private r_, a_ = simde__m128_to_private(a),277
- b_ = simde__m128_to_private(b);278
-279
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)280
- r_.neon_f32 = vaddq_f32(a_.neon_f32, b_.neon_f32);281
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)282
- r_.wasm_v128 = wasm_f32x4_add(a_.wasm_v128, b_.wasm_v128);283
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)284
- r_.altivec_f32 = vec_add(a_.altivec_f32, b_.altivec_f32);285
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)286
- r_.f32 = a_.f32 + b_.f32;287
-#else288
- SIMDE_VECTORIZE289
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {290
- r_.f32[i] = a_.f32[i] + b_.f32[i];291
- }292
-#endif293
-294
- return simde__m128_from_private(r_);295
-#endif296
-}297
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)298
-#define _mm_add_ps(a, b) simde_mm_add_ps((a), (b))299
-#endif300
-301
-SIMDE_FUNCTION_ATTRIBUTES302
-simde__m128 simde_mm_add_ss(simde__m128 a, simde__m128 b)303
-{304
-#if defined(SIMDE_X86_SSE_NATIVE)305
- return _mm_add_ss(a, b);306
-#elif defined(SIMDE_ASSUME_VECTORIZATION)307
- return simde_mm_move_ss(a, simde_mm_add_ps(a, b));308
-#else309
- simde__m128_private r_, a_ = simde__m128_to_private(a),310
- b_ = simde__m128_to_private(b);311
-312
- r_.f32[0] = a_.f32[0] + b_.f32[0];313
- r_.f32[1] = a_.f32[1];314
- r_.f32[2] = a_.f32[2];315
- r_.f32[3] = a_.f32[3];316
-317
- return simde__m128_from_private(r_);318
-#endif319
-}320
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)321
-#define _mm_add_ss(a, b) simde_mm_add_ss((a), (b))322
-#endif323
-324
-SIMDE_FUNCTION_ATTRIBUTES325
-simde__m128 simde_mm_and_ps(simde__m128 a, simde__m128 b)326
-{327
-#if defined(SIMDE_X86_SSE_NATIVE)328
- return _mm_and_ps(a, b);329
-#else330
- simde__m128_private r_, a_ = simde__m128_to_private(a),331
- b_ = simde__m128_to_private(b);332
-333
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)334
- r_.neon_i32 = vandq_s32(a_.neon_i32, b_.neon_i32);335
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)336
- r_.wasm_v128 = wasm_v128_and(a_.wasm_v128, b_.wasm_v128);337
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)338
- r_.i32 = a_.i32 & b_.i32;339
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)340
- r_.altivec_f32 = vec_and(a_.altivec_f32, b_.altivec_f32);341
-#else342
- SIMDE_VECTORIZE343
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {344
- r_.i32[i] = a_.i32[i] & b_.i32[i];345
- }346
-#endif347
-348
- return simde__m128_from_private(r_);349
-#endif350
-}351
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)352
-#define _mm_and_ps(a, b) simde_mm_and_ps((a), (b))353
-#endif354
-355
-SIMDE_FUNCTION_ATTRIBUTES356
-simde__m128 simde_mm_andnot_ps(simde__m128 a, simde__m128 b)357
-{358
-#if defined(SIMDE_X86_SSE_NATIVE)359
- return _mm_andnot_ps(a, b);360
-#else361
- simde__m128_private r_, a_ = simde__m128_to_private(a),362
- b_ = simde__m128_to_private(b);363
-364
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)365
- r_.neon_i32 = vbicq_s32(b_.neon_i32, a_.neon_i32);366
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)367
- r_.wasm_v128 = wasm_v128_andnot(b_.wasm_v128, a_.wasm_v128);368
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)369
- r_.altivec_f32 = vec_andc(b_.altivec_f32, a_.altivec_f32);370
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)371
- r_.i32 = ~a_.i32 & b_.i32;372
-#else373
- SIMDE_VECTORIZE374
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {375
- r_.i32[i] = ~(a_.i32[i]) & b_.i32[i];376
- }377
-#endif378
-379
- return simde__m128_from_private(r_);380
-#endif381
-}382
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)383
-#define _mm_andnot_ps(a, b) simde_mm_andnot_ps((a), (b))384
-#endif385
-386
-SIMDE_FUNCTION_ATTRIBUTES387
-simde__m64 simde_mm_avg_pu16(simde__m64 a, simde__m64 b)388
-{389
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)390
- return _mm_avg_pu16(a, b);391
-#else392
- simde__m64_private r_, a_ = simde__m64_to_private(a),393
- b_ = simde__m64_to_private(b);394
-395
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)396
- r_.neon_u16 = vrhadd_u16(b_.neon_u16, a_.neon_u16);397
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS) && \398
- defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \399
- defined(SIMDE_CONVERT_VECTOR_)400
- uint32_t wa SIMDE_VECTOR(16);401
- uint32_t wb SIMDE_VECTOR(16);402
- uint32_t wr SIMDE_VECTOR(16);403
- SIMDE_CONVERT_VECTOR_(wa, a_.u16);404
- SIMDE_CONVERT_VECTOR_(wb, b_.u16);405
- wr = (wa + wb + 1) >> 1;406
- SIMDE_CONVERT_VECTOR_(r_.u16, wr);407
-#else408
- SIMDE_VECTORIZE409
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {410
- r_.u16[i] = (a_.u16[i] + b_.u16[i] + 1) >> 1;411
- }412
-#endif413
-414
- return simde__m64_from_private(r_);415
-#endif416
-}417
-#define simde_m_pavgw(a, b) simde_mm_avg_pu16(a, b)418
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)419
-#define _mm_avg_pu16(a, b) simde_mm_avg_pu16(a, b)420
-#define _m_pavgw(a, b) simde_mm_avg_pu16(a, b)421
-#endif422
-423
-SIMDE_FUNCTION_ATTRIBUTES424
-simde__m64 simde_mm_avg_pu8(simde__m64 a, simde__m64 b)425
-{426
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)427
- return _mm_avg_pu8(a, b);428
-#else429
- simde__m64_private r_, a_ = simde__m64_to_private(a),430
- b_ = simde__m64_to_private(b);431
-432
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)433
- r_.neon_u8 = vrhadd_u8(b_.neon_u8, a_.neon_u8);434
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS) && \435
- defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \436
- defined(SIMDE_CONVERT_VECTOR_)437
- uint16_t wa SIMDE_VECTOR(16);438
- uint16_t wb SIMDE_VECTOR(16);439
- uint16_t wr SIMDE_VECTOR(16);440
- SIMDE_CONVERT_VECTOR_(wa, a_.u8);441
- SIMDE_CONVERT_VECTOR_(wb, b_.u8);442
- wr = (wa + wb + 1) >> 1;443
- SIMDE_CONVERT_VECTOR_(r_.u8, wr);444
-#else445
- SIMDE_VECTORIZE446
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {447
- r_.u8[i] = (a_.u8[i] + b_.u8[i] + 1) >> 1;448
- }449
-#endif450
-451
- return simde__m64_from_private(r_);452
-#endif453
-}454
-#define simde_m_pavgb(a, b) simde_mm_avg_pu8(a, b)455
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)456
-#define _mm_avg_pu8(a, b) simde_mm_avg_pu8(a, b)457
-#define _m_pavgb(a, b) simde_mm_avg_pu8(a, b)458
-#endif459
-460
-SIMDE_FUNCTION_ATTRIBUTES461
-simde__m128 simde_mm_cmpeq_ps(simde__m128 a, simde__m128 b)462
-{463
-#if defined(SIMDE_X86_SSE_NATIVE)464
- return _mm_cmpeq_ps(a, b);465
-#else466
- simde__m128_private r_, a_ = simde__m128_to_private(a),467
- b_ = simde__m128_to_private(b);468
-469
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)470
- r_.neon_u32 = vceqq_f32(a_.neon_f32, b_.neon_f32);471
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)472
- r_.wasm_v128 = wasm_f32x4_eq(a_.wasm_v128, b_.wasm_v128);473
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)474
- r_.altivec_f32 = (SIMDE_POWER_ALTIVEC_VECTOR(float))vec_cmpeq(475
- a_.altivec_f32, b_.altivec_f32);476
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)477
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), a_.f32 == b_.f32);478
-#else479
- SIMDE_VECTORIZE480
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {481
- r_.u32[i] = (a_.f32[i] == b_.f32[i]) ? ~UINT32_C(0)482
- : UINT32_C(0);483
- }484
-#endif485
-486
- return simde__m128_from_private(r_);487
-#endif488
-}489
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)490
-#define _mm_cmpeq_ps(a, b) simde_mm_cmpeq_ps((a), (b))491
-#endif492
-493
-SIMDE_FUNCTION_ATTRIBUTES494
-simde__m128 simde_mm_cmpeq_ss(simde__m128 a, simde__m128 b)495
-{496
-#if defined(SIMDE_X86_SSE_NATIVE)497
- return _mm_cmpeq_ss(a, b);498
-#elif defined(SIMDE_ASSUME_VECTORIZATION)499
- return simde_mm_move_ss(a, simde_mm_cmpeq_ps(a, b));500
-#else501
- simde__m128_private r_, a_ = simde__m128_to_private(a),502
- b_ = simde__m128_to_private(b);503
-504
- r_.u32[0] = (a_.f32[0] == b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);505
- SIMDE_VECTORIZE506
- for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {507
- r_.u32[i] = a_.u32[i];508
- }509
-510
- return simde__m128_from_private(r_);511
-#endif512
-}513
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)514
-#define _mm_cmpeq_ss(a, b) simde_mm_cmpeq_ss((a), (b))515
-#endif516
-517
-SIMDE_FUNCTION_ATTRIBUTES518
-simde__m128 simde_mm_cmpge_ps(simde__m128 a, simde__m128 b)519
-{520
-#if defined(SIMDE_X86_SSE_NATIVE)521
- return _mm_cmpge_ps(a, b);522
-#else523
- simde__m128_private r_, a_ = simde__m128_to_private(a),524
- b_ = simde__m128_to_private(b);525
-526
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)527
- r_.neon_u32 = vcgeq_f32(a_.neon_f32, b_.neon_f32);528
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)529
- r_.wasm_v128 = wasm_f32x4_ge(a_.wasm_v128, b_.wasm_v128);530
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)531
- r_.altivec_f32 = (SIMDE_POWER_ALTIVEC_VECTOR(float))vec_cmpge(532
- a_.altivec_f32, b_.altivec_f32);533
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)534
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 >= b_.f32));535
-#else536
- SIMDE_VECTORIZE537
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {538
- r_.u32[i] = (a_.f32[i] >= b_.f32[i]) ? ~UINT32_C(0)539
- : UINT32_C(0);540
- }541
-#endif542
-543
- return simde__m128_from_private(r_);544
-#endif545
-}546
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)547
-#define _mm_cmpge_ps(a, b) simde_mm_cmpge_ps((a), (b))548
-#endif549
-550
-SIMDE_FUNCTION_ATTRIBUTES551
-simde__m128 simde_mm_cmpge_ss(simde__m128 a, simde__m128 b)552
-{553
-#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)554
- return _mm_cmpge_ss(a, b);555
-#elif defined(SIMDE_ASSUME_VECTORIZATION)556
- return simde_mm_move_ss(a, simde_mm_cmpge_ps(a, b));557
-#else558
- simde__m128_private r_, a_ = simde__m128_to_private(a),559
- b_ = simde__m128_to_private(b);560
-561
- r_.u32[0] = (a_.f32[0] >= b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);562
- SIMDE_VECTORIZE563
- for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {564
- r_.u32[i] = a_.u32[i];565
- }566
-567
- return simde__m128_from_private(r_);568
-#endif569
-}570
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)571
-#define _mm_cmpge_ss(a, b) simde_mm_cmpge_ss((a), (b))572
-#endif573
-574
-SIMDE_FUNCTION_ATTRIBUTES575
-simde__m128 simde_mm_cmpgt_ps(simde__m128 a, simde__m128 b)576
-{577
-#if defined(SIMDE_X86_SSE_NATIVE)578
- return _mm_cmpgt_ps(a, b);579
-#else580
- simde__m128_private r_, a_ = simde__m128_to_private(a),581
- b_ = simde__m128_to_private(b);582
-583
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)584
- r_.neon_u32 = vcgtq_f32(a_.neon_f32, b_.neon_f32);585
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)586
- r_.wasm_v128 = wasm_f32x4_gt(a_.wasm_v128, b_.wasm_v128);587
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)588
- r_.altivec_f32 = (SIMDE_POWER_ALTIVEC_VECTOR(float))vec_cmpgt(589
- a_.altivec_f32, b_.altivec_f32);590
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)591
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 > b_.f32));592
-#else593
- SIMDE_VECTORIZE594
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {595
- r_.u32[i] = (a_.f32[i] > b_.f32[i]) ? ~UINT32_C(0)596
- : UINT32_C(0);597
- }598
-#endif599
-600
- return simde__m128_from_private(r_);601
-#endif602
-}603
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)604
-#define _mm_cmpgt_ps(a, b) simde_mm_cmpgt_ps((a), (b))605
-#endif606
-607
-SIMDE_FUNCTION_ATTRIBUTES608
-simde__m128 simde_mm_cmpgt_ss(simde__m128 a, simde__m128 b)609
-{610
-#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)611
- return _mm_cmpgt_ss(a, b);612
-#elif defined(SIMDE_ASSUME_VECTORIZATION)613
- return simde_mm_move_ss(a, simde_mm_cmpgt_ps(a, b));614
-#else615
- simde__m128_private r_, a_ = simde__m128_to_private(a),616
- b_ = simde__m128_to_private(b);617
-618
- r_.u32[0] = (a_.f32[0] > b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);619
- SIMDE_VECTORIZE620
- for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {621
- r_.u32[i] = a_.u32[i];622
- }623
-624
- return simde__m128_from_private(r_);625
-#endif626
-}627
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)628
-#define _mm_cmpgt_ss(a, b) simde_mm_cmpgt_ss((a), (b))629
-#endif630
-631
-SIMDE_FUNCTION_ATTRIBUTES632
-simde__m128 simde_mm_cmple_ps(simde__m128 a, simde__m128 b)633
-{634
-#if defined(SIMDE_X86_SSE_NATIVE)635
- return _mm_cmple_ps(a, b);636
-#else637
- simde__m128_private r_, a_ = simde__m128_to_private(a),638
- b_ = simde__m128_to_private(b);639
-640
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)641
- r_.neon_u32 = vcleq_f32(a_.neon_f32, b_.neon_f32);642
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)643
- r_.wasm_v128 = wasm_f32x4_le(a_.wasm_v128, b_.wasm_v128);644
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)645
- r_.altivec_f32 = (SIMDE_POWER_ALTIVEC_VECTOR(float))vec_cmple(646
- a_.altivec_f32, b_.altivec_f32);647
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)648
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 <= b_.f32));649
-#else650
- SIMDE_VECTORIZE651
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {652
- r_.u32[i] = (a_.f32[i] <= b_.f32[i]) ? ~UINT32_C(0)653
- : UINT32_C(0);654
- }655
-#endif656
-657
- return simde__m128_from_private(r_);658
-#endif659
-}660
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)661
-#define _mm_cmple_ps(a, b) simde_mm_cmple_ps((a), (b))662
-#endif663
-664
-SIMDE_FUNCTION_ATTRIBUTES665
-simde__m128 simde_mm_cmple_ss(simde__m128 a, simde__m128 b)666
-{667
-#if defined(SIMDE_X86_SSE_NATIVE)668
- return _mm_cmple_ss(a, b);669
-#elif defined(SIMDE_ASSUME_VECTORIZATION)670
- return simde_mm_move_ss(a, simde_mm_cmple_ps(a, b));671
-#else672
- simde__m128_private r_, a_ = simde__m128_to_private(a),673
- b_ = simde__m128_to_private(b);674
-675
- r_.u32[0] = (a_.f32[0] <= b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);676
- SIMDE_VECTORIZE677
- for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {678
- r_.u32[i] = a_.u32[i];679
- }680
-681
- return simde__m128_from_private(r_);682
-#endif683
-}684
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)685
-#define _mm_cmple_ss(a, b) simde_mm_cmple_ss((a), (b))686
-#endif687
-688
-SIMDE_FUNCTION_ATTRIBUTES689
-simde__m128 simde_mm_cmplt_ps(simde__m128 a, simde__m128 b)690
-{691
-#if defined(SIMDE_X86_SSE_NATIVE)692
- return _mm_cmplt_ps(a, b);693
-#else694
- simde__m128_private r_, a_ = simde__m128_to_private(a),695
- b_ = simde__m128_to_private(b);696
-697
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)698
- r_.neon_u32 = vcltq_f32(a_.neon_f32, b_.neon_f32);699
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)700
- r_.wasm_v128 = wasm_f32x4_lt(a_.wasm_v128, b_.wasm_v128);701
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)702
- r_.altivec_f32 = (SIMDE_POWER_ALTIVEC_VECTOR(float))vec_cmplt(703
- a_.altivec_f32, b_.altivec_f32);704
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)705
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 < b_.f32));706
-#else707
- SIMDE_VECTORIZE708
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {709
- r_.u32[i] = (a_.f32[i] < b_.f32[i]) ? ~UINT32_C(0)710
- : UINT32_C(0);711
- }712
-#endif713
-714
- return simde__m128_from_private(r_);715
-#endif716
-}717
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)718
-#define _mm_cmplt_ps(a, b) simde_mm_cmplt_ps((a), (b))719
-#endif720
-721
-SIMDE_FUNCTION_ATTRIBUTES722
-simde__m128 simde_mm_cmplt_ss(simde__m128 a, simde__m128 b)723
-{724
-#if defined(SIMDE_X86_SSE_NATIVE)725
- return _mm_cmplt_ss(a, b);726
-#elif defined(SIMDE_ASSUME_VECTORIZATION)727
- return simde_mm_move_ss(a, simde_mm_cmplt_ps(a, b));728
-#else729
- simde__m128_private r_, a_ = simde__m128_to_private(a),730
- b_ = simde__m128_to_private(b);731
-732
- r_.u32[0] = (a_.f32[0] < b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);733
- SIMDE_VECTORIZE734
- for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {735
- r_.u32[i] = a_.u32[i];736
- }737
-738
- return simde__m128_from_private(r_);739
-#endif740
-}741
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)742
-#define _mm_cmplt_ss(a, b) simde_mm_cmplt_ss((a), (b))743
-#endif744
-745
-SIMDE_FUNCTION_ATTRIBUTES746
-simde__m128 simde_mm_cmpneq_ps(simde__m128 a, simde__m128 b)747
-{748
-#if defined(SIMDE_X86_SSE_NATIVE)749
- return _mm_cmpneq_ps(a, b);750
-#else751
- simde__m128_private r_, a_ = simde__m128_to_private(a),752
- b_ = simde__m128_to_private(b);753
-754
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)755
- r_.neon_u32 = vmvnq_u32(vceqq_f32(a_.neon_f32, b_.neon_f32));756
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)757
- r_.wasm_v128 = wasm_f32x4_ne(a_.wasm_v128, b_.wasm_v128);758
-#elif defined(SIMDE_POWER_ALTIVEC_P9_NATIVE) && SIMDE_ARCH_POWER_CHECK(900) && \759
- !defined(HEDLEY_IBM_VERSION)760
- /* vec_cmpne(vector float, vector float) is missing from XL C/C++ v16.1.1,761
- though the documentation (table 89 on page 432 of the IBM XL C/C++ for762
- Linux Compiler Reference, Version 16.1.1) shows that it should be763
- present. Both GCC and clang support it. */764
- r_.altivec_f32 = (SIMDE_POWER_ALTIVEC_VECTOR(float))vec_cmpne(765
- a_.altivec_f32, b_.altivec_f32);766
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)767
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 != b_.f32));768
-#else769
- SIMDE_VECTORIZE770
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {771
- r_.u32[i] = (a_.f32[i] != b_.f32[i]) ? ~UINT32_C(0)772
- : UINT32_C(0);773
- }774
-#endif775
-776
- return simde__m128_from_private(r_);777
-#endif778
-}779
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)780
-#define _mm_cmpneq_ps(a, b) simde_mm_cmpneq_ps((a), (b))781
-#endif782
-783
-SIMDE_FUNCTION_ATTRIBUTES784
-simde__m128 simde_mm_cmpneq_ss(simde__m128 a, simde__m128 b)785
-{786
-#if defined(SIMDE_X86_SSE_NATIVE)787
- return _mm_cmpneq_ss(a, b);788
-#elif defined(SIMDE_ASSUME_VECTORIZATION)789
- return simde_mm_move_ss(a, simde_mm_cmpneq_ps(a, b));790
-#else791
- simde__m128_private r_, a_ = simde__m128_to_private(a),792
- b_ = simde__m128_to_private(b);793
-794
- r_.u32[0] = (a_.f32[0] != b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);795
- SIMDE_VECTORIZE796
- for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {797
- r_.u32[i] = a_.u32[i];798
- }799
-800
- return simde__m128_from_private(r_);801
-#endif802
-}803
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)804
-#define _mm_cmpneq_ss(a, b) simde_mm_cmpneq_ss((a), (b))805
-#endif806
-807
-SIMDE_FUNCTION_ATTRIBUTES808
-simde__m128 simde_mm_cmpnge_ps(simde__m128 a, simde__m128 b)809
-{810
- return simde_mm_cmplt_ps(a, b);811
-}812
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)813
-#define _mm_cmpnge_ps(a, b) simde_mm_cmpnge_ps((a), (b))814
-#endif815
-816
-SIMDE_FUNCTION_ATTRIBUTES817
-simde__m128 simde_mm_cmpnge_ss(simde__m128 a, simde__m128 b)818
-{819
- return simde_mm_cmplt_ss(a, b);820
-}821
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)822
-#define _mm_cmpnge_ss(a, b) simde_mm_cmpnge_ss((a), (b))823
-#endif824
-825
-SIMDE_FUNCTION_ATTRIBUTES826
-simde__m128 simde_mm_cmpngt_ps(simde__m128 a, simde__m128 b)827
-{828
- return simde_mm_cmple_ps(a, b);829
-}830
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)831
-#define _mm_cmpngt_ps(a, b) simde_mm_cmpngt_ps((a), (b))832
-#endif833
-834
-SIMDE_FUNCTION_ATTRIBUTES835
-simde__m128 simde_mm_cmpngt_ss(simde__m128 a, simde__m128 b)836
-{837
- return simde_mm_cmple_ss(a, b);838
-}839
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)840
-#define _mm_cmpngt_ss(a, b) simde_mm_cmpngt_ss((a), (b))841
-#endif842
-843
-SIMDE_FUNCTION_ATTRIBUTES844
-simde__m128 simde_mm_cmpnle_ps(simde__m128 a, simde__m128 b)845
-{846
- return simde_mm_cmpgt_ps(a, b);847
-}848
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)849
-#define _mm_cmpnle_ps(a, b) simde_mm_cmpnle_ps((a), (b))850
-#endif851
-852
-SIMDE_FUNCTION_ATTRIBUTES853
-simde__m128 simde_mm_cmpnle_ss(simde__m128 a, simde__m128 b)854
-{855
- return simde_mm_cmpgt_ss(a, b);856
-}857
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)858
-#define _mm_cmpnle_ss(a, b) simde_mm_cmpnle_ss((a), (b))859
-#endif860
-861
-SIMDE_FUNCTION_ATTRIBUTES862
-simde__m128 simde_mm_cmpnlt_ps(simde__m128 a, simde__m128 b)863
-{864
- return simde_mm_cmpge_ps(a, b);865
-}866
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)867
-#define _mm_cmpnlt_ps(a, b) simde_mm_cmpnlt_ps((a), (b))868
-#endif869
-870
-SIMDE_FUNCTION_ATTRIBUTES871
-simde__m128 simde_mm_cmpnlt_ss(simde__m128 a, simde__m128 b)872
-{873
- return simde_mm_cmpge_ss(a, b);874
-}875
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)876
-#define _mm_cmpnlt_ss(a, b) simde_mm_cmpnlt_ss((a), (b))877
-#endif878
-879
-SIMDE_FUNCTION_ATTRIBUTES880
-simde__m128 simde_mm_cmpord_ps(simde__m128 a, simde__m128 b)881
-{882
-#if defined(SIMDE_X86_SSE_NATIVE)883
- return _mm_cmpord_ps(a, b);884
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)885
- return wasm_v128_and(wasm_f32x4_eq(a, a), wasm_f32x4_eq(b, b));886
-#else887
- simde__m128_private r_, a_ = simde__m128_to_private(a),888
- b_ = simde__m128_to_private(b);889
-890
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)891
- /* Note: NEON does not have ordered compare builtin892
- Need to compare a eq a and b eq b to check for NaN893
- Do AND of results to get final */894
- uint32x4_t ceqaa = vceqq_f32(a_.neon_f32, a_.neon_f32);895
- uint32x4_t ceqbb = vceqq_f32(b_.neon_f32, b_.neon_f32);896
- r_.neon_u32 = vandq_u32(ceqaa, ceqbb);897
-#elif defined(simde_math_isnanf)898
- SIMDE_VECTORIZE899
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {900
- r_.u32[i] = (simde_math_isnanf(a_.f32[i]) ||901
- simde_math_isnanf(b_.f32[i]))902
- ? UINT32_C(0)903
- : ~UINT32_C(0);904
- }905
-#else906
- HEDLEY_UNREACHABLE();907
-#endif908
-909
- return simde__m128_from_private(r_);910
-#endif911
-}912
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)913
-#define _mm_cmpord_ps(a, b) simde_mm_cmpord_ps((a), (b))914
-#endif915
-916
-SIMDE_FUNCTION_ATTRIBUTES917
-simde__m128 simde_mm_cmpunord_ps(simde__m128 a, simde__m128 b)918
-{919
-#if defined(SIMDE_X86_SSE_NATIVE)920
- return _mm_cmpunord_ps(a, b);921
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)922
- return wasm_v128_or(wasm_f32x4_ne(a, a), wasm_f32x4_ne(b, b));923
-#else924
- simde__m128_private r_, a_ = simde__m128_to_private(a),925
- b_ = simde__m128_to_private(b);926
-927
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)928
- uint32x4_t ceqaa = vceqq_f32(a_.neon_f32, a_.neon_f32);929
- uint32x4_t ceqbb = vceqq_f32(b_.neon_f32, b_.neon_f32);930
- r_.neon_u32 = vmvnq_u32(vandq_u32(ceqaa, ceqbb));931
-#elif defined(simde_math_isnanf)932
- SIMDE_VECTORIZE933
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {934
- r_.u32[i] = (simde_math_isnanf(a_.f32[i]) ||935
- simde_math_isnanf(b_.f32[i]))936
- ? ~UINT32_C(0)937
- : UINT32_C(0);938
- }939
-#else940
- HEDLEY_UNREACHABLE();941
-#endif942
-943
- return simde__m128_from_private(r_);944
-#endif945
-}946
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)947
-#define _mm_cmpunord_ps(a, b) simde_mm_cmpunord_ps((a), (b))948
-#endif949
-950
-SIMDE_FUNCTION_ATTRIBUTES951
-simde__m128 simde_mm_cmpunord_ss(simde__m128 a, simde__m128 b)952
-{953
-#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)954
- return _mm_cmpunord_ss(a, b);955
-#elif defined(SIMDE_ASSUME_VECTORIZATION)956
- return simde_mm_move_ss(a, simde_mm_cmpunord_ps(a, b));957
-#else958
- simde__m128_private r_, a_ = simde__m128_to_private(a),959
- b_ = simde__m128_to_private(b);960
-961
-#if defined(simde_math_isnanf)962
- r_.u32[0] =963
- (simde_math_isnanf(a_.f32[0]) || simde_math_isnanf(b_.f32[0]))964
- ? ~UINT32_C(0)965
- : UINT32_C(0);966
- SIMDE_VECTORIZE967
- for (size_t i = 1; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {968
- r_.u32[i] = a_.u32[i];969
- }970
-#else971
- HEDLEY_UNREACHABLE();972
-#endif973
-974
- return simde__m128_from_private(r_);975
-#endif976
-}977
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)978
-#define _mm_cmpunord_ss(a, b) simde_mm_cmpunord_ss((a), (b))979
-#endif980
-981
-SIMDE_FUNCTION_ATTRIBUTES982
-int simde_mm_comieq_ss(simde__m128 a, simde__m128 b)983
-{984
-#if defined(SIMDE_X86_SSE_NATIVE)985
- return _mm_comieq_ss(a, b);986
-#else987
- simde__m128_private a_ = simde__m128_to_private(a),988
- b_ = simde__m128_to_private(b);989
-990
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)991
- uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);992
- uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);993
- uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));994
- uint32x4_t a_eq_b = vceqq_f32(a_.neon_f32, b_.neon_f32);995
- return !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_eq_b), 0) != 0);996
-#else997
- return a_.f32[0] == b_.f32[0];998
-#endif999
-#endif1000
-}1001
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1002
-#define _mm_comieq_ss(a, b) simde_mm_comieq_ss((a), (b))1003
-#endif1004
-1005
-SIMDE_FUNCTION_ATTRIBUTES1006
-int simde_mm_comige_ss(simde__m128 a, simde__m128 b)1007
-{1008
-#if defined(SIMDE_X86_SSE_NATIVE)1009
- return _mm_comige_ss(a, b);1010
-#else1011
- simde__m128_private a_ = simde__m128_to_private(a),1012
- b_ = simde__m128_to_private(b);1013
-1014
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1015
- uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1016
- uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1017
- uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);1018
- uint32x4_t a_ge_b = vcgeq_f32(a_.neon_f32, b_.neon_f32);1019
- return !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_ge_b), 0) != 0);1020
-#else1021
- return a_.f32[0] >= b_.f32[0];1022
-#endif1023
-#endif1024
-}1025
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1026
-#define _mm_comige_ss(a, b) simde_mm_comige_ss((a), (b))1027
-#endif1028
-1029
-SIMDE_FUNCTION_ATTRIBUTES1030
-int simde_mm_comigt_ss(simde__m128 a, simde__m128 b)1031
-{1032
-#if defined(SIMDE_X86_SSE_NATIVE)1033
- return _mm_comigt_ss(a, b);1034
-#else1035
- simde__m128_private a_ = simde__m128_to_private(a),1036
- b_ = simde__m128_to_private(b);1037
-1038
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1039
- uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1040
- uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1041
- uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);1042
- uint32x4_t a_gt_b = vcgtq_f32(a_.neon_f32, b_.neon_f32);1043
- return !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_gt_b), 0) != 0);1044
-#else1045
- return a_.f32[0] > b_.f32[0];1046
-#endif1047
-#endif1048
-}1049
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1050
-#define _mm_comigt_ss(a, b) simde_mm_comigt_ss((a), (b))1051
-#endif1052
-1053
-SIMDE_FUNCTION_ATTRIBUTES1054
-int simde_mm_comile_ss(simde__m128 a, simde__m128 b)1055
-{1056
-#if defined(SIMDE_X86_SSE_NATIVE)1057
- return _mm_comile_ss(a, b);1058
-#else1059
- simde__m128_private a_ = simde__m128_to_private(a),1060
- b_ = simde__m128_to_private(b);1061
-1062
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1063
- uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1064
- uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1065
- uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));1066
- uint32x4_t a_le_b = vcleq_f32(a_.neon_f32, b_.neon_f32);1067
- return !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_le_b), 0) != 0);1068
-#else1069
- return a_.f32[0] <= b_.f32[0];1070
-#endif1071
-#endif1072
-}1073
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1074
-#define _mm_comile_ss(a, b) simde_mm_comile_ss((a), (b))1075
-#endif1076
-1077
-SIMDE_FUNCTION_ATTRIBUTES1078
-int simde_mm_comilt_ss(simde__m128 a, simde__m128 b)1079
-{1080
-#if defined(SIMDE_X86_SSE_NATIVE)1081
- return _mm_comilt_ss(a, b);1082
-#else1083
- simde__m128_private a_ = simde__m128_to_private(a),1084
- b_ = simde__m128_to_private(b);1085
-1086
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1087
- uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1088
- uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1089
- uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));1090
- uint32x4_t a_lt_b = vcltq_f32(a_.neon_f32, b_.neon_f32);1091
- return !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_lt_b), 0) != 0);1092
-#else1093
- return a_.f32[0] < b_.f32[0];1094
-#endif1095
-#endif1096
-}1097
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1098
-#define _mm_comilt_ss(a, b) simde_mm_comilt_ss((a), (b))1099
-#endif1100
-1101
-SIMDE_FUNCTION_ATTRIBUTES1102
-int simde_mm_comineq_ss(simde__m128 a, simde__m128 b)1103
-{1104
-#if defined(SIMDE_X86_SSE_NATIVE)1105
- return _mm_comineq_ss(a, b);1106
-#else1107
- simde__m128_private a_ = simde__m128_to_private(a),1108
- b_ = simde__m128_to_private(b);1109
-1110
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1111
- uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1112
- uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1113
- uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);1114
- uint32x4_t a_neq_b = vmvnq_u32(vceqq_f32(a_.neon_f32, b_.neon_f32));1115
- return !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_neq_b), 0) != 0);1116
-#else1117
- return a_.f32[0] != b_.f32[0];1118
-#endif1119
-#endif1120
-}1121
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1122
-#define _mm_comineq_ss(a, b) simde_mm_comineq_ss((a), (b))1123
-#endif1124
-1125
-SIMDE_FUNCTION_ATTRIBUTES1126
-simde__m128 simde_mm_cvt_pi2ps(simde__m128 a, simde__m64 b)1127
-{1128
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1129
- return _mm_cvt_pi2ps(a, b);1130
-#else1131
- simde__m128_private r_, a_ = simde__m128_to_private(a);1132
- simde__m64_private b_ = simde__m64_to_private(b);1133
-1134
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1135
- r_.neon_f32 = vcombine_f32(vcvt_f32_s32(b_.neon_i32),1136
- vget_high_f32(a_.neon_f32));1137
-#elif defined(SIMDE_CONVERT_VECTOR_)1138
- SIMDE_CONVERT_VECTOR_(r_.m64_private[0].f32, b_.i32);1139
- r_.m64_private[1] = a_.m64_private[1];1140
-1141
-#else1142
- r_.f32[0] = (simde_float32)b_.i32[0];1143
- r_.f32[1] = (simde_float32)b_.i32[1];1144
- r_.i32[2] = a_.i32[2];1145
- r_.i32[3] = a_.i32[3];1146
-#endif1147
-1148
- return simde__m128_from_private(r_);1149
-#endif1150
-}1151
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1152
-#define _mm_cvt_pi2ps(a, b) simde_mm_cvt_pi2ps((a), b)1153
-#endif1154
-1155
-SIMDE_FUNCTION_ATTRIBUTES1156
-simde__m64 simde_mm_cvt_ps2pi(simde__m128 a)1157
-{1158
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1159
- return _mm_cvt_ps2pi(a);1160
-#else1161
- simde__m64_private r_;1162
- simde__m128_private a_ = simde__m128_to_private(a);1163
-1164
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1165
- r_.neon_i32 = vcvt_s32_f32(vget_low_f32(a_.neon_f32));1166
-#elif defined(SIMDE_CONVERT_VECTOR_) && !defined(__clang__)1167
- SIMDE_CONVERT_VECTOR_(r_.i32, a_.m64_private[0].f32);1168
-#else1169
- SIMDE_VECTORIZE1170
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1171
- r_.i32[i] = HEDLEY_STATIC_CAST(int32_t, a_.f32[i]);1172
- }1173
-#endif1174
-1175
- return simde__m64_from_private(r_);1176
-#endif1177
-}1178
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1179
-#define _mm_cvt_ps2pi(a) simde_mm_cvt_ps2pi((a))1180
-#endif1181
-1182
-SIMDE_FUNCTION_ATTRIBUTES1183
-simde__m128 simde_mm_cvt_si2ss(simde__m128 a, int32_t b)1184
-{1185
-#if defined(SIMDE_X86_SSE_NATIVE)1186
- return _mm_cvt_si2ss(a, b);1187
-#else1188
- simde__m128_private r_, a_ = simde__m128_to_private(a);1189
-1190
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1191
- r_.neon_f32 = vsetq_lane_f32((float)b, a_.neon_f32, 0);1192
-#else1193
- r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, b);1194
- r_.i32[1] = a_.i32[1];1195
- r_.i32[2] = a_.i32[2];1196
- r_.i32[3] = a_.i32[3];1197
-#endif1198
-1199
- return simde__m128_from_private(r_);1200
-#endif1201
-}1202
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1203
-#define _mm_cvt_si2ss(a, b) simde_mm_cvt_si2ss((a), b)1204
-#endif1205
-1206
-SIMDE_FUNCTION_ATTRIBUTES1207
-int32_t simde_mm_cvt_ss2si(simde__m128 a)1208
-{1209
-#if defined(SIMDE_X86_SSE_NATIVE)1210
- return _mm_cvt_ss2si(a);1211
-#else1212
- simde__m128_private a_ = simde__m128_to_private(a);1213
-1214
-#if defined(SIMDE_ARM_NEON_A32V8_NATIVE) && !defined(SIMDE_BUG_GCC_95399)1215
- return vgetq_lane_s32(vcvtnq_s32_f32(a_.neon_f32), 0);1216
-#elif defined(simde_math_nearbyintf)1217
- return SIMDE_CONVERT_FTOI(int32_t, simde_math_nearbyintf(a_.f32[0]));1218
-#else1219
- HEDLEY_UNREACHABLE();1220
-#endif1221
-#endif1222
-}1223
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1224
-#define _mm_cvt_ss2si(a) simde_mm_cvt_ss2si((a))1225
-#endif1226
-1227
-SIMDE_FUNCTION_ATTRIBUTES1228
-simde__m128 simde_mm_cvtpi16_ps(simde__m64 a)1229
-{1230
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1231
- return _mm_cvtpi16_ps(a);1232
-#else1233
- simde__m128_private r_;1234
- simde__m64_private a_ = simde__m64_to_private(a);1235
-1236
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE) && 0 /* TODO */1237
- r_.neon_f32 = vmovl_s16(1238
- vget_low_s16(vuzp1q_s16(a_.neon_i16, vmovq_n_s16(0))));1239
-#elif defined(SIMDE_CONVERT_VECTOR_)1240
- SIMDE_CONVERT_VECTOR_(r_.f32, a_.i16);1241
-#else1242
- SIMDE_VECTORIZE1243
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1244
- simde_float32 v = a_.i16[i];1245
- r_.f32[i] = v;1246
- }1247
-#endif1248
-1249
- return simde__m128_from_private(r_);1250
-#endif1251
-}1252
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1253
-#define _mm_cvtpi16_ps(a) simde_mm_cvtpi16_ps(a)1254
-#endif1255
-1256
-SIMDE_FUNCTION_ATTRIBUTES1257
-simde__m128 simde_mm_cvtpi32_ps(simde__m128 a, simde__m64 b)1258
-{1259
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1260
- return _mm_cvtpi32_ps(a, b);1261
-#else1262
- simde__m128_private r_, a_ = simde__m128_to_private(a);1263
- simde__m64_private b_ = simde__m64_to_private(b);1264
-1265
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1266
- r_.neon_f32 = vcombine_f32(vcvt_f32_s32(b_.neon_i32),1267
- vget_high_f32(a_.neon_f32));1268
-#elif defined(SIMDE_CONVERT_VECTOR_)1269
- SIMDE_CONVERT_VECTOR_(r_.m64_private[0].f32, b_.i32);1270
- r_.m64_private[1] = a_.m64_private[1];1271
-#else1272
- r_.f32[0] = (simde_float32)b_.i32[0];1273
- r_.f32[1] = (simde_float32)b_.i32[1];1274
- r_.i32[2] = a_.i32[2];1275
- r_.i32[3] = a_.i32[3];1276
-#endif1277
-1278
- return simde__m128_from_private(r_);1279
-#endif1280
-}1281
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1282
-#define _mm_cvtpi32_ps(a, b) simde_mm_cvtpi32_ps((a), b)1283
-#endif1284
-1285
-SIMDE_FUNCTION_ATTRIBUTES1286
-simde__m128 simde_mm_cvtpi32x2_ps(simde__m64 a, simde__m64 b)1287
-{1288
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1289
- return _mm_cvtpi32x2_ps(a, b);1290
-#else1291
- simde__m128_private r_;1292
- simde__m64_private a_ = simde__m64_to_private(a),1293
- b_ = simde__m64_to_private(b);1294
-1295
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1296
- r_.neon_f32 = vcvtq_f32_s32(vcombine_s32(a_.neon_i32, b_.neon_i32));1297
-#elif defined(SIMDE_CONVERT_VECTOR_)1298
- SIMDE_CONVERT_VECTOR_(r_.m64_private[0].f32, a_.i32);1299
- SIMDE_CONVERT_VECTOR_(r_.m64_private[1].f32, b_.i32);1300
-#else1301
- r_.f32[0] = (simde_float32)a_.i32[0];1302
- r_.f32[1] = (simde_float32)a_.i32[1];1303
- r_.f32[2] = (simde_float32)b_.i32[0];1304
- r_.f32[3] = (simde_float32)b_.i32[1];1305
-#endif1306
-1307
- return simde__m128_from_private(r_);1308
-#endif1309
-}1310
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1311
-#define _mm_cvtpi32x2_ps(a, b) simde_mm_cvtpi32x2_ps(a, b)1312
-#endif1313
-1314
-SIMDE_FUNCTION_ATTRIBUTES1315
-simde__m128 simde_mm_cvtpi8_ps(simde__m64 a)1316
-{1317
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1318
- return _mm_cvtpi8_ps(a);1319
-#else1320
- simde__m128_private r_;1321
- simde__m64_private a_ = simde__m64_to_private(a);1322
-1323
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1324
- r_.neon_f32 =1325
- vcvtq_f32_s32(vmovl_s16(vget_low_s16(vmovl_s8(a_.neon_i8))));1326
-#else1327
- r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, a_.i8[0]);1328
- r_.f32[1] = HEDLEY_STATIC_CAST(simde_float32, a_.i8[1]);1329
- r_.f32[2] = HEDLEY_STATIC_CAST(simde_float32, a_.i8[2]);1330
- r_.f32[3] = HEDLEY_STATIC_CAST(simde_float32, a_.i8[3]);1331
-#endif1332
-1333
- return simde__m128_from_private(r_);1334
-#endif1335
-}1336
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1337
-#define _mm_cvtpi8_ps(a) simde_mm_cvtpi8_ps(a)1338
-#endif1339
-1340
-SIMDE_FUNCTION_ATTRIBUTES1341
-simde__m64 simde_mm_cvtps_pi16(simde__m128 a)1342
-{1343
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1344
- return _mm_cvtps_pi16(a);1345
-#else1346
- simde__m64_private r_;1347
- simde__m128_private a_ = simde__m128_to_private(a);1348
-1349
-#if defined(SIMDE_CONVERT_VECTOR_)1350
- SIMDE_CONVERT_VECTOR_(r_.i16, a_.f32);1351
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1352
- r_.neon_i16 = vmovn_s32(vcvtq_s32_f32(a_.neon_f32));1353
-#else1354
- SIMDE_VECTORIZE1355
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1356
- r_.i16[i] = SIMDE_CONVERT_FTOI(int16_t, a_.f32[i]);1357
- }1358
-#endif1359
-1360
- return simde__m64_from_private(r_);1361
-#endif1362
-}1363
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1364
-#define _mm_cvtps_pi16(a) simde_mm_cvtps_pi16((a))1365
-#endif1366
-1367
-SIMDE_FUNCTION_ATTRIBUTES1368
-simde__m64 simde_mm_cvtps_pi32(simde__m128 a)1369
-{1370
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1371
- return _mm_cvtps_pi32(a);1372
-#else1373
- simde__m64_private r_;1374
- simde__m128_private a_ = simde__m128_to_private(a);1375
-1376
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1377
- r_.neon_i32 = vcvt_s32_f32(vget_low_f32(a_.neon_f32));1378
-#elif defined(SIMDE_CONVERT_VECTOR_)1379
- SIMDE_CONVERT_VECTOR_(r_.i32, a_.m64_private[0].f32);1380
-#else1381
- SIMDE_VECTORIZE1382
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1383
- r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, a_.f32[i]);1384
- }1385
-#endif1386
-1387
- return simde__m64_from_private(r_);1388
-#endif1389
-}1390
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1391
-#define _mm_cvtps_pi32(a) simde_mm_cvtps_pi32((a))1392
-#endif1393
-1394
-SIMDE_FUNCTION_ATTRIBUTES1395
-simde__m64 simde_mm_cvtps_pi8(simde__m128 a)1396
-{1397
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1398
- return _mm_cvtps_pi8(a);1399
-#else1400
- simde__m64_private r_;1401
- simde__m128_private a_ = simde__m128_to_private(a);1402
-1403
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1404
- int16x4_t b = vmovn_s32(vcvtq_s32_f32(a_.neon_f32));1405
- int16x8_t c = vcombine_s16(b, vmov_n_s16(0));1406
- r_.neon_i8 = vmovn_s16(c);1407
-#else1408
- SIMDE_VECTORIZE1409
- for (size_t i = 0; i < (sizeof(a_.f32) / sizeof(a_.f32[0])); i++) {1410
- r_.i8[i] = SIMDE_CONVERT_FTOI(int8_t, a_.f32[i]);1411
- }1412
- /* Note: the upper half is undefined */1413
-#endif1414
-1415
- return simde__m64_from_private(r_);1416
-#endif1417
-}1418
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1419
-#define _mm_cvtps_pi8(a) simde_mm_cvtps_pi8((a))1420
-#endif1421
-1422
-SIMDE_FUNCTION_ATTRIBUTES1423
-simde__m128 simde_mm_cvtpu16_ps(simde__m64 a)1424
-{1425
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1426
- return _mm_cvtpu16_ps(a);1427
-#else1428
- simde__m128_private r_;1429
- simde__m64_private a_ = simde__m64_to_private(a);1430
-1431
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1432
- r_.neon_f32 = vcvtq_f32_u32(vmovl_u16(a_.neon_u16));1433
-#elif defined(SIMDE_CONVERT_VECTOR_)1434
- SIMDE_CONVERT_VECTOR_(r_.f32, a_.u16);1435
-#else1436
- SIMDE_VECTORIZE1437
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1438
- r_.f32[i] = (simde_float32)a_.u16[i];1439
- }1440
-#endif1441
-1442
- return simde__m128_from_private(r_);1443
-#endif1444
-}1445
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1446
-#define _mm_cvtpu16_ps(a) simde_mm_cvtpu16_ps(a)1447
-#endif1448
-1449
-SIMDE_FUNCTION_ATTRIBUTES1450
-simde__m128 simde_mm_cvtpu8_ps(simde__m64 a)1451
-{1452
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1453
- return _mm_cvtpu8_ps(a);1454
-#else1455
- simde__m128_private r_;1456
- simde__m64_private a_ = simde__m64_to_private(a);1457
-1458
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1459
- r_.neon_f32 =1460
- vcvtq_f32_u32(vmovl_u16(vget_low_u16(vmovl_u8(a_.neon_u8))));1461
-#else1462
- SIMDE_VECTORIZE1463
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1464
- r_.f32[i] = HEDLEY_STATIC_CAST(simde_float32, a_.u8[i]);1465
- }1466
-#endif1467
-1468
- return simde__m128_from_private(r_);1469
-#endif1470
-}1471
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1472
-#define _mm_cvtpu8_ps(a) simde_mm_cvtpu8_ps(a)1473
-#endif1474
-1475
-SIMDE_FUNCTION_ATTRIBUTES1476
-simde__m128 simde_mm_cvtsi32_ss(simde__m128 a, int32_t b)1477
-{1478
-#if defined(SIMDE_X86_SSE_NATIVE)1479
- return _mm_cvtsi32_ss(a, b);1480
-#else1481
- simde__m128_private r_;1482
- simde__m128_private a_ = simde__m128_to_private(a);1483
-1484
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1485
- r_.neon_f32 = vsetq_lane_f32((simde_float32)b, a_.neon_f32, 0);1486
-#else1487
- r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, b);1488
- SIMDE_VECTORIZE1489
- for (size_t i = 1; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1490
- r_.i32[i] = a_.i32[i];1491
- }1492
-#endif1493
-1494
- return simde__m128_from_private(r_);1495
-#endif1496
-}1497
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1498
-#define _mm_cvtsi32_ss(a, b) simde_mm_cvtsi32_ss((a), b)1499
-#endif1500
-1501
-SIMDE_FUNCTION_ATTRIBUTES1502
-simde__m128 simde_mm_cvtsi64_ss(simde__m128 a, int64_t b)1503
-{1504
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_ARCH_AMD64)1505
-#if !defined(__PGI)1506
- return _mm_cvtsi64_ss(a, b);1507
-#else1508
- return _mm_cvtsi64x_ss(a, b);1509
-#endif1510
-#else1511
- simde__m128_private r_;1512
- simde__m128_private a_ = simde__m128_to_private(a);1513
-1514
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1515
- r_.neon_f32 = vsetq_lane_f32((simde_float32)b, a_.neon_f32, 0);1516
-#else1517
- r_ = a_;1518
- r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, b);1519
-#endif1520
-1521
- return simde__m128_from_private(r_);1522
-#endif1523
-}1524
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1525
-#define _mm_cvtsi64_ss(a, b) simde_mm_cvtsi64_ss((a), b)1526
-#endif1527
-1528
-SIMDE_FUNCTION_ATTRIBUTES1529
-simde_float32 simde_mm_cvtss_f32(simde__m128 a)1530
-{1531
-#if defined(SIMDE_X86_SSE_NATIVE)1532
- return _mm_cvtss_f32(a);1533
-#else1534
- simde__m128_private a_ = simde__m128_to_private(a);1535
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1536
- return vgetq_lane_f32(a_.neon_f32, 0);1537
-#else1538
- return a_.f32[0];1539
-#endif1540
-#endif1541
-}1542
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1543
-#define _mm_cvtss_f32(a) simde_mm_cvtss_f32((a))1544
-#endif1545
-1546
-SIMDE_FUNCTION_ATTRIBUTES1547
-int32_t simde_mm_cvtss_si32(simde__m128 a)1548
-{1549
- return simde_mm_cvt_ss2si(a);1550
-}1551
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1552
-#define _mm_cvtss_si32(a) simde_mm_cvtss_si32((a))1553
-#endif1554
-1555
-SIMDE_FUNCTION_ATTRIBUTES1556
-int64_t simde_mm_cvtss_si64(simde__m128 a)1557
-{1558
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_ARCH_AMD64)1559
-#if !defined(__PGI)1560
- return _mm_cvtss_si64(a);1561
-#else1562
- return _mm_cvtss_si64x(a);1563
-#endif1564
-#else1565
- simde__m128_private a_ = simde__m128_to_private(a);1566
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1567
- return SIMDE_CONVERT_FTOI(int64_t, vgetq_lane_f32(a_.neon_f32, 0));1568
-#else1569
- return SIMDE_CONVERT_FTOI(int64_t, a_.f32[0]);1570
-#endif1571
-#endif1572
-}1573
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1574
-#define _mm_cvtss_si64(a) simde_mm_cvtss_si64((a))1575
-#endif1576
-1577
-SIMDE_FUNCTION_ATTRIBUTES1578
-simde__m64 simde_mm_cvtt_ps2pi(simde__m128 a)1579
-{1580
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1581
- return _mm_cvtt_ps2pi(a);1582
-#else1583
- simde__m64_private r_;1584
- simde__m128_private a_ = simde__m128_to_private(a);1585
-1586
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1587
- r_.neon_i32 = vcvt_s32_f32(vget_low_f32(a_.neon_f32));1588
-#elif defined(SIMDE_CONVERT_VECTOR_)1589
- SIMDE_CONVERT_VECTOR_(r_.i32, a_.m64_private[0].f32);1590
-#else1591
- SIMDE_VECTORIZE1592
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1593
- r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, a_.f32[i]);1594
- }1595
-#endif1596
-1597
- return simde__m64_from_private(r_);1598
-#endif1599
-}1600
-#define simde_mm_cvttps_pi32(a) simde_mm_cvtt_ps2pi(a)1601
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1602
-#define _mm_cvtt_ps2pi(a) simde_mm_cvtt_ps2pi((a))1603
-#define _mm_cvttps_pi32(a) simde_mm_cvttps_pi32((a))1604
-#endif1605
-1606
-SIMDE_FUNCTION_ATTRIBUTES1607
-int32_t simde_mm_cvtt_ss2si(simde__m128 a)1608
-{1609
-#if defined(SIMDE_X86_SSE_NATIVE)1610
- return _mm_cvtt_ss2si(a);1611
-#else1612
- simde__m128_private a_ = simde__m128_to_private(a);1613
-1614
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1615
- return SIMDE_CONVERT_FTOI(int32_t, vgetq_lane_f32(a_.neon_f32, 0));1616
-#else1617
- return SIMDE_CONVERT_FTOI(int32_t, a_.f32[0]);1618
-#endif1619
-#endif1620
-}1621
-#define simde_mm_cvttss_si32(a) simde_mm_cvtt_ss2si((a))1622
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1623
-#define _mm_cvtt_ss2si(a) simde_mm_cvtt_ss2si((a))1624
-#define _mm_cvttss_si32(a) simde_mm_cvtt_ss2si((a))1625
-#endif1626
-1627
-SIMDE_FUNCTION_ATTRIBUTES1628
-int64_t simde_mm_cvttss_si64(simde__m128 a)1629
-{1630
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_ARCH_AMD64) && \1631
- !defined(_MSC_VER)1632
-#if defined(__PGI)1633
- return _mm_cvttss_si64x(a);1634
-#else1635
- return _mm_cvttss_si64(a);1636
-#endif1637
-#else1638
- simde__m128_private a_ = simde__m128_to_private(a);1639
-1640
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1641
- return SIMDE_CONVERT_FTOI(int64_t, vgetq_lane_f32(a_.neon_f32, 0));1642
-#else1643
- return SIMDE_CONVERT_FTOI(int64_t, a_.f32[0]);1644
-#endif1645
-#endif1646
-}1647
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1648
-#define _mm_cvttss_si64(a) simde_mm_cvttss_si64((a))1649
-#endif1650
-1651
-SIMDE_FUNCTION_ATTRIBUTES1652
-simde__m128 simde_mm_cmpord_ss(simde__m128 a, simde__m128 b)1653
-{1654
-#if defined(SIMDE_X86_SSE_NATIVE)1655
- return _mm_cmpord_ss(a, b);1656
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1657
- return simde_mm_move_ss(a, simde_mm_cmpord_ps(a, b));1658
-#else1659
- simde__m128_private r_, a_ = simde__m128_to_private(a);1660
-1661
-#if defined(simde_math_isnanf)1662
- r_.u32[0] = (simde_math_isnanf(simde_mm_cvtss_f32(a)) ||1663
- simde_math_isnanf(simde_mm_cvtss_f32(b)))1664
- ? UINT32_C(0)1665
- : ~UINT32_C(0);1666
- SIMDE_VECTORIZE1667
- for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1668
- r_.u32[i] = a_.u32[i];1669
- }1670
-#else1671
- HEDLEY_UNREACHABLE();1672
-#endif1673
-1674
- return simde__m128_from_private(r_);1675
-#endif1676
-}1677
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1678
-#define _mm_cmpord_ss(a, b) simde_mm_cmpord_ss((a), (b))1679
-#endif1680
-1681
-SIMDE_FUNCTION_ATTRIBUTES1682
-simde__m128 simde_mm_div_ps(simde__m128 a, simde__m128 b)1683
-{1684
-#if defined(SIMDE_X86_SSE_NATIVE)1685
- return _mm_div_ps(a, b);1686
-#else1687
- simde__m128_private r_, a_ = simde__m128_to_private(a),1688
- b_ = simde__m128_to_private(b);1689
-1690
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1691
- r_.neon_f32 = vdivq_f32(a_.neon_f32, b_.neon_f32);1692
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1693
- float32x4_t recip0 = vrecpeq_f32(b_.neon_f32);1694
- float32x4_t recip1 =1695
- vmulq_f32(recip0, vrecpsq_f32(recip0, b_.neon_f32));1696
- r_.neon_f32 = vmulq_f32(a_.neon_f32, recip1);1697
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1698
- r_.wasm_v128 = wasm_f32x4_div(a_.wasm_v128, b_.wasm_v128);1699
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1700
- r_.f32 = a_.f32 / b_.f32;1701
-#else1702
- SIMDE_VECTORIZE1703
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1704
- r_.f32[i] = a_.f32[i] / b_.f32[i];1705
- }1706
-#endif1707
-1708
- return simde__m128_from_private(r_);1709
-#endif1710
-}1711
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1712
-#define _mm_div_ps(a, b) simde_mm_div_ps((a), (b))1713
-#endif1714
-1715
-SIMDE_FUNCTION_ATTRIBUTES1716
-simde__m128 simde_mm_div_ss(simde__m128 a, simde__m128 b)1717
-{1718
-#if defined(SIMDE_X86_SSE_NATIVE)1719
- return _mm_div_ss(a, b);1720
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1721
- return simde_mm_move_ss(a, simde_mm_div_ps(a, b));1722
-#else1723
- simde__m128_private r_, a_ = simde__m128_to_private(a),1724
- b_ = simde__m128_to_private(b);1725
-1726
- r_.f32[0] = a_.f32[0] / b_.f32[0];1727
- SIMDE_VECTORIZE1728
- for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1729
- r_.f32[i] = a_.f32[i];1730
- }1731
-1732
- return simde__m128_from_private(r_);1733
-#endif1734
-}1735
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1736
-#define _mm_div_ss(a, b) simde_mm_div_ss((a), (b))1737
-#endif1738
-1739
-SIMDE_FUNCTION_ATTRIBUTES1740
-int16_t simde_mm_extract_pi16(simde__m64 a, const int imm8)1741
- SIMDE_REQUIRE_RANGE(imm8, 0, 3)1742
-{1743
- simde__m64_private a_ = simde__m64_to_private(a);1744
- return a_.i16[imm8];1745
-}1746
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \1747
- !defined(HEDLEY_PGI_VERSION)1748
-#if HEDLEY_HAS_WARNING("-Wvector-conversion")1749
-/* https://bugs.llvm.org/show_bug.cgi?id=44589 */1750
-#define simde_mm_extract_pi16(a, imm8) \1751
- (HEDLEY_DIAGNOSTIC_PUSH _Pragma( \1752
- "clang diagnostic ignored \"-Wvector-conversion\"") \1753
- HEDLEY_STATIC_CAST(int16_t, _mm_extract_pi16((a), (imm8))) \1754
- HEDLEY_DIAGNOSTIC_POP)1755
-#else1756
-#define simde_mm_extract_pi16(a, imm8) \1757
- HEDLEY_STATIC_CAST(int16_t, _mm_extract_pi16(a, imm8))1758
-#endif1759
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1760
-#define simde_mm_extract_pi16(a, imm8) \1761
- vget_lane_s16(simde__m64_to_private(a).neon_i16, imm8)1762
-#endif1763
-#define simde_m_pextrw(a, imm8) simde_mm_extract_pi16(a, imm8)1764
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1765
-#define _mm_extract_pi16(a, imm8) simde_mm_extract_pi16((a), (imm8))1766
-#endif1767
-1768
-enum {1769
-#if defined(SIMDE_X86_SSE_NATIVE)1770
- SIMDE_MM_ROUND_NEAREST = _MM_ROUND_NEAREST,1771
- SIMDE_MM_ROUND_DOWN = _MM_ROUND_DOWN,1772
- SIMDE_MM_ROUND_UP = _MM_ROUND_UP,1773
- SIMDE_MM_ROUND_TOWARD_ZERO = _MM_ROUND_TOWARD_ZERO1774
-#else1775
- SIMDE_MM_ROUND_NEAREST1776
-#if defined(FE_TONEAREST)1777
- = FE_TONEAREST1778
-#endif1779
- ,1780
-1781
- SIMDE_MM_ROUND_DOWN1782
-#if defined(FE_DOWNWARD)1783
- = FE_DOWNWARD1784
-#endif1785
- ,1786
-1787
- SIMDE_MM_ROUND_UP1788
-#if defined(FE_UPWARD)1789
- = FE_UPWARD1790
-#endif1791
- ,1792
-1793
- SIMDE_MM_ROUND_TOWARD_ZERO1794
-#if defined(FE_TOWARDZERO)1795
- = FE_TOWARDZERO1796
-#endif1797
-#endif1798
-};1799
-1800
-SIMDE_FUNCTION_ATTRIBUTES1801
-unsigned int SIMDE_MM_GET_ROUNDING_MODE(void)1802
-{1803
-#if defined(SIMDE_X86_SSE_NATIVE)1804
- return _MM_GET_ROUNDING_MODE();1805
-#elif defined(SIMDE_HAVE_FENV_H)1806
- return HEDLEY_STATIC_CAST(unsigned int, fegetround());1807
-#else1808
- HEDLEY_UNREACHABLE();1809
-#endif1810
-}1811
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1812
-#define _MM_GET_ROUNDING_MODE() SIMDE_MM_GET_ROUNDING_MODE()1813
-#endif1814
-1815
-SIMDE_FUNCTION_ATTRIBUTES1816
-void SIMDE_MM_SET_ROUNDING_MODE(unsigned int a)1817
-{1818
-#if defined(SIMDE_X86_SSE_NATIVE)1819
- _MM_SET_ROUNDING_MODE(a);1820
-#elif defined(SIMDE_HAVE_FENV_H)1821
- fesetround(HEDLEY_STATIC_CAST(int, a));1822
-#endif1823
-}1824
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1825
-#define _MM_SET_ROUNDING_MODE(a) SIMDE_MM_SET_ROUNDING_MODE(a)1826
-#endif1827
-1828
-SIMDE_FUNCTION_ATTRIBUTES1829
-simde__m64 simde_mm_insert_pi16(simde__m64 a, int16_t i, const int imm8)1830
- SIMDE_REQUIRE_RANGE(imm8, 0, 3)1831
-{1832
- simde__m64_private r_, a_ = simde__m64_to_private(a);1833
-1834
- r_.i64[0] = a_.i64[0];1835
- r_.i16[imm8] = i;1836
-1837
- return simde__m64_from_private(r_);1838
-}1839
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \1840
- !defined(__PGI)1841
-#if HEDLEY_HAS_WARNING("-Wvector-conversion")1842
-/* https://bugs.llvm.org/show_bug.cgi?id=44589 */1843
-#define ssimde_mm_insert_pi16(a, i, imm8) \1844
- (HEDLEY_DIAGNOSTIC_PUSH _Pragma( \1845
- "clang diagnostic ignored \"-Wvector-conversion\"")( \1846
- _mm_insert_pi16((a), (i), (imm8))) HEDLEY_DIAGNOSTIC_POP)1847
-#else1848
-#define simde_mm_insert_pi16(a, i, imm8) _mm_insert_pi16(a, i, imm8)1849
-#endif1850
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1851
-#define simde_mm_insert_pi16(a, i, imm8) \1852
- simde__m64_from_neon_i16( \1853
- vset_lane_s16((i), simde__m64_to_neon_i16(a), (imm8)))1854
-#endif1855
-#define simde_m_pinsrw(a, i, imm8) (simde_mm_insert_pi16(a, i, imm8))1856
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1857
-#define _mm_insert_pi16(a, i, imm8) simde_mm_insert_pi16(a, i, imm8)1858
-#define _m_pinsrw(a, i, imm8) simde_mm_insert_pi16(a, i, imm8)1859
-#endif1860
-1861
-SIMDE_FUNCTION_ATTRIBUTES1862
-simde__m1281863
-simde_mm_load_ps(simde_float32 const mem_addr[HEDLEY_ARRAY_PARAM(4)])1864
-{1865
- simde_assert_aligned(16, mem_addr);1866
-1867
-#if defined(SIMDE_X86_SSE_NATIVE)1868
- return _mm_load_ps(mem_addr);1869
-#else1870
- simde__m128_private r_;1871
-1872
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1873
- r_.neon_f32 = vld1q_f32(mem_addr);1874
-#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)1875
- r_.altivec_f32 = vec_vsx_ld(0, mem_addr);1876
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1877
- r_.altivec_f32 = vec_ld(0, mem_addr);1878
-#else1879
- r_ = *SIMDE_ALIGN_CAST(simde__m128_private const *, mem_addr);1880
-#endif1881
-1882
- return simde__m128_from_private(r_);1883
-#endif1884
-}1885
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1886
-#define _mm_load_ps(mem_addr) simde_mm_load_ps(mem_addr)1887
-#endif1888
-1889
-SIMDE_FUNCTION_ATTRIBUTES1890
-simde__m128 simde_mm_load_ps1(simde_float32 const *mem_addr)1891
-{1892
-#if defined(SIMDE_X86_SSE_NATIVE)1893
- return _mm_load_ps1(mem_addr);1894
-#else1895
- simde__m128_private r_;1896
-1897
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1898
- r_.neon_f32 = vld1q_dup_f32(mem_addr);1899
-#else1900
- r_ = simde__m128_to_private(simde_mm_set1_ps(*mem_addr));1901
-#endif1902
-1903
- return simde__m128_from_private(r_);1904
-#endif1905
-}1906
-#define simde_mm_load1_ps(mem_addr) simde_mm_load_ps1(mem_addr)1907
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1908
-#define _mm_load_ps1(mem_addr) simde_mm_load_ps1(mem_addr)1909
-#define _mm_load1_ps(mem_addr) simde_mm_load_ps1(mem_addr)1910
-#endif1911
-1912
-SIMDE_FUNCTION_ATTRIBUTES1913
-simde__m128 simde_mm_load_ss(simde_float32 const *mem_addr)1914
-{1915
-#if defined(SIMDE_X86_SSE_NATIVE)1916
- return _mm_load_ss(mem_addr);1917
-#else1918
- simde__m128_private r_;1919
-1920
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1921
- r_.neon_f32 = vsetq_lane_f32(*mem_addr, vdupq_n_f32(0), 0);1922
-#else1923
- r_.f32[0] = *mem_addr;1924
- r_.i32[1] = 0;1925
- r_.i32[2] = 0;1926
- r_.i32[3] = 0;1927
-#endif1928
-1929
- return simde__m128_from_private(r_);1930
-#endif1931
-}1932
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1933
-#define _mm_load_ss(mem_addr) simde_mm_load_ss(mem_addr)1934
-#endif1935
-1936
-SIMDE_FUNCTION_ATTRIBUTES1937
-simde__m128 simde_mm_loadh_pi(simde__m128 a, simde__m64 const *mem_addr)1938
-{1939
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1940
- return _mm_loadh_pi(a,1941
- HEDLEY_REINTERPRET_CAST(__m64 const *, mem_addr));1942
-#else1943
- simde__m128_private r_, a_ = simde__m128_to_private(a);1944
-1945
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1946
- r_.neon_f32 = vcombine_f32(1947
- vget_low_f32(a_.neon_f32),1948
- vld1_f32(HEDLEY_REINTERPRET_CAST(const float32_t *, mem_addr)));1949
-#else1950
- simde__m64_private b_ =1951
- *HEDLEY_REINTERPRET_CAST(simde__m64_private const *, mem_addr);1952
- r_.f32[0] = a_.f32[0];1953
- r_.f32[1] = a_.f32[1];1954
- r_.f32[2] = b_.f32[0];1955
- r_.f32[3] = b_.f32[1];1956
-#endif1957
-1958
- return simde__m128_from_private(r_);1959
-#endif1960
-}1961
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1962
-#define _mm_loadh_pi(a, mem_addr) \1963
- simde_mm_loadh_pi((a), (simde__m64 const *)(mem_addr))1964
-#endif1965
-1966
-/* The SSE documentation says that there are no alignment requirements1967
- for mem_addr. Unfortunately they used the __m64 type for the argument1968
- which is supposed to be 8-byte aligned, so some compilers (like clang1969
- with -Wcast-align) will generate a warning if you try to cast, say,1970
- a simde_float32* to a simde__m64* for this function.1971
-1972
- I think the choice of argument type is unfortunate, but I do think we1973
- need to stick to it here. If there is demand I can always add something1974
- like simde_x_mm_loadl_f32(simde__m128, simde_float32 mem_addr[2]) */1975
-SIMDE_FUNCTION_ATTRIBUTES1976
-simde__m128 simde_mm_loadl_pi(simde__m128 a, simde__m64 const *mem_addr)1977
-{1978
-#if defined(SIMDE_X86_SSE_NATIVE)1979
- return _mm_loadl_pi(a,1980
- HEDLEY_REINTERPRET_CAST(__m64 const *, mem_addr));1981
-#else1982
- simde__m128_private r_, a_ = simde__m128_to_private(a);1983
-1984
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1985
- r_.neon_f32 = vcombine_f32(1986
- vld1_f32(HEDLEY_REINTERPRET_CAST(const float32_t *, mem_addr)),1987
- vget_high_f32(a_.neon_f32));1988
-#else1989
- simde__m64_private b_;1990
- simde_memcpy(&b_, mem_addr, sizeof(b_));1991
- r_.i32[0] = b_.i32[0];1992
- r_.i32[1] = b_.i32[1];1993
- r_.i32[2] = a_.i32[2];1994
- r_.i32[3] = a_.i32[3];1995
-#endif1996
-1997
- return simde__m128_from_private(r_);1998
-#endif1999
-}2000
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2001
-#define _mm_loadl_pi(a, mem_addr) \2002
- simde_mm_loadl_pi((a), (simde__m64 const *)(mem_addr))2003
-#endif2004
-2005
-SIMDE_FUNCTION_ATTRIBUTES2006
-simde__m1282007
-simde_mm_loadr_ps(simde_float32 const mem_addr[HEDLEY_ARRAY_PARAM(4)])2008
-{2009
- simde_assert_aligned(16, mem_addr);2010
-2011
-#if defined(SIMDE_X86_SSE_NATIVE)2012
- return _mm_loadr_ps(mem_addr);2013
-#else2014
- simde__m128_private r_,2015
- v_ = simde__m128_to_private(simde_mm_load_ps(mem_addr));2016
-2017
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2018
- r_.neon_f32 = vrev64q_f32(v_.neon_f32);2019
- r_.neon_f32 = vextq_f32(r_.neon_f32, r_.neon_f32, 2);2020
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2021
- r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, v_.f32, v_.f32, 3, 2, 1, 0);2022
-#else2023
- r_.f32[0] = v_.f32[3];2024
- r_.f32[1] = v_.f32[2];2025
- r_.f32[2] = v_.f32[1];2026
- r_.f32[3] = v_.f32[0];2027
-#endif2028
-2029
- return simde__m128_from_private(r_);2030
-#endif2031
-}2032
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2033
-#define _mm_loadr_ps(mem_addr) simde_mm_loadr_ps(mem_addr)2034
-#endif2035
-2036
-SIMDE_FUNCTION_ATTRIBUTES2037
-simde__m1282038
-simde_mm_loadu_ps(simde_float32 const mem_addr[HEDLEY_ARRAY_PARAM(4)])2039
-{2040
-#if defined(SIMDE_X86_SSE_NATIVE)2041
- return _mm_loadu_ps(mem_addr);2042
-#else2043
- simde__m128_private r_;2044
-2045
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2046
- r_.neon_f32 =2047
- vld1q_f32(HEDLEY_REINTERPRET_CAST(const float32_t *, mem_addr));2048
-#else2049
- r_.f32[0] = mem_addr[0];2050
- r_.f32[1] = mem_addr[1];2051
- r_.f32[2] = mem_addr[2];2052
- r_.f32[3] = mem_addr[3];2053
-#endif2054
-2055
- return simde__m128_from_private(r_);2056
-#endif2057
-}2058
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2059
-#define _mm_loadu_ps(mem_addr) simde_mm_loadu_ps(mem_addr)2060
-#endif2061
-2062
-SIMDE_FUNCTION_ATTRIBUTES2063
-void simde_mm_maskmove_si64(simde__m64 a, simde__m64 mask, int8_t *mem_addr)2064
-{2065
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2066
- _mm_maskmove_si64(a, mask, HEDLEY_REINTERPRET_CAST(char *, mem_addr));2067
-#else2068
- simde__m64_private a_ = simde__m64_to_private(a),2069
- mask_ = simde__m64_to_private(mask);2070
-2071
- SIMDE_VECTORIZE2072
- for (size_t i = 0; i < (sizeof(a_.i8) / sizeof(a_.i8[0])); i++)2073
- if (mask_.i8[i] < 0)2074
- mem_addr[i] = a_.i8[i];2075
-#endif2076
-}2077
-#define simde_m_maskmovq(a, mask, mem_addr) \2078
- simde_mm_maskmove_si64(a, mask, mem_addr)2079
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2080
-#define _mm_maskmove_si64(a, mask, mem_addr) \2081
- simde_mm_maskmove_si64( \2082
- (a), (mask), \2083
- SIMDE_CHECKED_REINTERPRET_CAST(int8_t *, char *, (mem_addr)))2084
-#endif2085
-2086
-SIMDE_FUNCTION_ATTRIBUTES2087
-simde__m64 simde_mm_max_pi16(simde__m64 a, simde__m64 b)2088
-{2089
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2090
- return _mm_max_pi16(a, b);2091
-#else2092
- simde__m64_private r_, a_ = simde__m64_to_private(a),2093
- b_ = simde__m64_to_private(b);2094
-2095
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2096
- r_.neon_i16 = vmax_s16(a_.neon_i16, b_.neon_i16);2097
-#else2098
- SIMDE_VECTORIZE2099
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {2100
- r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? a_.i16[i] : b_.i16[i];2101
- }2102
-#endif2103
-2104
- return simde__m64_from_private(r_);2105
-#endif2106
-}2107
-#define simde_m_pmaxsw(a, b) simde_mm_max_pi16(a, b)2108
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2109
-#define _mm_max_pi16(a, b) simde_mm_max_pi16(a, b)2110
-#define _m_pmaxsw(a, b) simde_mm_max_pi16(a, b)2111
-#endif2112
-2113
-SIMDE_FUNCTION_ATTRIBUTES2114
-simde__m128 simde_mm_max_ps(simde__m128 a, simde__m128 b)2115
-{2116
-#if defined(SIMDE_X86_SSE_NATIVE)2117
- return _mm_max_ps(a, b);2118
-#else2119
- simde__m128_private r_, a_ = simde__m128_to_private(a),2120
- b_ = simde__m128_to_private(b);2121
-2122
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2123
- r_.neon_f32 = vmaxq_f32(a_.neon_f32, b_.neon_f32);2124
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2125
- r_.altivec_f32 = vec_max(a_.altivec_f32, b_.altivec_f32);2126
-#else2127
- SIMDE_VECTORIZE2128
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2129
- r_.f32[i] = (a_.f32[i] > b_.f32[i]) ? a_.f32[i] : b_.f32[i];2130
- }2131
-#endif2132
-2133
- return simde__m128_from_private(r_);2134
-#endif2135
-}2136
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2137
-#define _mm_max_ps(a, b) simde_mm_max_ps((a), (b))2138
-#endif2139
-2140
-SIMDE_FUNCTION_ATTRIBUTES2141
-simde__m64 simde_mm_max_pu8(simde__m64 a, simde__m64 b)2142
-{2143
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2144
- return _mm_max_pu8(a, b);2145
-#else2146
- simde__m64_private r_, a_ = simde__m64_to_private(a),2147
- b_ = simde__m64_to_private(b);2148
-2149
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2150
- r_.neon_u8 = vmax_u8(a_.neon_u8, b_.neon_u8);2151
-#else2152
- SIMDE_VECTORIZE2153
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {2154
- r_.u8[i] = (a_.u8[i] > b_.u8[i]) ? a_.u8[i] : b_.u8[i];2155
- }2156
-#endif2157
-2158
- return simde__m64_from_private(r_);2159
-#endif2160
-}2161
-#define simde_m_pmaxub(a, b) simde_mm_max_pu8(a, b)2162
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2163
-#define _mm_max_pu8(a, b) simde_mm_max_pu8(a, b)2164
-#define _m_pmaxub(a, b) simde_mm_max_pu8(a, b)2165
-#endif2166
-2167
-SIMDE_FUNCTION_ATTRIBUTES2168
-simde__m128 simde_mm_max_ss(simde__m128 a, simde__m128 b)2169
-{2170
-#if defined(SIMDE_X86_SSE_NATIVE)2171
- return _mm_max_ss(a, b);2172
-#elif defined(SIMDE_ASSUME_VECTORIZATION)2173
- return simde_mm_move_ss(a, simde_mm_max_ps(a, b));2174
-#else2175
- simde__m128_private r_, a_ = simde__m128_to_private(a),2176
- b_ = simde__m128_to_private(b);2177
-2178
- r_.f32[0] = (a_.f32[0] > b_.f32[0]) ? a_.f32[0] : b_.f32[0];2179
- r_.f32[1] = a_.f32[1];2180
- r_.f32[2] = a_.f32[2];2181
- r_.f32[3] = a_.f32[3];2182
-2183
- return simde__m128_from_private(r_);2184
-#endif2185
-}2186
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2187
-#define _mm_max_ss(a, b) simde_mm_max_ss((a), (b))2188
-#endif2189
-2190
-SIMDE_FUNCTION_ATTRIBUTES2191
-simde__m64 simde_mm_min_pi16(simde__m64 a, simde__m64 b)2192
-{2193
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2194
- return _mm_min_pi16(a, b);2195
-#else2196
- simde__m64_private r_, a_ = simde__m64_to_private(a),2197
- b_ = simde__m64_to_private(b);2198
-2199
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2200
- r_.neon_i16 = vmin_s16(a_.neon_i16, b_.neon_i16);2201
-#else2202
- SIMDE_VECTORIZE2203
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {2204
- r_.i16[i] = (a_.i16[i] < b_.i16[i]) ? a_.i16[i] : b_.i16[i];2205
- }2206
-#endif2207
-2208
- return simde__m64_from_private(r_);2209
-#endif2210
-}2211
-#define simde_m_pminsw(a, b) simde_mm_min_pi16(a, b)2212
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2213
-#define _mm_min_pi16(a, b) simde_mm_min_pi16(a, b)2214
-#define _m_pminsw(a, b) simde_mm_min_pi16(a, b)2215
-#endif2216
-2217
-SIMDE_FUNCTION_ATTRIBUTES2218
-simde__m128 simde_mm_min_ps(simde__m128 a, simde__m128 b)2219
-{2220
-#if defined(SIMDE_X86_SSE_NATIVE)2221
- return _mm_min_ps(a, b);2222
-#else2223
- simde__m128_private r_, a_ = simde__m128_to_private(a),2224
- b_ = simde__m128_to_private(b);2225
-2226
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2227
- r_.neon_f32 = vminq_f32(a_.neon_f32, b_.neon_f32);2228
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2229
- r_.altivec_f32 = vec_min(a_.altivec_f32, b_.altivec_f32);2230
-#else2231
- SIMDE_VECTORIZE2232
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2233
- r_.f32[i] = (a_.f32[i] < b_.f32[i]) ? a_.f32[i] : b_.f32[i];2234
- }2235
-#endif2236
-2237
- return simde__m128_from_private(r_);2238
-#endif2239
-}2240
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2241
-#define _mm_min_ps(a, b) simde_mm_min_ps((a), (b))2242
-#endif2243
-2244
-SIMDE_FUNCTION_ATTRIBUTES2245
-simde__m64 simde_mm_min_pu8(simde__m64 a, simde__m64 b)2246
-{2247
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2248
- return _mm_min_pu8(a, b);2249
-#else2250
- simde__m64_private r_, a_ = simde__m64_to_private(a),2251
- b_ = simde__m64_to_private(b);2252
-2253
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2254
- r_.neon_u8 = vmin_u8(a_.neon_u8, b_.neon_u8);2255
-#else2256
- SIMDE_VECTORIZE2257
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {2258
- r_.u8[i] = (a_.u8[i] < b_.u8[i]) ? a_.u8[i] : b_.u8[i];2259
- }2260
-#endif2261
-2262
- return simde__m64_from_private(r_);2263
-#endif2264
-}2265
-#define simde_m_pminub(a, b) simde_mm_min_pu8(a, b)2266
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2267
-#define _mm_min_pu8(a, b) simde_mm_min_pu8(a, b)2268
-#define _m_pminub(a, b) simde_mm_min_pu8(a, b)2269
-#endif2270
-2271
-SIMDE_FUNCTION_ATTRIBUTES2272
-simde__m128 simde_mm_min_ss(simde__m128 a, simde__m128 b)2273
-{2274
-#if defined(SIMDE_X86_SSE_NATIVE)2275
- return _mm_min_ss(a, b);2276
-#elif defined(SIMDE_ASSUME_VECTORIZATION)2277
- return simde_mm_move_ss(a, simde_mm_min_ps(a, b));2278
-#else2279
- simde__m128_private r_, a_ = simde__m128_to_private(a),2280
- b_ = simde__m128_to_private(b);2281
-2282
- r_.f32[0] = (a_.f32[0] < b_.f32[0]) ? a_.f32[0] : b_.f32[0];2283
- r_.f32[1] = a_.f32[1];2284
- r_.f32[2] = a_.f32[2];2285
- r_.f32[3] = a_.f32[3];2286
-2287
- return simde__m128_from_private(r_);2288
-#endif2289
-}2290
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2291
-#define _mm_min_ss(a, b) simde_mm_min_ss((a), (b))2292
-#endif2293
-2294
-SIMDE_FUNCTION_ATTRIBUTES2295
-simde__m128 simde_mm_movehl_ps(simde__m128 a, simde__m128 b)2296
-{2297
-#if defined(SIMDE_X86_SSE_NATIVE)2298
- return _mm_movehl_ps(a, b);2299
-#else2300
- simde__m128_private r_, a_ = simde__m128_to_private(a),2301
- b_ = simde__m128_to_private(b);2302
-2303
-#if defined(SIMDE_SHUFFLE_VECTOR_)2304
- r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 6, 7, 2, 3);2305
-#else2306
- r_.f32[0] = b_.f32[2];2307
- r_.f32[1] = b_.f32[3];2308
- r_.f32[2] = a_.f32[2];2309
- r_.f32[3] = a_.f32[3];2310
-#endif2311
-2312
- return simde__m128_from_private(r_);2313
-#endif2314
-}2315
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2316
-#define _mm_movehl_ps(a, b) simde_mm_movehl_ps((a), (b))2317
-#endif2318
-2319
-SIMDE_FUNCTION_ATTRIBUTES2320
-simde__m128 simde_mm_movelh_ps(simde__m128 a, simde__m128 b)2321
-{2322
-#if defined(SIMDE_X86_SSE_NATIVE)2323
- return _mm_movelh_ps(a, b);2324
-#else2325
- simde__m128_private r_, a_ = simde__m128_to_private(a),2326
- b_ = simde__m128_to_private(b);2327
-2328
-#if defined(SIMDE_SHUFFLE_VECTOR_)2329
- r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 0, 1, 4, 5);2330
-#else2331
- r_.f32[0] = a_.f32[0];2332
- r_.f32[1] = a_.f32[1];2333
- r_.f32[2] = b_.f32[0];2334
- r_.f32[3] = b_.f32[1];2335
-#endif2336
-2337
- return simde__m128_from_private(r_);2338
-#endif2339
-}2340
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2341
-#define _mm_movelh_ps(a, b) simde_mm_movelh_ps((a), (b))2342
-#endif2343
-2344
-SIMDE_FUNCTION_ATTRIBUTES2345
-int simde_mm_movemask_pi8(simde__m64 a)2346
-{2347
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2348
- return _mm_movemask_pi8(a);2349
-#else2350
- simde__m64_private a_ = simde__m64_to_private(a);2351
- int r = 0;2352
- const size_t nmemb = sizeof(a_.i8) / sizeof(a_.i8[0]);2353
-2354
- SIMDE_VECTORIZE_REDUCTION(| : r)2355
- for (size_t i = 0; i < nmemb; i++) {2356
- r |= (a_.u8[nmemb - 1 - i] >> 7) << (nmemb - 1 - i);2357
- }2358
-2359
- return r;2360
-#endif2361
-}2362
-#define simde_m_pmovmskb(a, b) simde_mm_movemask_pi8(a, b)2363
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2364
-#define _mm_movemask_pi8(a) simde_mm_movemask_pi8(a)2365
-#endif2366
-2367
-SIMDE_FUNCTION_ATTRIBUTES2368
-int simde_mm_movemask_ps(simde__m128 a)2369
-{2370
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2371
- return _mm_movemask_ps(a);2372
-#else2373
- int r = 0;2374
- simde__m128_private a_ = simde__m128_to_private(a);2375
-2376
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2377
- /* TODO: check to see if NEON version is faster than the portable version */2378
- static const uint32x4_t movemask = {1, 2, 4, 8};2379
- static const uint32x4_t highbit = {0x80000000, 0x80000000, 0x80000000,2380
- 0x80000000};2381
- uint32x4_t t0 = a_.neon_u32;2382
- uint32x4_t t1 = vtstq_u32(t0, highbit);2383
- uint32x4_t t2 = vandq_u32(t1, movemask);2384
- uint32x2_t t3 = vorr_u32(vget_low_u32(t2), vget_high_u32(t2));2385
- r = vget_lane_u32(t3, 0) | vget_lane_u32(t3, 1);2386
-#else2387
- SIMDE_VECTORIZE_REDUCTION(| : r)2388
- for (size_t i = 0; i < sizeof(a_.u32) / sizeof(a_.u32[0]); i++) {2389
- r |= (a_.u32[i] >> ((sizeof(a_.u32[i]) * CHAR_BIT) - 1)) << i;2390
- }2391
-#endif2392
-2393
- return r;2394
-#endif2395
-}2396
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2397
-#define _mm_movemask_ps(a) simde_mm_movemask_ps((a))2398
-#endif2399
-2400
-SIMDE_FUNCTION_ATTRIBUTES2401
-simde__m128 simde_mm_mul_ps(simde__m128 a, simde__m128 b)2402
-{2403
-#if defined(SIMDE_X86_SSE_NATIVE)2404
- return _mm_mul_ps(a, b);2405
-#else2406
- simde__m128_private r_, a_ = simde__m128_to_private(a),2407
- b_ = simde__m128_to_private(b);2408
-2409
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2410
- r_.neon_f32 = vmulq_f32(a_.neon_f32, b_.neon_f32);2411
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)2412
- r_.wasm_v128 = wasm_f32x4_mul(a_.wasm_v128, b_.wasm_v128);2413
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2414
- r_.f32 = a_.f32 * b_.f32;2415
-#else2416
- SIMDE_VECTORIZE2417
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2418
- r_.f32[i] = a_.f32[i] * b_.f32[i];2419
- }2420
-#endif2421
-2422
- return simde__m128_from_private(r_);2423
-#endif2424
-}2425
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2426
-#define _mm_mul_ps(a, b) simde_mm_mul_ps((a), (b))2427
-#endif2428
-2429
-SIMDE_FUNCTION_ATTRIBUTES2430
-simde__m128 simde_mm_mul_ss(simde__m128 a, simde__m128 b)2431
-{2432
-#if defined(SIMDE_X86_SSE_NATIVE)2433
- return _mm_mul_ss(a, b);2434
-#elif defined(SIMDE_ASSUME_VECTORIZATION)2435
- return simde_mm_move_ss(a, simde_mm_mul_ps(a, b));2436
-#else2437
- simde__m128_private r_, a_ = simde__m128_to_private(a),2438
- b_ = simde__m128_to_private(b);2439
-2440
- r_.f32[0] = a_.f32[0] * b_.f32[0];2441
- r_.f32[1] = a_.f32[1];2442
- r_.f32[2] = a_.f32[2];2443
- r_.f32[3] = a_.f32[3];2444
-2445
- return simde__m128_from_private(r_);2446
-#endif2447
-}2448
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2449
-#define _mm_mul_ss(a, b) simde_mm_mul_ss((a), (b))2450
-#endif2451
-2452
-SIMDE_FUNCTION_ATTRIBUTES2453
-simde__m64 simde_mm_mulhi_pu16(simde__m64 a, simde__m64 b)2454
-{2455
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2456
- return _mm_mulhi_pu16(a, b);2457
-#else2458
- simde__m64_private r_, a_ = simde__m64_to_private(a),2459
- b_ = simde__m64_to_private(b);2460
-2461
- SIMDE_VECTORIZE2462
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {2463
- r_.u16[i] = HEDLEY_STATIC_CAST(2464
- uint16_t, ((HEDLEY_STATIC_CAST(uint32_t, a_.u16[i]) *2465
- HEDLEY_STATIC_CAST(uint32_t, b_.u16[i])) >>2466
- UINT32_C(16)));2467
- }2468
-2469
- return simde__m64_from_private(r_);2470
-#endif2471
-}2472
-#define simde_m_pmulhuw(a, b) simde_mm_mulhi_pu16(a, b)2473
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2474
-#define _mm_mulhi_pu16(a, b) simde_mm_mulhi_pu16(a, b)2475
-#endif2476
-2477
-SIMDE_FUNCTION_ATTRIBUTES2478
-simde__m128 simde_mm_or_ps(simde__m128 a, simde__m128 b)2479
-{2480
-#if defined(SIMDE_X86_SSE_NATIVE)2481
- return _mm_or_ps(a, b);2482
-#else2483
- simde__m128_private r_, a_ = simde__m128_to_private(a),2484
- b_ = simde__m128_to_private(b);2485
-2486
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2487
- r_.neon_i32 = vorrq_s32(a_.neon_i32, b_.neon_i32);2488
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2489
- r_.altivec_i32 = vec_or(a_.altivec_i32, b_.altivec_i32);2490
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2491
- r_.i32f = a_.i32f | b_.i32f;2492
-#else2493
- SIMDE_VECTORIZE2494
- for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {2495
- r_.u32[i] = a_.u32[i] | b_.u32[i];2496
- }2497
-#endif2498
-2499
- return simde__m128_from_private(r_);2500
-#endif2501
-}2502
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2503
-#define _mm_or_ps(a, b) simde_mm_or_ps((a), (b))2504
-#endif2505
-2506
-SIMDE_FUNCTION_ATTRIBUTES2507
-void simde_mm_prefetch(char const *p, int i)2508
-{2509
- (void)p;2510
- (void)i;2511
-}2512
-#if defined(SIMDE_X86_SSE_NATIVE)2513
-#define simde_mm_prefetch(p, i) _mm_prefetch(p, i)2514
-#endif2515
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2516
-#define _mm_prefetch(p, i) simde_mm_prefetch(p, i)2517
-#endif2518
-2519
-SIMDE_FUNCTION_ATTRIBUTES2520
-simde__m128 simde_mm_rcp_ps(simde__m128 a)2521
-{2522
-#if defined(SIMDE_X86_SSE_NATIVE)2523
- return _mm_rcp_ps(a);2524
-#else2525
- simde__m128_private r_, a_ = simde__m128_to_private(a);2526
-2527
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2528
- float32x4_t recip = vrecpeq_f32(a_.neon_f32);2529
-2530
-#if SIMDE_ACCURACY_PREFERENCE > 02531
- for (int i = 0; i < SIMDE_ACCURACY_PREFERENCE; ++i) {2532
- recip = vmulq_f32(recip, vrecpsq_f32(recip, a_.neon_f32));2533
- }2534
-#endif2535
-2536
- r_.neon_f32 = recip;2537
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2538
- r_.altivec_f32 = vec_re(a_.altivec_f32);2539
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)2540
- r_.f32 = 1.0f / a_.f32;2541
-#else2542
- SIMDE_VECTORIZE2543
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2544
- r_.f32[i] = 1.0f / a_.f32[i];2545
- }2546
-#endif2547
-2548
- return simde__m128_from_private(r_);2549
-#endif2550
-}2551
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2552
-#define _mm_rcp_ps(a) simde_mm_rcp_ps((a))2553
-#endif2554
-2555
-SIMDE_FUNCTION_ATTRIBUTES2556
-simde__m128 simde_mm_rcp_ss(simde__m128 a)2557
-{2558
-#if defined(SIMDE_X86_SSE_NATIVE)2559
- return _mm_rcp_ss(a);2560
-#elif defined(SIMDE_ASSUME_VECTORIZATION)2561
- return simde_mm_move_ss(a, simde_mm_rcp_ps(a));2562
-#else2563
- simde__m128_private r_, a_ = simde__m128_to_private(a);2564
-2565
- r_.f32[0] = 1.0f / a_.f32[0];2566
- r_.f32[1] = a_.f32[1];2567
- r_.f32[2] = a_.f32[2];2568
- r_.f32[3] = a_.f32[3];2569
-2570
- return simde__m128_from_private(r_);2571
-#endif2572
-}2573
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2574
-#define _mm_rcp_ss(a) simde_mm_rcp_ss((a))2575
-#endif2576
-2577
-SIMDE_FUNCTION_ATTRIBUTES2578
-simde__m128 simde_mm_rsqrt_ps(simde__m128 a)2579
-{2580
-#if defined(SIMDE_X86_SSE_NATIVE)2581
- return _mm_rsqrt_ps(a);2582
-#else2583
- simde__m128_private r_, a_ = simde__m128_to_private(a);2584
-2585
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2586
- r_.neon_f32 = vrsqrteq_f32(a_.neon_f32);2587
-#elif defined(__STDC_IEC_559__)2588
- /* https://basesandframes.files.wordpress.com/2020/04/even_faster_math_functions_green_2020.pdf2589
- Pages 100 - 103 */2590
- SIMDE_VECTORIZE2591
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2592
-#if SIMDE_ACCURACY_PREFERENCE <= 02593
- r_.i32[i] = INT32_C(0x5F37624F) - (a_.i32[i] >> 1);2594
-#else2595
- simde_float32 x = a_.f32[i];2596
- simde_float32 xhalf = SIMDE_FLOAT32_C(0.5) * x;2597
- int32_t ix;2598
-2599
- simde_memcpy(&ix, &x, sizeof(ix));2600
-2601
-#if SIMDE_ACCURACY_PREFERENCE == 12602
- ix = INT32_C(0x5F375A82) - (ix >> 1);2603
-#else2604
- ix = INT32_C(0x5F37599E) - (ix >> 1);2605
-#endif2606
-2607
- simde_memcpy(&x, &ix, sizeof(x));2608
-2609
-#if SIMDE_ACCURACY_PREFERENCE >= 22610
- x = x * (SIMDE_FLOAT32_C(1.5008909) - xhalf * x * x);2611
-#endif2612
- x = x * (SIMDE_FLOAT32_C(1.5008909) - xhalf * x * x);2613
-2614
- r_.f32[i] = x;2615
-#endif2616
- }2617
-#elif defined(simde_math_sqrtf)2618
- SIMDE_VECTORIZE2619
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2620
- r_.f32[i] = 1.0f / simde_math_sqrtf(a_.f32[i]);2621
- }2622
-#else2623
- HEDLEY_UNREACHABLE();2624
-#endif2625
-2626
- return simde__m128_from_private(r_);2627
-#endif2628
-}2629
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2630
-#define _mm_rsqrt_ps(a) simde_mm_rsqrt_ps((a))2631
-#endif2632
-2633
-SIMDE_FUNCTION_ATTRIBUTES2634
-simde__m128 simde_mm_rsqrt_ss(simde__m128 a)2635
-{2636
-#if defined(SIMDE_X86_SSE_NATIVE)2637
- return _mm_rsqrt_ss(a);2638
-#elif defined(SIMDE_ASSUME_VECTORIZATION)2639
- return simde_mm_move_ss(a, simde_mm_rsqrt_ps(a));2640
-#else2641
- simde__m128_private r_, a_ = simde__m128_to_private(a);2642
-2643
-#if defined(__STDC_IEC_559__)2644
- {2645
-#if SIMDE_ACCURACY_PREFERENCE <= 02646
- r_.i32[0] = INT32_C(0x5F37624F) - (a_.i32[0] >> 1);2647
-#else2648
- simde_float32 x = a_.f32[0];2649
- simde_float32 xhalf = SIMDE_FLOAT32_C(0.5) * x;2650
- int32_t ix;2651
-2652
- simde_memcpy(&ix, &x, sizeof(ix));2653
-2654
-#if SIMDE_ACCURACY_PREFERENCE == 12655
- ix = INT32_C(0x5F375A82) - (ix >> 1);2656
-#else2657
- ix = INT32_C(0x5F37599E) - (ix >> 1);2658
-#endif2659
-2660
- simde_memcpy(&x, &ix, sizeof(x));2661
-2662
-#if SIMDE_ACCURACY_PREFERENCE >= 22663
- x = x * (SIMDE_FLOAT32_C(1.5008909) - xhalf * x * x);2664
-#endif2665
- x = x * (SIMDE_FLOAT32_C(1.5008909) - xhalf * x * x);2666
-2667
- r_.f32[0] = x;2668
-#endif2669
- }2670
- r_.f32[1] = a_.f32[1];2671
- r_.f32[2] = a_.f32[2];2672
- r_.f32[3] = a_.f32[3];2673
-#elif defined(simde_math_sqrtf)2674
- r_.f32[0] = 1.0f / simde_math_sqrtf(a_.f32[0]);2675
- r_.f32[1] = a_.f32[1];2676
- r_.f32[2] = a_.f32[2];2677
- r_.f32[3] = a_.f32[3];2678
-#else2679
- HEDLEY_UNREACHABLE();2680
-#endif2681
-2682
- return simde__m128_from_private(r_);2683
-#endif2684
-}2685
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2686
-#define _mm_rsqrt_ss(a) simde_mm_rsqrt_ss((a))2687
-#endif2688
-2689
-SIMDE_FUNCTION_ATTRIBUTES2690
-simde__m64 simde_mm_sad_pu8(simde__m64 a, simde__m64 b)2691
-{2692
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2693
- return _mm_sad_pu8(a, b);2694
-#else2695
- simde__m64_private r_, a_ = simde__m64_to_private(a),2696
- b_ = simde__m64_to_private(b);2697
- uint16_t sum = 0;2698
-2699
-#if defined(SIMDE_HAVE_STDLIB_H)2700
- SIMDE_VECTORIZE_REDUCTION(+ : sum)2701
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {2702
- sum += HEDLEY_STATIC_CAST(uint8_t, abs(a_.u8[i] - b_.u8[i]));2703
- }2704
-2705
- r_.i16[0] = HEDLEY_STATIC_CAST(int16_t, sum);2706
- r_.i16[1] = 0;2707
- r_.i16[2] = 0;2708
- r_.i16[3] = 0;2709
-#else2710
- HEDLEY_UNREACHABLE();2711
-#endif2712
-2713
- return simde__m64_from_private(r_);2714
-#endif2715
-}2716
-#define simde_m_psadbw(a, b) simde_mm_sad_pu8(a, b)2717
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2718
-#define _mm_sad_pu8(a, b) simde_mm_sad_pu8(a, b)2719
-#define _m_psadbw(a, b) simde_mm_sad_pu8(a, b)2720
-#endif2721
-2722
-SIMDE_FUNCTION_ATTRIBUTES2723
-simde__m128 simde_mm_set_ss(simde_float32 a)2724
-{2725
-#if defined(SIMDE_X86_SSE_NATIVE)2726
- return _mm_set_ss(a);2727
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2728
- return vsetq_lane_f32(a, vdupq_n_f32(SIMDE_FLOAT32_C(0.0)), 0);2729
-#else2730
- return simde_mm_set_ps(SIMDE_FLOAT32_C(0.0), SIMDE_FLOAT32_C(0.0),2731
- SIMDE_FLOAT32_C(0.0), a);2732
-#endif2733
-}2734
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2735
-#define _mm_set_ss(a) simde_mm_set_ss(a)2736
-#endif2737
-2738
-SIMDE_FUNCTION_ATTRIBUTES2739
-simde__m128 simde_mm_setr_ps(simde_float32 e3, simde_float32 e2,2740
- simde_float32 e1, simde_float32 e0)2741
-{2742
-#if defined(SIMDE_X86_SSE_NATIVE)2743
- return _mm_setr_ps(e3, e2, e1, e0);2744
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2745
- SIMDE_ALIGN(16) simde_float32 data[4] = {e3, e2, e1, e0};2746
- return vld1q_f32(data);2747
-#else2748
- return simde_mm_set_ps(e0, e1, e2, e3);2749
-#endif2750
-}2751
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2752
-#define _mm_setr_ps(e3, e2, e1, e0) simde_mm_setr_ps(e3, e2, e1, e0)2753
-#endif2754
-2755
-SIMDE_FUNCTION_ATTRIBUTES2756
-simde__m128 simde_mm_setzero_ps(void)2757
-{2758
-#if defined(SIMDE_X86_SSE_NATIVE)2759
- return _mm_setzero_ps();2760
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2761
- return vdupq_n_f32(SIMDE_FLOAT32_C(0.0));2762
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2763
- return vec_splats((float)0);2764
-#else2765
- simde__m128 r;2766
- simde_memset(&r, 0, sizeof(r));2767
- return r;2768
-#endif2769
-}2770
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2771
-#define _mm_setzero_ps() simde_mm_setzero_ps()2772
-#endif2773
-2774
-#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)2775
-HEDLEY_DIAGNOSTIC_PUSH2776
-SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_2777
-#endif2778
-2779
-SIMDE_FUNCTION_ATTRIBUTES2780
-simde__m128 simde_mm_undefined_ps(void)2781
-{2782
- simde__m128_private r_;2783
-2784
-#if defined(SIMDE_HAVE_UNDEFINED128)2785
- r_.n = _mm_undefined_ps();2786
-#elif !defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)2787
- r_ = simde__m128_to_private(simde_mm_setzero_ps());2788
-#endif2789
-2790
- return simde__m128_from_private(r_);2791
-}2792
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2793
-#define _mm_undefined_ps() simde_mm_undefined_ps()2794
-#endif2795
-2796
-#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)2797
-HEDLEY_DIAGNOSTIC_POP2798
-#endif2799
-2800
-SIMDE_FUNCTION_ATTRIBUTES2801
-simde__m128 simde_x_mm_setone_ps(void)2802
-{2803
- simde__m128 t = simde_mm_setzero_ps();2804
- return simde_mm_cmpeq_ps(t, t);2805
-}2806
-2807
-SIMDE_FUNCTION_ATTRIBUTES2808
-void simde_mm_sfence(void)2809
-{2810
- /* TODO: Use Hedley. */2811
-#if defined(SIMDE_X86_SSE_NATIVE)2812
- _mm_sfence();2813
-#elif defined(__GNUC__) && \2814
- ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 7))2815
- __atomic_thread_fence(__ATOMIC_SEQ_CST);2816
-#elif !defined(__INTEL_COMPILER) && defined(__STDC_VERSION__) && \2817
- (__STDC_VERSION__ >= 201112L) && !defined(__STDC_NO_ATOMICS__)2818
-#if defined(__GNUC__) && (__GNUC__ == 4) && (__GNUC_MINOR__ < 9)2819
- __atomic_thread_fence(__ATOMIC_SEQ_CST);2820
-#else2821
- atomic_thread_fence(memory_order_seq_cst);2822
-#endif2823
-#elif defined(_MSC_VER)2824
- MemoryBarrier();2825
-#elif HEDLEY_HAS_EXTENSION(c_atomic)2826
- __c11_atomic_thread_fence(__ATOMIC_SEQ_CST);2827
-#elif defined(__GNUC__) && \2828
- ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 1))2829
- __sync_synchronize();2830
-#elif defined(_OPENMP)2831
-#pragma omp critical(simde_mm_sfence_)2832
- {2833
- }2834
-#endif2835
-}2836
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2837
-#define _mm_sfence() simde_mm_sfence()2838
-#endif2839
-2840
-#define SIMDE_MM_SHUFFLE(z, y, x, w) \2841
- (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))2842
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2843
-#define _MM_SHUFFLE(z, y, x, w) SIMDE_MM_SHUFFLE(z, y, x, w)2844
-#endif2845
-2846
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \2847
- !defined(__PGI)2848
-#define simde_mm_shuffle_pi16(a, imm8) _mm_shuffle_pi16(a, imm8)2849
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2850
-#define simde_mm_shuffle_pi16(a, imm8) \2851
- (__extension__({ \2852
- const simde__m64_private simde__tmp_a_ = \2853
- simde__m64_to_private(a); \2854
- simde__m64_from_private((simde__m64_private){ \2855
- .i16 = SIMDE_SHUFFLE_VECTOR_( \2856
- 16, 8, (simde__tmp_a_).i16, \2857
- (simde__tmp_a_).i16, (((imm8)) & 3), \2858
- (((imm8) >> 2) & 3), (((imm8) >> 4) & 3), \2859
- (((imm8) >> 6) & 3))}); \2860
- }))2861
-#else2862
-SIMDE_FUNCTION_ATTRIBUTES2863
-simde__m64 simde_mm_shuffle_pi16(simde__m64 a, const int imm8)2864
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)2865
-{2866
- simde__m64_private r_;2867
- simde__m64_private a_ = simde__m64_to_private(a);2868
-2869
- for (size_t i = 0; i < sizeof(r_.i16) / sizeof(r_.i16[0]); i++) {2870
- r_.i16[i] = a_.i16[(imm8 >> (i * 2)) & 3];2871
- }2872
-2873
- HEDLEY_DIAGNOSTIC_PUSH2874
-#if HEDLEY_HAS_WARNING("-Wconditional-uninitialized")2875
-#pragma clang diagnostic ignored "-Wconditional-uninitialized"2876
-#endif2877
- return simde__m64_from_private(r_);2878
- HEDLEY_DIAGNOSTIC_POP2879
-}2880
-#endif2881
-#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)2882
-#define simde_m_pshufw(a, imm8) _m_pshufw(a, imm8)2883
-#else2884
-#define simde_m_pshufw(a, imm8) simde_mm_shuffle_pi16(a, imm8)2885
-#endif2886
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2887
-#define _mm_shuffle_pi16(a, imm8) simde_mm_shuffle_pi16(a, imm8)2888
-#define _m_pshufw(a, imm8) simde_mm_shuffle_pi16(a, imm8)2889
-#endif2890
-2891
-#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)2892
-#define simde_mm_shuffle_ps(a, b, imm8) _mm_shuffle_ps(a, b, imm8)2893
-#elif defined(SIMDE_SHUFFLE_VECTOR_)2894
-#define simde_mm_shuffle_ps(a, b, imm8) \2895
- (__extension__({ \2896
- simde__m128_from_private((simde__m128_private){ \2897
- .f32 = SIMDE_SHUFFLE_VECTOR_( \2898
- 32, 16, simde__m128_to_private(a).f32, \2899
- simde__m128_to_private(b).f32, (((imm8)) & 3), \2900
- (((imm8) >> 2) & 3), (((imm8) >> 4) & 3) + 4, \2901
- (((imm8) >> 6) & 3) + 4)}); \2902
- }))2903
-#else2904
-SIMDE_FUNCTION_ATTRIBUTES2905
-simde__m128 simde_mm_shuffle_ps(simde__m128 a, simde__m128 b, const int imm8)2906
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)2907
-{2908
- simde__m128_private r_, a_ = simde__m128_to_private(a),2909
- b_ = simde__m128_to_private(b);2910
-2911
- r_.f32[0] = a_.f32[(imm8 >> 0) & 3];2912
- r_.f32[1] = a_.f32[(imm8 >> 2) & 3];2913
- r_.f32[2] = b_.f32[(imm8 >> 4) & 3];2914
- r_.f32[3] = b_.f32[(imm8 >> 6) & 3];2915
-2916
- return simde__m128_from_private(r_);2917
-}2918
-#endif2919
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2920
-#define _mm_shuffle_ps(a, b, imm8) simde_mm_shuffle_ps((a), (b), imm8)2921
-#endif2922
-2923
-SIMDE_FUNCTION_ATTRIBUTES2924
-simde__m128 simde_mm_sqrt_ps(simde__m128 a)2925
-{2926
-#if defined(SIMDE_X86_SSE_NATIVE)2927
- return _mm_sqrt_ps(a);2928
-#else2929
- simde__m128_private r_, a_ = simde__m128_to_private(a);2930
-2931
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2932
- float32x4_t recipsq = vrsqrteq_f32(a_.neon_f32);2933
- float32x4_t sq = vrecpeq_f32(recipsq);2934
- /* ??? use step versions of both sqrt and recip for better accuracy? */2935
- r_.neon_f32 = sq;2936
-#elif defined(simde_math_sqrt)2937
- SIMDE_VECTORIZE2938
- for (size_t i = 0; i < sizeof(r_.f32) / sizeof(r_.f32[0]); i++) {2939
- r_.f32[i] = simde_math_sqrtf(a_.f32[i]);2940
- }2941
-#else2942
- HEDLEY_UNREACHABLE();2943
-#endif2944
-2945
- return simde__m128_from_private(r_);2946
-#endif2947
-}2948
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2949
-#define _mm_sqrt_ps(a) simde_mm_sqrt_ps((a))2950
-#endif2951
-2952
-SIMDE_FUNCTION_ATTRIBUTES2953
-simde__m128 simde_mm_sqrt_ss(simde__m128 a)2954
-{2955
-#if defined(SIMDE_X86_SSE_NATIVE)2956
- return _mm_sqrt_ss(a);2957
-#elif defined(SIMDE_ASSUME_VECTORIZATION)2958
- return simde_mm_move_ss(a, simde_mm_sqrt_ps(a));2959
-#else2960
- simde__m128_private r_, a_ = simde__m128_to_private(a);2961
-2962
-#if defined(simde_math_sqrtf)2963
- r_.f32[0] = simde_math_sqrtf(a_.f32[0]);2964
- r_.f32[1] = a_.f32[1];2965
- r_.f32[2] = a_.f32[2];2966
- r_.f32[3] = a_.f32[3];2967
-#else2968
- HEDLEY_UNREACHABLE();2969
-#endif2970
-2971
- return simde__m128_from_private(r_);2972
-#endif2973
-}2974
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2975
-#define _mm_sqrt_ss(a) simde_mm_sqrt_ss((a))2976
-#endif2977
-2978
-SIMDE_FUNCTION_ATTRIBUTES2979
-void simde_mm_store_ps(simde_float32 mem_addr[4], simde__m128 a)2980
-{2981
- simde_assert_aligned(16, mem_addr);2982
-2983
-#if defined(SIMDE_X86_SSE_NATIVE)2984
- _mm_store_ps(mem_addr, a);2985
-#else2986
- simde__m128_private a_ = simde__m128_to_private(a);2987
-2988
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2989
- vst1q_f32(mem_addr, a_.neon_f32);2990
-#elif defined(SIMDE_POWER_ALTIVE_P7_NATIVE)2991
- vec_vsx_st(a_.altivec_32, 0, mem_addr);2992
-#elif defined(SIMDE_POWER_ALTIVE_P5_NATIVE)2993
- vec_st(a_.altivec_32, 0, mem_addr);2994
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)2995
- wasm_v128_store(mem_addr, a_.wasm_v128);2996
-#else2997
- SIMDE_VECTORIZE_ALIGNED(mem_addr : 16)2998
- for (size_t i = 0; i < sizeof(a_.f32) / sizeof(a_.f32[0]); i++) {2999
- mem_addr[i] = a_.f32[i];3000
- }3001
-#endif3002
-#endif3003
-}3004
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3005
-#define _mm_store_ps(mem_addr, a) \3006
- simde_mm_store_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3007
- float *, simde_float32 *, mem_addr), \3008
- (a))3009
-#endif3010
-3011
-SIMDE_FUNCTION_ATTRIBUTES3012
-void simde_mm_store_ps1(simde_float32 mem_addr[4], simde__m128 a)3013
-{3014
- simde_assert_aligned(16, mem_addr);3015
-3016
-#if defined(SIMDE_X86_SSE_NATIVE)3017
- _mm_store_ps1(mem_addr, a);3018
-#else3019
- simde__m128_private a_ = simde__m128_to_private(a);3020
-3021
- SIMDE_VECTORIZE_ALIGNED(mem_addr : 16)3022
- for (size_t i = 0; i < sizeof(a_.f32) / sizeof(a_.f32[0]); i++) {3023
- mem_addr[i] = a_.f32[0];3024
- }3025
-#endif3026
-}3027
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3028
-#define _mm_store_ps1(mem_addr, a) \3029
- simde_mm_store_ps1(SIMDE_CHECKED_REINTERPRET_CAST( \3030
- float *, simde_float32 *, mem_addr), \3031
- (a))3032
-#endif3033
-3034
-SIMDE_FUNCTION_ATTRIBUTES3035
-void simde_mm_store_ss(simde_float32 *mem_addr, simde__m128 a)3036
-{3037
-#if defined(SIMDE_X86_SSE_NATIVE)3038
- _mm_store_ss(mem_addr, a);3039
-#else3040
- simde__m128_private a_ = simde__m128_to_private(a);3041
-3042
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3043
- vst1q_lane_f32(mem_addr, a_.neon_f32, 0);3044
-#else3045
- *mem_addr = a_.f32[0];3046
-#endif3047
-#endif3048
-}3049
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3050
-#define _mm_store_ss(mem_addr, a) \3051
- simde_mm_store_ss(SIMDE_CHECKED_REINTERPRET_CAST( \3052
- float *, simde_float32 *, mem_addr), \3053
- (a))3054
-#endif3055
-3056
-SIMDE_FUNCTION_ATTRIBUTES3057
-void simde_mm_store1_ps(simde_float32 mem_addr[4], simde__m128 a)3058
-{3059
- simde_assert_aligned(16, mem_addr);3060
-3061
-#if defined(SIMDE_X86_SSE_NATIVE)3062
- _mm_store1_ps(mem_addr, a);3063
-#else3064
- simde_mm_store_ps1(mem_addr, a);3065
-#endif3066
-}3067
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3068
-#define _mm_store1_ps(mem_addr, a) \3069
- simde_mm_store1_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3070
- float *, simde_float32 *, mem_addr), \3071
- (a))3072
-#endif3073
-3074
-SIMDE_FUNCTION_ATTRIBUTES3075
-void simde_mm_storeh_pi(simde__m64 *mem_addr, simde__m128 a)3076
-{3077
-#if defined(SIMDE_X86_SSE_NATIVE)3078
- _mm_storeh_pi(HEDLEY_REINTERPRET_CAST(__m64 *, mem_addr), a);3079
-#else3080
- simde__m64_private *dest_ =3081
- HEDLEY_REINTERPRET_CAST(simde__m64_private *, mem_addr);3082
- simde__m128_private a_ = simde__m128_to_private(a);3083
-3084
- dest_->f32[0] = a_.f32[2];3085
- dest_->f32[1] = a_.f32[3];3086
-#endif3087
-}3088
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3089
-#define _mm_storeh_pi(mem_addr, a) simde_mm_storeh_pi(mem_addr, (a))3090
-#endif3091
-3092
-SIMDE_FUNCTION_ATTRIBUTES3093
-void simde_mm_storel_pi(simde__m64 *mem_addr, simde__m128 a)3094
-{3095
-#if defined(SIMDE_X86_SSE_NATIVE)3096
- _mm_storel_pi(HEDLEY_REINTERPRET_CAST(__m64 *, mem_addr), a);3097
-#else3098
- simde__m64_private *dest_ =3099
- HEDLEY_REINTERPRET_CAST(simde__m64_private *, mem_addr);3100
- simde__m128_private a_ = simde__m128_to_private(a);3101
-3102
- dest_->f32[0] = a_.f32[0];3103
- dest_->f32[1] = a_.f32[1];3104
-#endif3105
-}3106
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3107
-#define _mm_storel_pi(mem_addr, a) simde_mm_storel_pi(mem_addr, (a))3108
-#endif3109
-3110
-SIMDE_FUNCTION_ATTRIBUTES3111
-void simde_mm_storer_ps(simde_float32 mem_addr[4], simde__m128 a)3112
-{3113
- simde_assert_aligned(16, mem_addr);3114
-3115
-#if defined(SIMDE_X86_SSE_NATIVE)3116
- _mm_storer_ps(mem_addr, a);3117
-#else3118
- simde__m128_private a_ = simde__m128_to_private(a);3119
-3120
-#if defined(SIMDE_SHUFFLE_VECTOR_)3121
- a_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, a_.f32, 3, 2, 1, 0);3122
- simde_mm_store_ps(mem_addr, simde__m128_from_private(a_));3123
-#else3124
- SIMDE_VECTORIZE_ALIGNED(mem_addr : 16)3125
- for (size_t i = 0; i < sizeof(a_.f32) / sizeof(a_.f32[0]); i++) {3126
- mem_addr[i] =3127
- a_.f32[((sizeof(a_.f32) / sizeof(a_.f32[0])) - 1) - i];3128
- }3129
-#endif3130
-#endif3131
-}3132
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3133
-#define _mm_storer_ps(mem_addr, a) \3134
- simde_mm_storer_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3135
- float *, simde_float32 *, mem_addr), \3136
- (a))3137
-#endif3138
-3139
-SIMDE_FUNCTION_ATTRIBUTES3140
-void simde_mm_storeu_ps(simde_float32 mem_addr[4], simde__m128 a)3141
-{3142
-#if defined(SIMDE_X86_SSE_NATIVE)3143
- _mm_storeu_ps(mem_addr, a);3144
-#else3145
- simde__m128_private a_ = simde__m128_to_private(a);3146
-3147
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3148
- vst1q_f32(mem_addr, a_.neon_f32);3149
-#else3150
- simde_memcpy(mem_addr, &a_, sizeof(a_));3151
-#endif3152
-#endif3153
-}3154
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3155
-#define _mm_storeu_ps(mem_addr, a) \3156
- simde_mm_storeu_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3157
- float *, simde_float32 *, mem_addr), \3158
- (a))3159
-#endif3160
-3161
-SIMDE_FUNCTION_ATTRIBUTES3162
-simde__m128 simde_mm_sub_ps(simde__m128 a, simde__m128 b)3163
-{3164
-#if defined(SIMDE_X86_SSE_NATIVE)3165
- return _mm_sub_ps(a, b);3166
-#else3167
- simde__m128_private r_, a_ = simde__m128_to_private(a),3168
- b_ = simde__m128_to_private(b);3169
-3170
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3171
- r_.neon_f32 = vsubq_f32(a_.neon_f32, b_.neon_f32);3172
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)3173
- r_.wasm_v128 = wasm_f32x4_sub(a_.wasm_v128, b_.wasm_v128);3174
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3175
- r_.f32 = a_.f32 - b_.f32;3176
-#else3177
- SIMDE_VECTORIZE3178
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {3179
- r_.f32[i] = a_.f32[i] - b_.f32[i];3180
- }3181
-#endif3182
-3183
- return simde__m128_from_private(r_);3184
-#endif3185
-}3186
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3187
-#define _mm_sub_ps(a, b) simde_mm_sub_ps((a), (b))3188
-#endif3189
-3190
-SIMDE_FUNCTION_ATTRIBUTES3191
-simde__m128 simde_mm_sub_ss(simde__m128 a, simde__m128 b)3192
-{3193
-#if defined(SIMDE_X86_SSE_NATIVE)3194
- return _mm_sub_ss(a, b);3195
-#elif defined(SIMDE_ASSUME_VECTORIZATION)3196
- return simde_mm_move_ss(a, simde_mm_sub_ps(a, b));3197
-#else3198
- simde__m128_private r_, a_ = simde__m128_to_private(a),3199
- b_ = simde__m128_to_private(b);3200
-3201
- r_.f32[0] = a_.f32[0] - b_.f32[0];3202
- r_.f32[1] = a_.f32[1];3203
- r_.f32[2] = a_.f32[2];3204
- r_.f32[3] = a_.f32[3];3205
-3206
- return simde__m128_from_private(r_);3207
-#endif3208
-}3209
-3210
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3211
-#define _mm_sub_ss(a, b) simde_mm_sub_ss((a), (b))3212
-#endif3213
-3214
-SIMDE_FUNCTION_ATTRIBUTES3215
-int simde_mm_ucomieq_ss(simde__m128 a, simde__m128 b)3216
-{3217
-#if defined(SIMDE_X86_SSE_NATIVE)3218
- return _mm_ucomieq_ss(a, b);3219
-#else3220
- simde__m128_private a_ = simde__m128_to_private(a),3221
- b_ = simde__m128_to_private(b);3222
- int r;3223
-3224
-#if defined(SIMDE_HAVE_FENV_H)3225
- fenv_t envp;3226
- int x = feholdexcept(&envp);3227
- r = a_.f32[0] == b_.f32[0];3228
- if (HEDLEY_LIKELY(x == 0))3229
- fesetenv(&envp);3230
-#else3231
- r = a_.f32[0] == b_.f32[0];3232
-#endif3233
-3234
- return r;3235
-#endif3236
-}3237
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3238
-#define _mm_ucomieq_ss(a, b) simde_mm_ucomieq_ss((a), (b))3239
-#endif3240
-3241
-SIMDE_FUNCTION_ATTRIBUTES3242
-int simde_mm_ucomige_ss(simde__m128 a, simde__m128 b)3243
-{3244
-#if defined(SIMDE_X86_SSE_NATIVE)3245
- return _mm_ucomige_ss(a, b);3246
-#else3247
- simde__m128_private a_ = simde__m128_to_private(a),3248
- b_ = simde__m128_to_private(b);3249
- int r;3250
-3251
-#if defined(SIMDE_HAVE_FENV_H)3252
- fenv_t envp;3253
- int x = feholdexcept(&envp);3254
- r = a_.f32[0] >= b_.f32[0];3255
- if (HEDLEY_LIKELY(x == 0))3256
- fesetenv(&envp);3257
-#else3258
- r = a_.f32[0] >= b_.f32[0];3259
-#endif3260
-3261
- return r;3262
-#endif3263
-}3264
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3265
-#define _mm_ucomige_ss(a, b) simde_mm_ucomige_ss((a), (b))3266
-#endif3267
-3268
-SIMDE_FUNCTION_ATTRIBUTES3269
-int simde_mm_ucomigt_ss(simde__m128 a, simde__m128 b)3270
-{3271
-#if defined(SIMDE_X86_SSE_NATIVE)3272
- return _mm_ucomigt_ss(a, b);3273
-#else3274
- simde__m128_private a_ = simde__m128_to_private(a),3275
- b_ = simde__m128_to_private(b);3276
- int r;3277
-3278
-#if defined(SIMDE_HAVE_FENV_H)3279
- fenv_t envp;3280
- int x = feholdexcept(&envp);3281
- r = a_.f32[0] > b_.f32[0];3282
- if (HEDLEY_LIKELY(x == 0))3283
- fesetenv(&envp);3284
-#else3285
- r = a_.f32[0] > b_.f32[0];3286
-#endif3287
-3288
- return r;3289
-#endif3290
-}3291
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3292
-#define _mm_ucomigt_ss(a, b) simde_mm_ucomigt_ss((a), (b))3293
-#endif3294
-3295
-SIMDE_FUNCTION_ATTRIBUTES3296
-int simde_mm_ucomile_ss(simde__m128 a, simde__m128 b)3297
-{3298
-#if defined(SIMDE_X86_SSE_NATIVE)3299
- return _mm_ucomile_ss(a, b);3300
-#else3301
- simde__m128_private a_ = simde__m128_to_private(a),3302
- b_ = simde__m128_to_private(b);3303
- int r;3304
-3305
-#if defined(SIMDE_HAVE_FENV_H)3306
- fenv_t envp;3307
- int x = feholdexcept(&envp);3308
- r = a_.f32[0] <= b_.f32[0];3309
- if (HEDLEY_LIKELY(x == 0))3310
- fesetenv(&envp);3311
-#else3312
- r = a_.f32[0] <= b_.f32[0];3313
-#endif3314
-3315
- return r;3316
-#endif3317
-}3318
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3319
-#define _mm_ucomile_ss(a, b) simde_mm_ucomile_ss((a), (b))3320
-#endif3321
-3322
-SIMDE_FUNCTION_ATTRIBUTES3323
-int simde_mm_ucomilt_ss(simde__m128 a, simde__m128 b)3324
-{3325
-#if defined(SIMDE_X86_SSE_NATIVE)3326
- return _mm_ucomilt_ss(a, b);3327
-#else3328
- simde__m128_private a_ = simde__m128_to_private(a),3329
- b_ = simde__m128_to_private(b);3330
- int r;3331
-3332
-#if defined(SIMDE_HAVE_FENV_H)3333
- fenv_t envp;3334
- int x = feholdexcept(&envp);3335
- r = a_.f32[0] < b_.f32[0];3336
- if (HEDLEY_LIKELY(x == 0))3337
- fesetenv(&envp);3338
-#else3339
- r = a_.f32[0] < b_.f32[0];3340
-#endif3341
-3342
- return r;3343
-#endif3344
-}3345
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3346
-#define _mm_ucomilt_ss(a, b) simde_mm_ucomilt_ss((a), (b))3347
-#endif3348
-3349
-SIMDE_FUNCTION_ATTRIBUTES3350
-int simde_mm_ucomineq_ss(simde__m128 a, simde__m128 b)3351
-{3352
-#if defined(SIMDE_X86_SSE_NATIVE)3353
- return _mm_ucomineq_ss(a, b);3354
-#else3355
- simde__m128_private a_ = simde__m128_to_private(a),3356
- b_ = simde__m128_to_private(b);3357
- int r;3358
-3359
-#if defined(SIMDE_HAVE_FENV_H)3360
- fenv_t envp;3361
- int x = feholdexcept(&envp);3362
- r = a_.f32[0] != b_.f32[0];3363
- if (HEDLEY_LIKELY(x == 0))3364
- fesetenv(&envp);3365
-#else3366
- r = a_.f32[0] != b_.f32[0];3367
-#endif3368
-3369
- return r;3370
-#endif3371
-}3372
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3373
-#define _mm_ucomineq_ss(a, b) simde_mm_ucomineq_ss((a), (b))3374
-#endif3375
-3376
-#if defined(SIMDE_X86_SSE_NATIVE)3377
-#if defined(__has_builtin)3378
-#if __has_builtin(__builtin_ia32_undef128)3379
-#define SIMDE_HAVE_UNDEFINED1283380
-#endif3381
-#elif !defined(__PGI) && !defined(SIMDE_BUG_GCC_REV_208793) && \3382
- !defined(_MSC_VER)3383
-#define SIMDE_HAVE_UNDEFINED1283384
-#endif3385
-#endif3386
-3387
-#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)3388
-HEDLEY_DIAGNOSTIC_PUSH3389
-SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_3390
-#endif3391
-3392
-SIMDE_FUNCTION_ATTRIBUTES3393
-simde__m128 simde_mm_unpackhi_ps(simde__m128 a, simde__m128 b)3394
-{3395
-#if defined(SIMDE_X86_SSE_NATIVE)3396
- return _mm_unpackhi_ps(a, b);3397
-#else3398
- simde__m128_private r_, a_ = simde__m128_to_private(a),3399
- b_ = simde__m128_to_private(b);3400
-3401
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3402
- float32x2_t a1 = vget_high_f32(a_.neon_f32);3403
- float32x2_t b1 = vget_high_f32(b_.neon_f32);3404
- float32x2x2_t result = vzip_f32(a1, b1);3405
- r_.neon_f32 = vcombine_f32(result.val[0], result.val[1]);3406
-#elif defined(SIMDE_SHUFFLE_VECTOR_)3407
- r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 2, 6, 3, 7);3408
-#else3409
- r_.f32[0] = a_.f32[2];3410
- r_.f32[1] = b_.f32[2];3411
- r_.f32[2] = a_.f32[3];3412
- r_.f32[3] = b_.f32[3];3413
-#endif3414
-3415
- return simde__m128_from_private(r_);3416
-#endif3417
-}3418
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3419
-#define _mm_unpackhi_ps(a, b) simde_mm_unpackhi_ps((a), (b))3420
-#endif3421
-3422
-SIMDE_FUNCTION_ATTRIBUTES3423
-simde__m128 simde_mm_unpacklo_ps(simde__m128 a, simde__m128 b)3424
-{3425
-#if defined(SIMDE_X86_SSE_NATIVE)3426
- return _mm_unpacklo_ps(a, b);3427
-#else3428
- simde__m128_private r_, a_ = simde__m128_to_private(a),3429
- b_ = simde__m128_to_private(b);3430
-3431
-#if defined(SIMDE_SHUFFLE_VECTOR_)3432
- r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 0, 4, 1, 5);3433
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3434
- float32x2_t a1 = vget_low_f32(a_.neon_f32);3435
- float32x2_t b1 = vget_low_f32(b_.neon_f32);3436
- float32x2x2_t result = vzip_f32(a1, b1);3437
- r_.neon_f32 = vcombine_f32(result.val[0], result.val[1]);3438
-#else3439
- r_.f32[0] = a_.f32[0];3440
- r_.f32[1] = b_.f32[0];3441
- r_.f32[2] = a_.f32[1];3442
- r_.f32[3] = b_.f32[1];3443
-#endif3444
-3445
- return simde__m128_from_private(r_);3446
-#endif3447
-}3448
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3449
-#define _mm_unpacklo_ps(a, b) simde_mm_unpacklo_ps((a), (b))3450
-#endif3451
-3452
-SIMDE_FUNCTION_ATTRIBUTES3453
-simde__m128 simde_mm_xor_ps(simde__m128 a, simde__m128 b)3454
-{3455
-#if defined(SIMDE_X86_SSE_NATIVE)3456
- return _mm_xor_ps(a, b);3457
-#else3458
- simde__m128_private r_, a_ = simde__m128_to_private(a),3459
- b_ = simde__m128_to_private(b);3460
-3461
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3462
- r_.neon_i32 = veorq_s32(a_.neon_i32, b_.neon_i32);3463
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)3464
- r_.altivec_i32 = vec_xor(a_.altivec_i32, b_.altivec_i32);3465
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3466
- r_.i32f = a_.i32f ^ b_.i32f;3467
-#else3468
- SIMDE_VECTORIZE3469
- for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {3470
- r_.u32[i] = a_.u32[i] ^ b_.u32[i];3471
- }3472
-#endif3473
-3474
- return simde__m128_from_private(r_);3475
-#endif3476
-}3477
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3478
-#define _mm_xor_ps(a, b) simde_mm_xor_ps((a), (b))3479
-#endif3480
-3481
-SIMDE_FUNCTION_ATTRIBUTES3482
-void simde_mm_stream_pi(simde__m64 *mem_addr, simde__m64 a)3483
-{3484
-#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3485
- _mm_stream_pi(HEDLEY_REINTERPRET_CAST(__m64 *, mem_addr), a);3486
-#else3487
- simde__m64_private *dest = HEDLEY_REINTERPRET_CAST(simde__m64_private *,3488
- mem_addr),3489
- a_ = simde__m64_to_private(a);3490
-3491
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3492
- dest->i64[0] = vget_lane_s64(a_.neon_i64, 0);3493
-#else3494
- dest->i64[0] = a_.i64[0];3495
-#endif3496
-#endif3497
-}3498
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3499
-#define _mm_stream_pi(mem_addr, a) simde_mm_stream_pi(mem_addr, (a))3500
-#endif3501
-3502
-SIMDE_FUNCTION_ATTRIBUTES3503
-void simde_mm_stream_ps(simde_float32 mem_addr[4], simde__m128 a)3504
-{3505
- simde_assert_aligned(16, mem_addr);3506
-3507
-#if defined(SIMDE_X86_SSE_NATIVE)3508
- _mm_stream_ps(mem_addr, a);3509
-#else3510
- simde__m128_private a_ = simde__m128_to_private(a);3511
-3512
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3513
- vst1q_f32(SIMDE_ASSUME_ALIGNED(16, mem_addr), a_.neon_f32);3514
-#else3515
- simde_memcpy(SIMDE_ASSUME_ALIGNED(16, mem_addr), &a_, sizeof(a_));3516
-#endif3517
-#endif3518
-}3519
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3520
-#define _mm_stream_ps(mem_addr, a) \3521
- simde_mm_stream_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3522
- float *, simde_float32 *, mem_addr), \3523
- (a))3524
-#endif3525
-3526
-SIMDE_FUNCTION_ATTRIBUTES3527
-uint32_t simde_mm_getcsr(void)3528
-{3529
-#if defined(SIMDE_X86_SSE_NATIVE)3530
- return _mm_getcsr();3531
-#else3532
- uint32_t r = 0;3533
-3534
-#if defined(SIMDE_HAVE_FENV_H)3535
- int rounding_mode = fegetround();3536
-3537
- switch (rounding_mode) {3538
-#if defined(FE_TONEAREST)3539
- case FE_TONEAREST:3540
- break;3541
-#endif3542
-#if defined(FE_UPWARD)3543
- case FE_UPWARD:3544
- r |= 2 << 13;3545
- break;3546
-#endif3547
-#if defined(FE_DOWNWARD)3548
- case FE_DOWNWARD:3549
- r |= 1 << 13;3550
- break;3551
-#endif3552
-#if defined(FE_TOWARDZERO)3553
- case FE_TOWARDZERO:3554
- r = 3 << 13;3555
- break;3556
-#endif3557
- }3558
-#else3559
- HEDLEY_UNREACHABLE();3560
-#endif3561
-3562
- return r;3563
-#endif3564
-}3565
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3566
-#define _mm_getcsr() simde_mm_getcsr()3567
-#endif3568
-3569
-SIMDE_FUNCTION_ATTRIBUTES3570
-void simde_mm_setcsr(uint32_t a)3571
-{3572
-#if defined(SIMDE_X86_SSE_NATIVE)3573
- _mm_setcsr(a);3574
-#else3575
- switch ((a >> 13) & 3) {3576
-#if defined(FE_TONEAREST)3577
- case 0:3578
- fesetround(FE_TONEAREST);3579
-#endif3580
-#if defined(FE_DOWNWARD)3581
- break;3582
- case 1:3583
- fesetround(FE_DOWNWARD);3584
-#endif3585
-#if defined(FE_UPWARD)3586
- break;3587
- case 2:3588
- fesetround(FE_UPWARD);3589
-#endif3590
-#if defined(FE_TOWARDZERO)3591
- break;3592
- case 3:3593
- fesetround(FE_TOWARDZERO);3594
- break;3595
-#endif3596
- }3597
-#endif3598
-}3599
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3600
-#define _mm_setcsr(a) simde_mm_setcsr(a)3601
-#endif3602
-3603
-#define SIMDE_MM_TRANSPOSE4_PS(row0, row1, row2, row3) \3604
- do { \3605
- simde__m128 tmp3, tmp2, tmp1, tmp0; \3606
- tmp0 = simde_mm_unpacklo_ps((row0), (row1)); \3607
- tmp2 = simde_mm_unpacklo_ps((row2), (row3)); \3608
- tmp1 = simde_mm_unpackhi_ps((row0), (row1)); \3609
- tmp3 = simde_mm_unpackhi_ps((row2), (row3)); \3610
- row0 = simde_mm_movelh_ps(tmp0, tmp2); \3611
- row1 = simde_mm_movehl_ps(tmp2, tmp0); \3612
- row2 = simde_mm_movelh_ps(tmp1, tmp3); \3613
- row3 = simde_mm_movehl_ps(tmp3, tmp1); \3614
- } while (0)3615
-3616
-#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3617
-#define _MM_TRANSPOSE4_PS(row0, row1, row2, row3) \3618
- SIMDE_MM_TRANSPOSE4_PS(row0, row1, row2, row3)3619
-#endif3620
-3621
-#if defined(_MM_EXCEPT_INVALID)3622
-#define SIMDE_MM_EXCEPT_INVALID _MM_EXCEPT_INVALID3623
-#else3624
-#define SIMDE_MM_EXCEPT_INVALID (0x0001)3625
-#endif3626
-#if defined(_MM_EXCEPT_DENORM)3627
-#define SIMDE_MM_EXCEPT_DENORM _MM_EXCEPT_DENORM3628
-#else3629
-#define SIMDE_MM_EXCEPT_DENORM (0x0002)3630
-#endif3631
-#if defined(_MM_EXCEPT_DIV_ZERO)3632
-#define SIMDE_MM_EXCEPT_DIV_ZERO _MM_EXCEPT_DIV_ZERO3633
-#else3634
-#define SIMDE_MM_EXCEPT_DIV_ZERO (0x0004)3635
-#endif3636
-#if defined(_MM_EXCEPT_OVERFLOW)3637
-#define SIMDE_MM_EXCEPT_OVERFLOW _MM_EXCEPT_OVERFLOW3638
-#else3639
-#define SIMDE_MM_EXCEPT_OVERFLOW (0x0008)3640
-#endif3641
-#if defined(_MM_EXCEPT_UNDERFLOW)3642
-#define SIMDE_MM_EXCEPT_UNDERFLOW _MM_EXCEPT_UNDERFLOW3643
-#else3644
-#define SIMDE_MM_EXCEPT_UNDERFLOW (0x0010)3645
-#endif3646
-#if defined(_MM_EXCEPT_INEXACT)3647
-#define SIMDE_MM_EXCEPT_INEXACT _MM_EXCEPT_INEXACT3648
-#else3649
-#define SIMDE_MM_EXCEPT_INEXACT (0x0020)3650
-#endif3651
-#if defined(_MM_EXCEPT_MASK)3652
-#define SIMDE_MM_EXCEPT_MASK _MM_EXCEPT_MASK3653
-#else3654
-#define SIMDE_MM_EXCEPT_MASK \3655
- (SIMDE_MM_EXCEPT_INVALID | SIMDE_MM_EXCEPT_DENORM | \3656
- SIMDE_MM_EXCEPT_DIV_ZERO | SIMDE_MM_EXCEPT_OVERFLOW | \3657
- SIMDE_MM_EXCEPT_UNDERFLOW | SIMDE_MM_EXCEPT_INEXACT)3658
-#endif3659
-3660
-#if defined(_MM_MASK_INVALID)3661
-#define SIMDE_MM_MASK_INVALID _MM_MASK_INVALID3662
-#else3663
-#define SIMDE_MM_MASK_INVALID (0x0080)3664
-#endif3665
-#if defined(_MM_MASK_DENORM)3666
-#define SIMDE_MM_MASK_DENORM _MM_MASK_DENORM3667
-#else3668
-#define SIMDE_MM_MASK_DENORM (0x0100)3669
-#endif3670
-#if defined(_MM_MASK_DIV_ZERO)3671
-#define SIMDE_MM_MASK_DIV_ZERO _MM_MASK_DIV_ZERO3672
-#else3673
-#define SIMDE_MM_MASK_DIV_ZERO (0x0200)3674
-#endif3675
-#if defined(_MM_MASK_OVERFLOW)3676
-#define SIMDE_MM_MASK_OVERFLOW _MM_MASK_OVERFLOW3677
-#else3678
-#define SIMDE_MM_MASK_OVERFLOW (0x0400)3679
-#endif3680
-#if defined(_MM_MASK_UNDERFLOW)3681
-#define SIMDE_MM_MASK_UNDERFLOW _MM_MASK_UNDERFLOW3682
-#else3683
-#define SIMDE_MM_MASK_UNDERFLOW (0x0800)3684
-#endif3685
-#if defined(_MM_MASK_INEXACT)3686
-#define SIMDE_MM_MASK_INEXACT _MM_MASK_INEXACT3687
-#else3688
-#define SIMDE_MM_MASK_INEXACT (0x1000)3689
-#endif3690
-#if defined(_MM_MASK_MASK)3691
-#define SIMDE_MM_MASK_MASK _MM_MASK_MASK3692
-#else3693
-#define SIMDE_MM_MASK_MASK \3694
- (SIMDE_MM_MASK_INVALID | SIMDE_MM_MASK_DENORM | \3695
- SIMDE_MM_MASK_DIV_ZERO | SIMDE_MM_MASK_OVERFLOW | \3696
- SIMDE_MM_MASK_UNDERFLOW | SIMDE_MM_MASK_INEXACT)3697
-#endif3698
-3699
-#if defined(_MM_FLUSH_ZERO_MASK)3700
-#define SIMDE_MM_FLUSH_ZERO_MASK _MM_FLUSH_ZERO_MASK3701
-#else3702
-#define SIMDE_MM_FLUSH_ZERO_MASK (0x8000)3703
-#endif3704
-#if defined(_MM_FLUSH_ZERO_ON)3705
-#define SIMDE_MM_FLUSH_ZERO_ON _MM_FLUSH_ZERO_ON3706
-#else3707
-#define SIMDE_MM_FLUSH_ZERO_ON (0x8000)3708
-#endif3709
-#if defined(_MM_FLUSH_ZERO_OFF)3710
-#define SIMDE_MM_FLUSH_ZERO_OFF _MM_FLUSH_ZERO_OFF3711
-#else3712
-#define SIMDE_MM_FLUSH_ZERO_OFF (0x0000)3713
-#endif3714
-3715
-SIMDE_END_DECLS_3716
-3717
-HEDLEY_DIAGNOSTIC_POP3718
-3719
-#endif /* !defined(SIMDE_X86_SSE_H) */3720
obs-studio-26.1.0.tar.xz/libobs/util/simde/sse2.h
Deleted
6216
1
2
-/* SPDX-License-Identifier: MIT3
- *4
- * Permission is hereby granted, free of charge, to any person5
- * obtaining a copy of this software and associated documentation6
- * files (the "Software"), to deal in the Software without7
- * restriction, including without limitation the rights to use, copy,8
- * modify, merge, publish, distribute, sublicense, and/or sell copies9
- * of the Software, and to permit persons to whom the Software is10
- * furnished to do so, subject to the following conditions:11
- *12
- * The above copyright notice and this permission notice shall be13
- * included in all copies or substantial portions of the Software.14
- *15
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,16
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF17
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND18
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS19
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN20
- * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN21
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE22
- * SOFTWARE.23
- *24
- * Copyright:25
- * 2017-2020 Evan Nemerson <evan@nemerson.com>26
- * 2015-2017 John W. Ratcliff <jratcliffscarab@gmail.com>27
- * 2015 Brandon Rowlett <browlett@nvidia.com>28
- * 2015 Ken Fast <kfast@gdeb.com>29
- * 2017 Hasindu Gamaarachchi <hasindu@unsw.edu.au>30
- * 2018 Jeff Daily <jeff.daily@amd.com>31
- */32
-33
-#if !defined(SIMDE_X86_SSE2_H)34
-#define SIMDE_X86_SSE2_H35
-36
-#include "sse.h"37
-38
-#if !defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ENABLE_NATIVE_ALIASES)39
-#define SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES40
-#endif41
-42
-HEDLEY_DIAGNOSTIC_PUSH43
-SIMDE_DISABLE_UNWANTED_DIAGNOSTICS44
-SIMDE_BEGIN_DECLS_45
-46
-typedef union {47
-#if defined(SIMDE_VECTOR_SUBSCRIPT)48
- SIMDE_ALIGN(16) int8_t i8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;49
- SIMDE_ALIGN(16) int16_t i16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;50
- SIMDE_ALIGN(16) int32_t i32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;51
- SIMDE_ALIGN(16) int64_t i64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;52
- SIMDE_ALIGN(16) uint8_t u8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;53
- SIMDE_ALIGN(16) uint16_t u16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;54
- SIMDE_ALIGN(16) uint32_t u32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;55
- SIMDE_ALIGN(16) uint64_t u64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;56
-#if defined(SIMDE_HAVE_INT128_)57
- SIMDE_ALIGN(16) simde_int128 i128 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;58
- SIMDE_ALIGN(16) simde_uint128 u128 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;59
-#endif60
- SIMDE_ALIGN(16) simde_float32 f32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;61
- SIMDE_ALIGN(16) simde_float64 f64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;62
-63
- SIMDE_ALIGN(16) int_fast32_t i32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;64
- SIMDE_ALIGN(16) uint_fast32_t u32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;65
-#else66
- SIMDE_ALIGN(16) int8_t i8[16];67
- SIMDE_ALIGN(16) int16_t i16[8];68
- SIMDE_ALIGN(16) int32_t i32[4];69
- SIMDE_ALIGN(16) int64_t i64[2];70
- SIMDE_ALIGN(16) uint8_t u8[16];71
- SIMDE_ALIGN(16) uint16_t u16[8];72
- SIMDE_ALIGN(16) uint32_t u32[4];73
- SIMDE_ALIGN(16) uint64_t u64[2];74
-#if defined(SIMDE_HAVE_INT128_)75
- SIMDE_ALIGN(16) simde_int128 i128[1];76
- SIMDE_ALIGN(16) simde_uint128 u128[1];77
-#endif78
- SIMDE_ALIGN(16) simde_float32 f32[4];79
- SIMDE_ALIGN(16) simde_float64 f64[2];80
-81
- SIMDE_ALIGN(16) int_fast32_t i32f[16 / sizeof(int_fast32_t)];82
- SIMDE_ALIGN(16) uint_fast32_t u32f[16 / sizeof(uint_fast32_t)];83
-#endif84
-85
- SIMDE_ALIGN(16) simde__m64_private m64_private[2];86
- SIMDE_ALIGN(16) simde__m64 m64[2];87
-88
-#if defined(SIMDE_X86_SSE2_NATIVE)89
- SIMDE_ALIGN(16) __m128i n;90
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)91
- SIMDE_ALIGN(16) int8x16_t neon_i8;92
- SIMDE_ALIGN(16) int16x8_t neon_i16;93
- SIMDE_ALIGN(16) int32x4_t neon_i32;94
- SIMDE_ALIGN(16) int64x2_t neon_i64;95
- SIMDE_ALIGN(16) uint8x16_t neon_u8;96
- SIMDE_ALIGN(16) uint16x8_t neon_u16;97
- SIMDE_ALIGN(16) uint32x4_t neon_u32;98
- SIMDE_ALIGN(16) uint64x2_t neon_u64;99
- SIMDE_ALIGN(16) float32x4_t neon_f32;100
-#if defined(SIMDE_ARCH_AARCH64)101
- SIMDE_ALIGN(16) float64x2_t neon_f64;102
-#endif103
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)104
- SIMDE_ALIGN(16) v128_t wasm_v128;105
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)106
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed char) altivec_i8;107
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed short) altivec_i16;108
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32;109
-#if defined(__UINT_FAST32_TYPE__)110
- SIMDE_ALIGN(16)111
- SIMDE_POWER_ALTIVEC_VECTOR(__INT_FAST32_TYPE__) altivec_i32f;112
-#else113
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32f;114
-#endif115
- SIMDE_ALIGN(16)116
- SIMDE_POWER_ALTIVEC_VECTOR(signed long long) altivec_i64;117
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned char) altivec_u8;118
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned short) altivec_u16;119
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32;120
-#if defined(__UINT_FAST32_TYPE__)121
- SIMDE_ALIGN(16) vector __UINT_FAST32_TYPE__ altivec_u32f;122
-#else123
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32f;124
-#endif125
- SIMDE_ALIGN(16)126
- SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long) altivec_u64;127
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(float) altivec_f32;128
-#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)129
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(double) altivec_f64;130
-#endif131
-#endif132
-} simde__m128i_private;133
-134
-typedef union {135
-#if defined(SIMDE_VECTOR_SUBSCRIPT)136
- SIMDE_ALIGN(16) int8_t i8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;137
- SIMDE_ALIGN(16) int16_t i16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;138
- SIMDE_ALIGN(16) int32_t i32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;139
- SIMDE_ALIGN(16) int64_t i64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;140
- SIMDE_ALIGN(16) uint8_t u8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;141
- SIMDE_ALIGN(16) uint16_t u16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;142
- SIMDE_ALIGN(16) uint32_t u32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;143
- SIMDE_ALIGN(16) uint64_t u64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;144
- SIMDE_ALIGN(16) simde_float32 f32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;145
- SIMDE_ALIGN(16) simde_float64 f64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;146
- SIMDE_ALIGN(16) int_fast32_t i32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;147
- SIMDE_ALIGN(16) uint_fast32_t u32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;148
-#else149
- SIMDE_ALIGN(16) int8_t i8[16];150
- SIMDE_ALIGN(16) int16_t i16[8];151
- SIMDE_ALIGN(16) int32_t i32[4];152
- SIMDE_ALIGN(16) int64_t i64[2];153
- SIMDE_ALIGN(16) uint8_t u8[16];154
- SIMDE_ALIGN(16) uint16_t u16[8];155
- SIMDE_ALIGN(16) uint32_t u32[4];156
- SIMDE_ALIGN(16) uint64_t u64[2];157
- SIMDE_ALIGN(16) simde_float32 f32[4];158
- SIMDE_ALIGN(16) simde_float64 f64[2];159
- SIMDE_ALIGN(16) int_fast32_t i32f[16 / sizeof(int_fast32_t)];160
- SIMDE_ALIGN(16) uint_fast32_t u32f[16 / sizeof(uint_fast32_t)];161
-#endif162
-163
- SIMDE_ALIGN(16) simde__m64_private m64_private[2];164
- SIMDE_ALIGN(16) simde__m64 m64[2];165
-166
-#if defined(SIMDE_X86_SSE2_NATIVE)167
- SIMDE_ALIGN(16) __m128d n;168
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)169
- SIMDE_ALIGN(16) int8x16_t neon_i8;170
- SIMDE_ALIGN(16) int16x8_t neon_i16;171
- SIMDE_ALIGN(16) int32x4_t neon_i32;172
- SIMDE_ALIGN(16) int64x2_t neon_i64;173
- SIMDE_ALIGN(16) uint8x16_t neon_u8;174
- SIMDE_ALIGN(16) uint16x8_t neon_u16;175
- SIMDE_ALIGN(16) uint32x4_t neon_u32;176
- SIMDE_ALIGN(16) uint64x2_t neon_u64;177
- SIMDE_ALIGN(16) float32x4_t neon_f32;178
-#if defined(SIMDE_ARCH_AARCH64)179
- SIMDE_ALIGN(16) float64x2_t neon_f64;180
-#endif181
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)182
- SIMDE_ALIGN(16) v128_t wasm_v128;183
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)184
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed char) altivec_i8;185
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed short) altivec_i16;186
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32;187
-#if defined(__INT_FAST32_TYPE__)188
- SIMDE_ALIGN(16)189
- SIMDE_POWER_ALTIVEC_VECTOR(__INT_FAST32_TYPE__) altivec_i32f;190
-#else191
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32f;192
-#endif193
- SIMDE_ALIGN(16)194
- SIMDE_POWER_ALTIVEC_VECTOR(signed long long) altivec_i64;195
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned char) altivec_u8;196
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned short) altivec_u16;197
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32;198
-#if defined(__UINT_FAST32_TYPE__)199
- SIMDE_ALIGN(16) vector __UINT_FAST32_TYPE__ altivec_u32f;200
-#else201
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32f;202
-#endif203
- SIMDE_ALIGN(16)204
- SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long) altivec_u64;205
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(float) altivec_f32;206
-#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)207
- SIMDE_ALIGN(16) SIMDE_POWER_ALTIVEC_VECTOR(double) altivec_f64;208
-#endif209
-#endif210
-} simde__m128d_private;211
-212
-#if defined(SIMDE_X86_SSE2_NATIVE)213
-typedef __m128i simde__m128i;214
-typedef __m128d simde__m128d;215
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)216
-typedef int64x2_t simde__m128i;217
-#if defined(SIMDE_ARCH_AARCH64)218
-typedef float64x2_t simde__m128d;219
-#elif defined(SIMDE_VECTOR_SUBSCRIPT)220
-typedef simde_float64 simde__m128d SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;221
-#else222
-typedef simde__m128d_private simde__m128d;223
-#endif224
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)225
-typedef v128_t simde__m128i;226
-typedef v128_t simde__m128d;227
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)228
-typedef SIMDE_POWER_ALTIVEC_VECTOR(float) simde__m128i;229
-typedef SIMDE_POWER_ALTIVEC_VECTOR(double) simde__m128d;230
-#elif defined(SIMDE_VECTOR_SUBSCRIPT)231
-typedef int_fast32_t simde__m128i SIMDE_ALIGN(16)232
- SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;233
-typedef simde_float64 simde__m128d SIMDE_ALIGN(16)234
- SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;235
-#else236
-typedef simde__m128i_private simde__m128i;237
-typedef simde__m128d_private simde__m128d;238
-#endif239
-240
-#if !defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ENABLE_NATIVE_ALIASES)241
-#define SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES242
-typedef simde__m128i __m128i;243
-typedef simde__m128d __m128d;244
-#endif245
-246
-HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128i), "simde__m128i size incorrect");247
-HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128i_private),248
- "simde__m128i_private size incorrect");249
-HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128d), "simde__m128d size incorrect");250
-HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128d_private),251
- "simde__m128d_private size incorrect");252
-#if defined(SIMDE_CHECK_ALIGNMENT) && defined(SIMDE_ALIGN_OF)253
-HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128i) == 16,254
- "simde__m128i is not 16-byte aligned");255
-HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128i_private) == 16,256
- "simde__m128i_private is not 16-byte aligned");257
-HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128d) == 16,258
- "simde__m128d is not 16-byte aligned");259
-HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128d_private) == 16,260
- "simde__m128d_private is not 16-byte aligned");261
-#endif262
-263
-SIMDE_FUNCTION_ATTRIBUTES264
-simde__m128i simde__m128i_from_private(simde__m128i_private v)265
-{266
- simde__m128i r;267
- simde_memcpy(&r, &v, sizeof(r));268
- return r;269
-}270
-271
-SIMDE_FUNCTION_ATTRIBUTES272
-simde__m128i_private simde__m128i_to_private(simde__m128i v)273
-{274
- simde__m128i_private r;275
- simde_memcpy(&r, &v, sizeof(r));276
- return r;277
-}278
-279
-SIMDE_FUNCTION_ATTRIBUTES280
-simde__m128d simde__m128d_from_private(simde__m128d_private v)281
-{282
- simde__m128d r;283
- simde_memcpy(&r, &v, sizeof(r));284
- return r;285
-}286
-287
-SIMDE_FUNCTION_ATTRIBUTES288
-simde__m128d_private simde__m128d_to_private(simde__m128d v)289
-{290
- simde__m128d_private r;291
- simde_memcpy(&r, &v, sizeof(r));292
- return r;293
-}294
-295
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)296
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, int8x16_t, neon, i8)297
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, int16x8_t, neon, i16)298
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, int32x4_t, neon, i32)299
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, int64x2_t, neon, i64)300
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, uint8x16_t, neon, u8)301
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, uint16x8_t, neon, u16)302
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, uint32x4_t, neon, u32)303
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, uint64x2_t, neon, u64)304
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, float32x4_t, neon, f32)305
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)306
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, float64x2_t, neon, f64)307
-#endif308
-#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */309
-310
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)311
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, int8x16_t, neon, i8)312
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, int16x8_t, neon, i16)313
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, int32x4_t, neon, i32)314
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, int64x2_t, neon, i64)315
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, uint8x16_t, neon, u8)316
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, uint16x8_t, neon, u16)317
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, uint32x4_t, neon, u32)318
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, uint64x2_t, neon, u64)319
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, float32x4_t, neon, f32)320
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)321
-SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, float64x2_t, neon, f64)322
-#endif323
-#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */324
-325
-SIMDE_FUNCTION_ATTRIBUTES326
-simde__m128i simde_mm_add_epi8(simde__m128i a, simde__m128i b)327
-{328
-#if defined(SIMDE_X86_SSE2_NATIVE)329
- return _mm_add_epi8(a, b);330
-#else331
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),332
- b_ = simde__m128i_to_private(b);333
-334
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)335
- r_.neon_i8 = vaddq_s8(a_.neon_i8, b_.neon_i8);336
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)337
- r_.altivec_i8 = vec_add(a_.altivec_i8, b_.altivec_i8);338
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)339
- r_.i8 = a_.i8 + b_.i8;340
-#else341
- SIMDE_VECTORIZE342
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {343
- r_.i8[i] = a_.i8[i] + b_.i8[i];344
- }345
-#endif346
-347
- return simde__m128i_from_private(r_);348
-#endif349
-}350
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)351
-#define _mm_add_epi8(a, b) simde_mm_add_epi8(a, b)352
-#endif353
-354
-SIMDE_FUNCTION_ATTRIBUTES355
-simde__m128i simde_mm_add_epi16(simde__m128i a, simde__m128i b)356
-{357
-#if defined(SIMDE_X86_SSE2_NATIVE)358
- return _mm_add_epi16(a, b);359
-#else360
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),361
- b_ = simde__m128i_to_private(b);362
-363
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)364
- r_.neon_i16 = vaddq_s16(a_.neon_i16, b_.neon_i16);365
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)366
- r_.altivec_i16 = vec_add(a_.altivec_i16, b_.altivec_i16);367
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)368
- r_.i16 = a_.i16 + b_.i16;369
-#else370
- SIMDE_VECTORIZE371
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {372
- r_.i16[i] = a_.i16[i] + b_.i16[i];373
- }374
-#endif375
-376
- return simde__m128i_from_private(r_);377
-#endif378
-}379
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)380
-#define _mm_add_epi16(a, b) simde_mm_add_epi16(a, b)381
-#endif382
-383
-SIMDE_FUNCTION_ATTRIBUTES384
-simde__m128i simde_mm_add_epi32(simde__m128i a, simde__m128i b)385
-{386
-#if defined(SIMDE_X86_SSE2_NATIVE)387
- return _mm_add_epi32(a, b);388
-#else389
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),390
- b_ = simde__m128i_to_private(b);391
-392
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)393
- r_.neon_i32 = vaddq_s32(a_.neon_i32, b_.neon_i32);394
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)395
- r_.altivec_i32 = vec_add(a_.altivec_i32, b_.altivec_i32);396
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)397
- r_.i32 = a_.i32 + b_.i32;398
-#else399
- SIMDE_VECTORIZE400
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {401
- r_.i32[i] = a_.i32[i] + b_.i32[i];402
- }403
-#endif404
-405
- return simde__m128i_from_private(r_);406
-#endif407
-}408
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)409
-#define _mm_add_epi32(a, b) simde_mm_add_epi32(a, b)410
-#endif411
-412
-SIMDE_FUNCTION_ATTRIBUTES413
-simde__m128i simde_mm_add_epi64(simde__m128i a, simde__m128i b)414
-{415
-#if defined(SIMDE_X86_SSE2_NATIVE)416
- return _mm_add_epi64(a, b);417
-#else418
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),419
- b_ = simde__m128i_to_private(b);420
-421
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)422
- r_.neon_i64 = vaddq_s64(a_.neon_i64, b_.neon_i64);423
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)424
- r_.altivec_i64 = vec_add(a_.altivec_i64, b_.altivec_i64);425
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)426
- r_.i64 = a_.i64 + b_.i64;427
-#else428
- SIMDE_VECTORIZE429
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {430
- r_.i64[i] = a_.i64[i] + b_.i64[i];431
- }432
-#endif433
-434
- return simde__m128i_from_private(r_);435
-#endif436
-}437
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)438
-#define _mm_add_epi64(a, b) simde_mm_add_epi64(a, b)439
-#endif440
-441
-SIMDE_FUNCTION_ATTRIBUTES442
-simde__m128d simde_mm_add_pd(simde__m128d a, simde__m128d b)443
-{444
-#if defined(SIMDE_X86_SSE2_NATIVE)445
- return _mm_add_pd(a, b);446
-#else447
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),448
- b_ = simde__m128d_to_private(b);449
-450
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)451
- r_.neon_f64 = vaddq_f64(a_.neon_f64, b_.neon_f64);452
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)453
- r_.wasm_v128 = wasm_f64x2_add(a_.wasm_v128, b_.wasm_v128);454
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)455
- r_.altivec_f64 = vec_add(a_.altivec_f64, b_.altivec_f64);456
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)457
- r_.f64 = a_.f64 + b_.f64;458
-#else459
- SIMDE_VECTORIZE460
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {461
- r_.f64[i] = a_.f64[i] + b_.f64[i];462
- }463
-#endif464
-465
- return simde__m128d_from_private(r_);466
-#endif467
-}468
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)469
-#define _mm_add_pd(a, b) simde_mm_add_pd(a, b)470
-#endif471
-472
-SIMDE_FUNCTION_ATTRIBUTES473
-simde__m128d simde_mm_move_sd(simde__m128d a, simde__m128d b)474
-{475
-#if defined(SIMDE_X86_SSE2_NATIVE)476
- return _mm_move_sd(a, b);477
-#else478
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),479
- b_ = simde__m128d_to_private(b);480
-481
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)482
- r_.neon_f64 =483
- vsetq_lane_f64(vgetq_lane_f64(b_.neon_f64, 0), a_.neon_f64, 0);484
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)485
- SIMDE_POWER_ALTIVEC_VECTOR(unsigned char)486
- m = {16, 17, 18, 19, 20, 21, 22, 23, 8, 9, 10, 11, 12, 13, 14, 15};487
- r_.altivec_f64 = vec_perm(a_.altivec_f64, b_.altivec_f64, m);488
-#elif defined(SIMDE_SHUFFLE_VECTOR_)489
- r_.f64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.f64, b_.f64, 2, 1);490
-#else491
- r_.f64[0] = b_.f64[0];492
- r_.f64[1] = a_.f64[1];493
-#endif494
-495
- return simde__m128d_from_private(r_);496
-#endif497
-}498
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)499
-#define _mm_move_sd(a, b) simde_mm_move_sd(a, b)500
-#endif501
-502
-SIMDE_FUNCTION_ATTRIBUTES503
-simde__m128d simde_mm_add_sd(simde__m128d a, simde__m128d b)504
-{505
-#if defined(SIMDE_X86_SSE2_NATIVE)506
- return _mm_add_sd(a, b);507
-#else508
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),509
- b_ = simde__m128d_to_private(b);510
-511
- r_.f64[0] = a_.f64[0] + b_.f64[0];512
- r_.f64[1] = a_.f64[1];513
-514
-#if defined(SIMDE_ASSUME_VECTORIZATION)515
- return simde_mm_move_sd(a, simde_mm_add_pd(a, b));516
-#else517
- r_.f64[0] = a_.f64[0] + b_.f64[0];518
- r_.f64[1] = a_.f64[1];519
-#endif520
-521
- return simde__m128d_from_private(r_);522
-#endif523
-}524
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)525
-#define _mm_add_sd(a, b) simde_mm_add_sd(a, b)526
-#endif527
-528
-SIMDE_FUNCTION_ATTRIBUTES529
-simde__m64 simde_mm_add_si64(simde__m64 a, simde__m64 b)530
-{531
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)532
- return _mm_add_si64(a, b);533
-#else534
- simde__m64_private r_, a_ = simde__m64_to_private(a),535
- b_ = simde__m64_to_private(b);536
-537
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)538
- r_.neon_i64 = vadd_s64(a_.neon_i64, b_.neon_i64);539
-#else540
- r_.i64[0] = a_.i64[0] + b_.i64[0];541
-#endif542
-543
- return simde__m64_from_private(r_);544
-#endif545
-}546
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)547
-#define _mm_add_si64(a, b) simde_mm_add_si64(a, b)548
-#endif549
-550
-SIMDE_FUNCTION_ATTRIBUTES551
-simde__m128i simde_mm_adds_epi8(simde__m128i a, simde__m128i b)552
-{553
-#if defined(SIMDE_X86_SSE2_NATIVE)554
- return _mm_adds_epi8(a, b);555
-#else556
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),557
- b_ = simde__m128i_to_private(b);558
-559
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)560
- r_.neon_i8 = vqaddq_s8(a_.neon_i8, b_.neon_i8);561
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)562
- r_.altivec_i8 = vec_adds(a_.altivec_i8, b_.altivec_i8);563
-#else564
- SIMDE_VECTORIZE565
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {566
- const int32_t tmp = HEDLEY_STATIC_CAST(int16_t, a_.i8[i]) +567
- HEDLEY_STATIC_CAST(int16_t, b_.i8[i]);568
- r_.i8[i] = HEDLEY_STATIC_CAST(569
- int8_t,570
- ((tmp < INT8_MAX) ? ((tmp > INT8_MIN) ? tmp : INT8_MIN)571
- : INT8_MAX));572
- }573
-#endif574
-575
- return simde__m128i_from_private(r_);576
-#endif577
-}578
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)579
-#define _mm_adds_epi8(a, b) simde_mm_adds_epi8(a, b)580
-#endif581
-582
-SIMDE_FUNCTION_ATTRIBUTES583
-simde__m128i simde_mm_adds_epi16(simde__m128i a, simde__m128i b)584
-{585
-#if defined(SIMDE_X86_SSE2_NATIVE)586
- return _mm_adds_epi16(a, b);587
-#else588
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),589
- b_ = simde__m128i_to_private(b);590
-591
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)592
- r_.neon_i16 = vqaddq_s16(a_.neon_i16, b_.neon_i16);593
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)594
- r_.altivec_i16 = vec_adds(a_.altivec_i16, b_.altivec_i16);595
-#else596
- SIMDE_VECTORIZE597
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {598
- const int32_t tmp = HEDLEY_STATIC_CAST(int32_t, a_.i16[i]) +599
- HEDLEY_STATIC_CAST(int32_t, b_.i16[i]);600
- r_.i16[i] = HEDLEY_STATIC_CAST(601
- int16_t,602
- ((tmp < INT16_MAX)603
- ? ((tmp > INT16_MIN) ? tmp : INT16_MIN)604
- : INT16_MAX));605
- }606
-#endif607
-608
- return simde__m128i_from_private(r_);609
-#endif610
-}611
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)612
-#define _mm_adds_epi16(a, b) simde_mm_adds_epi16(a, b)613
-#endif614
-615
-SIMDE_FUNCTION_ATTRIBUTES616
-simde__m128i simde_mm_adds_epu8(simde__m128i a, simde__m128i b)617
-{618
-#if defined(SIMDE_X86_SSE2_NATIVE)619
- return _mm_adds_epu8(a, b);620
-#else621
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),622
- b_ = simde__m128i_to_private(b);623
-624
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)625
- r_.neon_u8 = vqaddq_u8(a_.neon_u8, b_.neon_u8);626
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)627
- r_.altivec_u8 = vec_adds(a_.altivec_u8, b_.altivec_u8);628
-#else629
- SIMDE_VECTORIZE630
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {631
- r_.u8[i] = ((UINT8_MAX - a_.u8[i]) > b_.u8[i])632
- ? (a_.u8[i] + b_.u8[i])633
- : UINT8_MAX;634
- }635
-#endif636
-637
- return simde__m128i_from_private(r_);638
-#endif639
-}640
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)641
-#define _mm_adds_epu8(a, b) simde_mm_adds_epu8(a, b)642
-#endif643
-644
-SIMDE_FUNCTION_ATTRIBUTES645
-simde__m128i simde_mm_adds_epu16(simde__m128i a, simde__m128i b)646
-{647
-#if defined(SIMDE_X86_SSE2_NATIVE)648
- return _mm_adds_epu16(a, b);649
-#else650
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),651
- b_ = simde__m128i_to_private(b);652
-653
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)654
- r_.neon_u16 = vqaddq_u16(a_.neon_u16, b_.neon_u16);655
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)656
- r_.altivec_u16 = vec_adds(a_.altivec_u16, b_.altivec_u16);657
-#else658
- SIMDE_VECTORIZE659
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {660
- r_.u16[i] = ((UINT16_MAX - a_.u16[i]) > b_.u16[i])661
- ? (a_.u16[i] + b_.u16[i])662
- : UINT16_MAX;663
- }664
-#endif665
-666
- return simde__m128i_from_private(r_);667
-#endif668
-}669
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)670
-#define _mm_adds_epu16(a, b) simde_mm_adds_epu16(a, b)671
-#endif672
-673
-SIMDE_FUNCTION_ATTRIBUTES674
-simde__m128d simde_mm_and_pd(simde__m128d a, simde__m128d b)675
-{676
-#if defined(SIMDE_X86_SSE2_NATIVE)677
- return _mm_and_pd(a, b);678
-#else679
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),680
- b_ = simde__m128d_to_private(b);681
-682
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)683
- r_.neon_i32 = vandq_s32(a_.neon_i32, b_.neon_i32);684
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)685
- r_.wasm_v128 = wasm_v128_and(a_.wasm_v128, b_.wasm_v128);686
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)687
- r_.altivec_f64 = vec_and(a_.altivec_f64, b_.altivec_f64);688
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)689
- r_.i32f = a_.i32f & b_.i32f;690
-#else691
- SIMDE_VECTORIZE692
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {693
- r_.i32f[i] = a_.i32f[i] & b_.i32f[i];694
- }695
-#endif696
-697
- return simde__m128d_from_private(r_);698
-#endif699
-}700
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)701
-#define _mm_and_pd(a, b) simde_mm_and_pd(a, b)702
-#endif703
-704
-SIMDE_FUNCTION_ATTRIBUTES705
-simde__m128i simde_mm_and_si128(simde__m128i a, simde__m128i b)706
-{707
-#if defined(SIMDE_X86_SSE2_NATIVE)708
- return _mm_and_si128(a, b);709
-#else710
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),711
- b_ = simde__m128i_to_private(b);712
-713
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)714
- r_.neon_i32 = vandq_s32(b_.neon_i32, a_.neon_i32);715
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)716
- r_.altivec_u32f = vec_and(a_.altivec_u32f, b_.altivec_u32f);717
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)718
- r_.i32f = a_.i32f & b_.i32f;719
-#else720
- SIMDE_VECTORIZE721
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {722
- r_.i32f[i] = a_.i32f[i] & b_.i32f[i];723
- }724
-#endif725
-726
- return simde__m128i_from_private(r_);727
-#endif728
-}729
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)730
-#define _mm_and_si128(a, b) simde_mm_and_si128(a, b)731
-#endif732
-733
-SIMDE_FUNCTION_ATTRIBUTES734
-simde__m128d simde_mm_andnot_pd(simde__m128d a, simde__m128d b)735
-{736
-#if defined(SIMDE_X86_SSE2_NATIVE)737
- return _mm_andnot_pd(a, b);738
-#else739
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),740
- b_ = simde__m128d_to_private(b);741
-742
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)743
- r_.neon_i32 = vbicq_s32(a_.neon_i32, b_.neon_i32);744
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)745
- r_.wasm_v128 = wasm_v128_andnot(b_.wasm_v128, a_.wasm_v128);746
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)747
- r_.altivec_i32f = vec_andc(a_.altivec_i32f, b_.altivec_i32f);748
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)749
- r_.i32f = ~a_.i32f & b_.i32f;750
-#else751
- SIMDE_VECTORIZE752
- for (size_t i = 0; i < (sizeof(r_.u64) / sizeof(r_.u64[0])); i++) {753
- r_.u64[i] = ~a_.u64[i] & b_.u64[i];754
- }755
-#endif756
-757
- return simde__m128d_from_private(r_);758
-#endif759
-}760
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)761
-#define _mm_andnot_pd(a, b) simde_mm_andnot_pd(a, b)762
-#endif763
-764
-SIMDE_FUNCTION_ATTRIBUTES765
-simde__m128i simde_mm_andnot_si128(simde__m128i a, simde__m128i b)766
-{767
-#if defined(SIMDE_X86_SSE2_NATIVE)768
- return _mm_andnot_si128(a, b);769
-#else770
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),771
- b_ = simde__m128i_to_private(b);772
-773
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)774
- r_.neon_i32 = vbicq_s32(b_.neon_i32, a_.neon_i32);775
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)776
- r_.altivec_i32 = vec_andc(b_.altivec_i32, a_.altivec_i32);777
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)778
- r_.i32f = ~a_.i32f & b_.i32f;779
-#else780
- SIMDE_VECTORIZE781
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {782
- r_.i32f[i] = ~(a_.i32f[i]) & b_.i32f[i];783
- }784
-#endif785
-786
- return simde__m128i_from_private(r_);787
-#endif788
-}789
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)790
-#define _mm_andnot_si128(a, b) simde_mm_andnot_si128(a, b)791
-#endif792
-793
-SIMDE_FUNCTION_ATTRIBUTES794
-simde__m128i simde_mm_avg_epu8(simde__m128i a, simde__m128i b)795
-{796
-#if defined(SIMDE_X86_SSE2_NATIVE)797
- return _mm_avg_epu8(a, b);798
-#else799
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),800
- b_ = simde__m128i_to_private(b);801
-802
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)803
- r_.neon_u8 = vrhaddq_u8(b_.neon_u8, a_.neon_u8);804
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)805
- r_.altivec_u8 = vec_avg(a_.altivec_u8, b_.altivec_u8);806
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS) && \807
- defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \808
- defined(SIMDE_CONVERT_VECTOR_)809
- uint16_t wa SIMDE_VECTOR(32);810
- uint16_t wb SIMDE_VECTOR(32);811
- uint16_t wr SIMDE_VECTOR(32);812
- SIMDE_CONVERT_VECTOR_(wa, a_.u8);813
- SIMDE_CONVERT_VECTOR_(wb, b_.u8);814
- wr = (wa + wb + 1) >> 1;815
- SIMDE_CONVERT_VECTOR_(r_.u8, wr);816
-#else817
- SIMDE_VECTORIZE818
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {819
- r_.u8[i] = (a_.u8[i] + b_.u8[i] + 1) >> 1;820
- }821
-#endif822
-823
- return simde__m128i_from_private(r_);824
-#endif825
-}826
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)827
-#define _mm_avg_epu8(a, b) simde_mm_avg_epu8(a, b)828
-#endif829
-830
-SIMDE_FUNCTION_ATTRIBUTES831
-simde__m128i simde_mm_avg_epu16(simde__m128i a, simde__m128i b)832
-{833
-#if defined(SIMDE_X86_SSE2_NATIVE)834
- return _mm_avg_epu16(a, b);835
-#else836
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),837
- b_ = simde__m128i_to_private(b);838
-839
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)840
- r_.neon_u16 = vrhaddq_u16(b_.neon_u16, a_.neon_u16);841
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)842
- r_.altivec_u16 = vec_avg(a_.altivec_u16, b_.altivec_u16);843
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS) && \844
- defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \845
- defined(SIMDE_CONVERT_VECTOR_)846
- uint32_t wa SIMDE_VECTOR(32);847
- uint32_t wb SIMDE_VECTOR(32);848
- uint32_t wr SIMDE_VECTOR(32);849
- SIMDE_CONVERT_VECTOR_(wa, a_.u16);850
- SIMDE_CONVERT_VECTOR_(wb, b_.u16);851
- wr = (wa + wb + 1) >> 1;852
- SIMDE_CONVERT_VECTOR_(r_.u16, wr);853
-#else854
- SIMDE_VECTORIZE855
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {856
- r_.u16[i] = (a_.u16[i] + b_.u16[i] + 1) >> 1;857
- }858
-#endif859
-860
- return simde__m128i_from_private(r_);861
-#endif862
-}863
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)864
-#define _mm_avg_epu16(a, b) simde_mm_avg_epu16(a, b)865
-#endif866
-867
-SIMDE_FUNCTION_ATTRIBUTES868
-simde__m128i simde_mm_setzero_si128(void)869
-{870
-#if defined(SIMDE_X86_SSE2_NATIVE)871
- return _mm_setzero_si128();872
-#else873
- simde__m128i_private r_;874
-875
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)876
- r_.neon_i32 = vdupq_n_s32(0);877
-#else878
- SIMDE_VECTORIZE879
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {880
- r_.i32f[i] = 0;881
- }882
-#endif883
-884
- return simde__m128i_from_private(r_);885
-#endif886
-}887
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)888
-#define _mm_setzero_si128() (simde_mm_setzero_si128())889
-#endif890
-891
-SIMDE_FUNCTION_ATTRIBUTES892
-simde__m128i simde_mm_bslli_si128(simde__m128i a, const int imm8)893
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)894
-{895
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);896
-897
- if (HEDLEY_UNLIKELY((imm8 & ~15))) {898
- return simde_mm_setzero_si128();899
- }900
-901
-#if defined(SIMDE_HAVE_INT128_) && defined(__BYTE_ORDER__) && \902
- (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) && 0903
- r_.u128[0] = a_.u128[0] << s;904
-#else905
- r_ = simde__m128i_to_private(simde_mm_setzero_si128());906
- for (int i = imm8;907
- i < HEDLEY_STATIC_CAST(int, sizeof(r_.i8) / sizeof(r_.i8[0]));908
- i++) {909
- r_.i8[i] = a_.i8[i - imm8];910
- }911
-#endif912
-913
- return simde__m128i_from_private(r_);914
-}915
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)916
-#define simde_mm_bslli_si128(a, imm8) _mm_slli_si128(a, imm8)917
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) && !defined(__clang__)918
-#define simde_mm_bslli_si128(a, imm8) \919
- simde__m128i_from_neon_i8( \920
- ((imm8) <= 0) \921
- ? simde__m128i_to_neon_i8(a) \922
- : (((imm8) > 15) \923
- ? (vdupq_n_s8(0)) \924
- : (vextq_s8(vdupq_n_s8(0), \925
- simde__m128i_to_neon_i8(a), \926
- 16 - (imm8)))))927
-#elif defined(SIMDE_SHUFFLE_VECTOR_)928
-#define simde_mm_bslli_si128(a, imm8) \929
- (__extension__({ \930
- const simde__m128i_private simde__tmp_a_ = \931
- simde__m128i_to_private(a); \932
- const simde__m128i_private simde__tmp_z_ = \933
- simde__m128i_to_private(simde_mm_setzero_si128()); \934
- simde__m128i_private simde__tmp_r_; \935
- if (HEDLEY_UNLIKELY(imm8 > 15)) { \936
- simde__tmp_r_ = simde__m128i_to_private( \937
- simde_mm_setzero_si128()); \938
- } else { \939
- simde__tmp_r_.i8 = SIMDE_SHUFFLE_VECTOR_( \940
- 8, 16, simde__tmp_z_.i8, (simde__tmp_a_).i8, \941
- HEDLEY_STATIC_CAST(int8_t, (16 - imm8) & 31), \942
- HEDLEY_STATIC_CAST(int8_t, (17 - imm8) & 31), \943
- HEDLEY_STATIC_CAST(int8_t, (18 - imm8) & 31), \944
- HEDLEY_STATIC_CAST(int8_t, (19 - imm8) & 31), \945
- HEDLEY_STATIC_CAST(int8_t, (20 - imm8) & 31), \946
- HEDLEY_STATIC_CAST(int8_t, (21 - imm8) & 31), \947
- HEDLEY_STATIC_CAST(int8_t, (22 - imm8) & 31), \948
- HEDLEY_STATIC_CAST(int8_t, (23 - imm8) & 31), \949
- HEDLEY_STATIC_CAST(int8_t, (24 - imm8) & 31), \950
- HEDLEY_STATIC_CAST(int8_t, (25 - imm8) & 31), \951
- HEDLEY_STATIC_CAST(int8_t, (26 - imm8) & 31), \952
- HEDLEY_STATIC_CAST(int8_t, (27 - imm8) & 31), \953
- HEDLEY_STATIC_CAST(int8_t, (28 - imm8) & 31), \954
- HEDLEY_STATIC_CAST(int8_t, (29 - imm8) & 31), \955
- HEDLEY_STATIC_CAST(int8_t, (30 - imm8) & 31), \956
- HEDLEY_STATIC_CAST(int8_t, (31 - imm8) & 31)); \957
- } \958
- simde__m128i_from_private(simde__tmp_r_); \959
- }))960
-#endif961
-#define simde_mm_slli_si128(a, imm8) simde_mm_bslli_si128(a, imm8)962
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)963
-#define _mm_bslli_si128(a, b) simde_mm_bslli_si128(a, b)964
-#define _mm_slli_si128(a, b) simde_mm_bslli_si128(a, b)965
-#endif966
-967
-SIMDE_FUNCTION_ATTRIBUTES968
-simde__m128i simde_mm_bsrli_si128(simde__m128i a, const int imm8)969
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)970
-{971
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);972
-973
- SIMDE_VECTORIZE974
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {975
- const int e = HEDLEY_STATIC_CAST(int, i) + imm8;976
- r_.i8[i] = (e < 16) ? a_.i8[e] : 0;977
- }978
-979
- return simde__m128i_from_private(r_);980
-}981
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)982
-#define simde_mm_bsrli_si128(a, imm8) _mm_srli_si128(a, imm8)983
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) && !defined(__clang__)984
-#define simde_mm_bsrli_si128(a, imm8) \985
- simde__m128i_from_neon_i8( \986
- ((imm8 < 0) || (imm8 > 15)) \987
- ? vdupq_n_s8(0) \988
- : (vextq_s8(simde__m128i_to_private(a).neon_i8, \989
- vdupq_n_s8(0), \990
- ((imm8 & 15) != 0) ? imm8 : (imm8 & 15))))991
-#elif defined(SIMDE_SHUFFLE_VECTOR_)992
-#define simde_mm_bsrli_si128(a, imm8) \993
- (__extension__({ \994
- const simde__m128i_private simde__tmp_a_ = \995
- simde__m128i_to_private(a); \996
- const simde__m128i_private simde__tmp_z_ = \997
- simde__m128i_to_private(simde_mm_setzero_si128()); \998
- simde__m128i_private simde__tmp_r_ = \999
- simde__m128i_to_private(a); \1000
- if (HEDLEY_UNLIKELY(imm8 > 15)) { \1001
- simde__tmp_r_ = simde__m128i_to_private( \1002
- simde_mm_setzero_si128()); \1003
- } else { \1004
- simde__tmp_r_.i8 = SIMDE_SHUFFLE_VECTOR_( \1005
- 8, 16, simde__tmp_z_.i8, (simde__tmp_a_).i8, \1006
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 16) & 31), \1007
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 17) & 31), \1008
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 18) & 31), \1009
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 19) & 31), \1010
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 20) & 31), \1011
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 21) & 31), \1012
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 22) & 31), \1013
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 23) & 31), \1014
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 24) & 31), \1015
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 25) & 31), \1016
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 26) & 31), \1017
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 27) & 31), \1018
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 28) & 31), \1019
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 29) & 31), \1020
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 30) & 31), \1021
- HEDLEY_STATIC_CAST(int8_t, (imm8 + 31) & 31)); \1022
- } \1023
- simde__m128i_from_private(simde__tmp_r_); \1024
- }))1025
-#endif1026
-#define simde_mm_srli_si128(a, imm8) simde_mm_bsrli_si128((a), (imm8))1027
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1028
-#define _mm_bsrli_si128(a, imm8) simde_mm_bsrli_si128((a), (imm8))1029
-#define _mm_srli_si128(a, imm8) simde_mm_bsrli_si128((a), (imm8))1030
-#endif1031
-1032
-SIMDE_FUNCTION_ATTRIBUTES1033
-void simde_mm_clflush(void const *p)1034
-{1035
-#if defined(SIMDE_X86_SSE2_NATIVE)1036
- _mm_clflush(p);1037
-#else1038
- (void)p;1039
-#endif1040
-}1041
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1042
-#define _mm_clflush(a, b) simde_mm_clflush()1043
-#endif1044
-1045
-SIMDE_FUNCTION_ATTRIBUTES1046
-int simde_mm_comieq_sd(simde__m128d a, simde__m128d b)1047
-{1048
-#if defined(SIMDE_X86_SSE2_NATIVE)1049
- return _mm_comieq_sd(a, b);1050
-#else1051
- simde__m128d_private a_ = simde__m128d_to_private(a),1052
- b_ = simde__m128d_to_private(b);1053
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1054
- return !!vgetq_lane_u64(vceqq_f64(a_.neon_f64, b_.neon_f64), 0);1055
-#else1056
- return a_.f64[0] == b_.f64[0];1057
-#endif1058
-#endif1059
-}1060
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1061
-#define _mm_comieq_sd(a, b) simde_mm_comieq_sd(a, b)1062
-#endif1063
-1064
-SIMDE_FUNCTION_ATTRIBUTES1065
-int simde_mm_comige_sd(simde__m128d a, simde__m128d b)1066
-{1067
-#if defined(SIMDE_X86_SSE2_NATIVE)1068
- return _mm_comige_sd(a, b);1069
-#else1070
- simde__m128d_private a_ = simde__m128d_to_private(a),1071
- b_ = simde__m128d_to_private(b);1072
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1073
- return !!vgetq_lane_u64(vcgeq_f64(a_.neon_f64, b_.neon_f64), 0);1074
-#else1075
- return a_.f64[0] >= b_.f64[0];1076
-#endif1077
-#endif1078
-}1079
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1080
-#define _mm_comige_sd(a, b) simde_mm_comige_sd(a, b)1081
-#endif1082
-1083
-SIMDE_FUNCTION_ATTRIBUTES1084
-int simde_mm_comigt_sd(simde__m128d a, simde__m128d b)1085
-{1086
-#if defined(SIMDE_X86_SSE2_NATIVE)1087
- return _mm_comigt_sd(a, b);1088
-#else1089
- simde__m128d_private a_ = simde__m128d_to_private(a),1090
- b_ = simde__m128d_to_private(b);1091
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1092
- return !!vgetq_lane_u64(vcgtq_f64(a_.neon_f64, b_.neon_f64), 0);1093
-#else1094
- return a_.f64[0] > b_.f64[0];1095
-#endif1096
-#endif1097
-}1098
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1099
-#define _mm_comigt_sd(a, b) simde_mm_comigt_sd(a, b)1100
-#endif1101
-1102
-SIMDE_FUNCTION_ATTRIBUTES1103
-int simde_mm_comile_sd(simde__m128d a, simde__m128d b)1104
-{1105
-#if defined(SIMDE_X86_SSE2_NATIVE)1106
- return _mm_comile_sd(a, b);1107
-#else1108
- simde__m128d_private a_ = simde__m128d_to_private(a),1109
- b_ = simde__m128d_to_private(b);1110
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1111
- return !!vgetq_lane_u64(vcleq_f64(a_.neon_f64, b_.neon_f64), 0);1112
-#else1113
- return a_.f64[0] <= b_.f64[0];1114
-#endif1115
-#endif1116
-}1117
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1118
-#define _mm_comile_sd(a, b) simde_mm_comile_sd(a, b)1119
-#endif1120
-1121
-SIMDE_FUNCTION_ATTRIBUTES1122
-int simde_mm_comilt_sd(simde__m128d a, simde__m128d b)1123
-{1124
-#if defined(SIMDE_X86_SSE2_NATIVE)1125
- return _mm_comilt_sd(a, b);1126
-#else1127
- simde__m128d_private a_ = simde__m128d_to_private(a),1128
- b_ = simde__m128d_to_private(b);1129
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1130
- return !!vgetq_lane_u64(vcltq_f64(a_.neon_f64, b_.neon_f64), 0);1131
-#else1132
- return a_.f64[0] < b_.f64[0];1133
-#endif1134
-#endif1135
-}1136
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1137
-#define _mm_comilt_sd(a, b) simde_mm_comilt_sd(a, b)1138
-#endif1139
-1140
-SIMDE_FUNCTION_ATTRIBUTES1141
-int simde_mm_comineq_sd(simde__m128d a, simde__m128d b)1142
-{1143
-#if defined(SIMDE_X86_SSE2_NATIVE)1144
- return _mm_comineq_sd(a, b);1145
-#else1146
- simde__m128d_private a_ = simde__m128d_to_private(a),1147
- b_ = simde__m128d_to_private(b);1148
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1149
- return !vgetq_lane_u64(vceqq_f64(a_.neon_f64, b_.neon_f64), 0);1150
-#else1151
- return a_.f64[0] != b_.f64[0];1152
-#endif1153
-#endif1154
-}1155
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1156
-#define _mm_comineq_sd(a, b) simde_mm_comineq_sd(a, b)1157
-#endif1158
-1159
-SIMDE_FUNCTION_ATTRIBUTES1160
-simde__m128 simde_mm_castpd_ps(simde__m128d a)1161
-{1162
-#if defined(SIMDE_X86_SSE2_NATIVE)1163
- return _mm_castpd_ps(a);1164
-#else1165
- simde__m128 r;1166
- simde_memcpy(&r, &a, sizeof(a));1167
- return r;1168
-#endif1169
-}1170
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1171
-#define _mm_castpd_ps(a) simde_mm_castpd_ps(a)1172
-#endif1173
-1174
-SIMDE_FUNCTION_ATTRIBUTES1175
-simde__m128i simde_mm_castpd_si128(simde__m128d a)1176
-{1177
-#if defined(SIMDE_X86_SSE2_NATIVE)1178
- return _mm_castpd_si128(a);1179
-#else1180
- simde__m128i r;1181
- simde_memcpy(&r, &a, sizeof(a));1182
- return r;1183
-#endif1184
-}1185
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1186
-#define _mm_castpd_si128(a) simde_mm_castpd_si128(a)1187
-#endif1188
-1189
-SIMDE_FUNCTION_ATTRIBUTES1190
-simde__m128d simde_mm_castps_pd(simde__m128 a)1191
-{1192
-#if defined(SIMDE_X86_SSE2_NATIVE)1193
- return _mm_castps_pd(a);1194
-#else1195
- simde__m128d r;1196
- simde_memcpy(&r, &a, sizeof(a));1197
- return r;1198
-#endif1199
-}1200
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1201
-#define _mm_castps_pd(a) simde_mm_castps_pd(a)1202
-#endif1203
-1204
-SIMDE_FUNCTION_ATTRIBUTES1205
-simde__m128i simde_mm_castps_si128(simde__m128 a)1206
-{1207
-#if defined(SIMDE_X86_SSE2_NATIVE)1208
- return _mm_castps_si128(a);1209
-#else1210
- simde__m128i r;1211
- simde_memcpy(&r, &a, sizeof(a));1212
- return r;1213
-#endif1214
-}1215
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1216
-#define _mm_castps_si128(a) simde_mm_castps_si128(a)1217
-#endif1218
-1219
-SIMDE_FUNCTION_ATTRIBUTES1220
-simde__m128d simde_mm_castsi128_pd(simde__m128i a)1221
-{1222
-#if defined(SIMDE_X86_SSE2_NATIVE)1223
- return _mm_castsi128_pd(a);1224
-#else1225
- simde__m128d r;1226
- simde_memcpy(&r, &a, sizeof(a));1227
- return r;1228
-#endif1229
-}1230
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1231
-#define _mm_castsi128_pd(a) simde_mm_castsi128_pd(a)1232
-#endif1233
-1234
-SIMDE_FUNCTION_ATTRIBUTES1235
-simde__m128 simde_mm_castsi128_ps(simde__m128i a)1236
-{1237
-#if defined(SIMDE_X86_SSE2_NATIVE)1238
- return _mm_castsi128_ps(a);1239
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1240
- return a;1241
-#else1242
- simde__m128 r;1243
- simde_memcpy(&r, &a, sizeof(a));1244
- return r;1245
-#endif1246
-}1247
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1248
-#define _mm_castsi128_ps(a) simde_mm_castsi128_ps(a)1249
-#endif1250
-1251
-SIMDE_FUNCTION_ATTRIBUTES1252
-simde__m128i simde_mm_cmpeq_epi8(simde__m128i a, simde__m128i b)1253
-{1254
-#if defined(SIMDE_X86_SSE2_NATIVE)1255
- return _mm_cmpeq_epi8(a, b);1256
-#else1257
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1258
- b_ = simde__m128i_to_private(b);1259
-1260
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1261
- r_.neon_i8 = vreinterpretq_s8_u8(vceqq_s8(b_.neon_i8, a_.neon_i8));1262
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1263
- r_.wasm_v128 = wasm_i8x16_eq(a_.wasm_v128, b_.wasm_v128);1264
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1265
- r_.altivec_i8 = (SIMDE_POWER_ALTIVEC_VECTOR(signed char))vec_cmpeq(1266
- a_.altivec_i8, b_.altivec_i8);1267
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1268
- r_.i8 = HEDLEY_STATIC_CAST(__typeof__(r_.i8), (a_.i8 == b_.i8));1269
-#else1270
- SIMDE_VECTORIZE1271
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1272
- r_.i8[i] = (a_.i8[i] == b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);1273
- }1274
-#endif1275
-1276
- return simde__m128i_from_private(r_);1277
-#endif1278
-}1279
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1280
-#define _mm_cmpeq_epi8(a, b) simde_mm_cmpeq_epi8(a, b)1281
-#endif1282
-1283
-SIMDE_FUNCTION_ATTRIBUTES1284
-simde__m128i simde_mm_cmpeq_epi16(simde__m128i a, simde__m128i b)1285
-{1286
-#if defined(SIMDE_X86_SSE2_NATIVE)1287
- return _mm_cmpeq_epi16(a, b);1288
-#else1289
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1290
- b_ = simde__m128i_to_private(b);1291
-1292
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1293
- r_.neon_i16 =1294
- vreinterpretq_s16_u16(vceqq_s16(b_.neon_i16, a_.neon_i16));1295
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1296
- r_.wasm_v128 = wasm_i16x8_eq(a_.wasm_v128, b_.wasm_v128);1297
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1298
- r_.altivec_i16 = (SIMDE_POWER_ALTIVEC_VECTOR(signed short))vec_cmpeq(1299
- a_.altivec_i16, b_.altivec_i16);1300
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1301
- r_.i16 = (a_.i16 == b_.i16);1302
-#else1303
- SIMDE_VECTORIZE1304
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1305
- r_.i16[i] = (a_.i16[i] == b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);1306
- }1307
-#endif1308
-1309
- return simde__m128i_from_private(r_);1310
-#endif1311
-}1312
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1313
-#define _mm_cmpeq_epi16(a, b) simde_mm_cmpeq_epi16(a, b)1314
-#endif1315
-1316
-SIMDE_FUNCTION_ATTRIBUTES1317
-simde__m128i simde_mm_cmpeq_epi32(simde__m128i a, simde__m128i b)1318
-{1319
-#if defined(SIMDE_X86_SSE2_NATIVE)1320
- return _mm_cmpeq_epi32(a, b);1321
-#else1322
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1323
- b_ = simde__m128i_to_private(b);1324
-1325
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1326
- r_.neon_i32 =1327
- vreinterpretq_s32_u32(vceqq_s32(b_.neon_i32, a_.neon_i32));1328
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1329
- r_.wasm_v128 = wasm_i32x4_eq(a_.wasm_v128, b_.wasm_v128);1330
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1331
- r_.altivec_i32 = (SIMDE_POWER_ALTIVEC_VECTOR(signed int))vec_cmpeq(1332
- a_.altivec_i32, b_.altivec_i32);1333
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1334
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), a_.i32 == b_.i32);1335
-#else1336
- SIMDE_VECTORIZE1337
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1338
- r_.i32[i] = (a_.i32[i] == b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);1339
- }1340
-#endif1341
-1342
- return simde__m128i_from_private(r_);1343
-#endif1344
-}1345
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1346
-#define _mm_cmpeq_epi32(a, b) simde_mm_cmpeq_epi32(a, b)1347
-#endif1348
-1349
-SIMDE_FUNCTION_ATTRIBUTES1350
-simde__m128d simde_mm_cmpeq_pd(simde__m128d a, simde__m128d b)1351
-{1352
-#if defined(SIMDE_X86_SSE2_NATIVE)1353
- return _mm_cmpeq_pd(a, b);1354
-#else1355
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1356
- b_ = simde__m128d_to_private(b);1357
-1358
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1359
- r_.neon_i64 = vreinterpretq_s64_u64(1360
- vceqq_s64(vreinterpretq_s64_f64(b_.neon_f64),1361
- vreinterpretq_s64_f64(a_.neon_f64)));1362
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1363
- r_.wasm_v128 = wasm_f64x2_eq(a_.wasm_v128, b_.wasm_v128);1364
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1365
- r_.altivec_f64 = (SIMDE_POWER_ALTIVEC_VECTOR(double))vec_cmpeq(1366
- a_.altivec_f64, b_.altivec_f64);1367
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1368
- r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 == b_.f64));1369
-#else1370
- SIMDE_VECTORIZE1371
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1372
- r_.u64[i] = (a_.f64[i] == b_.f64[i]) ? ~UINT64_C(0)1373
- : UINT64_C(0);1374
- }1375
-#endif1376
-1377
- return simde__m128d_from_private(r_);1378
-#endif1379
-}1380
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1381
-#define _mm_cmpeq_pd(a, b) simde_mm_cmpeq_pd(a, b)1382
-#endif1383
-1384
-SIMDE_FUNCTION_ATTRIBUTES1385
-simde__m128d simde_mm_cmpeq_sd(simde__m128d a, simde__m128d b)1386
-{1387
-#if defined(SIMDE_X86_SSE2_NATIVE)1388
- return _mm_cmpeq_sd(a, b);1389
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1390
- return simde_mm_move_sd(a, simde_mm_cmpeq_pd(a, b));1391
-#else1392
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1393
- b_ = simde__m128d_to_private(b);1394
-1395
- r_.u64[0] = (a_.u64[0] == b_.u64[0]) ? ~UINT64_C(0) : 0;1396
- r_.u64[1] = a_.u64[1];1397
-1398
- return simde__m128d_from_private(r_);1399
-#endif1400
-}1401
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1402
-#define _mm_cmpeq_sd(a, b) simde_mm_cmpeq_sd(a, b)1403
-#endif1404
-1405
-SIMDE_FUNCTION_ATTRIBUTES1406
-simde__m128d simde_mm_cmpneq_pd(simde__m128d a, simde__m128d b)1407
-{1408
-#if defined(SIMDE_X86_SSE2_NATIVE)1409
- return _mm_cmpneq_pd(a, b);1410
-#else1411
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1412
- b_ = simde__m128d_to_private(b);1413
-1414
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1415
- r_.neon_f32 = vreinterpretq_f32_u16(1416
- vmvnq_u16(vceqq_s16(b_.neon_i16, a_.neon_i16)));1417
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1418
- r_.wasm_v128 = wasm_f64x2_ne(a_.wasm_v128, b_.wasm_v128);1419
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1420
- r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 != b_.f64));1421
-#else1422
- SIMDE_VECTORIZE1423
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1424
- r_.u64[i] = (a_.f64[i] != b_.f64[i]) ? ~UINT64_C(0)1425
- : UINT64_C(0);1426
- }1427
-#endif1428
-1429
- return simde__m128d_from_private(r_);1430
-#endif1431
-}1432
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1433
-#define _mm_cmpneq_pd(a, b) simde_mm_cmpneq_pd(a, b)1434
-#endif1435
-1436
-SIMDE_FUNCTION_ATTRIBUTES1437
-simde__m128d simde_mm_cmpneq_sd(simde__m128d a, simde__m128d b)1438
-{1439
-#if defined(SIMDE_X86_SSE2_NATIVE)1440
- return _mm_cmpneq_sd(a, b);1441
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1442
- return simde_mm_move_sd(a, simde_mm_cmpneq_pd(a, b));1443
-#else1444
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1445
- b_ = simde__m128d_to_private(b);1446
-1447
- r_.u64[0] = (a_.f64[0] != b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);1448
- r_.u64[1] = a_.u64[1];1449
-1450
- return simde__m128d_from_private(r_);1451
-#endif1452
-}1453
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1454
-#define _mm_cmpneq_sd(a, b) simde_mm_cmpneq_sd(a, b)1455
-#endif1456
-1457
-SIMDE_FUNCTION_ATTRIBUTES1458
-simde__m128i simde_mm_cmplt_epi8(simde__m128i a, simde__m128i b)1459
-{1460
-#if defined(SIMDE_X86_SSE2_NATIVE)1461
- return _mm_cmplt_epi8(a, b);1462
-#else1463
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1464
- b_ = simde__m128i_to_private(b);1465
-1466
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1467
- r_.neon_i8 = vreinterpretq_s8_u8(vcltq_s8(a_.neon_i8, b_.neon_i8));1468
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1469
- r_.altivec_i8 = HEDLEY_REINTERPRET_CAST(1470
- SIMDE_POWER_ALTIVEC_VECTOR(signed char),1471
- vec_cmplt(a_.altivec_i8, b_.altivec_i8));1472
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1473
- r_.wasm_v128 = wasm_i8x16_lt(a_.wasm_v128, b_.wasm_v128);1474
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1475
- r_.i8 = HEDLEY_STATIC_CAST(__typeof__(r_.i8), (a_.i8 < b_.i8));1476
-#else1477
- SIMDE_VECTORIZE1478
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1479
- r_.i8[i] = (a_.i8[i] < b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);1480
- }1481
-#endif1482
-1483
- return simde__m128i_from_private(r_);1484
-#endif1485
-}1486
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1487
-#define _mm_cmplt_epi8(a, b) simde_mm_cmplt_epi8(a, b)1488
-#endif1489
-1490
-SIMDE_FUNCTION_ATTRIBUTES1491
-simde__m128i simde_mm_cmplt_epi16(simde__m128i a, simde__m128i b)1492
-{1493
-#if defined(SIMDE_X86_SSE2_NATIVE)1494
- return _mm_cmplt_epi16(a, b);1495
-#else1496
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1497
- b_ = simde__m128i_to_private(b);1498
-1499
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1500
- r_.neon_i16 =1501
- vreinterpretq_s16_u16(vcltq_s16(a_.neon_i16, b_.neon_i16));1502
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1503
- r_.altivec_i16 = HEDLEY_REINTERPRET_CAST(1504
- SIMDE_POWER_ALTIVEC_VECTOR(signed short),1505
- vec_cmplt(a_.altivec_i16, b_.altivec_i16));1506
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1507
- r_.wasm_v128 = wasm_i16x8_lt(a_.wasm_v128, b_.wasm_v128);1508
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1509
- r_.i16 = HEDLEY_STATIC_CAST(__typeof__(r_.i16), (a_.i16 < b_.i16));1510
-#else1511
- SIMDE_VECTORIZE1512
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1513
- r_.i16[i] = (a_.i16[i] < b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);1514
- }1515
-#endif1516
-1517
- return simde__m128i_from_private(r_);1518
-#endif1519
-}1520
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1521
-#define _mm_cmplt_epi16(a, b) simde_mm_cmplt_epi16(a, b)1522
-#endif1523
-1524
-SIMDE_FUNCTION_ATTRIBUTES1525
-simde__m128i simde_mm_cmplt_epi32(simde__m128i a, simde__m128i b)1526
-{1527
-#if defined(SIMDE_X86_SSE2_NATIVE)1528
- return _mm_cmplt_epi32(a, b);1529
-#else1530
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1531
- b_ = simde__m128i_to_private(b);1532
-1533
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1534
- r_.neon_i32 =1535
- vreinterpretq_s32_u32(vcltq_s32(a_.neon_i32, b_.neon_i32));1536
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1537
- r_.altivec_i32 = HEDLEY_REINTERPRET_CAST(1538
- SIMDE_POWER_ALTIVEC_VECTOR(signed int),1539
- vec_cmplt(a_.altivec_i32, b_.altivec_i32));1540
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1541
- r_.wasm_v128 = wasm_i32x4_lt(a_.wasm_v128, b_.wasm_v128);1542
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1543
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.i32 < b_.i32));1544
-#else1545
- SIMDE_VECTORIZE1546
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1547
- r_.i32[i] = (a_.i32[i] < b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);1548
- }1549
-#endif1550
-1551
- return simde__m128i_from_private(r_);1552
-#endif1553
-}1554
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1555
-#define _mm_cmplt_epi32(a, b) simde_mm_cmplt_epi32(a, b)1556
-#endif1557
-1558
-SIMDE_FUNCTION_ATTRIBUTES1559
-simde__m128d simde_mm_cmplt_pd(simde__m128d a, simde__m128d b)1560
-{1561
-#if defined(SIMDE_X86_SSE2_NATIVE)1562
- return _mm_cmplt_pd(a, b);1563
-#else1564
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1565
- b_ = simde__m128d_to_private(b);1566
-1567
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1568
- r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 < b_.f64));1569
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1570
- r_.wasm_v128 = wasm_f64x2_lt(a_.wasm_v128, b_.wasm_v128);1571
-#else1572
- SIMDE_VECTORIZE1573
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1574
- r_.u64[i] = (a_.f64[i] < b_.f64[i]) ? ~UINT64_C(0)1575
- : UINT64_C(0);1576
- }1577
-#endif1578
-1579
- return simde__m128d_from_private(r_);1580
-#endif1581
-}1582
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1583
-#define _mm_cmplt_pd(a, b) simde_mm_cmplt_pd(a, b)1584
-#endif1585
-1586
-SIMDE_FUNCTION_ATTRIBUTES1587
-simde__m128d simde_mm_cmplt_sd(simde__m128d a, simde__m128d b)1588
-{1589
-#if defined(SIMDE_X86_SSE2_NATIVE)1590
- return _mm_cmplt_sd(a, b);1591
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1592
- return simde_mm_move_sd(a, simde_mm_cmplt_pd(a, b));1593
-#else1594
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1595
- b_ = simde__m128d_to_private(b);1596
-1597
- r_.u64[0] = (a_.f64[0] < b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);1598
- r_.u64[1] = a_.u64[1];1599
-1600
- return simde__m128d_from_private(r_);1601
-#endif1602
-}1603
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1604
-#define _mm_cmplt_sd(a, b) simde_mm_cmplt_sd(a, b)1605
-#endif1606
-1607
-SIMDE_FUNCTION_ATTRIBUTES1608
-simde__m128d simde_mm_cmple_pd(simde__m128d a, simde__m128d b)1609
-{1610
-#if defined(SIMDE_X86_SSE2_NATIVE)1611
- return _mm_cmple_pd(a, b);1612
-#else1613
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1614
- b_ = simde__m128d_to_private(b);1615
-1616
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1617
- r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 <= b_.f64));1618
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1619
- r_.wasm_v128 = wasm_f64x2_le(a_.wasm_v128, b_.wasm_v128);1620
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1621
- r_.altivec_f64 = (SIMDE_POWER_ALTIVEC_VECTOR(double))vec_cmple(1622
- a_.altivec_f64, b_.altivec_f64);1623
-#else1624
- SIMDE_VECTORIZE1625
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1626
- r_.u64[i] = (a_.f64[i] <= b_.f64[i]) ? ~UINT64_C(0)1627
- : UINT64_C(0);1628
- }1629
-#endif1630
-1631
- return simde__m128d_from_private(r_);1632
-#endif1633
-}1634
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1635
-#define _mm_cmple_pd(a, b) simde_mm_cmple_pd(a, b)1636
-#endif1637
-1638
-SIMDE_FUNCTION_ATTRIBUTES1639
-simde__m128d simde_mm_cmple_sd(simde__m128d a, simde__m128d b)1640
-{1641
-#if defined(SIMDE_X86_SSE2_NATIVE)1642
- return _mm_cmple_sd(a, b);1643
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1644
- return simde_mm_move_sd(a, simde_mm_cmple_pd(a, b));1645
-#else1646
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1647
- b_ = simde__m128d_to_private(b);1648
-1649
- r_.u64[0] = (a_.f64[0] <= b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);1650
- r_.u64[1] = a_.u64[1];1651
-1652
- return simde__m128d_from_private(r_);1653
-#endif1654
-}1655
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1656
-#define _mm_cmple_sd(a, b) simde_mm_cmple_sd(a, b)1657
-#endif1658
-1659
-SIMDE_FUNCTION_ATTRIBUTES1660
-simde__m128i simde_mm_cmpgt_epi8(simde__m128i a, simde__m128i b)1661
-{1662
-#if defined(SIMDE_X86_SSE2_NATIVE)1663
- return _mm_cmpgt_epi8(a, b);1664
-#else1665
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1666
- b_ = simde__m128i_to_private(b);1667
-1668
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1669
- r_.neon_i8 = vreinterpretq_s8_u8(vcgtq_s8(a_.neon_i8, b_.neon_i8));1670
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1671
- r_.wasm_v128 = wasm_i8x16_gt(a_.wasm_v128, b_.wasm_v128);1672
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1673
- r_.altivec_i8 = (SIMDE_POWER_ALTIVEC_VECTOR(signed char))vec_cmpgt(1674
- a_.altivec_i8, b_.altivec_i8);1675
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1676
- r_.i8 = HEDLEY_STATIC_CAST(__typeof__(r_.i8), (a_.i8 > b_.i8));1677
-#else1678
- SIMDE_VECTORIZE1679
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1680
- r_.i8[i] = (a_.i8[i] > b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);1681
- }1682
-#endif1683
-1684
- return simde__m128i_from_private(r_);1685
-#endif1686
-}1687
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1688
-#define _mm_cmpgt_epi8(a, b) simde_mm_cmpgt_epi8(a, b)1689
-#endif1690
-1691
-SIMDE_FUNCTION_ATTRIBUTES1692
-simde__m128i simde_mm_cmpgt_epi16(simde__m128i a, simde__m128i b)1693
-{1694
-#if defined(SIMDE_X86_SSE2_NATIVE)1695
- return _mm_cmpgt_epi16(a, b);1696
-#else1697
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1698
- b_ = simde__m128i_to_private(b);1699
-1700
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1701
- r_.neon_i16 =1702
- vreinterpretq_s16_u16(vcgtq_s16(a_.neon_i16, b_.neon_i16));1703
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1704
- r_.wasm_v128 = wasm_i16x8_gt(a_.wasm_v128, b_.wasm_v128);1705
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1706
- r_.altivec_i16 = HEDLEY_REINTERPRET_CAST(1707
- SIMDE_POWER_ALTIVEC_VECTOR(signed short),1708
- vec_cmpgt(a_.altivec_i16, b_.altivec_i16));1709
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1710
- r_.i16 = HEDLEY_STATIC_CAST(__typeof__(r_.i16), (a_.i16 > b_.i16));1711
-#else1712
- SIMDE_VECTORIZE1713
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1714
- r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);1715
- }1716
-#endif1717
-1718
- return simde__m128i_from_private(r_);1719
-#endif1720
-}1721
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1722
-#define _mm_cmpgt_epi16(a, b) simde_mm_cmpgt_epi16(a, b)1723
-#endif1724
-1725
-SIMDE_FUNCTION_ATTRIBUTES1726
-simde__m128i simde_mm_cmpgt_epi32(simde__m128i a, simde__m128i b)1727
-{1728
-#if defined(SIMDE_X86_SSE2_NATIVE)1729
- return _mm_cmpgt_epi32(a, b);1730
-#else1731
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),1732
- b_ = simde__m128i_to_private(b);1733
-1734
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1735
- r_.neon_i32 =1736
- vreinterpretq_s32_u32(vcgtq_s32(a_.neon_i32, b_.neon_i32));1737
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1738
- r_.wasm_v128 = wasm_i32x4_gt(a_.wasm_v128, b_.wasm_v128);1739
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1740
- r_.altivec_i32 = (SIMDE_POWER_ALTIVEC_VECTOR(signed int))vec_cmpgt(1741
- a_.altivec_i32, b_.altivec_i32);1742
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1743
- r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.i32 > b_.i32));1744
-#else1745
- SIMDE_VECTORIZE1746
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1747
- r_.i32[i] = (a_.i32[i] > b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);1748
- }1749
-#endif1750
-1751
- return simde__m128i_from_private(r_);1752
-#endif1753
-}1754
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1755
-#define _mm_cmpgt_epi32(a, b) simde_mm_cmpgt_epi32(a, b)1756
-#endif1757
-1758
-SIMDE_FUNCTION_ATTRIBUTES1759
-simde__m128d simde_mm_cmpgt_pd(simde__m128d a, simde__m128d b)1760
-{1761
-#if defined(SIMDE_X86_SSE2_NATIVE)1762
- return _mm_cmpgt_pd(a, b);1763
-#else1764
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1765
- b_ = simde__m128d_to_private(b);1766
-1767
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1768
- r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 > b_.f64));1769
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1770
- r_.wasm_v128 = wasm_f64x2_gt(a_.wasm_v128, b_.wasm_v128);1771
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1772
- r_.altivec_f64 =1773
- HEDLEY_STATIC_CAST(SIMDE_POWER_ALTIVEC_VECTOR(double),1774
- vec_cmpgt(a_.altivec_f64, b_.altivec_f64));1775
-#else1776
- SIMDE_VECTORIZE1777
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1778
- r_.u64[i] = (a_.f64[i] > b_.f64[i]) ? ~UINT64_C(0)1779
- : UINT64_C(0);1780
- }1781
-#endif1782
-1783
- return simde__m128d_from_private(r_);1784
-#endif1785
-}1786
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1787
-#define _mm_cmpgt_pd(a, b) simde_mm_cmpgt_pd(a, b)1788
-#endif1789
-1790
-SIMDE_FUNCTION_ATTRIBUTES1791
-simde__m128d simde_mm_cmpgt_sd(simde__m128d a, simde__m128d b)1792
-{1793
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)1794
- return _mm_cmpgt_sd(a, b);1795
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1796
- return simde_mm_move_sd(a, simde_mm_cmpgt_pd(a, b));1797
-#else1798
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1799
- b_ = simde__m128d_to_private(b);1800
-1801
- r_.u64[0] = (a_.f64[0] > b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);1802
- r_.u64[1] = a_.u64[1];1803
-1804
- return simde__m128d_from_private(r_);1805
-#endif1806
-}1807
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1808
-#define _mm_cmpgt_sd(a, b) simde_mm_cmpgt_sd(a, b)1809
-#endif1810
-1811
-SIMDE_FUNCTION_ATTRIBUTES1812
-simde__m128d simde_mm_cmpge_pd(simde__m128d a, simde__m128d b)1813
-{1814
-#if defined(SIMDE_X86_SSE2_NATIVE)1815
- return _mm_cmpge_pd(a, b);1816
-#else1817
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1818
- b_ = simde__m128d_to_private(b);1819
-1820
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1821
- r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 >= b_.f64));1822
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)1823
- r_.wasm_v128 = wasm_f64x2_ge(a_.wasm_v128, b_.wasm_v128);1824
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)1825
- r_.altivec_f64 =1826
- HEDLEY_STATIC_CAST(SIMDE_POWER_ALTIVEC_VECTOR(double),1827
- vec_cmpge(a_.altivec_f64, b_.altivec_f64));1828
-#else1829
- SIMDE_VECTORIZE1830
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1831
- r_.u64[i] = (a_.f64[i] >= b_.f64[i]) ? ~UINT64_C(0)1832
- : UINT64_C(0);1833
- }1834
-#endif1835
-1836
- return simde__m128d_from_private(r_);1837
-#endif1838
-}1839
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1840
-#define _mm_cmpge_pd(a, b) simde_mm_cmpge_pd(a, b)1841
-#endif1842
-1843
-SIMDE_FUNCTION_ATTRIBUTES1844
-simde__m128d simde_mm_cmpge_sd(simde__m128d a, simde__m128d b)1845
-{1846
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)1847
- return _mm_cmpge_sd(a, b);1848
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1849
- return simde_mm_move_sd(a, simde_mm_cmpge_pd(a, b));1850
-#else1851
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1852
- b_ = simde__m128d_to_private(b);1853
-1854
- r_.u64[0] = (a_.f64[0] >= b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);1855
- r_.u64[1] = a_.u64[1];1856
-1857
- return simde__m128d_from_private(r_);1858
-#endif1859
-}1860
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1861
-#define _mm_cmpge_sd(a, b) simde_mm_cmpge_sd(a, b)1862
-#endif1863
-1864
-SIMDE_FUNCTION_ATTRIBUTES1865
-simde__m128d simde_mm_cmpnge_pd(simde__m128d a, simde__m128d b)1866
-{1867
-#if defined(SIMDE_X86_SSE2_NATIVE)1868
- return _mm_cmpnge_pd(a, b);1869
-#else1870
- return simde_mm_cmplt_pd(a, b);1871
-#endif1872
-}1873
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1874
-#define _mm_cmpnge_pd(a, b) simde_mm_cmpnge_pd(a, b)1875
-#endif1876
-1877
-SIMDE_FUNCTION_ATTRIBUTES1878
-simde__m128d simde_mm_cmpnge_sd(simde__m128d a, simde__m128d b)1879
-{1880
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)1881
- return _mm_cmpnge_sd(a, b);1882
-#else1883
- return simde_mm_cmplt_sd(a, b);1884
-#endif1885
-}1886
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1887
-#define _mm_cmpnge_sd(a, b) simde_mm_cmpnge_sd(a, b)1888
-#endif1889
-1890
-SIMDE_FUNCTION_ATTRIBUTES1891
-simde__m128d simde_mm_cmpnlt_pd(simde__m128d a, simde__m128d b)1892
-{1893
-#if defined(SIMDE_X86_SSE2_NATIVE)1894
- return _mm_cmpnlt_pd(a, b);1895
-#else1896
- return simde_mm_cmpge_pd(a, b);1897
-#endif1898
-}1899
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1900
-#define _mm_cmpnlt_pd(a, b) simde_mm_cmpnlt_pd(a, b)1901
-#endif1902
-1903
-SIMDE_FUNCTION_ATTRIBUTES1904
-simde__m128d simde_mm_cmpnlt_sd(simde__m128d a, simde__m128d b)1905
-{1906
-#if defined(SIMDE_X86_SSE2_NATIVE)1907
- return _mm_cmpnlt_sd(a, b);1908
-#else1909
- return simde_mm_cmpge_sd(a, b);1910
-#endif1911
-}1912
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1913
-#define _mm_cmpnlt_sd(a, b) simde_mm_cmpnlt_sd(a, b)1914
-#endif1915
-1916
-SIMDE_FUNCTION_ATTRIBUTES1917
-simde__m128d simde_mm_cmpnle_pd(simde__m128d a, simde__m128d b)1918
-{1919
-#if defined(SIMDE_X86_SSE2_NATIVE)1920
- return _mm_cmpnle_pd(a, b);1921
-#else1922
- return simde_mm_cmpgt_pd(a, b);1923
-#endif1924
-}1925
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1926
-#define _mm_cmpnle_pd(a, b) simde_mm_cmpnle_pd(a, b)1927
-#endif1928
-1929
-SIMDE_FUNCTION_ATTRIBUTES1930
-simde__m128d simde_mm_cmpnle_sd(simde__m128d a, simde__m128d b)1931
-{1932
-#if defined(SIMDE_X86_SSE2_NATIVE)1933
- return _mm_cmpnle_sd(a, b);1934
-#else1935
- return simde_mm_cmpgt_sd(a, b);1936
-#endif1937
-}1938
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1939
-#define _mm_cmpnle_sd(a, b) simde_mm_cmpnle_sd(a, b)1940
-#endif1941
-1942
-SIMDE_FUNCTION_ATTRIBUTES1943
-simde__m128d simde_mm_cmpord_pd(simde__m128d a, simde__m128d b)1944
-{1945
-#if defined(SIMDE_X86_SSE2_NATIVE)1946
- return _mm_cmpord_pd(a, b);1947
-#else1948
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1949
- b_ = simde__m128d_to_private(b);1950
-1951
-#if defined(simde_math_isnan)1952
- SIMDE_VECTORIZE1953
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1954
- r_.u64[i] = (!simde_math_isnan(a_.f64[i]) &&1955
- !simde_math_isnan(b_.f64[i]))1956
- ? ~UINT64_C(0)1957
- : UINT64_C(0);1958
- }1959
-#else1960
- HEDLEY_UNREACHABLE();1961
-#endif1962
-1963
- return simde__m128d_from_private(r_);1964
-#endif1965
-}1966
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1967
-#define _mm_cmpord_pd(a, b) simde_mm_cmpord_pd(a, b)1968
-#endif1969
-1970
-SIMDE_FUNCTION_ATTRIBUTES1971
-simde_float64 simde_mm_cvtsd_f64(simde__m128d a)1972
-{1973
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)1974
- return _mm_cvtsd_f64(a);1975
-#else1976
- simde__m128d_private a_ = simde__m128d_to_private(a);1977
- return a_.f64[0];1978
-#endif1979
-}1980
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1981
-#define _mm_cvtsd_f64(a) simde_mm_cvtsd_f64(a)1982
-#endif1983
-1984
-SIMDE_FUNCTION_ATTRIBUTES1985
-simde__m128d simde_mm_cmpord_sd(simde__m128d a, simde__m128d b)1986
-{1987
-#if defined(SIMDE_X86_SSE2_NATIVE)1988
- return _mm_cmpord_sd(a, b);1989
-#elif defined(SIMDE_ASSUME_VECTORIZATION)1990
- return simde_mm_move_sd(a, simde_mm_cmpord_pd(a, b));1991
-#else1992
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),1993
- b_ = simde__m128d_to_private(b);1994
-1995
-#if defined(simde_math_isnan)1996
- r_.u64[0] =1997
- (!simde_math_isnan(a_.f64[0]) && !simde_math_isnan(b_.f64[0]))1998
- ? ~UINT64_C(0)1999
- : UINT64_C(0);2000
- r_.u64[1] = a_.u64[1];2001
-#else2002
- HEDLEY_UNREACHABLE();2003
-#endif2004
-2005
- return simde__m128d_from_private(r_);2006
-#endif2007
-}2008
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2009
-#define _mm_cmpord_sd(a, b) simde_mm_cmpord_sd(a, b)2010
-#endif2011
-2012
-SIMDE_FUNCTION_ATTRIBUTES2013
-simde__m128d simde_mm_cmpunord_pd(simde__m128d a, simde__m128d b)2014
-{2015
-#if defined(SIMDE_X86_SSE2_NATIVE)2016
- return _mm_cmpunord_pd(a, b);2017
-#else2018
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),2019
- b_ = simde__m128d_to_private(b);2020
-2021
-#if defined(simde_math_isnan)2022
- SIMDE_VECTORIZE2023
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2024
- r_.u64[i] = (simde_math_isnan(a_.f64[i]) ||2025
- simde_math_isnan(b_.f64[i]))2026
- ? ~UINT64_C(0)2027
- : UINT64_C(0);2028
- }2029
-#else2030
- HEDLEY_UNREACHABLE();2031
-#endif2032
-2033
- return simde__m128d_from_private(r_);2034
-#endif2035
-}2036
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2037
-#define _mm_cmpunord_pd(a, b) simde_mm_cmpunord_pd(a, b)2038
-#endif2039
-2040
-SIMDE_FUNCTION_ATTRIBUTES2041
-simde__m128d simde_mm_cmpunord_sd(simde__m128d a, simde__m128d b)2042
-{2043
-#if defined(SIMDE_X86_SSE2_NATIVE)2044
- return _mm_cmpunord_sd(a, b);2045
-#elif defined(SIMDE_ASSUME_VECTORIZATION)2046
- return simde_mm_move_sd(a, simde_mm_cmpunord_pd(a, b));2047
-#else2048
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),2049
- b_ = simde__m128d_to_private(b);2050
-2051
-#if defined(simde_math_isnan)2052
- r_.u64[0] = (simde_math_isnan(a_.f64[0]) || simde_math_isnan(b_.f64[0]))2053
- ? ~UINT64_C(0)2054
- : UINT64_C(0);2055
- r_.u64[1] = a_.u64[1];2056
-2057
-#else2058
- HEDLEY_UNREACHABLE();2059
-#endif2060
-2061
- return simde__m128d_from_private(r_);2062
-#endif2063
-}2064
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2065
-#define _mm_cmpunord_sd(a, b) simde_mm_cmpunord_sd(a, b)2066
-#endif2067
-2068
-SIMDE_FUNCTION_ATTRIBUTES2069
-simde__m128d simde_mm_cvtepi32_pd(simde__m128i a)2070
-{2071
-#if defined(SIMDE_X86_SSE2_NATIVE)2072
- return _mm_cvtepi32_pd(a);2073
-#else2074
- simde__m128d_private r_;2075
- simde__m128i_private a_ = simde__m128i_to_private(a);2076
-2077
-#if defined(SIMDE_CONVERT_VECTOR_)2078
- SIMDE_CONVERT_VECTOR_(r_.f64, a_.m64_private[0].i32);2079
-#else2080
- SIMDE_VECTORIZE2081
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2082
- r_.f64[i] = (simde_float64)a_.i32[i];2083
- }2084
-#endif2085
-2086
- return simde__m128d_from_private(r_);2087
-#endif2088
-}2089
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2090
-#define _mm_cvtepi32_pd(a) simde_mm_cvtepi32_pd(a)2091
-#endif2092
-2093
-SIMDE_FUNCTION_ATTRIBUTES2094
-simde__m128 simde_mm_cvtepi32_ps(simde__m128i a)2095
-{2096
-#if defined(SIMDE_X86_SSE2_NATIVE)2097
- return _mm_cvtepi32_ps(a);2098
-#else2099
- simde__m128_private r_;2100
- simde__m128i_private a_ = simde__m128i_to_private(a);2101
-2102
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2103
- r_.neon_f32 = vcvtq_f32_s32(a_.neon_i32);2104
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2105
- r_.altivec_f32 = vec_ctf(a_.altivec_i32, 0);2106
-#elif defined(SIMDE_CONVERT_VECTOR_)2107
- SIMDE_CONVERT_VECTOR_(r_.f32, a_.i32);2108
-#else2109
- SIMDE_VECTORIZE2110
- for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2111
- r_.f32[i] = (simde_float32)a_.i32[i];2112
- }2113
-#endif2114
-2115
- return simde__m128_from_private(r_);2116
-#endif2117
-}2118
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2119
-#define _mm_cvtepi32_ps(a) simde_mm_cvtepi32_ps(a)2120
-#endif2121
-2122
-SIMDE_FUNCTION_ATTRIBUTES2123
-simde__m128i simde_mm_cvtpd_epi32(simde__m128d a)2124
-{2125
-#if defined(SIMDE_X86_SSE2_NATIVE)2126
- return _mm_cvtpd_epi32(a);2127
-#else2128
- simde__m128i_private r_;2129
- simde__m128d_private a_ = simde__m128d_to_private(a);2130
-2131
-#if defined(SIMDE_CONVERT_VECTOR_)2132
- SIMDE_CONVERT_VECTOR_(r_.m64_private[0].i32, a_.f64);2133
- r_.m64_private[1] = simde__m64_to_private(simde_mm_setzero_si64());2134
-#else2135
- SIMDE_VECTORIZE2136
- for (size_t i = 0; i < (sizeof(a_.f64) / sizeof(a_.f64[0])); i++) {2137
- r_.i32[i] = HEDLEY_STATIC_CAST(int32_t, a_.f64[i]);2138
- }2139
- simde_memset(&(r_.m64_private[1]), 0, sizeof(r_.m64_private[1]));2140
-#endif2141
-2142
- return simde__m128i_from_private(r_);2143
-#endif2144
-}2145
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2146
-#define _mm_cvtpd_epi32(a) simde_mm_cvtpd_epi32(a)2147
-#endif2148
-2149
-SIMDE_FUNCTION_ATTRIBUTES2150
-simde__m64 simde_mm_cvtpd_pi32(simde__m128d a)2151
-{2152
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2153
- return _mm_cvtpd_pi32(a);2154
-#else2155
- simde__m64_private r_;2156
- simde__m128d_private a_ = simde__m128d_to_private(a);2157
-2158
-#if defined(SIMDE_CONVERT_VECTOR_)2159
- SIMDE_CONVERT_VECTOR_(r_.i32, a_.f64);2160
-#else2161
- SIMDE_VECTORIZE2162
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {2163
- r_.i32[i] = HEDLEY_STATIC_CAST(int32_t, a_.f64[i]);2164
- }2165
-#endif2166
-2167
- return simde__m64_from_private(r_);2168
-#endif2169
-}2170
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2171
-#define _mm_cvtpd_pi32(a) simde_mm_cvtpd_pi32(a)2172
-#endif2173
-2174
-SIMDE_FUNCTION_ATTRIBUTES2175
-simde__m128 simde_mm_cvtpd_ps(simde__m128d a)2176
-{2177
-#if defined(SIMDE_X86_SSE2_NATIVE)2178
- return _mm_cvtpd_ps(a);2179
-#else2180
- simde__m128_private r_;2181
- simde__m128d_private a_ = simde__m128d_to_private(a);2182
-2183
-#if defined(SIMDE_CONVERT_VECTOR_)2184
- SIMDE_CONVERT_VECTOR_(r_.m64_private[0].f32, a_.f64);2185
- r_.m64_private[1] = simde__m64_to_private(simde_mm_setzero_si64());2186
-#else2187
- SIMDE_VECTORIZE2188
- for (size_t i = 0; i < (sizeof(a_.f64) / sizeof(a_.f64[0])); i++) {2189
- r_.f32[i] = (simde_float32)a_.f64[i];2190
- }2191
- simde_memset(&(r_.m64_private[1]), 0, sizeof(r_.m64_private[1]));2192
-#endif2193
-2194
- return simde__m128_from_private(r_);2195
-#endif2196
-}2197
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2198
-#define _mm_cvtpd_ps(a) simde_mm_cvtpd_ps(a)2199
-#endif2200
-2201
-SIMDE_FUNCTION_ATTRIBUTES2202
-simde__m128d simde_mm_cvtpi32_pd(simde__m64 a)2203
-{2204
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2205
- return _mm_cvtpi32_pd(a);2206
-#else2207
- simde__m128d_private r_;2208
- simde__m64_private a_ = simde__m64_to_private(a);2209
-2210
-#if defined(SIMDE_CONVERT_VECTOR_)2211
- SIMDE_CONVERT_VECTOR_(r_.f64, a_.i32);2212
-#else2213
- SIMDE_VECTORIZE2214
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2215
- r_.f64[i] = (simde_float64)a_.i32[i];2216
- }2217
-#endif2218
-2219
- return simde__m128d_from_private(r_);2220
-#endif2221
-}2222
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2223
-#define _mm_cvtpi32_pd(a) simde_mm_cvtpi32_pd(a)2224
-#endif2225
-2226
-SIMDE_FUNCTION_ATTRIBUTES2227
-simde__m128i simde_mm_cvtps_epi32(simde__m128 a)2228
-{2229
-#if defined(SIMDE_X86_SSE2_NATIVE)2230
- return _mm_cvtps_epi32(a);2231
-#else2232
- simde__m128i_private r_;2233
- simde__m128_private a_ = simde__m128_to_private(a);2234
-2235
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2236
-/* The default rounding mode on SSE is 'round to even', which ArmV72237
- does not support! It is supported on ARMv8 however. */2238
-#if defined(SIMDE_ARCH_AARCH64)2239
- r_.neon_i32 = vcvtnq_s32_f32(a_.neon_f32);2240
-#else2241
- uint32x4_t signmask = vdupq_n_u32(0x80000000);2242
- float32x4_t half = vbslq_f32(signmask, a_.neon_f32,2243
- vdupq_n_f32(0.5f)); /* +/- 0.5 */2244
- int32x4_t r_normal = vcvtq_s32_f32(2245
- vaddq_f32(a_.neon_f32, half)); /* round to integer: [a + 0.5]*/2246
- int32x4_t r_trunc =2247
- vcvtq_s32_f32(a_.neon_f32); /* truncate to integer: [a] */2248
- int32x4_t plusone = vshrq_n_s32(vnegq_s32(r_trunc), 31); /* 1 or 0 */2249
- int32x4_t r_even = vbicq_s32(vaddq_s32(r_trunc, plusone),2250
- vdupq_n_s32(1)); /* ([a] + {0,1}) & ~1 */2251
- float32x4_t delta = vsubq_f32(2252
- a_.neon_f32,2253
- vcvtq_f32_s32(r_trunc)); /* compute delta: delta = (a - [a]) */2254
- uint32x4_t is_delta_half =2255
- vceqq_f32(delta, half); /* delta == +/- 0.5 */2256
- r_.neon_i32 = vbslq_s32(is_delta_half, r_even, r_normal);2257
-#endif2258
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2259
- r_.altivec_i32 = vec_cts(a_.altivec_f32, 0);2260
-#elif defined(SIMDE_CONVERT_VECTOR_)2261
- SIMDE_CONVERT_VECTOR_(r_.i32, a_.f32);2262
-#else2263
- SIMDE_VECTORIZE2264
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {2265
- r_.i32[i] = HEDLEY_STATIC_CAST(int32_t, a_.f32[i]);2266
- }2267
-#endif2268
-2269
- return simde__m128i_from_private(r_);2270
-#endif2271
-}2272
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2273
-#define _mm_cvtps_epi32(a) simde_mm_cvtps_epi32(a)2274
-#endif2275
-2276
-SIMDE_FUNCTION_ATTRIBUTES2277
-simde__m128d simde_mm_cvtps_pd(simde__m128 a)2278
-{2279
-#if defined(SIMDE_X86_SSE2_NATIVE)2280
- return _mm_cvtps_pd(a);2281
-#else2282
- simde__m128d_private r_;2283
- simde__m128_private a_ = simde__m128_to_private(a);2284
-2285
-#if defined(SIMDE_CONVERT_VECTOR_)2286
- SIMDE_CONVERT_VECTOR_(r_.f64, a_.m64_private[0].f32);2287
-#else2288
- SIMDE_VECTORIZE2289
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2290
- r_.f64[i] = a_.f32[i];2291
- }2292
-#endif2293
-2294
- return simde__m128d_from_private(r_);2295
-#endif2296
-}2297
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2298
-#define _mm_cvtps_pd(a) simde_mm_cvtps_pd(a)2299
-#endif2300
-2301
-SIMDE_FUNCTION_ATTRIBUTES2302
-int32_t simde_mm_cvtsd_si32(simde__m128d a)2303
-{2304
-#if defined(SIMDE_X86_SSE2_NATIVE)2305
- return _mm_cvtsd_si32(a);2306
-#else2307
- simde__m128d_private a_ = simde__m128d_to_private(a);2308
- return SIMDE_CONVERT_FTOI(int32_t, a_.f64[0]);2309
-#endif2310
-}2311
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2312
-#define _mm_cvtsd_si32(a) simde_mm_cvtsd_si32(a)2313
-#endif2314
-2315
-SIMDE_FUNCTION_ATTRIBUTES2316
-int64_t simde_mm_cvtsd_si64(simde__m128d a)2317
-{2318
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2319
-#if defined(__PGI)2320
- return _mm_cvtsd_si64x(a);2321
-#else2322
- return _mm_cvtsd_si64(a);2323
-#endif2324
-#else2325
- simde__m128d_private a_ = simde__m128d_to_private(a);2326
- return SIMDE_CONVERT_FTOI(int64_t, a_.f64[0]);2327
-#endif2328
-}2329
-#define simde_mm_cvtsd_si64x(a) simde_mm_cvtsd_si64(a)2330
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2331
-#define _mm_cvtsd_si64(a) simde_mm_cvtsd_si64(a)2332
-#define _mm_cvtsd_si64x(a) simde_mm_cvtsd_si64x(a)2333
-#endif2334
-2335
-SIMDE_FUNCTION_ATTRIBUTES2336
-simde__m128 simde_mm_cvtsd_ss(simde__m128 a, simde__m128d b)2337
-{2338
-#if defined(SIMDE_X86_SSE2_NATIVE)2339
- return _mm_cvtsd_ss(a, b);2340
-#else2341
- simde__m128_private r_, a_ = simde__m128_to_private(a);2342
- simde__m128d_private b_ = simde__m128d_to_private(b);2343
-2344
- r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, b_.f64[0]);2345
-2346
- SIMDE_VECTORIZE2347
- for (size_t i = 1; i < (sizeof(r_) / sizeof(r_.i32[0])); i++) {2348
- r_.i32[i] = a_.i32[i];2349
- }2350
-2351
- return simde__m128_from_private(r_);2352
-#endif2353
-}2354
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2355
-#define _mm_cvtsd_ss(a, b) simde_mm_cvtsd_ss(a, b)2356
-#endif2357
-2358
-SIMDE_FUNCTION_ATTRIBUTES2359
-int32_t simde_mm_cvtsi128_si32(simde__m128i a)2360
-{2361
-#if defined(SIMDE_X86_SSE2_NATIVE)2362
- return _mm_cvtsi128_si32(a);2363
-#else2364
- simde__m128i_private a_ = simde__m128i_to_private(a);2365
-2366
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2367
- return vgetq_lane_s32(a_.neon_i32, 0);2368
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2369
-#if defined(SIMDE_BUG_GCC_95227)2370
- (void)a_;2371
-#endif2372
- return vec_extract(a_.altivec_i32, 0);2373
-#else2374
- return a_.i32[0];2375
-#endif2376
-#endif2377
-}2378
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2379
-#define _mm_cvtsi128_si32(a) simde_mm_cvtsi128_si32(a)2380
-#endif2381
-2382
-SIMDE_FUNCTION_ATTRIBUTES2383
-int64_t simde_mm_cvtsi128_si64(simde__m128i a)2384
-{2385
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2386
-#if defined(__PGI)2387
- return _mm_cvtsi128_si64x(a);2388
-#else2389
- return _mm_cvtsi128_si64(a);2390
-#endif2391
-#else2392
- simde__m128i_private a_ = simde__m128i_to_private(a);2393
-#if defined(SIMDE_POWER_ALTIVEC_P5_NATIVE) && !defined(HEDLEY_IBM_VERSION)2394
- return vec_extract(a_.i64, 0);2395
-#endif2396
- return a_.i64[0];2397
-#endif2398
-}2399
-#define simde_mm_cvtsi128_si64x(a) simde_mm_cvtsi128_si64(a)2400
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2401
-#define _mm_cvtsi128_si64(a) simde_mm_cvtsi128_si64(a)2402
-#define _mm_cvtsi128_si64x(a) simde_mm_cvtsi128_si64x(a)2403
-#endif2404
-2405
-SIMDE_FUNCTION_ATTRIBUTES2406
-simde__m128d simde_mm_cvtsi32_sd(simde__m128d a, int32_t b)2407
-{2408
-2409
-#if defined(SIMDE_X86_SSE2_NATIVE)2410
- return _mm_cvtsi32_sd(a, b);2411
-#else2412
- simde__m128d_private r_;2413
- simde__m128d_private a_ = simde__m128d_to_private(a);2414
-2415
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE) && defined(SIMDE_ARCH_AMD64)2416
- r_.neon_f64 = vsetq_lane_f64((simde_float64)b, a_.neon_f64, 0);2417
-#else2418
- r_.f64[0] = HEDLEY_STATIC_CAST(simde_float64, b);2419
- r_.i64[1] = a_.i64[1];2420
-#endif2421
-2422
- return simde__m128d_from_private(r_);2423
-#endif2424
-}2425
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2426
-#define _mm_cvtsi32_sd(a, b) simde_mm_cvtsi32_sd(a, b)2427
-#endif2428
-2429
-SIMDE_FUNCTION_ATTRIBUTES2430
-simde__m128i simde_mm_cvtsi32_si128(int32_t a)2431
-{2432
-#if defined(SIMDE_X86_SSE2_NATIVE)2433
- return _mm_cvtsi32_si128(a);2434
-#else2435
- simde__m128i_private r_;2436
-2437
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2438
- r_.neon_i32 = vsetq_lane_s32(a, vdupq_n_s32(0), 0);2439
-#else2440
- r_.i32[0] = a;2441
- r_.i32[1] = 0;2442
- r_.i32[2] = 0;2443
- r_.i32[3] = 0;2444
-#endif2445
-2446
- return simde__m128i_from_private(r_);2447
-#endif2448
-}2449
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2450
-#define _mm_cvtsi32_si128(a) simde_mm_cvtsi32_si128(a)2451
-#endif2452
-2453
-SIMDE_FUNCTION_ATTRIBUTES2454
-simde__m128d simde_mm_cvtsi64_sd(simde__m128d a, int64_t b)2455
-{2456
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2457
-#if !defined(__PGI)2458
- return _mm_cvtsi64_sd(a, b);2459
-#else2460
- return _mm_cvtsi64x_sd(a, b);2461
-#endif2462
-#else2463
- simde__m128d_private r_, a_ = simde__m128d_to_private(a);2464
-2465
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2466
- r_.neon_f64 = vsetq_lane_f64((simde_float64)b, a_.neon_f64, 0);2467
-#else2468
- r_.f64[0] = HEDLEY_STATIC_CAST(simde_float64, b);2469
- r_.f64[1] = a_.f64[1];2470
-#endif2471
-2472
- return simde__m128d_from_private(r_);2473
-#endif2474
-}2475
-#define simde_mm_cvtsi64x_sd(a, b) simde_mm_cvtsi64_sd(a, b)2476
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2477
-#define _mm_cvtsi64_sd(a, b) simde_mm_cvtsi64_sd(a, b)2478
-#define _mm_cvtsi64x_sd(a, b) simde_mm_cvtsi64x_sd(a, b)2479
-#endif2480
-2481
-SIMDE_FUNCTION_ATTRIBUTES2482
-simde__m128i simde_mm_cvtsi64_si128(int64_t a)2483
-{2484
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2485
-#if !defined(__PGI)2486
- return _mm_cvtsi64_si128(a);2487
-#else2488
- return _mm_cvtsi64x_si128(a);2489
-#endif2490
-#else2491
- simde__m128i_private r_;2492
-2493
- r_.i64[0] = a;2494
- r_.i64[1] = 0;2495
-2496
- return simde__m128i_from_private(r_);2497
-#endif2498
-}2499
-#define simde_mm_cvtsi64x_si128(a) simde_mm_cvtsi64_si128(a)2500
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2501
-#define _mm_cvtsi64_si128(a) simde_mm_cvtsi64_si128(a)2502
-#define _mm_cvtsi64x_si128(a) simde_mm_cvtsi64x_si128(a)2503
-#endif2504
-2505
-SIMDE_FUNCTION_ATTRIBUTES2506
-simde__m128d simde_mm_cvtss_sd(simde__m128d a, simde__m128 b)2507
-{2508
-#if defined(SIMDE_X86_SSE2_NATIVE)2509
- return _mm_cvtss_sd(a, b);2510
-#else2511
- simde__m128d_private a_ = simde__m128d_to_private(a);2512
- simde__m128_private b_ = simde__m128_to_private(b);2513
-2514
- a_.f64[0] = HEDLEY_STATIC_CAST(simde_float64, b_.f32[0]);2515
-2516
- return simde__m128d_from_private(a_);2517
-#endif2518
-}2519
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2520
-#define _mm_cvtss_sd(a, b) simde_mm_cvtss_sd(a, b)2521
-#endif2522
-2523
-SIMDE_FUNCTION_ATTRIBUTES2524
-simde__m128i simde_mm_cvttpd_epi32(simde__m128d a)2525
-{2526
-#if defined(SIMDE_X86_SSE2_NATIVE)2527
- return _mm_cvttpd_epi32(a);2528
-#else2529
- simde__m128i_private r_;2530
- simde__m128d_private a_ = simde__m128d_to_private(a);2531
-2532
- for (size_t i = 0; i < (sizeof(a_.f64) / sizeof(a_.f64[0])); i++) {2533
- r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, a_.f64[i]);2534
- }2535
- simde_memset(&(r_.m64_private[1]), 0, sizeof(r_.m64_private[1]));2536
-2537
- return simde__m128i_from_private(r_);2538
-#endif2539
-}2540
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2541
-#define _mm_cvttpd_epi32(a) simde_mm_cvttpd_epi32(a)2542
-#endif2543
-2544
-SIMDE_FUNCTION_ATTRIBUTES2545
-simde__m64 simde_mm_cvttpd_pi32(simde__m128d a)2546
-{2547
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2548
- return _mm_cvttpd_pi32(a);2549
-#else2550
- simde__m64_private r_;2551
- simde__m128d_private a_ = simde__m128d_to_private(a);2552
-2553
-#if defined(SIMDE_CONVERT_VECTOR_)2554
- SIMDE_CONVERT_VECTOR_(r_.i32, a_.f64);2555
-#else2556
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {2557
- r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, a_.f64[i]);2558
- }2559
-#endif2560
-2561
- return simde__m64_from_private(r_);2562
-#endif2563
-}2564
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2565
-#define _mm_cvttpd_pi32(a) simde_mm_cvttpd_pi32(a)2566
-#endif2567
-2568
-SIMDE_FUNCTION_ATTRIBUTES2569
-simde__m128i simde_mm_cvttps_epi32(simde__m128 a)2570
-{2571
-#if defined(SIMDE_X86_SSE2_NATIVE)2572
- return _mm_cvttps_epi32(a);2573
-#else2574
- simde__m128i_private r_;2575
- simde__m128_private a_ = simde__m128_to_private(a);2576
-2577
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2578
- r_.neon_i32 = vcvtq_s32_f32(a_.neon_f32);2579
-#elif defined(SIMDE_CONVERT_VECTOR_)2580
- SIMDE_CONVERT_VECTOR_(r_.i32, a_.f32);2581
-#else2582
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {2583
- r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, a_.f32[i]);2584
- }2585
-#endif2586
-2587
- return simde__m128i_from_private(r_);2588
-#endif2589
-}2590
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2591
-#define _mm_cvttps_epi32(a) simde_mm_cvttps_epi32(a)2592
-#endif2593
-2594
-SIMDE_FUNCTION_ATTRIBUTES2595
-int32_t simde_mm_cvttsd_si32(simde__m128d a)2596
-{2597
-#if defined(SIMDE_X86_SSE2_NATIVE)2598
- return _mm_cvttsd_si32(a);2599
-#else2600
- simde__m128d_private a_ = simde__m128d_to_private(a);2601
- return SIMDE_CONVERT_FTOI(int32_t, a_.f64[0]);2602
-#endif2603
-}2604
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2605
-#define _mm_cvttsd_si32(a) simde_mm_cvttsd_si32(a)2606
-#endif2607
-2608
-SIMDE_FUNCTION_ATTRIBUTES2609
-int64_t simde_mm_cvttsd_si64(simde__m128d a)2610
-{2611
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2612
-#if !defined(__PGI)2613
- return _mm_cvttsd_si64(a);2614
-#else2615
- return _mm_cvttsd_si64x(a);2616
-#endif2617
-#else2618
- simde__m128d_private a_ = simde__m128d_to_private(a);2619
- return SIMDE_CONVERT_FTOI(int64_t, a_.f64[0]);2620
-#endif2621
-}2622
-#define simde_mm_cvttsd_si64x(a) simde_mm_cvttsd_si64(a)2623
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2624
-#define _mm_cvttsd_si64(a) simde_mm_cvttsd_si64(a)2625
-#define _mm_cvttsd_si64x(a) simde_mm_cvttsd_si64x(a)2626
-#endif2627
-2628
-SIMDE_FUNCTION_ATTRIBUTES2629
-simde__m128d simde_mm_div_pd(simde__m128d a, simde__m128d b)2630
-{2631
-#if defined(SIMDE_X86_SSE2_NATIVE)2632
- return _mm_div_pd(a, b);2633
-#else2634
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),2635
- b_ = simde__m128d_to_private(b);2636
-2637
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2638
- r_.f64 = a_.f64 / b_.f64;2639
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)2640
- r_.wasm_v128 = wasm_f64x2_div(a_.wasm_v128, b_.wasm_v128);2641
-#else2642
- SIMDE_VECTORIZE2643
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2644
- r_.f64[i] = a_.f64[i] / b_.f64[i];2645
- }2646
-#endif2647
-2648
- return simde__m128d_from_private(r_);2649
-#endif2650
-}2651
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2652
-#define _mm_div_pd(a, b) simde_mm_div_pd(a, b)2653
-#endif2654
-2655
-SIMDE_FUNCTION_ATTRIBUTES2656
-simde__m128d simde_mm_div_sd(simde__m128d a, simde__m128d b)2657
-{2658
-#if defined(SIMDE_X86_SSE2_NATIVE)2659
- return _mm_div_sd(a, b);2660
-#elif defined(SIMDE_ASSUME_VECTORIZATION)2661
- return simde_mm_move_sd(a, simde_mm_div_pd(a, b));2662
-#else2663
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),2664
- b_ = simde__m128d_to_private(b);2665
-2666
- r_.f64[0] = a_.f64[0] / b_.f64[0];2667
- r_.f64[1] = a_.f64[1];2668
-2669
- return simde__m128d_from_private(r_);2670
-#endif2671
-}2672
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2673
-#define _mm_div_sd(a, b) simde_mm_div_sd(a, b)2674
-#endif2675
-2676
-SIMDE_FUNCTION_ATTRIBUTES2677
-int32_t simde_mm_extract_epi16(simde__m128i a, const int imm8)2678
- SIMDE_REQUIRE_RANGE(imm8, 0, 7)2679
-{2680
- uint16_t r;2681
- simde__m128i_private a_ = simde__m128i_to_private(a);2682
-2683
-#if defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2684
-#if defined(SIMDE_BUG_GCC_95227)2685
- (void)a_;2686
- (void)imm8;2687
-#endif2688
- r = vec_extract(a_.altivec_i16, imm8);2689
-#else2690
- r = a_.u16[imm8 & 7];2691
-#endif2692
-2693
- return HEDLEY_STATIC_CAST(int32_t, r);2694
-}2695
-#if defined(SIMDE_X86_SSE2_NATIVE) && \2696
- (!defined(HEDLEY_GCC_VERSION) || HEDLEY_GCC_VERSION_CHECK(4, 6, 0))2697
-#define simde_mm_extract_epi16(a, imm8) _mm_extract_epi16(a, imm8)2698
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2699
-#define simde_mm_extract_epi16(a, imm8) \2700
- HEDLEY_STATIC_CAST(int32_t, \2701
- vgetq_lane_s16(simde__m128i_to_private(a).neon_i16, \2702
- (imm8)) & \2703
- (UINT32_C(0x0000ffff)))2704
-#endif2705
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2706
-#define _mm_extract_epi16(a, imm8) simde_mm_extract_epi16(a, imm8)2707
-#endif2708
-2709
-SIMDE_FUNCTION_ATTRIBUTES2710
-simde__m128i simde_mm_insert_epi16(simde__m128i a, int16_t i, const int imm8)2711
- SIMDE_REQUIRE_RANGE(imm8, 0, 7)2712
-{2713
- simde__m128i_private a_ = simde__m128i_to_private(a);2714
- a_.i16[imm8 & 7] = i;2715
- return simde__m128i_from_private(a_);2716
-}2717
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)2718
-#define simde_mm_insert_epi16(a, i, imm8) _mm_insert_epi16((a), (i), (imm8))2719
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2720
-#define simde_mm_insert_epi16(a, i, imm8) \2721
- simde__m128i_from_neon_i16( \2722
- vsetq_lane_s16((i), simde__m128i_to_neon_i16(a), (imm8)))2723
-#endif2724
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2725
-#define _mm_insert_epi16(a, i, imm8) simde_mm_insert_epi16(a, i, imm8)2726
-#endif2727
-2728
-SIMDE_FUNCTION_ATTRIBUTES2729
-simde__m128d2730
-simde_mm_load_pd(simde_float64 const mem_addr[HEDLEY_ARRAY_PARAM(2)])2731
-{2732
- simde_assert_aligned(16, mem_addr);2733
-2734
-#if defined(SIMDE_X86_SSE2_NATIVE)2735
- return _mm_load_pd(mem_addr);2736
-#else2737
- simde__m128d_private r_;2738
-2739
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2740
- r_.neon_u32 =2741
- vld1q_u32(HEDLEY_REINTERPRET_CAST(uint32_t const *, mem_addr));2742
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE) && !defined(HEDLEY_IBM_VERSION)2743
- r_.altivec_f64 = vec_ld(2744
- 0, HEDLEY_REINTERPRET_CAST(SIMDE_POWER_ALTIVEC_VECTOR(double)2745
- const *,2746
- mem_addr));2747
-#else2748
- r_ = *SIMDE_ALIGN_CAST(simde__m128d_private const *, mem_addr);2749
-#endif2750
-2751
- return simde__m128d_from_private(r_);2752
-#endif2753
-}2754
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2755
-#define _mm_load_pd(mem_addr) simde_mm_load_pd(mem_addr)2756
-#endif2757
-2758
-SIMDE_FUNCTION_ATTRIBUTES2759
-simde__m128d simde_mm_load_pd1(simde_float64 const *mem_addr)2760
-{2761
-#if defined(SIMDE_X86_SSE2_NATIVE)2762
- return _mm_load1_pd(mem_addr);2763
-#else2764
- simde__m128d_private r_;2765
-2766
- r_.f64[0] = *mem_addr;2767
- r_.f64[1] = *mem_addr;2768
-2769
- return simde__m128d_from_private(r_);2770
-#endif2771
-}2772
-#define simde_mm_load1_pd(mem_addr) simde_mm_load_pd1(mem_addr)2773
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2774
-#define _mm_load_pd1(mem_addr) simde_mm_load_pd1(mem_addr)2775
-#define _mm_load1_pd(mem_addr) simde_mm_load1_pd(mem_addr)2776
-#endif2777
-2778
-SIMDE_FUNCTION_ATTRIBUTES2779
-simde__m128d simde_mm_load_sd(simde_float64 const *mem_addr)2780
-{2781
-#if defined(SIMDE_X86_SSE2_NATIVE)2782
- return _mm_load_sd(mem_addr);2783
-#else2784
- simde__m128d_private r_;2785
-2786
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2787
- r_.neon_f64 = vsetq_lane_f64(*mem_addr, vdupq_n_f64(0), 0);2788
-#else2789
- r_.f64[0] = *mem_addr;2790
- r_.u64[1] = UINT64_C(0);2791
-#endif2792
-2793
- return simde__m128d_from_private(r_);2794
-#endif2795
-}2796
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2797
-#define _mm_load_sd(mem_addr) simde_mm_load_sd(mem_addr)2798
-#endif2799
-2800
-SIMDE_FUNCTION_ATTRIBUTES2801
-simde__m128i simde_mm_load_si128(simde__m128i const *mem_addr)2802
-{2803
- simde_assert_aligned(16, mem_addr);2804
-2805
-#if defined(SIMDE_X86_SSE2_NATIVE)2806
- return _mm_load_si128(2807
- HEDLEY_REINTERPRET_CAST(__m128i const *, mem_addr));2808
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2809
- simde__m128i_private r_;2810
-2811
-#if defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)2812
- r_.altivec_i32 = vec_ld(2813
- 0, HEDLEY_REINTERPRET_CAST(2814
- SIMDE_POWER_ALTIVEC_VECTOR(int) const *, mem_addr));2815
-#else2816
- r_.neon_i32 = vld1q_s32((int32_t const *)mem_addr);2817
-#endif2818
-2819
- return simde__m128i_from_private(r_);2820
-#else2821
- return *mem_addr;2822
-#endif2823
-}2824
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2825
-#define _mm_load_si128(mem_addr) simde_mm_load_si128(mem_addr)2826
-#endif2827
-2828
-SIMDE_FUNCTION_ATTRIBUTES2829
-simde__m128d simde_mm_loadh_pd(simde__m128d a, simde_float64 const *mem_addr)2830
-{2831
-#if defined(SIMDE_X86_SSE2_NATIVE)2832
- return _mm_loadh_pd(a, mem_addr);2833
-#else2834
- simde__m128d_private r_, a_ = simde__m128d_to_private(a);2835
- simde_float64 t;2836
-2837
- simde_memcpy(&t, mem_addr, sizeof(t));2838
- r_.f64[0] = a_.f64[0];2839
- r_.f64[1] = t;2840
-2841
- return simde__m128d_from_private(r_);2842
-#endif2843
-}2844
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2845
-#define _mm_loadh_pd(a, mem_addr) simde_mm_loadh_pd(a, mem_addr)2846
-#endif2847
-2848
-SIMDE_FUNCTION_ATTRIBUTES2849
-simde__m128i simde_mm_loadl_epi64(simde__m128i const *mem_addr)2850
-{2851
- simde_assert_aligned(16, mem_addr);2852
-2853
-#if defined(SIMDE_X86_SSE2_NATIVE)2854
- return _mm_loadl_epi64(2855
- HEDLEY_REINTERPRET_CAST(__m128i const *, mem_addr));2856
-#else2857
- simde__m128i_private r_;2858
-2859
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2860
- r_.neon_i32 = vcombine_s32(vld1_s32((int32_t const *)mem_addr),2861
- vcreate_s32(0));2862
-#else2863
- r_.i64[0] = *HEDLEY_REINTERPRET_CAST(int64_t const *, mem_addr);2864
- r_.i64[1] = 0;2865
-#endif2866
-2867
- return simde__m128i_from_private(r_);2868
-#endif2869
-}2870
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2871
-#define _mm_loadl_epi64(mem_addr) simde_mm_loadl_epi64(mem_addr)2872
-#endif2873
-2874
-SIMDE_FUNCTION_ATTRIBUTES2875
-simde__m128d simde_mm_loadl_pd(simde__m128d a, simde_float64 const *mem_addr)2876
-{2877
-#if defined(SIMDE_X86_SSE2_NATIVE)2878
- return _mm_loadl_pd(a, mem_addr);2879
-#else2880
- simde__m128d_private r_, a_ = simde__m128d_to_private(a);2881
-2882
- r_.f64[0] = *mem_addr;2883
- r_.u64[1] = a_.u64[1];2884
-2885
- return simde__m128d_from_private(r_);2886
-#endif2887
-}2888
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2889
-#define _mm_loadl_pd(a, mem_addr) simde_mm_loadl_pd(a, mem_addr)2890
-#endif2891
-2892
-SIMDE_FUNCTION_ATTRIBUTES2893
-simde__m128d2894
-simde_mm_loadr_pd(simde_float64 const mem_addr[HEDLEY_ARRAY_PARAM(2)])2895
-{2896
- simde_assert_aligned(16, mem_addr);2897
-2898
-#if defined(SIMDE_X86_SSE2_NATIVE)2899
- return _mm_loadr_pd(mem_addr);2900
-#else2901
- simde__m128d_private r_;2902
-2903
- r_.f64[0] = mem_addr[1];2904
- r_.f64[1] = mem_addr[0];2905
-2906
- return simde__m128d_from_private(r_);2907
-#endif2908
-}2909
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2910
-#define _mm_loadr_pd(mem_addr) simde_mm_loadr_pd(mem_addr)2911
-#endif2912
-2913
-SIMDE_FUNCTION_ATTRIBUTES2914
-simde__m128d2915
-simde_mm_loadu_pd(simde_float64 const mem_addr[HEDLEY_ARRAY_PARAM(2)])2916
-{2917
-#if defined(SIMDE_X86_SSE2_NATIVE)2918
- return _mm_loadu_pd(mem_addr);2919
-#else2920
- simde__m128d_private r_;2921
-2922
- simde_memcpy(&r_, mem_addr, sizeof(r_));2923
-2924
- return simde__m128d_from_private(r_);2925
-#endif2926
-}2927
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2928
-#define _mm_loadu_pd(mem_addr) simde_mm_loadu_pd(mem_addr)2929
-#endif2930
-2931
-SIMDE_FUNCTION_ATTRIBUTES2932
-simde__m128i simde_mm_loadu_si128(simde__m128i const *mem_addr)2933
-{2934
-#if defined(SIMDE_X86_SSE2_NATIVE)2935
- return _mm_loadu_si128(HEDLEY_STATIC_CAST(__m128i const *, mem_addr));2936
-#else2937
- simde__m128i_private r_;2938
-2939
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2940
- r_.neon_i32 = vld1q_s32((int32_t const *)mem_addr);2941
-#else2942
- simde_memcpy(&r_, mem_addr, sizeof(r_));2943
-#endif2944
-2945
- return simde__m128i_from_private(r_);2946
-#endif2947
-}2948
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2949
-#define _mm_loadu_si128(mem_addr) simde_mm_loadu_si128(mem_addr)2950
-#endif2951
-2952
-SIMDE_FUNCTION_ATTRIBUTES2953
-simde__m128i simde_mm_madd_epi16(simde__m128i a, simde__m128i b)2954
-{2955
-#if defined(SIMDE_X86_SSE2_NATIVE)2956
- return _mm_madd_epi16(a, b);2957
-#else2958
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),2959
- b_ = simde__m128i_to_private(b);2960
-2961
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2962
- int32x4_t pl =2963
- vmull_s16(vget_low_s16(a_.neon_i16), vget_low_s16(b_.neon_i16));2964
- int32x4_t ph = vmull_s16(vget_high_s16(a_.neon_i16),2965
- vget_high_s16(b_.neon_i16));2966
- int32x2_t rl = vpadd_s32(vget_low_s32(pl), vget_high_s32(pl));2967
- int32x2_t rh = vpadd_s32(vget_low_s32(ph), vget_high_s32(ph));2968
- r_.neon_i32 = vcombine_s32(rl, rh);2969
-#else2970
- SIMDE_VECTORIZE2971
- for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i16[0])); i += 2) {2972
- r_.i32[i / 2] = (a_.i16[i] * b_.i16[i]) +2973
- (a_.i16[i + 1] * b_.i16[i + 1]);2974
- }2975
-#endif2976
-2977
- return simde__m128i_from_private(r_);2978
-#endif2979
-}2980
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2981
-#define _mm_madd_epi16(a, b) simde_mm_madd_epi16(a, b)2982
-#endif2983
-2984
-SIMDE_FUNCTION_ATTRIBUTES2985
-void simde_mm_maskmoveu_si128(simde__m128i a, simde__m128i mask,2986
- int8_t mem_addr[HEDLEY_ARRAY_PARAM(16)])2987
-{2988
-#if defined(SIMDE_X86_SSE2_NATIVE)2989
- _mm_maskmoveu_si128(a, mask, HEDLEY_REINTERPRET_CAST(char *, mem_addr));2990
-#else2991
- simde__m128i_private a_ = simde__m128i_to_private(a),2992
- mask_ = simde__m128i_to_private(mask);2993
-2994
- for (size_t i = 0; i < (sizeof(a_.i8) / sizeof(a_.i8[0])); i++) {2995
- if (mask_.u8[i] & 0x80) {2996
- mem_addr[i] = a_.i8[i];2997
- }2998
- }2999
-#endif3000
-}3001
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3002
-#define _mm_maskmoveu_si128(a, mask, mem_addr) \3003
- simde_mm_maskmoveu_si128( \3004
- (a), (mask), \3005
- SIMDE_CHECKED_REINTERPRET_CAST(int8_t *, char *, (mem_addr)))3006
-#endif3007
-3008
-SIMDE_FUNCTION_ATTRIBUTES3009
-int32_t simde_mm_movemask_epi8(simde__m128i a)3010
-{3011
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__INTEL_COMPILER)3012
- /* ICC has trouble with _mm_movemask_epi8 at -O2 and above: */3013
- return _mm_movemask_epi8(a);3014
-#else3015
- int32_t r = 0;3016
- simde__m128i_private a_ = simde__m128i_to_private(a);3017
-3018
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3019
- uint8x16_t input = a_.neon_u8;3020
- SIMDE_ALIGN_AS(16, int8x8_t)3021
- static const int8_t xr[8] = {-7, -6, -5, -4, -3, -2, -1, 0};3022
- uint8x8_t mask_and = vdup_n_u8(0x80);3023
- int8x8_t mask_shift = vld1_s8(xr);3024
-3025
- uint8x8_t lo = vget_low_u8(input);3026
- uint8x8_t hi = vget_high_u8(input);3027
-3028
- lo = vand_u8(lo, mask_and);3029
- lo = vshl_u8(lo, mask_shift);3030
-3031
- hi = vand_u8(hi, mask_and);3032
- hi = vshl_u8(hi, mask_shift);3033
-3034
- lo = vpadd_u8(lo, lo);3035
- lo = vpadd_u8(lo, lo);3036
- lo = vpadd_u8(lo, lo);3037
-3038
- hi = vpadd_u8(hi, hi);3039
- hi = vpadd_u8(hi, hi);3040
- hi = vpadd_u8(hi, hi);3041
-3042
- r = ((hi[0] << 8) | (lo[0] & 0xFF));3043
-#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE) && !defined(HEDLEY_IBM_VERSION)3044
- static const SIMDE_POWER_ALTIVEC_VECTOR(unsigned char)3045
- perm = {120, 112, 104, 96, 88, 80, 72, 64,3046
- 56, 48, 40, 32, 24, 16, 8, 0};3047
- r = HEDLEY_STATIC_CAST(3048
- int32_t, vec_extract(vec_vbpermq(a_.altivec_u8, perm), 1));3049
-#else3050
- SIMDE_VECTORIZE_REDUCTION(| : r)3051
- for (size_t i = 0; i < (sizeof(a_.u8) / sizeof(a_.u8[0])); i++) {3052
- r |= (a_.u8[15 - i] >> 7) << (15 - i);3053
- }3054
-#endif3055
-3056
- return r;3057
-#endif3058
-}3059
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3060
-#define _mm_movemask_epi8(a) simde_mm_movemask_epi8(a)3061
-#endif3062
-3063
-SIMDE_FUNCTION_ATTRIBUTES3064
-int32_t simde_mm_movemask_pd(simde__m128d a)3065
-{3066
-#if defined(SIMDE_X86_SSE2_NATIVE)3067
- return _mm_movemask_pd(a);3068
-#else3069
- int32_t r = 0;3070
- simde__m128d_private a_ = simde__m128d_to_private(a);3071
-3072
- SIMDE_VECTORIZE3073
- for (size_t i = 0; i < (sizeof(a_.u64) / sizeof(a_.u64[0])); i++) {3074
- r |= (a_.u64[i] >> 63) << i;3075
- }3076
-3077
- return r;3078
-#endif3079
-}3080
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3081
-#define _mm_movemask_pd(a) simde_mm_movemask_pd(a)3082
-#endif3083
-3084
-SIMDE_FUNCTION_ATTRIBUTES3085
-simde__m64 simde_mm_movepi64_pi64(simde__m128i a)3086
-{3087
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3088
- return _mm_movepi64_pi64(a);3089
-#else3090
- simde__m64_private r_;3091
- simde__m128i_private a_ = simde__m128i_to_private(a);3092
-3093
- r_.i64[0] = a_.i64[0];3094
-3095
- return simde__m64_from_private(r_);3096
-#endif3097
-}3098
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3099
-#define _mm_movepi64_pi64(a) simde_mm_movepi64_pi64(a)3100
-#endif3101
-3102
-SIMDE_FUNCTION_ATTRIBUTES3103
-simde__m128i simde_mm_movpi64_epi64(simde__m64 a)3104
-{3105
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3106
- return _mm_movpi64_epi64(a);3107
-#else3108
- simde__m128i_private r_;3109
- simde__m64_private a_ = simde__m64_to_private(a);3110
-3111
- r_.i64[0] = a_.i64[0];3112
- r_.i64[1] = 0;3113
-3114
- return simde__m128i_from_private(r_);3115
-#endif3116
-}3117
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3118
-#define _mm_movpi64_epi64(a) simde_mm_movpi64_epi64(a)3119
-#endif3120
-3121
-SIMDE_FUNCTION_ATTRIBUTES3122
-simde__m128i simde_mm_min_epi16(simde__m128i a, simde__m128i b)3123
-{3124
-#if defined(SIMDE_X86_SSE2_NATIVE)3125
- return _mm_min_epi16(a, b);3126
-#else3127
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3128
- b_ = simde__m128i_to_private(b);3129
-3130
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3131
- r_.neon_i16 = vminq_s16(a_.neon_i16, b_.neon_i16);3132
-#else3133
- SIMDE_VECTORIZE3134
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {3135
- r_.i16[i] = (a_.i16[i] < b_.i16[i]) ? a_.i16[i] : b_.i16[i];3136
- }3137
-#endif3138
-3139
- return simde__m128i_from_private(r_);3140
-#endif3141
-}3142
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3143
-#define _mm_min_epi16(a, b) simde_mm_min_epi16(a, b)3144
-#endif3145
-3146
-SIMDE_FUNCTION_ATTRIBUTES3147
-simde__m128i simde_mm_min_epu8(simde__m128i a, simde__m128i b)3148
-{3149
-#if defined(SIMDE_X86_SSE2_NATIVE)3150
- return _mm_min_epu8(a, b);3151
-#else3152
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3153
- b_ = simde__m128i_to_private(b);3154
-3155
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3156
- r_.neon_u8 = vminq_u8(a_.neon_u8, b_.neon_u8);3157
-#else3158
- SIMDE_VECTORIZE3159
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {3160
- r_.u8[i] = (a_.u8[i] < b_.u8[i]) ? a_.u8[i] : b_.u8[i];3161
- }3162
-#endif3163
-3164
- return simde__m128i_from_private(r_);3165
-#endif3166
-}3167
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3168
-#define _mm_min_epu8(a, b) simde_mm_min_epu8(a, b)3169
-#endif3170
-3171
-SIMDE_FUNCTION_ATTRIBUTES3172
-simde__m128d simde_mm_min_pd(simde__m128d a, simde__m128d b)3173
-{3174
-#if defined(SIMDE_X86_SSE2_NATIVE)3175
- return _mm_min_pd(a, b);3176
-#else3177
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),3178
- b_ = simde__m128d_to_private(b);3179
-3180
- SIMDE_VECTORIZE3181
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {3182
- r_.f64[i] = (a_.f64[i] < b_.f64[i]) ? a_.f64[i] : b_.f64[i];3183
- }3184
-3185
- return simde__m128d_from_private(r_);3186
-#endif3187
-}3188
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3189
-#define _mm_min_pd(a, b) simde_mm_min_pd(a, b)3190
-#endif3191
-3192
-SIMDE_FUNCTION_ATTRIBUTES3193
-simde__m128d simde_mm_min_sd(simde__m128d a, simde__m128d b)3194
-{3195
-#if defined(SIMDE_X86_SSE2_NATIVE)3196
- return _mm_min_sd(a, b);3197
-#elif defined(SIMDE_ASSUME_VECTORIZATION)3198
- return simde_mm_move_sd(a, simde_mm_min_pd(a, b));3199
-#else3200
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),3201
- b_ = simde__m128d_to_private(b);3202
-3203
- r_.f64[0] = (a_.f64[0] < b_.f64[0]) ? a_.f64[0] : b_.f64[0];3204
- r_.f64[1] = a_.f64[1];3205
-3206
- return simde__m128d_from_private(r_);3207
-#endif3208
-}3209
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3210
-#define _mm_min_sd(a, b) simde_mm_min_sd(a, b)3211
-#endif3212
-3213
-SIMDE_FUNCTION_ATTRIBUTES3214
-simde__m128i simde_mm_max_epi16(simde__m128i a, simde__m128i b)3215
-{3216
-#if defined(SIMDE_X86_SSE2_NATIVE)3217
- return _mm_max_epi16(a, b);3218
-#else3219
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3220
- b_ = simde__m128i_to_private(b);3221
-3222
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3223
- r_.neon_i16 = vmaxq_s16(a_.neon_i16, b_.neon_i16);3224
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)3225
- r_.altivec_i16 = vec_max(a_.altivec_i16, b_.altivec_i16);3226
-#else3227
- SIMDE_VECTORIZE3228
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {3229
- r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? a_.i16[i] : b_.i16[i];3230
- }3231
-#endif3232
-3233
- return simde__m128i_from_private(r_);3234
-#endif3235
-}3236
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3237
-#define _mm_max_epi16(a, b) simde_mm_max_epi16(a, b)3238
-#endif3239
-3240
-SIMDE_FUNCTION_ATTRIBUTES3241
-simde__m128i simde_mm_max_epu8(simde__m128i a, simde__m128i b)3242
-{3243
-#if defined(SIMDE_X86_SSE2_NATIVE)3244
- return _mm_max_epu8(a, b);3245
-#else3246
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3247
- b_ = simde__m128i_to_private(b);3248
-3249
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3250
- r_.neon_u8 = vmaxq_u8(a_.neon_u8, b_.neon_u8);3251
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)3252
- r_.altivec_u8 = vec_max(a_.altivec_u8, b_.altivec_u8);3253
-#else3254
- SIMDE_VECTORIZE3255
- for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {3256
- r_.u8[i] = (a_.u8[i] > b_.u8[i]) ? a_.u8[i] : b_.u8[i];3257
- }3258
-#endif3259
-3260
- return simde__m128i_from_private(r_);3261
-#endif3262
-}3263
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3264
-#define _mm_max_epu8(a, b) simde_mm_max_epu8(a, b)3265
-#endif3266
-3267
-SIMDE_FUNCTION_ATTRIBUTES3268
-simde__m128d simde_mm_max_pd(simde__m128d a, simde__m128d b)3269
-{3270
-#if defined(SIMDE_X86_SSE2_NATIVE)3271
- return _mm_max_pd(a, b);3272
-#else3273
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),3274
- b_ = simde__m128d_to_private(b);3275
-3276
-#if defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)3277
- r_.altivec_f64 = vec_max(a_.altivec_f64, b_.altivec_f64);3278
-#else3279
- SIMDE_VECTORIZE3280
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {3281
- r_.f64[i] = (a_.f64[i] > b_.f64[i]) ? a_.f64[i] : b_.f64[i];3282
- }3283
-#endif3284
-3285
- return simde__m128d_from_private(r_);3286
-#endif3287
-}3288
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3289
-#define _mm_max_pd(a, b) simde_mm_max_pd(a, b)3290
-#endif3291
-3292
-SIMDE_FUNCTION_ATTRIBUTES3293
-simde__m128d simde_mm_max_sd(simde__m128d a, simde__m128d b)3294
-{3295
-#if defined(SIMDE_X86_SSE2_NATIVE)3296
- return _mm_max_sd(a, b);3297
-#elif defined(SIMDE_ASSUME_VECTORIZATION)3298
- return simde_mm_move_sd(a, simde_mm_max_pd(a, b));3299
-#else3300
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),3301
- b_ = simde__m128d_to_private(b);3302
-3303
- r_.f64[0] = (a_.f64[0] > b_.f64[0]) ? a_.f64[0] : b_.f64[0];3304
- r_.f64[1] = a_.f64[1];3305
-3306
- return simde__m128d_from_private(r_);3307
-#endif3308
-}3309
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3310
-#define _mm_max_sd(a, b) simde_mm_max_sd(a, b)3311
-#endif3312
-3313
-SIMDE_FUNCTION_ATTRIBUTES3314
-simde__m128i simde_mm_move_epi64(simde__m128i a)3315
-{3316
-#if defined(SIMDE_X86_SSE2_NATIVE)3317
- return _mm_move_epi64(a);3318
-#else3319
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);3320
-3321
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3322
- r_.neon_i64 = vsetq_lane_s64(0, a_.neon_i64, 1);3323
-#else3324
- r_.i64[0] = a_.i64[0];3325
- r_.i64[1] = 0;3326
-#endif3327
-3328
- return simde__m128i_from_private(r_);3329
-#endif3330
-}3331
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3332
-#define _mm_move_epi64(a) simde_mm_move_epi64(a)3333
-#endif3334
-3335
-SIMDE_FUNCTION_ATTRIBUTES3336
-simde__m128i simde_mm_mul_epu32(simde__m128i a, simde__m128i b)3337
-{3338
-#if defined(SIMDE_X86_SSE2_NATIVE)3339
- return _mm_mul_epu32(a, b);3340
-#else3341
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3342
- b_ = simde__m128i_to_private(b);3343
-3344
- SIMDE_VECTORIZE3345
- for (size_t i = 0; i < (sizeof(r_.u64) / sizeof(r_.u64[0])); i++) {3346
- r_.u64[i] = HEDLEY_STATIC_CAST(uint64_t, a_.u32[i * 2]) *3347
- HEDLEY_STATIC_CAST(uint64_t, b_.u32[i * 2]);3348
- }3349
-3350
- return simde__m128i_from_private(r_);3351
-#endif3352
-}3353
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3354
-#define _mm_mul_epu32(a, b) simde_mm_mul_epu32(a, b)3355
-#endif3356
-3357
-SIMDE_FUNCTION_ATTRIBUTES3358
-simde__m128i simde_x_mm_mul_epi64(simde__m128i a, simde__m128i b)3359
-{3360
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3361
- b_ = simde__m128i_to_private(b);3362
-3363
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3364
- r_.i64 = a_.i64 * b_.i64;3365
-#else3366
- SIMDE_VECTORIZE3367
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {3368
- r_.i64[i] = a_.i64[i] * b_.i64[i];3369
- }3370
-#endif3371
-3372
- return simde__m128i_from_private(r_);3373
-}3374
-3375
-SIMDE_FUNCTION_ATTRIBUTES3376
-simde__m128i simde_x_mm_mod_epi64(simde__m128i a, simde__m128i b)3377
-{3378
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3379
- b_ = simde__m128i_to_private(b);3380
-3381
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3382
- r_.i64 = a_.i64 % b_.i64;3383
-#else3384
- SIMDE_VECTORIZE3385
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {3386
- r_.i64[i] = a_.i64[i] % b_.i64[i];3387
- }3388
-#endif3389
-3390
- return simde__m128i_from_private(r_);3391
-}3392
-3393
-SIMDE_FUNCTION_ATTRIBUTES3394
-simde__m128d simde_mm_mul_pd(simde__m128d a, simde__m128d b)3395
-{3396
-#if defined(SIMDE_X86_SSE2_NATIVE)3397
- return _mm_mul_pd(a, b);3398
-#else3399
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),3400
- b_ = simde__m128d_to_private(b);3401
-3402
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3403
- r_.f64 = a_.f64 * b_.f64;3404
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)3405
- r_.wasm_v128 = wasm_f64x2_mul(a_.wasm_v128, b_.wasm_v128);3406
-#else3407
- SIMDE_VECTORIZE3408
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {3409
- r_.f64[i] = a_.f64[i] * b_.f64[i];3410
- }3411
-#endif3412
-3413
- return simde__m128d_from_private(r_);3414
-#endif3415
-}3416
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3417
-#define _mm_mul_pd(a, b) simde_mm_mul_pd(a, b)3418
-#endif3419
-3420
-SIMDE_FUNCTION_ATTRIBUTES3421
-simde__m128d simde_mm_mul_sd(simde__m128d a, simde__m128d b)3422
-{3423
-#if defined(SIMDE_X86_SSE2_NATIVE)3424
- return _mm_mul_sd(a, b);3425
-#elif defined(SIMDE_ASSUME_VECTORIZATION)3426
- return simde_mm_move_sd(a, simde_mm_mul_pd(a, b));3427
-#else3428
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),3429
- b_ = simde__m128d_to_private(b);3430
-3431
- r_.f64[0] = a_.f64[0] * b_.f64[0];3432
- r_.f64[1] = a_.f64[1];3433
-3434
- return simde__m128d_from_private(r_);3435
-#endif3436
-}3437
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3438
-#define _mm_mul_sd(a, b) simde_mm_mul_sd(a, b)3439
-#endif3440
-3441
-SIMDE_FUNCTION_ATTRIBUTES3442
-simde__m64 simde_mm_mul_su32(simde__m64 a, simde__m64 b)3443
-{3444
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \3445
- !defined(__PGI)3446
- return _mm_mul_su32(a, b);3447
-#else3448
- simde__m64_private r_, a_ = simde__m64_to_private(a),3449
- b_ = simde__m64_to_private(b);3450
-3451
- r_.u64[0] = HEDLEY_STATIC_CAST(uint64_t, a_.u32[0]) *3452
- HEDLEY_STATIC_CAST(uint64_t, b_.u32[0]);3453
-3454
- return simde__m64_from_private(r_);3455
-#endif3456
-}3457
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3458
-#define _mm_mul_su32(a, b) simde_mm_mul_su32(a, b)3459
-#endif3460
-3461
-SIMDE_FUNCTION_ATTRIBUTES3462
-simde__m128i simde_mm_mulhi_epi16(simde__m128i a, simde__m128i b)3463
-{3464
-#if defined(SIMDE_X86_SSE2_NATIVE)3465
- return _mm_mulhi_epi16(a, b);3466
-#else3467
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3468
- b_ = simde__m128i_to_private(b);3469
-3470
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3471
- int16x4_t a3210 = vget_low_s16(a_.neon_i16);3472
- int16x4_t b3210 = vget_low_s16(b_.neon_i16);3473
- int32x4_t ab3210 = vmull_s16(a3210, b3210); /* 3333222211110000 */3474
- int16x4_t a7654 = vget_high_s16(a_.neon_i16);3475
- int16x4_t b7654 = vget_high_s16(b_.neon_i16);3476
- int32x4_t ab7654 = vmull_s16(a7654, b7654); /* 7777666655554444 */3477
- uint16x8x2_t rv = vuzpq_u16(vreinterpretq_u16_s32(ab3210),3478
- vreinterpretq_u16_s32(ab7654));3479
- r_.neon_u16 = rv.val[1];3480
-#else3481
- SIMDE_VECTORIZE3482
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {3483
- r_.u16[i] = HEDLEY_STATIC_CAST(3484
- uint16_t,3485
- (HEDLEY_STATIC_CAST(3486
- uint32_t,3487
- HEDLEY_STATIC_CAST(int32_t, a_.i16[i]) *3488
- HEDLEY_STATIC_CAST(int32_t,3489
- b_.i16[i])) >>3490
- 16));3491
- }3492
-#endif3493
-3494
- return simde__m128i_from_private(r_);3495
-#endif3496
-}3497
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3498
-#define _mm_mulhi_epi16(a, b) simde_mm_mulhi_epi16(a, b)3499
-#endif3500
-3501
-SIMDE_FUNCTION_ATTRIBUTES3502
-simde__m128i simde_mm_mulhi_epu16(simde__m128i a, simde__m128i b)3503
-{3504
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)3505
- return _mm_mulhi_epu16(a, b);3506
-#else3507
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3508
- b_ = simde__m128i_to_private(b);3509
-3510
- SIMDE_VECTORIZE3511
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {3512
- r_.u16[i] = HEDLEY_STATIC_CAST(3513
- uint16_t,3514
- HEDLEY_STATIC_CAST(uint32_t, a_.u16[i]) *3515
- HEDLEY_STATIC_CAST(uint32_t,3516
- b_.u16[i]) >>3517
- 16);3518
- }3519
-3520
- return simde__m128i_from_private(r_);3521
-#endif3522
-}3523
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3524
-#define _mm_mulhi_epu16(a, b) simde_mm_mulhi_epu16(a, b)3525
-#endif3526
-3527
-SIMDE_FUNCTION_ATTRIBUTES3528
-simde__m128i simde_mm_mullo_epi16(simde__m128i a, simde__m128i b)3529
-{3530
-#if defined(SIMDE_X86_SSE2_NATIVE)3531
- return _mm_mullo_epi16(a, b);3532
-#else3533
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3534
- b_ = simde__m128i_to_private(b);3535
-3536
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3537
- r_.neon_i16 = vmulq_s16(a_.neon_i16, b_.neon_i16);3538
-#else3539
- SIMDE_VECTORIZE3540
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {3541
- r_.u16[i] = HEDLEY_STATIC_CAST(3542
- uint16_t,3543
- HEDLEY_STATIC_CAST(uint32_t, a_.u16[i]) *3544
- HEDLEY_STATIC_CAST(uint32_t, b_.u16[i]));3545
- }3546
-#endif3547
-3548
- return simde__m128i_from_private(r_);3549
-#endif3550
-}3551
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3552
-#define _mm_mullo_epi16(a, b) simde_mm_mullo_epi16(a, b)3553
-#endif3554
-3555
-SIMDE_FUNCTION_ATTRIBUTES3556
-simde__m128d simde_mm_or_pd(simde__m128d a, simde__m128d b)3557
-{3558
-#if defined(SIMDE_X86_SSE2_NATIVE)3559
- return _mm_or_pd(a, b);3560
-#else3561
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),3562
- b_ = simde__m128d_to_private(b);3563
-3564
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3565
- r_.i32f = a_.i32f | b_.i32f;3566
-#else3567
- SIMDE_VECTORIZE3568
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {3569
- r_.i32f[i] = a_.i32f[i] | b_.i32f[i];3570
- }3571
-#endif3572
-3573
- return simde__m128d_from_private(r_);3574
-#endif3575
-}3576
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3577
-#define _mm_or_pd(a, b) simde_mm_or_pd(a, b)3578
-#endif3579
-3580
-SIMDE_FUNCTION_ATTRIBUTES3581
-simde__m128i simde_mm_or_si128(simde__m128i a, simde__m128i b)3582
-{3583
-#if defined(SIMDE_X86_SSE2_NATIVE)3584
- return _mm_or_si128(a, b);3585
-#else3586
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3587
- b_ = simde__m128i_to_private(b);3588
-3589
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3590
- r_.neon_i32 = vorrq_s32(a_.neon_i32, b_.neon_i32);3591
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)3592
- r_.altivec_i32 = vec_or(a_.altivec_i32, b_.altivec_i32);3593
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3594
- r_.i32f = a_.i32f | b_.i32f;3595
-#else3596
- SIMDE_VECTORIZE3597
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {3598
- r_.i32f[i] = a_.i32f[i] | b_.i32f[i];3599
- }3600
-#endif3601
-3602
- return simde__m128i_from_private(r_);3603
-#endif3604
-}3605
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3606
-#define _mm_or_si128(a, b) simde_mm_or_si128(a, b)3607
-#endif3608
-3609
-SIMDE_FUNCTION_ATTRIBUTES3610
-simde__m128i simde_mm_packs_epi16(simde__m128i a, simde__m128i b)3611
-{3612
-#if defined(SIMDE_X86_SSE2_NATIVE)3613
- return _mm_packs_epi16(a, b);3614
-#else3615
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3616
- b_ = simde__m128i_to_private(b);3617
-3618
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3619
- r_.neon_i8 =3620
- vcombine_s8(vqmovn_s16(a_.neon_i16), vqmovn_s16(b_.neon_i16));3621
-#else3622
- SIMDE_VECTORIZE3623
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {3624
- r_.i8[i] = (a_.i16[i] > INT8_MAX)3625
- ? INT8_MAX3626
- : ((a_.i16[i] < INT8_MIN)3627
- ? INT8_MIN3628
- : HEDLEY_STATIC_CAST(int8_t,3629
- a_.i16[i]));3630
- r_.i8[i + 8] = (b_.i16[i] > INT8_MAX)3631
- ? INT8_MAX3632
- : ((b_.i16[i] < INT8_MIN)3633
- ? INT8_MIN3634
- : HEDLEY_STATIC_CAST(3635
- int8_t, b_.i16[i]));3636
- }3637
-#endif3638
-3639
- return simde__m128i_from_private(r_);3640
-#endif3641
-}3642
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3643
-#define _mm_packs_epi16(a, b) simde_mm_packs_epi16(a, b)3644
-#endif3645
-3646
-SIMDE_FUNCTION_ATTRIBUTES3647
-simde__m128i simde_mm_packs_epi32(simde__m128i a, simde__m128i b)3648
-{3649
-#if defined(SIMDE_X86_SSE2_NATIVE)3650
- return _mm_packs_epi32(a, b);3651
-#else3652
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3653
- b_ = simde__m128i_to_private(b);3654
-3655
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3656
- r_.neon_i16 =3657
- vcombine_s16(vqmovn_s32(a_.neon_i32), vqmovn_s32(b_.neon_i32));3658
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)3659
- r_.altivec_i16 = vec_packs(a_.altivec_i32, b_.altivec_i32);3660
-#else3661
- SIMDE_VECTORIZE3662
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {3663
- r_.i16[i] = (a_.i32[i] > INT16_MAX)3664
- ? INT16_MAX3665
- : ((a_.i32[i] < INT16_MIN)3666
- ? INT16_MIN3667
- : HEDLEY_STATIC_CAST(int16_t,3668
- a_.i32[i]));3669
- r_.i16[i + 4] =3670
- (b_.i32[i] > INT16_MAX)3671
- ? INT16_MAX3672
- : ((b_.i32[i] < INT16_MIN)3673
- ? INT16_MIN3674
- : HEDLEY_STATIC_CAST(int16_t,3675
- b_.i32[i]));3676
- }3677
-#endif3678
-3679
- return simde__m128i_from_private(r_);3680
-#endif3681
-}3682
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3683
-#define _mm_packs_epi32(a, b) simde_mm_packs_epi32(a, b)3684
-#endif3685
-3686
-SIMDE_FUNCTION_ATTRIBUTES3687
-simde__m128i simde_mm_packus_epi16(simde__m128i a, simde__m128i b)3688
-{3689
-#if defined(SIMDE_X86_SSE2_NATIVE)3690
- return _mm_packus_epi16(a, b);3691
-#else3692
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3693
- b_ = simde__m128i_to_private(b);3694
-3695
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3696
- r_.neon_u8 =3697
- vcombine_u8(vqmovun_s16(a_.neon_i16), vqmovun_s16(b_.neon_i16));3698
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)3699
- r_.altivec_u8 = vec_packsu(a_.altivec_i16, b_.altivec_i16);3700
-#else3701
- SIMDE_VECTORIZE3702
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {3703
- r_.u8[i] = (a_.i16[i] > UINT8_MAX)3704
- ? UINT8_MAX3705
- : ((a_.i16[i] < 0)3706
- ? UINT8_C(0)3707
- : HEDLEY_STATIC_CAST(uint8_t,3708
- a_.i16[i]));3709
- r_.u8[i + 8] =3710
- (b_.i16[i] > UINT8_MAX)3711
- ? UINT8_MAX3712
- : ((b_.i16[i] < 0)3713
- ? UINT8_C(0)3714
- : HEDLEY_STATIC_CAST(uint8_t,3715
- b_.i16[i]));3716
- }3717
-#endif3718
-3719
- return simde__m128i_from_private(r_);3720
-#endif3721
-}3722
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3723
-#define _mm_packus_epi16(a, b) simde_mm_packus_epi16(a, b)3724
-#endif3725
-3726
-SIMDE_FUNCTION_ATTRIBUTES3727
-void simde_mm_pause(void)3728
-{3729
-#if defined(SIMDE_X86_SSE2_NATIVE)3730
- _mm_pause();3731
-#endif3732
-}3733
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3734
-#define _mm_pause() (simde_mm_pause())3735
-#endif3736
-3737
-SIMDE_FUNCTION_ATTRIBUTES3738
-simde__m128i simde_mm_sad_epu8(simde__m128i a, simde__m128i b)3739
-{3740
-#if defined(SIMDE_X86_SSE2_NATIVE)3741
- return _mm_sad_epu8(a, b);3742
-#else3743
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),3744
- b_ = simde__m128i_to_private(b);3745
-3746
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {3747
- uint16_t tmp = 0;3748
- SIMDE_VECTORIZE_REDUCTION(+ : tmp)3749
- for (size_t j = 0; j < ((sizeof(r_.u8) / sizeof(r_.u8[0])) / 2);3750
- j++) {3751
- const size_t e = j + (i * 8);3752
- tmp += (a_.u8[e] > b_.u8[e]) ? (a_.u8[e] - b_.u8[e])3753
- : (b_.u8[e] - a_.u8[e]);3754
- }3755
- r_.i64[i] = tmp;3756
- }3757
-3758
- return simde__m128i_from_private(r_);3759
-#endif3760
-}3761
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3762
-#define _mm_sad_epu8(a, b) simde_mm_sad_epu8(a, b)3763
-#endif3764
-3765
-SIMDE_FUNCTION_ATTRIBUTES3766
-simde__m128i simde_mm_set_epi8(int8_t e15, int8_t e14, int8_t e13, int8_t e12,3767
- int8_t e11, int8_t e10, int8_t e9, int8_t e8,3768
- int8_t e7, int8_t e6, int8_t e5, int8_t e4,3769
- int8_t e3, int8_t e2, int8_t e1, int8_t e0)3770
-{3771
-3772
-#if defined(SIMDE_X86_SSE2_NATIVE)3773
- return _mm_set_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5,3774
- e4, e3, e2, e1, e0);3775
-#else3776
- simde__m128i_private r_;3777
-3778
-#if defined(SIMDE_WASM_SIMD128_NATIVE)3779
- r_.wasm_v128 = wasm_i8x16_make(e0, e1, e2, e3, e4, e5, e6, e7, e8, e9,3780
- e10, e11, e12, e13, e14, e15);3781
-#else3782
- r_.i8[0] = e0;3783
- r_.i8[1] = e1;3784
- r_.i8[2] = e2;3785
- r_.i8[3] = e3;3786
- r_.i8[4] = e4;3787
- r_.i8[5] = e5;3788
- r_.i8[6] = e6;3789
- r_.i8[7] = e7;3790
- r_.i8[8] = e8;3791
- r_.i8[9] = e9;3792
- r_.i8[10] = e10;3793
- r_.i8[11] = e11;3794
- r_.i8[12] = e12;3795
- r_.i8[13] = e13;3796
- r_.i8[14] = e14;3797
- r_.i8[15] = e15;3798
-#endif3799
-3800
- return simde__m128i_from_private(r_);3801
-#endif3802
-}3803
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3804
-#define _mm_set_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5, e4, e3, \3805
- e2, e1, e0) \3806
- simde_mm_set_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5, \3807
- e4, e3, e2, e1, e0)3808
-#endif3809
-3810
-SIMDE_FUNCTION_ATTRIBUTES3811
-simde__m128i simde_mm_set_epi16(int16_t e7, int16_t e6, int16_t e5, int16_t e4,3812
- int16_t e3, int16_t e2, int16_t e1, int16_t e0)3813
-{3814
-#if defined(SIMDE_X86_SSE2_NATIVE)3815
- return _mm_set_epi16(e7, e6, e5, e4, e3, e2, e1, e0);3816
-#else3817
- simde__m128i_private r_;3818
-3819
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3820
- SIMDE_ALIGN_AS(16, int16x8_t)3821
- int16_t data[8] = {e0, e1, e2, e3, e4, e5, e6, e7};3822
- r_.neon_i16 = vld1q_s16(data);3823
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)3824
- r_.wasm_v128 = wasm_i16x8_make(e0, e1, e2, e3, e4, e5, e6, e7);3825
-#else3826
- r_.i16[0] = e0;3827
- r_.i16[1] = e1;3828
- r_.i16[2] = e2;3829
- r_.i16[3] = e3;3830
- r_.i16[4] = e4;3831
- r_.i16[5] = e5;3832
- r_.i16[6] = e6;3833
- r_.i16[7] = e7;3834
-#endif3835
-3836
- return simde__m128i_from_private(r_);3837
-#endif3838
-}3839
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3840
-#define _mm_set_epi16(e7, e6, e5, e4, e3, e2, e1, e0) \3841
- simde_mm_set_epi16(e7, e6, e5, e4, e3, e2, e1, e0)3842
-#endif3843
-3844
-SIMDE_FUNCTION_ATTRIBUTES3845
-simde__m128i simde_mm_set_epi32(int32_t e3, int32_t e2, int32_t e1, int32_t e0)3846
-{3847
-#if defined(SIMDE_X86_SSE2_NATIVE)3848
- return _mm_set_epi32(e3, e2, e1, e0);3849
-#else3850
- simde__m128i_private r_;3851
-3852
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3853
- SIMDE_ALIGN_AS(16, int32x4_t) int32_t data[4] = {e0, e1, e2, e3};3854
- r_.neon_i32 = vld1q_s32(data);3855
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)3856
- r_.wasm_v128 = wasm_i32x4_make(e0, e1, e2, e3);3857
-#else3858
- r_.i32[0] = e0;3859
- r_.i32[1] = e1;3860
- r_.i32[2] = e2;3861
- r_.i32[3] = e3;3862
-#endif3863
-3864
- return simde__m128i_from_private(r_);3865
-#endif3866
-}3867
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3868
-#define _mm_set_epi32(e3, e2, e1, e0) simde_mm_set_epi32(e3, e2, e1, e0)3869
-#endif3870
-3871
-SIMDE_FUNCTION_ATTRIBUTES3872
-simde__m128i simde_mm_set_epi64(simde__m64 e1, simde__m64 e0)3873
-{3874
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3875
- return _mm_set_epi64(e1, e0);3876
-#else3877
- simde__m128i_private r_;3878
-3879
- r_.m64_private[0] = simde__m64_to_private(e0);3880
- r_.m64_private[1] = simde__m64_to_private(e1);3881
-3882
- return simde__m128i_from_private(r_);3883
-#endif3884
-}3885
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3886
-#define _mm_set_epi64(e1, e0) (simde_mm_set_epi64((e1), (e0)))3887
-#endif3888
-3889
-SIMDE_FUNCTION_ATTRIBUTES3890
-simde__m128i simde_mm_set_epi64x(int64_t e1, int64_t e0)3891
-{3892
-#if defined(SIMDE_X86_SSE2_NATIVE) && \3893
- (!defined(HEDLEY_MSVC_VERSION) || HEDLEY_MSVC_VERSION_CHECK(19, 0, 0))3894
- return _mm_set_epi64x(e1, e0);3895
-#else3896
- simde__m128i_private r_;3897
-3898
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3899
- r_.neon_i64 = vcombine_s64(vdup_n_s64(e0), vdup_n_s64(e1));3900
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)3901
- r_.wasm_v128 = wasm_i64x2_make(e0, e1);3902
-#else3903
- r_.i64[0] = e0;3904
- r_.i64[1] = e1;3905
-#endif3906
-3907
- return simde__m128i_from_private(r_);3908
-#endif3909
-}3910
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3911
-#define _mm_set_epi64x(e1, e0) simde_mm_set_epi64x(e1, e0)3912
-#endif3913
-3914
-SIMDE_FUNCTION_ATTRIBUTES3915
-simde__m128i simde_x_mm_set_epu8(uint8_t e15, uint8_t e14, uint8_t e13,3916
- uint8_t e12, uint8_t e11, uint8_t e10,3917
- uint8_t e9, uint8_t e8, uint8_t e7, uint8_t e6,3918
- uint8_t e5, uint8_t e4, uint8_t e3, uint8_t e2,3919
- uint8_t e1, uint8_t e0)3920
-{3921
-#if defined(SIMDE_X86_SSE2_NATIVE)3922
- return _mm_set_epi8(3923
- HEDLEY_STATIC_CAST(char, e15), HEDLEY_STATIC_CAST(char, e14),3924
- HEDLEY_STATIC_CAST(char, e13), HEDLEY_STATIC_CAST(char, e12),3925
- HEDLEY_STATIC_CAST(char, e11), HEDLEY_STATIC_CAST(char, e10),3926
- HEDLEY_STATIC_CAST(char, e9), HEDLEY_STATIC_CAST(char, e8),3927
- HEDLEY_STATIC_CAST(char, e7), HEDLEY_STATIC_CAST(char, e6),3928
- HEDLEY_STATIC_CAST(char, e5), HEDLEY_STATIC_CAST(char, e4),3929
- HEDLEY_STATIC_CAST(char, e3), HEDLEY_STATIC_CAST(char, e2),3930
- HEDLEY_STATIC_CAST(char, e1), HEDLEY_STATIC_CAST(char, e0));3931
-#else3932
- simde__m128i_private r_;3933
-3934
- r_.u8[0] = e0;3935
- r_.u8[1] = e1;3936
- r_.u8[2] = e2;3937
- r_.u8[3] = e3;3938
- r_.u8[4] = e4;3939
- r_.u8[5] = e5;3940
- r_.u8[6] = e6;3941
- r_.u8[7] = e7;3942
- r_.u8[8] = e8;3943
- r_.u8[9] = e9;3944
- r_.u8[10] = e10;3945
- r_.u8[11] = e11;3946
- r_.u8[12] = e12;3947
- r_.u8[13] = e13;3948
- r_.u8[14] = e14;3949
- r_.u8[15] = e15;3950
-3951
- return simde__m128i_from_private(r_);3952
-#endif3953
-}3954
-3955
-SIMDE_FUNCTION_ATTRIBUTES3956
-simde__m128i simde_x_mm_set_epu16(uint16_t e7, uint16_t e6, uint16_t e5,3957
- uint16_t e4, uint16_t e3, uint16_t e2,3958
- uint16_t e1, uint16_t e0)3959
-{3960
-#if defined(SIMDE_X86_SSE2_NATIVE)3961
- return _mm_set_epi16(3962
- HEDLEY_STATIC_CAST(short, e7), HEDLEY_STATIC_CAST(short, e6),3963
- HEDLEY_STATIC_CAST(short, e5), HEDLEY_STATIC_CAST(short, e4),3964
- HEDLEY_STATIC_CAST(short, e3), HEDLEY_STATIC_CAST(short, e2),3965
- HEDLEY_STATIC_CAST(short, e1), HEDLEY_STATIC_CAST(short, e0));3966
-#else3967
- simde__m128i_private r_;3968
-3969
- r_.u16[0] = e0;3970
- r_.u16[1] = e1;3971
- r_.u16[2] = e2;3972
- r_.u16[3] = e3;3973
- r_.u16[4] = e4;3974
- r_.u16[5] = e5;3975
- r_.u16[6] = e6;3976
- r_.u16[7] = e7;3977
-3978
- return simde__m128i_from_private(r_);3979
-#endif3980
-}3981
-3982
-SIMDE_FUNCTION_ATTRIBUTES3983
-simde__m128i simde_x_mm_set_epu32(uint32_t e3, uint32_t e2, uint32_t e1,3984
- uint32_t e0)3985
-{3986
-#if defined(SIMDE_X86_SSE2_NATIVE)3987
- return _mm_set_epi32(HEDLEY_STATIC_CAST(int, e3),3988
- HEDLEY_STATIC_CAST(int, e2),3989
- HEDLEY_STATIC_CAST(int, e1),3990
- HEDLEY_STATIC_CAST(int, e0));3991
-#else3992
- simde__m128i_private r_;3993
-3994
- r_.u32[0] = e0;3995
- r_.u32[1] = e1;3996
- r_.u32[2] = e2;3997
- r_.u32[3] = e3;3998
-3999
- return simde__m128i_from_private(r_);4000
-#endif4001
-}4002
-4003
-SIMDE_FUNCTION_ATTRIBUTES4004
-simde__m128i simde_x_mm_set_epu64x(uint64_t e1, uint64_t e0)4005
-{4006
-#if defined(SIMDE_X86_SSE2_NATIVE) && \4007
- (!defined(HEDLEY_MSVC_VERSION) || HEDLEY_MSVC_VERSION_CHECK(19, 0, 0))4008
- return _mm_set_epi64x(HEDLEY_STATIC_CAST(int64_t, e1),4009
- HEDLEY_STATIC_CAST(int64_t, e0));4010
-#else4011
- simde__m128i_private r_;4012
-4013
- r_.u64[0] = e0;4014
- r_.u64[1] = e1;4015
-4016
- return simde__m128i_from_private(r_);4017
-#endif4018
-}4019
-4020
-SIMDE_FUNCTION_ATTRIBUTES4021
-simde__m128d simde_mm_set_pd(simde_float64 e1, simde_float64 e0)4022
-{4023
-#if defined(SIMDE_X86_SSE2_NATIVE)4024
- return _mm_set_pd(e1, e0);4025
-#else4026
- simde__m128d_private r_;4027
-4028
-#if defined(SIMDE_WASM_SIMD128_NATIVE)4029
- r_.wasm_v128 = wasm_f64x2_make(e0, e1);4030
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)4031
- r_.wasm_v128 = wasm_f64x2_make(e0, e1);4032
-#else4033
- r_.f64[0] = e0;4034
- r_.f64[1] = e1;4035
-#endif4036
-4037
- return simde__m128d_from_private(r_);4038
-#endif4039
-}4040
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4041
-#define _mm_set_pd(e1, e0) simde_mm_set_pd(e1, e0)4042
-#endif4043
-4044
-SIMDE_FUNCTION_ATTRIBUTES4045
-simde__m128d simde_mm_set_pd1(simde_float64 a)4046
-{4047
-#if defined(SIMDE_X86_SSE2_NATIVE)4048
- return _mm_set1_pd(a);4049
-#else4050
- simde__m128d_private r_;4051
-4052
- r_.f64[0] = a;4053
- r_.f64[1] = a;4054
-4055
- return simde__m128d_from_private(r_);4056
-#endif4057
-}4058
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4059
-#define _mm_set_pd1(a) simde_mm_set1_pd(a)4060
-#endif4061
-4062
-SIMDE_FUNCTION_ATTRIBUTES4063
-simde__m128d simde_mm_set_sd(simde_float64 a)4064
-{4065
-#if defined(SIMDE_X86_SSE2_NATIVE)4066
- return _mm_set_sd(a);4067
-#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)4068
- return vsetq_lane_f64(a, vdupq_n_f64(SIMDE_FLOAT32_C(0.0)), 0);4069
-#else4070
- return simde_mm_set_pd(SIMDE_FLOAT64_C(0.0), a);4071
-4072
-#endif4073
-}4074
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4075
-#define _mm_set_sd(a) simde_mm_set_sd(a)4076
-#endif4077
-4078
-SIMDE_FUNCTION_ATTRIBUTES4079
-simde__m128i simde_mm_set1_epi8(int8_t a)4080
-{4081
-#if defined(SIMDE_X86_SSE2_NATIVE)4082
- return _mm_set1_epi8(a);4083
-#else4084
- simde__m128i_private r_;4085
-4086
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4087
- r_.neon_i8 = vdupq_n_s8(a);4088
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)4089
- r_.wasm_v128 = wasm_i8x16_splat(a);4090
-#else4091
- SIMDE_VECTORIZE4092
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {4093
- r_.i8[i] = a;4094
- }4095
-#endif4096
-4097
- return simde__m128i_from_private(r_);4098
-#endif4099
-}4100
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4101
-#define _mm_set1_epi8(a) simde_mm_set1_epi8(a)4102
-#endif4103
-4104
-SIMDE_FUNCTION_ATTRIBUTES4105
-simde__m128i simde_mm_set1_epi16(int16_t a)4106
-{4107
-#if defined(SIMDE_X86_SSE2_NATIVE)4108
- return _mm_set1_epi16(a);4109
-#else4110
- simde__m128i_private r_;4111
-4112
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4113
- r_.neon_i16 = vdupq_n_s16(a);4114
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)4115
- r_.wasm_v128 = wasm_i16x8_splat(a);4116
-#else4117
- SIMDE_VECTORIZE4118
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4119
- r_.i16[i] = a;4120
- }4121
-#endif4122
-4123
- return simde__m128i_from_private(r_);4124
-#endif4125
-}4126
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4127
-#define _mm_set1_epi16(a) simde_mm_set1_epi16(a)4128
-#endif4129
-4130
-SIMDE_FUNCTION_ATTRIBUTES4131
-simde__m128i simde_mm_set1_epi32(int32_t a)4132
-{4133
-#if defined(SIMDE_X86_SSE2_NATIVE)4134
- return _mm_set1_epi32(a);4135
-#else4136
- simde__m128i_private r_;4137
-4138
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4139
- r_.neon_i32 = vdupq_n_s32(a);4140
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)4141
- r_.wasm_v128 = wasm_i32x4_splat(a);4142
-#else4143
- SIMDE_VECTORIZE4144
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {4145
- r_.i32[i] = a;4146
- }4147
-#endif4148
-4149
- return simde__m128i_from_private(r_);4150
-#endif4151
-}4152
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4153
-#define _mm_set1_epi32(a) simde_mm_set1_epi32(a)4154
-#endif4155
-4156
-SIMDE_FUNCTION_ATTRIBUTES4157
-simde__m128i simde_mm_set1_epi64x(int64_t a)4158
-{4159
-#if defined(SIMDE_X86_SSE2_NATIVE) && \4160
- (!defined(HEDLEY_MSVC_VERSION) || HEDLEY_MSVC_VERSION_CHECK(19, 0, 0))4161
- return _mm_set1_epi64x(a);4162
-#else4163
- simde__m128i_private r_;4164
-4165
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4166
- r_.neon_i64 = vmovq_n_s64(a);4167
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)4168
- r_.wasm_v128 = wasm_i64x2_splat(a);4169
-#else4170
- SIMDE_VECTORIZE4171
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {4172
- r_.i64[i] = a;4173
- }4174
-#endif4175
-4176
- return simde__m128i_from_private(r_);4177
-#endif4178
-}4179
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4180
-#define _mm_set1_epi64x(a) simde_mm_set1_epi64x(a)4181
-#endif4182
-4183
-SIMDE_FUNCTION_ATTRIBUTES4184
-simde__m128i simde_mm_set1_epi64(simde__m64 a)4185
-{4186
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)4187
- return _mm_set1_epi64(a);4188
-#else4189
- simde__m64_private a_ = simde__m64_to_private(a);4190
- return simde_mm_set1_epi64x(a_.i64[0]);4191
-#endif4192
-}4193
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4194
-#define _mm_set1_epi64(a) simde_mm_set1_epi64(a)4195
-#endif4196
-4197
-SIMDE_FUNCTION_ATTRIBUTES4198
-simde__m128i simde_x_mm_set1_epu8(uint8_t value)4199
-{4200
- return simde_mm_set1_epi8(HEDLEY_STATIC_CAST(int8_t, value));4201
-}4202
-4203
-SIMDE_FUNCTION_ATTRIBUTES4204
-simde__m128i simde_x_mm_set1_epu16(uint16_t value)4205
-{4206
- return simde_mm_set1_epi16(HEDLEY_STATIC_CAST(int16_t, value));4207
-}4208
-4209
-SIMDE_FUNCTION_ATTRIBUTES4210
-simde__m128i simde_x_mm_set1_epu32(uint32_t value)4211
-{4212
- return simde_mm_set1_epi32(HEDLEY_STATIC_CAST(int32_t, value));4213
-}4214
-4215
-SIMDE_FUNCTION_ATTRIBUTES4216
-simde__m128i simde_x_mm_set1_epu64(uint64_t value)4217
-{4218
- return simde_mm_set1_epi64x(HEDLEY_STATIC_CAST(int64_t, value));4219
-}4220
-4221
-SIMDE_FUNCTION_ATTRIBUTES4222
-simde__m128d simde_mm_set1_pd(simde_float64 a)4223
-{4224
-#if defined(SIMDE_X86_SSE2_NATIVE)4225
- return _mm_set1_pd(a);4226
-#else4227
- simde__m128d_private r_;4228
-4229
-#if defined(SIMDE_WASM_SIMD128_NATIVE)4230
- r_.wasm_v128 = wasm_f64x2_splat(a);4231
-#else4232
- SIMDE_VECTORIZE4233
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {4234
- r_.f64[i] = a;4235
- }4236
-#endif4237
-4238
- return simde__m128d_from_private(r_);4239
-#endif4240
-}4241
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4242
-#define _mm_set1_pd(a) simde_mm_set1_pd(a)4243
-#endif4244
-4245
-SIMDE_FUNCTION_ATTRIBUTES4246
-simde__m128i simde_mm_setr_epi8(int8_t e15, int8_t e14, int8_t e13, int8_t e12,4247
- int8_t e11, int8_t e10, int8_t e9, int8_t e8,4248
- int8_t e7, int8_t e6, int8_t e5, int8_t e4,4249
- int8_t e3, int8_t e2, int8_t e1, int8_t e0)4250
-{4251
-#if defined(SIMDE_X86_SSE2_NATIVE)4252
- return _mm_setr_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5,4253
- e4, e3, e2, e1, e0);4254
-#else4255
- return simde_mm_set_epi8(e0, e1, e2, e3, e4, e5, e6, e7, e8, e9, e10,4256
- e11, e12, e13, e14, e15);4257
-#endif4258
-}4259
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4260
-#define _mm_setr_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5, e4, \4261
- e3, e2, e1, e0) \4262
- simde_mm_setr_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5, \4263
- e4, e3, e2, e1, e0)4264
-#endif4265
-4266
-SIMDE_FUNCTION_ATTRIBUTES4267
-simde__m128i simde_mm_setr_epi16(int16_t e7, int16_t e6, int16_t e5, int16_t e4,4268
- int16_t e3, int16_t e2, int16_t e1, int16_t e0)4269
-{4270
-#if defined(SIMDE_X86_SSE2_NATIVE)4271
- return _mm_setr_epi16(e7, e6, e5, e4, e3, e2, e1, e0);4272
-#else4273
- return simde_mm_set_epi16(e0, e1, e2, e3, e4, e5, e6, e7);4274
-#endif4275
-}4276
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4277
-#define _mm_setr_epi16(e7, e6, e5, e4, e3, e2, e1, e0) \4278
- simde_mm_setr_epi16(e7, e6, e5, e4, e3, e2, e1, e0)4279
-#endif4280
-4281
-SIMDE_FUNCTION_ATTRIBUTES4282
-simde__m128i simde_mm_setr_epi32(int32_t e3, int32_t e2, int32_t e1, int32_t e0)4283
-{4284
-#if defined(SIMDE_X86_SSE2_NATIVE)4285
- return _mm_setr_epi32(e3, e2, e1, e0);4286
-#else4287
- return simde_mm_set_epi32(e0, e1, e2, e3);4288
-#endif4289
-}4290
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4291
-#define _mm_setr_epi32(e3, e2, e1, e0) simde_mm_setr_epi32(e3, e2, e1, e0)4292
-#endif4293
-4294
-SIMDE_FUNCTION_ATTRIBUTES4295
-simde__m128i simde_mm_setr_epi64(simde__m64 e1, simde__m64 e0)4296
-{4297
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)4298
- return _mm_setr_epi64(e1, e0);4299
-#else4300
- return simde_mm_set_epi64(e0, e1);4301
-#endif4302
-}4303
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4304
-#define _mm_setr_epi64(e1, e0) (simde_mm_setr_epi64((e1), (e0)))4305
-#endif4306
-4307
-SIMDE_FUNCTION_ATTRIBUTES4308
-simde__m128d simde_mm_setr_pd(simde_float64 e1, simde_float64 e0)4309
-{4310
-#if defined(SIMDE_X86_SSE2_NATIVE)4311
- return _mm_setr_pd(e1, e0);4312
-#else4313
- return simde_mm_set_pd(e0, e1);4314
-#endif4315
-}4316
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4317
-#define _mm_setr_pd(e1, e0) simde_mm_setr_pd(e1, e0)4318
-#endif4319
-4320
-SIMDE_FUNCTION_ATTRIBUTES4321
-simde__m128d simde_mm_setzero_pd(void)4322
-{4323
-#if defined(SIMDE_X86_SSE2_NATIVE)4324
- return _mm_setzero_pd();4325
-#else4326
- simde__m128d_private r_;4327
-4328
- SIMDE_VECTORIZE4329
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {4330
- r_.i32f[i] = 0;4331
- }4332
-4333
- return simde__m128d_from_private(r_);4334
-#endif4335
-}4336
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4337
-#define _mm_setzero_pd() simde_mm_setzero_pd()4338
-#endif4339
-4340
-#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)4341
-HEDLEY_DIAGNOSTIC_PUSH4342
-SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_4343
-#endif4344
-4345
-SIMDE_FUNCTION_ATTRIBUTES4346
-simde__m128d simde_mm_undefined_pd(void)4347
-{4348
- simde__m128d_private r_;4349
-4350
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE__HAVE_UNDEFINED128)4351
- r_.n = _mm_undefined_pd();4352
-#elif !defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)4353
- r_ = simde__m128d_to_private(simde_mm_setzero_pd());4354
-#endif4355
-4356
- return simde__m128d_from_private(r_);4357
-}4358
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4359
-#define _mm_undefined_pd() simde_mm_undefined_pd()4360
-#endif4361
-4362
-SIMDE_FUNCTION_ATTRIBUTES4363
-simde__m128i simde_mm_undefined_si128(void)4364
-{4365
- simde__m128i_private r_;4366
-4367
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE__HAVE_UNDEFINED128)4368
- r_.n = _mm_undefined_si128();4369
-#elif !defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)4370
- r_ = simde__m128i_to_private(simde_mm_setzero_si128());4371
-#endif4372
-4373
- return simde__m128i_from_private(r_);4374
-}4375
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4376
-#define _mm_undefined_si128() (simde_mm_undefined_si128())4377
-#endif4378
-4379
-#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)4380
-HEDLEY_DIAGNOSTIC_POP4381
-#endif4382
-4383
-SIMDE_FUNCTION_ATTRIBUTES4384
-simde__m128d simde_x_mm_setone_pd(void)4385
-{4386
- return simde_mm_castps_pd(simde_x_mm_setone_ps());4387
-}4388
-4389
-SIMDE_FUNCTION_ATTRIBUTES4390
-simde__m128i simde_x_mm_setone_si128(void)4391
-{4392
- return simde_mm_castps_si128(simde_x_mm_setone_ps());4393
-}4394
-4395
-SIMDE_FUNCTION_ATTRIBUTES4396
-simde__m128i simde_mm_shuffle_epi32(simde__m128i a, const int imm8)4397
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)4398
-{4399
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4400
-4401
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {4402
- r_.i32[i] = a_.i32[(imm8 >> (i * 2)) & 3];4403
- }4404
-4405
- return simde__m128i_from_private(r_);4406
-}4407
-#if defined(SIMDE_X86_SSE2_NATIVE)4408
-#define simde_mm_shuffle_epi32(a, imm8) _mm_shuffle_epi32((a), (imm8))4409
-#elif defined(SIMDE_SHUFFLE_VECTOR_)4410
-#define simde_mm_shuffle_epi32(a, imm8) \4411
- (__extension__({ \4412
- const simde__m128i_private simde__tmp_a_ = \4413
- simde__m128i_to_private(a); \4414
- simde__m128i_from_private((simde__m128i_private){ \4415
- .i32 = SIMDE_SHUFFLE_VECTOR_( \4416
- 32, 16, (simde__tmp_a_).i32, \4417
- (simde__tmp_a_).i32, ((imm8)) & 3, \4418
- ((imm8) >> 2) & 3, ((imm8) >> 4) & 3, \4419
- ((imm8) >> 6) & 3)}); \4420
- }))4421
-#endif4422
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4423
-#define _mm_shuffle_epi32(a, imm8) simde_mm_shuffle_epi32(a, imm8)4424
-#endif4425
-4426
-SIMDE_FUNCTION_ATTRIBUTES4427
-simde__m128d simde_mm_shuffle_pd(simde__m128d a, simde__m128d b, const int imm8)4428
- SIMDE_REQUIRE_RANGE(imm8, 0, 3)4429
-{4430
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),4431
- b_ = simde__m128d_to_private(b);4432
-4433
- r_.f64[0] = ((imm8 & 1) == 0) ? a_.f64[0] : a_.f64[1];4434
- r_.f64[1] = ((imm8 & 2) == 0) ? b_.f64[0] : b_.f64[1];4435
-4436
- return simde__m128d_from_private(r_);4437
-}4438
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)4439
-#define simde_mm_shuffle_pd(a, b, imm8) _mm_shuffle_pd((a), (b), (imm8))4440
-#elif defined(SIMDE_SHUFFLE_VECTOR_)4441
-#define simde_mm_shuffle_pd(a, b, imm8) \4442
- (__extension__({ \4443
- simde__m128d_from_private((simde__m128d_private){ \4444
- .f64 = SIMDE_SHUFFLE_VECTOR_( \4445
- 64, 16, simde__m128d_to_private(a).f64, \4446
- simde__m128d_to_private(b).f64, \4447
- (((imm8)) & 1), (((imm8) >> 1) & 1) + 2)}); \4448
- }))4449
-#endif4450
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4451
-#define _mm_shuffle_pd(a, b, imm8) simde_mm_shuffle_pd(a, b, imm8)4452
-#endif4453
-4454
-SIMDE_FUNCTION_ATTRIBUTES4455
-simde__m128i simde_mm_shufflehi_epi16(simde__m128i a, const int imm8)4456
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)4457
-{4458
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4459
-4460
- SIMDE_VECTORIZE4461
- for (size_t i = 0; i < ((sizeof(a_.i16) / sizeof(a_.i16[0])) / 2);4462
- i++) {4463
- r_.i16[i] = a_.i16[i];4464
- }4465
- for (size_t i = ((sizeof(a_.i16) / sizeof(a_.i16[0])) / 2);4466
- i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4467
- r_.i16[i] = a_.i16[((imm8 >> ((i - 4) * 2)) & 3) + 4];4468
- }4469
-4470
- return simde__m128i_from_private(r_);4471
-}4472
-#if defined(SIMDE_X86_SSE2_NATIVE)4473
-#define simde_mm_shufflehi_epi16(a, imm8) _mm_shufflehi_epi16((a), (imm8))4474
-#elif defined(SIMDE_SHUFFLE_VECTOR_)4475
-#define simde_mm_shufflehi_epi16(a, imm8) \4476
- (__extension__({ \4477
- const simde__m128i_private simde__tmp_a_ = \4478
- simde__m128i_to_private(a); \4479
- simde__m128i_from_private((simde__m128i_private){ \4480
- .i16 = SIMDE_SHUFFLE_VECTOR_( \4481
- 16, 16, (simde__tmp_a_).i16, \4482
- (simde__tmp_a_).i16, 0, 1, 2, 3, \4483
- (((imm8)) & 3) + 4, (((imm8) >> 2) & 3) + 4, \4484
- (((imm8) >> 4) & 3) + 4, \4485
- (((imm8) >> 6) & 3) + 4)}); \4486
- }))4487
-#endif4488
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4489
-#define _mm_shufflehi_epi16(a, imm8) simde_mm_shufflehi_epi16(a, imm8)4490
-#endif4491
-4492
-SIMDE_FUNCTION_ATTRIBUTES4493
-simde__m128i simde_mm_shufflelo_epi16(simde__m128i a, const int imm8)4494
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)4495
-{4496
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4497
-4498
- for (size_t i = 0; i < ((sizeof(r_.i16) / sizeof(r_.i16[0])) / 2);4499
- i++) {4500
- r_.i16[i] = a_.i16[((imm8 >> (i * 2)) & 3)];4501
- }4502
- SIMDE_VECTORIZE4503
- for (size_t i = ((sizeof(a_.i16) / sizeof(a_.i16[0])) / 2);4504
- i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4505
- r_.i16[i] = a_.i16[i];4506
- }4507
-4508
- return simde__m128i_from_private(r_);4509
-}4510
-#if defined(SIMDE_X86_SSE2_NATIVE)4511
-#define simde_mm_shufflelo_epi16(a, imm8) _mm_shufflelo_epi16((a), (imm8))4512
-#elif defined(SIMDE_SHUFFLE_VECTOR_)4513
-#define simde_mm_shufflelo_epi16(a, imm8) \4514
- (__extension__({ \4515
- const simde__m128i_private simde__tmp_a_ = \4516
- simde__m128i_to_private(a); \4517
- simde__m128i_from_private((simde__m128i_private){ \4518
- .i16 = SIMDE_SHUFFLE_VECTOR_( \4519
- 16, 16, (simde__tmp_a_).i16, \4520
- (simde__tmp_a_).i16, (((imm8)) & 3), \4521
- (((imm8) >> 2) & 3), (((imm8) >> 4) & 3), \4522
- (((imm8) >> 6) & 3), 4, 5, 6, 7)}); \4523
- }))4524
-#endif4525
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4526
-#define _mm_shufflelo_epi16(a, imm8) simde_mm_shufflelo_epi16(a, imm8)4527
-#endif4528
-4529
-SIMDE_FUNCTION_ATTRIBUTES4530
-simde__m128i simde_mm_sll_epi16(simde__m128i a, simde__m128i count)4531
-{4532
-#if defined(SIMDE_X86_SSE2_NATIVE)4533
- return _mm_sll_epi16(a, count);4534
-#else4535
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),4536
- count_ = simde__m128i_to_private(count);4537
-4538
- if (count_.u64[0] > 15)4539
- return simde_mm_setzero_si128();4540
-4541
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)4542
- r_.u16 = (a_.u16 << count_.u64[0]);4543
-#else4544
- SIMDE_VECTORIZE4545
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {4546
- r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t,4547
- (a_.u16[i] << count_.u64[0]));4548
- }4549
-#endif4550
-4551
- return simde__m128i_from_private(r_);4552
-#endif4553
-}4554
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4555
-#define _mm_sll_epi16(a, count) simde_mm_sll_epi16((a), (count))4556
-#endif4557
-4558
-SIMDE_FUNCTION_ATTRIBUTES4559
-simde__m128i simde_mm_sll_epi32(simde__m128i a, simde__m128i count)4560
-{4561
-#if defined(SIMDE_X86_SSE2_NATIVE)4562
- return _mm_sll_epi32(a, count);4563
-#else4564
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),4565
- count_ = simde__m128i_to_private(count);4566
-4567
- if (count_.u64[0] > 31)4568
- return simde_mm_setzero_si128();4569
-4570
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)4571
- r_.u32 = (a_.u32 << count_.u64[0]);4572
-#else4573
- SIMDE_VECTORIZE4574
- for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {4575
- r_.u32[i] = HEDLEY_STATIC_CAST(uint32_t,4576
- (a_.u32[i] << count_.u64[0]));4577
- }4578
-#endif4579
-4580
- return simde__m128i_from_private(r_);4581
-#endif4582
-}4583
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4584
-#define _mm_sll_epi32(a, count) (simde_mm_sll_epi32(a, (count)))4585
-#endif4586
-4587
-SIMDE_FUNCTION_ATTRIBUTES4588
-simde__m128i simde_mm_sll_epi64(simde__m128i a, simde__m128i count)4589
-{4590
-#if defined(SIMDE_X86_SSE2_NATIVE)4591
- return _mm_sll_epi64(a, count);4592
-#else4593
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),4594
- count_ = simde__m128i_to_private(count);4595
-4596
- if (count_.u64[0] > 63)4597
- return simde_mm_setzero_si128();4598
-4599
- const int_fast16_t s = HEDLEY_STATIC_CAST(int_fast16_t, count_.u64[0]);4600
-#if !defined(SIMDE_BUG_GCC_94488)4601
- SIMDE_VECTORIZE4602
-#endif4603
- for (size_t i = 0; i < (sizeof(r_.u64) / sizeof(r_.u64[0])); i++) {4604
- r_.u64[i] = a_.u64[i] << s;4605
- }4606
-4607
- return simde__m128i_from_private(r_);4608
-#endif4609
-}4610
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4611
-#define _mm_sll_epi64(a, count) (simde_mm_sll_epi64(a, (count)))4612
-#endif4613
-4614
-SIMDE_FUNCTION_ATTRIBUTES4615
-simde__m128d simde_mm_sqrt_pd(simde__m128d a)4616
-{4617
-#if defined(SIMDE_X86_SSE2_NATIVE)4618
- return _mm_sqrt_pd(a);4619
-#else4620
- simde__m128d_private r_, a_ = simde__m128d_to_private(a);4621
-4622
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)4623
- r_.neon_f64 = vsqrtq_f64(a_.neon_f64);4624
-#elif defined(simde_math_sqrt)4625
- SIMDE_VECTORIZE4626
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {4627
- r_.f64[i] = simde_math_sqrt(a_.f64[i]);4628
- }4629
-#else4630
- HEDLEY_UNREACHABLE();4631
-#endif4632
-4633
- return simde__m128d_from_private(r_);4634
-#endif4635
-}4636
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4637
-#define _mm_sqrt_pd(a) simde_mm_sqrt_pd(a)4638
-#endif4639
-4640
-SIMDE_FUNCTION_ATTRIBUTES4641
-simde__m128d simde_mm_sqrt_sd(simde__m128d a, simde__m128d b)4642
-{4643
-#if defined(SIMDE_X86_SSE2_NATIVE)4644
- return _mm_sqrt_sd(a, b);4645
-#elif defined(SIMDE_ASSUME_VECTORIZATION)4646
- return simde_mm_move_sd(a, simde_mm_sqrt_pd(b));4647
-#else4648
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),4649
- b_ = simde__m128d_to_private(b);4650
-4651
-#if defined(simde_math_sqrt)4652
- r_.f64[0] = simde_math_sqrt(b_.f64[0]);4653
- r_.f64[1] = a_.f64[1];4654
-#else4655
- HEDLEY_UNREACHABLE();4656
-#endif4657
-4658
- return simde__m128d_from_private(r_);4659
-#endif4660
-}4661
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4662
-#define _mm_sqrt_sd(a, b) simde_mm_sqrt_sd(a, b)4663
-#endif4664
-4665
-SIMDE_FUNCTION_ATTRIBUTES4666
-simde__m128i simde_mm_srl_epi16(simde__m128i a, simde__m128i count)4667
-{4668
-#if defined(SIMDE_X86_SSE2_NATIVE)4669
- return _mm_srl_epi16(a, count);4670
-#else4671
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),4672
- count_ = simde__m128i_to_private(count);4673
-4674
- const int cnt = HEDLEY_STATIC_CAST(4675
- int, (count_.i64[0] > 16 ? 16 : count_.i64[0]));4676
-4677
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4678
- r_.neon_u16 = vshlq_u16(a_.neon_u16,4679
- vdupq_n_s16(HEDLEY_STATIC_CAST(int16_t, -cnt)));4680
-#else4681
- SIMDE_VECTORIZE4682
- for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {4683
- r_.u16[i] = a_.u16[i] >> cnt;4684
- }4685
-#endif4686
-4687
- return simde__m128i_from_private(r_);4688
-#endif4689
-}4690
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4691
-#define _mm_srl_epi16(a, count) (simde_mm_srl_epi16(a, (count)))4692
-#endif4693
-4694
-SIMDE_FUNCTION_ATTRIBUTES4695
-simde__m128i simde_mm_srl_epi32(simde__m128i a, simde__m128i count)4696
-{4697
-#if defined(SIMDE_X86_SSE2_NATIVE)4698
- return _mm_srl_epi32(a, count);4699
-#else4700
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),4701
- count_ = simde__m128i_to_private(count);4702
-4703
- const int cnt = HEDLEY_STATIC_CAST(4704
- int, (count_.i64[0] > 32 ? 32 : count_.i64[0]));4705
-4706
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4707
- r_.neon_u32 = vshlq_u32(a_.neon_u32,4708
- vdupq_n_s32(HEDLEY_STATIC_CAST(int32_t, -cnt)));4709
-#else4710
- SIMDE_VECTORIZE4711
- for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {4712
- r_.u32[i] = a_.u32[i] >> cnt;4713
- }4714
-#endif4715
-4716
- return simde__m128i_from_private(r_);4717
-#endif4718
-}4719
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4720
-#define _mm_srl_epi32(a, count) (simde_mm_srl_epi32(a, (count)))4721
-#endif4722
-4723
-SIMDE_FUNCTION_ATTRIBUTES4724
-simde__m128i simde_mm_srl_epi64(simde__m128i a, simde__m128i count)4725
-{4726
-#if defined(SIMDE_X86_SSE2_NATIVE)4727
- return _mm_srl_epi64(a, count);4728
-#else4729
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),4730
- count_ = simde__m128i_to_private(count);4731
-4732
- const int cnt = HEDLEY_STATIC_CAST(4733
- int, (count_.i64[0] > 64 ? 64 : count_.i64[0]));4734
-4735
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4736
- r_.neon_u64 = vshlq_u64(a_.neon_u64,4737
- vdupq_n_s64(HEDLEY_STATIC_CAST(int64_t, -cnt)));4738
-#else4739
-#if !defined(SIMDE_BUG_GCC_94488)4740
- SIMDE_VECTORIZE4741
-#endif4742
- for (size_t i = 0; i < (sizeof(r_.u64) / sizeof(r_.u64[0])); i++) {4743
- r_.u64[i] = a_.u64[i] >> cnt;4744
- }4745
-#endif4746
-4747
- return simde__m128i_from_private(r_);4748
-#endif4749
-}4750
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4751
-#define _mm_srl_epi64(a, count) (simde_mm_srl_epi64(a, (count)))4752
-#endif4753
-4754
-SIMDE_FUNCTION_ATTRIBUTES4755
-simde__m128i simde_mm_srai_epi16(simde__m128i a, const int imm8)4756
- SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)4757
-{4758
- /* MSVC requires a range of (0, 255). */4759
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4760
-4761
- const int cnt = (imm8 & ~15) ? 15 : imm8;4762
-4763
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4764
- r_.neon_i16 = vshlq_s16(a_.neon_i16, vdupq_n_s16(-cnt));4765
-#else4766
- SIMDE_VECTORIZE4767
- for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i16[0])); i++) {4768
- r_.i16[i] = a_.i16[i] >> cnt;4769
- }4770
-#endif4771
-4772
- return simde__m128i_from_private(r_);4773
-}4774
-#if defined(SIMDE_X86_SSE2_NATIVE)4775
-#define simde_mm_srai_epi16(a, imm8) _mm_srai_epi16((a), (imm8))4776
-#endif4777
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4778
-#define _mm_srai_epi16(a, imm8) simde_mm_srai_epi16(a, imm8)4779
-#endif4780
-4781
-SIMDE_FUNCTION_ATTRIBUTES4782
-simde__m128i simde_mm_srai_epi32(simde__m128i a, const int imm8)4783
- SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)4784
-{4785
- /* MSVC requires a range of (0, 255). */4786
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4787
-4788
- const int cnt = (imm8 & ~31) ? 31 : imm8;4789
-4790
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4791
- r_.neon_i32 = vshlq_s32(a_.neon_i32, vdupq_n_s32(-cnt));4792
-#else4793
- SIMDE_VECTORIZE4794
- for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i32[0])); i++) {4795
- r_.i32[i] = a_.i32[i] >> cnt;4796
- }4797
-#endif4798
-4799
- return simde__m128i_from_private(r_);4800
-}4801
-#if defined(SIMDE_X86_SSE2_NATIVE)4802
-#define simde_mm_srai_epi32(a, imm8) _mm_srai_epi32((a), (imm8))4803
-#endif4804
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4805
-#define _mm_srai_epi32(a, imm8) simde_mm_srai_epi32(a, imm8)4806
-#endif4807
-4808
-SIMDE_FUNCTION_ATTRIBUTES4809
-simde__m128i simde_mm_sra_epi16(simde__m128i a, simde__m128i count)4810
-{4811
-#if defined(SIMDE_X86_SSE2_NATIVE)4812
- return _mm_sra_epi16(a, count);4813
-#else4814
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),4815
- count_ = simde__m128i_to_private(count);4816
-4817
- const int cnt = HEDLEY_STATIC_CAST(4818
- int, (count_.i64[0] > 15 ? 15 : count_.i64[0]));4819
-4820
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4821
- r_.neon_i16 = vshlq_s16(a_.neon_i16,4822
- vdupq_n_s16(HEDLEY_STATIC_CAST(int16_t, -cnt)));4823
-#else4824
- SIMDE_VECTORIZE4825
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4826
- r_.i16[i] = a_.i16[i] >> cnt;4827
- }4828
-#endif4829
-4830
- return simde__m128i_from_private(r_);4831
-#endif4832
-}4833
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4834
-#define _mm_sra_epi16(a, count) (simde_mm_sra_epi16(a, count))4835
-#endif4836
-4837
-SIMDE_FUNCTION_ATTRIBUTES4838
-simde__m128i simde_mm_sra_epi32(simde__m128i a, simde__m128i count)4839
-{4840
-#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(SIMDE_BUG_GCC_BAD_MM_SRA_EPI32)4841
- return _mm_sra_epi32(a, count);4842
-#else4843
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),4844
- count_ = simde__m128i_to_private(count);4845
-4846
- const int cnt = count_.u64[0] > 314847
- ? 314848
- : HEDLEY_STATIC_CAST(int, count_.u64[0]);4849
-4850
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4851
- r_.neon_i32 = vshlq_s32(a_.neon_i32,4852
- vdupq_n_s32(HEDLEY_STATIC_CAST(int32_t, -cnt)));4853
-#else4854
- SIMDE_VECTORIZE4855
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {4856
- r_.i32[i] = a_.i32[i] >> cnt;4857
- }4858
-#endif4859
-4860
- return simde__m128i_from_private(r_);4861
-#endif4862
-}4863
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4864
-#define _mm_sra_epi32(a, count) (simde_mm_sra_epi32(a, (count)))4865
-#endif4866
-4867
-SIMDE_FUNCTION_ATTRIBUTES4868
-simde__m128i simde_mm_slli_epi16(simde__m128i a, const int imm8)4869
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)4870
-{4871
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4872
-4873
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)4874
- r_.i16 = a_.i16 << (imm8 & 0xff);4875
-#else4876
- const int s =4877
- (imm8 >4878
- HEDLEY_STATIC_CAST(int, sizeof(r_.i16[0]) * CHAR_BIT) - 1)4879
- ? 04880
- : imm8;4881
- SIMDE_VECTORIZE4882
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4883
- r_.i16[i] = HEDLEY_STATIC_CAST(int16_t, a_.i16[i] << s);4884
- }4885
-#endif4886
-4887
- return simde__m128i_from_private(r_);4888
-}4889
-#if defined(SIMDE_X86_SSE2_NATIVE)4890
-#define simde_mm_slli_epi16(a, imm8) _mm_slli_epi16(a, imm8)4891
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)4892
-#define simde_mm_slli_epi16(a, imm8) \4893
- simde__m128i_from_neon_u16( \4894
- vshlq_n_u16(simde__m128i_to_neon_u16(a), (imm8)))4895
-#endif4896
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4897
-#define _mm_slli_epi16(a, imm8) simde_mm_slli_epi16(a, imm8)4898
-#endif4899
-4900
-SIMDE_FUNCTION_ATTRIBUTES4901
-simde__m128i simde_mm_slli_epi32(simde__m128i a, const int imm8)4902
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)4903
-{4904
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4905
-4906
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)4907
- r_.i32 = a_.i32 << imm8;4908
-#else4909
- SIMDE_VECTORIZE4910
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {4911
- r_.i32[i] = a_.i32[i] << (imm8 & 0xff);4912
- }4913
-#endif4914
-4915
- return simde__m128i_from_private(r_);4916
-}4917
-#if defined(SIMDE_X86_SSE2_NATIVE)4918
-#define simde_mm_slli_epi32(a, imm8) _mm_slli_epi32(a, imm8)4919
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)4920
-#define simde_mm_slli_epi32(a, imm8) \4921
- simde__m128i_from_neon_u32( \4922
- vshlq_n_u32(simde__m128i_to_neon_u32(a), (imm8)))4923
-#endif4924
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4925
-#define _mm_slli_epi32(a, imm8) simde_mm_slli_epi32(a, imm8)4926
-#endif4927
-4928
-SIMDE_FUNCTION_ATTRIBUTES4929
-simde__m128i simde_mm_slli_epi64(simde__m128i a, const int imm8)4930
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)4931
-{4932
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4933
-4934
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)4935
- r_.i64 = a_.i64 << imm8;4936
-#else4937
- SIMDE_VECTORIZE4938
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {4939
- r_.i64[i] = a_.i64[i] << (imm8 & 0xff);4940
- }4941
-#endif4942
-4943
- return simde__m128i_from_private(r_);4944
-}4945
-#if defined(SIMDE_X86_SSE2_NATIVE)4946
-#define simde_mm_slli_epi64(a, imm8) _mm_slli_epi64(a, imm8)4947
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)4948
-#define simde_mm_slli_epi64(a, imm8) \4949
- simde__m128i_from_neon_u64( \4950
- vshlq_n_u64(simde__m128i_to_neon_u64(a), (imm8)))4951
-#endif4952
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4953
-#define _mm_slli_epi64(a, imm8) simde_mm_slli_epi64(a, imm8)4954
-#endif4955
-4956
-SIMDE_FUNCTION_ATTRIBUTES4957
-simde__m128i simde_mm_srli_epi16(simde__m128i a, const int imm8)4958
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)4959
-{4960
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4961
-4962
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)4963
- r_.u16 = a_.u16 >> imm8;4964
-#else4965
- SIMDE_VECTORIZE4966
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4967
- r_.u16[i] = a_.u16[i] >> (imm8 & 0xff);4968
- }4969
-#endif4970
-4971
- return simde__m128i_from_private(r_);4972
-}4973
-#if defined(SIMDE_X86_SSE2_NATIVE)4974
-#define simde_mm_srli_epi16(a, imm8) _mm_srli_epi16(a, imm8)4975
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)4976
-#define simde_mm_srli_epi16(a, imm8) \4977
- simde__m128i_from_neon_u16( \4978
- vshrq_n_u16(simde__m128i_to_neon_u16(a), imm8))4979
-#endif4980
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4981
-#define _mm_srli_epi16(a, imm8) simde_mm_srli_epi16(a, imm8)4982
-#endif4983
-4984
-SIMDE_FUNCTION_ATTRIBUTES4985
-simde__m128i simde_mm_srli_epi32(simde__m128i a, const int imm8)4986
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)4987
-{4988
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);4989
-4990
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)4991
- r_.u32 = a_.u32 >> (imm8 & 0xff);4992
-#else4993
- SIMDE_VECTORIZE4994
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {4995
- r_.u32[i] = a_.u32[i] >> (imm8 & 0xff);4996
- }4997
-#endif4998
-4999
- return simde__m128i_from_private(r_);5000
-}5001
-#if defined(SIMDE_X86_SSE2_NATIVE)5002
-#define simde_mm_srli_epi32(a, imm8) _mm_srli_epi32(a, imm8)5003
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5004
-#define simde_mm_srli_epi32(a, imm8) \5005
- simde__m128i_from_neon_u32( \5006
- vshrq_n_u32(simde__m128i_to_neon_u32(a), imm8))5007
-#endif5008
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5009
-#define _mm_srli_epi32(a, imm8) simde_mm_srli_epi32(a, imm8)5010
-#endif5011
-5012
-SIMDE_FUNCTION_ATTRIBUTES5013
-simde__m128i simde_mm_srli_epi64(simde__m128i a, const int imm8)5014
- SIMDE_REQUIRE_RANGE(imm8, 0, 255)5015
-{5016
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);5017
-5018
- if (HEDLEY_UNLIKELY((imm8 & 63) != imm8))5019
- return simde_mm_setzero_si128();5020
-5021
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5022
- r_.neon_u64 = vshlq_u64(a_.neon_u64, vdupq_n_s64(-imm8));5023
-#else5024
-#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && !defined(SIMDE_BUG_GCC_94488)5025
- r_.u64 = a_.u64 >> imm8;5026
-#else5027
- SIMDE_VECTORIZE5028
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {5029
- r_.u64[i] = a_.u64[i] >> imm8;5030
- }5031
-#endif5032
-#endif5033
-5034
- return simde__m128i_from_private(r_);5035
-}5036
-#if defined(SIMDE_X86_SSE2_NATIVE)5037
-#define simde_mm_srli_epi64(a, imm8) _mm_srli_epi64(a, imm8)5038
-#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) && !defined(__clang__)5039
-#define simde_mm_srli_epi64(a, imm8) \5040
- ((imm8 == 0) ? (a) \5041
- : (simde__m128i_from_neon_u64(vshrq_n_u64( \5042
- simde__m128i_to_neon_u64(a), imm8))))5043
-#endif5044
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5045
-#define _mm_srli_epi64(a, imm8) simde_mm_srli_epi64(a, imm8)5046
-#endif5047
-5048
-SIMDE_FUNCTION_ATTRIBUTES5049
-void simde_mm_store_pd(simde_float64 mem_addr[HEDLEY_ARRAY_PARAM(2)],5050
- simde__m128d a)5051
-{5052
- simde_assert_aligned(16, mem_addr);5053
-5054
-#if defined(SIMDE_X86_SSE2_NATIVE)5055
- _mm_store_pd(mem_addr, a);5056
-#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)5057
- vst1q_f64(mem_addr, simde__m128d_to_private(a).neon_f64);5058
-#else5059
- simde_memcpy(mem_addr, &a, sizeof(a));5060
-#endif5061
-}5062
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5063
-#define _mm_store_pd(mem_addr, a) \5064
- simde_mm_store_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5065
-#endif5066
-5067
-SIMDE_FUNCTION_ATTRIBUTES5068
-void simde_mm_store1_pd(simde_float64 mem_addr[HEDLEY_ARRAY_PARAM(2)],5069
- simde__m128d a)5070
-{5071
- simde_assert_aligned(16, mem_addr);5072
-5073
-#if defined(SIMDE_X86_SSE2_NATIVE)5074
- _mm_store1_pd(mem_addr, a);5075
-#else5076
- simde__m128d_private a_ = simde__m128d_to_private(a);5077
-5078
- mem_addr[0] = a_.f64[0];5079
- mem_addr[1] = a_.f64[0];5080
-#endif5081
-}5082
-#define simde_mm_store_pd1(mem_addr, a) \5083
- simde_mm_store1_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5084
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5085
-#define _mm_store1_pd(mem_addr, a) \5086
- simde_mm_store1_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5087
-#define _mm_store_pd1(mem_addr, a) \5088
- simde_mm_store_pd1(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5089
-#endif5090
-5091
-SIMDE_FUNCTION_ATTRIBUTES5092
-void simde_mm_store_sd(simde_float64 *mem_addr, simde__m128d a)5093
-{5094
-#if defined(SIMDE_X86_SSE2_NATIVE)5095
- _mm_store_sd(mem_addr, a);5096
-#else5097
- simde__m128d_private a_ = simde__m128d_to_private(a);5098
-5099
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)5100
- simde_float64 v = vgetq_lane_f64(a_.neon_f64, 0);5101
- simde_memcpy(mem_addr, &v, sizeof(simde_float64));5102
-#else5103
- simde_float64 v = a_.f64[0];5104
- simde_memcpy(mem_addr, &v, sizeof(simde_float64));5105
-#endif5106
-#endif5107
-}5108
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5109
-#define _mm_store_sd(mem_addr, a) \5110
- simde_mm_store_sd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5111
-#endif5112
-5113
-SIMDE_FUNCTION_ATTRIBUTES5114
-void simde_mm_store_si128(simde__m128i *mem_addr, simde__m128i a)5115
-{5116
-#if defined(SIMDE_X86_SSE2_NATIVE)5117
- _mm_store_si128(HEDLEY_STATIC_CAST(__m128i *, mem_addr), a);5118
-#else5119
- simde__m128i_private a_ = simde__m128i_to_private(a);5120
-5121
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5122
- vst1q_s32(HEDLEY_REINTERPRET_CAST(int32_t *, mem_addr), a_.neon_i32);5123
-#else5124
- simde_memcpy(SIMDE_ASSUME_ALIGNED(16, mem_addr), &a_, sizeof(a_));5125
-#endif5126
-#endif5127
-}5128
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5129
-#define _mm_store_si128(mem_addr, a) simde_mm_store_si128(mem_addr, a)5130
-#endif5131
-5132
-SIMDE_FUNCTION_ATTRIBUTES5133
-void simde_mm_storeh_pd(simde_float64 *mem_addr, simde__m128d a)5134
-{5135
-#if defined(SIMDE_X86_SSE2_NATIVE)5136
- _mm_storeh_pd(mem_addr, a);5137
-#else5138
- simde__m128d_private a_ = simde__m128d_to_private(a);5139
-5140
-#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)5141
- *mem_addr = vgetq_lane_f64(a_.neon_f64, 1);5142
-#else5143
- *mem_addr = a_.f64[1];5144
-#endif5145
-#endif5146
-}5147
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5148
-#define _mm_storeh_pd(mem_addr, a) \5149
- simde_mm_storeh_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5150
-#endif5151
-5152
-SIMDE_FUNCTION_ATTRIBUTES5153
-void simde_mm_storel_epi64(simde__m128i *mem_addr, simde__m128i a)5154
-{5155
-#if defined(SIMDE_X86_SSE2_NATIVE)5156
- _mm_storel_epi64(HEDLEY_STATIC_CAST(__m128i *, mem_addr), a);5157
-#else5158
- simde__m128i_private a_ = simde__m128i_to_private(a);5159
- int64_t tmp;5160
-5161
- /* memcpy to prevent aliasing, tmp because we can't take the5162
- * address of a vector element. */5163
-5164
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5165
- tmp = vgetq_lane_s64(a_.neon_i64, 0);5166
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)5167
-#if defined(SIMDE_BUG_GCC_95227)5168
- (void)a_;5169
-#endif5170
- tmp = vec_extract(a_.altivec_i64, 0);5171
-#else5172
- tmp = a_.i64[0];5173
-#endif5174
-5175
- simde_memcpy(mem_addr, &tmp, sizeof(tmp));5176
-#endif5177
-}5178
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5179
-#define _mm_storel_epi64(mem_addr, a) simde_mm_storel_epi64(mem_addr, a)5180
-#endif5181
-5182
-SIMDE_FUNCTION_ATTRIBUTES5183
-void simde_mm_storel_pd(simde_float64 *mem_addr, simde__m128d a)5184
-{5185
-#if defined(SIMDE_X86_SSE2_NATIVE)5186
- _mm_storel_pd(mem_addr, a);5187
-#else5188
- simde__m128d_private a_ = simde__m128d_to_private(a);5189
-5190
- *mem_addr = a_.f64[0];5191
-#endif5192
-}5193
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5194
-#define _mm_storel_pd(mem_addr, a) \5195
- simde_mm_storel_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5196
-#endif5197
-5198
-SIMDE_FUNCTION_ATTRIBUTES5199
-void simde_mm_storer_pd(simde_float64 mem_addr[2], simde__m128d a)5200
-{5201
- simde_assert_aligned(16, mem_addr);5202
-5203
-#if defined(SIMDE_X86_SSE2_NATIVE)5204
- _mm_storer_pd(mem_addr, a);5205
-#else5206
- simde__m128d_private a_ = simde__m128d_to_private(a);5207
-5208
- mem_addr[0] = a_.f64[1];5209
- mem_addr[1] = a_.f64[0];5210
-#endif5211
-}5212
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5213
-#define _mm_storer_pd(mem_addr, a) \5214
- simde_mm_storer_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5215
-#endif5216
-5217
-SIMDE_FUNCTION_ATTRIBUTES5218
-void simde_mm_storeu_pd(simde_float64 *mem_addr, simde__m128d a)5219
-{5220
-#if defined(SIMDE_X86_SSE2_NATIVE)5221
- _mm_storeu_pd(mem_addr, a);5222
-#else5223
- simde_memcpy(mem_addr, &a, sizeof(a));5224
-#endif5225
-}5226
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5227
-#define _mm_storeu_pd(mem_addr, a) \5228
- simde_mm_storeu_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5229
-#endif5230
-5231
-SIMDE_FUNCTION_ATTRIBUTES5232
-void simde_mm_storeu_si128(simde__m128i *mem_addr, simde__m128i a)5233
-{5234
-#if defined(SIMDE_X86_SSE2_NATIVE)5235
- _mm_storeu_si128(HEDLEY_STATIC_CAST(__m128i *, mem_addr), a);5236
-#else5237
- simde__m128i_private a_ = simde__m128i_to_private(a);5238
-5239
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5240
- vst1q_s32(HEDLEY_REINTERPRET_CAST(int32_t *, mem_addr), a_.neon_i32);5241
-#else5242
- simde_memcpy(mem_addr, &a_, sizeof(a_));5243
-#endif5244
-#endif5245
-}5246
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5247
-#define _mm_storeu_si128(mem_addr, a) simde_mm_storeu_si128(mem_addr, a)5248
-#endif5249
-5250
-SIMDE_FUNCTION_ATTRIBUTES5251
-void simde_mm_stream_pd(simde_float64 mem_addr[HEDLEY_ARRAY_PARAM(2)],5252
- simde__m128d a)5253
-{5254
- simde_assert_aligned(16, mem_addr);5255
-5256
-#if defined(SIMDE_X86_SSE2_NATIVE)5257
- _mm_stream_pd(mem_addr, a);5258
-#else5259
- simde_memcpy(mem_addr, &a, sizeof(a));5260
-#endif5261
-}5262
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5263
-#define _mm_stream_pd(mem_addr, a) \5264
- simde_mm_stream_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)5265
-#endif5266
-5267
-SIMDE_FUNCTION_ATTRIBUTES5268
-void simde_mm_stream_si128(simde__m128i *mem_addr, simde__m128i a)5269
-{5270
- simde_assert_aligned(16, mem_addr);5271
-5272
-#if defined(SIMDE_X86_SSE2_NATIVE)5273
- _mm_stream_si128(HEDLEY_STATIC_CAST(__m128i *, mem_addr), a);5274
-#else5275
- simde_memcpy(mem_addr, &a, sizeof(a));5276
-#endif5277
-}5278
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5279
-#define _mm_stream_si128(mem_addr, a) simde_mm_stream_si128(mem_addr, a)5280
-#endif5281
-5282
-SIMDE_FUNCTION_ATTRIBUTES5283
-void simde_mm_stream_si32(int32_t *mem_addr, int32_t a)5284
-{5285
-#if defined(SIMDE_X86_SSE2_NATIVE)5286
- _mm_stream_si32(mem_addr, a);5287
-#else5288
- *mem_addr = a;5289
-#endif5290
-}5291
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5292
-#define _mm_stream_si32(mem_addr, a) simde_mm_stream_si32(mem_addr, a)5293
-#endif5294
-5295
-SIMDE_FUNCTION_ATTRIBUTES5296
-void simde_mm_stream_si64(int64_t *mem_addr, int64_t a)5297
-{5298
- *mem_addr = a;5299
-}5300
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5301
-#define _mm_stream_si64(mem_addr, a) \5302
- simde_mm_stream_si64(SIMDE_CHECKED_REINTERPRET_CAST( \5303
- int64_t *, __int64 *, mem_addr), \5304
- a)5305
-#endif5306
-5307
-SIMDE_FUNCTION_ATTRIBUTES5308
-simde__m128i simde_mm_sub_epi8(simde__m128i a, simde__m128i b)5309
-{5310
-#if defined(SIMDE_X86_SSE2_NATIVE)5311
- return _mm_sub_epi8(a, b);5312
-#else5313
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5314
- b_ = simde__m128i_to_private(b);5315
-5316
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5317
- r_.neon_i8 = vsubq_s8(a_.neon_i8, b_.neon_i8);5318
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)5319
- r_.i8 = a_.i8 - b_.i8;5320
-#else5321
- SIMDE_VECTORIZE5322
- for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {5323
- r_.i8[i] = a_.i8[i] - b_.i8[i];5324
- }5325
-#endif5326
-5327
- return simde__m128i_from_private(r_);5328
-#endif5329
-}5330
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5331
-#define _mm_sub_epi8(a, b) simde_mm_sub_epi8(a, b)5332
-#endif5333
-5334
-SIMDE_FUNCTION_ATTRIBUTES5335
-simde__m128i simde_mm_sub_epi16(simde__m128i a, simde__m128i b)5336
-{5337
-#if defined(SIMDE_X86_SSE2_NATIVE)5338
- return _mm_sub_epi16(a, b);5339
-#else5340
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5341
- b_ = simde__m128i_to_private(b);5342
-5343
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5344
- r_.neon_i16 = vsubq_s16(a_.neon_i16, b_.neon_i16);5345
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)5346
- r_.i16 = a_.i16 - b_.i16;5347
-#else5348
- SIMDE_VECTORIZE5349
- for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {5350
- r_.i16[i] = a_.i16[i] - b_.i16[i];5351
- }5352
-#endif5353
-5354
- return simde__m128i_from_private(r_);5355
-#endif5356
-}5357
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5358
-#define _mm_sub_epi16(a, b) simde_mm_sub_epi16(a, b)5359
-#endif5360
-5361
-SIMDE_FUNCTION_ATTRIBUTES5362
-simde__m128i simde_mm_sub_epi32(simde__m128i a, simde__m128i b)5363
-{5364
-#if defined(SIMDE_X86_SSE2_NATIVE)5365
- return _mm_sub_epi32(a, b);5366
-#else5367
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5368
- b_ = simde__m128i_to_private(b);5369
-5370
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5371
- r_.neon_i32 = vsubq_s32(a_.neon_i32, b_.neon_i32);5372
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)5373
- r_.i32 = a_.i32 - b_.i32;5374
-#else5375
- SIMDE_VECTORIZE5376
- for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {5377
- r_.i32[i] = a_.i32[i] - b_.i32[i];5378
- }5379
-#endif5380
-5381
- return simde__m128i_from_private(r_);5382
-#endif5383
-}5384
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5385
-#define _mm_sub_epi32(a, b) simde_mm_sub_epi32(a, b)5386
-#endif5387
-5388
-SIMDE_FUNCTION_ATTRIBUTES5389
-simde__m128i simde_mm_sub_epi64(simde__m128i a, simde__m128i b)5390
-{5391
-#if defined(SIMDE_X86_SSE2_NATIVE)5392
- return _mm_sub_epi64(a, b);5393
-#else5394
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5395
- b_ = simde__m128i_to_private(b);5396
-5397
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5398
- r_.neon_i64 = vsubq_s64(a_.neon_i64, b_.neon_i64);5399
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)5400
- r_.i64 = a_.i64 - b_.i64;5401
-#else5402
- SIMDE_VECTORIZE5403
- for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {5404
- r_.i64[i] = a_.i64[i] - b_.i64[i];5405
- }5406
-#endif5407
-5408
- return simde__m128i_from_private(r_);5409
-#endif5410
-}5411
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5412
-#define _mm_sub_epi64(a, b) simde_mm_sub_epi64(a, b)5413
-#endif5414
-5415
-SIMDE_FUNCTION_ATTRIBUTES5416
-simde__m128i simde_x_mm_sub_epu32(simde__m128i a, simde__m128i b)5417
-{5418
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5419
- b_ = simde__m128i_to_private(b);5420
-5421
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)5422
- r_.u32 = a_.u32 - b_.u32;5423
-#else5424
- SIMDE_VECTORIZE5425
- for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {5426
- r_.u32[i] = a_.u32[i] - b_.u32[i];5427
- }5428
-#endif5429
-5430
- return simde__m128i_from_private(r_);5431
-}5432
-5433
-SIMDE_FUNCTION_ATTRIBUTES5434
-simde__m128d simde_mm_sub_pd(simde__m128d a, simde__m128d b)5435
-{5436
-#if defined(SIMDE_X86_SSE2_NATIVE)5437
- return _mm_sub_pd(a, b);5438
-#else5439
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),5440
- b_ = simde__m128d_to_private(b);5441
-5442
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)5443
- r_.f64 = a_.f64 - b_.f64;5444
-#elif defined(SIMDE_WASM_SIMD128_NATIVE)5445
- r_.wasm_v128 = wasm_f64x2_sub(a_.wasm_v128, b_.wasm_v128);5446
-#else5447
- SIMDE_VECTORIZE5448
- for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {5449
- r_.f64[i] = a_.f64[i] - b_.f64[i];5450
- }5451
-#endif5452
-5453
- return simde__m128d_from_private(r_);5454
-#endif5455
-}5456
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5457
-#define _mm_sub_pd(a, b) simde_mm_sub_pd(a, b)5458
-#endif5459
-5460
-SIMDE_FUNCTION_ATTRIBUTES5461
-simde__m128d simde_mm_sub_sd(simde__m128d a, simde__m128d b)5462
-{5463
-#if defined(SIMDE_X86_SSE2_NATIVE)5464
- return _mm_sub_sd(a, b);5465
-#elif defined(SIMDE_ASSUME_VECTORIZATION)5466
- return simde_mm_move_sd(a, simde_mm_sub_pd(a, b));5467
-#else5468
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),5469
- b_ = simde__m128d_to_private(b);5470
-5471
- r_.f64[0] = a_.f64[0] - b_.f64[0];5472
- r_.f64[1] = a_.f64[1];5473
-5474
- return simde__m128d_from_private(r_);5475
-#endif5476
-}5477
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5478
-#define _mm_sub_sd(a, b) simde_mm_sub_sd(a, b)5479
-#endif5480
-5481
-SIMDE_FUNCTION_ATTRIBUTES5482
-simde__m64 simde_mm_sub_si64(simde__m64 a, simde__m64 b)5483
-{5484
-#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)5485
- return _mm_sub_si64(a, b);5486
-#else5487
- simde__m64_private r_, a_ = simde__m64_to_private(a),5488
- b_ = simde__m64_to_private(b);5489
-5490
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)5491
- r_.i64 = a_.i64 - b_.i64;5492
-#else5493
- r_.i64[0] = a_.i64[0] - b_.i64[0];5494
-#endif5495
-5496
- return simde__m64_from_private(r_);5497
-#endif5498
-}5499
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5500
-#define _mm_sub_si64(a, b) simde_mm_sub_si64(a, b)5501
-#endif5502
-5503
-SIMDE_FUNCTION_ATTRIBUTES5504
-simde__m128i simde_mm_subs_epi8(simde__m128i a, simde__m128i b)5505
-{5506
-#if defined(SIMDE_X86_SSE2_NATIVE)5507
- return _mm_subs_epi8(a, b);5508
-#else5509
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5510
- b_ = simde__m128i_to_private(b);5511
-5512
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5513
- r_.neon_i8 = vqsubq_s8(a_.neon_i8, b_.neon_i8);5514
-#else5515
- SIMDE_VECTORIZE5516
- for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i8[0])); i++) {5517
- if (((b_.i8[i]) > 0 && (a_.i8[i]) < INT8_MIN + (b_.i8[i]))) {5518
- r_.i8[i] = INT8_MIN;5519
- } else if ((b_.i8[i]) < 0 &&5520
- (a_.i8[i]) > INT8_MAX + (b_.i8[i])) {5521
- r_.i8[i] = INT8_MAX;5522
- } else {5523
- r_.i8[i] = (a_.i8[i]) - (b_.i8[i]);5524
- }5525
- }5526
-#endif5527
-5528
- return simde__m128i_from_private(r_);5529
-#endif5530
-}5531
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5532
-#define _mm_subs_epi8(a, b) simde_mm_subs_epi8(a, b)5533
-#endif5534
-5535
-SIMDE_FUNCTION_ATTRIBUTES5536
-simde__m128i simde_mm_subs_epi16(simde__m128i a, simde__m128i b)5537
-{5538
-#if defined(SIMDE_X86_SSE2_NATIVE)5539
- return _mm_subs_epi16(a, b);5540
-#else5541
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5542
- b_ = simde__m128i_to_private(b);5543
-5544
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5545
- r_.neon_i16 = vqsubq_s16(a_.neon_i16, b_.neon_i16);5546
-#else5547
- SIMDE_VECTORIZE5548
- for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i16[0])); i++) {5549
- if (((b_.i16[i]) > 0 &&5550
- (a_.i16[i]) < INT16_MIN + (b_.i16[i]))) {5551
- r_.i16[i] = INT16_MIN;5552
- } else if ((b_.i16[i]) < 0 &&5553
- (a_.i16[i]) > INT16_MAX + (b_.i16[i])) {5554
- r_.i16[i] = INT16_MAX;5555
- } else {5556
- r_.i16[i] = (a_.i16[i]) - (b_.i16[i]);5557
- }5558
- }5559
-#endif5560
-5561
- return simde__m128i_from_private(r_);5562
-#endif5563
-}5564
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5565
-#define _mm_subs_epi16(a, b) simde_mm_subs_epi16(a, b)5566
-#endif5567
-5568
-SIMDE_FUNCTION_ATTRIBUTES5569
-simde__m128i simde_mm_subs_epu8(simde__m128i a, simde__m128i b)5570
-{5571
-#if defined(SIMDE_X86_SSE2_NATIVE)5572
- return _mm_subs_epu8(a, b);5573
-#else5574
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5575
- b_ = simde__m128i_to_private(b);5576
-5577
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5578
- r_.neon_u8 = vqsubq_u8(a_.neon_u8, b_.neon_u8);5579
-#else5580
- SIMDE_VECTORIZE5581
- for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i8[0])); i++) {5582
- const int32_t x = a_.u8[i] - b_.u8[i];5583
- if (x < 0) {5584
- r_.u8[i] = 0;5585
- } else if (x > UINT8_MAX) {5586
- r_.u8[i] = UINT8_MAX;5587
- } else {5588
- r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, x);5589
- }5590
- }5591
-#endif5592
-5593
- return simde__m128i_from_private(r_);5594
-#endif5595
-}5596
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5597
-#define _mm_subs_epu8(a, b) simde_mm_subs_epu8(a, b)5598
-#endif5599
-5600
-SIMDE_FUNCTION_ATTRIBUTES5601
-simde__m128i simde_mm_subs_epu16(simde__m128i a, simde__m128i b)5602
-{5603
-#if defined(SIMDE_X86_SSE2_NATIVE)5604
- return _mm_subs_epu16(a, b);5605
-#else5606
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5607
- b_ = simde__m128i_to_private(b);5608
-5609
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5610
- r_.neon_u16 = vqsubq_u16(a_.neon_u16, b_.neon_u16);5611
-#else5612
- SIMDE_VECTORIZE5613
- for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i16[0])); i++) {5614
- const int32_t x = a_.u16[i] - b_.u16[i];5615
- if (x < 0) {5616
- r_.u16[i] = 0;5617
- } else if (x > UINT16_MAX) {5618
- r_.u16[i] = UINT16_MAX;5619
- } else {5620
- r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, x);5621
- }5622
- }5623
-#endif5624
-5625
- return simde__m128i_from_private(r_);5626
-#endif5627
-}5628
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5629
-#define _mm_subs_epu16(a, b) simde_mm_subs_epu16(a, b)5630
-#endif5631
-5632
-SIMDE_FUNCTION_ATTRIBUTES5633
-int simde_mm_ucomieq_sd(simde__m128d a, simde__m128d b)5634
-{5635
-#if defined(SIMDE_X86_SSE2_NATIVE)5636
- return _mm_ucomieq_sd(a, b);5637
-#else5638
- simde__m128d_private a_ = simde__m128d_to_private(a),5639
- b_ = simde__m128d_to_private(b);5640
- int r;5641
-5642
-#if defined(SIMDE_HAVE_FENV_H)5643
- fenv_t envp;5644
- int x = feholdexcept(&envp);5645
- r = a_.f64[0] == b_.f64[0];5646
- if (HEDLEY_LIKELY(x == 0))5647
- fesetenv(&envp);5648
-#else5649
- r = a_.f64[0] == b_.f64[0];5650
-#endif5651
-5652
- return r;5653
-#endif5654
-}5655
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5656
-#define _mm_ucomieq_sd(a, b) simde_mm_ucomieq_sd(a, b)5657
-#endif5658
-5659
-SIMDE_FUNCTION_ATTRIBUTES5660
-int simde_mm_ucomige_sd(simde__m128d a, simde__m128d b)5661
-{5662
-#if defined(SIMDE_X86_SSE2_NATIVE)5663
- return _mm_ucomige_sd(a, b);5664
-#else5665
- simde__m128d_private a_ = simde__m128d_to_private(a),5666
- b_ = simde__m128d_to_private(b);5667
- int r;5668
-5669
-#if defined(SIMDE_HAVE_FENV_H)5670
- fenv_t envp;5671
- int x = feholdexcept(&envp);5672
- r = a_.f64[0] >= b_.f64[0];5673
- if (HEDLEY_LIKELY(x == 0))5674
- fesetenv(&envp);5675
-#else5676
- r = a_.f64[0] >= b_.f64[0];5677
-#endif5678
-5679
- return r;5680
-#endif5681
-}5682
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5683
-#define _mm_ucomige_sd(a, b) simde_mm_ucomige_sd(a, b)5684
-#endif5685
-5686
-SIMDE_FUNCTION_ATTRIBUTES5687
-int simde_mm_ucomigt_sd(simde__m128d a, simde__m128d b)5688
-{5689
-#if defined(SIMDE_X86_SSE2_NATIVE)5690
- return _mm_ucomigt_sd(a, b);5691
-#else5692
- simde__m128d_private a_ = simde__m128d_to_private(a),5693
- b_ = simde__m128d_to_private(b);5694
- int r;5695
-5696
-#if defined(SIMDE_HAVE_FENV_H)5697
- fenv_t envp;5698
- int x = feholdexcept(&envp);5699
- r = a_.f64[0] > b_.f64[0];5700
- if (HEDLEY_LIKELY(x == 0))5701
- fesetenv(&envp);5702
-#else5703
- r = a_.f64[0] > b_.f64[0];5704
-#endif5705
-5706
- return r;5707
-#endif5708
-}5709
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5710
-#define _mm_ucomigt_sd(a, b) simde_mm_ucomigt_sd(a, b)5711
-#endif5712
-5713
-SIMDE_FUNCTION_ATTRIBUTES5714
-int simde_mm_ucomile_sd(simde__m128d a, simde__m128d b)5715
-{5716
-#if defined(SIMDE_X86_SSE2_NATIVE)5717
- return _mm_ucomile_sd(a, b);5718
-#else5719
- simde__m128d_private a_ = simde__m128d_to_private(a),5720
- b_ = simde__m128d_to_private(b);5721
- int r;5722
-5723
-#if defined(SIMDE_HAVE_FENV_H)5724
- fenv_t envp;5725
- int x = feholdexcept(&envp);5726
- r = a_.f64[0] <= b_.f64[0];5727
- if (HEDLEY_LIKELY(x == 0))5728
- fesetenv(&envp);5729
-#else5730
- r = a_.f64[0] <= b_.f64[0];5731
-#endif5732
-5733
- return r;5734
-#endif5735
-}5736
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5737
-#define _mm_ucomile_sd(a, b) simde_mm_ucomile_sd(a, b)5738
-#endif5739
-5740
-SIMDE_FUNCTION_ATTRIBUTES5741
-int simde_mm_ucomilt_sd(simde__m128d a, simde__m128d b)5742
-{5743
-#if defined(SIMDE_X86_SSE2_NATIVE)5744
- return _mm_ucomilt_sd(a, b);5745
-#else5746
- simde__m128d_private a_ = simde__m128d_to_private(a),5747
- b_ = simde__m128d_to_private(b);5748
- int r;5749
-5750
-#if defined(SIMDE_HAVE_FENV_H)5751
- fenv_t envp;5752
- int x = feholdexcept(&envp);5753
- r = a_.f64[0] < b_.f64[0];5754
- if (HEDLEY_LIKELY(x == 0))5755
- fesetenv(&envp);5756
-#else5757
- r = a_.f64[0] < b_.f64[0];5758
-#endif5759
-5760
- return r;5761
-#endif5762
-}5763
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5764
-#define _mm_ucomilt_sd(a, b) simde_mm_ucomilt_sd(a, b)5765
-#endif5766
-5767
-SIMDE_FUNCTION_ATTRIBUTES5768
-int simde_mm_ucomineq_sd(simde__m128d a, simde__m128d b)5769
-{5770
-#if defined(SIMDE_X86_SSE2_NATIVE)5771
- return _mm_ucomineq_sd(a, b);5772
-#else5773
- simde__m128d_private a_ = simde__m128d_to_private(a),5774
- b_ = simde__m128d_to_private(b);5775
- int r;5776
-5777
-#if defined(SIMDE_HAVE_FENV_H)5778
- fenv_t envp;5779
- int x = feholdexcept(&envp);5780
- r = a_.f64[0] != b_.f64[0];5781
- if (HEDLEY_LIKELY(x == 0))5782
- fesetenv(&envp);5783
-#else5784
- r = a_.f64[0] != b_.f64[0];5785
-#endif5786
-5787
- return r;5788
-#endif5789
-}5790
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5791
-#define _mm_ucomineq_sd(a, b) simde_mm_ucomineq_sd(a, b)5792
-#endif5793
-5794
-#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)5795
-HEDLEY_DIAGNOSTIC_PUSH5796
-SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_5797
-#endif5798
-5799
-#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)5800
-HEDLEY_DIAGNOSTIC_POP5801
-#endif5802
-5803
-SIMDE_FUNCTION_ATTRIBUTES5804
-void simde_mm_lfence(void)5805
-{5806
-#if defined(SIMDE_X86_SSE2_NATIVE)5807
- _mm_lfence();5808
-#else5809
- simde_mm_sfence();5810
-#endif5811
-}5812
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5813
-#define _mm_lfence() simde_mm_lfence()5814
-#endif5815
-5816
-SIMDE_FUNCTION_ATTRIBUTES5817
-void simde_mm_mfence(void)5818
-{5819
-#if defined(SIMDE_X86_SSE2_NATIVE)5820
- _mm_mfence();5821
-#else5822
- simde_mm_sfence();5823
-#endif5824
-}5825
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5826
-#define _mm_mfence() simde_mm_mfence()5827
-#endif5828
-5829
-SIMDE_FUNCTION_ATTRIBUTES5830
-simde__m128i simde_mm_unpackhi_epi8(simde__m128i a, simde__m128i b)5831
-{5832
-#if defined(SIMDE_X86_SSE2_NATIVE)5833
- return _mm_unpackhi_epi8(a, b);5834
-#else5835
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5836
- b_ = simde__m128i_to_private(b);5837
-5838
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5839
- int8x8_t a1 = vreinterpret_s8_s16(vget_high_s16(a_.neon_i16));5840
- int8x8_t b1 = vreinterpret_s8_s16(vget_high_s16(b_.neon_i16));5841
- int8x8x2_t result = vzip_s8(a1, b1);5842
- r_.neon_i8 = vcombine_s8(result.val[0], result.val[1]);5843
-#elif defined(SIMDE_SHUFFLE_VECTOR_)5844
- r_.i8 = SIMDE_SHUFFLE_VECTOR_(8, 16, a_.i8, b_.i8, 8, 24, 9, 25, 10, 26,5845
- 11, 27, 12, 28, 13, 29, 14, 30, 15, 31);5846
-#else5847
- SIMDE_VECTORIZE5848
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i8[0])) / 2); i++) {5849
- r_.i8[(i * 2)] =5850
- a_.i8[i + ((sizeof(r_) / sizeof(r_.i8[0])) / 2)];5851
- r_.i8[(i * 2) + 1] =5852
- b_.i8[i + ((sizeof(r_) / sizeof(r_.i8[0])) / 2)];5853
- }5854
-#endif5855
-5856
- return simde__m128i_from_private(r_);5857
-#endif5858
-}5859
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5860
-#define _mm_unpackhi_epi8(a, b) simde_mm_unpackhi_epi8(a, b)5861
-#endif5862
-5863
-SIMDE_FUNCTION_ATTRIBUTES5864
-simde__m128i simde_mm_unpackhi_epi16(simde__m128i a, simde__m128i b)5865
-{5866
-#if defined(SIMDE_X86_SSE2_NATIVE)5867
- return _mm_unpackhi_epi16(a, b);5868
-#else5869
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5870
- b_ = simde__m128i_to_private(b);5871
-5872
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5873
- int16x4_t a1 = vget_high_s16(a_.neon_i16);5874
- int16x4_t b1 = vget_high_s16(b_.neon_i16);5875
- int16x4x2_t result = vzip_s16(a1, b1);5876
- r_.neon_i16 = vcombine_s16(result.val[0], result.val[1]);5877
-#elif defined(SIMDE_SHUFFLE_VECTOR_)5878
- r_.i16 = SIMDE_SHUFFLE_VECTOR_(16, 16, a_.i16, b_.i16, 4, 12, 5, 13, 6,5879
- 14, 7, 15);5880
-#else5881
- SIMDE_VECTORIZE5882
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i16[0])) / 2); i++) {5883
- r_.i16[(i * 2)] =5884
- a_.i16[i + ((sizeof(r_) / sizeof(r_.i16[0])) / 2)];5885
- r_.i16[(i * 2) + 1] =5886
- b_.i16[i + ((sizeof(r_) / sizeof(r_.i16[0])) / 2)];5887
- }5888
-#endif5889
-5890
- return simde__m128i_from_private(r_);5891
-#endif5892
-}5893
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5894
-#define _mm_unpackhi_epi16(a, b) simde_mm_unpackhi_epi16(a, b)5895
-#endif5896
-5897
-SIMDE_FUNCTION_ATTRIBUTES5898
-simde__m128i simde_mm_unpackhi_epi32(simde__m128i a, simde__m128i b)5899
-{5900
-#if defined(SIMDE_X86_SSE2_NATIVE)5901
- return _mm_unpackhi_epi32(a, b);5902
-#else5903
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5904
- b_ = simde__m128i_to_private(b);5905
-5906
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5907
- int32x2_t a1 = vget_high_s32(a_.neon_i32);5908
- int32x2_t b1 = vget_high_s32(b_.neon_i32);5909
- int32x2x2_t result = vzip_s32(a1, b1);5910
- r_.neon_i32 = vcombine_s32(result.val[0], result.val[1]);5911
-#elif defined(SIMDE_SHUFFLE_VECTOR_)5912
- r_.i32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.i32, b_.i32, 2, 6, 3, 7);5913
-#else5914
- SIMDE_VECTORIZE5915
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i32[0])) / 2); i++) {5916
- r_.i32[(i * 2)] =5917
- a_.i32[i + ((sizeof(r_) / sizeof(r_.i32[0])) / 2)];5918
- r_.i32[(i * 2) + 1] =5919
- b_.i32[i + ((sizeof(r_) / sizeof(r_.i32[0])) / 2)];5920
- }5921
-#endif5922
-5923
- return simde__m128i_from_private(r_);5924
-#endif5925
-}5926
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5927
-#define _mm_unpackhi_epi32(a, b) simde_mm_unpackhi_epi32(a, b)5928
-#endif5929
-5930
-SIMDE_FUNCTION_ATTRIBUTES5931
-simde__m128i simde_mm_unpackhi_epi64(simde__m128i a, simde__m128i b)5932
-{5933
-#if defined(SIMDE_X86_SSE2_NATIVE)5934
- return _mm_unpackhi_epi64(a, b);5935
-#else5936
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5937
- b_ = simde__m128i_to_private(b);5938
-5939
-#if defined(SIMDE_SHUFFLE_VECTOR_)5940
- r_.i64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.i64, b_.i64, 1, 3);5941
-#else5942
- SIMDE_VECTORIZE5943
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i64[0])) / 2); i++) {5944
- r_.i64[(i * 2)] =5945
- a_.i64[i + ((sizeof(r_) / sizeof(r_.i64[0])) / 2)];5946
- r_.i64[(i * 2) + 1] =5947
- b_.i64[i + ((sizeof(r_) / sizeof(r_.i64[0])) / 2)];5948
- }5949
-#endif5950
-5951
- return simde__m128i_from_private(r_);5952
-#endif5953
-}5954
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5955
-#define _mm_unpackhi_epi64(a, b) simde_mm_unpackhi_epi64(a, b)5956
-#endif5957
-5958
-SIMDE_FUNCTION_ATTRIBUTES5959
-simde__m128d simde_mm_unpackhi_pd(simde__m128d a, simde__m128d b)5960
-{5961
-#if defined(SIMDE_X86_SSE2_NATIVE)5962
- return _mm_unpackhi_pd(a, b);5963
-#else5964
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),5965
- b_ = simde__m128d_to_private(b);5966
-5967
-#if defined(SIMDE_SHUFFLE_VECTOR_)5968
- r_.f64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.f64, b_.f64, 1, 3);5969
-#else5970
- SIMDE_VECTORIZE5971
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.f64[0])) / 2); i++) {5972
- r_.f64[(i * 2)] =5973
- a_.f64[i + ((sizeof(r_) / sizeof(r_.f64[0])) / 2)];5974
- r_.f64[(i * 2) + 1] =5975
- b_.f64[i + ((sizeof(r_) / sizeof(r_.f64[0])) / 2)];5976
- }5977
-#endif5978
-5979
- return simde__m128d_from_private(r_);5980
-#endif5981
-}5982
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5983
-#define _mm_unpackhi_pd(a, b) simde_mm_unpackhi_pd(a, b)5984
-#endif5985
-5986
-SIMDE_FUNCTION_ATTRIBUTES5987
-simde__m128i simde_mm_unpacklo_epi8(simde__m128i a, simde__m128i b)5988
-{5989
-#if defined(SIMDE_X86_SSE2_NATIVE)5990
- return _mm_unpacklo_epi8(a, b);5991
-#else5992
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),5993
- b_ = simde__m128i_to_private(b);5994
-5995
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5996
- int8x8_t a1 = vreinterpret_s8_s16(vget_low_s16(a_.neon_i16));5997
- int8x8_t b1 = vreinterpret_s8_s16(vget_low_s16(b_.neon_i16));5998
- int8x8x2_t result = vzip_s8(a1, b1);5999
- r_.neon_i8 = vcombine_s8(result.val[0], result.val[1]);6000
-#elif defined(SIMDE_SHUFFLE_VECTOR_)6001
- r_.i8 = SIMDE_SHUFFLE_VECTOR_(8, 16, a_.i8, b_.i8, 0, 16, 1, 17, 2, 18,6002
- 3, 19, 4, 20, 5, 21, 6, 22, 7, 23);6003
-#else6004
- SIMDE_VECTORIZE6005
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i8[0])) / 2); i++) {6006
- r_.i8[(i * 2)] = a_.i8[i];6007
- r_.i8[(i * 2) + 1] = b_.i8[i];6008
- }6009
-#endif6010
-6011
- return simde__m128i_from_private(r_);6012
-#endif6013
-}6014
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6015
-#define _mm_unpacklo_epi8(a, b) simde_mm_unpacklo_epi8(a, b)6016
-#endif6017
-6018
-SIMDE_FUNCTION_ATTRIBUTES6019
-simde__m128i simde_mm_unpacklo_epi16(simde__m128i a, simde__m128i b)6020
-{6021
-#if defined(SIMDE_X86_SSE2_NATIVE)6022
- return _mm_unpacklo_epi16(a, b);6023
-#else6024
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),6025
- b_ = simde__m128i_to_private(b);6026
-6027
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6028
- int16x4_t a1 = vget_low_s16(a_.neon_i16);6029
- int16x4_t b1 = vget_low_s16(b_.neon_i16);6030
- int16x4x2_t result = vzip_s16(a1, b1);6031
- r_.neon_i16 = vcombine_s16(result.val[0], result.val[1]);6032
-#elif defined(SIMDE_SHUFFLE_VECTOR_)6033
- r_.i16 = SIMDE_SHUFFLE_VECTOR_(16, 16, a_.i16, b_.i16, 0, 8, 1, 9, 2,6034
- 10, 3, 11);6035
-#else6036
- SIMDE_VECTORIZE6037
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i16[0])) / 2); i++) {6038
- r_.i16[(i * 2)] = a_.i16[i];6039
- r_.i16[(i * 2) + 1] = b_.i16[i];6040
- }6041
-#endif6042
-6043
- return simde__m128i_from_private(r_);6044
-#endif6045
-}6046
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6047
-#define _mm_unpacklo_epi16(a, b) simde_mm_unpacklo_epi16(a, b)6048
-#endif6049
-6050
-SIMDE_FUNCTION_ATTRIBUTES6051
-simde__m128i simde_mm_unpacklo_epi32(simde__m128i a, simde__m128i b)6052
-{6053
-#if defined(SIMDE_X86_SSE2_NATIVE)6054
- return _mm_unpacklo_epi32(a, b);6055
-#else6056
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),6057
- b_ = simde__m128i_to_private(b);6058
-6059
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6060
- int32x2_t a1 = vget_low_s32(a_.neon_i32);6061
- int32x2_t b1 = vget_low_s32(b_.neon_i32);6062
- int32x2x2_t result = vzip_s32(a1, b1);6063
- r_.neon_i32 = vcombine_s32(result.val[0], result.val[1]);6064
-#elif defined(SIMDE_SHUFFLE_VECTOR_)6065
- r_.i32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.i32, b_.i32, 0, 4, 1, 5);6066
-#else6067
- SIMDE_VECTORIZE6068
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i32[0])) / 2); i++) {6069
- r_.i32[(i * 2)] = a_.i32[i];6070
- r_.i32[(i * 2) + 1] = b_.i32[i];6071
- }6072
-#endif6073
-6074
- return simde__m128i_from_private(r_);6075
-#endif6076
-}6077
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6078
-#define _mm_unpacklo_epi32(a, b) simde_mm_unpacklo_epi32(a, b)6079
-#endif6080
-6081
-SIMDE_FUNCTION_ATTRIBUTES6082
-simde__m128i simde_mm_unpacklo_epi64(simde__m128i a, simde__m128i b)6083
-{6084
-#if defined(SIMDE_X86_SSE2_NATIVE)6085
- return _mm_unpacklo_epi64(a, b);6086
-#else6087
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),6088
- b_ = simde__m128i_to_private(b);6089
-6090
-#if defined(SIMDE_SHUFFLE_VECTOR_)6091
- r_.i64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.i64, b_.i64, 0, 2);6092
-#else6093
- SIMDE_VECTORIZE6094
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i64[0])) / 2); i++) {6095
- r_.i64[(i * 2)] = a_.i64[i];6096
- r_.i64[(i * 2) + 1] = b_.i64[i];6097
- }6098
-#endif6099
-6100
- return simde__m128i_from_private(r_);6101
-#endif6102
-}6103
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6104
-#define _mm_unpacklo_epi64(a, b) simde_mm_unpacklo_epi64(a, b)6105
-#endif6106
-6107
-SIMDE_FUNCTION_ATTRIBUTES6108
-simde__m128d simde_mm_unpacklo_pd(simde__m128d a, simde__m128d b)6109
-{6110
-#if defined(SIMDE_X86_SSE2_NATIVE)6111
- return _mm_unpacklo_pd(a, b);6112
-#else6113
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),6114
- b_ = simde__m128d_to_private(b);6115
-6116
-#if defined(SIMDE_SHUFFLE_VECTOR_)6117
- r_.f64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.f64, b_.f64, 0, 2);6118
-#else6119
- SIMDE_VECTORIZE6120
- for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.f64[0])) / 2); i++) {6121
- r_.f64[(i * 2)] = a_.f64[i];6122
- r_.f64[(i * 2) + 1] = b_.f64[i];6123
- }6124
-#endif6125
-6126
- return simde__m128d_from_private(r_);6127
-#endif6128
-}6129
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6130
-#define _mm_unpacklo_pd(a, b) simde_mm_unpacklo_pd(a, b)6131
-#endif6132
-6133
-SIMDE_FUNCTION_ATTRIBUTES6134
-simde__m128d simde_mm_xor_pd(simde__m128d a, simde__m128d b)6135
-{6136
-#if defined(SIMDE_X86_SSE2_NATIVE)6137
- return _mm_xor_pd(a, b);6138
-#else6139
- simde__m128d_private r_, a_ = simde__m128d_to_private(a),6140
- b_ = simde__m128d_to_private(b);6141
-6142
-#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6143
- r_.i32f = a_.i32f ^ b_.i32f;6144
-#else6145
- SIMDE_VECTORIZE6146
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {6147
- r_.i32f[i] = a_.i32f[i] ^ b_.i32f[i];6148
- }6149
-#endif6150
-6151
- return simde__m128d_from_private(r_);6152
-#endif6153
-}6154
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6155
-#define _mm_xor_pd(a, b) simde_mm_xor_pd(a, b)6156
-#endif6157
-6158
-SIMDE_FUNCTION_ATTRIBUTES6159
-simde__m128i simde_mm_xor_si128(simde__m128i a, simde__m128i b)6160
-{6161
-#if defined(SIMDE_X86_SSE2_NATIVE)6162
- return _mm_xor_si128(a, b);6163
-#else6164
- simde__m128i_private r_, a_ = simde__m128i_to_private(a),6165
- b_ = simde__m128i_to_private(b);6166
-6167
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6168
- r_.neon_i32 = veorq_s32(a_.neon_i32, b_.neon_i32);6169
-#elif defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)6170
- r_.altivec_i32 = vec_xor(a_.altivec_i32, b_.altivec_i32);6171
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6172
- r_.i32f = a_.i32f ^ b_.i32f;6173
-#else6174
- SIMDE_VECTORIZE6175
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {6176
- r_.i32f[i] = a_.i32f[i] ^ b_.i32f[i];6177
- }6178
-#endif6179
-6180
- return simde__m128i_from_private(r_);6181
-#endif6182
-}6183
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6184
-#define _mm_xor_si128(a, b) simde_mm_xor_si128(a, b)6185
-#endif6186
-6187
-SIMDE_FUNCTION_ATTRIBUTES6188
-simde__m128i simde_x_mm_not_si128(simde__m128i a)6189
-{6190
- simde__m128i_private r_, a_ = simde__m128i_to_private(a);6191
-6192
-#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6193
- r_.neon_i32 = vmvnq_s32(a_.neon_i32);6194
-#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6195
- r_.i32f = ~(a_.i32f);6196
-#else6197
- SIMDE_VECTORIZE6198
- for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {6199
- r_.i32f[i] = ~(a_.i32f[i]);6200
- }6201
-#endif6202
-6203
- return simde__m128i_from_private(r_);6204
-}6205
-6206
-#define SIMDE_MM_SHUFFLE2(x, y) (((x) << 1) | (y))6207
-#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6208
-#define _MM_SHUFFLE2(x, y) SIMDE_MM_SHUFFLE2(x, y)6209
-#endif6210
-6211
-SIMDE_END_DECLS_6212
-6213
-HEDLEY_DIAGNOSTIC_POP6214
-6215
-#endif /* !defined(SIMDE_X86_SSE2_H) */6216
obs-studio-26.1.0.tar.xz/libobs/util/sse2neon.h
Deleted
4209
1
2
-#ifndef SSE2NEON_H3
-#define SSE2NEON_H4
-5
-// This header file provides a simple API translation layer6
-// between SSE intrinsics to their corresponding Arm/Aarch64 NEON versions7
-//8
-// This header file does not yet translate all of the SSE intrinsics.9
-//10
-// Contributors to this work are:11
-// John W. Ratcliff <jratcliffscarab@gmail.com>12
-// Brandon Rowlett <browlett@nvidia.com>13
-// Ken Fast <kfast@gdeb.com>14
-// Eric van Beurden <evanbeurden@nvidia.com>15
-// Alexander Potylitsin <apotylitsin@nvidia.com>16
-// Hasindu Gamaarachchi <hasindu2008@gmail.com>17
-// Jim Huang <jserv@biilabs.io>18
-// Mark Cheng <marktwtn@biilabs.io>19
-// Malcolm James MacLeod <malcolm@gulden.com>20
-// Devin Hussey (easyaspi314) <husseydevin@gmail.com>21
-// Sebastian Pop <spop@amazon.com>22
-// Developer Ecosystem Engineering <DeveloperEcosystemEngineering@apple.com>23
-// Danila Kutenin <danilak@google.com>24
-25
-/*26
- * sse2neon is freely redistributable under the MIT License.27
- *28
- * Permission is hereby granted, free of charge, to any person obtaining a copy29
- * of this software and associated documentation files (the "Software"), to deal30
- * in the Software without restriction, including without limitation the rights31
- * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell32
- * copies of the Software, and to permit persons to whom the Software is33
- * furnished to do so, subject to the following conditions:34
- *35
- * The above copyright notice and this permission notice shall be included in36
- * all copies or substantial portions of the Software.37
- *38
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR39
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,40
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE41
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER42
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,43
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE44
- * SOFTWARE.45
- */46
-47
-#if defined(__GNUC__) || defined(__clang__)48
-#pragma push_macro("FORCE_INLINE")49
-#pragma push_macro("ALIGN_STRUCT")50
-#define FORCE_INLINE static inline __attribute__((always_inline))51
-#define ALIGN_STRUCT(x) __attribute__((aligned(x)))52
-#else53
-#error "Macro name collisions may happen with unsupported compiler."54
-#ifdef FORCE_INLINE55
-#undef FORCE_INLINE56
-#endif57
-#define FORCE_INLINE static inline58
-#ifndef ALIGN_STRUCT59
-#define ALIGN_STRUCT(x) __declspec(align(x))60
-#endif61
-#endif62
-63
-#include <stdint.h>64
-#include <stdlib.h>65
-66
-#include <arm_neon.h>67
-68
-/* "__has_builtin" can be used to query support for built-in functions69
- * provided by gcc/clang and other compilers that support it.70
- */71
-#ifndef __has_builtin /* GCC prior to 10 or non-clang compilers */72
-/* Compatibility with gcc <= 9 */73
-#if __GNUC__ <= 974
-#define __has_builtin(x) HAS##x75
-#define HAS__builtin_popcount 176
-#define HAS__builtin_popcountll 177
-#else78
-#define __has_builtin(x) 079
-#endif80
-#endif81
-82
-/**83
- * MACRO for shuffle parameter for _mm_shuffle_ps().84
- * Argument fp3 is a digit[0123] that represents the fp from argument "b"85
- * of mm_shuffle_ps that will be placed in fp3 of result. fp2 is the same86
- * for fp2 in result. fp1 is a digit[0123] that represents the fp from87
- * argument "a" of mm_shuffle_ps that will be places in fp1 of result.88
- * fp0 is the same for fp0 of result.89
- */90
-#define _MM_SHUFFLE(fp3, fp2, fp1, fp0) \91
- (((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | ((fp0)))92
-93
-/* indicate immediate constant argument in a given range */94
-#define __constrange(a, b) const95
-96
-/* A few intrinsics accept traditional data types like ints or floats, but97
- * most operate on data types that are specific to SSE.98
- * If a vector type ends in d, it contains doubles, and if it does not have99
- * a suffix, it contains floats. An integer vector type can contain any type100
- * of integer, from chars to shorts to unsigned long longs.101
- */102
-typedef float32x2_t __m64;103
-typedef float32x4_t __m128; /* 128-bit vector containing 4 floats */104
-// On ARM 32-bit architecture, the float64x2_t is not supported.105
-// The data type __m128d should be represented in a different way for related106
-// intrinsic conversion.107
-#if defined(__aarch64__)108
-typedef float64x2_t __m128d; /* 128-bit vector containing 2 doubles */109
-#else110
-typedef float32x4_t __m128d;111
-#endif112
-typedef int64x1_t __m64i;113
-typedef int64x2_t __m128i; /* 128-bit vector containing integers */114
-115
-/* type-safe casting between types */116
-117
-#define vreinterpretq_m128_f16(x) vreinterpretq_f32_f16(x)118
-#define vreinterpretq_m128_f32(x) (x)119
-#define vreinterpretq_m128_f64(x) vreinterpretq_f32_f64(x)120
-121
-#define vreinterpretq_m128_u8(x) vreinterpretq_f32_u8(x)122
-#define vreinterpretq_m128_u16(x) vreinterpretq_f32_u16(x)123
-#define vreinterpretq_m128_u32(x) vreinterpretq_f32_u32(x)124
-#define vreinterpretq_m128_u64(x) vreinterpretq_f32_u64(x)125
-126
-#define vreinterpretq_m128_s8(x) vreinterpretq_f32_s8(x)127
-#define vreinterpretq_m128_s16(x) vreinterpretq_f32_s16(x)128
-#define vreinterpretq_m128_s32(x) vreinterpretq_f32_s32(x)129
-#define vreinterpretq_m128_s64(x) vreinterpretq_f32_s64(x)130
-131
-#define vreinterpretq_f16_m128(x) vreinterpretq_f16_f32(x)132
-#define vreinterpretq_f32_m128(x) (x)133
-#define vreinterpretq_f64_m128(x) vreinterpretq_f64_f32(x)134
-135
-#define vreinterpretq_u8_m128(x) vreinterpretq_u8_f32(x)136
-#define vreinterpretq_u16_m128(x) vreinterpretq_u16_f32(x)137
-#define vreinterpretq_u32_m128(x) vreinterpretq_u32_f32(x)138
-#define vreinterpretq_u64_m128(x) vreinterpretq_u64_f32(x)139
-140
-#define vreinterpretq_s8_m128(x) vreinterpretq_s8_f32(x)141
-#define vreinterpretq_s16_m128(x) vreinterpretq_s16_f32(x)142
-#define vreinterpretq_s32_m128(x) vreinterpretq_s32_f32(x)143
-#define vreinterpretq_s64_m128(x) vreinterpretq_s64_f32(x)144
-145
-#define vreinterpretq_m128i_s8(x) vreinterpretq_s64_s8(x)146
-#define vreinterpretq_m128i_s16(x) vreinterpretq_s64_s16(x)147
-#define vreinterpretq_m128i_s32(x) vreinterpretq_s64_s32(x)148
-#define vreinterpretq_m128i_s64(x) (x)149
-150
-#define vreinterpretq_m128i_u8(x) vreinterpretq_s64_u8(x)151
-#define vreinterpretq_m128i_u16(x) vreinterpretq_s64_u16(x)152
-#define vreinterpretq_m128i_u32(x) vreinterpretq_s64_u32(x)153
-#define vreinterpretq_m128i_u64(x) vreinterpretq_s64_u64(x)154
-155
-#define vreinterpretq_s8_m128i(x) vreinterpretq_s8_s64(x)156
-#define vreinterpretq_s16_m128i(x) vreinterpretq_s16_s64(x)157
-#define vreinterpretq_s32_m128i(x) vreinterpretq_s32_s64(x)158
-#define vreinterpretq_s64_m128i(x) (x)159
-160
-#define vreinterpretq_u8_m128i(x) vreinterpretq_u8_s64(x)161
-#define vreinterpretq_u16_m128i(x) vreinterpretq_u16_s64(x)162
-#define vreinterpretq_u32_m128i(x) vreinterpretq_u32_s64(x)163
-#define vreinterpretq_u64_m128i(x) vreinterpretq_u64_s64(x)164
-165
-#define vreinterpret_m64i_s8(x) vreinterpret_s64_s8(x)166
-#define vreinterpret_m64i_s16(x) vreinterpret_s64_s16(x)167
-#define vreinterpret_m64i_s32(x) vreinterpret_s64_s32(x)168
-#define vreinterpret_m64i_s64(x) (x)169
-170
-#define vreinterpret_m64i_u8(x) vreinterpret_s64_u8(x)171
-#define vreinterpret_m64i_u16(x) vreinterpret_s64_u16(x)172
-#define vreinterpret_m64i_u32(x) vreinterpret_s64_u32(x)173
-#define vreinterpret_m64i_u64(x) vreinterpret_s64_u64(x)174
-175
-#define vreinterpret_u8_m64i(x) vreinterpret_u8_s64(x)176
-#define vreinterpret_u16_m64i(x) vreinterpret_u16_s64(x)177
-#define vreinterpret_u32_m64i(x) vreinterpret_u32_s64(x)178
-#define vreinterpret_u64_m64i(x) vreinterpret_u64_s64(x)179
-180
-#define vreinterpret_s8_m64i(x) vreinterpret_s8_s64(x)181
-#define vreinterpret_s16_m64i(x) vreinterpret_s16_s64(x)182
-#define vreinterpret_s32_m64i(x) vreinterpret_s32_s64(x)183
-#define vreinterpret_s64_m64i(x) (x)184
-185
-// A struct is defined in this header file called 'SIMDVec' which can be used186
-// by applications which attempt to access the contents of an _m128 struct187
-// directly. It is important to note that accessing the __m128 struct directly188
-// is bad coding practice by Microsoft: @see:189
-// https://msdn.microsoft.com/en-us/library/ayeb3ayc.aspx190
-//191
-// However, some legacy source code may try to access the contents of an __m128192
-// struct directly so the developer can use the SIMDVec as an alias for it. Any193
-// casting must be done manually by the developer, as you cannot cast or194
-// otherwise alias the base NEON data type for intrinsic operations.195
-//196
-// union intended to allow direct access to an __m128 variable using the names197
-// that the MSVC compiler provides. This union should really only be used when198
-// trying to access the members of the vector as integer values. GCC/clang199
-// allow native access to the float members through a simple array access200
-// operator (in C since 4.6, in C++ since 4.8).201
-//202
-// Ideally direct accesses to SIMD vectors should not be used since it can cause203
-// a performance hit. If it really is needed however, the original __m128204
-// variable can be aliased with a pointer to this union and used to access205
-// individual components. The use of this union should be hidden behind a macro206
-// that is used throughout the codebase to access the members instead of always207
-// declaring this type of variable.208
-typedef union ALIGN_STRUCT(16) SIMDVec {209
- float m128_f32[4]; // as floats - DON'T USE. Added for convenience.210
- int8_t m128_i8[16]; // as signed 8-bit integers.211
- int16_t m128_i16[8]; // as signed 16-bit integers.212
- int32_t m128_i32[4]; // as signed 32-bit integers.213
- int64_t m128_i64[2]; // as signed 64-bit integers.214
- uint8_t m128_u8[16]; // as unsigned 8-bit integers.215
- uint16_t m128_u16[8]; // as unsigned 16-bit integers.216
- uint32_t m128_u32[4]; // as unsigned 32-bit integers.217
- uint64_t m128_u64[2]; // as unsigned 64-bit integers.218
-} SIMDVec;219
-220
-// casting using SIMDVec221
-#define vreinterpretq_nth_u64_m128i(x, n) (((SIMDVec *)&x)->m128_u64[n])222
-#define vreinterpretq_nth_u32_m128i(x, n) (((SIMDVec *)&x)->m128_u32[n])223
-224
-/* Backwards compatibility for compilers with lack of specific type support */225
-226
-// Older gcc does not define vld1q_u8_x4 type227
-#if defined(__GNUC__) && !defined(__clang__)228
-#if __GNUC__ <= 9229
-FORCE_INLINE uint8x16x4_t vld1q_u8_x4(const uint8_t *p)230
-{231
- uint8x16x4_t ret;232
- ret.val[0] = vld1q_u8(p + 0);233
- ret.val[1] = vld1q_u8(p + 16);234
- ret.val[2] = vld1q_u8(p + 32);235
- ret.val[3] = vld1q_u8(p + 48);236
- return ret;237
-}238
-#endif239
-#endif240
-241
-/* Function Naming Conventions242
- * The naming convention of SSE intrinsics is straightforward. A generic SSE243
- * intrinsic function is given as follows:244
- * _mm_<name>_<data_type>245
- *246
- * The parts of this format are given as follows:247
- * 1. <name> describes the operation performed by the intrinsic248
- * 2. <data_type> identifies the data type of the function's primary arguments249
- *250
- * This last part, <data_type>, is a little complicated. It identifies the251
- * content of the input values, and can be set to any of the following values:252
- * + ps - vectors contain floats (ps stands for packed single-precision)253
- * + pd - vectors cantain doubles (pd stands for packed double-precision)254
- * + epi8/epi16/epi32/epi64 - vectors contain 8-bit/16-bit/32-bit/64-bit255
- * signed integers256
- * + epu8/epu16/epu32/epu64 - vectors contain 8-bit/16-bit/32-bit/64-bit257
- * unsigned integers258
- * + si128 - unspecified 128-bit vector or 256-bit vector259
- * + m128/m128i/m128d - identifies input vector types when they are different260
- * than the type of the returned vector261
- *262
- * For example, _mm_setzero_ps. The _mm implies that the function returns263
- * a 128-bit vector. The _ps at the end implies that the argument vectors264
- * contain floats.265
- *266
- * A complete example: Byte Shuffle - pshufb (_mm_shuffle_epi8)267
- * // Set packed 16-bit integers. 128 bits, 8 short, per 16 bits268
- * __m128i v_in = _mm_setr_epi16(1, 2, 3, 4, 5, 6, 7, 8);269
- * // Set packed 8-bit integers270
- * // 128 bits, 16 chars, per 8 bits271
- * __m128i v_perm = _mm_setr_epi8(1, 0, 2, 3, 8, 9, 10, 11,272
- * 4, 5, 12, 13, 6, 7, 14, 15);273
- * // Shuffle packed 8-bit integers274
- * __m128i v_out = _mm_shuffle_epi8(v_in, v_perm); // pshufb275
- *276
- * Data (Number, Binary, Byte Index):277
- +------+------+-------------+------+------+-------------+278
- | 1 | 2 | 3 | 4 | Number279
- +------+------+------+------+------+------+------+------+280
- | 0000 | 0001 | 0000 | 0010 | 0000 | 0011 | 0000 | 0100 | Binary281
- +------+------+------+------+------+------+------+------+282
- | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | Index283
- +------+------+------+------+------+------+------+------+284
-285
- +------+------+------+------+------+------+------+------+286
- | 5 | 6 | 7 | 8 | Number287
- +------+------+------+------+------+------+------+------+288
- | 0000 | 0101 | 0000 | 0110 | 0000 | 0111 | 0000 | 1000 | Binary289
- +------+------+------+------+------+------+------+------+290
- | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | Index291
- +------+------+------+------+------+------+------+------+292
- * Index (Byte Index):293
- +------+------+------+------+------+------+------+------+294
- | 1 | 0 | 2 | 3 | 8 | 9 | 10 | 11 |295
- +------+------+------+------+------+------+------+------+296
-297
- +------+------+------+------+------+------+------+------+298
- | 4 | 5 | 12 | 13 | 6 | 7 | 14 | 15 |299
- +------+------+------+------+------+------+------+------+300
- * Result:301
- +------+------+------+------+------+------+------+------+302
- | 1 | 0 | 2 | 3 | 8 | 9 | 10 | 11 | Index303
- +------+------+------+------+------+------+------+------+304
- | 0001 | 0000 | 0000 | 0010 | 0000 | 0101 | 0000 | 0110 | Binary305
- +------+------+------+------+------+------+------+------+306
- | 256 | 2 | 5 | 6 | Number307
- +------+------+------+------+------+------+------+------+308
-309
- +------+------+------+------+------+------+------+------+310
- | 4 | 5 | 12 | 13 | 6 | 7 | 14 | 15 | Index311
- +------+------+------+------+------+------+------+------+312
- | 0000 | 0011 | 0000 | 0111 | 0000 | 0100 | 0000 | 1000 | Binary313
- +------+------+------+------+------+------+------+------+314
- | 3 | 7 | 4 | 8 | Number315
- +------+------+------+------+------+------+-------------+316
- */317
-318
-/* Set/get methods */319
-320
-/* Constants for use with _mm_prefetch. */321
-enum _mm_hint {322
- _MM_HINT_NTA = 0, /* load data to L1 and L2 cache, mark it as NTA */323
- _MM_HINT_T0 = 1, /* load data to L1 and L2 cache */324
- _MM_HINT_T1 = 2, /* load data to L2 cache only */325
- _MM_HINT_T2 = 3, /* load data to L2 cache only, mark it as NTA */326
- _MM_HINT_ENTA = 4, /* exclusive version of _MM_HINT_NTA */327
- _MM_HINT_ET0 = 5, /* exclusive version of _MM_HINT_T0 */328
- _MM_HINT_ET1 = 6, /* exclusive version of _MM_HINT_T1 */329
- _MM_HINT_ET2 = 7 /* exclusive version of _MM_HINT_T2 */330
-};331
-332
-// Loads one cache line of data from address p to a location closer to the333
-// processor. https://msdn.microsoft.com/en-us/library/84szxsww(v=vs.100).aspx334
-FORCE_INLINE void _mm_prefetch(const void *p, int i)335
-{336
- (void)i;337
- __builtin_prefetch(p);338
-}339
-340
-// extracts the lower order floating point value from the parameter :341
-// https://msdn.microsoft.com/en-us/library/bb514059%28v=vs.120%29.aspx?f=255&MSPPError=-2147217396342
-FORCE_INLINE float _mm_cvtss_f32(__m128 a)343
-{344
- return vgetq_lane_f32(vreinterpretq_f32_m128(a), 0);345
-}346
-347
-// Sets the 128-bit value to zero348
-// https://msdn.microsoft.com/en-us/library/vstudio/ys7dw0kh(v=vs.100).aspx349
-FORCE_INLINE __m128i _mm_setzero_si128(void)350
-{351
- return vreinterpretq_m128i_s32(vdupq_n_s32(0));352
-}353
-354
-// Clears the four single-precision, floating-point values.355
-// https://msdn.microsoft.com/en-us/library/vstudio/tk1t2tbz(v=vs.100).aspx356
-FORCE_INLINE __m128 _mm_setzero_ps(void)357
-{358
- return vreinterpretq_m128_f32(vdupq_n_f32(0));359
-}360
-361
-// Sets the four single-precision, floating-point values to w.362
-//363
-// r0 := r1 := r2 := r3 := w364
-//365
-// https://msdn.microsoft.com/en-us/library/vstudio/2x1se8ha(v=vs.100).aspx366
-FORCE_INLINE __m128 _mm_set1_ps(float _w)367
-{368
- return vreinterpretq_m128_f32(vdupq_n_f32(_w));369
-}370
-371
-// Sets the four single-precision, floating-point values to w.372
-// https://msdn.microsoft.com/en-us/library/vstudio/2x1se8ha(v=vs.100).aspx373
-FORCE_INLINE __m128 _mm_set_ps1(float _w)374
-{375
- return vreinterpretq_m128_f32(vdupq_n_f32(_w));376
-}377
-378
-// Sets the four single-precision, floating-point values to the four inputs.379
-// https://msdn.microsoft.com/en-us/library/vstudio/afh0zf75(v=vs.100).aspx380
-FORCE_INLINE __m128 _mm_set_ps(float w, float z, float y, float x)381
-{382
- float ALIGN_STRUCT(16) data[4] = {x, y, z, w};383
- return vreinterpretq_m128_f32(vld1q_f32(data));384
-}385
-386
-// Copy single-precision (32-bit) floating-point element a to the lower element387
-// of dst, and zero the upper 3 elements.388
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_set_ss&expand=4901,4895,4901389
-FORCE_INLINE __m128 _mm_set_ss(float a)390
-{391
- float ALIGN_STRUCT(16) data[4] = {a, 0, 0, 0};392
- return vreinterpretq_m128_f32(vld1q_f32(data));393
-}394
-395
-// Sets the four single-precision, floating-point values to the four inputs in396
-// reverse order.397
-// https://msdn.microsoft.com/en-us/library/vstudio/d2172ct3(v=vs.100).aspx398
-FORCE_INLINE __m128 _mm_setr_ps(float w, float z, float y, float x)399
-{400
- float ALIGN_STRUCT(16) data[4] = {w, z, y, x};401
- return vreinterpretq_m128_f32(vld1q_f32(data));402
-}403
-404
-// Sets the 8 signed 16-bit integer values in reverse order.405
-//406
-// Return Value407
-// r0 := w0408
-// r1 := w1409
-// ...410
-// r7 := w7411
-FORCE_INLINE __m128i _mm_setr_epi16(short w0, short w1, short w2, short w3,412
- short w4, short w5, short w6, short w7)413
-{414
- int16_t ALIGN_STRUCT(16) data[8] = {w0, w1, w2, w3, w4, w5, w6, w7};415
- return vreinterpretq_m128i_s16(vld1q_s16((int16_t *)data));416
-}417
-418
-// Sets the 4 signed 32-bit integer values in reverse order419
-// https://technet.microsoft.com/en-us/library/security/27yb3ee5(v=vs.90).aspx420
-FORCE_INLINE __m128i _mm_setr_epi32(int i3, int i2, int i1, int i0)421
-{422
- int32_t ALIGN_STRUCT(16) data[4] = {i3, i2, i1, i0};423
- return vreinterpretq_m128i_s32(vld1q_s32(data));424
-}425
-426
-// Sets the 16 signed 8-bit integer values to b.427
-//428
-// r0 := b429
-// r1 := b430
-// ...431
-// r15 := b432
-//433
-// https://msdn.microsoft.com/en-us/library/6e14xhyf(v=vs.100).aspx434
-FORCE_INLINE __m128i _mm_set1_epi8(signed char w)435
-{436
- return vreinterpretq_m128i_s8(vdupq_n_s8(w));437
-}438
-439
-// Sets the 8 signed 16-bit integer values to w.440
-//441
-// r0 := w442
-// r1 := w443
-// ...444
-// r7 := w445
-//446
-// https://msdn.microsoft.com/en-us/library/k0ya3x0e(v=vs.90).aspx447
-FORCE_INLINE __m128i _mm_set1_epi16(short w)448
-{449
- return vreinterpretq_m128i_s16(vdupq_n_s16(w));450
-}451
-452
-// Sets the 16 signed 8-bit integer values.453
-// https://msdn.microsoft.com/en-us/library/x0cx8zd3(v=vs.90).aspx454
-FORCE_INLINE __m128i455
-_mm_set_epi8(signed char b15, signed char b14, signed char b13, signed char b12,456
- signed char b11, signed char b10, signed char b9, signed char b8,457
- signed char b7, signed char b6, signed char b5, signed char b4,458
- signed char b3, signed char b2, signed char b1, signed char b0)459
-{460
- int8_t ALIGN_STRUCT(16)461
- data[16] = {(int8_t)b0, (int8_t)b1, (int8_t)b2, (int8_t)b3,462
- (int8_t)b4, (int8_t)b5, (int8_t)b6, (int8_t)b7,463
- (int8_t)b8, (int8_t)b9, (int8_t)b10, (int8_t)b11,464
- (int8_t)b12, (int8_t)b13, (int8_t)b14, (int8_t)b15};465
- return (__m128i)vld1q_s8(data);466
-}467
-468
-// Sets the 8 signed 16-bit integer values.469
-// https://msdn.microsoft.com/en-au/library/3e0fek84(v=vs.90).aspx470
-FORCE_INLINE __m128i _mm_set_epi16(short i7, short i6, short i5, short i4,471
- short i3, short i2, short i1, short i0)472
-{473
- int16_t ALIGN_STRUCT(16) data[8] = {i0, i1, i2, i3, i4, i5, i6, i7};474
- return vreinterpretq_m128i_s16(vld1q_s16(data));475
-}476
-477
-// Sets the 16 signed 8-bit integer values in reverse order.478
-// https://msdn.microsoft.com/en-us/library/2khb9c7k(v=vs.90).aspx479
-FORCE_INLINE __m128i _mm_setr_epi8(480
- signed char b0, signed char b1, signed char b2, signed char b3,481
- signed char b4, signed char b5, signed char b6, signed char b7,482
- signed char b8, signed char b9, signed char b10, signed char b11,483
- signed char b12, signed char b13, signed char b14, signed char b15)484
-{485
- int8_t ALIGN_STRUCT(16)486
- data[16] = {(int8_t)b0, (int8_t)b1, (int8_t)b2, (int8_t)b3,487
- (int8_t)b4, (int8_t)b5, (int8_t)b6, (int8_t)b7,488
- (int8_t)b8, (int8_t)b9, (int8_t)b10, (int8_t)b11,489
- (int8_t)b12, (int8_t)b13, (int8_t)b14, (int8_t)b15};490
- return (__m128i)vld1q_s8(data);491
-}492
-493
-// Sets the 4 signed 32-bit integer values to i.494
-//495
-// r0 := i496
-// r1 := i497
-// r2 := i498
-// r3 := I499
-//500
-// https://msdn.microsoft.com/en-us/library/vstudio/h4xscxat(v=vs.100).aspx501
-FORCE_INLINE __m128i _mm_set1_epi32(int _i)502
-{503
- return vreinterpretq_m128i_s32(vdupq_n_s32(_i));504
-}505
-506
-// Sets the 2 signed 64-bit integer values to i.507
-// https://docs.microsoft.com/en-us/previous-versions/visualstudio/visual-studio-2010/whtfzhzk(v=vs.100)508
-FORCE_INLINE __m128i _mm_set1_epi64(int64_t _i)509
-{510
- return vreinterpretq_m128i_s64(vdupq_n_s64(_i));511
-}512
-513
-// Sets the 2 signed 64-bit integer values to i.514
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_set1_epi64x&expand=4961515
-FORCE_INLINE __m128i _mm_set1_epi64x(int64_t _i)516
-{517
- return vreinterpretq_m128i_s64(vdupq_n_s64(_i));518
-}519
-520
-// Sets the 4 signed 32-bit integer values.521
-// https://msdn.microsoft.com/en-us/library/vstudio/019beekt(v=vs.100).aspx522
-FORCE_INLINE __m128i _mm_set_epi32(int i3, int i2, int i1, int i0)523
-{524
- int32_t ALIGN_STRUCT(16) data[4] = {i0, i1, i2, i3};525
- return vreinterpretq_m128i_s32(vld1q_s32(data));526
-}527
-528
-// Returns the __m128i structure with its two 64-bit integer values529
-// initialized to the values of the two 64-bit integers passed in.530
-// https://msdn.microsoft.com/en-us/library/dk2sdw0h(v=vs.120).aspx531
-FORCE_INLINE __m128i _mm_set_epi64x(int64_t i1, int64_t i2)532
-{533
- int64_t ALIGN_STRUCT(16) data[2] = {i2, i1};534
- return vreinterpretq_m128i_s64(vld1q_s64(data));535
-}536
-537
-// Stores four single-precision, floating-point values.538
-// https://msdn.microsoft.com/en-us/library/vstudio/s3h4ay6y(v=vs.100).aspx539
-FORCE_INLINE void _mm_store_ps(float *p, __m128 a)540
-{541
- vst1q_f32(p, vreinterpretq_f32_m128(a));542
-}543
-544
-// Stores four single-precision, floating-point values.545
-// https://msdn.microsoft.com/en-us/library/44e30x22(v=vs.100).aspx546
-FORCE_INLINE void _mm_storeu_ps(float *p, __m128 a)547
-{548
- vst1q_f32(p, vreinterpretq_f32_m128(a));549
-}550
-551
-// Stores four 32-bit integer values as (as a __m128i value) at the address p.552
-// https://msdn.microsoft.com/en-us/library/vstudio/edk11s13(v=vs.100).aspx553
-FORCE_INLINE void _mm_store_si128(__m128i *p, __m128i a)554
-{555
- vst1q_s32((int32_t *)p, vreinterpretq_s32_m128i(a));556
-}557
-558
-// Stores four 32-bit integer values as (as a __m128i value) at the address p.559
-// https://msdn.microsoft.com/en-us/library/vstudio/edk11s13(v=vs.100).aspx560
-FORCE_INLINE void _mm_storeu_si128(__m128i *p, __m128i a)561
-{562
- vst1q_s32((int32_t *)p, vreinterpretq_s32_m128i(a));563
-}564
-565
-// Stores the lower single - precision, floating - point value.566
-// https://msdn.microsoft.com/en-us/library/tzz10fbx(v=vs.100).aspx567
-FORCE_INLINE void _mm_store_ss(float *p, __m128 a)568
-{569
- vst1q_lane_f32(p, vreinterpretq_f32_m128(a), 0);570
-}571
-572
-// Reads the lower 64 bits of b and stores them into the lower 64 bits of a.573
-// https://msdn.microsoft.com/en-us/library/hhwf428f%28v=vs.90%29.aspx574
-FORCE_INLINE void _mm_storel_epi64(__m128i *a, __m128i b)575
-{576
- uint64x1_t hi = vget_high_u64(vreinterpretq_u64_m128i(*a));577
- uint64x1_t lo = vget_low_u64(vreinterpretq_u64_m128i(b));578
- *a = vreinterpretq_m128i_u64(vcombine_u64(lo, hi));579
-}580
-581
-// Stores the lower two single-precision floating point values of a to the582
-// address p.583
-//584
-// *p0 := a0585
-// *p1 := a1586
-//587
-// https://msdn.microsoft.com/en-us/library/h54t98ks(v=vs.90).aspx588
-FORCE_INLINE void _mm_storel_pi(__m64 *p, __m128 a)589
-{590
- *p = vget_low_f32(a);591
-}592
-593
-// Stores the upper two single-precision, floating-point values of a to the594
-// address p.595
-//596
-// *p0 := a2597
-// *p1 := a3598
-//599
-// https://msdn.microsoft.com/en-us/library/a7525fs8(v%3dvs.90).aspx600
-FORCE_INLINE void _mm_storeh_pi(__m64 *p, __m128 a)601
-{602
- *p = vget_high_f32(a);603
-}604
-605
-// Loads a single single-precision, floating-point value, copying it into all606
-// four words607
-// https://msdn.microsoft.com/en-us/library/vstudio/5cdkf716(v=vs.100).aspx608
-FORCE_INLINE __m128 _mm_load1_ps(const float *p)609
-{610
- return vreinterpretq_m128_f32(vld1q_dup_f32(p));611
-}612
-#define _mm_load_ps1 _mm_load1_ps613
-614
-// Sets the lower two single-precision, floating-point values with 64615
-// bits of data loaded from the address p; the upper two values are passed616
-// through from a.617
-//618
-// Return Value619
-// r0 := *p0620
-// r1 := *p1621
-// r2 := a2622
-// r3 := a3623
-//624
-// https://msdn.microsoft.com/en-us/library/s57cyak2(v=vs.100).aspx625
-FORCE_INLINE __m128 _mm_loadl_pi(__m128 a, __m64 const *p)626
-{627
- return vreinterpretq_m128_f32(628
- vcombine_f32(vld1_f32((const float32_t *)p), vget_high_f32(a)));629
-}630
-631
-// Sets the upper two single-precision, floating-point values with 64632
-// bits of data loaded from the address p; the lower two values are passed633
-// through from a.634
-//635
-// r0 := a0636
-// r1 := a1637
-// r2 := *p0638
-// r3 := *p1639
-//640
-// https://msdn.microsoft.com/en-us/library/w92wta0x(v%3dvs.100).aspx641
-FORCE_INLINE __m128 _mm_loadh_pi(__m128 a, __m64 const *p)642
-{643
- return vreinterpretq_m128_f32(644
- vcombine_f32(vget_low_f32(a), vld1_f32((const float32_t *)p)));645
-}646
-647
-// Loads four single-precision, floating-point values.648
-// https://msdn.microsoft.com/en-us/library/vstudio/zzd50xxt(v=vs.100).aspx649
-FORCE_INLINE __m128 _mm_load_ps(const float *p)650
-{651
- return vreinterpretq_m128_f32(vld1q_f32(p));652
-}653
-654
-// Loads four single-precision, floating-point values.655
-// https://msdn.microsoft.com/en-us/library/x1b16s7z%28v=vs.90%29.aspx656
-FORCE_INLINE __m128 _mm_loadu_ps(const float *p)657
-{658
- // for neon, alignment doesn't matter, so _mm_load_ps and _mm_loadu_ps are659
- // equivalent for neon660
- return vreinterpretq_m128_f32(vld1q_f32(p));661
-}662
-663
-// Loads a double-precision, floating-point value.664
-// The upper double-precision, floating-point is set to zero. The address p does665
-// not need to be 16-byte aligned.666
-// https://docs.microsoft.com/en-us/previous-versions/visualstudio/visual-studio-2010/574w9fdd(v%3dvs.100)667
-FORCE_INLINE __m128d _mm_load_sd(const double *p)668
-{669
-#if defined(__aarch64__)670
- return vsetq_lane_f64(*p, vdupq_n_f64(0), 0);671
-#else672
- const float *fp = (const float *)p;673
- float ALIGN_STRUCT(16) data[4] = {fp[0], fp[1], 0, 0};674
- return vld1q_f32(data);675
-#endif676
-}677
-678
-// Loads an single - precision, floating - point value into the low word and679
-// clears the upper three words.680
-// https://msdn.microsoft.com/en-us/library/548bb9h4%28v=vs.90%29.aspx681
-FORCE_INLINE __m128 _mm_load_ss(const float *p)682
-{683
- return vreinterpretq_m128_f32(vsetq_lane_f32(*p, vdupq_n_f32(0), 0));684
-}685
-686
-FORCE_INLINE __m128i _mm_loadl_epi64(__m128i const *p)687
-{688
- /* Load the lower 64 bits of the value pointed to by p into the689
- * lower 64 bits of the result, zeroing the upper 64 bits of the result.690
- */691
- return vreinterpretq_m128i_s32(692
- vcombine_s32(vld1_s32((int32_t const *)p), vcreate_s32(0)));693
-}694
-695
-/* Logic/Binary operations */696
-697
-// Compares for inequality.698
-// https://msdn.microsoft.com/en-us/library/sf44thbx(v=vs.100).aspx699
-FORCE_INLINE __m128 _mm_cmpneq_ps(__m128 a, __m128 b)700
-{701
- return vreinterpretq_m128_u32(vmvnq_u32(vceqq_f32(702
- vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(b))));703
-}704
-705
-// Computes the bitwise AND-NOT of the four single-precision, floating-point706
-// values of a and b.707
-//708
-// r0 := ~a0 & b0709
-// r1 := ~a1 & b1710
-// r2 := ~a2 & b2711
-// r3 := ~a3 & b3712
-//713
-// https://msdn.microsoft.com/en-us/library/vstudio/68h7wd02(v=vs.100).aspx714
-FORCE_INLINE __m128 _mm_andnot_ps(__m128 a, __m128 b)715
-{716
- return vreinterpretq_m128_s32(717
- vbicq_s32(vreinterpretq_s32_m128(b),718
- vreinterpretq_s32_m128(a))); // *NOTE* argument swap719
-}720
-721
-// Computes the bitwise AND of the 128-bit value in b and the bitwise NOT of the722
-// 128-bit value in a.723
-//724
-// r := (~a) & b725
-//726
-// https://msdn.microsoft.com/en-us/library/vstudio/1beaceh8(v=vs.100).aspx727
-FORCE_INLINE __m128i _mm_andnot_si128(__m128i a, __m128i b)728
-{729
- return vreinterpretq_m128i_s32(730
- vbicq_s32(vreinterpretq_s32_m128i(b),731
- vreinterpretq_s32_m128i(a))); // *NOTE* argument swap732
-}733
-734
-// Computes the bitwise AND of the 128-bit value in a and the 128-bit value in735
-// b.736
-//737
-// r := a & b738
-//739
-// https://msdn.microsoft.com/en-us/library/vstudio/6d1txsa8(v=vs.100).aspx740
-FORCE_INLINE __m128i _mm_and_si128(__m128i a, __m128i b)741
-{742
- return vreinterpretq_m128i_s32(vandq_s32(vreinterpretq_s32_m128i(a),743
- vreinterpretq_s32_m128i(b)));744
-}745
-746
-// Computes the bitwise AND of the four single-precision, floating-point values747
-// of a and b.748
-//749
-// r0 := a0 & b0750
-// r1 := a1 & b1751
-// r2 := a2 & b2752
-// r3 := a3 & b3753
-//754
-// https://msdn.microsoft.com/en-us/library/vstudio/73ck1xc5(v=vs.100).aspx755
-FORCE_INLINE __m128 _mm_and_ps(__m128 a, __m128 b)756
-{757
- return vreinterpretq_m128_s32(vandq_s32(vreinterpretq_s32_m128(a),758
- vreinterpretq_s32_m128(b)));759
-}760
-761
-// Computes the bitwise OR of the four single-precision, floating-point values762
-// of a and b.763
-// https://msdn.microsoft.com/en-us/library/vstudio/7ctdsyy0(v=vs.100).aspx764
-FORCE_INLINE __m128 _mm_or_ps(__m128 a, __m128 b)765
-{766
- return vreinterpretq_m128_s32(vorrq_s32(vreinterpretq_s32_m128(a),767
- vreinterpretq_s32_m128(b)));768
-}769
-770
-// Computes bitwise EXOR (exclusive-or) of the four single-precision,771
-// floating-point values of a and b.772
-// https://msdn.microsoft.com/en-us/library/ss6k3wk8(v=vs.100).aspx773
-FORCE_INLINE __m128 _mm_xor_ps(__m128 a, __m128 b)774
-{775
- return vreinterpretq_m128_s32(veorq_s32(vreinterpretq_s32_m128(a),776
- vreinterpretq_s32_m128(b)));777
-}778
-779
-// Computes the bitwise OR of the 128-bit value in a and the 128-bit value in b.780
-//781
-// r := a | b782
-//783
-// https://msdn.microsoft.com/en-us/library/vstudio/ew8ty0db(v=vs.100).aspx784
-FORCE_INLINE __m128i _mm_or_si128(__m128i a, __m128i b)785
-{786
- return vreinterpretq_m128i_s32(vorrq_s32(vreinterpretq_s32_m128i(a),787
- vreinterpretq_s32_m128i(b)));788
-}789
-790
-// Computes the bitwise XOR of the 128-bit value in a and the 128-bit value in791
-// b. https://msdn.microsoft.com/en-us/library/fzt08www(v=vs.100).aspx792
-FORCE_INLINE __m128i _mm_xor_si128(__m128i a, __m128i b)793
-{794
- return vreinterpretq_m128i_s32(veorq_s32(vreinterpretq_s32_m128i(a),795
- vreinterpretq_s32_m128i(b)));796
-}797
-798
-// Moves the upper two values of B into the lower two values of A.799
-//800
-// r3 := a3801
-// r2 := a2802
-// r1 := b3803
-// r0 := b2804
-FORCE_INLINE __m128 _mm_movehl_ps(__m128 __A, __m128 __B)805
-{806
- float32x2_t a32 = vget_high_f32(vreinterpretq_f32_m128(__A));807
- float32x2_t b32 = vget_high_f32(vreinterpretq_f32_m128(__B));808
- return vreinterpretq_m128_f32(vcombine_f32(b32, a32));809
-}810
-811
-// Moves the lower two values of B into the upper two values of A.812
-//813
-// r3 := b1814
-// r2 := b0815
-// r1 := a1816
-// r0 := a0817
-FORCE_INLINE __m128 _mm_movelh_ps(__m128 __A, __m128 __B)818
-{819
- float32x2_t a10 = vget_low_f32(vreinterpretq_f32_m128(__A));820
- float32x2_t b10 = vget_low_f32(vreinterpretq_f32_m128(__B));821
- return vreinterpretq_m128_f32(vcombine_f32(a10, b10));822
-}823
-824
-FORCE_INLINE __m128i _mm_abs_epi32(__m128i a)825
-{826
- return vreinterpretq_m128i_s32(vabsq_s32(vreinterpretq_s32_m128i(a)));827
-}828
-829
-FORCE_INLINE __m128i _mm_abs_epi16(__m128i a)830
-{831
- return vreinterpretq_m128i_s16(vabsq_s16(vreinterpretq_s16_m128i(a)));832
-}833
-834
-FORCE_INLINE __m128i _mm_abs_epi8(__m128i a)835
-{836
- return vreinterpretq_m128i_s8(vabsq_s8(vreinterpretq_s8_m128i(a)));837
-}838
-839
-// Takes the upper 64 bits of a and places it in the low end of the result840
-// Takes the lower 64 bits of b and places it into the high end of the result.841
-FORCE_INLINE __m128 _mm_shuffle_ps_1032(__m128 a, __m128 b)842
-{843
- float32x2_t a32 = vget_high_f32(vreinterpretq_f32_m128(a));844
- float32x2_t b10 = vget_low_f32(vreinterpretq_f32_m128(b));845
- return vreinterpretq_m128_f32(vcombine_f32(a32, b10));846
-}847
-848
-// takes the lower two 32-bit values from a and swaps them and places in high849
-// end of result takes the higher two 32 bit values from b and swaps them and850
-// places in low end of result.851
-FORCE_INLINE __m128 _mm_shuffle_ps_2301(__m128 a, __m128 b)852
-{853
- float32x2_t a01 = vrev64_f32(vget_low_f32(vreinterpretq_f32_m128(a)));854
- float32x2_t b23 = vrev64_f32(vget_high_f32(vreinterpretq_f32_m128(b)));855
- return vreinterpretq_m128_f32(vcombine_f32(a01, b23));856
-}857
-858
-FORCE_INLINE __m128 _mm_shuffle_ps_0321(__m128 a, __m128 b)859
-{860
- float32x2_t a21 = vget_high_f32(vextq_f32(861
- vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(a), 3));862
- float32x2_t b03 = vget_low_f32(vextq_f32(vreinterpretq_f32_m128(b),863
- vreinterpretq_f32_m128(b), 3));864
- return vreinterpretq_m128_f32(vcombine_f32(a21, b03));865
-}866
-867
-FORCE_INLINE __m128 _mm_shuffle_ps_2103(__m128 a, __m128 b)868
-{869
- float32x2_t a03 = vget_low_f32(vextq_f32(vreinterpretq_f32_m128(a),870
- vreinterpretq_f32_m128(a), 3));871
- float32x2_t b21 = vget_high_f32(vextq_f32(872
- vreinterpretq_f32_m128(b), vreinterpretq_f32_m128(b), 3));873
- return vreinterpretq_m128_f32(vcombine_f32(a03, b21));874
-}875
-876
-FORCE_INLINE __m128 _mm_shuffle_ps_1010(__m128 a, __m128 b)877
-{878
- float32x2_t a10 = vget_low_f32(vreinterpretq_f32_m128(a));879
- float32x2_t b10 = vget_low_f32(vreinterpretq_f32_m128(b));880
- return vreinterpretq_m128_f32(vcombine_f32(a10, b10));881
-}882
-883
-FORCE_INLINE __m128 _mm_shuffle_ps_1001(__m128 a, __m128 b)884
-{885
- float32x2_t a01 = vrev64_f32(vget_low_f32(vreinterpretq_f32_m128(a)));886
- float32x2_t b10 = vget_low_f32(vreinterpretq_f32_m128(b));887
- return vreinterpretq_m128_f32(vcombine_f32(a01, b10));888
-}889
-890
-FORCE_INLINE __m128 _mm_shuffle_ps_0101(__m128 a, __m128 b)891
-{892
- float32x2_t a01 = vrev64_f32(vget_low_f32(vreinterpretq_f32_m128(a)));893
- float32x2_t b01 = vrev64_f32(vget_low_f32(vreinterpretq_f32_m128(b)));894
- return vreinterpretq_m128_f32(vcombine_f32(a01, b01));895
-}896
-897
-// keeps the low 64 bits of b in the low and puts the high 64 bits of a in the898
-// high899
-FORCE_INLINE __m128 _mm_shuffle_ps_3210(__m128 a, __m128 b)900
-{901
- float32x2_t a10 = vget_low_f32(vreinterpretq_f32_m128(a));902
- float32x2_t b32 = vget_high_f32(vreinterpretq_f32_m128(b));903
- return vreinterpretq_m128_f32(vcombine_f32(a10, b32));904
-}905
-906
-FORCE_INLINE __m128 _mm_shuffle_ps_0011(__m128 a, __m128 b)907
-{908
- float32x2_t a11 =909
- vdup_lane_f32(vget_low_f32(vreinterpretq_f32_m128(a)), 1);910
- float32x2_t b00 =911
- vdup_lane_f32(vget_low_f32(vreinterpretq_f32_m128(b)), 0);912
- return vreinterpretq_m128_f32(vcombine_f32(a11, b00));913
-}914
-915
-FORCE_INLINE __m128 _mm_shuffle_ps_0022(__m128 a, __m128 b)916
-{917
- float32x2_t a22 =918
- vdup_lane_f32(vget_high_f32(vreinterpretq_f32_m128(a)), 0);919
- float32x2_t b00 =920
- vdup_lane_f32(vget_low_f32(vreinterpretq_f32_m128(b)), 0);921
- return vreinterpretq_m128_f32(vcombine_f32(a22, b00));922
-}923
-924
-FORCE_INLINE __m128 _mm_shuffle_ps_2200(__m128 a, __m128 b)925
-{926
- float32x2_t a00 =927
- vdup_lane_f32(vget_low_f32(vreinterpretq_f32_m128(a)), 0);928
- float32x2_t b22 =929
- vdup_lane_f32(vget_high_f32(vreinterpretq_f32_m128(b)), 0);930
- return vreinterpretq_m128_f32(vcombine_f32(a00, b22));931
-}932
-933
-FORCE_INLINE __m128 _mm_shuffle_ps_3202(__m128 a, __m128 b)934
-{935
- float32_t a0 = vgetq_lane_f32(vreinterpretq_f32_m128(a), 0);936
- float32x2_t a22 =937
- vdup_lane_f32(vget_high_f32(vreinterpretq_f32_m128(a)), 0);938
- float32x2_t a02 = vset_lane_f32(a0, a22, 1); /* TODO: use vzip ?*/939
- float32x2_t b32 = vget_high_f32(vreinterpretq_f32_m128(b));940
- return vreinterpretq_m128_f32(vcombine_f32(a02, b32));941
-}942
-943
-FORCE_INLINE __m128 _mm_shuffle_ps_1133(__m128 a, __m128 b)944
-{945
- float32x2_t a33 =946
- vdup_lane_f32(vget_high_f32(vreinterpretq_f32_m128(a)), 1);947
- float32x2_t b11 =948
- vdup_lane_f32(vget_low_f32(vreinterpretq_f32_m128(b)), 1);949
- return vreinterpretq_m128_f32(vcombine_f32(a33, b11));950
-}951
-952
-FORCE_INLINE __m128 _mm_shuffle_ps_2010(__m128 a, __m128 b)953
-{954
- float32x2_t a10 = vget_low_f32(vreinterpretq_f32_m128(a));955
- float32_t b2 = vgetq_lane_f32(vreinterpretq_f32_m128(b), 2);956
- float32x2_t b00 =957
- vdup_lane_f32(vget_low_f32(vreinterpretq_f32_m128(b)), 0);958
- float32x2_t b20 = vset_lane_f32(b2, b00, 1);959
- return vreinterpretq_m128_f32(vcombine_f32(a10, b20));960
-}961
-962
-FORCE_INLINE __m128 _mm_shuffle_ps_2001(__m128 a, __m128 b)963
-{964
- float32x2_t a01 = vrev64_f32(vget_low_f32(vreinterpretq_f32_m128(a)));965
- float32_t b2 = vgetq_lane_f32(b, 2);966
- float32x2_t b00 =967
- vdup_lane_f32(vget_low_f32(vreinterpretq_f32_m128(b)), 0);968
- float32x2_t b20 = vset_lane_f32(b2, b00, 1);969
- return vreinterpretq_m128_f32(vcombine_f32(a01, b20));970
-}971
-972
-FORCE_INLINE __m128 _mm_shuffle_ps_2032(__m128 a, __m128 b)973
-{974
- float32x2_t a32 = vget_high_f32(vreinterpretq_f32_m128(a));975
- float32_t b2 = vgetq_lane_f32(b, 2);976
- float32x2_t b00 =977
- vdup_lane_f32(vget_low_f32(vreinterpretq_f32_m128(b)), 0);978
- float32x2_t b20 = vset_lane_f32(b2, b00, 1);979
- return vreinterpretq_m128_f32(vcombine_f32(a32, b20));980
-}981
-982
-// NEON does not support a general purpose permute intrinsic983
-// Selects four specific single-precision, floating-point values from a and b,984
-// based on the mask i.985
-// https://msdn.microsoft.com/en-us/library/vstudio/5f0858x0(v=vs.100).aspx986
-#if 0 /* C version */987
-FORCE_INLINE __m128 _mm_shuffle_ps_default(__m128 a,988
- __m128 b,989
- __constrange(0, 255) int imm)990
-{991
- __m128 ret;992
- ret[0] = a[imm & 0x3];993
- ret[1] = a[(imm >> 2) & 0x3];994
- ret[2] = b[(imm >> 4) & 0x03];995
- ret[3] = b[(imm >> 6) & 0x03];996
- return ret;997
-}998
-#endif999
-#define _mm_shuffle_ps_default(a, b, imm) \1000
- __extension__({ \1001
- float32x4_t ret; \1002
- ret = vmovq_n_f32(vgetq_lane_f32(vreinterpretq_f32_m128(a), \1003
- (imm) & (0x3))); \1004
- ret = vsetq_lane_f32(vgetq_lane_f32(vreinterpretq_f32_m128(a), \1005
- ((imm) >> 2) & 0x3), \1006
- ret, 1); \1007
- ret = vsetq_lane_f32(vgetq_lane_f32(vreinterpretq_f32_m128(b), \1008
- ((imm) >> 4) & 0x3), \1009
- ret, 2); \1010
- ret = vsetq_lane_f32(vgetq_lane_f32(vreinterpretq_f32_m128(b), \1011
- ((imm) >> 6) & 0x3), \1012
- ret, 3); \1013
- vreinterpretq_m128_f32(ret); \1014
- })1015
-1016
-// FORCE_INLINE __m128 _mm_shuffle_ps(__m128 a, __m128 b, __constrange(0,255)1017
-// int imm)1018
-#if __has_builtin(__builtin_shufflevector)1019
-#define _mm_shuffle_ps(a, b, imm) \1020
- __extension__({ \1021
- float32x4_t _input1 = vreinterpretq_f32_m128(a); \1022
- float32x4_t _input2 = vreinterpretq_f32_m128(b); \1023
- float32x4_t _shuf = __builtin_shufflevector( \1024
- _input1, _input2, (imm) & (0x3), ((imm) >> 2) & 0x3, \1025
- (((imm) >> 4) & 0x3) + 4, (((imm) >> 6) & 0x3) + 4); \1026
- vreinterpretq_m128_f32(_shuf); \1027
- })1028
-#else // generic1029
-#define _mm_shuffle_ps(a, b, imm) \1030
- __extension__({ \1031
- __m128 ret; \1032
- switch (imm) { \1033
- case _MM_SHUFFLE(1, 0, 3, 2): \1034
- ret = _mm_shuffle_ps_1032((a), (b)); \1035
- break; \1036
- case _MM_SHUFFLE(2, 3, 0, 1): \1037
- ret = _mm_shuffle_ps_2301((a), (b)); \1038
- break; \1039
- case _MM_SHUFFLE(0, 3, 2, 1): \1040
- ret = _mm_shuffle_ps_0321((a), (b)); \1041
- break; \1042
- case _MM_SHUFFLE(2, 1, 0, 3): \1043
- ret = _mm_shuffle_ps_2103((a), (b)); \1044
- break; \1045
- case _MM_SHUFFLE(1, 0, 1, 0): \1046
- ret = _mm_movelh_ps((a), (b)); \1047
- break; \1048
- case _MM_SHUFFLE(1, 0, 0, 1): \1049
- ret = _mm_shuffle_ps_1001((a), (b)); \1050
- break; \1051
- case _MM_SHUFFLE(0, 1, 0, 1): \1052
- ret = _mm_shuffle_ps_0101((a), (b)); \1053
- break; \1054
- case _MM_SHUFFLE(3, 2, 1, 0): \1055
- ret = _mm_shuffle_ps_3210((a), (b)); \1056
- break; \1057
- case _MM_SHUFFLE(0, 0, 1, 1): \1058
- ret = _mm_shuffle_ps_0011((a), (b)); \1059
- break; \1060
- case _MM_SHUFFLE(0, 0, 2, 2): \1061
- ret = _mm_shuffle_ps_0022((a), (b)); \1062
- break; \1063
- case _MM_SHUFFLE(2, 2, 0, 0): \1064
- ret = _mm_shuffle_ps_2200((a), (b)); \1065
- break; \1066
- case _MM_SHUFFLE(3, 2, 0, 2): \1067
- ret = _mm_shuffle_ps_3202((a), (b)); \1068
- break; \1069
- case _MM_SHUFFLE(3, 2, 3, 2): \1070
- ret = _mm_movehl_ps((b), (a)); \1071
- break; \1072
- case _MM_SHUFFLE(1, 1, 3, 3): \1073
- ret = _mm_shuffle_ps_1133((a), (b)); \1074
- break; \1075
- case _MM_SHUFFLE(2, 0, 1, 0): \1076
- ret = _mm_shuffle_ps_2010((a), (b)); \1077
- break; \1078
- case _MM_SHUFFLE(2, 0, 0, 1): \1079
- ret = _mm_shuffle_ps_2001((a), (b)); \1080
- break; \1081
- case _MM_SHUFFLE(2, 0, 3, 2): \1082
- ret = _mm_shuffle_ps_2032((a), (b)); \1083
- break; \1084
- default: \1085
- ret = _mm_shuffle_ps_default((a), (b), (imm)); \1086
- break; \1087
- } \1088
- ret; \1089
- })1090
-#endif1091
-1092
-// Takes the upper 64 bits of a and places it in the low end of the result1093
-// Takes the lower 64 bits of a and places it into the high end of the result.1094
-FORCE_INLINE __m128i _mm_shuffle_epi_1032(__m128i a)1095
-{1096
- int32x2_t a32 = vget_high_s32(vreinterpretq_s32_m128i(a));1097
- int32x2_t a10 = vget_low_s32(vreinterpretq_s32_m128i(a));1098
- return vreinterpretq_m128i_s32(vcombine_s32(a32, a10));1099
-}1100
-1101
-// takes the lower two 32-bit values from a and swaps them and places in low end1102
-// of result takes the higher two 32 bit values from a and swaps them and places1103
-// in high end of result.1104
-FORCE_INLINE __m128i _mm_shuffle_epi_2301(__m128i a)1105
-{1106
- int32x2_t a01 = vrev64_s32(vget_low_s32(vreinterpretq_s32_m128i(a)));1107
- int32x2_t a23 = vrev64_s32(vget_high_s32(vreinterpretq_s32_m128i(a)));1108
- return vreinterpretq_m128i_s32(vcombine_s32(a01, a23));1109
-}1110
-1111
-// rotates the least significant 32 bits into the most signficant 32 bits, and1112
-// shifts the rest down1113
-FORCE_INLINE __m128i _mm_shuffle_epi_0321(__m128i a)1114
-{1115
- return vreinterpretq_m128i_s32(vextq_s32(1116
- vreinterpretq_s32_m128i(a), vreinterpretq_s32_m128i(a), 1));1117
-}1118
-1119
-// rotates the most significant 32 bits into the least signficant 32 bits, and1120
-// shifts the rest up1121
-FORCE_INLINE __m128i _mm_shuffle_epi_2103(__m128i a)1122
-{1123
- return vreinterpretq_m128i_s32(vextq_s32(1124
- vreinterpretq_s32_m128i(a), vreinterpretq_s32_m128i(a), 3));1125
-}1126
-1127
-// gets the lower 64 bits of a, and places it in the upper 64 bits1128
-// gets the lower 64 bits of a and places it in the lower 64 bits1129
-FORCE_INLINE __m128i _mm_shuffle_epi_1010(__m128i a)1130
-{1131
- int32x2_t a10 = vget_low_s32(vreinterpretq_s32_m128i(a));1132
- return vreinterpretq_m128i_s32(vcombine_s32(a10, a10));1133
-}1134
-1135
-// gets the lower 64 bits of a, swaps the 0 and 1 elements, and places it in the1136
-// lower 64 bits gets the lower 64 bits of a, and places it in the upper 64 bits1137
-FORCE_INLINE __m128i _mm_shuffle_epi_1001(__m128i a)1138
-{1139
- int32x2_t a01 = vrev64_s32(vget_low_s32(vreinterpretq_s32_m128i(a)));1140
- int32x2_t a10 = vget_low_s32(vreinterpretq_s32_m128i(a));1141
- return vreinterpretq_m128i_s32(vcombine_s32(a01, a10));1142
-}1143
-1144
-// gets the lower 64 bits of a, swaps the 0 and 1 elements and places it in the1145
-// upper 64 bits gets the lower 64 bits of a, swaps the 0 and 1 elements, and1146
-// places it in the lower 64 bits1147
-FORCE_INLINE __m128i _mm_shuffle_epi_0101(__m128i a)1148
-{1149
- int32x2_t a01 = vrev64_s32(vget_low_s32(vreinterpretq_s32_m128i(a)));1150
- return vreinterpretq_m128i_s32(vcombine_s32(a01, a01));1151
-}1152
-1153
-FORCE_INLINE __m128i _mm_shuffle_epi_2211(__m128i a)1154
-{1155
- int32x2_t a11 =1156
- vdup_lane_s32(vget_low_s32(vreinterpretq_s32_m128i(a)), 1);1157
- int32x2_t a22 =1158
- vdup_lane_s32(vget_high_s32(vreinterpretq_s32_m128i(a)), 0);1159
- return vreinterpretq_m128i_s32(vcombine_s32(a11, a22));1160
-}1161
-1162
-FORCE_INLINE __m128i _mm_shuffle_epi_0122(__m128i a)1163
-{1164
- int32x2_t a22 =1165
- vdup_lane_s32(vget_high_s32(vreinterpretq_s32_m128i(a)), 0);1166
- int32x2_t a01 = vrev64_s32(vget_low_s32(vreinterpretq_s32_m128i(a)));1167
- return vreinterpretq_m128i_s32(vcombine_s32(a22, a01));1168
-}1169
-1170
-FORCE_INLINE __m128i _mm_shuffle_epi_3332(__m128i a)1171
-{1172
- int32x2_t a32 = vget_high_s32(vreinterpretq_s32_m128i(a));1173
- int32x2_t a33 =1174
- vdup_lane_s32(vget_high_s32(vreinterpretq_s32_m128i(a)), 1);1175
- return vreinterpretq_m128i_s32(vcombine_s32(a32, a33));1176
-}1177
-1178
-// Shuffle packed 8-bit integers in a according to shuffle control mask in the1179
-// corresponding 8-bit element of b, and store the results in dst.1180
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_shuffle_epi8&expand=51461181
-FORCE_INLINE __m128i _mm_shuffle_epi8(__m128i a, __m128i b)1182
-{1183
- int8x16_t tbl = vreinterpretq_s8_m128i(a); // input a1184
- uint8x16_t idx = vreinterpretq_u8_m128i(b); // input b1185
- uint8x16_t idx_masked =1186
- vandq_u8(idx, vdupq_n_u8(0x8F)); // avoid using meaningless bits1187
-#if defined(__aarch64__)1188
- return vreinterpretq_m128i_s8(vqtbl1q_s8(tbl, idx_masked));1189
-#elif defined(__GNUC__)1190
- int8x16_t ret;1191
- // %e and %f represent the even and odd D registers1192
- // respectively.1193
- __asm__ __volatile__("vtbl.8 %e[ret], {%e[tbl], %f[tbl]}, %e[idx]\n"1194
- "vtbl.8 %f[ret], {%e[tbl], %f[tbl]}, %f[idx]\n"1195
- : [ret] "=&w"(ret)1196
- : [tbl] "w"(tbl), [idx] "w"(idx_masked));1197
- return vreinterpretq_m128i_s8(ret);1198
-#else1199
- // use this line if testing on aarch641200
- int8x8x2_t a_split = {vget_low_s8(tbl), vget_high_s8(tbl)};1201
- return vreinterpretq_m128i_s8(1202
- vcombine_s8(vtbl2_s8(a_split, vget_low_u8(idx_masked)),1203
- vtbl2_s8(a_split, vget_high_u8(idx_masked))));1204
-#endif1205
-}1206
-1207
-#if 0 /* C version */1208
-FORCE_INLINE __m128i _mm_shuffle_epi32_default(__m128i a,1209
- __constrange(0, 255) int imm)1210
-{1211
- __m128i ret;1212
- ret[0] = a[imm & 0x3];1213
- ret[1] = a[(imm >> 2) & 0x3];1214
- ret[2] = a[(imm >> 4) & 0x03];1215
- ret[3] = a[(imm >> 6) & 0x03];1216
- return ret;1217
-}1218
-#endif1219
-#define _mm_shuffle_epi32_default(a, imm) \1220
- __extension__({ \1221
- int32x4_t ret; \1222
- ret = vmovq_n_s32(vgetq_lane_s32(vreinterpretq_s32_m128i(a), \1223
- (imm) & (0x3))); \1224
- ret = vsetq_lane_s32( \1225
- vgetq_lane_s32(vreinterpretq_s32_m128i(a), \1226
- ((imm) >> 2) & 0x3), \1227
- ret, 1); \1228
- ret = vsetq_lane_s32( \1229
- vgetq_lane_s32(vreinterpretq_s32_m128i(a), \1230
- ((imm) >> 4) & 0x3), \1231
- ret, 2); \1232
- ret = vsetq_lane_s32( \1233
- vgetq_lane_s32(vreinterpretq_s32_m128i(a), \1234
- ((imm) >> 6) & 0x3), \1235
- ret, 3); \1236
- vreinterpretq_m128i_s32(ret); \1237
- })1238
-1239
-// FORCE_INLINE __m128i _mm_shuffle_epi32_splat(__m128i a, __constrange(0,255)1240
-// int imm)1241
-#if defined(__aarch64__)1242
-#define _mm_shuffle_epi32_splat(a, imm) \1243
- __extension__({ \1244
- vreinterpretq_m128i_s32( \1245
- vdupq_laneq_s32(vreinterpretq_s32_m128i(a), (imm))); \1246
- })1247
-#else1248
-#define _mm_shuffle_epi32_splat(a, imm) \1249
- __extension__({ \1250
- vreinterpretq_m128i_s32(vdupq_n_s32( \1251
- vgetq_lane_s32(vreinterpretq_s32_m128i(a), (imm)))); \1252
- })1253
-#endif1254
-1255
-// Shuffles the 4 signed or unsigned 32-bit integers in a as specified by imm.1256
-// https://msdn.microsoft.com/en-us/library/56f67xbk%28v=vs.90%29.aspx1257
-// FORCE_INLINE __m128i _mm_shuffle_epi32(__m128i a,1258
-// __constrange(0,255) int imm)1259
-#if __has_builtin(__builtin_shufflevector)1260
-#define _mm_shuffle_epi32(a, imm) \1261
- __extension__({ \1262
- int32x4_t _input = vreinterpretq_s32_m128i(a); \1263
- int32x4_t _shuf = __builtin_shufflevector( \1264
- _input, _input, (imm) & (0x3), ((imm) >> 2) & 0x3, \1265
- ((imm) >> 4) & 0x3, ((imm) >> 6) & 0x3); \1266
- vreinterpretq_m128i_s32(_shuf); \1267
- })1268
-#else // generic1269
-#define _mm_shuffle_epi32(a, imm) \1270
- __extension__({ \1271
- __m128i ret; \1272
- switch (imm) { \1273
- case _MM_SHUFFLE(1, 0, 3, 2): \1274
- ret = _mm_shuffle_epi_1032((a)); \1275
- break; \1276
- case _MM_SHUFFLE(2, 3, 0, 1): \1277
- ret = _mm_shuffle_epi_2301((a)); \1278
- break; \1279
- case _MM_SHUFFLE(0, 3, 2, 1): \1280
- ret = _mm_shuffle_epi_0321((a)); \1281
- break; \1282
- case _MM_SHUFFLE(2, 1, 0, 3): \1283
- ret = _mm_shuffle_epi_2103((a)); \1284
- break; \1285
- case _MM_SHUFFLE(1, 0, 1, 0): \1286
- ret = _mm_shuffle_epi_1010((a)); \1287
- break; \1288
- case _MM_SHUFFLE(1, 0, 0, 1): \1289
- ret = _mm_shuffle_epi_1001((a)); \1290
- break; \1291
- case _MM_SHUFFLE(0, 1, 0, 1): \1292
- ret = _mm_shuffle_epi_0101((a)); \1293
- break; \1294
- case _MM_SHUFFLE(2, 2, 1, 1): \1295
- ret = _mm_shuffle_epi_2211((a)); \1296
- break; \1297
- case _MM_SHUFFLE(0, 1, 2, 2): \1298
- ret = _mm_shuffle_epi_0122((a)); \1299
- break; \1300
- case _MM_SHUFFLE(3, 3, 3, 2): \1301
- ret = _mm_shuffle_epi_3332((a)); \1302
- break; \1303
- case _MM_SHUFFLE(0, 0, 0, 0): \1304
- ret = _mm_shuffle_epi32_splat((a), 0); \1305
- break; \1306
- case _MM_SHUFFLE(1, 1, 1, 1): \1307
- ret = _mm_shuffle_epi32_splat((a), 1); \1308
- break; \1309
- case _MM_SHUFFLE(2, 2, 2, 2): \1310
- ret = _mm_shuffle_epi32_splat((a), 2); \1311
- break; \1312
- case _MM_SHUFFLE(3, 3, 3, 3): \1313
- ret = _mm_shuffle_epi32_splat((a), 3); \1314
- break; \1315
- default: \1316
- ret = _mm_shuffle_epi32_default((a), (imm)); \1317
- break; \1318
- } \1319
- ret; \1320
- })1321
-#endif1322
-1323
-// Shuffles the lower 4 signed or unsigned 16-bit integers in a as specified1324
-// by imm.1325
-// https://docs.microsoft.com/en-us/previous-versions/visualstudio/visual-studio-2010/y41dkk37(v=vs.100)1326
-// FORCE_INLINE __m128i _mm_shufflelo_epi16_function(__m128i a,1327
-// __constrange(0,255) int1328
-// imm)1329
-#define _mm_shufflelo_epi16_function(a, imm) \1330
- __extension__({ \1331
- int16x8_t ret = vreinterpretq_s16_m128i(a); \1332
- int16x4_t lowBits = vget_low_s16(ret); \1333
- ret = vsetq_lane_s16(vget_lane_s16(lowBits, (imm) & (0x3)), \1334
- ret, 0); \1335
- ret = vsetq_lane_s16( \1336
- vget_lane_s16(lowBits, ((imm) >> 2) & 0x3), ret, 1); \1337
- ret = vsetq_lane_s16( \1338
- vget_lane_s16(lowBits, ((imm) >> 4) & 0x3), ret, 2); \1339
- ret = vsetq_lane_s16( \1340
- vget_lane_s16(lowBits, ((imm) >> 6) & 0x3), ret, 3); \1341
- vreinterpretq_m128i_s16(ret); \1342
- })1343
-1344
-// FORCE_INLINE __m128i _mm_shufflelo_epi16(__m128i a,1345
-// __constrange(0,255) int imm)1346
-#if __has_builtin(__builtin_shufflevector)1347
-#define _mm_shufflelo_epi16(a, imm) \1348
- __extension__({ \1349
- int16x8_t _input = vreinterpretq_s16_m128i(a); \1350
- int16x8_t _shuf = __builtin_shufflevector( \1351
- _input, _input, ((imm) & (0x3)), (((imm) >> 2) & 0x3), \1352
- (((imm) >> 4) & 0x3), (((imm) >> 6) & 0x3), 4, 5, 6, \1353
- 7); \1354
- vreinterpretq_m128i_s16(_shuf); \1355
- })1356
-#else // generic1357
-#define _mm_shufflelo_epi16(a, imm) _mm_shufflelo_epi16_function((a), (imm))1358
-#endif1359
-1360
-// Shuffles the upper 4 signed or unsigned 16-bit integers in a as specified1361
-// by imm.1362
-// https://msdn.microsoft.com/en-us/library/13ywktbs(v=vs.100).aspx1363
-// FORCE_INLINE __m128i _mm_shufflehi_epi16_function(__m128i a,1364
-// __constrange(0,255) int1365
-// imm)1366
-#define _mm_shufflehi_epi16_function(a, imm) \1367
- __extension__({ \1368
- int16x8_t ret = vreinterpretq_s16_m128i(a); \1369
- int16x4_t highBits = vget_high_s16(ret); \1370
- ret = vsetq_lane_s16(vget_lane_s16(highBits, (imm) & (0x3)), \1371
- ret, 4); \1372
- ret = vsetq_lane_s16( \1373
- vget_lane_s16(highBits, ((imm) >> 2) & 0x3), ret, 5); \1374
- ret = vsetq_lane_s16( \1375
- vget_lane_s16(highBits, ((imm) >> 4) & 0x3), ret, 6); \1376
- ret = vsetq_lane_s16( \1377
- vget_lane_s16(highBits, ((imm) >> 6) & 0x3), ret, 7); \1378
- vreinterpretq_m128i_s16(ret); \1379
- })1380
-1381
-// FORCE_INLINE __m128i _mm_shufflehi_epi16(__m128i a,1382
-// __constrange(0,255) int imm)1383
-#if __has_builtin(__builtin_shufflevector)1384
-#define _mm_shufflehi_epi16(a, imm) \1385
- __extension__({ \1386
- int16x8_t _input = vreinterpretq_s16_m128i(a); \1387
- int16x8_t _shuf = __builtin_shufflevector( \1388
- _input, _input, 0, 1, 2, 3, ((imm) & (0x3)) + 4, \1389
- (((imm) >> 2) & 0x3) + 4, (((imm) >> 4) & 0x3) + 4, \1390
- (((imm) >> 6) & 0x3) + 4); \1391
- vreinterpretq_m128i_s16(_shuf); \1392
- })1393
-#else // generic1394
-#define _mm_shufflehi_epi16(a, imm) _mm_shufflehi_epi16_function((a), (imm))1395
-#endif1396
-1397
-// Blend packed 16-bit integers from a and b using control mask imm8, and store1398
-// the results in dst.1399
-//1400
-// FOR j := 0 to 71401
-// i := j*161402
-// IF imm8[j]1403
-// dst[i+15:i] := b[i+15:i]1404
-// ELSE1405
-// dst[i+15:i] := a[i+15:i]1406
-// FI1407
-// ENDFOR1408
-// FORCE_INLINE __m128i _mm_blend_epi16(__m128i a, __m128i b,1409
-// __constrange(0,255) int imm)1410
-#define _mm_blend_epi16(a, b, imm) \1411
- __extension__({ \1412
- const uint16_t _mask[8] = { \1413
- ((imm) & (1 << 0)) ? 0xFFFF : 0x0000, \1414
- ((imm) & (1 << 1)) ? 0xFFFF : 0x0000, \1415
- ((imm) & (1 << 2)) ? 0xFFFF : 0x0000, \1416
- ((imm) & (1 << 3)) ? 0xFFFF : 0x0000, \1417
- ((imm) & (1 << 4)) ? 0xFFFF : 0x0000, \1418
- ((imm) & (1 << 5)) ? 0xFFFF : 0x0000, \1419
- ((imm) & (1 << 6)) ? 0xFFFF : 0x0000, \1420
- ((imm) & (1 << 7)) ? 0xFFFF : 0x0000}; \1421
- uint16x8_t _mask_vec = vld1q_u16(_mask); \1422
- uint16x8_t _a = vreinterpretq_u16_m128i(a); \1423
- uint16x8_t _b = vreinterpretq_u16_m128i(b); \1424
- vreinterpretq_m128i_u16(vbslq_u16(_mask_vec, _b, _a)); \1425
- })1426
-1427
-// Blend packed 8-bit integers from a and b using mask, and store the results in1428
-// dst.1429
-//1430
-// FOR j := 0 to 151431
-// i := j*81432
-// IF mask[i+7]1433
-// dst[i+7:i] := b[i+7:i]1434
-// ELSE1435
-// dst[i+7:i] := a[i+7:i]1436
-// FI1437
-// ENDFOR1438
-FORCE_INLINE __m128i _mm_blendv_epi8(__m128i _a, __m128i _b, __m128i _mask)1439
-{1440
- // Use a signed shift right to create a mask with the sign bit1441
- uint8x16_t mask = vreinterpretq_u8_s8(1442
- vshrq_n_s8(vreinterpretq_s8_m128i(_mask), 7));1443
- uint8x16_t a = vreinterpretq_u8_m128i(_a);1444
- uint8x16_t b = vreinterpretq_u8_m128i(_b);1445
- return vreinterpretq_m128i_u8(vbslq_u8(mask, b, a));1446
-}1447
-1448
-/* Shifts */1449
-1450
-// Shifts the 4 signed 32-bit integers in a right by count bits while shifting1451
-// in the sign bit.1452
-//1453
-// r0 := a0 >> count1454
-// r1 := a1 >> count1455
-// r2 := a2 >> count1456
-// r3 := a3 >> count immediate1457
-FORCE_INLINE __m128i _mm_srai_epi32(__m128i a, int count)1458
-{1459
- return (__m128i)vshlq_s32((int32x4_t)a, vdupq_n_s32(-count));1460
-}1461
-1462
-// Shifts the 8 signed 16-bit integers in a right by count bits while shifting1463
-// in the sign bit.1464
-//1465
-// r0 := a0 >> count1466
-// r1 := a1 >> count1467
-// ...1468
-// r7 := a7 >> count1469
-FORCE_INLINE __m128i _mm_srai_epi16(__m128i a, int count)1470
-{1471
- return (__m128i)vshlq_s16((int16x8_t)a, vdupq_n_s16(-count));1472
-}1473
-1474
-// Shifts the 8 signed or unsigned 16-bit integers in a left by count bits while1475
-// shifting in zeros.1476
-//1477
-// r0 := a0 << count1478
-// r1 := a1 << count1479
-// ...1480
-// r7 := a7 << count1481
-//1482
-// https://msdn.microsoft.com/en-us/library/es73bcsy(v=vs.90).aspx1483
-#define _mm_slli_epi16(a, imm) \1484
- __extension__({ \1485
- __m128i ret; \1486
- if ((imm) <= 0) { \1487
- ret = a; \1488
- } else if ((imm) > 31) { \1489
- ret = _mm_setzero_si128(); \1490
- } else { \1491
- ret = vreinterpretq_m128i_s16(vshlq_n_s16( \1492
- vreinterpretq_s16_m128i(a), (imm))); \1493
- } \1494
- ret; \1495
- })1496
-1497
-// Shifts the 4 signed or unsigned 32-bit integers in a left by count bits while1498
-// shifting in zeros. :1499
-// https://msdn.microsoft.com/en-us/library/z2k3bbtb%28v=vs.90%29.aspx1500
-// FORCE_INLINE __m128i _mm_slli_epi32(__m128i a, __constrange(0,255) int imm)1501
-#define _mm_slli_epi32(a, imm) \1502
- __extension__({ \1503
- __m128i ret; \1504
- if ((imm) <= 0) { \1505
- ret = a; \1506
- } else if ((imm) > 31) { \1507
- ret = _mm_setzero_si128(); \1508
- } else { \1509
- ret = vreinterpretq_m128i_s32(vshlq_n_s32( \1510
- vreinterpretq_s32_m128i(a), (imm))); \1511
- } \1512
- ret; \1513
- })1514
-1515
-// Shift packed 64-bit integers in a left by imm8 while shifting in zeros, and1516
-// store the results in dst.1517
-#define _mm_slli_epi64(a, imm) \1518
- __extension__({ \1519
- __m128i ret; \1520
- if ((imm) <= 0) { \1521
- ret = a; \1522
- } else if ((imm) > 63) { \1523
- ret = _mm_setzero_si128(); \1524
- } else { \1525
- ret = vreinterpretq_m128i_s64(vshlq_n_s64( \1526
- vreinterpretq_s64_m128i(a), (imm))); \1527
- } \1528
- ret; \1529
- })1530
-1531
-// Shifts the 8 signed or unsigned 16-bit integers in a right by count bits1532
-// while shifting in zeros.1533
-//1534
-// r0 := srl(a0, count)1535
-// r1 := srl(a1, count)1536
-// ...1537
-// r7 := srl(a7, count)1538
-//1539
-// https://msdn.microsoft.com/en-us/library/6tcwd38t(v=vs.90).aspx1540
-#define _mm_srli_epi16(a, imm) \1541
- __extension__({ \1542
- __m128i ret; \1543
- if ((imm) <= 0) { \1544
- ret = a; \1545
- } else if ((imm) > 31) { \1546
- ret = _mm_setzero_si128(); \1547
- } else { \1548
- ret = vreinterpretq_m128i_u16(vshrq_n_u16( \1549
- vreinterpretq_u16_m128i(a), (imm))); \1550
- } \1551
- ret; \1552
- })1553
-1554
-// Shifts the 4 signed or unsigned 32-bit integers in a right by count bits1555
-// while shifting in zeros.1556
-// https://msdn.microsoft.com/en-us/library/w486zcfa(v=vs.100).aspx FORCE_INLINE1557
-// __m128i _mm_srli_epi32(__m128i a, __constrange(0,255) int imm)1558
-#define _mm_srli_epi32(a, imm) \1559
- __extension__({ \1560
- __m128i ret; \1561
- if ((imm) <= 0) { \1562
- ret = a; \1563
- } else if ((imm) > 31) { \1564
- ret = _mm_setzero_si128(); \1565
- } else { \1566
- ret = vreinterpretq_m128i_u32(vshrq_n_u32( \1567
- vreinterpretq_u32_m128i(a), (imm))); \1568
- } \1569
- ret; \1570
- })1571
-1572
-// Shift packed 64-bit integers in a right by imm8 while shifting in zeros, and1573
-// store the results in dst.1574
-#define _mm_srli_epi64(a, imm) \1575
- __extension__({ \1576
- __m128i ret; \1577
- if ((imm) <= 0) { \1578
- ret = a; \1579
- } else if ((imm) > 63) { \1580
- ret = _mm_setzero_si128(); \1581
- } else { \1582
- ret = vreinterpretq_m128i_u64(vshrq_n_u64( \1583
- vreinterpretq_u64_m128i(a), (imm))); \1584
- } \1585
- ret; \1586
- })1587
-1588
-// Shifts the 4 signed 32 - bit integers in a right by count bits while shifting1589
-// in the sign bit.1590
-// https://msdn.microsoft.com/en-us/library/z1939387(v=vs.100).aspx1591
-// FORCE_INLINE __m128i _mm_srai_epi32(__m128i a, __constrange(0,255) int imm)1592
-#define _mm_srai_epi32(a, imm) \1593
- __extension__({ \1594
- __m128i ret; \1595
- if ((imm) <= 0) { \1596
- ret = a; \1597
- } else if ((imm) > 31) { \1598
- ret = vreinterpretq_m128i_s32( \1599
- vshrq_n_s32(vreinterpretq_s32_m128i(a), 16)); \1600
- ret = vreinterpretq_m128i_s32(vshrq_n_s32( \1601
- vreinterpretq_s32_m128i(ret), 16)); \1602
- } else { \1603
- ret = vreinterpretq_m128i_s32(vshrq_n_s32( \1604
- vreinterpretq_s32_m128i(a), (imm))); \1605
- } \1606
- ret; \1607
- })1608
-1609
-// Shifts the 128 - bit value in a right by imm bytes while shifting in1610
-// zeros.imm must be an immediate.1611
-//1612
-// r := srl(a, imm*8)1613
-//1614
-// https://msdn.microsoft.com/en-us/library/305w28yz(v=vs.100).aspx1615
-// FORCE_INLINE _mm_srli_si128(__m128i a, __constrange(0,255) int imm)1616
-#define _mm_srli_si128(a, imm) \1617
- __extension__({ \1618
- __m128i ret; \1619
- if ((imm) <= 0) { \1620
- ret = a; \1621
- } else if ((imm) > 15) { \1622
- ret = _mm_setzero_si128(); \1623
- } else { \1624
- ret = vreinterpretq_m128i_s8( \1625
- vextq_s8(vreinterpretq_s8_m128i(a), \1626
- vdupq_n_s8(0), (imm))); \1627
- } \1628
- ret; \1629
- })1630
-1631
-// Shifts the 128-bit value in a left by imm bytes while shifting in zeros. imm1632
-// must be an immediate.1633
-//1634
-// r := a << (imm * 8)1635
-//1636
-// https://msdn.microsoft.com/en-us/library/34d3k2kt(v=vs.100).aspx1637
-// FORCE_INLINE __m128i _mm_slli_si128(__m128i a, __constrange(0,255) int imm)1638
-#define _mm_slli_si128(a, imm) \1639
- __extension__({ \1640
- __m128i ret; \1641
- if ((imm) <= 0) { \1642
- ret = a; \1643
- } else if ((imm) > 15) { \1644
- ret = _mm_setzero_si128(); \1645
- } else { \1646
- ret = vreinterpretq_m128i_s8(vextq_s8( \1647
- vdupq_n_s8(0), vreinterpretq_s8_m128i(a), \1648
- 16 - (imm))); \1649
- } \1650
- ret; \1651
- })1652
-1653
-// Shifts the 8 signed or unsigned 16-bit integers in a left by count bits while1654
-// shifting in zeros.1655
-//1656
-// r0 := a0 << count1657
-// r1 := a1 << count1658
-// ...1659
-// r7 := a7 << count1660
-//1661
-// https://msdn.microsoft.com/en-us/library/c79w388h(v%3dvs.90).aspx1662
-FORCE_INLINE __m128i _mm_sll_epi16(__m128i a, __m128i count)1663
-{1664
- uint64_t c = vreinterpretq_nth_u64_m128i(count, 0);1665
- if (c > 15)1666
- return _mm_setzero_si128();1667
-1668
- int16x8_t vc = vdupq_n_s16((int16_t)c);1669
- return vreinterpretq_m128i_s16(1670
- vshlq_s16(vreinterpretq_s16_m128i(a), vc));1671
-}1672
-1673
-// Shifts the 4 signed or unsigned 32-bit integers in a left by count bits while1674
-// shifting in zeros.1675
-//1676
-// r0 := a0 << count1677
-// r1 := a1 << count1678
-// r2 := a2 << count1679
-// r3 := a3 << count1680
-//1681
-// https://msdn.microsoft.com/en-us/library/6fe5a6s9(v%3dvs.90).aspx1682
-FORCE_INLINE __m128i _mm_sll_epi32(__m128i a, __m128i count)1683
-{1684
- uint64_t c = vreinterpretq_nth_u64_m128i(count, 0);1685
- if (c > 31)1686
- return _mm_setzero_si128();1687
-1688
- int32x4_t vc = vdupq_n_s32((int32_t)c);1689
- return vreinterpretq_m128i_s32(1690
- vshlq_s32(vreinterpretq_s32_m128i(a), vc));1691
-}1692
-1693
-// Shifts the 2 signed or unsigned 64-bit integers in a left by count bits while1694
-// shifting in zeros.1695
-//1696
-// r0 := a0 << count1697
-// r1 := a1 << count1698
-//1699
-// https://msdn.microsoft.com/en-us/library/6ta9dffd(v%3dvs.90).aspx1700
-FORCE_INLINE __m128i _mm_sll_epi64(__m128i a, __m128i count)1701
-{1702
- uint64_t c = vreinterpretq_nth_u64_m128i(count, 0);1703
- if (c > 63)1704
- return _mm_setzero_si128();1705
-1706
- int64x2_t vc = vdupq_n_s64((int64_t)c);1707
- return vreinterpretq_m128i_s64(1708
- vshlq_s64(vreinterpretq_s64_m128i(a), vc));1709
-}1710
-1711
-// Shifts the 8 signed or unsigned 16-bit integers in a right by count bits1712
-// while shifting in zeros.1713
-//1714
-// r0 := srl(a0, count)1715
-// r1 := srl(a1, count)1716
-// ...1717
-// r7 := srl(a7, count)1718
-//1719
-// https://msdn.microsoft.com/en-us/library/wd5ax830(v%3dvs.90).aspx1720
-FORCE_INLINE __m128i _mm_srl_epi16(__m128i a, __m128i count)1721
-{1722
- uint64_t c = vreinterpretq_nth_u64_m128i(count, 0);1723
- if (c > 15)1724
- return _mm_setzero_si128();1725
-1726
- int16x8_t vc = vdupq_n_s16(-(int16_t)c);1727
- return vreinterpretq_m128i_u16(1728
- vshlq_u16(vreinterpretq_u16_m128i(a), vc));1729
-}1730
-1731
-// Shifts the 4 signed or unsigned 32-bit integers in a right by count bits1732
-// while shifting in zeros.1733
-//1734
-// r0 := srl(a0, count)1735
-// r1 := srl(a1, count)1736
-// r2 := srl(a2, count)1737
-// r3 := srl(a3, count)1738
-//1739
-// https://msdn.microsoft.com/en-us/library/a9cbttf4(v%3dvs.90).aspx1740
-FORCE_INLINE __m128i _mm_srl_epi32(__m128i a, __m128i count)1741
-{1742
- uint64_t c = vreinterpretq_nth_u64_m128i(count, 0);1743
- if (c > 31)1744
- return _mm_setzero_si128();1745
-1746
- int32x4_t vc = vdupq_n_s32(-(int32_t)c);1747
- return vreinterpretq_m128i_u32(1748
- vshlq_u32(vreinterpretq_u32_m128i(a), vc));1749
-}1750
-1751
-// Shifts the 2 signed or unsigned 64-bit integers in a right by count bits1752
-// while shifting in zeros.1753
-//1754
-// r0 := srl(a0, count)1755
-// r1 := srl(a1, count)1756
-//1757
-// https://msdn.microsoft.com/en-us/library/yf6cf9k8(v%3dvs.90).aspx1758
-FORCE_INLINE __m128i _mm_srl_epi64(__m128i a, __m128i count)1759
-{1760
- uint64_t c = vreinterpretq_nth_u64_m128i(count, 0);1761
- if (c > 63)1762
- return _mm_setzero_si128();1763
-1764
- int64x2_t vc = vdupq_n_s64(-(int64_t)c);1765
- return vreinterpretq_m128i_u64(1766
- vshlq_u64(vreinterpretq_u64_m128i(a), vc));1767
-}1768
-1769
-// NEON does not provide a version of this function.1770
-// Creates a 16-bit mask from the most significant bits of the 16 signed or1771
-// unsigned 8-bit integers in a and zero extends the upper bits.1772
-// https://msdn.microsoft.com/en-us/library/vstudio/s090c8fk(v=vs.100).aspx1773
-FORCE_INLINE int _mm_movemask_epi8(__m128i a)1774
-{1775
-#if defined(__aarch64__)1776
- uint8x16_t input = vreinterpretq_u8_m128i(a);1777
- const int8_t ALIGN_STRUCT(16) xr[16] = {-7, -6, -5, -4, -3, -2, -1, 0,1778
- -7, -6, -5, -4, -3, -2, -1, 0};1779
- const uint8x16_t mask_and = vdupq_n_u8(0x80);1780
- const int8x16_t mask_shift = vld1q_s8(xr);1781
- const uint8x16_t mask_result =1782
- vshlq_u8(vandq_u8(input, mask_and), mask_shift);1783
- uint8x8_t lo = vget_low_u8(mask_result);1784
- uint8x8_t hi = vget_high_u8(mask_result);1785
-1786
- return vaddv_u8(lo) + (vaddv_u8(hi) << 8);1787
-#else1788
- // Use increasingly wide shifts+adds to collect the sign bits1789
- // together.1790
- // Since the widening shifts would be rather confusing to follow in little1791
- // endian, everything will be illustrated in big endian order instead. This1792
- // has a different result - the bits would actually be reversed on a big1793
- // endian machine.1794
-1795
- // Starting input (only half the elements are shown):1796
- // 89 ff 1d c0 00 10 99 331797
- uint8x16_t input = vreinterpretq_u8_m128i(a);1798
-1799
- // Shift out everything but the sign bits with an unsigned shift right.1800
- //1801
- // Bytes of the vector::1802
- // 89 ff 1d c0 00 10 99 331803
- // \ \ \ \ \ \ \ \ high_bits = (uint16x4_t)(input >> 7)1804
- // | | | | | | | |1805
- // 01 01 00 01 00 00 01 001806
- //1807
- // Bits of first important lane(s):1808
- // 10001001 (89)1809
- // \______1810
- // |1811
- // 00000001 (01)1812
- uint16x8_t high_bits = vreinterpretq_u16_u8(vshrq_n_u8(input, 7));1813
-1814
- // Merge the even lanes together with a 16-bit unsigned shift right + add.1815
- // 'xx' represents garbage data which will be ignored in the final result.1816
- // In the important bytes, the add functions like a binary OR.1817
- //1818
- // 01 01 00 01 00 00 01 001819
- // \_ | \_ | \_ | \_ | paired16 = (uint32x4_t)(input + (input >> 7))1820
- // \| \| \| \|1821
- // xx 03 xx 01 xx 00 xx 021822
- //1823
- // 00000001 00000001 (01 01)1824
- // \_______ |1825
- // \|1826
- // xxxxxxxx xxxxxx11 (xx 03)1827
- uint32x4_t paired16 =1828
- vreinterpretq_u32_u16(vsraq_n_u16(high_bits, high_bits, 7));1829
-1830
- // Repeat with a wider 32-bit shift + add.1831
- // xx 03 xx 01 xx 00 xx 021832
- // \____ | \____ | paired32 = (uint64x1_t)(paired16 + (paired16 >>1833
- // 14))1834
- // \| \|1835
- // xx xx xx 0d xx xx xx 021836
- //1837
- // 00000011 00000001 (03 01)1838
- // \\_____ ||1839
- // '----.\||1840
- // xxxxxxxx xxxx1101 (xx 0d)1841
- uint64x2_t paired32 =1842
- vreinterpretq_u64_u32(vsraq_n_u32(paired16, paired16, 14));1843
-1844
- // Last, an even wider 64-bit shift + add to get our result in the low 8 bit1845
- // lanes. xx xx xx 0d xx xx xx 021846
- // \_________ | paired64 = (uint8x8_t)(paired32 + (paired32 >>1847
- // 28))1848
- // \|1849
- // xx xx xx xx xx xx xx d21850
- //1851
- // 00001101 00000010 (0d 02)1852
- // \ \___ | |1853
- // '---. \| |1854
- // xxxxxxxx 11010010 (xx d2)1855
- uint8x16_t paired64 =1856
- vreinterpretq_u8_u64(vsraq_n_u64(paired32, paired32, 28));1857
-1858
- // Extract the low 8 bits from each 64-bit lane with 2 8-bit extracts.1859
- // xx xx xx xx xx xx xx d21860
- // || return paired64[0]1861
- // d21862
- // Note: Little endian would return the correct value 4b (01001011) instead.1863
- return vgetq_lane_u8(paired64, 0) |1864
- ((int)vgetq_lane_u8(paired64, 8) << 8);1865
-#endif1866
-}1867
-1868
-// NEON does not provide this method1869
-// Creates a 4-bit mask from the most significant bits of the four1870
-// single-precision, floating-point values.1871
-// https://msdn.microsoft.com/en-us/library/vstudio/4490ys29(v=vs.100).aspx1872
-FORCE_INLINE int _mm_movemask_ps(__m128 a)1873
-{1874
- uint32x4_t input = vreinterpretq_u32_m128(a);1875
-#if defined(__aarch64__)1876
- static const int32x4_t shift = {-31, -30, -29, -28};1877
- static const uint32x4_t highbit = {0x80000000, 0x80000000, 0x80000000,1878
- 0x80000000};1879
- return vaddvq_u32(vshlq_u32(vandq_u32(input, highbit), shift));1880
-#else1881
- // Uses the exact same method as _mm_movemask_epi8, see that for details.1882
- // Shift out everything but the sign bits with a 32-bit unsigned shift1883
- // right.1884
- uint64x2_t high_bits = vreinterpretq_u64_u32(vshrq_n_u32(input, 31));1885
- // Merge the two pairs together with a 64-bit unsigned shift right + add.1886
- uint8x16_t paired =1887
- vreinterpretq_u8_u64(vsraq_n_u64(high_bits, high_bits, 31));1888
- // Extract the result.1889
- return vgetq_lane_u8(paired, 0) | (vgetq_lane_u8(paired, 8) << 2);1890
-#endif1891
-}1892
-1893
-// Compute the bitwise AND of 128 bits (representing integer data) in a and1894
-// mask, and return 1 if the result is zero, otherwise return 0.1895
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_test_all_zeros&expand=58711896
-FORCE_INLINE int _mm_test_all_zeros(__m128i a, __m128i mask)1897
-{1898
- int64x2_t a_and_mask = vandq_s64(vreinterpretq_s64_m128i(a),1899
- vreinterpretq_s64_m128i(mask));1900
- return (vgetq_lane_s64(a_and_mask, 0) | vgetq_lane_s64(a_and_mask, 1))1901
- ? 01902
- : 1;1903
-}1904
-1905
-/* Math operations */1906
-1907
-// Subtracts the four single-precision, floating-point values of a and b.1908
-//1909
-// r0 := a0 - b01910
-// r1 := a1 - b11911
-// r2 := a2 - b21912
-// r3 := a3 - b31913
-//1914
-// https://msdn.microsoft.com/en-us/library/vstudio/1zad2k61(v=vs.100).aspx1915
-FORCE_INLINE __m128 _mm_sub_ps(__m128 a, __m128 b)1916
-{1917
- return vreinterpretq_m128_f32(vsubq_f32(vreinterpretq_f32_m128(a),1918
- vreinterpretq_f32_m128(b)));1919
-}1920
-1921
-// Subtract 2 packed 64-bit integers in b from 2 packed 64-bit integers in a,1922
-// and store the results in dst.1923
-// r0 := a0 - b01924
-// r1 := a1 - b11925
-FORCE_INLINE __m128i _mm_sub_epi64(__m128i a, __m128i b)1926
-{1927
- return vreinterpretq_m128i_s64(vsubq_s64(vreinterpretq_s64_m128i(a),1928
- vreinterpretq_s64_m128i(b)));1929
-}1930
-1931
-// Subtracts the 4 signed or unsigned 32-bit integers of b from the 4 signed or1932
-// unsigned 32-bit integers of a.1933
-//1934
-// r0 := a0 - b01935
-// r1 := a1 - b11936
-// r2 := a2 - b21937
-// r3 := a3 - b31938
-//1939
-// https://msdn.microsoft.com/en-us/library/vstudio/fhh866h0(v=vs.100).aspx1940
-FORCE_INLINE __m128i _mm_sub_epi32(__m128i a, __m128i b)1941
-{1942
- return vreinterpretq_m128i_s32(vsubq_s32(vreinterpretq_s32_m128i(a),1943
- vreinterpretq_s32_m128i(b)));1944
-}1945
-1946
-FORCE_INLINE __m128i _mm_sub_epi16(__m128i a, __m128i b)1947
-{1948
- return vreinterpretq_m128i_s16(vsubq_s16(vreinterpretq_s16_m128i(a),1949
- vreinterpretq_s16_m128i(b)));1950
-}1951
-1952
-FORCE_INLINE __m128i _mm_sub_epi8(__m128i a, __m128i b)1953
-{1954
- return vreinterpretq_m128i_s8(1955
- vsubq_s8(vreinterpretq_s8_m128i(a), vreinterpretq_s8_m128i(b)));1956
-}1957
-1958
-// Subtracts the 8 unsigned 16-bit integers of bfrom the 8 unsigned 16-bit1959
-// integers of a and saturates..1960
-// https://technet.microsoft.com/en-us/subscriptions/index/f44y0s19(v=vs.90).aspx1961
-FORCE_INLINE __m128i _mm_subs_epu16(__m128i a, __m128i b)1962
-{1963
- return vreinterpretq_m128i_u16(vqsubq_u16(vreinterpretq_u16_m128i(a),1964
- vreinterpretq_u16_m128i(b)));1965
-}1966
-1967
-// Subtracts the 16 unsigned 8-bit integers of b from the 16 unsigned 8-bit1968
-// integers of a and saturates.1969
-//1970
-// r0 := UnsignedSaturate(a0 - b0)1971
-// r1 := UnsignedSaturate(a1 - b1)1972
-// ...1973
-// r15 := UnsignedSaturate(a15 - b15)1974
-//1975
-// https://technet.microsoft.com/en-us/subscriptions/yadkxc18(v=vs.90)1976
-FORCE_INLINE __m128i _mm_subs_epu8(__m128i a, __m128i b)1977
-{1978
- return vreinterpretq_m128i_u8(vqsubq_u8(vreinterpretq_u8_m128i(a),1979
- vreinterpretq_u8_m128i(b)));1980
-}1981
-1982
-// Subtracts the 16 signed 8-bit integers of b from the 16 signed 8-bit integers1983
-// of a and saturates.1984
-//1985
-// r0 := SignedSaturate(a0 - b0)1986
-// r1 := SignedSaturate(a1 - b1)1987
-// ...1988
-// r15 := SignedSaturate(a15 - b15)1989
-//1990
-// https://technet.microsoft.com/en-us/subscriptions/by7kzks1(v=vs.90)1991
-FORCE_INLINE __m128i _mm_subs_epi8(__m128i a, __m128i b)1992
-{1993
- return vreinterpretq_m128i_s8(vqsubq_s8(vreinterpretq_s8_m128i(a),1994
- vreinterpretq_s8_m128i(b)));1995
-}1996
-1997
-// Subtracts the 8 signed 16-bit integers of b from the 8 signed 16-bit integers1998
-// of a and saturates.1999
-//2000
-// r0 := SignedSaturate(a0 - b0)2001
-// r1 := SignedSaturate(a1 - b1)2002
-// ...2003
-// r7 := SignedSaturate(a7 - b7)2004
-//2005
-// https://technet.microsoft.com/en-us/subscriptions/3247z5b8(v=vs.90)2006
-FORCE_INLINE __m128i _mm_subs_epi16(__m128i a, __m128i b)2007
-{2008
- return vreinterpretq_m128i_s16(vqsubq_s16(vreinterpretq_s16_m128i(a),2009
- vreinterpretq_s16_m128i(b)));2010
-}2011
-2012
-FORCE_INLINE __m128i _mm_adds_epu16(__m128i a, __m128i b)2013
-{2014
- return vreinterpretq_m128i_u16(vqaddq_u16(vreinterpretq_u16_m128i(a),2015
- vreinterpretq_u16_m128i(b)));2016
-}2017
-2018
-// Negate packed 8-bit integers in a when the corresponding signed2019
-// 8-bit integer in b is negative, and store the results in dst.2020
-// Element in dst are zeroed out when the corresponding element2021
-// in b is zero.2022
-//2023
-// for i in 0..152024
-// if b[i] < 02025
-// r[i] := -a[i]2026
-// else if b[i] == 02027
-// r[i] := 02028
-// else2029
-// r[i] := a[i]2030
-// fi2031
-// done2032
-FORCE_INLINE __m128i _mm_sign_epi8(__m128i _a, __m128i _b)2033
-{2034
- int8x16_t a = vreinterpretq_s8_m128i(_a);2035
- int8x16_t b = vreinterpretq_s8_m128i(_b);2036
-2037
- int8x16_t zero = vdupq_n_s8(0);2038
- // signed shift right: faster than vclt2039
- // (b < 0) ? 0xFF : 02040
- uint8x16_t ltMask = vreinterpretq_u8_s8(vshrq_n_s8(b, 7));2041
- // (b == 0) ? 0xFF : 02042
- int8x16_t zeroMask = vreinterpretq_s8_u8(vceqq_s8(b, zero));2043
- // -a2044
- int8x16_t neg = vnegq_s8(a);2045
- // bitwise select either a or neg based on ltMask2046
- int8x16_t masked = vbslq_s8(ltMask, a, neg);2047
- // res = masked & (~zeroMask)2048
- int8x16_t res = vbicq_s8(masked, zeroMask);2049
- return vreinterpretq_m128i_s8(res);2050
-}2051
-2052
-// Negate packed 16-bit integers in a when the corresponding signed2053
-// 16-bit integer in b is negative, and store the results in dst.2054
-// Element in dst are zeroed out when the corresponding element2055
-// in b is zero.2056
-//2057
-// for i in 0..72058
-// if b[i] < 02059
-// r[i] := -a[i]2060
-// else if b[i] == 02061
-// r[i] := 02062
-// else2063
-// r[i] := a[i]2064
-// fi2065
-// done2066
-FORCE_INLINE __m128i _mm_sign_epi16(__m128i _a, __m128i _b)2067
-{2068
- int16x8_t a = vreinterpretq_s16_m128i(_a);2069
- int16x8_t b = vreinterpretq_s16_m128i(_b);2070
-2071
- int16x8_t zero = vdupq_n_s16(0);2072
- // signed shift right: faster than vclt2073
- // (b < 0) ? 0xFFFF : 02074
- uint16x8_t ltMask = vreinterpretq_u16_s16(vshrq_n_s16(b, 15));2075
- // (b == 0) ? 0xFFFF : 02076
- int16x8_t zeroMask = vreinterpretq_s16_u16(vceqq_s16(b, zero));2077
- // -a2078
- int16x8_t neg = vnegq_s16(a);2079
- // bitwise select either a or neg based on ltMask2080
- int16x8_t masked = vbslq_s16(ltMask, a, neg);2081
- // res = masked & (~zeroMask)2082
- int16x8_t res = vbicq_s16(masked, zeroMask);2083
- return vreinterpretq_m128i_s16(res);2084
-}2085
-2086
-// Negate packed 32-bit integers in a when the corresponding signed2087
-// 32-bit integer in b is negative, and store the results in dst.2088
-// Element in dst are zeroed out when the corresponding element2089
-// in b is zero.2090
-//2091
-// for i in 0..32092
-// if b[i] < 02093
-// r[i] := -a[i]2094
-// else if b[i] == 02095
-// r[i] := 02096
-// else2097
-// r[i] := a[i]2098
-// fi2099
-// done2100
-FORCE_INLINE __m128i _mm_sign_epi32(__m128i _a, __m128i _b)2101
-{2102
- int32x4_t a = vreinterpretq_s32_m128i(_a);2103
- int32x4_t b = vreinterpretq_s32_m128i(_b);2104
-2105
- int32x4_t zero = vdupq_n_s32(0);2106
- // signed shift right: faster than vclt2107
- // (b < 0) ? 0xFFFFFFFF : 02108
- uint32x4_t ltMask = vreinterpretq_u32_s32(vshrq_n_s32(b, 31));2109
- // (b == 0) ? 0xFFFFFFFF : 02110
- int32x4_t zeroMask = vreinterpretq_s32_u32(vceqq_s32(b, zero));2111
- // neg = -a2112
- int32x4_t neg = vnegq_s32(a);2113
- // bitwise select either a or neg based on ltMask2114
- int32x4_t masked = vbslq_s32(ltMask, a, neg);2115
- // res = masked & (~zeroMask)2116
- int32x4_t res = vbicq_s32(masked, zeroMask);2117
- return vreinterpretq_m128i_s32(res);2118
-}2119
-2120
-// Computes the average of the 16 unsigned 8-bit integers in a and the 162121
-// unsigned 8-bit integers in b and rounds.2122
-//2123
-// r0 := (a0 + b0) / 22124
-// r1 := (a1 + b1) / 22125
-// ...2126
-// r15 := (a15 + b15) / 22127
-//2128
-// https://msdn.microsoft.com/en-us/library/vstudio/8zwh554a(v%3dvs.90).aspx2129
-FORCE_INLINE __m128i _mm_avg_epu8(__m128i a, __m128i b)2130
-{2131
- return vreinterpretq_m128i_u8(vrhaddq_u8(vreinterpretq_u8_m128i(a),2132
- vreinterpretq_u8_m128i(b)));2133
-}2134
-2135
-// Computes the average of the 8 unsigned 16-bit integers in a and the 82136
-// unsigned 16-bit integers in b and rounds.2137
-//2138
-// r0 := (a0 + b0) / 22139
-// r1 := (a1 + b1) / 22140
-// ...2141
-// r7 := (a7 + b7) / 22142
-//2143
-// https://msdn.microsoft.com/en-us/library/vstudio/y13ca3c8(v=vs.90).aspx2144
-FORCE_INLINE __m128i _mm_avg_epu16(__m128i a, __m128i b)2145
-{2146
- return (__m128i)vrhaddq_u16(vreinterpretq_u16_m128i(a),2147
- vreinterpretq_u16_m128i(b));2148
-}2149
-2150
-// Adds the four single-precision, floating-point values of a and b.2151
-//2152
-// r0 := a0 + b02153
-// r1 := a1 + b12154
-// r2 := a2 + b22155
-// r3 := a3 + b32156
-//2157
-// https://msdn.microsoft.com/en-us/library/vstudio/c9848chc(v=vs.100).aspx2158
-FORCE_INLINE __m128 _mm_add_ps(__m128 a, __m128 b)2159
-{2160
- return vreinterpretq_m128_f32(vaddq_f32(vreinterpretq_f32_m128(a),2161
- vreinterpretq_f32_m128(b)));2162
-}2163
-2164
-// adds the scalar single-precision floating point values of a and b.2165
-// https://msdn.microsoft.com/en-us/library/be94x2y6(v=vs.100).aspx2166
-FORCE_INLINE __m128 _mm_add_ss(__m128 a, __m128 b)2167
-{2168
- float32_t b0 = vgetq_lane_f32(vreinterpretq_f32_m128(b), 0);2169
- float32x4_t value = vsetq_lane_f32(b0, vdupq_n_f32(0), 0);2170
- // the upper values in the result must be the remnants of <a>.2171
- return vreinterpretq_m128_f32(vaddq_f32(a, value));2172
-}2173
-2174
-// Adds the 4 signed or unsigned 64-bit integers in a to the 4 signed or2175
-// unsigned 32-bit integers in b.2176
-// https://msdn.microsoft.com/en-us/library/vstudio/09xs4fkk(v=vs.100).aspx2177
-FORCE_INLINE __m128i _mm_add_epi64(__m128i a, __m128i b)2178
-{2179
- return vreinterpretq_m128i_s64(vaddq_s64(vreinterpretq_s64_m128i(a),2180
- vreinterpretq_s64_m128i(b)));2181
-}2182
-2183
-// Adds the 4 signed or unsigned 32-bit integers in a to the 4 signed or2184
-// unsigned 32-bit integers in b.2185
-//2186
-// r0 := a0 + b02187
-// r1 := a1 + b12188
-// r2 := a2 + b22189
-// r3 := a3 + b32190
-//2191
-// https://msdn.microsoft.com/en-us/library/vstudio/09xs4fkk(v=vs.100).aspx2192
-FORCE_INLINE __m128i _mm_add_epi32(__m128i a, __m128i b)2193
-{2194
- return vreinterpretq_m128i_s32(vaddq_s32(vreinterpretq_s32_m128i(a),2195
- vreinterpretq_s32_m128i(b)));2196
-}2197
-2198
-// Adds the 8 signed or unsigned 16-bit integers in a to the 8 signed or2199
-// unsigned 16-bit integers in b.2200
-// https://msdn.microsoft.com/en-us/library/fceha5k4(v=vs.100).aspx2201
-FORCE_INLINE __m128i _mm_add_epi16(__m128i a, __m128i b)2202
-{2203
- return vreinterpretq_m128i_s16(vaddq_s16(vreinterpretq_s16_m128i(a),2204
- vreinterpretq_s16_m128i(b)));2205
-}2206
-2207
-// Adds the 16 signed or unsigned 8-bit integers in a to the 16 signed or2208
-// unsigned 8-bit integers in b.2209
-// https://technet.microsoft.com/en-us/subscriptions/yc7tcyzs(v=vs.90)2210
-FORCE_INLINE __m128i _mm_add_epi8(__m128i a, __m128i b)2211
-{2212
- return vreinterpretq_m128i_s8(2213
- vaddq_s8(vreinterpretq_s8_m128i(a), vreinterpretq_s8_m128i(b)));2214
-}2215
-2216
-// Adds the 8 signed 16-bit integers in a to the 8 signed 16-bit integers in b2217
-// and saturates.2218
-//2219
-// r0 := SignedSaturate(a0 + b0)2220
-// r1 := SignedSaturate(a1 + b1)2221
-// ...2222
-// r7 := SignedSaturate(a7 + b7)2223
-//2224
-// https://msdn.microsoft.com/en-us/library/1a306ef8(v=vs.100).aspx2225
-FORCE_INLINE __m128i _mm_adds_epi16(__m128i a, __m128i b)2226
-{2227
- return vreinterpretq_m128i_s16(vqaddq_s16(vreinterpretq_s16_m128i(a),2228
- vreinterpretq_s16_m128i(b)));2229
-}2230
-2231
-// Adds the 16 unsigned 8-bit integers in a to the 16 unsigned 8-bit integers in2232
-// b and saturates..2233
-// https://msdn.microsoft.com/en-us/library/9hahyddy(v=vs.100).aspx2234
-FORCE_INLINE __m128i _mm_adds_epu8(__m128i a, __m128i b)2235
-{2236
- return vreinterpretq_m128i_u8(vqaddq_u8(vreinterpretq_u8_m128i(a),2237
- vreinterpretq_u8_m128i(b)));2238
-}2239
-2240
-// Multiplies the 8 signed or unsigned 16-bit integers from a by the 8 signed or2241
-// unsigned 16-bit integers from b.2242
-//2243
-// r0 := (a0 * b0)[15:0]2244
-// r1 := (a1 * b1)[15:0]2245
-// ...2246
-// r7 := (a7 * b7)[15:0]2247
-//2248
-// https://msdn.microsoft.com/en-us/library/vstudio/9ks1472s(v=vs.100).aspx2249
-FORCE_INLINE __m128i _mm_mullo_epi16(__m128i a, __m128i b)2250
-{2251
- return vreinterpretq_m128i_s16(vmulq_s16(vreinterpretq_s16_m128i(a),2252
- vreinterpretq_s16_m128i(b)));2253
-}2254
-2255
-// Multiplies the 4 signed or unsigned 32-bit integers from a by the 4 signed or2256
-// unsigned 32-bit integers from b.2257
-// https://msdn.microsoft.com/en-us/library/vstudio/bb531409(v=vs.100).aspx2258
-FORCE_INLINE __m128i _mm_mullo_epi32(__m128i a, __m128i b)2259
-{2260
- return vreinterpretq_m128i_s32(vmulq_s32(vreinterpretq_s32_m128i(a),2261
- vreinterpretq_s32_m128i(b)));2262
-}2263
-2264
-// Multiplies the four single-precision, floating-point values of a and b.2265
-//2266
-// r0 := a0 * b02267
-// r1 := a1 * b12268
-// r2 := a2 * b22269
-// r3 := a3 * b32270
-//2271
-// https://msdn.microsoft.com/en-us/library/vstudio/22kbk6t9(v=vs.100).aspx2272
-FORCE_INLINE __m128 _mm_mul_ps(__m128 a, __m128 b)2273
-{2274
- return vreinterpretq_m128_f32(vmulq_f32(vreinterpretq_f32_m128(a),2275
- vreinterpretq_f32_m128(b)));2276
-}2277
-2278
-// Multiply the low unsigned 32-bit integers from each packed 64-bit element in2279
-// a and b, and store the unsigned 64-bit results in dst.2280
-//2281
-// r0 := (a0 & 0xFFFFFFFF) * (b0 & 0xFFFFFFFF)2282
-// r1 := (a2 & 0xFFFFFFFF) * (b2 & 0xFFFFFFFF)2283
-FORCE_INLINE __m128i _mm_mul_epu32(__m128i a, __m128i b)2284
-{2285
- // vmull_u32 upcasts instead of masking, so we downcast.2286
- uint32x2_t a_lo = vmovn_u64(vreinterpretq_u64_m128i(a));2287
- uint32x2_t b_lo = vmovn_u64(vreinterpretq_u64_m128i(b));2288
- return vreinterpretq_m128i_u64(vmull_u32(a_lo, b_lo));2289
-}2290
-2291
-// Multiply the low signed 32-bit integers from each packed 64-bit element in2292
-// a and b, and store the signed 64-bit results in dst.2293
-//2294
-// r0 := (int64_t)(int32_t)a0 * (int64_t)(int32_t)b02295
-// r1 := (int64_t)(int32_t)a2 * (int64_t)(int32_t)b22296
-FORCE_INLINE __m128i _mm_mul_epi32(__m128i a, __m128i b)2297
-{2298
- // vmull_s32 upcasts instead of masking, so we downcast.2299
- int32x2_t a_lo = vmovn_s64(vreinterpretq_s64_m128i(a));2300
- int32x2_t b_lo = vmovn_s64(vreinterpretq_s64_m128i(b));2301
- return vreinterpretq_m128i_s64(vmull_s32(a_lo, b_lo));2302
-}2303
-2304
-// Multiplies the 8 signed 16-bit integers from a by the 8 signed 16-bit2305
-// integers from b.2306
-//2307
-// r0 := (a0 * b0) + (a1 * b1)2308
-// r1 := (a2 * b2) + (a3 * b3)2309
-// r2 := (a4 * b4) + (a5 * b5)2310
-// r3 := (a6 * b6) + (a7 * b7)2311
-// https://msdn.microsoft.com/en-us/library/yht36sa6(v=vs.90).aspx2312
-FORCE_INLINE __m128i _mm_madd_epi16(__m128i a, __m128i b)2313
-{2314
- int32x4_t low = vmull_s16(vget_low_s16(vreinterpretq_s16_m128i(a)),2315
- vget_low_s16(vreinterpretq_s16_m128i(b)));2316
- int32x4_t high = vmull_s16(vget_high_s16(vreinterpretq_s16_m128i(a)),2317
- vget_high_s16(vreinterpretq_s16_m128i(b)));2318
-2319
- int32x2_t low_sum = vpadd_s32(vget_low_s32(low), vget_high_s32(low));2320
- int32x2_t high_sum = vpadd_s32(vget_low_s32(high), vget_high_s32(high));2321
-2322
- return vreinterpretq_m128i_s32(vcombine_s32(low_sum, high_sum));2323
-}2324
-2325
-// Multiply packed signed 16-bit integers in a and b, producing intermediate2326
-// signed 32-bit integers. Shift right by 15 bits while rounding up, and store2327
-// the packed 16-bit integers in dst.2328
-//2329
-// r0 := Round(((int32_t)a0 * (int32_t)b0) >> 15)2330
-// r1 := Round(((int32_t)a1 * (int32_t)b1) >> 15)2331
-// r2 := Round(((int32_t)a2 * (int32_t)b2) >> 15)2332
-// ...2333
-// r7 := Round(((int32_t)a7 * (int32_t)b7) >> 15)2334
-FORCE_INLINE __m128i _mm_mulhrs_epi16(__m128i a, __m128i b)2335
-{2336
- // Has issues due to saturation2337
- // return vreinterpretq_m128i_s16(vqrdmulhq_s16(a, b));2338
-2339
- // Multiply2340
- int32x4_t mul_lo = vmull_s16(vget_low_s16(vreinterpretq_s16_m128i(a)),2341
- vget_low_s16(vreinterpretq_s16_m128i(b)));2342
- int32x4_t mul_hi = vmull_s16(vget_high_s16(vreinterpretq_s16_m128i(a)),2343
- vget_high_s16(vreinterpretq_s16_m128i(b)));2344
-2345
- // Rounding narrowing shift right2346
- // narrow = (int16_t)((mul + 16384) >> 15);2347
- int16x4_t narrow_lo = vrshrn_n_s32(mul_lo, 15);2348
- int16x4_t narrow_hi = vrshrn_n_s32(mul_hi, 15);2349
-2350
- // Join together2351
- return vreinterpretq_m128i_s16(vcombine_s16(narrow_lo, narrow_hi));2352
-}2353
-2354
-// Vertically multiply each unsigned 8-bit integer from a with the corresponding2355
-// signed 8-bit integer from b, producing intermediate signed 16-bit integers.2356
-// Horizontally add adjacent pairs of intermediate signed 16-bit integers,2357
-// and pack the saturated results in dst.2358
-//2359
-// FOR j := 0 to 72360
-// i := j*162361
-// dst[i+15:i] := Saturate_To_Int16( a[i+15:i+8]*b[i+15:i+8] +2362
-// a[i+7:i]*b[i+7:i] )2363
-// ENDFOR2364
-FORCE_INLINE __m128i _mm_maddubs_epi16(__m128i _a, __m128i _b)2365
-{2366
- // This would be much simpler if x86 would choose to zero extend OR sign2367
- // extend, not both. This could probably be optimized better.2368
- uint16x8_t a = vreinterpretq_u16_m128i(_a);2369
- int16x8_t b = vreinterpretq_s16_m128i(_b);2370
-2371
- // Zero extend a2372
- int16x8_t a_odd = vreinterpretq_s16_u16(vshrq_n_u16(a, 8));2373
- int16x8_t a_even =2374
- vreinterpretq_s16_u16(vbicq_u16(a, vdupq_n_u16(0xff00)));2375
-2376
- // Sign extend by shifting left then shifting right.2377
- int16x8_t b_even = vshrq_n_s16(vshlq_n_s16(b, 8), 8);2378
- int16x8_t b_odd = vshrq_n_s16(b, 8);2379
-2380
- // multiply2381
- int16x8_t prod1 = vmulq_s16(a_even, b_even);2382
- int16x8_t prod2 = vmulq_s16(a_odd, b_odd);2383
-2384
- // saturated add2385
- return vreinterpretq_m128i_s16(vqaddq_s16(prod1, prod2));2386
-}2387
-2388
-// Computes the absolute difference of the 16 unsigned 8-bit integers from a2389
-// and the 16 unsigned 8-bit integers from b.2390
-//2391
-// Return Value2392
-// Sums the upper 8 differences and lower 8 differences and packs the2393
-// resulting 2 unsigned 16-bit integers into the upper and lower 64-bit2394
-// elements.2395
-//2396
-// r0 := abs(a0 - b0) + abs(a1 - b1) +...+ abs(a7 - b7)2397
-// r1 := 0x02398
-// r2 := 0x02399
-// r3 := 0x02400
-// r4 := abs(a8 - b8) + abs(a9 - b9) +...+ abs(a15 - b15)2401
-// r5 := 0x02402
-// r6 := 0x02403
-// r7 := 0x02404
-FORCE_INLINE __m128i _mm_sad_epu8(__m128i a, __m128i b)2405
-{2406
- uint16x8_t t = vpaddlq_u8(vabdq_u8((uint8x16_t)a, (uint8x16_t)b));2407
- uint16_t r0 = t[0] + t[1] + t[2] + t[3];2408
- uint16_t r4 = t[4] + t[5] + t[6] + t[7];2409
- uint16x8_t r = vsetq_lane_u16(r0, vdupq_n_u16(0), 0);2410
- return (__m128i)vsetq_lane_u16(r4, r, 4);2411
-}2412
-2413
-// Divides the four single-precision, floating-point values of a and b.2414
-//2415
-// r0 := a0 / b02416
-// r1 := a1 / b12417
-// r2 := a2 / b22418
-// r3 := a3 / b32419
-//2420
-// https://msdn.microsoft.com/en-us/library/edaw8147(v=vs.100).aspx2421
-FORCE_INLINE __m128 _mm_div_ps(__m128 a, __m128 b)2422
-{2423
- float32x4_t recip0 = vrecpeq_f32(vreinterpretq_f32_m128(b));2424
- float32x4_t recip1 = vmulq_f32(2425
- recip0, vrecpsq_f32(recip0, vreinterpretq_f32_m128(b)));2426
- return vreinterpretq_m128_f32(2427
- vmulq_f32(vreinterpretq_f32_m128(a), recip1));2428
-}2429
-2430
-// Divides the scalar single-precision floating point value of a by b.2431
-// https://msdn.microsoft.com/en-us/library/4y73xa49(v=vs.100).aspx2432
-FORCE_INLINE __m128 _mm_div_ss(__m128 a, __m128 b)2433
-{2434
- float32_t value =2435
- vgetq_lane_f32(vreinterpretq_f32_m128(_mm_div_ps(a, b)), 0);2436
- return vreinterpretq_m128_f32(2437
- vsetq_lane_f32(value, vreinterpretq_f32_m128(a), 0));2438
-}2439
-2440
-// Computes the approximations of reciprocals of the four single-precision,2441
-// floating-point values of a.2442
-// https://msdn.microsoft.com/en-us/library/vstudio/796k1tty(v=vs.100).aspx2443
-FORCE_INLINE __m128 _mm_rcp_ps(__m128 in)2444
-{2445
- float32x4_t recip = vrecpeq_f32(vreinterpretq_f32_m128(in));2446
- recip = vmulq_f32(recip,2447
- vrecpsq_f32(recip, vreinterpretq_f32_m128(in)));2448
- return vreinterpretq_m128_f32(recip);2449
-}2450
-2451
-// Computes the approximations of square roots of the four single-precision,2452
-// floating-point values of a. First computes reciprocal square roots and then2453
-// reciprocals of the four values.2454
-//2455
-// r0 := sqrt(a0)2456
-// r1 := sqrt(a1)2457
-// r2 := sqrt(a2)2458
-// r3 := sqrt(a3)2459
-//2460
-// https://msdn.microsoft.com/en-us/library/vstudio/8z67bwwk(v=vs.100).aspx2461
-FORCE_INLINE __m128 _mm_sqrt_ps(__m128 in)2462
-{2463
-#if defined(__aarch64__)2464
- return vreinterpretq_m128_f32(vsqrtq_f32(vreinterpretq_f32_m128(in)));2465
-#else2466
- float32x4_t recipsq = vrsqrteq_f32(vreinterpretq_f32_m128(in));2467
- float32x4_t sq = vrecpeq_f32(recipsq);2468
- // ??? use step versions of both sqrt and recip for better accuracy?2469
- return vreinterpretq_m128_f32(sq);2470
-#endif2471
-}2472
-2473
-// Computes the approximation of the square root of the scalar single-precision2474
-// floating point value of in.2475
-// https://msdn.microsoft.com/en-us/library/ahfsc22d(v=vs.100).aspx2476
-FORCE_INLINE __m128 _mm_sqrt_ss(__m128 in)2477
-{2478
- float32_t value =2479
- vgetq_lane_f32(vreinterpretq_f32_m128(_mm_sqrt_ps(in)), 0);2480
- return vreinterpretq_m128_f32(2481
- vsetq_lane_f32(value, vreinterpretq_f32_m128(in), 0));2482
-}2483
-2484
-// Computes the approximations of the reciprocal square roots of the four2485
-// single-precision floating point values of in.2486
-// https://msdn.microsoft.com/en-us/library/22hfsh53(v=vs.100).aspx2487
-FORCE_INLINE __m128 _mm_rsqrt_ps(__m128 in)2488
-{2489
- return vreinterpretq_m128_f32(vrsqrteq_f32(vreinterpretq_f32_m128(in)));2490
-}2491
-2492
-// Compute the approximate reciprocal square root of the lower single-precision2493
-// (32-bit) floating-point element in a, store the result in the lower element2494
-// of dst, and copy the upper 3 packed elements from a to the upper elements of2495
-// dst.2496
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_rsqrt_ss2497
-FORCE_INLINE __m128 _mm_rsqrt_ss(__m128 in)2498
-{2499
- return vsetq_lane_f32(vgetq_lane_f32(_mm_rsqrt_ps(in), 0), in, 0);2500
-}2501
-2502
-// Computes the maximums of the four single-precision, floating-point values of2503
-// a and b.2504
-// https://msdn.microsoft.com/en-us/library/vstudio/ff5d607a(v=vs.100).aspx2505
-FORCE_INLINE __m128 _mm_max_ps(__m128 a, __m128 b)2506
-{2507
- return vreinterpretq_m128_f32(vmaxq_f32(vreinterpretq_f32_m128(a),2508
- vreinterpretq_f32_m128(b)));2509
-}2510
-2511
-// Computes the minima of the four single-precision, floating-point values of a2512
-// and b.2513
-// https://msdn.microsoft.com/en-us/library/vstudio/wh13kadz(v=vs.100).aspx2514
-FORCE_INLINE __m128 _mm_min_ps(__m128 a, __m128 b)2515
-{2516
- return vreinterpretq_m128_f32(vminq_f32(vreinterpretq_f32_m128(a),2517
- vreinterpretq_f32_m128(b)));2518
-}2519
-2520
-// Computes the maximum of the two lower scalar single-precision floating point2521
-// values of a and b.2522
-// https://msdn.microsoft.com/en-us/library/s6db5esz(v=vs.100).aspx2523
-FORCE_INLINE __m128 _mm_max_ss(__m128 a, __m128 b)2524
-{2525
- float32_t value = vgetq_lane_f32(vmaxq_f32(vreinterpretq_f32_m128(a),2526
- vreinterpretq_f32_m128(b)),2527
- 0);2528
- return vreinterpretq_m128_f32(2529
- vsetq_lane_f32(value, vreinterpretq_f32_m128(a), 0));2530
-}2531
-2532
-// Computes the minimum of the two lower scalar single-precision floating point2533
-// values of a and b.2534
-// https://msdn.microsoft.com/en-us/library/0a9y7xaa(v=vs.100).aspx2535
-FORCE_INLINE __m128 _mm_min_ss(__m128 a, __m128 b)2536
-{2537
- float32_t value = vgetq_lane_f32(vminq_f32(vreinterpretq_f32_m128(a),2538
- vreinterpretq_f32_m128(b)),2539
- 0);2540
- return vreinterpretq_m128_f32(2541
- vsetq_lane_f32(value, vreinterpretq_f32_m128(a), 0));2542
-}2543
-2544
-// Computes the pairwise maxima of the 16 unsigned 8-bit integers from a and the2545
-// 16 unsigned 8-bit integers from b.2546
-// https://msdn.microsoft.com/en-us/library/st6634za(v=vs.100).aspx2547
-FORCE_INLINE __m128i _mm_max_epu8(__m128i a, __m128i b)2548
-{2549
- return vreinterpretq_m128i_u8(2550
- vmaxq_u8(vreinterpretq_u8_m128i(a), vreinterpretq_u8_m128i(b)));2551
-}2552
-2553
-// Computes the pairwise minima of the 16 unsigned 8-bit integers from a and the2554
-// 16 unsigned 8-bit integers from b.2555
-// https://msdn.microsoft.com/ko-kr/library/17k8cf58(v=vs.100).aspxx2556
-FORCE_INLINE __m128i _mm_min_epu8(__m128i a, __m128i b)2557
-{2558
- return vreinterpretq_m128i_u8(2559
- vminq_u8(vreinterpretq_u8_m128i(a), vreinterpretq_u8_m128i(b)));2560
-}2561
-2562
-// Computes the pairwise minima of the 8 signed 16-bit integers from a and the 82563
-// signed 16-bit integers from b.2564
-// https://msdn.microsoft.com/en-us/library/vstudio/6te997ew(v=vs.100).aspx2565
-FORCE_INLINE __m128i _mm_min_epi16(__m128i a, __m128i b)2566
-{2567
- return vreinterpretq_m128i_s16(vminq_s16(vreinterpretq_s16_m128i(a),2568
- vreinterpretq_s16_m128i(b)));2569
-}2570
-2571
-// Computes the pairwise maxima of the 8 signed 16-bit integers from a and the 82572
-// signed 16-bit integers from b.2573
-// https://msdn.microsoft.com/en-us/LIBRary/3x060h7c(v=vs.100).aspx2574
-FORCE_INLINE __m128i _mm_max_epi16(__m128i a, __m128i b)2575
-{2576
- return vreinterpretq_m128i_s16(vmaxq_s16(vreinterpretq_s16_m128i(a),2577
- vreinterpretq_s16_m128i(b)));2578
-}2579
-2580
-// epi versions of min/max2581
-// Computes the pariwise maximums of the four signed 32-bit integer values of a2582
-// and b.2583
-//2584
-// A 128-bit parameter that can be defined with the following equations:2585
-// r0 := (a0 > b0) ? a0 : b02586
-// r1 := (a1 > b1) ? a1 : b12587
-// r2 := (a2 > b2) ? a2 : b22588
-// r3 := (a3 > b3) ? a3 : b32589
-//2590
-// https://msdn.microsoft.com/en-us/library/vstudio/bb514055(v=vs.100).aspx2591
-FORCE_INLINE __m128i _mm_max_epi32(__m128i a, __m128i b)2592
-{2593
- return vreinterpretq_m128i_s32(vmaxq_s32(vreinterpretq_s32_m128i(a),2594
- vreinterpretq_s32_m128i(b)));2595
-}2596
-2597
-// Computes the pariwise minima of the four signed 32-bit integer values of a2598
-// and b.2599
-//2600
-// A 128-bit parameter that can be defined with the following equations:2601
-// r0 := (a0 < b0) ? a0 : b02602
-// r1 := (a1 < b1) ? a1 : b12603
-// r2 := (a2 < b2) ? a2 : b22604
-// r3 := (a3 < b3) ? a3 : b32605
-//2606
-// https://msdn.microsoft.com/en-us/library/vstudio/bb531476(v=vs.100).aspx2607
-FORCE_INLINE __m128i _mm_min_epi32(__m128i a, __m128i b)2608
-{2609
- return vreinterpretq_m128i_s32(vminq_s32(vreinterpretq_s32_m128i(a),2610
- vreinterpretq_s32_m128i(b)));2611
-}2612
-2613
-// Multiplies the 8 signed 16-bit integers from a by the 8 signed 16-bit2614
-// integers from b.2615
-//2616
-// r0 := (a0 * b0)[31:16]2617
-// r1 := (a1 * b1)[31:16]2618
-// ...2619
-// r7 := (a7 * b7)[31:16]2620
-//2621
-// https://msdn.microsoft.com/en-us/library/vstudio/59hddw1d(v=vs.100).aspx2622
-FORCE_INLINE __m128i _mm_mulhi_epi16(__m128i a, __m128i b)2623
-{2624
- /* FIXME: issue with large values because of result saturation */2625
- // int16x8_t ret = vqdmulhq_s16(vreinterpretq_s16_m128i(a),2626
- // vreinterpretq_s16_m128i(b)); /* =2*a*b */ return2627
- // vreinterpretq_m128i_s16(vshrq_n_s16(ret, 1));2628
- int16x4_t a3210 = vget_low_s16(vreinterpretq_s16_m128i(a));2629
- int16x4_t b3210 = vget_low_s16(vreinterpretq_s16_m128i(b));2630
- int32x4_t ab3210 = vmull_s16(a3210, b3210); /* 3333222211110000 */2631
- int16x4_t a7654 = vget_high_s16(vreinterpretq_s16_m128i(a));2632
- int16x4_t b7654 = vget_high_s16(vreinterpretq_s16_m128i(b));2633
- int32x4_t ab7654 = vmull_s16(a7654, b7654); /* 7777666655554444 */2634
- uint16x8x2_t r = vuzpq_u16(vreinterpretq_u16_s32(ab3210),2635
- vreinterpretq_u16_s32(ab7654));2636
- return vreinterpretq_m128i_u16(r.val[1]);2637
-}2638
-2639
-// Computes pairwise add of each argument as single-precision, floating-point2640
-// values a and b.2641
-// https://msdn.microsoft.com/en-us/library/yd9wecaa.aspx2642
-FORCE_INLINE __m128 _mm_hadd_ps(__m128 a, __m128 b)2643
-{2644
-#if defined(__aarch64__)2645
- return vreinterpretq_m128_f32(vpaddq_f32(vreinterpretq_f32_m128(a),2646
- vreinterpretq_f32_m128(b)));2647
-#else2648
- float32x2_t a10 = vget_low_f32(vreinterpretq_f32_m128(a));2649
- float32x2_t a32 = vget_high_f32(vreinterpretq_f32_m128(a));2650
- float32x2_t b10 = vget_low_f32(vreinterpretq_f32_m128(b));2651
- float32x2_t b32 = vget_high_f32(vreinterpretq_f32_m128(b));2652
- return vreinterpretq_m128_f32(2653
- vcombine_f32(vpadd_f32(a10, a32), vpadd_f32(b10, b32)));2654
-#endif2655
-}2656
-2657
-// Computes pairwise add of each argument as a 16-bit signed or unsigned integer2658
-// values a and b.2659
-FORCE_INLINE __m128i _mm_hadd_epi16(__m128i _a, __m128i _b)2660
-{2661
- int16x8_t a = vreinterpretq_s16_m128i(_a);2662
- int16x8_t b = vreinterpretq_s16_m128i(_b);2663
-#if defined(__aarch64__)2664
- return vreinterpretq_m128i_s16(vpaddq_s16(a, b));2665
-#else2666
- return vreinterpretq_m128i_s16(2667
- vcombine_s16(vpadd_s16(vget_low_s16(a), vget_high_s16(a)),2668
- vpadd_s16(vget_low_s16(b), vget_high_s16(b))));2669
-#endif2670
-}2671
-2672
-// Computes pairwise difference of each argument as a 16-bit signed or unsigned2673
-// integer values a and b.2674
-FORCE_INLINE __m128i _mm_hsub_epi16(__m128i _a, __m128i _b)2675
-{2676
- int32x4_t a = vreinterpretq_s32_m128i(_a);2677
- int32x4_t b = vreinterpretq_s32_m128i(_b);2678
- // Interleave using vshrn/vmovn2679
- // [a0|a2|a4|a6|b0|b2|b4|b6]2680
- // [a1|a3|a5|a7|b1|b3|b5|b7]2681
- int16x8_t ab0246 = vcombine_s16(vmovn_s32(a), vmovn_s32(b));2682
- int16x8_t ab1357 = vcombine_s16(vshrn_n_s32(a, 16), vshrn_n_s32(b, 16));2683
- // Subtract2684
- return vreinterpretq_m128i_s16(vsubq_s16(ab0246, ab1357));2685
-}2686
-2687
-// Computes saturated pairwise sub of each argument as a 16-bit signed2688
-// integer values a and b.2689
-FORCE_INLINE __m128i _mm_hadds_epi16(__m128i _a, __m128i _b)2690
-{2691
- int32x4_t a = vreinterpretq_s32_m128i(_a);2692
- int32x4_t b = vreinterpretq_s32_m128i(_b);2693
- // Interleave using vshrn/vmovn2694
- // [a0|a2|a4|a6|b0|b2|b4|b6]2695
- // [a1|a3|a5|a7|b1|b3|b5|b7]2696
- int16x8_t ab0246 = vcombine_s16(vmovn_s32(a), vmovn_s32(b));2697
- int16x8_t ab1357 = vcombine_s16(vshrn_n_s32(a, 16), vshrn_n_s32(b, 16));2698
- // Saturated add2699
- return vreinterpretq_m128i_s16(vqaddq_s16(ab0246, ab1357));2700
-}2701
-2702
-// Computes saturated pairwise difference of each argument as a 16-bit signed2703
-// integer values a and b.2704
-FORCE_INLINE __m128i _mm_hsubs_epi16(__m128i _a, __m128i _b)2705
-{2706
- int32x4_t a = vreinterpretq_s32_m128i(_a);2707
- int32x4_t b = vreinterpretq_s32_m128i(_b);2708
- // Interleave using vshrn/vmovn2709
- // [a0|a2|a4|a6|b0|b2|b4|b6]2710
- // [a1|a3|a5|a7|b1|b3|b5|b7]2711
- int16x8_t ab0246 = vcombine_s16(vmovn_s32(a), vmovn_s32(b));2712
- int16x8_t ab1357 = vcombine_s16(vshrn_n_s32(a, 16), vshrn_n_s32(b, 16));2713
- // Saturated subtract2714
- return vreinterpretq_m128i_s16(vqsubq_s16(ab0246, ab1357));2715
-}2716
-2717
-// Computes pairwise add of each argument as a 32-bit signed or unsigned integer2718
-// values a and b.2719
-FORCE_INLINE __m128i _mm_hadd_epi32(__m128i _a, __m128i _b)2720
-{2721
- int32x4_t a = vreinterpretq_s32_m128i(_a);2722
- int32x4_t b = vreinterpretq_s32_m128i(_b);2723
- return vreinterpretq_m128i_s32(2724
- vcombine_s32(vpadd_s32(vget_low_s32(a), vget_high_s32(a)),2725
- vpadd_s32(vget_low_s32(b), vget_high_s32(b))));2726
-}2727
-2728
-// Computes pairwise difference of each argument as a 32-bit signed or unsigned2729
-// integer values a and b.2730
-FORCE_INLINE __m128i _mm_hsub_epi32(__m128i _a, __m128i _b)2731
-{2732
- int64x2_t a = vreinterpretq_s64_m128i(_a);2733
- int64x2_t b = vreinterpretq_s64_m128i(_b);2734
- // Interleave using vshrn/vmovn2735
- // [a0|a2|b0|b2]2736
- // [a1|a2|b1|b3]2737
- int32x4_t ab02 = vcombine_s32(vmovn_s64(a), vmovn_s64(b));2738
- int32x4_t ab13 = vcombine_s32(vshrn_n_s64(a, 32), vshrn_n_s64(b, 32));2739
- // Subtract2740
- return vreinterpretq_m128i_s32(vsubq_s32(ab02, ab13));2741
-}2742
-2743
-/* Compare operations */2744
-2745
-// Compares for less than2746
-// https://msdn.microsoft.com/en-us/library/vstudio/f330yhc8(v=vs.100).aspx2747
-FORCE_INLINE __m128 _mm_cmplt_ps(__m128 a, __m128 b)2748
-{2749
- return vreinterpretq_m128_u32(vcltq_f32(vreinterpretq_f32_m128(a),2750
- vreinterpretq_f32_m128(b)));2751
-}2752
-2753
-// Compares for greater than.2754
-//2755
-// r0 := (a0 > b0) ? 0xffffffff : 0x02756
-// r1 := (a1 > b1) ? 0xffffffff : 0x02757
-// r2 := (a2 > b2) ? 0xffffffff : 0x02758
-// r3 := (a3 > b3) ? 0xffffffff : 0x02759
-//2760
-// https://msdn.microsoft.com/en-us/library/vstudio/11dy102s(v=vs.100).aspx2761
-FORCE_INLINE __m128 _mm_cmpgt_ps(__m128 a, __m128 b)2762
-{2763
- return vreinterpretq_m128_u32(vcgtq_f32(vreinterpretq_f32_m128(a),2764
- vreinterpretq_f32_m128(b)));2765
-}2766
-2767
-// Compares for greater than or equal.2768
-// https://msdn.microsoft.com/en-us/library/vstudio/fs813y2t(v=vs.100).aspx2769
-FORCE_INLINE __m128 _mm_cmpge_ps(__m128 a, __m128 b)2770
-{2771
- return vreinterpretq_m128_u32(vcgeq_f32(vreinterpretq_f32_m128(a),2772
- vreinterpretq_f32_m128(b)));2773
-}2774
-2775
-// Compares for less than or equal.2776
-//2777
-// r0 := (a0 <= b0) ? 0xffffffff : 0x02778
-// r1 := (a1 <= b1) ? 0xffffffff : 0x02779
-// r2 := (a2 <= b2) ? 0xffffffff : 0x02780
-// r3 := (a3 <= b3) ? 0xffffffff : 0x02781
-//2782
-// https://msdn.microsoft.com/en-us/library/vstudio/1s75w83z(v=vs.100).aspx2783
-FORCE_INLINE __m128 _mm_cmple_ps(__m128 a, __m128 b)2784
-{2785
- return vreinterpretq_m128_u32(vcleq_f32(vreinterpretq_f32_m128(a),2786
- vreinterpretq_f32_m128(b)));2787
-}2788
-2789
-// Compares for equality.2790
-// https://msdn.microsoft.com/en-us/library/vstudio/36aectz5(v=vs.100).aspx2791
-FORCE_INLINE __m128 _mm_cmpeq_ps(__m128 a, __m128 b)2792
-{2793
- return vreinterpretq_m128_u32(vceqq_f32(vreinterpretq_f32_m128(a),2794
- vreinterpretq_f32_m128(b)));2795
-}2796
-2797
-// Compares the 16 signed or unsigned 8-bit integers in a and the 16 signed or2798
-// unsigned 8-bit integers in b for equality.2799
-// https://msdn.microsoft.com/en-us/library/windows/desktop/bz5xk21a(v=vs.90).aspx2800
-FORCE_INLINE __m128i _mm_cmpeq_epi8(__m128i a, __m128i b)2801
-{2802
- return vreinterpretq_m128i_u8(2803
- vceqq_s8(vreinterpretq_s8_m128i(a), vreinterpretq_s8_m128i(b)));2804
-}2805
-2806
-// Compares the 8 signed or unsigned 16-bit integers in a and the 8 signed or2807
-// unsigned 16-bit integers in b for equality.2808
-// https://msdn.microsoft.com/en-us/library/2ay060te(v=vs.100).aspx2809
-FORCE_INLINE __m128i _mm_cmpeq_epi16(__m128i a, __m128i b)2810
-{2811
- return vreinterpretq_m128i_u16(vceqq_s16(vreinterpretq_s16_m128i(a),2812
- vreinterpretq_s16_m128i(b)));2813
-}2814
-2815
-// Compare packed 32-bit integers in a and b for equality, and store the results2816
-// in dst2817
-FORCE_INLINE __m128i _mm_cmpeq_epi32(__m128i a, __m128i b)2818
-{2819
- return vreinterpretq_m128i_u32(vceqq_s32(vreinterpretq_s32_m128i(a),2820
- vreinterpretq_s32_m128i(b)));2821
-}2822
-2823
-// Compare packed 64-bit integers in a and b for equality, and store the results2824
-// in dst2825
-FORCE_INLINE __m128i _mm_cmpeq_epi64(__m128i a, __m128i b)2826
-{2827
-#if defined(__aarch64__)2828
- return vreinterpretq_m128i_u64(vceqq_u64(vreinterpretq_u64_m128i(a),2829
- vreinterpretq_u64_m128i(b)));2830
-#else2831
- // ARMv7 lacks vceqq_u642832
- // (a == b) -> (a_lo == b_lo) && (a_hi == b_hi)2833
- uint32x4_t cmp = vceqq_u32(vreinterpretq_u32_m128i(a),2834
- vreinterpretq_u32_m128i(b));2835
- uint32x4_t swapped = vrev64q_u32(cmp);2836
- return vreinterpretq_m128i_u32(vandq_u32(cmp, swapped));2837
-#endif2838
-}2839
-2840
-// Compares the 16 signed 8-bit integers in a and the 16 signed 8-bit integers2841
-// in b for lesser than.2842
-// https://msdn.microsoft.com/en-us/library/windows/desktop/9s46csht(v=vs.90).aspx2843
-FORCE_INLINE __m128i _mm_cmplt_epi8(__m128i a, __m128i b)2844
-{2845
- return vreinterpretq_m128i_u8(2846
- vcltq_s8(vreinterpretq_s8_m128i(a), vreinterpretq_s8_m128i(b)));2847
-}2848
-2849
-// Compares the 16 signed 8-bit integers in a and the 16 signed 8-bit integers2850
-// in b for greater than.2851
-//2852
-// r0 := (a0 > b0) ? 0xff : 0x02853
-// r1 := (a1 > b1) ? 0xff : 0x02854
-// ...2855
-// r15 := (a15 > b15) ? 0xff : 0x02856
-//2857
-// https://msdn.microsoft.com/zh-tw/library/wf45zt2b(v=vs.100).aspx2858
-FORCE_INLINE __m128i _mm_cmpgt_epi8(__m128i a, __m128i b)2859
-{2860
- return vreinterpretq_m128i_u8(2861
- vcgtq_s8(vreinterpretq_s8_m128i(a), vreinterpretq_s8_m128i(b)));2862
-}2863
-2864
-// Compares the 8 signed 16-bit integers in a and the 8 signed 16-bit integers2865
-// in b for less than.2866
-//2867
-// r0 := (a0 < b0) ? 0xffff : 0x02868
-// r1 := (a1 < b1) ? 0xffff : 0x02869
-// ...2870
-// r7 := (a7 < b7) ? 0xffff : 0x02871
-//2872
-// https://technet.microsoft.com/en-us/library/t863edb2(v=vs.100).aspx2873
-FORCE_INLINE __m128i _mm_cmplt_epi16(__m128i a, __m128i b)2874
-{2875
- return vreinterpretq_m128i_u16(vcltq_s16(vreinterpretq_s16_m128i(a),2876
- vreinterpretq_s16_m128i(b)));2877
-}2878
-2879
-// Compares the 8 signed 16-bit integers in a and the 8 signed 16-bit integers2880
-// in b for greater than.2881
-//2882
-// r0 := (a0 > b0) ? 0xffff : 0x02883
-// r1 := (a1 > b1) ? 0xffff : 0x02884
-// ...2885
-// r7 := (a7 > b7) ? 0xffff : 0x02886
-//2887
-// https://technet.microsoft.com/en-us/library/xd43yfsa(v=vs.100).aspx2888
-FORCE_INLINE __m128i _mm_cmpgt_epi16(__m128i a, __m128i b)2889
-{2890
- return vreinterpretq_m128i_u16(vcgtq_s16(vreinterpretq_s16_m128i(a),2891
- vreinterpretq_s16_m128i(b)));2892
-}2893
-2894
-// Compares the 4 signed 32-bit integers in a and the 4 signed 32-bit integers2895
-// in b for less than.2896
-// https://msdn.microsoft.com/en-us/library/vstudio/4ak0bf5d(v=vs.100).aspx2897
-FORCE_INLINE __m128i _mm_cmplt_epi32(__m128i a, __m128i b)2898
-{2899
- return vreinterpretq_m128i_u32(vcltq_s32(vreinterpretq_s32_m128i(a),2900
- vreinterpretq_s32_m128i(b)));2901
-}2902
-2903
-// Compares the 4 signed 32-bit integers in a and the 4 signed 32-bit integers2904
-// in b for greater than.2905
-// https://msdn.microsoft.com/en-us/library/vstudio/1s9f2z0y(v=vs.100).aspx2906
-FORCE_INLINE __m128i _mm_cmpgt_epi32(__m128i a, __m128i b)2907
-{2908
- return vreinterpretq_m128i_u32(vcgtq_s32(vreinterpretq_s32_m128i(a),2909
- vreinterpretq_s32_m128i(b)));2910
-}2911
-2912
-// Compares the 2 signed 64-bit integers in a and the 2 signed 64-bit integers2913
-// in b for greater than.2914
-FORCE_INLINE __m128i _mm_cmpgt_epi64(__m128i a, __m128i b)2915
-{2916
-#if defined(__aarch64__)2917
- return vreinterpretq_m128i_u64(vcgtq_s64(vreinterpretq_s64_m128i(a),2918
- vreinterpretq_s64_m128i(b)));2919
-#else2920
- // ARMv7 lacks vcgtq_s64.2921
- // This is based off of Clang's SSE2 polyfill:2922
- // (a > b) -> ((a_hi > b_hi) || (a_lo > b_lo && a_hi == b_hi))2923
-2924
- // Mask the sign bit out since we need a signed AND an unsigned comparison2925
- // and it is ugly to try and split them.2926
- int32x4_t mask = vreinterpretq_s32_s64(vdupq_n_s64(0x80000000ull));2927
- int32x4_t a_mask = veorq_s32(vreinterpretq_s32_m128i(a), mask);2928
- int32x4_t b_mask = veorq_s32(vreinterpretq_s32_m128i(b), mask);2929
- // Check if a > b2930
- int64x2_t greater = vreinterpretq_s64_u32(vcgtq_s32(a_mask, b_mask));2931
- // Copy upper mask to lower mask2932
- // a_hi > b_hi2933
- int64x2_t gt_hi = vshrq_n_s64(greater, 63);2934
- // Copy lower mask to upper mask2935
- // a_lo > b_lo2936
- int64x2_t gt_lo = vsliq_n_s64(greater, greater, 32);2937
- // Compare for equality2938
- int64x2_t equal = vreinterpretq_s64_u32(vceqq_s32(a_mask, b_mask));2939
- // Copy upper mask to lower mask2940
- // a_hi == b_hi2941
- int64x2_t eq_hi = vshrq_n_s64(equal, 63);2942
- // a_hi > b_hi || (a_lo > b_lo && a_hi == b_hi)2943
- int64x2_t ret = vorrq_s64(gt_hi, vandq_s64(gt_lo, eq_hi));2944
- return vreinterpretq_m128i_s64(ret);2945
-#endif2946
-}2947
-2948
-// Compares the four 32-bit floats in a and b to check if any values are NaN.2949
-// Ordered compare between each value returns true for "orderable" and false for2950
-// "not orderable" (NaN).2951
-// https://msdn.microsoft.com/en-us/library/vstudio/0h9w00fx(v=vs.100).aspx see2952
-// also:2953
-// http://stackoverflow.com/questions/8627331/what-does-ordered-unordered-comparison-mean2954
-// http://stackoverflow.com/questions/29349621/neon-isnanval-intrinsics2955
-FORCE_INLINE __m128 _mm_cmpord_ps(__m128 a, __m128 b)2956
-{2957
- // Note: NEON does not have ordered compare builtin2958
- // Need to compare a eq a and b eq b to check for NaN2959
- // Do AND of results to get final2960
- uint32x4_t ceqaa =2961
- vceqq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(a));2962
- uint32x4_t ceqbb =2963
- vceqq_f32(vreinterpretq_f32_m128(b), vreinterpretq_f32_m128(b));2964
- return vreinterpretq_m128_u32(vandq_u32(ceqaa, ceqbb));2965
-}2966
-2967
-// Compares the lower single-precision floating point scalar values of a and b2968
-// using a less than operation. :2969
-// https://msdn.microsoft.com/en-us/library/2kwe606b(v=vs.90).aspx Important2970
-// note!! The documentation on MSDN is incorrect! If either of the values is a2971
-// NAN the docs say you will get a one, but in fact, it will return a zero!!2972
-FORCE_INLINE int _mm_comilt_ss(__m128 a, __m128 b)2973
-{2974
- uint32x4_t a_not_nan =2975
- vceqq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(a));2976
- uint32x4_t b_not_nan =2977
- vceqq_f32(vreinterpretq_f32_m128(b), vreinterpretq_f32_m128(b));2978
- uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);2979
- uint32x4_t a_lt_b =2980
- vcltq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(b));2981
- return (vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_lt_b), 0) != 0) ? 12982
- : 0;2983
-}2984
-2985
-// Compares the lower single-precision floating point scalar values of a and b2986
-// using a greater than operation. :2987
-// https://msdn.microsoft.com/en-us/library/b0738e0t(v=vs.100).aspx2988
-FORCE_INLINE int _mm_comigt_ss(__m128 a, __m128 b)2989
-{2990
- // return vgetq_lane_u32(vcgtq_f32(vreinterpretq_f32_m128(a),2991
- // vreinterpretq_f32_m128(b)), 0);2992
- uint32x4_t a_not_nan =2993
- vceqq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(a));2994
- uint32x4_t b_not_nan =2995
- vceqq_f32(vreinterpretq_f32_m128(b), vreinterpretq_f32_m128(b));2996
- uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);2997
- uint32x4_t a_gt_b =2998
- vcgtq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(b));2999
- return (vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_gt_b), 0) != 0) ? 13000
- : 0;3001
-}3002
-3003
-// Compares the lower single-precision floating point scalar values of a and b3004
-// using a less than or equal operation. :3005
-// https://msdn.microsoft.com/en-us/library/1w4t7c57(v=vs.90).aspx3006
-FORCE_INLINE int _mm_comile_ss(__m128 a, __m128 b)3007
-{3008
- // return vgetq_lane_u32(vcleq_f32(vreinterpretq_f32_m128(a),3009
- // vreinterpretq_f32_m128(b)), 0);3010
- uint32x4_t a_not_nan =3011
- vceqq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(a));3012
- uint32x4_t b_not_nan =3013
- vceqq_f32(vreinterpretq_f32_m128(b), vreinterpretq_f32_m128(b));3014
- uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);3015
- uint32x4_t a_le_b =3016
- vcleq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(b));3017
- return (vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_le_b), 0) != 0) ? 13018
- : 0;3019
-}3020
-3021
-// Compares the lower single-precision floating point scalar values of a and b3022
-// using a greater than or equal operation. :3023
-// https://msdn.microsoft.com/en-us/library/8t80des6(v=vs.100).aspx3024
-FORCE_INLINE int _mm_comige_ss(__m128 a, __m128 b)3025
-{3026
- // return vgetq_lane_u32(vcgeq_f32(vreinterpretq_f32_m128(a),3027
- // vreinterpretq_f32_m128(b)), 0);3028
- uint32x4_t a_not_nan =3029
- vceqq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(a));3030
- uint32x4_t b_not_nan =3031
- vceqq_f32(vreinterpretq_f32_m128(b), vreinterpretq_f32_m128(b));3032
- uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);3033
- uint32x4_t a_ge_b =3034
- vcgeq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(b));3035
- return (vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_ge_b), 0) != 0) ? 13036
- : 0;3037
-}3038
-3039
-// Compares the lower single-precision floating point scalar values of a and b3040
-// using an equality operation. :3041
-// https://msdn.microsoft.com/en-us/library/93yx2h2b(v=vs.100).aspx3042
-FORCE_INLINE int _mm_comieq_ss(__m128 a, __m128 b)3043
-{3044
- // return vgetq_lane_u32(vceqq_f32(vreinterpretq_f32_m128(a),3045
- // vreinterpretq_f32_m128(b)), 0);3046
- uint32x4_t a_not_nan =3047
- vceqq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(a));3048
- uint32x4_t b_not_nan =3049
- vceqq_f32(vreinterpretq_f32_m128(b), vreinterpretq_f32_m128(b));3050
- uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);3051
- uint32x4_t a_eq_b =3052
- vceqq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(b));3053
- return (vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_eq_b), 0) != 0) ? 13054
- : 0;3055
-}3056
-3057
-// Compares the lower single-precision floating point scalar values of a and b3058
-// using an inequality operation. :3059
-// https://msdn.microsoft.com/en-us/library/bafh5e0a(v=vs.90).aspx3060
-FORCE_INLINE int _mm_comineq_ss(__m128 a, __m128 b)3061
-{3062
- // return !vgetq_lane_u32(vceqq_f32(vreinterpretq_f32_m128(a),3063
- // vreinterpretq_f32_m128(b)), 0);3064
- uint32x4_t a_not_nan =3065
- vceqq_f32(vreinterpretq_f32_m128(a), vreinterpretq_f32_m128(a));3066
- uint32x4_t b_not_nan =3067
- vceqq_f32(vreinterpretq_f32_m128(b), vreinterpretq_f32_m128(b));3068
- uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));3069
- uint32x4_t a_neq_b = vmvnq_u32(vceqq_f32(vreinterpretq_f32_m128(a),3070
- vreinterpretq_f32_m128(b)));3071
- return (vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_neq_b), 0) != 0) ? 1 : 0;3072
-}3073
-3074
-// according to the documentation, these intrinsics behave the same as the3075
-// non-'u' versions. We'll just alias them here.3076
-#define _mm_ucomilt_ss _mm_comilt_ss3077
-#define _mm_ucomile_ss _mm_comile_ss3078
-#define _mm_ucomigt_ss _mm_comigt_ss3079
-#define _mm_ucomige_ss _mm_comige_ss3080
-#define _mm_ucomieq_ss _mm_comieq_ss3081
-#define _mm_ucomineq_ss _mm_comineq_ss3082
-3083
-/* Conversions */3084
-3085
-// Converts the four single-precision, floating-point values of a to signed3086
-// 32-bit integer values using truncate.3087
-// https://msdn.microsoft.com/en-us/library/vstudio/1h005y6x(v=vs.100).aspx3088
-FORCE_INLINE __m128i _mm_cvttps_epi32(__m128 a)3089
-{3090
- return vreinterpretq_m128i_s32(3091
- vcvtq_s32_f32(vreinterpretq_f32_m128(a)));3092
-}3093
-3094
-// Converts the four signed 32-bit integer values of a to single-precision,3095
-// floating-point values3096
-// https://msdn.microsoft.com/en-us/library/vstudio/36bwxcx5(v=vs.100).aspx3097
-FORCE_INLINE __m128 _mm_cvtepi32_ps(__m128i a)3098
-{3099
- return vreinterpretq_m128_f32(3100
- vcvtq_f32_s32(vreinterpretq_s32_m128i(a)));3101
-}3102
-3103
-// Converts the four unsigned 8-bit integers in the lower 16 bits to four3104
-// unsigned 32-bit integers.3105
-FORCE_INLINE __m128i _mm_cvtepu8_epi16(__m128i a)3106
-{3107
- uint8x16_t u8x16 = vreinterpretq_u8_m128i(a); /* xxxx xxxx xxxx DCBA */3108
- uint16x8_t u16x8 =3109
- vmovl_u8(vget_low_u8(u8x16)); /* 0x0x 0x0x 0D0C 0B0A */3110
- return vreinterpretq_m128i_u16(u16x8);3111
-}3112
-3113
-// Converts the four unsigned 8-bit integers in the lower 32 bits to four3114
-// unsigned 32-bit integers.3115
-// https://msdn.microsoft.com/en-us/library/bb531467%28v=vs.100%29.aspx3116
-FORCE_INLINE __m128i _mm_cvtepu8_epi32(__m128i a)3117
-{3118
- uint8x16_t u8x16 = vreinterpretq_u8_m128i(a); /* xxxx xxxx xxxx DCBA */3119
- uint16x8_t u16x8 =3120
- vmovl_u8(vget_low_u8(u8x16)); /* 0x0x 0x0x 0D0C 0B0A */3121
- uint32x4_t u32x4 =3122
- vmovl_u16(vget_low_u16(u16x8)); /* 000D 000C 000B 000A */3123
- return vreinterpretq_m128i_u32(u32x4);3124
-}3125
-3126
-// Converts the two unsigned 8-bit integers in the lower 16 bits to two3127
-// unsigned 64-bit integers.3128
-FORCE_INLINE __m128i _mm_cvtepu8_epi64(__m128i a)3129
-{3130
- uint8x16_t u8x16 = vreinterpretq_u8_m128i(a); /* xxxx xxxx xxxx xxBA */3131
- uint16x8_t u16x8 =3132
- vmovl_u8(vget_low_u8(u8x16)); /* 0x0x 0x0x 0x0x 0B0A */3133
- uint32x4_t u32x4 =3134
- vmovl_u16(vget_low_u16(u16x8)); /* 000x 000x 000B 000A */3135
- uint64x2_t u64x2 =3136
- vmovl_u32(vget_low_u32(u32x4)); /* 0000 000B 0000 000A */3137
- return vreinterpretq_m128i_u64(u64x2);3138
-}3139
-3140
-// Converts the four unsigned 8-bit integers in the lower 16 bits to four3141
-// unsigned 32-bit integers.3142
-FORCE_INLINE __m128i _mm_cvtepi8_epi16(__m128i a)3143
-{3144
- int8x16_t s8x16 = vreinterpretq_s8_m128i(a); /* xxxx xxxx xxxx DCBA */3145
- int16x8_t s16x8 =3146
- vmovl_s8(vget_low_s8(s8x16)); /* 0x0x 0x0x 0D0C 0B0A */3147
- return vreinterpretq_m128i_s16(s16x8);3148
-}3149
-3150
-// Converts the four unsigned 8-bit integers in the lower 32 bits to four3151
-// unsigned 32-bit integers.3152
-FORCE_INLINE __m128i _mm_cvtepi8_epi32(__m128i a)3153
-{3154
- int8x16_t s8x16 = vreinterpretq_s8_m128i(a); /* xxxx xxxx xxxx DCBA */3155
- int16x8_t s16x8 =3156
- vmovl_s8(vget_low_s8(s8x16)); /* 0x0x 0x0x 0D0C 0B0A */3157
- int32x4_t s32x4 =3158
- vmovl_s16(vget_low_s16(s16x8)); /* 000D 000C 000B 000A */3159
- return vreinterpretq_m128i_s32(s32x4);3160
-}3161
-3162
-// Converts the two signed 8-bit integers in the lower 32 bits to four3163
-// signed 64-bit integers.3164
-FORCE_INLINE __m128i _mm_cvtepi8_epi64(__m128i a)3165
-{3166
- int8x16_t s8x16 = vreinterpretq_s8_m128i(a); /* xxxx xxxx xxxx xxBA */3167
- int16x8_t s16x8 =3168
- vmovl_s8(vget_low_s8(s8x16)); /* 0x0x 0x0x 0x0x 0B0A */3169
- int32x4_t s32x4 =3170
- vmovl_s16(vget_low_s16(s16x8)); /* 000x 000x 000B 000A */3171
- int64x2_t s64x2 =3172
- vmovl_s32(vget_low_s32(s32x4)); /* 0000 000B 0000 000A */3173
- return vreinterpretq_m128i_s64(s64x2);3174
-}3175
-3176
-// Converts the four signed 16-bit integers in the lower 64 bits to four signed3177
-// 32-bit integers.3178
-FORCE_INLINE __m128i _mm_cvtepi16_epi32(__m128i a)3179
-{3180
- return vreinterpretq_m128i_s32(3181
- vmovl_s16(vget_low_s16(vreinterpretq_s16_m128i(a))));3182
-}3183
-3184
-// Converts the two signed 16-bit integers in the lower 32 bits two signed3185
-// 32-bit integers.3186
-FORCE_INLINE __m128i _mm_cvtepi16_epi64(__m128i a)3187
-{3188
- int16x8_t s16x8 = vreinterpretq_s16_m128i(a); /* xxxx xxxx xxxx 0B0A */3189
- int32x4_t s32x4 =3190
- vmovl_s16(vget_low_s16(s16x8)); /* 000x 000x 000B 000A */3191
- int64x2_t s64x2 =3192
- vmovl_s32(vget_low_s32(s32x4)); /* 0000 000B 0000 000A */3193
- return vreinterpretq_m128i_s64(s64x2);3194
-}3195
-3196
-// Converts the four unsigned 16-bit integers in the lower 64 bits to four3197
-// unsigned 32-bit integers.3198
-FORCE_INLINE __m128i _mm_cvtepu16_epi32(__m128i a)3199
-{3200
- return vreinterpretq_m128i_u32(3201
- vmovl_u16(vget_low_u16(vreinterpretq_u16_m128i(a))));3202
-}3203
-3204
-// Converts the two unsigned 16-bit integers in the lower 32 bits to two3205
-// unsigned 64-bit integers.3206
-FORCE_INLINE __m128i _mm_cvtepu16_epi64(__m128i a)3207
-{3208
- uint16x8_t u16x8 = vreinterpretq_u16_m128i(a); /* xxxx xxxx xxxx 0B0A */3209
- uint32x4_t u32x4 =3210
- vmovl_u16(vget_low_u16(u16x8)); /* 000x 000x 000B 000A */3211
- uint64x2_t u64x2 =3212
- vmovl_u32(vget_low_u32(u32x4)); /* 0000 000B 0000 000A */3213
- return vreinterpretq_m128i_u64(u64x2);3214
-}3215
-3216
-// Converts the two unsigned 32-bit integers in the lower 64 bits to two3217
-// unsigned 64-bit integers.3218
-FORCE_INLINE __m128i _mm_cvtepu32_epi64(__m128i a)3219
-{3220
- return vreinterpretq_m128i_u64(3221
- vmovl_u32(vget_low_u32(vreinterpretq_u32_m128i(a))));3222
-}3223
-3224
-// Converts the two signed 32-bit integers in the lower 64 bits to two signed3225
-// 64-bit integers.3226
-FORCE_INLINE __m128i _mm_cvtepi32_epi64(__m128i a)3227
-{3228
- return vreinterpretq_m128i_s64(3229
- vmovl_s32(vget_low_s32(vreinterpretq_s32_m128i(a))));3230
-}3231
-3232
-// Converts the four single-precision, floating-point values of a to signed3233
-// 32-bit integer values.3234
-//3235
-// r0 := (int) a03236
-// r1 := (int) a13237
-// r2 := (int) a23238
-// r3 := (int) a33239
-//3240
-// https://msdn.microsoft.com/en-us/library/vstudio/xdc42k5e(v=vs.100).aspx3241
-// *NOTE*. The default rounding mode on SSE is 'round to even', which ARMv7-A3242
-// does not support! It is supported on ARMv8-A however.3243
-FORCE_INLINE __m128i _mm_cvtps_epi32(__m128 a)3244
-{3245
-#if defined(__aarch64__)3246
- return vreinterpretq_m128i_s32(vcvtnq_s32_f32(a));3247
-#else3248
- uint32x4_t signmask = vdupq_n_u32(0x80000000);3249
- float32x4_t half = vbslq_f32(signmask, vreinterpretq_f32_m128(a),3250
- vdupq_n_f32(0.5f)); /* +/- 0.5 */3251
- int32x4_t r_normal =3252
- vcvtq_s32_f32(vaddq_f32(vreinterpretq_f32_m128(a),3253
- half)); /* round to integer: [a + 0.5]*/3254
- int32x4_t r_trunc = vcvtq_s32_f32(3255
- vreinterpretq_f32_m128(a)); /* truncate to integer: [a] */3256
- int32x4_t plusone = vreinterpretq_s32_u32(vshrq_n_u32(3257
- vreinterpretq_u32_s32(vnegq_s32(r_trunc)), 31)); /* 1 or 0 */3258
- int32x4_t r_even = vbicq_s32(vaddq_s32(r_trunc, plusone),3259
- vdupq_n_s32(1)); /* ([a] + {0,1}) & ~1 */3260
- float32x4_t delta = vsubq_f32(3261
- vreinterpretq_f32_m128(a),3262
- vcvtq_f32_s32(r_trunc)); /* compute delta: delta = (a - [a]) */3263
- uint32x4_t is_delta_half =3264
- vceqq_f32(delta, half); /* delta == +/- 0.5 */3265
- return vreinterpretq_m128i_s32(3266
- vbslq_s32(is_delta_half, r_even, r_normal));3267
-#endif3268
-}3269
-3270
-// Moves the least significant 32 bits of a to a 32-bit integer.3271
-// https://msdn.microsoft.com/en-us/library/5z7a9642%28v=vs.90%29.aspx3272
-FORCE_INLINE int _mm_cvtsi128_si32(__m128i a)3273
-{3274
- return vgetq_lane_s32(vreinterpretq_s32_m128i(a), 0);3275
-}3276
-3277
-// Extracts the low order 64-bit integer from the parameter.3278
-// https://msdn.microsoft.com/en-us/library/bb531384(v=vs.120).aspx3279
-FORCE_INLINE uint64_t _mm_cvtsi128_si64(__m128i a)3280
-{3281
- return vgetq_lane_s64(vreinterpretq_s64_m128i(a), 0);3282
-}3283
-3284
-// Moves 32-bit integer a to the least significant 32 bits of an __m128 object,3285
-// zero extending the upper bits.3286
-//3287
-// r0 := a3288
-// r1 := 0x03289
-// r2 := 0x03290
-// r3 := 0x03291
-//3292
-// https://msdn.microsoft.com/en-us/library/ct3539ha%28v=vs.90%29.aspx3293
-FORCE_INLINE __m128i _mm_cvtsi32_si128(int a)3294
-{3295
- return vreinterpretq_m128i_s32(vsetq_lane_s32(a, vdupq_n_s32(0), 0));3296
-}3297
-3298
-// Moves 64-bit integer a to the least significant 64 bits of an __m128 object,3299
-// zero extending the upper bits.3300
-//3301
-// r0 := a3302
-// r1 := 0x03303
-FORCE_INLINE __m128i _mm_cvtsi64_si128(int64_t a)3304
-{3305
- return vreinterpretq_m128i_s64(vsetq_lane_s64(a, vdupq_n_s64(0), 0));3306
-}3307
-3308
-// Applies a type cast to reinterpret four 32-bit floating point values passed3309
-// in as a 128-bit parameter as packed 32-bit integers.3310
-// https://msdn.microsoft.com/en-us/library/bb514099.aspx3311
-FORCE_INLINE __m128i _mm_castps_si128(__m128 a)3312
-{3313
- return vreinterpretq_m128i_s32(vreinterpretq_s32_m128(a));3314
-}3315
-3316
-// Applies a type cast to reinterpret four 32-bit integers passed in as a3317
-// 128-bit parameter as packed 32-bit floating point values.3318
-// https://msdn.microsoft.com/en-us/library/bb514029.aspx3319
-FORCE_INLINE __m128 _mm_castsi128_ps(__m128i a)3320
-{3321
- return vreinterpretq_m128_s32(vreinterpretq_s32_m128i(a));3322
-}3323
-3324
-// Loads 128-bit value. :3325
-// https://msdn.microsoft.com/en-us/library/atzzad1h(v=vs.80).aspx3326
-FORCE_INLINE __m128i _mm_load_si128(const __m128i *p)3327
-{3328
- return vreinterpretq_m128i_s32(vld1q_s32((const int32_t *)p));3329
-}3330
-3331
-// Loads 128-bit value. :3332
-// https://msdn.microsoft.com/zh-cn/library/f4k12ae8(v=vs.90).aspx3333
-FORCE_INLINE __m128i _mm_loadu_si128(const __m128i *p)3334
-{3335
- return vreinterpretq_m128i_s32(vld1q_s32((const int32_t *)p));3336
-}3337
-3338
-// _mm_lddqu_si128 functions the same as _mm_loadu_si128.3339
-#define _mm_lddqu_si128 _mm_loadu_si1283340
-3341
-/* Miscellaneous Operations */3342
-3343
-// Shifts the 8 signed 16-bit integers in a right by count bits while shifting3344
-// in the sign bit.3345
-//3346
-// r0 := a0 >> count3347
-// r1 := a1 >> count3348
-// ...3349
-// r7 := a7 >> count3350
-//3351
-// https://msdn.microsoft.com/en-us/library/3c9997dk(v%3dvs.90).aspx3352
-FORCE_INLINE __m128i _mm_sra_epi16(__m128i a, __m128i count)3353
-{3354
- int64_t c = (int64_t)vget_low_s64((int64x2_t)count);3355
- if (c > 15)3356
- return _mm_cmplt_epi16(a, _mm_setzero_si128());3357
- return vreinterpretq_m128i_s16(3358
- vshlq_s16((int16x8_t)a, vdupq_n_s16(-c)));3359
-}3360
-3361
-// Shifts the 4 signed 32-bit integers in a right by count bits while shifting3362
-// in the sign bit.3363
-//3364
-// r0 := a0 >> count3365
-// r1 := a1 >> count3366
-// r2 := a2 >> count3367
-// r3 := a3 >> count3368
-//3369
-// https://msdn.microsoft.com/en-us/library/ce40009e(v%3dvs.100).aspx3370
-FORCE_INLINE __m128i _mm_sra_epi32(__m128i a, __m128i count)3371
-{3372
- int64_t c = (int64_t)vget_low_s64((int64x2_t)count);3373
- if (c > 31)3374
- return _mm_cmplt_epi32(a, _mm_setzero_si128());3375
- return vreinterpretq_m128i_s32(3376
- vshlq_s32((int32x4_t)a, vdupq_n_s32(-c)));3377
-}3378
-3379
-// Packs the 16 signed 16-bit integers from a and b into 8-bit integers and3380
-// saturates.3381
-// https://msdn.microsoft.com/en-us/library/k4y4f7w5%28v=vs.90%29.aspx3382
-FORCE_INLINE __m128i _mm_packs_epi16(__m128i a, __m128i b)3383
-{3384
- return vreinterpretq_m128i_s8(3385
- vcombine_s8(vqmovn_s16(vreinterpretq_s16_m128i(a)),3386
- vqmovn_s16(vreinterpretq_s16_m128i(b))));3387
-}3388
-3389
-// Packs the 16 signed 16 - bit integers from a and b into 8 - bit unsigned3390
-// integers and saturates.3391
-//3392
-// r0 := UnsignedSaturate(a0)3393
-// r1 := UnsignedSaturate(a1)3394
-// ...3395
-// r7 := UnsignedSaturate(a7)3396
-// r8 := UnsignedSaturate(b0)3397
-// r9 := UnsignedSaturate(b1)3398
-// ...3399
-// r15 := UnsignedSaturate(b7)3400
-//3401
-// https://msdn.microsoft.com/en-us/library/07ad1wx4(v=vs.100).aspx3402
-FORCE_INLINE __m128i _mm_packus_epi16(const __m128i a, const __m128i b)3403
-{3404
- return vreinterpretq_m128i_u8(3405
- vcombine_u8(vqmovun_s16(vreinterpretq_s16_m128i(a)),3406
- vqmovun_s16(vreinterpretq_s16_m128i(b))));3407
-}3408
-3409
-// Packs the 8 signed 32-bit integers from a and b into signed 16-bit integers3410
-// and saturates.3411
-//3412
-// r0 := SignedSaturate(a0)3413
-// r1 := SignedSaturate(a1)3414
-// r2 := SignedSaturate(a2)3415
-// r3 := SignedSaturate(a3)3416
-// r4 := SignedSaturate(b0)3417
-// r5 := SignedSaturate(b1)3418
-// r6 := SignedSaturate(b2)3419
-// r7 := SignedSaturate(b3)3420
-//3421
-// https://msdn.microsoft.com/en-us/library/393t56f9%28v=vs.90%29.aspx3422
-FORCE_INLINE __m128i _mm_packs_epi32(__m128i a, __m128i b)3423
-{3424
- return vreinterpretq_m128i_s16(3425
- vcombine_s16(vqmovn_s32(vreinterpretq_s32_m128i(a)),3426
- vqmovn_s32(vreinterpretq_s32_m128i(b))));3427
-}3428
-3429
-// Packs the 8 unsigned 32-bit integers from a and b into unsigned 16-bit3430
-// integers and saturates.3431
-//3432
-// r0 := UnsignedSaturate(a0)3433
-// r1 := UnsignedSaturate(a1)3434
-// r2 := UnsignedSaturate(a2)3435
-// r3 := UnsignedSaturate(a3)3436
-// r4 := UnsignedSaturate(b0)3437
-// r5 := UnsignedSaturate(b1)3438
-// r6 := UnsignedSaturate(b2)3439
-// r7 := UnsignedSaturate(b3)3440
-FORCE_INLINE __m128i _mm_packus_epi32(__m128i a, __m128i b)3441
-{3442
- return vreinterpretq_m128i_u16(3443
- vcombine_u16(vqmovn_u32(vreinterpretq_u32_m128i(a)),3444
- vqmovn_u32(vreinterpretq_u32_m128i(b))));3445
-}3446
-3447
-// Interleaves the lower 8 signed or unsigned 8-bit integers in a with the lower3448
-// 8 signed or unsigned 8-bit integers in b.3449
-//3450
-// r0 := a03451
-// r1 := b03452
-// r2 := a13453
-// r3 := b13454
-// ...3455
-// r14 := a73456
-// r15 := b73457
-//3458
-// https://msdn.microsoft.com/en-us/library/xf7k860c%28v=vs.90%29.aspx3459
-FORCE_INLINE __m128i _mm_unpacklo_epi8(__m128i a, __m128i b)3460
-{3461
-#if defined(__aarch64__)3462
- return vreinterpretq_m128i_s8(vzip1q_s8(vreinterpretq_s8_m128i(a),3463
- vreinterpretq_s8_m128i(b)));3464
-#else3465
- int8x8_t a1 =3466
- vreinterpret_s8_s16(vget_low_s16(vreinterpretq_s16_m128i(a)));3467
- int8x8_t b1 =3468
- vreinterpret_s8_s16(vget_low_s16(vreinterpretq_s16_m128i(b)));3469
- int8x8x2_t result = vzip_s8(a1, b1);3470
- return vreinterpretq_m128i_s8(3471
- vcombine_s8(result.val[0], result.val[1]));3472
-#endif3473
-}3474
-3475
-// Interleaves the lower 4 signed or unsigned 16-bit integers in a with the3476
-// lower 4 signed or unsigned 16-bit integers in b.3477
-//3478
-// r0 := a03479
-// r1 := b03480
-// r2 := a13481
-// r3 := b13482
-// r4 := a23483
-// r5 := b23484
-// r6 := a33485
-// r7 := b33486
-//3487
-// https://msdn.microsoft.com/en-us/library/btxb17bw%28v=vs.90%29.aspx3488
-FORCE_INLINE __m128i _mm_unpacklo_epi16(__m128i a, __m128i b)3489
-{3490
-#if defined(__aarch64__)3491
- return vreinterpretq_m128i_s16(vzip1q_s16(vreinterpretq_s16_m128i(a),3492
- vreinterpretq_s16_m128i(b)));3493
-#else3494
- int16x4_t a1 = vget_low_s16(vreinterpretq_s16_m128i(a));3495
- int16x4_t b1 = vget_low_s16(vreinterpretq_s16_m128i(b));3496
- int16x4x2_t result = vzip_s16(a1, b1);3497
- return vreinterpretq_m128i_s16(3498
- vcombine_s16(result.val[0], result.val[1]));3499
-#endif3500
-}3501
-3502
-// Interleaves the lower 2 signed or unsigned 32 - bit integers in a with the3503
-// lower 2 signed or unsigned 32 - bit integers in b.3504
-//3505
-// r0 := a03506
-// r1 := b03507
-// r2 := a13508
-// r3 := b13509
-//3510
-// https://msdn.microsoft.com/en-us/library/x8atst9d(v=vs.100).aspx3511
-FORCE_INLINE __m128i _mm_unpacklo_epi32(__m128i a, __m128i b)3512
-{3513
-#if defined(__aarch64__)3514
- return vreinterpretq_m128i_s32(vzip1q_s32(vreinterpretq_s32_m128i(a),3515
- vreinterpretq_s32_m128i(b)));3516
-#else3517
- int32x2_t a1 = vget_low_s32(vreinterpretq_s32_m128i(a));3518
- int32x2_t b1 = vget_low_s32(vreinterpretq_s32_m128i(b));3519
- int32x2x2_t result = vzip_s32(a1, b1);3520
- return vreinterpretq_m128i_s32(3521
- vcombine_s32(result.val[0], result.val[1]));3522
-#endif3523
-}3524
-3525
-FORCE_INLINE __m128i _mm_unpacklo_epi64(__m128i a, __m128i b)3526
-{3527
- int64x1_t a_l = vget_low_s64(vreinterpretq_s64_m128i(a));3528
- int64x1_t b_l = vget_low_s64(vreinterpretq_s64_m128i(b));3529
- return vreinterpretq_m128i_s64(vcombine_s64(a_l, b_l));3530
-}3531
-3532
-// Selects and interleaves the lower two single-precision, floating-point values3533
-// from a and b.3534
-//3535
-// r0 := a03536
-// r1 := b03537
-// r2 := a13538
-// r3 := b13539
-//3540
-// https://msdn.microsoft.com/en-us/library/25st103b%28v=vs.90%29.aspx3541
-FORCE_INLINE __m128 _mm_unpacklo_ps(__m128 a, __m128 b)3542
-{3543
-#if defined(__aarch64__)3544
- return vreinterpretq_m128_f32(vzip1q_f32(vreinterpretq_f32_m128(a),3545
- vreinterpretq_f32_m128(b)));3546
-#else3547
- float32x2_t a1 = vget_low_f32(vreinterpretq_f32_m128(a));3548
- float32x2_t b1 = vget_low_f32(vreinterpretq_f32_m128(b));3549
- float32x2x2_t result = vzip_f32(a1, b1);3550
- return vreinterpretq_m128_f32(3551
- vcombine_f32(result.val[0], result.val[1]));3552
-#endif3553
-}3554
-3555
-// Selects and interleaves the upper two single-precision, floating-point values3556
-// from a and b.3557
-//3558
-// r0 := a23559
-// r1 := b23560
-// r2 := a33561
-// r3 := b33562
-//3563
-// https://msdn.microsoft.com/en-us/library/skccxx7d%28v=vs.90%29.aspx3564
-FORCE_INLINE __m128 _mm_unpackhi_ps(__m128 a, __m128 b)3565
-{3566
-#if defined(__aarch64__)3567
- return vreinterpretq_m128_f32(vzip2q_f32(vreinterpretq_f32_m128(a),3568
- vreinterpretq_f32_m128(b)));3569
-#else3570
- float32x2_t a1 = vget_high_f32(vreinterpretq_f32_m128(a));3571
- float32x2_t b1 = vget_high_f32(vreinterpretq_f32_m128(b));3572
- float32x2x2_t result = vzip_f32(a1, b1);3573
- return vreinterpretq_m128_f32(3574
- vcombine_f32(result.val[0], result.val[1]));3575
-#endif3576
-}3577
-3578
-// Interleaves the upper 8 signed or unsigned 8-bit integers in a with the upper3579
-// 8 signed or unsigned 8-bit integers in b.3580
-//3581
-// r0 := a83582
-// r1 := b83583
-// r2 := a93584
-// r3 := b93585
-// ...3586
-// r14 := a153587
-// r15 := b153588
-//3589
-// https://msdn.microsoft.com/en-us/library/t5h7783k(v=vs.100).aspx3590
-FORCE_INLINE __m128i _mm_unpackhi_epi8(__m128i a, __m128i b)3591
-{3592
-#if defined(__aarch64__)3593
- return vreinterpretq_m128i_s8(vzip2q_s8(vreinterpretq_s8_m128i(a),3594
- vreinterpretq_s8_m128i(b)));3595
-#else3596
- int8x8_t a1 =3597
- vreinterpret_s8_s16(vget_high_s16(vreinterpretq_s16_m128i(a)));3598
- int8x8_t b1 =3599
- vreinterpret_s8_s16(vget_high_s16(vreinterpretq_s16_m128i(b)));3600
- int8x8x2_t result = vzip_s8(a1, b1);3601
- return vreinterpretq_m128i_s8(3602
- vcombine_s8(result.val[0], result.val[1]));3603
-#endif3604
-}3605
-3606
-// Interleaves the upper 4 signed or unsigned 16-bit integers in a with the3607
-// upper 4 signed or unsigned 16-bit integers in b.3608
-//3609
-// r0 := a43610
-// r1 := b43611
-// r2 := a53612
-// r3 := b53613
-// r4 := a63614
-// r5 := b63615
-// r6 := a73616
-// r7 := b73617
-//3618
-// https://msdn.microsoft.com/en-us/library/03196cz7(v=vs.100).aspx3619
-FORCE_INLINE __m128i _mm_unpackhi_epi16(__m128i a, __m128i b)3620
-{3621
-#if defined(__aarch64__)3622
- return vreinterpretq_m128i_s16(vzip2q_s16(vreinterpretq_s16_m128i(a),3623
- vreinterpretq_s16_m128i(b)));3624
-#else3625
- int16x4_t a1 = vget_high_s16(vreinterpretq_s16_m128i(a));3626
- int16x4_t b1 = vget_high_s16(vreinterpretq_s16_m128i(b));3627
- int16x4x2_t result = vzip_s16(a1, b1);3628
- return vreinterpretq_m128i_s16(3629
- vcombine_s16(result.val[0], result.val[1]));3630
-#endif3631
-}3632
-3633
-// Interleaves the upper 2 signed or unsigned 32-bit integers in a with the3634
-// upper 2 signed or unsigned 32-bit integers in b.3635
-// https://msdn.microsoft.com/en-us/library/65sa7cbs(v=vs.100).aspx3636
-FORCE_INLINE __m128i _mm_unpackhi_epi32(__m128i a, __m128i b)3637
-{3638
-#if defined(__aarch64__)3639
- return vreinterpretq_m128i_s32(vzip2q_s32(vreinterpretq_s32_m128i(a),3640
- vreinterpretq_s32_m128i(b)));3641
-#else3642
- int32x2_t a1 = vget_high_s32(vreinterpretq_s32_m128i(a));3643
- int32x2_t b1 = vget_high_s32(vreinterpretq_s32_m128i(b));3644
- int32x2x2_t result = vzip_s32(a1, b1);3645
- return vreinterpretq_m128i_s32(3646
- vcombine_s32(result.val[0], result.val[1]));3647
-#endif3648
-}3649
-3650
-// Interleaves the upper signed or unsigned 64-bit integer in a with the3651
-// upper signed or unsigned 64-bit integer in b.3652
-//3653
-// r0 := a13654
-// r1 := b13655
-FORCE_INLINE __m128i _mm_unpackhi_epi64(__m128i a, __m128i b)3656
-{3657
- int64x1_t a_h = vget_high_s64(vreinterpretq_s64_m128i(a));3658
- int64x1_t b_h = vget_high_s64(vreinterpretq_s64_m128i(b));3659
- return vreinterpretq_m128i_s64(vcombine_s64(a_h, b_h));3660
-}3661
-3662
-// Horizontally compute the minimum amongst the packed unsigned 16-bit integers3663
-// in a, store the minimum and index in dst, and zero the remaining bits in dst.3664
-//3665
-// index[2:0] := 03666
-// min[15:0] := a[15:0]3667
-// FOR j := 0 to 73668
-// i := j*163669
-// IF a[i+15:i] < min[15:0]3670
-// index[2:0] := j3671
-// min[15:0] := a[i+15:i]3672
-// FI3673
-// ENDFOR3674
-// dst[15:0] := min[15:0]3675
-// dst[18:16] := index[2:0]3676
-// dst[127:19] := 03677
-//3678
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_minpos_epu16&expand=37893679
-FORCE_INLINE __m128i _mm_minpos_epu16(__m128i a)3680
-{3681
- __m128i dst;3682
- uint16_t min, idx = 0;3683
- // Find the minimum value3684
-#if defined(__aarch64__)3685
- min = vminvq_u16(vreinterpretq_u16_m128i(a));3686
-#else3687
- __m64i tmp;3688
- tmp = vreinterpret_m64i_u16(3689
- vmin_u16(vget_low_u16(vreinterpretq_u16_m128i(a)),3690
- vget_high_u16(vreinterpretq_u16_m128i(a))));3691
- tmp = vreinterpret_m64i_u16(vpmin_u16(vreinterpret_u16_m64i(tmp),3692
- vreinterpret_u16_m64i(tmp)));3693
- tmp = vreinterpret_m64i_u16(vpmin_u16(vreinterpret_u16_m64i(tmp),3694
- vreinterpret_u16_m64i(tmp)));3695
- min = vget_lane_u16(vreinterpret_u16_m64i(tmp), 0);3696
-#endif3697
- // Get the index of the minimum value3698
- int i;3699
- for (i = 0; i < 8; i++) {3700
- if (min == vgetq_lane_u16(vreinterpretq_u16_m128i(a), 0)) {3701
- idx = (uint16_t)i;3702
- break;3703
- }3704
- a = _mm_srli_si128(a, 2);3705
- }3706
- // Generate result3707
- dst = _mm_setzero_si128();3708
- dst = vreinterpretq_m128i_u16(3709
- vsetq_lane_u16(min, vreinterpretq_u16_m128i(dst), 0));3710
- dst = vreinterpretq_m128i_u16(3711
- vsetq_lane_u16(idx, vreinterpretq_u16_m128i(dst), 1));3712
- return dst;3713
-}3714
-3715
-// shift to right3716
-// https://msdn.microsoft.com/en-us/library/bb514041(v=vs.120).aspx3717
-// http://blog.csdn.net/hemmingway/article/details/448283033718
-// Clang requires a macro here, as it is extremely picky about c being a3719
-// literal.3720
-#define _mm_alignr_epi8(a, b, c) \3721
- ((__m128i)vextq_s8((int8x16_t)(b), (int8x16_t)(a), (c)))3722
-3723
-// Extracts the selected signed or unsigned 8-bit integer from a and zero3724
-// extends.3725
-// FORCE_INLINE int _mm_extract_epi8(__m128i a, __constrange(0,16) int imm)3726
-#define _mm_extract_epi8(a, imm) vgetq_lane_u8(vreinterpretq_u8_m128i(a), (imm))3727
-3728
-// Inserts the least significant 8 bits of b into the selected 8-bit integer3729
-// of a.3730
-// FORCE_INLINE __m128i _mm_insert_epi8(__m128i a, int b,3731
-// __constrange(0,16) int imm)3732
-#define _mm_insert_epi8(a, b, imm) \3733
- __extension__({ \3734
- vreinterpretq_m128i_s8( \3735
- vsetq_lane_s8((b), vreinterpretq_s8_m128i(a), (imm))); \3736
- })3737
-3738
-// Extracts the selected signed or unsigned 16-bit integer from a and zero3739
-// extends.3740
-// https://msdn.microsoft.com/en-us/library/6dceta0c(v=vs.100).aspx3741
-// FORCE_INLINE int _mm_extract_epi16(__m128i a, __constrange(0,8) int imm)3742
-#define _mm_extract_epi16(a, imm) \3743
- vgetq_lane_u16(vreinterpretq_u16_m128i(a), (imm))3744
-3745
-// Inserts the least significant 16 bits of b into the selected 16-bit integer3746
-// of a.3747
-// https://msdn.microsoft.com/en-us/library/kaze8hz1%28v=vs.100%29.aspx3748
-// FORCE_INLINE __m128i _mm_insert_epi16(__m128i a, int b,3749
-// __constrange(0,8) int imm)3750
-#define _mm_insert_epi16(a, b, imm) \3751
- __extension__({ \3752
- vreinterpretq_m128i_s16(vsetq_lane_s16( \3753
- (b), vreinterpretq_s16_m128i(a), (imm))); \3754
- })3755
-3756
-// Extracts the selected signed or unsigned 32-bit integer from a and zero3757
-// extends.3758
-// FORCE_INLINE int _mm_extract_epi32(__m128i a, __constrange(0,4) int imm)3759
-#define _mm_extract_epi32(a, imm) \3760
- vgetq_lane_s32(vreinterpretq_s32_m128i(a), (imm))3761
-3762
-// Extracts the selected single-precision (32-bit) floating-point from a.3763
-// FORCE_INLINE int _mm_extract_ps(__m128 a, __constrange(0,4) int imm)3764
-#define _mm_extract_ps(a, imm) vgetq_lane_s32(vreinterpretq_s32_m128(a), (imm))3765
-3766
-// Inserts the least significant 32 bits of b into the selected 32-bit integer3767
-// of a.3768
-// FORCE_INLINE __m128i _mm_insert_epi32(__m128i a, int b,3769
-// __constrange(0,4) int imm)3770
-#define _mm_insert_epi32(a, b, imm) \3771
- __extension__({ \3772
- vreinterpretq_m128i_s32(vsetq_lane_s32( \3773
- (b), vreinterpretq_s32_m128i(a), (imm))); \3774
- })3775
-3776
-// Extracts the selected signed or unsigned 64-bit integer from a and zero3777
-// extends.3778
-// FORCE_INLINE __int64 _mm_extract_epi64(__m128i a, __constrange(0,2) int imm)3779
-#define _mm_extract_epi64(a, imm) \3780
- vgetq_lane_s64(vreinterpretq_s64_m128i(a), (imm))3781
-3782
-// Inserts the least significant 64 bits of b into the selected 64-bit integer3783
-// of a.3784
-// FORCE_INLINE __m128i _mm_insert_epi64(__m128i a, __int64 b,3785
-// __constrange(0,2) int imm)3786
-#define _mm_insert_epi64(a, b, imm) \3787
- __extension__({ \3788
- vreinterpretq_m128i_s64(vsetq_lane_s64( \3789
- (b), vreinterpretq_s64_m128i(a), (imm))); \3790
- })3791
-3792
-// Count the number of bits set to 1 in unsigned 32-bit integer a, and3793
-// return that count in dst.3794
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_popcnt_u323795
-FORCE_INLINE int _mm_popcnt_u32(unsigned int a)3796
-{3797
-#if defined(__aarch64__)3798
-#if __has_builtin(__builtin_popcount)3799
- return __builtin_popcount(a);3800
-#else3801
- return (int)vaddlv_u8(vcnt_u8(vcreate_u8((uint64_t)a)));3802
-#endif3803
-#else3804
- uint32_t count = 0;3805
- uint8x8_t input_val, count8x8_val;3806
- uint16x4_t count16x4_val;3807
- uint32x2_t count32x2_val;3808
-3809
- input_val = vld1_u8((uint8_t *)&a);3810
- count8x8_val = vcnt_u8(input_val);3811
- count16x4_val = vpaddl_u8(count8x8_val);3812
- count32x2_val = vpaddl_u16(count16x4_val);3813
-3814
- vst1_u32(&count, count32x2_val);3815
- return count;3816
-#endif3817
-}3818
-3819
-// Count the number of bits set to 1 in unsigned 64-bit integer a, and3820
-// return that count in dst.3821
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_popcnt_u643822
-FORCE_INLINE int64_t _mm_popcnt_u64(uint64_t a)3823
-{3824
-#if defined(__aarch64__)3825
-#if __has_builtin(__builtin_popcountll)3826
- return __builtin_popcountll(a);3827
-#else3828
- return (int64_t)vaddlv_u8(vcnt_u8(vcreate_u8(a)));3829
-#endif3830
-#else3831
- uint64_t count = 0;3832
- uint8x8_t input_val, count8x8_val;3833
- uint16x4_t count16x4_val;3834
- uint32x2_t count32x2_val;3835
- uint64x1_t count64x1_val;3836
-3837
- input_val = vld1_u8((uint8_t *)&a);3838
- count8x8_val = vcnt_u8(input_val);3839
- count16x4_val = vpaddl_u8(count8x8_val);3840
- count32x2_val = vpaddl_u16(count16x4_val);3841
- count64x1_val = vpaddl_u32(count32x2_val);3842
- vst1_u64(&count, count64x1_val);3843
- return count;3844
-#endif3845
-}3846
-3847
-// Macro: Transpose the 4x4 matrix formed by the 4 rows of single-precision3848
-// (32-bit) floating-point elements in row0, row1, row2, and row3, and store the3849
-// transposed matrix in these vectors (row0 now contains column 0, etc.).3850
-// https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=MM_TRANSPOSE4_PS&expand=59493851
-#define _MM_TRANSPOSE4_PS(row0, row1, row2, row3) \3852
- do { \3853
- __m128 tmp0, tmp1, tmp2, tmp3; \3854
- tmp0 = _mm_unpacklo_ps(row0, row1); \3855
- tmp2 = _mm_unpacklo_ps(row2, row3); \3856
- tmp1 = _mm_unpackhi_ps(row0, row1); \3857
- tmp3 = _mm_unpackhi_ps(row2, row3); \3858
- row0 = _mm_movelh_ps(tmp0, tmp2); \3859
- row1 = _mm_movehl_ps(tmp2, tmp0); \3860
- row2 = _mm_movelh_ps(tmp1, tmp3); \3861
- row3 = _mm_movehl_ps(tmp3, tmp1); \3862
- } while (0)3863
-3864
-/* Crypto Extensions */3865
-3866
-#if defined(__ARM_FEATURE_CRYPTO)3867
-// Wraps vmull_p643868
-FORCE_INLINE uint64x2_t _sse2neon_vmull_p64(uint64x1_t _a, uint64x1_t _b)3869
-{3870
- poly64_t a = vget_lane_p64(vreinterpret_p64_u64(_a), 0);3871
- poly64_t b = vget_lane_p64(vreinterpret_p64_u64(_b), 0);3872
- return vreinterpretq_u64_p128(vmull_p64(a, b));3873
-}3874
-#else // ARMv7 polyfill3875
-// ARMv7/some A64 lacks vmull_p64, but it has vmull_p8.3876
-//3877
-// vmull_p8 calculates 8 8-bit->16-bit polynomial multiplies, but we need a3878
-// 64-bit->128-bit polynomial multiply.3879
-//3880
-// It needs some work and is somewhat slow, but it is still faster than all3881
-// known scalar methods.3882
-//3883
-// Algorithm adapted to C from3884
-// https://www.workofard.com/2017/07/ghash-for-low-end-cores/, which is adapted3885
-// from "Fast Software Polynomial Multiplication on ARM Processors Using the3886
-// NEON Engine" by Danilo Camara, Conrado Gouvea, Julio Lopez and Ricardo Dahab3887
-// (https://hal.inria.fr/hal-01506572)3888
-static uint64x2_t _sse2neon_vmull_p64(uint64x1_t _a, uint64x1_t _b)3889
-{3890
- poly8x8_t a = vreinterpret_p8_u64(_a);3891
- poly8x8_t b = vreinterpret_p8_u64(_b);3892
-3893
- // Masks3894
- uint8x16_t k48_32 = vcombine_u8(vcreate_u8(0x0000ffffffffffff),3895
- vcreate_u8(0x00000000ffffffff));3896
- uint8x16_t k16_00 = vcombine_u8(vcreate_u8(0x000000000000ffff),3897
- vcreate_u8(0x0000000000000000));3898
-3899
- // Do the multiplies, rotating with vext to get all combinations3900
- uint8x16_t d = vreinterpretq_u8_p16(vmull_p8(a, b)); // D = A0 * B03901
- uint8x16_t e = vreinterpretq_u8_p16(3902
- vmull_p8(a, vext_p8(b, b, 1))); // E = A0 * B13903
- uint8x16_t f = vreinterpretq_u8_p16(3904
- vmull_p8(vext_p8(a, a, 1), b)); // F = A1 * B03905
- uint8x16_t g = vreinterpretq_u8_p16(3906
- vmull_p8(a, vext_p8(b, b, 2))); // G = A0 * B23907
- uint8x16_t h = vreinterpretq_u8_p16(3908
- vmull_p8(vext_p8(a, a, 2), b)); // H = A2 * B03909
- uint8x16_t i = vreinterpretq_u8_p16(3910
- vmull_p8(a, vext_p8(b, b, 3))); // I = A0 * B33911
- uint8x16_t j = vreinterpretq_u8_p16(3912
- vmull_p8(vext_p8(a, a, 3), b)); // J = A3 * B03913
- uint8x16_t k = vreinterpretq_u8_p16(3914
- vmull_p8(a, vext_p8(b, b, 4))); // L = A0 * B43915
-3916
- // Add cross products3917
- uint8x16_t l = veorq_u8(e, f); // L = E + F3918
- uint8x16_t m = veorq_u8(g, h); // M = G + H3919
- uint8x16_t n = veorq_u8(i, j); // N = I + J3920
-3921
- // Interleave. Using vzip1 and vzip2 prevents Clang from emitting TBL3922
- // instructions.3923
-#if defined(__aarch64__)3924
- uint8x16_t lm_p0 = vreinterpretq_u8_u64(3925
- vzip1q_u64(vreinterpretq_u64_u8(l), vreinterpretq_u64_u8(m)));3926
- uint8x16_t lm_p1 = vreinterpretq_u8_u64(3927
- vzip2q_u64(vreinterpretq_u64_u8(l), vreinterpretq_u64_u8(m)));3928
- uint8x16_t nk_p0 = vreinterpretq_u8_u64(3929
- vzip1q_u64(vreinterpretq_u64_u8(n), vreinterpretq_u64_u8(k)));3930
- uint8x16_t nk_p1 = vreinterpretq_u8_u64(3931
- vzip2q_u64(vreinterpretq_u64_u8(n), vreinterpretq_u64_u8(k)));3932
-#else3933
- uint8x16_t lm_p0 = vcombine_u8(vget_low_u8(l), vget_low_u8(m));3934
- uint8x16_t lm_p1 = vcombine_u8(vget_high_u8(l), vget_high_u8(m));3935
- uint8x16_t nk_p0 = vcombine_u8(vget_low_u8(n), vget_low_u8(k));3936
- uint8x16_t nk_p1 = vcombine_u8(vget_high_u8(n), vget_high_u8(k));3937
-#endif3938
- // t0 = (L) (P0 + P1) << 83939
- // t1 = (M) (P2 + P3) << 163940
- uint8x16_t t0t1_tmp = veorq_u8(lm_p0, lm_p1);3941
- uint8x16_t t0t1_h = vandq_u8(lm_p1, k48_32);3942
- uint8x16_t t0t1_l = veorq_u8(t0t1_tmp, t0t1_h);3943
-3944
- // t2 = (N) (P4 + P5) << 243945
- // t3 = (K) (P6 + P7) << 323946
- uint8x16_t t2t3_tmp = veorq_u8(nk_p0, nk_p1);3947
- uint8x16_t t2t3_h = vandq_u8(nk_p1, k16_00);3948
- uint8x16_t t2t3_l = veorq_u8(t2t3_tmp, t2t3_h);3949
-3950
- // De-interleave3951
-#if defined(__aarch64__)3952
- uint8x16_t t0 = vreinterpretq_u8_u64(vuzp1q_u64(3953
- vreinterpretq_u64_u8(t0t1_l), vreinterpretq_u64_u8(t0t1_h)));3954
- uint8x16_t t1 = vreinterpretq_u8_u64(vuzp2q_u64(3955
- vreinterpretq_u64_u8(t0t1_l), vreinterpretq_u64_u8(t0t1_h)));3956
- uint8x16_t t2 = vreinterpretq_u8_u64(vuzp1q_u64(3957
- vreinterpretq_u64_u8(t2t3_l), vreinterpretq_u64_u8(t2t3_h)));3958
- uint8x16_t t3 = vreinterpretq_u8_u64(vuzp2q_u64(3959
- vreinterpretq_u64_u8(t2t3_l), vreinterpretq_u64_u8(t2t3_h)));3960
-#else3961
- uint8x16_t t1 = vcombine_u8(vget_high_u8(t0t1_l), vget_high_u8(t0t1_h));3962
- uint8x16_t t0 = vcombine_u8(vget_low_u8(t0t1_l), vget_low_u8(t0t1_h));3963
- uint8x16_t t3 = vcombine_u8(vget_high_u8(t2t3_l), vget_high_u8(t2t3_h));3964
- uint8x16_t t2 = vcombine_u8(vget_low_u8(t2t3_l), vget_low_u8(t2t3_h));3965
-#endif3966
- // Shift the cross products3967
- uint8x16_t t0_shift = vextq_u8(t0, t0, 15); // t0 << 83968
- uint8x16_t t1_shift = vextq_u8(t1, t1, 14); // t1 << 163969
- uint8x16_t t2_shift = vextq_u8(t2, t2, 13); // t2 << 243970
- uint8x16_t t3_shift = vextq_u8(t3, t3, 12); // t3 << 323971
-3972
- // Accumulate the products3973
- uint8x16_t cross1 = veorq_u8(t0_shift, t1_shift);3974
- uint8x16_t cross2 = veorq_u8(t2_shift, t3_shift);3975
- uint8x16_t mix = veorq_u8(d, cross1);3976
- uint8x16_t r = veorq_u8(mix, cross2);3977
- return vreinterpretq_u64_u8(r);3978
-}3979
-#endif // ARMv7 polyfill3980
-3981
-FORCE_INLINE __m128i _mm_clmulepi64_si128(__m128i _a, __m128i _b, const int imm)3982
-{3983
- uint64x2_t a = vreinterpretq_u64_m128i(_a);3984
- uint64x2_t b = vreinterpretq_u64_m128i(_b);3985
- switch (imm & 0x11) {3986
- case 0x00:3987
- return vreinterpretq_m128i_u64(3988
- _sse2neon_vmull_p64(vget_low_u64(a), vget_low_u64(b)));3989
- case 0x01:3990
- return vreinterpretq_m128i_u64(3991
- _sse2neon_vmull_p64(vget_high_u64(a), vget_low_u64(b)));3992
- case 0x10:3993
- return vreinterpretq_m128i_u64(3994
- _sse2neon_vmull_p64(vget_low_u64(a), vget_high_u64(b)));3995
- case 0x11:3996
- return vreinterpretq_m128i_u64(_sse2neon_vmull_p64(3997
- vget_high_u64(a), vget_high_u64(b)));3998
- default:3999
- abort();4000
- }4001
-}4002
-4003
-#if !defined(__ARM_FEATURE_CRYPTO) && defined(__aarch64__)4004
-// In the absence of crypto extensions, implement aesenc using regular neon4005
-// intrinsics instead. See:4006
-// https://www.workofard.com/2017/01/accelerated-aes-for-the-arm64-linux-kernel/4007
-// https://www.workofard.com/2017/07/ghash-for-low-end-cores/ and4008
-// https://github.com/ColinIanKing/linux-next-mirror/blob/b5f466091e130caaf0735976648f72bd5e09aa84/crypto/aegis128-neon-inner.c#L524009
-// for more information Reproduced with permission of the author.4010
-FORCE_INLINE __m128i _mm_aesenc_si128(__m128i EncBlock, __m128i RoundKey)4011
-{4012
- static const uint8_t crypto_aes_sbox[256] = {4013
- 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5, 0x30, 0x01,4014
- 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76, 0xca, 0x82, 0xc9, 0x7d,4015
- 0xfa, 0x59, 0x47, 0xf0, 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4,4016
- 0x72, 0xc0, 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc,4017
- 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15, 0x04, 0xc7,4018
- 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a, 0x07, 0x12, 0x80, 0xe2,4019
- 0xeb, 0x27, 0xb2, 0x75, 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e,4020
- 0x5a, 0xa0, 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,4021
- 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b, 0x6a, 0xcb,4022
- 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf, 0xd0, 0xef, 0xaa, 0xfb,4023
- 0x43, 0x4d, 0x33, 0x85, 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c,4024
- 0x9f, 0xa8, 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5,4025
- 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2, 0xcd, 0x0c,4026
- 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17, 0xc4, 0xa7, 0x7e, 0x3d,4027
- 0x64, 0x5d, 0x19, 0x73, 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a,4028
- 0x90, 0x88, 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,4029
- 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c, 0xc2, 0xd3,4030
- 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79, 0xe7, 0xc8, 0x37, 0x6d,4031
- 0x8d, 0xd5, 0x4e, 0xa9, 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a,4032
- 0xae, 0x08, 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6,4033
- 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a, 0x70, 0x3e,4034
- 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e, 0x61, 0x35, 0x57, 0xb9,4035
- 0x86, 0xc1, 0x1d, 0x9e, 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9,4036
- 0x8e, 0x94, 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,4037
- 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68, 0x41, 0x99,4038
- 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16};4039
- static const uint8_t shift_rows[] = {0x0, 0x5, 0xa, 0xf, 0x4, 0x9,4040
- 0xe, 0x3, 0x8, 0xd, 0x2, 0x7,4041
- 0xc, 0x1, 0x6, 0xb};4042
- static const uint8_t ror32by8[] = {0x1, 0x2, 0x3, 0x0, 0x5, 0x6,4043
- 0x7, 0x4, 0x9, 0xa, 0xb, 0x8,4044
- 0xd, 0xe, 0xf, 0xc};4045
-4046
- uint8x16_t v;4047
- uint8x16_t w = vreinterpretq_u8_m128i(EncBlock);4048
-4049
- // shift rows4050
- w = vqtbl1q_u8(w, vld1q_u8(shift_rows));4051
-4052
- // sub bytes4053
- v = vqtbl4q_u8(vld1q_u8_x4(crypto_aes_sbox), w);4054
- v = vqtbx4q_u8(v, vld1q_u8_x4(crypto_aes_sbox + 0x40), w - 0x40);4055
- v = vqtbx4q_u8(v, vld1q_u8_x4(crypto_aes_sbox + 0x80), w - 0x80);4056
- v = vqtbx4q_u8(v, vld1q_u8_x4(crypto_aes_sbox + 0xc0), w - 0xc0);4057
-4058
- // mix columns4059
- w = (v << 1) ^ (uint8x16_t)(((int8x16_t)v >> 7) & 0x1b);4060
- w ^= (uint8x16_t)vrev32q_u16((uint16x8_t)v);4061
- w ^= vqtbl1q_u8(v ^ w, vld1q_u8(ror32by8));4062
-4063
- // add round key4064
- return vreinterpretq_m128i_u8(w) ^ RoundKey;4065
-}4066
-#elif defined(__ARM_FEATURE_CRYPTO)4067
-// Implements equivalent of 'aesenc' by combining AESE (with an empty key) and4068
-// AESMC and then manually applying the real key as an xor operation This4069
-// unfortunately means an additional xor op; the compiler should be able to4070
-// optimise this away for repeated calls however See4071
-// https://blog.michaelbrase.com/2018/05/08/emulating-x86-aes-intrinsics-on-armv8-a4072
-// for more details.4073
-inline __m128i _mm_aesenc_si128(__m128i a, __m128i b)4074
-{4075
- return vreinterpretq_m128i_u8(4076
- vaesmcq_u8(4077
- vaeseq_u8(vreinterpretq_u8_m128i(a), vdupq_n_u8(0))) ^4078
- vreinterpretq_u8_m128i(b));4079
-}4080
-#endif4081
-4082
-/* Streaming Extensions */4083
-4084
-// Guarantees that every preceding store is globally visible before any4085
-// subsequent store.4086
-// https://msdn.microsoft.com/en-us/library/5h2w73d1%28v=vs.90%29.aspx4087
-FORCE_INLINE void _mm_sfence(void)4088
-{4089
- __sync_synchronize();4090
-}4091
-4092
-// Stores the data in a to the address p without polluting the caches. If the4093
-// cache line containing address p is already in the cache, the cache will be4094
-// updated.Address p must be 16 - byte aligned.4095
-// https://msdn.microsoft.com/en-us/library/ba08y07y%28v=vs.90%29.aspx4096
-FORCE_INLINE void _mm_stream_si128(__m128i *p, __m128i a)4097
-{4098
-#if __has_builtin(__builtin_nontemporal_store)4099
- __builtin_nontemporal_store(a, p);4100
-#else4101
- vst1q_s64((int64_t *)p, vreinterpretq_s64_m128i(a));4102
-#endif4103
-}4104
-4105
-// Cache line containing p is flushed and invalidated from all caches in the4106
-// coherency domain. :4107
-// https://msdn.microsoft.com/en-us/library/ba08y07y(v=vs.100).aspx4108
-FORCE_INLINE void _mm_clflush(void const *p)4109
-{4110
- (void)p;4111
- // no corollary for Neon?4112
-}4113
-4114
-// Allocate aligned blocks of memory.4115
-// https://software.intel.com/en-us/4116
-// cpp-compiler-developer-guide-and-reference-allocating-and-freeing-aligned-memory-blocks4117
-FORCE_INLINE void *_mm_malloc(size_t size, size_t align)4118
-{4119
- void *ptr;4120
- if (align == 1)4121
- return malloc(size);4122
- if (align == 2 || (sizeof(void *) == 8 && align == 4))4123
- align = sizeof(void *);4124
- if (!posix_memalign(&ptr, align, size))4125
- return ptr;4126
- return NULL;4127
-}4128
-4129
-FORCE_INLINE void _mm_free(void *addr)4130
-{4131
- free(addr);4132
-}4133
-4134
-// Starting with the initial value in crc, accumulates a CRC32 value for4135
-// unsigned 8-bit integer v.4136
-// https://msdn.microsoft.com/en-us/library/bb514036(v=vs.100)4137
-FORCE_INLINE uint32_t _mm_crc32_u8(uint32_t crc, uint8_t v)4138
-{4139
-#if defined(__aarch64__) && defined(__ARM_FEATURE_CRC32)4140
- __asm__ __volatile__("crc32cb %w[c], %w[c], %w[v]\n\t"4141
- : [c] "+r"(crc)4142
- : [v] "r"(v));4143
-#else4144
- crc ^= v;4145
- for (int bit = 0; bit < 8; bit++) {4146
- if (crc & 1)4147
- crc = (crc >> 1) ^ UINT32_C(0x82f63b78);4148
- else4149
- crc = (crc >> 1);4150
- }4151
-#endif4152
- return crc;4153
-}4154
-4155
-// Starting with the initial value in crc, accumulates a CRC32 value for4156
-// unsigned 16-bit integer v.4157
-// https://msdn.microsoft.com/en-us/library/bb531411(v=vs.100)4158
-FORCE_INLINE uint32_t _mm_crc32_u16(uint32_t crc, uint16_t v)4159
-{4160
-#if defined(__aarch64__) && defined(__ARM_FEATURE_CRC32)4161
- __asm__ __volatile__("crc32ch %w[c], %w[c], %w[v]\n\t"4162
- : [c] "+r"(crc)4163
- : [v] "r"(v));4164
-#else4165
- crc = _mm_crc32_u8(crc, v & 0xff);4166
- crc = _mm_crc32_u8(crc, (v >> 8) & 0xff);4167
-#endif4168
- return crc;4169
-}4170
-4171
-// Starting with the initial value in crc, accumulates a CRC32 value for4172
-// unsigned 32-bit integer v.4173
-// https://msdn.microsoft.com/en-us/library/bb531394(v=vs.100)4174
-FORCE_INLINE uint32_t _mm_crc32_u32(uint32_t crc, uint32_t v)4175
-{4176
-#if defined(__aarch64__) && defined(__ARM_FEATURE_CRC32)4177
- __asm__ __volatile__("crc32cw %w[c], %w[c], %w[v]\n\t"4178
- : [c] "+r"(crc)4179
- : [v] "r"(v));4180
-#else4181
- crc = _mm_crc32_u16(crc, v & 0xffff);4182
- crc = _mm_crc32_u16(crc, (v >> 16) & 0xffff);4183
-#endif4184
- return crc;4185
-}4186
-4187
-// Starting with the initial value in crc, accumulates a CRC32 value for4188
-// unsigned 64-bit integer v.4189
-// https://msdn.microsoft.com/en-us/library/bb514033(v=vs.100)4190
-FORCE_INLINE uint64_t _mm_crc32_u64(uint64_t crc, uint64_t v)4191
-{4192
-#if defined(__aarch64__) && defined(__ARM_FEATURE_CRC32)4193
- __asm__ __volatile__("crc32cx %w[c], %w[c], %x[v]\n\t"4194
- : [c] "+r"(crc)4195
- : [v] "r"(v));4196
-#else4197
- crc = _mm_crc32_u32((uint32_t)(crc), v & 0xffffffff);4198
- crc = _mm_crc32_u32((uint32_t)(crc), (v >> 32) & 0xffffffff);4199
-#endif4200
- return crc;4201
-}4202
-4203
-#if defined(__GNUC__) || defined(__clang__)4204
-#pragma pop_macro("ALIGN_STRUCT")4205
-#pragma pop_macro("FORCE_INLINE")4206
-#endif4207
-4208
-#endif4209
obs-studio-26.1.0.tar.xz/.github/workflows/main.yml -> obs-studio-26.1.1.tar.xz/.github/workflows/main.yml
Changed
153
1
2
runs-on: [macos-latest]3
env:4
MIN_MACOS_VERSION: '10.13'5
- MACOS_DEPS_VERSION: '2020-12-11'6
+ MACOS_DEPS_VERSION: '2020-12-22'7
VLC_VERSION: '3.0.8'8
SPARKLE_VERSION: '1.23.0'9
QT_VERSION: '5.15.2'10
11
shell: bash12
run: |13
if [ -d /usr/local/opt/openssl@1.0.2t ]; then14
- brew uninstall openssl@1.0.2t15
- brew untap local/openssl16
+ brew uninstall openssl@1.0.2t17
+ brew untap local/openssl18
fi19
20
if [ -d /usr/local/opt/python@2.7.17 ]; then21
- brew uninstall python@2.7.1722
- brew untap local/python223
+ brew uninstall python@2.7.1724
+ brew untap local/python225
+ fi26
+27
+ if [ -d /usr/local/opt/speexdsp ]; then28
+ brew unlink speexdsp29
fi30
brew bundle --file ./CI/scripts/macos/Brewfile31
- name: 'Restore Chromium Embedded Framework from cache'32
33
run: |34
mkdir ./build35
cd ./build36
- cmake -DENABLE_UNIT_TESTS=YES -DENABLE_SPARKLE_UPDATER=ON -DDISABLE_PYTHON=ON -DCMAKE_OSX_DEPLOYMENT_TARGET=${{ env.MIN_MACOS_VERSION }} -DQTDIR="/tmp/obsdeps" -DSWIGDIR="/tmp/obsdeps" -DDepsPath="/tmp/obsdeps" -DVLCPath="${{ github.workspace }}/cmbuild/vlc-${{ env.VLC_VERSION }}" -DENABLE_VLC=ON -DBUILD_BROWSER=ON -DBROWSER_DEPLOY=ON -DWITH_RTMPS=ON -DCEF_ROOT_DIR="${{ github.workspace }}/cmbuild/cef_binary_${{ env.CEF_BUILD_VERSION }}_macosx64" ..37
+ LEGACY_BROWSER="$(test "${{ env.CEF_BUILD_VERSION }}" -le 3770 && echo "ON" || echo "OFF")"38
+ cmake -DENABLE_UNIT_TESTS=YES -DENABLE_SPARKLE_UPDATER=ON -DDISABLE_PYTHON=ON -DCMAKE_OSX_DEPLOYMENT_TARGET=${{ env.MIN_MACOS_VERSION }} -DQTDIR="/tmp/obsdeps" -DSWIGDIR="/tmp/obsdeps" -DDepsPath="/tmp/obsdeps" -DVLCPath="${{ github.workspace }}/cmbuild/vlc-${{ env.VLC_VERSION }}" -DENABLE_VLC=ON -DBUILD_BROWSER=ON -DBROWSER_LEGACY=$LEGACY_BROWSER -DWITH_RTMPS=ON -DCEF_ROOT_DIR="${{ github.workspace }}/cmbuild/cef_binary_${{ env.CEF_BUILD_VERSION }}_macosx64" ..39
- name: 'Build'40
shell: bash41
working-directory: ${{ github.workspace }}/build42
43
mkdir -p OBS.app/Contents/MacOS44
mkdir OBS.app/Contents/PlugIns45
mkdir OBS.app/Contents/Resources46
+ mkdir OBS.app/Contents/Frameworks47
48
cp rundir/RelWithDebInfo/bin/obs ./OBS.app/Contents/MacOS49
cp rundir/RelWithDebInfo/bin/obs-ffmpeg-mux ./OBS.app/Contents/MacOS50
+ if ! [ "${{ env.CEF_BUILD_VERSION }}" -le 3770 ]; then51
+ cp -R "rundir/RelWithDebInfo/bin/OBS Helper.app" "./OBS.app/Contents/Frameworks/OBS Helper.app"52
+ cp -R "rundir/RelWithDebInfo/bin/OBS Helper (GPU).app" "./OBS.app/Contents/Frameworks/OBS Helper (GPU).app"53
+ cp -R "rundir/RelWithDebInfo/bin/OBS Helper (Plugin).app" "./OBS.app/Contents/Frameworks/OBS Helper (Plugin).app"54
+ cp -R "rundir/RelWithDebInfo/bin/OBS Helper (Renderer).app" "./OBS.app/Contents/Frameworks/OBS Helper (Renderer).app"55
+ fi56
cp rundir/RelWithDebInfo/bin/libobsglad.0.dylib ./OBS.app/Contents/MacOS57
cp -R rundir/RelWithDebInfo/data ./OBS.app/Contents/Resources58
cp ../CI/scripts/macos/app/AppIcon.icns ./OBS.app/Contents/Resources59
60
rm -rf ./OBS.app/Contents/Resources/data/obs-scripting/61
fi62
63
+ BUNDLE_PLUGINS=(64
+ ./OBS.app/Contents/PlugIns/coreaudio-encoder.so65
+ ./OBS.app/Contents/PlugIns/decklink-ouput-ui.so66
+ ./OBS.app/Contents/PlugIns/decklink-captions.so67
+ ./OBS.app/Contents/PlugIns/frontend-tools.so68
+ ./OBS.app/Contents/PlugIns/image-source.so69
+ ./OBS.app/Contents/PlugIns/mac-avcapture.so70
+ ./OBS.app/Contents/PlugIns/mac-capture.so71
+ ./OBS.app/Contents/PlugIns/mac-decklink.so72
+ ./OBS.app/Contents/PlugIns/mac-syphon.so73
+ ./OBS.app/Contents/PlugIns/mac-vth264.so74
+ ./OBS.app/Contents/PlugIns/mac-virtualcam.so75
+ ./OBS.app/Contents/PlugIns/obs-browser.so76
+ ./OBS.app/Contents/PlugIns/obs-ffmpeg.so77
+ ./OBS.app/Contents/PlugIns/obs-filters.so78
+ ./OBS.app/Contents/PlugIns/obs-transitions.so79
+ ./OBS.app/Contents/PlugIns/obs-vst.so80
+ ./OBS.app/Contents/PlugIns/rtmp-services.so81
+ ./OBS.app/Contents/MacOS/obs-ffmpeg-mux82
+ ./OBS.app/Contents/MacOS/obslua.so83
+ ./OBS.app/Contents/PlugIns/obs-x264.so84
+ ./OBS.app/Contents/PlugIns/text-freetype2.so85
+ ./OBS.app/Contents/PlugIns/obs-outputs.so86
+ )87
+88
+ if ! [ "${{ env.CEF_BUILD_VERSION }}" -le 3770 ]; then89
../CI/scripts/macos/app/dylibBundler -cd -of -a ./OBS.app -q -f \90
-s ./OBS.app/Contents/MacOS \91
-s "${{ github.workspace }}/cmbuild/sparkle/Sparkle.framework" \92
-s ./rundir/RelWithDebInfo/bin \93
- -x ./OBS.app/Contents/PlugIns/coreaudio-encoder.so \94
- -x ./OBS.app/Contents/PlugIns/decklink-ouput-ui.so \95
- -x ./OBS.app/Contents/PlugIns/decklink-captions.so \96
- -x ./OBS.app/Contents/PlugIns/frontend-tools.so \97
- -x ./OBS.app/Contents/PlugIns/image-source.so \98
- -x ./OBS.app/Contents/PlugIns/linux-jack.so \99
- -x ./OBS.app/Contents/PlugIns/mac-avcapture.so \100
- -x ./OBS.app/Contents/PlugIns/mac-capture.so \101
- -x ./OBS.app/Contents/PlugIns/mac-decklink.so \102
- -x ./OBS.app/Contents/PlugIns/mac-syphon.so \103
- -x ./OBS.app/Contents/PlugIns/mac-vth264.so \104
- -x ./OBS.app/Contents/PlugIns/mac-virtualcam.so \105
- -x ./OBS.app/Contents/PlugIns/obs-browser.so \106
- -x ./OBS.app/Contents/PlugIns/obs-browser-page \107
- -x ./OBS.app/Contents/PlugIns/obs-ffmpeg.so \108
- -x ./OBS.app/Contents/PlugIns/obs-filters.so \109
- -x ./OBS.app/Contents/PlugIns/obs-transitions.so \110
- -x ./OBS.app/Contents/PlugIns/obs-vst.so \111
- -x ./OBS.app/Contents/PlugIns/rtmp-services.so \112
- -x ./OBS.app/Contents/MacOS/obs-ffmpeg-mux \113
- -x ./OBS.app/Contents/MacOS/obslua.so \114
- -x ./OBS.app/Contents/PlugIns/obs-x264.so \115
- -x ./OBS.app/Contents/PlugIns/text-freetype2.so \116
- -x ./OBS.app/Contents/PlugIns/obs-libfdk.so \117
- -x ./OBS.app/Contents/PlugIns/obs-outputs.so118
+ $(echo "${BUNDLE_PLUGINS[@]/#/-x }")119
+ else120
+ ../CI/scripts/macos/app/dylibBundler -cd -of -a ./OBS.app -q -f \121
+ -s ./OBS.app/Contents/MacOS \122
+ -s "${{ github.workspace }}/cmbuild/sparkle/Sparkle.framework" \123
+ -s ./rundir/RelWithDebInfo/bin \124
+ $(echo "${BUNDLE_PLUGINS[@]/#/-x }") \125
+ -x ./OBS.app/Contents/PlugIns/obs-browser-page126
+ fi127
128
mv ./libobs-opengl/libobs-opengl.so ./OBS.app/Contents/Frameworks129
130
131
codesign --force --options runtime --sign "${SIGN_IDENTITY:--}" "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework/Libraries/libswiftshader_libEGL.dylib"132
codesign --force --options runtime --sign "${SIGN_IDENTITY:--}" "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework/Libraries/libGLESv2.dylib"133
codesign --force --options runtime --sign "${SIGN_IDENTITY:--}" "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework/Libraries/libswiftshader_libGLESv2.dylib"134
+ if ! [ "${{ env.CEF_BUILD_VERSION }}" -le 3770 ]; then135
+ codesign --force --options runtime --sign "${SIGN_IDENTITY:--}" "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework/Libraries/libvk_swiftshader.dylib"136
+ fi137
codesign --force --options runtime --sign "${SIGN_IDENTITY:--}" --deep "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework"138
139
codesign --force --options runtime --deep --sign "${SIGN_IDENTITY:--}" "./OBS.app/Contents/Resources/data/obs-mac-virtualcam.plugin"140
141
codesign --force --options runtime --entitlements "../CI/scripts/macos/app/entitlements.plist" --sign "${SIGN_IDENTITY:--}" --deep ./OBS.app142
143
+ if ! [ "${{ env.CEF_BUILD_VERSION }}" -le 3770 ]; then144
+ codesign --force --options runtime --sign "${SIGN_IDENTITY:--}" --deep "./OBS.app/Contents/Frameworks/OBS Helper.app"145
+ codesign --force --options runtime --entitlements "../CI/scripts/macos/helpers/helper-gpu-entitlements.plist" --sign "${SIGN_IDENTITY:--}" --deep "./OBS.app/Contents/Frameworks/OBS Helper (GPU).app"146
+ codesign --force --options runtime --entitlements "../CI/scripts/macos/helpers/helper-plugin-entitlements.plist" --sign "${SIGN_IDENTITY:--}" --deep "./OBS.app/Contents/Frameworks/OBS Helper (Plugin).app"147
+ codesign --force --options runtime --entitlements "../CI/scripts/macos/helpers/helper-renderer-entitlements.plist" --sign "${SIGN_IDENTITY:--}" --deep "./OBS.app/Contents/Frameworks/OBS Helper (Renderer).app"148
+ fi149
+150
codesign -dvv ./OBS.app151
- name: 'Package'152
if: success() && (github.event_name != 'pull_request' || env.SEEKING_TESTERS == '1')153
obs-studio-26.1.0.tar.xz/CI/full-build-macos.sh -> obs-studio-26.1.1.tar.xz/CI/full-build-macos.sh
Changed
162
1
2
CI_SPARKLE_VERSION=$(cat ${CI_WORKFLOW} | sed -En "s/[ ]+SPARKLE_VERSION: '([0-9\.]+)'/\1/p")3
CI_QT_VERSION=$(cat ${CI_WORKFLOW} | sed -En "s/[ ]+QT_VERSION: '([0-9\.]+)'/\1/p" | head -1)4
CI_MIN_MACOS_VERSION=$(cat ${CI_WORKFLOW} | sed -En "s/[ ]+MIN_MACOS_VERSION: '([0-9\.]+)'/\1/p")5
+NPROC="${NPROC:-$(sysctl -n hw.ncpu)}"6
7
BUILD_DEPS=(8
"obs-deps ${MACOS_DEPS_VERSION:-${CI_DEPS_VERSION}}"9
10
-DCMAKE_OSX_DEPLOYMENT_TARGET=${MIN_MACOS_VERSION:-${CI_MIN_MACOS_VERSION}} \11
..12
step "Build..."13
- make -j414
+ make -j${NPROC}15
if [ ! -d libcef_dll ]; then mkdir libcef_dll; fi16
}17
18
19
-DDepsPath="/tmp/obsdeps" \20
-DVLCPath="${DEPS_BUILD_DIR}/vlc-${VLC_VERSION:-${CI_VLC_VERSION}}" \21
-DBUILD_BROWSER=ON \22
- -DBROWSER_DEPLOY=ON \23
+ -DBROWSER_LEGACY="$(test "${CEF_BUILD_VERSION:-${CI_CEF_VERSION}}" -le 3770 && echo "ON" || echo "OFF")" \24
-DWITH_RTMPS=ON \25
-DCEF_ROOT_DIR="${DEPS_BUILD_DIR}/cef_binary_${CEF_BUILD_VERSION:-${CI_CEF_VERSION}}_macosx64" \26
-DCMAKE_BUILD_TYPE="${BUILD_CONFIG}" \27
28
run_obs_build() {29
ensure_dir "${CHECKOUT_DIR}/${BUILD_DIR}"30
hr "Build OBS..."31
- make -j432
+ make -j${NPROC}33
}34
35
## OBS BUNDLE AS MACOS APPLICATION ##36
37
hr "Bundle dylibs for macOS application"38
39
step "Run dylibBundler.."40
- ${CI_SCRIPTS}/app/dylibbundler -cd -of -a ./OBS.app -q -f \41
- -s ./OBS.app/Contents/MacOS \42
- -s "${DEPS_BUILD_DIR}/sparkle/Sparkle.framework" \43
- -s ./rundir/${BUILD_CONFIG}/bin/ \44
- -x ./OBS.app/Contents/PlugIns/coreaudio-encoder.so \45
- -x ./OBS.app/Contents/PlugIns/decklink-ouput-ui.so \46
- -x ./OBS.app/Contents/PlugIns/decklink-captions.so \47
- -x ./OBS.app/Contents/PlugIns/frontend-tools.so \48
- -x ./OBS.app/Contents/PlugIns/image-source.so \49
- -x ./OBS.app/Contents/PlugIns/linux-jack.so \50
- -x ./OBS.app/Contents/PlugIns/mac-avcapture.so \51
- -x ./OBS.app/Contents/PlugIns/mac-capture.so \52
- -x ./OBS.app/Contents/PlugIns/mac-decklink.so \53
- -x ./OBS.app/Contents/PlugIns/mac-syphon.so \54
- -x ./OBS.app/Contents/PlugIns/mac-vth264.so \55
- -x ./OBS.app/Contents/PlugIns/mac-virtualcam.so \56
- -x ./OBS.app/Contents/PlugIns/obs-browser.so \57
- -x ./OBS.app/Contents/PlugIns/obs-browser-page \58
- -x ./OBS.app/Contents/PlugIns/obs-ffmpeg.so \59
- -x ./OBS.app/Contents/PlugIns/obs-filters.so \60
- -x ./OBS.app/Contents/PlugIns/obs-transitions.so \61
- -x ./OBS.app/Contents/PlugIns/obs-vst.so \62
- -x ./OBS.app/Contents/PlugIns/rtmp-services.so \63
- -x ./OBS.app/Contents/MacOS/obs-ffmpeg-mux \64
- -x ./OBS.app/Contents/MacOS/obslua.so \65
- -x ./OBS.app/Contents/PlugIns/obs-x264.so \66
- -x ./OBS.app/Contents/PlugIns/text-freetype2.so \67
- -x ./OBS.app/Contents/PlugIns/obs-libfdk.so \68
- -x ./OBS.app/Contents/PlugIns/obs-outputs.so69
- step "Move libobs-opengl to final destination"70
71
+ BUNDLE_PLUGINS=(72
+ ./OBS.app/Contents/PlugIns/coreaudio-encoder.so73
+ ./OBS.app/Contents/PlugIns/decklink-ouput-ui.so74
+ ./OBS.app/Contents/PlugIns/decklink-captions.so75
+ ./OBS.app/Contents/PlugIns/frontend-tools.so76
+ ./OBS.app/Contents/PlugIns/image-source.so77
+ ./OBS.app/Contents/PlugIns/mac-avcapture.so78
+ ./OBS.app/Contents/PlugIns/mac-capture.so79
+ ./OBS.app/Contents/PlugIns/mac-decklink.so80
+ ./OBS.app/Contents/PlugIns/mac-syphon.so81
+ ./OBS.app/Contents/PlugIns/mac-vth264.so82
+ ./OBS.app/Contents/PlugIns/mac-virtualcam.so83
+ ./OBS.app/Contents/PlugIns/obs-browser.so84
+ ./OBS.app/Contents/PlugIns/obs-ffmpeg.so85
+ ./OBS.app/Contents/PlugIns/obs-filters.so86
+ ./OBS.app/Contents/PlugIns/obs-transitions.so87
+ ./OBS.app/Contents/PlugIns/obs-vst.so88
+ ./OBS.app/Contents/PlugIns/rtmp-services.so89
+ ./OBS.app/Contents/MacOS/obs-ffmpeg-mux90
+ ./OBS.app/Contents/MacOS/obslua.so91
+ ./OBS.app/Contents/PlugIns/obs-x264.so92
+ ./OBS.app/Contents/PlugIns/text-freetype2.so93
+ ./OBS.app/Contents/PlugIns/obs-outputs.so94
+ )95
+ if ! [ "${CEF_BUILD_VERSION:-${CI_CEF_VERSION}}" -le 3770 ]; then96
+ ${CI_SCRIPTS}/app/dylibbundler -cd -of -a ./OBS.app -q -f \97
+ -s ./OBS.app/Contents/MacOS \98
+ -s "${DEPS_BUILD_DIR}/sparkle/Sparkle.framework" \99
+ -s ./rundir/${BUILD_CONFIG}/bin/ \100
+ $(echo "${BUNDLE_PLUGINS[@]/#/-x }")101
+ else102
+ ${CI_SCRIPTS}/app/dylibbundler -cd -of -a ./OBS.app -q -f \103
+ -s ./OBS.app/Contents/MacOS \104
+ -s "${DEPS_BUILD_DIR}/sparkle/Sparkle.framework" \105
+ -s ./rundir/${BUILD_CONFIG}/bin/ \106
+ $(echo "${BUNDLE_PLUGINS[@]/#/-x }") \107
+ -x ./OBS.app/Contents/PlugIns/obs-browser-page108
+ fi109
+110
+ step "Move libobs-opengl to final destination"111
if [ -f "./libobs-opengl/libobs-opengl.so" ]; then112
cp ./libobs-opengl/libobs-opengl.so ./OBS.app/Contents/Frameworks113
else114
115
mkdir -p OBS.app/Contents/MacOS116
mkdir OBS.app/Contents/PlugIns117
mkdir OBS.app/Contents/Resources118
+ mkdir OBS.app/Contents/Frameworks119
120
cp rundir/${BUILD_CONFIG}/bin/obs ./OBS.app/Contents/MacOS121
cp rundir/${BUILD_CONFIG}/bin/obs-ffmpeg-mux ./OBS.app/Contents/MacOS122
cp rundir/${BUILD_CONFIG}/bin/libobsglad.0.dylib ./OBS.app/Contents/MacOS123
+ if ! [ "${CEF_BUILD_VERSION:-${CI_CEF_VERSION}}" -le 3770 ]; then124
+ cp -R "rundir/${BUILD_CONFIG}/bin/OBS Helper.app" "./OBS.app/Contents/Frameworks/OBS Helper.app"125
+ cp -R "rundir/${BUILD_CONFIG}/bin/OBS Helper (GPU).app" "./OBS.app/Contents/Frameworks/OBS Helper (GPU).app"126
+ cp -R "rundir/${BUILD_CONFIG}/bin/OBS Helper (Plugin).app" "./OBS.app/Contents/Frameworks/OBS Helper (Plugin).app"127
+ cp -R "rundir/${BUILD_CONFIG}/bin/OBS Helper (Renderer).app" "./OBS.app/Contents/Frameworks/OBS Helper (Renderer).app"128
+ fi129
cp -R rundir/${BUILD_CONFIG}/data ./OBS.app/Contents/Resources130
cp ${CI_SCRIPTS}/app/AppIcon.icns ./OBS.app/Contents/Resources131
cp -R rundir/${BUILD_CONFIG}/obs-plugins/ ./OBS.app/Contents/PlugIns132
133
codesign --force --options runtime --sign "${CODESIGN_IDENT}" "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework/Libraries/libswiftshader_libEGL.dylib"134
codesign --force --options runtime --sign "${CODESIGN_IDENT}" "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework/Libraries/libGLESv2.dylib"135
codesign --force --options runtime --sign "${CODESIGN_IDENT}" "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework/Libraries/libswiftshader_libGLESv2.dylib"136
- codesign --force --options runtime --sign "${CODESIGN_IDENT}" --deep "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework"137
+ if ! [ "${CEF_BUILD_VERSION:-${CI_CEF_VERSION}}" -le 3770 ]; then138
+ codesign --force --options runtime --sign "${CODESIGN_IDENT}" "./OBS.app/Contents/Frameworks/Chromium Embedded Framework.framework/Libraries/libvk_swiftshader.dylib"139
+ fi140
+141
echo -n "${COLOR_RESET}"142
143
step "Code-sign DAL Plugin..."144
145
echo -n "${COLOR_ORANGE}"146
codesign --force --options runtime --entitlements "${CI_SCRIPTS}/app/entitlements.plist" --sign "${CODESIGN_IDENT}" --deep ./OBS.app147
echo -n "${COLOR_RESET}"148
+149
+ if ! [ "${CEF_BUILD_VERSION:-${CI_CEF_VERSION}}" -le 3770 ]; then150
+ step "Code-sign CEF helper apps..."151
+ echo -n "${COLOR_ORANGE}"152
+ codesign --force --options runtime --sign "${CODESIGN_IDENT}" --deep "./OBS.app/Contents/Frameworks/OBS Helper.app"153
+ codesign --force --options runtime --entitlements "${CI_SCRIPTS}/helpers/helper-gpu-entitlements.plist" --sign "${CODESIGN_IDENT}" --deep "./OBS.app/Contents/Frameworks/OBS Helper (GPU).app"154
+ codesign --force --options runtime --entitlements "${CI_SCRIPTS}/helpers/helper-plugin-entitlements.plist" --sign "${CODESIGN_IDENT}" --deep "./OBS.app/Contents/Frameworks/OBS Helper (Plugin).app"155
+ codesign --force --options runtime --entitlements "${CI_SCRIPTS}/helpers/helper-renderer-entitlements.plist" --sign "${CODESIGN_IDENT}" --deep "./OBS.app/Contents/Frameworks/OBS Helper (Renderer).app"156
+ echo -n "${COLOR_RESET}"157
+ fi158
+159
step "Check code-sign result..."160
codesign -dvv ./OBS.app161
}162
obs-studio-26.1.0.tar.xz/CI/scripts/macos/Brewfile -> obs-studio-26.1.1.tar.xz/CI/scripts/macos/Brewfile
Changed
11
1
2
tap "akeru-inc/tap"3
-brew "jack"4
-brew "speexdsp"5
brew "cmake"6
brew "freetype"7
-brew "fdk-aac"8
brew "cmocka"9
brew "akeru-inc/tap/xcnotary"10
\ No newline at end of file11
obs-studio-26.1.1.tar.xz/CI/scripts/macos/helpers
Added
2
1
+(directory)2
obs-studio-26.1.1.tar.xz/CI/scripts/macos/helpers/helper-gpu-entitlements.plist
Added
11
1
2
+<?xml version="1.0" encoding="UTF-8"?>3
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">4
+<plist version="1.0">5
+<dict>6
+ <key>com.apple.security.cs.allow-jit</key>7
+ <true/>8
+</dict>9
+</plist>10
\ No newline at end of file11
obs-studio-26.1.1.tar.xz/CI/scripts/macos/helpers/helper-plugin-entitlements.plist
Added
13
1
2
+<?xml version="1.0" encoding="UTF-8"?>3
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">4
+<plist version="1.0">5
+<dict>6
+ <key>com.apple.security.cs.allow-unsigned-executable-memory</key>7
+ <true/>8
+ <key>com.apple.security.cs.disable-library-validation</key>9
+ <true/>10
+</dict>11
+</plist>12
\ No newline at end of file13
obs-studio-26.1.1.tar.xz/CI/scripts/macos/helpers/helper-renderer-entitlements.plist
Added
11
1
2
+<?xml version="1.0" encoding="UTF-8"?>3
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">4
+<plist version="1.0">5
+<dict>6
+ <key>com.apple.security.cs.allow-jit</key>7
+ <true/>8
+</dict>9
+</plist>10
\ No newline at end of file11
obs-studio-26.1.0.tar.xz/CMakeLists.txt -> obs-studio-26.1.1.tar.xz/CMakeLists.txt
Changed
20
1
2
endif ()3
4
if(LOWERCASE_CMAKE_SYSTEM_PROCESSOR MATCHES "(i[3-6]86|x86|x64|x86_64|amd64|e2k)")5
- set(NEEDS_SIMDE "0")6
if(NOT MSVC)7
set(ARCH_SIMD_FLAGS "-mmmx" "-msse" "-msse2")8
endif()9
elseif(LOWERCASE_CMAKE_SYSTEM_PROCESSOR MATCHES "^(powerpc|ppc)64(le)?")10
- set(NEEDS_SIMDE "0")11
+ set(ARCH_SIMD_DEFINES "-DNO_WARN_X86_INTRINSICS")12
set(ARCH_SIMD_FLAGS "-mvsx")13
add_compile_definitions(NO_WARN_X86_INTRINSICS)14
else()15
- set(NEEDS_SIMDE "1")16
- add_definitions(-DNEEDS_SIMDE=1)17
if(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX)18
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DSIMDE_ENABLE_OPENMP")19
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -DSIMDE_ENABLE_OPENMP")20
obs-studio-26.1.0.tar.xz/UI/data/locale.ini -> obs-studio-26.1.1.tar.xz/UI/data/locale.ini
Changed
7
1
2
Name=Esperanto3
4
[kab-KAB]5
-Name=Taglizit6
+Name=Taqbaylit7
obs-studio-26.1.0.tar.xz/UI/data/themes/Acri.qss -> obs-studio-26.1.1.tar.xz/UI/data/themes/Acri.qss
Changed
12
1
2
max-height: 40px;3
}4
5
-#contextContainer QPushButton[themeID2=contextBarButton] {6
- padding: 0px;7
+#contextContainer QPushButton {8
+ padding: 0px 12px;9
}10
11
QPushButton#sourcePropertiesButton {12
obs-studio-26.1.0.tar.xz/UI/installer/mp-installer.nsi -> obs-studio-26.1.1.tar.xz/UI/installer/mp-installer.nsi
Changed
9
1
2
ClearErrors3
GetDLLVersion "vcruntime140.DLL" $R0 $R14
GetDLLVersion "msvcp140.DLL" $R0 $R15
+ GetDLLVersion "msvcp140_1.DLL" $R0 $R16
IfErrors vs2019Missing_32 vs2019OK_327
vs2019Missing_32:8
MessageBox MB_YESNO|MB_ICONEXCLAMATION "Your system is missing runtime components that ${APPNAME} requires. Would you like to download them?" IDYES vs2019true_32 IDNO vs2019false_329
obs-studio-26.1.0.tar.xz/UI/win-update/updater/updater.cpp -> obs-studio-26.1.1.tar.xz/UI/win-update/updater/updater.cpp
Changed
119
1
2
} else {3
DeleteFile(outputPath.c_str());4
}5
+ if (state == STATE_INSTALL_FAILED)6
+ DeleteFile(tempPath.c_str());7
} else if (state == STATE_DOWNLOADED) {8
DeleteFile(tempPath.c_str());9
}10
11
12
bool DownloadWorkerThread()13
{14
- const DWORD tlsProtocols = WINHTTP_FLAG_SECURE_PROTOCOL_TLS1_2;15
+ const DWORD tlsProtocols = WINHTTP_FLAG_SECURE_PROTOCOL_TLS1_2 |16
+ WINHTTP_FLAG_SECURE_PROTOCOL_TLS1_3;17
+18
+ const DWORD enableHTTP2Flag = WINHTTP_PROTOCOL_FLAG_HTTP2;19
20
HttpHandle hSession = WinHttpOpen(L"OBS Studio Updater/2.1",21
WINHTTP_ACCESS_TYPE_DEFAULT_PROXY,22
23
WinHttpSetOption(hSession, WINHTTP_OPTION_SECURE_PROTOCOLS,24
(LPVOID)&tlsProtocols, sizeof(tlsProtocols));25
26
+ WinHttpSetOption(hSession, WINHTTP_OPTION_ENABLE_HTTP_PROTOCOL,27
+ (LPVOID)&enableHTTP2Flag, sizeof(enableHTTP2Flag));28
+29
HttpHandle hConnect = WinHttpConnect(hSession,30
L"cdn-fastly.obsproject.com",31
INTERNET_DEFAULT_HTTPS_PORT, 0);32
33
}34
}35
36
+static bool MoveInUseFileAway(update_t &file)37
+{38
+ _TCHAR deleteMeName[MAX_PATH];39
+ _TCHAR randomStr[MAX_PATH];40
+41
+ BYTE junk[40];42
+ BYTE hash[BLAKE2_HASH_LENGTH];43
+44
+ CryptGenRandom(hProvider, sizeof(junk), junk);45
+ blake2b(hash, sizeof(hash), junk, sizeof(junk), NULL, 0);46
+ HashToString(hash, randomStr);47
+ randomStr[8] = 0;48
+49
+ StringCbCopy(deleteMeName, sizeof(deleteMeName),50
+ file.outputPath.c_str());51
+52
+ StringCbCat(deleteMeName, sizeof(deleteMeName), L".");53
+ StringCbCat(deleteMeName, sizeof(deleteMeName), randomStr);54
+ StringCbCat(deleteMeName, sizeof(deleteMeName), L".deleteme");55
+56
+ if (MoveFile(file.outputPath.c_str(), deleteMeName)) {57
+58
+ if (MyCopyFile(deleteMeName, file.outputPath.c_str())) {59
+ MoveFileEx(deleteMeName, NULL,60
+ MOVEFILE_DELAY_UNTIL_REBOOT);61
+62
+ return true;63
+ } else {64
+ MoveFile(deleteMeName, file.outputPath.c_str());65
+ }66
+ }67
+68
+ return false;69
+}70
+71
static bool UpdateFile(update_t &file)72
{73
wchar_t oldFileRenamedPath[MAX_PATH];74
75
76
int error_code;77
bool installed_ok;78
+ bool already_tried_to_move = false;79
+80
+ retryAfterMovingFile:81
82
if (file.patchable) {83
error_code = ApplyPatch(file.tempPath.c_str(),84
85
int is_sharing_violation =86
(error_code == ERROR_SHARING_VIOLATION);87
88
- if (is_sharing_violation)89
+ if (is_sharing_violation) {90
+ if (!already_tried_to_move) {91
+ already_tried_to_move = true;92
+93
+ if (MoveInUseFileAway(file))94
+ goto retryAfterMovingFile;95
+ }96
+97
Status(L"Update failed: %s is still in use. "98
L"Close all "99
L"programs and try again.",100
curFileName);101
- else102
+ } else {103
Status(L"Update failed: Couldn't update %s "104
L"(error %d)",105
curFileName, GetLastError());106
+ }107
108
file.state = STATE_INSTALL_FAILED;109
return false;110
111
/* ------------------------------------- *112
* Download Updates */113
114
- if (!RunDownloadWorkers(2))115
+ if (!RunDownloadWorkers(4))116
return false;117
118
if ((size_t)completedUpdates != updates.size()) {119
obs-studio-26.1.0.tar.xz/UI/window-basic-auto-config.cpp -> obs-studio-26.1.1.tar.xz/UI/window-basic-auto-config.cpp
Changed
41
1
2
if (!wiz->customServer) {3
if (wiz->serviceName == "Twitch")4
wiz->service = AutoConfig::Service::Twitch;5
- else if (wiz->serviceName == "Smashcast")6
- wiz->service = AutoConfig::Service::Smashcast;7
else8
wiz->service = AutoConfig::Service::Other;9
} else {10
11
return;12
13
std::string service = QT_TO_UTF8(ui->service->currentText());14
- bool regionBased = service == "Twitch" || service == "Smashcast";15
+ bool regionBased = service == "Twitch";16
bool testBandwidth = ui->doBandwidthTest->isChecked();17
bool custom = IsCustomService();18
19
20
} else if (regionOther) {21
return true;22
}23
- } else if (service == Service::Smashcast) {24
- if (strcmp(server, "Default") == 0) {25
- return true;26
- } else if (astrcmp_n(server, "US-West:", 8) == 0 ||27
- astrcmp_n(server, "US-East:", 8) == 0) {28
- return regionUS;29
- } else if (astrcmp_n(server, "EU-", 3) == 0) {30
- return regionEU;31
- } else if (astrcmp_n(server, "South Korea:", 12) == 0 ||32
- astrcmp_n(server, "Asia:", 5) == 0 ||33
- astrcmp_n(server, "China:", 6) == 0) {34
- return regionAsia;35
- } else if (regionOther) {36
- return true;37
- }38
} else {39
return true;40
}41
obs-studio-26.1.0.tar.xz/UI/window-basic-auto-config.hpp -> obs-studio-26.1.1.tar.xz/UI/window-basic-auto-config.hpp
Changed
9
1
2
3
enum class Service {4
Twitch,5
- Smashcast,6
Other,7
};8
9
obs-studio-26.1.0.tar.xz/UI/window-basic-main.cpp -> obs-studio-26.1.1.tar.xz/UI/window-basic-main.cpp
Changed
16
1
2
uint32_t cx = primaryScreen->size().width();3
uint32_t cy = primaryScreen->size().height();4
5
+#ifdef SUPPORTS_FRACTIONAL_SCALING6
+ cx *= devicePixelRatioF();7
+ cy *= devicePixelRatioF();8
+#elif9
+ cx *= devicePixelRatio();10
+ cy *= devicePixelRatio();11
+#endif12
+13
bool oldResolutionDefaults = config_get_bool(14
App()->GlobalConfig(), "General", "Pre19Defaults");15
16
obs-studio-26.1.0.tar.xz/cmake/Modules/FindLibcurl.cmake -> obs-studio-26.1.1.tar.xz/cmake/Modules/FindLibcurl.cmake
Changed
70
1
2
PATH_SUFFIXES3
include)4
5
-find_library(CURL_LIB6
- NAMES ${_CURL_LIBRARIES} curl libcurl7
- HINTS8
- ENV curlPath${_lib_suffix}9
- ENV curlPath10
- ENV DepsPath${_lib_suffix}11
- ENV DepsPath12
- ${curlPath${_lib_suffix}}13
- ${curlPath}14
- ${DepsPath${_lib_suffix}}15
- ${DepsPath}16
- ${_CURL_LIBRARY_DIRS}17
- PATHS18
- /usr/lib /usr/local/lib /opt/local/lib /sw/lib19
- PATH_SUFFIXES20
- lib${_lib_suffix} lib21
- libs${_lib_suffix} libs22
- bin${_lib_suffix} bin23
- ../lib${_lib_suffix} ../lib24
- ../libs${_lib_suffix} ../libs25
- ../bin${_lib_suffix} ../bin26
- "build/Win${_lib_suffix}/VC12/DLL Release - DLL Windows SSPI"27
- "../build/Win${_lib_suffix}/VC12/DLL Release - DLL Windows SSPI")28
+if(APPLE)29
+ find_library(CURL_LIB30
+ NAMES ${_CURL_LIBRARIES} curl libcurl31
+ HINTS32
+ ENV curlPath${_lib_suffix}33
+ ENV curlPath34
+ ENV DepsPath${_lib_suffix}35
+ ENV DepsPath36
+ ${curlPath${_lib_suffix}}37
+ ${curlPath}38
+ ${DepsPath${_lib_suffix}}39
+ ${DepsPath}40
+ ${_CURL_LIBRARY_DIRS}41
+ )42
+else()43
+ find_library(CURL_LIB44
+ NAMES ${_CURL_LIBRARIES} curl libcurl45
+ HINTS46
+ ENV curlPath${_lib_suffix}47
+ ENV curlPath48
+ ENV DepsPath${_lib_suffix}49
+ ENV DepsPath50
+ ${curlPath${_lib_suffix}}51
+ ${curlPath}52
+ ${DepsPath${_lib_suffix}}53
+ ${DepsPath}54
+ ${_CURL_LIBRARY_DIRS}55
+ PATHS56
+ /usr/lib /usr/local/lib /opt/local/lib /sw/lib57
+ PATH_SUFFIXES58
+ lib${_lib_suffix} lib59
+ libs${_lib_suffix} libs60
+ bin${_lib_suffix} bin61
+ ../lib${_lib_suffix} ../lib62
+ ../libs${_lib_suffix} ../libs63
+ ../bin${_lib_suffix} ../bin64
+ "build/Win${_lib_suffix}/VC12/DLL Release - DLL Windows SSPI"65
+ "../build/Win${_lib_suffix}/VC12/DLL Release - DLL Windows SSPI")66
+endif()67
68
include(FindPackageHandleStandardArgs)69
find_package_handle_standard_args(Libcurl DEFAULT_MSG CURL_LIB CURL_INCLUDE_DIR)70
obs-studio-26.1.0.tar.xz/docs/sphinx/reference-frontend-api.rst -> obs-studio-26.1.1.tar.xz/docs/sphinx/reference-frontend-api.rst
Changed
17
1
2
3
---------------------------------------4
5
+.. function:: void obs_frontend_open_projector(const char *type, int monitor, const char *geometry, const char *name)6
+7
+ :param type: "Preview", "Source", "Scene", "StudioProgram", or "Multiview" (case insensitive).8
+ :param monitor: Monitor to open the projector on. If -1, opens a window.9
+ :param geometry: If *monitor* is -1, size and position of the projector window. Encoded in Base64 using Qt's geometry encoding.10
+ :param name: If *type* is "Source" or "Scene", name of the source or scene to be displayed.11
+12
+---------------------------------------13
+14
.. function:: void obs_frontend_save(void)15
16
Saves the current scene collection.17
obs-studio-26.1.0.tar.xz/libobs/CMakeLists.txt -> obs-studio-26.1.1.tar.xz/libobs/CMakeLists.txt
Changed
64
1
2
util/pipe-posix.c3
util/platform-nix.c)4
5
- if(NEEDS_SIMDE)6
- set(libobs_PLATFORM_HEADERS7
- util/simde/check.h8
- util/simde/hedley.h9
- util/simde/mmx.h10
- util/simde/simde-arch.h11
- util/simde/simde-common.h12
- util/simde/sse.h13
- util/simde/sse2.h14
- util/threading-posix.h)15
- else()16
- set(libobs_PLATFORM_HEADERS17
- util/threading-posix.h)18
- endif()19
+ set(libobs_PLATFORM_HEADERS20
+ util/threading-posix.h)21
22
if(HAVE_PULSEAUDIO)23
set(libobs_audio_monitoring_HEADERS24
25
set(libobs_util_HEADERS26
util/curl/curl-helper.h27
util/sse-intrin.h28
- util/sse2neon.h29
util/array-serializer.h30
util/file-serializer.h31
util/utf8.h32
33
obs-video-gpu-encode.c34
obs-video.c)35
set(libobs_libobs_HEADERS36
+ util/simde/check.h37
+ util/simde/debug-trap.h38
+ util/simde/hedley.h39
+ util/simde/simde-align.h40
+ util/simde/simde-arch.h41
+ util/simde/simde-common.h42
+ util/simde/simde-constify.h43
+ util/simde/simde-detect-clang.h44
+ util/simde/simde-diagnostic.h45
+ util/simde/simde-features.h46
+ util/simde/simde-math.h47
+ util/simde/x86/mmx.h48
+ util/simde/x86/sse2.h49
+ util/simde/x86/sse.h50
${libobs_PLATFORM_HEADERS}51
obs-audio-controls.h52
obs-defs.h53
54
PUBLIC55
HAVE_OBSCONFIG_H)56
57
+target_compile_definitions(libobs58
+ PUBLIC59
+ ${ARCH_SIMD_DEFINES})60
+61
target_compile_options(libobs62
PUBLIC63
${ARCH_SIMD_FLAGS})64
obs-studio-26.1.0.tar.xz/libobs/media-io/media-remux.c -> obs-studio-26.1.1.tar.xz/libobs/media-io/media-remux.c
Changed
10
1
2
3
/* Treat "Invalid data found when processing input" and4
* "Invalid argument" as non-fatal */5
- if (ret == AVERROR_INVALIDDATA || ret == EINVAL)6
+ if (ret == AVERROR_INVALIDDATA || ret == -EINVAL)7
continue;8
9
break;10
obs-studio-26.1.0.tar.xz/libobs/obs-config.h -> obs-studio-26.1.1.tar.xz/libobs/obs-config.h
Changed
10
1
2
*3
* Reset to zero each major or minor version4
*/5
-#define LIBOBS_API_PATCH_VER 06
+#define LIBOBS_API_PATCH_VER 17
8
#define MAKE_SEMANTIC_VERSION(major, minor, patch) \9
((major << 24) | (minor << 16) | patch)10
obs-studio-26.1.0.tar.xz/libobs/obs-scene.c -> obs-studio-26.1.1.tar.xz/libobs/obs-scene.c
Changed
65
1
2
}3
4
static void apply_scene_item_audio_actions(struct obs_scene_item *item,5
- float **p_buf, uint64_t ts,6
+ float *buf, uint64_t ts,7
size_t sample_rate)8
{9
bool cur_visible = item->visible;10
uint64_t frame_num = 0;11
size_t deref_count = 0;12
- float *buf = NULL;13
-14
- if (p_buf) {15
- if (!*p_buf)16
- *p_buf = malloc(AUDIO_OUTPUT_FRAMES * sizeof(float));17
- buf = *p_buf;18
- }19
20
pthread_mutex_lock(&item->actions_mutex);21
22
23
}24
}25
26
-static bool apply_scene_item_volume(struct obs_scene_item *item, float **buf,27
+static bool apply_scene_item_volume(struct obs_scene_item *item, float *buf,28
uint64_t ts, size_t sample_rate)29
{30
bool actions_pending;31
32
size_t sample_rate)33
{34
uint64_t timestamp = 0;35
- float *buf = NULL;36
+ float buf[AUDIO_OUTPUT_FRAMES];37
struct obs_source_audio_mix child_audio;38
struct obs_scene *scene = data;39
struct obs_scene_item *item;40
41
size_t pos, count;42
bool apply_buf;43
44
- apply_buf = apply_scene_item_volume(item, &buf, timestamp,45
+ apply_buf = apply_scene_item_volume(item, buf, timestamp,46
sample_rate);47
48
if (obs_source_audio_pending(item->source)) {49
50
*ts_out = timestamp;51
audio_unlock(scene);52
53
- free(buf);54
return true;55
}56
57
58
}59
60
obs_sceneitem_set_crop(dst, &src->crop);61
+ obs_sceneitem_set_locked(dst, src->locked);62
63
if (defer_texture_update) {64
os_atomic_set_bool(&dst->update_transform, true);65
obs-studio-26.1.0.tar.xz/libobs/obs-source.c -> obs-studio-26.1.1.tar.xz/libobs/obs-source.c
Changed
19
1
2
static void apply_audio_actions(obs_source_t *source, size_t channels,3
size_t sample_rate)4
{5
- float *vol_data = malloc(sizeof(float) * AUDIO_OUTPUT_FRAMES);6
+ float vol_data[AUDIO_OUTPUT_FRAMES];7
float cur_vol = get_source_volume(source, source->audio_ts);8
size_t frame_num = 0;9
10
11
if ((source->audio_mixers & (1 << mix)) != 0)12
multiply_vol_data(source, mix, channels, vol_data);13
}14
-15
- free(vol_data);16
}17
18
static void apply_audio_volume(obs_source_t *source, uint32_t mixers,19
obs-studio-26.1.0.tar.xz/libobs/obsconfig.h.in -> obs-studio-26.1.1.tar.xz/libobs/obsconfig.h.in
Changed
9
1
2
#define HAVE_DBUS @HAVE_DBUS@3
#define HAVE_PULSEAUDIO @HAVE_PULSEAUDIO@4
#define USE_XINPUT @USE_XINPUT@5
-#define NEEDS_SIMDE @NEEDS_SIMDE@6
#define LIBOBS_IMAGEMAGICK_DIR_STYLE_6L 67
#define LIBOBS_IMAGEMAGICK_DIR_STYLE_7GE 78
#define LIBOBS_IMAGEMAGICK_DIR_STYLE @LIBOBS_IMAGEMAGICK_DIR_STYLE@9
obs-studio-26.1.0.tar.xz/libobs/util/simde/README.libobs -> obs-studio-26.1.1.tar.xz/libobs/util/simde/README.libobs
Changed
10
1
2
-This is a slightly modified version of https://github.com/nemequ/simde/commit/cafec4b952fa5a31a51a10326f97c2e7c90677713
-sse{,2}.h and mmx.h was moved down from the original "x86" subdirectory,4
-subsequently the '#include "../simde-common.h"' line in mmx.h was changed to '#include "simde-common.h"'5
+This is a slightly modified version of the simde directory in6
+https://github.com/simd-everywhere/simde/commit/c3d7abfaba6729a8b11d09a314b34a4db628911d7
+Unused files have removed.8
9
Then the code was reformatted using the "formatcode.sh" script in the root of this repository.10
obs-studio-26.1.0.tar.xz/libobs/util/simde/check.h -> obs-studio-26.1.1.tar.xz/libobs/util/simde/check.h
Changed
9
1
2
#endif3
4
#include "hedley.h"5
+#include "simde-diagnostic.h"6
#include <stdint.h>7
8
#if !defined(_WIN32)9
obs-studio-26.1.0.tar.xz/libobs/util/simde/hedley.h -> obs-studio-26.1.1.tar.xz/libobs/util/simde/hedley.h
Changed
564
1
2
* SPDX-License-Identifier: CC0-1.03
*/4
5
-#if !defined(HEDLEY_VERSION) || (HEDLEY_VERSION < 12)6
+#if !defined(HEDLEY_VERSION) || (HEDLEY_VERSION < 14)7
#if defined(HEDLEY_VERSION)8
#undef HEDLEY_VERSION9
#endif10
-#define HEDLEY_VERSION 1211
+#define HEDLEY_VERSION 1412
13
#if defined(HEDLEY_STRINGIFY_EX)14
#undef HEDLEY_STRINGIFY_EX15
16
#endif17
#define HEDLEY_CONCAT(a, b) HEDLEY_CONCAT_EX(a, b)18
19
+#if defined(HEDLEY_CONCAT3_EX)20
+#undef HEDLEY_CONCAT3_EX21
+#endif22
+#define HEDLEY_CONCAT3_EX(a, b, c) a##b##c23
+24
+#if defined(HEDLEY_CONCAT3)25
+#undef HEDLEY_CONCAT326
+#endif27
+#define HEDLEY_CONCAT3(a, b, c) HEDLEY_CONCAT3_EX(a, b, c)28
+29
#if defined(HEDLEY_VERSION_ENCODE)30
#undef HEDLEY_VERSION_ENCODE31
#endif32
33
#if defined(HEDLEY_MSVC_VERSION)34
#undef HEDLEY_MSVC_VERSION35
#endif36
-#if defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 140000000)37
+#if defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 140000000) && !defined(__ICL)38
#define HEDLEY_MSVC_VERSION \39
HEDLEY_VERSION_ENCODE(_MSC_FULL_VER / 10000000, \40
(_MSC_FULL_VER % 10000000) / 100000, \41
(_MSC_FULL_VER % 100000) / 100)42
-#elif defined(_MSC_FULL_VER)43
+#elif defined(_MSC_FULL_VER) && !defined(__ICL)44
#define HEDLEY_MSVC_VERSION \45
HEDLEY_VERSION_ENCODE(_MSC_FULL_VER / 1000000, \46
(_MSC_FULL_VER % 1000000) / 10000, \47
(_MSC_FULL_VER % 10000) / 10)48
-#elif defined(_MSC_VER)49
+#elif defined(_MSC_VER) && !defined(__ICL)50
#define HEDLEY_MSVC_VERSION \51
HEDLEY_VERSION_ENCODE(_MSC_VER / 100, _MSC_VER % 100, 0)52
#endif53
54
#if defined(HEDLEY_MSVC_VERSION_CHECK)55
#undef HEDLEY_MSVC_VERSION_CHECK56
#endif57
-#if !defined(_MSC_VER)58
+#if !defined(HEDLEY_MSVC_VERSION)59
#define HEDLEY_MSVC_VERSION_CHECK(major, minor, patch) (0)60
#elif defined(_MSC_VER) && (_MSC_VER >= 1400)61
#define HEDLEY_MSVC_VERSION_CHECK(major, minor, patch) \62
63
#if defined(HEDLEY_INTEL_VERSION)64
#undef HEDLEY_INTEL_VERSION65
#endif66
-#if defined(__INTEL_COMPILER) && defined(__INTEL_COMPILER_UPDATE)67
+#if defined(__INTEL_COMPILER) && defined(__INTEL_COMPILER_UPDATE) && \68
+ !defined(__ICL)69
#define HEDLEY_INTEL_VERSION \70
HEDLEY_VERSION_ENCODE(__INTEL_COMPILER / 100, __INTEL_COMPILER % 100, \71
__INTEL_COMPILER_UPDATE)72
-#elif defined(__INTEL_COMPILER)73
+#elif defined(__INTEL_COMPILER) && !defined(__ICL)74
#define HEDLEY_INTEL_VERSION \75
HEDLEY_VERSION_ENCODE(__INTEL_COMPILER / 100, __INTEL_COMPILER % 100, 0)76
#endif77
78
#define HEDLEY_INTEL_VERSION_CHECK(major, minor, patch) (0)79
#endif80
81
+#if defined(HEDLEY_INTEL_CL_VERSION)82
+#undef HEDLEY_INTEL_CL_VERSION83
+#endif84
+#if defined(__INTEL_COMPILER) && defined(__INTEL_COMPILER_UPDATE) && \85
+ defined(__ICL)86
+#define HEDLEY_INTEL_CL_VERSION \87
+ HEDLEY_VERSION_ENCODE(__INTEL_COMPILER, __INTEL_COMPILER_UPDATE, 0)88
+#endif89
+90
+#if defined(HEDLEY_INTEL_CL_VERSION_CHECK)91
+#undef HEDLEY_INTEL_CL_VERSION_CHECK92
+#endif93
+#if defined(HEDLEY_INTEL_CL_VERSION)94
+#define HEDLEY_INTEL_CL_VERSION_CHECK(major, minor, patch) \95
+ (HEDLEY_INTEL_CL_VERSION >= HEDLEY_VERSION_ENCODE(major, minor, patch))96
+#else97
+#define HEDLEY_INTEL_CL_VERSION_CHECK(major, minor, patch) (0)98
+#endif99
+100
#if defined(HEDLEY_PGI_VERSION)101
#undef HEDLEY_PGI_VERSION102
#endif103
104
HEDLEY_GCC_VERSION_CHECK(major, minor, patch)105
#endif106
107
+#if (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) || \108
+ defined(__clang__) || HEDLEY_GCC_VERSION_CHECK(3, 0, 0) || \109
+ HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \110
+ HEDLEY_IAR_VERSION_CHECK(8, 0, 0) || \111
+ HEDLEY_PGI_VERSION_CHECK(18, 4, 0) || \112
+ HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \113
+ HEDLEY_TI_VERSION_CHECK(15, 12, 0) || \114
+ HEDLEY_TI_ARMCL_VERSION_CHECK(4, 7, 0) || \115
+ HEDLEY_TI_CL430_VERSION_CHECK(2, 0, 1) || \116
+ HEDLEY_TI_CL2000_VERSION_CHECK(6, 1, 0) || \117
+ HEDLEY_TI_CL6X_VERSION_CHECK(7, 0, 0) || \118
+ HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \119
+ HEDLEY_TI_CLPRU_VERSION_CHECK(2, 1, 0) || \120
+ HEDLEY_CRAY_VERSION_CHECK(5, 0, 0) || \121
+ HEDLEY_TINYC_VERSION_CHECK(0, 9, 17) || \122
+ HEDLEY_SUNPRO_VERSION_CHECK(8, 0, 0) || \123
+ (HEDLEY_IBM_VERSION_CHECK(10, 1, 0) && defined(__C99_PRAGMA_OPERATOR))124
+#define HEDLEY_PRAGMA(value) _Pragma(#value)125
+#elif HEDLEY_MSVC_VERSION_CHECK(15, 0, 0)126
+#define HEDLEY_PRAGMA(value) __pragma(value)127
+#else128
+#define HEDLEY_PRAGMA(value)129
+#endif130
+131
+#if defined(HEDLEY_DIAGNOSTIC_PUSH)132
+#undef HEDLEY_DIAGNOSTIC_PUSH133
+#endif134
+#if defined(HEDLEY_DIAGNOSTIC_POP)135
+#undef HEDLEY_DIAGNOSTIC_POP136
+#endif137
+#if defined(__clang__)138
+#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("clang diagnostic push")139
+#define HEDLEY_DIAGNOSTIC_POP _Pragma("clang diagnostic pop")140
+#elif HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)141
+#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("warning(push)")142
+#define HEDLEY_DIAGNOSTIC_POP _Pragma("warning(pop)")143
+#elif HEDLEY_GCC_VERSION_CHECK(4, 6, 0)144
+#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("GCC diagnostic push")145
+#define HEDLEY_DIAGNOSTIC_POP _Pragma("GCC diagnostic pop")146
+#elif HEDLEY_MSVC_VERSION_CHECK(15, 0, 0) || \147
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)148
+#define HEDLEY_DIAGNOSTIC_PUSH __pragma(warning(push))149
+#define HEDLEY_DIAGNOSTIC_POP __pragma(warning(pop))150
+#elif HEDLEY_ARM_VERSION_CHECK(5, 6, 0)151
+#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("push")152
+#define HEDLEY_DIAGNOSTIC_POP _Pragma("pop")153
+#elif HEDLEY_TI_VERSION_CHECK(15, 12, 0) || \154
+ HEDLEY_TI_ARMCL_VERSION_CHECK(5, 2, 0) || \155
+ HEDLEY_TI_CL430_VERSION_CHECK(4, 4, 0) || \156
+ HEDLEY_TI_CL6X_VERSION_CHECK(8, 1, 0) || \157
+ HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \158
+ HEDLEY_TI_CLPRU_VERSION_CHECK(2, 1, 0)159
+#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("diag_push")160
+#define HEDLEY_DIAGNOSTIC_POP _Pragma("diag_pop")161
+#elif HEDLEY_PELLES_VERSION_CHECK(2, 90, 0)162
+#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("warning(push)")163
+#define HEDLEY_DIAGNOSTIC_POP _Pragma("warning(pop)")164
+#else165
+#define HEDLEY_DIAGNOSTIC_PUSH166
+#define HEDLEY_DIAGNOSTIC_POP167
+#endif168
+169
/* HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_ is for170
HEDLEY INTERNAL USE ONLY. API subject to change without notice. */171
#if defined(HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_)172
173
#if defined(__cplusplus)174
#if HEDLEY_HAS_WARNING("-Wc++98-compat")175
#if HEDLEY_HAS_WARNING("-Wc++17-extensions")176
+#if HEDLEY_HAS_WARNING("-Wc++1z-extensions")177
+#define HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(xpr) \178
+ HEDLEY_DIAGNOSTIC_PUSH \179
+ _Pragma("clang diagnostic ignored \"-Wc++98-compat\"") _Pragma( \180
+ "clang diagnostic ignored \"-Wc++17-extensions\"") \181
+ _Pragma("clang diagnostic ignored \"-Wc++1z-extensions\"") \182
+ xpr HEDLEY_DIAGNOSTIC_POP183
+#else184
#define HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(xpr) \185
HEDLEY_DIAGNOSTIC_PUSH \186
_Pragma("clang diagnostic ignored \"-Wc++98-compat\"") \187
_Pragma("clang diagnostic ignored \"-Wc++17-extensions\"") \188
xpr HEDLEY_DIAGNOSTIC_POP189
+#endif190
#else191
#define HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(xpr) \192
HEDLEY_DIAGNOSTIC_PUSH \193
194
#elif HEDLEY_IAR_VERSION_CHECK(8, 3, 0)195
#define HEDLEY_CPP_CAST(T, expr) \196
HEDLEY_DIAGNOSTIC_PUSH \197
- _Pragma("diag_suppress=Pe137") HEDLEY_DIAGNOSTIC_POP #else198
+ _Pragma("diag_suppress=Pe137") HEDLEY_DIAGNOSTIC_POP199
+#else200
#define HEDLEY_CPP_CAST(T, expr) ((T)(expr))201
#endif202
#else203
#define HEDLEY_CPP_CAST(T, expr) (expr)204
#endif205
206
-#if (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) || \207
- defined(__clang__) || HEDLEY_GCC_VERSION_CHECK(3, 0, 0) || \208
- HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \209
- HEDLEY_IAR_VERSION_CHECK(8, 0, 0) || \210
- HEDLEY_PGI_VERSION_CHECK(18, 4, 0) || \211
- HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \212
- HEDLEY_TI_VERSION_CHECK(15, 12, 0) || \213
- HEDLEY_TI_ARMCL_VERSION_CHECK(4, 7, 0) || \214
- HEDLEY_TI_CL430_VERSION_CHECK(2, 0, 1) || \215
- HEDLEY_TI_CL2000_VERSION_CHECK(6, 1, 0) || \216
- HEDLEY_TI_CL6X_VERSION_CHECK(7, 0, 0) || \217
- HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \218
- HEDLEY_TI_CLPRU_VERSION_CHECK(2, 1, 0) || \219
- HEDLEY_CRAY_VERSION_CHECK(5, 0, 0) || \220
- HEDLEY_TINYC_VERSION_CHECK(0, 9, 17) || \221
- HEDLEY_SUNPRO_VERSION_CHECK(8, 0, 0) || \222
- (HEDLEY_IBM_VERSION_CHECK(10, 1, 0) && defined(__C99_PRAGMA_OPERATOR))223
-#define HEDLEY_PRAGMA(value) _Pragma(#value)224
-#elif HEDLEY_MSVC_VERSION_CHECK(15, 0, 0)225
-#define HEDLEY_PRAGMA(value) __pragma(value)226
-#else227
-#define HEDLEY_PRAGMA(value)228
-#endif229
-230
-#if defined(HEDLEY_DIAGNOSTIC_PUSH)231
-#undef HEDLEY_DIAGNOSTIC_PUSH232
-#endif233
-#if defined(HEDLEY_DIAGNOSTIC_POP)234
-#undef HEDLEY_DIAGNOSTIC_POP235
-#endif236
-#if defined(__clang__)237
-#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("clang diagnostic push")238
-#define HEDLEY_DIAGNOSTIC_POP _Pragma("clang diagnostic pop")239
-#elif HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)240
-#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("warning(push)")241
-#define HEDLEY_DIAGNOSTIC_POP _Pragma("warning(pop)")242
-#elif HEDLEY_GCC_VERSION_CHECK(4, 6, 0)243
-#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("GCC diagnostic push")244
-#define HEDLEY_DIAGNOSTIC_POP _Pragma("GCC diagnostic pop")245
-#elif HEDLEY_MSVC_VERSION_CHECK(15, 0, 0)246
-#define HEDLEY_DIAGNOSTIC_PUSH __pragma(warning(push))247
-#define HEDLEY_DIAGNOSTIC_POP __pragma(warning(pop))248
-#elif HEDLEY_ARM_VERSION_CHECK(5, 6, 0)249
-#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("push")250
-#define HEDLEY_DIAGNOSTIC_POP _Pragma("pop")251
-#elif HEDLEY_TI_VERSION_CHECK(15, 12, 0) || \252
- HEDLEY_TI_ARMCL_VERSION_CHECK(5, 2, 0) || \253
- HEDLEY_TI_CL430_VERSION_CHECK(4, 4, 0) || \254
- HEDLEY_TI_CL6X_VERSION_CHECK(8, 1, 0) || \255
- HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \256
- HEDLEY_TI_CLPRU_VERSION_CHECK(2, 1, 0)257
-#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("diag_push")258
-#define HEDLEY_DIAGNOSTIC_POP _Pragma("diag_pop")259
-#elif HEDLEY_PELLES_VERSION_CHECK(2, 90, 0)260
-#define HEDLEY_DIAGNOSTIC_PUSH _Pragma("warning(push)")261
-#define HEDLEY_DIAGNOSTIC_POP _Pragma("warning(pop)")262
-#else263
-#define HEDLEY_DIAGNOSTIC_PUSH264
-#define HEDLEY_DIAGNOSTIC_POP265
-#endif266
-267
#if defined(HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED)268
#undef HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED269
#endif270
271
#elif HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)272
#define HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED \273
_Pragma("warning(disable:1478 1786)")274
+#elif HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)275
+#define HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED \276
+ __pragma(warning(disable : 1478 1786))277
+#elif HEDLEY_PGI_VERSION_CHECK(20, 7, 0)278
+#define HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED \279
+ _Pragma("diag_suppress 1215,1216,1444,1445")280
#elif HEDLEY_PGI_VERSION_CHECK(17, 10, 0)281
#define HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress 1215,1444")282
#elif HEDLEY_GCC_VERSION_CHECK(4, 3, 0)283
284
#elif HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)285
#define HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS \286
_Pragma("warning(disable:161)")287
+#elif HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)288
+#define HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS \289
+ __pragma(warning(disable : 161))290
#elif HEDLEY_PGI_VERSION_CHECK(17, 10, 0)291
#define HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress 1675")292
#elif HEDLEY_GCC_VERSION_CHECK(4, 3, 0)293
294
#elif HEDLEY_INTEL_VERSION_CHECK(17, 0, 0)295
#define HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES \296
_Pragma("warning(disable:1292)")297
+#elif HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)298
+#define HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES \299
+ __pragma(warning(disable : 1292))300
#elif HEDLEY_MSVC_VERSION_CHECK(19, 0, 0)301
#define HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES \302
__pragma(warning(disable : 5030))303
+#elif HEDLEY_PGI_VERSION_CHECK(20, 7, 0)304
+#define HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES \305
+ _Pragma("diag_suppress 1097,1098")306
#elif HEDLEY_PGI_VERSION_CHECK(17, 10, 0)307
#define HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES \308
_Pragma("diag_suppress 1097")309
310
#if defined(HEDLEY_DEPRECATED_FOR)311
#undef HEDLEY_DEPRECATED_FOR312
#endif313
-#if defined(__cplusplus) && (__cplusplus >= 201402L)314
-#define HEDLEY_DEPRECATED(since) \315
- HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_( \316
- [[deprecated("Since " #since)]])317
-#define HEDLEY_DEPRECATED_FOR(since, replacement) \318
- HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_( \319
- [[deprecated("Since " #since "; use " #replacement)]])320
+#if HEDLEY_MSVC_VERSION_CHECK(14, 0, 0) || \321
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)322
+#define HEDLEY_DEPRECATED(since) __declspec(deprecated("Since " #since))323
+#define HEDLEY_DEPRECATED_FOR(since, replacement) \324
+ __declspec(deprecated("Since " #since "; use " #replacement))325
#elif HEDLEY_HAS_EXTENSION(attribute_deprecated_with_message) || \326
HEDLEY_GCC_VERSION_CHECK(4, 5, 0) || \327
HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \328
329
__attribute__((__deprecated__("Since " #since)))330
#define HEDLEY_DEPRECATED_FOR(since, replacement) \331
__attribute__((__deprecated__("Since " #since "; use " #replacement)))332
+#elif defined(__cplusplus) && (__cplusplus >= 201402L)333
+#define HEDLEY_DEPRECATED(since) \334
+ HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_( \335
+ [[deprecated("Since " #since)]])336
+#define HEDLEY_DEPRECATED_FOR(since, replacement) \337
+ HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_( \338
+ [[deprecated("Since " #since "; use " #replacement)]])339
#elif HEDLEY_HAS_ATTRIBUTE(deprecated) || HEDLEY_GCC_VERSION_CHECK(3, 1, 0) || \340
HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \341
HEDLEY_TI_VERSION_CHECK(15, 12, 0) || \342
343
#define HEDLEY_DEPRECATED(since) __attribute__((__deprecated__))344
#define HEDLEY_DEPRECATED_FOR(since, replacement) \345
__attribute__((__deprecated__))346
-#elif HEDLEY_MSVC_VERSION_CHECK(14, 0, 0)347
-#define HEDLEY_DEPRECATED(since) __declspec(deprecated("Since " #since))348
-#define HEDLEY_DEPRECATED_FOR(since, replacement) \349
- __declspec(deprecated("Since " #since "; use " #replacement))350
-#elif HEDLEY_MSVC_VERSION_CHECK(13, 10, 0) || \351
- HEDLEY_PELLES_VERSION_CHECK(6, 50, 0)352
+#elif HEDLEY_MSVC_VERSION_CHECK(13, 10, 0) || \353
+ HEDLEY_PELLES_VERSION_CHECK(6, 50, 0) || \354
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)355
#define HEDLEY_DEPRECATED(since) __declspec(deprecated)356
#define HEDLEY_DEPRECATED_FOR(since, replacement) __declspec(deprecated)357
#elif HEDLEY_IAR_VERSION_CHECK(8, 0, 0)358
359
#if defined(HEDLEY_WARN_UNUSED_RESULT_MSG)360
#undef HEDLEY_WARN_UNUSED_RESULT_MSG361
#endif362
-#if (HEDLEY_HAS_CPP_ATTRIBUTE(nodiscard) >= 201907L)363
-#define HEDLEY_WARN_UNUSED_RESULT \364
- HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])365
-#define HEDLEY_WARN_UNUSED_RESULT_MSG(msg) \366
- HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard(msg)]])367
-#elif HEDLEY_HAS_CPP_ATTRIBUTE(nodiscard)368
-#define HEDLEY_WARN_UNUSED_RESULT \369
- HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])370
-#define HEDLEY_WARN_UNUSED_RESULT_MSG(msg) \371
- HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])372
-#elif HEDLEY_HAS_ATTRIBUTE(warn_unused_result) || \373
+#if HEDLEY_HAS_ATTRIBUTE(warn_unused_result) || \374
HEDLEY_GCC_VERSION_CHECK(3, 4, 0) || \375
HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \376
HEDLEY_TI_VERSION_CHECK(15, 12, 0) || \377
378
#define HEDLEY_WARN_UNUSED_RESULT __attribute__((__warn_unused_result__))379
#define HEDLEY_WARN_UNUSED_RESULT_MSG(msg) \380
__attribute__((__warn_unused_result__))381
+#elif (HEDLEY_HAS_CPP_ATTRIBUTE(nodiscard) >= 201907L)382
+#define HEDLEY_WARN_UNUSED_RESULT \383
+ HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])384
+#define HEDLEY_WARN_UNUSED_RESULT_MSG(msg) \385
+ HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard(msg)]])386
+#elif HEDLEY_HAS_CPP_ATTRIBUTE(nodiscard)387
+#define HEDLEY_WARN_UNUSED_RESULT \388
+ HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])389
+#define HEDLEY_WARN_UNUSED_RESULT_MSG(msg) \390
+ HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])391
#elif defined(_Check_return_) /* SAL */392
#define HEDLEY_WARN_UNUSED_RESULT _Check_return_393
#define HEDLEY_WARN_UNUSED_RESULT_MSG(msg) _Check_return_394
395
#define HEDLEY_NO_RETURN __attribute__((__noreturn__))396
#elif HEDLEY_SUNPRO_VERSION_CHECK(5, 10, 0)397
#define HEDLEY_NO_RETURN _Pragma("does_not_return")398
-#elif HEDLEY_MSVC_VERSION_CHECK(13, 10, 0)399
+#elif HEDLEY_MSVC_VERSION_CHECK(13, 10, 0) || \400
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)401
#define HEDLEY_NO_RETURN __declspec(noreturn)402
#elif HEDLEY_TI_CL6X_VERSION_CHECK(6, 0, 0) && defined(__cplusplus)403
#define HEDLEY_NO_RETURN _Pragma("FUNC_NEVER_RETURNS;")404
405
#if defined(HEDLEY_ASSUME)406
#undef HEDLEY_ASSUME407
#endif408
-#if HEDLEY_MSVC_VERSION_CHECK(13, 10, 0) || HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)409
+#if HEDLEY_MSVC_VERSION_CHECK(13, 10, 0) || \410
+ HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \411
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)412
#define HEDLEY_ASSUME(expr) __assume(expr)413
#elif HEDLEY_HAS_BUILTIN(__builtin_assume)414
#define HEDLEY_ASSUME(expr) __builtin_assume(expr)415
416
#if HEDLEY_HAS_BUILTIN(__builtin_unpredictable)417
#define HEDLEY_UNPREDICTABLE(expr) __builtin_unpredictable((expr))418
#endif419
-#if HEDLEY_HAS_BUILTIN(__builtin_expect_with_probability) || \420
+#if (HEDLEY_HAS_BUILTIN(__builtin_expect_with_probability) && \421
+ !defined(HEDLEY_PGI_VERSION)) || \422
HEDLEY_GCC_VERSION_CHECK(9, 0, 0)423
#define HEDLEY_PREDICT(expr, value, probability) \424
__builtin_expect_with_probability((expr), (value), (probability))425
426
__builtin_expect_with_probability(!!(expr), 0, (probability))427
#define HEDLEY_LIKELY(expr) __builtin_expect(!!(expr), 1)428
#define HEDLEY_UNLIKELY(expr) __builtin_expect(!!(expr), 0)429
-#elif HEDLEY_HAS_BUILTIN(__builtin_expect) || \430
+#elif (HEDLEY_HAS_BUILTIN(__builtin_expect) && \431
+ !defined(HEDLEY_INTEL_CL_VERSION)) || \432
HEDLEY_GCC_VERSION_CHECK(3, 0, 0) || \433
HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \434
(HEDLEY_SUNPRO_VERSION_CHECK(5, 15, 0) && defined(__cplusplus)) || \435
436
#define HEDLEY_MALLOC __attribute__((__malloc__))437
#elif HEDLEY_SUNPRO_VERSION_CHECK(5, 10, 0)438
#define HEDLEY_MALLOC _Pragma("returns_new_memory")439
-#elif HEDLEY_MSVC_VERSION_CHECK(14, 0, 0)440
+#elif HEDLEY_MSVC_VERSION_CHECK(14, 0, 0) || \441
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)442
#define HEDLEY_MALLOC __declspec(restrict)443
#else444
#define HEDLEY_MALLOC445
446
#elif HEDLEY_GCC_VERSION_CHECK(3, 1, 0) || \447
HEDLEY_MSVC_VERSION_CHECK(14, 0, 0) || \448
HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \449
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0) || \450
HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \451
HEDLEY_IBM_VERSION_CHECK(10, 1, 0) || \452
HEDLEY_PGI_VERSION_CHECK(17, 10, 0) || \453
454
#define HEDLEY_INLINE inline455
#elif defined(HEDLEY_GCC_VERSION) || HEDLEY_ARM_VERSION_CHECK(6, 2, 0)456
#define HEDLEY_INLINE __inline__457
-#elif HEDLEY_MSVC_VERSION_CHECK(12, 0, 0) || \458
- HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \459
- HEDLEY_TI_ARMCL_VERSION_CHECK(5, 1, 0) || \460
- HEDLEY_TI_CL430_VERSION_CHECK(3, 1, 0) || \461
- HEDLEY_TI_CL2000_VERSION_CHECK(6, 2, 0) || \462
- HEDLEY_TI_CL6X_VERSION_CHECK(8, 0, 0) || \463
- HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \464
+#elif HEDLEY_MSVC_VERSION_CHECK(12, 0, 0) || \465
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0) || \466
+ HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \467
+ HEDLEY_TI_ARMCL_VERSION_CHECK(5, 1, 0) || \468
+ HEDLEY_TI_CL430_VERSION_CHECK(3, 1, 0) || \469
+ HEDLEY_TI_CL2000_VERSION_CHECK(6, 2, 0) || \470
+ HEDLEY_TI_CL6X_VERSION_CHECK(8, 0, 0) || \471
+ HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \472
HEDLEY_TI_CLPRU_VERSION_CHECK(2, 1, 0)473
#define HEDLEY_INLINE __inline474
#else475
476
HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \477
HEDLEY_TI_CLPRU_VERSION_CHECK(2, 1, 0)478
#define HEDLEY_ALWAYS_INLINE __attribute__((__always_inline__)) HEDLEY_INLINE479
-#elif HEDLEY_MSVC_VERSION_CHECK(12, 0, 0)480
+#elif HEDLEY_MSVC_VERSION_CHECK(12, 0, 0) || \481
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)482
#define HEDLEY_ALWAYS_INLINE __forceinline483
#elif defined(__cplusplus) && (HEDLEY_TI_ARMCL_VERSION_CHECK(5, 2, 0) || \484
HEDLEY_TI_CL430_VERSION_CHECK(4, 3, 0) || \485
486
HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \487
HEDLEY_TI_CLPRU_VERSION_CHECK(2, 1, 0)488
#define HEDLEY_NEVER_INLINE __attribute__((__noinline__))489
-#elif HEDLEY_MSVC_VERSION_CHECK(13, 10, 0)490
+#elif HEDLEY_MSVC_VERSION_CHECK(13, 10, 0) || \491
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)492
#define HEDLEY_NEVER_INLINE __declspec(noinline)493
#elif HEDLEY_PGI_VERSION_CHECK(10, 2, 0)494
#define HEDLEY_NEVER_INLINE _Pragma("noinline")495
496
#if HEDLEY_HAS_ATTRIBUTE(nothrow) || HEDLEY_GCC_VERSION_CHECK(3, 3, 0) || \497
HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)498
#define HEDLEY_NO_THROW __attribute__((__nothrow__))499
-#elif HEDLEY_MSVC_VERSION_CHECK(13, 1, 0) || HEDLEY_ARM_VERSION_CHECK(4, 1, 0)500
+#elif HEDLEY_MSVC_VERSION_CHECK(13, 1, 0) || \501
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0) || \502
+ HEDLEY_ARM_VERSION_CHECK(4, 1, 0)503
#define HEDLEY_NO_THROW __declspec(nothrow)504
#else505
#define HEDLEY_NO_THROW506
507
#if defined(HEDLEY_FALL_THROUGH)508
#undef HEDLEY_FALL_THROUGH509
#endif510
-#if HEDLEY_GNUC_HAS_ATTRIBUTE(fallthrough, 7, 0, 0) && \511
- !defined(HEDLEY_PGI_VERSION)512
+#if HEDLEY_HAS_ATTRIBUTE(fallthrough) || HEDLEY_GCC_VERSION_CHECK(7, 0, 0)513
#define HEDLEY_FALL_THROUGH __attribute__((__fallthrough__))514
#elif HEDLEY_HAS_CPP_ATTRIBUTE_NS(clang, fallthrough)515
#define HEDLEY_FALL_THROUGH \516
517
#endif518
#if !defined(__cplusplus) && \519
((defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L)) || \520
- HEDLEY_HAS_FEATURE(c_static_assert) || \521
+ (HEDLEY_HAS_FEATURE(c_static_assert) && \522
+ !defined(HEDLEY_INTEL_CL_VERSION)) || \523
HEDLEY_GCC_VERSION_CHECK(6, 0, 0) || \524
HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || defined(_Static_assert))525
#define HEDLEY_STATIC_ASSERT(expr, message) _Static_assert(expr, message)526
#elif (defined(__cplusplus) && (__cplusplus >= 201103L)) || \527
- HEDLEY_MSVC_VERSION_CHECK(16, 0, 0)528
+ HEDLEY_MSVC_VERSION_CHECK(16, 0, 0) || \529
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)530
#define HEDLEY_STATIC_ASSERT(expr, message) \531
HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_( \532
static_assert(expr, message))533
534
HEDLEY_PGI_VERSION_CHECK(18, 4, 0) || \535
HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)536
#define HEDLEY_WARNING(msg) HEDLEY_PRAGMA(GCC warning msg)537
-#elif HEDLEY_MSVC_VERSION_CHECK(15, 0, 0)538
+#elif HEDLEY_MSVC_VERSION_CHECK(15, 0, 0) || \539
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)540
#define HEDLEY_WARNING(msg) HEDLEY_PRAGMA(message(msg))541
#else542
#define HEDLEY_WARNING(msg) HEDLEY_MESSAGE(msg)543
544
#endif545
#if HEDLEY_HAS_ATTRIBUTE(flag_enum)546
#define HEDLEY_FLAGS __attribute__((__flag_enum__))547
+#else548
+#define HEDLEY_FLAGS549
#endif550
551
#if defined(HEDLEY_FLAGS_CAST)552
553
#if defined(HEDLEY_EMPTY_BASES)554
#undef HEDLEY_EMPTY_BASES555
#endif556
-#if HEDLEY_MSVC_VERSION_CHECK(19, 0, 23918) && \557
- !HEDLEY_MSVC_VERSION_CHECK(20, 0, 0)558
+#if (HEDLEY_MSVC_VERSION_CHECK(19, 0, 23918) && \559
+ !HEDLEY_MSVC_VERSION_CHECK(20, 0, 0)) || \560
+ HEDLEY_INTEL_CL_VERSION_CHECK(2021, 1, 0)561
#define HEDLEY_EMPTY_BASES __declspec(empty_bases)562
#else563
#define HEDLEY_EMPTY_BASES564
obs-studio-26.1.1.tar.xz/libobs/util/simde/simde-align.h
Added
483
1
2
+/* Alignment3
+ * Created by Evan Nemerson <evan@nemerson.com>4
+ *5
+ * To the extent possible under law, the authors have waived all6
+ * copyright and related or neighboring rights to this code. For7
+ * details, see the Creative Commons Zero 1.0 Universal license at8
+ * <https://creativecommons.org/publicdomain/zero/1.0/>9
+ *10
+ * SPDX-License-Identifier: CC0-1.011
+ *12
+ **********************************************************************13
+ *14
+ * This is portability layer which should help iron out some15
+ * differences across various compilers, as well as various verisons of16
+ * C and C++.17
+ *18
+ * It was originally developed for SIMD Everywhere19
+ * (<https://github.com/simd-everywhere/simde>), but since its only20
+ * dependency is Hedley (<https://nemequ.github.io/hedley>, also CC0)21
+ * it can easily be used in other projects, so please feel free to do22
+ * so.23
+ *24
+ * If you do use this in your project, please keep a link to SIMDe in25
+ * your code to remind you where to report any bugs and/or check for26
+ * updated versions.27
+ *28
+ * # API Overview29
+ *30
+ * The API has several parts, and most macros have a few variations.31
+ * There are APIs for declaring aligned fields/variables, optimization32
+ * hints, and run-time alignment checks.33
+ *34
+ * Briefly, macros ending with "_TO" take numeric values and are great35
+ * when you know the value you would like to use. Macros ending with36
+ * "_LIKE", on the other hand, accept a type and are used when you want37
+ * to use the alignment of a type instead of hardcoding a value.38
+ *39
+ * Documentation for each section of the API is inline.40
+ *41
+ * True to form, MSVC is the main problem and imposes several42
+ * limitations on the effectiveness of the APIs. Detailed descriptions43
+ * of the limitations of each macro are inline, but in general:44
+ *45
+ * * On C11+ or C++11+ code written using this API will work. The46
+ * ASSUME macros may or may not generate a hint to the compiler, but47
+ * that is only an optimization issue and will not actually cause48
+ * failures.49
+ * * If you're using pretty much any compiler other than MSVC,50
+ * everything should basically work as well as in C11/C++11.51
+ */52
+53
+#if !defined(SIMDE_ALIGN_H)54
+#define SIMDE_ALIGN_H55
+56
+#include "hedley.h"57
+58
+/* I know this seems a little silly, but some non-hosted compilers59
+ * don't have stddef.h, so we try to accomodate them. */60
+#if !defined(SIMDE_ALIGN_SIZE_T_)61
+#if defined(__SIZE_TYPE__)62
+#define SIMDE_ALIGN_SIZE_T_ __SIZE_TYPE__63
+#elif defined(__SIZE_T_TYPE__)64
+#define SIMDE_ALIGN_SIZE_T_ __SIZE_TYPE__65
+#elif defined(__cplusplus)66
+#include <cstddef>67
+#define SIMDE_ALIGN_SIZE_T_ size_t68
+#else69
+#include <stddef.h>70
+#define SIMDE_ALIGN_SIZE_T_ size_t71
+#endif72
+#endif73
+74
+#if !defined(SIMDE_ALIGN_INTPTR_T_)75
+#if defined(__INTPTR_TYPE__)76
+#define SIMDE_ALIGN_INTPTR_T_ __INTPTR_TYPE__77
+#elif defined(__PTRDIFF_TYPE__)78
+#define SIMDE_ALIGN_INTPTR_T_ __PTRDIFF_TYPE__79
+#elif defined(__PTRDIFF_T_TYPE__)80
+#define SIMDE_ALIGN_INTPTR_T_ __PTRDIFF_T_TYPE__81
+#elif defined(__cplusplus)82
+#include <cstddef>83
+#define SIMDE_ALIGN_INTPTR_T_ ptrdiff_t84
+#else85
+#include <stddef.h>86
+#define SIMDE_ALIGN_INTPTR_T_ ptrdiff_t87
+#endif88
+#endif89
+90
+#if defined(SIMDE_ALIGN_DEBUG)91
+#if defined(__cplusplus)92
+#include <cstdio>93
+#else94
+#include <stdio.h>95
+#endif96
+#endif97
+98
+/* SIMDE_ALIGN_OF(Type)99
+ *100
+ * The SIMDE_ALIGN_OF macro works like alignof, or _Alignof, or101
+ * __alignof, or __alignof__, or __ALIGNOF__, depending on the compiler.102
+ * It isn't defined everywhere (only when the compiler has some alignof-103
+ * like feature we can use to implement it), but it should work in most104
+ * modern compilers, as well as C11 and C++11.105
+ *106
+ * If we can't find an implementation for SIMDE_ALIGN_OF then the macro107
+ * will not be defined, so if you can handle that situation sensibly108
+ * you may need to sprinkle some ifdefs into your code.109
+ */110
+#if (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L)) || \111
+ (0 && HEDLEY_HAS_FEATURE(c_alignof))112
+#define SIMDE_ALIGN_OF(Type) _Alignof(Type)113
+#elif (defined(__cplusplus) && (__cplusplus >= 201103L)) || \114
+ (0 && HEDLEY_HAS_FEATURE(cxx_alignof))115
+#define SIMDE_ALIGN_OF(Type) alignof(Type)116
+#elif HEDLEY_GCC_VERSION_CHECK(2, 95, 0) || \117
+ HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \118
+ HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \119
+ HEDLEY_SUNPRO_VERSION_CHECK(5, 13, 0) || \120
+ HEDLEY_TINYC_VERSION_CHECK(0, 9, 24) || \121
+ HEDLEY_PGI_VERSION_CHECK(19, 10, 0) || \122
+ HEDLEY_CRAY_VERSION_CHECK(10, 0, 0) || \123
+ HEDLEY_TI_ARMCL_VERSION_CHECK(16, 9, 0) || \124
+ HEDLEY_TI_CL2000_VERSION_CHECK(16, 9, 0) || \125
+ HEDLEY_TI_CL6X_VERSION_CHECK(8, 0, 0) || \126
+ HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \127
+ HEDLEY_TI_CL430_VERSION_CHECK(16, 9, 0) || \128
+ HEDLEY_TI_CLPRU_VERSION_CHECK(2, 3, 2) || defined(__IBM__ALIGNOF__) || \129
+ defined(__clang__)130
+#define SIMDE_ALIGN_OF(Type) __alignof__(Type)131
+#elif HEDLEY_IAR_VERSION_CHECK(8, 40, 0)132
+#define SIMDE_ALIGN_OF(Type) __ALIGNOF__(Type)133
+#elif HEDLEY_MSVC_VERSION_CHECK(19, 0, 0)134
+/* Probably goes back much further, but MS takes down their old docs.135
+ * If you can verify that this works in earlier versions please let136
+ * me know! */137
+#define SIMDE_ALIGN_OF(Type) __alignof(Type)138
+#endif139
+140
+/* SIMDE_ALIGN_MAXIMUM:141
+ *142
+ * This is the maximum alignment that the compiler supports. You can143
+ * define the value prior to including SIMDe if necessary, but in that144
+ * case *please* submit an issue so we can add the platform to the145
+ * detection code.146
+ *147
+ * Most compilers are okay with types which are aligned beyond what148
+ * they think is the maximum, as long as the alignment is a power149
+ * of two. MSVC is the exception (of course), so we need to cap the150
+ * alignment requests at values that the implementation supports.151
+ *152
+ * XL C/C++ will accept values larger than 16 (which is the alignment153
+ * of an AltiVec vector), but will not reliably align to the larger154
+ * value, so so we cap the value at 16 there.155
+ *156
+ * If the compiler accepts any power-of-two value within reason then157
+ * this macro should be left undefined, and the SIMDE_ALIGN_CAP158
+ * macro will just return the value passed to it. */159
+#if !defined(SIMDE_ALIGN_MAXIMUM)160
+#if defined(HEDLEY_MSVC_VERSION)161
+#if defined(_M_IX86) || defined(_M_AMD64)162
+#if HEDLEY_MSVC_VERSION_CHECK(19, 14, 0)163
+#define SIMDE_ALIGN_PLATFORM_MAXIMUM 64164
+#elif HEDLEY_MSVC_VERSION_CHECK(16, 0, 0)165
+/* VS 2010 is really a guess based on Wikipedia; if anyone can166
+ * test with old VS versions I'd really appreciate it. */167
+#define SIMDE_ALIGN_PLATFORM_MAXIMUM 32168
+#else169
+#define SIMDE_ALIGN_PLATFORM_MAXIMUM 16170
+#endif171
+#elif defined(_M_ARM) || defined(_M_ARM64)172
+#define SIMDE_ALIGN_PLATFORM_MAXIMUM 8173
+#endif174
+#elif defined(HEDLEY_IBM_VERSION)175
+#define SIMDE_ALIGN_PLATFORM_MAXIMUM 16176
+#endif177
+#endif178
+179
+/* You can mostly ignore these; they're intended for internal use.180
+ * If you do need to use them please let me know; if they fulfill181
+ * a common use case I'll probably drop the trailing underscore182
+ * and make them part of the public API. */183
+#if defined(SIMDE_ALIGN_PLATFORM_MAXIMUM)184
+#if SIMDE_ALIGN_PLATFORM_MAXIMUM >= 64185
+#define SIMDE_ALIGN_64_ 64186
+#define SIMDE_ALIGN_32_ 32187
+#define SIMDE_ALIGN_16_ 16188
+#define SIMDE_ALIGN_8_ 8189
+#elif SIMDE_ALIGN_PLATFORM_MAXIMUM >= 32190
+#define SIMDE_ALIGN_64_ 32191
+#define SIMDE_ALIGN_32_ 32192
+#define SIMDE_ALIGN_16_ 16193
+#define SIMDE_ALIGN_8_ 8194
+#elif SIMDE_ALIGN_PLATFORM_MAXIMUM >= 16195
+#define SIMDE_ALIGN_64_ 16196
+#define SIMDE_ALIGN_32_ 16197
+#define SIMDE_ALIGN_16_ 16198
+#define SIMDE_ALIGN_8_ 8199
+#elif SIMDE_ALIGN_PLATFORM_MAXIMUM >= 8200
+#define SIMDE_ALIGN_64_ 8201
+#define SIMDE_ALIGN_32_ 8202
+#define SIMDE_ALIGN_16_ 8203
+#define SIMDE_ALIGN_8_ 8204
+#else205
+#error Max alignment expected to be >= 8206
+#endif207
+#else208
+#define SIMDE_ALIGN_64_ 64209
+#define SIMDE_ALIGN_32_ 32210
+#define SIMDE_ALIGN_16_ 16211
+#define SIMDE_ALIGN_8_ 8212
+#endif213
+214
+/**215
+ * SIMDE_ALIGN_CAP(Alignment)216
+ *217
+ * Returns the minimum of Alignment or SIMDE_ALIGN_MAXIMUM.218
+ */219
+#if defined(SIMDE_ALIGN_MAXIMUM)220
+#define SIMDE_ALIGN_CAP(Alignment) \221
+ (((Alignment) < (SIMDE_ALIGN_PLATFORM_MAXIMUM)) \222
+ ? (Alignment) \223
+ : (SIMDE_ALIGN_PLATFORM_MAXIMUM))224
+#else225
+#define SIMDE_ALIGN_CAP(Alignment) (Alignment)226
+#endif227
+228
+/* SIMDE_ALIGN_TO(Alignment)229
+ *230
+ * SIMDE_ALIGN_TO is used to declare types or variables. It basically231
+ * maps to the align attribute in most compilers, the align declspec232
+ * in MSVC, or _Alignas/alignas in C11/C++11.233
+ *234
+ * Example:235
+ *236
+ * struct i32x4 {237
+ * SIMDE_ALIGN_TO(16) int32_t values[4];238
+ * }239
+ *240
+ * Limitations:241
+ *242
+ * MSVC requires that the Alignment parameter be numeric; you can't do243
+ * something like `SIMDE_ALIGN_TO(SIMDE_ALIGN_OF(int))`. This is244
+ * unfortunate because that's really how the LIKE macros are245
+ * implemented, and I am not aware of a way to get anything like this246
+ * to work without using the C11/C++11 keywords.247
+ *248
+ * It also means that we can't use SIMDE_ALIGN_CAP to limit the249
+ * alignment to the value specified, which MSVC also requires, so on250
+ * MSVC you should use the `SIMDE_ALIGN_TO_8/16/32/64` macros instead.251
+ * They work like `SIMDE_ALIGN_TO(SIMDE_ALIGN_CAP(Alignment))` would,252
+ * but should be safe to use on MSVC.253
+ *254
+ * All this is to say that, if you want your code to work on MSVC, you255
+ * should use the SIMDE_ALIGN_TO_8/16/32/64 macros below instead of256
+ * SIMDE_ALIGN_TO(8/16/32/64).257
+ */258
+#if HEDLEY_HAS_ATTRIBUTE(aligned) || HEDLEY_GCC_VERSION_CHECK(2, 95, 0) || \259
+ HEDLEY_CRAY_VERSION_CHECK(8, 4, 0) || \260
+ HEDLEY_IBM_VERSION_CHECK(11, 1, 0) || \261
+ HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \262
+ HEDLEY_PGI_VERSION_CHECK(19, 4, 0) || \263
+ HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \264
+ HEDLEY_TINYC_VERSION_CHECK(0, 9, 24) || \265
+ HEDLEY_TI_ARMCL_VERSION_CHECK(16, 9, 0) || \266
+ HEDLEY_TI_CL2000_VERSION_CHECK(16, 9, 0) || \267
+ HEDLEY_TI_CL6X_VERSION_CHECK(8, 0, 0) || \268
+ HEDLEY_TI_CL7X_VERSION_CHECK(1, 2, 0) || \269
+ HEDLEY_TI_CL430_VERSION_CHECK(16, 9, 0) || \270
+ HEDLEY_TI_CLPRU_VERSION_CHECK(2, 3, 2)271
+#define SIMDE_ALIGN_TO(Alignment) \272
+ __attribute__((__aligned__(SIMDE_ALIGN_CAP(Alignment))))273
+#elif (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L))274
+#define SIMDE_ALIGN_TO(Alignment) _Alignas(SIMDE_ALIGN_CAP(Alignment))275
+#elif (defined(__cplusplus) && (__cplusplus >= 201103L))276
+#define SIMDE_ALIGN_TO(Alignment) alignas(SIMDE_ALIGN_CAP(Alignment))277
+#elif defined(HEDLEY_MSVC_VERSION)278
+#define SIMDE_ALIGN_TO(Alignment) __declspec(align(Alignment))279
+/* Unfortunately MSVC can't handle __declspec(align(__alignof(Type)));280
+ * the alignment passed to the declspec has to be an integer. */281
+#define SIMDE_ALIGN_OF_UNUSABLE_FOR_LIKE282
+#endif283
+#define SIMDE_ALIGN_TO_64 SIMDE_ALIGN_TO(SIMDE_ALIGN_64_)284
+#define SIMDE_ALIGN_TO_32 SIMDE_ALIGN_TO(SIMDE_ALIGN_32_)285
+#define SIMDE_ALIGN_TO_16 SIMDE_ALIGN_TO(SIMDE_ALIGN_16_)286
+#define SIMDE_ALIGN_TO_8 SIMDE_ALIGN_TO(SIMDE_ALIGN_8_)287
+288
+/* SIMDE_ALIGN_ASSUME_TO(Pointer, Alignment)289
+ *290
+ * SIMDE_ALIGN_ASSUME_TO is semantically similar to C++20's291
+ * std::assume_aligned, or __builtin_assume_aligned. It tells the292
+ * compiler to assume that the provided pointer is aligned to an293
+ * `Alignment`-byte boundary.294
+ *295
+ * If you define SIMDE_ALIGN_DEBUG prior to including this header then296
+ * SIMDE_ALIGN_ASSUME_TO will turn into a runtime check. We don't297
+ * integrate with NDEBUG in this header, but it may be a good idea to298
+ * put something like this in your code:299
+ *300
+ * #if !defined(NDEBUG)301
+ * #define SIMDE_ALIGN_DEBUG302
+ * #endif303
+ * #include <.../simde-align.h>304
+ */305
+#if HEDLEY_HAS_BUILTIN(__builtin_assume_aligned) || \306
+ HEDLEY_GCC_VERSION_CHECK(4, 7, 0)307
+#define SIMDE_ALIGN_ASSUME_TO_UNCHECKED(Pointer, Alignment) \308
+ HEDLEY_REINTERPRET_CAST( \309
+ __typeof__(Pointer), \310
+ __builtin_assume_aligned( \311
+ HEDLEY_CONST_CAST( \312
+ void *, HEDLEY_REINTERPRET_CAST(const void *, \313
+ Pointer)), \314
+ Alignment))315
+#elif HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)316
+#define SIMDE_ALIGN_ASSUME_TO_UNCHECKED(Pointer, Alignment) \317
+ (__extension__({ \318
+ __typeof__(v) simde_assume_aligned_t_ = (Pointer); \319
+ __assume_aligned(simde_assume_aligned_t_, Alignment); \320
+ simde_assume_aligned_t_; \321
+ }))322
+#elif defined(__cplusplus) && (__cplusplus > 201703L)323
+#include <memory>324
+#define SIMDE_ALIGN_ASSUME_TO_UNCHECKED(Pointer, Alignment) \325
+ std::assume_aligned<Alignment>(Pointer)326
+#else327
+#if defined(__cplusplus)328
+template<typename T>329
+HEDLEY_ALWAYS_INLINE static T *330
+simde_align_assume_to_unchecked(T *ptr, const size_t alignment)331
+#else332
+HEDLEY_ALWAYS_INLINE static void *333
+simde_align_assume_to_unchecked(void *ptr, const size_t alignment)334
+#endif335
+{336
+ HEDLEY_ASSUME((HEDLEY_REINTERPRET_CAST(size_t, (ptr)) %337
+ SIMDE_ALIGN_CAP(alignment)) == 0);338
+ return ptr;339
+}340
+#if defined(__cplusplus)341
+#define SIMDE_ALIGN_ASSUME_TO_UNCHECKED(Pointer, Alignment) \342
+ simde_align_assume_to_unchecked((Pointer), (Alignment))343
+#else344
+#define SIMDE_ALIGN_ASSUME_TO_UNCHECKED(Pointer, Alignment) \345
+ simde_align_assume_to_unchecked( \346
+ HEDLEY_CONST_CAST(void *, HEDLEY_REINTERPRET_CAST( \347
+ const void *, Pointer)), \348
+ (Alignment))349
+#endif350
+#endif351
+352
+#if !defined(SIMDE_ALIGN_DEBUG)353
+#define SIMDE_ALIGN_ASSUME_TO(Pointer, Alignment) \354
+ SIMDE_ALIGN_ASSUME_TO_UNCHECKED(Pointer, Alignment)355
+#else356
+#include <stdio.h>357
+#if defined(__cplusplus)358
+template<typename T>359
+static HEDLEY_ALWAYS_INLINE T *360
+simde_align_assume_to_checked_uncapped(T *ptr, const size_t alignment,361
+ const char *file, int line,362
+ const char *ptrname)363
+#else364
+static HEDLEY_ALWAYS_INLINE void *365
+simde_align_assume_to_checked_uncapped(void *ptr, const size_t alignment,366
+ const char *file, int line,367
+ const char *ptrname)368
+#endif369
+{370
+ if (HEDLEY_UNLIKELY(371
+ (HEDLEY_REINTERPRET_CAST(SIMDE_ALIGN_INTPTR_T_, (ptr)) %372
+ HEDLEY_STATIC_CAST(SIMDE_ALIGN_INTPTR_T_,373
+ SIMDE_ALIGN_CAP(alignment))) != 0)) {374
+ fprintf(stderr,375
+ "%s:%d: alignment check failed for `%s' (%p %% %u == %u)\n",376
+ file, line, ptrname,377
+ HEDLEY_REINTERPRET_CAST(const void *, ptr),378
+ HEDLEY_STATIC_CAST(unsigned int,379
+ SIMDE_ALIGN_CAP(alignment)),380
+ HEDLEY_STATIC_CAST(381
+ unsigned int,382
+ HEDLEY_REINTERPRET_CAST(SIMDE_ALIGN_INTPTR_T_,383
+ (ptr)) %384
+ HEDLEY_STATIC_CAST(385
+ SIMDE_ALIGN_INTPTR_T_,386
+ SIMDE_ALIGN_CAP(alignment))));387
+ }388
+389
+ return ptr;390
+}391
+392
+#if defined(__cplusplus)393
+#define SIMDE_ALIGN_ASSUME_TO(Pointer, Alignment) \394
+ simde_align_assume_to_checked_uncapped((Pointer), (Alignment), \395
+ __FILE__, __LINE__, #Pointer)396
+#else397
+#define SIMDE_ALIGN_ASSUME_TO(Pointer, Alignment) \398
+ simde_align_assume_to_checked_uncapped( \399
+ HEDLEY_CONST_CAST(void *, HEDLEY_REINTERPRET_CAST( \400
+ const void *, Pointer)), \401
+ (Alignment), __FILE__, __LINE__, #Pointer)402
+#endif403
+#endif404
+405
+/* SIMDE_ALIGN_LIKE(Type)406
+ * SIMDE_ALIGN_LIKE_#(Type)407
+ *408
+ * The SIMDE_ALIGN_LIKE macros are similar to the SIMDE_ALIGN_TO macros409
+ * except instead of an integer they take a type; basically, it's just410
+ * a more convenient way to do something like:411
+ *412
+ * SIMDE_ALIGN_TO(SIMDE_ALIGN_OF(Type))413
+ *414
+ * The versions with a numeric suffix will fall back on using a numeric415
+ * value in the event we can't use SIMDE_ALIGN_OF(Type). This is416
+ * mainly for MSVC, where __declspec(align()) can't handle anything417
+ * other than hard-coded numeric values.418
+ */419
+#if defined(SIMDE_ALIGN_OF) && defined(SIMDE_ALIGN_TO) && \420
+ !defined(SIMDE_ALIGN_OF_UNUSABLE_FOR_LIKE)421
+#define SIMDE_ALIGN_LIKE(Type) SIMDE_ALIGN_TO(SIMDE_ALIGN_OF(Type))422
+#define SIMDE_ALIGN_LIKE_64(Type) SIMDE_ALIGN_LIKE(Type)423
+#define SIMDE_ALIGN_LIKE_32(Type) SIMDE_ALIGN_LIKE(Type)424
+#define SIMDE_ALIGN_LIKE_16(Type) SIMDE_ALIGN_LIKE(Type)425
+#define SIMDE_ALIGN_LIKE_8(Type) SIMDE_ALIGN_LIKE(Type)426
+#else427
+#define SIMDE_ALIGN_LIKE_64(Type) SIMDE_ALIGN_TO_64428
+#define SIMDE_ALIGN_LIKE_32(Type) SIMDE_ALIGN_TO_32429
+#define SIMDE_ALIGN_LIKE_16(Type) SIMDE_ALIGN_TO_16430
+#define SIMDE_ALIGN_LIKE_8(Type) SIMDE_ALIGN_TO_8431
+#endif432
+433
+/* SIMDE_ALIGN_ASSUME_LIKE(Pointer, Type)434
+ *435
+ * Tihs is similar to SIMDE_ALIGN_ASSUME_TO, except that it takes a436
+ * type instead of a numeric value. */437
+#if defined(SIMDE_ALIGN_OF) && defined(SIMDE_ALIGN_ASSUME_TO)438
+#define SIMDE_ALIGN_ASSUME_LIKE(Pointer, Type) \439
+ SIMDE_ALIGN_ASSUME_TO(Pointer, SIMDE_ALIGN_OF(Type))440
+#endif441
+442
+/* SIMDE_ALIGN_CAST(Type, Pointer)443
+ *444
+ * SIMDE_ALIGN_CAST is like C++'s reinterpret_cast, but it will try445
+ * to silence warnings that some compilers may produce if you try446
+ * to assign to a type with increased alignment requirements.447
+ *448
+ * Note that it does *not* actually attempt to tell the compiler that449
+ * the pointer is aligned like the destination should be; that's the450
+ * job of the next macro. This macro is necessary for stupid APIs451
+ * like _mm_loadu_si128 where the input is a __m128i* but the function452
+ * is specifically for data which isn't necessarily aligned to453
+ * _Alignof(__m128i).454
+ */455
+#if HEDLEY_HAS_WARNING("-Wcast-align") || defined(__clang__) || \456
+ HEDLEY_GCC_VERSION_CHECK(3, 4, 0)457
+#define SIMDE_ALIGN_CAST(Type, Pointer) \458
+ (__extension__({ \459
+ HEDLEY_DIAGNOSTIC_PUSH \460
+ _Pragma("GCC diagnostic ignored \"-Wcast-align\"") \461
+ Type simde_r_ = \462
+ HEDLEY_REINTERPRET_CAST(Type, Pointer); \463
+ HEDLEY_DIAGNOSTIC_POP \464
+ simde_r_; \465
+ }))466
+#else467
+#define SIMDE_ALIGN_CAST(Type, Pointer) HEDLEY_REINTERPRET_CAST(Type, Pointer)468
+#endif469
+470
+/* SIMDE_ALIGN_ASSUME_CAST(Type, Pointer)471
+ *472
+ * This is sort of like a combination of a reinterpret_cast and a473
+ * SIMDE_ALIGN_ASSUME_LIKE. It uses SIMDE_ALIGN_ASSUME_LIKE to tell474
+ * the compiler that the pointer is aligned like the specified type475
+ * and casts the pointer to the specified type while suppressing any476
+ * warnings from the compiler about casting to a type with greater477
+ * alignment requirements.478
+ */479
+#define SIMDE_ALIGN_ASSUME_CAST(Type, Pointer) \480
+ SIMDE_ALIGN_ASSUME_LIKE(SIMDE_ALIGN_CAST(Type, Pointer), Type)481
+482
+#endif /* !defined(SIMDE_ALIGN_H) */483
obs-studio-26.1.0.tar.xz/libobs/util/simde/simde-arch.h -> obs-studio-26.1.1.tar.xz/libobs/util/simde/simde-arch.h
Changed
74
1
2
* an undefined macro being used (e.g., GCC with -Wundef).3
*4
* This was originally created for SIMDe5
- * <https://github.com/nemequ/simde> (hence the prefix), but this6
+ * <https://github.com/simd-everywhere/simde> (hence the prefix), but this7
* header has no dependencies and may be used anywhere. It is8
* originally based on information from9
* <https://sourceforge.net/p/predef/wiki/Architectures/>, though it10
* has been enhanced with additional information.11
*12
* If you improve this file, or find a bug, please file the issue at13
- * <https://github.com/nemequ/simde/issues>. If you copy this into14
+ * <https://github.com/simd-everywhere/simde/issues>. If you copy this into15
* your project, even if you change the prefix, please keep the links16
* to SIMDe intact so others know where to report issues, submit17
* enhancements, and find the latest version. */18
19
/* AMD64 / x86_6420
<https://en.wikipedia.org/wiki/X86-64> */21
#if defined(__amd64__) || defined(__amd64) || defined(__x86_64__) || \22
- defined(__x86_64) || defined(_M_X66) || defined(_M_AMD64)23
+ defined(__x86_64) || defined(_M_X64) || defined(_M_AMD64)24
#define SIMDE_ARCH_AMD64 100025
#endif26
27
28
#define SIMDE_ARCH_ARM_NEON SIMDE_ARCH_ARM29
#endif30
#endif31
+#if defined(__ARM_FEATURE_SVE)32
+#define SIMDE_ARCH_ARM_SVE33
+#endif34
35
/* Blackfin36
<https://en.wikipedia.org/wiki/Blackfin> */37
38
#define SIMDE_ARCH_X86_AVX 139
#endif40
#endif41
+#if defined(__AVX512VP2INTERSECT__)42
+#define SIMDE_ARCH_X86_AVX512VP2INTERSECT 143
+#endif44
+#if defined(__AVX512VBMI__)45
+#define SIMDE_ARCH_X86_AVX512VBMI 146
+#endif47
#if defined(__AVX512BW__)48
#define SIMDE_ARCH_X86_AVX512BW 149
#endif50
51
#if defined(__GFNI__)52
#define SIMDE_ARCH_X86_GFNI 153
#endif54
+#if defined(__PCLMUL__)55
+#define SIMDE_ARCH_X86_PCLMUL 156
+#endif57
+#if defined(__VPCLMULQDQ__)58
+#define SIMDE_ARCH_X86_VPCLMULQDQ 159
+#endif60
#endif61
62
/* Itanium63
64
#define SIMDE_ARCH_MIPS_CHECK(version) (0)65
#endif66
67
+#if defined(__mips_loongson_mmi)68
+#define SIMDE_ARCH_MIPS_LOONGSON_MMI 169
+#endif70
+71
/* Matsushita MN1030072
<https://en.wikipedia.org/wiki/MN103> */73
#if defined(__MN10300__) || defined(__mn10300__)74
obs-studio-26.1.0.tar.xz/libobs/util/simde/simde-common.h -> obs-studio-26.1.1.tar.xz/libobs/util/simde/simde-common.h
Changed
674
1
2
#include "hedley.h"3
4
#define SIMDE_VERSION_MAJOR 05
-#define SIMDE_VERSION_MINOR 56
-#define SIMDE_VERSION_MICRO 07
+#define SIMDE_VERSION_MINOR 78
+#define SIMDE_VERSION_MICRO 19
#define SIMDE_VERSION \10
HEDLEY_VERSION_ENCODE(SIMDE_VERSION_MAJOR, SIMDE_VERSION_MINOR, \11
SIMDE_VERSION_MICRO)12
13
-#include "simde-arch.h"14
-#include "simde-features.h"15
-#include "simde-diagnostic.h"16
-17
#include <stddef.h>18
#include <stdint.h>19
20
-#if HEDLEY_HAS_ATTRIBUTE(aligned) || HEDLEY_GCC_VERSION_CHECK(2, 95, 0) || \21
- HEDLEY_CRAY_VERSION_CHECK(8, 4, 0) || \22
- HEDLEY_IBM_VERSION_CHECK(11, 1, 0) || \23
- HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \24
- HEDLEY_PGI_VERSION_CHECK(19, 4, 0) || \25
- HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \26
- HEDLEY_TINYC_VERSION_CHECK(0, 9, 24) || \27
- HEDLEY_TI_VERSION_CHECK(8, 1, 0)28
-#define SIMDE_ALIGN(alignment) __attribute__((aligned(alignment)))29
-#elif defined(_MSC_VER) && !(defined(_M_ARM) && !defined(_M_ARM64))30
-#define SIMDE_ALIGN(alignment) __declspec(align(alignment))31
-#elif defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L)32
-#define SIMDE_ALIGN(alignment) _Alignas(alignment)33
-#elif defined(__cplusplus) && (__cplusplus >= 201103L)34
-#define SIMDE_ALIGN(alignment) alignas(alignment)35
-#else36
-#define SIMDE_ALIGN(alignment)37
-#endif38
-39
-#if HEDLEY_GNUC_VERSION_CHECK(2, 95, 0) || \40
- HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \41
- HEDLEY_IBM_VERSION_CHECK(11, 1, 0)42
-#define SIMDE_ALIGN_OF(T) (__alignof__(T))43
-#elif (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L)) || \44
- HEDLEY_HAS_FEATURE(c11_alignof)45
-#define SIMDE_ALIGN_OF(T) (_Alignof(T))46
-#elif (defined(__cplusplus) && (__cplusplus >= 201103L)) || \47
- HEDLEY_HAS_FEATURE(cxx_alignof)48
-#define SIMDE_ALIGN_OF(T) (alignof(T))49
-#endif50
-51
-#if defined(SIMDE_ALIGN_OF)52
-#define SIMDE_ALIGN_AS(N, T) SIMDE_ALIGN(SIMDE_ALIGN_OF(T))53
-#else54
-#define SIMDE_ALIGN_AS(N, T) SIMDE_ALIGN(N)55
+#include "simde-detect-clang.h"56
+#include "simde-arch.h"57
+#include "simde-features.h"58
+#include "simde-diagnostic.h"59
+#include "simde-math.h"60
+#include "simde-constify.h"61
+#include "simde-align.h"62
+63
+/* In some situations, SIMDe has to make large performance sacrifices64
+ * for small increases in how faithfully it reproduces an API, but65
+ * only a relatively small number of users will actually need the API66
+ * to be completely accurate. The SIMDE_FAST_* options can be used to67
+ * disable these trade-offs.68
+ *69
+ * They can be enabled by passing -DSIMDE_FAST_MATH to the compiler, or70
+ * the individual defines (e.g., -DSIMDE_FAST_NANS) if you only want to71
+ * enable some optimizations. Using -ffast-math and/or72
+ * -ffinite-math-only will also enable the relevant options. If you73
+ * don't want that you can pass -DSIMDE_NO_FAST_* to disable them. */74
+75
+/* Most programs avoid NaNs by never passing values which can result in76
+ * a NaN; for example, if you only pass non-negative values to the sqrt77
+ * functions, it won't generate a NaN. On some platforms, similar78
+ * functions handle NaNs differently; for example, the _mm_min_ps SSE79
+ * function will return 0.0 if you pass it (0.0, NaN), but the NEON80
+ * vminq_f32 function will return NaN. Making them behave like one81
+ * another is expensive; it requires generating a mask of all lanes82
+ * with NaNs, then performing the operation (e.g., vminq_f32), then83
+ * blending together the result with another vector using the mask.84
+ *85
+ * If you don't want SIMDe to worry about the differences between how86
+ * NaNs are handled on the two platforms, define this (or pass87
+ * -ffinite-math-only) */88
+#if !defined(SIMDE_FAST_MATH) && !defined(SIMDE_NO_FAST_MATH) && \89
+ defined(__FAST_MATH__)90
+#define SIMDE_FAST_MATH91
+#endif92
+93
+#if !defined(SIMDE_FAST_NANS) && !defined(SIMDE_NO_FAST_NANS)94
+#if defined(SIMDE_FAST_MATH)95
+#define SIMDE_FAST_NANS96
+#elif defined(__FINITE_MATH_ONLY__)97
+#if __FINITE_MATH_ONLY__98
+#define SIMDE_FAST_NANS99
+#endif100
+#endif101
+#endif102
+103
+/* Many functions are defined as using the current rounding mode104
+ * (i.e., the SIMD version of fegetround()) when converting to105
+ * an integer. For example, _mm_cvtpd_epi32. Unfortunately,106
+ * on some platforms (such as ARMv8+ where round-to-nearest is107
+ * always used, regardless of the FPSCR register) this means we108
+ * have to first query the current rounding mode, then choose109
+ * the proper function (rounnd110
+ , ceil, floor, etc.) */111
+#if !defined(SIMDE_FAST_ROUND_MODE) && !defined(SIMDE_NO_FAST_ROUND_MODE) && \112
+ defined(SIMDE_FAST_MATH)113
+#define SIMDE_FAST_ROUND_MODE114
+#endif115
+116
+/* This controls how ties are rounded. For example, does 10.5 round to117
+ * 10 or 11? IEEE 754 specifies round-towards-even, but ARMv7 (for118
+ * example) doesn't support it and it must be emulated (which is rather119
+ * slow). If you're okay with just using the default for whatever arch120
+ * you're on, you should definitely define this.121
+ *122
+ * Note that we don't use this macro to avoid correct implementations123
+ * in functions which are explicitly about rounding (such as vrnd* on124
+ * NEON, _mm_round_* on x86, etc.); it is only used for code where125
+ * rounding is a component in another function, and even then it isn't126
+ * usually a problem since such functions will use the current rounding127
+ * mode. */128
+#if !defined(SIMDE_FAST_ROUND_TIES) && !defined(SIMDE_NO_FAST_ROUND_TIES) && \129
+ defined(SIMDE_FAST_MATH)130
+#define SIMDE_FAST_ROUND_TIES131
+#endif132
+133
+/* For functions which convert from one type to another (mostly from134
+ * floating point to integer types), sometimes we need to do a range135
+ * check and potentially return a different result if the value136
+ * falls outside that range. Skipping this check can provide a137
+ * performance boost, at the expense of faithfulness to the API we're138
+ * emulating. */139
+#if !defined(SIMDE_FAST_CONVERSION_RANGE) && \140
+ !defined(SIMDE_NO_FAST_CONVERSION_RANGE) && defined(SIMDE_FAST_MATH)141
+#define SIMDE_FAST_CONVERSION_RANGE142
#endif143
144
-#define simde_assert_aligned(alignment, val) \145
- simde_assert_int(HEDLEY_REINTERPRET_CAST( \146
- uintptr_t, HEDLEY_REINTERPRET_CAST( \147
- const void *, (val))) % \148
- (alignment), \149
- ==, 0)150
-151
#if HEDLEY_HAS_BUILTIN(__builtin_constant_p) || \152
HEDLEY_GCC_VERSION_CHECK(3, 4, 0) || \153
HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \154
155
#define SIMDE_CHECK_CONSTANT_(expr) (std::is_constant_evaluated())156
#endif157
158
-/* diagnose_if + __builtin_constant_p was broken until clang 9,159
- * which is when __FILE_NAME__ was added. */160
-#if defined(SIMDE_CHECK_CONSTANT_) && defined(__FILE_NAME__)161
+#if !defined(SIMDE_NO_CHECK_IMMEDIATE_CONSTANT)162
+#if defined(SIMDE_CHECK_CONSTANT_) && \163
+ SIMDE_DETECT_CLANG_VERSION_CHECK(9, 0, 0) && \164
+ (!defined(__apple_build_version__) || \165
+ ((__apple_build_version__ < 11000000) || \166
+ (__apple_build_version__ >= 12000000)))167
#define SIMDE_REQUIRE_CONSTANT(arg) \168
HEDLEY_REQUIRE_MSG(SIMDE_CHECK_CONSTANT_(arg), \169
"`" #arg "' must be constant")170
#else171
#define SIMDE_REQUIRE_CONSTANT(arg)172
#endif173
+#else174
+#define SIMDE_REQUIRE_CONSTANT(arg)175
+#endif176
177
#define SIMDE_REQUIRE_RANGE(arg, min, max) \178
HEDLEY_REQUIRE_MSG((((arg) >= (min)) && ((arg) <= (max))), \179
180
SIMDE_REQUIRE_CONSTANT(arg) \181
SIMDE_REQUIRE_RANGE(arg, min, max)182
183
-/* SIMDE_ASSUME_ALIGNED allows you to (try to) tell the compiler184
- * that a pointer is aligned to an `alignment`-byte boundary. */185
-#if HEDLEY_HAS_BUILTIN(__builtin_assume_aligned) || \186
- HEDLEY_GCC_VERSION_CHECK(4, 7, 0)187
-#define SIMDE_ASSUME_ALIGNED(alignment, v) \188
- HEDLEY_REINTERPRET_CAST(__typeof__(v), \189
- __builtin_assume_aligned(v, alignment))190
-#elif defined(__cplusplus) && (__cplusplus > 201703L)191
-#define SIMDE_ASSUME_ALIGNED(alignment, v) std::assume_aligned<alignment>(v)192
-#elif HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)193
-#define SIMDE_ASSUME_ALIGNED(alignment, v) \194
- (__extension__({ \195
- __typeof__(v) simde_assume_aligned_t_ = (v); \196
- __assume_aligned(simde_assume_aligned_t_, alignment); \197
- simde_assume_aligned_t_; \198
- }))199
-#else200
-#define SIMDE_ASSUME_ALIGNED(alignment, v) (v)201
-#endif202
-203
-/* SIMDE_ALIGN_CAST allows you to convert to a type with greater204
- * aligment requirements without triggering a warning. */205
-#if HEDLEY_HAS_WARNING("-Wcast-align")206
-#define SIMDE_ALIGN_CAST(T, v) \207
- (__extension__({ \208
- HEDLEY_DIAGNOSTIC_PUSH \209
- _Pragma("clang diagnostic ignored \"-Wcast-align\"") \210
- T simde_r_ = HEDLEY_REINTERPRET_CAST(T, v); \211
- HEDLEY_DIAGNOSTIC_POP \212
- simde_r_; \213
- }))214
-#else215
-#define SIMDE_ALIGN_CAST(T, v) HEDLEY_REINTERPRET_CAST(T, v)216
+/* A copy of HEDLEY_STATIC_ASSERT, except we don't define an empty217
+ * fallback if we can't find an implementation; instead we have to218
+ * check if SIMDE_STATIC_ASSERT is defined before using it. */219
+#if !defined(__cplusplus) && \220
+ ((defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L)) || \221
+ HEDLEY_HAS_FEATURE(c_static_assert) || \222
+ HEDLEY_GCC_VERSION_CHECK(6, 0, 0) || \223
+ HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || defined(_Static_assert))224
+#define SIMDE_STATIC_ASSERT(expr, message) _Static_assert(expr, message)225
+#elif (defined(__cplusplus) && (__cplusplus >= 201103L)) || \226
+ HEDLEY_MSVC_VERSION_CHECK(16, 0, 0)227
+#define SIMDE_STATIC_ASSERT(expr, message) \228
+ HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_( \229
+ static_assert(expr, message))230
#endif231
232
#if (HEDLEY_HAS_ATTRIBUTE(may_alias) && !defined(HEDLEY_SUNPRO_VERSION)) || \233
234
235
* SIMDE_VECTOR - Declaring a vector.236
* SIMDE_VECTOR_OPS - basic operations (binary and unary).237
+ * SIMDE_VECTOR_NEGATE - negating a vector238
* SIMDE_VECTOR_SCALAR - For binary operators, the second argument239
can be a scalar, in which case the result is as if that scalar240
had been broadcast to all lanes of a vector.241
242
#if HEDLEY_GCC_VERSION_CHECK(4, 8, 0)243
#define SIMDE_VECTOR(size) __attribute__((__vector_size__(size)))244
#define SIMDE_VECTOR_OPS245
+#define SIMDE_VECTOR_NEGATE246
#define SIMDE_VECTOR_SCALAR247
#define SIMDE_VECTOR_SUBSCRIPT248
#elif HEDLEY_INTEL_VERSION_CHECK(16, 0, 0)249
#define SIMDE_VECTOR(size) __attribute__((__vector_size__(size)))250
#define SIMDE_VECTOR_OPS251
+#define SIMDE_VECTOR_NEGATE252
/* ICC only supports SIMDE_VECTOR_SCALAR for constants */253
#define SIMDE_VECTOR_SUBSCRIPT254
#elif HEDLEY_GCC_VERSION_CHECK(4, 1, 0) || HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)255
256
#elif HEDLEY_HAS_ATTRIBUTE(vector_size)257
#define SIMDE_VECTOR(size) __attribute__((__vector_size__(size)))258
#define SIMDE_VECTOR_OPS259
+#define SIMDE_VECTOR_NEGATE260
#define SIMDE_VECTOR_SUBSCRIPT261
-#if HEDLEY_HAS_ATTRIBUTE(diagnose_if) /* clang 4.0 */262
+#if SIMDE_DETECT_CLANG_VERSION_CHECK(5, 0, 0)263
#define SIMDE_VECTOR_SCALAR264
#endif265
#endif266
267
#endif268
269
#if defined(SIMDE_ENABLE_OPENMP)270
-#define SIMDE_VECTORIZE _Pragma("omp simd")271
+#define SIMDE_VECTORIZE HEDLEY_PRAGMA(omp simd)272
#define SIMDE_VECTORIZE_SAFELEN(l) HEDLEY_PRAGMA(omp simd safelen(l))273
+#if defined(__clang__)274
+#define SIMDE_VECTORIZE_REDUCTION(r) \275
+ HEDLEY_DIAGNOSTIC_PUSH \276
+ _Pragma("clang diagnostic ignored \"-Wsign-conversion\"") \277
+ HEDLEY_PRAGMA(omp simd reduction(r)) HEDLEY_DIAGNOSTIC_POP278
+#else279
#define SIMDE_VECTORIZE_REDUCTION(r) HEDLEY_PRAGMA(omp simd reduction(r))280
+#endif281
#define SIMDE_VECTORIZE_ALIGNED(a) HEDLEY_PRAGMA(omp simd aligned(a))282
#elif defined(SIMDE_ENABLE_CILKPLUS)283
-#define SIMDE_VECTORIZE _Pragma("simd")284
+#define SIMDE_VECTORIZE HEDLEY_PRAGMA(simd)285
#define SIMDE_VECTORIZE_SAFELEN(l) HEDLEY_PRAGMA(simd vectorlength(l))286
#define SIMDE_VECTORIZE_REDUCTION(r) HEDLEY_PRAGMA(simd reduction(r))287
#define SIMDE_VECTORIZE_ALIGNED(a) HEDLEY_PRAGMA(simd aligned(a))288
#elif defined(__clang__) && !defined(HEDLEY_IBM_VERSION)289
-#define SIMDE_VECTORIZE _Pragma("clang loop vectorize(enable)")290
+#define SIMDE_VECTORIZE HEDLEY_PRAGMA(clang loop vectorize(enable))291
#define SIMDE_VECTORIZE_SAFELEN(l) HEDLEY_PRAGMA(clang loop vectorize_width(l))292
#define SIMDE_VECTORIZE_REDUCTION(r) SIMDE_VECTORIZE293
#define SIMDE_VECTORIZE_ALIGNED(a)294
#elif HEDLEY_GCC_VERSION_CHECK(4, 9, 0)295
-#define SIMDE_VECTORIZE _Pragma("GCC ivdep")296
+#define SIMDE_VECTORIZE HEDLEY_PRAGMA(GCC ivdep)297
#define SIMDE_VECTORIZE_SAFELEN(l) SIMDE_VECTORIZE298
#define SIMDE_VECTORIZE_REDUCTION(r) SIMDE_VECTORIZE299
#define SIMDE_VECTORIZE_ALIGNED(a)300
#elif HEDLEY_CRAY_VERSION_CHECK(5, 0, 0)301
-#define SIMDE_VECTORIZE _Pragma("_CRI ivdep")302
+#define SIMDE_VECTORIZE HEDLEY_PRAGMA(_CRI ivdep)303
#define SIMDE_VECTORIZE_SAFELEN(l) SIMDE_VECTORIZE304
#define SIMDE_VECTORIZE_REDUCTION(r) SIMDE_VECTORIZE305
#define SIMDE_VECTORIZE_ALIGNED(a)306
307
HEDLEY_DIAGNOSTIC_POP308
#endif309
310
-#if HEDLEY_HAS_WARNING("-Wpedantic")311
-#define SIMDE_DIAGNOSTIC_DISABLE_INT128 \312
- _Pragma("clang diagnostic ignored \"-Wpedantic\"")313
-#elif defined(HEDLEY_GCC_VERSION)314
-#define SIMDE_DIAGNOSTIC_DISABLE_INT128 \315
- _Pragma("GCC diagnostic ignored \"-Wpedantic\"")316
-#else317
-#define SIMDE_DIAGNOSTIC_DISABLE_INT128318
-#endif319
-320
#if defined(__SIZEOF_INT128__)321
#define SIMDE_HAVE_INT128_322
HEDLEY_DIAGNOSTIC_PUSH323
-SIMDE_DIAGNOSTIC_DISABLE_INT128324
+SIMDE_DIAGNOSTIC_DISABLE_PEDANTIC_325
typedef __int128 simde_int128;326
typedef unsigned __int128 simde_uint128;327
HEDLEY_DIAGNOSTIC_POP328
329
#endif330
typedef SIMDE_FLOAT64_TYPE simde_float64;331
332
-/* Whether to assume that the compiler can auto-vectorize reasonably333
- well. This will cause SIMDe to attempt to compose vector334
- operations using more simple vector operations instead of minimize335
- serial work.336
-337
- As an example, consider the _mm_add_ss(a, b) function from SSE,338
- which returns { a0 + b0, a1, a2, a3 }. This pattern is repeated339
- for other operations (sub, mul, etc.).340
-341
- The naïve implementation would result in loading a0 and b0, adding342
- them into a temporary variable, then splicing that value into a new343
- vector with the remaining elements from a.344
-345
- On platforms which support vectorization, it's generally faster to346
- simply perform the operation on the entire vector to avoid having347
- to move data between SIMD registers and non-SIMD registers.348
- Basically, instead of the temporary variable being (a0 + b0) it349
- would be a vector of (a + b), which is then combined with a to form350
- the result.351
-352
- By default, SIMDe will prefer the pure-vector versions if we detect353
- a vector ISA extension, but this can be overridden by defining354
- SIMDE_NO_ASSUME_VECTORIZATION. You can also define355
- SIMDE_ASSUME_VECTORIZATION if you want to force SIMDe to use the356
- vectorized version. */357
-#if !defined(SIMDE_NO_ASSUME_VECTORIZATION) && \358
- !defined(SIMDE_ASSUME_VECTORIZATION)359
-#if defined(__SSE__) || defined(__ARM_NEON) || defined(__mips_msa) || \360
- defined(__ALTIVEC__) || defined(__wasm_simd128__)361
-#define SIMDE_ASSUME_VECTORIZATION362
-#endif363
-#endif364
-365
#if HEDLEY_HAS_WARNING("-Wbad-function-cast")366
#define SIMDE_CONVERT_FTOI(T, v) \367
HEDLEY_DIAGNOSTIC_PUSH \368
369
#define SIMDE_CONVERT_FTOI(T, v) ((T)(v))370
#endif371
372
+/* TODO: detect compilers which support this outside of C11 mode */373
#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L)374
#define SIMDE_CHECKED_REINTERPRET_CAST(to, from, value) \375
- (_Generic((value), to : (value), from : ((to)(value))))376
+ _Generic((value), to \377
+ : (value), default \378
+ : (_Generic((value), from \379
+ : ((to)(value)))))380
#define SIMDE_CHECKED_STATIC_CAST(to, from, value) \381
- (_Generic((value), to : (value), from : ((to)(value))))382
+ _Generic((value), to \383
+ : (value), default \384
+ : (_Generic((value), from \385
+ : ((to)(value)))))386
#else387
#define SIMDE_CHECKED_REINTERPRET_CAST(to, from, value) \388
HEDLEY_REINTERPRET_CAST(to, value)389
390
#if defined(__STDC_HOSTED__)391
#define SIMDE_STDC_HOSTED __STDC_HOSTED__392
#else393
-#if defined(HEDLEY_PGI_VERSION_CHECK) || defined(HEDLEY_MSVC_VERSION_CHECK)394
+#if defined(HEDLEY_PGI_VERSION) || defined(HEDLEY_MSVC_VERSION)395
#define SIMDE_STDC_HOSTED 1396
#else397
#define SIMDE_STDC_HOSTED 0398
399
#endif400
401
/* Try to deal with environments without a standard library. */402
-#if !defined(simde_memcpy) || !defined(simde_memset)403
-#if !defined(SIMDE_NO_STRING_H) && defined(__has_include)404
-#if __has_include(<string.h>)405
-#include <string.h>406
#if !defined(simde_memcpy)407
-#define simde_memcpy(dest, src, n) memcpy(dest, src, n)408
+#if HEDLEY_HAS_BUILTIN(__builtin_memcpy)409
+#define simde_memcpy(dest, src, n) __builtin_memcpy(dest, src, n)410
+#endif411
#endif412
#if !defined(simde_memset)413
-#define simde_memset(s, c, n) memset(s, c, n)414
+#if HEDLEY_HAS_BUILTIN(__builtin_memset)415
+#define simde_memset(s, c, n) __builtin_memset(s, c, n)416
#endif417
-#else418
+#endif419
+#if !defined(simde_memcmp)420
+#if HEDLEY_HAS_BUILTIN(__builtin_memcmp)421
+#define simde_memcmp(s1, s2, n) __builtin_memcmp(s1, s2, n)422
+#endif423
+#endif424
+425
+#if !defined(simde_memcpy) || !defined(simde_memset) || !defined(simde_memcmp)426
+#if !defined(SIMDE_NO_STRING_H)427
+#if defined(__has_include)428
+#if !__has_include(<string.h>)429
#define SIMDE_NO_STRING_H430
#endif431
+#elif (SIMDE_STDC_HOSTED == 0)432
+#define SIMDE_NO_STRING_H433
#endif434
#endif435
-#if !defined(simde_memcpy) || !defined(simde_memset)436
-#if !defined(SIMDE_NO_STRING_H) && (SIMDE_STDC_HOSTED == 1)437
+438
+#if !defined(SIMDE_NO_STRING_H)439
#include <string.h>440
#if !defined(simde_memcpy)441
#define simde_memcpy(dest, src, n) memcpy(dest, src, n)442
443
#if !defined(simde_memset)444
#define simde_memset(s, c, n) memset(s, c, n)445
#endif446
-#elif (HEDLEY_HAS_BUILTIN(__builtin_memcpy) && \447
- HEDLEY_HAS_BUILTIN(__builtin_memset)) || \448
- HEDLEY_GCC_VERSION_CHECK(4, 2, 0)449
-#if !defined(simde_memcpy)450
-#define simde_memcpy(dest, src, n) __builtin_memcpy(dest, src, n)451
-#endif452
-#if !defined(simde_memset)453
-#define simde_memset(s, c, n) __builtin_memset(s, c, n)454
+#if !defined(simde_memcmp)455
+#define simde_memcmp(s1, s2, n) memcmp(s1, s2, n)456
#endif457
#else458
/* These are meant to be portable, not fast. If you're hitting them you459
460
}461
#define simde_memset(s, c, n) simde_memset_(s, c, n)462
#endif463
-#endif /* !defined(SIMDE_NO_STRING_H) && (SIMDE_STDC_HOSTED == 1) */464
-#endif /* !defined(simde_memcpy) || !defined(simde_memset) */465
466
-#include "simde-math.h"467
+#if !defined(simde_memcmp)468
+SIMDE_FUCTION_ATTRIBUTES469
+int simde_memcmp_(const void *s1, const void *s2, size_t n)470
+{471
+ unsigned char *s1_ = HEDLEY_STATIC_CAST(unsigned char *, s1);472
+ unsigned char *s2_ = HEDLEY_STATIC_CAST(unsigned char *, s2);473
+ for (size_t i = 0; i < len; i++) {474
+ if (s1_[i] != s2_[i]) {475
+ return (int)(s1_[i] - s2_[i]);476
+ }477
+ }478
+ return 0;479
+}480
+#define simde_memcmp(s1, s2, n) simde_memcmp_(s1, s2, n)481
+#endif482
+#endif483
+#endif484
485
#if defined(FE_ALL_EXCEPT)486
#define SIMDE_HAVE_FENV_H487
488
489
#include "check.h"490
491
+/* GCC/clang have a bunch of functionality in builtins which we would492
+ * like to access, but the suffixes indicate whether the operate on493
+ * int, long, or long long, not fixed width types (e.g., int32_t).494
+ * we use these macros to attempt to map from fixed-width to the495
+ * names GCC uses. Note that you should still cast the input(s) and496
+ * return values (to/from SIMDE_BUILTIN_TYPE_*_) since often even if497
+ * types are the same size they may not be compatible according to the498
+ * compiler. For example, on x86 long and long lonsg are generally499
+ * both 64 bits, but platforms vary on whether an int64_t is mapped500
+ * to a long or long long. */501
+502
+#include <limits.h>503
+504
+HEDLEY_DIAGNOSTIC_PUSH505
+SIMDE_DIAGNOSTIC_DISABLE_CPP98_COMPAT_PEDANTIC_506
+507
+#if (INT8_MAX == INT_MAX) && (INT8_MIN == INT_MIN)508
+#define SIMDE_BUILTIN_SUFFIX_8_509
+#define SIMDE_BUILTIN_TYPE_8_ int510
+#elif (INT8_MAX == LONG_MAX) && (INT8_MIN == LONG_MIN)511
+#define SIMDE_BUILTIN_SUFFIX_8_ l512
+#define SIMDE_BUILTIN_TYPE_8_ long513
+#elif (INT8_MAX == LLONG_MAX) && (INT8_MIN == LLONG_MIN)514
+#define SIMDE_BUILTIN_SUFFIX_8_ ll515
+#define SIMDE_BUILTIN_TYPE_8_ long long516
+#endif517
+518
+#if (INT16_MAX == INT_MAX) && (INT16_MIN == INT_MIN)519
+#define SIMDE_BUILTIN_SUFFIX_16_520
+#define SIMDE_BUILTIN_TYPE_16_ int521
+#elif (INT16_MAX == LONG_MAX) && (INT16_MIN == LONG_MIN)522
+#define SIMDE_BUILTIN_SUFFIX_16_ l523
+#define SIMDE_BUILTIN_TYPE_16_ long524
+#elif (INT16_MAX == LLONG_MAX) && (INT16_MIN == LLONG_MIN)525
+#define SIMDE_BUILTIN_SUFFIX_16_ ll526
+#define SIMDE_BUILTIN_TYPE_16_ long long527
+#endif528
+529
+#if (INT32_MAX == INT_MAX) && (INT32_MIN == INT_MIN)530
+#define SIMDE_BUILTIN_SUFFIX_32_531
+#define SIMDE_BUILTIN_TYPE_32_ int532
+#elif (INT32_MAX == LONG_MAX) && (INT32_MIN == LONG_MIN)533
+#define SIMDE_BUILTIN_SUFFIX_32_ l534
+#define SIMDE_BUILTIN_TYPE_32_ long535
+#elif (INT32_MAX == LLONG_MAX) && (INT32_MIN == LLONG_MIN)536
+#define SIMDE_BUILTIN_SUFFIX_32_ ll537
+#define SIMDE_BUILTIN_TYPE_32_ long long538
+#endif539
+540
+#if (INT64_MAX == INT_MAX) && (INT64_MIN == INT_MIN)541
+#define SIMDE_BUILTIN_SUFFIX_64_542
+#define SIMDE_BUILTIN_TYPE_64_ int543
+#elif (INT64_MAX == LONG_MAX) && (INT64_MIN == LONG_MIN)544
+#define SIMDE_BUILTIN_SUFFIX_64_ l545
+#define SIMDE_BUILTIN_TYPE_64_ long546
+#elif (INT64_MAX == LLONG_MAX) && (INT64_MIN == LLONG_MIN)547
+#define SIMDE_BUILTIN_SUFFIX_64_ ll548
+#define SIMDE_BUILTIN_TYPE_64_ long long549
+#endif550
+551
+#if defined(SIMDE_BUILTIN_SUFFIX_8_)552
+#define SIMDE_BUILTIN_8_(name) \553
+ HEDLEY_CONCAT3(__builtin_, name, SIMDE_BUILTIN_SUFFIX_8_)554
+#define SIMDE_BUILTIN_HAS_8_(name) \555
+ HEDLEY_HAS_BUILTIN( \556
+ HEDLEY_CONCAT3(__builtin_, name, SIMDE_BUILTIN_SUFFIX_8_))557
+#else558
+#define SIMDE_BUILTIN_HAS_8_(name) 0559
+#endif560
+#if defined(SIMDE_BUILTIN_SUFFIX_16_)561
+#define SIMDE_BUILTIN_16_(name) \562
+ HEDLEY_CONCAT3(__builtin_, name, SIMDE_BUILTIN_SUFFIX_16_)563
+#define SIMDE_BUILTIN_HAS_16_(name) \564
+ HEDLEY_HAS_BUILTIN( \565
+ HEDLEY_CONCAT3(__builtin_, name, SIMDE_BUILTIN_SUFFIX_16_))566
+#else567
+#define SIMDE_BUILTIN_HAS_16_(name) 0568
+#endif569
+#if defined(SIMDE_BUILTIN_SUFFIX_32_)570
+#define SIMDE_BUILTIN_32_(name) \571
+ HEDLEY_CONCAT3(__builtin_, name, SIMDE_BUILTIN_SUFFIX_32_)572
+#define SIMDE_BUILTIN_HAS_32_(name) \573
+ HEDLEY_HAS_BUILTIN( \574
+ HEDLEY_CONCAT3(__builtin_, name, SIMDE_BUILTIN_SUFFIX_32_))575
+#else576
+#define SIMDE_BUILTIN_HAS_32_(name) 0577
+#endif578
+#if defined(SIMDE_BUILTIN_SUFFIX_64_)579
+#define SIMDE_BUILTIN_64_(name) \580
+ HEDLEY_CONCAT3(__builtin_, name, SIMDE_BUILTIN_SUFFIX_64_)581
+#define SIMDE_BUILTIN_HAS_64_(name) \582
+ HEDLEY_HAS_BUILTIN( \583
+ HEDLEY_CONCAT3(__builtin_, name, SIMDE_BUILTIN_SUFFIX_64_))584
+#else585
+#define SIMDE_BUILTIN_HAS_64_(name) 0586
+#endif587
+588
+HEDLEY_DIAGNOSTIC_POP589
+590
/* Sometimes we run into problems with specific versions of compilers591
which make the native versions unusable for us. Often this is due592
to missing functions, sometimes buggy implementations, etc. These593
594
#if defined(SIMDE_ARCH_X86) && !defined(SIMDE_ARCH_AMD64)595
#define SIMDE_BUG_GCC_94482596
#endif597
+#if (defined(SIMDE_ARCH_X86) && !defined(SIMDE_ARCH_AMD64)) || \598
+ defined(SIMDE_ARCH_SYSTEMZ)599
+#define SIMDE_BUG_GCC_53784600
+#endif601
+#if defined(SIMDE_ARCH_X86) || defined(SIMDE_ARCH_AMD64)602
+#if HEDLEY_GCC_VERSION_CHECK(4, 3, 0) /* -Wsign-conversion */603
+#define SIMDE_BUG_GCC_95144604
+#endif605
+#endif606
#if !HEDLEY_GCC_VERSION_CHECK(9, 4, 0) && defined(SIMDE_ARCH_AARCH64)607
#define SIMDE_BUG_GCC_94488608
#endif609
-#if defined(SIMDE_ARCH_POWER)610
+#if defined(SIMDE_ARCH_ARM)611
+#define SIMDE_BUG_GCC_95399612
+#define SIMDE_BUG_GCC_95471613
+#elif defined(SIMDE_ARCH_POWER)614
#define SIMDE_BUG_GCC_95227615
+#define SIMDE_BUG_GCC_95782616
+#elif defined(SIMDE_ARCH_X86) || defined(SIMDE_ARCH_AMD64)617
+#if !HEDLEY_GCC_VERSION_CHECK(10, 2, 0) && !defined(__OPTIMIZE__)618
+#define SIMDE_BUG_GCC_96174619
+#endif620
#endif621
#define SIMDE_BUG_GCC_95399622
#elif defined(__clang__)623
#if defined(SIMDE_ARCH_AARCH64)624
#define SIMDE_BUG_CLANG_45541625
+#define SIMDE_BUG_CLANG_46844626
+#define SIMDE_BUG_CLANG_48257627
+#if SIMDE_DETECT_CLANG_VERSION_CHECK(10, 0, 0) && \628
+ SIMDE_DETECT_CLANG_VERSION_NOT(11, 0, 0)629
+#define SIMDE_BUG_CLANG_BAD_VI64_OPS630
+#endif631
+#endif632
+#if defined(SIMDE_ARCH_POWER)633
+#define SIMDE_BUG_CLANG_46770634
+#endif635
+#if defined(_ARCH_PWR9) && !SIMDE_DETECT_CLANG_VERSION_CHECK(12, 0, 0) && \636
+ !defined(__OPTIMIZE__)637
+#define SIMDE_BUG_CLANG_POWER9_16x4_BAD_SHIFT638
+#endif639
+#if defined(SIMDE_ARCH_X86) || defined(SIMDE_ARCH_AMD64)640
+#if HEDLEY_HAS_WARNING("-Wsign-conversion") && \641
+ SIMDE_DETECT_CLANG_VERSION_NOT(11, 0, 0)642
+#define SIMDE_BUG_CLANG_45931643
+#endif644
+#if HEDLEY_HAS_WARNING("-Wvector-conversion") && \645
+ SIMDE_DETECT_CLANG_VERSION_NOT(11, 0, 0)646
+#define SIMDE_BUG_CLANG_44589647
+#endif648
#endif649
+#define SIMDE_BUG_CLANG_45959650
+#elif defined(HEDLEY_MSVC_VERSION)651
+#if defined(SIMDE_ARCH_X86)652
+#define SIMDE_BUG_MSVC_ROUND_EXTRACT653
#endif654
-#if defined(HEDLEY_EMSCRIPTEN_VERSION)655
-#define SIMDE_BUG_EMSCRIPTEN_MISSING_IMPL /* Placeholder for (as yet) unfiled issues. */656
-#define SIMDE_BUG_EMSCRIPTEN_5242657
+#elif defined(HEDLEY_INTEL_VERSION)658
+#define SIMDE_BUG_INTEL_857088659
#endif660
#endif661
662
/* GCC and Clang both have the same issue:663
* https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95144664
* https://bugs.llvm.org/show_bug.cgi?id=45931665
+ * This is just an easy way to work around it.666
*/667
-#if HEDLEY_HAS_WARNING("-Wsign-conversion") || HEDLEY_GCC_VERSION_CHECK(4, 3, 0)668
+#if (HEDLEY_HAS_WARNING("-Wsign-conversion") && \669
+ SIMDE_DETECT_CLANG_VERSION_NOT(11, 0, 0)) || \670
+ HEDLEY_GCC_VERSION_CHECK(4, 3, 0)671
#define SIMDE_BUG_IGNORE_SIGN_CONVERSION(expr) \672
(__extension__({ \673
HEDLEY_DIAGNOSTIC_PUSH \674
obs-studio-26.1.1.tar.xz/libobs/util/simde/simde-constify.h
Added
927
1
2
+/* SPDX-License-Identifier: MIT3
+ *4
+ * Permission is hereby granted, free of charge, to any person5
+ * obtaining a copy of this software and associated documentation6
+ * files (the "Software"), to deal in the Software without7
+ * restriction, including without limitation the rights to use, copy,8
+ * modify, merge, publish, distribute, sublicense, and/or sell copies9
+ * of the Software, and to permit persons to whom the Software is10
+ * furnished to do so, subject to the following conditions:11
+ *12
+ * The above copyright notice and this permission notice shall be13
+ * included in all copies or substantial portions of the Software.14
+ *15
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,16
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF17
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND18
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS19
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN20
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN21
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE22
+ * SOFTWARE.23
+ *24
+ * Copyright:25
+ * 2020 Evan Nemerson <evan@nemerson.com>26
+ */27
+28
+/* Constify macros. For internal use only.29
+ *30
+ * These are used to make it possible to call a function which takes31
+ * an Integer Constant Expression (ICE) using a compile time constant.32
+ * Technically it would also be possible to use a value not trivially33
+ * known by the compiler, but there would be a siginficant performance34
+ * hit (a switch switch is used).35
+ *36
+ * The basic idea is pretty simple; we just emit a do while loop which37
+ * contains a switch with a case for every possible value of the38
+ * constant.39
+ *40
+ * As long as the value you pass to the function in constant, pretty41
+ * much any copmiler shouldn't have a problem generating exactly the42
+ * same code as if you had used an ICE.43
+ *44
+ * This is intended to be used in the SIMDe implementations of45
+ * functions the compilers require to be an ICE, but the other benefit46
+ * is that if we also disable the warnings from47
+ * SIMDE_REQUIRE_CONSTANT_RANGE we can actually just allow the tests48
+ * to use non-ICE parameters49
+ */50
+51
+#if !defined(SIMDE_CONSTIFY_H)52
+#define SIMDE_CONSTIFY_H53
+54
+#include "simde-diagnostic.h"55
+56
+HEDLEY_DIAGNOSTIC_PUSH57
+SIMDE_DIAGNOSTIC_DISABLE_VARIADIC_MACROS_58
+SIMDE_DIAGNOSTIC_DISABLE_CPP98_COMPAT_PEDANTIC_59
+60
+#define SIMDE_CONSTIFY_2_(func_name, result, default_case, imm, ...) \61
+ do { \62
+ switch (imm) { \63
+ case 0: \64
+ result = func_name(__VA_ARGS__, 0); \65
+ break; \66
+ case 1: \67
+ result = func_name(__VA_ARGS__, 1); \68
+ break; \69
+ default: \70
+ result = default_case; \71
+ break; \72
+ } \73
+ } while (0)74
+75
+#define SIMDE_CONSTIFY_4_(func_name, result, default_case, imm, ...) \76
+ do { \77
+ switch (imm) { \78
+ case 0: \79
+ result = func_name(__VA_ARGS__, 0); \80
+ break; \81
+ case 1: \82
+ result = func_name(__VA_ARGS__, 1); \83
+ break; \84
+ case 2: \85
+ result = func_name(__VA_ARGS__, 2); \86
+ break; \87
+ case 3: \88
+ result = func_name(__VA_ARGS__, 3); \89
+ break; \90
+ default: \91
+ result = default_case; \92
+ break; \93
+ } \94
+ } while (0)95
+96
+#define SIMDE_CONSTIFY_8_(func_name, result, default_case, imm, ...) \97
+ do { \98
+ switch (imm) { \99
+ case 0: \100
+ result = func_name(__VA_ARGS__, 0); \101
+ break; \102
+ case 1: \103
+ result = func_name(__VA_ARGS__, 1); \104
+ break; \105
+ case 2: \106
+ result = func_name(__VA_ARGS__, 2); \107
+ break; \108
+ case 3: \109
+ result = func_name(__VA_ARGS__, 3); \110
+ break; \111
+ case 4: \112
+ result = func_name(__VA_ARGS__, 4); \113
+ break; \114
+ case 5: \115
+ result = func_name(__VA_ARGS__, 5); \116
+ break; \117
+ case 6: \118
+ result = func_name(__VA_ARGS__, 6); \119
+ break; \120
+ case 7: \121
+ result = func_name(__VA_ARGS__, 7); \122
+ break; \123
+ default: \124
+ result = default_case; \125
+ break; \126
+ } \127
+ } while (0)128
+129
+#define SIMDE_CONSTIFY_16_(func_name, result, default_case, imm, ...) \130
+ do { \131
+ switch (imm) { \132
+ case 0: \133
+ result = func_name(__VA_ARGS__, 0); \134
+ break; \135
+ case 1: \136
+ result = func_name(__VA_ARGS__, 1); \137
+ break; \138
+ case 2: \139
+ result = func_name(__VA_ARGS__, 2); \140
+ break; \141
+ case 3: \142
+ result = func_name(__VA_ARGS__, 3); \143
+ break; \144
+ case 4: \145
+ result = func_name(__VA_ARGS__, 4); \146
+ break; \147
+ case 5: \148
+ result = func_name(__VA_ARGS__, 5); \149
+ break; \150
+ case 6: \151
+ result = func_name(__VA_ARGS__, 6); \152
+ break; \153
+ case 7: \154
+ result = func_name(__VA_ARGS__, 7); \155
+ break; \156
+ case 8: \157
+ result = func_name(__VA_ARGS__, 8); \158
+ break; \159
+ case 9: \160
+ result = func_name(__VA_ARGS__, 9); \161
+ break; \162
+ case 10: \163
+ result = func_name(__VA_ARGS__, 10); \164
+ break; \165
+ case 11: \166
+ result = func_name(__VA_ARGS__, 11); \167
+ break; \168
+ case 12: \169
+ result = func_name(__VA_ARGS__, 12); \170
+ break; \171
+ case 13: \172
+ result = func_name(__VA_ARGS__, 13); \173
+ break; \174
+ case 14: \175
+ result = func_name(__VA_ARGS__, 14); \176
+ break; \177
+ case 15: \178
+ result = func_name(__VA_ARGS__, 15); \179
+ break; \180
+ default: \181
+ result = default_case; \182
+ break; \183
+ } \184
+ } while (0)185
+186
+#define SIMDE_CONSTIFY_32_(func_name, result, default_case, imm, ...) \187
+ do { \188
+ switch (imm) { \189
+ case 0: \190
+ result = func_name(__VA_ARGS__, 0); \191
+ break; \192
+ case 1: \193
+ result = func_name(__VA_ARGS__, 1); \194
+ break; \195
+ case 2: \196
+ result = func_name(__VA_ARGS__, 2); \197
+ break; \198
+ case 3: \199
+ result = func_name(__VA_ARGS__, 3); \200
+ break; \201
+ case 4: \202
+ result = func_name(__VA_ARGS__, 4); \203
+ break; \204
+ case 5: \205
+ result = func_name(__VA_ARGS__, 5); \206
+ break; \207
+ case 6: \208
+ result = func_name(__VA_ARGS__, 6); \209
+ break; \210
+ case 7: \211
+ result = func_name(__VA_ARGS__, 7); \212
+ break; \213
+ case 8: \214
+ result = func_name(__VA_ARGS__, 8); \215
+ break; \216
+ case 9: \217
+ result = func_name(__VA_ARGS__, 9); \218
+ break; \219
+ case 10: \220
+ result = func_name(__VA_ARGS__, 10); \221
+ break; \222
+ case 11: \223
+ result = func_name(__VA_ARGS__, 11); \224
+ break; \225
+ case 12: \226
+ result = func_name(__VA_ARGS__, 12); \227
+ break; \228
+ case 13: \229
+ result = func_name(__VA_ARGS__, 13); \230
+ break; \231
+ case 14: \232
+ result = func_name(__VA_ARGS__, 14); \233
+ break; \234
+ case 15: \235
+ result = func_name(__VA_ARGS__, 15); \236
+ break; \237
+ case 16: \238
+ result = func_name(__VA_ARGS__, 16); \239
+ break; \240
+ case 17: \241
+ result = func_name(__VA_ARGS__, 17); \242
+ break; \243
+ case 18: \244
+ result = func_name(__VA_ARGS__, 18); \245
+ break; \246
+ case 19: \247
+ result = func_name(__VA_ARGS__, 19); \248
+ break; \249
+ case 20: \250
+ result = func_name(__VA_ARGS__, 20); \251
+ break; \252
+ case 21: \253
+ result = func_name(__VA_ARGS__, 21); \254
+ break; \255
+ case 22: \256
+ result = func_name(__VA_ARGS__, 22); \257
+ break; \258
+ case 23: \259
+ result = func_name(__VA_ARGS__, 23); \260
+ break; \261
+ case 24: \262
+ result = func_name(__VA_ARGS__, 24); \263
+ break; \264
+ case 25: \265
+ result = func_name(__VA_ARGS__, 25); \266
+ break; \267
+ case 26: \268
+ result = func_name(__VA_ARGS__, 26); \269
+ break; \270
+ case 27: \271
+ result = func_name(__VA_ARGS__, 27); \272
+ break; \273
+ case 28: \274
+ result = func_name(__VA_ARGS__, 28); \275
+ break; \276
+ case 29: \277
+ result = func_name(__VA_ARGS__, 29); \278
+ break; \279
+ case 30: \280
+ result = func_name(__VA_ARGS__, 30); \281
+ break; \282
+ case 31: \283
+ result = func_name(__VA_ARGS__, 31); \284
+ break; \285
+ default: \286
+ result = default_case; \287
+ break; \288
+ } \289
+ } while (0)290
+291
+#define SIMDE_CONSTIFY_64_(func_name, result, default_case, imm, ...) \292
+ do { \293
+ switch (imm) { \294
+ case 0: \295
+ result = func_name(__VA_ARGS__, 0); \296
+ break; \297
+ case 1: \298
+ result = func_name(__VA_ARGS__, 1); \299
+ break; \300
+ case 2: \301
+ result = func_name(__VA_ARGS__, 2); \302
+ break; \303
+ case 3: \304
+ result = func_name(__VA_ARGS__, 3); \305
+ break; \306
+ case 4: \307
+ result = func_name(__VA_ARGS__, 4); \308
+ break; \309
+ case 5: \310
+ result = func_name(__VA_ARGS__, 5); \311
+ break; \312
+ case 6: \313
+ result = func_name(__VA_ARGS__, 6); \314
+ break; \315
+ case 7: \316
+ result = func_name(__VA_ARGS__, 7); \317
+ break; \318
+ case 8: \319
+ result = func_name(__VA_ARGS__, 8); \320
+ break; \321
+ case 9: \322
+ result = func_name(__VA_ARGS__, 9); \323
+ break; \324
+ case 10: \325
+ result = func_name(__VA_ARGS__, 10); \326
+ break; \327
+ case 11: \328
+ result = func_name(__VA_ARGS__, 11); \329
+ break; \330
+ case 12: \331
+ result = func_name(__VA_ARGS__, 12); \332
+ break; \333
+ case 13: \334
+ result = func_name(__VA_ARGS__, 13); \335
+ break; \336
+ case 14: \337
+ result = func_name(__VA_ARGS__, 14); \338
+ break; \339
+ case 15: \340
+ result = func_name(__VA_ARGS__, 15); \341
+ break; \342
+ case 16: \343
+ result = func_name(__VA_ARGS__, 16); \344
+ break; \345
+ case 17: \346
+ result = func_name(__VA_ARGS__, 17); \347
+ break; \348
+ case 18: \349
+ result = func_name(__VA_ARGS__, 18); \350
+ break; \351
+ case 19: \352
+ result = func_name(__VA_ARGS__, 19); \353
+ break; \354
+ case 20: \355
+ result = func_name(__VA_ARGS__, 20); \356
+ break; \357
+ case 21: \358
+ result = func_name(__VA_ARGS__, 21); \359
+ break; \360
+ case 22: \361
+ result = func_name(__VA_ARGS__, 22); \362
+ break; \363
+ case 23: \364
+ result = func_name(__VA_ARGS__, 23); \365
+ break; \366
+ case 24: \367
+ result = func_name(__VA_ARGS__, 24); \368
+ break; \369
+ case 25: \370
+ result = func_name(__VA_ARGS__, 25); \371
+ break; \372
+ case 26: \373
+ result = func_name(__VA_ARGS__, 26); \374
+ break; \375
+ case 27: \376
+ result = func_name(__VA_ARGS__, 27); \377
+ break; \378
+ case 28: \379
+ result = func_name(__VA_ARGS__, 28); \380
+ break; \381
+ case 29: \382
+ result = func_name(__VA_ARGS__, 29); \383
+ break; \384
+ case 30: \385
+ result = func_name(__VA_ARGS__, 30); \386
+ break; \387
+ case 31: \388
+ result = func_name(__VA_ARGS__, 31); \389
+ break; \390
+ case 32: \391
+ result = func_name(__VA_ARGS__, 32); \392
+ break; \393
+ case 33: \394
+ result = func_name(__VA_ARGS__, 33); \395
+ break; \396
+ case 34: \397
+ result = func_name(__VA_ARGS__, 34); \398
+ break; \399
+ case 35: \400
+ result = func_name(__VA_ARGS__, 35); \401
+ break; \402
+ case 36: \403
+ result = func_name(__VA_ARGS__, 36); \404
+ break; \405
+ case 37: \406
+ result = func_name(__VA_ARGS__, 37); \407
+ break; \408
+ case 38: \409
+ result = func_name(__VA_ARGS__, 38); \410
+ break; \411
+ case 39: \412
+ result = func_name(__VA_ARGS__, 39); \413
+ break; \414
+ case 40: \415
+ result = func_name(__VA_ARGS__, 40); \416
+ break; \417
+ case 41: \418
+ result = func_name(__VA_ARGS__, 41); \419
+ break; \420
+ case 42: \421
+ result = func_name(__VA_ARGS__, 42); \422
+ break; \423
+ case 43: \424
+ result = func_name(__VA_ARGS__, 43); \425
+ break; \426
+ case 44: \427
+ result = func_name(__VA_ARGS__, 44); \428
+ break; \429
+ case 45: \430
+ result = func_name(__VA_ARGS__, 45); \431
+ break; \432
+ case 46: \433
+ result = func_name(__VA_ARGS__, 46); \434
+ break; \435
+ case 47: \436
+ result = func_name(__VA_ARGS__, 47); \437
+ break; \438
+ case 48: \439
+ result = func_name(__VA_ARGS__, 48); \440
+ break; \441
+ case 49: \442
+ result = func_name(__VA_ARGS__, 49); \443
+ break; \444
+ case 50: \445
+ result = func_name(__VA_ARGS__, 50); \446
+ break; \447
+ case 51: \448
+ result = func_name(__VA_ARGS__, 51); \449
+ break; \450
+ case 52: \451
+ result = func_name(__VA_ARGS__, 52); \452
+ break; \453
+ case 53: \454
+ result = func_name(__VA_ARGS__, 53); \455
+ break; \456
+ case 54: \457
+ result = func_name(__VA_ARGS__, 54); \458
+ break; \459
+ case 55: \460
+ result = func_name(__VA_ARGS__, 55); \461
+ break; \462
+ case 56: \463
+ result = func_name(__VA_ARGS__, 56); \464
+ break; \465
+ case 57: \466
+ result = func_name(__VA_ARGS__, 57); \467
+ break; \468
+ case 58: \469
+ result = func_name(__VA_ARGS__, 58); \470
+ break; \471
+ case 59: \472
+ result = func_name(__VA_ARGS__, 59); \473
+ break; \474
+ case 60: \475
+ result = func_name(__VA_ARGS__, 60); \476
+ break; \477
+ case 61: \478
+ result = func_name(__VA_ARGS__, 61); \479
+ break; \480
+ case 62: \481
+ result = func_name(__VA_ARGS__, 62); \482
+ break; \483
+ case 63: \484
+ result = func_name(__VA_ARGS__, 63); \485
+ break; \486
+ default: \487
+ result = default_case; \488
+ break; \489
+ } \490
+ } while (0)491
+492
+#define SIMDE_CONSTIFY_2_NO_RESULT_(func_name, default_case, imm, ...) \493
+ do { \494
+ switch (imm) { \495
+ case 0: \496
+ func_name(__VA_ARGS__, 0); \497
+ break; \498
+ case 1: \499
+ func_name(__VA_ARGS__, 1); \500
+ break; \501
+ default: \502
+ default_case; \503
+ break; \504
+ } \505
+ } while (0)506
+507
+#define SIMDE_CONSTIFY_4_NO_RESULT_(func_name, default_case, imm, ...) \508
+ do { \509
+ switch (imm) { \510
+ case 0: \511
+ func_name(__VA_ARGS__, 0); \512
+ break; \513
+ case 1: \514
+ func_name(__VA_ARGS__, 1); \515
+ break; \516
+ case 2: \517
+ func_name(__VA_ARGS__, 2); \518
+ break; \519
+ case 3: \520
+ func_name(__VA_ARGS__, 3); \521
+ break; \522
+ default: \523
+ default_case; \524
+ break; \525
+ } \526
+ } while (0)527
+528
+#define SIMDE_CONSTIFY_8_NO_RESULT_(func_name, default_case, imm, ...) \529
+ do { \530
+ switch (imm) { \531
+ case 0: \532
+ func_name(__VA_ARGS__, 0); \533
+ break; \534
+ case 1: \535
+ func_name(__VA_ARGS__, 1); \536
+ break; \537
+ case 2: \538
+ func_name(__VA_ARGS__, 2); \539
+ break; \540
+ case 3: \541
+ func_name(__VA_ARGS__, 3); \542
+ break; \543
+ case 4: \544
+ func_name(__VA_ARGS__, 4); \545
+ break; \546
+ case 5: \547
+ func_name(__VA_ARGS__, 5); \548
+ break; \549
+ case 6: \550
+ func_name(__VA_ARGS__, 6); \551
+ break; \552
+ case 7: \553
+ func_name(__VA_ARGS__, 7); \554
+ break; \555
+ default: \556
+ default_case; \557
+ break; \558
+ } \559
+ } while (0)560
+561
+#define SIMDE_CONSTIFY_16_NO_RESULT_(func_name, default_case, imm, ...) \562
+ do { \563
+ switch (imm) { \564
+ case 0: \565
+ func_name(__VA_ARGS__, 0); \566
+ break; \567
+ case 1: \568
+ func_name(__VA_ARGS__, 1); \569
+ break; \570
+ case 2: \571
+ func_name(__VA_ARGS__, 2); \572
+ break; \573
+ case 3: \574
+ func_name(__VA_ARGS__, 3); \575
+ break; \576
+ case 4: \577
+ func_name(__VA_ARGS__, 4); \578
+ break; \579
+ case 5: \580
+ func_name(__VA_ARGS__, 5); \581
+ break; \582
+ case 6: \583
+ func_name(__VA_ARGS__, 6); \584
+ break; \585
+ case 7: \586
+ func_name(__VA_ARGS__, 7); \587
+ break; \588
+ case 8: \589
+ func_name(__VA_ARGS__, 8); \590
+ break; \591
+ case 9: \592
+ func_name(__VA_ARGS__, 9); \593
+ break; \594
+ case 10: \595
+ func_name(__VA_ARGS__, 10); \596
+ break; \597
+ case 11: \598
+ func_name(__VA_ARGS__, 11); \599
+ break; \600
+ case 12: \601
+ func_name(__VA_ARGS__, 12); \602
+ break; \603
+ case 13: \604
+ func_name(__VA_ARGS__, 13); \605
+ break; \606
+ case 14: \607
+ func_name(__VA_ARGS__, 14); \608
+ break; \609
+ case 15: \610
+ func_name(__VA_ARGS__, 15); \611
+ break; \612
+ default: \613
+ default_case; \614
+ break; \615
+ } \616
+ } while (0)617
+618
+#define SIMDE_CONSTIFY_32_NO_RESULT_(func_name, default_case, imm, ...) \619
+ do { \620
+ switch (imm) { \621
+ case 0: \622
+ func_name(__VA_ARGS__, 0); \623
+ break; \624
+ case 1: \625
+ func_name(__VA_ARGS__, 1); \626
+ break; \627
+ case 2: \628
+ func_name(__VA_ARGS__, 2); \629
+ break; \630
+ case 3: \631
+ func_name(__VA_ARGS__, 3); \632
+ break; \633
+ case 4: \634
+ func_name(__VA_ARGS__, 4); \635
+ break; \636
+ case 5: \637
+ func_name(__VA_ARGS__, 5); \638
+ break; \639
+ case 6: \640
+ func_name(__VA_ARGS__, 6); \641
+ break; \642
+ case 7: \643
+ func_name(__VA_ARGS__, 7); \644
+ break; \645
+ case 8: \646
+ func_name(__VA_ARGS__, 8); \647
+ break; \648
+ case 9: \649
+ func_name(__VA_ARGS__, 9); \650
+ break; \651
+ case 10: \652
+ func_name(__VA_ARGS__, 10); \653
+ break; \654
+ case 11: \655
+ func_name(__VA_ARGS__, 11); \656
+ break; \657
+ case 12: \658
+ func_name(__VA_ARGS__, 12); \659
+ break; \660
+ case 13: \661
+ func_name(__VA_ARGS__, 13); \662
+ break; \663
+ case 14: \664
+ func_name(__VA_ARGS__, 14); \665
+ break; \666
+ case 15: \667
+ func_name(__VA_ARGS__, 15); \668
+ break; \669
+ case 16: \670
+ func_name(__VA_ARGS__, 16); \671
+ break; \672
+ case 17: \673
+ func_name(__VA_ARGS__, 17); \674
+ break; \675
+ case 18: \676
+ func_name(__VA_ARGS__, 18); \677
+ break; \678
+ case 19: \679
+ func_name(__VA_ARGS__, 19); \680
+ break; \681
+ case 20: \682
+ func_name(__VA_ARGS__, 20); \683
+ break; \684
+ case 21: \685
+ func_name(__VA_ARGS__, 21); \686
+ break; \687
+ case 22: \688
+ func_name(__VA_ARGS__, 22); \689
+ break; \690
+ case 23: \691
+ func_name(__VA_ARGS__, 23); \692
+ break; \693
+ case 24: \694
+ func_name(__VA_ARGS__, 24); \695
+ break; \696
+ case 25: \697
+ func_name(__VA_ARGS__, 25); \698
+ break; \699
+ case 26: \700
+ func_name(__VA_ARGS__, 26); \701
+ break; \702
+ case 27: \703
+ func_name(__VA_ARGS__, 27); \704
+ break; \705
+ case 28: \706
+ func_name(__VA_ARGS__, 28); \707
+ break; \708
+ case 29: \709
+ func_name(__VA_ARGS__, 29); \710
+ break; \711
+ case 30: \712
+ func_name(__VA_ARGS__, 30); \713
+ break; \714
+ case 31: \715
+ func_name(__VA_ARGS__, 31); \716
+ break; \717
+ default: \718
+ default_case; \719
+ break; \720
+ } \721
+ } while (0)722
+723
+#define SIMDE_CONSTIFY_64_NO_RESULT_(func_name, default_case, imm, ...) \724
+ do { \725
+ switch (imm) { \726
+ case 0: \727
+ func_name(__VA_ARGS__, 0); \728
+ break; \729
+ case 1: \730
+ func_name(__VA_ARGS__, 1); \731
+ break; \732
+ case 2: \733
+ func_name(__VA_ARGS__, 2); \734
+ break; \735
+ case 3: \736
+ func_name(__VA_ARGS__, 3); \737
+ break; \738
+ case 4: \739
+ func_name(__VA_ARGS__, 4); \740
+ break; \741
+ case 5: \742
+ func_name(__VA_ARGS__, 5); \743
+ break; \744
+ case 6: \745
+ func_name(__VA_ARGS__, 6); \746
+ break; \747
+ case 7: \748
+ func_name(__VA_ARGS__, 7); \749
+ break; \750
+ case 8: \751
+ func_name(__VA_ARGS__, 8); \752
+ break; \753
+ case 9: \754
+ func_name(__VA_ARGS__, 9); \755
+ break; \756
+ case 10: \757
+ func_name(__VA_ARGS__, 10); \758
+ break; \759
+ case 11: \760
+ func_name(__VA_ARGS__, 11); \761
+ break; \762
+ case 12: \763
+ func_name(__VA_ARGS__, 12); \764
+ break; \765
+ case 13: \766
+ func_name(__VA_ARGS__, 13); \767
+ break; \768
+ case 14: \769
+ func_name(__VA_ARGS__, 14); \770
+ break; \771
+ case 15: \772
+ func_name(__VA_ARGS__, 15); \773
+ break; \774
+ case 16: \775
+ func_name(__VA_ARGS__, 16); \776
+ break; \777
+ case 17: \778
+ func_name(__VA_ARGS__, 17); \779
+ break; \780
+ case 18: \781
+ func_name(__VA_ARGS__, 18); \782
+ break; \783
+ case 19: \784
+ func_name(__VA_ARGS__, 19); \785
+ break; \786
+ case 20: \787
+ func_name(__VA_ARGS__, 20); \788
+ break; \789
+ case 21: \790
+ func_name(__VA_ARGS__, 21); \791
+ break; \792
+ case 22: \793
+ func_name(__VA_ARGS__, 22); \794
+ break; \795
+ case 23: \796
+ func_name(__VA_ARGS__, 23); \797
+ break; \798
+ case 24: \799
+ func_name(__VA_ARGS__, 24); \800
+ break; \801
+ case 25: \802
+ func_name(__VA_ARGS__, 25); \803
+ break; \804
+ case 26: \805
+ func_name(__VA_ARGS__, 26); \806
+ break; \807
+ case 27: \808
+ func_name(__VA_ARGS__, 27); \809
+ break; \810
+ case 28: \811
+ func_name(__VA_ARGS__, 28); \812
+ break; \813
+ case 29: \814
+ func_name(__VA_ARGS__, 29); \815
+ break; \816
+ case 30: \817
+ func_name(__VA_ARGS__, 30); \818
+ break; \819
+ case 31: \820
+ func_name(__VA_ARGS__, 31); \821
+ break; \822
+ case 32: \823
+ func_name(__VA_ARGS__, 32); \824
+ break; \825
+ case 33: \826
+ func_name(__VA_ARGS__, 33); \827
+ break; \828
+ case 34: \829
+ func_name(__VA_ARGS__, 34); \830
+ break; \831
+ case 35: \832
+ func_name(__VA_ARGS__, 35); \833
+ break; \834
+ case 36: \835
+ func_name(__VA_ARGS__, 36); \836
+ break; \837
+ case 37: \838
+ func_name(__VA_ARGS__, 37); \839
+ break; \840
+ case 38: \841
+ func_name(__VA_ARGS__, 38); \842
+ break; \843
+ case 39: \844
+ func_name(__VA_ARGS__, 39); \845
+ break; \846
+ case 40: \847
+ func_name(__VA_ARGS__, 40); \848
+ break; \849
+ case 41: \850
+ func_name(__VA_ARGS__, 41); \851
+ break; \852
+ case 42: \853
+ func_name(__VA_ARGS__, 42); \854
+ break; \855
+ case 43: \856
+ func_name(__VA_ARGS__, 43); \857
+ break; \858
+ case 44: \859
+ func_name(__VA_ARGS__, 44); \860
+ break; \861
+ case 45: \862
+ func_name(__VA_ARGS__, 45); \863
+ break; \864
+ case 46: \865
+ func_name(__VA_ARGS__, 46); \866
+ break; \867
+ case 47: \868
+ func_name(__VA_ARGS__, 47); \869
+ break; \870
+ case 48: \871
+ func_name(__VA_ARGS__, 48); \872
+ break; \873
+ case 49: \874
+ func_name(__VA_ARGS__, 49); \875
+ break; \876
+ case 50: \877
+ func_name(__VA_ARGS__, 50); \878
+ break; \879
+ case 51: \880
+ func_name(__VA_ARGS__, 51); \881
+ break; \882
+ case 52: \883
+ func_name(__VA_ARGS__, 52); \884
+ break; \885
+ case 53: \886
+ func_name(__VA_ARGS__, 53); \887
+ break; \888
+ case 54: \889
+ func_name(__VA_ARGS__, 54); \890
+ break; \891
+ case 55: \892
+ func_name(__VA_ARGS__, 55); \893
+ break; \894
+ case 56: \895
+ func_name(__VA_ARGS__, 56); \896
+ break; \897
+ case 57: \898
+ func_name(__VA_ARGS__, 57); \899
+ break; \900
+ case 58: \901
+ func_name(__VA_ARGS__, 58); \902
+ break; \903
+ case 59: \904
+ func_name(__VA_ARGS__, 59); \905
+ break; \906
+ case 60: \907
+ func_name(__VA_ARGS__, 60); \908
+ break; \909
+ case 61: \910
+ func_name(__VA_ARGS__, 61); \911
+ break; \912
+ case 62: \913
+ func_name(__VA_ARGS__, 62); \914
+ break; \915
+ case 63: \916
+ func_name(__VA_ARGS__, 63); \917
+ break; \918
+ default: \919
+ default_case; \920
+ break; \921
+ } \922
+ } while (0)923
+924
+HEDLEY_DIAGNOSTIC_POP925
+926
+#endif927
obs-studio-26.1.1.tar.xz/libobs/util/simde/simde-detect-clang.h
Added
116
1
2
+/* Detect Clang Version3
+ * Created by Evan Nemerson <evan@nemerson.com>4
+ *5
+ * To the extent possible under law, the author(s) have dedicated all6
+ * copyright and related and neighboring rights to this software to7
+ * the public domain worldwide. This software is distributed without8
+ * any warranty.9
+ *10
+ * For details, see <http://creativecommons.org/publicdomain/zero/1.0/>.11
+ * SPDX-License-Identifier: CC0-1.012
+ */13
+14
+/* This file was originally part of SIMDe15
+ * (<https://github.com/simd-everywhere/simde>). You're free to do with it as16
+ * you please, but I do have a few small requests:17
+ *18
+ * * If you make improvements, please submit them back to SIMDe19
+ * (at <https://github.com/simd-everywhere/simde/issues>) so others can20
+ * benefit from them.21
+ * * Please keep a link to SIMDe intact so people know where to submit22
+ * improvements.23
+ * * If you expose it publicly, please change the SIMDE_ prefix to24
+ * something specific to your project.25
+ *26
+ * The version numbers clang exposes (in the ___clang_major__,27
+ * __clang_minor__, and __clang_patchlevel__ macros) are unreliable.28
+ * Vendors such as Apple will define these values to their version29
+ * numbers; for example, "Apple Clang 4.0" is really clang 3.1, but30
+ * __clang_major__ and __clang_minor__ are defined to 4 and 031
+ * respectively, instead of 3 and 1.32
+ *33
+ * The solution is *usually* to use clang's feature detection macros34
+ * (<https://clang.llvm.org/docs/LanguageExtensions.html#feature-checking-macros>)35
+ * to determine if the feature you're interested in is available. This36
+ * generally works well, and it should probably be the first thing you37
+ * try. Unfortunately, it's not possible to check for everything. In38
+ * particular, compiler bugs.39
+ *40
+ * This file just uses the feature checking macros to detect features41
+ * added in specific versions of clang to identify which version of42
+ * clang the compiler is based on.43
+ *44
+ * Right now it only goes back to 3.6, but I'm happy to accept patches45
+ * to go back further. And, of course, newer versions are welcome if46
+ * they're not already present, and if you find a way to detect a point47
+ * release that would be great, too!48
+ */49
+50
+#if !defined(SIMDE_DETECT_CLANG_H)51
+#define SIMDE_DETECT_CLANG_H 152
+53
+/* Attempt to detect the upstream clang version number. I usually only54
+ * worry about major version numbers (at least for 4.0+), but if you55
+ * need more resolution I'm happy to accept patches that are able to56
+ * detect minor versions as well. That said, you'll probably have a57
+ * hard time with detection since AFAIK most minor releases don't add58
+ * anything we can detect. */59
+60
+#if defined(__clang__) && !defined(SIMDE_DETECT_CLANG_VERSION)61
+#if __has_warning("-Wformat-insufficient-args")62
+#define SIMDE_DETECT_CLANG_VERSION 12000063
+#elif __has_warning("-Wimplicit-const-int-float-conversion")64
+#define SIMDE_DETECT_CLANG_VERSION 11000065
+#elif __has_warning("-Wmisleading-indentation")66
+#define SIMDE_DETECT_CLANG_VERSION 10000067
+#elif defined(__FILE_NAME__)68
+#define SIMDE_DETECT_CLANG_VERSION 9000069
+#elif __has_warning("-Wextra-semi-stmt") || \70
+ __has_builtin(__builtin_rotateleft32)71
+#define SIMDE_DETECT_CLANG_VERSION 8000072
+#elif __has_warning("-Wc++98-compat-extra-semi")73
+#define SIMDE_DETECT_CLANG_VERSION 7000074
+#elif __has_warning("-Wpragma-pack")75
+#define SIMDE_DETECT_CLANG_VERSION 6000076
+#elif __has_warning("-Wbitfield-enum-conversion")77
+#define SIMDE_DETECT_CLANG_VERSION 5000078
+#elif __has_attribute(diagnose_if)79
+#define SIMDE_DETECT_CLANG_VERSION 4000080
+#elif __has_warning("-Wcast-calling-convention")81
+#define SIMDE_DETECT_CLANG_VERSION 3090082
+#elif __has_warning("-WCL4")83
+#define SIMDE_DETECT_CLANG_VERSION 3080084
+#elif __has_warning("-WIndependentClass-attribute")85
+#define SIMDE_DETECT_CLANG_VERSION 3070086
+#elif __has_warning("-Wambiguous-ellipsis")87
+#define SIMDE_DETECT_CLANG_VERSION 3060088
+#else89
+#define SIMDE_DETECT_CLANG_VERSION 190
+#endif91
+#endif /* defined(__clang__) && !defined(SIMDE_DETECT_CLANG_VERSION) */92
+93
+/* The SIMDE_DETECT_CLANG_VERSION_CHECK macro is pretty94
+ * straightforward; it returns true if the compiler is a derivative95
+ * of clang >= the specified version.96
+ *97
+ * Since this file is often (primarily?) useful for working around bugs98
+ * it is also helpful to have a macro which returns true if only if the99
+ * compiler is a version of clang *older* than the specified version to100
+ * make it a bit easier to ifdef regions to add code for older versions,101
+ * such as pragmas to disable a specific warning. */102
+103
+#if defined(SIMDE_DETECT_CLANG_VERSION)104
+#define SIMDE_DETECT_CLANG_VERSION_CHECK(major, minor, revision) \105
+ (SIMDE_DETECT_CLANG_VERSION >= \106
+ ((major * 10000) + (minor * 1000) + (revision)))107
+#define SIMDE_DETECT_CLANG_VERSION_NOT(major, minor, revision) \108
+ (SIMDE_DETECT_CLANG_VERSION < \109
+ ((major * 10000) + (minor * 1000) + (revision)))110
+#else111
+#define SIMDE_DETECT_CLANG_VERSION_CHECK(major, minor, revision) (0)112
+#define SIMDE_DETECT_CLANG_VERSION_NOT(major, minor, revision) (1)113
+#endif114
+115
+#endif /* !defined(SIMDE_DETECT_CLANG_H) */116
obs-studio-26.1.0.tar.xz/libobs/util/simde/simde-diagnostic.h -> obs-studio-26.1.1.tar.xz/libobs/util/simde/simde-diagnostic.h
Changed
262
1
2
*/3
4
#if !defined(SIMDE_DIAGNOSTIC_H)5
+#define SIMDE_DIAGNOSTIC_H6
7
#include "hedley.h"8
+#include "simde-detect-clang.h"9
10
/* This is only to help us implement functions like _mm_undefined_ps. */11
#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)12
13
#define SIMDE_DIAGNOSTIC_DISABLE_SIMD_PRAGMA_DEPRECATED_14
#endif15
16
+/* MSVC emits a diagnostic when we call a function (like17
+ * simde_mm_set_epi32) while initializing a struct. We currently do18
+ * this a *lot* in the tests. */19
#if defined(HEDLEY_MSVC_VERSION)20
#define SIMDE_DIAGNOSTIC_DISABLE_NON_CONSTANT_AGGREGATE_INITIALIZER_ \21
__pragma(warning(disable : 4204))22
23
#define SIMDE_DIAGNOSTIC_DISABLE_VARIADIC_MACROS_24
#endif25
26
+/* emscripten requires us to use a __wasm_unimplemented_simd128__ macro27
+ * before we can access certain SIMD intrinsics, but this diagnostic28
+ * warns about it being a reserved name. It is a reserved name, but29
+ * it's reserved for the compiler and we are using it to convey30
+ * information to the compiler.31
+ *32
+ * This is also used when enabling native aliases since we don't get to33
+ * choose the macro names. */34
+#if HEDLEY_HAS_WARNING("-Wdouble-promotion")35
+#define SIMDE_DIAGNOSTIC_DISABLE_RESERVED_ID_MACRO_ \36
+ _Pragma("clang diagnostic ignored \"-Wreserved-id-macro\"")37
+#else38
+#define SIMDE_DIAGNOSTIC_DISABLE_RESERVED_ID_MACRO_39
+#endif40
+41
+/* clang 3.8 warns about the packed attribute being unnecessary when42
+ * used in the _mm_loadu_* functions. That *may* be true for version43
+ * 3.8, but for later versions it is crucial in order to make unaligned44
+ * access safe. */45
+#if HEDLEY_HAS_WARNING("-Wpacked")46
+#define SIMDE_DIAGNOSTIC_DISABLE_PACKED_ \47
+ _Pragma("clang diagnostic ignored \"-Wpacked\"")48
+#else49
+#define SIMDE_DIAGNOSTIC_DISABLE_PACKED_50
+#endif51
+52
/* Triggered when assigning a float to a double implicitly. We use53
* explicit casts in SIMDe, this is only used in the test suite. */54
#if HEDLEY_HAS_WARNING("-Wdouble-promotion")55
56
57
/* Several compilers treat conformant array parameters as VLAs. We58
* test to make sure we're in C mode (C++ doesn't support CAPs), and59
- * that the version of the standard supports CAPs. We also blacklist60
+ * that the version of the standard supports CAPs. We also reject61
* some buggy compilers like MSVC (the logic is in Hedley if you want62
* to take a look), but with certain warnings enabled some compilers63
* still like to emit a diagnostic. */64
65
#elif HEDLEY_GCC_VERSION_CHECK(3, 4, 0)66
#define SIMDE_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION_ \67
_Pragma("GCC diagnostic ignored \"-Wunused-function\"")68
+#elif HEDLEY_MSVC_VERSION_CHECK(19, 0, 0) /* Likely goes back further */69
+#define SIMDE_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION_ \70
+ __pragma(warning(disable : 4505))71
#else72
#define SIMDE_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION_73
#endif74
75
#define SIMDE_DIAGNOSTIC_DISABLE_PASS_FAILED_76
#endif77
78
-/* https://github.com/nemequ/simde/issues/277 */79
+#if HEDLEY_HAS_WARNING("-Wpadded")80
+#define SIMDE_DIAGNOSTIC_DISABLE_PADDED_ \81
+ _Pragma("clang diagnostic ignored \"-Wpadded\"")82
+#elif HEDLEY_MSVC_VERSION_CHECK(19, 0, 0) /* Likely goes back further */83
+#define SIMDE_DIAGNOSTIC_DISABLE_PADDED_ __pragma(warning(disable : 4324))84
+#else85
+#define SIMDE_DIAGNOSTIC_DISABLE_PADDED_86
+#endif87
+88
+#if HEDLEY_HAS_WARNING("-Wzero-as-null-pointer-constant")89
+#define SIMDE_DIAGNOSTIC_DISABLE_ZERO_AS_NULL_POINTER_CONSTANT_ \90
+ _Pragma("clang diagnostic ignored \"-Wzero-as-null-pointer-constant\"")91
+#else92
+#define SIMDE_DIAGNOSTIC_DISABLE_ZERO_AS_NULL_POINTER_CONSTANT_93
+#endif94
+95
+#if HEDLEY_HAS_WARNING("-Wold-style-cast")96
+#define SIMDE_DIAGNOSTIC_DISABLE_OLD_STYLE_CAST_ \97
+ _Pragma("clang diagnostic ignored \"-Wold-style-cast\"")98
+#else99
+#define SIMDE_DIAGNOSTIC_DISABLE_OLD_STYLE_CAST_100
+#endif101
+102
+#if HEDLEY_HAS_WARNING("-Wcast-function-type") || \103
+ HEDLEY_GCC_VERSION_CHECK(8, 0, 0)104
+#define SIMDE_DIAGNOSTIC_DISABLE_CAST_FUNCTION_TYPE_ \105
+ _Pragma("GCC diagnostic ignored \"-Wcast-function-type\"")106
+#else107
+#define SIMDE_DIAGNOSTIC_DISABLE_CAST_FUNCTION_TYPE_108
+#endif109
+110
+/* clang will emit this warning when we use C99 extensions whan not in111
+ * C99 mode, even though it does support this. In such cases we check112
+ * the compiler and version first, so we know it's not a problem. */113
+#if HEDLEY_HAS_WARNING("-Wc99-extensions")114
+#define SIMDE_DIAGNOSTIC_DISABLE_C99_EXTENSIONS_ \115
+ _Pragma("clang diagnostic ignored \"-Wc99-extensions\"")116
+#else117
+#define SIMDE_DIAGNOSTIC_DISABLE_C99_EXTENSIONS_118
+#endif119
+120
+/* https://github.com/simd-everywhere/simde/issues/277 */121
#if defined(HEDLEY_GCC_VERSION) && HEDLEY_GCC_VERSION_CHECK(4, 6, 0) && \122
- !HEDLEY_GCC_VERSION_CHECK(6, 0, 0) && defined(__cplusplus)123
-#define SIMDE_DIAGNOSTIC_DISABLE_BUGGY_UNUSED_BUT_SET_VARIBALE \124
+ !HEDLEY_GCC_VERSION_CHECK(6, 4, 0) && defined(__cplusplus)125
+#define SIMDE_DIAGNOSTIC_DISABLE_BUGGY_UNUSED_BUT_SET_VARIBALE_ \126
_Pragma("GCC diagnostic ignored \"-Wunused-but-set-variable\"")127
#else128
-#define SIMDE_DIAGNOSTIC_DISABLE_BUGGY_UNUSED_BUT_SET_VARIBALE129
+#define SIMDE_DIAGNOSTIC_DISABLE_BUGGY_UNUSED_BUT_SET_VARIBALE_130
+#endif131
+132
+/* This is the warning that you normally define _CRT_SECURE_NO_WARNINGS133
+ * to silence, but you have to do that before including anything and134
+ * that would require reordering includes. */135
+#if defined(_MSC_VER)136
+#define SIMDE_DIAGNOSTIC_DISABLE_ANNEX_K_ __pragma(warning(disable : 4996))137
+#else138
+#define SIMDE_DIAGNOSTIC_DISABLE_ANNEX_K_139
#endif140
141
/* Some compilers, such as clang, may use `long long` for 64-bit142
143
* -Wc++98-compat-pedantic which says 'long long' is incompatible with144
* C++98. */145
#if HEDLEY_HAS_WARNING("-Wc++98-compat-pedantic")146
-#define SIMDE_DIAGNOSTIC_DISABLE_CPP98_COMPAT_PEDANTIC \147
+#define SIMDE_DIAGNOSTIC_DISABLE_CPP98_COMPAT_PEDANTIC_ \148
_Pragma("clang diagnostic ignored \"-Wc++98-compat-pedantic\"")149
#else150
-#define SIMDE_DIAGNOSTIC_DISABLE_CPP98_COMPAT_PEDANTIC151
+#define SIMDE_DIAGNOSTIC_DISABLE_CPP98_COMPAT_PEDANTIC_152
+#endif153
+154
+/* Some problem as above */155
+#if HEDLEY_HAS_WARNING("-Wc++11-long-long")156
+#define SIMDE_DIAGNOSTIC_DISABLE_CPP11_LONG_LONG_ \157
+ _Pragma("clang diagnostic ignored \"-Wc++11-long-long\"")158
+#else159
+#define SIMDE_DIAGNOSTIC_DISABLE_CPP11_LONG_LONG_160
+#endif161
+162
+/* emscripten emits this whenever stdin/stdout/stderr is used in a163
+ * macro. */164
+#if HEDLEY_HAS_WARNING("-Wdisabled-macro-expansion")165
+#define SIMDE_DIAGNOSTIC_DISABLE_DISABLED_MACRO_EXPANSION_ \166
+ _Pragma("clang diagnostic ignored \"-Wdisabled-macro-expansion\"")167
+#else168
+#define SIMDE_DIAGNOSTIC_DISABLE_DISABLED_MACRO_EXPANSION_169
+#endif170
+171
+/* Clang uses C11 generic selections to implement some AltiVec172
+ * functions, which triggers this diagnostic when not compiling173
+ * in C11 mode */174
+#if HEDLEY_HAS_WARNING("-Wc11-extensions")175
+#define SIMDE_DIAGNOSTIC_DISABLE_C11_EXTENSIONS_ \176
+ _Pragma("clang diagnostic ignored \"-Wc11-extensions\"")177
+#else178
+#define SIMDE_DIAGNOSTIC_DISABLE_C11_EXTENSIONS_179
+#endif180
+181
+/* Clang sometimes triggers this warning in macros in the AltiVec and182
+ * NEON headers, or due to missing functions. */183
+#if HEDLEY_HAS_WARNING("-Wvector-conversion")184
+#define SIMDE_DIAGNOSTIC_DISABLE_VECTOR_CONVERSION_ \185
+ _Pragma("clang diagnostic ignored \"-Wvector-conversion\"")186
+/* For NEON, the situation with -Wvector-conversion in clang < 10 is187
+ * bad enough that we just disable the warning altogether. */188
+#if defined(SIMDE_ARCH_ARM) && SIMDE_DETECT_CLANG_VERSION_NOT(10, 0, 0)189
+#define SIMDE_DIAGNOSTIC_DISABLE_BUGGY_VECTOR_CONVERSION_ \190
+ SIMDE_DIAGNOSTIC_DISABLE_VECTOR_CONVERSION_191
+#endif192
+#else193
+#define SIMDE_DIAGNOSTIC_DISABLE_VECTOR_CONVERSION_194
+#endif195
+#if !defined(SIMDE_DIAGNOSTIC_DISABLE_BUGGY_VECTOR_CONVERSION_)196
+#define SIMDE_DIAGNOSTIC_DISABLE_BUGGY_VECTOR_CONVERSION_197
+#endif198
+199
+/* SLEEF triggers this a *lot* in their headers */200
+#if HEDLEY_HAS_WARNING("-Wignored-qualifiers")201
+#define SIMDE_DIAGNOSTIC_DISABLE_IGNORED_QUALIFIERS_ \202
+ _Pragma("clang diagnostic ignored \"-Wignored-qualifiers\"")203
+#elif HEDLEY_GCC_VERSION_CHECK(4, 3, 0)204
+#define SIMDE_DIAGNOSTIC_DISABLE_IGNORED_QUALIFIERS_ \205
+ _Pragma("GCC diagnostic ignored \"-Wignored-qualifiers\"")206
+#else207
+#define SIMDE_DIAGNOSTIC_DISABLE_IGNORED_QUALIFIERS_208
+#endif209
+210
+/* GCC emits this under some circumstances when using __int128 */211
+#if HEDLEY_GCC_VERSION_CHECK(4, 8, 0)212
+#define SIMDE_DIAGNOSTIC_DISABLE_PEDANTIC_ \213
+ _Pragma("GCC diagnostic ignored \"-Wpedantic\"")214
+#else215
+#define SIMDE_DIAGNOSTIC_DISABLE_PEDANTIC_216
+#endif217
+218
+/* MSVC doesn't like (__assume(0), code) and will warn about code being219
+ * unreachable, but we want it there because not all compilers220
+ * understand the unreachable macro and will complain if it is missing.221
+ * I'm planning on adding a new macro to Hedley to handle this a bit222
+ * more elegantly, but until then... */223
+#if defined(HEDLEY_MSVC_VERSION)224
+#define SIMDE_DIAGNOSTIC_DISABLE_UNREACHABLE_ __pragma(warning(disable : 4702))225
+#else226
+#define SIMDE_DIAGNOSTIC_DISABLE_UNREACHABLE_227
+#endif228
+229
+/* This is a false positive from GCC in a few places. */230
+#if HEDLEY_GCC_VERSION_CHECK(4, 7, 0)231
+#define SIMDE_DIAGNOSTIC_DISABLE_MAYBE_UNINITIAZILED_ \232
+ _Pragma("GCC diagnostic ignored \"-Wmaybe-uninitialized\"")233
+#else234
+#define SIMDE_DIAGNOSTIC_DISABLE_MAYBE_UNINITIAZILED_235
+#endif236
+237
+#if defined(SIMDE_ENABLE_NATIVE_ALIASES)238
+#define SIMDE_DISABLE_UNWANTED_DIAGNOSTICS_NATIVE_ALIASES_ \239
+ SIMDE_DIAGNOSTIC_DISABLE_RESERVED_ID_MACRO_240
+#else241
+#define SIMDE_DISABLE_UNWANTED_DIAGNOSTICS_NATIVE_ALIASES_242
#endif243
244
#define SIMDE_DISABLE_UNWANTED_DIAGNOSTICS \245
+ SIMDE_DISABLE_UNWANTED_DIAGNOSTICS_NATIVE_ALIASES_ \246
SIMDE_DIAGNOSTIC_DISABLE_PSABI_ \247
SIMDE_DIAGNOSTIC_DISABLE_NO_EMMS_INSTRUCTION_ \248
SIMDE_DIAGNOSTIC_DISABLE_SIMD_PRAGMA_DEPRECATED_ \249
250
SIMDE_DIAGNOSTIC_DISABLE_USED_BUT_MARKED_UNUSED_ \251
SIMDE_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION_ \252
SIMDE_DIAGNOSTIC_DISABLE_PASS_FAILED_ \253
- SIMDE_DIAGNOSTIC_DISABLE_CPP98_COMPAT_PEDANTIC \254
- SIMDE_DIAGNOSTIC_DISABLE_BUGGY_UNUSED_BUT_SET_VARIBALE255
+ SIMDE_DIAGNOSTIC_DISABLE_CPP98_COMPAT_PEDANTIC_ \256
+ SIMDE_DIAGNOSTIC_DISABLE_CPP11_LONG_LONG_ \257
+ SIMDE_DIAGNOSTIC_DISABLE_BUGGY_UNUSED_BUT_SET_VARIBALE_ \258
+ SIMDE_DIAGNOSTIC_DISABLE_BUGGY_VECTOR_CONVERSION_259
260
-#endif261
+#endif /* !defined(SIMDE_DIAGNOSTIC_H) */262
obs-studio-26.1.0.tar.xz/libobs/util/simde/simde-features.h -> obs-studio-26.1.1.tar.xz/libobs/util/simde/simde-features.h
Changed
292
1
2
#define SIMDE_FEATURES_H3
4
#include "simde-arch.h"5
+#include "simde-diagnostic.h"6
7
#if !defined(SIMDE_X86_SVML_NATIVE) && !defined(SIMDE_X86_SVML_NO_NATIVE) && \8
!defined(SIMDE_NO_NATIVE)9
10
#define SIMDE_X86_AVX512F_NATIVE11
#endif12
13
+#if !defined(SIMDE_X86_AVX512VP2INTERSECT_NATIVE) && \14
+ !defined(SIMDE_X86_AVX512VP2INTERSECT_NO_NATIVE) && \15
+ !defined(SIMDE_NO_NATIVE)16
+#if defined(SIMDE_ARCH_X86_AVX512VP2INTERSECT)17
+#define SIMDE_X86_AVX512VP2INTERSECT_NATIVE18
+#endif19
+#endif20
+#if defined(SIMDE_X86_AVX512VP2INTERSECT_NATIVE) && \21
+ !defined(SIMDE_X86_AVX512F_NATIVE)22
+#define SIMDE_X86_AVX512F_NATIVE23
+#endif24
+25
+#if !defined(SIMDE_X86_AVX512VBMI_NATIVE) && \26
+ !defined(SIMDE_X86_AVX512VBMI_NO_NATIVE) && !defined(SIMDE_NO_NATIVE)27
+#if defined(SIMDE_ARCH_X86_AVX512VBMI)28
+#define SIMDE_X86_AVX512VBMI_NATIVE29
+#endif30
+#endif31
+#if defined(SIMDE_X86_AVX512VBMI_NATIVE) && !defined(SIMDE_X86_AVX512F_NATIVE)32
+#define SIMDE_X86_AVX512F_NATIVE33
+#endif34
+35
#if !defined(SIMDE_X86_AVX512CD_NATIVE) && \36
!defined(SIMDE_X86_AVX512CD_NO_NATIVE) && !defined(SIMDE_NO_NATIVE)37
#if defined(SIMDE_ARCH_X86_AVX512CD)38
39
#endif40
#endif41
42
+#if !defined(SIMDE_X86_PCLMUL_NATIVE) && \43
+ !defined(SIMDE_X86_PCLMUL_NO_NATIVE) && !defined(SIMDE_NO_NATIVE)44
+#if defined(SIMDE_ARCH_X86_PCLMUL)45
+#define SIMDE_X86_PCLMUL_NATIVE46
+#endif47
+#endif48
+49
+#if !defined(SIMDE_X86_VPCLMULQDQ_NATIVE) && \50
+ !defined(SIMDE_X86_VPCLMULQDQ_NO_NATIVE) && !defined(SIMDE_NO_NATIVE)51
+#if defined(SIMDE_ARCH_X86_VPCLMULQDQ)52
+#define SIMDE_X86_VPCLMULQDQ_NATIVE53
+#endif54
+#endif55
+56
#if !defined(SIMDE_X86_SVML_NATIVE) && !defined(SIMDE_X86_SVML_NO_NATIVE) && \57
!defined(SIMDE_NO_NATIVE)58
#if defined(__INTEL_COMPILER)59
60
#pragma warning(disable : 4799)61
#endif62
63
-#if defined(SIMDE_X86_AVX_NATIVE) || defined(SIMDE_X86_GFNI_NATIVE) || \64
- defined(SIMDE_X86_SVML_NATIVE)65
+#if defined(SIMDE_X86_AVX_NATIVE) || defined(SIMDE_X86_GFNI_NATIVE)66
#include <immintrin.h>67
#elif defined(SIMDE_X86_SSE4_2_NATIVE)68
#include <nmmintrin.h>69
70
71
#if !defined(SIMDE_ARM_NEON_A32V8_NATIVE) && \72
!defined(SIMDE_ARM_NEON_A32V8_NO_NATIVE) && !defined(SIMDE_NO_NATIVE)73
-#if defined(SIMDE_ARCH_ARM_NEON) && SIMDE_ARCH_ARM_CHECK(80)74
+#if defined(SIMDE_ARCH_ARM_NEON) && SIMDE_ARCH_ARM_CHECK(80) && \75
+ (__ARM_NEON_FP & 0x02)76
#define SIMDE_ARM_NEON_A32V8_NATIVE77
#endif78
#endif79
80
#include <arm_neon.h>81
#endif82
83
+#if !defined(SIMDE_ARM_SVE_NATIVE) && !defined(SIMDE_ARM_SVE_NO_NATIVE) && \84
+ !defined(SIMDE_NO_NATIVE)85
+#if defined(SIMDE_ARCH_ARM_SVE)86
+#define SIMDE_ARM_SVE_NATIVE87
+#include <arm_sve.h>88
+#endif89
+#endif90
+91
#if !defined(SIMDE_WASM_SIMD128_NATIVE) && \92
!defined(SIMDE_WASM_SIMD128_NO_NATIVE) && !defined(SIMDE_NO_NATIVE)93
#if defined(SIMDE_ARCH_WASM_SIMD128)94
95
#endif96
#if defined(SIMDE_WASM_SIMD128_NATIVE)97
#if !defined(__wasm_unimplemented_simd128__)98
+HEDLEY_DIAGNOSTIC_PUSH99
+SIMDE_DIAGNOSTIC_DISABLE_RESERVED_ID_MACRO_100
#define __wasm_unimplemented_simd128__101
+HEDLEY_DIAGNOSTIC_POP102
#endif103
#include <wasm_simd128.h>104
#endif105
106
#define SIMDE_POWER_ALTIVEC_P5_NATIVE107
#endif108
#endif109
-#if defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)110
-/* stdbool.h conflicts with the bool in altivec.h */111
-#if defined(bool) && !defined(SIMDE_POWER_ALTIVEC_NO_UNDEF_BOOL_)112
+113
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)114
+/* AltiVec conflicts with lots of stuff. The bool keyword conflicts115
+ * with the bool keyword in C++ and the bool macro in C99+ (defined116
+ * in stdbool.h). The vector keyword conflicts with std::vector in117
+ * C++ if you are `using std;`.118
+ *119
+ * Luckily AltiVec allows you to use `__vector`/`__bool`/`__pixel`120
+ * instead, but altivec.h will unconditionally define121
+ * `vector`/`bool`/`pixel` so we need to work around that.122
+ *123
+ * Unfortunately this means that if your code uses AltiVec directly124
+ * it may break. If this is the case you'll want to define125
+ * `SIMDE_POWER_ALTIVEC_NO_UNDEF` before including SIMDe. Or, even126
+ * better, port your code to use the double-underscore versions. */127
+#if defined(bool)128
#undef bool129
#endif130
+131
#include <altivec.h>132
-/* GCC allows you to undefine these macros to prevent conflicts with133
- * standard types as they become context-sensitive keywords. */134
-#if defined(__cplusplus)135
+136
+#if !defined(SIMDE_POWER_ALTIVEC_NO_UNDEF)137
#if defined(vector)138
#undef vector139
#endif140
141
#if defined(bool)142
#undef bool143
#endif144
-#define SIMDE_POWER_ALTIVEC_VECTOR(T) vector T145
-#define SIMDE_POWER_ALTIVEC_PIXEL pixel146
-#define SIMDE_POWER_ALTIVEC_BOOL bool147
-#else148
+#endif /* !defined(SIMDE_POWER_ALTIVEC_NO_UNDEF) */149
+150
+/* Use these intsead of vector/pixel/bool in SIMDe. */151
#define SIMDE_POWER_ALTIVEC_VECTOR(T) __vector T152
#define SIMDE_POWER_ALTIVEC_PIXEL __pixel153
#define SIMDE_POWER_ALTIVEC_BOOL __bool154
-#endif /* defined(__cplusplus) */155
+156
+/* Re-define bool if we're using stdbool.h */157
+#if !defined(__cplusplus) && defined(__bool_true_false_are_defined) && \158
+ !defined(SIMDE_POWER_ALTIVEC_NO_UNDEF)159
+#define bool _Bool160
+#endif161
+#endif162
+163
+#if !defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE) && \164
+ !defined(SIMDE_MIPS_LOONGSON_MMI_NO_NATIVE) && \165
+ !defined(SIMDE_NO_NATIVE)166
+#if defined(SIMDE_ARCH_MIPS_LOONGSON_MMI)167
+#define SIMDE_MIPS_LOONGSON_MMI_NATIVE 1168
+#endif169
+#endif170
+#if defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)171
+#include <loongson-mmiintrin.h>172
+#endif173
+174
+/* This is used to determine whether or not to fall back on a vector175
+ * function in an earlier ISA extensions, as well as whether176
+ * we expected any attempts at vectorization to be fruitful or if we177
+ * expect to always be running serial code. */178
+179
+#if !defined(SIMDE_NATURAL_VECTOR_SIZE)180
+#if defined(SIMDE_X86_AVX512F_NATIVE)181
+#define SIMDE_NATURAL_VECTOR_SIZE (512)182
+#elif defined(SIMDE_X86_AVX_NATIVE)183
+#define SIMDE_NATURAL_VECTOR_SIZE (256)184
+#elif defined(SIMDE_X86_SSE_NATIVE) || defined(SIMDE_ARM_NEON_A32V7_NATIVE) || \185
+ defined(SIMDE_WASM_SIMD128_NATIVE) || \186
+ defined(SIMDE_POWER_ALTIVEC_P5_NATIVE)187
+#define SIMDE_NATURAL_VECTOR_SIZE (128)188
+#endif189
+190
+#if !defined(SIMDE_NATURAL_VECTOR_SIZE)191
+#define SIMDE_NATURAL_VECTOR_SIZE (0)192
+#endif193
+#endif194
+195
+#define SIMDE_NATURAL_VECTOR_SIZE_LE(x) \196
+ ((SIMDE_NATURAL_VECTOR_SIZE > 0) && (SIMDE_NATURAL_VECTOR_SIZE <= (x)))197
+#define SIMDE_NATURAL_VECTOR_SIZE_GE(x) \198
+ ((SIMDE_NATURAL_VECTOR_SIZE > 0) && (SIMDE_NATURAL_VECTOR_SIZE >= (x)))199
+200
+/* Native aliases */201
+#if defined(SIMDE_ENABLE_NATIVE_ALIASES)202
+#if !defined(SIMDE_X86_MMX_NATIVE)203
+#define SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES204
+#endif205
+#if !defined(SIMDE_X86_SSE_NATIVE)206
+#define SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES207
+#endif208
+#if !defined(SIMDE_X86_SSE2_NATIVE)209
+#define SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES210
+#endif211
+#if !defined(SIMDE_X86_SSE3_NATIVE)212
+#define SIMDE_X86_SSE3_ENABLE_NATIVE_ALIASES213
+#endif214
+#if !defined(SIMDE_X86_SSSE3_NATIVE)215
+#define SIMDE_X86_SSSE3_ENABLE_NATIVE_ALIASES216
+#endif217
+#if !defined(SIMDE_X86_SSE4_1_NATIVE)218
+#define SIMDE_X86_SSE4_1_ENABLE_NATIVE_ALIASES219
+#endif220
+#if !defined(SIMDE_X86_SSE4_2_NATIVE)221
+#define SIMDE_X86_SSE4_2_ENABLE_NATIVE_ALIASES222
+#endif223
+#if !defined(SIMDE_X86_AVX_NATIVE)224
+#define SIMDE_X86_AVX_ENABLE_NATIVE_ALIASES225
+#endif226
+#if !defined(SIMDE_X86_AVX2_NATIVE)227
+#define SIMDE_X86_AVX2_ENABLE_NATIVE_ALIASES228
+#endif229
+#if !defined(SIMDE_X86_FMA_NATIVE)230
+#define SIMDE_X86_FMA_ENABLE_NATIVE_ALIASES231
+#endif232
+#if !defined(SIMDE_X86_AVX512F_NATIVE)233
+#define SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES234
+#endif235
+#if !defined(SIMDE_X86_AVX512VL_NATIVE)236
+#define SIMDE_X86_AVX512VL_ENABLE_NATIVE_ALIASES237
+#endif238
+#if !defined(SIMDE_X86_AVX512BW_NATIVE)239
+#define SIMDE_X86_AVX512BW_ENABLE_NATIVE_ALIASES240
+#endif241
+#if !defined(SIMDE_X86_AVX512DQ_NATIVE)242
+#define SIMDE_X86_AVX512DQ_ENABLE_NATIVE_ALIASES243
+#endif244
+#if !defined(SIMDE_X86_AVX512CD_NATIVE)245
+#define SIMDE_X86_AVX512CD_ENABLE_NATIVE_ALIASES246
+#endif247
+#if !defined(SIMDE_X86_GFNI_NATIVE)248
+#define SIMDE_X86_GFNI_ENABLE_NATIVE_ALIASES249
+#endif250
+#if !defined(SIMDE_X86_PCLMUL_NATIVE)251
+#define SIMDE_X86_PCLMUL_ENABLE_NATIVE_ALIASES252
+#endif253
+#if !defined(SIMDE_X86_VPCLMULQDQ_NATIVE)254
+#define SIMDE_X86_VPCLMULQDQ_ENABLE_NATIVE_ALIASES255
+#endif256
+257
+#if !defined(SIMDE_ARM_NEON_A32V7_NATIVE)258
+#define SIMDE_ARM_NEON_A32V7_ENABLE_NATIVE_ALIASES259
+#endif260
+#if !defined(SIMDE_ARM_NEON_A32V8_NATIVE)261
+#define SIMDE_ARM_NEON_A32V8_ENABLE_NATIVE_ALIASES262
+#endif263
+#if !defined(SIMDE_ARM_NEON_A64V8_NATIVE)264
+#define SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES265
+#endif266
+#endif267
+268
+/* Are floating point values stored using IEEE 754? Knowing269
+ * this at during preprocessing is a bit tricky, mostly because what270
+ * we're curious about is how values are stored and not whether the271
+ * implementation is fully conformant in terms of rounding, NaN272
+ * handling, etc.273
+ *274
+ * For example, if you use -ffast-math or -Ofast on275
+ * GCC or clang IEEE 754 isn't strictly followed, therefore IEE 754276
+ * support is not advertised (by defining __STDC_IEC_559__).277
+ *278
+ * However, what we care about is whether it is safe to assume that279
+ * floating point values are stored in IEEE 754 format, in which case280
+ * we can provide faster implementations of some functions.281
+ *282
+ * Luckily every vaugely modern architecture I'm aware of uses IEEE 754-283
+ * so we just assume IEEE 754 for now. There is a test which verifies284
+ * this, if that test fails sowewhere please let us know and we'll add285
+ * an exception for that platform. Meanwhile, you can define286
+ * SIMDE_NO_IEEE754_STORAGE. */287
+#if !defined(SIMDE_IEEE754_STORAGE) && !defined(SIMDE_NO_IEE754_STORAGE)288
+#define SIMDE_IEEE754_STORAGE289
#endif290
291
#endif /* !defined(SIMDE_FEATURES_H) */292
obs-studio-26.1.0.tar.xz/libobs/util/simde/simde-math.h -> obs-studio-26.1.1.tar.xz/libobs/util/simde/simde-math.h
Changed
674
1
2
#include "hedley.h"3
#include "simde-features.h"4
5
+#include <stdint.h>6
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7
+#include <arm_neon.h>8
+#endif9
+10
+HEDLEY_DIAGNOSTIC_PUSH11
+SIMDE_DISABLE_UNWANTED_DIAGNOSTICS12
+13
+/* SLEEF support14
+ * https://sleef.org/15
+ *16
+ * If you include <sleef.h> prior to including SIMDe, SIMDe will use17
+ * SLEEF. You can also define SIMDE_MATH_SLEEF_ENABLE prior to18
+ * including SIMDe to force the issue.19
+ *20
+ * Note that SLEEF does requires linking to libsleef.21
+ *22
+ * By default, SIMDe will use the 1 ULP functions, but if you use23
+ * SIMDE_ACCURACY_PREFERENCE of 0 we will use up to 4 ULP. This is24
+ * only the case for the simde_math_* functions; for code in other25
+ * SIMDe headers which calls SLEEF directly we may use functions with26
+ * greater error if the API we're implementing is less precise (for27
+ * example, SVML guarantees 4 ULP, so we will generally use the 3.528
+ * ULP functions from SLEEF). */29
+#if !defined(SIMDE_MATH_SLEEF_DISABLE)30
+#if defined(__SLEEF_H__)31
+#define SIMDE_MATH_SLEEF_ENABLE32
+#endif33
+#endif34
+35
+#if defined(SIMDE_MATH_SLEEF_ENABLE) && !defined(__SLEEF_H__)36
+HEDLEY_DIAGNOSTIC_PUSH37
+SIMDE_DIAGNOSTIC_DISABLE_IGNORED_QUALIFIERS_38
+#include <sleef.h>39
+HEDLEY_DIAGNOSTIC_POP40
+#endif41
+42
+#if defined(SIMDE_MATH_SLEEF_ENABLE) && defined(__SLEEF_H__)43
+#if defined(SLEEF_VERSION_MAJOR)44
+#define SIMDE_MATH_SLEEF_VERSION_CHECK(major, minor, patch) \45
+ (HEDLEY_VERSION_ENCODE(SLEEF_VERSION_MAJOR, SLEEF_VERSION_MINOR, \46
+ SLEEF_VERSION_PATCHLEVEL) >= \47
+ HEDLEY_VERSION_ENCODE(major, minor, patch))48
+#else49
+#define SIMDE_MATH_SLEEF_VERSION_CHECK(major, minor, patch) \50
+ (HEDLEY_VERSION_ENCODE(3, 0, 0) >= \51
+ HEDLEY_VERSION_ENCODE(major, minor, patch))52
+#endif53
+#else54
+#define SIMDE_MATH_SLEEF_VERSION_CHECK(major, minor, patch) (0)55
+#endif56
+57
#if defined(__has_builtin)58
#define SIMDE_MATH_BUILTIN_LIBM(func) __has_builtin(__builtin_##func)59
#elif HEDLEY_INTEL_VERSION_CHECK(13, 0, 0) || \60
61
#endif62
#endif63
64
-#if !defined(__cplusplus)65
-/* If this is a problem we *might* be able to avoid including66
- * <complex.h> on some compilers (gcc, clang, and others which67
- * implement builtins like __builtin_cexpf). If you don't have68
- * a <complex.h> please file an issue and we'll take a look. */69
+/* Try to avoid including <complex> since it pulls in a *lot* of code. */70
+#if HEDLEY_HAS_BUILTIN(__builtin_creal) || \71
+ HEDLEY_GCC_VERSION_CHECK(4, 7, 0) || \72
+ HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)73
+HEDLEY_DIAGNOSTIC_PUSH74
+SIMDE_DIAGNOSTIC_DISABLE_C99_EXTENSIONS_75
+typedef __complex__ float simde_cfloat32;76
+typedef __complex__ double simde_cfloat64;77
+HEDLEY_DIAGNOSTIC_POP78
+#define SIMDE_MATH_CMPLX(x, y) \79
+ (HEDLEY_STATIC_CAST(double, x) + \80
+ HEDLEY_STATIC_CAST(double, y) * (__extension__ 1.0j))81
+#define SIMDE_MATH_CMPLXF(x, y) \82
+ (HEDLEY_STATIC_CAST(float, x) + \83
+ HEDLEY_STATIC_CAST(float, y) * (__extension__ 1.0fj))84
+85
+#if !defined(simde_math_creal)86
+#define simde_math_crealf(z) __builtin_crealf(z)87
+#endif88
+#if !defined(simde_math_crealf)89
+#define simde_math_creal(z) __builtin_creal(z)90
+#endif91
+#if !defined(simde_math_cimag)92
+#define simde_math_cimagf(z) __builtin_cimagf(z)93
+#endif94
+#if !defined(simde_math_cimagf)95
+#define simde_math_cimag(z) __builtin_cimag(z)96
+#endif97
+#elif !defined(__cplusplus)98
#include <complex.h>99
100
#if !defined(HEDLEY_MSVC_VERSION)101
102
typedef _Fcomplex simde_cfloat32;103
typedef _Dcomplex simde_cfloat64;104
#endif105
-#if HEDLEY_HAS_BUILTIN(__builtin_complex) || \106
- HEDLEY_GCC_VERSION_CHECK(4, 7, 0) || \107
- HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)108
-#define SIMDE_MATH_CMPLX(x, y) __builtin_complex((double)(x), (double)(y))109
-#define SIMDE_MATH_CMPLXF(x, y) __builtin_complex((float)(x), (float)(y))110
-#elif defined(HEDLEY_MSVC_VERSION)111
+112
+#if defined(HEDLEY_MSVC_VERSION)113
#define SIMDE_MATH_CMPLX(x, y) ((simde_cfloat64){(x), (y)})114
#define SIMDE_MATH_CMPLXF(x, y) ((simde_cfloat32){(x), (y)})115
#elif defined(CMPLX) && defined(CMPLXF)116
#define SIMDE_MATH_CMPLX(x, y) CMPLX(x, y)117
#define SIMDE_MATH_CMPLXF(x, y) CMPLXF(x, y)118
#else119
-/* CMPLX / CMPLXF are in C99, but these seem to be necessary in120
- * some compilers that aren't even MSVC. */121
#define SIMDE_MATH_CMPLX(x, y) \122
(HEDLEY_STATIC_CAST(double, x) + HEDLEY_STATIC_CAST(double, y) * I)123
#define SIMDE_MATH_CMPLXF(x, y) \124
125
#endif126
127
#if !defined(simde_math_creal)128
-#if SIMDE_MATH_BUILTIN_LIBM(creal)129
-#define simde_math_creal(z) __builtin_creal(z)130
-#else131
#define simde_math_creal(z) creal(z)132
#endif133
-#endif134
-135
#if !defined(simde_math_crealf)136
-#if SIMDE_MATH_BUILTIN_LIBM(crealf)137
-#define simde_math_crealf(z) __builtin_crealf(z)138
-#else139
#define simde_math_crealf(z) crealf(z)140
#endif141
-#endif142
-143
#if !defined(simde_math_cimag)144
-#if SIMDE_MATH_BUILTIN_LIBM(cimag)145
-#define simde_math_cimag(z) __builtin_cimag(z)146
-#else147
#define simde_math_cimag(z) cimag(z)148
#endif149
-#endif150
-151
#if !defined(simde_math_cimagf)152
-#if SIMDE_MATH_BUILTIN_LIBM(cimagf)153
-#define simde_math_cimagf(z) __builtin_cimagf(z)154
-#else155
#define simde_math_cimagf(z) cimagf(z)156
#endif157
-#endif158
#else159
-160
HEDLEY_DIAGNOSTIC_PUSH161
#if defined(HEDLEY_MSVC_VERSION)162
#pragma warning(disable : 4530)163
164
#endif165
#endif166
167
+#if !defined(SIMDE_MATH_PI_OVER_180)168
+#define SIMDE_MATH_PI_OVER_180 \169
+ 0.0174532925199432957692369076848861271344287188854172545609719144170
+#endif171
+172
+#if !defined(SIMDE_MATH_PI_OVER_180F)173
+#define SIMDE_MATH_PI_OVER_180F \174
+ 0.0174532925199432957692369076848861271344287188854172545609719144f175
+#endif176
+177
+#if !defined(SIMDE_MATH_180_OVER_PI)178
+#define SIMDE_MATH_180_OVER_PI \179
+ 57.295779513082320876798154814105170332405472466564321549160243861180
+#endif181
+182
+#if !defined(SIMDE_MATH_180_OVER_PIF)183
+#define SIMDE_MATH_180_OVER_PIF \184
+ 57.295779513082320876798154814105170332405472466564321549160243861f185
+#endif186
+187
#if !defined(SIMDE_MATH_FLT_MIN)188
#if defined(FLT_MIN)189
#define SIMDE_MATH_FLT_MIN FLT_MIN190
191
#endif192
#endif193
194
+/*** Manipulation functions ***/195
+196
+#if !defined(simde_math_nextafter)197
+#if (HEDLEY_HAS_BUILTIN(__builtin_nextafter) && \198
+ !defined(HEDLEY_IBM_VERSION)) || \199
+ HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \200
+ HEDLEY_GCC_VERSION_CHECK(3, 4, 0) || \201
+ HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)202
+#define simde_math_nextafter(x, y) __builtin_nextafter(x, y)203
+#elif defined(SIMDE_MATH_HAVE_CMATH)204
+#define simde_math_nextafter(x, y) std::nextafter(x, y)205
+#elif defined(SIMDE_MATH_HAVE_MATH_H)206
+#define simde_math_nextafter(x, y) nextafter(x, y)207
+#endif208
+#endif209
+210
+#if !defined(simde_math_nextafterf)211
+#if (HEDLEY_HAS_BUILTIN(__builtin_nextafterf) && \212
+ !defined(HEDLEY_IBM_VERSION)) || \213
+ HEDLEY_ARM_VERSION_CHECK(4, 1, 0) || \214
+ HEDLEY_GCC_VERSION_CHECK(3, 4, 0) || \215
+ HEDLEY_INTEL_VERSION_CHECK(13, 0, 0)216
+#define simde_math_nextafterf(x, y) __builtin_nextafterf(x, y)217
+#elif defined(SIMDE_MATH_HAVE_CMATH)218
+#define simde_math_nextafterf(x, y) std::nextafter(x, y)219
+#elif defined(SIMDE_MATH_HAVE_MATH_H)220
+#define simde_math_nextafterf(x, y) nextafterf(x, y)221
+#endif222
+#endif223
+224
/*** Functions from C99 ***/225
226
#if !defined(simde_math_abs)227
228
#endif229
#endif230
231
-#if !defined(simde_math_absf)232
-#if SIMDE_MATH_BUILTIN_LIBM(absf)233
-#define simde_math_absf(v) __builtin_absf(v)234
+#if !defined(simde_math_fabsf)235
+#if SIMDE_MATH_BUILTIN_LIBM(fabsf)236
+#define simde_math_fabsf(v) __builtin_fabsf(v)237
#elif defined(SIMDE_MATH_HAVE_CMATH)238
-#define simde_math_absf(v) std::abs(v)239
+#define simde_math_fabsf(v) std::abs(v)240
#elif defined(SIMDE_MATH_HAVE_MATH_H)241
-#define simde_math_absf(v) absf(v)242
+#define simde_math_fabsf(v) fabsf(v)243
#endif244
#endif245
246
247
#endif248
249
#if !defined(simde_math_cosf)250
-#if SIMDE_MATH_BUILTIN_LIBM(cosf)251
+#if defined(SIMDE_MATH_SLEEF_ENABLE)252
+#if SIMDE_ACCURACY_PREFERENCE < 1253
+#define simde_math_cosf(v) Sleef_cosf_u35(v)254
+#else255
+#define simde_math_cosf(v) Sleef_cosf_u10(v)256
+#endif257
+#elif SIMDE_MATH_BUILTIN_LIBM(cosf)258
#define simde_math_cosf(v) __builtin_cosf(v)259
#elif defined(SIMDE_MATH_HAVE_CMATH)260
#define simde_math_cosf(v) std::cos(v)261
262
#endif263
#endif264
265
+#if !defined(simde_math_fma)266
+#if SIMDE_MATH_BUILTIN_LIBM(fma)267
+#define simde_math_fma(x, y, z) __builtin_fma(x, y, z)268
+#elif defined(SIMDE_MATH_HAVE_CMATH)269
+#define simde_math_fma(x, y, z) std::fma(x, y, z)270
+#elif defined(SIMDE_MATH_HAVE_MATH_H)271
+#define simde_math_fma(x, y, z) fma(x, y, z)272
+#endif273
+#endif274
+275
+#if !defined(simde_math_fmaf)276
+#if SIMDE_MATH_BUILTIN_LIBM(fmaf)277
+#define simde_math_fmaf(x, y, z) __builtin_fmaf(x, y, z)278
+#elif defined(SIMDE_MATH_HAVE_CMATH)279
+#define simde_math_fmaf(x, y, z) std::fma(x, y, z)280
+#elif defined(SIMDE_MATH_HAVE_MATH_H)281
+#define simde_math_fmaf(x, y, z) fmaf(x, y, z)282
+#endif283
+#endif284
+285
+#if !defined(simde_math_fmax)286
+#if SIMDE_MATH_BUILTIN_LIBM(fmax)287
+#define simde_math_fmax(x, y, z) __builtin_fmax(x, y, z)288
+#elif defined(SIMDE_MATH_HAVE_CMATH)289
+#define simde_math_fmax(x, y, z) std::fmax(x, y, z)290
+#elif defined(SIMDE_MATH_HAVE_MATH_H)291
+#define simde_math_fmax(x, y, z) fmax(x, y, z)292
+#endif293
+#endif294
+295
+#if !defined(simde_math_fmaxf)296
+#if SIMDE_MATH_BUILTIN_LIBM(fmaxf)297
+#define simde_math_fmaxf(x, y, z) __builtin_fmaxf(x, y, z)298
+#elif defined(SIMDE_MATH_HAVE_CMATH)299
+#define simde_math_fmaxf(x, y, z) std::fmax(x, y, z)300
+#elif defined(SIMDE_MATH_HAVE_MATH_H)301
+#define simde_math_fmaxf(x, y, z) fmaxf(x, y, z)302
+#endif303
+#endif304
+305
#if !defined(simde_math_hypot)306
#if SIMDE_MATH_BUILTIN_LIBM(hypot)307
#define simde_math_hypot(y, x) __builtin_hypot(y, x)308
309
#endif310
#endif311
312
+#if !defined(simde_math_modf)313
+#if SIMDE_MATH_BUILTIN_LIBM(modf)314
+#define simde_math_modf(x, iptr) __builtin_modf(x, iptr)315
+#elif defined(SIMDE_MATH_HAVE_CMATH)316
+#define simde_math_modf(x, iptr) std::modf(x, iptr)317
+#elif defined(SIMDE_MATH_HAVE_MATH_H)318
+#define simde_math_modf(x, iptr) modf(x, iptr)319
+#endif320
+#endif321
+322
+#if !defined(simde_math_modff)323
+#if SIMDE_MATH_BUILTIN_LIBM(modff)324
+#define simde_math_modff(x, iptr) __builtin_modff(x, iptr)325
+#elif defined(SIMDE_MATH_HAVE_CMATH)326
+#define simde_math_modff(x, iptr) std::modf(x, iptr)327
+#elif defined(SIMDE_MATH_HAVE_MATH_H)328
+#define simde_math_modff(x, iptr) modff(x, iptr)329
+#endif330
+#endif331
+332
#if !defined(simde_math_nearbyint)333
#if SIMDE_MATH_BUILTIN_LIBM(nearbyint)334
#define simde_math_nearbyint(v) __builtin_nearbyint(v)335
336
#endif337
#endif338
339
+#if !defined(simde_math_roundeven)340
+#if HEDLEY_HAS_BUILTIN(__builtin_roundeven) || \341
+ HEDLEY_GCC_VERSION_CHECK(10, 0, 0)342
+#define simde_math_roundeven(v) __builtin_roundeven(v)343
+#elif defined(simde_math_round) && defined(simde_math_fabs)344
+static HEDLEY_INLINE double simde_math_roundeven(double v)345
+{346
+ double rounded = simde_math_round(v);347
+ double diff = rounded - v;348
+ if (HEDLEY_UNLIKELY(simde_math_fabs(diff) == 0.5) &&349
+ (HEDLEY_STATIC_CAST(int64_t, rounded) & 1)) {350
+ rounded = v - diff;351
+ }352
+ return rounded;353
+}354
+#define simde_math_roundeven simde_math_roundeven355
+#endif356
+#endif357
+358
+#if !defined(simde_math_roundevenf)359
+#if HEDLEY_HAS_BUILTIN(__builtin_roundevenf) || \360
+ HEDLEY_GCC_VERSION_CHECK(10, 0, 0)361
+#define simde_math_roundevenf(v) __builtin_roundevenf(v)362
+#elif defined(simde_math_roundf) && defined(simde_math_fabsf)363
+static HEDLEY_INLINE float simde_math_roundevenf(float v)364
+{365
+ float rounded = simde_math_roundf(v);366
+ float diff = rounded - v;367
+ if (HEDLEY_UNLIKELY(simde_math_fabsf(diff) == 0.5f) &&368
+ (HEDLEY_STATIC_CAST(int32_t, rounded) & 1)) {369
+ rounded = v - diff;370
+ }371
+ return rounded;372
+}373
+#define simde_math_roundevenf simde_math_roundevenf374
+#endif375
+#endif376
+377
#if !defined(simde_math_sin)378
#if SIMDE_MATH_BUILTIN_LIBM(sin)379
#define simde_math_sin(v) __builtin_sin(v)380
381
/*** Complex functions ***/382
383
#if !defined(simde_math_cexp)384
-#if defined(__cplusplus)385
-#define simde_math_cexp(v) std::cexp(v)386
-#elif SIMDE_MATH_BUILTIN_LIBM(cexp)387
+#if SIMDE_MATH_BUILTIN_LIBM(cexp)388
#define simde_math_cexp(v) __builtin_cexp(v)389
+#elif defined(__cplusplus)390
+#define simde_math_cexp(v) std::cexp(v)391
#elif defined(SIMDE_MATH_HAVE_MATH_H)392
#define simde_math_cexp(v) cexp(v)393
#endif394
#endif395
396
#if !defined(simde_math_cexpf)397
-#if defined(__cplusplus)398
-#define simde_math_cexpf(v) std::exp(v)399
-#elif SIMDE_MATH_BUILTIN_LIBM(cexpf)400
+#if SIMDE_MATH_BUILTIN_LIBM(cexpf)401
#define simde_math_cexpf(v) __builtin_cexpf(v)402
+#elif defined(__cplusplus)403
+#define simde_math_cexpf(v) std::exp(v)404
#elif defined(SIMDE_MATH_HAVE_MATH_H)405
#define simde_math_cexpf(v) cexpf(v)406
#endif407
408
409
static HEDLEY_INLINE double simde_math_rad2deg(double radians)410
{411
- return radians * (180.0 / SIMDE_MATH_PI);412
+ return radians * SIMDE_MATH_180_OVER_PI;413
}414
415
static HEDLEY_INLINE float simde_math_rad2degf(float radians)416
{417
- return radians * (180.0f / SIMDE_MATH_PIF);418
+ return radians * SIMDE_MATH_180_OVER_PIF;419
}420
421
static HEDLEY_INLINE double simde_math_deg2rad(double degrees)422
{423
- return degrees * (SIMDE_MATH_PI / 180.0);424
+ return degrees * SIMDE_MATH_PI_OVER_180;425
}426
427
static HEDLEY_INLINE float simde_math_deg2radf(float degrees)428
{429
- return degrees * (SIMDE_MATH_PIF / 180.0f);430
+ return degrees * (SIMDE_MATH_PI_OVER_180F);431
}432
433
+/*** Saturated arithmetic ***/434
+435
+static HEDLEY_INLINE int8_t simde_math_adds_i8(int8_t a, int8_t b)436
+{437
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)438
+ return vqaddb_s8(a, b);439
+#else440
+ uint8_t a_ = HEDLEY_STATIC_CAST(uint8_t, a);441
+ uint8_t b_ = HEDLEY_STATIC_CAST(uint8_t, b);442
+ uint8_t r_ = a_ + b_;443
+444
+ a_ = (a_ >> ((8 * sizeof(r_)) - 1)) + INT8_MAX;445
+ if (HEDLEY_STATIC_CAST(int8_t, ((a_ ^ b_) | ~(b_ ^ r_))) >= 0) {446
+ r_ = a_;447
+ }448
+449
+ return HEDLEY_STATIC_CAST(int8_t, r_);450
+#endif451
+}452
+453
+static HEDLEY_INLINE int16_t simde_math_adds_i16(int16_t a, int16_t b)454
+{455
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)456
+ return vqaddh_s16(a, b);457
+#else458
+ uint16_t a_ = HEDLEY_STATIC_CAST(uint16_t, a);459
+ uint16_t b_ = HEDLEY_STATIC_CAST(uint16_t, b);460
+ uint16_t r_ = a_ + b_;461
+462
+ a_ = (a_ >> ((8 * sizeof(r_)) - 1)) + INT16_MAX;463
+ if (HEDLEY_STATIC_CAST(int16_t, ((a_ ^ b_) | ~(b_ ^ r_))) >= 0) {464
+ r_ = a_;465
+ }466
+467
+ return HEDLEY_STATIC_CAST(int16_t, r_);468
+#endif469
+}470
+471
+static HEDLEY_INLINE int32_t simde_math_adds_i32(int32_t a, int32_t b)472
+{473
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)474
+ return vqadds_s32(a, b);475
+#else476
+ uint32_t a_ = HEDLEY_STATIC_CAST(uint32_t, a);477
+ uint32_t b_ = HEDLEY_STATIC_CAST(uint32_t, b);478
+ uint32_t r_ = a_ + b_;479
+480
+ a_ = (a_ >> ((8 * sizeof(r_)) - 1)) + INT32_MAX;481
+ if (HEDLEY_STATIC_CAST(int32_t, ((a_ ^ b_) | ~(b_ ^ r_))) >= 0) {482
+ r_ = a_;483
+ }484
+485
+ return HEDLEY_STATIC_CAST(int32_t, r_);486
+#endif487
+}488
+489
+static HEDLEY_INLINE int64_t simde_math_adds_i64(int64_t a, int64_t b)490
+{491
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)492
+ return vqaddd_s64(a, b);493
+#else494
+ uint64_t a_ = HEDLEY_STATIC_CAST(uint64_t, a);495
+ uint64_t b_ = HEDLEY_STATIC_CAST(uint64_t, b);496
+ uint64_t r_ = a_ + b_;497
+498
+ a_ = (a_ >> ((8 * sizeof(r_)) - 1)) + INT64_MAX;499
+ if (HEDLEY_STATIC_CAST(int64_t, ((a_ ^ b_) | ~(b_ ^ r_))) >= 0) {500
+ r_ = a_;501
+ }502
+503
+ return HEDLEY_STATIC_CAST(int64_t, r_);504
+#endif505
+}506
+507
+static HEDLEY_INLINE uint8_t simde_math_adds_u8(uint8_t a, uint8_t b)508
+{509
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)510
+ return vqaddb_u8(a, b);511
+#else512
+ uint8_t r = a + b;513
+ r |= -(r < a);514
+ return r;515
+#endif516
+}517
+518
+static HEDLEY_INLINE uint16_t simde_math_adds_u16(uint16_t a, uint16_t b)519
+{520
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)521
+ return vqaddh_u16(a, b);522
+#else523
+ uint16_t r = a + b;524
+ r |= -(r < a);525
+ return r;526
+#endif527
+}528
+529
+static HEDLEY_INLINE uint32_t simde_math_adds_u32(uint32_t a, uint32_t b)530
+{531
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)532
+ return vqadds_u32(a, b);533
+#else534
+ uint32_t r = a + b;535
+ r |= -(r < a);536
+ return r;537
+#endif538
+}539
+540
+static HEDLEY_INLINE uint64_t simde_math_adds_u64(uint64_t a, uint64_t b)541
+{542
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)543
+ return vqaddd_u64(a, b);544
+#else545
+ uint64_t r = a + b;546
+ r |= -(r < a);547
+ return r;548
+#endif549
+}550
+551
+static HEDLEY_INLINE int8_t simde_math_subs_i8(int8_t a, int8_t b)552
+{553
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)554
+ return vqsubb_s8(a, b);555
+#else556
+ uint8_t a_ = HEDLEY_STATIC_CAST(uint8_t, a);557
+ uint8_t b_ = HEDLEY_STATIC_CAST(uint8_t, b);558
+ uint8_t r_ = a_ - b_;559
+560
+ a_ = (a_ >> 7) + INT8_MAX;561
+562
+ if (HEDLEY_STATIC_CAST(int8_t, (a_ ^ b_) & (a_ ^ r_)) < 0) {563
+ r_ = a_;564
+ }565
+566
+ return HEDLEY_STATIC_CAST(int8_t, r_);567
+#endif568
+}569
+570
+static HEDLEY_INLINE int16_t simde_math_subs_i16(int16_t a, int16_t b)571
+{572
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)573
+ return vqsubh_s16(a, b);574
+#else575
+ uint16_t a_ = HEDLEY_STATIC_CAST(uint16_t, a);576
+ uint16_t b_ = HEDLEY_STATIC_CAST(uint16_t, b);577
+ uint16_t r_ = a_ - b_;578
+579
+ a_ = (a_ >> 15) + INT16_MAX;580
+581
+ if (HEDLEY_STATIC_CAST(int16_t, (a_ ^ b_) & (a_ ^ r_)) < 0) {582
+ r_ = a_;583
+ }584
+585
+ return HEDLEY_STATIC_CAST(int16_t, r_);586
+#endif587
+}588
+589
+static HEDLEY_INLINE int32_t simde_math_subs_i32(int32_t a, int32_t b)590
+{591
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)592
+ return vqsubs_s32(a, b);593
+#else594
+ uint32_t a_ = HEDLEY_STATIC_CAST(uint32_t, a);595
+ uint32_t b_ = HEDLEY_STATIC_CAST(uint32_t, b);596
+ uint32_t r_ = a_ - b_;597
+598
+ a_ = (a_ >> 31) + INT32_MAX;599
+600
+ if (HEDLEY_STATIC_CAST(int32_t, (a_ ^ b_) & (a_ ^ r_)) < 0) {601
+ r_ = a_;602
+ }603
+604
+ return HEDLEY_STATIC_CAST(int32_t, r_);605
+#endif606
+}607
+608
+static HEDLEY_INLINE int64_t simde_math_subs_i64(int64_t a, int64_t b)609
+{610
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)611
+ return vqsubd_s64(a, b);612
+#else613
+ uint64_t a_ = HEDLEY_STATIC_CAST(uint64_t, a);614
+ uint64_t b_ = HEDLEY_STATIC_CAST(uint64_t, b);615
+ uint64_t r_ = a_ - b_;616
+617
+ a_ = (a_ >> 63) + INT64_MAX;618
+619
+ if (HEDLEY_STATIC_CAST(int64_t, (a_ ^ b_) & (a_ ^ r_)) < 0) {620
+ r_ = a_;621
+ }622
+623
+ return HEDLEY_STATIC_CAST(int64_t, r_);624
+#endif625
+}626
+627
+static HEDLEY_INLINE uint8_t simde_math_subs_u8(uint8_t a, uint8_t b)628
+{629
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)630
+ return vqsubb_u8(a, b);631
+#else632
+ uint8_t res = a - b;633
+ res &= -(res <= a);634
+ return res;635
+#endif636
+}637
+638
+static HEDLEY_INLINE uint16_t simde_math_subs_u16(uint16_t a, uint16_t b)639
+{640
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)641
+ return vqsubh_u16(a, b);642
+#else643
+ uint16_t res = a - b;644
+ res &= -(res <= a);645
+ return res;646
+#endif647
+}648
+649
+static HEDLEY_INLINE uint32_t simde_math_subs_u32(uint32_t a, uint32_t b)650
+{651
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)652
+ return vqsubs_u32(a, b);653
+#else654
+ uint32_t res = a - b;655
+ res &= -(res <= a);656
+ return res;657
+#endif658
+}659
+660
+static HEDLEY_INLINE uint64_t simde_math_subs_u64(uint64_t a, uint64_t b)661
+{662
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)663
+ return vqsubd_u64(a, b);664
+#else665
+ uint64_t res = a - b;666
+ res &= -(res <= a);667
+ return res;668
+#endif669
+}670
+671
+HEDLEY_DIAGNOSTIC_POP672
+673
#endif /* !defined(SIMDE_MATH_H) */674
obs-studio-26.1.1.tar.xz/libobs/util/simde/x86
Added
2
1
+(directory)2
obs-studio-26.1.1.tar.xz/libobs/util/simde/x86/mmx.h
Added
2458
1
2
+/* SPDX-License-Identifier: MIT3
+ *4
+ * Permission is hereby granted, free of charge, to any person5
+ * obtaining a copy of this software and associated documentation6
+ * files (the "Software"), to deal in the Software without7
+ * restriction, including without limitation the rights to use, copy,8
+ * modify, merge, publish, distribute, sublicense, and/or sell copies9
+ * of the Software, and to permit persons to whom the Software is10
+ * furnished to do so, subject to the following conditions:11
+ *12
+ * The above copyright notice and this permission notice shall be13
+ * included in all copies or substantial portions of the Software.14
+ *15
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,16
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF17
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND18
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS19
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN20
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN21
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE22
+ * SOFTWARE.23
+ *24
+ * Copyright:25
+ * 2017-2020 Evan Nemerson <evan@nemerson.com>26
+ */27
+28
+#if !defined(SIMDE_X86_MMX_H)29
+#define SIMDE_X86_MMX_H30
+31
+#include "../simde-common.h"32
+33
+HEDLEY_DIAGNOSTIC_PUSH34
+SIMDE_DISABLE_UNWANTED_DIAGNOSTICS35
+36
+#if defined(SIMDE_X86_MMX_NATIVE)37
+#define SIMDE_X86_MMX_USE_NATIVE_TYPE38
+#elif defined(SIMDE_X86_SSE_NATIVE)39
+#define SIMDE_X86_MMX_USE_NATIVE_TYPE40
+#endif41
+42
+#if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)43
+#include <mmintrin.h>44
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)45
+#include <arm_neon.h>46
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)47
+#include <loongson-mmiintrin.h>48
+#endif49
+50
+#include <stdint.h>51
+#include <limits.h>52
+53
+SIMDE_BEGIN_DECLS_54
+55
+typedef union {56
+#if defined(SIMDE_VECTOR_SUBSCRIPT)57
+ SIMDE_ALIGN_TO_8 int8_t i8 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;58
+ SIMDE_ALIGN_TO_8 int16_t i16 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;59
+ SIMDE_ALIGN_TO_8 int32_t i32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;60
+ SIMDE_ALIGN_TO_8 int64_t i64 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;61
+ SIMDE_ALIGN_TO_8 uint8_t u8 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;62
+ SIMDE_ALIGN_TO_8 uint16_t u16 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;63
+ SIMDE_ALIGN_TO_8 uint32_t u32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;64
+ SIMDE_ALIGN_TO_8 uint64_t u64 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;65
+ SIMDE_ALIGN_TO_8 simde_float32 f32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;66
+ SIMDE_ALIGN_TO_8 int_fast32_t i32f SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;67
+ SIMDE_ALIGN_TO_8 uint_fast32_t u32f SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;68
+#else69
+ SIMDE_ALIGN_TO_8 int8_t i8[8];70
+ SIMDE_ALIGN_TO_8 int16_t i16[4];71
+ SIMDE_ALIGN_TO_8 int32_t i32[2];72
+ SIMDE_ALIGN_TO_8 int64_t i64[1];73
+ SIMDE_ALIGN_TO_8 uint8_t u8[8];74
+ SIMDE_ALIGN_TO_8 uint16_t u16[4];75
+ SIMDE_ALIGN_TO_8 uint32_t u32[2];76
+ SIMDE_ALIGN_TO_8 uint64_t u64[1];77
+ SIMDE_ALIGN_TO_8 simde_float32 f32[2];78
+ SIMDE_ALIGN_TO_8 int_fast32_t i32f[8 / sizeof(int_fast32_t)];79
+ SIMDE_ALIGN_TO_8 uint_fast32_t u32f[8 / sizeof(uint_fast32_t)];80
+#endif81
+82
+#if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)83
+ __m64 n;84
+#endif85
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)86
+ int8x8_t neon_i8;87
+ int16x4_t neon_i16;88
+ int32x2_t neon_i32;89
+ int64x1_t neon_i64;90
+ uint8x8_t neon_u8;91
+ uint16x4_t neon_u16;92
+ uint32x2_t neon_u32;93
+ uint64x1_t neon_u64;94
+ float32x2_t neon_f32;95
+#endif96
+#if defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)97
+ int8x8_t mmi_i8;98
+ int16x4_t mmi_i16;99
+ int32x2_t mmi_i32;100
+ int64_t mmi_i64;101
+ uint8x8_t mmi_u8;102
+ uint16x4_t mmi_u16;103
+ uint32x2_t mmi_u32;104
+ uint64_t mmi_u64;105
+#endif106
+} simde__m64_private;107
+108
+#if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)109
+typedef __m64 simde__m64;110
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)111
+typedef int32x2_t simde__m64;112
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)113
+typedef int32x2_t simde__m64;114
+#elif defined(SIMDE_VECTOR_SUBSCRIPT)115
+typedef int32_t simde__m64 SIMDE_ALIGN_TO_8 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;116
+#else117
+typedef simde__m64_private simde__m64;118
+#endif119
+120
+#if !defined(SIMDE_X86_MMX_USE_NATIVE_TYPE) && \121
+ defined(SIMDE_ENABLE_NATIVE_ALIASES)122
+#define SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES123
+typedef simde__m64 __m64;124
+#endif125
+126
+HEDLEY_STATIC_ASSERT(8 == sizeof(simde__m64), "__m64 size incorrect");127
+HEDLEY_STATIC_ASSERT(8 == sizeof(simde__m64_private), "__m64 size incorrect");128
+#if defined(SIMDE_CHECK_ALIGNMENT) && defined(SIMDE_ALIGN_OF)129
+HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m64) == 8,130
+ "simde__m64 is not 8-byte aligned");131
+HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m64_private) == 8,132
+ "simde__m64_private is not 8-byte aligned");133
+#endif134
+135
+SIMDE_FUNCTION_ATTRIBUTES136
+simde__m64 simde__m64_from_private(simde__m64_private v)137
+{138
+ simde__m64 r;139
+ simde_memcpy(&r, &v, sizeof(r));140
+ return r;141
+}142
+143
+SIMDE_FUNCTION_ATTRIBUTES144
+simde__m64_private simde__m64_to_private(simde__m64 v)145
+{146
+ simde__m64_private r;147
+ simde_memcpy(&r, &v, sizeof(r));148
+ return r;149
+}150
+151
+#define SIMDE_X86_GENERATE_CONVERSION_FUNCTION(simde_type, source_type, isax, \152
+ fragment) \153
+ SIMDE_FUNCTION_ATTRIBUTES \154
+ simde__##simde_type simde__##simde_type##_from_##isax##_##fragment( \155
+ source_type value) \156
+ { \157
+ simde__##simde_type##_private r_; \158
+ r_.isax##_##fragment = value; \159
+ return simde__##simde_type##_from_private(r_); \160
+ } \161
+ \162
+ SIMDE_FUNCTION_ATTRIBUTES \163
+ source_type simde__##simde_type##_to_##isax##_##fragment( \164
+ simde__##simde_type value) \165
+ { \166
+ simde__##simde_type##_private r_ = \167
+ simde__##simde_type##_to_private(value); \168
+ return r_.isax##_##fragment; \169
+ }170
+171
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)172
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int8x8_t, neon, i8)173
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int16x4_t, neon, i16)174
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int32x2_t, neon, i32)175
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int64x1_t, neon, i64)176
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint8x8_t, neon, u8)177
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint16x4_t, neon, u16)178
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint32x2_t, neon, u32)179
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint64x1_t, neon, u64)180
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, float32x2_t, neon, f32)181
+#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */182
+183
+#if defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)184
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int8x8_t, mmi, i8)185
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int16x4_t, mmi, i16)186
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int32x2_t, mmi, i32)187
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int64_t, mmi, i64)188
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint8x8_t, mmi, u8)189
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint16x4_t, mmi, u16)190
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint32x2_t, mmi, u32)191
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint64_t, mmi, u64)192
+#endif /* defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE) */193
+194
+SIMDE_FUNCTION_ATTRIBUTES195
+simde__m64 simde_mm_add_pi8(simde__m64 a, simde__m64 b)196
+{197
+#if defined(SIMDE_X86_MMX_NATIVE)198
+ return _mm_add_pi8(a, b);199
+#else200
+ simde__m64_private r_;201
+ simde__m64_private a_ = simde__m64_to_private(a);202
+ simde__m64_private b_ = simde__m64_to_private(b);203
+204
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)205
+ r_.neon_i8 = vadd_s8(a_.neon_i8, b_.neon_i8);206
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)207
+ r_.mmi_i8 = paddb_s(a_.mmi_i8, b_.mmi_i8);208
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)209
+ r_.i8 = a_.i8 + b_.i8;210
+#else211
+ SIMDE_VECTORIZE212
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {213
+ r_.i8[i] = a_.i8[i] + b_.i8[i];214
+ }215
+#endif216
+217
+ return simde__m64_from_private(r_);218
+#endif219
+}220
+#define simde_m_paddb(a, b) simde_mm_add_pi8(a, b)221
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)222
+#define _mm_add_pi8(a, b) simde_mm_add_pi8(a, b)223
+#define _m_paddb(a, b) simde_m_paddb(a, b)224
+#endif225
+226
+SIMDE_FUNCTION_ATTRIBUTES227
+simde__m64 simde_mm_add_pi16(simde__m64 a, simde__m64 b)228
+{229
+#if defined(SIMDE_X86_MMX_NATIVE)230
+ return _mm_add_pi16(a, b);231
+#else232
+ simde__m64_private r_;233
+ simde__m64_private a_ = simde__m64_to_private(a);234
+ simde__m64_private b_ = simde__m64_to_private(b);235
+236
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)237
+ r_.neon_i16 = vadd_s16(a_.neon_i16, b_.neon_i16);238
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)239
+ r_.mmi_i16 = paddh_s(a_.mmi_i16, b_.mmi_i16);240
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)241
+ r_.i16 = a_.i16 + b_.i16;242
+#else243
+ SIMDE_VECTORIZE244
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {245
+ r_.i16[i] = a_.i16[i] + b_.i16[i];246
+ }247
+#endif248
+249
+ return simde__m64_from_private(r_);250
+#endif251
+}252
+#define simde_m_paddw(a, b) simde_mm_add_pi16(a, b)253
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)254
+#define _mm_add_pi16(a, b) simde_mm_add_pi16(a, b)255
+#define _m_paddw(a, b) simde_mm_add_pi16(a, b)256
+#endif257
+258
+SIMDE_FUNCTION_ATTRIBUTES259
+simde__m64 simde_mm_add_pi32(simde__m64 a, simde__m64 b)260
+{261
+#if defined(SIMDE_X86_MMX_NATIVE)262
+ return _mm_add_pi32(a, b);263
+#else264
+ simde__m64_private r_;265
+ simde__m64_private a_ = simde__m64_to_private(a);266
+ simde__m64_private b_ = simde__m64_to_private(b);267
+268
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)269
+ r_.neon_i32 = vadd_s32(a_.neon_i32, b_.neon_i32);270
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)271
+ r_.mmi_i32 = paddw_s(a_.mmi_i32, b_.mmi_i32);272
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)273
+ r_.i32 = a_.i32 + b_.i32;274
+#else275
+ SIMDE_VECTORIZE276
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {277
+ r_.i32[i] = a_.i32[i] + b_.i32[i];278
+ }279
+#endif280
+281
+ return simde__m64_from_private(r_);282
+#endif283
+}284
+#define simde_m_paddd(a, b) simde_mm_add_pi32(a, b)285
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)286
+#define _mm_add_pi32(a, b) simde_mm_add_pi32(a, b)287
+#define _m_paddd(a, b) simde_mm_add_pi32(a, b)288
+#endif289
+290
+SIMDE_FUNCTION_ATTRIBUTES291
+simde__m64 simde_mm_adds_pi8(simde__m64 a, simde__m64 b)292
+{293
+#if defined(SIMDE_X86_MMX_NATIVE)294
+ return _mm_adds_pi8(a, b);295
+#else296
+ simde__m64_private r_, a_ = simde__m64_to_private(a),297
+ b_ = simde__m64_to_private(b);298
+299
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)300
+ r_.neon_i8 = vqadd_s8(a_.neon_i8, b_.neon_i8);301
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)302
+ r_.mmi_i8 = paddsb(a_.mmi_i8, b_.mmi_i8);303
+#else304
+ SIMDE_VECTORIZE305
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {306
+ if ((((b_.i8[i]) > 0) &&307
+ ((a_.i8[i]) > (INT8_MAX - (b_.i8[i]))))) {308
+ r_.i8[i] = INT8_MAX;309
+ } else if ((((b_.i8[i]) < 0) &&310
+ ((a_.i8[i]) < (INT8_MIN - (b_.i8[i]))))) {311
+ r_.i8[i] = INT8_MIN;312
+ } else {313
+ r_.i8[i] = (a_.i8[i]) + (b_.i8[i]);314
+ }315
+ }316
+#endif317
+318
+ return simde__m64_from_private(r_);319
+#endif320
+}321
+#define simde_m_paddsb(a, b) simde_mm_adds_pi8(a, b)322
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)323
+#define _mm_adds_pi8(a, b) simde_mm_adds_pi8(a, b)324
+#define _m_paddsb(a, b) simde_mm_adds_pi8(a, b)325
+#endif326
+327
+SIMDE_FUNCTION_ATTRIBUTES328
+simde__m64 simde_mm_adds_pu8(simde__m64 a, simde__m64 b)329
+{330
+#if defined(SIMDE_X86_MMX_NATIVE)331
+ return _mm_adds_pu8(a, b);332
+#else333
+ simde__m64_private r_;334
+ simde__m64_private a_ = simde__m64_to_private(a);335
+ simde__m64_private b_ = simde__m64_to_private(b);336
+337
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)338
+ r_.neon_u8 = vqadd_u8(a_.neon_u8, b_.neon_u8);339
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)340
+ r_.mmi_u8 = paddusb(a_.mmi_u8, b_.mmi_u8);341
+#else342
+ SIMDE_VECTORIZE343
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {344
+ const uint_fast16_t x =345
+ HEDLEY_STATIC_CAST(uint_fast16_t, a_.u8[i]) +346
+ HEDLEY_STATIC_CAST(uint_fast16_t, b_.u8[i]);347
+ if (x > UINT8_MAX)348
+ r_.u8[i] = UINT8_MAX;349
+ else350
+ r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, x);351
+ }352
+#endif353
+354
+ return simde__m64_from_private(r_);355
+#endif356
+}357
+#define simde_m_paddusb(a, b) simde_mm_adds_pu8(a, b)358
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)359
+#define _mm_adds_pu8(a, b) simde_mm_adds_pu8(a, b)360
+#define _m_paddusb(a, b) simde_mm_adds_pu8(a, b)361
+#endif362
+363
+SIMDE_FUNCTION_ATTRIBUTES364
+simde__m64 simde_mm_adds_pi16(simde__m64 a, simde__m64 b)365
+{366
+#if defined(SIMDE_X86_MMX_NATIVE)367
+ return _mm_adds_pi16(a, b);368
+#else369
+ simde__m64_private r_;370
+ simde__m64_private a_ = simde__m64_to_private(a);371
+ simde__m64_private b_ = simde__m64_to_private(b);372
+373
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)374
+ r_.neon_i16 = vqadd_s16(a_.neon_i16, b_.neon_i16);375
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)376
+ r_.mmi_i16 = paddsh(a_.mmi_i16, b_.mmi_i16);377
+#else378
+ SIMDE_VECTORIZE379
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {380
+ if ((((b_.i16[i]) > 0) &&381
+ ((a_.i16[i]) > (INT16_MAX - (b_.i16[i]))))) {382
+ r_.i16[i] = INT16_MAX;383
+ } else if ((((b_.i16[i]) < 0) &&384
+ ((a_.i16[i]) < (SHRT_MIN - (b_.i16[i]))))) {385
+ r_.i16[i] = SHRT_MIN;386
+ } else {387
+ r_.i16[i] = (a_.i16[i]) + (b_.i16[i]);388
+ }389
+ }390
+#endif391
+392
+ return simde__m64_from_private(r_);393
+#endif394
+}395
+#define simde_m_paddsw(a, b) simde_mm_adds_pi16(a, b)396
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)397
+#define _mm_adds_pi16(a, b) simde_mm_adds_pi16(a, b)398
+#define _m_paddsw(a, b) simde_mm_adds_pi16(a, b)399
+#endif400
+401
+SIMDE_FUNCTION_ATTRIBUTES402
+simde__m64 simde_mm_adds_pu16(simde__m64 a, simde__m64 b)403
+{404
+#if defined(SIMDE_X86_MMX_NATIVE)405
+ return _mm_adds_pu16(a, b);406
+#else407
+ simde__m64_private r_;408
+ simde__m64_private a_ = simde__m64_to_private(a);409
+ simde__m64_private b_ = simde__m64_to_private(b);410
+411
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)412
+ r_.neon_u16 = vqadd_u16(a_.neon_u16, b_.neon_u16);413
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)414
+ r_.mmi_u16 = paddush(a_.mmi_u16, b_.mmi_u16);415
+#else416
+ SIMDE_VECTORIZE417
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {418
+ const uint32_t x = a_.u16[i] + b_.u16[i];419
+ if (x > UINT16_MAX)420
+ r_.u16[i] = UINT16_MAX;421
+ else422
+ r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, x);423
+ }424
+#endif425
+426
+ return simde__m64_from_private(r_);427
+#endif428
+}429
+#define simde_m_paddusw(a, b) simde_mm_adds_pu16(a, b)430
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)431
+#define _mm_adds_pu16(a, b) simde_mm_adds_pu16(a, b)432
+#define _m_paddusw(a, b) simde_mm_adds_pu16(a, b)433
+#endif434
+435
+SIMDE_FUNCTION_ATTRIBUTES436
+simde__m64 simde_mm_and_si64(simde__m64 a, simde__m64 b)437
+{438
+#if defined(SIMDE_X86_MMX_NATIVE)439
+ return _mm_and_si64(a, b);440
+#else441
+ simde__m64_private r_;442
+ simde__m64_private a_ = simde__m64_to_private(a);443
+ simde__m64_private b_ = simde__m64_to_private(b);444
+445
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)446
+ r_.neon_i32 = vand_s32(a_.neon_i32, b_.neon_i32);447
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)448
+ r_.i64 = a_.i64 & b_.i64;449
+#else450
+ r_.i64[0] = a_.i64[0] & b_.i64[0];451
+#endif452
+453
+ return simde__m64_from_private(r_);454
+#endif455
+}456
+#define simde_m_pand(a, b) simde_mm_and_si64(a, b)457
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)458
+#define _mm_and_si64(a, b) simde_mm_and_si64(a, b)459
+#define _m_pand(a, b) simde_mm_and_si64(a, b)460
+#endif461
+462
+SIMDE_FUNCTION_ATTRIBUTES463
+simde__m64 simde_mm_andnot_si64(simde__m64 a, simde__m64 b)464
+{465
+#if defined(SIMDE_X86_MMX_NATIVE)466
+ return _mm_andnot_si64(a, b);467
+#else468
+ simde__m64_private r_;469
+ simde__m64_private a_ = simde__m64_to_private(a);470
+ simde__m64_private b_ = simde__m64_to_private(b);471
+472
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)473
+ r_.neon_i32 = vbic_s32(b_.neon_i32, a_.neon_i32);474
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)475
+ r_.mmi_i32 = pandn_sw(a_.mmi_i32, b_.mmi_i32);476
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)477
+ r_.i32f = ~a_.i32f & b_.i32f;478
+#else479
+ r_.u64[0] = (~(a_.u64[0])) & (b_.u64[0]);480
+#endif481
+482
+ return simde__m64_from_private(r_);483
+#endif484
+}485
+#define simde_m_pandn(a, b) simde_mm_andnot_si64(a, b)486
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)487
+#define _mm_andnot_si64(a, b) simde_mm_andnot_si64(a, b)488
+#define _m_pandn(a, b) simde_mm_andnot_si64(a, b)489
+#endif490
+491
+SIMDE_FUNCTION_ATTRIBUTES492
+simde__m64 simde_mm_cmpeq_pi8(simde__m64 a, simde__m64 b)493
+{494
+#if defined(SIMDE_X86_MMX_NATIVE)495
+ return _mm_cmpeq_pi8(a, b);496
+#else497
+ simde__m64_private r_;498
+ simde__m64_private a_ = simde__m64_to_private(a);499
+ simde__m64_private b_ = simde__m64_to_private(b);500
+501
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)502
+ r_.neon_u8 = vceq_s8(a_.neon_i8, b_.neon_i8);503
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)504
+ r_.mmi_i8 = pcmpeqb_s(a_.mmi_i8, b_.mmi_i8);505
+#else506
+ SIMDE_VECTORIZE507
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {508
+ r_.i8[i] = (a_.i8[i] == b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);509
+ }510
+#endif511
+512
+ return simde__m64_from_private(r_);513
+#endif514
+}515
+#define simde_m_pcmpeqb(a, b) simde_mm_cmpeq_pi8(a, b)516
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)517
+#define _mm_cmpeq_pi8(a, b) simde_mm_cmpeq_pi8(a, b)518
+#define _m_pcmpeqb(a, b) simde_mm_cmpeq_pi8(a, b)519
+#endif520
+521
+SIMDE_FUNCTION_ATTRIBUTES522
+simde__m64 simde_mm_cmpeq_pi16(simde__m64 a, simde__m64 b)523
+{524
+#if defined(SIMDE_X86_MMX_NATIVE)525
+ return _mm_cmpeq_pi16(a, b);526
+#else527
+ simde__m64_private r_;528
+ simde__m64_private a_ = simde__m64_to_private(a);529
+ simde__m64_private b_ = simde__m64_to_private(b);530
+531
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)532
+ r_.neon_u16 = vceq_s16(a_.neon_i16, b_.neon_i16);533
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)534
+ r_.mmi_i16 = pcmpeqh_s(a_.mmi_i16, b_.mmi_i16);535
+#else536
+ SIMDE_VECTORIZE537
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {538
+ r_.i16[i] = (a_.i16[i] == b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);539
+ }540
+#endif541
+542
+ return simde__m64_from_private(r_);543
+#endif544
+}545
+#define simde_m_pcmpeqw(a, b) simde_mm_cmpeq_pi16(a, b)546
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)547
+#define _mm_cmpeq_pi16(a, b) simde_mm_cmpeq_pi16(a, b)548
+#define _m_pcmpeqw(a, b) simde_mm_cmpeq_pi16(a, b)549
+#endif550
+551
+SIMDE_FUNCTION_ATTRIBUTES552
+simde__m64 simde_mm_cmpeq_pi32(simde__m64 a, simde__m64 b)553
+{554
+#if defined(SIMDE_X86_MMX_NATIVE)555
+ return _mm_cmpeq_pi32(a, b);556
+#else557
+ simde__m64_private r_;558
+ simde__m64_private a_ = simde__m64_to_private(a);559
+ simde__m64_private b_ = simde__m64_to_private(b);560
+561
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)562
+ r_.neon_u32 = vceq_s32(a_.neon_i32, b_.neon_i32);563
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)564
+ r_.mmi_i32 = pcmpeqw_s(a_.mmi_i32, b_.mmi_i32);565
+#else566
+ SIMDE_VECTORIZE567
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {568
+ r_.i32[i] = (a_.i32[i] == b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);569
+ }570
+#endif571
+572
+ return simde__m64_from_private(r_);573
+#endif574
+}575
+#define simde_m_pcmpeqd(a, b) simde_mm_cmpeq_pi32(a, b)576
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)577
+#define _mm_cmpeq_pi32(a, b) simde_mm_cmpeq_pi32(a, b)578
+#define _m_pcmpeqd(a, b) simde_mm_cmpeq_pi32(a, b)579
+#endif580
+581
+SIMDE_FUNCTION_ATTRIBUTES582
+simde__m64 simde_mm_cmpgt_pi8(simde__m64 a, simde__m64 b)583
+{584
+#if defined(SIMDE_X86_MMX_NATIVE)585
+ return _mm_cmpgt_pi8(a, b);586
+#else587
+ simde__m64_private r_;588
+ simde__m64_private a_ = simde__m64_to_private(a);589
+ simde__m64_private b_ = simde__m64_to_private(b);590
+591
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)592
+ r_.neon_u8 = vcgt_s8(a_.neon_i8, b_.neon_i8);593
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)594
+ r_.mmi_i8 = pcmpgtb_s(a_.mmi_i8, b_.mmi_i8);595
+#else596
+ SIMDE_VECTORIZE597
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {598
+ r_.i8[i] = (a_.i8[i] > b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);599
+ }600
+#endif601
+602
+ return simde__m64_from_private(r_);603
+#endif604
+}605
+#define simde_m_pcmpgtb(a, b) simde_mm_cmpgt_pi8(a, b)606
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)607
+#define _mm_cmpgt_pi8(a, b) simde_mm_cmpgt_pi8(a, b)608
+#define _m_pcmpgtb(a, b) simde_mm_cmpgt_pi8(a, b)609
+#endif610
+611
+SIMDE_FUNCTION_ATTRIBUTES612
+simde__m64 simde_mm_cmpgt_pi16(simde__m64 a, simde__m64 b)613
+{614
+#if defined(SIMDE_X86_MMX_NATIVE)615
+ return _mm_cmpgt_pi16(a, b);616
+#else617
+ simde__m64_private r_;618
+ simde__m64_private a_ = simde__m64_to_private(a);619
+ simde__m64_private b_ = simde__m64_to_private(b);620
+621
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)622
+ r_.neon_u16 = vcgt_s16(a_.neon_i16, b_.neon_i16);623
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)624
+ r_.mmi_i16 = pcmpgth_s(a_.mmi_i16, b_.mmi_i16);625
+#else626
+ SIMDE_VECTORIZE627
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {628
+ r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);629
+ }630
+#endif631
+632
+ return simde__m64_from_private(r_);633
+#endif634
+}635
+#define simde_m_pcmpgtw(a, b) simde_mm_cmpgt_pi16(a, b)636
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)637
+#define _mm_cmpgt_pi16(a, b) simde_mm_cmpgt_pi16(a, b)638
+#define _m_pcmpgtw(a, b) simde_mm_cmpgt_pi16(a, b)639
+#endif640
+641
+SIMDE_FUNCTION_ATTRIBUTES642
+simde__m64 simde_mm_cmpgt_pi32(simde__m64 a, simde__m64 b)643
+{644
+#if defined(SIMDE_X86_MMX_NATIVE)645
+ return _mm_cmpgt_pi32(a, b);646
+#else647
+ simde__m64_private r_;648
+ simde__m64_private a_ = simde__m64_to_private(a);649
+ simde__m64_private b_ = simde__m64_to_private(b);650
+651
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)652
+ r_.neon_u32 = vcgt_s32(a_.neon_i32, b_.neon_i32);653
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)654
+ r_.mmi_i32 = pcmpgtw_s(a_.mmi_i32, b_.mmi_i32);655
+#else656
+ SIMDE_VECTORIZE657
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {658
+ r_.i32[i] = (a_.i32[i] > b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);659
+ }660
+#endif661
+662
+ return simde__m64_from_private(r_);663
+#endif664
+}665
+#define simde_m_pcmpgtd(a, b) simde_mm_cmpgt_pi32(a, b)666
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)667
+#define _mm_cmpgt_pi32(a, b) simde_mm_cmpgt_pi32(a, b)668
+#define _m_pcmpgtd(a, b) simde_mm_cmpgt_pi32(a, b)669
+#endif670
+671
+SIMDE_FUNCTION_ATTRIBUTES672
+int64_t simde_mm_cvtm64_si64(simde__m64 a)673
+{674
+#if defined(SIMDE_X86_MMX_NATIVE) && defined(SIMDE_ARCH_AMD64) && \675
+ !defined(__PGI)676
+ return _mm_cvtm64_si64(a);677
+#else678
+ simde__m64_private a_ = simde__m64_to_private(a);679
+680
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)681
+ HEDLEY_DIAGNOSTIC_PUSH682
+#if HEDLEY_HAS_WARNING("-Wvector-conversion") && \683
+ SIMDE_DETECT_CLANG_VERSION_NOT(10, 0, 0)684
+#pragma clang diagnostic ignored "-Wvector-conversion"685
+#endif686
+ return vget_lane_s64(a_.neon_i64, 0);687
+ HEDLEY_DIAGNOSTIC_POP688
+#else689
+ return a_.i64[0];690
+#endif691
+#endif692
+}693
+#define simde_m_to_int64(a) simde_mm_cvtm64_si64(a)694
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)695
+#define _mm_cvtm64_si64(a) simde_mm_cvtm64_si64(a)696
+#define _m_to_int64(a) simde_mm_cvtm64_si64(a)697
+#endif698
+699
+SIMDE_FUNCTION_ATTRIBUTES700
+simde__m64 simde_mm_cvtsi32_si64(int32_t a)701
+{702
+#if defined(SIMDE_X86_MMX_NATIVE)703
+ return _mm_cvtsi32_si64(a);704
+#else705
+ simde__m64_private r_;706
+707
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)708
+ const int32_t av[sizeof(r_.neon_i32) / sizeof(r_.neon_i32[0])] = {a, 0};709
+ r_.neon_i32 = vld1_s32(av);710
+#else711
+ r_.i32[0] = a;712
+ r_.i32[1] = 0;713
+#endif714
+715
+ return simde__m64_from_private(r_);716
+#endif717
+}718
+#define simde_m_from_int(a) simde_mm_cvtsi32_si64(a)719
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)720
+#define _mm_cvtsi32_si64(a) simde_mm_cvtsi32_si64(a)721
+#define _m_from_int(a) simde_mm_cvtsi32_si64(a)722
+#endif723
+724
+SIMDE_FUNCTION_ATTRIBUTES725
+simde__m64 simde_mm_cvtsi64_m64(int64_t a)726
+{727
+#if defined(SIMDE_X86_MMX_NATIVE) && defined(SIMDE_ARCH_AMD64) && \728
+ !defined(__PGI)729
+ return _mm_cvtsi64_m64(a);730
+#else731
+ simde__m64_private r_;732
+733
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)734
+ r_.neon_i64 = vld1_s64(&a);735
+#else736
+ r_.i64[0] = a;737
+#endif738
+739
+ return simde__m64_from_private(r_);740
+#endif741
+}742
+#define simde_m_from_int64(a) simde_mm_cvtsi64_m64(a)743
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)744
+#define _mm_cvtsi64_m64(a) simde_mm_cvtsi64_m64(a)745
+#define _m_from_int64(a) simde_mm_cvtsi64_m64(a)746
+#endif747
+748
+SIMDE_FUNCTION_ATTRIBUTES749
+int32_t simde_mm_cvtsi64_si32(simde__m64 a)750
+{751
+#if defined(SIMDE_X86_MMX_NATIVE)752
+ return _mm_cvtsi64_si32(a);753
+#else754
+ simde__m64_private a_ = simde__m64_to_private(a);755
+756
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)757
+ HEDLEY_DIAGNOSTIC_PUSH758
+#if HEDLEY_HAS_WARNING("-Wvector-conversion") && \759
+ SIMDE_DETECT_CLANG_VERSION_NOT(10, 0, 0)760
+#pragma clang diagnostic ignored "-Wvector-conversion"761
+#endif762
+ return vget_lane_s32(a_.neon_i32, 0);763
+ HEDLEY_DIAGNOSTIC_POP764
+#else765
+ return a_.i32[0];766
+#endif767
+#endif768
+}769
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)770
+#define _mm_cvtsi64_si32(a) simde_mm_cvtsi64_si32(a)771
+#endif772
+773
+SIMDE_FUNCTION_ATTRIBUTES774
+void simde_mm_empty(void)775
+{776
+#if defined(SIMDE_X86_MMX_NATIVE)777
+ _mm_empty();778
+#else779
+ /* noop */780
+#endif781
+}782
+#define simde_m_empty() simde_mm_empty()783
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)784
+#define _mm_empty() simde_mm_empty()785
+#define _m_empty() simde_mm_empty()786
+#endif787
+788
+SIMDE_FUNCTION_ATTRIBUTES789
+simde__m64 simde_mm_madd_pi16(simde__m64 a, simde__m64 b)790
+{791
+#if defined(SIMDE_X86_MMX_NATIVE)792
+ return _mm_madd_pi16(a, b);793
+#else794
+ simde__m64_private r_;795
+ simde__m64_private a_ = simde__m64_to_private(a);796
+ simde__m64_private b_ = simde__m64_to_private(b);797
+798
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)799
+ int32x4_t i1 = vmull_s16(a_.neon_i16, b_.neon_i16);800
+ r_.neon_i32 = vpadd_s32(vget_low_s32(i1), vget_high_s32(i1));801
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)802
+ r_.mmi_i32 = pmaddhw(a_.mmi_i16, b_.mmi_i16);803
+#else804
+ SIMDE_VECTORIZE805
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i += 2) {806
+ r_.i32[i / 2] = (a_.i16[i] * b_.i16[i]) +807
+ (a_.i16[i + 1] * b_.i16[i + 1]);808
+ }809
+#endif810
+811
+ return simde__m64_from_private(r_);812
+#endif813
+}814
+#define simde_m_pmaddwd(a, b) simde_mm_madd_pi16(a, b)815
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)816
+#define _mm_madd_pi16(a, b) simde_mm_madd_pi16(a, b)817
+#define _m_pmaddwd(a, b) simde_mm_madd_pi16(a, b)818
+#endif819
+820
+SIMDE_FUNCTION_ATTRIBUTES821
+simde__m64 simde_mm_mulhi_pi16(simde__m64 a, simde__m64 b)822
+{823
+#if defined(SIMDE_X86_MMX_NATIVE)824
+ return _mm_mulhi_pi16(a, b);825
+#else826
+ simde__m64_private r_;827
+ simde__m64_private a_ = simde__m64_to_private(a);828
+ simde__m64_private b_ = simde__m64_to_private(b);829
+830
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)831
+ const int32x4_t t1 = vmull_s16(a_.neon_i16, b_.neon_i16);832
+ const uint32x4_t t2 = vshrq_n_u32(vreinterpretq_u32_s32(t1), 16);833
+ const uint16x4_t t3 = vmovn_u32(t2);834
+ r_.neon_u16 = t3;835
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)836
+ r_.mmi_i16 = pmulhh(a_.mmi_i16, b_.mmi_i16);837
+#else838
+ SIMDE_VECTORIZE839
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {840
+ r_.i16[i] = HEDLEY_STATIC_CAST(int16_t,841
+ ((a_.i16[i] * b_.i16[i]) >> 16));842
+ }843
+#endif844
+845
+ return simde__m64_from_private(r_);846
+#endif847
+}848
+#define simde_m_pmulhw(a, b) simde_mm_mulhi_pi16(a, b)849
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)850
+#define _mm_mulhi_pi16(a, b) simde_mm_mulhi_pi16(a, b)851
+#define _m_pmulhw(a, b) simde_mm_mulhi_pi16(a, b)852
+#endif853
+854
+SIMDE_FUNCTION_ATTRIBUTES855
+simde__m64 simde_mm_mullo_pi16(simde__m64 a, simde__m64 b)856
+{857
+#if defined(SIMDE_X86_MMX_NATIVE)858
+ return _mm_mullo_pi16(a, b);859
+#else860
+ simde__m64_private r_;861
+ simde__m64_private a_ = simde__m64_to_private(a);862
+ simde__m64_private b_ = simde__m64_to_private(b);863
+864
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)865
+ const int32x4_t t1 = vmull_s16(a_.neon_i16, b_.neon_i16);866
+ const uint16x4_t t2 = vmovn_u32(vreinterpretq_u32_s32(t1));867
+ r_.neon_u16 = t2;868
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)869
+ r_.mmi_i16 = pmullh(a_.mmi_i16, b_.mmi_i16);870
+#else871
+ SIMDE_VECTORIZE872
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {873
+ r_.i16[i] = HEDLEY_STATIC_CAST(874
+ int16_t, ((a_.i16[i] * b_.i16[i]) & 0xffff));875
+ }876
+#endif877
+878
+ return simde__m64_from_private(r_);879
+#endif880
+}881
+#define simde_m_pmullw(a, b) simde_mm_mullo_pi16(a, b)882
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)883
+#define _mm_mullo_pi16(a, b) simde_mm_mullo_pi16(a, b)884
+#define _m_pmullw(a, b) simde_mm_mullo_pi16(a, b)885
+#endif886
+887
+SIMDE_FUNCTION_ATTRIBUTES888
+simde__m64 simde_mm_or_si64(simde__m64 a, simde__m64 b)889
+{890
+#if defined(SIMDE_X86_MMX_NATIVE)891
+ return _mm_or_si64(a, b);892
+#else893
+ simde__m64_private r_;894
+ simde__m64_private a_ = simde__m64_to_private(a);895
+ simde__m64_private b_ = simde__m64_to_private(b);896
+897
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)898
+ r_.neon_i32 = vorr_s32(a_.neon_i32, b_.neon_i32);899
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)900
+ r_.i64 = a_.i64 | b_.i64;901
+#else902
+ r_.i64[0] = a_.i64[0] | b_.i64[0];903
+#endif904
+905
+ return simde__m64_from_private(r_);906
+#endif907
+}908
+#define simde_m_por(a, b) simde_mm_or_si64(a, b)909
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)910
+#define _mm_or_si64(a, b) simde_mm_or_si64(a, b)911
+#define _m_por(a, b) simde_mm_or_si64(a, b)912
+#endif913
+914
+SIMDE_FUNCTION_ATTRIBUTES915
+simde__m64 simde_mm_packs_pi16(simde__m64 a, simde__m64 b)916
+{917
+#if defined(SIMDE_X86_MMX_NATIVE)918
+ return _mm_packs_pi16(a, b);919
+#else920
+ simde__m64_private r_;921
+ simde__m64_private a_ = simde__m64_to_private(a);922
+ simde__m64_private b_ = simde__m64_to_private(b);923
+924
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)925
+ r_.neon_i8 = vqmovn_s16(vcombine_s16(a_.neon_i16, b_.neon_i16));926
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)927
+ r_.mmi_i8 = packsshb(a_.mmi_i16, b_.mmi_i16);928
+#else929
+ SIMDE_VECTORIZE930
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {931
+ if (a_.i16[i] < INT8_MIN) {932
+ r_.i8[i] = INT8_MIN;933
+ } else if (a_.i16[i] > INT8_MAX) {934
+ r_.i8[i] = INT8_MAX;935
+ } else {936
+ r_.i8[i] = HEDLEY_STATIC_CAST(int8_t, a_.i16[i]);937
+ }938
+ }939
+940
+ SIMDE_VECTORIZE941
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {942
+ if (b_.i16[i] < INT8_MIN) {943
+ r_.i8[i + 4] = INT8_MIN;944
+ } else if (b_.i16[i] > INT8_MAX) {945
+ r_.i8[i + 4] = INT8_MAX;946
+ } else {947
+ r_.i8[i + 4] = HEDLEY_STATIC_CAST(int8_t, b_.i16[i]);948
+ }949
+ }950
+#endif951
+952
+ return simde__m64_from_private(r_);953
+#endif954
+}955
+#define simde_m_packsswb(a, b) simde_mm_packs_pi16(a, b)956
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)957
+#define _mm_packs_pi16(a, b) simde_mm_packs_pi16(a, b)958
+#define _m_packsswb(a, b) simde_mm_packs_pi16(a, b)959
+#endif960
+961
+SIMDE_FUNCTION_ATTRIBUTES962
+simde__m64 simde_mm_packs_pi32(simde__m64 a, simde__m64 b)963
+{964
+#if defined(SIMDE_X86_MMX_NATIVE)965
+ return _mm_packs_pi32(a, b);966
+#else967
+ simde__m64_private r_;968
+ simde__m64_private a_ = simde__m64_to_private(a);969
+ simde__m64_private b_ = simde__m64_to_private(b);970
+971
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)972
+ r_.neon_i16 = vqmovn_s32(vcombine_s32(a_.neon_i32, b_.neon_i32));973
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)974
+ r_.mmi_i16 = packsswh(a_.mmi_i32, b_.mmi_i32);975
+#else976
+ SIMDE_VECTORIZE977
+ for (size_t i = 0; i < (8 / sizeof(a_.i32[0])); i++) {978
+ if (a_.i32[i] < SHRT_MIN) {979
+ r_.i16[i] = SHRT_MIN;980
+ } else if (a_.i32[i] > INT16_MAX) {981
+ r_.i16[i] = INT16_MAX;982
+ } else {983
+ r_.i16[i] = HEDLEY_STATIC_CAST(int16_t, a_.i32[i]);984
+ }985
+ }986
+987
+ SIMDE_VECTORIZE988
+ for (size_t i = 0; i < (8 / sizeof(b_.i32[0])); i++) {989
+ if (b_.i32[i] < SHRT_MIN) {990
+ r_.i16[i + 2] = SHRT_MIN;991
+ } else if (b_.i32[i] > INT16_MAX) {992
+ r_.i16[i + 2] = INT16_MAX;993
+ } else {994
+ r_.i16[i + 2] = HEDLEY_STATIC_CAST(int16_t, b_.i32[i]);995
+ }996
+ }997
+#endif998
+999
+ return simde__m64_from_private(r_);1000
+#endif1001
+}1002
+#define simde_m_packssdw(a, b) simde_mm_packs_pi32(a, b)1003
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1004
+#define _mm_packs_pi32(a, b) simde_mm_packs_pi32(a, b)1005
+#define _m_packssdw(a, b) simde_mm_packs_pi32(a, b)1006
+#endif1007
+1008
+SIMDE_FUNCTION_ATTRIBUTES1009
+simde__m64 simde_mm_packs_pu16(simde__m64 a, simde__m64 b)1010
+{1011
+#if defined(SIMDE_X86_MMX_NATIVE)1012
+ return _mm_packs_pu16(a, b);1013
+#else1014
+ simde__m64_private r_;1015
+ simde__m64_private a_ = simde__m64_to_private(a);1016
+ simde__m64_private b_ = simde__m64_to_private(b);1017
+1018
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1019
+ const int16x8_t t1 = vcombine_s16(a_.neon_i16, b_.neon_i16);1020
+1021
+ /* Set elements which are < 0 to 0 */1022
+ const int16x8_t t2 =1023
+ vandq_s16(t1, vreinterpretq_s16_u16(vcgezq_s16(t1)));1024
+1025
+ /* Vector with all s16 elements set to UINT8_MAX */1026
+ const int16x8_t vmax =1027
+ vmovq_n_s16(HEDLEY_STATIC_CAST(int16_t, UINT8_MAX));1028
+1029
+ /* Elements which are within the acceptable range */1030
+ const int16x8_t le_max =1031
+ vandq_s16(t2, vreinterpretq_s16_u16(vcleq_s16(t2, vmax)));1032
+ const int16x8_t gt_max =1033
+ vandq_s16(vmax, vreinterpretq_s16_u16(vcgtq_s16(t2, vmax)));1034
+1035
+ /* Final values as 16-bit integers */1036
+ const int16x8_t values = vorrq_s16(le_max, gt_max);1037
+1038
+ r_.neon_u8 = vmovn_u16(vreinterpretq_u16_s16(values));1039
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)1040
+ r_.mmi_u8 = packushb(a_.mmi_u16, b_.mmi_u16);1041
+#else1042
+ SIMDE_VECTORIZE1043
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1044
+ if (a_.i16[i] > UINT8_MAX) {1045
+ r_.u8[i] = UINT8_MAX;1046
+ } else if (a_.i16[i] < 0) {1047
+ r_.u8[i] = 0;1048
+ } else {1049
+ r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, a_.i16[i]);1050
+ }1051
+ }1052
+1053
+ SIMDE_VECTORIZE1054
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1055
+ if (b_.i16[i] > UINT8_MAX) {1056
+ r_.u8[i + 4] = UINT8_MAX;1057
+ } else if (b_.i16[i] < 0) {1058
+ r_.u8[i + 4] = 0;1059
+ } else {1060
+ r_.u8[i + 4] = HEDLEY_STATIC_CAST(uint8_t, b_.i16[i]);1061
+ }1062
+ }1063
+#endif1064
+1065
+ return simde__m64_from_private(r_);1066
+#endif1067
+}1068
+#define simde_m_packuswb(a, b) simde_mm_packs_pu16(a, b)1069
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1070
+#define _mm_packs_pu16(a, b) simde_mm_packs_pu16(a, b)1071
+#define _m_packuswb(a, b) simde_mm_packs_pu16(a, b)1072
+#endif1073
+1074
+SIMDE_FUNCTION_ATTRIBUTES1075
+simde__m64 simde_mm_set_pi8(int8_t e7, int8_t e6, int8_t e5, int8_t e4,1076
+ int8_t e3, int8_t e2, int8_t e1, int8_t e0)1077
+{1078
+#if defined(SIMDE_X86_MMX_NATIVE)1079
+ return _mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0);1080
+#else1081
+ simde__m64_private r_;1082
+1083
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1084
+ const int8_t v[sizeof(r_.i8) / sizeof(r_.i8[0])] = {e0, e1, e2, e3,1085
+ e4, e5, e6, e7};1086
+ r_.neon_i8 = vld1_s8(v);1087
+#else1088
+ r_.i8[0] = e0;1089
+ r_.i8[1] = e1;1090
+ r_.i8[2] = e2;1091
+ r_.i8[3] = e3;1092
+ r_.i8[4] = e4;1093
+ r_.i8[5] = e5;1094
+ r_.i8[6] = e6;1095
+ r_.i8[7] = e7;1096
+#endif1097
+1098
+ return simde__m64_from_private(r_);1099
+#endif1100
+}1101
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1102
+#define _mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0) \1103
+ simde_mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0)1104
+#endif1105
+1106
+SIMDE_FUNCTION_ATTRIBUTES1107
+simde__m64 simde_x_mm_set_pu8(uint8_t e7, uint8_t e6, uint8_t e5, uint8_t e4,1108
+ uint8_t e3, uint8_t e2, uint8_t e1, uint8_t e0)1109
+{1110
+ simde__m64_private r_;1111
+1112
+#if defined(SIMDE_X86_MMX_NATIVE)1113
+ r_.n = _mm_set_pi8(1114
+ HEDLEY_STATIC_CAST(int8_t, e7), HEDLEY_STATIC_CAST(int8_t, e6),1115
+ HEDLEY_STATIC_CAST(int8_t, e5), HEDLEY_STATIC_CAST(int8_t, e4),1116
+ HEDLEY_STATIC_CAST(int8_t, e3), HEDLEY_STATIC_CAST(int8_t, e2),1117
+ HEDLEY_STATIC_CAST(int8_t, e1), HEDLEY_STATIC_CAST(int8_t, e0));1118
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1119
+ const uint8_t v[sizeof(r_.u8) / sizeof(r_.u8[0])] = {e0, e1, e2, e3,1120
+ e4, e5, e6, e7};1121
+ r_.neon_u8 = vld1_u8(v);1122
+#else1123
+ r_.u8[0] = e0;1124
+ r_.u8[1] = e1;1125
+ r_.u8[2] = e2;1126
+ r_.u8[3] = e3;1127
+ r_.u8[4] = e4;1128
+ r_.u8[5] = e5;1129
+ r_.u8[6] = e6;1130
+ r_.u8[7] = e7;1131
+#endif1132
+1133
+ return simde__m64_from_private(r_);1134
+}1135
+1136
+SIMDE_FUNCTION_ATTRIBUTES1137
+simde__m64 simde_mm_set_pi16(int16_t e3, int16_t e2, int16_t e1, int16_t e0)1138
+{1139
+#if defined(SIMDE_X86_MMX_NATIVE)1140
+ return _mm_set_pi16(e3, e2, e1, e0);1141
+#else1142
+ simde__m64_private r_;1143
+1144
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1145
+ const int16_t v[sizeof(r_.i16) / sizeof(r_.i16[0])] = {e0, e1, e2, e3};1146
+ r_.neon_i16 = vld1_s16(v);1147
+#else1148
+ r_.i16[0] = e0;1149
+ r_.i16[1] = e1;1150
+ r_.i16[2] = e2;1151
+ r_.i16[3] = e3;1152
+#endif1153
+1154
+ return simde__m64_from_private(r_);1155
+#endif1156
+}1157
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1158
+#define _mm_set_pi16(e3, e2, e1, e0) simde_mm_set_pi16(e3, e2, e1, e0)1159
+#endif1160
+1161
+SIMDE_FUNCTION_ATTRIBUTES1162
+simde__m64 simde_x_mm_set_pu16(uint16_t e3, uint16_t e2, uint16_t e1,1163
+ uint16_t e0)1164
+{1165
+ simde__m64_private r_;1166
+1167
+#if defined(SIMDE_X86_MMX_NATIVE)1168
+ r_.n = _mm_set_pi16(HEDLEY_STATIC_CAST(int16_t, e3),1169
+ HEDLEY_STATIC_CAST(int16_t, e2),1170
+ HEDLEY_STATIC_CAST(int16_t, e1),1171
+ HEDLEY_STATIC_CAST(int16_t, e0));1172
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1173
+ const uint16_t v[sizeof(r_.u16) / sizeof(r_.u16[0])] = {e0, e1, e2, e3};1174
+ r_.neon_u16 = vld1_u16(v);1175
+#else1176
+ r_.u16[0] = e0;1177
+ r_.u16[1] = e1;1178
+ r_.u16[2] = e2;1179
+ r_.u16[3] = e3;1180
+#endif1181
+1182
+ return simde__m64_from_private(r_);1183
+}1184
+1185
+SIMDE_FUNCTION_ATTRIBUTES1186
+simde__m64 simde_x_mm_set_pu32(uint32_t e1, uint32_t e0)1187
+{1188
+ simde__m64_private r_;1189
+1190
+#if defined(SIMDE_X86_MMX_NATIVE)1191
+ r_.n = _mm_set_pi32(HEDLEY_STATIC_CAST(int32_t, e1),1192
+ HEDLEY_STATIC_CAST(int32_t, e0));1193
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1194
+ const uint32_t v[sizeof(r_.u32) / sizeof(r_.u32[0])] = {e0, e1};1195
+ r_.neon_u32 = vld1_u32(v);1196
+#else1197
+ r_.u32[0] = e0;1198
+ r_.u32[1] = e1;1199
+#endif1200
+1201
+ return simde__m64_from_private(r_);1202
+}1203
+1204
+SIMDE_FUNCTION_ATTRIBUTES1205
+simde__m64 simde_mm_set_pi32(int32_t e1, int32_t e0)1206
+{1207
+ simde__m64_private r_;1208
+1209
+#if defined(SIMDE_X86_MMX_NATIVE)1210
+ r_.n = _mm_set_pi32(e1, e0);1211
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1212
+ const int32_t v[sizeof(r_.i32) / sizeof(r_.i32[0])] = {e0, e1};1213
+ r_.neon_i32 = vld1_s32(v);1214
+#else1215
+ r_.i32[0] = e0;1216
+ r_.i32[1] = e1;1217
+#endif1218
+1219
+ return simde__m64_from_private(r_);1220
+}1221
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1222
+#define _mm_set_pi32(e1, e0) simde_mm_set_pi32(e1, e0)1223
+#endif1224
+1225
+SIMDE_FUNCTION_ATTRIBUTES1226
+simde__m64 simde_x_mm_set_pi64(int64_t e0)1227
+{1228
+ simde__m64_private r_;1229
+1230
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1231
+ const int64_t v[sizeof(r_.i64) / sizeof(r_.i64[0])] = {e0};1232
+ r_.neon_i64 = vld1_s64(v);1233
+#else1234
+ r_.i64[0] = e0;1235
+#endif1236
+1237
+ return simde__m64_from_private(r_);1238
+}1239
+1240
+SIMDE_FUNCTION_ATTRIBUTES1241
+simde__m64 simde_x_mm_set_f32x2(simde_float32 e1, simde_float32 e0)1242
+{1243
+ simde__m64_private r_;1244
+1245
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1246
+ const simde_float32 v[sizeof(r_.f32) / sizeof(r_.f32[0])] = {e0, e1};1247
+ r_.neon_f32 = vld1_f32(v);1248
+#else1249
+ r_.f32[0] = e0;1250
+ r_.f32[1] = e1;1251
+#endif1252
+1253
+ return simde__m64_from_private(r_);1254
+}1255
+1256
+SIMDE_FUNCTION_ATTRIBUTES1257
+simde__m64 simde_mm_set1_pi8(int8_t a)1258
+{1259
+#if defined(SIMDE_X86_MMX_NATIVE)1260
+ return _mm_set1_pi8(a);1261
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1262
+ simde__m64_private r_;1263
+ r_.neon_i8 = vmov_n_s8(a);1264
+ return simde__m64_from_private(r_);1265
+#else1266
+ return simde_mm_set_pi8(a, a, a, a, a, a, a, a);1267
+#endif1268
+}1269
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1270
+#define _mm_set1_pi8(a) simde_mm_set1_pi8(a)1271
+#endif1272
+1273
+SIMDE_FUNCTION_ATTRIBUTES1274
+simde__m64 simde_mm_set1_pi16(int16_t a)1275
+{1276
+#if defined(SIMDE_X86_MMX_NATIVE)1277
+ return _mm_set1_pi16(a);1278
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1279
+ simde__m64_private r_;1280
+ r_.neon_i16 = vmov_n_s16(a);1281
+ return simde__m64_from_private(r_);1282
+#else1283
+ return simde_mm_set_pi16(a, a, a, a);1284
+#endif1285
+}1286
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1287
+#define _mm_set1_pi16(a) simde_mm_set1_pi16(a)1288
+#endif1289
+1290
+SIMDE_FUNCTION_ATTRIBUTES1291
+simde__m64 simde_mm_set1_pi32(int32_t a)1292
+{1293
+#if defined(SIMDE_X86_MMX_NATIVE)1294
+ return _mm_set1_pi32(a);1295
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1296
+ simde__m64_private r_;1297
+ r_.neon_i32 = vmov_n_s32(a);1298
+ return simde__m64_from_private(r_);1299
+#else1300
+ return simde_mm_set_pi32(a, a);1301
+#endif1302
+}1303
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1304
+#define _mm_set1_pi32(a) simde_mm_set1_pi32(a)1305
+#endif1306
+1307
+SIMDE_FUNCTION_ATTRIBUTES1308
+simde__m64 simde_mm_setr_pi8(int8_t e7, int8_t e6, int8_t e5, int8_t e4,1309
+ int8_t e3, int8_t e2, int8_t e1, int8_t e0)1310
+{1311
+#if defined(SIMDE_X86_MMX_NATIVE)1312
+ return _mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0);1313
+#else1314
+ return simde_mm_set_pi8(e0, e1, e2, e3, e4, e5, e6, e7);1315
+#endif1316
+}1317
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1318
+#define _mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0) \1319
+ simde_mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0)1320
+#endif1321
+1322
+SIMDE_FUNCTION_ATTRIBUTES1323
+simde__m64 simde_mm_setr_pi16(int16_t e3, int16_t e2, int16_t e1, int16_t e0)1324
+{1325
+#if defined(SIMDE_X86_MMX_NATIVE)1326
+ return _mm_setr_pi16(e3, e2, e1, e0);1327
+#else1328
+ return simde_mm_set_pi16(e0, e1, e2, e3);1329
+#endif1330
+}1331
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1332
+#define _mm_setr_pi16(e3, e2, e1, e0) simde_mm_setr_pi16(e3, e2, e1, e0)1333
+#endif1334
+1335
+SIMDE_FUNCTION_ATTRIBUTES1336
+simde__m64 simde_mm_setr_pi32(int32_t e1, int32_t e0)1337
+{1338
+#if defined(SIMDE_X86_MMX_NATIVE)1339
+ return _mm_setr_pi32(e1, e0);1340
+#else1341
+ return simde_mm_set_pi32(e0, e1);1342
+#endif1343
+}1344
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1345
+#define _mm_setr_pi32(e1, e0) simde_mm_setr_pi32(e1, e0)1346
+#endif1347
+1348
+SIMDE_FUNCTION_ATTRIBUTES1349
+simde__m64 simde_mm_setzero_si64(void)1350
+{1351
+#if defined(SIMDE_X86_MMX_NATIVE)1352
+ return _mm_setzero_si64();1353
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1354
+ simde__m64_private r_;1355
+ r_.neon_u32 = vmov_n_u32(0);1356
+ return simde__m64_from_private(r_);1357
+#else1358
+ return simde_mm_set_pi32(0, 0);1359
+#endif1360
+}1361
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1362
+#define _mm_setzero_si64() simde_mm_setzero_si64()1363
+#endif1364
+1365
+SIMDE_FUNCTION_ATTRIBUTES1366
+simde__m64 simde_x_mm_load_si64(const void *mem_addr)1367
+{1368
+ simde__m64 r;1369
+ simde_memcpy(&r, SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m64),1370
+ sizeof(r));1371
+ return r;1372
+}1373
+1374
+SIMDE_FUNCTION_ATTRIBUTES1375
+simde__m64 simde_x_mm_loadu_si64(const void *mem_addr)1376
+{1377
+ simde__m64 r;1378
+ simde_memcpy(&r, mem_addr, sizeof(r));1379
+ return r;1380
+}1381
+1382
+SIMDE_FUNCTION_ATTRIBUTES1383
+void simde_x_mm_store_si64(void *mem_addr, simde__m64 value)1384
+{1385
+ simde_memcpy(SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m64), &value,1386
+ sizeof(value));1387
+}1388
+1389
+SIMDE_FUNCTION_ATTRIBUTES1390
+void simde_x_mm_storeu_si64(void *mem_addr, simde__m64 value)1391
+{1392
+ simde_memcpy(mem_addr, &value, sizeof(value));1393
+}1394
+1395
+SIMDE_FUNCTION_ATTRIBUTES1396
+simde__m64 simde_x_mm_setone_si64(void)1397
+{1398
+ return simde_mm_set1_pi32(~INT32_C(0));1399
+}1400
+1401
+SIMDE_FUNCTION_ATTRIBUTES1402
+simde__m64 simde_mm_sll_pi16(simde__m64 a, simde__m64 count)1403
+{1404
+#if defined(SIMDE_X86_MMX_NATIVE)1405
+ return _mm_sll_pi16(a, count);1406
+#else1407
+ simde__m64_private r_;1408
+ simde__m64_private a_ = simde__m64_to_private(a);1409
+ simde__m64_private count_ = simde__m64_to_private(count);1410
+1411
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1412
+ HEDLEY_DIAGNOSTIC_PUSH1413
+#if HEDLEY_HAS_WARNING("-Wvector-conversion") && \1414
+ SIMDE_DETECT_CLANG_VERSION_NOT(10, 0, 0)1415
+#pragma clang diagnostic ignored "-Wvector-conversion"1416
+#endif1417
+ r_.neon_i16 =1418
+ vshl_s16(a_.neon_i16,1419
+ vmov_n_s16(HEDLEY_STATIC_CAST(1420
+ int16_t, vget_lane_u64(count_.neon_u64, 0))));1421
+ HEDLEY_DIAGNOSTIC_POP1422
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \1423
+ defined(SIMDE_BUG_CLANG_POWER9_16x4_BAD_SHIFT)1424
+ if (HEDLEY_UNLIKELY(count_.u64[0] > 15))1425
+ return simde_mm_setzero_si64();1426
+1427
+ r_.i16 = a_.i16 << HEDLEY_STATIC_CAST(int16_t, count_.u64[0]);1428
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1429
+ r_.i16 = a_.i16 << count_.u64[0];1430
+#else1431
+ if (HEDLEY_UNLIKELY(count_.u64[0] > 15)) {1432
+ simde_memset(&r_, 0, sizeof(r_));1433
+ return simde__m64_from_private(r_);1434
+ }1435
+1436
+ SIMDE_VECTORIZE1437
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {1438
+ r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t,1439
+ a_.u16[i] << count_.u64[0]);1440
+ }1441
+#endif1442
+1443
+ return simde__m64_from_private(r_);1444
+#endif1445
+}1446
+#define simde_m_psllw(a, count) simde_mm_sll_pi16(a, count)1447
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1448
+#define _mm_sll_pi16(a, count) simde_mm_sll_pi16(a, count)1449
+#define _m_psllw(a, count) simde_mm_sll_pi16(a, count)1450
+#endif1451
+1452
+SIMDE_FUNCTION_ATTRIBUTES1453
+simde__m64 simde_mm_sll_pi32(simde__m64 a, simde__m64 count)1454
+{1455
+#if defined(SIMDE_X86_MMX_NATIVE)1456
+ return _mm_sll_pi32(a, count);1457
+#else1458
+ simde__m64_private r_;1459
+ simde__m64_private a_ = simde__m64_to_private(a);1460
+ simde__m64_private count_ = simde__m64_to_private(count);1461
+1462
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1463
+ HEDLEY_DIAGNOSTIC_PUSH1464
+#if HEDLEY_HAS_WARNING("-Wvector-conversion") && \1465
+ SIMDE_DETECT_CLANG_VERSION_NOT(10, 0, 0)1466
+#pragma clang diagnostic ignored "-Wvector-conversion"1467
+#endif1468
+ r_.neon_i32 =1469
+ vshl_s32(a_.neon_i32,1470
+ vmov_n_s32(HEDLEY_STATIC_CAST(1471
+ int32_t, vget_lane_u64(count_.neon_u64, 0))));1472
+ HEDLEY_DIAGNOSTIC_POP1473
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1474
+ r_.i32 = a_.i32 << count_.u64[0];1475
+#else1476
+ if (HEDLEY_UNLIKELY(count_.u64[0] > 31)) {1477
+ simde_memset(&r_, 0, sizeof(r_));1478
+ return simde__m64_from_private(r_);1479
+ }1480
+1481
+ SIMDE_VECTORIZE1482
+ for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {1483
+ r_.u32[i] = a_.u32[i] << count_.u64[0];1484
+ }1485
+#endif1486
+1487
+ return simde__m64_from_private(r_);1488
+#endif1489
+}1490
+#define simde_m_pslld(a, count) simde_mm_sll_pi32(a, count)1491
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1492
+#define _mm_sll_pi32(a, count) simde_mm_sll_pi32(a, count)1493
+#define _m_pslld(a, count) simde_mm_sll_pi32(a, count)1494
+#endif1495
+1496
+SIMDE_FUNCTION_ATTRIBUTES1497
+simde__m64 simde_mm_slli_pi16(simde__m64 a, int count)1498
+{1499
+#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1500
+ return _mm_slli_pi16(a, count);1501
+#else1502
+ simde__m64_private r_;1503
+ simde__m64_private a_ = simde__m64_to_private(a);1504
+1505
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \1506
+ defined(SIMDE_BUG_CLANG_POWER9_16x4_BAD_SHIFT)1507
+ if (HEDLEY_UNLIKELY(count > 15))1508
+ return simde_mm_setzero_si64();1509
+1510
+ r_.i16 = a_.i16 << HEDLEY_STATIC_CAST(int16_t, count);1511
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1512
+ r_.i16 = a_.i16 << count;1513
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1514
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1515
+ r_.neon_i16 = vshl_s16(a_.neon_i16, vmov_n_s16((int16_t)count));1516
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)1517
+ r_.mmi_i16 = psllh_s(a_.mmi_i16, b_.mmi_i16);1518
+#else1519
+ SIMDE_VECTORIZE1520
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {1521
+ r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, a_.u16[i] << count);1522
+ }1523
+#endif1524
+1525
+ return simde__m64_from_private(r_);1526
+#endif1527
+}1528
+#define simde_m_psllwi(a, count) simde_mm_slli_pi16(a, count)1529
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1530
+#define _mm_slli_pi16(a, count) simde_mm_slli_pi16(a, count)1531
+#define _m_psllwi(a, count) simde_mm_slli_pi16(a, count)1532
+#endif1533
+1534
+SIMDE_FUNCTION_ATTRIBUTES1535
+simde__m64 simde_mm_slli_pi32(simde__m64 a, int count)1536
+{1537
+#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1538
+ return _mm_slli_pi32(a, count);1539
+#else1540
+ simde__m64_private r_;1541
+ simde__m64_private a_ = simde__m64_to_private(a);1542
+1543
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1544
+ r_.i32 = a_.i32 << count;1545
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1546
+ r_.neon_i32 = vshl_s32(a_.neon_i32, vmov_n_s32((int32_t)count));1547
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)1548
+ r_.mmi_i32 = psllw_s(a_.mmi_i32, b_.mmi_i32);1549
+#else1550
+ SIMDE_VECTORIZE1551
+ for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {1552
+ r_.u32[i] = a_.u32[i] << count;1553
+ }1554
+#endif1555
+1556
+ return simde__m64_from_private(r_);1557
+#endif1558
+}1559
+#define simde_m_pslldi(a, b) simde_mm_slli_pi32(a, b)1560
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1561
+#define _mm_slli_pi32(a, count) simde_mm_slli_pi32(a, count)1562
+#define _m_pslldi(a, count) simde_mm_slli_pi32(a, count)1563
+#endif1564
+1565
+SIMDE_FUNCTION_ATTRIBUTES1566
+simde__m64 simde_mm_slli_si64(simde__m64 a, int count)1567
+{1568
+#if defined(SIMDE_X86_MMX_NATIVE)1569
+ return _mm_slli_si64(a, count);1570
+#else1571
+ simde__m64_private r_;1572
+ simde__m64_private a_ = simde__m64_to_private(a);1573
+1574
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1575
+ r_.i64 = a_.i64 << count;1576
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1577
+ r_.neon_i64 = vshl_s64(a_.neon_i64, vmov_n_s64((int64_t)count));1578
+#else1579
+ r_.u64[0] = a_.u64[0] << count;1580
+#endif1581
+1582
+ return simde__m64_from_private(r_);1583
+#endif1584
+}1585
+#define simde_m_psllqi(a, count) simde_mm_slli_si64(a, count)1586
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1587
+#define _mm_slli_si64(a, count) simde_mm_slli_si64(a, count)1588
+#define _m_psllqi(a, count) simde_mm_slli_si64(a, count)1589
+#endif1590
+1591
+SIMDE_FUNCTION_ATTRIBUTES1592
+simde__m64 simde_mm_sll_si64(simde__m64 a, simde__m64 count)1593
+{1594
+#if defined(SIMDE_X86_MMX_NATIVE)1595
+ return _mm_sll_si64(a, count);1596
+#else1597
+ simde__m64_private r_;1598
+ simde__m64_private a_ = simde__m64_to_private(a);1599
+ simde__m64_private count_ = simde__m64_to_private(count);1600
+1601
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1602
+ r_.neon_i64 = vshl_s64(a_.neon_i64, count_.neon_i64);1603
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1604
+ r_.i64 = a_.i64 << count_.i64;1605
+#else1606
+ if (HEDLEY_UNLIKELY(count_.u64[0] > 63)) {1607
+ simde_memset(&r_, 0, sizeof(r_));1608
+ return simde__m64_from_private(r_);1609
+ }1610
+1611
+ r_.u64[0] = a_.u64[0] << count_.u64[0];1612
+#endif1613
+1614
+ return simde__m64_from_private(r_);1615
+#endif1616
+}1617
+#define simde_m_psllq(a, count) simde_mm_sll_si64(a, count)1618
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1619
+#define _mm_sll_si64(a, count) simde_mm_sll_si64(a, count)1620
+#define _m_psllq(a, count) simde_mm_sll_si64(a, count)1621
+#endif1622
+1623
+SIMDE_FUNCTION_ATTRIBUTES1624
+simde__m64 simde_mm_srl_pi16(simde__m64 a, simde__m64 count)1625
+{1626
+#if defined(SIMDE_X86_MMX_NATIVE)1627
+ return _mm_srl_pi16(a, count);1628
+#else1629
+ simde__m64_private r_;1630
+ simde__m64_private a_ = simde__m64_to_private(a);1631
+ simde__m64_private count_ = simde__m64_to_private(count);1632
+1633
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \1634
+ defined(SIMDE_BUG_CLANG_POWER9_16x4_BAD_SHIFT)1635
+ if (HEDLEY_UNLIKELY(count_.u64[0] > 15))1636
+ return simde_mm_setzero_si64();1637
+1638
+ r_.i16 = a_.i16 >> HEDLEY_STATIC_CAST(int16_t, count_.u64[0]);1639
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1640
+ r_.u16 = a_.u16 >> count_.u64[0];1641
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1642
+ r_.neon_u16 = vshl_u16(1643
+ a_.neon_u16,1644
+ vmov_n_s16(-((int16_t)vget_lane_u64(count_.neon_u64, 0))));1645
+#else1646
+ if (HEDLEY_UNLIKELY(count_.u64[0] > 15)) {1647
+ simde_memset(&r_, 0, sizeof(r_));1648
+ return simde__m64_from_private(r_);1649
+ }1650
+1651
+ SIMDE_VECTORIZE1652
+ for (size_t i = 0; i < sizeof(r_.u16) / sizeof(r_.u16[0]); i++) {1653
+ r_.u16[i] = a_.u16[i] >> count_.u64[0];1654
+ }1655
+#endif1656
+1657
+ return simde__m64_from_private(r_);1658
+#endif1659
+}1660
+#define simde_m_psrlw(a, count) simde_mm_srl_pi16(a, count)1661
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1662
+#define _mm_srl_pi16(a, count) simde_mm_srl_pi16(a, count)1663
+#define _m_psrlw(a, count) simde_mm_srl_pi16(a, count)1664
+#endif1665
+1666
+SIMDE_FUNCTION_ATTRIBUTES1667
+simde__m64 simde_mm_srl_pi32(simde__m64 a, simde__m64 count)1668
+{1669
+#if defined(SIMDE_X86_MMX_NATIVE)1670
+ return _mm_srl_pi32(a, count);1671
+#else1672
+ simde__m64_private r_;1673
+ simde__m64_private a_ = simde__m64_to_private(a);1674
+ simde__m64_private count_ = simde__m64_to_private(count);1675
+1676
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1677
+ r_.u32 = a_.u32 >> count_.u64[0];1678
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1679
+ r_.neon_u32 = vshl_u32(1680
+ a_.neon_u32,1681
+ vmov_n_s32(-((int32_t)vget_lane_u64(count_.neon_u64, 0))));1682
+#else1683
+ if (HEDLEY_UNLIKELY(count_.u64[0] > 31)) {1684
+ simde_memset(&r_, 0, sizeof(r_));1685
+ return simde__m64_from_private(r_);1686
+ }1687
+1688
+ SIMDE_VECTORIZE1689
+ for (size_t i = 0; i < sizeof(r_.u32) / sizeof(r_.u32[0]); i++) {1690
+ r_.u32[i] = a_.u32[i] >> count_.u64[0];1691
+ }1692
+#endif1693
+1694
+ return simde__m64_from_private(r_);1695
+#endif1696
+}1697
+#define simde_m_psrld(a, count) simde_mm_srl_pi32(a, count)1698
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1699
+#define _mm_srl_pi32(a, count) simde_mm_srl_pi32(a, count)1700
+#define _m_psrld(a, count) simde_mm_srl_pi32(a, count)1701
+#endif1702
+1703
+SIMDE_FUNCTION_ATTRIBUTES1704
+simde__m64 simde_mm_srli_pi16(simde__m64 a, int count)1705
+{1706
+#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1707
+ return _mm_srli_pi16(a, count);1708
+#else1709
+ simde__m64_private r_;1710
+ simde__m64_private a_ = simde__m64_to_private(a);1711
+1712
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1713
+ r_.u16 = a_.u16 >> count;1714
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1715
+ r_.neon_u16 = vshl_u16(a_.neon_u16, vmov_n_s16(-((int16_t)count)));1716
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)1717
+ r_.mmi_i16 = psrlh_s(a_.mmi_i16, b_.mmi_i16);1718
+#else1719
+ SIMDE_VECTORIZE1720
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {1721
+ r_.u16[i] = a_.u16[i] >> count;1722
+ }1723
+#endif1724
+1725
+ return simde__m64_from_private(r_);1726
+#endif1727
+}1728
+#define simde_m_psrlwi(a, count) simde_mm_srli_pi16(a, count)1729
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1730
+#define _mm_srli_pi16(a, count) simde_mm_srli_pi16(a, count)1731
+#define _m_psrlwi(a, count) simde_mm_srli_pi16(a, count)1732
+#endif1733
+1734
+SIMDE_FUNCTION_ATTRIBUTES1735
+simde__m64 simde_mm_srli_pi32(simde__m64 a, int count)1736
+{1737
+#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1738
+ return _mm_srli_pi32(a, count);1739
+#else1740
+ simde__m64_private r_;1741
+ simde__m64_private a_ = simde__m64_to_private(a);1742
+1743
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1744
+ r_.u32 = a_.u32 >> count;1745
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1746
+ r_.neon_u32 = vshl_u32(a_.neon_u32, vmov_n_s32(-((int32_t)count)));1747
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)1748
+ r_.mmi_i32 = psrlw_s(a_.mmi_i32, b_.mmi_i32);1749
+#else1750
+ SIMDE_VECTORIZE1751
+ for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {1752
+ r_.u32[i] = a_.u32[i] >> count;1753
+ }1754
+#endif1755
+1756
+ return simde__m64_from_private(r_);1757
+#endif1758
+}1759
+#define simde_m_psrldi(a, count) simde_mm_srli_pi32(a, count)1760
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1761
+#define _mm_srli_pi32(a, count) simde_mm_srli_pi32(a, count)1762
+#define _m_psrldi(a, count) simde_mm_srli_pi32(a, count)1763
+#endif1764
+1765
+SIMDE_FUNCTION_ATTRIBUTES1766
+simde__m64 simde_mm_srli_si64(simde__m64 a, int count)1767
+{1768
+#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1769
+ return _mm_srli_si64(a, count);1770
+#else1771
+ simde__m64_private r_;1772
+ simde__m64_private a_ = simde__m64_to_private(a);1773
+1774
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1775
+ r_.neon_u64 = vshl_u64(a_.neon_u64, vmov_n_s64(-count));1776
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1777
+ r_.u64 = a_.u64 >> count;1778
+#else1779
+ r_.u64[0] = a_.u64[0] >> count;1780
+#endif1781
+1782
+ return simde__m64_from_private(r_);1783
+#endif1784
+}1785
+#define simde_m_psrlqi(a, count) simde_mm_srli_si64(a, count)1786
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1787
+#define _mm_srli_si64(a, count) simde_mm_srli_si64(a, count)1788
+#define _m_psrlqi(a, count) simde_mm_srli_si64(a, count)1789
+#endif1790
+1791
+SIMDE_FUNCTION_ATTRIBUTES1792
+simde__m64 simde_mm_srl_si64(simde__m64 a, simde__m64 count)1793
+{1794
+#if defined(SIMDE_X86_MMX_NATIVE)1795
+ return _mm_srl_si64(a, count);1796
+#else1797
+ simde__m64_private r_;1798
+ simde__m64_private a_ = simde__m64_to_private(a);1799
+ simde__m64_private count_ = simde__m64_to_private(count);1800
+1801
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1802
+ r_.neon_u64 = vshl_u64(a_.neon_u64, vneg_s64(count_.neon_i64));1803
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1804
+ r_.u64 = a_.u64 >> count_.u64;1805
+#else1806
+ if (HEDLEY_UNLIKELY(count_.u64[0] > 63)) {1807
+ simde_memset(&r_, 0, sizeof(r_));1808
+ return simde__m64_from_private(r_);1809
+ }1810
+1811
+ r_.u64[0] = a_.u64[0] >> count_.u64[0];1812
+#endif1813
+1814
+ return simde__m64_from_private(r_);1815
+#endif1816
+}1817
+#define simde_m_psrlq(a, count) simde_mm_srl_si64(a, count)1818
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1819
+#define _mm_srl_si64(a, count) simde_mm_srl_si64(a, count)1820
+#define _m_psrlq(a, count) simde_mm_srl_si64(a, count)1821
+#endif1822
+1823
+SIMDE_FUNCTION_ATTRIBUTES1824
+simde__m64 simde_mm_srai_pi16(simde__m64 a, int count)1825
+{1826
+#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1827
+ return _mm_srai_pi16(a, count);1828
+#else1829
+ simde__m64_private r_;1830
+ simde__m64_private a_ = simde__m64_to_private(a);1831
+1832
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1833
+ r_.i16 = a_.i16 >> (count & 0xff);1834
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1835
+ r_.neon_i16 = vshl_s16(a_.neon_i16,1836
+ vmov_n_s16(-HEDLEY_STATIC_CAST(int16_t, count)));1837
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)1838
+ r_.mmi_i16 = psrah_s(a_.mmi_i16, count);1839
+#else1840
+ SIMDE_VECTORIZE1841
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1842
+ r_.i16[i] = a_.i16[i] >> (count & 0xff);1843
+ }1844
+#endif1845
+1846
+ return simde__m64_from_private(r_);1847
+#endif1848
+}1849
+#define simde_m_psrawi(a, count) simde_mm_srai_pi16(a, count)1850
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1851
+#define _mm_srai_pi16(a, count) simde_mm_srai_pi16(a, count)1852
+#define _m_psrawi(a, count) simde_mm_srai_pi16(a, count)1853
+#endif1854
+1855
+SIMDE_FUNCTION_ATTRIBUTES1856
+simde__m64 simde_mm_srai_pi32(simde__m64 a, int count)1857
+{1858
+#if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)1859
+ return _mm_srai_pi32(a, count);1860
+#else1861
+ simde__m64_private r_;1862
+ simde__m64_private a_ = simde__m64_to_private(a);1863
+1864
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1865
+ r_.i32 = a_.i32 >> (count & 0xff);1866
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1867
+ r_.neon_i32 = vshl_s32(a_.neon_i32,1868
+ vmov_n_s32(-HEDLEY_STATIC_CAST(int32_t, count)));1869
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)1870
+ r_.mmi_i32 = psraw_s(a_.mmi_i32, count);1871
+#else1872
+ SIMDE_VECTORIZE1873
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1874
+ r_.i32[i] = a_.i32[i] >> (count & 0xff);1875
+ }1876
+#endif1877
+1878
+ return simde__m64_from_private(r_);1879
+#endif1880
+}1881
+#define simde_m_psradi(a, count) simde_mm_srai_pi32(a, count)1882
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1883
+#define _mm_srai_pi32(a, count) simde_mm_srai_pi32(a, count)1884
+#define _m_psradi(a, count) simde_mm_srai_pi32(a, count)1885
+#endif1886
+1887
+SIMDE_FUNCTION_ATTRIBUTES1888
+simde__m64 simde_mm_sra_pi16(simde__m64 a, simde__m64 count)1889
+{1890
+#if defined(SIMDE_X86_MMX_NATIVE)1891
+ return _mm_sra_pi16(a, count);1892
+#else1893
+ simde__m64_private r_;1894
+ simde__m64_private a_ = simde__m64_to_private(a);1895
+ simde__m64_private count_ = simde__m64_to_private(count);1896
+ const int cnt = HEDLEY_STATIC_CAST(1897
+ int, (count_.i64[0] > 15 ? 15 : count_.i64[0]));1898
+1899
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1900
+ r_.i16 = a_.i16 >> cnt;1901
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1902
+ r_.neon_i16 =1903
+ vshl_s16(a_.neon_i16,1904
+ vmov_n_s16(-HEDLEY_STATIC_CAST(1905
+ int16_t, vget_lane_u64(count_.neon_u64, 0))));1906
+#else1907
+ SIMDE_VECTORIZE1908
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1909
+ r_.i16[i] = a_.i16[i] >> cnt;1910
+ }1911
+#endif1912
+1913
+ return simde__m64_from_private(r_);1914
+#endif1915
+}1916
+#define simde_m_psraw(a, count) simde_mm_sra_pi16(a, count)1917
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1918
+#define _mm_sra_pi16(a, count) simde_mm_sra_pi16(a, count)1919
+#define _m_psraw(a, count) simde_mm_sra_pi16(a, count)1920
+#endif1921
+1922
+SIMDE_FUNCTION_ATTRIBUTES1923
+simde__m64 simde_mm_sra_pi32(simde__m64 a, simde__m64 count)1924
+{1925
+#if defined(SIMDE_X86_MMX_NATIVE)1926
+ return _mm_sra_pi32(a, count);1927
+#else1928
+ simde__m64_private r_;1929
+ simde__m64_private a_ = simde__m64_to_private(a);1930
+ simde__m64_private count_ = simde__m64_to_private(count);1931
+ const int32_t cnt =1932
+ (count_.u64[0] > 31)1933
+ ? 311934
+ : HEDLEY_STATIC_CAST(int32_t, count_.u64[0]);1935
+1936
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)1937
+ r_.i32 = a_.i32 >> cnt;1938
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1939
+ r_.neon_i32 =1940
+ vshl_s32(a_.neon_i32,1941
+ vmov_n_s32(-HEDLEY_STATIC_CAST(1942
+ int32_t, vget_lane_u64(count_.neon_u64, 0))));1943
+#else1944
+ SIMDE_VECTORIZE1945
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1946
+ r_.i32[i] = a_.i32[i] >> cnt;1947
+ }1948
+#endif1949
+1950
+ return simde__m64_from_private(r_);1951
+#endif1952
+}1953
+#define simde_m_psrad(a, b) simde_mm_sra_pi32(a, b)1954
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1955
+#define _mm_sra_pi32(a, count) simde_mm_sra_pi32(a, count)1956
+#define _m_psrad(a, count) simde_mm_sra_pi32(a, count)1957
+#endif1958
+1959
+SIMDE_FUNCTION_ATTRIBUTES1960
+simde__m64 simde_mm_sub_pi8(simde__m64 a, simde__m64 b)1961
+{1962
+#if defined(SIMDE_X86_MMX_NATIVE)1963
+ return _mm_sub_pi8(a, b);1964
+#else1965
+ simde__m64_private r_;1966
+ simde__m64_private a_ = simde__m64_to_private(a);1967
+ simde__m64_private b_ = simde__m64_to_private(b);1968
+1969
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1970
+ r_.neon_i8 = vsub_s8(a_.neon_i8, b_.neon_i8);1971
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)1972
+ r_.mmi_i8 = psubb_s(a_.mmi_i8, b_.mmi_i8);1973
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1974
+ r_.i8 = a_.i8 - b_.i8;1975
+#else1976
+ SIMDE_VECTORIZE1977
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1978
+ r_.i8[i] = a_.i8[i] - b_.i8[i];1979
+ }1980
+#endif1981
+1982
+ return simde__m64_from_private(r_);1983
+#endif1984
+}1985
+#define simde_m_psubb(a, b) simde_mm_sub_pi8(a, b)1986
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)1987
+#define _mm_sub_pi8(a, b) simde_mm_sub_pi8(a, b)1988
+#define _m_psubb(a, b) simde_mm_sub_pi8(a, b)1989
+#endif1990
+1991
+SIMDE_FUNCTION_ATTRIBUTES1992
+simde__m64 simde_mm_sub_pi16(simde__m64 a, simde__m64 b)1993
+{1994
+#if defined(SIMDE_X86_MMX_NATIVE)1995
+ return _mm_sub_pi16(a, b);1996
+#else1997
+ simde__m64_private r_;1998
+ simde__m64_private a_ = simde__m64_to_private(a);1999
+ simde__m64_private b_ = simde__m64_to_private(b);2000
+2001
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2002
+ r_.neon_i16 = vsub_s16(a_.neon_i16, b_.neon_i16);2003
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2004
+ r_.mmi_i16 = psubh_s(a_.mmi_i16, b_.mmi_i16);2005
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2006
+ r_.i16 = a_.i16 - b_.i16;2007
+#else2008
+ SIMDE_VECTORIZE2009
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {2010
+ r_.i16[i] = a_.i16[i] - b_.i16[i];2011
+ }2012
+#endif2013
+2014
+ return simde__m64_from_private(r_);2015
+#endif2016
+}2017
+#define simde_m_psubw(a, b) simde_mm_sub_pi16(a, b)2018
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2019
+#define _mm_sub_pi16(a, b) simde_mm_sub_pi16(a, b)2020
+#define _m_psubw(a, b) simde_mm_sub_pi16(a, b)2021
+#endif2022
+2023
+SIMDE_FUNCTION_ATTRIBUTES2024
+simde__m64 simde_mm_sub_pi32(simde__m64 a, simde__m64 b)2025
+{2026
+#if defined(SIMDE_X86_MMX_NATIVE)2027
+ return _mm_sub_pi32(a, b);2028
+#else2029
+ simde__m64_private r_;2030
+ simde__m64_private a_ = simde__m64_to_private(a);2031
+ simde__m64_private b_ = simde__m64_to_private(b);2032
+2033
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2034
+ r_.neon_i32 = vsub_s32(a_.neon_i32, b_.neon_i32);2035
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2036
+ r_.mmi_i32 = psubw_s(a_.mmi_i32, b_.mmi_i32);2037
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2038
+ r_.i32 = a_.i32 - b_.i32;2039
+#else2040
+ SIMDE_VECTORIZE2041
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {2042
+ r_.i32[i] = a_.i32[i] - b_.i32[i];2043
+ }2044
+#endif2045
+2046
+ return simde__m64_from_private(r_);2047
+#endif2048
+}2049
+#define simde_m_psubd(a, b) simde_mm_sub_pi32(a, b)2050
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2051
+#define _mm_sub_pi32(a, b) simde_mm_sub_pi32(a, b)2052
+#define _m_psubd(a, b) simde_mm_sub_pi32(a, b)2053
+#endif2054
+2055
+SIMDE_FUNCTION_ATTRIBUTES2056
+simde__m64 simde_mm_subs_pi8(simde__m64 a, simde__m64 b)2057
+{2058
+#if defined(SIMDE_X86_MMX_NATIVE)2059
+ return _mm_subs_pi8(a, b);2060
+#else2061
+ simde__m64_private r_;2062
+ simde__m64_private a_ = simde__m64_to_private(a);2063
+ simde__m64_private b_ = simde__m64_to_private(b);2064
+2065
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2066
+ r_.neon_i8 = vqsub_s8(a_.neon_i8, b_.neon_i8);2067
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2068
+ r_.mmi_i8 = psubsb(a_.mmi_i8, b_.mmi_i8);2069
+#else2070
+ SIMDE_VECTORIZE2071
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {2072
+ if (((b_.i8[i]) > 0 && (a_.i8[i]) < INT8_MIN + (b_.i8[i]))) {2073
+ r_.i8[i] = INT8_MIN;2074
+ } else if ((b_.i8[i]) < 0 &&2075
+ (a_.i8[i]) > INT8_MAX + (b_.i8[i])) {2076
+ r_.i8[i] = INT8_MAX;2077
+ } else {2078
+ r_.i8[i] = (a_.i8[i]) - (b_.i8[i]);2079
+ }2080
+ }2081
+#endif2082
+2083
+ return simde__m64_from_private(r_);2084
+#endif2085
+}2086
+#define simde_m_psubsb(a, b) simde_mm_subs_pi8(a, b)2087
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2088
+#define _mm_subs_pi8(a, b) simde_mm_subs_pi8(a, b)2089
+#define _m_psubsb(a, b) simde_mm_subs_pi8(a, b)2090
+#endif2091
+2092
+SIMDE_FUNCTION_ATTRIBUTES2093
+simde__m64 simde_mm_subs_pu8(simde__m64 a, simde__m64 b)2094
+{2095
+#if defined(SIMDE_X86_MMX_NATIVE)2096
+ return _mm_subs_pu8(a, b);2097
+#else2098
+ simde__m64_private r_;2099
+ simde__m64_private a_ = simde__m64_to_private(a);2100
+ simde__m64_private b_ = simde__m64_to_private(b);2101
+2102
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2103
+ r_.neon_u8 = vqsub_u8(a_.neon_u8, b_.neon_u8);2104
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2105
+ r_.mmi_u8 = psubusb(a_.mmi_u8, b_.mmi_u8);2106
+#else2107
+ SIMDE_VECTORIZE2108
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {2109
+ const int32_t x = a_.u8[i] - b_.u8[i];2110
+ if (x < 0) {2111
+ r_.u8[i] = 0;2112
+ } else if (x > UINT8_MAX) {2113
+ r_.u8[i] = UINT8_MAX;2114
+ } else {2115
+ r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, x);2116
+ }2117
+ }2118
+#endif2119
+2120
+ return simde__m64_from_private(r_);2121
+#endif2122
+}2123
+#define simde_m_psubusb(a, b) simde_mm_subs_pu8(a, b)2124
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2125
+#define _mm_subs_pu8(a, b) simde_mm_subs_pu8(a, b)2126
+#define _m_psubusb(a, b) simde_mm_subs_pu8(a, b)2127
+#endif2128
+2129
+SIMDE_FUNCTION_ATTRIBUTES2130
+simde__m64 simde_mm_subs_pi16(simde__m64 a, simde__m64 b)2131
+{2132
+#if defined(SIMDE_X86_MMX_NATIVE)2133
+ return _mm_subs_pi16(a, b);2134
+#else2135
+ simde__m64_private r_;2136
+ simde__m64_private a_ = simde__m64_to_private(a);2137
+ simde__m64_private b_ = simde__m64_to_private(b);2138
+2139
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2140
+ r_.neon_i16 = vqsub_s16(a_.neon_i16, b_.neon_i16);2141
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2142
+ r_.mmi_i16 = psubsh(a_.mmi_i16, b_.mmi_i16);2143
+#else2144
+ SIMDE_VECTORIZE2145
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {2146
+ if (((b_.i16[i]) > 0 && (a_.i16[i]) < SHRT_MIN + (b_.i16[i]))) {2147
+ r_.i16[i] = SHRT_MIN;2148
+ } else if ((b_.i16[i]) < 0 &&2149
+ (a_.i16[i]) > INT16_MAX + (b_.i16[i])) {2150
+ r_.i16[i] = INT16_MAX;2151
+ } else {2152
+ r_.i16[i] = (a_.i16[i]) - (b_.i16[i]);2153
+ }2154
+ }2155
+#endif2156
+2157
+ return simde__m64_from_private(r_);2158
+#endif2159
+}2160
+#define simde_m_psubsw(a, b) simde_mm_subs_pi16(a, b)2161
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2162
+#define _mm_subs_pi16(a, b) simde_mm_subs_pi16(a, b)2163
+#define _m_psubsw(a, b) simde_mm_subs_pi16(a, b)2164
+#endif2165
+2166
+SIMDE_FUNCTION_ATTRIBUTES2167
+simde__m64 simde_mm_subs_pu16(simde__m64 a, simde__m64 b)2168
+{2169
+#if defined(SIMDE_X86_MMX_NATIVE)2170
+ return _mm_subs_pu16(a, b);2171
+#else2172
+ simde__m64_private r_;2173
+ simde__m64_private a_ = simde__m64_to_private(a);2174
+ simde__m64_private b_ = simde__m64_to_private(b);2175
+2176
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2177
+ r_.neon_u16 = vqsub_u16(a_.neon_u16, b_.neon_u16);2178
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2179
+ r_.mmi_u16 = psubush(a_.mmi_u16, b_.mmi_u16);2180
+#else2181
+ SIMDE_VECTORIZE2182
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {2183
+ const int x = a_.u16[i] - b_.u16[i];2184
+ if (x < 0) {2185
+ r_.u16[i] = 0;2186
+ } else if (x > UINT16_MAX) {2187
+ r_.u16[i] = UINT16_MAX;2188
+ } else {2189
+ r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, x);2190
+ }2191
+ }2192
+#endif2193
+2194
+ return simde__m64_from_private(r_);2195
+#endif2196
+}2197
+#define simde_m_psubusw(a, b) simde_mm_subs_pu16(a, b)2198
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2199
+#define _mm_subs_pu16(a, b) simde_mm_subs_pu16(a, b)2200
+#define _m_psubusw(a, b) simde_mm_subs_pu16(a, b)2201
+#endif2202
+2203
+SIMDE_FUNCTION_ATTRIBUTES2204
+simde__m64 simde_mm_unpackhi_pi8(simde__m64 a, simde__m64 b)2205
+{2206
+#if defined(SIMDE_X86_MMX_NATIVE)2207
+ return _mm_unpackhi_pi8(a, b);2208
+#else2209
+ simde__m64_private r_;2210
+ simde__m64_private a_ = simde__m64_to_private(a);2211
+ simde__m64_private b_ = simde__m64_to_private(b);2212
+2213
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2214
+ r_.neon_i8 = vzip2_s8(a_.neon_i8, b_.neon_i8);2215
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2216
+ r_.i8 = SIMDE_SHUFFLE_VECTOR_(8, 8, a_.i8, b_.i8, 4, 12, 5, 13, 6, 14,2217
+ 7, 15);2218
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2219
+ r_.mmi_i8 = punpckhbh_s(a_.mmi_i8, b_.mmi_i8);2220
+#else2221
+ r_.i8[0] = a_.i8[4];2222
+ r_.i8[1] = b_.i8[4];2223
+ r_.i8[2] = a_.i8[5];2224
+ r_.i8[3] = b_.i8[5];2225
+ r_.i8[4] = a_.i8[6];2226
+ r_.i8[5] = b_.i8[6];2227
+ r_.i8[6] = a_.i8[7];2228
+ r_.i8[7] = b_.i8[7];2229
+#endif2230
+2231
+ return simde__m64_from_private(r_);2232
+#endif2233
+}2234
+#define simde_m_punpckhbw(a, b) simde_mm_unpackhi_pi8(a, b)2235
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2236
+#define _mm_unpackhi_pi8(a, b) simde_mm_unpackhi_pi8(a, b)2237
+#define _m_punpckhbw(a, b) simde_mm_unpackhi_pi8(a, b)2238
+#endif2239
+2240
+SIMDE_FUNCTION_ATTRIBUTES2241
+simde__m64 simde_mm_unpackhi_pi16(simde__m64 a, simde__m64 b)2242
+{2243
+#if defined(SIMDE_X86_MMX_NATIVE)2244
+ return _mm_unpackhi_pi16(a, b);2245
+#else2246
+ simde__m64_private r_;2247
+ simde__m64_private a_ = simde__m64_to_private(a);2248
+ simde__m64_private b_ = simde__m64_to_private(b);2249
+2250
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2251
+ r_.neon_i16 = vzip2_s16(a_.neon_i16, b_.neon_i16);2252
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2253
+ r_.mmi_i16 = punpckhhw_s(a_.mmi_i16, b_.mmi_i16);2254
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2255
+ r_.i16 = SIMDE_SHUFFLE_VECTOR_(16, 8, a_.i16, b_.i16, 2, 6, 3, 7);2256
+#else2257
+ r_.i16[0] = a_.i16[2];2258
+ r_.i16[1] = b_.i16[2];2259
+ r_.i16[2] = a_.i16[3];2260
+ r_.i16[3] = b_.i16[3];2261
+#endif2262
+2263
+ return simde__m64_from_private(r_);2264
+#endif2265
+}2266
+#define simde_m_punpckhwd(a, b) simde_mm_unpackhi_pi16(a, b)2267
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2268
+#define _mm_unpackhi_pi16(a, b) simde_mm_unpackhi_pi16(a, b)2269
+#define _m_punpckhwd(a, b) simde_mm_unpackhi_pi16(a, b)2270
+#endif2271
+2272
+SIMDE_FUNCTION_ATTRIBUTES2273
+simde__m64 simde_mm_unpackhi_pi32(simde__m64 a, simde__m64 b)2274
+{2275
+#if defined(SIMDE_X86_MMX_NATIVE)2276
+ return _mm_unpackhi_pi32(a, b);2277
+#else2278
+ simde__m64_private r_;2279
+ simde__m64_private a_ = simde__m64_to_private(a);2280
+ simde__m64_private b_ = simde__m64_to_private(b);2281
+2282
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2283
+ r_.neon_i32 = vzip2_s32(a_.neon_i32, b_.neon_i32);2284
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2285
+ r_.mmi_i32 = punpckhwd_s(a_.mmi_i32, b_.mmi_i32);2286
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2287
+ r_.i32 = SIMDE_SHUFFLE_VECTOR_(32, 8, a_.i32, b_.i32, 1, 3);2288
+#else2289
+ r_.i32[0] = a_.i32[1];2290
+ r_.i32[1] = b_.i32[1];2291
+#endif2292
+2293
+ return simde__m64_from_private(r_);2294
+#endif2295
+}2296
+#define simde_m_punpckhdq(a, b) simde_mm_unpackhi_pi32(a, b)2297
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2298
+#define _mm_unpackhi_pi32(a, b) simde_mm_unpackhi_pi32(a, b)2299
+#define _m_punpckhdq(a, b) simde_mm_unpackhi_pi32(a, b)2300
+#endif2301
+2302
+SIMDE_FUNCTION_ATTRIBUTES2303
+simde__m64 simde_mm_unpacklo_pi8(simde__m64 a, simde__m64 b)2304
+{2305
+#if defined(SIMDE_X86_MMX_NATIVE)2306
+ return _mm_unpacklo_pi8(a, b);2307
+#else2308
+ simde__m64_private r_;2309
+ simde__m64_private a_ = simde__m64_to_private(a);2310
+ simde__m64_private b_ = simde__m64_to_private(b);2311
+2312
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2313
+ r_.neon_i8 = vzip1_s8(a_.neon_i8, b_.neon_i8);2314
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2315
+ r_.mmi_i8 = punpcklbh_s(a_.mmi_i8, b_.mmi_i8);2316
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2317
+ r_.i8 = SIMDE_SHUFFLE_VECTOR_(8, 8, a_.i8, b_.i8, 0, 8, 1, 9, 2, 10, 3,2318
+ 11);2319
+#else2320
+ r_.i8[0] = a_.i8[0];2321
+ r_.i8[1] = b_.i8[0];2322
+ r_.i8[2] = a_.i8[1];2323
+ r_.i8[3] = b_.i8[1];2324
+ r_.i8[4] = a_.i8[2];2325
+ r_.i8[5] = b_.i8[2];2326
+ r_.i8[6] = a_.i8[3];2327
+ r_.i8[7] = b_.i8[3];2328
+#endif2329
+2330
+ return simde__m64_from_private(r_);2331
+#endif2332
+}2333
+#define simde_m_punpcklbw(a, b) simde_mm_unpacklo_pi8(a, b)2334
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2335
+#define _mm_unpacklo_pi8(a, b) simde_mm_unpacklo_pi8(a, b)2336
+#define _m_punpcklbw(a, b) simde_mm_unpacklo_pi8(a, b)2337
+#endif2338
+2339
+SIMDE_FUNCTION_ATTRIBUTES2340
+simde__m64 simde_mm_unpacklo_pi16(simde__m64 a, simde__m64 b)2341
+{2342
+#if defined(SIMDE_X86_MMX_NATIVE)2343
+ return _mm_unpacklo_pi16(a, b);2344
+#else2345
+ simde__m64_private r_;2346
+ simde__m64_private a_ = simde__m64_to_private(a);2347
+ simde__m64_private b_ = simde__m64_to_private(b);2348
+2349
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2350
+ r_.neon_i16 = vzip1_s16(a_.neon_i16, b_.neon_i16);2351
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2352
+ r_.mmi_i16 = punpcklhw_s(a_.mmi_i16, b_.mmi_i16);2353
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2354
+ r_.i16 = SIMDE_SHUFFLE_VECTOR_(16, 8, a_.i16, b_.i16, 0, 4, 1, 5);2355
+#else2356
+ r_.i16[0] = a_.i16[0];2357
+ r_.i16[1] = b_.i16[0];2358
+ r_.i16[2] = a_.i16[1];2359
+ r_.i16[3] = b_.i16[1];2360
+#endif2361
+2362
+ return simde__m64_from_private(r_);2363
+#endif2364
+}2365
+#define simde_m_punpcklwd(a, b) simde_mm_unpacklo_pi16(a, b)2366
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2367
+#define _mm_unpacklo_pi16(a, b) simde_mm_unpacklo_pi16(a, b)2368
+#define _m_punpcklwd(a, b) simde_mm_unpacklo_pi16(a, b)2369
+#endif2370
+2371
+SIMDE_FUNCTION_ATTRIBUTES2372
+simde__m64 simde_mm_unpacklo_pi32(simde__m64 a, simde__m64 b)2373
+{2374
+#if defined(SIMDE_X86_MMX_NATIVE)2375
+ return _mm_unpacklo_pi32(a, b);2376
+#else2377
+ simde__m64_private r_;2378
+ simde__m64_private a_ = simde__m64_to_private(a);2379
+ simde__m64_private b_ = simde__m64_to_private(b);2380
+2381
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2382
+ r_.neon_i32 = vzip1_s32(a_.neon_i32, b_.neon_i32);2383
+#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)2384
+ r_.mmi_i32 = punpcklwd_s(a_.mmi_i32, b_.mmi_i32);2385
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2386
+ r_.i32 = SIMDE_SHUFFLE_VECTOR_(32, 8, a_.i32, b_.i32, 0, 2);2387
+#else2388
+ r_.i32[0] = a_.i32[0];2389
+ r_.i32[1] = b_.i32[0];2390
+#endif2391
+2392
+ return simde__m64_from_private(r_);2393
+#endif2394
+}2395
+#define simde_m_punpckldq(a, b) simde_mm_unpacklo_pi32(a, b)2396
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2397
+#define _mm_unpacklo_pi32(a, b) simde_mm_unpacklo_pi32(a, b)2398
+#define _m_punpckldq(a, b) simde_mm_unpacklo_pi32(a, b)2399
+#endif2400
+2401
+SIMDE_FUNCTION_ATTRIBUTES2402
+simde__m64 simde_mm_xor_si64(simde__m64 a, simde__m64 b)2403
+{2404
+#if defined(SIMDE_X86_MMX_NATIVE)2405
+ return _mm_xor_si64(a, b);2406
+#else2407
+ simde__m64_private r_;2408
+ simde__m64_private a_ = simde__m64_to_private(a);2409
+ simde__m64_private b_ = simde__m64_to_private(b);2410
+2411
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2412
+ r_.neon_i32 = veor_s32(a_.neon_i32, b_.neon_i32);2413
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2414
+ r_.i32f = a_.i32f ^ b_.i32f;2415
+#else2416
+ r_.u64[0] = a_.u64[0] ^ b_.u64[0];2417
+#endif2418
+2419
+ return simde__m64_from_private(r_);2420
+#endif2421
+}2422
+#define simde_m_pxor(a, b) simde_mm_xor_si64(a, b)2423
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2424
+#define _mm_xor_si64(a, b) simde_mm_xor_si64(a, b)2425
+#define _m_pxor(a, b) simde_mm_xor_si64(a, b)2426
+#endif2427
+2428
+SIMDE_FUNCTION_ATTRIBUTES2429
+int32_t simde_m_to_int(simde__m64 a)2430
+{2431
+#if defined(SIMDE_X86_MMX_NATIVE)2432
+ return _m_to_int(a);2433
+#else2434
+ simde__m64_private a_ = simde__m64_to_private(a);2435
+2436
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2437
+ HEDLEY_DIAGNOSTIC_PUSH2438
+#if HEDLEY_HAS_WARNING("-Wvector-conversion") && \2439
+ SIMDE_DETECT_CLANG_VERSION_NOT(10, 0, 0)2440
+#pragma clang diagnostic ignored "-Wvector-conversion"2441
+#endif2442
+ return vget_lane_s32(a_.neon_i32, 0);2443
+ HEDLEY_DIAGNOSTIC_POP2444
+#else2445
+ return a_.i32[0];2446
+#endif2447
+#endif2448
+}2449
+#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)2450
+#define _m_to_int(a) simde_m_to_int(a)2451
+#endif2452
+2453
+SIMDE_END_DECLS_2454
+2455
+HEDLEY_DIAGNOSTIC_POP2456
+2457
+#endif /* !defined(SIMDE_X86_MMX_H) */2458
obs-studio-26.1.1.tar.xz/libobs/util/simde/x86/sse.h
Added
4481
1
2
+/* SPDX-License-Identifier: MIT3
+ *4
+ * Permission is hereby granted, free of charge, to any person5
+ * obtaining a copy of this software and associated documentation6
+ * files (the "Software"), to deal in the Software without7
+ * restriction, including without limitation the rights to use, copy,8
+ * modify, merge, publish, distribute, sublicense, and/or sell copies9
+ * of the Software, and to permit persons to whom the Software is10
+ * furnished to do so, subject to the following conditions:11
+ *12
+ * The above copyright notice and this permission notice shall be13
+ * included in all copies or substantial portions of the Software.14
+ *15
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,16
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF17
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND18
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS19
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN20
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN21
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE22
+ * SOFTWARE.23
+ *24
+ * Copyright:25
+ * 2017-2020 Evan Nemerson <evan@nemerson.com>26
+ * 2015-2017 John W. Ratcliff <jratcliffscarab@gmail.com>27
+ * 2015 Brandon Rowlett <browlett@nvidia.com>28
+ * 2015 Ken Fast <kfast@gdeb.com>29
+ */30
+31
+#if !defined(SIMDE_X86_SSE_H)32
+#define SIMDE_X86_SSE_H33
+34
+#include "mmx.h"35
+36
+#if defined(_WIN32)37
+#include <windows.h>38
+#endif39
+40
+HEDLEY_DIAGNOSTIC_PUSH41
+SIMDE_DISABLE_UNWANTED_DIAGNOSTICS42
+SIMDE_BEGIN_DECLS_43
+44
+typedef union {45
+#if defined(SIMDE_VECTOR_SUBSCRIPT)46
+ SIMDE_ALIGN_TO_16 int8_t i8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;47
+ SIMDE_ALIGN_TO_16 int16_t i16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;48
+ SIMDE_ALIGN_TO_16 int32_t i32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;49
+ SIMDE_ALIGN_TO_16 int64_t i64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;50
+ SIMDE_ALIGN_TO_16 uint8_t u8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;51
+ SIMDE_ALIGN_TO_16 uint16_t u16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;52
+ SIMDE_ALIGN_TO_16 uint32_t u32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;53
+ SIMDE_ALIGN_TO_16 uint64_t u64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;54
+#if defined(SIMDE_HAVE_INT128_)55
+ SIMDE_ALIGN_TO_16 simde_int128 i128 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;56
+ SIMDE_ALIGN_TO_16 simde_uint128 u128 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;57
+#endif58
+ SIMDE_ALIGN_TO_16 simde_float32 f32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;59
+ SIMDE_ALIGN_TO_16 int_fast32_t i32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;60
+ SIMDE_ALIGN_TO_16 uint_fast32_t u32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;61
+#else62
+ SIMDE_ALIGN_TO_16 int8_t i8[16];63
+ SIMDE_ALIGN_TO_16 int16_t i16[8];64
+ SIMDE_ALIGN_TO_16 int32_t i32[4];65
+ SIMDE_ALIGN_TO_16 int64_t i64[2];66
+ SIMDE_ALIGN_TO_16 uint8_t u8[16];67
+ SIMDE_ALIGN_TO_16 uint16_t u16[8];68
+ SIMDE_ALIGN_TO_16 uint32_t u32[4];69
+ SIMDE_ALIGN_TO_16 uint64_t u64[2];70
+#if defined(SIMDE_HAVE_INT128_)71
+ SIMDE_ALIGN_TO_16 simde_int128 i128[1];72
+ SIMDE_ALIGN_TO_16 simde_uint128 u128[1];73
+#endif74
+ SIMDE_ALIGN_TO_16 simde_float32 f32[4];75
+ SIMDE_ALIGN_TO_16 int_fast32_t i32f[16 / sizeof(int_fast32_t)];76
+ SIMDE_ALIGN_TO_16 uint_fast32_t u32f[16 / sizeof(uint_fast32_t)];77
+#endif78
+79
+ SIMDE_ALIGN_TO_16 simde__m64_private m64_private[2];80
+ SIMDE_ALIGN_TO_16 simde__m64 m64[2];81
+82
+#if defined(SIMDE_X86_SSE_NATIVE)83
+ SIMDE_ALIGN_TO_16 __m128 n;84
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)85
+ SIMDE_ALIGN_TO_16 int8x16_t neon_i8;86
+ SIMDE_ALIGN_TO_16 int16x8_t neon_i16;87
+ SIMDE_ALIGN_TO_16 int32x4_t neon_i32;88
+ SIMDE_ALIGN_TO_16 int64x2_t neon_i64;89
+ SIMDE_ALIGN_TO_16 uint8x16_t neon_u8;90
+ SIMDE_ALIGN_TO_16 uint16x8_t neon_u16;91
+ SIMDE_ALIGN_TO_16 uint32x4_t neon_u32;92
+ SIMDE_ALIGN_TO_16 uint64x2_t neon_u64;93
+ SIMDE_ALIGN_TO_16 float32x4_t neon_f32;94
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)95
+ SIMDE_ALIGN_TO_16 float64x2_t neon_f64;96
+#endif97
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)98
+ SIMDE_ALIGN_TO_16 v128_t wasm_v128;99
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)100
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(unsigned char) altivec_u8;101
+ SIMDE_ALIGN_TO_16102
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned short) altivec_u16;103
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32;104
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed char) altivec_i8;105
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed short) altivec_i16;106
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32;107
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(float) altivec_f32;108
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)109
+ SIMDE_ALIGN_TO_16110
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long) altivec_u64;111
+ SIMDE_ALIGN_TO_16112
+ SIMDE_POWER_ALTIVEC_VECTOR(signed long long) altivec_i64;113
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(double) altivec_f64;114
+#endif115
+#endif116
+} simde__m128_private;117
+118
+#if defined(SIMDE_X86_SSE_NATIVE)119
+typedef __m128 simde__m128;120
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)121
+typedef float32x4_t simde__m128;122
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)123
+typedef v128_t simde__m128;124
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)125
+typedef SIMDE_POWER_ALTIVEC_VECTOR(float) simde__m128;126
+#elif defined(SIMDE_VECTOR_SUBSCRIPT)127
+typedef simde_float32128
+ simde__m128 SIMDE_ALIGN_TO_16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;129
+#else130
+typedef simde__m128_private simde__m128;131
+#endif132
+133
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)134
+typedef simde__m128 __m128;135
+#endif136
+137
+HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128), "simde__m128 size incorrect");138
+HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128_private),139
+ "simde__m128_private size incorrect");140
+#if defined(SIMDE_CHECK_ALIGNMENT) && defined(SIMDE_ALIGN_OF)141
+HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128) == 16,142
+ "simde__m128 is not 16-byte aligned");143
+HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128_private) == 16,144
+ "simde__m128_private is not 16-byte aligned");145
+#endif146
+147
+SIMDE_FUNCTION_ATTRIBUTES148
+simde__m128 simde__m128_from_private(simde__m128_private v)149
+{150
+ simde__m128 r;151
+ simde_memcpy(&r, &v, sizeof(r));152
+ return r;153
+}154
+155
+SIMDE_FUNCTION_ATTRIBUTES156
+simde__m128_private simde__m128_to_private(simde__m128 v)157
+{158
+ simde__m128_private r;159
+ simde_memcpy(&r, &v, sizeof(r));160
+ return r;161
+}162
+163
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)164
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, int8x16_t, neon, i8)165
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, int16x8_t, neon, i16)166
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, int32x4_t, neon, i32)167
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, int64x2_t, neon, i64)168
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, uint8x16_t, neon, u8)169
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, uint16x8_t, neon, u16)170
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, uint32x4_t, neon, u32)171
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, uint64x2_t, neon, u64)172
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, float32x4_t, neon, f32)173
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)174
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, float64x2_t, neon, f64)175
+#endif176
+#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */177
+178
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)179
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128,180
+ SIMDE_POWER_ALTIVEC_VECTOR(signed char),181
+ altivec, i8)182
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128,183
+ SIMDE_POWER_ALTIVEC_VECTOR(signed short),184
+ altivec, i16)185
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128,186
+ SIMDE_POWER_ALTIVEC_VECTOR(signed int),187
+ altivec, i32)188
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(189
+ m128, SIMDE_POWER_ALTIVEC_VECTOR(unsigned char), altivec, u8)190
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(191
+ m128, SIMDE_POWER_ALTIVEC_VECTOR(unsigned short), altivec, u16)192
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128,193
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned int),194
+ altivec, u32)195
+196
+#if defined(SIMDE_BUG_GCC_95782)197
+SIMDE_FUNCTION_ATTRIBUTES198
+SIMDE_POWER_ALTIVEC_VECTOR(float)199
+simde__m128_to_altivec_f32(simde__m128 value)200
+{201
+ simde__m128_private r_ = simde__m128_to_private(value);202
+ return r_.altivec_f32;203
+}204
+205
+SIMDE_FUNCTION_ATTRIBUTES206
+simde__m128 simde__m128_from_altivec_f32(SIMDE_POWER_ALTIVEC_VECTOR(float)207
+ value)208
+{209
+ simde__m128_private r_;210
+ r_.altivec_f32 = value;211
+ return simde__m128_from_private(r_);212
+}213
+#else214
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, SIMDE_POWER_ALTIVEC_VECTOR(float),215
+ altivec, f32)216
+#endif217
+218
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)219
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(220
+ m128, SIMDE_POWER_ALTIVEC_VECTOR(signed long long), altivec, i64)221
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(222
+ m128, SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long), altivec, u64)223
+#endif224
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)225
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128, v128_t, wasm, v128);226
+#endif /* defined(SIMDE_POWER_ALTIVEC_P6_NATIVE) */227
+228
+enum {229
+#if defined(SIMDE_X86_SSE_NATIVE)230
+ SIMDE_MM_ROUND_NEAREST = _MM_ROUND_NEAREST,231
+ SIMDE_MM_ROUND_DOWN = _MM_ROUND_DOWN,232
+ SIMDE_MM_ROUND_UP = _MM_ROUND_UP,233
+ SIMDE_MM_ROUND_TOWARD_ZERO = _MM_ROUND_TOWARD_ZERO234
+#else235
+ SIMDE_MM_ROUND_NEAREST = 0x0000,236
+ SIMDE_MM_ROUND_DOWN = 0x2000,237
+ SIMDE_MM_ROUND_UP = 0x4000,238
+ SIMDE_MM_ROUND_TOWARD_ZERO = 0x6000239
+#endif240
+};241
+242
+#if defined(_MM_FROUND_TO_NEAREST_INT)243
+#define SIMDE_MM_FROUND_TO_NEAREST_INT _MM_FROUND_TO_NEAREST_INT244
+#define SIMDE_MM_FROUND_TO_NEG_INF _MM_FROUND_TO_NEG_INF245
+#define SIMDE_MM_FROUND_TO_POS_INF _MM_FROUND_TO_POS_INF246
+#define SIMDE_MM_FROUND_TO_ZERO _MM_FROUND_TO_ZERO247
+#define SIMDE_MM_FROUND_CUR_DIRECTION _MM_FROUND_CUR_DIRECTION248
+249
+#define SIMDE_MM_FROUND_RAISE_EXC _MM_FROUND_RAISE_EXC250
+#define SIMDE_MM_FROUND_NO_EXC _MM_FROUND_NO_EXC251
+#else252
+#define SIMDE_MM_FROUND_TO_NEAREST_INT 0x00253
+#define SIMDE_MM_FROUND_TO_NEG_INF 0x01254
+#define SIMDE_MM_FROUND_TO_POS_INF 0x02255
+#define SIMDE_MM_FROUND_TO_ZERO 0x03256
+#define SIMDE_MM_FROUND_CUR_DIRECTION 0x04257
+258
+#define SIMDE_MM_FROUND_RAISE_EXC 0x00259
+#define SIMDE_MM_FROUND_NO_EXC 0x08260
+#endif261
+262
+#define SIMDE_MM_FROUND_NINT \263
+ (SIMDE_MM_FROUND_TO_NEAREST_INT | SIMDE_MM_FROUND_RAISE_EXC)264
+#define SIMDE_MM_FROUND_FLOOR \265
+ (SIMDE_MM_FROUND_TO_NEG_INF | SIMDE_MM_FROUND_RAISE_EXC)266
+#define SIMDE_MM_FROUND_CEIL \267
+ (SIMDE_MM_FROUND_TO_POS_INF | SIMDE_MM_FROUND_RAISE_EXC)268
+#define SIMDE_MM_FROUND_TRUNC \269
+ (SIMDE_MM_FROUND_TO_ZERO | SIMDE_MM_FROUND_RAISE_EXC)270
+#define SIMDE_MM_FROUND_RINT \271
+ (SIMDE_MM_FROUND_CUR_DIRECTION | SIMDE_MM_FROUND_RAISE_EXC)272
+#define SIMDE_MM_FROUND_NEARBYINT \273
+ (SIMDE_MM_FROUND_CUR_DIRECTION | SIMDE_MM_FROUND_NO_EXC)274
+275
+#if defined(SIMDE_X86_SSE4_1_ENABLE_NATIVE_ALIASES) && \276
+ !defined(_MM_FROUND_TO_NEAREST_INT)277
+#define _MM_FROUND_TO_NEAREST_INT SIMDE_MM_FROUND_TO_NEAREST_INT278
+#define _MM_FROUND_TO_NEG_INF SIMDE_MM_FROUND_TO_NEG_INF279
+#define _MM_FROUND_TO_POS_INF SIMDE_MM_FROUND_TO_POS_INF280
+#define _MM_FROUND_TO_ZERO SIMDE_MM_FROUND_TO_ZERO281
+#define _MM_FROUND_CUR_DIRECTION SIMDE_MM_FROUND_CUR_DIRECTION282
+#define _MM_FROUND_RAISE_EXC SIMDE_MM_FROUND_RAISE_EXC283
+#define _MM_FROUND_NINT SIMDE_MM_FROUND_NINT284
+#define _MM_FROUND_FLOOR SIMDE_MM_FROUND_FLOOR285
+#define _MM_FROUND_CEIL SIMDE_MM_FROUND_CEIL286
+#define _MM_FROUND_TRUNC SIMDE_MM_FROUND_TRUNC287
+#define _MM_FROUND_RINT SIMDE_MM_FROUND_RINT288
+#define _MM_FROUND_NEARBYINT SIMDE_MM_FROUND_NEARBYINT289
+#endif290
+291
+SIMDE_FUNCTION_ATTRIBUTES292
+unsigned int SIMDE_MM_GET_ROUNDING_MODE(void)293
+{294
+#if defined(SIMDE_X86_SSE_NATIVE)295
+ return _MM_GET_ROUNDING_MODE();296
+#elif defined(SIMDE_HAVE_FENV_H)297
+ unsigned int vfe_mode;298
+299
+ switch (fegetround()) {300
+#if defined(FE_TONEAREST)301
+ case FE_TONEAREST:302
+ vfe_mode = SIMDE_MM_ROUND_NEAREST;303
+ break;304
+#endif305
+306
+#if defined(FE_TOWARDZERO)307
+ case FE_TOWARDZERO:308
+ vfe_mode = SIMDE_MM_ROUND_DOWN;309
+ break;310
+#endif311
+312
+#if defined(FE_UPWARD)313
+ case FE_UPWARD:314
+ vfe_mode = SIMDE_MM_ROUND_UP;315
+ break;316
+#endif317
+318
+#if defined(FE_DOWNWARD)319
+ case FE_DOWNWARD:320
+ vfe_mode = SIMDE_MM_ROUND_TOWARD_ZERO;321
+ break;322
+#endif323
+324
+ default:325
+ vfe_mode = SIMDE_MM_ROUND_NEAREST;326
+ break;327
+ }328
+329
+ return vfe_mode;330
+#else331
+ return SIMDE_MM_ROUND_NEAREST;332
+#endif333
+}334
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)335
+#define _MM_GET_ROUNDING_MODE() SIMDE_MM_GET_ROUNDING_MODE()336
+#endif337
+338
+SIMDE_FUNCTION_ATTRIBUTES339
+void SIMDE_MM_SET_ROUNDING_MODE(unsigned int a)340
+{341
+#if defined(SIMDE_X86_SSE_NATIVE)342
+ _MM_SET_ROUNDING_MODE(a);343
+#elif defined(SIMDE_HAVE_FENV_H)344
+ int fe_mode = FE_TONEAREST;345
+346
+ switch (a) {347
+#if defined(FE_TONEAREST)348
+ case SIMDE_MM_ROUND_NEAREST:349
+ fe_mode = FE_TONEAREST;350
+ break;351
+#endif352
+353
+#if defined(FE_TOWARDZERO)354
+ case SIMDE_MM_ROUND_TOWARD_ZERO:355
+ fe_mode = FE_TOWARDZERO;356
+ break;357
+#endif358
+359
+#if defined(FE_DOWNWARD)360
+ case SIMDE_MM_ROUND_DOWN:361
+ fe_mode = FE_DOWNWARD;362
+ break;363
+#endif364
+365
+#if defined(FE_UPWARD)366
+ case SIMDE_MM_ROUND_UP:367
+ fe_mode = FE_UPWARD;368
+ break;369
+#endif370
+371
+ default:372
+ return;373
+ }374
+375
+ fesetround(fe_mode);376
+#else377
+ (void)a;378
+#endif379
+}380
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)381
+#define _MM_SET_ROUNDING_MODE(a) SIMDE_MM_SET_ROUNDING_MODE(a)382
+#endif383
+384
+SIMDE_FUNCTION_ATTRIBUTES385
+uint32_t simde_mm_getcsr(void)386
+{387
+#if defined(SIMDE_X86_SSE_NATIVE)388
+ return _mm_getcsr();389
+#else390
+ return SIMDE_MM_GET_ROUNDING_MODE();391
+#endif392
+}393
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)394
+#define _mm_getcsr() simde_mm_getcsr()395
+#endif396
+397
+SIMDE_FUNCTION_ATTRIBUTES398
+void simde_mm_setcsr(uint32_t a)399
+{400
+#if defined(SIMDE_X86_SSE_NATIVE)401
+ _mm_setcsr(a);402
+#else403
+ SIMDE_MM_SET_ROUNDING_MODE(HEDLEY_STATIC_CAST(unsigned int, a));404
+#endif405
+}406
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)407
+#define _mm_setcsr(a) simde_mm_setcsr(a)408
+#endif409
+410
+SIMDE_FUNCTION_ATTRIBUTES411
+simde__m128 simde_x_mm_round_ps(simde__m128 a, int rounding, int lax_rounding)412
+ SIMDE_REQUIRE_CONSTANT_RANGE(rounding, 0, 15)413
+ SIMDE_REQUIRE_CONSTANT_RANGE(lax_rounding, 0, 1)414
+{415
+ simde__m128_private r_, a_ = simde__m128_to_private(a);416
+417
+ (void)lax_rounding;418
+419
+/* For architectures which lack a current direction SIMD instruction.420
+ *421
+ * Note that NEON actually has a current rounding mode instruction,422
+ * but in ARMv8+ the rounding mode is ignored and nearest is always423
+ * used, so we treat ARMv7 as having a rounding mode but ARMv8 as424
+ * not. */425
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE) || defined(SIMDE_ARM_NEON_A32V8)426
+ if ((rounding & 7) == SIMDE_MM_FROUND_CUR_DIRECTION)427
+ rounding = HEDLEY_STATIC_CAST(int, SIMDE_MM_GET_ROUNDING_MODE())428
+ << 13;429
+#endif430
+431
+ switch (rounding & ~SIMDE_MM_FROUND_NO_EXC) {432
+ case SIMDE_MM_FROUND_CUR_DIRECTION:433
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)434
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(435
+ SIMDE_POWER_ALTIVEC_VECTOR(float),436
+ vec_round(a_.altivec_f32));437
+#elif defined(SIMDE_ARM_NEON_A32V8_NATIVE) && !defined(SIMDE_BUG_GCC_95399)438
+ r_.neon_f32 = vrndiq_f32(a_.neon_f32);439
+#elif defined(simde_math_nearbyintf)440
+ SIMDE_VECTORIZE441
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0]));442
+ i++) {443
+ r_.f32[i] = simde_math_nearbyintf(a_.f32[i]);444
+ }445
+#else446
+ HEDLEY_UNREACHABLE_RETURN(simde_mm_undefined_pd());447
+#endif448
+ break;449
+450
+ case SIMDE_MM_FROUND_TO_NEAREST_INT:451
+#if defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)452
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(453
+ SIMDE_POWER_ALTIVEC_VECTOR(float),454
+ vec_rint(a_.altivec_f32));455
+#elif defined(SIMDE_ARM_NEON_A32V8_NATIVE)456
+ r_.neon_f32 = vrndnq_f32(a_.neon_f32);457
+#elif defined(simde_math_roundevenf)458
+ SIMDE_VECTORIZE459
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0]));460
+ i++) {461
+ r_.f32[i] = simde_math_roundevenf(a_.f32[i]);462
+ }463
+#else464
+ HEDLEY_UNREACHABLE_RETURN(simde_mm_undefined_pd());465
+#endif466
+ break;467
+468
+ case SIMDE_MM_FROUND_TO_NEG_INF:469
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)470
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(471
+ SIMDE_POWER_ALTIVEC_VECTOR(float),472
+ vec_floor(a_.altivec_f32));473
+#elif defined(SIMDE_ARM_NEON_A32V8_NATIVE)474
+ r_.neon_f32 = vrndmq_f32(a_.neon_f32);475
+#elif defined(simde_math_floorf)476
+ SIMDE_VECTORIZE477
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0]));478
+ i++) {479
+ r_.f32[i] = simde_math_floorf(a_.f32[i]);480
+ }481
+#else482
+ HEDLEY_UNREACHABLE_RETURN(simde_mm_undefined_pd());483
+#endif484
+ break;485
+486
+ case SIMDE_MM_FROUND_TO_POS_INF:487
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)488
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(489
+ SIMDE_POWER_ALTIVEC_VECTOR(float),490
+ vec_ceil(a_.altivec_f32));491
+#elif defined(SIMDE_ARM_NEON_A32V8_NATIVE)492
+ r_.neon_f32 = vrndpq_f32(a_.neon_f32);493
+#elif defined(simde_math_ceilf)494
+ SIMDE_VECTORIZE495
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0]));496
+ i++) {497
+ r_.f32[i] = simde_math_ceilf(a_.f32[i]);498
+ }499
+#else500
+ HEDLEY_UNREACHABLE_RETURN(simde_mm_undefined_pd());501
+#endif502
+ break;503
+504
+ case SIMDE_MM_FROUND_TO_ZERO:505
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)506
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(507
+ SIMDE_POWER_ALTIVEC_VECTOR(float),508
+ vec_trunc(a_.altivec_f32));509
+#elif defined(SIMDE_ARM_NEON_A32V8_NATIVE)510
+ r_.neon_f32 = vrndq_f32(a_.neon_f32);511
+#elif defined(simde_math_truncf)512
+ SIMDE_VECTORIZE513
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0]));514
+ i++) {515
+ r_.f32[i] = simde_math_truncf(a_.f32[i]);516
+ }517
+#else518
+ HEDLEY_UNREACHABLE_RETURN(simde_mm_undefined_pd());519
+#endif520
+ break;521
+522
+ default:523
+ HEDLEY_UNREACHABLE_RETURN(simde_mm_undefined_pd());524
+ }525
+526
+ return simde__m128_from_private(r_);527
+}528
+#if defined(SIMDE_X86_SSE4_1_NATIVE)529
+#define simde_mm_round_ps(a, rounding) _mm_round_ps((a), (rounding))530
+#else531
+#define simde_mm_round_ps(a, rounding) simde_x_mm_round_ps((a), (rounding), 0)532
+#endif533
+#if defined(SIMDE_X86_SSE4_1_ENABLE_NATIVE_ALIASES)534
+#define _mm_round_ps(a, rounding) simde_mm_round_ps((a), (rounding))535
+#endif536
+537
+SIMDE_FUNCTION_ATTRIBUTES538
+simde__m128 simde_mm_set_ps(simde_float32 e3, simde_float32 e2,539
+ simde_float32 e1, simde_float32 e0)540
+{541
+#if defined(SIMDE_X86_SSE_NATIVE)542
+ return _mm_set_ps(e3, e2, e1, e0);543
+#else544
+ simde__m128_private r_;545
+546
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)547
+ SIMDE_ALIGN_TO_16 simde_float32 data[4] = {e0, e1, e2, e3};548
+ r_.neon_f32 = vld1q_f32(data);549
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)550
+ r_.wasm_v128 = wasm_f32x4_make(e0, e1, e2, e3);551
+#else552
+ r_.f32[0] = e0;553
+ r_.f32[1] = e1;554
+ r_.f32[2] = e2;555
+ r_.f32[3] = e3;556
+#endif557
+558
+ return simde__m128_from_private(r_);559
+#endif560
+}561
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)562
+#define _mm_set_ps(e3, e2, e1, e0) simde_mm_set_ps(e3, e2, e1, e0)563
+#endif564
+565
+SIMDE_FUNCTION_ATTRIBUTES566
+simde__m128 simde_mm_set_ps1(simde_float32 a)567
+{568
+#if defined(SIMDE_X86_SSE_NATIVE)569
+ return _mm_set_ps1(a);570
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)571
+ return vdupq_n_f32(a);572
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)573
+ (void)a;574
+ return vec_splats(a);575
+#else576
+ return simde_mm_set_ps(a, a, a, a);577
+#endif578
+}579
+#define simde_mm_set1_ps(a) simde_mm_set_ps1(a)580
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)581
+#define _mm_set_ps1(a) simde_mm_set_ps1(a)582
+#define _mm_set1_ps(a) simde_mm_set1_ps(a)583
+#endif584
+585
+SIMDE_FUNCTION_ATTRIBUTES586
+simde__m128 simde_mm_move_ss(simde__m128 a, simde__m128 b)587
+{588
+#if defined(SIMDE_X86_SSE_NATIVE)589
+ return _mm_move_ss(a, b);590
+#else591
+ simde__m128_private r_, a_ = simde__m128_to_private(a),592
+ b_ = simde__m128_to_private(b);593
+594
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)595
+ r_.neon_f32 =596
+ vsetq_lane_f32(vgetq_lane_f32(b_.neon_f32, 0), a_.neon_f32, 0);597
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)598
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned char)599
+ m = {16, 17, 18, 19, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};600
+ r_.altivec_f32 = vec_perm(a_.altivec_f32, b_.altivec_f32, m);601
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)602
+ r_.wasm_v128 = wasm_v8x16_shuffle(b_.wasm_v128, a_.wasm_v128, 0, 1, 2,603
+ 3, 20, 21, 22, 23, 24, 25, 26, 27, 28,604
+ 29, 30, 31);605
+#elif defined(SIMDE_SHUFFLE_VECTOR_)606
+ r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 4, 1, 2, 3);607
+#else608
+ r_.f32[0] = b_.f32[0];609
+ r_.f32[1] = a_.f32[1];610
+ r_.f32[2] = a_.f32[2];611
+ r_.f32[3] = a_.f32[3];612
+#endif613
+614
+ return simde__m128_from_private(r_);615
+#endif616
+}617
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)618
+#define _mm_move_ss(a, b) simde_mm_move_ss((a), (b))619
+#endif620
+621
+SIMDE_FUNCTION_ATTRIBUTES622
+simde__m128 simde_mm_add_ps(simde__m128 a, simde__m128 b)623
+{624
+#if defined(SIMDE_X86_SSE_NATIVE)625
+ return _mm_add_ps(a, b);626
+#else627
+ simde__m128_private r_, a_ = simde__m128_to_private(a),628
+ b_ = simde__m128_to_private(b);629
+630
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)631
+ r_.neon_f32 = vaddq_f32(a_.neon_f32, b_.neon_f32);632
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)633
+ r_.wasm_v128 = wasm_f32x4_add(a_.wasm_v128, b_.wasm_v128);634
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)635
+ r_.altivec_f32 = vec_add(a_.altivec_f32, b_.altivec_f32);636
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)637
+ r_.f32 = a_.f32 + b_.f32;638
+#else639
+ SIMDE_VECTORIZE640
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {641
+ r_.f32[i] = a_.f32[i] + b_.f32[i];642
+ }643
+#endif644
+645
+ return simde__m128_from_private(r_);646
+#endif647
+}648
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)649
+#define _mm_add_ps(a, b) simde_mm_add_ps((a), (b))650
+#endif651
+652
+SIMDE_FUNCTION_ATTRIBUTES653
+simde__m128 simde_mm_add_ss(simde__m128 a, simde__m128 b)654
+{655
+#if defined(SIMDE_X86_SSE_NATIVE)656
+ return _mm_add_ss(a, b);657
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)658
+ return simde_mm_move_ss(a, simde_mm_add_ps(a, b));659
+#else660
+ simde__m128_private r_, a_ = simde__m128_to_private(a),661
+ b_ = simde__m128_to_private(b);662
+663
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)664
+ float32_t b0 = vgetq_lane_f32(b_.neon_f32, 0);665
+ float32x4_t value = vsetq_lane_f32(b0, vdupq_n_f32(0), 0);666
+ // the upper values in the result must be the remnants of <a>.667
+ r_.neon_f32 = vaddq_f32(a_.neon_f32, value);668
+#else669
+ r_.f32[0] = a_.f32[0] + b_.f32[0];670
+ r_.f32[1] = a_.f32[1];671
+ r_.f32[2] = a_.f32[2];672
+ r_.f32[3] = a_.f32[3];673
+#endif674
+675
+ return simde__m128_from_private(r_);676
+#endif677
+}678
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)679
+#define _mm_add_ss(a, b) simde_mm_add_ss((a), (b))680
+#endif681
+682
+SIMDE_FUNCTION_ATTRIBUTES683
+simde__m128 simde_mm_and_ps(simde__m128 a, simde__m128 b)684
+{685
+#if defined(SIMDE_X86_SSE_NATIVE)686
+ return _mm_and_ps(a, b);687
+#else688
+ simde__m128_private r_, a_ = simde__m128_to_private(a),689
+ b_ = simde__m128_to_private(b);690
+691
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)692
+ r_.neon_i32 = vandq_s32(a_.neon_i32, b_.neon_i32);693
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)694
+ r_.wasm_v128 = wasm_v128_and(a_.wasm_v128, b_.wasm_v128);695
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)696
+ r_.i32 = a_.i32 & b_.i32;697
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)698
+ r_.altivec_f32 = vec_and(a_.altivec_f32, b_.altivec_f32);699
+#else700
+ SIMDE_VECTORIZE701
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {702
+ r_.i32[i] = a_.i32[i] & b_.i32[i];703
+ }704
+#endif705
+706
+ return simde__m128_from_private(r_);707
+#endif708
+}709
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)710
+#define _mm_and_ps(a, b) simde_mm_and_ps((a), (b))711
+#endif712
+713
+SIMDE_FUNCTION_ATTRIBUTES714
+simde__m128 simde_mm_andnot_ps(simde__m128 a, simde__m128 b)715
+{716
+#if defined(SIMDE_X86_SSE_NATIVE)717
+ return _mm_andnot_ps(a, b);718
+#else719
+ simde__m128_private r_, a_ = simde__m128_to_private(a),720
+ b_ = simde__m128_to_private(b);721
+722
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)723
+ r_.neon_i32 = vbicq_s32(b_.neon_i32, a_.neon_i32);724
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)725
+ r_.wasm_v128 = wasm_v128_andnot(b_.wasm_v128, a_.wasm_v128);726
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)727
+ r_.altivec_f32 = vec_andc(b_.altivec_f32, a_.altivec_f32);728
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)729
+ r_.i32 = ~a_.i32 & b_.i32;730
+#else731
+ SIMDE_VECTORIZE732
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {733
+ r_.i32[i] = ~(a_.i32[i]) & b_.i32[i];734
+ }735
+#endif736
+737
+ return simde__m128_from_private(r_);738
+#endif739
+}740
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)741
+#define _mm_andnot_ps(a, b) simde_mm_andnot_ps((a), (b))742
+#endif743
+744
+SIMDE_FUNCTION_ATTRIBUTES745
+simde__m128 simde_mm_xor_ps(simde__m128 a, simde__m128 b)746
+{747
+#if defined(SIMDE_X86_SSE_NATIVE)748
+ return _mm_xor_ps(a, b);749
+#else750
+ simde__m128_private r_, a_ = simde__m128_to_private(a),751
+ b_ = simde__m128_to_private(b);752
+753
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)754
+ r_.neon_i32 = veorq_s32(a_.neon_i32, b_.neon_i32);755
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)756
+ r_.wasm_v128 = wasm_v128_xor(a_.wasm_v128, b_.wasm_v128);757
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)758
+ r_.altivec_i32 = vec_xor(a_.altivec_i32, b_.altivec_i32);759
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)760
+ r_.i32f = a_.i32f ^ b_.i32f;761
+#else762
+ SIMDE_VECTORIZE763
+ for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {764
+ r_.u32[i] = a_.u32[i] ^ b_.u32[i];765
+ }766
+#endif767
+768
+ return simde__m128_from_private(r_);769
+#endif770
+}771
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)772
+#define _mm_xor_ps(a, b) simde_mm_xor_ps((a), (b))773
+#endif774
+775
+SIMDE_FUNCTION_ATTRIBUTES776
+simde__m128 simde_mm_or_ps(simde__m128 a, simde__m128 b)777
+{778
+#if defined(SIMDE_X86_SSE_NATIVE)779
+ return _mm_or_ps(a, b);780
+#else781
+ simde__m128_private r_, a_ = simde__m128_to_private(a),782
+ b_ = simde__m128_to_private(b);783
+784
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)785
+ r_.neon_i32 = vorrq_s32(a_.neon_i32, b_.neon_i32);786
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)787
+ r_.wasm_v128 = wasm_v128_or(a_.wasm_v128, b_.wasm_v128);788
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)789
+ r_.altivec_i32 = vec_or(a_.altivec_i32, b_.altivec_i32);790
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)791
+ r_.i32f = a_.i32f | b_.i32f;792
+#else793
+ SIMDE_VECTORIZE794
+ for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {795
+ r_.u32[i] = a_.u32[i] | b_.u32[i];796
+ }797
+#endif798
+799
+ return simde__m128_from_private(r_);800
+#endif801
+}802
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)803
+#define _mm_or_ps(a, b) simde_mm_or_ps((a), (b))804
+#endif805
+806
+SIMDE_FUNCTION_ATTRIBUTES807
+simde__m128 simde_x_mm_not_ps(simde__m128 a)808
+{809
+#if defined(SIMDE_X86_AVX512VL_NATIVE)810
+ __m128i ai = _mm_castps_si128(a);811
+ return _mm_castsi128_ps(_mm_ternarylogic_epi32(ai, ai, ai, 0x55));812
+#elif defined(SIMDE_X86_SSE2_NATIVE)813
+ /* Note: we use ints instead of floats because we don't want cmpeq814
+ * to return false for (NaN, NaN) */815
+ __m128i ai = _mm_castps_si128(a);816
+ return _mm_castsi128_ps(_mm_andnot_si128(ai, _mm_cmpeq_epi32(ai, ai)));817
+#else818
+ simde__m128_private r_, a_ = simde__m128_to_private(a);819
+820
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)821
+ r_.neon_i32 = vmvnq_s32(a_.neon_i32);822
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)823
+ r_.altivec_i32 = vec_nor(a_.altivec_i32, a_.altivec_i32);824
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)825
+ r_.wasm_v128 = wasm_v128_not(a_.wasm_v128);826
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)827
+ r_.i32 = ~a_.i32;828
+#else829
+ SIMDE_VECTORIZE830
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {831
+ r_.i32[i] = ~(a_.i32[i]);832
+ }833
+#endif834
+835
+ return simde__m128_from_private(r_);836
+#endif837
+}838
+839
+SIMDE_FUNCTION_ATTRIBUTES840
+simde__m128 simde_x_mm_select_ps(simde__m128 a, simde__m128 b, simde__m128 mask)841
+{842
+/* This function is for when you want to blend two elements together843
+ * according to a mask. It is similar to _mm_blendv_ps, except that844
+ * it is undefined whether the blend is based on the highest bit in845
+ * each lane (like blendv) or just bitwise operations. This allows846
+ * us to implement the function efficiently everywhere.847
+ *848
+ * Basically, you promise that all the lanes in mask are either 0 or849
+ * ~0. */850
+#if defined(SIMDE_X86_SSE4_1_NATIVE)851
+ return _mm_blendv_ps(a, b, mask);852
+#else853
+ simde__m128_private r_, a_ = simde__m128_to_private(a),854
+ b_ = simde__m128_to_private(b),855
+ mask_ = simde__m128_to_private(mask);856
+857
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)858
+ r_.neon_i32 = vbslq_s32(mask_.neon_u32, b_.neon_i32, a_.neon_i32);859
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)860
+ r_.wasm_v128 = wasm_v128_bitselect(b_.wasm_v128, a_.wasm_v128,861
+ mask_.wasm_v128);862
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)863
+ r_.altivec_i32 =864
+ vec_sel(a_.altivec_i32, b_.altivec_i32, mask_.altivec_u32);865
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)866
+ r_.i32 = a_.i32 ^ ((a_.i32 ^ b_.i32) & mask_.i32);867
+#else868
+ SIMDE_VECTORIZE869
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {870
+ r_.i32[i] = a_.i32[i] ^871
+ ((a_.i32[i] ^ b_.i32[i]) & mask_.i32[i]);872
+ }873
+#endif874
+875
+ return simde__m128_from_private(r_);876
+#endif877
+}878
+879
+SIMDE_FUNCTION_ATTRIBUTES880
+simde__m64 simde_mm_avg_pu16(simde__m64 a, simde__m64 b)881
+{882
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)883
+ return _mm_avg_pu16(a, b);884
+#else885
+ simde__m64_private r_, a_ = simde__m64_to_private(a),886
+ b_ = simde__m64_to_private(b);887
+888
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)889
+ r_.neon_u16 = vrhadd_u16(b_.neon_u16, a_.neon_u16);890
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS) && \891
+ defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \892
+ defined(SIMDE_CONVERT_VECTOR_)893
+ uint32_t wa SIMDE_VECTOR(16);894
+ uint32_t wb SIMDE_VECTOR(16);895
+ uint32_t wr SIMDE_VECTOR(16);896
+ SIMDE_CONVERT_VECTOR_(wa, a_.u16);897
+ SIMDE_CONVERT_VECTOR_(wb, b_.u16);898
+ wr = (wa + wb + 1) >> 1;899
+ SIMDE_CONVERT_VECTOR_(r_.u16, wr);900
+#else901
+ SIMDE_VECTORIZE902
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {903
+ r_.u16[i] = (a_.u16[i] + b_.u16[i] + 1) >> 1;904
+ }905
+#endif906
+907
+ return simde__m64_from_private(r_);908
+#endif909
+}910
+#define simde_m_pavgw(a, b) simde_mm_avg_pu16(a, b)911
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)912
+#define _mm_avg_pu16(a, b) simde_mm_avg_pu16(a, b)913
+#define _m_pavgw(a, b) simde_mm_avg_pu16(a, b)914
+#endif915
+916
+SIMDE_FUNCTION_ATTRIBUTES917
+simde__m64 simde_mm_avg_pu8(simde__m64 a, simde__m64 b)918
+{919
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)920
+ return _mm_avg_pu8(a, b);921
+#else922
+ simde__m64_private r_, a_ = simde__m64_to_private(a),923
+ b_ = simde__m64_to_private(b);924
+925
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)926
+ r_.neon_u8 = vrhadd_u8(b_.neon_u8, a_.neon_u8);927
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS) && \928
+ defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \929
+ defined(SIMDE_CONVERT_VECTOR_)930
+ uint16_t wa SIMDE_VECTOR(16);931
+ uint16_t wb SIMDE_VECTOR(16);932
+ uint16_t wr SIMDE_VECTOR(16);933
+ SIMDE_CONVERT_VECTOR_(wa, a_.u8);934
+ SIMDE_CONVERT_VECTOR_(wb, b_.u8);935
+ wr = (wa + wb + 1) >> 1;936
+ SIMDE_CONVERT_VECTOR_(r_.u8, wr);937
+#else938
+ SIMDE_VECTORIZE939
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {940
+ r_.u8[i] = (a_.u8[i] + b_.u8[i] + 1) >> 1;941
+ }942
+#endif943
+944
+ return simde__m64_from_private(r_);945
+#endif946
+}947
+#define simde_m_pavgb(a, b) simde_mm_avg_pu8(a, b)948
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)949
+#define _mm_avg_pu8(a, b) simde_mm_avg_pu8(a, b)950
+#define _m_pavgb(a, b) simde_mm_avg_pu8(a, b)951
+#endif952
+953
+SIMDE_FUNCTION_ATTRIBUTES954
+simde__m128 simde_x_mm_abs_ps(simde__m128 a)955
+{956
+#if defined(SIMDE_X86_AVX512F_NATIVE) && \957
+ (!defined(HEDLEY_GCC_VERSION) || HEDLEY_GCC_VERSION_CHECK(7, 1, 0))958
+ return _mm512_castps512_ps128(_mm512_abs_ps(_mm512_castps128_ps512(a)));959
+#else960
+ simde__m128_private r_, a_ = simde__m128_to_private(a);961
+962
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)963
+ r_.neon_f32 = vabsq_f32(a_.neon_f32);964
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)965
+ r_.altivec_f32 = vec_abs(a_.altivec_f32);966
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)967
+ r_.wasm_v128 = wasm_f32x4_abs(a_.wasm_v128);968
+#else969
+ SIMDE_VECTORIZE970
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {971
+ r_.f32[i] = simde_math_fabsf(a_.f32[i]);972
+ }973
+#endif974
+975
+ return simde__m128_from_private(r_);976
+#endif977
+}978
+979
+SIMDE_FUNCTION_ATTRIBUTES980
+simde__m128 simde_mm_cmpeq_ps(simde__m128 a, simde__m128 b)981
+{982
+#if defined(SIMDE_X86_SSE_NATIVE)983
+ return _mm_cmpeq_ps(a, b);984
+#else985
+ simde__m128_private r_, a_ = simde__m128_to_private(a),986
+ b_ = simde__m128_to_private(b);987
+988
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)989
+ r_.neon_u32 = vceqq_f32(a_.neon_f32, b_.neon_f32);990
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)991
+ r_.wasm_v128 = wasm_f32x4_eq(a_.wasm_v128, b_.wasm_v128);992
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)993
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(994
+ SIMDE_POWER_ALTIVEC_VECTOR(float),995
+ vec_cmpeq(a_.altivec_f32, b_.altivec_f32));996
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)997
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), a_.f32 == b_.f32);998
+#else999
+ SIMDE_VECTORIZE1000
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1001
+ r_.u32[i] = (a_.f32[i] == b_.f32[i]) ? ~UINT32_C(0)1002
+ : UINT32_C(0);1003
+ }1004
+#endif1005
+1006
+ return simde__m128_from_private(r_);1007
+#endif1008
+}1009
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1010
+#define _mm_cmpeq_ps(a, b) simde_mm_cmpeq_ps((a), (b))1011
+#endif1012
+1013
+SIMDE_FUNCTION_ATTRIBUTES1014
+simde__m128 simde_mm_cmpeq_ss(simde__m128 a, simde__m128 b)1015
+{1016
+#if defined(SIMDE_X86_SSE_NATIVE)1017
+ return _mm_cmpeq_ss(a, b);1018
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1019
+ return simde_mm_move_ss(a, simde_mm_cmpeq_ps(a, b));1020
+#else1021
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1022
+ b_ = simde__m128_to_private(b);1023
+1024
+ r_.u32[0] = (a_.f32[0] == b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);1025
+ SIMDE_VECTORIZE1026
+ for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1027
+ r_.u32[i] = a_.u32[i];1028
+ }1029
+1030
+ return simde__m128_from_private(r_);1031
+#endif1032
+}1033
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1034
+#define _mm_cmpeq_ss(a, b) simde_mm_cmpeq_ss((a), (b))1035
+#endif1036
+1037
+SIMDE_FUNCTION_ATTRIBUTES1038
+simde__m128 simde_mm_cmpge_ps(simde__m128 a, simde__m128 b)1039
+{1040
+#if defined(SIMDE_X86_SSE_NATIVE)1041
+ return _mm_cmpge_ps(a, b);1042
+#else1043
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1044
+ b_ = simde__m128_to_private(b);1045
+1046
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1047
+ r_.neon_u32 = vcgeq_f32(a_.neon_f32, b_.neon_f32);1048
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1049
+ r_.wasm_v128 = wasm_f32x4_ge(a_.wasm_v128, b_.wasm_v128);1050
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1051
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1052
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1053
+ vec_cmpge(a_.altivec_f32, b_.altivec_f32));1054
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1055
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 >= b_.f32));1056
+#else1057
+ SIMDE_VECTORIZE1058
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1059
+ r_.u32[i] = (a_.f32[i] >= b_.f32[i]) ? ~UINT32_C(0)1060
+ : UINT32_C(0);1061
+ }1062
+#endif1063
+1064
+ return simde__m128_from_private(r_);1065
+#endif1066
+}1067
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1068
+#define _mm_cmpge_ps(a, b) simde_mm_cmpge_ps((a), (b))1069
+#endif1070
+1071
+SIMDE_FUNCTION_ATTRIBUTES1072
+simde__m128 simde_mm_cmpge_ss(simde__m128 a, simde__m128 b)1073
+{1074
+#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)1075
+ return _mm_cmpge_ss(a, b);1076
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1077
+ return simde_mm_move_ss(a, simde_mm_cmpge_ps(a, b));1078
+#else1079
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1080
+ b_ = simde__m128_to_private(b);1081
+1082
+ r_.u32[0] = (a_.f32[0] >= b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);1083
+ SIMDE_VECTORIZE1084
+ for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1085
+ r_.u32[i] = a_.u32[i];1086
+ }1087
+1088
+ return simde__m128_from_private(r_);1089
+#endif1090
+}1091
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1092
+#define _mm_cmpge_ss(a, b) simde_mm_cmpge_ss((a), (b))1093
+#endif1094
+1095
+SIMDE_FUNCTION_ATTRIBUTES1096
+simde__m128 simde_mm_cmpgt_ps(simde__m128 a, simde__m128 b)1097
+{1098
+#if defined(SIMDE_X86_SSE_NATIVE)1099
+ return _mm_cmpgt_ps(a, b);1100
+#else1101
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1102
+ b_ = simde__m128_to_private(b);1103
+1104
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1105
+ r_.neon_u32 = vcgtq_f32(a_.neon_f32, b_.neon_f32);1106
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1107
+ r_.wasm_v128 = wasm_f32x4_gt(a_.wasm_v128, b_.wasm_v128);1108
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1109
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1110
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1111
+ vec_cmpgt(a_.altivec_f32, b_.altivec_f32));1112
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1113
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 > b_.f32));1114
+#else1115
+ SIMDE_VECTORIZE1116
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1117
+ r_.u32[i] = (a_.f32[i] > b_.f32[i]) ? ~UINT32_C(0)1118
+ : UINT32_C(0);1119
+ }1120
+#endif1121
+1122
+ return simde__m128_from_private(r_);1123
+#endif1124
+}1125
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1126
+#define _mm_cmpgt_ps(a, b) simde_mm_cmpgt_ps((a), (b))1127
+#endif1128
+1129
+SIMDE_FUNCTION_ATTRIBUTES1130
+simde__m128 simde_mm_cmpgt_ss(simde__m128 a, simde__m128 b)1131
+{1132
+#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)1133
+ return _mm_cmpgt_ss(a, b);1134
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1135
+ return simde_mm_move_ss(a, simde_mm_cmpgt_ps(a, b));1136
+#else1137
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1138
+ b_ = simde__m128_to_private(b);1139
+1140
+ r_.u32[0] = (a_.f32[0] > b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);1141
+ SIMDE_VECTORIZE1142
+ for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1143
+ r_.u32[i] = a_.u32[i];1144
+ }1145
+1146
+ return simde__m128_from_private(r_);1147
+#endif1148
+}1149
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1150
+#define _mm_cmpgt_ss(a, b) simde_mm_cmpgt_ss((a), (b))1151
+#endif1152
+1153
+SIMDE_FUNCTION_ATTRIBUTES1154
+simde__m128 simde_mm_cmple_ps(simde__m128 a, simde__m128 b)1155
+{1156
+#if defined(SIMDE_X86_SSE_NATIVE)1157
+ return _mm_cmple_ps(a, b);1158
+#else1159
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1160
+ b_ = simde__m128_to_private(b);1161
+1162
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1163
+ r_.neon_u32 = vcleq_f32(a_.neon_f32, b_.neon_f32);1164
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1165
+ r_.wasm_v128 = wasm_f32x4_le(a_.wasm_v128, b_.wasm_v128);1166
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1167
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1168
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1169
+ vec_cmple(a_.altivec_f32, b_.altivec_f32));1170
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1171
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 <= b_.f32));1172
+#else1173
+ SIMDE_VECTORIZE1174
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1175
+ r_.u32[i] = (a_.f32[i] <= b_.f32[i]) ? ~UINT32_C(0)1176
+ : UINT32_C(0);1177
+ }1178
+#endif1179
+1180
+ return simde__m128_from_private(r_);1181
+#endif1182
+}1183
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1184
+#define _mm_cmple_ps(a, b) simde_mm_cmple_ps((a), (b))1185
+#endif1186
+1187
+SIMDE_FUNCTION_ATTRIBUTES1188
+simde__m128 simde_mm_cmple_ss(simde__m128 a, simde__m128 b)1189
+{1190
+#if defined(SIMDE_X86_SSE_NATIVE)1191
+ return _mm_cmple_ss(a, b);1192
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1193
+ return simde_mm_move_ss(a, simde_mm_cmple_ps(a, b));1194
+#else1195
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1196
+ b_ = simde__m128_to_private(b);1197
+1198
+ r_.u32[0] = (a_.f32[0] <= b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);1199
+ SIMDE_VECTORIZE1200
+ for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1201
+ r_.u32[i] = a_.u32[i];1202
+ }1203
+1204
+ return simde__m128_from_private(r_);1205
+#endif1206
+}1207
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1208
+#define _mm_cmple_ss(a, b) simde_mm_cmple_ss((a), (b))1209
+#endif1210
+1211
+SIMDE_FUNCTION_ATTRIBUTES1212
+simde__m128 simde_mm_cmplt_ps(simde__m128 a, simde__m128 b)1213
+{1214
+#if defined(SIMDE_X86_SSE_NATIVE)1215
+ return _mm_cmplt_ps(a, b);1216
+#else1217
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1218
+ b_ = simde__m128_to_private(b);1219
+1220
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1221
+ r_.neon_u32 = vcltq_f32(a_.neon_f32, b_.neon_f32);1222
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1223
+ r_.wasm_v128 = wasm_f32x4_lt(a_.wasm_v128, b_.wasm_v128);1224
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1225
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1226
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1227
+ vec_cmplt(a_.altivec_f32, b_.altivec_f32));1228
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1229
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 < b_.f32));1230
+#else1231
+ SIMDE_VECTORIZE1232
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1233
+ r_.u32[i] = (a_.f32[i] < b_.f32[i]) ? ~UINT32_C(0)1234
+ : UINT32_C(0);1235
+ }1236
+#endif1237
+1238
+ return simde__m128_from_private(r_);1239
+#endif1240
+}1241
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1242
+#define _mm_cmplt_ps(a, b) simde_mm_cmplt_ps((a), (b))1243
+#endif1244
+1245
+SIMDE_FUNCTION_ATTRIBUTES1246
+simde__m128 simde_mm_cmplt_ss(simde__m128 a, simde__m128 b)1247
+{1248
+#if defined(SIMDE_X86_SSE_NATIVE)1249
+ return _mm_cmplt_ss(a, b);1250
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1251
+ return simde_mm_move_ss(a, simde_mm_cmplt_ps(a, b));1252
+#else1253
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1254
+ b_ = simde__m128_to_private(b);1255
+1256
+ r_.u32[0] = (a_.f32[0] < b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);1257
+ SIMDE_VECTORIZE1258
+ for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1259
+ r_.u32[i] = a_.u32[i];1260
+ }1261
+1262
+ return simde__m128_from_private(r_);1263
+#endif1264
+}1265
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1266
+#define _mm_cmplt_ss(a, b) simde_mm_cmplt_ss((a), (b))1267
+#endif1268
+1269
+SIMDE_FUNCTION_ATTRIBUTES1270
+simde__m128 simde_mm_cmpneq_ps(simde__m128 a, simde__m128 b)1271
+{1272
+#if defined(SIMDE_X86_SSE_NATIVE)1273
+ return _mm_cmpneq_ps(a, b);1274
+#else1275
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1276
+ b_ = simde__m128_to_private(b);1277
+1278
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1279
+ r_.neon_u32 = vmvnq_u32(vceqq_f32(a_.neon_f32, b_.neon_f32));1280
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1281
+ r_.wasm_v128 = wasm_f32x4_ne(a_.wasm_v128, b_.wasm_v128);1282
+#elif defined(SIMDE_POWER_ALTIVEC_P9_NATIVE) && SIMDE_ARCH_POWER_CHECK(900) && \1283
+ !defined(HEDLEY_IBM_VERSION)1284
+ /* vec_cmpne(SIMDE_POWER_ALTIVEC_VECTOR(float), SIMDE_POWER_ALTIVEC_VECTOR(float))1285
+ is missing from XL C/C++ v16.1.1,1286
+ though the documentation (table 89 on page 432 of the IBM XL C/C++ for1287
+ Linux Compiler Reference, Version 16.1.1) shows that it should be1288
+ present. Both GCC and clang support it. */1289
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1290
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1291
+ vec_cmpne(a_.altivec_f32, b_.altivec_f32));1292
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)1293
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1294
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1295
+ vec_cmpeq(a_.altivec_f32, b_.altivec_f32));1296
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1297
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1298
+ vec_nor(r_.altivec_f32, r_.altivec_f32));1299
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1300
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.f32 != b_.f32));1301
+#else1302
+ SIMDE_VECTORIZE1303
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1304
+ r_.u32[i] = (a_.f32[i] != b_.f32[i]) ? ~UINT32_C(0)1305
+ : UINT32_C(0);1306
+ }1307
+#endif1308
+1309
+ return simde__m128_from_private(r_);1310
+#endif1311
+}1312
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1313
+#define _mm_cmpneq_ps(a, b) simde_mm_cmpneq_ps((a), (b))1314
+#endif1315
+1316
+SIMDE_FUNCTION_ATTRIBUTES1317
+simde__m128 simde_mm_cmpneq_ss(simde__m128 a, simde__m128 b)1318
+{1319
+#if defined(SIMDE_X86_SSE_NATIVE)1320
+ return _mm_cmpneq_ss(a, b);1321
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1322
+ return simde_mm_move_ss(a, simde_mm_cmpneq_ps(a, b));1323
+#else1324
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1325
+ b_ = simde__m128_to_private(b);1326
+1327
+ r_.u32[0] = (a_.f32[0] != b_.f32[0]) ? ~UINT32_C(0) : UINT32_C(0);1328
+ SIMDE_VECTORIZE1329
+ for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1330
+ r_.u32[i] = a_.u32[i];1331
+ }1332
+1333
+ return simde__m128_from_private(r_);1334
+#endif1335
+}1336
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1337
+#define _mm_cmpneq_ss(a, b) simde_mm_cmpneq_ss((a), (b))1338
+#endif1339
+1340
+SIMDE_FUNCTION_ATTRIBUTES1341
+simde__m128 simde_mm_cmpnge_ps(simde__m128 a, simde__m128 b)1342
+{1343
+ return simde_mm_cmplt_ps(a, b);1344
+}1345
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1346
+#define _mm_cmpnge_ps(a, b) simde_mm_cmpnge_ps((a), (b))1347
+#endif1348
+1349
+SIMDE_FUNCTION_ATTRIBUTES1350
+simde__m128 simde_mm_cmpnge_ss(simde__m128 a, simde__m128 b)1351
+{1352
+ return simde_mm_cmplt_ss(a, b);1353
+}1354
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1355
+#define _mm_cmpnge_ss(a, b) simde_mm_cmpnge_ss((a), (b))1356
+#endif1357
+1358
+SIMDE_FUNCTION_ATTRIBUTES1359
+simde__m128 simde_mm_cmpngt_ps(simde__m128 a, simde__m128 b)1360
+{1361
+ return simde_mm_cmple_ps(a, b);1362
+}1363
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1364
+#define _mm_cmpngt_ps(a, b) simde_mm_cmpngt_ps((a), (b))1365
+#endif1366
+1367
+SIMDE_FUNCTION_ATTRIBUTES1368
+simde__m128 simde_mm_cmpngt_ss(simde__m128 a, simde__m128 b)1369
+{1370
+ return simde_mm_cmple_ss(a, b);1371
+}1372
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1373
+#define _mm_cmpngt_ss(a, b) simde_mm_cmpngt_ss((a), (b))1374
+#endif1375
+1376
+SIMDE_FUNCTION_ATTRIBUTES1377
+simde__m128 simde_mm_cmpnle_ps(simde__m128 a, simde__m128 b)1378
+{1379
+ return simde_mm_cmpgt_ps(a, b);1380
+}1381
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1382
+#define _mm_cmpnle_ps(a, b) simde_mm_cmpnle_ps((a), (b))1383
+#endif1384
+1385
+SIMDE_FUNCTION_ATTRIBUTES1386
+simde__m128 simde_mm_cmpnle_ss(simde__m128 a, simde__m128 b)1387
+{1388
+ return simde_mm_cmpgt_ss(a, b);1389
+}1390
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1391
+#define _mm_cmpnle_ss(a, b) simde_mm_cmpnle_ss((a), (b))1392
+#endif1393
+1394
+SIMDE_FUNCTION_ATTRIBUTES1395
+simde__m128 simde_mm_cmpnlt_ps(simde__m128 a, simde__m128 b)1396
+{1397
+ return simde_mm_cmpge_ps(a, b);1398
+}1399
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1400
+#define _mm_cmpnlt_ps(a, b) simde_mm_cmpnlt_ps((a), (b))1401
+#endif1402
+1403
+SIMDE_FUNCTION_ATTRIBUTES1404
+simde__m128 simde_mm_cmpnlt_ss(simde__m128 a, simde__m128 b)1405
+{1406
+ return simde_mm_cmpge_ss(a, b);1407
+}1408
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1409
+#define _mm_cmpnlt_ss(a, b) simde_mm_cmpnlt_ss((a), (b))1410
+#endif1411
+1412
+SIMDE_FUNCTION_ATTRIBUTES1413
+simde__m128 simde_mm_cmpord_ps(simde__m128 a, simde__m128 b)1414
+{1415
+#if defined(SIMDE_X86_SSE_NATIVE)1416
+ return _mm_cmpord_ps(a, b);1417
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1418
+ return wasm_v128_and(wasm_f32x4_eq(a, a), wasm_f32x4_eq(b, b));1419
+#else1420
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1421
+ b_ = simde__m128_to_private(b);1422
+1423
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1424
+ /* Note: NEON does not have ordered compare builtin1425
+ Need to compare a eq a and b eq b to check for NaN1426
+ Do AND of results to get final */1427
+ uint32x4_t ceqaa = vceqq_f32(a_.neon_f32, a_.neon_f32);1428
+ uint32x4_t ceqbb = vceqq_f32(b_.neon_f32, b_.neon_f32);1429
+ r_.neon_u32 = vandq_u32(ceqaa, ceqbb);1430
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1431
+ r_.wasm_v128 = wasm_v128_and(wasm_f32x4_eq(a_.wasm_v128, a_.wasm_v128),1432
+ wasm_f32x4_eq(b_.wasm_v128, b_.wasm_v128));1433
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1434
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1435
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1436
+ vec_and(vec_cmpeq(a_.altivec_f32, a_.altivec_f32),1437
+ vec_cmpeq(b_.altivec_f32, b_.altivec_f32)));1438
+#elif defined(simde_math_isnanf)1439
+ SIMDE_VECTORIZE1440
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1441
+ r_.u32[i] = (simde_math_isnanf(a_.f32[i]) ||1442
+ simde_math_isnanf(b_.f32[i]))1443
+ ? UINT32_C(0)1444
+ : ~UINT32_C(0);1445
+ }1446
+#else1447
+ HEDLEY_UNREACHABLE();1448
+#endif1449
+1450
+ return simde__m128_from_private(r_);1451
+#endif1452
+}1453
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1454
+#define _mm_cmpord_ps(a, b) simde_mm_cmpord_ps((a), (b))1455
+#endif1456
+1457
+SIMDE_FUNCTION_ATTRIBUTES1458
+simde__m128 simde_mm_cmpunord_ps(simde__m128 a, simde__m128 b)1459
+{1460
+#if defined(SIMDE_X86_SSE_NATIVE)1461
+ return _mm_cmpunord_ps(a, b);1462
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1463
+ return wasm_v128_or(wasm_f32x4_ne(a, a), wasm_f32x4_ne(b, b));1464
+#else1465
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1466
+ b_ = simde__m128_to_private(b);1467
+1468
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1469
+ uint32x4_t ceqaa = vceqq_f32(a_.neon_f32, a_.neon_f32);1470
+ uint32x4_t ceqbb = vceqq_f32(b_.neon_f32, b_.neon_f32);1471
+ r_.neon_u32 = vmvnq_u32(vandq_u32(ceqaa, ceqbb));1472
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1473
+ r_.wasm_v128 = wasm_v128_or(wasm_f32x4_ne(a_.wasm_v128, a_.wasm_v128),1474
+ wasm_f32x4_ne(b_.wasm_v128, b_.wasm_v128));1475
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)1476
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1477
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1478
+ vec_nand(vec_cmpeq(a_.altivec_f32, a_.altivec_f32),1479
+ vec_cmpeq(b_.altivec_f32, b_.altivec_f32)));1480
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1481
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(1482
+ SIMDE_POWER_ALTIVEC_VECTOR(float),1483
+ vec_and(vec_cmpeq(a_.altivec_f32, a_.altivec_f32),1484
+ vec_cmpeq(b_.altivec_f32, b_.altivec_f32)));1485
+ r_.altivec_f32 = vec_nor(r_.altivec_f32, r_.altivec_f32);1486
+#elif defined(simde_math_isnanf)1487
+ SIMDE_VECTORIZE1488
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1489
+ r_.u32[i] = (simde_math_isnanf(a_.f32[i]) ||1490
+ simde_math_isnanf(b_.f32[i]))1491
+ ? ~UINT32_C(0)1492
+ : UINT32_C(0);1493
+ }1494
+#else1495
+ HEDLEY_UNREACHABLE();1496
+#endif1497
+1498
+ return simde__m128_from_private(r_);1499
+#endif1500
+}1501
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1502
+#define _mm_cmpunord_ps(a, b) simde_mm_cmpunord_ps((a), (b))1503
+#endif1504
+1505
+SIMDE_FUNCTION_ATTRIBUTES1506
+simde__m128 simde_mm_cmpunord_ss(simde__m128 a, simde__m128 b)1507
+{1508
+#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)1509
+ return _mm_cmpunord_ss(a, b);1510
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1511
+ return simde_mm_move_ss(a, simde_mm_cmpunord_ps(a, b));1512
+#else1513
+ simde__m128_private r_, a_ = simde__m128_to_private(a),1514
+ b_ = simde__m128_to_private(b);1515
+1516
+#if defined(simde_math_isnanf)1517
+ r_.u32[0] =1518
+ (simde_math_isnanf(a_.f32[0]) || simde_math_isnanf(b_.f32[0]))1519
+ ? ~UINT32_C(0)1520
+ : UINT32_C(0);1521
+ SIMDE_VECTORIZE1522
+ for (size_t i = 1; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {1523
+ r_.u32[i] = a_.u32[i];1524
+ }1525
+#else1526
+ HEDLEY_UNREACHABLE();1527
+#endif1528
+1529
+ return simde__m128_from_private(r_);1530
+#endif1531
+}1532
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1533
+#define _mm_cmpunord_ss(a, b) simde_mm_cmpunord_ss((a), (b))1534
+#endif1535
+1536
+SIMDE_FUNCTION_ATTRIBUTES1537
+int simde_mm_comieq_ss(simde__m128 a, simde__m128 b)1538
+{1539
+#if defined(SIMDE_X86_SSE_NATIVE)1540
+ return _mm_comieq_ss(a, b);1541
+#else1542
+ simde__m128_private a_ = simde__m128_to_private(a),1543
+ b_ = simde__m128_to_private(b);1544
+1545
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1546
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1547
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1548
+ uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));1549
+ uint32x4_t a_eq_b = vceqq_f32(a_.neon_f32, b_.neon_f32);1550
+ return !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_eq_b), 0) != 0);1551
+#else1552
+ return a_.f32[0] == b_.f32[0];1553
+#endif1554
+#endif1555
+}1556
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1557
+#define _mm_comieq_ss(a, b) simde_mm_comieq_ss((a), (b))1558
+#endif1559
+1560
+SIMDE_FUNCTION_ATTRIBUTES1561
+int simde_mm_comige_ss(simde__m128 a, simde__m128 b)1562
+{1563
+#if defined(SIMDE_X86_SSE_NATIVE)1564
+ return _mm_comige_ss(a, b);1565
+#else1566
+ simde__m128_private a_ = simde__m128_to_private(a),1567
+ b_ = simde__m128_to_private(b);1568
+1569
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1570
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1571
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1572
+ uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);1573
+ uint32x4_t a_ge_b = vcgeq_f32(a_.neon_f32, b_.neon_f32);1574
+ return !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_ge_b), 0) != 0);1575
+#else1576
+ return a_.f32[0] >= b_.f32[0];1577
+#endif1578
+#endif1579
+}1580
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1581
+#define _mm_comige_ss(a, b) simde_mm_comige_ss((a), (b))1582
+#endif1583
+1584
+SIMDE_FUNCTION_ATTRIBUTES1585
+int simde_mm_comigt_ss(simde__m128 a, simde__m128 b)1586
+{1587
+#if defined(SIMDE_X86_SSE_NATIVE)1588
+ return _mm_comigt_ss(a, b);1589
+#else1590
+ simde__m128_private a_ = simde__m128_to_private(a),1591
+ b_ = simde__m128_to_private(b);1592
+1593
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1594
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1595
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1596
+ uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);1597
+ uint32x4_t a_gt_b = vcgtq_f32(a_.neon_f32, b_.neon_f32);1598
+ return !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_gt_b), 0) != 0);1599
+#else1600
+ return a_.f32[0] > b_.f32[0];1601
+#endif1602
+#endif1603
+}1604
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1605
+#define _mm_comigt_ss(a, b) simde_mm_comigt_ss((a), (b))1606
+#endif1607
+1608
+SIMDE_FUNCTION_ATTRIBUTES1609
+int simde_mm_comile_ss(simde__m128 a, simde__m128 b)1610
+{1611
+#if defined(SIMDE_X86_SSE_NATIVE)1612
+ return _mm_comile_ss(a, b);1613
+#else1614
+ simde__m128_private a_ = simde__m128_to_private(a),1615
+ b_ = simde__m128_to_private(b);1616
+1617
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1618
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1619
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1620
+ uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));1621
+ uint32x4_t a_le_b = vcleq_f32(a_.neon_f32, b_.neon_f32);1622
+ return !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_le_b), 0) != 0);1623
+#else1624
+ return a_.f32[0] <= b_.f32[0];1625
+#endif1626
+#endif1627
+}1628
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1629
+#define _mm_comile_ss(a, b) simde_mm_comile_ss((a), (b))1630
+#endif1631
+1632
+SIMDE_FUNCTION_ATTRIBUTES1633
+int simde_mm_comilt_ss(simde__m128 a, simde__m128 b)1634
+{1635
+#if defined(SIMDE_X86_SSE_NATIVE)1636
+ return _mm_comilt_ss(a, b);1637
+#else1638
+ simde__m128_private a_ = simde__m128_to_private(a),1639
+ b_ = simde__m128_to_private(b);1640
+1641
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1642
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1643
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1644
+ uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));1645
+ uint32x4_t a_lt_b = vcltq_f32(a_.neon_f32, b_.neon_f32);1646
+ return !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_lt_b), 0) != 0);1647
+#else1648
+ return a_.f32[0] < b_.f32[0];1649
+#endif1650
+#endif1651
+}1652
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1653
+#define _mm_comilt_ss(a, b) simde_mm_comilt_ss((a), (b))1654
+#endif1655
+1656
+SIMDE_FUNCTION_ATTRIBUTES1657
+int simde_mm_comineq_ss(simde__m128 a, simde__m128 b)1658
+{1659
+#if defined(SIMDE_X86_SSE_NATIVE)1660
+ return _mm_comineq_ss(a, b);1661
+#else1662
+ simde__m128_private a_ = simde__m128_to_private(a),1663
+ b_ = simde__m128_to_private(b);1664
+1665
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1666
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);1667
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);1668
+ uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);1669
+ uint32x4_t a_neq_b = vmvnq_u32(vceqq_f32(a_.neon_f32, b_.neon_f32));1670
+ return !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_neq_b), 0) != 0);1671
+#else1672
+ return a_.f32[0] != b_.f32[0];1673
+#endif1674
+#endif1675
+}1676
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1677
+#define _mm_comineq_ss(a, b) simde_mm_comineq_ss((a), (b))1678
+#endif1679
+1680
+SIMDE_FUNCTION_ATTRIBUTES1681
+simde__m128 simde_x_mm_copysign_ps(simde__m128 dest, simde__m128 src)1682
+{1683
+ simde__m128_private r_, dest_ = simde__m128_to_private(dest),1684
+ src_ = simde__m128_to_private(src);1685
+1686
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1687
+ const uint32x4_t sign_pos =1688
+ vreinterpretq_u32_f32(vdupq_n_f32(-SIMDE_FLOAT32_C(0.0)));1689
+ r_.neon_u32 = vbslq_u32(sign_pos, src_.neon_u32, dest_.neon_u32);1690
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1691
+ const v128_t sign_pos = wasm_f32x4_splat(-0.0f);1692
+ r_.wasm_v128 =1693
+ wasm_v128_bitselect(src_.wasm_v128, dest_.wasm_v128, sign_pos);1694
+#elif defined(SIMDE_POWER_ALTIVEC_P9_NATIVE)1695
+#if !defined(HEDLEY_IBM_VERSION)1696
+ r_.altivec_f32 = vec_cpsgn(dest_.altivec_f32, src_.altivec_f32);1697
+#else1698
+ r_.altivec_f32 = vec_cpsgn(src_.altivec_f32, dest_.altivec_f32);1699
+#endif1700
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1701
+ const SIMDE_POWER_ALTIVEC_VECTOR(unsigned int)1702
+ sign_pos = HEDLEY_REINTERPRET_CAST(1703
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned int),1704
+ vec_splats(-0.0f));1705
+ r_.altivec_f32 = vec_sel(dest_.altivec_f32, src_.altivec_f32, sign_pos);1706
+#elif defined(SIMDE_IEEE754_STORAGE)1707
+ (void)src_;1708
+ (void)dest_;1709
+ simde__m128 sign_pos = simde_mm_set1_ps(-0.0f);1710
+ r_ = simde__m128_to_private(simde_mm_xor_ps(1711
+ dest, simde_mm_and_ps(simde_mm_xor_ps(dest, src), sign_pos)));1712
+#else1713
+ SIMDE_VECTORIZE1714
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1715
+ r_.f32[i] = simde_math_copysignf(dest_.f32[i], src_.f32[i]);1716
+ }1717
+#endif1718
+1719
+ return simde__m128_from_private(r_);1720
+}1721
+1722
+SIMDE_FUNCTION_ATTRIBUTES1723
+simde__m128 simde_x_mm_xorsign_ps(simde__m128 dest, simde__m128 src)1724
+{1725
+ return simde_mm_xor_ps(simde_mm_and_ps(simde_mm_set1_ps(-0.0f), src),1726
+ dest);1727
+}1728
+1729
+SIMDE_FUNCTION_ATTRIBUTES1730
+simde__m128 simde_mm_cvt_pi2ps(simde__m128 a, simde__m64 b)1731
+{1732
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1733
+ return _mm_cvt_pi2ps(a, b);1734
+#else1735
+ simde__m128_private r_, a_ = simde__m128_to_private(a);1736
+ simde__m64_private b_ = simde__m64_to_private(b);1737
+1738
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1739
+ r_.neon_f32 = vcombine_f32(vcvt_f32_s32(b_.neon_i32),1740
+ vget_high_f32(a_.neon_f32));1741
+#elif defined(SIMDE_CONVERT_VECTOR_)1742
+ SIMDE_CONVERT_VECTOR_(r_.m64_private[0].f32, b_.i32);1743
+ r_.m64_private[1] = a_.m64_private[1];1744
+#else1745
+ r_.f32[0] = (simde_float32)b_.i32[0];1746
+ r_.f32[1] = (simde_float32)b_.i32[1];1747
+ r_.i32[2] = a_.i32[2];1748
+ r_.i32[3] = a_.i32[3];1749
+#endif1750
+1751
+ return simde__m128_from_private(r_);1752
+#endif1753
+}1754
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1755
+#define _mm_cvt_pi2ps(a, b) simde_mm_cvt_pi2ps((a), (b))1756
+#endif1757
+1758
+SIMDE_FUNCTION_ATTRIBUTES1759
+simde__m64 simde_mm_cvt_ps2pi(simde__m128 a)1760
+{1761
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1762
+ return _mm_cvt_ps2pi(a);1763
+#else1764
+ simde__m64_private r_;1765
+ simde__m128_private a_;1766
+1767
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1768
+ a_ = simde__m128_to_private(1769
+ simde_mm_round_ps(a, SIMDE_MM_FROUND_CUR_DIRECTION));1770
+ r_.neon_i32 = vcvt_s32_f32(vget_low_f32(a_.neon_f32));1771
+#elif defined(SIMDE_CONVERT_VECTOR_) && SIMDE_NATURAL_VECTOR_SIZE_GE(128)1772
+ a_ = simde__m128_to_private(1773
+ simde_mm_round_ps(a, SIMDE_MM_FROUND_CUR_DIRECTION));1774
+ SIMDE_CONVERT_VECTOR_(r_.i32, a_.m64_private[0].f32);1775
+#else1776
+ a_ = simde__m128_to_private(a);1777
+1778
+ SIMDE_VECTORIZE1779
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1780
+ r_.i32[i] = HEDLEY_STATIC_CAST(1781
+ int32_t, simde_math_nearbyintf(a_.f32[i]));1782
+ }1783
+#endif1784
+1785
+ return simde__m64_from_private(r_);1786
+#endif1787
+}1788
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1789
+#define _mm_cvt_ps2pi(a) simde_mm_cvt_ps2pi((a))1790
+#endif1791
+1792
+SIMDE_FUNCTION_ATTRIBUTES1793
+simde__m128 simde_mm_cvt_si2ss(simde__m128 a, int32_t b)1794
+{1795
+#if defined(SIMDE_X86_SSE_NATIVE)1796
+ return _mm_cvt_si2ss(a, b);1797
+#else1798
+ simde__m128_private r_, a_ = simde__m128_to_private(a);1799
+1800
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1801
+ r_.neon_f32 =1802
+ vsetq_lane_f32(HEDLEY_STATIC_CAST(float, b), a_.neon_f32, 0);1803
+#else1804
+ r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, b);1805
+ r_.i32[1] = a_.i32[1];1806
+ r_.i32[2] = a_.i32[2];1807
+ r_.i32[3] = a_.i32[3];1808
+#endif1809
+1810
+ return simde__m128_from_private(r_);1811
+#endif1812
+}1813
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1814
+#define _mm_cvt_si2ss(a, b) simde_mm_cvt_si2ss((a), b)1815
+#endif1816
+1817
+SIMDE_FUNCTION_ATTRIBUTES1818
+int32_t simde_mm_cvt_ss2si(simde__m128 a)1819
+{1820
+#if defined(SIMDE_X86_SSE_NATIVE)1821
+ return _mm_cvt_ss2si(a);1822
+#elif defined(SIMDE_ARM_NEON_A32V8_NATIVE) && !defined(SIMDE_BUG_GCC_95399)1823
+ return vgetq_lane_s32(vcvtnq_s32_f32(simde__m128_to_neon_f32(a)), 0);1824
+#else1825
+ simde__m128_private a_ = simde__m128_to_private(1826
+ simde_mm_round_ps(a, SIMDE_MM_FROUND_CUR_DIRECTION));1827
+ return SIMDE_CONVERT_FTOI(int32_t, a_.f32[0]);1828
+#endif1829
+}1830
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1831
+#define _mm_cvt_ss2si(a) simde_mm_cvt_ss2si((a))1832
+#endif1833
+1834
+SIMDE_FUNCTION_ATTRIBUTES1835
+simde__m128 simde_mm_cvtpi16_ps(simde__m64 a)1836
+{1837
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1838
+ return _mm_cvtpi16_ps(a);1839
+#else1840
+ simde__m128_private r_;1841
+ simde__m64_private a_ = simde__m64_to_private(a);1842
+1843
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1844
+ r_.neon_f32 = vcvtq_f32_s32(vmovl_s16(a_.neon_i16));1845
+#elif defined(SIMDE_CONVERT_VECTOR_)1846
+ SIMDE_CONVERT_VECTOR_(r_.f32, a_.i16);1847
+#else1848
+ SIMDE_VECTORIZE1849
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {1850
+ simde_float32 v = a_.i16[i];1851
+ r_.f32[i] = v;1852
+ }1853
+#endif1854
+1855
+ return simde__m128_from_private(r_);1856
+#endif1857
+}1858
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1859
+#define _mm_cvtpi16_ps(a) simde_mm_cvtpi16_ps(a)1860
+#endif1861
+1862
+SIMDE_FUNCTION_ATTRIBUTES1863
+simde__m128 simde_mm_cvtpi32_ps(simde__m128 a, simde__m64 b)1864
+{1865
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1866
+ return _mm_cvtpi32_ps(a, b);1867
+#else1868
+ simde__m128_private r_, a_ = simde__m128_to_private(a);1869
+ simde__m64_private b_ = simde__m64_to_private(b);1870
+1871
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1872
+ r_.neon_f32 = vcombine_f32(vcvt_f32_s32(b_.neon_i32),1873
+ vget_high_f32(a_.neon_f32));1874
+#elif defined(SIMDE_CONVERT_VECTOR_)1875
+ SIMDE_CONVERT_VECTOR_(r_.m64_private[0].f32, b_.i32);1876
+ r_.m64_private[1] = a_.m64_private[1];1877
+#else1878
+ r_.f32[0] = (simde_float32)b_.i32[0];1879
+ r_.f32[1] = (simde_float32)b_.i32[1];1880
+ r_.i32[2] = a_.i32[2];1881
+ r_.i32[3] = a_.i32[3];1882
+#endif1883
+1884
+ return simde__m128_from_private(r_);1885
+#endif1886
+}1887
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1888
+#define _mm_cvtpi32_ps(a, b) simde_mm_cvtpi32_ps((a), b)1889
+#endif1890
+1891
+SIMDE_FUNCTION_ATTRIBUTES1892
+simde__m128 simde_mm_cvtpi32x2_ps(simde__m64 a, simde__m64 b)1893
+{1894
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1895
+ return _mm_cvtpi32x2_ps(a, b);1896
+#else1897
+ simde__m128_private r_;1898
+ simde__m64_private a_ = simde__m64_to_private(a),1899
+ b_ = simde__m64_to_private(b);1900
+1901
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1902
+ r_.neon_f32 = vcvtq_f32_s32(vcombine_s32(a_.neon_i32, b_.neon_i32));1903
+#elif defined(SIMDE_CONVERT_VECTOR_)1904
+ SIMDE_CONVERT_VECTOR_(r_.m64_private[0].f32, a_.i32);1905
+ SIMDE_CONVERT_VECTOR_(r_.m64_private[1].f32, b_.i32);1906
+#else1907
+ r_.f32[0] = (simde_float32)a_.i32[0];1908
+ r_.f32[1] = (simde_float32)a_.i32[1];1909
+ r_.f32[2] = (simde_float32)b_.i32[0];1910
+ r_.f32[3] = (simde_float32)b_.i32[1];1911
+#endif1912
+1913
+ return simde__m128_from_private(r_);1914
+#endif1915
+}1916
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1917
+#define _mm_cvtpi32x2_ps(a, b) simde_mm_cvtpi32x2_ps(a, b)1918
+#endif1919
+1920
+SIMDE_FUNCTION_ATTRIBUTES1921
+simde__m128 simde_mm_cvtpi8_ps(simde__m64 a)1922
+{1923
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1924
+ return _mm_cvtpi8_ps(a);1925
+#else1926
+ simde__m128_private r_;1927
+ simde__m64_private a_ = simde__m64_to_private(a);1928
+1929
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1930
+ r_.neon_f32 =1931
+ vcvtq_f32_s32(vmovl_s16(vget_low_s16(vmovl_s8(a_.neon_i8))));1932
+#else1933
+ r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, a_.i8[0]);1934
+ r_.f32[1] = HEDLEY_STATIC_CAST(simde_float32, a_.i8[1]);1935
+ r_.f32[2] = HEDLEY_STATIC_CAST(simde_float32, a_.i8[2]);1936
+ r_.f32[3] = HEDLEY_STATIC_CAST(simde_float32, a_.i8[3]);1937
+#endif1938
+1939
+ return simde__m128_from_private(r_);1940
+#endif1941
+}1942
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1943
+#define _mm_cvtpi8_ps(a) simde_mm_cvtpi8_ps(a)1944
+#endif1945
+1946
+SIMDE_FUNCTION_ATTRIBUTES1947
+simde__m64 simde_mm_cvtps_pi16(simde__m128 a)1948
+{1949
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1950
+ return _mm_cvtps_pi16(a);1951
+#else1952
+ simde__m64_private r_;1953
+ simde__m128_private a_ = simde__m128_to_private(a);1954
+1955
+#if defined(SIMDE_ARM_NEON_A32V8_NATIVE) && !defined(SIMDE_BUG_GCC_95399)1956
+ r_.neon_i16 = vmovn_s32(vcvtq_s32_f32(vrndiq_f32(a_.neon_f32)));1957
+#else1958
+ SIMDE_VECTORIZE1959
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1960
+ r_.i16[i] = SIMDE_CONVERT_FTOI(int16_t,1961
+ simde_math_roundf(a_.f32[i]));1962
+ }1963
+#endif1964
+1965
+ return simde__m64_from_private(r_);1966
+#endif1967
+}1968
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)1969
+#define _mm_cvtps_pi16(a) simde_mm_cvtps_pi16((a))1970
+#endif1971
+1972
+SIMDE_FUNCTION_ATTRIBUTES1973
+simde__m64 simde_mm_cvtps_pi32(simde__m128 a)1974
+{1975
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)1976
+ return _mm_cvtps_pi32(a);1977
+#else1978
+ simde__m64_private r_;1979
+ simde__m128_private a_ = simde__m128_to_private(a);1980
+1981
+#if defined(SIMDE_ARM_NEON_A32V8_NATIVE) && \1982
+ defined(SIMDE_FAST_CONVERSION_RANGE) && !defined(SIMDE_BUG_GCC_95399)1983
+ r_.neon_i32 = vcvt_s32_f32(vget_low_f32(vrndiq_f32(a_.neon_f32)));1984
+#else1985
+ SIMDE_VECTORIZE1986
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1987
+ simde_float32 v = simde_math_roundf(a_.f32[i]);1988
+#if !defined(SIMDE_FAST_CONVERSION_RANGE)1989
+ r_.i32[i] =1990
+ ((v > HEDLEY_STATIC_CAST(simde_float32, INT32_MIN)) &&1991
+ (v < HEDLEY_STATIC_CAST(simde_float32, INT32_MAX)))1992
+ ? SIMDE_CONVERT_FTOI(int32_t, v)1993
+ : INT32_MIN;1994
+#else1995
+ r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, v);1996
+#endif1997
+ }1998
+#endif1999
+2000
+ return simde__m64_from_private(r_);2001
+#endif2002
+}2003
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2004
+#define _mm_cvtps_pi32(a) simde_mm_cvtps_pi32((a))2005
+#endif2006
+2007
+SIMDE_FUNCTION_ATTRIBUTES2008
+simde__m64 simde_mm_cvtps_pi8(simde__m128 a)2009
+{2010
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2011
+ return _mm_cvtps_pi8(a);2012
+#else2013
+ simde__m64_private r_;2014
+ simde__m128_private a_ = simde__m128_to_private(a);2015
+2016
+#if defined(SIMDE_ARM_NEON_A32V8_NATIVE) && !defined(SIMDE_BUG_GCC_95471)2017
+ /* Clamp the input to [INT8_MIN, INT8_MAX], round, convert to i32, narrow to2018
+ * i16, combine with an all-zero vector of i16 (which will become the upper2019
+ * half), narrow to i8. */2020
+ float32x4_t max =2021
+ vdupq_n_f32(HEDLEY_STATIC_CAST(simde_float32, INT8_MAX));2022
+ float32x4_t min =2023
+ vdupq_n_f32(HEDLEY_STATIC_CAST(simde_float32, INT8_MIN));2024
+ float32x4_t values =2025
+ vrndnq_f32(vmaxq_f32(vminq_f32(max, a_.neon_f32), min));2026
+ r_.neon_i8 = vmovn_s16(2027
+ vcombine_s16(vmovn_s32(vcvtq_s32_f32(values)), vdup_n_s16(0)));2028
+#else2029
+ SIMDE_VECTORIZE2030
+ for (size_t i = 0; i < (sizeof(a_.f32) / sizeof(a_.f32[0])); i++) {2031
+ if (a_.f32[i] > HEDLEY_STATIC_CAST(simde_float32, INT8_MAX))2032
+ r_.i8[i] = INT8_MAX;2033
+ else if (a_.f32[i] <2034
+ HEDLEY_STATIC_CAST(simde_float32, INT8_MIN))2035
+ r_.i8[i] = INT8_MIN;2036
+ else2037
+ r_.i8[i] = SIMDE_CONVERT_FTOI(2038
+ int8_t, simde_math_roundf(a_.f32[i]));2039
+ }2040
+ /* Note: the upper half is undefined */2041
+#endif2042
+2043
+ return simde__m64_from_private(r_);2044
+#endif2045
+}2046
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2047
+#define _mm_cvtps_pi8(a) simde_mm_cvtps_pi8((a))2048
+#endif2049
+2050
+SIMDE_FUNCTION_ATTRIBUTES2051
+simde__m128 simde_mm_cvtpu16_ps(simde__m64 a)2052
+{2053
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2054
+ return _mm_cvtpu16_ps(a);2055
+#else2056
+ simde__m128_private r_;2057
+ simde__m64_private a_ = simde__m64_to_private(a);2058
+2059
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2060
+ r_.neon_f32 = vcvtq_f32_u32(vmovl_u16(a_.neon_u16));2061
+#elif defined(SIMDE_CONVERT_VECTOR_)2062
+ SIMDE_CONVERT_VECTOR_(r_.f32, a_.u16);2063
+#else2064
+ SIMDE_VECTORIZE2065
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2066
+ r_.f32[i] = (simde_float32)a_.u16[i];2067
+ }2068
+#endif2069
+2070
+ return simde__m128_from_private(r_);2071
+#endif2072
+}2073
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2074
+#define _mm_cvtpu16_ps(a) simde_mm_cvtpu16_ps(a)2075
+#endif2076
+2077
+SIMDE_FUNCTION_ATTRIBUTES2078
+simde__m128 simde_mm_cvtpu8_ps(simde__m64 a)2079
+{2080
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2081
+ return _mm_cvtpu8_ps(a);2082
+#else2083
+ simde__m128_private r_;2084
+ simde__m64_private a_ = simde__m64_to_private(a);2085
+2086
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2087
+ r_.neon_f32 =2088
+ vcvtq_f32_u32(vmovl_u16(vget_low_u16(vmovl_u8(a_.neon_u8))));2089
+#else2090
+ SIMDE_VECTORIZE2091
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2092
+ r_.f32[i] = HEDLEY_STATIC_CAST(simde_float32, a_.u8[i]);2093
+ }2094
+#endif2095
+2096
+ return simde__m128_from_private(r_);2097
+#endif2098
+}2099
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2100
+#define _mm_cvtpu8_ps(a) simde_mm_cvtpu8_ps(a)2101
+#endif2102
+2103
+SIMDE_FUNCTION_ATTRIBUTES2104
+simde__m128 simde_mm_cvtsi32_ss(simde__m128 a, int32_t b)2105
+{2106
+#if defined(SIMDE_X86_SSE_NATIVE)2107
+ return _mm_cvtsi32_ss(a, b);2108
+#else2109
+ simde__m128_private r_;2110
+ simde__m128_private a_ = simde__m128_to_private(a);2111
+2112
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2113
+ r_.neon_f32 = vsetq_lane_f32(HEDLEY_STATIC_CAST(float32_t, b),2114
+ a_.neon_f32, 0);2115
+#else2116
+ r_ = a_;2117
+ r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, b);2118
+#endif2119
+2120
+ return simde__m128_from_private(r_);2121
+#endif2122
+}2123
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2124
+#define _mm_cvtsi32_ss(a, b) simde_mm_cvtsi32_ss((a), b)2125
+#endif2126
+2127
+SIMDE_FUNCTION_ATTRIBUTES2128
+simde__m128 simde_mm_cvtsi64_ss(simde__m128 a, int64_t b)2129
+{2130
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_ARCH_AMD64)2131
+#if !defined(__PGI)2132
+ return _mm_cvtsi64_ss(a, b);2133
+#else2134
+ return _mm_cvtsi64x_ss(a, b);2135
+#endif2136
+#else2137
+ simde__m128_private r_;2138
+ simde__m128_private a_ = simde__m128_to_private(a);2139
+2140
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2141
+ r_.neon_f32 = vsetq_lane_f32(HEDLEY_STATIC_CAST(float32_t, b),2142
+ a_.neon_f32, 0);2143
+#else2144
+ r_ = a_;2145
+ r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, b);2146
+#endif2147
+2148
+ return simde__m128_from_private(r_);2149
+#endif2150
+}2151
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2152
+#define _mm_cvtsi64_ss(a, b) simde_mm_cvtsi64_ss((a), b)2153
+#endif2154
+2155
+SIMDE_FUNCTION_ATTRIBUTES2156
+simde_float32 simde_mm_cvtss_f32(simde__m128 a)2157
+{2158
+#if defined(SIMDE_X86_SSE_NATIVE)2159
+ return _mm_cvtss_f32(a);2160
+#else2161
+ simde__m128_private a_ = simde__m128_to_private(a);2162
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2163
+ return vgetq_lane_f32(a_.neon_f32, 0);2164
+#else2165
+ return a_.f32[0];2166
+#endif2167
+#endif2168
+}2169
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2170
+#define _mm_cvtss_f32(a) simde_mm_cvtss_f32((a))2171
+#endif2172
+2173
+SIMDE_FUNCTION_ATTRIBUTES2174
+int32_t simde_mm_cvtss_si32(simde__m128 a)2175
+{2176
+ return simde_mm_cvt_ss2si(a);2177
+}2178
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2179
+#define _mm_cvtss_si32(a) simde_mm_cvtss_si32((a))2180
+#endif2181
+2182
+SIMDE_FUNCTION_ATTRIBUTES2183
+int64_t simde_mm_cvtss_si64(simde__m128 a)2184
+{2185
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_ARCH_AMD64)2186
+#if !defined(__PGI)2187
+ return _mm_cvtss_si64(a);2188
+#else2189
+ return _mm_cvtss_si64x(a);2190
+#endif2191
+#else2192
+ simde__m128_private a_ = simde__m128_to_private(a);2193
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2194
+ return SIMDE_CONVERT_FTOI(2195
+ int64_t, simde_math_roundf(vgetq_lane_f32(a_.neon_f32, 0)));2196
+#else2197
+ return SIMDE_CONVERT_FTOI(int64_t, simde_math_roundf(a_.f32[0]));2198
+#endif2199
+#endif2200
+}2201
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2202
+#define _mm_cvtss_si64(a) simde_mm_cvtss_si64((a))2203
+#endif2204
+2205
+SIMDE_FUNCTION_ATTRIBUTES2206
+simde__m64 simde_mm_cvtt_ps2pi(simde__m128 a)2207
+{2208
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2209
+ return _mm_cvtt_ps2pi(a);2210
+#else2211
+ simde__m64_private r_;2212
+ simde__m128_private a_ = simde__m128_to_private(a);2213
+2214
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE) && defined(SIMDE_FAST_CONVERSION_RANGE)2215
+ r_.neon_i32 = vcvt_s32_f32(vget_low_f32(a_.neon_f32));2216
+#else2217
+ SIMDE_VECTORIZE2218
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2219
+ simde_float32 v = a_.f32[i];2220
+#if !defined(SIMDE_FAST_CONVERSION_RANGE)2221
+ r_.i32[i] =2222
+ ((v > HEDLEY_STATIC_CAST(simde_float32, INT32_MIN)) &&2223
+ (v < HEDLEY_STATIC_CAST(simde_float32, INT32_MAX)))2224
+ ? SIMDE_CONVERT_FTOI(int32_t, v)2225
+ : INT32_MIN;2226
+#else2227
+ r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, v);2228
+#endif2229
+ }2230
+#endif2231
+2232
+ return simde__m64_from_private(r_);2233
+#endif2234
+}2235
+#define simde_mm_cvttps_pi32(a) simde_mm_cvtt_ps2pi(a)2236
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2237
+#define _mm_cvtt_ps2pi(a) simde_mm_cvtt_ps2pi((a))2238
+#define _mm_cvttps_pi32(a) simde_mm_cvttps_pi32((a))2239
+#endif2240
+2241
+SIMDE_FUNCTION_ATTRIBUTES2242
+int32_t simde_mm_cvtt_ss2si(simde__m128 a)2243
+{2244
+#if defined(SIMDE_X86_SSE_NATIVE)2245
+ return _mm_cvtt_ss2si(a);2246
+#else2247
+ simde__m128_private a_ = simde__m128_to_private(a);2248
+2249
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE) && defined(SIMDE_FAST_CONVERSION_RANGE)2250
+ return SIMDE_CONVERT_FTOI(int32_t, vgetq_lane_f32(a_.neon_f32, 0));2251
+#else2252
+ simde_float32 v = a_.f32[0];2253
+#if !defined(SIMDE_FAST_CONVERSION_RANGE)2254
+ return ((v > HEDLEY_STATIC_CAST(simde_float32, INT32_MIN)) &&2255
+ (v < HEDLEY_STATIC_CAST(simde_float32, INT32_MAX)))2256
+ ? SIMDE_CONVERT_FTOI(int32_t, v)2257
+ : INT32_MIN;2258
+#else2259
+ return SIMDE_CONVERT_FTOI(int32_t, v);2260
+#endif2261
+#endif2262
+#endif2263
+}2264
+#define simde_mm_cvttss_si32(a) simde_mm_cvtt_ss2si((a))2265
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2266
+#define _mm_cvtt_ss2si(a) simde_mm_cvtt_ss2si((a))2267
+#define _mm_cvttss_si32(a) simde_mm_cvtt_ss2si((a))2268
+#endif2269
+2270
+SIMDE_FUNCTION_ATTRIBUTES2271
+int64_t simde_mm_cvttss_si64(simde__m128 a)2272
+{2273
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_ARCH_AMD64) && \2274
+ !defined(_MSC_VER)2275
+#if defined(__PGI)2276
+ return _mm_cvttss_si64x(a);2277
+#else2278
+ return _mm_cvttss_si64(a);2279
+#endif2280
+#else2281
+ simde__m128_private a_ = simde__m128_to_private(a);2282
+2283
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2284
+ return SIMDE_CONVERT_FTOI(int64_t, vgetq_lane_f32(a_.neon_f32, 0));2285
+#else2286
+ return SIMDE_CONVERT_FTOI(int64_t, a_.f32[0]);2287
+#endif2288
+#endif2289
+}2290
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2291
+#define _mm_cvttss_si64(a) simde_mm_cvttss_si64((a))2292
+#endif2293
+2294
+SIMDE_FUNCTION_ATTRIBUTES2295
+simde__m128 simde_mm_cmpord_ss(simde__m128 a, simde__m128 b)2296
+{2297
+#if defined(SIMDE_X86_SSE_NATIVE)2298
+ return _mm_cmpord_ss(a, b);2299
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2300
+ return simde_mm_move_ss(a, simde_mm_cmpord_ps(a, b));2301
+#else2302
+ simde__m128_private r_, a_ = simde__m128_to_private(a);2303
+2304
+#if defined(simde_math_isnanf)2305
+ r_.u32[0] = (simde_math_isnanf(simde_mm_cvtss_f32(a)) ||2306
+ simde_math_isnanf(simde_mm_cvtss_f32(b)))2307
+ ? UINT32_C(0)2308
+ : ~UINT32_C(0);2309
+ SIMDE_VECTORIZE2310
+ for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2311
+ r_.u32[i] = a_.u32[i];2312
+ }2313
+#else2314
+ HEDLEY_UNREACHABLE();2315
+#endif2316
+2317
+ return simde__m128_from_private(r_);2318
+#endif2319
+}2320
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2321
+#define _mm_cmpord_ss(a, b) simde_mm_cmpord_ss((a), (b))2322
+#endif2323
+2324
+SIMDE_FUNCTION_ATTRIBUTES2325
+simde__m128 simde_mm_div_ps(simde__m128 a, simde__m128 b)2326
+{2327
+#if defined(SIMDE_X86_SSE_NATIVE)2328
+ return _mm_div_ps(a, b);2329
+#else2330
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2331
+ b_ = simde__m128_to_private(b);2332
+2333
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2334
+ r_.neon_f32 = vdivq_f32(a_.neon_f32, b_.neon_f32);2335
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2336
+ float32x4_t recip0 = vrecpeq_f32(b_.neon_f32);2337
+ float32x4_t recip1 =2338
+ vmulq_f32(recip0, vrecpsq_f32(recip0, b_.neon_f32));2339
+ r_.neon_f32 = vmulq_f32(a_.neon_f32, recip1);2340
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2341
+ r_.wasm_v128 = wasm_f32x4_div(a_.wasm_v128, b_.wasm_v128);2342
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)2343
+ r_.altivec_f32 = vec_div(a_.altivec_f32, b_.altivec_f32);2344
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2345
+ r_.f32 = a_.f32 / b_.f32;2346
+#else2347
+ SIMDE_VECTORIZE2348
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2349
+ r_.f32[i] = a_.f32[i] / b_.f32[i];2350
+ }2351
+#endif2352
+2353
+ return simde__m128_from_private(r_);2354
+#endif2355
+}2356
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2357
+#define _mm_div_ps(a, b) simde_mm_div_ps((a), (b))2358
+#endif2359
+2360
+SIMDE_FUNCTION_ATTRIBUTES2361
+simde__m128 simde_mm_div_ss(simde__m128 a, simde__m128 b)2362
+{2363
+#if defined(SIMDE_X86_SSE_NATIVE)2364
+ return _mm_div_ss(a, b);2365
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2366
+ return simde_mm_move_ss(a, simde_mm_div_ps(a, b));2367
+#else2368
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2369
+ b_ = simde__m128_to_private(b);2370
+2371
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2372
+ float32_t value = vgetq_lane_f32(2373
+ simde__m128_to_private(simde_mm_div_ps(a, b)).neon_f32, 0);2374
+ r_.neon_f32 = vsetq_lane_f32(value, a_.neon_f32, 0);2375
+#else2376
+ r_.f32[0] = a_.f32[0] / b_.f32[0];2377
+ SIMDE_VECTORIZE2378
+ for (size_t i = 1; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2379
+ r_.f32[i] = a_.f32[i];2380
+ }2381
+#endif2382
+2383
+ return simde__m128_from_private(r_);2384
+#endif2385
+}2386
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2387
+#define _mm_div_ss(a, b) simde_mm_div_ss((a), (b))2388
+#endif2389
+2390
+SIMDE_FUNCTION_ATTRIBUTES2391
+int16_t simde_mm_extract_pi16(simde__m64 a, const int imm8)2392
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 3)2393
+{2394
+ simde__m64_private a_ = simde__m64_to_private(a);2395
+ return a_.i16[imm8];2396
+}2397
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \2398
+ !defined(HEDLEY_PGI_VERSION)2399
+#if defined(SIMDE_BUG_CLANG_44589)2400
+#define simde_mm_extract_pi16(a, imm8) \2401
+ (HEDLEY_DIAGNOSTIC_PUSH _Pragma( \2402
+ "clang diagnostic ignored \"-Wvector-conversion\"") \2403
+ HEDLEY_STATIC_CAST(int16_t, _mm_extract_pi16((a), (imm8))) \2404
+ HEDLEY_DIAGNOSTIC_POP)2405
+#else2406
+#define simde_mm_extract_pi16(a, imm8) \2407
+ HEDLEY_STATIC_CAST(int16_t, _mm_extract_pi16(a, imm8))2408
+#endif2409
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2410
+#define simde_mm_extract_pi16(a, imm8) \2411
+ vget_lane_s16(simde__m64_to_private(a).neon_i16, imm8)2412
+#endif2413
+#define simde_m_pextrw(a, imm8) simde_mm_extract_pi16(a, imm8)2414
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2415
+#define _mm_extract_pi16(a, imm8) simde_mm_extract_pi16((a), (imm8))2416
+#define _m_pextrw(a, imm8) simde_mm_extract_pi16((a), (imm8))2417
+#endif2418
+2419
+SIMDE_FUNCTION_ATTRIBUTES2420
+simde__m64 simde_mm_insert_pi16(simde__m64 a, int16_t i, const int imm8)2421
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 3)2422
+{2423
+ simde__m64_private r_, a_ = simde__m64_to_private(a);2424
+2425
+ r_.i64[0] = a_.i64[0];2426
+ r_.i16[imm8] = i;2427
+2428
+ return simde__m64_from_private(r_);2429
+}2430
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \2431
+ !defined(__PGI)2432
+#if defined(SIMDE_BUG_CLANG_44589)2433
+#define ssimde_mm_insert_pi16(a, i, imm8) \2434
+ (HEDLEY_DIAGNOSTIC_PUSH _Pragma( \2435
+ "clang diagnostic ignored \"-Wvector-conversion\"")( \2436
+ _mm_insert_pi16((a), (i), (imm8))) HEDLEY_DIAGNOSTIC_POP)2437
+#else2438
+#define simde_mm_insert_pi16(a, i, imm8) _mm_insert_pi16(a, i, imm8)2439
+#endif2440
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2441
+#define simde_mm_insert_pi16(a, i, imm8) \2442
+ simde__m64_from_neon_i16( \2443
+ vset_lane_s16((i), simde__m64_to_neon_i16(a), (imm8)))2444
+#endif2445
+#define simde_m_pinsrw(a, i, imm8) (simde_mm_insert_pi16(a, i, imm8))2446
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2447
+#define _mm_insert_pi16(a, i, imm8) simde_mm_insert_pi16(a, i, imm8)2448
+#define _m_pinsrw(a, i, imm8) simde_mm_insert_pi16(a, i, imm8)2449
+#endif2450
+2451
+SIMDE_FUNCTION_ATTRIBUTES2452
+simde__m1282453
+simde_mm_load_ps(simde_float32 const mem_addr[HEDLEY_ARRAY_PARAM(4)])2454
+{2455
+#if defined(SIMDE_X86_SSE_NATIVE)2456
+ return _mm_load_ps(mem_addr);2457
+#else2458
+ simde__m128_private r_;2459
+2460
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2461
+ r_.neon_f32 = vld1q_f32(mem_addr);2462
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)2463
+ r_.altivec_f32 = vec_vsx_ld(0, mem_addr);2464
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2465
+ r_.altivec_f32 = vec_ld(0, mem_addr);2466
+#else2467
+ simde_memcpy(&r_, SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m128),2468
+ sizeof(r_));2469
+#endif2470
+2471
+ return simde__m128_from_private(r_);2472
+#endif2473
+}2474
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2475
+#define _mm_load_ps(mem_addr) simde_mm_load_ps(mem_addr)2476
+#endif2477
+2478
+SIMDE_FUNCTION_ATTRIBUTES2479
+simde__m128 simde_mm_load1_ps(simde_float32 const *mem_addr)2480
+{2481
+#if defined(SIMDE_X86_SSE_NATIVE)2482
+ return _mm_load_ps1(mem_addr);2483
+#else2484
+ simde__m128_private r_;2485
+2486
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2487
+ r_.neon_f32 = vld1q_dup_f32(mem_addr);2488
+#else2489
+ r_ = simde__m128_to_private(simde_mm_set1_ps(*mem_addr));2490
+#endif2491
+2492
+ return simde__m128_from_private(r_);2493
+#endif2494
+}2495
+#define simde_mm_load_ps1(mem_addr) simde_mm_load1_ps(mem_addr)2496
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2497
+#define _mm_load_ps1(mem_addr) simde_mm_load1_ps(mem_addr)2498
+#define _mm_load1_ps(mem_addr) simde_mm_load1_ps(mem_addr)2499
+#endif2500
+2501
+SIMDE_FUNCTION_ATTRIBUTES2502
+simde__m128 simde_mm_load_ss(simde_float32 const *mem_addr)2503
+{2504
+#if defined(SIMDE_X86_SSE_NATIVE)2505
+ return _mm_load_ss(mem_addr);2506
+#else2507
+ simde__m128_private r_;2508
+2509
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2510
+ r_.neon_f32 = vsetq_lane_f32(*mem_addr, vdupq_n_f32(0), 0);2511
+#else2512
+ r_.f32[0] = *mem_addr;2513
+ r_.i32[1] = 0;2514
+ r_.i32[2] = 0;2515
+ r_.i32[3] = 0;2516
+#endif2517
+2518
+ return simde__m128_from_private(r_);2519
+#endif2520
+}2521
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2522
+#define _mm_load_ss(mem_addr) simde_mm_load_ss(mem_addr)2523
+#endif2524
+2525
+SIMDE_FUNCTION_ATTRIBUTES2526
+simde__m128 simde_mm_loadh_pi(simde__m128 a, simde__m64 const *mem_addr)2527
+{2528
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2529
+ return _mm_loadh_pi(a,2530
+ HEDLEY_REINTERPRET_CAST(__m64 const *, mem_addr));2531
+#else2532
+ simde__m128_private r_, a_ = simde__m128_to_private(a);2533
+2534
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2535
+ r_.neon_f32 = vcombine_f32(2536
+ vget_low_f32(a_.neon_f32),2537
+ vld1_f32(HEDLEY_REINTERPRET_CAST(const float32_t *, mem_addr)));2538
+#else2539
+ simde__m64_private b_ =2540
+ *HEDLEY_REINTERPRET_CAST(simde__m64_private const *, mem_addr);2541
+ r_.f32[0] = a_.f32[0];2542
+ r_.f32[1] = a_.f32[1];2543
+ r_.f32[2] = b_.f32[0];2544
+ r_.f32[3] = b_.f32[1];2545
+#endif2546
+2547
+ return simde__m128_from_private(r_);2548
+#endif2549
+}2550
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2551
+#if HEDLEY_HAS_WARNING("-Wold-style-cast")2552
+#define _mm_loadh_pi(a, mem_addr) \2553
+ simde_mm_loadh_pi((a), HEDLEY_REINTERPRET_CAST(simde__m64 const *, \2554
+ (mem_addr)))2555
+#else2556
+#define _mm_loadh_pi(a, mem_addr) \2557
+ simde_mm_loadh_pi((a), (simde__m64 const *)(mem_addr))2558
+#endif2559
+#endif2560
+2561
+/* The SSE documentation says that there are no alignment requirements2562
+ for mem_addr. Unfortunately they used the __m64 type for the argument2563
+ which is supposed to be 8-byte aligned, so some compilers (like clang2564
+ with -Wcast-align) will generate a warning if you try to cast, say,2565
+ a simde_float32* to a simde__m64* for this function.2566
+2567
+ I think the choice of argument type is unfortunate, but I do think we2568
+ need to stick to it here. If there is demand I can always add something2569
+ like simde_x_mm_loadl_f32(simde__m128, simde_float32 mem_addr[2]) */2570
+SIMDE_FUNCTION_ATTRIBUTES2571
+simde__m128 simde_mm_loadl_pi(simde__m128 a, simde__m64 const *mem_addr)2572
+{2573
+#if defined(SIMDE_X86_SSE_NATIVE)2574
+ return _mm_loadl_pi(a,2575
+ HEDLEY_REINTERPRET_CAST(__m64 const *, mem_addr));2576
+#else2577
+ simde__m128_private r_, a_ = simde__m128_to_private(a);2578
+2579
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2580
+ r_.neon_f32 = vcombine_f32(2581
+ vld1_f32(HEDLEY_REINTERPRET_CAST(const float32_t *, mem_addr)),2582
+ vget_high_f32(a_.neon_f32));2583
+#else2584
+ simde__m64_private b_;2585
+ simde_memcpy(&b_, mem_addr, sizeof(b_));2586
+ r_.i32[0] = b_.i32[0];2587
+ r_.i32[1] = b_.i32[1];2588
+ r_.i32[2] = a_.i32[2];2589
+ r_.i32[3] = a_.i32[3];2590
+#endif2591
+2592
+ return simde__m128_from_private(r_);2593
+#endif2594
+}2595
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2596
+#if HEDLEY_HAS_WARNING("-Wold-style-cast")2597
+#define _mm_loadl_pi(a, mem_addr) \2598
+ simde_mm_loadl_pi((a), HEDLEY_REINTERPRET_CAST(simde__m64 const *, \2599
+ (mem_addr)))2600
+#else2601
+#define _mm_loadl_pi(a, mem_addr) \2602
+ simde_mm_loadl_pi((a), (simde__m64 const *)(mem_addr))2603
+#endif2604
+#endif2605
+2606
+SIMDE_FUNCTION_ATTRIBUTES2607
+simde__m1282608
+simde_mm_loadr_ps(simde_float32 const mem_addr[HEDLEY_ARRAY_PARAM(4)])2609
+{2610
+#if defined(SIMDE_X86_SSE_NATIVE)2611
+ return _mm_loadr_ps(mem_addr);2612
+#else2613
+ simde__m128_private r_,2614
+ v_ = simde__m128_to_private(simde_mm_load_ps(mem_addr));2615
+2616
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2617
+ r_.neon_f32 = vrev64q_f32(v_.neon_f32);2618
+ r_.neon_f32 = vextq_f32(r_.neon_f32, r_.neon_f32, 2);2619
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE) && defined(__PPC64__)2620
+ r_.altivec_f32 = vec_reve(v_.altivec_f32);2621
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2622
+ r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, v_.f32, v_.f32, 3, 2, 1, 0);2623
+#else2624
+ r_.f32[0] = v_.f32[3];2625
+ r_.f32[1] = v_.f32[2];2626
+ r_.f32[2] = v_.f32[1];2627
+ r_.f32[3] = v_.f32[0];2628
+#endif2629
+2630
+ return simde__m128_from_private(r_);2631
+#endif2632
+}2633
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2634
+#define _mm_loadr_ps(mem_addr) simde_mm_loadr_ps(mem_addr)2635
+#endif2636
+2637
+SIMDE_FUNCTION_ATTRIBUTES2638
+simde__m1282639
+simde_mm_loadu_ps(simde_float32 const mem_addr[HEDLEY_ARRAY_PARAM(4)])2640
+{2641
+#if defined(SIMDE_X86_SSE_NATIVE)2642
+ return _mm_loadu_ps(mem_addr);2643
+#else2644
+ simde__m128_private r_;2645
+2646
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2647
+ r_.neon_f32 =2648
+ vld1q_f32(HEDLEY_REINTERPRET_CAST(const float32_t *, mem_addr));2649
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2650
+ r_.wasm_v128 = wasm_v128_load(mem_addr);2651
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE) && defined(__PPC64__)2652
+ r_.altivec_f32 = vec_vsx_ld(0, mem_addr);2653
+#else2654
+ simde_memcpy(&r_, mem_addr, sizeof(r_));2655
+#endif2656
+2657
+ return simde__m128_from_private(r_);2658
+#endif2659
+}2660
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2661
+#define _mm_loadu_ps(mem_addr) simde_mm_loadu_ps(mem_addr)2662
+#endif2663
+2664
+SIMDE_FUNCTION_ATTRIBUTES2665
+void simde_mm_maskmove_si64(simde__m64 a, simde__m64 mask, int8_t *mem_addr)2666
+{2667
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2668
+ _mm_maskmove_si64(a, mask, HEDLEY_REINTERPRET_CAST(char *, mem_addr));2669
+#else2670
+ simde__m64_private a_ = simde__m64_to_private(a),2671
+ mask_ = simde__m64_to_private(mask);2672
+2673
+ SIMDE_VECTORIZE2674
+ for (size_t i = 0; i < (sizeof(a_.i8) / sizeof(a_.i8[0])); i++)2675
+ if (mask_.i8[i] < 0)2676
+ mem_addr[i] = a_.i8[i];2677
+#endif2678
+}2679
+#define simde_m_maskmovq(a, mask, mem_addr) \2680
+ simde_mm_maskmove_si64(a, mask, mem_addr)2681
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2682
+#define _mm_maskmove_si64(a, mask, mem_addr) \2683
+ simde_mm_maskmove_si64( \2684
+ (a), (mask), \2685
+ SIMDE_CHECKED_REINTERPRET_CAST(int8_t *, char *, (mem_addr)))2686
+#define _m_maskmovq(a, mask, mem_addr) \2687
+ simde_mm_maskmove_si64( \2688
+ (a), (mask), \2689
+ SIMDE_CHECKED_REINTERPRET_CAST(int8_t *, char *, (mem_addr)))2690
+#endif2691
+2692
+SIMDE_FUNCTION_ATTRIBUTES2693
+simde__m64 simde_mm_max_pi16(simde__m64 a, simde__m64 b)2694
+{2695
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2696
+ return _mm_max_pi16(a, b);2697
+#else2698
+ simde__m64_private r_, a_ = simde__m64_to_private(a),2699
+ b_ = simde__m64_to_private(b);2700
+2701
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2702
+ r_.neon_i16 = vmax_s16(a_.neon_i16, b_.neon_i16);2703
+#else2704
+ SIMDE_VECTORIZE2705
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {2706
+ r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? a_.i16[i] : b_.i16[i];2707
+ }2708
+#endif2709
+2710
+ return simde__m64_from_private(r_);2711
+#endif2712
+}2713
+#define simde_m_pmaxsw(a, b) simde_mm_max_pi16(a, b)2714
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2715
+#define _mm_max_pi16(a, b) simde_mm_max_pi16(a, b)2716
+#define _m_pmaxsw(a, b) simde_mm_max_pi16(a, b)2717
+#endif2718
+2719
+SIMDE_FUNCTION_ATTRIBUTES2720
+simde__m128 simde_mm_max_ps(simde__m128 a, simde__m128 b)2721
+{2722
+#if defined(SIMDE_X86_SSE_NATIVE)2723
+ return _mm_max_ps(a, b);2724
+#else2725
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2726
+ b_ = simde__m128_to_private(b);2727
+2728
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE) && defined(SIMDE_FAST_NANS)2729
+ r_.neon_f32 = vmaxq_f32(a_.neon_f32, b_.neon_f32);2730
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2731
+ r_.neon_f32 = vbslq_f32(vcgtq_f32(a_.neon_f32, b_.neon_f32),2732
+ a_.neon_f32, b_.neon_f32);2733
+#elif defined(SIMDE_WASM_SIMD128_NATIVE) && defined(SIMDE_FAST_NANS)2734
+ r_.wasm_v128 = wasm_f32x4_max(a_.wasm_v128, b_.wasm_v128);2735
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2736
+ r_.wasm_v128 =2737
+ wasm_v128_bitselect(a_.wasm_v128, b_.wasm_v128,2738
+ wasm_f32x4_gt(a_.wasm_v128, b_.wasm_v128));2739
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE) && defined(SIMDE_FAST_NANS)2740
+ r_.altivec_f32 = vec_max(a_.altivec_f32, b_.altivec_f32);2741
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2742
+ r_.altivec_f32 = vec_sel(b_.altivec_f32, a_.altivec_f32,2743
+ vec_cmpgt(a_.altivec_f32, b_.altivec_f32));2744
+#else2745
+ SIMDE_VECTORIZE2746
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2747
+ r_.f32[i] = (a_.f32[i] > b_.f32[i]) ? a_.f32[i] : b_.f32[i];2748
+ }2749
+#endif2750
+2751
+ return simde__m128_from_private(r_);2752
+#endif2753
+}2754
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2755
+#define _mm_max_ps(a, b) simde_mm_max_ps((a), (b))2756
+#endif2757
+2758
+SIMDE_FUNCTION_ATTRIBUTES2759
+simde__m64 simde_mm_max_pu8(simde__m64 a, simde__m64 b)2760
+{2761
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2762
+ return _mm_max_pu8(a, b);2763
+#else2764
+ simde__m64_private r_, a_ = simde__m64_to_private(a),2765
+ b_ = simde__m64_to_private(b);2766
+2767
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2768
+ r_.neon_u8 = vmax_u8(a_.neon_u8, b_.neon_u8);2769
+#else2770
+ SIMDE_VECTORIZE2771
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {2772
+ r_.u8[i] = (a_.u8[i] > b_.u8[i]) ? a_.u8[i] : b_.u8[i];2773
+ }2774
+#endif2775
+2776
+ return simde__m64_from_private(r_);2777
+#endif2778
+}2779
+#define simde_m_pmaxub(a, b) simde_mm_max_pu8(a, b)2780
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2781
+#define _mm_max_pu8(a, b) simde_mm_max_pu8(a, b)2782
+#define _m_pmaxub(a, b) simde_mm_max_pu8(a, b)2783
+#endif2784
+2785
+SIMDE_FUNCTION_ATTRIBUTES2786
+simde__m128 simde_mm_max_ss(simde__m128 a, simde__m128 b)2787
+{2788
+#if defined(SIMDE_X86_SSE_NATIVE)2789
+ return _mm_max_ss(a, b);2790
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2791
+ return simde_mm_move_ss(a, simde_mm_max_ps(a, b));2792
+#else2793
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2794
+ b_ = simde__m128_to_private(b);2795
+2796
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2797
+ float32_t value = vgetq_lane_f32(maxq_f32(a_.neon_f32, b_.neon_f32), 0);2798
+ r_.neon_f32 = vsetq_lane_f32(value, a_.neon_f32, 0);2799
+#else2800
+ r_.f32[0] = (a_.f32[0] > b_.f32[0]) ? a_.f32[0] : b_.f32[0];2801
+ r_.f32[1] = a_.f32[1];2802
+ r_.f32[2] = a_.f32[2];2803
+ r_.f32[3] = a_.f32[3];2804
+#endif2805
+2806
+ return simde__m128_from_private(r_);2807
+#endif2808
+}2809
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2810
+#define _mm_max_ss(a, b) simde_mm_max_ss((a), (b))2811
+#endif2812
+2813
+SIMDE_FUNCTION_ATTRIBUTES2814
+simde__m64 simde_mm_min_pi16(simde__m64 a, simde__m64 b)2815
+{2816
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2817
+ return _mm_min_pi16(a, b);2818
+#else2819
+ simde__m64_private r_, a_ = simde__m64_to_private(a),2820
+ b_ = simde__m64_to_private(b);2821
+2822
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2823
+ r_.neon_i16 = vmin_s16(a_.neon_i16, b_.neon_i16);2824
+#else2825
+ SIMDE_VECTORIZE2826
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {2827
+ r_.i16[i] = (a_.i16[i] < b_.i16[i]) ? a_.i16[i] : b_.i16[i];2828
+ }2829
+#endif2830
+2831
+ return simde__m64_from_private(r_);2832
+#endif2833
+}2834
+#define simde_m_pminsw(a, b) simde_mm_min_pi16(a, b)2835
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2836
+#define _mm_min_pi16(a, b) simde_mm_min_pi16(a, b)2837
+#define _m_pminsw(a, b) simde_mm_min_pi16(a, b)2838
+#endif2839
+2840
+SIMDE_FUNCTION_ATTRIBUTES2841
+simde__m128 simde_mm_min_ps(simde__m128 a, simde__m128 b)2842
+{2843
+#if defined(SIMDE_X86_SSE_NATIVE)2844
+ return _mm_min_ps(a, b);2845
+#elif defined(SIMDE_FAST_NANS) && defined(SIMDE_ARM_NEON_A32V7_NATIVE)2846
+ return simde__m128_from_neon_f32(vminq_f32(simde__m128_to_neon_f32(a),2847
+ simde__m128_to_neon_f32(b)));2848
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2849
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2850
+ b_ = simde__m128_to_private(b);2851
+#if defined(SIMDE_FAST_NANS)2852
+ r_.wasm_v128 = wasm_f32x4_min(a_.wasm_v128, b_.wasm_v128);2853
+#else2854
+ r_.wasm_v128 =2855
+ wasm_v128_bitselect(a_.wasm_v128, b_.wasm_v128,2856
+ wasm_f32x4_lt(a_.wasm_v128, b_.wasm_v128));2857
+#endif2858
+ return simde__m128_from_private(r_);2859
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2860
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2861
+ b_ = simde__m128_to_private(b);2862
+2863
+#if defined(SIMDE_FAST_NANS)2864
+ r_.altivec_f32 = vec_min(a_.altivec_f32, b_.altivec_f32);2865
+#else2866
+ r_.altivec_f32 = vec_sel(b_.altivec_f32, a_.altivec_f32,2867
+ vec_cmpgt(b_.altivec_f32, a_.altivec_f32));2868
+#endif2869
+2870
+ return simde__m128_from_private(r_);2871
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2872
+ simde__m128 mask = simde_mm_cmplt_ps(a, b);2873
+ return simde_mm_or_ps(simde_mm_and_ps(mask, a),2874
+ simde_mm_andnot_ps(mask, b));2875
+#else2876
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2877
+ b_ = simde__m128_to_private(b);2878
+2879
+ SIMDE_VECTORIZE2880
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2881
+ r_.f32[i] = (a_.f32[i] < b_.f32[i]) ? a_.f32[i] : b_.f32[i];2882
+ }2883
+2884
+ return simde__m128_from_private(r_);2885
+#endif2886
+}2887
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2888
+#define _mm_min_ps(a, b) simde_mm_min_ps((a), (b))2889
+#endif2890
+2891
+SIMDE_FUNCTION_ATTRIBUTES2892
+simde__m64 simde_mm_min_pu8(simde__m64 a, simde__m64 b)2893
+{2894
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2895
+ return _mm_min_pu8(a, b);2896
+#else2897
+ simde__m64_private r_, a_ = simde__m64_to_private(a),2898
+ b_ = simde__m64_to_private(b);2899
+2900
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2901
+ r_.neon_u8 = vmin_u8(a_.neon_u8, b_.neon_u8);2902
+#else2903
+ SIMDE_VECTORIZE2904
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {2905
+ r_.u8[i] = (a_.u8[i] < b_.u8[i]) ? a_.u8[i] : b_.u8[i];2906
+ }2907
+#endif2908
+2909
+ return simde__m64_from_private(r_);2910
+#endif2911
+}2912
+#define simde_m_pminub(a, b) simde_mm_min_pu8(a, b)2913
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2914
+#define _mm_min_pu8(a, b) simde_mm_min_pu8(a, b)2915
+#define _m_pminub(a, b) simde_mm_min_pu8(a, b)2916
+#endif2917
+2918
+SIMDE_FUNCTION_ATTRIBUTES2919
+simde__m128 simde_mm_min_ss(simde__m128 a, simde__m128 b)2920
+{2921
+#if defined(SIMDE_X86_SSE_NATIVE)2922
+ return _mm_min_ss(a, b);2923
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2924
+ return simde_mm_move_ss(a, simde_mm_min_ps(a, b));2925
+#else2926
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2927
+ b_ = simde__m128_to_private(b);2928
+2929
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2930
+ float32_t value =2931
+ vgetq_lane_f32(vminq_f32(a_.neon_f32, b_.neon_f32), 0);2932
+ r_.neon_f32 = vsetq_lane_f32(value, a_.neon_f32, 0);2933
+#else2934
+ r_.f32[0] = (a_.f32[0] < b_.f32[0]) ? a_.f32[0] : b_.f32[0];2935
+ r_.f32[1] = a_.f32[1];2936
+ r_.f32[2] = a_.f32[2];2937
+ r_.f32[3] = a_.f32[3];2938
+#endif2939
+2940
+ return simde__m128_from_private(r_);2941
+#endif2942
+}2943
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2944
+#define _mm_min_ss(a, b) simde_mm_min_ss((a), (b))2945
+#endif2946
+2947
+SIMDE_FUNCTION_ATTRIBUTES2948
+simde__m128 simde_mm_movehl_ps(simde__m128 a, simde__m128 b)2949
+{2950
+#if defined(SIMDE_X86_SSE_NATIVE)2951
+ return _mm_movehl_ps(a, b);2952
+#else2953
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2954
+ b_ = simde__m128_to_private(b);2955
+2956
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2957
+ float32x2_t a32 = vget_high_f32(a_.neon_f32);2958
+ float32x2_t b32 = vget_high_f32(b_.neon_f32);2959
+ r_.neon_f32 = vcombine_f32(b32, a32);2960
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)2961
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(2962
+ SIMDE_POWER_ALTIVEC_VECTOR(float),2963
+ vec_mergel(b_.altivec_i64, a_.altivec_i64));2964
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2965
+ r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 6, 7, 2, 3);2966
+#else2967
+ r_.f32[0] = b_.f32[2];2968
+ r_.f32[1] = b_.f32[3];2969
+ r_.f32[2] = a_.f32[2];2970
+ r_.f32[3] = a_.f32[3];2971
+#endif2972
+2973
+ return simde__m128_from_private(r_);2974
+#endif2975
+}2976
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)2977
+#define _mm_movehl_ps(a, b) simde_mm_movehl_ps((a), (b))2978
+#endif2979
+2980
+SIMDE_FUNCTION_ATTRIBUTES2981
+simde__m128 simde_mm_movelh_ps(simde__m128 a, simde__m128 b)2982
+{2983
+#if defined(SIMDE_X86_SSE_NATIVE)2984
+ return _mm_movelh_ps(a, b);2985
+#else2986
+ simde__m128_private r_, a_ = simde__m128_to_private(a),2987
+ b_ = simde__m128_to_private(b);2988
+2989
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2990
+ float32x2_t a10 = vget_low_f32(a_.neon_f32);2991
+ float32x2_t b10 = vget_low_f32(b_.neon_f32);2992
+ r_.neon_f32 = vcombine_f32(a10, b10);2993
+#elif defined(SIMDE_SHUFFLE_VECTOR_)2994
+ r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 0, 1, 4, 5);2995
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)2996
+ r_.altivec_f32 = HEDLEY_REINTERPRET_CAST(2997
+ SIMDE_POWER_ALTIVEC_VECTOR(float),2998
+ vec_mergeh(a_.altivec_i64, b_.altivec_i64));2999
+#else3000
+ r_.f32[0] = a_.f32[0];3001
+ r_.f32[1] = a_.f32[1];3002
+ r_.f32[2] = b_.f32[0];3003
+ r_.f32[3] = b_.f32[1];3004
+#endif3005
+3006
+ return simde__m128_from_private(r_);3007
+#endif3008
+}3009
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3010
+#define _mm_movelh_ps(a, b) simde_mm_movelh_ps((a), (b))3011
+#endif3012
+3013
+SIMDE_FUNCTION_ATTRIBUTES3014
+int simde_mm_movemask_pi8(simde__m64 a)3015
+{3016
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3017
+ return _mm_movemask_pi8(a);3018
+#else3019
+ simde__m64_private a_ = simde__m64_to_private(a);3020
+ int r = 0;3021
+3022
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3023
+ uint8x8_t input = a_.neon_u8;3024
+ const int8_t xr[8] = {-7, -6, -5, -4, -3, -2, -1, 0};3025
+ const uint8x8_t mask_and = vdup_n_u8(0x80);3026
+ const int8x8_t mask_shift = vld1_s8(xr);3027
+ const uint8x8_t mask_result =3028
+ vshl_u8(vand_u8(input, mask_and), mask_shift);3029
+ uint8x8_t lo = mask_result;3030
+ r = vaddv_u8(lo);3031
+#else3032
+ const size_t nmemb = sizeof(a_.i8) / sizeof(a_.i8[0]);3033
+ SIMDE_VECTORIZE_REDUCTION(| : r)3034
+ for (size_t i = 0; i < nmemb; i++) {3035
+ r |= (a_.u8[nmemb - 1 - i] >> 7) << (nmemb - 1 - i);3036
+ }3037
+#endif3038
+3039
+ return r;3040
+#endif3041
+}3042
+#define simde_m_pmovmskb(a) simde_mm_movemask_pi8(a)3043
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3044
+#define _mm_movemask_pi8(a) simde_mm_movemask_pi8(a)3045
+#define _m_pmovmskb(a) simde_mm_movemask_pi8(a)3046
+#endif3047
+3048
+SIMDE_FUNCTION_ATTRIBUTES3049
+int simde_mm_movemask_ps(simde__m128 a)3050
+{3051
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3052
+ return _mm_movemask_ps(a);3053
+#else3054
+ int r = 0;3055
+ simde__m128_private a_ = simde__m128_to_private(a);3056
+3057
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3058
+ static const int32_t shift_amount[] = {0, 1, 2, 3};3059
+ const int32x4_t shift = vld1q_s32(shift_amount);3060
+ uint32x4_t tmp = vshrq_n_u32(a_.neon_u32, 31);3061
+ return HEDLEY_STATIC_CAST(int, vaddvq_u32(vshlq_u32(tmp, shift)));3062
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3063
+ // Shift out everything but the sign bits with a 32-bit unsigned shift right.3064
+ uint64x2_t high_bits =3065
+ vreinterpretq_u64_u32(vshrq_n_u32(a_.neon_u32, 31));3066
+ // Merge the two pairs together with a 64-bit unsigned shift right + add.3067
+ uint8x16_t paired =3068
+ vreinterpretq_u8_u64(vsraq_n_u64(high_bits, high_bits, 31));3069
+ // Extract the result.3070
+ return vgetq_lane_u8(paired, 0) | (vgetq_lane_u8(paired, 8) << 2);3071
+#else3072
+ SIMDE_VECTORIZE_REDUCTION(| : r)3073
+ for (size_t i = 0; i < sizeof(a_.u32) / sizeof(a_.u32[0]); i++) {3074
+ r |= (a_.u32[i] >> ((sizeof(a_.u32[i]) * CHAR_BIT) - 1)) << i;3075
+ }3076
+#endif3077
+3078
+ return r;3079
+#endif3080
+}3081
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3082
+#define _mm_movemask_ps(a) simde_mm_movemask_ps((a))3083
+#endif3084
+3085
+SIMDE_FUNCTION_ATTRIBUTES3086
+simde__m128 simde_mm_mul_ps(simde__m128 a, simde__m128 b)3087
+{3088
+#if defined(SIMDE_X86_SSE_NATIVE)3089
+ return _mm_mul_ps(a, b);3090
+#else3091
+ simde__m128_private r_, a_ = simde__m128_to_private(a),3092
+ b_ = simde__m128_to_private(b);3093
+3094
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3095
+ r_.neon_f32 = vmulq_f32(a_.neon_f32, b_.neon_f32);3096
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3097
+ r_.wasm_v128 = wasm_f32x4_mul(a_.wasm_v128, b_.wasm_v128);3098
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3099
+ r_.f32 = a_.f32 * b_.f32;3100
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)3101
+ r_.altivec_f32 = vec_mul(a_.altivec_f32, b_.altivec_f32);3102
+#else3103
+ SIMDE_VECTORIZE3104
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {3105
+ r_.f32[i] = a_.f32[i] * b_.f32[i];3106
+ }3107
+#endif3108
+3109
+ return simde__m128_from_private(r_);3110
+#endif3111
+}3112
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3113
+#define _mm_mul_ps(a, b) simde_mm_mul_ps((a), (b))3114
+#endif3115
+3116
+SIMDE_FUNCTION_ATTRIBUTES3117
+simde__m128 simde_mm_mul_ss(simde__m128 a, simde__m128 b)3118
+{3119
+#if defined(SIMDE_X86_SSE_NATIVE)3120
+ return _mm_mul_ss(a, b);3121
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)3122
+ return simde_mm_move_ss(a, simde_mm_mul_ps(a, b));3123
+#else3124
+ simde__m128_private r_, a_ = simde__m128_to_private(a),3125
+ b_ = simde__m128_to_private(b);3126
+3127
+ r_.f32[0] = a_.f32[0] * b_.f32[0];3128
+ r_.f32[1] = a_.f32[1];3129
+ r_.f32[2] = a_.f32[2];3130
+ r_.f32[3] = a_.f32[3];3131
+3132
+ return simde__m128_from_private(r_);3133
+#endif3134
+}3135
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3136
+#define _mm_mul_ss(a, b) simde_mm_mul_ss((a), (b))3137
+#endif3138
+3139
+SIMDE_FUNCTION_ATTRIBUTES3140
+simde__m64 simde_mm_mulhi_pu16(simde__m64 a, simde__m64 b)3141
+{3142
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3143
+ return _mm_mulhi_pu16(a, b);3144
+#else3145
+ simde__m64_private r_, a_ = simde__m64_to_private(a),3146
+ b_ = simde__m64_to_private(b);3147
+3148
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3149
+ const uint32x4_t t1 = vmull_u16(a_.neon_u16, b_.neon_u16);3150
+ const uint32x4_t t2 = vshrq_n_u32(t1, 16);3151
+ const uint16x4_t t3 = vmovn_u32(t2);3152
+ r_.neon_u16 = t3;3153
+#else3154
+ SIMDE_VECTORIZE3155
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {3156
+ r_.u16[i] = HEDLEY_STATIC_CAST(3157
+ uint16_t, ((HEDLEY_STATIC_CAST(uint32_t, a_.u16[i]) *3158
+ HEDLEY_STATIC_CAST(uint32_t, b_.u16[i])) >>3159
+ UINT32_C(16)));3160
+ }3161
+#endif3162
+3163
+ return simde__m64_from_private(r_);3164
+#endif3165
+}3166
+#define simde_m_pmulhuw(a, b) simde_mm_mulhi_pu16(a, b)3167
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3168
+#define _mm_mulhi_pu16(a, b) simde_mm_mulhi_pu16(a, b)3169
+#define _m_pmulhuw(a, b) simde_mm_mulhi_pu16(a, b)3170
+#endif3171
+3172
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(HEDLEY_GCC_VERSION)3173
+#define SIMDE_MM_HINT_NTA HEDLEY_STATIC_CAST(enum _mm_hint, 0)3174
+#define SIMDE_MM_HINT_T0 HEDLEY_STATIC_CAST(enum _mm_hint, 1)3175
+#define SIMDE_MM_HINT_T1 HEDLEY_STATIC_CAST(enum _mm_hint, 2)3176
+#define SIMDE_MM_HINT_T2 HEDLEY_STATIC_CAST(enum _mm_hint, 3)3177
+#define SIMDE_MM_HINT_ENTA HEDLEY_STATIC_CAST(enum _mm_hint, 4)3178
+#define SIMDE_MM_HINT_ET0 HEDLEY_STATIC_CAST(enum _mm_hint, 5)3179
+#define SIMDE_MM_HINT_ET1 HEDLEY_STATIC_CAST(enum _mm_hint, 6)3180
+#define SIMDE_MM_HINT_ET2 HEDLEY_STATIC_CAST(enum _mm_hint, 7)3181
+#else3182
+#define SIMDE_MM_HINT_NTA 03183
+#define SIMDE_MM_HINT_T0 13184
+#define SIMDE_MM_HINT_T1 23185
+#define SIMDE_MM_HINT_T2 33186
+#define SIMDE_MM_HINT_ENTA 43187
+#define SIMDE_MM_HINT_ET0 53188
+#define SIMDE_MM_HINT_ET1 63189
+#define SIMDE_MM_HINT_ET2 73190
+#endif3191
+3192
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3193
+HEDLEY_DIAGNOSTIC_PUSH3194
+#if HEDLEY_HAS_WARNING("-Wreserved-id-macro")3195
+_Pragma("clang diagnostic ignored \"-Wreserved-id-macro\"")3196
+#endif3197
+#undef _MM_HINT_NTA3198
+#define _MM_HINT_NTA SIMDE_MM_HINT_NTA3199
+#undef _MM_HINT_T03200
+#define _MM_HINT_T0 SIMDE_MM_HINT_T03201
+#undef _MM_HINT_T13202
+#define _MM_HINT_T1 SIMDE_MM_HINT_T13203
+#undef _MM_HINT_T23204
+#define _MM_HINT_T2 SIMDE_MM_HINT_T23205
+#undef _MM_HINT_ETNA3206
+#define _MM_HINT_ETNA SIMDE_MM_HINT_ETNA3207
+#undef _MM_HINT_ET03208
+#define _MM_HINT_ET0 SIMDE_MM_HINT_ET03209
+#undef _MM_HINT_ET13210
+#define _MM_HINT_ET1 SIMDE_MM_HINT_ET13211
+#undef _MM_HINT_ET13212
+#define _MM_HINT_ET2 SIMDE_MM_HINT_ET23213
+ HEDLEY_DIAGNOSTIC_POP3214
+#endif3215
+3216
+ SIMDE_FUNCTION_ATTRIBUTES void simde_mm_prefetch(char const *p, int i)3217
+{3218
+#if defined(HEDLEY_GCC_VERSION)3219
+ __builtin_prefetch(p);3220
+#else3221
+ (void)p;3222
+#endif3223
+3224
+ (void)i;3225
+}3226
+#if defined(SIMDE_X86_SSE_NATIVE)3227
+#if defined(__clang__) && \3228
+ !SIMDE_DETECT_CLANG_VERSION_CHECK( \3229
+ 10, 0, 0) /* https://reviews.llvm.org/D71718 */3230
+#define simde_mm_prefetch(p, i) \3231
+ (__extension__({ \3232
+ HEDLEY_DIAGNOSTIC_PUSH \3233
+ HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL \3234
+ _mm_prefetch((p), (i)); \3235
+ HEDLEY_DIAGNOSTIC_POP \3236
+ }))3237
+#else3238
+#define simde_mm_prefetch(p, i) _mm_prefetch(p, i)3239
+#endif3240
+#endif3241
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3242
+#define _mm_prefetch(p, i) simde_mm_prefetch(p, i)3243
+#endif3244
+3245
+SIMDE_FUNCTION_ATTRIBUTES3246
+simde__m128 simde_x_mm_negate_ps(simde__m128 a)3247
+{3248
+#if defined(SIMDE_X86_SSE_NATIVE)3249
+ return simde_mm_xor_ps(a, _mm_set1_ps(SIMDE_FLOAT32_C(-0.0)));3250
+#else3251
+ simde__m128_private r_, a_ = simde__m128_to_private(a);3252
+3253
+#if defined(SIMDE_POWER_ALTIVEC_P8_NATIVE) && \3254
+ (!defined(HEDLEY_GCC_VERSION) || HEDLEY_GCC_VERSION_CHECK(8, 1, 0))3255
+ r_.altivec_f32 = vec_neg(a_.altivec_f32);3256
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3257
+ r_.neon_f32 = vnegq_f32(a_.neon_f32);3258
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3259
+ r_.wasm_v128 = wasm_f32x4_neg(a_.wasm_v128);3260
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)3261
+ r_.altivec_f32 = vec_neg(a_.altivec_f32);3262
+#elif defined(SIMDE_VECTOR_NEGATE)3263
+ r_.f32 = -a_.f32;3264
+#else3265
+ SIMDE_VECTORIZE3266
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {3267
+ r_.f32[i] = -a_.f32[i];3268
+ }3269
+#endif3270
+3271
+ return simde__m128_from_private(r_);3272
+#endif3273
+}3274
+3275
+SIMDE_FUNCTION_ATTRIBUTES3276
+simde__m128 simde_mm_rcp_ps(simde__m128 a)3277
+{3278
+#if defined(SIMDE_X86_SSE_NATIVE)3279
+ return _mm_rcp_ps(a);3280
+#else3281
+ simde__m128_private r_, a_ = simde__m128_to_private(a);3282
+3283
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3284
+ float32x4_t recip = vrecpeq_f32(a_.neon_f32);3285
+3286
+#if SIMDE_ACCURACY_PREFERENCE > 03287
+ for (int i = 0; i < SIMDE_ACCURACY_PREFERENCE; ++i) {3288
+ recip = vmulq_f32(recip, vrecpsq_f32(recip, a_.neon_f32));3289
+ }3290
+#endif3291
+3292
+ r_.neon_f32 = recip;3293
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3294
+ r_.wasm_v128 = wasm_f32x4_div(simde_mm_set1_ps(1.0f), a_.wasm_v128);3295
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3296
+ r_.altivec_f32 = vec_re(a_.altivec_f32);3297
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)3298
+ r_.f32 = 1.0f / a_.f32;3299
+#elif defined(SIMDE_IEEE754_STORAGE)3300
+ /* https://stackoverflow.com/questions/12227126/division-as-multiply-and-lut-fast-float-division-reciprocal/12228234#12228234 */3301
+ SIMDE_VECTORIZE3302
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {3303
+ int32_t ix;3304
+ simde_float32 fx = a_.f32[i];3305
+ simde_memcpy(&ix, &fx, sizeof(ix));3306
+ int32_t x = INT32_C(0x7EF311C3) - ix;3307
+ simde_float32 temp;3308
+ simde_memcpy(&temp, &x, sizeof(temp));3309
+ r_.f32[i] = temp * (SIMDE_FLOAT32_C(2.0) - temp * fx);3310
+ }3311
+#else3312
+ SIMDE_VECTORIZE3313
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {3314
+ r_.f32[i] = 1.0f / a_.f32[i];3315
+ }3316
+#endif3317
+3318
+ return simde__m128_from_private(r_);3319
+#endif3320
+}3321
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3322
+#define _mm_rcp_ps(a) simde_mm_rcp_ps((a))3323
+#endif3324
+3325
+SIMDE_FUNCTION_ATTRIBUTES3326
+simde__m128 simde_mm_rcp_ss(simde__m128 a)3327
+{3328
+#if defined(SIMDE_X86_SSE_NATIVE)3329
+ return _mm_rcp_ss(a);3330
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)3331
+ return simde_mm_move_ss(a, simde_mm_rcp_ps(a));3332
+#else3333
+ simde__m128_private r_, a_ = simde__m128_to_private(a);3334
+3335
+ r_.f32[0] = 1.0f / a_.f32[0];3336
+ r_.f32[1] = a_.f32[1];3337
+ r_.f32[2] = a_.f32[2];3338
+ r_.f32[3] = a_.f32[3];3339
+3340
+ return simde__m128_from_private(r_);3341
+#endif3342
+}3343
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3344
+#define _mm_rcp_ss(a) simde_mm_rcp_ss((a))3345
+#endif3346
+3347
+SIMDE_FUNCTION_ATTRIBUTES3348
+simde__m128 simde_mm_rsqrt_ps(simde__m128 a)3349
+{3350
+#if defined(SIMDE_X86_SSE_NATIVE)3351
+ return _mm_rsqrt_ps(a);3352
+#else3353
+ simde__m128_private r_, a_ = simde__m128_to_private(a);3354
+3355
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3356
+ r_.neon_f32 = vrsqrteq_f32(a_.neon_f32);3357
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3358
+ r_.altivec_f32 = vec_rsqrte(a_.altivec_f32);3359
+#elif defined(SIMDE_IEEE754_STORAGE)3360
+ /* https://basesandframes.files.wordpress.com/2020/04/even_faster_math_functions_green_2020.pdf3361
+ Pages 100 - 103 */3362
+ SIMDE_VECTORIZE3363
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {3364
+#if SIMDE_ACCURACY_PREFERENCE <= 03365
+ r_.i32[i] = INT32_C(0x5F37624F) - (a_.i32[i] >> 1);3366
+#else3367
+ simde_float32 x = a_.f32[i];3368
+ simde_float32 xhalf = SIMDE_FLOAT32_C(0.5) * x;3369
+ int32_t ix;3370
+3371
+ simde_memcpy(&ix, &x, sizeof(ix));3372
+3373
+#if SIMDE_ACCURACY_PREFERENCE == 13374
+ ix = INT32_C(0x5F375A82) - (ix >> 1);3375
+#else3376
+ ix = INT32_C(0x5F37599E) - (ix >> 1);3377
+#endif3378
+3379
+ simde_memcpy(&x, &ix, sizeof(x));3380
+3381
+#if SIMDE_ACCURACY_PREFERENCE >= 23382
+ x = x * (SIMDE_FLOAT32_C(1.5008909) - xhalf * x * x);3383
+#endif3384
+ x = x * (SIMDE_FLOAT32_C(1.5008909) - xhalf * x * x);3385
+3386
+ r_.f32[i] = x;3387
+#endif3388
+ }3389
+#elif defined(simde_math_sqrtf)3390
+ SIMDE_VECTORIZE3391
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {3392
+ r_.f32[i] = 1.0f / simde_math_sqrtf(a_.f32[i]);3393
+ }3394
+#else3395
+ HEDLEY_UNREACHABLE();3396
+#endif3397
+3398
+ return simde__m128_from_private(r_);3399
+#endif3400
+}3401
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3402
+#define _mm_rsqrt_ps(a) simde_mm_rsqrt_ps((a))3403
+#endif3404
+3405
+SIMDE_FUNCTION_ATTRIBUTES3406
+simde__m128 simde_mm_rsqrt_ss(simde__m128 a)3407
+{3408
+#if defined(SIMDE_X86_SSE_NATIVE)3409
+ return _mm_rsqrt_ss(a);3410
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)3411
+ return simde_mm_move_ss(a, simde_mm_rsqrt_ps(a));3412
+#else3413
+ simde__m128_private r_, a_ = simde__m128_to_private(a);3414
+3415
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3416
+ r_.neon_f32 =3417
+ vsetq_lane_f32(vgetq_lane_f32(simde_mm_rsqrt_ps(a).neon_f32, 0),3418
+ a_.neon_f32, 0);3419
+#elif defined(SIMDE_IEEE754_STORAGE)3420
+ {3421
+#if SIMDE_ACCURACY_PREFERENCE <= 03422
+ r_.i32[0] = INT32_C(0x5F37624F) - (a_.i32[0] >> 1);3423
+#else3424
+ simde_float32 x = a_.f32[0];3425
+ simde_float32 xhalf = SIMDE_FLOAT32_C(0.5) * x;3426
+ int32_t ix;3427
+3428
+ simde_memcpy(&ix, &x, sizeof(ix));3429
+3430
+#if SIMDE_ACCURACY_PREFERENCE == 13431
+ ix = INT32_C(0x5F375A82) - (ix >> 1);3432
+#else3433
+ ix = INT32_C(0x5F37599E) - (ix >> 1);3434
+#endif3435
+3436
+ simde_memcpy(&x, &ix, sizeof(x));3437
+3438
+#if SIMDE_ACCURACY_PREFERENCE >= 23439
+ x = x * (SIMDE_FLOAT32_C(1.5008909) - xhalf * x * x);3440
+#endif3441
+ x = x * (SIMDE_FLOAT32_C(1.5008909) - xhalf * x * x);3442
+3443
+ r_.f32[0] = x;3444
+#endif3445
+ }3446
+ r_.f32[1] = a_.f32[1];3447
+ r_.f32[2] = a_.f32[2];3448
+ r_.f32[3] = a_.f32[3];3449
+#elif defined(simde_math_sqrtf)3450
+ r_.f32[0] = 1.0f / simde_math_sqrtf(a_.f32[0]);3451
+ r_.f32[1] = a_.f32[1];3452
+ r_.f32[2] = a_.f32[2];3453
+ r_.f32[3] = a_.f32[3];3454
+#else3455
+ HEDLEY_UNREACHABLE();3456
+#endif3457
+3458
+ return simde__m128_from_private(r_);3459
+#endif3460
+}3461
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3462
+#define _mm_rsqrt_ss(a) simde_mm_rsqrt_ss((a))3463
+#endif3464
+3465
+SIMDE_FUNCTION_ATTRIBUTES3466
+simde__m64 simde_mm_sad_pu8(simde__m64 a, simde__m64 b)3467
+{3468
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3469
+ return _mm_sad_pu8(a, b);3470
+#else3471
+ simde__m64_private r_, a_ = simde__m64_to_private(a),3472
+ b_ = simde__m64_to_private(b);3473
+3474
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3475
+ uint16x4_t t = vpaddl_u8(vabd_u8(a_.neon_u8, b_.neon_u8));3476
+ uint16_t r0 = t[0] + t[1] + t[2] + t[3];3477
+ r_.neon_u16 = vset_lane_u16(r0, vdup_n_u16(0), 0);3478
+#else3479
+ uint16_t sum = 0;3480
+3481
+#if defined(SIMDE_HAVE_STDLIB_H)3482
+ SIMDE_VECTORIZE_REDUCTION(+ : sum)3483
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {3484
+ sum += HEDLEY_STATIC_CAST(uint8_t, abs(a_.u8[i] - b_.u8[i]));3485
+ }3486
+3487
+ r_.i16[0] = HEDLEY_STATIC_CAST(int16_t, sum);3488
+ r_.i16[1] = 0;3489
+ r_.i16[2] = 0;3490
+ r_.i16[3] = 0;3491
+#else3492
+ HEDLEY_UNREACHABLE();3493
+#endif3494
+#endif3495
+3496
+ return simde__m64_from_private(r_);3497
+#endif3498
+}3499
+#define simde_m_psadbw(a, b) simde_mm_sad_pu8(a, b)3500
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3501
+#define _mm_sad_pu8(a, b) simde_mm_sad_pu8(a, b)3502
+#define _m_psadbw(a, b) simde_mm_sad_pu8(a, b)3503
+#endif3504
+3505
+SIMDE_FUNCTION_ATTRIBUTES3506
+simde__m128 simde_mm_set_ss(simde_float32 a)3507
+{3508
+#if defined(SIMDE_X86_SSE_NATIVE)3509
+ return _mm_set_ss(a);3510
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3511
+ return vsetq_lane_f32(a, vdupq_n_f32(SIMDE_FLOAT32_C(0.0)), 0);3512
+#else3513
+ return simde_mm_set_ps(SIMDE_FLOAT32_C(0.0), SIMDE_FLOAT32_C(0.0),3514
+ SIMDE_FLOAT32_C(0.0), a);3515
+#endif3516
+}3517
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3518
+#define _mm_set_ss(a) simde_mm_set_ss(a)3519
+#endif3520
+3521
+SIMDE_FUNCTION_ATTRIBUTES3522
+simde__m128 simde_mm_setr_ps(simde_float32 e3, simde_float32 e2,3523
+ simde_float32 e1, simde_float32 e0)3524
+{3525
+#if defined(SIMDE_X86_SSE_NATIVE)3526
+ return _mm_setr_ps(e3, e2, e1, e0);3527
+#else3528
+ return simde_mm_set_ps(e0, e1, e2, e3);3529
+#endif3530
+}3531
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3532
+#define _mm_setr_ps(e3, e2, e1, e0) simde_mm_setr_ps(e3, e2, e1, e0)3533
+#endif3534
+3535
+SIMDE_FUNCTION_ATTRIBUTES3536
+simde__m128 simde_mm_setzero_ps(void)3537
+{3538
+#if defined(SIMDE_X86_SSE_NATIVE)3539
+ return _mm_setzero_ps();3540
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3541
+ return vdupq_n_f32(SIMDE_FLOAT32_C(0.0));3542
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3543
+ return vec_splats(SIMDE_FLOAT32_C(0.0));3544
+#else3545
+ simde__m128 r;3546
+ simde_memset(&r, 0, sizeof(r));3547
+ return r;3548
+#endif3549
+}3550
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3551
+#define _mm_setzero_ps() simde_mm_setzero_ps()3552
+#endif3553
+3554
+#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)3555
+HEDLEY_DIAGNOSTIC_PUSH3556
+SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_3557
+#endif3558
+3559
+SIMDE_FUNCTION_ATTRIBUTES3560
+simde__m128 simde_mm_undefined_ps(void)3561
+{3562
+ simde__m128_private r_;3563
+3564
+#if defined(SIMDE_HAVE_UNDEFINED128)3565
+ r_.n = _mm_undefined_ps();3566
+#elif !defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)3567
+ r_ = simde__m128_to_private(simde_mm_setzero_ps());3568
+#endif3569
+3570
+ return simde__m128_from_private(r_);3571
+}3572
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3573
+#define _mm_undefined_ps() simde_mm_undefined_ps()3574
+#endif3575
+3576
+#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)3577
+HEDLEY_DIAGNOSTIC_POP3578
+#endif3579
+3580
+SIMDE_FUNCTION_ATTRIBUTES3581
+simde__m128 simde_x_mm_setone_ps(void)3582
+{3583
+ simde__m128 t = simde_mm_setzero_ps();3584
+ return simde_mm_cmpeq_ps(t, t);3585
+}3586
+3587
+SIMDE_FUNCTION_ATTRIBUTES3588
+void simde_mm_sfence(void)3589
+{3590
+ /* TODO: Use Hedley. */3591
+#if defined(SIMDE_X86_SSE_NATIVE)3592
+ _mm_sfence();3593
+#elif defined(__GNUC__) && \3594
+ ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 7))3595
+ __atomic_thread_fence(__ATOMIC_SEQ_CST);3596
+#elif !defined(__INTEL_COMPILER) && defined(__STDC_VERSION__) && \3597
+ (__STDC_VERSION__ >= 201112L) && !defined(__STDC_NO_ATOMICS__)3598
+#if defined(__GNUC__) && (__GNUC__ == 4) && (__GNUC_MINOR__ < 9)3599
+ __atomic_thread_fence(__ATOMIC_SEQ_CST);3600
+#else3601
+ atomic_thread_fence(memory_order_seq_cst);3602
+#endif3603
+#elif defined(_MSC_VER)3604
+ MemoryBarrier();3605
+#elif HEDLEY_HAS_EXTENSION(c_atomic)3606
+ __c11_atomic_thread_fence(__ATOMIC_SEQ_CST);3607
+#elif defined(__GNUC__) && \3608
+ ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 1))3609
+ __sync_synchronize();3610
+#elif defined(_OPENMP)3611
+#pragma omp critical(simde_mm_sfence_)3612
+ {3613
+ }3614
+#endif3615
+}3616
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3617
+#define _mm_sfence() simde_mm_sfence()3618
+#endif3619
+3620
+#define SIMDE_MM_SHUFFLE(z, y, x, w) \3621
+ (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))3622
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3623
+#define _MM_SHUFFLE(z, y, x, w) SIMDE_MM_SHUFFLE(z, y, x, w)3624
+#endif3625
+3626
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \3627
+ !defined(__PGI)3628
+#define simde_mm_shuffle_pi16(a, imm8) _mm_shuffle_pi16(a, imm8)3629
+#elif defined(SIMDE_SHUFFLE_VECTOR_)3630
+#define simde_mm_shuffle_pi16(a, imm8) \3631
+ (__extension__({ \3632
+ const simde__m64_private simde__tmp_a_ = \3633
+ simde__m64_to_private(a); \3634
+ simde__m64_from_private((simde__m64_private){ \3635
+ .i16 = SIMDE_SHUFFLE_VECTOR_( \3636
+ 16, 8, (simde__tmp_a_).i16, \3637
+ (simde__tmp_a_).i16, (((imm8)) & 3), \3638
+ (((imm8) >> 2) & 3), (((imm8) >> 4) & 3), \3639
+ (((imm8) >> 6) & 3))}); \3640
+ }))3641
+#else3642
+SIMDE_FUNCTION_ATTRIBUTES3643
+simde__m64 simde_mm_shuffle_pi16(simde__m64 a, const int imm8)3644
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)3645
+{3646
+ simde__m64_private r_;3647
+ simde__m64_private a_ = simde__m64_to_private(a);3648
+3649
+ for (size_t i = 0; i < sizeof(r_.i16) / sizeof(r_.i16[0]); i++) {3650
+ r_.i16[i] = a_.i16[(imm8 >> (i * 2)) & 3];3651
+ }3652
+3653
+ HEDLEY_DIAGNOSTIC_PUSH3654
+#if HEDLEY_HAS_WARNING("-Wconditional-uninitialized")3655
+#pragma clang diagnostic ignored "-Wconditional-uninitialized"3656
+#endif3657
+ return simde__m64_from_private(r_);3658
+ HEDLEY_DIAGNOSTIC_POP3659
+}3660
+#endif3661
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \3662
+ !defined(__PGI)3663
+#define simde_m_pshufw(a, imm8) _m_pshufw(a, imm8)3664
+#else3665
+#define simde_m_pshufw(a, imm8) simde_mm_shuffle_pi16(a, imm8)3666
+#endif3667
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3668
+#define _mm_shuffle_pi16(a, imm8) simde_mm_shuffle_pi16(a, imm8)3669
+#define _m_pshufw(a, imm8) simde_mm_shuffle_pi16(a, imm8)3670
+#endif3671
+3672
+#if defined(SIMDE_X86_SSE_NATIVE) && !defined(__PGI)3673
+#define simde_mm_shuffle_ps(a, b, imm8) _mm_shuffle_ps(a, b, imm8)3674
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3675
+#define simde_mm_shuffle_ps(a, b, imm8) \3676
+ __extension__({ \3677
+ float32x4_t ret; \3678
+ ret = vmovq_n_f32(vgetq_lane_f32(a, (imm8) & (0x3))); \3679
+ ret = vsetq_lane_f32(vgetq_lane_f32(a, ((imm8) >> 2) & 0x3), \3680
+ ret, 1); \3681
+ ret = vsetq_lane_f32(vgetq_lane_f32(b, ((imm8) >> 4) & 0x3), \3682
+ ret, 2); \3683
+ ret = vsetq_lane_f32(vgetq_lane_f32(b, ((imm8) >> 6) & 0x3), \3684
+ ret, 3); \3685
+ })3686
+#elif defined(SIMDE_SHUFFLE_VECTOR_)3687
+#define simde_mm_shuffle_ps(a, b, imm8) \3688
+ (__extension__({ \3689
+ simde__m128_from_private((simde__m128_private){ \3690
+ .f32 = SIMDE_SHUFFLE_VECTOR_( \3691
+ 32, 16, simde__m128_to_private(a).f32, \3692
+ simde__m128_to_private(b).f32, (((imm8)) & 3), \3693
+ (((imm8) >> 2) & 3), (((imm8) >> 4) & 3) + 4, \3694
+ (((imm8) >> 6) & 3) + 4)}); \3695
+ }))3696
+#else3697
+SIMDE_FUNCTION_ATTRIBUTES3698
+simde__m128 simde_mm_shuffle_ps(simde__m128 a, simde__m128 b, const int imm8)3699
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)3700
+{3701
+ simde__m128_private r_, a_ = simde__m128_to_private(a),3702
+ b_ = simde__m128_to_private(b);3703
+3704
+ r_.f32[0] = a_.f32[(imm8 >> 0) & 3];3705
+ r_.f32[1] = a_.f32[(imm8 >> 2) & 3];3706
+ r_.f32[2] = b_.f32[(imm8 >> 4) & 3];3707
+ r_.f32[3] = b_.f32[(imm8 >> 6) & 3];3708
+3709
+ return simde__m128_from_private(r_);3710
+}3711
+#endif3712
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3713
+#define _mm_shuffle_ps(a, b, imm8) simde_mm_shuffle_ps((a), (b), imm8)3714
+#endif3715
+3716
+SIMDE_FUNCTION_ATTRIBUTES3717
+simde__m128 simde_mm_sqrt_ps(simde__m128 a)3718
+{3719
+#if defined(SIMDE_X86_SSE_NATIVE)3720
+ return _mm_sqrt_ps(a);3721
+#else3722
+ simde__m128_private r_, a_ = simde__m128_to_private(a);3723
+3724
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3725
+ r_.neon_f32 = vsqrtq_f32(a_.neon_f32);3726
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3727
+ float32x4_t est = vrsqrteq_f32(a_.neon_f32);3728
+ for (int i = 0; i <= SIMDE_ACCURACY_PREFERENCE; i++) {3729
+ est = vmulq_f32(vrsqrtsq_f32(vmulq_f32(a_.neon_f32, est), est),3730
+ est);3731
+ }3732
+ r_.neon_f32 = vmulq_f32(a_.neon_f32, est);3733
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3734
+ r_.wasm_v128 = wasm_f32x4_sqrt(a_.wasm_v128);3735
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)3736
+ r_.altivec_f32 = vec_sqrt(a_.altivec_f32);3737
+#elif defined(simde_math_sqrt)3738
+ SIMDE_VECTORIZE3739
+ for (size_t i = 0; i < sizeof(r_.f32) / sizeof(r_.f32[0]); i++) {3740
+ r_.f32[i] = simde_math_sqrtf(a_.f32[i]);3741
+ }3742
+#else3743
+ HEDLEY_UNREACHABLE();3744
+#endif3745
+3746
+ return simde__m128_from_private(r_);3747
+#endif3748
+}3749
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3750
+#define _mm_sqrt_ps(a) simde_mm_sqrt_ps((a))3751
+#endif3752
+3753
+SIMDE_FUNCTION_ATTRIBUTES3754
+simde__m128 simde_mm_sqrt_ss(simde__m128 a)3755
+{3756
+#if defined(SIMDE_X86_SSE_NATIVE)3757
+ return _mm_sqrt_ss(a);3758
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)3759
+ return simde_mm_move_ss(a, simde_mm_sqrt_ps(a));3760
+#else3761
+ simde__m128_private r_, a_ = simde__m128_to_private(a);3762
+3763
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3764
+ float32_t value = vgetq_lane_f32(3765
+ simde__m128_to_private(simde_mm_sqrt_ps(a)).neon_f32, 0);3766
+ r_.neon_f32 = vsetq_lane_f32(value, a_.neon_f32, 0);3767
+#elif defined(simde_math_sqrtf)3768
+ r_.f32[0] = simde_math_sqrtf(a_.f32[0]);3769
+ r_.f32[1] = a_.f32[1];3770
+ r_.f32[2] = a_.f32[2];3771
+ r_.f32[3] = a_.f32[3];3772
+#else3773
+ HEDLEY_UNREACHABLE();3774
+#endif3775
+3776
+ return simde__m128_from_private(r_);3777
+#endif3778
+}3779
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3780
+#define _mm_sqrt_ss(a) simde_mm_sqrt_ss((a))3781
+#endif3782
+3783
+SIMDE_FUNCTION_ATTRIBUTES3784
+void simde_mm_store_ps(simde_float32 mem_addr[4], simde__m128 a)3785
+{3786
+#if defined(SIMDE_X86_SSE_NATIVE)3787
+ _mm_store_ps(mem_addr, a);3788
+#else3789
+ simde__m128_private a_ = simde__m128_to_private(a);3790
+3791
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3792
+ vst1q_f32(mem_addr, a_.neon_f32);3793
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3794
+ vec_st(a_.altivec_f32, 0, mem_addr);3795
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3796
+ wasm_v128_store(mem_addr, a_.wasm_v128);3797
+#else3798
+ simde_memcpy(mem_addr, &a_, sizeof(a));3799
+#endif3800
+#endif3801
+}3802
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3803
+#define _mm_store_ps(mem_addr, a) \3804
+ simde_mm_store_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3805
+ float *, simde_float32 *, mem_addr), \3806
+ (a))3807
+#endif3808
+3809
+SIMDE_FUNCTION_ATTRIBUTES3810
+void simde_mm_store1_ps(simde_float32 mem_addr[4], simde__m128 a)3811
+{3812
+ simde_float32 *mem_addr_ =3813
+ SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m128);3814
+3815
+#if defined(SIMDE_X86_SSE_NATIVE)3816
+ _mm_store_ps1(mem_addr_, a);3817
+#else3818
+ simde__m128_private a_ = simde__m128_to_private(a);3819
+3820
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3821
+ vst1q_f32(mem_addr_, vdupq_lane_f32(vget_low_f32(a_.neon_f32), 0));3822
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3823
+ wasm_v128_store(mem_addr_,3824
+ wasm_v32x4_shuffle(a_.wasm_v128, a_.wasm_v128, 0, 0, 0,3825
+ 0));3826
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3827
+ vec_st(vec_splat(a_.altivec_f32, 0), 0, mem_addr_);3828
+#elif defined(SIMDE_SHUFFLE_VECTOR_)3829
+ simde__m128_private tmp_;3830
+ tmp_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, a_.f32, 0, 0, 0, 0);3831
+ simde_mm_store_ps(mem_addr_, tmp_.f32);3832
+#else3833
+ SIMDE_VECTORIZE_ALIGNED(mem_addr_ : 16)3834
+ for (size_t i = 0; i < sizeof(a_.f32) / sizeof(a_.f32[0]); i++) {3835
+ mem_addr_[i] = a_.f32[0];3836
+ }3837
+#endif3838
+#endif3839
+}3840
+#define simde_mm_store_ps1(mem_addr, a) simde_mm_store1_ps(mem_addr, a)3841
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3842
+#define _mm_store_ps1(mem_addr, a) \3843
+ simde_mm_store1_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3844
+ float *, simde_float32 *, mem_addr), \3845
+ (a))3846
+#define _mm_store1_ps(mem_addr, a) \3847
+ simde_mm_store1_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3848
+ float *, simde_float32 *, mem_addr), \3849
+ (a))3850
+#endif3851
+3852
+SIMDE_FUNCTION_ATTRIBUTES3853
+void simde_mm_store_ss(simde_float32 *mem_addr, simde__m128 a)3854
+{3855
+#if defined(SIMDE_X86_SSE_NATIVE)3856
+ _mm_store_ss(mem_addr, a);3857
+#else3858
+ simde__m128_private a_ = simde__m128_to_private(a);3859
+3860
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3861
+ vst1q_lane_f32(mem_addr, a_.neon_f32, 0);3862
+#else3863
+ *mem_addr = a_.f32[0];3864
+#endif3865
+#endif3866
+}3867
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3868
+#define _mm_store_ss(mem_addr, a) \3869
+ simde_mm_store_ss(SIMDE_CHECKED_REINTERPRET_CAST( \3870
+ float *, simde_float32 *, mem_addr), \3871
+ (a))3872
+#endif3873
+3874
+SIMDE_FUNCTION_ATTRIBUTES3875
+void simde_mm_storeh_pi(simde__m64 *mem_addr, simde__m128 a)3876
+{3877
+#if defined(SIMDE_X86_SSE_NATIVE)3878
+ _mm_storeh_pi(HEDLEY_REINTERPRET_CAST(__m64 *, mem_addr), a);3879
+#else3880
+ simde__m128_private a_ = simde__m128_to_private(a);3881
+3882
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3883
+ vst1_f32(HEDLEY_REINTERPRET_CAST(float32_t *, mem_addr),3884
+ vget_high_f32(a_.neon_f32));3885
+#else3886
+ simde_memcpy(mem_addr, &(a_.m64[1]), sizeof(a_.m64[1]));3887
+#endif3888
+#endif3889
+}3890
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3891
+#define _mm_storeh_pi(mem_addr, a) simde_mm_storeh_pi(mem_addr, (a))3892
+#endif3893
+3894
+SIMDE_FUNCTION_ATTRIBUTES3895
+void simde_mm_storel_pi(simde__m64 *mem_addr, simde__m128 a)3896
+{3897
+#if defined(SIMDE_X86_SSE_NATIVE)3898
+ _mm_storel_pi(HEDLEY_REINTERPRET_CAST(__m64 *, mem_addr), a);3899
+#else3900
+ simde__m64_private *dest_ =3901
+ HEDLEY_REINTERPRET_CAST(simde__m64_private *, mem_addr);3902
+ simde__m128_private a_ = simde__m128_to_private(a);3903
+3904
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3905
+ dest_->neon_f32 = vget_low_f32(a_.neon_f32);3906
+#else3907
+ dest_->f32[0] = a_.f32[0];3908
+ dest_->f32[1] = a_.f32[1];3909
+#endif3910
+#endif3911
+}3912
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3913
+#define _mm_storel_pi(mem_addr, a) simde_mm_storel_pi(mem_addr, (a))3914
+#endif3915
+3916
+SIMDE_FUNCTION_ATTRIBUTES3917
+void simde_mm_storer_ps(simde_float32 mem_addr[4], simde__m128 a)3918
+{3919
+#if defined(SIMDE_X86_SSE_NATIVE)3920
+ _mm_storer_ps(mem_addr, a);3921
+#else3922
+ simde__m128_private a_ = simde__m128_to_private(a);3923
+3924
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3925
+ vec_st(vec_reve(a_.altivec_f32), 0, mem_addr);3926
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3927
+ float32x4_t tmp = vrev64q_f32(a_.neon_f32);3928
+ vst1q_f32(mem_addr, vextq_f32(tmp, tmp, 2));3929
+#elif defined(SIMDE_SHUFFLE_VECTOR_)3930
+ a_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, a_.f32, 3, 2, 1, 0);3931
+ simde_mm_store_ps(mem_addr, simde__m128_from_private(a_));3932
+#else3933
+ SIMDE_VECTORIZE_ALIGNED(mem_addr : 16)3934
+ for (size_t i = 0; i < sizeof(a_.f32) / sizeof(a_.f32[0]); i++) {3935
+ mem_addr[i] =3936
+ a_.f32[((sizeof(a_.f32) / sizeof(a_.f32[0])) - 1) - i];3937
+ }3938
+#endif3939
+#endif3940
+}3941
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3942
+#define _mm_storer_ps(mem_addr, a) \3943
+ simde_mm_storer_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3944
+ float *, simde_float32 *, mem_addr), \3945
+ (a))3946
+#endif3947
+3948
+SIMDE_FUNCTION_ATTRIBUTES3949
+void simde_mm_storeu_ps(simde_float32 mem_addr[4], simde__m128 a)3950
+{3951
+#if defined(SIMDE_X86_SSE_NATIVE)3952
+ _mm_storeu_ps(mem_addr, a);3953
+#else3954
+ simde__m128_private a_ = simde__m128_to_private(a);3955
+3956
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3957
+ vst1q_f32(mem_addr, a_.neon_f32);3958
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)3959
+ vec_vsx_st(a_.altivec_f32, 0, mem_addr);3960
+#else3961
+ simde_memcpy(mem_addr, &a_, sizeof(a_));3962
+#endif3963
+#endif3964
+}3965
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)3966
+#define _mm_storeu_ps(mem_addr, a) \3967
+ simde_mm_storeu_ps(SIMDE_CHECKED_REINTERPRET_CAST( \3968
+ float *, simde_float32 *, mem_addr), \3969
+ (a))3970
+#endif3971
+3972
+SIMDE_FUNCTION_ATTRIBUTES3973
+simde__m128 simde_mm_sub_ps(simde__m128 a, simde__m128 b)3974
+{3975
+#if defined(SIMDE_X86_SSE_NATIVE)3976
+ return _mm_sub_ps(a, b);3977
+#else3978
+ simde__m128_private r_, a_ = simde__m128_to_private(a),3979
+ b_ = simde__m128_to_private(b);3980
+3981
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3982
+ r_.neon_f32 = vsubq_f32(a_.neon_f32, b_.neon_f32);3983
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3984
+ r_.wasm_v128 = wasm_f32x4_sub(a_.wasm_v128, b_.wasm_v128);3985
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3986
+ r_.altivec_f32 = vec_sub(a_.altivec_f32, b_.altivec_f32);3987
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3988
+ r_.f32 = a_.f32 - b_.f32;3989
+#else3990
+ SIMDE_VECTORIZE3991
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {3992
+ r_.f32[i] = a_.f32[i] - b_.f32[i];3993
+ }3994
+#endif3995
+3996
+ return simde__m128_from_private(r_);3997
+#endif3998
+}3999
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4000
+#define _mm_sub_ps(a, b) simde_mm_sub_ps((a), (b))4001
+#endif4002
+4003
+SIMDE_FUNCTION_ATTRIBUTES4004
+simde__m128 simde_mm_sub_ss(simde__m128 a, simde__m128 b)4005
+{4006
+#if defined(SIMDE_X86_SSE_NATIVE)4007
+ return _mm_sub_ss(a, b);4008
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)4009
+ return simde_mm_move_ss(a, simde_mm_sub_ps(a, b));4010
+#else4011
+ simde__m128_private r_, a_ = simde__m128_to_private(a),4012
+ b_ = simde__m128_to_private(b);4013
+4014
+ r_.f32[0] = a_.f32[0] - b_.f32[0];4015
+ r_.f32[1] = a_.f32[1];4016
+ r_.f32[2] = a_.f32[2];4017
+ r_.f32[3] = a_.f32[3];4018
+4019
+ return simde__m128_from_private(r_);4020
+#endif4021
+}4022
+4023
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4024
+#define _mm_sub_ss(a, b) simde_mm_sub_ss((a), (b))4025
+#endif4026
+4027
+SIMDE_FUNCTION_ATTRIBUTES4028
+int simde_mm_ucomieq_ss(simde__m128 a, simde__m128 b)4029
+{4030
+#if defined(SIMDE_X86_SSE_NATIVE)4031
+ return _mm_ucomieq_ss(a, b);4032
+#else4033
+ simde__m128_private a_ = simde__m128_to_private(a),4034
+ b_ = simde__m128_to_private(b);4035
+ int r;4036
+4037
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4038
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);4039
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);4040
+ uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));4041
+ uint32x4_t a_eq_b = vceqq_f32(a_.neon_f32, b_.neon_f32);4042
+ r = !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_eq_b), 0) != 0);4043
+#elif defined(SIMDE_HAVE_FENV_H)4044
+ fenv_t envp;4045
+ int x = feholdexcept(&envp);4046
+ r = a_.f32[0] == b_.f32[0];4047
+ if (HEDLEY_LIKELY(x == 0))4048
+ fesetenv(&envp);4049
+#else4050
+ r = a_.f32[0] == b_.f32[0];4051
+#endif4052
+4053
+ return r;4054
+#endif4055
+}4056
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4057
+#define _mm_ucomieq_ss(a, b) simde_mm_ucomieq_ss((a), (b))4058
+#endif4059
+4060
+SIMDE_FUNCTION_ATTRIBUTES4061
+int simde_mm_ucomige_ss(simde__m128 a, simde__m128 b)4062
+{4063
+#if defined(SIMDE_X86_SSE_NATIVE)4064
+ return _mm_ucomige_ss(a, b);4065
+#else4066
+ simde__m128_private a_ = simde__m128_to_private(a),4067
+ b_ = simde__m128_to_private(b);4068
+ int r;4069
+4070
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4071
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);4072
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);4073
+ uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);4074
+ uint32x4_t a_ge_b = vcgeq_f32(a_.neon_f32, b_.neon_f32);4075
+ r = !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_ge_b), 0) != 0);4076
+#elif defined(SIMDE_HAVE_FENV_H)4077
+ fenv_t envp;4078
+ int x = feholdexcept(&envp);4079
+ r = a_.f32[0] >= b_.f32[0];4080
+ if (HEDLEY_LIKELY(x == 0))4081
+ fesetenv(&envp);4082
+#else4083
+ r = a_.f32[0] >= b_.f32[0];4084
+#endif4085
+4086
+ return r;4087
+#endif4088
+}4089
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4090
+#define _mm_ucomige_ss(a, b) simde_mm_ucomige_ss((a), (b))4091
+#endif4092
+4093
+SIMDE_FUNCTION_ATTRIBUTES4094
+int simde_mm_ucomigt_ss(simde__m128 a, simde__m128 b)4095
+{4096
+#if defined(SIMDE_X86_SSE_NATIVE)4097
+ return _mm_ucomigt_ss(a, b);4098
+#else4099
+ simde__m128_private a_ = simde__m128_to_private(a),4100
+ b_ = simde__m128_to_private(b);4101
+ int r;4102
+4103
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4104
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);4105
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);4106
+ uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);4107
+ uint32x4_t a_gt_b = vcgtq_f32(a_.neon_f32, b_.neon_f32);4108
+ r = !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_gt_b), 0) != 0);4109
+#elif defined(SIMDE_HAVE_FENV_H)4110
+ fenv_t envp;4111
+ int x = feholdexcept(&envp);4112
+ r = a_.f32[0] > b_.f32[0];4113
+ if (HEDLEY_LIKELY(x == 0))4114
+ fesetenv(&envp);4115
+#else4116
+ r = a_.f32[0] > b_.f32[0];4117
+#endif4118
+4119
+ return r;4120
+#endif4121
+}4122
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4123
+#define _mm_ucomigt_ss(a, b) simde_mm_ucomigt_ss((a), (b))4124
+#endif4125
+4126
+SIMDE_FUNCTION_ATTRIBUTES4127
+int simde_mm_ucomile_ss(simde__m128 a, simde__m128 b)4128
+{4129
+#if defined(SIMDE_X86_SSE_NATIVE)4130
+ return _mm_ucomile_ss(a, b);4131
+#else4132
+ simde__m128_private a_ = simde__m128_to_private(a),4133
+ b_ = simde__m128_to_private(b);4134
+ int r;4135
+4136
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4137
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);4138
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);4139
+ uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));4140
+ uint32x4_t a_le_b = vcleq_f32(a_.neon_f32, b_.neon_f32);4141
+ r = !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_le_b), 0) != 0);4142
+#elif defined(SIMDE_HAVE_FENV_H)4143
+ fenv_t envp;4144
+ int x = feholdexcept(&envp);4145
+ r = a_.f32[0] <= b_.f32[0];4146
+ if (HEDLEY_LIKELY(x == 0))4147
+ fesetenv(&envp);4148
+#else4149
+ r = a_.f32[0] <= b_.f32[0];4150
+#endif4151
+4152
+ return r;4153
+#endif4154
+}4155
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4156
+#define _mm_ucomile_ss(a, b) simde_mm_ucomile_ss((a), (b))4157
+#endif4158
+4159
+SIMDE_FUNCTION_ATTRIBUTES4160
+int simde_mm_ucomilt_ss(simde__m128 a, simde__m128 b)4161
+{4162
+#if defined(SIMDE_X86_SSE_NATIVE)4163
+ return _mm_ucomilt_ss(a, b);4164
+#else4165
+ simde__m128_private a_ = simde__m128_to_private(a),4166
+ b_ = simde__m128_to_private(b);4167
+ int r;4168
+4169
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4170
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);4171
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);4172
+ uint32x4_t a_or_b_nan = vmvnq_u32(vandq_u32(a_not_nan, b_not_nan));4173
+ uint32x4_t a_lt_b = vcltq_f32(a_.neon_f32, b_.neon_f32);4174
+ r = !!(vgetq_lane_u32(vorrq_u32(a_or_b_nan, a_lt_b), 0) != 0);4175
+#elif defined(SIMDE_HAVE_FENV_H)4176
+ fenv_t envp;4177
+ int x = feholdexcept(&envp);4178
+ r = a_.f32[0] < b_.f32[0];4179
+ if (HEDLEY_LIKELY(x == 0))4180
+ fesetenv(&envp);4181
+#else4182
+ r = a_.f32[0] < b_.f32[0];4183
+#endif4184
+4185
+ return r;4186
+#endif4187
+}4188
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4189
+#define _mm_ucomilt_ss(a, b) simde_mm_ucomilt_ss((a), (b))4190
+#endif4191
+4192
+SIMDE_FUNCTION_ATTRIBUTES4193
+int simde_mm_ucomineq_ss(simde__m128 a, simde__m128 b)4194
+{4195
+#if defined(SIMDE_X86_SSE_NATIVE)4196
+ return _mm_ucomineq_ss(a, b);4197
+#else4198
+ simde__m128_private a_ = simde__m128_to_private(a),4199
+ b_ = simde__m128_to_private(b);4200
+ int r;4201
+4202
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4203
+ uint32x4_t a_not_nan = vceqq_f32(a_.neon_f32, a_.neon_f32);4204
+ uint32x4_t b_not_nan = vceqq_f32(b_.neon_f32, b_.neon_f32);4205
+ uint32x4_t a_and_b_not_nan = vandq_u32(a_not_nan, b_not_nan);4206
+ uint32x4_t a_neq_b = vmvnq_u32(vceqq_f32(a_.neon_f32, b_.neon_f32));4207
+ r = !!(vgetq_lane_u32(vandq_u32(a_and_b_not_nan, a_neq_b), 0) != 0);4208
+#elif defined(SIMDE_HAVE_FENV_H)4209
+ fenv_t envp;4210
+ int x = feholdexcept(&envp);4211
+ r = a_.f32[0] != b_.f32[0];4212
+ if (HEDLEY_LIKELY(x == 0))4213
+ fesetenv(&envp);4214
+#else4215
+ r = a_.f32[0] != b_.f32[0];4216
+#endif4217
+4218
+ return r;4219
+#endif4220
+}4221
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4222
+#define _mm_ucomineq_ss(a, b) simde_mm_ucomineq_ss((a), (b))4223
+#endif4224
+4225
+#if defined(SIMDE_X86_SSE_NATIVE)4226
+#if defined(__has_builtin)4227
+#if __has_builtin(__builtin_ia32_undef128)4228
+#define SIMDE_HAVE_UNDEFINED1284229
+#endif4230
+#elif !defined(__PGI) && !defined(SIMDE_BUG_GCC_REV_208793) && \4231
+ !defined(_MSC_VER)4232
+#define SIMDE_HAVE_UNDEFINED1284233
+#endif4234
+#endif4235
+4236
+#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)4237
+HEDLEY_DIAGNOSTIC_PUSH4238
+SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_4239
+#endif4240
+4241
+SIMDE_FUNCTION_ATTRIBUTES4242
+simde__m128 simde_mm_unpackhi_ps(simde__m128 a, simde__m128 b)4243
+{4244
+#if defined(SIMDE_X86_SSE_NATIVE)4245
+ return _mm_unpackhi_ps(a, b);4246
+#else4247
+ simde__m128_private r_, a_ = simde__m128_to_private(a),4248
+ b_ = simde__m128_to_private(b);4249
+4250
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)4251
+ r_.neon_f32 = vzip2q_f32(a_.neon_f32, b_.neon_f32);4252
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)4253
+ float32x2_t a1 = vget_high_f32(a_.neon_f32);4254
+ float32x2_t b1 = vget_high_f32(b_.neon_f32);4255
+ float32x2x2_t result = vzip_f32(a1, b1);4256
+ r_.neon_f32 = vcombine_f32(result.val[0], result.val[1]);4257
+#elif defined(SIMDE_SHUFFLE_VECTOR_)4258
+ r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 2, 6, 3, 7);4259
+#else4260
+ r_.f32[0] = a_.f32[2];4261
+ r_.f32[1] = b_.f32[2];4262
+ r_.f32[2] = a_.f32[3];4263
+ r_.f32[3] = b_.f32[3];4264
+#endif4265
+4266
+ return simde__m128_from_private(r_);4267
+#endif4268
+}4269
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4270
+#define _mm_unpackhi_ps(a, b) simde_mm_unpackhi_ps((a), (b))4271
+#endif4272
+4273
+SIMDE_FUNCTION_ATTRIBUTES4274
+simde__m128 simde_mm_unpacklo_ps(simde__m128 a, simde__m128 b)4275
+{4276
+#if defined(SIMDE_X86_SSE_NATIVE)4277
+ return _mm_unpacklo_ps(a, b);4278
+#else4279
+ simde__m128_private r_, a_ = simde__m128_to_private(a),4280
+ b_ = simde__m128_to_private(b);4281
+4282
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)4283
+ r_.neon_f32 = vzip1q_f32(a_.neon_f32, b_.neon_f32);4284
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)4285
+ r_.altivec_f32 = vec_mergeh(a_.altivec_f32, b_.altivec_f32);4286
+#elif defined(SIMDE_SHUFFLE_VECTOR_)4287
+ r_.f32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.f32, b_.f32, 0, 4, 1, 5);4288
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)4289
+ float32x2_t a1 = vget_low_f32(a_.neon_f32);4290
+ float32x2_t b1 = vget_low_f32(b_.neon_f32);4291
+ float32x2x2_t result = vzip_f32(a1, b1);4292
+ r_.neon_f32 = vcombine_f32(result.val[0], result.val[1]);4293
+#else4294
+ r_.f32[0] = a_.f32[0];4295
+ r_.f32[1] = b_.f32[0];4296
+ r_.f32[2] = a_.f32[1];4297
+ r_.f32[3] = b_.f32[1];4298
+#endif4299
+4300
+ return simde__m128_from_private(r_);4301
+#endif4302
+}4303
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4304
+#define _mm_unpacklo_ps(a, b) simde_mm_unpacklo_ps((a), (b))4305
+#endif4306
+4307
+SIMDE_FUNCTION_ATTRIBUTES4308
+void simde_mm_stream_pi(simde__m64 *mem_addr, simde__m64 a)4309
+{4310
+#if defined(SIMDE_X86_SSE_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)4311
+ _mm_stream_pi(HEDLEY_REINTERPRET_CAST(__m64 *, mem_addr), a);4312
+#else4313
+ simde__m64_private *dest = HEDLEY_REINTERPRET_CAST(simde__m64_private *,4314
+ mem_addr),4315
+ a_ = simde__m64_to_private(a);4316
+4317
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4318
+ dest->i64[0] = vget_lane_s64(a_.neon_i64, 0);4319
+#else4320
+ dest->i64[0] = a_.i64[0];4321
+#endif4322
+#endif4323
+}4324
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4325
+#define _mm_stream_pi(mem_addr, a) simde_mm_stream_pi(mem_addr, (a))4326
+#endif4327
+4328
+SIMDE_FUNCTION_ATTRIBUTES4329
+void simde_mm_stream_ps(simde_float32 mem_addr[4], simde__m128 a)4330
+{4331
+#if defined(SIMDE_X86_SSE_NATIVE)4332
+ _mm_stream_ps(mem_addr, a);4333
+#elif HEDLEY_HAS_BUILTIN(__builtin_nontemporal_store) && \4334
+ defined(SIMDE_VECTOR_SUBSCRIPT_OPS)4335
+ simde__m128_private a_ = simde__m128_to_private(a);4336
+ __builtin_nontemporal_store(4337
+ a_.f32, SIMDE_ALIGN_CAST(__typeof__(a_.f32) *, mem_addr));4338
+#else4339
+ simde_mm_store_ps(mem_addr, a);4340
+#endif4341
+}4342
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4343
+#define _mm_stream_ps(mem_addr, a) \4344
+ simde_mm_stream_ps(SIMDE_CHECKED_REINTERPRET_CAST( \4345
+ float *, simde_float32 *, mem_addr), \4346
+ (a))4347
+#endif4348
+4349
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4350
+#define SIMDE_MM_TRANSPOSE4_PS(row0, row1, row2, row3) \4351
+ do { \4352
+ float32x4x2_t ROW01 = vtrnq_f32(row0, row1); \4353
+ float32x4x2_t ROW23 = vtrnq_f32(row2, row3); \4354
+ row0 = vcombine_f32(vget_low_f32(ROW01.val[0]), \4355
+ vget_low_f32(ROW23.val[0])); \4356
+ row1 = vcombine_f32(vget_low_f32(ROW01.val[1]), \4357
+ vget_low_f32(ROW23.val[1])); \4358
+ row2 = vcombine_f32(vget_high_f32(ROW01.val[0]), \4359
+ vget_high_f32(ROW23.val[0])); \4360
+ row3 = vcombine_f32(vget_high_f32(ROW01.val[1]), \4361
+ vget_high_f32(ROW23.val[1])); \4362
+ } while (0)4363
+#else4364
+#define SIMDE_MM_TRANSPOSE4_PS(row0, row1, row2, row3) \4365
+ do { \4366
+ simde__m128 tmp3, tmp2, tmp1, tmp0; \4367
+ tmp0 = simde_mm_unpacklo_ps((row0), (row1)); \4368
+ tmp2 = simde_mm_unpacklo_ps((row2), (row3)); \4369
+ tmp1 = simde_mm_unpackhi_ps((row0), (row1)); \4370
+ tmp3 = simde_mm_unpackhi_ps((row2), (row3)); \4371
+ row0 = simde_mm_movelh_ps(tmp0, tmp2); \4372
+ row1 = simde_mm_movehl_ps(tmp2, tmp0); \4373
+ row2 = simde_mm_movelh_ps(tmp1, tmp3); \4374
+ row3 = simde_mm_movehl_ps(tmp3, tmp1); \4375
+ } while (0)4376
+#endif4377
+#if defined(SIMDE_X86_SSE_ENABLE_NATIVE_ALIASES)4378
+#define _MM_TRANSPOSE4_PS(row0, row1, row2, row3) \4379
+ SIMDE_MM_TRANSPOSE4_PS(row0, row1, row2, row3)4380
+#endif4381
+4382
+#if defined(_MM_EXCEPT_INVALID)4383
+#define SIMDE_MM_EXCEPT_INVALID _MM_EXCEPT_INVALID4384
+#else4385
+#define SIMDE_MM_EXCEPT_INVALID (0x0001)4386
+#endif4387
+#if defined(_MM_EXCEPT_DENORM)4388
+#define SIMDE_MM_EXCEPT_DENORM _MM_EXCEPT_DENORM4389
+#else4390
+#define SIMDE_MM_EXCEPT_DENORM (0x0002)4391
+#endif4392
+#if defined(_MM_EXCEPT_DIV_ZERO)4393
+#define SIMDE_MM_EXCEPT_DIV_ZERO _MM_EXCEPT_DIV_ZERO4394
+#else4395
+#define SIMDE_MM_EXCEPT_DIV_ZERO (0x0004)4396
+#endif4397
+#if defined(_MM_EXCEPT_OVERFLOW)4398
+#define SIMDE_MM_EXCEPT_OVERFLOW _MM_EXCEPT_OVERFLOW4399
+#else4400
+#define SIMDE_MM_EXCEPT_OVERFLOW (0x0008)4401
+#endif4402
+#if defined(_MM_EXCEPT_UNDERFLOW)4403
+#define SIMDE_MM_EXCEPT_UNDERFLOW _MM_EXCEPT_UNDERFLOW4404
+#else4405
+#define SIMDE_MM_EXCEPT_UNDERFLOW (0x0010)4406
+#endif4407
+#if defined(_MM_EXCEPT_INEXACT)4408
+#define SIMDE_MM_EXCEPT_INEXACT _MM_EXCEPT_INEXACT4409
+#else4410
+#define SIMDE_MM_EXCEPT_INEXACT (0x0020)4411
+#endif4412
+#if defined(_MM_EXCEPT_MASK)4413
+#define SIMDE_MM_EXCEPT_MASK _MM_EXCEPT_MASK4414
+#else4415
+#define SIMDE_MM_EXCEPT_MASK \4416
+ (SIMDE_MM_EXCEPT_INVALID | SIMDE_MM_EXCEPT_DENORM | \4417
+ SIMDE_MM_EXCEPT_DIV_ZERO | SIMDE_MM_EXCEPT_OVERFLOW | \4418
+ SIMDE_MM_EXCEPT_UNDERFLOW | SIMDE_MM_EXCEPT_INEXACT)4419
+#endif4420
+4421
+#if defined(_MM_MASK_INVALID)4422
+#define SIMDE_MM_MASK_INVALID _MM_MASK_INVALID4423
+#else4424
+#define SIMDE_MM_MASK_INVALID (0x0080)4425
+#endif4426
+#if defined(_MM_MASK_DENORM)4427
+#define SIMDE_MM_MASK_DENORM _MM_MASK_DENORM4428
+#else4429
+#define SIMDE_MM_MASK_DENORM (0x0100)4430
+#endif4431
+#if defined(_MM_MASK_DIV_ZERO)4432
+#define SIMDE_MM_MASK_DIV_ZERO _MM_MASK_DIV_ZERO4433
+#else4434
+#define SIMDE_MM_MASK_DIV_ZERO (0x0200)4435
+#endif4436
+#if defined(_MM_MASK_OVERFLOW)4437
+#define SIMDE_MM_MASK_OVERFLOW _MM_MASK_OVERFLOW4438
+#else4439
+#define SIMDE_MM_MASK_OVERFLOW (0x0400)4440
+#endif4441
+#if defined(_MM_MASK_UNDERFLOW)4442
+#define SIMDE_MM_MASK_UNDERFLOW _MM_MASK_UNDERFLOW4443
+#else4444
+#define SIMDE_MM_MASK_UNDERFLOW (0x0800)4445
+#endif4446
+#if defined(_MM_MASK_INEXACT)4447
+#define SIMDE_MM_MASK_INEXACT _MM_MASK_INEXACT4448
+#else4449
+#define SIMDE_MM_MASK_INEXACT (0x1000)4450
+#endif4451
+#if defined(_MM_MASK_MASK)4452
+#define SIMDE_MM_MASK_MASK _MM_MASK_MASK4453
+#else4454
+#define SIMDE_MM_MASK_MASK \4455
+ (SIMDE_MM_MASK_INVALID | SIMDE_MM_MASK_DENORM | \4456
+ SIMDE_MM_MASK_DIV_ZERO | SIMDE_MM_MASK_OVERFLOW | \4457
+ SIMDE_MM_MASK_UNDERFLOW | SIMDE_MM_MASK_INEXACT)4458
+#endif4459
+4460
+#if defined(_MM_FLUSH_ZERO_MASK)4461
+#define SIMDE_MM_FLUSH_ZERO_MASK _MM_FLUSH_ZERO_MASK4462
+#else4463
+#define SIMDE_MM_FLUSH_ZERO_MASK (0x8000)4464
+#endif4465
+#if defined(_MM_FLUSH_ZERO_ON)4466
+#define SIMDE_MM_FLUSH_ZERO_ON _MM_FLUSH_ZERO_ON4467
+#else4468
+#define SIMDE_MM_FLUSH_ZERO_ON (0x8000)4469
+#endif4470
+#if defined(_MM_FLUSH_ZERO_OFF)4471
+#define SIMDE_MM_FLUSH_ZERO_OFF _MM_FLUSH_ZERO_OFF4472
+#else4473
+#define SIMDE_MM_FLUSH_ZERO_OFF (0x0000)4474
+#endif4475
+4476
+SIMDE_END_DECLS_4477
+4478
+HEDLEY_DIAGNOSTIC_POP4479
+4480
+#endif /* !defined(SIMDE_X86_SSE_H) */4481
obs-studio-26.1.1.tar.xz/libobs/util/simde/x86/sse2.h
Added
7551
1
2
+/* SPDX-License-Identifier: MIT3
+ *4
+ * Permission is hereby granted, free of charge, to any person5
+ * obtaining a copy of this software and associated documentation6
+ * files (the "Software"), to deal in the Software without7
+ * restriction, including without limitation the rights to use, copy,8
+ * modify, merge, publish, distribute, sublicense, and/or sell copies9
+ * of the Software, and to permit persons to whom the Software is10
+ * furnished to do so, subject to the following conditions:11
+ *12
+ * The above copyright notice and this permission notice shall be13
+ * included in all copies or substantial portions of the Software.14
+ *15
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,16
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF17
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND18
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS19
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN20
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN21
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE22
+ * SOFTWARE.23
+ *24
+ * Copyright:25
+ * 2017-2020 Evan Nemerson <evan@nemerson.com>26
+ * 2015-2017 John W. Ratcliff <jratcliffscarab@gmail.com>27
+ * 2015 Brandon Rowlett <browlett@nvidia.com>28
+ * 2015 Ken Fast <kfast@gdeb.com>29
+ * 2017 Hasindu Gamaarachchi <hasindu@unsw.edu.au>30
+ * 2018 Jeff Daily <jeff.daily@amd.com>31
+ */32
+33
+#if !defined(SIMDE_X86_SSE2_H)34
+#define SIMDE_X86_SSE2_H35
+36
+#include "sse.h"37
+38
+HEDLEY_DIAGNOSTIC_PUSH39
+SIMDE_DISABLE_UNWANTED_DIAGNOSTICS40
+SIMDE_BEGIN_DECLS_41
+42
+typedef union {43
+#if defined(SIMDE_VECTOR_SUBSCRIPT)44
+ SIMDE_ALIGN_TO_16 int8_t i8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;45
+ SIMDE_ALIGN_TO_16 int16_t i16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;46
+ SIMDE_ALIGN_TO_16 int32_t i32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;47
+ SIMDE_ALIGN_TO_16 int64_t i64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;48
+ SIMDE_ALIGN_TO_16 uint8_t u8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;49
+ SIMDE_ALIGN_TO_16 uint16_t u16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;50
+ SIMDE_ALIGN_TO_16 uint32_t u32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;51
+ SIMDE_ALIGN_TO_16 uint64_t u64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;52
+#if defined(SIMDE_HAVE_INT128_)53
+ SIMDE_ALIGN_TO_16 simde_int128 i128 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;54
+ SIMDE_ALIGN_TO_16 simde_uint128 u128 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;55
+#endif56
+ SIMDE_ALIGN_TO_16 simde_float32 f32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;57
+ SIMDE_ALIGN_TO_16 simde_float64 f64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;58
+59
+ SIMDE_ALIGN_TO_16 int_fast32_t i32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;60
+ SIMDE_ALIGN_TO_16 uint_fast32_t u32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;61
+#else62
+ SIMDE_ALIGN_TO_16 int8_t i8[16];63
+ SIMDE_ALIGN_TO_16 int16_t i16[8];64
+ SIMDE_ALIGN_TO_16 int32_t i32[4];65
+ SIMDE_ALIGN_TO_16 int64_t i64[2];66
+ SIMDE_ALIGN_TO_16 uint8_t u8[16];67
+ SIMDE_ALIGN_TO_16 uint16_t u16[8];68
+ SIMDE_ALIGN_TO_16 uint32_t u32[4];69
+ SIMDE_ALIGN_TO_16 uint64_t u64[2];70
+#if defined(SIMDE_HAVE_INT128_)71
+ SIMDE_ALIGN_TO_16 simde_int128 i128[1];72
+ SIMDE_ALIGN_TO_16 simde_uint128 u128[1];73
+#endif74
+ SIMDE_ALIGN_TO_16 simde_float32 f32[4];75
+ SIMDE_ALIGN_TO_16 simde_float64 f64[2];76
+77
+ SIMDE_ALIGN_TO_16 int_fast32_t i32f[16 / sizeof(int_fast32_t)];78
+ SIMDE_ALIGN_TO_16 uint_fast32_t u32f[16 / sizeof(uint_fast32_t)];79
+#endif80
+81
+ SIMDE_ALIGN_TO_16 simde__m64_private m64_private[2];82
+ SIMDE_ALIGN_TO_16 simde__m64 m64[2];83
+84
+#if defined(SIMDE_X86_SSE2_NATIVE)85
+ SIMDE_ALIGN_TO_16 __m128i n;86
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)87
+ SIMDE_ALIGN_TO_16 int8x16_t neon_i8;88
+ SIMDE_ALIGN_TO_16 int16x8_t neon_i16;89
+ SIMDE_ALIGN_TO_16 int32x4_t neon_i32;90
+ SIMDE_ALIGN_TO_16 int64x2_t neon_i64;91
+ SIMDE_ALIGN_TO_16 uint8x16_t neon_u8;92
+ SIMDE_ALIGN_TO_16 uint16x8_t neon_u16;93
+ SIMDE_ALIGN_TO_16 uint32x4_t neon_u32;94
+ SIMDE_ALIGN_TO_16 uint64x2_t neon_u64;95
+ SIMDE_ALIGN_TO_16 float32x4_t neon_f32;96
+#if defined(SIMDE_ARCH_AARCH64)97
+ SIMDE_ALIGN_TO_16 float64x2_t neon_f64;98
+#endif99
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)100
+ SIMDE_ALIGN_TO_16 v128_t wasm_v128;101
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)102
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed char) altivec_i8;103
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed short) altivec_i16;104
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32;105
+#if defined(__UINT_FAST32_TYPE__) && defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)106
+ SIMDE_ALIGN_TO_16107
+ SIMDE_POWER_ALTIVEC_VECTOR(__INT_FAST32_TYPE__) altivec_i32f;108
+#else109
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32f;110
+#endif111
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(unsigned char) altivec_u8;112
+ SIMDE_ALIGN_TO_16113
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned short) altivec_u16;114
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32;115
+#if defined(__UINT_FAST32_TYPE__) && defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)116
+ SIMDE_ALIGN_TO_16117
+ SIMDE_POWER_ALTIVEC_VECTOR(__UINT_FAST32_TYPE__) altivec_u32f;118
+#else119
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32f;120
+#endif121
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(float) altivec_f32;122
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)123
+ SIMDE_ALIGN_TO_16124
+ SIMDE_POWER_ALTIVEC_VECTOR(signed long long) altivec_i64;125
+ SIMDE_ALIGN_TO_16126
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long) altivec_u64;127
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(double) altivec_f64;128
+#endif129
+#endif130
+} simde__m128i_private;131
+132
+typedef union {133
+#if defined(SIMDE_VECTOR_SUBSCRIPT)134
+ SIMDE_ALIGN_TO_16 int8_t i8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;135
+ SIMDE_ALIGN_TO_16 int16_t i16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;136
+ SIMDE_ALIGN_TO_16 int32_t i32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;137
+ SIMDE_ALIGN_TO_16 int64_t i64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;138
+ SIMDE_ALIGN_TO_16 uint8_t u8 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;139
+ SIMDE_ALIGN_TO_16 uint16_t u16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;140
+ SIMDE_ALIGN_TO_16 uint32_t u32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;141
+ SIMDE_ALIGN_TO_16 uint64_t u64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;142
+ SIMDE_ALIGN_TO_16 simde_float32 f32 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;143
+ SIMDE_ALIGN_TO_16 simde_float64 f64 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;144
+ SIMDE_ALIGN_TO_16 int_fast32_t i32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;145
+ SIMDE_ALIGN_TO_16 uint_fast32_t u32f SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;146
+#else147
+ SIMDE_ALIGN_TO_16 int8_t i8[16];148
+ SIMDE_ALIGN_TO_16 int16_t i16[8];149
+ SIMDE_ALIGN_TO_16 int32_t i32[4];150
+ SIMDE_ALIGN_TO_16 int64_t i64[2];151
+ SIMDE_ALIGN_TO_16 uint8_t u8[16];152
+ SIMDE_ALIGN_TO_16 uint16_t u16[8];153
+ SIMDE_ALIGN_TO_16 uint32_t u32[4];154
+ SIMDE_ALIGN_TO_16 uint64_t u64[2];155
+ SIMDE_ALIGN_TO_16 simde_float32 f32[4];156
+ SIMDE_ALIGN_TO_16 simde_float64 f64[2];157
+ SIMDE_ALIGN_TO_16 int_fast32_t i32f[16 / sizeof(int_fast32_t)];158
+ SIMDE_ALIGN_TO_16 uint_fast32_t u32f[16 / sizeof(uint_fast32_t)];159
+#endif160
+161
+ SIMDE_ALIGN_TO_16 simde__m64_private m64_private[2];162
+ SIMDE_ALIGN_TO_16 simde__m64 m64[2];163
+164
+#if defined(SIMDE_X86_SSE2_NATIVE)165
+ SIMDE_ALIGN_TO_16 __m128d n;166
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)167
+ SIMDE_ALIGN_TO_16 int8x16_t neon_i8;168
+ SIMDE_ALIGN_TO_16 int16x8_t neon_i16;169
+ SIMDE_ALIGN_TO_16 int32x4_t neon_i32;170
+ SIMDE_ALIGN_TO_16 int64x2_t neon_i64;171
+ SIMDE_ALIGN_TO_16 uint8x16_t neon_u8;172
+ SIMDE_ALIGN_TO_16 uint16x8_t neon_u16;173
+ SIMDE_ALIGN_TO_16 uint32x4_t neon_u32;174
+ SIMDE_ALIGN_TO_16 uint64x2_t neon_u64;175
+ SIMDE_ALIGN_TO_16 float32x4_t neon_f32;176
+#if defined(SIMDE_ARCH_AARCH64)177
+ SIMDE_ALIGN_TO_16 float64x2_t neon_f64;178
+#endif179
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)180
+ SIMDE_ALIGN_TO_16 v128_t wasm_v128;181
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)182
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed char) altivec_i8;183
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed short) altivec_i16;184
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32;185
+#if defined(__INT_FAST32_TYPE__) && defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)186
+ SIMDE_ALIGN_TO_16187
+ SIMDE_POWER_ALTIVEC_VECTOR(__INT_FAST32_TYPE__) altivec_i32f;188
+#else189
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(signed int) altivec_i32f;190
+#endif191
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(unsigned char) altivec_u8;192
+ SIMDE_ALIGN_TO_16193
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned short) altivec_u16;194
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32;195
+#if defined(__UINT_FAST32_TYPE__) && defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)196
+ SIMDE_ALIGN_TO_16197
+ SIMDE_POWER_ALTIVEC_VECTOR(__UINT_FAST32_TYPE__) altivec_u32f;198
+#else199
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(unsigned int) altivec_u32f;200
+#endif201
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(float) altivec_f32;202
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)203
+ SIMDE_ALIGN_TO_16204
+ SIMDE_POWER_ALTIVEC_VECTOR(signed long long) altivec_i64;205
+ SIMDE_ALIGN_TO_16206
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long) altivec_u64;207
+ SIMDE_ALIGN_TO_16 SIMDE_POWER_ALTIVEC_VECTOR(double) altivec_f64;208
+#endif209
+#endif210
+} simde__m128d_private;211
+212
+#if defined(SIMDE_X86_SSE2_NATIVE)213
+typedef __m128i simde__m128i;214
+typedef __m128d simde__m128d;215
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)216
+typedef int64x2_t simde__m128i;217
+#if defined(SIMDE_ARCH_AARCH64)218
+typedef float64x2_t simde__m128d;219
+#elif defined(SIMDE_VECTOR_SUBSCRIPT)220
+typedef simde_float64 simde__m128d SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;221
+#else222
+typedef simde__m128d_private simde__m128d;223
+#endif224
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)225
+typedef v128_t simde__m128i;226
+typedef v128_t simde__m128d;227
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)228
+typedef SIMDE_POWER_ALTIVEC_VECTOR(float) simde__m128i;229
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)230
+typedef SIMDE_POWER_ALTIVEC_VECTOR(double) simde__m128d;231
+#else232
+typedef simde__m128d_private simde__m128d;233
+#endif234
+#elif defined(SIMDE_VECTOR_SUBSCRIPT)235
+typedef int64_t simde__m128i SIMDE_ALIGN_TO_16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;236
+typedef simde_float64237
+ simde__m128d SIMDE_ALIGN_TO_16 SIMDE_VECTOR(16) SIMDE_MAY_ALIAS;238
+#else239
+typedef simde__m128i_private simde__m128i;240
+typedef simde__m128d_private simde__m128d;241
+#endif242
+243
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)244
+typedef simde__m128i __m128i;245
+typedef simde__m128d __m128d;246
+#endif247
+248
+HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128i), "simde__m128i size incorrect");249
+HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128i_private),250
+ "simde__m128i_private size incorrect");251
+HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128d), "simde__m128d size incorrect");252
+HEDLEY_STATIC_ASSERT(16 == sizeof(simde__m128d_private),253
+ "simde__m128d_private size incorrect");254
+#if defined(SIMDE_CHECK_ALIGNMENT) && defined(SIMDE_ALIGN_OF)255
+HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128i) == 16,256
+ "simde__m128i is not 16-byte aligned");257
+HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128i_private) == 16,258
+ "simde__m128i_private is not 16-byte aligned");259
+HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128d) == 16,260
+ "simde__m128d is not 16-byte aligned");261
+HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m128d_private) == 16,262
+ "simde__m128d_private is not 16-byte aligned");263
+#endif264
+265
+SIMDE_FUNCTION_ATTRIBUTES266
+simde__m128i simde__m128i_from_private(simde__m128i_private v)267
+{268
+ simde__m128i r;269
+ simde_memcpy(&r, &v, sizeof(r));270
+ return r;271
+}272
+273
+SIMDE_FUNCTION_ATTRIBUTES274
+simde__m128i_private simde__m128i_to_private(simde__m128i v)275
+{276
+ simde__m128i_private r;277
+ simde_memcpy(&r, &v, sizeof(r));278
+ return r;279
+}280
+281
+SIMDE_FUNCTION_ATTRIBUTES282
+simde__m128d simde__m128d_from_private(simde__m128d_private v)283
+{284
+ simde__m128d r;285
+ simde_memcpy(&r, &v, sizeof(r));286
+ return r;287
+}288
+289
+SIMDE_FUNCTION_ATTRIBUTES290
+simde__m128d_private simde__m128d_to_private(simde__m128d v)291
+{292
+ simde__m128d_private r;293
+ simde_memcpy(&r, &v, sizeof(r));294
+ return r;295
+}296
+297
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)298
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, int8x16_t, neon, i8)299
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, int16x8_t, neon, i16)300
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, int32x4_t, neon, i32)301
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, int64x2_t, neon, i64)302
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, uint8x16_t, neon, u8)303
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, uint16x8_t, neon, u16)304
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, uint32x4_t, neon, u32)305
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, uint64x2_t, neon, u64)306
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, float32x4_t, neon, f32)307
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)308
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, float64x2_t, neon, f64)309
+#endif310
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)311
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i,312
+ SIMDE_POWER_ALTIVEC_VECTOR(signed char),313
+ altivec, i8)314
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i,315
+ SIMDE_POWER_ALTIVEC_VECTOR(signed short),316
+ altivec, i16)317
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i,318
+ SIMDE_POWER_ALTIVEC_VECTOR(signed int),319
+ altivec, i32)320
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(321
+ m128i, SIMDE_POWER_ALTIVEC_VECTOR(unsigned char), altivec, u8)322
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(323
+ m128i, SIMDE_POWER_ALTIVEC_VECTOR(unsigned short), altivec, u16)324
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i,325
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned int),326
+ altivec, u32)327
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)328
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(329
+ m128i, SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long), altivec, u64)330
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(331
+ m128i, SIMDE_POWER_ALTIVEC_VECTOR(signed long long), altivec, i64)332
+#endif333
+#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */334
+335
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)336
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, int8x16_t, neon, i8)337
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, int16x8_t, neon, i16)338
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, int32x4_t, neon, i32)339
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, int64x2_t, neon, i64)340
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, uint8x16_t, neon, u8)341
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, uint16x8_t, neon, u16)342
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, uint32x4_t, neon, u32)343
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, uint64x2_t, neon, u64)344
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, float32x4_t, neon, f32)345
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)346
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, float64x2_t, neon, f64)347
+#endif348
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)349
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d,350
+ SIMDE_POWER_ALTIVEC_VECTOR(signed char),351
+ altivec, i8)352
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d,353
+ SIMDE_POWER_ALTIVEC_VECTOR(signed short),354
+ altivec, i16)355
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d,356
+ SIMDE_POWER_ALTIVEC_VECTOR(signed int),357
+ altivec, i32)358
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(359
+ m128d, SIMDE_POWER_ALTIVEC_VECTOR(unsigned char), altivec, u8)360
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(361
+ m128d, SIMDE_POWER_ALTIVEC_VECTOR(unsigned short), altivec, u16)362
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d,363
+ SIMDE_POWER_ALTIVEC_VECTOR(unsigned int),364
+ altivec, u32)365
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)366
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(367
+ m128d, SIMDE_POWER_ALTIVEC_VECTOR(unsigned long long), altivec, u64)368
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(369
+ m128d, SIMDE_POWER_ALTIVEC_VECTOR(signed long long), altivec, i64)370
+#if defined(SIMDE_BUG_GCC_95782)371
+SIMDE_FUNCTION_ATTRIBUTES372
+SIMDE_POWER_ALTIVEC_VECTOR(double)373
+simde__m128d_to_altivec_f64(simde__m128d value)374
+{375
+ simde__m128d_private r_ = simde__m128d_to_private(value);376
+ return r_.altivec_f64;377
+}378
+379
+SIMDE_FUNCTION_ATTRIBUTES380
+simde__m128d simde__m128d_from_altivec_f64(SIMDE_POWER_ALTIVEC_VECTOR(double)381
+ value)382
+{383
+ simde__m128d_private r_;384
+ r_.altivec_f64 = value;385
+ return simde__m128d_from_private(r_);386
+}387
+#else388
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d,389
+ SIMDE_POWER_ALTIVEC_VECTOR(double),390
+ altivec, f64)391
+#endif392
+#endif393
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)394
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128d, v128_t, wasm, v128);395
+SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m128i, v128_t, wasm, v128);396
+#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */397
+398
+SIMDE_FUNCTION_ATTRIBUTES399
+simde__m128d simde_mm_set_pd(simde_float64 e1, simde_float64 e0)400
+{401
+#if defined(SIMDE_X86_SSE2_NATIVE)402
+ return _mm_set_pd(e1, e0);403
+#else404
+ simde__m128d_private r_;405
+406
+#if defined(SIMDE_WASM_SIMD128_NATIVE)407
+ r_.wasm_v128 = wasm_f64x2_make(e0, e1);408
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)409
+ SIMDE_ALIGN_TO_16 simde_float64 data[2] = {e0, e1};410
+ r_.neon_f64 = vld1q_f64(data);411
+#else412
+ r_.f64[0] = e0;413
+ r_.f64[1] = e1;414
+#endif415
+416
+ return simde__m128d_from_private(r_);417
+#endif418
+}419
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)420
+#define _mm_set_pd(e1, e0) simde_mm_set_pd(e1, e0)421
+#endif422
+423
+SIMDE_FUNCTION_ATTRIBUTES424
+simde__m128d simde_mm_set1_pd(simde_float64 a)425
+{426
+#if defined(SIMDE_X86_SSE2_NATIVE)427
+ return _mm_set1_pd(a);428
+#else429
+ simde__m128d_private r_;430
+431
+#if defined(SIMDE_WASM_SIMD128_NATIVE)432
+ r_.wasm_v128 = wasm_f64x2_splat(a);433
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)434
+ r_.neon_f64 = vdupq_n_f64(a);435
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)436
+ r_.altivec_f64 = vec_splats(HEDLEY_STATIC_CAST(double, a));437
+#else438
+ SIMDE_VECTORIZE439
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {440
+ r_.f64[i] = a;441
+ }442
+#endif443
+444
+ return simde__m128d_from_private(r_);445
+#endif446
+}447
+#define simde_mm_set_pd1(a) simde_mm_set1_pd(a)448
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)449
+#define _mm_set1_pd(a) simde_mm_set1_pd(a)450
+#define _mm_set_pd1(a) simde_mm_set1_pd(a)451
+#endif452
+453
+SIMDE_FUNCTION_ATTRIBUTES454
+simde__m128d simde_x_mm_abs_pd(simde__m128d a)455
+{456
+#if defined(SIMDE_X86_AVX512F_NATIVE) && \457
+ (!defined(HEDLEY_GCC_VERSION) || HEDLEY_GCC_VERSION_CHECK(7, 4, 0))458
+ return _mm512_castpd512_pd128(_mm512_abs_pd(_mm512_castpd128_pd512(a)));459
+#else460
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a);461
+462
+#if defined(SIMDE_ARM_NEON_A32V8_NATIVE)463
+ r_.neon_f32 = vabsq_f32(a_.neon_f32);464
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)465
+ r_.altivec_f32 = vec_abs(a_.altivec_f32);466
+#else467
+ SIMDE_VECTORIZE468
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {469
+ r_.f64[i] = simde_math_fabs(a_.f64[i]);470
+ }471
+#endif472
+473
+ return simde__m128d_from_private(r_);474
+#endif475
+}476
+477
+SIMDE_FUNCTION_ATTRIBUTES478
+simde__m128d simde_x_mm_not_pd(simde__m128d a)479
+{480
+#if defined(SIMDE_X86_AVX512VL_NATIVE)481
+ __m128i ai = _mm_castpd_si128(a);482
+ return _mm_castsi128_pd(_mm_ternarylogic_epi64(ai, ai, ai, 0x55));483
+#else484
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a);485
+486
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)487
+ r_.neon_i32 = vmvnq_s32(a_.neon_i32);488
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)489
+ r_.altivec_f64 = vec_nor(a_.altivec_f64, a_.altivec_f64);490
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)491
+ r_.altivec_i32 = vec_nor(a_.altivec_i32, a_.altivec_i32);492
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)493
+ r_.wasm_v128 = wasm_v128_not(a_.wasm_v128);494
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)495
+ r_.i32f = ~a_.i32f;496
+#else497
+ SIMDE_VECTORIZE498
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {499
+ r_.i32f[i] = ~(a_.i32f[i]);500
+ }501
+#endif502
+503
+ return simde__m128d_from_private(r_);504
+#endif505
+}506
+507
+SIMDE_FUNCTION_ATTRIBUTES508
+simde__m128d simde_x_mm_select_pd(simde__m128d a, simde__m128d b,509
+ simde__m128d mask)510
+{511
+/* This function is for when you want to blend two elements together512
+ * according to a mask. It is similar to _mm_blendv_pd, except that513
+ * it is undefined whether the blend is based on the highest bit in514
+ * each lane (like blendv) or just bitwise operations. This allows515
+ * us to implement the function efficiently everywhere.516
+ *517
+ * Basically, you promise that all the lanes in mask are either 0 or518
+ * ~0. */519
+#if defined(SIMDE_X86_SSE4_1_NATIVE)520
+ return _mm_blendv_pd(a, b, mask);521
+#else522
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),523
+ b_ = simde__m128d_to_private(b),524
+ mask_ = simde__m128d_to_private(mask);525
+526
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)527
+ r_.i64 = a_.i64 ^ ((a_.i64 ^ b_.i64) & mask_.i64);528
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)529
+ r_.neon_i64 = vbslq_s64(mask_.neon_u64, b_.neon_i64, a_.neon_i64);530
+#else531
+ SIMDE_VECTORIZE532
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {533
+ r_.i64[i] = a_.i64[i] ^534
+ ((a_.i64[i] ^ b_.i64[i]) & mask_.i64[i]);535
+ }536
+#endif537
+538
+ return simde__m128d_from_private(r_);539
+#endif540
+}541
+542
+SIMDE_FUNCTION_ATTRIBUTES543
+simde__m128i simde_mm_add_epi8(simde__m128i a, simde__m128i b)544
+{545
+#if defined(SIMDE_X86_SSE2_NATIVE)546
+ return _mm_add_epi8(a, b);547
+#else548
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),549
+ b_ = simde__m128i_to_private(b);550
+551
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)552
+ r_.neon_i8 = vaddq_s8(a_.neon_i8, b_.neon_i8);553
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)554
+ r_.altivec_i8 = vec_add(a_.altivec_i8, b_.altivec_i8);555
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)556
+ r_.wasm_v128 = wasm_i8x16_add(a_.wasm_v128, b_.wasm_v128);557
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)558
+ r_.i8 = a_.i8 + b_.i8;559
+#else560
+ SIMDE_VECTORIZE561
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {562
+ r_.i8[i] = a_.i8[i] + b_.i8[i];563
+ }564
+#endif565
+566
+ return simde__m128i_from_private(r_);567
+#endif568
+}569
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)570
+#define _mm_add_epi8(a, b) simde_mm_add_epi8(a, b)571
+#endif572
+573
+SIMDE_FUNCTION_ATTRIBUTES574
+simde__m128i simde_mm_add_epi16(simde__m128i a, simde__m128i b)575
+{576
+#if defined(SIMDE_X86_SSE2_NATIVE)577
+ return _mm_add_epi16(a, b);578
+#else579
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),580
+ b_ = simde__m128i_to_private(b);581
+582
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)583
+ r_.neon_i16 = vaddq_s16(a_.neon_i16, b_.neon_i16);584
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)585
+ r_.altivec_i16 = vec_add(a_.altivec_i16, b_.altivec_i16);586
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)587
+ r_.wasm_v128 = wasm_i16x8_add(a_.wasm_v128, b_.wasm_v128);588
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)589
+ r_.i16 = a_.i16 + b_.i16;590
+#else591
+ SIMDE_VECTORIZE592
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {593
+ r_.i16[i] = a_.i16[i] + b_.i16[i];594
+ }595
+#endif596
+597
+ return simde__m128i_from_private(r_);598
+#endif599
+}600
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)601
+#define _mm_add_epi16(a, b) simde_mm_add_epi16(a, b)602
+#endif603
+604
+SIMDE_FUNCTION_ATTRIBUTES605
+simde__m128i simde_mm_add_epi32(simde__m128i a, simde__m128i b)606
+{607
+#if defined(SIMDE_X86_SSE2_NATIVE)608
+ return _mm_add_epi32(a, b);609
+#else610
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),611
+ b_ = simde__m128i_to_private(b);612
+613
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)614
+ r_.neon_i32 = vaddq_s32(a_.neon_i32, b_.neon_i32);615
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)616
+ r_.altivec_i32 = vec_add(a_.altivec_i32, b_.altivec_i32);617
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)618
+ r_.wasm_v128 = wasm_i32x4_add(a_.wasm_v128, b_.wasm_v128);619
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)620
+ r_.i32 = a_.i32 + b_.i32;621
+#else622
+ SIMDE_VECTORIZE623
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {624
+ r_.i32[i] = a_.i32[i] + b_.i32[i];625
+ }626
+#endif627
+628
+ return simde__m128i_from_private(r_);629
+#endif630
+}631
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)632
+#define _mm_add_epi32(a, b) simde_mm_add_epi32(a, b)633
+#endif634
+635
+SIMDE_FUNCTION_ATTRIBUTES636
+simde__m128i simde_mm_add_epi64(simde__m128i a, simde__m128i b)637
+{638
+#if defined(SIMDE_X86_SSE2_NATIVE)639
+ return _mm_add_epi64(a, b);640
+#else641
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),642
+ b_ = simde__m128i_to_private(b);643
+644
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)645
+ r_.neon_i64 = vaddq_s64(a_.neon_i64, b_.neon_i64);646
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)647
+ r_.altivec_i64 = vec_add(a_.altivec_i64, b_.altivec_i64);648
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)649
+ r_.wasm_v128 = wasm_i64x2_add(a_.wasm_v128, b_.wasm_v128);650
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)651
+ r_.i64 = a_.i64 + b_.i64;652
+#else653
+ SIMDE_VECTORIZE654
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {655
+ r_.i64[i] = a_.i64[i] + b_.i64[i];656
+ }657
+#endif658
+659
+ return simde__m128i_from_private(r_);660
+#endif661
+}662
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)663
+#define _mm_add_epi64(a, b) simde_mm_add_epi64(a, b)664
+#endif665
+666
+SIMDE_FUNCTION_ATTRIBUTES667
+simde__m128d simde_mm_add_pd(simde__m128d a, simde__m128d b)668
+{669
+#if defined(SIMDE_X86_SSE2_NATIVE)670
+ return _mm_add_pd(a, b);671
+#else672
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),673
+ b_ = simde__m128d_to_private(b);674
+675
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)676
+ r_.neon_f64 = vaddq_f64(a_.neon_f64, b_.neon_f64);677
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)678
+ r_.wasm_v128 = wasm_f64x2_add(a_.wasm_v128, b_.wasm_v128);679
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)680
+ r_.altivec_f64 = vec_add(a_.altivec_f64, b_.altivec_f64);681
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)682
+ r_.wasm_v128 = wasm_f64x2_add(a_.wasm_v128, b_.wasm_v128);683
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)684
+ r_.f64 = a_.f64 + b_.f64;685
+#else686
+ SIMDE_VECTORIZE687
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {688
+ r_.f64[i] = a_.f64[i] + b_.f64[i];689
+ }690
+#endif691
+692
+ return simde__m128d_from_private(r_);693
+#endif694
+}695
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)696
+#define _mm_add_pd(a, b) simde_mm_add_pd(a, b)697
+#endif698
+699
+SIMDE_FUNCTION_ATTRIBUTES700
+simde__m128d simde_mm_move_sd(simde__m128d a, simde__m128d b)701
+{702
+#if defined(SIMDE_X86_SSE2_NATIVE)703
+ return _mm_move_sd(a, b);704
+#else705
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),706
+ b_ = simde__m128d_to_private(b);707
+708
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)709
+ r_.neon_f64 =710
+ vsetq_lane_f64(vgetq_lane_f64(b_.neon_f64, 0), a_.neon_f64, 0);711
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)712
+#if defined(HEDLEY_IBM_VERSION)713
+ r_.altivec_f64 = vec_xxpermdi(a_.altivec_f64, b_.altivec_f64, 1);714
+#else715
+ r_.altivec_f64 = vec_xxpermdi(b_.altivec_f64, a_.altivec_f64, 1);716
+#endif717
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)718
+ r_.wasm_v128 = wasm_v64x2_shuffle(a_.wasm_v128, b_.wasm_v128, 2, 1);719
+#elif defined(SIMDE_SHUFFLE_VECTOR_)720
+ r_.f64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.f64, b_.f64, 2, 1);721
+#else722
+ r_.f64[0] = b_.f64[0];723
+ r_.f64[1] = a_.f64[1];724
+#endif725
+726
+ return simde__m128d_from_private(r_);727
+#endif728
+}729
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)730
+#define _mm_move_sd(a, b) simde_mm_move_sd(a, b)731
+#endif732
+733
+SIMDE_FUNCTION_ATTRIBUTES734
+simde__m128d simde_mm_add_sd(simde__m128d a, simde__m128d b)735
+{736
+#if defined(SIMDE_X86_SSE2_NATIVE)737
+ return _mm_add_sd(a, b);738
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)739
+ return simde_mm_move_sd(a, simde_mm_add_pd(a, b));740
+#else741
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),742
+ b_ = simde__m128d_to_private(b);743
+744
+ r_.f64[0] = a_.f64[0] + b_.f64[0];745
+ r_.f64[1] = a_.f64[1];746
+747
+ return simde__m128d_from_private(r_);748
+#endif749
+}750
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)751
+#define _mm_add_sd(a, b) simde_mm_add_sd(a, b)752
+#endif753
+754
+SIMDE_FUNCTION_ATTRIBUTES755
+simde__m64 simde_mm_add_si64(simde__m64 a, simde__m64 b)756
+{757
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)758
+ return _mm_add_si64(a, b);759
+#else760
+ simde__m64_private r_, a_ = simde__m64_to_private(a),761
+ b_ = simde__m64_to_private(b);762
+763
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)764
+ r_.neon_i64 = vadd_s64(a_.neon_i64, b_.neon_i64);765
+#else766
+ r_.i64[0] = a_.i64[0] + b_.i64[0];767
+#endif768
+769
+ return simde__m64_from_private(r_);770
+#endif771
+}772
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)773
+#define _mm_add_si64(a, b) simde_mm_add_si64(a, b)774
+#endif775
+776
+SIMDE_FUNCTION_ATTRIBUTES777
+simde__m128i simde_mm_adds_epi8(simde__m128i a, simde__m128i b)778
+{779
+#if defined(SIMDE_X86_SSE2_NATIVE)780
+ return _mm_adds_epi8(a, b);781
+#else782
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),783
+ b_ = simde__m128i_to_private(b);784
+785
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)786
+ r_.neon_i8 = vqaddq_s8(a_.neon_i8, b_.neon_i8);787
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)788
+ r_.wasm_v128 = wasm_i8x16_add_saturate(a_.wasm_v128, b_.wasm_v128);789
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)790
+ r_.altivec_i8 = vec_adds(a_.altivec_i8, b_.altivec_i8);791
+#else792
+ SIMDE_VECTORIZE793
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {794
+ const int_fast16_t tmp =795
+ HEDLEY_STATIC_CAST(int_fast16_t, a_.i8[i]) +796
+ HEDLEY_STATIC_CAST(int_fast16_t, b_.i8[i]);797
+ r_.i8[i] = HEDLEY_STATIC_CAST(798
+ int8_t,799
+ ((tmp < INT8_MAX) ? ((tmp > INT8_MIN) ? tmp : INT8_MIN)800
+ : INT8_MAX));801
+ }802
+#endif803
+804
+ return simde__m128i_from_private(r_);805
+#endif806
+}807
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)808
+#define _mm_adds_epi8(a, b) simde_mm_adds_epi8(a, b)809
+#endif810
+811
+SIMDE_FUNCTION_ATTRIBUTES812
+simde__m128i simde_mm_adds_epi16(simde__m128i a, simde__m128i b)813
+{814
+#if defined(SIMDE_X86_SSE2_NATIVE)815
+ return _mm_adds_epi16(a, b);816
+#else817
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),818
+ b_ = simde__m128i_to_private(b);819
+820
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)821
+ r_.neon_i16 = vqaddq_s16(a_.neon_i16, b_.neon_i16);822
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)823
+ r_.wasm_v128 = wasm_i16x8_add_saturate(a_.wasm_v128, b_.wasm_v128);824
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)825
+ r_.altivec_i16 = vec_adds(a_.altivec_i16, b_.altivec_i16);826
+#else827
+ SIMDE_VECTORIZE828
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {829
+ const int_fast32_t tmp =830
+ HEDLEY_STATIC_CAST(int_fast32_t, a_.i16[i]) +831
+ HEDLEY_STATIC_CAST(int_fast32_t, b_.i16[i]);832
+ r_.i16[i] = HEDLEY_STATIC_CAST(833
+ int16_t,834
+ ((tmp < INT16_MAX)835
+ ? ((tmp > INT16_MIN) ? tmp : INT16_MIN)836
+ : INT16_MAX));837
+ }838
+#endif839
+840
+ return simde__m128i_from_private(r_);841
+#endif842
+}843
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)844
+#define _mm_adds_epi16(a, b) simde_mm_adds_epi16(a, b)845
+#endif846
+847
+SIMDE_FUNCTION_ATTRIBUTES848
+simde__m128i simde_mm_adds_epu8(simde__m128i a, simde__m128i b)849
+{850
+#if defined(SIMDE_X86_SSE2_NATIVE)851
+ return _mm_adds_epu8(a, b);852
+#else853
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),854
+ b_ = simde__m128i_to_private(b);855
+856
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)857
+ r_.neon_u8 = vqaddq_u8(a_.neon_u8, b_.neon_u8);858
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)859
+ r_.wasm_v128 = wasm_u8x16_add_saturate(a_.wasm_v128, b_.wasm_v128);860
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)861
+ r_.altivec_u8 = vec_adds(a_.altivec_u8, b_.altivec_u8);862
+#else863
+ SIMDE_VECTORIZE864
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {865
+ r_.u8[i] = ((UINT8_MAX - a_.u8[i]) > b_.u8[i])866
+ ? (a_.u8[i] + b_.u8[i])867
+ : UINT8_MAX;868
+ }869
+#endif870
+871
+ return simde__m128i_from_private(r_);872
+#endif873
+}874
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)875
+#define _mm_adds_epu8(a, b) simde_mm_adds_epu8(a, b)876
+#endif877
+878
+SIMDE_FUNCTION_ATTRIBUTES879
+simde__m128i simde_mm_adds_epu16(simde__m128i a, simde__m128i b)880
+{881
+#if defined(SIMDE_X86_SSE2_NATIVE)882
+ return _mm_adds_epu16(a, b);883
+#else884
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),885
+ b_ = simde__m128i_to_private(b);886
+887
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)888
+ r_.neon_u16 = vqaddq_u16(a_.neon_u16, b_.neon_u16);889
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)890
+ r_.wasm_v128 = wasm_u16x8_add_saturate(a_.wasm_v128, b_.wasm_v128);891
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)892
+ r_.altivec_u16 = vec_adds(a_.altivec_u16, b_.altivec_u16);893
+#else894
+ SIMDE_VECTORIZE895
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {896
+ r_.u16[i] = ((UINT16_MAX - a_.u16[i]) > b_.u16[i])897
+ ? (a_.u16[i] + b_.u16[i])898
+ : UINT16_MAX;899
+ }900
+#endif901
+902
+ return simde__m128i_from_private(r_);903
+#endif904
+}905
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)906
+#define _mm_adds_epu16(a, b) simde_mm_adds_epu16(a, b)907
+#endif908
+909
+SIMDE_FUNCTION_ATTRIBUTES910
+simde__m128d simde_mm_and_pd(simde__m128d a, simde__m128d b)911
+{912
+#if defined(SIMDE_X86_SSE2_NATIVE)913
+ return _mm_and_pd(a, b);914
+#else915
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),916
+ b_ = simde__m128d_to_private(b);917
+918
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)919
+ r_.neon_i32 = vandq_s32(a_.neon_i32, b_.neon_i32);920
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)921
+ r_.wasm_v128 = wasm_v128_and(a_.wasm_v128, b_.wasm_v128);922
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)923
+ r_.altivec_f64 = vec_and(a_.altivec_f64, b_.altivec_f64);924
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)925
+ r_.i32f = a_.i32f & b_.i32f;926
+#else927
+ SIMDE_VECTORIZE928
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {929
+ r_.i32f[i] = a_.i32f[i] & b_.i32f[i];930
+ }931
+#endif932
+933
+ return simde__m128d_from_private(r_);934
+#endif935
+}936
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)937
+#define _mm_and_pd(a, b) simde_mm_and_pd(a, b)938
+#endif939
+940
+SIMDE_FUNCTION_ATTRIBUTES941
+simde__m128i simde_mm_and_si128(simde__m128i a, simde__m128i b)942
+{943
+#if defined(SIMDE_X86_SSE2_NATIVE)944
+ return _mm_and_si128(a, b);945
+#else946
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),947
+ b_ = simde__m128i_to_private(b);948
+949
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)950
+ r_.neon_i32 = vandq_s32(b_.neon_i32, a_.neon_i32);951
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)952
+ r_.altivec_u32f = vec_and(a_.altivec_u32f, b_.altivec_u32f);953
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)954
+ r_.i32f = a_.i32f & b_.i32f;955
+#else956
+ SIMDE_VECTORIZE957
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {958
+ r_.i32f[i] = a_.i32f[i] & b_.i32f[i];959
+ }960
+#endif961
+962
+ return simde__m128i_from_private(r_);963
+#endif964
+}965
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)966
+#define _mm_and_si128(a, b) simde_mm_and_si128(a, b)967
+#endif968
+969
+SIMDE_FUNCTION_ATTRIBUTES970
+simde__m128d simde_mm_andnot_pd(simde__m128d a, simde__m128d b)971
+{972
+#if defined(SIMDE_X86_SSE2_NATIVE)973
+ return _mm_andnot_pd(a, b);974
+#else975
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),976
+ b_ = simde__m128d_to_private(b);977
+978
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)979
+ r_.neon_i32 = vbicq_s32(b_.neon_i32, a_.neon_i32);980
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)981
+ r_.wasm_v128 = wasm_v128_andnot(b_.wasm_v128, a_.wasm_v128);982
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)983
+ r_.altivec_f64 = vec_andc(b_.altivec_f64, a_.altivec_f64);984
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)985
+ r_.altivec_i32f = vec_andc(b_.altivec_i32f, a_.altivec_i32f);986
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)987
+ r_.i32f = ~a_.i32f & b_.i32f;988
+#else989
+ SIMDE_VECTORIZE990
+ for (size_t i = 0; i < (sizeof(r_.u64) / sizeof(r_.u64[0])); i++) {991
+ r_.u64[i] = ~a_.u64[i] & b_.u64[i];992
+ }993
+#endif994
+995
+ return simde__m128d_from_private(r_);996
+#endif997
+}998
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)999
+#define _mm_andnot_pd(a, b) simde_mm_andnot_pd(a, b)1000
+#endif1001
+1002
+SIMDE_FUNCTION_ATTRIBUTES1003
+simde__m128i simde_mm_andnot_si128(simde__m128i a, simde__m128i b)1004
+{1005
+#if defined(SIMDE_X86_SSE2_NATIVE)1006
+ return _mm_andnot_si128(a, b);1007
+#else1008
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1009
+ b_ = simde__m128i_to_private(b);1010
+1011
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1012
+ r_.neon_i32 = vbicq_s32(b_.neon_i32, a_.neon_i32);1013
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1014
+ r_.altivec_i32 = vec_andc(b_.altivec_i32, a_.altivec_i32);1015
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1016
+ r_.i32f = ~a_.i32f & b_.i32f;1017
+#else1018
+ SIMDE_VECTORIZE1019
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {1020
+ r_.i32f[i] = ~(a_.i32f[i]) & b_.i32f[i];1021
+ }1022
+#endif1023
+1024
+ return simde__m128i_from_private(r_);1025
+#endif1026
+}1027
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1028
+#define _mm_andnot_si128(a, b) simde_mm_andnot_si128(a, b)1029
+#endif1030
+1031
+SIMDE_FUNCTION_ATTRIBUTES1032
+simde__m128d simde_mm_xor_pd(simde__m128d a, simde__m128d b)1033
+{1034
+#if defined(SIMDE_X86_SSE2_NATIVE)1035
+ return _mm_xor_pd(a, b);1036
+#else1037
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),1038
+ b_ = simde__m128d_to_private(b);1039
+1040
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1041
+ r_.i32f = a_.i32f ^ b_.i32f;1042
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1043
+ r_.wasm_v128 = wasm_v128_xor(a_.wasm_v128, b_.wasm_v128);1044
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1045
+ r_.neon_i64 = veorq_s64(a_.neon_i64, b_.neon_i64);1046
+#else1047
+ SIMDE_VECTORIZE1048
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {1049
+ r_.i32f[i] = a_.i32f[i] ^ b_.i32f[i];1050
+ }1051
+#endif1052
+1053
+ return simde__m128d_from_private(r_);1054
+#endif1055
+}1056
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1057
+#define _mm_xor_pd(a, b) simde_mm_xor_pd(a, b)1058
+#endif1059
+1060
+SIMDE_FUNCTION_ATTRIBUTES1061
+simde__m128i simde_mm_avg_epu8(simde__m128i a, simde__m128i b)1062
+{1063
+#if defined(SIMDE_X86_SSE2_NATIVE)1064
+ return _mm_avg_epu8(a, b);1065
+#else1066
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1067
+ b_ = simde__m128i_to_private(b);1068
+1069
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1070
+ r_.neon_u8 = vrhaddq_u8(b_.neon_u8, a_.neon_u8);1071
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1072
+ r_.wasm_v128 = wasm_u8x16_avgr(a_.wasm_v128, b_.wasm_v128);1073
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1074
+ r_.altivec_u8 = vec_avg(a_.altivec_u8, b_.altivec_u8);1075
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS) && \1076
+ defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \1077
+ defined(SIMDE_CONVERT_VECTOR_)1078
+ uint16_t wa SIMDE_VECTOR(32);1079
+ uint16_t wb SIMDE_VECTOR(32);1080
+ uint16_t wr SIMDE_VECTOR(32);1081
+ SIMDE_CONVERT_VECTOR_(wa, a_.u8);1082
+ SIMDE_CONVERT_VECTOR_(wb, b_.u8);1083
+ wr = (wa + wb + 1) >> 1;1084
+ SIMDE_CONVERT_VECTOR_(r_.u8, wr);1085
+#else1086
+ SIMDE_VECTORIZE1087
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {1088
+ r_.u8[i] = (a_.u8[i] + b_.u8[i] + 1) >> 1;1089
+ }1090
+#endif1091
+1092
+ return simde__m128i_from_private(r_);1093
+#endif1094
+}1095
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1096
+#define _mm_avg_epu8(a, b) simde_mm_avg_epu8(a, b)1097
+#endif1098
+1099
+SIMDE_FUNCTION_ATTRIBUTES1100
+simde__m128i simde_mm_avg_epu16(simde__m128i a, simde__m128i b)1101
+{1102
+#if defined(SIMDE_X86_SSE2_NATIVE)1103
+ return _mm_avg_epu16(a, b);1104
+#else1105
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1106
+ b_ = simde__m128i_to_private(b);1107
+1108
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1109
+ r_.neon_u16 = vrhaddq_u16(b_.neon_u16, a_.neon_u16);1110
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1111
+ r_.wasm_v128 = wasm_u16x8_avgr(a_.wasm_v128, b_.wasm_v128);1112
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1113
+ r_.altivec_u16 = vec_avg(a_.altivec_u16, b_.altivec_u16);1114
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS) && \1115
+ defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && \1116
+ defined(SIMDE_CONVERT_VECTOR_)1117
+ uint32_t wa SIMDE_VECTOR(32);1118
+ uint32_t wb SIMDE_VECTOR(32);1119
+ uint32_t wr SIMDE_VECTOR(32);1120
+ SIMDE_CONVERT_VECTOR_(wa, a_.u16);1121
+ SIMDE_CONVERT_VECTOR_(wb, b_.u16);1122
+ wr = (wa + wb + 1) >> 1;1123
+ SIMDE_CONVERT_VECTOR_(r_.u16, wr);1124
+#else1125
+ SIMDE_VECTORIZE1126
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {1127
+ r_.u16[i] = (a_.u16[i] + b_.u16[i] + 1) >> 1;1128
+ }1129
+#endif1130
+1131
+ return simde__m128i_from_private(r_);1132
+#endif1133
+}1134
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1135
+#define _mm_avg_epu16(a, b) simde_mm_avg_epu16(a, b)1136
+#endif1137
+1138
+SIMDE_FUNCTION_ATTRIBUTES1139
+simde__m128i simde_mm_setzero_si128(void)1140
+{1141
+#if defined(SIMDE_X86_SSE2_NATIVE)1142
+ return _mm_setzero_si128();1143
+#else1144
+ simde__m128i_private r_;1145
+1146
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1147
+ r_.neon_i32 = vdupq_n_s32(0);1148
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1149
+ r_.altivec_i32 = vec_splats(HEDLEY_STATIC_CAST(signed int, 0));1150
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1151
+ r_.wasm_v128 = wasm_i32x4_splat(INT32_C(0));1152
+#elif defined(SIMDE_VECTOR_SUBSCRIPT)1153
+ r_.i32 = __extension__(__typeof__(r_.i32)){0, 0, 0, 0};1154
+#else1155
+ SIMDE_VECTORIZE1156
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {1157
+ r_.i32f[i] = 0;1158
+ }1159
+#endif1160
+1161
+ return simde__m128i_from_private(r_);1162
+#endif1163
+}1164
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1165
+#define _mm_setzero_si128() (simde_mm_setzero_si128())1166
+#endif1167
+1168
+SIMDE_FUNCTION_ATTRIBUTES1169
+simde__m128i simde_mm_bslli_si128(simde__m128i a, const int imm8)1170
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)1171
+{1172
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);1173
+1174
+ if (HEDLEY_UNLIKELY((imm8 & ~15))) {1175
+ return simde_mm_setzero_si128();1176
+ }1177
+1178
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE) && defined(SIMDE_ENDIAN_ORDER)1179
+ r_.altivec_i8 =1180
+#if (SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_LITTLE)1181
+ vec_slo1182
+#else /* SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_BIG */1183
+ vec_sro1184
+#endif1185
+ (a_.altivec_i8,1186
+ vec_splats(HEDLEY_STATIC_CAST(unsigned char, imm8 * 8)));1187
+#elif defined(SIMDE_HAVE_INT128_) && (SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_LITTLE)1188
+ r_.u128[0] = a_.u128[0] << (imm8 * 8);1189
+#else1190
+ r_ = simde__m128i_to_private(simde_mm_setzero_si128());1191
+ for (int i = imm8;1192
+ i < HEDLEY_STATIC_CAST(int, sizeof(r_.i8) / sizeof(r_.i8[0]));1193
+ i++) {1194
+ r_.i8[i] = a_.i8[i - imm8];1195
+ }1196
+#endif1197
+1198
+ return simde__m128i_from_private(r_);1199
+}1200
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)1201
+#define simde_mm_bslli_si128(a, imm8) _mm_slli_si128(a, imm8)1202
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) && !defined(__clang__)1203
+#define simde_mm_bslli_si128(a, imm8) \1204
+ simde__m128i_from_neon_i8( \1205
+ ((imm8) <= 0) \1206
+ ? simde__m128i_to_neon_i8(a) \1207
+ : (((imm8) > 15) \1208
+ ? (vdupq_n_s8(0)) \1209
+ : (vextq_s8(vdupq_n_s8(0), \1210
+ simde__m128i_to_neon_i8(a), \1211
+ 16 - (imm8)))))1212
+#elif defined(SIMDE_SHUFFLE_VECTOR_) && !defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1213
+#define simde_mm_bslli_si128(a, imm8) \1214
+ (__extension__({ \1215
+ const simde__m128i_private simde__tmp_a_ = \1216
+ simde__m128i_to_private(a); \1217
+ const simde__m128i_private simde__tmp_z_ = \1218
+ simde__m128i_to_private(simde_mm_setzero_si128()); \1219
+ simde__m128i_private simde__tmp_r_; \1220
+ if (HEDLEY_UNLIKELY(imm8 > 15)) { \1221
+ simde__tmp_r_ = simde__m128i_to_private( \1222
+ simde_mm_setzero_si128()); \1223
+ } else { \1224
+ simde__tmp_r_.i8 = SIMDE_SHUFFLE_VECTOR_( \1225
+ 8, 16, simde__tmp_z_.i8, (simde__tmp_a_).i8, \1226
+ HEDLEY_STATIC_CAST(int8_t, (16 - imm8) & 31), \1227
+ HEDLEY_STATIC_CAST(int8_t, (17 - imm8) & 31), \1228
+ HEDLEY_STATIC_CAST(int8_t, (18 - imm8) & 31), \1229
+ HEDLEY_STATIC_CAST(int8_t, (19 - imm8) & 31), \1230
+ HEDLEY_STATIC_CAST(int8_t, (20 - imm8) & 31), \1231
+ HEDLEY_STATIC_CAST(int8_t, (21 - imm8) & 31), \1232
+ HEDLEY_STATIC_CAST(int8_t, (22 - imm8) & 31), \1233
+ HEDLEY_STATIC_CAST(int8_t, (23 - imm8) & 31), \1234
+ HEDLEY_STATIC_CAST(int8_t, (24 - imm8) & 31), \1235
+ HEDLEY_STATIC_CAST(int8_t, (25 - imm8) & 31), \1236
+ HEDLEY_STATIC_CAST(int8_t, (26 - imm8) & 31), \1237
+ HEDLEY_STATIC_CAST(int8_t, (27 - imm8) & 31), \1238
+ HEDLEY_STATIC_CAST(int8_t, (28 - imm8) & 31), \1239
+ HEDLEY_STATIC_CAST(int8_t, (29 - imm8) & 31), \1240
+ HEDLEY_STATIC_CAST(int8_t, (30 - imm8) & 31), \1241
+ HEDLEY_STATIC_CAST(int8_t, (31 - imm8) & 31)); \1242
+ } \1243
+ simde__m128i_from_private(simde__tmp_r_); \1244
+ }))1245
+#endif1246
+#define simde_mm_slli_si128(a, imm8) simde_mm_bslli_si128(a, imm8)1247
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1248
+#define _mm_bslli_si128(a, imm8) simde_mm_bslli_si128(a, imm8)1249
+#define _mm_slli_si128(a, imm8) simde_mm_bslli_si128(a, imm8)1250
+#endif1251
+1252
+SIMDE_FUNCTION_ATTRIBUTES1253
+simde__m128i simde_mm_bsrli_si128(simde__m128i a, const int imm8)1254
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)1255
+{1256
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);1257
+1258
+ if (HEDLEY_UNLIKELY((imm8 & ~15))) {1259
+ return simde_mm_setzero_si128();1260
+ }1261
+1262
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE) && defined(SIMDE_ENDIAN_ORDER)1263
+ r_.altivec_i8 =1264
+#if (SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_LITTLE)1265
+ vec_sro1266
+#else /* SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_BIG */1267
+ vec_slo1268
+#endif1269
+ (a_.altivec_i8,1270
+ vec_splats(HEDLEY_STATIC_CAST(unsigned char, imm8 * 8)));1271
+#else1272
+ SIMDE_VECTORIZE1273
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1274
+ const int e = HEDLEY_STATIC_CAST(int, i) + imm8;1275
+ r_.i8[i] = (e < 16) ? a_.i8[e] : 0;1276
+ }1277
+#endif1278
+1279
+ return simde__m128i_from_private(r_);1280
+}1281
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)1282
+#define simde_mm_bsrli_si128(a, imm8) _mm_srli_si128(a, imm8)1283
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) && !defined(__clang__)1284
+#define simde_mm_bsrli_si128(a, imm8) \1285
+ simde__m128i_from_neon_i8( \1286
+ ((imm8 < 0) || (imm8 > 15)) \1287
+ ? vdupq_n_s8(0) \1288
+ : (vextq_s8(simde__m128i_to_private(a).neon_i8, \1289
+ vdupq_n_s8(0), \1290
+ ((imm8 & 15) != 0) ? imm8 : (imm8 & 15))))1291
+#elif defined(SIMDE_SHUFFLE_VECTOR_) && !defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1292
+#define simde_mm_bsrli_si128(a, imm8) \1293
+ (__extension__({ \1294
+ const simde__m128i_private simde__tmp_a_ = \1295
+ simde__m128i_to_private(a); \1296
+ const simde__m128i_private simde__tmp_z_ = \1297
+ simde__m128i_to_private(simde_mm_setzero_si128()); \1298
+ simde__m128i_private simde__tmp_r_ = \1299
+ simde__m128i_to_private(a); \1300
+ if (HEDLEY_UNLIKELY(imm8 > 15)) { \1301
+ simde__tmp_r_ = simde__m128i_to_private( \1302
+ simde_mm_setzero_si128()); \1303
+ } else { \1304
+ simde__tmp_r_.i8 = SIMDE_SHUFFLE_VECTOR_( \1305
+ 8, 16, simde__tmp_z_.i8, (simde__tmp_a_).i8, \1306
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 16) & 31), \1307
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 17) & 31), \1308
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 18) & 31), \1309
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 19) & 31), \1310
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 20) & 31), \1311
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 21) & 31), \1312
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 22) & 31), \1313
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 23) & 31), \1314
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 24) & 31), \1315
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 25) & 31), \1316
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 26) & 31), \1317
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 27) & 31), \1318
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 28) & 31), \1319
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 29) & 31), \1320
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 30) & 31), \1321
+ HEDLEY_STATIC_CAST(int8_t, (imm8 + 31) & 31)); \1322
+ } \1323
+ simde__m128i_from_private(simde__tmp_r_); \1324
+ }))1325
+#endif1326
+#define simde_mm_srli_si128(a, imm8) simde_mm_bsrli_si128((a), (imm8))1327
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1328
+#define _mm_bsrli_si128(a, imm8) simde_mm_bsrli_si128((a), (imm8))1329
+#define _mm_srli_si128(a, imm8) simde_mm_bsrli_si128((a), (imm8))1330
+#endif1331
+1332
+SIMDE_FUNCTION_ATTRIBUTES1333
+void simde_mm_clflush(void const *p)1334
+{1335
+#if defined(SIMDE_X86_SSE2_NATIVE)1336
+ _mm_clflush(p);1337
+#else1338
+ (void)p;1339
+#endif1340
+}1341
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1342
+#define _mm_clflush(a, b) simde_mm_clflush()1343
+#endif1344
+1345
+SIMDE_FUNCTION_ATTRIBUTES1346
+int simde_mm_comieq_sd(simde__m128d a, simde__m128d b)1347
+{1348
+#if defined(SIMDE_X86_SSE2_NATIVE)1349
+ return _mm_comieq_sd(a, b);1350
+#else1351
+ simde__m128d_private a_ = simde__m128d_to_private(a),1352
+ b_ = simde__m128d_to_private(b);1353
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1354
+ return !!vgetq_lane_u64(vceqq_f64(a_.neon_f64, b_.neon_f64), 0);1355
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1356
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) ==1357
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);1358
+#else1359
+ return a_.f64[0] == b_.f64[0];1360
+#endif1361
+#endif1362
+}1363
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1364
+#define _mm_comieq_sd(a, b) simde_mm_comieq_sd(a, b)1365
+#endif1366
+1367
+SIMDE_FUNCTION_ATTRIBUTES1368
+int simde_mm_comige_sd(simde__m128d a, simde__m128d b)1369
+{1370
+#if defined(SIMDE_X86_SSE2_NATIVE)1371
+ return _mm_comige_sd(a, b);1372
+#else1373
+ simde__m128d_private a_ = simde__m128d_to_private(a),1374
+ b_ = simde__m128d_to_private(b);1375
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1376
+ return !!vgetq_lane_u64(vcgeq_f64(a_.neon_f64, b_.neon_f64), 0);1377
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1378
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) >=1379
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);1380
+#else1381
+ return a_.f64[0] >= b_.f64[0];1382
+#endif1383
+#endif1384
+}1385
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1386
+#define _mm_comige_sd(a, b) simde_mm_comige_sd(a, b)1387
+#endif1388
+1389
+SIMDE_FUNCTION_ATTRIBUTES1390
+int simde_mm_comigt_sd(simde__m128d a, simde__m128d b)1391
+{1392
+#if defined(SIMDE_X86_SSE2_NATIVE)1393
+ return _mm_comigt_sd(a, b);1394
+#else1395
+ simde__m128d_private a_ = simde__m128d_to_private(a),1396
+ b_ = simde__m128d_to_private(b);1397
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1398
+ return !!vgetq_lane_u64(vcgtq_f64(a_.neon_f64, b_.neon_f64), 0);1399
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1400
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) >1401
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);1402
+#else1403
+ return a_.f64[0] > b_.f64[0];1404
+#endif1405
+#endif1406
+}1407
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1408
+#define _mm_comigt_sd(a, b) simde_mm_comigt_sd(a, b)1409
+#endif1410
+1411
+SIMDE_FUNCTION_ATTRIBUTES1412
+int simde_mm_comile_sd(simde__m128d a, simde__m128d b)1413
+{1414
+#if defined(SIMDE_X86_SSE2_NATIVE)1415
+ return _mm_comile_sd(a, b);1416
+#else1417
+ simde__m128d_private a_ = simde__m128d_to_private(a),1418
+ b_ = simde__m128d_to_private(b);1419
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1420
+ return !!vgetq_lane_u64(vcleq_f64(a_.neon_f64, b_.neon_f64), 0);1421
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1422
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) <=1423
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);1424
+#else1425
+ return a_.f64[0] <= b_.f64[0];1426
+#endif1427
+#endif1428
+}1429
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1430
+#define _mm_comile_sd(a, b) simde_mm_comile_sd(a, b)1431
+#endif1432
+1433
+SIMDE_FUNCTION_ATTRIBUTES1434
+int simde_mm_comilt_sd(simde__m128d a, simde__m128d b)1435
+{1436
+#if defined(SIMDE_X86_SSE2_NATIVE)1437
+ return _mm_comilt_sd(a, b);1438
+#else1439
+ simde__m128d_private a_ = simde__m128d_to_private(a),1440
+ b_ = simde__m128d_to_private(b);1441
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1442
+ return !!vgetq_lane_u64(vcltq_f64(a_.neon_f64, b_.neon_f64), 0);1443
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1444
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) <1445
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);1446
+#else1447
+ return a_.f64[0] < b_.f64[0];1448
+#endif1449
+#endif1450
+}1451
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1452
+#define _mm_comilt_sd(a, b) simde_mm_comilt_sd(a, b)1453
+#endif1454
+1455
+SIMDE_FUNCTION_ATTRIBUTES1456
+int simde_mm_comineq_sd(simde__m128d a, simde__m128d b)1457
+{1458
+#if defined(SIMDE_X86_SSE2_NATIVE)1459
+ return _mm_comineq_sd(a, b);1460
+#else1461
+ simde__m128d_private a_ = simde__m128d_to_private(a),1462
+ b_ = simde__m128d_to_private(b);1463
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1464
+ return !vgetq_lane_u64(vceqq_f64(a_.neon_f64, b_.neon_f64), 0);1465
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1466
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) !=1467
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);1468
+#else1469
+ return a_.f64[0] != b_.f64[0];1470
+#endif1471
+#endif1472
+}1473
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1474
+#define _mm_comineq_sd(a, b) simde_mm_comineq_sd(a, b)1475
+#endif1476
+1477
+SIMDE_FUNCTION_ATTRIBUTES1478
+simde__m128d simde_x_mm_copysign_pd(simde__m128d dest, simde__m128d src)1479
+{1480
+ simde__m128d_private r_, dest_ = simde__m128d_to_private(dest),1481
+ src_ = simde__m128d_to_private(src);1482
+1483
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1484
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1485
+ uint64x2_t sign_pos =1486
+ vreinterpretq_u64_f64(vdupq_n_f64(-SIMDE_FLOAT64_C(0.0)));1487
+#else1488
+ simde_float64 dbl_nz = -SIMDE_FLOAT64_C(0.0);1489
+ uint64_t u64_nz;1490
+ simde_memcpy(&u64_nz, &dbl_nz, sizeof(u64_nz));1491
+ uint64x2_t sign_pos = vdupq_n_u64(u64_nz);1492
+#endif1493
+ r_.neon_u64 = vbslq_u64(sign_pos, src_.neon_u64, dest_.neon_u64);1494
+#elif defined(SIMDE_POWER_ALTIVEC_P9_NATIVE)1495
+#if !defined(HEDLEY_IBM_VERSION)1496
+ r_.altivec_f64 = vec_cpsgn(dest_.altivec_f64, src_.altivec_f64);1497
+#else1498
+ r_.altivec_f64 = vec_cpsgn(src_.altivec_f64, dest_.altivec_f64);1499
+#endif1500
+#elif defined(simde_math_copysign)1501
+ SIMDE_VECTORIZE1502
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1503
+ r_.f64[i] = simde_math_copysign(dest_.f64[i], src_.f64[i]);1504
+ }1505
+#else1506
+ simde__m128d sgnbit = simde_mm_set1_pd(-SIMDE_FLOAT64_C(0.0));1507
+ return simde_mm_xor_pd(simde_mm_and_pd(sgnbit, src),1508
+ simde_mm_andnot_pd(sgnbit, dest));1509
+#endif1510
+1511
+ return simde__m128d_from_private(r_);1512
+}1513
+1514
+SIMDE_FUNCTION_ATTRIBUTES1515
+simde__m128d simde_x_mm_xorsign_pd(simde__m128d dest, simde__m128d src)1516
+{1517
+ return simde_mm_xor_pd(simde_mm_and_pd(simde_mm_set1_pd(-0.0), src),1518
+ dest);1519
+}1520
+1521
+SIMDE_FUNCTION_ATTRIBUTES1522
+simde__m128 simde_mm_castpd_ps(simde__m128d a)1523
+{1524
+#if defined(SIMDE_X86_SSE2_NATIVE)1525
+ return _mm_castpd_ps(a);1526
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)1527
+ return vreinterpretq_f32_f64(a);1528
+#else1529
+ simde__m128 r;1530
+ simde_memcpy(&r, &a, sizeof(a));1531
+ return r;1532
+#endif1533
+}1534
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1535
+#define _mm_castpd_ps(a) simde_mm_castpd_ps(a)1536
+#endif1537
+1538
+SIMDE_FUNCTION_ATTRIBUTES1539
+simde__m128i simde_mm_castpd_si128(simde__m128d a)1540
+{1541
+#if defined(SIMDE_X86_SSE2_NATIVE)1542
+ return _mm_castpd_si128(a);1543
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)1544
+ return vreinterpretq_s64_f64(a);1545
+#else1546
+ simde__m128i r;1547
+ simde_memcpy(&r, &a, sizeof(a));1548
+ return r;1549
+#endif1550
+}1551
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1552
+#define _mm_castpd_si128(a) simde_mm_castpd_si128(a)1553
+#endif1554
+1555
+SIMDE_FUNCTION_ATTRIBUTES1556
+simde__m128d simde_mm_castps_pd(simde__m128 a)1557
+{1558
+#if defined(SIMDE_X86_SSE2_NATIVE)1559
+ return _mm_castps_pd(a);1560
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)1561
+ return vreinterpretq_f64_f32(a);1562
+#else1563
+ simde__m128d r;1564
+ simde_memcpy(&r, &a, sizeof(a));1565
+ return r;1566
+#endif1567
+}1568
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1569
+#define _mm_castps_pd(a) simde_mm_castps_pd(a)1570
+#endif1571
+1572
+SIMDE_FUNCTION_ATTRIBUTES1573
+simde__m128i simde_mm_castps_si128(simde__m128 a)1574
+{1575
+#if defined(SIMDE_X86_SSE2_NATIVE)1576
+ return _mm_castps_si128(a);1577
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1578
+ return simde__m128i_from_neon_i32(simde__m128_to_private(a).neon_i32);1579
+#else1580
+ simde__m128i r;1581
+ simde_memcpy(&r, &a, sizeof(a));1582
+ return r;1583
+#endif1584
+}1585
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1586
+#define _mm_castps_si128(a) simde_mm_castps_si128(a)1587
+#endif1588
+1589
+SIMDE_FUNCTION_ATTRIBUTES1590
+simde__m128d simde_mm_castsi128_pd(simde__m128i a)1591
+{1592
+#if defined(SIMDE_X86_SSE2_NATIVE)1593
+ return _mm_castsi128_pd(a);1594
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)1595
+ return vreinterpretq_f64_s64(a);1596
+#else1597
+ simde__m128d r;1598
+ simde_memcpy(&r, &a, sizeof(a));1599
+ return r;1600
+#endif1601
+}1602
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1603
+#define _mm_castsi128_pd(a) simde_mm_castsi128_pd(a)1604
+#endif1605
+1606
+SIMDE_FUNCTION_ATTRIBUTES1607
+simde__m128 simde_mm_castsi128_ps(simde__m128i a)1608
+{1609
+#if defined(SIMDE_X86_SSE2_NATIVE)1610
+ return _mm_castsi128_ps(a);1611
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1612
+ return HEDLEY_REINTERPRET_CAST(SIMDE_POWER_ALTIVEC_VECTOR(float), a);1613
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)1614
+ return simde__m128_from_neon_i32(simde__m128i_to_private(a).neon_i32);1615
+#else1616
+ simde__m128 r;1617
+ simde_memcpy(&r, &a, sizeof(a));1618
+ return r;1619
+#endif1620
+}1621
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1622
+#define _mm_castsi128_ps(a) simde_mm_castsi128_ps(a)1623
+#endif1624
+1625
+SIMDE_FUNCTION_ATTRIBUTES1626
+simde__m128i simde_mm_cmpeq_epi8(simde__m128i a, simde__m128i b)1627
+{1628
+#if defined(SIMDE_X86_SSE2_NATIVE)1629
+ return _mm_cmpeq_epi8(a, b);1630
+#else1631
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1632
+ b_ = simde__m128i_to_private(b);1633
+1634
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1635
+ r_.neon_u8 = vceqq_s8(b_.neon_i8, a_.neon_i8);1636
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1637
+ r_.wasm_v128 = wasm_i8x16_eq(a_.wasm_v128, b_.wasm_v128);1638
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1639
+ r_.altivec_i8 = HEDLEY_REINTERPRET_CAST(1640
+ SIMDE_POWER_ALTIVEC_VECTOR(signed char),1641
+ vec_cmpeq(a_.altivec_i8, b_.altivec_i8));1642
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1643
+ r_.i8 = HEDLEY_STATIC_CAST(__typeof__(r_.i8), (a_.i8 == b_.i8));1644
+#else1645
+ SIMDE_VECTORIZE1646
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1647
+ r_.i8[i] = (a_.i8[i] == b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);1648
+ }1649
+#endif1650
+1651
+ return simde__m128i_from_private(r_);1652
+#endif1653
+}1654
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1655
+#define _mm_cmpeq_epi8(a, b) simde_mm_cmpeq_epi8(a, b)1656
+#endif1657
+1658
+SIMDE_FUNCTION_ATTRIBUTES1659
+simde__m128i simde_mm_cmpeq_epi16(simde__m128i a, simde__m128i b)1660
+{1661
+#if defined(SIMDE_X86_SSE2_NATIVE)1662
+ return _mm_cmpeq_epi16(a, b);1663
+#else1664
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1665
+ b_ = simde__m128i_to_private(b);1666
+1667
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1668
+ r_.neon_u16 = vceqq_s16(b_.neon_i16, a_.neon_i16);1669
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1670
+ r_.wasm_v128 = wasm_i16x8_eq(a_.wasm_v128, b_.wasm_v128);1671
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1672
+ r_.altivec_i16 = HEDLEY_REINTERPRET_CAST(1673
+ SIMDE_POWER_ALTIVEC_VECTOR(signed short),1674
+ vec_cmpeq(a_.altivec_i16, b_.altivec_i16));1675
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1676
+ r_.i16 = (a_.i16 == b_.i16);1677
+#else1678
+ SIMDE_VECTORIZE1679
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1680
+ r_.i16[i] = (a_.i16[i] == b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);1681
+ }1682
+#endif1683
+1684
+ return simde__m128i_from_private(r_);1685
+#endif1686
+}1687
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1688
+#define _mm_cmpeq_epi16(a, b) simde_mm_cmpeq_epi16(a, b)1689
+#endif1690
+1691
+SIMDE_FUNCTION_ATTRIBUTES1692
+simde__m128i simde_mm_cmpeq_epi32(simde__m128i a, simde__m128i b)1693
+{1694
+#if defined(SIMDE_X86_SSE2_NATIVE)1695
+ return _mm_cmpeq_epi32(a, b);1696
+#else1697
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1698
+ b_ = simde__m128i_to_private(b);1699
+1700
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1701
+ r_.neon_u32 = vceqq_s32(b_.neon_i32, a_.neon_i32);1702
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1703
+ r_.wasm_v128 = wasm_i32x4_eq(a_.wasm_v128, b_.wasm_v128);1704
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1705
+ r_.altivec_i32 = HEDLEY_REINTERPRET_CAST(1706
+ SIMDE_POWER_ALTIVEC_VECTOR(signed int),1707
+ vec_cmpeq(a_.altivec_i32, b_.altivec_i32));1708
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1709
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), a_.i32 == b_.i32);1710
+#else1711
+ SIMDE_VECTORIZE1712
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1713
+ r_.i32[i] = (a_.i32[i] == b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);1714
+ }1715
+#endif1716
+1717
+ return simde__m128i_from_private(r_);1718
+#endif1719
+}1720
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1721
+#define _mm_cmpeq_epi32(a, b) simde_mm_cmpeq_epi32(a, b)1722
+#endif1723
+1724
+SIMDE_FUNCTION_ATTRIBUTES1725
+simde__m128d simde_mm_cmpeq_pd(simde__m128d a, simde__m128d b)1726
+{1727
+#if defined(SIMDE_X86_SSE2_NATIVE)1728
+ return _mm_cmpeq_pd(a, b);1729
+#else1730
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),1731
+ b_ = simde__m128d_to_private(b);1732
+1733
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1734
+ r_.neon_u64 = vceqq_s64(b_.neon_i64, a_.neon_i64);1735
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1736
+ r_.wasm_v128 = wasm_f64x2_eq(a_.wasm_v128, b_.wasm_v128);1737
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)1738
+ r_.altivec_f64 = HEDLEY_REINTERPRET_CAST(1739
+ SIMDE_POWER_ALTIVEC_VECTOR(double),1740
+ vec_cmpeq(a_.altivec_f64, b_.altivec_f64));1741
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1742
+ r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 == b_.f64));1743
+#else1744
+ SIMDE_VECTORIZE1745
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1746
+ r_.u64[i] = (a_.f64[i] == b_.f64[i]) ? ~UINT64_C(0)1747
+ : UINT64_C(0);1748
+ }1749
+#endif1750
+1751
+ return simde__m128d_from_private(r_);1752
+#endif1753
+}1754
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1755
+#define _mm_cmpeq_pd(a, b) simde_mm_cmpeq_pd(a, b)1756
+#endif1757
+1758
+SIMDE_FUNCTION_ATTRIBUTES1759
+simde__m128d simde_mm_cmpeq_sd(simde__m128d a, simde__m128d b)1760
+{1761
+#if defined(SIMDE_X86_SSE2_NATIVE)1762
+ return _mm_cmpeq_sd(a, b);1763
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1764
+ return simde_mm_move_sd(a, simde_mm_cmpeq_pd(a, b));1765
+#else1766
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),1767
+ b_ = simde__m128d_to_private(b);1768
+1769
+ r_.u64[0] = (a_.u64[0] == b_.u64[0]) ? ~UINT64_C(0) : 0;1770
+ r_.u64[1] = a_.u64[1];1771
+1772
+ return simde__m128d_from_private(r_);1773
+#endif1774
+}1775
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1776
+#define _mm_cmpeq_sd(a, b) simde_mm_cmpeq_sd(a, b)1777
+#endif1778
+1779
+SIMDE_FUNCTION_ATTRIBUTES1780
+simde__m128d simde_mm_cmpneq_pd(simde__m128d a, simde__m128d b)1781
+{1782
+#if defined(SIMDE_X86_SSE2_NATIVE)1783
+ return _mm_cmpneq_pd(a, b);1784
+#else1785
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),1786
+ b_ = simde__m128d_to_private(b);1787
+1788
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)1789
+ r_.neon_u32 = vmvnq_u32(1790
+ vreinterpretq_u32_u64(vceqq_f64(b_.neon_f64, a_.neon_f64)));1791
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1792
+ r_.wasm_v128 = wasm_f64x2_ne(a_.wasm_v128, b_.wasm_v128);1793
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1794
+ r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 != b_.f64));1795
+#else1796
+ SIMDE_VECTORIZE1797
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1798
+ r_.u64[i] = (a_.f64[i] != b_.f64[i]) ? ~UINT64_C(0)1799
+ : UINT64_C(0);1800
+ }1801
+#endif1802
+1803
+ return simde__m128d_from_private(r_);1804
+#endif1805
+}1806
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1807
+#define _mm_cmpneq_pd(a, b) simde_mm_cmpneq_pd(a, b)1808
+#endif1809
+1810
+SIMDE_FUNCTION_ATTRIBUTES1811
+simde__m128d simde_mm_cmpneq_sd(simde__m128d a, simde__m128d b)1812
+{1813
+#if defined(SIMDE_X86_SSE2_NATIVE)1814
+ return _mm_cmpneq_sd(a, b);1815
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1816
+ return simde_mm_move_sd(a, simde_mm_cmpneq_pd(a, b));1817
+#else1818
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),1819
+ b_ = simde__m128d_to_private(b);1820
+1821
+ r_.u64[0] = (a_.f64[0] != b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);1822
+ r_.u64[1] = a_.u64[1];1823
+1824
+ return simde__m128d_from_private(r_);1825
+#endif1826
+}1827
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1828
+#define _mm_cmpneq_sd(a, b) simde_mm_cmpneq_sd(a, b)1829
+#endif1830
+1831
+SIMDE_FUNCTION_ATTRIBUTES1832
+simde__m128i simde_mm_cmplt_epi8(simde__m128i a, simde__m128i b)1833
+{1834
+#if defined(SIMDE_X86_SSE2_NATIVE)1835
+ return _mm_cmplt_epi8(a, b);1836
+#else1837
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1838
+ b_ = simde__m128i_to_private(b);1839
+1840
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1841
+ r_.neon_u8 = vcltq_s8(a_.neon_i8, b_.neon_i8);1842
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1843
+ r_.altivec_i8 = HEDLEY_REINTERPRET_CAST(1844
+ SIMDE_POWER_ALTIVEC_VECTOR(signed char),1845
+ vec_cmplt(a_.altivec_i8, b_.altivec_i8));1846
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1847
+ r_.wasm_v128 = wasm_i8x16_lt(a_.wasm_v128, b_.wasm_v128);1848
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1849
+ r_.i8 = HEDLEY_STATIC_CAST(__typeof__(r_.i8), (a_.i8 < b_.i8));1850
+#else1851
+ SIMDE_VECTORIZE1852
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {1853
+ r_.i8[i] = (a_.i8[i] < b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);1854
+ }1855
+#endif1856
+1857
+ return simde__m128i_from_private(r_);1858
+#endif1859
+}1860
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1861
+#define _mm_cmplt_epi8(a, b) simde_mm_cmplt_epi8(a, b)1862
+#endif1863
+1864
+SIMDE_FUNCTION_ATTRIBUTES1865
+simde__m128i simde_mm_cmplt_epi16(simde__m128i a, simde__m128i b)1866
+{1867
+#if defined(SIMDE_X86_SSE2_NATIVE)1868
+ return _mm_cmplt_epi16(a, b);1869
+#else1870
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1871
+ b_ = simde__m128i_to_private(b);1872
+1873
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1874
+ r_.neon_u16 = vcltq_s16(a_.neon_i16, b_.neon_i16);1875
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1876
+ r_.altivec_i16 = HEDLEY_REINTERPRET_CAST(1877
+ SIMDE_POWER_ALTIVEC_VECTOR(signed short),1878
+ vec_cmplt(a_.altivec_i16, b_.altivec_i16));1879
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1880
+ r_.wasm_v128 = wasm_i16x8_lt(a_.wasm_v128, b_.wasm_v128);1881
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1882
+ r_.i16 = HEDLEY_STATIC_CAST(__typeof__(r_.i16), (a_.i16 < b_.i16));1883
+#else1884
+ SIMDE_VECTORIZE1885
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {1886
+ r_.i16[i] = (a_.i16[i] < b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);1887
+ }1888
+#endif1889
+1890
+ return simde__m128i_from_private(r_);1891
+#endif1892
+}1893
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1894
+#define _mm_cmplt_epi16(a, b) simde_mm_cmplt_epi16(a, b)1895
+#endif1896
+1897
+SIMDE_FUNCTION_ATTRIBUTES1898
+simde__m128i simde_mm_cmplt_epi32(simde__m128i a, simde__m128i b)1899
+{1900
+#if defined(SIMDE_X86_SSE2_NATIVE)1901
+ return _mm_cmplt_epi32(a, b);1902
+#else1903
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),1904
+ b_ = simde__m128i_to_private(b);1905
+1906
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)1907
+ r_.neon_u32 = vcltq_s32(a_.neon_i32, b_.neon_i32);1908
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1909
+ r_.altivec_i32 = HEDLEY_REINTERPRET_CAST(1910
+ SIMDE_POWER_ALTIVEC_VECTOR(signed int),1911
+ vec_cmplt(a_.altivec_i32, b_.altivec_i32));1912
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1913
+ r_.wasm_v128 = wasm_i32x4_lt(a_.wasm_v128, b_.wasm_v128);1914
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1915
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.i32 < b_.i32));1916
+#else1917
+ SIMDE_VECTORIZE1918
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {1919
+ r_.i32[i] = (a_.i32[i] < b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);1920
+ }1921
+#endif1922
+1923
+ return simde__m128i_from_private(r_);1924
+#endif1925
+}1926
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1927
+#define _mm_cmplt_epi32(a, b) simde_mm_cmplt_epi32(a, b)1928
+#endif1929
+1930
+SIMDE_FUNCTION_ATTRIBUTES1931
+simde__m128d simde_mm_cmplt_pd(simde__m128d a, simde__m128d b)1932
+{1933
+#if defined(SIMDE_X86_SSE2_NATIVE)1934
+ return _mm_cmplt_pd(a, b);1935
+#else1936
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),1937
+ b_ = simde__m128d_to_private(b);1938
+1939
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1940
+ r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 < b_.f64));1941
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)1942
+ r_.neon_u64 = vcltq_f64(a_.neon_f64, b_.neon_f64);1943
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1944
+ r_.wasm_v128 = wasm_f64x2_lt(a_.wasm_v128, b_.wasm_v128);1945
+#else1946
+ SIMDE_VECTORIZE1947
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {1948
+ r_.u64[i] = (a_.f64[i] < b_.f64[i]) ? ~UINT64_C(0)1949
+ : UINT64_C(0);1950
+ }1951
+#endif1952
+1953
+ return simde__m128d_from_private(r_);1954
+#endif1955
+}1956
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1957
+#define _mm_cmplt_pd(a, b) simde_mm_cmplt_pd(a, b)1958
+#endif1959
+1960
+SIMDE_FUNCTION_ATTRIBUTES1961
+simde__m128d simde_mm_cmplt_sd(simde__m128d a, simde__m128d b)1962
+{1963
+#if defined(SIMDE_X86_SSE2_NATIVE)1964
+ return _mm_cmplt_sd(a, b);1965
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)1966
+ return simde_mm_move_sd(a, simde_mm_cmplt_pd(a, b));1967
+#else1968
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),1969
+ b_ = simde__m128d_to_private(b);1970
+1971
+ r_.u64[0] = (a_.f64[0] < b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);1972
+ r_.u64[1] = a_.u64[1];1973
+1974
+ return simde__m128d_from_private(r_);1975
+#endif1976
+}1977
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)1978
+#define _mm_cmplt_sd(a, b) simde_mm_cmplt_sd(a, b)1979
+#endif1980
+1981
+SIMDE_FUNCTION_ATTRIBUTES1982
+simde__m128d simde_mm_cmple_pd(simde__m128d a, simde__m128d b)1983
+{1984
+#if defined(SIMDE_X86_SSE2_NATIVE)1985
+ return _mm_cmple_pd(a, b);1986
+#else1987
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),1988
+ b_ = simde__m128d_to_private(b);1989
+1990
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)1991
+ r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 <= b_.f64));1992
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)1993
+ r_.neon_u64 = vcleq_f64(a_.neon_f64, b_.neon_f64);1994
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)1995
+ r_.wasm_v128 = wasm_f64x2_le(a_.wasm_v128, b_.wasm_v128);1996
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)1997
+ r_.altivec_f64 = HEDLEY_REINTERPRET_CAST(1998
+ SIMDE_POWER_ALTIVEC_VECTOR(double),1999
+ vec_cmple(a_.altivec_f64, b_.altivec_f64));2000
+#else2001
+ SIMDE_VECTORIZE2002
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2003
+ r_.u64[i] = (a_.f64[i] <= b_.f64[i]) ? ~UINT64_C(0)2004
+ : UINT64_C(0);2005
+ }2006
+#endif2007
+2008
+ return simde__m128d_from_private(r_);2009
+#endif2010
+}2011
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2012
+#define _mm_cmple_pd(a, b) simde_mm_cmple_pd(a, b)2013
+#endif2014
+2015
+SIMDE_FUNCTION_ATTRIBUTES2016
+simde__m128d simde_mm_cmple_sd(simde__m128d a, simde__m128d b)2017
+{2018
+#if defined(SIMDE_X86_SSE2_NATIVE)2019
+ return _mm_cmple_sd(a, b);2020
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2021
+ return simde_mm_move_sd(a, simde_mm_cmple_pd(a, b));2022
+#else2023
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2024
+ b_ = simde__m128d_to_private(b);2025
+2026
+ r_.u64[0] = (a_.f64[0] <= b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);2027
+ r_.u64[1] = a_.u64[1];2028
+2029
+ return simde__m128d_from_private(r_);2030
+#endif2031
+}2032
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2033
+#define _mm_cmple_sd(a, b) simde_mm_cmple_sd(a, b)2034
+#endif2035
+2036
+SIMDE_FUNCTION_ATTRIBUTES2037
+simde__m128i simde_mm_cmpgt_epi8(simde__m128i a, simde__m128i b)2038
+{2039
+#if defined(SIMDE_X86_SSE2_NATIVE)2040
+ return _mm_cmpgt_epi8(a, b);2041
+#else2042
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),2043
+ b_ = simde__m128i_to_private(b);2044
+2045
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2046
+ r_.neon_u8 = vcgtq_s8(a_.neon_i8, b_.neon_i8);2047
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2048
+ r_.wasm_v128 = wasm_i8x16_gt(a_.wasm_v128, b_.wasm_v128);2049
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2050
+ r_.altivec_i8 = HEDLEY_REINTERPRET_CAST(2051
+ SIMDE_POWER_ALTIVEC_VECTOR(signed char),2052
+ vec_cmpgt(a_.altivec_i8, b_.altivec_i8));2053
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2054
+ r_.i8 = HEDLEY_STATIC_CAST(__typeof__(r_.i8), (a_.i8 > b_.i8));2055
+#else2056
+ SIMDE_VECTORIZE2057
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {2058
+ r_.i8[i] = (a_.i8[i] > b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);2059
+ }2060
+#endif2061
+2062
+ return simde__m128i_from_private(r_);2063
+#endif2064
+}2065
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2066
+#define _mm_cmpgt_epi8(a, b) simde_mm_cmpgt_epi8(a, b)2067
+#endif2068
+2069
+SIMDE_FUNCTION_ATTRIBUTES2070
+simde__m128i simde_mm_cmpgt_epi16(simde__m128i a, simde__m128i b)2071
+{2072
+#if defined(SIMDE_X86_SSE2_NATIVE)2073
+ return _mm_cmpgt_epi16(a, b);2074
+#else2075
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),2076
+ b_ = simde__m128i_to_private(b);2077
+2078
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2079
+ r_.neon_u16 = vcgtq_s16(a_.neon_i16, b_.neon_i16);2080
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2081
+ r_.wasm_v128 = wasm_i16x8_gt(a_.wasm_v128, b_.wasm_v128);2082
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2083
+ r_.altivec_i16 = HEDLEY_REINTERPRET_CAST(2084
+ SIMDE_POWER_ALTIVEC_VECTOR(signed short),2085
+ vec_cmpgt(a_.altivec_i16, b_.altivec_i16));2086
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2087
+ r_.i16 = HEDLEY_STATIC_CAST(__typeof__(r_.i16), (a_.i16 > b_.i16));2088
+#else2089
+ SIMDE_VECTORIZE2090
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {2091
+ r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);2092
+ }2093
+#endif2094
+2095
+ return simde__m128i_from_private(r_);2096
+#endif2097
+}2098
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2099
+#define _mm_cmpgt_epi16(a, b) simde_mm_cmpgt_epi16(a, b)2100
+#endif2101
+2102
+SIMDE_FUNCTION_ATTRIBUTES2103
+simde__m128i simde_mm_cmpgt_epi32(simde__m128i a, simde__m128i b)2104
+{2105
+#if defined(SIMDE_X86_SSE2_NATIVE)2106
+ return _mm_cmpgt_epi32(a, b);2107
+#else2108
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),2109
+ b_ = simde__m128i_to_private(b);2110
+2111
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2112
+ r_.neon_u32 = vcgtq_s32(a_.neon_i32, b_.neon_i32);2113
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2114
+ r_.wasm_v128 = wasm_i32x4_gt(a_.wasm_v128, b_.wasm_v128);2115
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2116
+ r_.altivec_i32 = HEDLEY_REINTERPRET_CAST(2117
+ SIMDE_POWER_ALTIVEC_VECTOR(signed int),2118
+ vec_cmpgt(a_.altivec_i32, b_.altivec_i32));2119
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2120
+ r_.i32 = HEDLEY_STATIC_CAST(__typeof__(r_.i32), (a_.i32 > b_.i32));2121
+#else2122
+ SIMDE_VECTORIZE2123
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {2124
+ r_.i32[i] = (a_.i32[i] > b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);2125
+ }2126
+#endif2127
+2128
+ return simde__m128i_from_private(r_);2129
+#endif2130
+}2131
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2132
+#define _mm_cmpgt_epi32(a, b) simde_mm_cmpgt_epi32(a, b)2133
+#endif2134
+2135
+SIMDE_FUNCTION_ATTRIBUTES2136
+simde__m128d simde_mm_cmpgt_pd(simde__m128d a, simde__m128d b)2137
+{2138
+#if defined(SIMDE_X86_SSE2_NATIVE)2139
+ return _mm_cmpgt_pd(a, b);2140
+#else2141
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2142
+ b_ = simde__m128d_to_private(b);2143
+2144
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2145
+ r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 > b_.f64));2146
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)2147
+ r_.neon_u64 = vcgtq_f64(a_.neon_f64, b_.neon_f64);2148
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2149
+ r_.wasm_v128 = wasm_f64x2_gt(a_.wasm_v128, b_.wasm_v128);2150
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2151
+ r_.altivec_f64 =2152
+ HEDLEY_STATIC_CAST(SIMDE_POWER_ALTIVEC_VECTOR(double),2153
+ vec_cmpgt(a_.altivec_f64, b_.altivec_f64));2154
+#else2155
+ SIMDE_VECTORIZE2156
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2157
+ r_.u64[i] = (a_.f64[i] > b_.f64[i]) ? ~UINT64_C(0)2158
+ : UINT64_C(0);2159
+ }2160
+#endif2161
+2162
+ return simde__m128d_from_private(r_);2163
+#endif2164
+}2165
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2166
+#define _mm_cmpgt_pd(a, b) simde_mm_cmpgt_pd(a, b)2167
+#endif2168
+2169
+SIMDE_FUNCTION_ATTRIBUTES2170
+simde__m128d simde_mm_cmpgt_sd(simde__m128d a, simde__m128d b)2171
+{2172
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)2173
+ return _mm_cmpgt_sd(a, b);2174
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2175
+ return simde_mm_move_sd(a, simde_mm_cmpgt_pd(a, b));2176
+#else2177
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2178
+ b_ = simde__m128d_to_private(b);2179
+2180
+ r_.u64[0] = (a_.f64[0] > b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);2181
+ r_.u64[1] = a_.u64[1];2182
+2183
+ return simde__m128d_from_private(r_);2184
+#endif2185
+}2186
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2187
+#define _mm_cmpgt_sd(a, b) simde_mm_cmpgt_sd(a, b)2188
+#endif2189
+2190
+SIMDE_FUNCTION_ATTRIBUTES2191
+simde__m128d simde_mm_cmpge_pd(simde__m128d a, simde__m128d b)2192
+{2193
+#if defined(SIMDE_X86_SSE2_NATIVE)2194
+ return _mm_cmpge_pd(a, b);2195
+#else2196
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2197
+ b_ = simde__m128d_to_private(b);2198
+2199
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)2200
+ r_.i64 = HEDLEY_STATIC_CAST(__typeof__(r_.i64), (a_.f64 >= b_.f64));2201
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)2202
+ r_.neon_u64 = vcgeq_f64(a_.neon_f64, b_.neon_f64);2203
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2204
+ r_.wasm_v128 = wasm_f64x2_ge(a_.wasm_v128, b_.wasm_v128);2205
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2206
+ r_.altivec_f64 =2207
+ HEDLEY_STATIC_CAST(SIMDE_POWER_ALTIVEC_VECTOR(double),2208
+ vec_cmpge(a_.altivec_f64, b_.altivec_f64));2209
+#else2210
+ SIMDE_VECTORIZE2211
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2212
+ r_.u64[i] = (a_.f64[i] >= b_.f64[i]) ? ~UINT64_C(0)2213
+ : UINT64_C(0);2214
+ }2215
+#endif2216
+2217
+ return simde__m128d_from_private(r_);2218
+#endif2219
+}2220
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2221
+#define _mm_cmpge_pd(a, b) simde_mm_cmpge_pd(a, b)2222
+#endif2223
+2224
+SIMDE_FUNCTION_ATTRIBUTES2225
+simde__m128d simde_mm_cmpge_sd(simde__m128d a, simde__m128d b)2226
+{2227
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)2228
+ return _mm_cmpge_sd(a, b);2229
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2230
+ return simde_mm_move_sd(a, simde_mm_cmpge_pd(a, b));2231
+#else2232
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2233
+ b_ = simde__m128d_to_private(b);2234
+2235
+ r_.u64[0] = (a_.f64[0] >= b_.f64[0]) ? ~UINT64_C(0) : UINT64_C(0);2236
+ r_.u64[1] = a_.u64[1];2237
+2238
+ return simde__m128d_from_private(r_);2239
+#endif2240
+}2241
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2242
+#define _mm_cmpge_sd(a, b) simde_mm_cmpge_sd(a, b)2243
+#endif2244
+2245
+SIMDE_FUNCTION_ATTRIBUTES2246
+simde__m128d simde_mm_cmpngt_pd(simde__m128d a, simde__m128d b)2247
+{2248
+#if defined(SIMDE_X86_SSE2_NATIVE)2249
+ return _mm_cmpngt_pd(a, b);2250
+#else2251
+ return simde_mm_cmple_pd(a, b);2252
+#endif2253
+}2254
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2255
+#define _mm_cmpngt_pd(a, b) simde_mm_cmpngt_pd(a, b)2256
+#endif2257
+2258
+SIMDE_FUNCTION_ATTRIBUTES2259
+simde__m128d simde_mm_cmpngt_sd(simde__m128d a, simde__m128d b)2260
+{2261
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)2262
+ return _mm_cmpngt_sd(a, b);2263
+#else2264
+ return simde_mm_cmple_sd(a, b);2265
+#endif2266
+}2267
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2268
+#define _mm_cmpngt_sd(a, b) simde_mm_cmpngt_sd(a, b)2269
+#endif2270
+2271
+SIMDE_FUNCTION_ATTRIBUTES2272
+simde__m128d simde_mm_cmpnge_pd(simde__m128d a, simde__m128d b)2273
+{2274
+#if defined(SIMDE_X86_SSE2_NATIVE)2275
+ return _mm_cmpnge_pd(a, b);2276
+#else2277
+ return simde_mm_cmplt_pd(a, b);2278
+#endif2279
+}2280
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2281
+#define _mm_cmpnge_pd(a, b) simde_mm_cmpnge_pd(a, b)2282
+#endif2283
+2284
+SIMDE_FUNCTION_ATTRIBUTES2285
+simde__m128d simde_mm_cmpnge_sd(simde__m128d a, simde__m128d b)2286
+{2287
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)2288
+ return _mm_cmpnge_sd(a, b);2289
+#else2290
+ return simde_mm_cmplt_sd(a, b);2291
+#endif2292
+}2293
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2294
+#define _mm_cmpnge_sd(a, b) simde_mm_cmpnge_sd(a, b)2295
+#endif2296
+2297
+SIMDE_FUNCTION_ATTRIBUTES2298
+simde__m128d simde_mm_cmpnlt_pd(simde__m128d a, simde__m128d b)2299
+{2300
+#if defined(SIMDE_X86_SSE2_NATIVE)2301
+ return _mm_cmpnlt_pd(a, b);2302
+#else2303
+ return simde_mm_cmpge_pd(a, b);2304
+#endif2305
+}2306
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2307
+#define _mm_cmpnlt_pd(a, b) simde_mm_cmpnlt_pd(a, b)2308
+#endif2309
+2310
+SIMDE_FUNCTION_ATTRIBUTES2311
+simde__m128d simde_mm_cmpnlt_sd(simde__m128d a, simde__m128d b)2312
+{2313
+#if defined(SIMDE_X86_SSE2_NATIVE)2314
+ return _mm_cmpnlt_sd(a, b);2315
+#else2316
+ return simde_mm_cmpge_sd(a, b);2317
+#endif2318
+}2319
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2320
+#define _mm_cmpnlt_sd(a, b) simde_mm_cmpnlt_sd(a, b)2321
+#endif2322
+2323
+SIMDE_FUNCTION_ATTRIBUTES2324
+simde__m128d simde_mm_cmpnle_pd(simde__m128d a, simde__m128d b)2325
+{2326
+#if defined(SIMDE_X86_SSE2_NATIVE)2327
+ return _mm_cmpnle_pd(a, b);2328
+#else2329
+ return simde_mm_cmpgt_pd(a, b);2330
+#endif2331
+}2332
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2333
+#define _mm_cmpnle_pd(a, b) simde_mm_cmpnle_pd(a, b)2334
+#endif2335
+2336
+SIMDE_FUNCTION_ATTRIBUTES2337
+simde__m128d simde_mm_cmpnle_sd(simde__m128d a, simde__m128d b)2338
+{2339
+#if defined(SIMDE_X86_SSE2_NATIVE)2340
+ return _mm_cmpnle_sd(a, b);2341
+#else2342
+ return simde_mm_cmpgt_sd(a, b);2343
+#endif2344
+}2345
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2346
+#define _mm_cmpnle_sd(a, b) simde_mm_cmpnle_sd(a, b)2347
+#endif2348
+2349
+SIMDE_FUNCTION_ATTRIBUTES2350
+simde__m128d simde_mm_cmpord_pd(simde__m128d a, simde__m128d b)2351
+{2352
+#if defined(SIMDE_X86_SSE2_NATIVE)2353
+ return _mm_cmpord_pd(a, b);2354
+#else2355
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2356
+ b_ = simde__m128d_to_private(b);2357
+2358
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2359
+ /* Note: NEON does not have ordered compare builtin2360
+ Need to compare a eq a and b eq b to check for NaN2361
+ Do AND of results to get final */2362
+ uint64x2_t ceqaa = vceqq_f64(a_.neon_f64, a_.neon_f64);2363
+ uint64x2_t ceqbb = vceqq_f64(b_.neon_f64, b_.neon_f64);2364
+ r_.neon_u64 = vandq_u64(ceqaa, ceqbb);2365
+#elif defined(simde_math_isnan)2366
+ SIMDE_VECTORIZE2367
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2368
+ r_.u64[i] = (!simde_math_isnan(a_.f64[i]) &&2369
+ !simde_math_isnan(b_.f64[i]))2370
+ ? ~UINT64_C(0)2371
+ : UINT64_C(0);2372
+ }2373
+#else2374
+ HEDLEY_UNREACHABLE();2375
+#endif2376
+2377
+ return simde__m128d_from_private(r_);2378
+#endif2379
+}2380
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2381
+#define _mm_cmpord_pd(a, b) simde_mm_cmpord_pd(a, b)2382
+#endif2383
+2384
+SIMDE_FUNCTION_ATTRIBUTES2385
+simde_float64 simde_mm_cvtsd_f64(simde__m128d a)2386
+{2387
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)2388
+ return _mm_cvtsd_f64(a);2389
+#else2390
+ simde__m128d_private a_ = simde__m128d_to_private(a);2391
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2392
+ return HEDLEY_STATIC_CAST(simde_float64,2393
+ vgetq_lane_f64(a_.neon_f64, 0));2394
+#else2395
+ return a_.f64[0];2396
+#endif2397
+#endif2398
+}2399
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2400
+#define _mm_cvtsd_f64(a) simde_mm_cvtsd_f64(a)2401
+#endif2402
+2403
+SIMDE_FUNCTION_ATTRIBUTES2404
+simde__m128d simde_mm_cmpord_sd(simde__m128d a, simde__m128d b)2405
+{2406
+#if defined(SIMDE_X86_SSE2_NATIVE)2407
+ return _mm_cmpord_sd(a, b);2408
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2409
+ return simde_mm_move_sd(a, simde_mm_cmpord_pd(a, b));2410
+#else2411
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2412
+ b_ = simde__m128d_to_private(b);2413
+2414
+#if defined(simde_math_isnan)2415
+ r_.u64[0] =2416
+ (!simde_math_isnan(a_.f64[0]) && !simde_math_isnan(b_.f64[0]))2417
+ ? ~UINT64_C(0)2418
+ : UINT64_C(0);2419
+ r_.u64[1] = a_.u64[1];2420
+#else2421
+ HEDLEY_UNREACHABLE();2422
+#endif2423
+2424
+ return simde__m128d_from_private(r_);2425
+#endif2426
+}2427
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2428
+#define _mm_cmpord_sd(a, b) simde_mm_cmpord_sd(a, b)2429
+#endif2430
+2431
+SIMDE_FUNCTION_ATTRIBUTES2432
+simde__m128d simde_mm_cmpunord_pd(simde__m128d a, simde__m128d b)2433
+{2434
+#if defined(SIMDE_X86_SSE2_NATIVE)2435
+ return _mm_cmpunord_pd(a, b);2436
+#else2437
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2438
+ b_ = simde__m128d_to_private(b);2439
+2440
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2441
+ uint64x2_t ceqaa = vceqq_f64(a_.neon_f64, a_.neon_f64);2442
+ uint64x2_t ceqbb = vceqq_f64(b_.neon_f64, b_.neon_f64);2443
+ r_.neon_u64 = vreinterpretq_u64_u32(2444
+ vmvnq_u32(vreinterpretq_u32_u64(vandq_u64(ceqaa, ceqbb))));2445
+#elif defined(simde_math_isnan)2446
+ SIMDE_VECTORIZE2447
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2448
+ r_.u64[i] = (simde_math_isnan(a_.f64[i]) ||2449
+ simde_math_isnan(b_.f64[i]))2450
+ ? ~UINT64_C(0)2451
+ : UINT64_C(0);2452
+ }2453
+#else2454
+ HEDLEY_UNREACHABLE();2455
+#endif2456
+2457
+ return simde__m128d_from_private(r_);2458
+#endif2459
+}2460
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2461
+#define _mm_cmpunord_pd(a, b) simde_mm_cmpunord_pd(a, b)2462
+#endif2463
+2464
+SIMDE_FUNCTION_ATTRIBUTES2465
+simde__m128d simde_mm_cmpunord_sd(simde__m128d a, simde__m128d b)2466
+{2467
+#if defined(SIMDE_X86_SSE2_NATIVE)2468
+ return _mm_cmpunord_sd(a, b);2469
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)2470
+ return simde_mm_move_sd(a, simde_mm_cmpunord_pd(a, b));2471
+#else2472
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),2473
+ b_ = simde__m128d_to_private(b);2474
+2475
+#if defined(simde_math_isnan)2476
+ r_.u64[0] = (simde_math_isnan(a_.f64[0]) || simde_math_isnan(b_.f64[0]))2477
+ ? ~UINT64_C(0)2478
+ : UINT64_C(0);2479
+ r_.u64[1] = a_.u64[1];2480
+#else2481
+ HEDLEY_UNREACHABLE();2482
+#endif2483
+2484
+ return simde__m128d_from_private(r_);2485
+#endif2486
+}2487
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2488
+#define _mm_cmpunord_sd(a, b) simde_mm_cmpunord_sd(a, b)2489
+#endif2490
+2491
+SIMDE_FUNCTION_ATTRIBUTES2492
+simde__m128d simde_mm_cvtepi32_pd(simde__m128i a)2493
+{2494
+#if defined(SIMDE_X86_SSE2_NATIVE)2495
+ return _mm_cvtepi32_pd(a);2496
+#else2497
+ simde__m128d_private r_;2498
+ simde__m128i_private a_ = simde__m128i_to_private(a);2499
+2500
+#if defined(SIMDE_CONVERT_VECTOR_)2501
+ SIMDE_CONVERT_VECTOR_(r_.f64, a_.m64_private[0].i32);2502
+#else2503
+ SIMDE_VECTORIZE2504
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2505
+ r_.f64[i] = (simde_float64)a_.i32[i];2506
+ }2507
+#endif2508
+2509
+ return simde__m128d_from_private(r_);2510
+#endif2511
+}2512
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2513
+#define _mm_cvtepi32_pd(a) simde_mm_cvtepi32_pd(a)2514
+#endif2515
+2516
+SIMDE_FUNCTION_ATTRIBUTES2517
+simde__m128 simde_mm_cvtepi32_ps(simde__m128i a)2518
+{2519
+#if defined(SIMDE_X86_SSE2_NATIVE)2520
+ return _mm_cvtepi32_ps(a);2521
+#else2522
+ simde__m128_private r_;2523
+ simde__m128i_private a_ = simde__m128i_to_private(a);2524
+2525
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2526
+ r_.neon_f32 = vcvtq_f32_s32(a_.neon_i32);2527
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2528
+ r_.wasm_v128 = wasm_f32x4_convert_i32x4(a_.wasm_v128);2529
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2530
+ HEDLEY_DIAGNOSTIC_PUSH2531
+#if HEDLEY_HAS_WARNING("-Wc11-extensions")2532
+#pragma clang diagnostic ignored "-Wc11-extensions"2533
+#endif2534
+ r_.altivec_f32 = vec_ctf(a_.altivec_i32, 0);2535
+ HEDLEY_DIAGNOSTIC_POP2536
+#elif defined(SIMDE_CONVERT_VECTOR_)2537
+ SIMDE_CONVERT_VECTOR_(r_.f32, a_.i32);2538
+#else2539
+ SIMDE_VECTORIZE2540
+ for (size_t i = 0; i < (sizeof(r_.f32) / sizeof(r_.f32[0])); i++) {2541
+ r_.f32[i] = (simde_float32)a_.i32[i];2542
+ }2543
+#endif2544
+2545
+ return simde__m128_from_private(r_);2546
+#endif2547
+}2548
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2549
+#define _mm_cvtepi32_ps(a) simde_mm_cvtepi32_ps(a)2550
+#endif2551
+2552
+SIMDE_FUNCTION_ATTRIBUTES2553
+simde__m64 simde_mm_cvtpd_pi32(simde__m128d a)2554
+{2555
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2556
+ return _mm_cvtpd_pi32(a);2557
+#else2558
+ simde__m64_private r_;2559
+ simde__m128d_private a_ = simde__m128d_to_private(a);2560
+2561
+ SIMDE_VECTORIZE2562
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {2563
+ simde_float64 v = simde_math_round(a_.f64[i]);2564
+#if defined(SIMDE_FAST_CONVERSION_RANGE)2565
+ r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, v);2566
+#else2567
+ r_.i32[i] =2568
+ ((v > HEDLEY_STATIC_CAST(simde_float64, INT32_MIN)) &&2569
+ (v < HEDLEY_STATIC_CAST(simde_float64, INT32_MAX)))2570
+ ? SIMDE_CONVERT_FTOI(int32_t, v)2571
+ : INT32_MIN;2572
+#endif2573
+ }2574
+2575
+ return simde__m64_from_private(r_);2576
+#endif2577
+}2578
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2579
+#define _mm_cvtpd_pi32(a) simde_mm_cvtpd_pi32(a)2580
+#endif2581
+2582
+SIMDE_FUNCTION_ATTRIBUTES2583
+simde__m128i simde_mm_cvtpd_epi32(simde__m128d a)2584
+{2585
+#if defined(SIMDE_X86_SSE2_NATIVE)2586
+ return _mm_cvtpd_epi32(a);2587
+#else2588
+ simde__m128i_private r_;2589
+2590
+ r_.m64[0] = simde_mm_cvtpd_pi32(a);2591
+ r_.m64[1] = simde_mm_setzero_si64();2592
+2593
+ return simde__m128i_from_private(r_);2594
+#endif2595
+}2596
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2597
+#define _mm_cvtpd_epi32(a) simde_mm_cvtpd_epi32(a)2598
+#endif2599
+2600
+SIMDE_FUNCTION_ATTRIBUTES2601
+simde__m128 simde_mm_cvtpd_ps(simde__m128d a)2602
+{2603
+#if defined(SIMDE_X86_SSE2_NATIVE)2604
+ return _mm_cvtpd_ps(a);2605
+#else2606
+ simde__m128_private r_;2607
+ simde__m128d_private a_ = simde__m128d_to_private(a);2608
+2609
+#if defined(SIMDE_CONVERT_VECTOR_)2610
+ SIMDE_CONVERT_VECTOR_(r_.m64_private[0].f32, a_.f64);2611
+ r_.m64_private[1] = simde__m64_to_private(simde_mm_setzero_si64());2612
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)2613
+ r_.neon_f32 = vreinterpretq_f32_f64(2614
+ vcombine_f64(vreinterpret_f64_f32(vcvtx_f32_f64(a_.neon_f64)),2615
+ vdup_n_f64(0)));2616
+#else2617
+ SIMDE_VECTORIZE2618
+ for (size_t i = 0; i < (sizeof(a_.f64) / sizeof(a_.f64[0])); i++) {2619
+ r_.f32[i] = (simde_float32)a_.f64[i];2620
+ }2621
+ simde_memset(&(r_.m64_private[1]), 0, sizeof(r_.m64_private[1]));2622
+#endif2623
+2624
+ return simde__m128_from_private(r_);2625
+#endif2626
+}2627
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2628
+#define _mm_cvtpd_ps(a) simde_mm_cvtpd_ps(a)2629
+#endif2630
+2631
+SIMDE_FUNCTION_ATTRIBUTES2632
+simde__m128d simde_mm_cvtpi32_pd(simde__m64 a)2633
+{2634
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)2635
+ return _mm_cvtpi32_pd(a);2636
+#else2637
+ simde__m128d_private r_;2638
+ simde__m64_private a_ = simde__m64_to_private(a);2639
+2640
+#if defined(SIMDE_CONVERT_VECTOR_)2641
+ SIMDE_CONVERT_VECTOR_(r_.f64, a_.i32);2642
+#else2643
+ SIMDE_VECTORIZE2644
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2645
+ r_.f64[i] = (simde_float64)a_.i32[i];2646
+ }2647
+#endif2648
+2649
+ return simde__m128d_from_private(r_);2650
+#endif2651
+}2652
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2653
+#define _mm_cvtpi32_pd(a) simde_mm_cvtpi32_pd(a)2654
+#endif2655
+2656
+SIMDE_FUNCTION_ATTRIBUTES2657
+simde__m128i simde_mm_cvtps_epi32(simde__m128 a)2658
+{2659
+#if defined(SIMDE_X86_SSE2_NATIVE)2660
+ return _mm_cvtps_epi32(a);2661
+#else2662
+ simde__m128i_private r_;2663
+ simde__m128_private a_ = simde__m128_to_private(a);2664
+2665
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE) && defined(SIMDE_FAST_CONVERSION_RANGE)2666
+ r_.neon_i32 = vcvtnq_s32_f32(a_.neon_f32);2667
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) && \2668
+ defined(SIMDE_FAST_CONVERSION_RANGE) && defined(SIMDE_FAST_ROUND_TIES)2669
+ r_.neon_i32 = vcvtnq_s32_f32(a_.neon_f32);2670
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE) && \2671
+ defined(SIMDE_FAST_CONVERSION_RANGE) && defined(SIMDE_FAST_ROUND_TIES)2672
+ HEDLEY_DIAGNOSTIC_PUSH2673
+ SIMDE_DIAGNOSTIC_DISABLE_C11_EXTENSIONS_2674
+ SIMDE_DIAGNOSTIC_DISABLE_VECTOR_CONVERSION_2675
+ r_.altivec_i32 = vec_cts(a_.altivec_f32, 1);2676
+ HEDLEY_DIAGNOSTIC_POP2677
+#else2678
+ a_ = simde__m128_to_private(2679
+ simde_x_mm_round_ps(a, SIMDE_MM_FROUND_TO_NEAREST_INT, 1));2680
+ SIMDE_VECTORIZE2681
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {2682
+ simde_float32 v = simde_math_roundf(a_.f32[i]);2683
+#if defined(SIMDE_FAST_CONVERSION_RANGE)2684
+ r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, v);2685
+#else2686
+ r_.i32[i] =2687
+ ((v > HEDLEY_STATIC_CAST(simde_float32, INT32_MIN)) &&2688
+ (v < HEDLEY_STATIC_CAST(simde_float32, INT32_MAX)))2689
+ ? SIMDE_CONVERT_FTOI(int32_t, v)2690
+ : INT32_MIN;2691
+#endif2692
+ }2693
+#endif2694
+2695
+ return simde__m128i_from_private(r_);2696
+#endif2697
+}2698
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2699
+#define _mm_cvtps_epi32(a) simde_mm_cvtps_epi32(a)2700
+#endif2701
+2702
+SIMDE_FUNCTION_ATTRIBUTES2703
+simde__m128d simde_mm_cvtps_pd(simde__m128 a)2704
+{2705
+#if defined(SIMDE_X86_SSE2_NATIVE)2706
+ return _mm_cvtps_pd(a);2707
+#else2708
+ simde__m128d_private r_;2709
+ simde__m128_private a_ = simde__m128_to_private(a);2710
+2711
+#if defined(SIMDE_CONVERT_VECTOR_)2712
+ SIMDE_CONVERT_VECTOR_(r_.f64, a_.m64_private[0].f32);2713
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)2714
+ r_.neon_f64 = vcvt_f64_f32(vget_low_f32(a_.neon_f32));2715
+#else2716
+ SIMDE_VECTORIZE2717
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {2718
+ r_.f64[i] = a_.f32[i];2719
+ }2720
+#endif2721
+2722
+ return simde__m128d_from_private(r_);2723
+#endif2724
+}2725
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2726
+#define _mm_cvtps_pd(a) simde_mm_cvtps_pd(a)2727
+#endif2728
+2729
+SIMDE_FUNCTION_ATTRIBUTES2730
+int32_t simde_mm_cvtsd_si32(simde__m128d a)2731
+{2732
+#if defined(SIMDE_X86_SSE2_NATIVE)2733
+ return _mm_cvtsd_si32(a);2734
+#else2735
+ simde__m128d_private a_ = simde__m128d_to_private(a);2736
+2737
+ simde_float64 v = simde_math_round(a_.f64[0]);2738
+#if defined(SIMDE_FAST_CONVERSION_RANGE)2739
+ return SIMDE_CONVERT_FTOI(int32_t, v);2740
+#else2741
+ return ((v > HEDLEY_STATIC_CAST(simde_float64, INT32_MIN)) &&2742
+ (v < HEDLEY_STATIC_CAST(simde_float64, INT32_MAX)))2743
+ ? SIMDE_CONVERT_FTOI(int32_t, v)2744
+ : INT32_MIN;2745
+#endif2746
+#endif2747
+}2748
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2749
+#define _mm_cvtsd_si32(a) simde_mm_cvtsd_si32(a)2750
+#endif2751
+2752
+SIMDE_FUNCTION_ATTRIBUTES2753
+int64_t simde_mm_cvtsd_si64(simde__m128d a)2754
+{2755
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2756
+#if defined(__PGI)2757
+ return _mm_cvtsd_si64x(a);2758
+#else2759
+ return _mm_cvtsd_si64(a);2760
+#endif2761
+#else2762
+ simde__m128d_private a_ = simde__m128d_to_private(a);2763
+ return SIMDE_CONVERT_FTOI(int64_t, simde_math_round(a_.f64[0]));2764
+#endif2765
+}2766
+#define simde_mm_cvtsd_si64x(a) simde_mm_cvtsd_si64(a)2767
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2768
+#define _mm_cvtsd_si64(a) simde_mm_cvtsd_si64(a)2769
+#define _mm_cvtsd_si64x(a) simde_mm_cvtsd_si64x(a)2770
+#endif2771
+2772
+SIMDE_FUNCTION_ATTRIBUTES2773
+simde__m128 simde_mm_cvtsd_ss(simde__m128 a, simde__m128d b)2774
+{2775
+#if defined(SIMDE_X86_SSE2_NATIVE)2776
+ return _mm_cvtsd_ss(a, b);2777
+#else2778
+ simde__m128_private r_, a_ = simde__m128_to_private(a);2779
+ simde__m128d_private b_ = simde__m128d_to_private(b);2780
+2781
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2782
+ r_.neon_f32 = vsetq_lane_f32(2783
+ vcvtxd_f32_f64(vgetq_lane_f64(b_.neon_f64, 0)), a_.neon_f32, 0);2784
+#else2785
+ r_.f32[0] = HEDLEY_STATIC_CAST(simde_float32, b_.f64[0]);2786
+2787
+ SIMDE_VECTORIZE2788
+ for (size_t i = 1; i < (sizeof(r_) / sizeof(r_.i32[0])); i++) {2789
+ r_.i32[i] = a_.i32[i];2790
+ }2791
+#endif2792
+ return simde__m128_from_private(r_);2793
+#endif2794
+}2795
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2796
+#define _mm_cvtsd_ss(a, b) simde_mm_cvtsd_ss(a, b)2797
+#endif2798
+2799
+SIMDE_FUNCTION_ATTRIBUTES2800
+int16_t simde_x_mm_cvtsi128_si16(simde__m128i a)2801
+{2802
+ simde__m128i_private a_ = simde__m128i_to_private(a);2803
+2804
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2805
+ return vgetq_lane_s16(a_.neon_i16, 0);2806
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2807
+ return HEDLEY_STATIC_CAST(int16_t,2808
+ wasm_i16x8_extract_lane(a_.wasm_v128, 0));2809
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2810
+#if defined(SIMDE_BUG_GCC_95227)2811
+ (void)a_;2812
+#endif2813
+ return vec_extract(a_.altivec_i16, 0);2814
+#else2815
+ return a_.i16[0];2816
+#endif2817
+}2818
+2819
+SIMDE_FUNCTION_ATTRIBUTES2820
+int32_t simde_mm_cvtsi128_si32(simde__m128i a)2821
+{2822
+#if defined(SIMDE_X86_SSE2_NATIVE)2823
+ return _mm_cvtsi128_si32(a);2824
+#else2825
+ simde__m128i_private a_ = simde__m128i_to_private(a);2826
+2827
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2828
+ return vgetq_lane_s32(a_.neon_i32, 0);2829
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2830
+ return HEDLEY_STATIC_CAST(int32_t,2831
+ wasm_i32x4_extract_lane(a_.wasm_v128, 0));2832
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)2833
+#if defined(SIMDE_BUG_GCC_95227)2834
+ (void)a_;2835
+#endif2836
+ return vec_extract(a_.altivec_i32, 0);2837
+#else2838
+ return a_.i32[0];2839
+#endif2840
+#endif2841
+}2842
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2843
+#define _mm_cvtsi128_si32(a) simde_mm_cvtsi128_si32(a)2844
+#endif2845
+2846
+SIMDE_FUNCTION_ATTRIBUTES2847
+int64_t simde_mm_cvtsi128_si64(simde__m128i a)2848
+{2849
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2850
+#if defined(__PGI)2851
+ return _mm_cvtsi128_si64x(a);2852
+#else2853
+ return _mm_cvtsi128_si64(a);2854
+#endif2855
+#else2856
+ simde__m128i_private a_ = simde__m128i_to_private(a);2857
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE) && !defined(HEDLEY_IBM_VERSION)2858
+ return vec_extract(HEDLEY_REINTERPRET_CAST(2859
+ SIMDE_POWER_ALTIVEC_VECTOR(signed long long),2860
+ a_.i64),2861
+ 0);2862
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)2863
+ return vgetq_lane_s64(a_.neon_i64, 0);2864
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2865
+ return HEDLEY_STATIC_CAST(int64_t,2866
+ wasm_i64x2_extract_lane(a_.wasm_v128, 0));2867
+#endif2868
+ return a_.i64[0];2869
+#endif2870
+}2871
+#define simde_mm_cvtsi128_si64x(a) simde_mm_cvtsi128_si64(a)2872
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2873
+#define _mm_cvtsi128_si64(a) simde_mm_cvtsi128_si64(a)2874
+#define _mm_cvtsi128_si64x(a) simde_mm_cvtsi128_si64x(a)2875
+#endif2876
+2877
+SIMDE_FUNCTION_ATTRIBUTES2878
+simde__m128d simde_mm_cvtsi32_sd(simde__m128d a, int32_t b)2879
+{2880
+#if defined(SIMDE_X86_SSE2_NATIVE)2881
+ return _mm_cvtsi32_sd(a, b);2882
+#else2883
+ simde__m128d_private r_;2884
+ simde__m128d_private a_ = simde__m128d_to_private(a);2885
+2886
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE) && defined(SIMDE_ARCH_AMD64)2887
+ r_.neon_f64 = vsetq_lane_f64(HEDLEY_STATIC_CAST(float64_t, b),2888
+ a_.neon_f64, 0);2889
+#else2890
+ r_.f64[0] = HEDLEY_STATIC_CAST(simde_float64, b);2891
+ r_.i64[1] = a_.i64[1];2892
+#endif2893
+2894
+ return simde__m128d_from_private(r_);2895
+#endif2896
+}2897
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2898
+#define _mm_cvtsi32_sd(a, b) simde_mm_cvtsi32_sd(a, b)2899
+#endif2900
+2901
+SIMDE_FUNCTION_ATTRIBUTES2902
+simde__m128i simde_x_mm_cvtsi16_si128(int16_t a)2903
+{2904
+ simde__m128i_private r_;2905
+2906
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2907
+ r_.neon_i16 = vsetq_lane_s16(a, vdupq_n_s16(0), 0);2908
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2909
+ r_.wasm_v128 = wasm_i16x8_make(a, 0, 0, 0, 0, 0, 0, 0);2910
+#else2911
+ r_.i16[0] = a;2912
+ r_.i16[1] = 0;2913
+ r_.i16[2] = 0;2914
+ r_.i16[3] = 0;2915
+ r_.i16[4] = 0;2916
+ r_.i16[5] = 0;2917
+ r_.i16[6] = 0;2918
+ r_.i16[7] = 0;2919
+#endif2920
+2921
+ return simde__m128i_from_private(r_);2922
+}2923
+2924
+SIMDE_FUNCTION_ATTRIBUTES2925
+simde__m128i simde_mm_cvtsi32_si128(int32_t a)2926
+{2927
+#if defined(SIMDE_X86_SSE2_NATIVE)2928
+ return _mm_cvtsi32_si128(a);2929
+#else2930
+ simde__m128i_private r_;2931
+2932
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2933
+ r_.neon_i32 = vsetq_lane_s32(a, vdupq_n_s32(0), 0);2934
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2935
+ r_.wasm_v128 = wasm_i32x4_make(a, 0, 0, 0);2936
+#else2937
+ r_.i32[0] = a;2938
+ r_.i32[1] = 0;2939
+ r_.i32[2] = 0;2940
+ r_.i32[3] = 0;2941
+#endif2942
+2943
+ return simde__m128i_from_private(r_);2944
+#endif2945
+}2946
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2947
+#define _mm_cvtsi32_si128(a) simde_mm_cvtsi32_si128(a)2948
+#endif2949
+2950
+SIMDE_FUNCTION_ATTRIBUTES2951
+simde__m128d simde_mm_cvtsi64_sd(simde__m128d a, int64_t b)2952
+{2953
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2954
+#if !defined(__PGI)2955
+ return _mm_cvtsi64_sd(a, b);2956
+#else2957
+ return _mm_cvtsi64x_sd(a, b);2958
+#endif2959
+#else2960
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a);2961
+2962
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)2963
+ r_.neon_f64 = vsetq_lane_f64(HEDLEY_STATIC_CAST(float64_t, b),2964
+ a_.neon_f64, 0);2965
+#else2966
+ r_.f64[0] = HEDLEY_STATIC_CAST(simde_float64, b);2967
+ r_.f64[1] = a_.f64[1];2968
+#endif2969
+2970
+ return simde__m128d_from_private(r_);2971
+#endif2972
+}2973
+#define simde_mm_cvtsi64x_sd(a, b) simde_mm_cvtsi64_sd(a, b)2974
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)2975
+#define _mm_cvtsi64_sd(a, b) simde_mm_cvtsi64_sd(a, b)2976
+#define _mm_cvtsi64x_sd(a, b) simde_mm_cvtsi64x_sd(a, b)2977
+#endif2978
+2979
+SIMDE_FUNCTION_ATTRIBUTES2980
+simde__m128i simde_mm_cvtsi64_si128(int64_t a)2981
+{2982
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)2983
+#if !defined(__PGI)2984
+ return _mm_cvtsi64_si128(a);2985
+#else2986
+ return _mm_cvtsi64x_si128(a);2987
+#endif2988
+#else2989
+ simde__m128i_private r_;2990
+2991
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)2992
+ r_.neon_i64 = vsetq_lane_s64(a, vdupq_n_s64(0), 0);2993
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)2994
+ r_.wasm_v128 = wasm_i64x2_make(a, 0);2995
+#else2996
+ r_.i64[0] = a;2997
+ r_.i64[1] = 0;2998
+#endif2999
+3000
+ return simde__m128i_from_private(r_);3001
+#endif3002
+}3003
+#define simde_mm_cvtsi64x_si128(a) simde_mm_cvtsi64_si128(a)3004
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3005
+#define _mm_cvtsi64_si128(a) simde_mm_cvtsi64_si128(a)3006
+#define _mm_cvtsi64x_si128(a) simde_mm_cvtsi64x_si128(a)3007
+#endif3008
+3009
+SIMDE_FUNCTION_ATTRIBUTES3010
+simde__m128d simde_mm_cvtss_sd(simde__m128d a, simde__m128 b)3011
+{3012
+#if defined(SIMDE_X86_SSE2_NATIVE)3013
+ return _mm_cvtss_sd(a, b);3014
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)3015
+ float64x2_t temp = vcvt_f64_f32(vset_lane_f32(3016
+ vgetq_lane_f32(simde__m128_to_private(b).neon_f32, 0),3017
+ vdup_n_f32(0), 0));3018
+ return vsetq_lane_f64(3019
+ vgetq_lane_f64(simde__m128d_to_private(a).neon_f64, 1), temp,3020
+ 1);3021
+#else3022
+ simde__m128d_private a_ = simde__m128d_to_private(a);3023
+ simde__m128_private b_ = simde__m128_to_private(b);3024
+3025
+ a_.f64[0] = HEDLEY_STATIC_CAST(simde_float64, b_.f32[0]);3026
+3027
+ return simde__m128d_from_private(a_);3028
+#endif3029
+}3030
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3031
+#define _mm_cvtss_sd(a, b) simde_mm_cvtss_sd(a, b)3032
+#endif3033
+3034
+SIMDE_FUNCTION_ATTRIBUTES3035
+simde__m64 simde_mm_cvttpd_pi32(simde__m128d a)3036
+{3037
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3038
+ return _mm_cvttpd_pi32(a);3039
+#else3040
+ simde__m64_private r_;3041
+ simde__m128d_private a_ = simde__m128d_to_private(a);3042
+3043
+#if defined(SIMDE_CONVERT_VECTOR_) && defined(SIMDE_FAST_CONVERSION_RANGE)3044
+ SIMDE_CONVERT_VECTOR_(r_.i32, a_.f64);3045
+#else3046
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {3047
+ simde_float64 v = a_.f64[i];3048
+#if defined(SIMDE_FAST_CONVERSION_RANGE)3049
+ r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, v);3050
+#else3051
+ r_.i32[i] =3052
+ ((v > HEDLEY_STATIC_CAST(simde_float64, INT32_MIN)) &&3053
+ (v < HEDLEY_STATIC_CAST(simde_float64, INT32_MAX)))3054
+ ? SIMDE_CONVERT_FTOI(int32_t, v)3055
+ : INT32_MIN;3056
+#endif3057
+ }3058
+#endif3059
+3060
+ return simde__m64_from_private(r_);3061
+#endif3062
+}3063
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3064
+#define _mm_cvttpd_pi32(a) simde_mm_cvttpd_pi32(a)3065
+#endif3066
+3067
+SIMDE_FUNCTION_ATTRIBUTES3068
+simde__m128i simde_mm_cvttpd_epi32(simde__m128d a)3069
+{3070
+#if defined(SIMDE_X86_SSE2_NATIVE)3071
+ return _mm_cvttpd_epi32(a);3072
+#else3073
+ simde__m128i_private r_;3074
+3075
+ r_.m64[0] = simde_mm_cvttpd_pi32(a);3076
+ r_.m64[1] = simde_mm_setzero_si64();3077
+3078
+ return simde__m128i_from_private(r_);3079
+#endif3080
+}3081
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3082
+#define _mm_cvttpd_epi32(a) simde_mm_cvttpd_epi32(a)3083
+#endif3084
+3085
+SIMDE_FUNCTION_ATTRIBUTES3086
+simde__m128i simde_mm_cvttps_epi32(simde__m128 a)3087
+{3088
+#if defined(SIMDE_X86_SSE2_NATIVE)3089
+ return _mm_cvttps_epi32(a);3090
+#else3091
+ simde__m128i_private r_;3092
+ simde__m128_private a_ = simde__m128_to_private(a);3093
+3094
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE) && defined(SIMDE_FAST_CONVERSION_RANGE)3095
+ r_.neon_i32 = vcvtq_s32_f32(a_.neon_f32);3096
+#elif defined(SIMDE_CONVERT_VECTOR_) && defined(SIMDE_FAST_CONVERSION_RANGE)3097
+ SIMDE_CONVERT_VECTOR_(r_.i32, a_.f32);3098
+#else3099
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {3100
+ simde_float32 v = a_.f32[i];3101
+#if defined(SIMDE_FAST_CONVERSION_RANGE)3102
+ r_.i32[i] = SIMDE_CONVERT_FTOI(int32_t, v);3103
+#else3104
+ r_.i32[i] =3105
+ ((v > HEDLEY_STATIC_CAST(simde_float32, INT32_MIN)) &&3106
+ (v < HEDLEY_STATIC_CAST(simde_float32, INT32_MAX)))3107
+ ? SIMDE_CONVERT_FTOI(int32_t, v)3108
+ : INT32_MIN;3109
+#endif3110
+ }3111
+#endif3112
+3113
+ return simde__m128i_from_private(r_);3114
+#endif3115
+}3116
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3117
+#define _mm_cvttps_epi32(a) simde_mm_cvttps_epi32(a)3118
+#endif3119
+3120
+SIMDE_FUNCTION_ATTRIBUTES3121
+int32_t simde_mm_cvttsd_si32(simde__m128d a)3122
+{3123
+#if defined(SIMDE_X86_SSE2_NATIVE)3124
+ return _mm_cvttsd_si32(a);3125
+#else3126
+ simde__m128d_private a_ = simde__m128d_to_private(a);3127
+ simde_float64 v = a_.f64[0];3128
+#if defined(SIMDE_FAST_CONVERSION_RANGE)3129
+ return SIMDE_CONVERT_FTOI(int32_t, v);3130
+#else3131
+ return ((v > HEDLEY_STATIC_CAST(simde_float64, INT32_MIN)) &&3132
+ (v < HEDLEY_STATIC_CAST(simde_float64, INT32_MAX)))3133
+ ? SIMDE_CONVERT_FTOI(int32_t, v)3134
+ : INT32_MIN;3135
+#endif3136
+#endif3137
+}3138
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3139
+#define _mm_cvttsd_si32(a) simde_mm_cvttsd_si32(a)3140
+#endif3141
+3142
+SIMDE_FUNCTION_ATTRIBUTES3143
+int64_t simde_mm_cvttsd_si64(simde__m128d a)3144
+{3145
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)3146
+#if !defined(__PGI)3147
+ return _mm_cvttsd_si64(a);3148
+#else3149
+ return _mm_cvttsd_si64x(a);3150
+#endif3151
+#else3152
+ simde__m128d_private a_ = simde__m128d_to_private(a);3153
+ return SIMDE_CONVERT_FTOI(int64_t, a_.f64[0]);3154
+#endif3155
+}3156
+#define simde_mm_cvttsd_si64x(a) simde_mm_cvttsd_si64(a)3157
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3158
+#define _mm_cvttsd_si64(a) simde_mm_cvttsd_si64(a)3159
+#define _mm_cvttsd_si64x(a) simde_mm_cvttsd_si64x(a)3160
+#endif3161
+3162
+SIMDE_FUNCTION_ATTRIBUTES3163
+simde__m128d simde_mm_div_pd(simde__m128d a, simde__m128d b)3164
+{3165
+#if defined(SIMDE_X86_SSE2_NATIVE)3166
+ return _mm_div_pd(a, b);3167
+#else3168
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),3169
+ b_ = simde__m128d_to_private(b);3170
+3171
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)3172
+ r_.f64 = a_.f64 / b_.f64;3173
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)3174
+ r_.neon_f64 = vdivq_f64(a_.neon_f64, b_.neon_f64);3175
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3176
+ r_.wasm_v128 = wasm_f64x2_div(a_.wasm_v128, b_.wasm_v128);3177
+#else3178
+ SIMDE_VECTORIZE3179
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {3180
+ r_.f64[i] = a_.f64[i] / b_.f64[i];3181
+ }3182
+#endif3183
+3184
+ return simde__m128d_from_private(r_);3185
+#endif3186
+}3187
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3188
+#define _mm_div_pd(a, b) simde_mm_div_pd(a, b)3189
+#endif3190
+3191
+SIMDE_FUNCTION_ATTRIBUTES3192
+simde__m128d simde_mm_div_sd(simde__m128d a, simde__m128d b)3193
+{3194
+#if defined(SIMDE_X86_SSE2_NATIVE)3195
+ return _mm_div_sd(a, b);3196
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)3197
+ return simde_mm_move_sd(a, simde_mm_div_pd(a, b));3198
+#else3199
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),3200
+ b_ = simde__m128d_to_private(b);3201
+3202
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3203
+ float64x2_t temp = vdivq_f64(a_.neon_f64, b_.neon_f64);3204
+ r_.neon_f64 = vsetq_lane_f64(vgetq_lane(a_.neon_f64, 1), temp, 1);3205
+#else3206
+ r_.f64[0] = a_.f64[0] / b_.f64[0];3207
+ r_.f64[1] = a_.f64[1];3208
+#endif3209
+3210
+ return simde__m128d_from_private(r_);3211
+#endif3212
+}3213
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3214
+#define _mm_div_sd(a, b) simde_mm_div_sd(a, b)3215
+#endif3216
+3217
+SIMDE_FUNCTION_ATTRIBUTES3218
+int32_t simde_mm_extract_epi16(simde__m128i a, const int imm8)3219
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 7)3220
+{3221
+ uint16_t r;3222
+ simde__m128i_private a_ = simde__m128i_to_private(a);3223
+3224
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3225
+#if defined(SIMDE_BUG_GCC_95227)3226
+ (void)a_;3227
+ (void)imm8;3228
+#endif3229
+ r = HEDLEY_STATIC_CAST(uint16_t, vec_extract(a_.altivec_i16, imm8));3230
+#else3231
+ r = a_.u16[imm8 & 7];3232
+#endif3233
+3234
+ return HEDLEY_STATIC_CAST(int32_t, r);3235
+}3236
+#if defined(SIMDE_X86_SSE2_NATIVE) && \3237
+ (!defined(HEDLEY_GCC_VERSION) || HEDLEY_GCC_VERSION_CHECK(4, 6, 0))3238
+#define simde_mm_extract_epi16(a, imm8) _mm_extract_epi16(a, imm8)3239
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3240
+#define simde_mm_extract_epi16(a, imm8) \3241
+ (HEDLEY_STATIC_CAST( \3242
+ int32_t, vgetq_lane_s16(simde__m128i_to_private(a).neon_i16, \3243
+ (imm8))) & \3244
+ (INT32_C(0x0000ffff)))3245
+#endif3246
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3247
+#define _mm_extract_epi16(a, imm8) simde_mm_extract_epi16(a, imm8)3248
+#endif3249
+3250
+SIMDE_FUNCTION_ATTRIBUTES3251
+simde__m128i simde_mm_insert_epi16(simde__m128i a, int16_t i, const int imm8)3252
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 7)3253
+{3254
+ simde__m128i_private a_ = simde__m128i_to_private(a);3255
+ a_.i16[imm8 & 7] = i;3256
+ return simde__m128i_from_private(a_);3257
+}3258
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)3259
+#define simde_mm_insert_epi16(a, i, imm8) _mm_insert_epi16((a), (i), (imm8))3260
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3261
+#define simde_mm_insert_epi16(a, i, imm8) \3262
+ simde__m128i_from_neon_i16( \3263
+ vsetq_lane_s16((i), simde__m128i_to_neon_i16(a), (imm8)))3264
+#endif3265
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3266
+#define _mm_insert_epi16(a, i, imm8) simde_mm_insert_epi16(a, i, imm8)3267
+#endif3268
+3269
+SIMDE_FUNCTION_ATTRIBUTES3270
+simde__m128d3271
+simde_mm_load_pd(simde_float64 const mem_addr[HEDLEY_ARRAY_PARAM(2)])3272
+{3273
+#if defined(SIMDE_X86_SSE2_NATIVE)3274
+ return _mm_load_pd(mem_addr);3275
+#else3276
+ simde__m128d_private r_;3277
+3278
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3279
+ r_.neon_f64 = vld1q_f64(mem_addr);3280
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3281
+ r_.neon_u32 =3282
+ vld1q_u32(HEDLEY_REINTERPRET_CAST(uint32_t const *, mem_addr));3283
+#else3284
+ simde_memcpy(&r_, SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m128d),3285
+ sizeof(r_));3286
+#endif3287
+3288
+ return simde__m128d_from_private(r_);3289
+#endif3290
+}3291
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3292
+#define _mm_load_pd(mem_addr) simde_mm_load_pd(mem_addr)3293
+#endif3294
+3295
+SIMDE_FUNCTION_ATTRIBUTES3296
+simde__m128d simde_mm_load1_pd(simde_float64 const *mem_addr)3297
+{3298
+#if defined(SIMDE_X86_SSE2_NATIVE)3299
+ return _mm_load1_pd(mem_addr);3300
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)3301
+ return simde__m128d_from_neon_f64(vld1q_dup_f64(mem_addr));3302
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3303
+ return simde__m128d_from_wasm_v128(wasm_v64x2_load_splat(mem_addr));3304
+#else3305
+ return simde_mm_set1_pd(*mem_addr);3306
+#endif3307
+}3308
+#define simde_mm_load_pd1(mem_addr) simde_mm_load1_pd(mem_addr)3309
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3310
+#define _mm_load_pd1(mem_addr) simde_mm_load1_pd(mem_addr)3311
+#define _mm_load1_pd(mem_addr) simde_mm_load1_pd(mem_addr)3312
+#endif3313
+3314
+SIMDE_FUNCTION_ATTRIBUTES3315
+simde__m128d simde_mm_load_sd(simde_float64 const *mem_addr)3316
+{3317
+#if defined(SIMDE_X86_SSE2_NATIVE)3318
+ return _mm_load_sd(mem_addr);3319
+#else3320
+ simde__m128d_private r_;3321
+3322
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3323
+ r_.neon_f64 = vsetq_lane_f64(*mem_addr, vdupq_n_f64(0), 0);3324
+#else3325
+ r_.f64[0] = *mem_addr;3326
+ r_.u64[1] = UINT64_C(0);3327
+#endif3328
+3329
+ return simde__m128d_from_private(r_);3330
+#endif3331
+}3332
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3333
+#define _mm_load_sd(mem_addr) simde_mm_load_sd(mem_addr)3334
+#endif3335
+3336
+SIMDE_FUNCTION_ATTRIBUTES3337
+simde__m128i simde_mm_load_si128(simde__m128i const *mem_addr)3338
+{3339
+#if defined(SIMDE_X86_SSE2_NATIVE)3340
+ return _mm_load_si128(3341
+ HEDLEY_REINTERPRET_CAST(__m128i const *, mem_addr));3342
+#else3343
+ simde__m128i_private r_;3344
+3345
+#if defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3346
+ r_.altivec_i32 = vec_ld(3347
+ 0, HEDLEY_REINTERPRET_CAST(3348
+ SIMDE_POWER_ALTIVEC_VECTOR(int) const *, mem_addr));3349
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3350
+ r_.neon_i32 =3351
+ vld1q_s32(HEDLEY_REINTERPRET_CAST(int32_t const *, mem_addr));3352
+#else3353
+ simde_memcpy(&r_, SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m128i),3354
+ sizeof(simde__m128i));3355
+#endif3356
+3357
+ return simde__m128i_from_private(r_);3358
+#endif3359
+}3360
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3361
+#define _mm_load_si128(mem_addr) simde_mm_load_si128(mem_addr)3362
+#endif3363
+3364
+SIMDE_FUNCTION_ATTRIBUTES3365
+simde__m128d simde_mm_loadh_pd(simde__m128d a, simde_float64 const *mem_addr)3366
+{3367
+#if defined(SIMDE_X86_SSE2_NATIVE)3368
+ return _mm_loadh_pd(a, mem_addr);3369
+#else3370
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a);3371
+3372
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3373
+ r_.neon_f64 = vcombine_f64(3374
+ vget_low_f64(a_.neon_f64),3375
+ vld1_f64(HEDLEY_REINTERPRET_CAST(const float64_t *, mem_addr)));3376
+#else3377
+ simde_float64 t;3378
+3379
+ simde_memcpy(&t, mem_addr, sizeof(t));3380
+ r_.f64[0] = a_.f64[0];3381
+ r_.f64[1] = t;3382
+#endif3383
+3384
+ return simde__m128d_from_private(r_);3385
+#endif3386
+}3387
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3388
+#define _mm_loadh_pd(a, mem_addr) simde_mm_loadh_pd(a, mem_addr)3389
+#endif3390
+3391
+SIMDE_FUNCTION_ATTRIBUTES3392
+simde__m128i simde_mm_loadl_epi64(simde__m128i const *mem_addr)3393
+{3394
+#if defined(SIMDE_X86_SSE2_NATIVE)3395
+ return _mm_loadl_epi64(mem_addr);3396
+#else3397
+ simde__m128i_private r_;3398
+3399
+ int64_t value;3400
+ simde_memcpy(&value, mem_addr, sizeof(value));3401
+3402
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3403
+ r_.neon_i64 = vcombine_s64(3404
+ vld1_s64(HEDLEY_REINTERPRET_CAST(int64_t const *, mem_addr)),3405
+ vdup_n_s64(0));3406
+#else3407
+ r_.i64[0] = value;3408
+ r_.i64[1] = 0;3409
+#endif3410
+3411
+ return simde__m128i_from_private(r_);3412
+#endif3413
+}3414
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3415
+#define _mm_loadl_epi64(mem_addr) simde_mm_loadl_epi64(mem_addr)3416
+#endif3417
+3418
+SIMDE_FUNCTION_ATTRIBUTES3419
+simde__m128d simde_mm_loadl_pd(simde__m128d a, simde_float64 const *mem_addr)3420
+{3421
+#if defined(SIMDE_X86_SSE2_NATIVE)3422
+ return _mm_loadl_pd(a, mem_addr);3423
+#else3424
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a);3425
+3426
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3427
+ r_.neon_f64 = vcombine_f64(3428
+ vld1_f64(HEDLEY_REINTERPRET_CAST(const float64_t *, mem_addr)),3429
+ vget_high_f64(a_.neon_f64));3430
+#else3431
+ r_.f64[0] = *mem_addr;3432
+ r_.u64[1] = a_.u64[1];3433
+#endif3434
+3435
+ return simde__m128d_from_private(r_);3436
+#endif3437
+}3438
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3439
+#define _mm_loadl_pd(a, mem_addr) simde_mm_loadl_pd(a, mem_addr)3440
+#endif3441
+3442
+SIMDE_FUNCTION_ATTRIBUTES3443
+simde__m128d3444
+simde_mm_loadr_pd(simde_float64 const mem_addr[HEDLEY_ARRAY_PARAM(2)])3445
+{3446
+#if defined(SIMDE_X86_SSE2_NATIVE)3447
+ return _mm_loadr_pd(mem_addr);3448
+#else3449
+ simde__m128d_private r_;3450
+3451
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3452
+ r_.neon_f64 = vld1q_f64(mem_addr);3453
+ r_.neon_f64 = vextq_f64(r_.neon_f64, r_.neon_f64, 1);3454
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3455
+ r_.neon_i64 =3456
+ vld1q_s64(HEDLEY_REINTERPRET_CAST(int64_t const *, mem_addr));3457
+ r_.neon_i64 = vextq_s64(r_.neon_i64, r_.neon_i64, 1);3458
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3459
+ v128_t tmp = wasm_v128_load(mem_addr);3460
+ r_.wasm_v128 = wasm_v64x2_shuffle(tmp, tmp, 1, 0);3461
+#else3462
+ r_.f64[0] = mem_addr[1];3463
+ r_.f64[1] = mem_addr[0];3464
+#endif3465
+3466
+ return simde__m128d_from_private(r_);3467
+#endif3468
+}3469
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3470
+#define _mm_loadr_pd(mem_addr) simde_mm_loadr_pd(mem_addr)3471
+#endif3472
+3473
+SIMDE_FUNCTION_ATTRIBUTES3474
+simde__m128d3475
+simde_mm_loadu_pd(simde_float64 const mem_addr[HEDLEY_ARRAY_PARAM(2)])3476
+{3477
+#if defined(SIMDE_X86_SSE2_NATIVE)3478
+ return _mm_loadu_pd(mem_addr);3479
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)3480
+ return vld1q_f64(mem_addr);3481
+#else3482
+ simde__m128d_private r_;3483
+3484
+ simde_memcpy(&r_, mem_addr, sizeof(r_));3485
+3486
+ return simde__m128d_from_private(r_);3487
+#endif3488
+}3489
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3490
+#define _mm_loadu_pd(mem_addr) simde_mm_loadu_pd(mem_addr)3491
+#endif3492
+3493
+SIMDE_FUNCTION_ATTRIBUTES3494
+simde__m128i simde_x_mm_loadu_epi8(int8_t const *mem_addr)3495
+{3496
+#if defined(SIMDE_X86_SSE2_NATIVE)3497
+ return _mm_loadu_si128(3498
+ SIMDE_ALIGN_CAST(simde__m128i const *, mem_addr));3499
+#else3500
+ simde__m128i_private r_;3501
+3502
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3503
+ r_.neon_i8 =3504
+ vld1q_s8(HEDLEY_REINTERPRET_CAST(int8_t const *, mem_addr));3505
+#else3506
+ simde_memcpy(&r_, mem_addr, sizeof(r_));3507
+#endif3508
+3509
+ return simde__m128i_from_private(r_);3510
+#endif3511
+}3512
+3513
+SIMDE_FUNCTION_ATTRIBUTES3514
+simde__m128i simde_x_mm_loadu_epi16(int16_t const *mem_addr)3515
+{3516
+#if defined(SIMDE_X86_SSE2_NATIVE)3517
+ return _mm_loadu_si128(3518
+ SIMDE_ALIGN_CAST(simde__m128i const *, mem_addr));3519
+#else3520
+ simde__m128i_private r_;3521
+3522
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3523
+ r_.neon_i16 =3524
+ vld1q_s16(HEDLEY_REINTERPRET_CAST(int16_t const *, mem_addr));3525
+#else3526
+ simde_memcpy(&r_, mem_addr, sizeof(r_));3527
+#endif3528
+3529
+ return simde__m128i_from_private(r_);3530
+#endif3531
+}3532
+3533
+SIMDE_FUNCTION_ATTRIBUTES3534
+simde__m128i simde_x_mm_loadu_epi32(int32_t const *mem_addr)3535
+{3536
+#if defined(SIMDE_X86_SSE2_NATIVE)3537
+ return _mm_loadu_si128(3538
+ SIMDE_ALIGN_CAST(simde__m128i const *, mem_addr));3539
+#else3540
+ simde__m128i_private r_;3541
+3542
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3543
+ r_.neon_i32 =3544
+ vld1q_s32(HEDLEY_REINTERPRET_CAST(int32_t const *, mem_addr));3545
+#else3546
+ simde_memcpy(&r_, mem_addr, sizeof(r_));3547
+#endif3548
+3549
+ return simde__m128i_from_private(r_);3550
+#endif3551
+}3552
+3553
+SIMDE_FUNCTION_ATTRIBUTES3554
+simde__m128i simde_x_mm_loadu_epi64(int64_t const *mem_addr)3555
+{3556
+#if defined(SIMDE_X86_SSE2_NATIVE)3557
+ return _mm_loadu_si128(3558
+ SIMDE_ALIGN_CAST(simde__m128i const *, mem_addr));3559
+#else3560
+ simde__m128i_private r_;3561
+3562
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3563
+ r_.neon_i64 =3564
+ vld1q_s64(HEDLEY_REINTERPRET_CAST(int64_t const *, mem_addr));3565
+#else3566
+ simde_memcpy(&r_, mem_addr, sizeof(r_));3567
+#endif3568
+3569
+ return simde__m128i_from_private(r_);3570
+#endif3571
+}3572
+3573
+SIMDE_FUNCTION_ATTRIBUTES3574
+simde__m128i simde_mm_loadu_si128(void const *mem_addr)3575
+{3576
+#if defined(SIMDE_X86_SSE2_NATIVE)3577
+ return _mm_loadu_si128(HEDLEY_STATIC_CAST(__m128i const *, mem_addr));3578
+#else3579
+ simde__m128i_private r_;3580
+3581
+#if HEDLEY_GNUC_HAS_ATTRIBUTE(may_alias, 3, 3, 0)3582
+ HEDLEY_DIAGNOSTIC_PUSH3583
+ SIMDE_DIAGNOSTIC_DISABLE_PACKED_3584
+ struct simde_mm_loadu_si128_s {3585
+ __typeof__(r_) v;3586
+ } __attribute__((__packed__, __may_alias__));3587
+ r_ = HEDLEY_REINTERPRET_CAST(const struct simde_mm_loadu_si128_s *,3588
+ mem_addr)3589
+ ->v;3590
+ HEDLEY_DIAGNOSTIC_POP3591
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3592
+ /* Note that this is a lower priority than the struct above since3593
+ * clang assumes mem_addr is aligned (since it is a __m128i*). */3594
+ r_.neon_i32 =3595
+ vld1q_s32(HEDLEY_REINTERPRET_CAST(int32_t const *, mem_addr));3596
+#else3597
+ simde_memcpy(&r_, mem_addr, sizeof(r_));3598
+#endif3599
+3600
+ return simde__m128i_from_private(r_);3601
+#endif3602
+}3603
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3604
+#define _mm_loadu_si128(mem_addr) simde_mm_loadu_si128(mem_addr)3605
+#endif3606
+3607
+SIMDE_FUNCTION_ATTRIBUTES3608
+simde__m128i simde_mm_madd_epi16(simde__m128i a, simde__m128i b)3609
+{3610
+#if defined(SIMDE_X86_SSE2_NATIVE)3611
+ return _mm_madd_epi16(a, b);3612
+#else3613
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),3614
+ b_ = simde__m128i_to_private(b);3615
+3616
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3617
+ int32x4_t pl =3618
+ vmull_s16(vget_low_s16(a_.neon_i16), vget_low_s16(b_.neon_i16));3619
+ int32x4_t ph = vmull_high_s16(a_.neon_i16, b_.neon_i16);3620
+ r_.neon_i32 = vpaddq_s32(pl, ph);3621
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3622
+ int32x4_t pl =3623
+ vmull_s16(vget_low_s16(a_.neon_i16), vget_low_s16(b_.neon_i16));3624
+ int32x4_t ph = vmull_s16(vget_high_s16(a_.neon_i16),3625
+ vget_high_s16(b_.neon_i16));3626
+ int32x2_t rl = vpadd_s32(vget_low_s32(pl), vget_high_s32(pl));3627
+ int32x2_t rh = vpadd_s32(vget_low_s32(ph), vget_high_s32(ph));3628
+ r_.neon_i32 = vcombine_s32(rl, rh);3629
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)3630
+ static const SIMDE_POWER_ALTIVEC_VECTOR(int) tz = {0, 0, 0, 0};3631
+ r_.altivec_i32 = vec_msum(a_.altivec_i16, b_.altivec_i16, tz);3632
+#else3633
+ SIMDE_VECTORIZE3634
+ for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i16[0])); i += 2) {3635
+ r_.i32[i / 2] = (a_.i16[i] * b_.i16[i]) +3636
+ (a_.i16[i + 1] * b_.i16[i + 1]);3637
+ }3638
+#endif3639
+3640
+ return simde__m128i_from_private(r_);3641
+#endif3642
+}3643
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3644
+#define _mm_madd_epi16(a, b) simde_mm_madd_epi16(a, b)3645
+#endif3646
+3647
+SIMDE_FUNCTION_ATTRIBUTES3648
+void simde_mm_maskmoveu_si128(simde__m128i a, simde__m128i mask,3649
+ int8_t mem_addr[HEDLEY_ARRAY_PARAM(16)])3650
+{3651
+#if defined(SIMDE_X86_SSE2_NATIVE)3652
+ _mm_maskmoveu_si128(a, mask, HEDLEY_REINTERPRET_CAST(char *, mem_addr));3653
+#else3654
+ simde__m128i_private a_ = simde__m128i_to_private(a),3655
+ mask_ = simde__m128i_to_private(mask);3656
+3657
+ for (size_t i = 0; i < (sizeof(a_.i8) / sizeof(a_.i8[0])); i++) {3658
+ if (mask_.u8[i] & 0x80) {3659
+ mem_addr[i] = a_.i8[i];3660
+ }3661
+ }3662
+#endif3663
+}3664
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3665
+#define _mm_maskmoveu_si128(a, mask, mem_addr) \3666
+ simde_mm_maskmoveu_si128( \3667
+ (a), (mask), \3668
+ SIMDE_CHECKED_REINTERPRET_CAST(int8_t *, char *, (mem_addr)))3669
+#endif3670
+3671
+SIMDE_FUNCTION_ATTRIBUTES3672
+int32_t simde_mm_movemask_epi8(simde__m128i a)3673
+{3674
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__INTEL_COMPILER)3675
+ /* ICC has trouble with _mm_movemask_epi8 at -O2 and above: */3676
+ return _mm_movemask_epi8(a);3677
+#else3678
+ int32_t r = 0;3679
+ simde__m128i_private a_ = simde__m128i_to_private(a);3680
+3681
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3682
+ uint8x16_t input = a_.neon_u8;3683
+ const int8_t xr[16] = {-7, -6, -5, -4, -3, -2, -1, 0,3684
+ -7, -6, -5, -4, -3, -2, -1, 0};3685
+ const uint8x16_t mask_and = vdupq_n_u8(0x80);3686
+ const int8x16_t mask_shift = vld1q_s8(xr);3687
+ const uint8x16_t mask_result =3688
+ vshlq_u8(vandq_u8(input, mask_and), mask_shift);3689
+ uint8x8_t lo = vget_low_u8(mask_result);3690
+ uint8x8_t hi = vget_high_u8(mask_result);3691
+ r = vaddv_u8(lo) + (vaddv_u8(hi) << 8);3692
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)3693
+ // Use increasingly wide shifts+adds to collect the sign bits3694
+ // together.3695
+ // Since the widening shifts would be rather confusing to follow in little endian, everything3696
+ // will be illustrated in big endian order instead. This has a different result - the bits3697
+ // would actually be reversed on a big endian machine.3698
+3699
+ // Starting input (only half the elements are shown):3700
+ // 89 ff 1d c0 00 10 99 333701
+ uint8x16_t input = a_.neon_u8;3702
+3703
+ // Shift out everything but the sign bits with an unsigned shift right.3704
+ //3705
+ // Bytes of the vector::3706
+ // 89 ff 1d c0 00 10 99 333707
+ // \ \ \ \ \ \ \ \ high_bits = (uint16x4_t)(input >> 7)3708
+ // | | | | | | | |3709
+ // 01 01 00 01 00 00 01 003710
+ //3711
+ // Bits of first important lane(s):3712
+ // 10001001 (89)3713
+ // \______3714
+ // |3715
+ // 00000001 (01)3716
+ uint16x8_t high_bits = vreinterpretq_u16_u8(vshrq_n_u8(input, 7));3717
+3718
+ // Merge the even lanes together with a 16-bit unsigned shift right + add.3719
+ // 'xx' represents garbage data which will be ignored in the final result.3720
+ // In the important bytes, the add functions like a binary OR.3721
+ //3722
+ // 01 01 00 01 00 00 01 003723
+ // \_ | \_ | \_ | \_ | paired16 = (uint32x4_t)(input + (input >> 7))3724
+ // \| \| \| \|3725
+ // xx 03 xx 01 xx 00 xx 023726
+ //3727
+ // 00000001 00000001 (01 01)3728
+ // \_______ |3729
+ // \|3730
+ // xxxxxxxx xxxxxx11 (xx 03)3731
+ uint32x4_t paired16 =3732
+ vreinterpretq_u32_u16(vsraq_n_u16(high_bits, high_bits, 7));3733
+3734
+ // Repeat with a wider 32-bit shift + add.3735
+ // xx 03 xx 01 xx 00 xx 023736
+ // \____ | \____ | paired32 = (uint64x1_t)(paired16 + (paired16 >> 14))3737
+ // \| \|3738
+ // xx xx xx 0d xx xx xx 023739
+ //3740
+ // 00000011 00000001 (03 01)3741
+ // \\_____ ||3742
+ // '----.\||3743
+ // xxxxxxxx xxxx1101 (xx 0d)3744
+ uint64x2_t paired32 =3745
+ vreinterpretq_u64_u32(vsraq_n_u32(paired16, paired16, 14));3746
+3747
+ // Last, an even wider 64-bit shift + add to get our result in the low 8 bit lanes.3748
+ // xx xx xx 0d xx xx xx 023749
+ // \_________ | paired64 = (uint8x8_t)(paired32 + (paired32 >> 28))3750
+ // \|3751
+ // xx xx xx xx xx xx xx d23752
+ //3753
+ // 00001101 00000010 (0d 02)3754
+ // \ \___ | |3755
+ // '---. \| |3756
+ // xxxxxxxx 11010010 (xx d2)3757
+ uint8x16_t paired64 =3758
+ vreinterpretq_u8_u64(vsraq_n_u64(paired32, paired32, 28));3759
+3760
+ // Extract the low 8 bits from each 64-bit lane with 2 8-bit extracts.3761
+ // xx xx xx xx xx xx xx d23762
+ // || return paired64[0]3763
+ // d23764
+ // Note: Little endian would return the correct value 4b (01001011) instead.3765
+ r = vgetq_lane_u8(paired64, 0) |3766
+ (HEDLEY_STATIC_CAST(int32_t, vgetq_lane_u8(paired64, 8)) << 8);3767
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE) && \3768
+ !defined(HEDLEY_IBM_VERSION) && \3769
+ (SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_LITTLE)3770
+ static const SIMDE_POWER_ALTIVEC_VECTOR(unsigned char)3771
+ perm = {120, 112, 104, 96, 88, 80, 72, 64,3772
+ 56, 48, 40, 32, 24, 16, 8, 0};3773
+ r = HEDLEY_STATIC_CAST(3774
+ int32_t, vec_extract(vec_vbpermq(a_.altivec_u8, perm), 1));3775
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE) && \3776
+ !defined(HEDLEY_IBM_VERSION) && \3777
+ (SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_BIG)3778
+ static const SIMDE_POWER_ALTIVEC_VECTOR(unsigned char)3779
+ perm = {120, 112, 104, 96, 88, 80, 72, 64,3780
+ 56, 48, 40, 32, 24, 16, 8, 0};3781
+ r = HEDLEY_STATIC_CAST(3782
+ int32_t, vec_extract(vec_vbpermq(a_.altivec_u8, perm), 14));3783
+#else3784
+ SIMDE_VECTORIZE_REDUCTION(| : r)3785
+ for (size_t i = 0; i < (sizeof(a_.u8) / sizeof(a_.u8[0])); i++) {3786
+ r |= (a_.u8[15 - i] >> 7) << (15 - i);3787
+ }3788
+#endif3789
+3790
+ return r;3791
+#endif3792
+}3793
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3794
+#define _mm_movemask_epi8(a) simde_mm_movemask_epi8(a)3795
+#endif3796
+3797
+SIMDE_FUNCTION_ATTRIBUTES3798
+int32_t simde_mm_movemask_pd(simde__m128d a)3799
+{3800
+#if defined(SIMDE_X86_SSE2_NATIVE)3801
+ return _mm_movemask_pd(a);3802
+#else3803
+ int32_t r = 0;3804
+ simde__m128d_private a_ = simde__m128d_to_private(a);3805
+3806
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3807
+ static const int64_t shift_amount[] = {0, 1};3808
+ const int64x2_t shift = vld1q_s64(shift_amount);3809
+ uint64x2_t tmp = vshrq_n_u64(a_.neon_u64, 63);3810
+ return HEDLEY_STATIC_CAST(int32_t, vaddvq_u64(vshlq_u64(tmp, shift)));3811
+#else3812
+ SIMDE_VECTORIZE_REDUCTION(| : r)3813
+ for (size_t i = 0; i < (sizeof(a_.u64) / sizeof(a_.u64[0])); i++) {3814
+ r |= (a_.u64[i] >> 63) << i;3815
+ }3816
+#endif3817
+3818
+ return r;3819
+#endif3820
+}3821
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3822
+#define _mm_movemask_pd(a) simde_mm_movemask_pd(a)3823
+#endif3824
+3825
+SIMDE_FUNCTION_ATTRIBUTES3826
+simde__m64 simde_mm_movepi64_pi64(simde__m128i a)3827
+{3828
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3829
+ return _mm_movepi64_pi64(a);3830
+#else3831
+ simde__m64_private r_;3832
+ simde__m128i_private a_ = simde__m128i_to_private(a);3833
+3834
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3835
+ r_.neon_i64 = vget_low_s64(a_.neon_i64);3836
+#else3837
+ r_.i64[0] = a_.i64[0];3838
+#endif3839
+3840
+ return simde__m64_from_private(r_);3841
+#endif3842
+}3843
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3844
+#define _mm_movepi64_pi64(a) simde_mm_movepi64_pi64(a)3845
+#endif3846
+3847
+SIMDE_FUNCTION_ATTRIBUTES3848
+simde__m128i simde_mm_movpi64_epi64(simde__m64 a)3849
+{3850
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)3851
+ return _mm_movpi64_epi64(a);3852
+#else3853
+ simde__m128i_private r_;3854
+ simde__m64_private a_ = simde__m64_to_private(a);3855
+3856
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3857
+ r_.neon_i64 = vcombine_s64(a_.neon_i64, vdup_n_s64(0));3858
+#else3859
+ r_.i64[0] = a_.i64[0];3860
+ r_.i64[1] = 0;3861
+#endif3862
+3863
+ return simde__m128i_from_private(r_);3864
+#endif3865
+}3866
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3867
+#define _mm_movpi64_epi64(a) simde_mm_movpi64_epi64(a)3868
+#endif3869
+3870
+SIMDE_FUNCTION_ATTRIBUTES3871
+simde__m128i simde_mm_min_epi16(simde__m128i a, simde__m128i b)3872
+{3873
+#if defined(SIMDE_X86_SSE2_NATIVE)3874
+ return _mm_min_epi16(a, b);3875
+#else3876
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),3877
+ b_ = simde__m128i_to_private(b);3878
+3879
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3880
+ r_.neon_i16 = vminq_s16(a_.neon_i16, b_.neon_i16);3881
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3882
+ r_.wasm_v128 = wasm_i16x8_min(a_.wasm_v128, b_.wasm_v128);3883
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3884
+ r_.altivec_i16 = vec_min(a_.altivec_i16, b_.altivec_i16);3885
+#else3886
+ SIMDE_VECTORIZE3887
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {3888
+ r_.i16[i] = (a_.i16[i] < b_.i16[i]) ? a_.i16[i] : b_.i16[i];3889
+ }3890
+#endif3891
+3892
+ return simde__m128i_from_private(r_);3893
+#endif3894
+}3895
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3896
+#define _mm_min_epi16(a, b) simde_mm_min_epi16(a, b)3897
+#endif3898
+3899
+SIMDE_FUNCTION_ATTRIBUTES3900
+simde__m128i simde_mm_min_epu8(simde__m128i a, simde__m128i b)3901
+{3902
+#if defined(SIMDE_X86_SSE2_NATIVE)3903
+ return _mm_min_epu8(a, b);3904
+#else3905
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),3906
+ b_ = simde__m128i_to_private(b);3907
+3908
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3909
+ r_.neon_u8 = vminq_u8(a_.neon_u8, b_.neon_u8);3910
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3911
+ r_.wasm_v128 = wasm_u8x16_min(a_.wasm_v128, b_.wasm_v128);3912
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3913
+ r_.altivec_u8 = vec_min(a_.altivec_u8, b_.altivec_u8);3914
+#else3915
+ SIMDE_VECTORIZE3916
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {3917
+ r_.u8[i] = (a_.u8[i] < b_.u8[i]) ? a_.u8[i] : b_.u8[i];3918
+ }3919
+#endif3920
+3921
+ return simde__m128i_from_private(r_);3922
+#endif3923
+}3924
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3925
+#define _mm_min_epu8(a, b) simde_mm_min_epu8(a, b)3926
+#endif3927
+3928
+SIMDE_FUNCTION_ATTRIBUTES3929
+simde__m128d simde_mm_min_pd(simde__m128d a, simde__m128d b)3930
+{3931
+#if defined(SIMDE_X86_SSE2_NATIVE)3932
+ return _mm_min_pd(a, b);3933
+#else3934
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),3935
+ b_ = simde__m128d_to_private(b);3936
+3937
+#if defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)3938
+ r_.altivec_f64 = vec_min(a_.altivec_f64, b_.altivec_f64);3939
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)3940
+ r_.neon_f64 = vminq_f64(a_.neon_f64, b_.neon_f64);3941
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3942
+ r_.wasm_v128 = wasm_f64x2_min(a_.wasm_v128, b_.wasm_v128);3943
+#else3944
+ SIMDE_VECTORIZE3945
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {3946
+ r_.f64[i] = (a_.f64[i] < b_.f64[i]) ? a_.f64[i] : b_.f64[i];3947
+ }3948
+#endif3949
+3950
+ return simde__m128d_from_private(r_);3951
+#endif3952
+}3953
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3954
+#define _mm_min_pd(a, b) simde_mm_min_pd(a, b)3955
+#endif3956
+3957
+SIMDE_FUNCTION_ATTRIBUTES3958
+simde__m128d simde_mm_min_sd(simde__m128d a, simde__m128d b)3959
+{3960
+#if defined(SIMDE_X86_SSE2_NATIVE)3961
+ return _mm_min_sd(a, b);3962
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)3963
+ return simde_mm_move_sd(a, simde_mm_min_pd(a, b));3964
+#else3965
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),3966
+ b_ = simde__m128d_to_private(b);3967
+3968
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)3969
+ float64x2_t temp = vminq_f64(a_.neon_f64, b_.neon_f64);3970
+ r_.neon_f64 = vsetq_lane_f64(vgetq_lane(a_.neon_f64, 1), temp, 1);3971
+#else3972
+ r_.f64[0] = (a_.f64[0] < b_.f64[0]) ? a_.f64[0] : b_.f64[0];3973
+ r_.f64[1] = a_.f64[1];3974
+#endif3975
+3976
+ return simde__m128d_from_private(r_);3977
+#endif3978
+}3979
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)3980
+#define _mm_min_sd(a, b) simde_mm_min_sd(a, b)3981
+#endif3982
+3983
+SIMDE_FUNCTION_ATTRIBUTES3984
+simde__m128i simde_mm_max_epi16(simde__m128i a, simde__m128i b)3985
+{3986
+#if defined(SIMDE_X86_SSE2_NATIVE)3987
+ return _mm_max_epi16(a, b);3988
+#else3989
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),3990
+ b_ = simde__m128i_to_private(b);3991
+3992
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)3993
+ r_.neon_i16 = vmaxq_s16(a_.neon_i16, b_.neon_i16);3994
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)3995
+ r_.wasm_v128 = wasm_i16x8_max(a_.wasm_v128, b_.wasm_v128);3996
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)3997
+ r_.altivec_i16 = vec_max(a_.altivec_i16, b_.altivec_i16);3998
+#else3999
+ SIMDE_VECTORIZE4000
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4001
+ r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? a_.i16[i] : b_.i16[i];4002
+ }4003
+#endif4004
+4005
+ return simde__m128i_from_private(r_);4006
+#endif4007
+}4008
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4009
+#define _mm_max_epi16(a, b) simde_mm_max_epi16(a, b)4010
+#endif4011
+4012
+SIMDE_FUNCTION_ATTRIBUTES4013
+simde__m128i simde_mm_max_epu8(simde__m128i a, simde__m128i b)4014
+{4015
+#if defined(SIMDE_X86_SSE2_NATIVE)4016
+ return _mm_max_epu8(a, b);4017
+#else4018
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4019
+ b_ = simde__m128i_to_private(b);4020
+4021
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4022
+ r_.neon_u8 = vmaxq_u8(a_.neon_u8, b_.neon_u8);4023
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)4024
+ r_.wasm_v128 = wasm_u8x16_max(a_.wasm_v128, b_.wasm_v128);4025
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)4026
+ r_.altivec_u8 = vec_max(a_.altivec_u8, b_.altivec_u8);4027
+#else4028
+ SIMDE_VECTORIZE4029
+ for (size_t i = 0; i < (sizeof(r_.u8) / sizeof(r_.u8[0])); i++) {4030
+ r_.u8[i] = (a_.u8[i] > b_.u8[i]) ? a_.u8[i] : b_.u8[i];4031
+ }4032
+#endif4033
+4034
+ return simde__m128i_from_private(r_);4035
+#endif4036
+}4037
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4038
+#define _mm_max_epu8(a, b) simde_mm_max_epu8(a, b)4039
+#endif4040
+4041
+SIMDE_FUNCTION_ATTRIBUTES4042
+simde__m128d simde_mm_max_pd(simde__m128d a, simde__m128d b)4043
+{4044
+#if defined(SIMDE_X86_SSE2_NATIVE)4045
+ return _mm_max_pd(a, b);4046
+#else4047
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),4048
+ b_ = simde__m128d_to_private(b);4049
+4050
+#if defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)4051
+ r_.altivec_f64 = vec_max(a_.altivec_f64, b_.altivec_f64);4052
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)4053
+ r_.wasm_v128 = wasm_f64x2_max(a_.wasm_v128, b_.wasm_v128);4054
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)4055
+ r_.neon_f64 = vmaxq_f64(a_.neon_f64, b_.neon_f64);4056
+#else4057
+ SIMDE_VECTORIZE4058
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {4059
+ r_.f64[i] = (a_.f64[i] > b_.f64[i]) ? a_.f64[i] : b_.f64[i];4060
+ }4061
+#endif4062
+4063
+ return simde__m128d_from_private(r_);4064
+#endif4065
+}4066
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4067
+#define _mm_max_pd(a, b) simde_mm_max_pd(a, b)4068
+#endif4069
+4070
+SIMDE_FUNCTION_ATTRIBUTES4071
+simde__m128d simde_mm_max_sd(simde__m128d a, simde__m128d b)4072
+{4073
+#if defined(SIMDE_X86_SSE2_NATIVE)4074
+ return _mm_max_sd(a, b);4075
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)4076
+ return simde_mm_move_sd(a, simde_mm_max_pd(a, b));4077
+#else4078
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),4079
+ b_ = simde__m128d_to_private(b);4080
+4081
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)4082
+ float64x2_t temp = vmaxq_f64(a_.neon_f64, b_.neon_f64);4083
+ r_.neon_f64 = vsetq_lane_f64(vgetq_lane(a_.neon_f64, 1), temp, 1);4084
+#else4085
+ r_.f64[0] = (a_.f64[0] > b_.f64[0]) ? a_.f64[0] : b_.f64[0];4086
+ r_.f64[1] = a_.f64[1];4087
+#endif4088
+4089
+ return simde__m128d_from_private(r_);4090
+#endif4091
+}4092
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4093
+#define _mm_max_sd(a, b) simde_mm_max_sd(a, b)4094
+#endif4095
+4096
+SIMDE_FUNCTION_ATTRIBUTES4097
+simde__m128i simde_mm_move_epi64(simde__m128i a)4098
+{4099
+#if defined(SIMDE_X86_SSE2_NATIVE)4100
+ return _mm_move_epi64(a);4101
+#else4102
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);4103
+4104
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4105
+ r_.neon_i64 = vsetq_lane_s64(0, a_.neon_i64, 1);4106
+#else4107
+ r_.i64[0] = a_.i64[0];4108
+ r_.i64[1] = 0;4109
+#endif4110
+4111
+ return simde__m128i_from_private(r_);4112
+#endif4113
+}4114
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4115
+#define _mm_move_epi64(a) simde_mm_move_epi64(a)4116
+#endif4117
+4118
+SIMDE_FUNCTION_ATTRIBUTES4119
+simde__m128i simde_mm_mul_epu32(simde__m128i a, simde__m128i b)4120
+{4121
+#if defined(SIMDE_X86_SSE2_NATIVE)4122
+ return _mm_mul_epu32(a, b);4123
+#else4124
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4125
+ b_ = simde__m128i_to_private(b);4126
+4127
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4128
+ uint32x2_t a_lo = vmovn_u64(a_.neon_u64);4129
+ uint32x2_t b_lo = vmovn_u64(b_.neon_u64);4130
+ r_.neon_u64 = vmull_u32(a_lo, b_lo);4131
+#elif defined(SIMDE_SHUFFLE_VECTOR_) && \4132
+ (SIMDE_ENDIAN_ORDER == SIMDE_ENDIAN_LITTLE)4133
+ __typeof__(a_.u32) z = {4134
+ 0,4135
+ };4136
+ a_.u32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.u32, z, 0, 4, 2, 6);4137
+ b_.u32 = SIMDE_SHUFFLE_VECTOR_(32, 16, b_.u32, z, 0, 4, 2, 6);4138
+ r_.u64 = HEDLEY_REINTERPRET_CAST(__typeof__(r_.u64), a_.u32) *4139
+ HEDLEY_REINTERPRET_CAST(__typeof__(r_.u64), b_.u32);4140
+#else4141
+ SIMDE_VECTORIZE4142
+ for (size_t i = 0; i < (sizeof(r_.u64) / sizeof(r_.u64[0])); i++) {4143
+ r_.u64[i] = HEDLEY_STATIC_CAST(uint64_t, a_.u32[i * 2]) *4144
+ HEDLEY_STATIC_CAST(uint64_t, b_.u32[i * 2]);4145
+ }4146
+#endif4147
+4148
+ return simde__m128i_from_private(r_);4149
+#endif4150
+}4151
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4152
+#define _mm_mul_epu32(a, b) simde_mm_mul_epu32(a, b)4153
+#endif4154
+4155
+SIMDE_FUNCTION_ATTRIBUTES4156
+simde__m128i simde_x_mm_mul_epi64(simde__m128i a, simde__m128i b)4157
+{4158
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4159
+ b_ = simde__m128i_to_private(b);4160
+4161
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)4162
+ r_.i64 = a_.i64 * b_.i64;4163
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)4164
+ r_.neon_f64 = vmulq_s64(a_.neon_f64, b_.neon_f64);4165
+#else4166
+ SIMDE_VECTORIZE4167
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {4168
+ r_.i64[i] = a_.i64[i] * b_.i64[i];4169
+ }4170
+#endif4171
+4172
+ return simde__m128i_from_private(r_);4173
+}4174
+4175
+SIMDE_FUNCTION_ATTRIBUTES4176
+simde__m128i simde_x_mm_mod_epi64(simde__m128i a, simde__m128i b)4177
+{4178
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4179
+ b_ = simde__m128i_to_private(b);4180
+4181
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)4182
+ r_.i64 = a_.i64 % b_.i64;4183
+#else4184
+ SIMDE_VECTORIZE4185
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {4186
+ r_.i64[i] = a_.i64[i] % b_.i64[i];4187
+ }4188
+#endif4189
+4190
+ return simde__m128i_from_private(r_);4191
+}4192
+4193
+SIMDE_FUNCTION_ATTRIBUTES4194
+simde__m128d simde_mm_mul_pd(simde__m128d a, simde__m128d b)4195
+{4196
+#if defined(SIMDE_X86_SSE2_NATIVE)4197
+ return _mm_mul_pd(a, b);4198
+#else4199
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),4200
+ b_ = simde__m128d_to_private(b);4201
+4202
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)4203
+ r_.f64 = a_.f64 * b_.f64;4204
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)4205
+ r_.neon_f64 = vmulq_f64(a_.neon_f64, b_.neon_f64);4206
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)4207
+ r_.wasm_v128 = wasm_f64x2_mul(a_.wasm_v128, b_.wasm_v128);4208
+#else4209
+ SIMDE_VECTORIZE4210
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {4211
+ r_.f64[i] = a_.f64[i] * b_.f64[i];4212
+ }4213
+#endif4214
+4215
+ return simde__m128d_from_private(r_);4216
+#endif4217
+}4218
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4219
+#define _mm_mul_pd(a, b) simde_mm_mul_pd(a, b)4220
+#endif4221
+4222
+SIMDE_FUNCTION_ATTRIBUTES4223
+simde__m128d simde_mm_mul_sd(simde__m128d a, simde__m128d b)4224
+{4225
+#if defined(SIMDE_X86_SSE2_NATIVE)4226
+ return _mm_mul_sd(a, b);4227
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)4228
+ return simde_mm_move_sd(a, simde_mm_mul_pd(a, b));4229
+#else4230
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),4231
+ b_ = simde__m128d_to_private(b);4232
+4233
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)4234
+ float64x2_t temp = vmulq_f64(a_.neon_f64, b_.neon_f64);4235
+ r_.neon_f64 = vsetq_lane_f64(vgetq_lane(a_.neon_f64, 1), temp, 1);4236
+#else4237
+ r_.f64[0] = a_.f64[0] * b_.f64[0];4238
+ r_.f64[1] = a_.f64[1];4239
+#endif4240
+4241
+ return simde__m128d_from_private(r_);4242
+#endif4243
+}4244
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4245
+#define _mm_mul_sd(a, b) simde_mm_mul_sd(a, b)4246
+#endif4247
+4248
+SIMDE_FUNCTION_ATTRIBUTES4249
+simde__m64 simde_mm_mul_su32(simde__m64 a, simde__m64 b)4250
+{4251
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE) && \4252
+ !defined(__PGI)4253
+ return _mm_mul_su32(a, b);4254
+#else4255
+ simde__m64_private r_, a_ = simde__m64_to_private(a),4256
+ b_ = simde__m64_to_private(b);4257
+4258
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4259
+ r_.u64[0] = vget_lane_u64(4260
+ vget_low_u64(vmull_u32(vreinterpret_u32_s64(a_.neon_i64),4261
+ vreinterpret_u32_s64(b_.neon_i64))),4262
+ 0);4263
+#else4264
+ r_.u64[0] = HEDLEY_STATIC_CAST(uint64_t, a_.u32[0]) *4265
+ HEDLEY_STATIC_CAST(uint64_t, b_.u32[0]);4266
+#endif4267
+4268
+ return simde__m64_from_private(r_);4269
+#endif4270
+}4271
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4272
+#define _mm_mul_su32(a, b) simde_mm_mul_su32(a, b)4273
+#endif4274
+4275
+SIMDE_FUNCTION_ATTRIBUTES4276
+simde__m128i simde_mm_mulhi_epi16(simde__m128i a, simde__m128i b)4277
+{4278
+#if defined(SIMDE_X86_SSE2_NATIVE)4279
+ return _mm_mulhi_epi16(a, b);4280
+#else4281
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4282
+ b_ = simde__m128i_to_private(b);4283
+4284
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4285
+ int16x4_t a3210 = vget_low_s16(a_.neon_i16);4286
+ int16x4_t b3210 = vget_low_s16(b_.neon_i16);4287
+ int32x4_t ab3210 = vmull_s16(a3210, b3210); /* 3333222211110000 */4288
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)4289
+ int32x4_t ab7654 = vmull_high_s16(a_.neon_i16, b_.neon_i16);4290
+ r_.neon_i16 = vuzp2q_s16(vreinterpretq_s16_s32(ab3210),4291
+ vreinterpretq_s16_s32(ab7654));4292
+#else4293
+ int16x4_t a7654 = vget_high_s16(a_.neon_i16);4294
+ int16x4_t b7654 = vget_high_s16(b_.neon_i16);4295
+ int32x4_t ab7654 = vmull_s16(a7654, b7654); /* 7777666655554444 */4296
+ uint16x8x2_t rv = vuzpq_u16(vreinterpretq_u16_s32(ab3210),4297
+ vreinterpretq_u16_s32(ab7654));4298
+ r_.neon_u16 = rv.val[1];4299
+#endif4300
+#else4301
+ SIMDE_VECTORIZE4302
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4303
+ r_.u16[i] = HEDLEY_STATIC_CAST(4304
+ uint16_t,4305
+ (HEDLEY_STATIC_CAST(4306
+ uint32_t,4307
+ HEDLEY_STATIC_CAST(int32_t, a_.i16[i]) *4308
+ HEDLEY_STATIC_CAST(int32_t,4309
+ b_.i16[i])) >>4310
+ 16));4311
+ }4312
+#endif4313
+4314
+ return simde__m128i_from_private(r_);4315
+#endif4316
+}4317
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4318
+#define _mm_mulhi_epi16(a, b) simde_mm_mulhi_epi16(a, b)4319
+#endif4320
+4321
+SIMDE_FUNCTION_ATTRIBUTES4322
+simde__m128i simde_mm_mulhi_epu16(simde__m128i a, simde__m128i b)4323
+{4324
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)4325
+ return _mm_mulhi_epu16(a, b);4326
+#else4327
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4328
+ b_ = simde__m128i_to_private(b);4329
+4330
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4331
+ uint16x4_t a3210 = vget_low_u16(a_.neon_u16);4332
+ uint16x4_t b3210 = vget_low_u16(b_.neon_u16);4333
+ uint32x4_t ab3210 = vmull_u16(a3210, b3210); /* 3333222211110000 */4334
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)4335
+ uint32x4_t ab7654 = vmull_high_u16(a_.neon_u16, b_.neon_u16);4336
+ r_.neon_u16 = vuzp2q_u16(vreinterpretq_u16_u32(ab3210),4337
+ vreinterpretq_u16_u32(ab7654));4338
+#else4339
+ uint16x4_t a7654 = vget_high_u16(a_.neon_u16);4340
+ uint16x4_t b7654 = vget_high_u16(b_.neon_u16);4341
+ uint32x4_t ab7654 = vmull_u16(a7654, b7654); /* 7777666655554444 */4342
+ uint16x8x2_t neon_r = vuzpq_u16(vreinterpretq_u16_u32(ab3210),4343
+ vreinterpretq_u16_u32(ab7654));4344
+ r_.neon_u16 = neon_r.val[1];4345
+#endif4346
+#else4347
+ SIMDE_VECTORIZE4348
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {4349
+ r_.u16[i] = HEDLEY_STATIC_CAST(4350
+ uint16_t,4351
+ HEDLEY_STATIC_CAST(uint32_t, a_.u16[i]) *4352
+ HEDLEY_STATIC_CAST(uint32_t,4353
+ b_.u16[i]) >>4354
+ 16);4355
+ }4356
+#endif4357
+4358
+ return simde__m128i_from_private(r_);4359
+#endif4360
+}4361
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4362
+#define _mm_mulhi_epu16(a, b) simde_mm_mulhi_epu16(a, b)4363
+#endif4364
+4365
+SIMDE_FUNCTION_ATTRIBUTES4366
+simde__m128i simde_mm_mullo_epi16(simde__m128i a, simde__m128i b)4367
+{4368
+#if defined(SIMDE_X86_SSE2_NATIVE)4369
+ return _mm_mullo_epi16(a, b);4370
+#else4371
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4372
+ b_ = simde__m128i_to_private(b);4373
+4374
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4375
+ r_.neon_i16 = vmulq_s16(a_.neon_i16, b_.neon_i16);4376
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)4377
+ (void)a_;4378
+ (void)b_;4379
+ r_.altivec_i16 = vec_mul(a_.altivec_i16, b_.altivec_i16);4380
+#else4381
+ SIMDE_VECTORIZE4382
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4383
+ r_.u16[i] = HEDLEY_STATIC_CAST(4384
+ uint16_t,4385
+ HEDLEY_STATIC_CAST(uint32_t, a_.u16[i]) *4386
+ HEDLEY_STATIC_CAST(uint32_t, b_.u16[i]));4387
+ }4388
+#endif4389
+4390
+ return simde__m128i_from_private(r_);4391
+#endif4392
+}4393
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4394
+#define _mm_mullo_epi16(a, b) simde_mm_mullo_epi16(a, b)4395
+#endif4396
+4397
+SIMDE_FUNCTION_ATTRIBUTES4398
+simde__m128d simde_mm_or_pd(simde__m128d a, simde__m128d b)4399
+{4400
+#if defined(SIMDE_X86_SSE2_NATIVE)4401
+ return _mm_or_pd(a, b);4402
+#else4403
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),4404
+ b_ = simde__m128d_to_private(b);4405
+4406
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)4407
+ r_.i32f = a_.i32f | b_.i32f;4408
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)4409
+ r_.wasm_v128 = wasm_v128_or(a_.wasm_v128, b_.wasm_v128);4410
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)4411
+ r_.neon_i64 = vorrq_s64(a_.neon_i64, b_.neon_i64);4412
+#else4413
+ SIMDE_VECTORIZE4414
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {4415
+ r_.i32f[i] = a_.i32f[i] | b_.i32f[i];4416
+ }4417
+#endif4418
+4419
+ return simde__m128d_from_private(r_);4420
+#endif4421
+}4422
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4423
+#define _mm_or_pd(a, b) simde_mm_or_pd(a, b)4424
+#endif4425
+4426
+SIMDE_FUNCTION_ATTRIBUTES4427
+simde__m128i simde_mm_or_si128(simde__m128i a, simde__m128i b)4428
+{4429
+#if defined(SIMDE_X86_SSE2_NATIVE)4430
+ return _mm_or_si128(a, b);4431
+#else4432
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4433
+ b_ = simde__m128i_to_private(b);4434
+4435
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4436
+ r_.neon_i32 = vorrq_s32(a_.neon_i32, b_.neon_i32);4437
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)4438
+ r_.altivec_i32 = vec_or(a_.altivec_i32, b_.altivec_i32);4439
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)4440
+ r_.i32f = a_.i32f | b_.i32f;4441
+#else4442
+ SIMDE_VECTORIZE4443
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {4444
+ r_.i32f[i] = a_.i32f[i] | b_.i32f[i];4445
+ }4446
+#endif4447
+4448
+ return simde__m128i_from_private(r_);4449
+#endif4450
+}4451
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4452
+#define _mm_or_si128(a, b) simde_mm_or_si128(a, b)4453
+#endif4454
+4455
+SIMDE_FUNCTION_ATTRIBUTES4456
+simde__m128i simde_mm_packs_epi16(simde__m128i a, simde__m128i b)4457
+{4458
+#if defined(SIMDE_X86_SSE2_NATIVE)4459
+ return _mm_packs_epi16(a, b);4460
+#else4461
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4462
+ b_ = simde__m128i_to_private(b);4463
+4464
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4465
+ r_.neon_i8 =4466
+ vcombine_s8(vqmovn_s16(a_.neon_i16), vqmovn_s16(b_.neon_i16));4467
+#else4468
+ SIMDE_VECTORIZE4469
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4470
+ r_.i8[i] = (a_.i16[i] > INT8_MAX)4471
+ ? INT8_MAX4472
+ : ((a_.i16[i] < INT8_MIN)4473
+ ? INT8_MIN4474
+ : HEDLEY_STATIC_CAST(int8_t,4475
+ a_.i16[i]));4476
+ r_.i8[i + 8] = (b_.i16[i] > INT8_MAX)4477
+ ? INT8_MAX4478
+ : ((b_.i16[i] < INT8_MIN)4479
+ ? INT8_MIN4480
+ : HEDLEY_STATIC_CAST(4481
+ int8_t, b_.i16[i]));4482
+ }4483
+#endif4484
+4485
+ return simde__m128i_from_private(r_);4486
+#endif4487
+}4488
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4489
+#define _mm_packs_epi16(a, b) simde_mm_packs_epi16(a, b)4490
+#endif4491
+4492
+SIMDE_FUNCTION_ATTRIBUTES4493
+simde__m128i simde_mm_packs_epi32(simde__m128i a, simde__m128i b)4494
+{4495
+#if defined(SIMDE_X86_SSE2_NATIVE)4496
+ return _mm_packs_epi32(a, b);4497
+#else4498
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4499
+ b_ = simde__m128i_to_private(b);4500
+4501
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4502
+ r_.neon_i16 =4503
+ vcombine_s16(vqmovn_s32(a_.neon_i32), vqmovn_s32(b_.neon_i32));4504
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)4505
+ r_.altivec_i16 = vec_packs(a_.altivec_i32, b_.altivec_i32);4506
+#else4507
+ SIMDE_VECTORIZE4508
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {4509
+ r_.i16[i] = (a_.i32[i] > INT16_MAX)4510
+ ? INT16_MAX4511
+ : ((a_.i32[i] < INT16_MIN)4512
+ ? INT16_MIN4513
+ : HEDLEY_STATIC_CAST(int16_t,4514
+ a_.i32[i]));4515
+ r_.i16[i + 4] =4516
+ (b_.i32[i] > INT16_MAX)4517
+ ? INT16_MAX4518
+ : ((b_.i32[i] < INT16_MIN)4519
+ ? INT16_MIN4520
+ : HEDLEY_STATIC_CAST(int16_t,4521
+ b_.i32[i]));4522
+ }4523
+#endif4524
+4525
+ return simde__m128i_from_private(r_);4526
+#endif4527
+}4528
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4529
+#define _mm_packs_epi32(a, b) simde_mm_packs_epi32(a, b)4530
+#endif4531
+4532
+SIMDE_FUNCTION_ATTRIBUTES4533
+simde__m128i simde_mm_packus_epi16(simde__m128i a, simde__m128i b)4534
+{4535
+#if defined(SIMDE_X86_SSE2_NATIVE)4536
+ return _mm_packus_epi16(a, b);4537
+#else4538
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4539
+ b_ = simde__m128i_to_private(b);4540
+4541
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4542
+ r_.neon_u8 =4543
+ vcombine_u8(vqmovun_s16(a_.neon_i16), vqmovun_s16(b_.neon_i16));4544
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)4545
+ r_.altivec_u8 = vec_packsu(a_.altivec_i16, b_.altivec_i16);4546
+#else4547
+ SIMDE_VECTORIZE4548
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {4549
+ r_.u8[i] = (a_.i16[i] > UINT8_MAX)4550
+ ? UINT8_MAX4551
+ : ((a_.i16[i] < 0)4552
+ ? UINT8_C(0)4553
+ : HEDLEY_STATIC_CAST(uint8_t,4554
+ a_.i16[i]));4555
+ r_.u8[i + 8] =4556
+ (b_.i16[i] > UINT8_MAX)4557
+ ? UINT8_MAX4558
+ : ((b_.i16[i] < 0)4559
+ ? UINT8_C(0)4560
+ : HEDLEY_STATIC_CAST(uint8_t,4561
+ b_.i16[i]));4562
+ }4563
+#endif4564
+4565
+ return simde__m128i_from_private(r_);4566
+#endif4567
+}4568
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4569
+#define _mm_packus_epi16(a, b) simde_mm_packus_epi16(a, b)4570
+#endif4571
+4572
+SIMDE_FUNCTION_ATTRIBUTES4573
+void simde_mm_pause(void)4574
+{4575
+#if defined(SIMDE_X86_SSE2_NATIVE)4576
+ _mm_pause();4577
+#endif4578
+}4579
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4580
+#define _mm_pause() (simde_mm_pause())4581
+#endif4582
+4583
+SIMDE_FUNCTION_ATTRIBUTES4584
+simde__m128i simde_mm_sad_epu8(simde__m128i a, simde__m128i b)4585
+{4586
+#if defined(SIMDE_X86_SSE2_NATIVE)4587
+ return _mm_sad_epu8(a, b);4588
+#else4589
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),4590
+ b_ = simde__m128i_to_private(b);4591
+4592
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4593
+ const uint16x8_t t = vpaddlq_u8(vabdq_u8(a_.neon_u8, b_.neon_u8));4594
+ r_.neon_u64 = vcombine_u64(vpaddl_u32(vpaddl_u16(vget_low_u16(t))),4595
+ vpaddl_u32(vpaddl_u16(vget_high_u16(t))));4596
+#else4597
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {4598
+ uint16_t tmp = 0;4599
+ SIMDE_VECTORIZE_REDUCTION(+ : tmp)4600
+ for (size_t j = 0; j < ((sizeof(r_.u8) / sizeof(r_.u8[0])) / 2);4601
+ j++) {4602
+ const size_t e = j + (i * 8);4603
+ tmp += (a_.u8[e] > b_.u8[e]) ? (a_.u8[e] - b_.u8[e])4604
+ : (b_.u8[e] - a_.u8[e]);4605
+ }4606
+ r_.i64[i] = tmp;4607
+ }4608
+#endif4609
+4610
+ return simde__m128i_from_private(r_);4611
+#endif4612
+}4613
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4614
+#define _mm_sad_epu8(a, b) simde_mm_sad_epu8(a, b)4615
+#endif4616
+4617
+SIMDE_FUNCTION_ATTRIBUTES4618
+simde__m128i simde_mm_set_epi8(int8_t e15, int8_t e14, int8_t e13, int8_t e12,4619
+ int8_t e11, int8_t e10, int8_t e9, int8_t e8,4620
+ int8_t e7, int8_t e6, int8_t e5, int8_t e4,4621
+ int8_t e3, int8_t e2, int8_t e1, int8_t e0)4622
+{4623
+4624
+#if defined(SIMDE_X86_SSE2_NATIVE)4625
+ return _mm_set_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5,4626
+ e4, e3, e2, e1, e0);4627
+#else4628
+ simde__m128i_private r_;4629
+4630
+#if defined(SIMDE_WASM_SIMD128_NATIVE)4631
+ r_.wasm_v128 = wasm_i8x16_make(e0, e1, e2, e3, e4, e5, e6, e7, e8, e9,4632
+ e10, e11, e12, e13, e14, e15);4633
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)4634
+ SIMDE_ALIGN_LIKE_16(int8x16_t)4635
+ int8_t data[16] = {e0, e1, e2, e3, e4, e5, e6, e7,4636
+ e8, e9, e10, e11, e12, e13, e14, e15};4637
+ r_.neon_i8 = vld1q_s8(data);4638
+#else4639
+ r_.i8[0] = e0;4640
+ r_.i8[1] = e1;4641
+ r_.i8[2] = e2;4642
+ r_.i8[3] = e3;4643
+ r_.i8[4] = e4;4644
+ r_.i8[5] = e5;4645
+ r_.i8[6] = e6;4646
+ r_.i8[7] = e7;4647
+ r_.i8[8] = e8;4648
+ r_.i8[9] = e9;4649
+ r_.i8[10] = e10;4650
+ r_.i8[11] = e11;4651
+ r_.i8[12] = e12;4652
+ r_.i8[13] = e13;4653
+ r_.i8[14] = e14;4654
+ r_.i8[15] = e15;4655
+#endif4656
+4657
+ return simde__m128i_from_private(r_);4658
+#endif4659
+}4660
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4661
+#define _mm_set_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5, e4, e3, \4662
+ e2, e1, e0) \4663
+ simde_mm_set_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5, \4664
+ e4, e3, e2, e1, e0)4665
+#endif4666
+4667
+SIMDE_FUNCTION_ATTRIBUTES4668
+simde__m128i simde_mm_set_epi16(int16_t e7, int16_t e6, int16_t e5, int16_t e4,4669
+ int16_t e3, int16_t e2, int16_t e1, int16_t e0)4670
+{4671
+#if defined(SIMDE_X86_SSE2_NATIVE)4672
+ return _mm_set_epi16(e7, e6, e5, e4, e3, e2, e1, e0);4673
+#else4674
+ simde__m128i_private r_;4675
+4676
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4677
+ SIMDE_ALIGN_LIKE_16(int16x8_t)4678
+ int16_t data[8] = {e0, e1, e2, e3, e4, e5, e6, e7};4679
+ r_.neon_i16 = vld1q_s16(data);4680
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)4681
+ r_.wasm_v128 = wasm_i16x8_make(e0, e1, e2, e3, e4, e5, e6, e7);4682
+#else4683
+ r_.i16[0] = e0;4684
+ r_.i16[1] = e1;4685
+ r_.i16[2] = e2;4686
+ r_.i16[3] = e3;4687
+ r_.i16[4] = e4;4688
+ r_.i16[5] = e5;4689
+ r_.i16[6] = e6;4690
+ r_.i16[7] = e7;4691
+#endif4692
+4693
+ return simde__m128i_from_private(r_);4694
+#endif4695
+}4696
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4697
+#define _mm_set_epi16(e7, e6, e5, e4, e3, e2, e1, e0) \4698
+ simde_mm_set_epi16(e7, e6, e5, e4, e3, e2, e1, e0)4699
+#endif4700
+4701
+SIMDE_FUNCTION_ATTRIBUTES4702
+simde__m128i simde_mm_loadu_si16(void const *mem_addr)4703
+{4704
+#if defined(SIMDE_X86_SSE2_NATIVE) && \4705
+ (SIMDE_DETECT_CLANG_VERSION_CHECK(8, 0, 0) || \4706
+ HEDLEY_GCC_VERSION_CHECK(11, 0, 0) || \4707
+ HEDLEY_INTEL_VERSION_CHECK(20, 21, 1))4708
+ return _mm_loadu_si16(mem_addr);4709
+#else4710
+ int16_t val;4711
+ simde_memcpy(&val, mem_addr, sizeof(val));4712
+ return simde_x_mm_cvtsi16_si128(val);4713
+#endif4714
+}4715
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4716
+#define _mm_loadu_si16(mem_addr) simde_mm_loadu_si16(mem_addr)4717
+#endif4718
+4719
+SIMDE_FUNCTION_ATTRIBUTES4720
+simde__m128i simde_mm_set_epi32(int32_t e3, int32_t e2, int32_t e1, int32_t e0)4721
+{4722
+#if defined(SIMDE_X86_SSE2_NATIVE)4723
+ return _mm_set_epi32(e3, e2, e1, e0);4724
+#else4725
+ simde__m128i_private r_;4726
+4727
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4728
+ SIMDE_ALIGN_LIKE_16(int32x4_t) int32_t data[4] = {e0, e1, e2, e3};4729
+ r_.neon_i32 = vld1q_s32(data);4730
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)4731
+ r_.wasm_v128 = wasm_i32x4_make(e0, e1, e2, e3);4732
+#else4733
+ r_.i32[0] = e0;4734
+ r_.i32[1] = e1;4735
+ r_.i32[2] = e2;4736
+ r_.i32[3] = e3;4737
+#endif4738
+4739
+ return simde__m128i_from_private(r_);4740
+#endif4741
+}4742
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4743
+#define _mm_set_epi32(e3, e2, e1, e0) simde_mm_set_epi32(e3, e2, e1, e0)4744
+#endif4745
+4746
+SIMDE_FUNCTION_ATTRIBUTES4747
+simde__m128i simde_mm_loadu_si32(void const *mem_addr)4748
+{4749
+#if defined(SIMDE_X86_SSE2_NATIVE) && \4750
+ (SIMDE_DETECT_CLANG_VERSION_CHECK(8, 0, 0) || \4751
+ HEDLEY_GCC_VERSION_CHECK(11, 0, 0) || \4752
+ HEDLEY_INTEL_VERSION_CHECK(20, 21, 1))4753
+ return _mm_loadu_si32(mem_addr);4754
+#else4755
+ int32_t val;4756
+ simde_memcpy(&val, mem_addr, sizeof(val));4757
+ return simde_mm_cvtsi32_si128(val);4758
+#endif4759
+}4760
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4761
+#define _mm_loadu_si32(mem_addr) simde_mm_loadu_si32(mem_addr)4762
+#endif4763
+4764
+SIMDE_FUNCTION_ATTRIBUTES4765
+simde__m128i simde_mm_set_epi64(simde__m64 e1, simde__m64 e0)4766
+{4767
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)4768
+ return _mm_set_epi64(e1, e0);4769
+#else4770
+ simde__m128i_private r_;4771
+4772
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4773
+ r_.neon_i64 = vcombine_s64(simde__m64_to_neon_i64(e0),4774
+ simde__m64_to_neon_i64(e1));4775
+#else4776
+ r_.m64[0] = e0;4777
+ r_.m64[1] = e1;4778
+#endif4779
+4780
+ return simde__m128i_from_private(r_);4781
+#endif4782
+}4783
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4784
+#define _mm_set_epi64(e1, e0) (simde_mm_set_epi64((e1), (e0)))4785
+#endif4786
+4787
+SIMDE_FUNCTION_ATTRIBUTES4788
+simde__m128i simde_mm_set_epi64x(int64_t e1, int64_t e0)4789
+{4790
+#if defined(SIMDE_X86_SSE2_NATIVE) && \4791
+ (!defined(HEDLEY_MSVC_VERSION) || HEDLEY_MSVC_VERSION_CHECK(19, 0, 0))4792
+ return _mm_set_epi64x(e1, e0);4793
+#else4794
+ simde__m128i_private r_;4795
+4796
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4797
+ SIMDE_ALIGN_LIKE_16(int64x2_t) int64_t data[2] = {e0, e1};4798
+ r_.neon_i64 = vld1q_s64(data);4799
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)4800
+ r_.wasm_v128 = wasm_i64x2_make(e0, e1);4801
+#else4802
+ r_.i64[0] = e0;4803
+ r_.i64[1] = e1;4804
+#endif4805
+4806
+ return simde__m128i_from_private(r_);4807
+#endif4808
+}4809
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4810
+#define _mm_set_epi64x(e1, e0) simde_mm_set_epi64x(e1, e0)4811
+#endif4812
+4813
+SIMDE_FUNCTION_ATTRIBUTES4814
+simde__m128i simde_mm_loadu_si64(void const *mem_addr)4815
+{4816
+#if defined(SIMDE_X86_SSE2_NATIVE) && \4817
+ (SIMDE_DETECT_CLANG_VERSION_CHECK(8, 0, 0) || \4818
+ HEDLEY_GCC_VERSION_CHECK(11, 0, 0) || \4819
+ HEDLEY_INTEL_VERSION_CHECK(20, 21, 1))4820
+ return _mm_loadu_si64(mem_addr);4821
+#else4822
+ int64_t val;4823
+ simde_memcpy(&val, mem_addr, sizeof(val));4824
+ return simde_mm_cvtsi64_si128(val);4825
+#endif4826
+}4827
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4828
+#define _mm_loadu_si64(mem_addr) simde_mm_loadu_si64(mem_addr)4829
+#endif4830
+4831
+SIMDE_FUNCTION_ATTRIBUTES4832
+simde__m128i simde_x_mm_set_epu8(uint8_t e15, uint8_t e14, uint8_t e13,4833
+ uint8_t e12, uint8_t e11, uint8_t e10,4834
+ uint8_t e9, uint8_t e8, uint8_t e7, uint8_t e6,4835
+ uint8_t e5, uint8_t e4, uint8_t e3, uint8_t e2,4836
+ uint8_t e1, uint8_t e0)4837
+{4838
+#if defined(SIMDE_X86_SSE2_NATIVE)4839
+ return _mm_set_epi8(4840
+ HEDLEY_STATIC_CAST(char, e15), HEDLEY_STATIC_CAST(char, e14),4841
+ HEDLEY_STATIC_CAST(char, e13), HEDLEY_STATIC_CAST(char, e12),4842
+ HEDLEY_STATIC_CAST(char, e11), HEDLEY_STATIC_CAST(char, e10),4843
+ HEDLEY_STATIC_CAST(char, e9), HEDLEY_STATIC_CAST(char, e8),4844
+ HEDLEY_STATIC_CAST(char, e7), HEDLEY_STATIC_CAST(char, e6),4845
+ HEDLEY_STATIC_CAST(char, e5), HEDLEY_STATIC_CAST(char, e4),4846
+ HEDLEY_STATIC_CAST(char, e3), HEDLEY_STATIC_CAST(char, e2),4847
+ HEDLEY_STATIC_CAST(char, e1), HEDLEY_STATIC_CAST(char, e0));4848
+#else4849
+ simde__m128i_private r_;4850
+4851
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4852
+ SIMDE_ALIGN_LIKE_16(uint8x16_t)4853
+ uint8_t data[16] = {e0, e1, e2, e3, e4, e5, e6, e7,4854
+ e8, e9, e10, e11, e12, e13, e14, e15};4855
+ r_.neon_u8 = vld1q_u8(data);4856
+#else4857
+ r_.u8[0] = e0;4858
+ r_.u8[1] = e1;4859
+ r_.u8[2] = e2;4860
+ r_.u8[3] = e3;4861
+ r_.u8[4] = e4;4862
+ r_.u8[5] = e5;4863
+ r_.u8[6] = e6;4864
+ r_.u8[7] = e7;4865
+ r_.u8[8] = e8;4866
+ r_.u8[9] = e9;4867
+ r_.u8[10] = e10;4868
+ r_.u8[11] = e11;4869
+ r_.u8[12] = e12;4870
+ r_.u8[13] = e13;4871
+ r_.u8[14] = e14;4872
+ r_.u8[15] = e15;4873
+#endif4874
+4875
+ return simde__m128i_from_private(r_);4876
+#endif4877
+}4878
+4879
+SIMDE_FUNCTION_ATTRIBUTES4880
+simde__m128i simde_x_mm_set_epu16(uint16_t e7, uint16_t e6, uint16_t e5,4881
+ uint16_t e4, uint16_t e3, uint16_t e2,4882
+ uint16_t e1, uint16_t e0)4883
+{4884
+#if defined(SIMDE_X86_SSE2_NATIVE)4885
+ return _mm_set_epi16(4886
+ HEDLEY_STATIC_CAST(short, e7), HEDLEY_STATIC_CAST(short, e6),4887
+ HEDLEY_STATIC_CAST(short, e5), HEDLEY_STATIC_CAST(short, e4),4888
+ HEDLEY_STATIC_CAST(short, e3), HEDLEY_STATIC_CAST(short, e2),4889
+ HEDLEY_STATIC_CAST(short, e1), HEDLEY_STATIC_CAST(short, e0));4890
+#else4891
+ simde__m128i_private r_;4892
+4893
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4894
+ SIMDE_ALIGN_LIKE_16(uint16x8_t)4895
+ uint16_t data[8] = {e0, e1, e2, e3, e4, e5, e6, e7};4896
+ r_.neon_u16 = vld1q_u16(data);4897
+#else4898
+ r_.u16[0] = e0;4899
+ r_.u16[1] = e1;4900
+ r_.u16[2] = e2;4901
+ r_.u16[3] = e3;4902
+ r_.u16[4] = e4;4903
+ r_.u16[5] = e5;4904
+ r_.u16[6] = e6;4905
+ r_.u16[7] = e7;4906
+#endif4907
+4908
+ return simde__m128i_from_private(r_);4909
+#endif4910
+}4911
+4912
+SIMDE_FUNCTION_ATTRIBUTES4913
+simde__m128i simde_x_mm_set_epu32(uint32_t e3, uint32_t e2, uint32_t e1,4914
+ uint32_t e0)4915
+{4916
+#if defined(SIMDE_X86_SSE2_NATIVE)4917
+ return _mm_set_epi32(HEDLEY_STATIC_CAST(int, e3),4918
+ HEDLEY_STATIC_CAST(int, e2),4919
+ HEDLEY_STATIC_CAST(int, e1),4920
+ HEDLEY_STATIC_CAST(int, e0));4921
+#else4922
+ simde__m128i_private r_;4923
+4924
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4925
+ SIMDE_ALIGN_LIKE_16(uint32x4_t) uint32_t data[4] = {e0, e1, e2, e3};4926
+ r_.neon_u32 = vld1q_u32(data);4927
+#else4928
+ r_.u32[0] = e0;4929
+ r_.u32[1] = e1;4930
+ r_.u32[2] = e2;4931
+ r_.u32[3] = e3;4932
+#endif4933
+4934
+ return simde__m128i_from_private(r_);4935
+#endif4936
+}4937
+4938
+SIMDE_FUNCTION_ATTRIBUTES4939
+simde__m128i simde_x_mm_set_epu64x(uint64_t e1, uint64_t e0)4940
+{4941
+#if defined(SIMDE_X86_SSE2_NATIVE) && \4942
+ (!defined(HEDLEY_MSVC_VERSION) || HEDLEY_MSVC_VERSION_CHECK(19, 0, 0))4943
+ return _mm_set_epi64x(HEDLEY_STATIC_CAST(int64_t, e1),4944
+ HEDLEY_STATIC_CAST(int64_t, e0));4945
+#else4946
+ simde__m128i_private r_;4947
+4948
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4949
+ SIMDE_ALIGN_LIKE_16(uint64x2_t) uint64_t data[2] = {e0, e1};4950
+ r_.neon_u64 = vld1q_u64(data);4951
+#else4952
+ r_.u64[0] = e0;4953
+ r_.u64[1] = e1;4954
+#endif4955
+4956
+ return simde__m128i_from_private(r_);4957
+#endif4958
+}4959
+4960
+SIMDE_FUNCTION_ATTRIBUTES4961
+simde__m128d simde_mm_set_sd(simde_float64 a)4962
+{4963
+#if defined(SIMDE_X86_SSE2_NATIVE)4964
+ return _mm_set_sd(a);4965
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)4966
+ return vsetq_lane_f64(a, vdupq_n_f64(SIMDE_FLOAT64_C(0.0)), 0);4967
+#else4968
+ return simde_mm_set_pd(SIMDE_FLOAT64_C(0.0), a);4969
+#endif4970
+}4971
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)4972
+#define _mm_set_sd(a) simde_mm_set_sd(a)4973
+#endif4974
+4975
+SIMDE_FUNCTION_ATTRIBUTES4976
+simde__m128i simde_mm_set1_epi8(int8_t a)4977
+{4978
+#if defined(SIMDE_X86_SSE2_NATIVE)4979
+ return _mm_set1_epi8(a);4980
+#else4981
+ simde__m128i_private r_;4982
+4983
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)4984
+ r_.neon_i8 = vdupq_n_s8(a);4985
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)4986
+ r_.wasm_v128 = wasm_i8x16_splat(a);4987
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)4988
+ r_.altivec_i8 = vec_splats(HEDLEY_STATIC_CAST(signed char, a));4989
+#else4990
+ SIMDE_VECTORIZE4991
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {4992
+ r_.i8[i] = a;4993
+ }4994
+#endif4995
+4996
+ return simde__m128i_from_private(r_);4997
+#endif4998
+}4999
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5000
+#define _mm_set1_epi8(a) simde_mm_set1_epi8(a)5001
+#endif5002
+5003
+SIMDE_FUNCTION_ATTRIBUTES5004
+simde__m128i simde_mm_set1_epi16(int16_t a)5005
+{5006
+#if defined(SIMDE_X86_SSE2_NATIVE)5007
+ return _mm_set1_epi16(a);5008
+#else5009
+ simde__m128i_private r_;5010
+5011
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5012
+ r_.neon_i16 = vdupq_n_s16(a);5013
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5014
+ r_.wasm_v128 = wasm_i16x8_splat(a);5015
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5016
+ r_.altivec_i16 = vec_splats(HEDLEY_STATIC_CAST(signed short, a));5017
+#else5018
+ SIMDE_VECTORIZE5019
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {5020
+ r_.i16[i] = a;5021
+ }5022
+#endif5023
+5024
+ return simde__m128i_from_private(r_);5025
+#endif5026
+}5027
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5028
+#define _mm_set1_epi16(a) simde_mm_set1_epi16(a)5029
+#endif5030
+5031
+SIMDE_FUNCTION_ATTRIBUTES5032
+simde__m128i simde_mm_set1_epi32(int32_t a)5033
+{5034
+#if defined(SIMDE_X86_SSE2_NATIVE)5035
+ return _mm_set1_epi32(a);5036
+#else5037
+ simde__m128i_private r_;5038
+5039
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5040
+ r_.neon_i32 = vdupq_n_s32(a);5041
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5042
+ r_.wasm_v128 = wasm_i32x4_splat(a);5043
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5044
+ r_.altivec_i32 = vec_splats(HEDLEY_STATIC_CAST(signed int, a));5045
+#else5046
+ SIMDE_VECTORIZE5047
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {5048
+ r_.i32[i] = a;5049
+ }5050
+#endif5051
+5052
+ return simde__m128i_from_private(r_);5053
+#endif5054
+}5055
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5056
+#define _mm_set1_epi32(a) simde_mm_set1_epi32(a)5057
+#endif5058
+5059
+SIMDE_FUNCTION_ATTRIBUTES5060
+simde__m128i simde_mm_set1_epi64x(int64_t a)5061
+{5062
+#if defined(SIMDE_X86_SSE2_NATIVE) && \5063
+ (!defined(HEDLEY_MSVC_VERSION) || HEDLEY_MSVC_VERSION_CHECK(19, 0, 0))5064
+ return _mm_set1_epi64x(a);5065
+#else5066
+ simde__m128i_private r_;5067
+5068
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5069
+ r_.neon_i64 = vdupq_n_s64(a);5070
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5071
+ r_.wasm_v128 = wasm_i64x2_splat(a);5072
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5073
+ r_.altivec_i64 = vec_splats(HEDLEY_STATIC_CAST(signed long long, a));5074
+#else5075
+ SIMDE_VECTORIZE5076
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {5077
+ r_.i64[i] = a;5078
+ }5079
+#endif5080
+5081
+ return simde__m128i_from_private(r_);5082
+#endif5083
+}5084
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5085
+#define _mm_set1_epi64x(a) simde_mm_set1_epi64x(a)5086
+#endif5087
+5088
+SIMDE_FUNCTION_ATTRIBUTES5089
+simde__m128i simde_mm_set1_epi64(simde__m64 a)5090
+{5091
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)5092
+ return _mm_set1_epi64(a);5093
+#else5094
+ simde__m64_private a_ = simde__m64_to_private(a);5095
+ return simde_mm_set1_epi64x(a_.i64[0]);5096
+#endif5097
+}5098
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5099
+#define _mm_set1_epi64(a) simde_mm_set1_epi64(a)5100
+#endif5101
+5102
+SIMDE_FUNCTION_ATTRIBUTES5103
+simde__m128i simde_x_mm_set1_epu8(uint8_t value)5104
+{5105
+#if defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5106
+ return simde__m128i_from_altivec_u8(5107
+ vec_splats(HEDLEY_STATIC_CAST(unsigned char, value)));5108
+#else5109
+ return simde_mm_set1_epi8(HEDLEY_STATIC_CAST(int8_t, value));5110
+#endif5111
+}5112
+5113
+SIMDE_FUNCTION_ATTRIBUTES5114
+simde__m128i simde_x_mm_set1_epu16(uint16_t value)5115
+{5116
+#if defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5117
+ return simde__m128i_from_altivec_u16(5118
+ vec_splats(HEDLEY_STATIC_CAST(unsigned short, value)));5119
+#else5120
+ return simde_mm_set1_epi16(HEDLEY_STATIC_CAST(int16_t, value));5121
+#endif5122
+}5123
+5124
+SIMDE_FUNCTION_ATTRIBUTES5125
+simde__m128i simde_x_mm_set1_epu32(uint32_t value)5126
+{5127
+#if defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5128
+ return simde__m128i_from_altivec_u32(5129
+ vec_splats(HEDLEY_STATIC_CAST(unsigned int, value)));5130
+#else5131
+ return simde_mm_set1_epi32(HEDLEY_STATIC_CAST(int32_t, value));5132
+#endif5133
+}5134
+5135
+SIMDE_FUNCTION_ATTRIBUTES5136
+simde__m128i simde_x_mm_set1_epu64(uint64_t value)5137
+{5138
+#if defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5139
+ return simde__m128i_from_altivec_u64(5140
+ vec_splats(HEDLEY_STATIC_CAST(unsigned long long, value)));5141
+#else5142
+ return simde_mm_set1_epi64x(HEDLEY_STATIC_CAST(int64_t, value));5143
+#endif5144
+}5145
+5146
+SIMDE_FUNCTION_ATTRIBUTES5147
+simde__m128i simde_mm_setr_epi8(int8_t e15, int8_t e14, int8_t e13, int8_t e12,5148
+ int8_t e11, int8_t e10, int8_t e9, int8_t e8,5149
+ int8_t e7, int8_t e6, int8_t e5, int8_t e4,5150
+ int8_t e3, int8_t e2, int8_t e1, int8_t e0)5151
+{5152
+#if defined(SIMDE_X86_SSE2_NATIVE)5153
+ return _mm_setr_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5,5154
+ e4, e3, e2, e1, e0);5155
+#else5156
+ return simde_mm_set_epi8(e0, e1, e2, e3, e4, e5, e6, e7, e8, e9, e10,5157
+ e11, e12, e13, e14, e15);5158
+#endif5159
+}5160
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5161
+#define _mm_setr_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5, e4, \5162
+ e3, e2, e1, e0) \5163
+ simde_mm_setr_epi8(e15, e14, e13, e12, e11, e10, e9, e8, e7, e6, e5, \5164
+ e4, e3, e2, e1, e0)5165
+#endif5166
+5167
+SIMDE_FUNCTION_ATTRIBUTES5168
+simde__m128i simde_mm_setr_epi16(int16_t e7, int16_t e6, int16_t e5, int16_t e4,5169
+ int16_t e3, int16_t e2, int16_t e1, int16_t e0)5170
+{5171
+#if defined(SIMDE_X86_SSE2_NATIVE)5172
+ return _mm_setr_epi16(e7, e6, e5, e4, e3, e2, e1, e0);5173
+#else5174
+ return simde_mm_set_epi16(e0, e1, e2, e3, e4, e5, e6, e7);5175
+#endif5176
+}5177
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5178
+#define _mm_setr_epi16(e7, e6, e5, e4, e3, e2, e1, e0) \5179
+ simde_mm_setr_epi16(e7, e6, e5, e4, e3, e2, e1, e0)5180
+#endif5181
+5182
+SIMDE_FUNCTION_ATTRIBUTES5183
+simde__m128i simde_mm_setr_epi32(int32_t e3, int32_t e2, int32_t e1, int32_t e0)5184
+{5185
+#if defined(SIMDE_X86_SSE2_NATIVE)5186
+ return _mm_setr_epi32(e3, e2, e1, e0);5187
+#else5188
+ return simde_mm_set_epi32(e0, e1, e2, e3);5189
+#endif5190
+}5191
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5192
+#define _mm_setr_epi32(e3, e2, e1, e0) simde_mm_setr_epi32(e3, e2, e1, e0)5193
+#endif5194
+5195
+SIMDE_FUNCTION_ATTRIBUTES5196
+simde__m128i simde_mm_setr_epi64(simde__m64 e1, simde__m64 e0)5197
+{5198
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)5199
+ return _mm_setr_epi64(e1, e0);5200
+#else5201
+ return simde_mm_set_epi64(e0, e1);5202
+#endif5203
+}5204
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5205
+#define _mm_setr_epi64(e1, e0) (simde_mm_setr_epi64((e1), (e0)))5206
+#endif5207
+5208
+SIMDE_FUNCTION_ATTRIBUTES5209
+simde__m128d simde_mm_setr_pd(simde_float64 e1, simde_float64 e0)5210
+{5211
+#if defined(SIMDE_X86_SSE2_NATIVE)5212
+ return _mm_setr_pd(e1, e0);5213
+#else5214
+ return simde_mm_set_pd(e0, e1);5215
+#endif5216
+}5217
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5218
+#define _mm_setr_pd(e1, e0) simde_mm_setr_pd(e1, e0)5219
+#endif5220
+5221
+SIMDE_FUNCTION_ATTRIBUTES5222
+simde__m128d simde_mm_setzero_pd(void)5223
+{5224
+#if defined(SIMDE_X86_SSE2_NATIVE)5225
+ return _mm_setzero_pd();5226
+#else5227
+ return simde_mm_castsi128_pd(simde_mm_setzero_si128());5228
+#endif5229
+}5230
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5231
+#define _mm_setzero_pd() simde_mm_setzero_pd()5232
+#endif5233
+5234
+#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)5235
+HEDLEY_DIAGNOSTIC_PUSH5236
+SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_5237
+#endif5238
+5239
+SIMDE_FUNCTION_ATTRIBUTES5240
+simde__m128d simde_mm_undefined_pd(void)5241
+{5242
+ simde__m128d_private r_;5243
+5244
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE__HAVE_UNDEFINED128)5245
+ r_.n = _mm_undefined_pd();5246
+#elif !defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)5247
+ r_ = simde__m128d_to_private(simde_mm_setzero_pd());5248
+#endif5249
+5250
+ return simde__m128d_from_private(r_);5251
+}5252
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5253
+#define _mm_undefined_pd() simde_mm_undefined_pd()5254
+#endif5255
+5256
+SIMDE_FUNCTION_ATTRIBUTES5257
+simde__m128i simde_mm_undefined_si128(void)5258
+{5259
+ simde__m128i_private r_;5260
+5261
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE__HAVE_UNDEFINED128)5262
+ r_.n = _mm_undefined_si128();5263
+#elif !defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)5264
+ r_ = simde__m128i_to_private(simde_mm_setzero_si128());5265
+#endif5266
+5267
+ return simde__m128i_from_private(r_);5268
+}5269
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5270
+#define _mm_undefined_si128() (simde_mm_undefined_si128())5271
+#endif5272
+5273
+#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)5274
+HEDLEY_DIAGNOSTIC_POP5275
+#endif5276
+5277
+SIMDE_FUNCTION_ATTRIBUTES5278
+simde__m128d simde_x_mm_setone_pd(void)5279
+{5280
+ return simde_mm_castps_pd(simde_x_mm_setone_ps());5281
+}5282
+5283
+SIMDE_FUNCTION_ATTRIBUTES5284
+simde__m128i simde_x_mm_setone_si128(void)5285
+{5286
+ return simde_mm_castps_si128(simde_x_mm_setone_ps());5287
+}5288
+5289
+SIMDE_FUNCTION_ATTRIBUTES5290
+simde__m128i simde_mm_shuffle_epi32(simde__m128i a, const int imm8)5291
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)5292
+{5293
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);5294
+5295
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {5296
+ r_.i32[i] = a_.i32[(imm8 >> (i * 2)) & 3];5297
+ }5298
+5299
+ return simde__m128i_from_private(r_);5300
+}5301
+#if defined(SIMDE_X86_SSE2_NATIVE)5302
+#define simde_mm_shuffle_epi32(a, imm8) _mm_shuffle_epi32((a), (imm8))5303
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5304
+#define simde_mm_shuffle_epi32(a, imm8) \5305
+ __extension__({ \5306
+ int32x4_t ret; \5307
+ ret = vmovq_n_s32(vgetq_lane_s32(vreinterpretq_s32_s64(a), \5308
+ (imm8) & (0x3))); \5309
+ ret = vsetq_lane_s32(vgetq_lane_s32(vreinterpretq_s32_s64(a), \5310
+ ((imm8) >> 2) & 0x3), \5311
+ ret, 1); \5312
+ ret = vsetq_lane_s32(vgetq_lane_s32(vreinterpretq_s32_s64(a), \5313
+ ((imm8) >> 4) & 0x3), \5314
+ ret, 2); \5315
+ ret = vsetq_lane_s32(vgetq_lane_s32(vreinterpretq_s32_s64(a), \5316
+ ((imm8) >> 6) & 0x3), \5317
+ ret, 3); \5318
+ vreinterpretq_s64_s32(ret); \5319
+ })5320
+#elif defined(SIMDE_SHUFFLE_VECTOR_)5321
+#define simde_mm_shuffle_epi32(a, imm8) \5322
+ (__extension__({ \5323
+ const simde__m128i_private simde__tmp_a_ = \5324
+ simde__m128i_to_private(a); \5325
+ simde__m128i_from_private((simde__m128i_private){ \5326
+ .i32 = SIMDE_SHUFFLE_VECTOR_( \5327
+ 32, 16, (simde__tmp_a_).i32, \5328
+ (simde__tmp_a_).i32, ((imm8)) & 3, \5329
+ ((imm8) >> 2) & 3, ((imm8) >> 4) & 3, \5330
+ ((imm8) >> 6) & 3)}); \5331
+ }))5332
+#endif5333
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5334
+#define _mm_shuffle_epi32(a, imm8) simde_mm_shuffle_epi32(a, imm8)5335
+#endif5336
+5337
+SIMDE_FUNCTION_ATTRIBUTES5338
+simde__m128d simde_mm_shuffle_pd(simde__m128d a, simde__m128d b, const int imm8)5339
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 3)5340
+{5341
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),5342
+ b_ = simde__m128d_to_private(b);5343
+5344
+ r_.f64[0] = ((imm8 & 1) == 0) ? a_.f64[0] : a_.f64[1];5345
+ r_.f64[1] = ((imm8 & 2) == 0) ? b_.f64[0] : b_.f64[1];5346
+5347
+ return simde__m128d_from_private(r_);5348
+}5349
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(__PGI)5350
+#define simde_mm_shuffle_pd(a, b, imm8) _mm_shuffle_pd((a), (b), (imm8))5351
+#elif defined(SIMDE_SHUFFLE_VECTOR_)5352
+#define simde_mm_shuffle_pd(a, b, imm8) \5353
+ (__extension__({ \5354
+ simde__m128d_from_private((simde__m128d_private){ \5355
+ .f64 = SIMDE_SHUFFLE_VECTOR_( \5356
+ 64, 16, simde__m128d_to_private(a).f64, \5357
+ simde__m128d_to_private(b).f64, \5358
+ (((imm8)) & 1), (((imm8) >> 1) & 1) + 2)}); \5359
+ }))5360
+#endif5361
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5362
+#define _mm_shuffle_pd(a, b, imm8) simde_mm_shuffle_pd(a, b, imm8)5363
+#endif5364
+5365
+SIMDE_FUNCTION_ATTRIBUTES5366
+simde__m128i simde_mm_shufflehi_epi16(simde__m128i a, const int imm8)5367
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)5368
+{5369
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);5370
+5371
+ SIMDE_VECTORIZE5372
+ for (size_t i = 0; i < ((sizeof(a_.i16) / sizeof(a_.i16[0])) / 2);5373
+ i++) {5374
+ r_.i16[i] = a_.i16[i];5375
+ }5376
+ for (size_t i = ((sizeof(a_.i16) / sizeof(a_.i16[0])) / 2);5377
+ i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {5378
+ r_.i16[i] = a_.i16[((imm8 >> ((i - 4) * 2)) & 3) + 4];5379
+ }5380
+5381
+ return simde__m128i_from_private(r_);5382
+}5383
+#if defined(SIMDE_X86_SSE2_NATIVE)5384
+#define simde_mm_shufflehi_epi16(a, imm8) _mm_shufflehi_epi16((a), (imm8))5385
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5386
+#define simde_mm_shufflehi_epi16(a, imm8) \5387
+ __extension__({ \5388
+ int16x8_t ret = vreinterpretq_s16_s64(a); \5389
+ int16x4_t highBits = vget_high_s16(ret); \5390
+ ret = vsetq_lane_s16(vget_lane_s16(highBits, (imm8) & (0x3)), \5391
+ ret, 4); \5392
+ ret = vsetq_lane_s16( \5393
+ vget_lane_s16(highBits, ((imm8) >> 2) & 0x3), ret, 5); \5394
+ ret = vsetq_lane_s16( \5395
+ vget_lane_s16(highBits, ((imm8) >> 4) & 0x3), ret, 6); \5396
+ ret = vsetq_lane_s16( \5397
+ vget_lane_s16(highBits, ((imm8) >> 6) & 0x3), ret, 7); \5398
+ vreinterpretq_s64_s16(ret); \5399
+ })5400
+#elif defined(SIMDE_SHUFFLE_VECTOR_)5401
+#define simde_mm_shufflehi_epi16(a, imm8) \5402
+ (__extension__({ \5403
+ const simde__m128i_private simde__tmp_a_ = \5404
+ simde__m128i_to_private(a); \5405
+ simde__m128i_from_private((simde__m128i_private){ \5406
+ .i16 = SIMDE_SHUFFLE_VECTOR_( \5407
+ 16, 16, (simde__tmp_a_).i16, \5408
+ (simde__tmp_a_).i16, 0, 1, 2, 3, \5409
+ (((imm8)) & 3) + 4, (((imm8) >> 2) & 3) + 4, \5410
+ (((imm8) >> 4) & 3) + 4, \5411
+ (((imm8) >> 6) & 3) + 4)}); \5412
+ }))5413
+#endif5414
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5415
+#define _mm_shufflehi_epi16(a, imm8) simde_mm_shufflehi_epi16(a, imm8)5416
+#endif5417
+5418
+SIMDE_FUNCTION_ATTRIBUTES5419
+simde__m128i simde_mm_shufflelo_epi16(simde__m128i a, const int imm8)5420
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)5421
+{5422
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);5423
+5424
+ for (size_t i = 0; i < ((sizeof(r_.i16) / sizeof(r_.i16[0])) / 2);5425
+ i++) {5426
+ r_.i16[i] = a_.i16[((imm8 >> (i * 2)) & 3)];5427
+ }5428
+ SIMDE_VECTORIZE5429
+ for (size_t i = ((sizeof(a_.i16) / sizeof(a_.i16[0])) / 2);5430
+ i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {5431
+ r_.i16[i] = a_.i16[i];5432
+ }5433
+5434
+ return simde__m128i_from_private(r_);5435
+}5436
+#if defined(SIMDE_X86_SSE2_NATIVE)5437
+#define simde_mm_shufflelo_epi16(a, imm8) _mm_shufflelo_epi16((a), (imm8))5438
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5439
+#define simde_mm_shufflelo_epi16(a, imm8) \5440
+ __extension__({ \5441
+ int16x8_t ret = vreinterpretq_s16_s64(a); \5442
+ int16x4_t lowBits = vget_low_s16(ret); \5443
+ ret = vsetq_lane_s16(vget_lane_s16(lowBits, (imm8) & (0x3)), \5444
+ ret, 0); \5445
+ ret = vsetq_lane_s16( \5446
+ vget_lane_s16(lowBits, ((imm8) >> 2) & 0x3), ret, 1); \5447
+ ret = vsetq_lane_s16( \5448
+ vget_lane_s16(lowBits, ((imm8) >> 4) & 0x3), ret, 2); \5449
+ ret = vsetq_lane_s16( \5450
+ vget_lane_s16(lowBits, ((imm8) >> 6) & 0x3), ret, 3); \5451
+ vreinterpretq_s64_s16(ret); \5452
+ })5453
+#elif defined(SIMDE_SHUFFLE_VECTOR_)5454
+#define simde_mm_shufflelo_epi16(a, imm8) \5455
+ (__extension__({ \5456
+ const simde__m128i_private simde__tmp_a_ = \5457
+ simde__m128i_to_private(a); \5458
+ simde__m128i_from_private((simde__m128i_private){ \5459
+ .i16 = SIMDE_SHUFFLE_VECTOR_( \5460
+ 16, 16, (simde__tmp_a_).i16, \5461
+ (simde__tmp_a_).i16, (((imm8)) & 3), \5462
+ (((imm8) >> 2) & 3), (((imm8) >> 4) & 3), \5463
+ (((imm8) >> 6) & 3), 4, 5, 6, 7)}); \5464
+ }))5465
+#endif5466
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5467
+#define _mm_shufflelo_epi16(a, imm8) simde_mm_shufflelo_epi16(a, imm8)5468
+#endif5469
+5470
+SIMDE_FUNCTION_ATTRIBUTES5471
+simde__m128i simde_mm_sll_epi16(simde__m128i a, simde__m128i count)5472
+{5473
+#if defined(SIMDE_X86_SSE2_NATIVE)5474
+ return _mm_sll_epi16(a, count);5475
+#else5476
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),5477
+ count_ = simde__m128i_to_private(count);5478
+5479
+ if (count_.u64[0] > 15)5480
+ return simde_mm_setzero_si128();5481
+5482
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)5483
+ r_.u16 = (a_.u16 << count_.u64[0]);5484
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5485
+ r_.neon_u16 = vshlq_u16(a_.neon_u16, vdupq_n_s16(HEDLEY_STATIC_CAST(5486
+ int16_t, count_.u64[0])));5487
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5488
+ r_.wasm_v128 =5489
+ ((wasm_i64x2_extract_lane(count_.wasm_v128, 0) < 16)5490
+ ? wasm_i16x8_shl(a_.wasm_v128,5491
+ HEDLEY_STATIC_CAST(5492
+ int32_t,5493
+ wasm_i64x2_extract_lane(5494
+ count_.wasm_v128, 0)))5495
+ : wasm_i16x8_const(0, 0, 0, 0, 0, 0, 0, 0));5496
+#else5497
+ SIMDE_VECTORIZE5498
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {5499
+ r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t,5500
+ (a_.u16[i] << count_.u64[0]));5501
+ }5502
+#endif5503
+5504
+ return simde__m128i_from_private(r_);5505
+#endif5506
+}5507
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5508
+#define _mm_sll_epi16(a, count) simde_mm_sll_epi16((a), (count))5509
+#endif5510
+5511
+SIMDE_FUNCTION_ATTRIBUTES5512
+simde__m128i simde_mm_sll_epi32(simde__m128i a, simde__m128i count)5513
+{5514
+#if defined(SIMDE_X86_SSE2_NATIVE)5515
+ return _mm_sll_epi32(a, count);5516
+#else5517
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),5518
+ count_ = simde__m128i_to_private(count);5519
+5520
+ if (count_.u64[0] > 31)5521
+ return simde_mm_setzero_si128();5522
+5523
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)5524
+ r_.u32 = (a_.u32 << count_.u64[0]);5525
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5526
+ r_.neon_u32 = vshlq_u32(a_.neon_u32, vdupq_n_s32(HEDLEY_STATIC_CAST(5527
+ int32_t, count_.u64[0])));5528
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5529
+ r_.wasm_v128 =5530
+ ((wasm_i64x2_extract_lane(count_.wasm_v128, 0) < 32)5531
+ ? wasm_i32x4_shl(a_.wasm_v128,5532
+ HEDLEY_STATIC_CAST(5533
+ int32_t,5534
+ wasm_i64x2_extract_lane(5535
+ count_.wasm_v128, 0)))5536
+ : wasm_i32x4_const(0, 0, 0, 0));5537
+#else5538
+ SIMDE_VECTORIZE5539
+ for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {5540
+ r_.u32[i] = HEDLEY_STATIC_CAST(uint32_t,5541
+ (a_.u32[i] << count_.u64[0]));5542
+ }5543
+#endif5544
+5545
+ return simde__m128i_from_private(r_);5546
+#endif5547
+}5548
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5549
+#define _mm_sll_epi32(a, count) (simde_mm_sll_epi32(a, (count)))5550
+#endif5551
+5552
+SIMDE_FUNCTION_ATTRIBUTES5553
+simde__m128i simde_mm_sll_epi64(simde__m128i a, simde__m128i count)5554
+{5555
+#if defined(SIMDE_X86_SSE2_NATIVE)5556
+ return _mm_sll_epi64(a, count);5557
+#else5558
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),5559
+ count_ = simde__m128i_to_private(count);5560
+5561
+ if (count_.u64[0] > 63)5562
+ return simde_mm_setzero_si128();5563
+5564
+ const int_fast16_t s = HEDLEY_STATIC_CAST(int_fast16_t, count_.u64[0]);5565
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5566
+ r_.neon_u64 = vshlq_u64(a_.neon_u64,5567
+ vdupq_n_s64(HEDLEY_STATIC_CAST(int64_t, s)));5568
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5569
+ r_.wasm_v128 = (s < 64) ? wasm_i64x2_shl(a_.wasm_v128, s)5570
+ : wasm_i64x2_const(0, 0);5571
+#else5572
+#if !defined(SIMDE_BUG_GCC_94488)5573
+ SIMDE_VECTORIZE5574
+#endif5575
+ for (size_t i = 0; i < (sizeof(r_.u64) / sizeof(r_.u64[0])); i++) {5576
+ r_.u64[i] = a_.u64[i] << s;5577
+ }5578
+#endif5579
+5580
+ return simde__m128i_from_private(r_);5581
+#endif5582
+}5583
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5584
+#define _mm_sll_epi64(a, count) (simde_mm_sll_epi64(a, (count)))5585
+#endif5586
+5587
+SIMDE_FUNCTION_ATTRIBUTES5588
+simde__m128d simde_mm_sqrt_pd(simde__m128d a)5589
+{5590
+#if defined(SIMDE_X86_SSE2_NATIVE)5591
+ return _mm_sqrt_pd(a);5592
+#else5593
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a);5594
+5595
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)5596
+ r_.neon_f64 = vsqrtq_f64(a_.neon_f64);5597
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5598
+ r_.wasm_v128 = wasm_f64x2_sqrt(a_.wasm_v128);5599
+#elif defined(simde_math_sqrt)5600
+ SIMDE_VECTORIZE5601
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {5602
+ r_.f64[i] = simde_math_sqrt(a_.f64[i]);5603
+ }5604
+#else5605
+ HEDLEY_UNREACHABLE();5606
+#endif5607
+5608
+ return simde__m128d_from_private(r_);5609
+#endif5610
+}5611
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5612
+#define _mm_sqrt_pd(a) simde_mm_sqrt_pd(a)5613
+#endif5614
+5615
+SIMDE_FUNCTION_ATTRIBUTES5616
+simde__m128d simde_mm_sqrt_sd(simde__m128d a, simde__m128d b)5617
+{5618
+#if defined(SIMDE_X86_SSE2_NATIVE)5619
+ return _mm_sqrt_sd(a, b);5620
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)5621
+ return simde_mm_move_sd(a, simde_mm_sqrt_pd(b));5622
+#else5623
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),5624
+ b_ = simde__m128d_to_private(b);5625
+5626
+#if defined(simde_math_sqrt)5627
+ r_.f64[0] = simde_math_sqrt(b_.f64[0]);5628
+ r_.f64[1] = a_.f64[1];5629
+#else5630
+ HEDLEY_UNREACHABLE();5631
+#endif5632
+5633
+ return simde__m128d_from_private(r_);5634
+#endif5635
+}5636
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5637
+#define _mm_sqrt_sd(a, b) simde_mm_sqrt_sd(a, b)5638
+#endif5639
+5640
+SIMDE_FUNCTION_ATTRIBUTES5641
+simde__m128i simde_mm_srl_epi16(simde__m128i a, simde__m128i count)5642
+{5643
+#if defined(SIMDE_X86_SSE2_NATIVE)5644
+ return _mm_srl_epi16(a, count);5645
+#else5646
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),5647
+ count_ = simde__m128i_to_private(count);5648
+5649
+ const int cnt = HEDLEY_STATIC_CAST(5650
+ int, (count_.i64[0] > 16 ? 16 : count_.i64[0]));5651
+5652
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5653
+ r_.neon_u16 = vshlq_u16(a_.neon_u16,5654
+ vdupq_n_s16(HEDLEY_STATIC_CAST(int16_t, -cnt)));5655
+#else5656
+ SIMDE_VECTORIZE5657
+ for (size_t i = 0; i < (sizeof(r_.u16) / sizeof(r_.u16[0])); i++) {5658
+ r_.u16[i] = a_.u16[i] >> cnt;5659
+ }5660
+#endif5661
+5662
+ return simde__m128i_from_private(r_);5663
+#endif5664
+}5665
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5666
+#define _mm_srl_epi16(a, count) (simde_mm_srl_epi16(a, (count)))5667
+#endif5668
+5669
+SIMDE_FUNCTION_ATTRIBUTES5670
+simde__m128i simde_mm_srl_epi32(simde__m128i a, simde__m128i count)5671
+{5672
+#if defined(SIMDE_X86_SSE2_NATIVE)5673
+ return _mm_srl_epi32(a, count);5674
+#else5675
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),5676
+ count_ = simde__m128i_to_private(count);5677
+5678
+ const int cnt = HEDLEY_STATIC_CAST(5679
+ int, (count_.i64[0] > 32 ? 32 : count_.i64[0]));5680
+5681
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5682
+ r_.neon_u32 = vshlq_u32(a_.neon_u32,5683
+ vdupq_n_s32(HEDLEY_STATIC_CAST(int32_t, -cnt)));5684
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5685
+ r_.wasm_v128 = wasm_u32x4_shr(a_.wasm_v128, cnt);5686
+#else5687
+ SIMDE_VECTORIZE5688
+ for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {5689
+ r_.u32[i] = a_.u32[i] >> cnt;5690
+ }5691
+#endif5692
+5693
+ return simde__m128i_from_private(r_);5694
+#endif5695
+}5696
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5697
+#define _mm_srl_epi32(a, count) (simde_mm_srl_epi32(a, (count)))5698
+#endif5699
+5700
+SIMDE_FUNCTION_ATTRIBUTES5701
+simde__m128i simde_mm_srl_epi64(simde__m128i a, simde__m128i count)5702
+{5703
+#if defined(SIMDE_X86_SSE2_NATIVE)5704
+ return _mm_srl_epi64(a, count);5705
+#else5706
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),5707
+ count_ = simde__m128i_to_private(count);5708
+5709
+ const int cnt = HEDLEY_STATIC_CAST(5710
+ int, (count_.i64[0] > 64 ? 64 : count_.i64[0]));5711
+5712
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5713
+ r_.neon_u64 = vshlq_u64(a_.neon_u64,5714
+ vdupq_n_s64(HEDLEY_STATIC_CAST(int64_t, -cnt)));5715
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5716
+ r_.wasm_v128 = wasm_u64x2_shr(a_.wasm_v128, cnt);5717
+#else5718
+#if !defined(SIMDE_BUG_GCC_94488)5719
+ SIMDE_VECTORIZE5720
+#endif5721
+ for (size_t i = 0; i < (sizeof(r_.u64) / sizeof(r_.u64[0])); i++) {5722
+ r_.u64[i] = a_.u64[i] >> cnt;5723
+ }5724
+#endif5725
+5726
+ return simde__m128i_from_private(r_);5727
+#endif5728
+}5729
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5730
+#define _mm_srl_epi64(a, count) (simde_mm_srl_epi64(a, (count)))5731
+#endif5732
+5733
+SIMDE_FUNCTION_ATTRIBUTES5734
+simde__m128i simde_mm_srai_epi16(simde__m128i a, const int imm8)5735
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)5736
+{5737
+ /* MSVC requires a range of (0, 255). */5738
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);5739
+5740
+ const int cnt = (imm8 & ~15) ? 15 : imm8;5741
+5742
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5743
+ r_.neon_i16 = vshlq_s16(a_.neon_i16,5744
+ vdupq_n_s16(HEDLEY_STATIC_CAST(int16_t, -cnt)));5745
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5746
+ r_.wasm_v128 = wasm_i16x8_shr(a_.wasm_v128, cnt);5747
+#else5748
+ SIMDE_VECTORIZE5749
+ for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i16[0])); i++) {5750
+ r_.i16[i] = a_.i16[i] >> cnt;5751
+ }5752
+#endif5753
+5754
+ return simde__m128i_from_private(r_);5755
+}5756
+#if defined(SIMDE_X86_SSE2_NATIVE)5757
+#define simde_mm_srai_epi16(a, imm8) _mm_srai_epi16((a), (imm8))5758
+#endif5759
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5760
+#define _mm_srai_epi16(a, imm8) simde_mm_srai_epi16(a, imm8)5761
+#endif5762
+5763
+SIMDE_FUNCTION_ATTRIBUTES5764
+simde__m128i simde_mm_srai_epi32(simde__m128i a, const int imm8)5765
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)5766
+{5767
+ /* MSVC requires a range of (0, 255). */5768
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);5769
+5770
+ const int cnt = (imm8 & ~31) ? 31 : imm8;5771
+5772
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5773
+ r_.neon_i32 = vshlq_s32(a_.neon_i32, vdupq_n_s32(-cnt));5774
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5775
+ r_.wasm_v128 = wasm_i32x4_shr(a_.wasm_v128, cnt);5776
+#else5777
+ SIMDE_VECTORIZE5778
+ for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i32[0])); i++) {5779
+ r_.i32[i] = a_.i32[i] >> cnt;5780
+ }5781
+#endif5782
+5783
+ return simde__m128i_from_private(r_);5784
+}5785
+#if defined(SIMDE_X86_SSE2_NATIVE)5786
+#define simde_mm_srai_epi32(a, imm8) _mm_srai_epi32((a), (imm8))5787
+#endif5788
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5789
+#define _mm_srai_epi32(a, imm8) simde_mm_srai_epi32(a, imm8)5790
+#endif5791
+5792
+SIMDE_FUNCTION_ATTRIBUTES5793
+simde__m128i simde_mm_sra_epi16(simde__m128i a, simde__m128i count)5794
+{5795
+#if defined(SIMDE_X86_SSE2_NATIVE)5796
+ return _mm_sra_epi16(a, count);5797
+#else5798
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),5799
+ count_ = simde__m128i_to_private(count);5800
+5801
+ const int cnt = HEDLEY_STATIC_CAST(5802
+ int, (count_.i64[0] > 15 ? 15 : count_.i64[0]));5803
+5804
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5805
+ r_.neon_i16 = vshlq_s16(a_.neon_i16,5806
+ vdupq_n_s16(HEDLEY_STATIC_CAST(int16_t, -cnt)));5807
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5808
+ r_.wasm_v128 = wasm_i16x8_shr(a_.wasm_v128, cnt);5809
+#else5810
+ SIMDE_VECTORIZE5811
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {5812
+ r_.i16[i] = a_.i16[i] >> cnt;5813
+ }5814
+#endif5815
+5816
+ return simde__m128i_from_private(r_);5817
+#endif5818
+}5819
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5820
+#define _mm_sra_epi16(a, count) (simde_mm_sra_epi16(a, count))5821
+#endif5822
+5823
+SIMDE_FUNCTION_ATTRIBUTES5824
+simde__m128i simde_mm_sra_epi32(simde__m128i a, simde__m128i count)5825
+{5826
+#if defined(SIMDE_X86_SSE2_NATIVE) && !defined(SIMDE_BUG_GCC_BAD_MM_SRA_EPI32)5827
+ return _mm_sra_epi32(a, count);5828
+#else5829
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),5830
+ count_ = simde__m128i_to_private(count);5831
+5832
+ const int cnt = count_.u64[0] > 315833
+ ? 315834
+ : HEDLEY_STATIC_CAST(int, count_.u64[0]);5835
+5836
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)5837
+ r_.neon_i32 = vshlq_s32(a_.neon_i32,5838
+ vdupq_n_s32(HEDLEY_STATIC_CAST(int32_t, -cnt)));5839
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5840
+ r_.wasm_v128 = wasm_i32x4_shr(a_.wasm_v128, cnt);5841
+#else5842
+ SIMDE_VECTORIZE5843
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {5844
+ r_.i32[i] = a_.i32[i] >> cnt;5845
+ }5846
+#endif5847
+5848
+ return simde__m128i_from_private(r_);5849
+#endif5850
+}5851
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5852
+#define _mm_sra_epi32(a, count) (simde_mm_sra_epi32(a, (count)))5853
+#endif5854
+5855
+SIMDE_FUNCTION_ATTRIBUTES5856
+simde__m128i simde_mm_slli_epi16(simde__m128i a, const int imm8)5857
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)5858
+{5859
+ if (HEDLEY_UNLIKELY((imm8 > 15))) {5860
+ return simde_mm_setzero_si128();5861
+ }5862
+5863
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);5864
+5865
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)5866
+ r_.i16 = a_.i16 << (imm8 & 0xff);5867
+#else5868
+ const int s =5869
+ (imm8 >5870
+ HEDLEY_STATIC_CAST(int, sizeof(r_.i16[0]) * CHAR_BIT) - 1)5871
+ ? 05872
+ : imm8;5873
+ SIMDE_VECTORIZE5874
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {5875
+ r_.i16[i] = HEDLEY_STATIC_CAST(int16_t, a_.i16[i] << s);5876
+ }5877
+#endif5878
+5879
+ return simde__m128i_from_private(r_);5880
+}5881
+#if defined(SIMDE_X86_SSE2_NATIVE)5882
+#define simde_mm_slli_epi16(a, imm8) _mm_slli_epi16(a, imm8)5883
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5884
+#define simde_mm_slli_epi16(a, imm8) \5885
+ (__extension__({ \5886
+ simde__m128i ret; \5887
+ if ((imm8) <= 0) { \5888
+ ret = a; \5889
+ } else if ((imm8) > 15) { \5890
+ ret = simde_mm_setzero_si128(); \5891
+ } else { \5892
+ ret = simde__m128i_from_neon_i16(vshlq_n_s16( \5893
+ simde__m128i_to_neon_i16(a), ((imm8)&15))); \5894
+ } \5895
+ ret; \5896
+ }))5897
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5898
+#define simde_mm_slli_epi16(a, imm8) \5899
+ ((imm8 < 16) \5900
+ ? wasm_i16x8_shl(simde__m128i_to_private(a).wasm_v128, imm8) \5901
+ : wasm_i16x8_const(0, 0, 0, 0, 0, 0, 0, 0))5902
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5903
+#define simde_mm_slli_epi16(a, imm8) \5904
+ ((imm8 & ~15) ? simde_mm_setzero_si128() \5905
+ : simde__m128i_from_altivec_i16( \5906
+ vec_sl(simde__m128i_to_altivec_i16(a), \5907
+ vec_splat_u16(HEDLEY_STATIC_CAST( \5908
+ unsigned short, imm8)))))5909
+#endif5910
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5911
+#define _mm_slli_epi16(a, imm8) simde_mm_slli_epi16(a, imm8)5912
+#endif5913
+5914
+SIMDE_FUNCTION_ATTRIBUTES5915
+simde__m128i simde_mm_slli_epi32(simde__m128i a, const int imm8)5916
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)5917
+{5918
+ if (HEDLEY_UNLIKELY((imm8 > 31))) {5919
+ return simde_mm_setzero_si128();5920
+ }5921
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);5922
+5923
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)5924
+ r_.i32 = a_.i32 << imm8;5925
+#else5926
+ SIMDE_VECTORIZE5927
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {5928
+ r_.i32[i] = a_.i32[i] << (imm8 & 0xff);5929
+ }5930
+#endif5931
+5932
+ return simde__m128i_from_private(r_);5933
+}5934
+#if defined(SIMDE_X86_SSE2_NATIVE)5935
+#define simde_mm_slli_epi32(a, imm8) _mm_slli_epi32(a, imm8)5936
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5937
+#define simde_mm_slli_epi32(a, imm8) \5938
+ (__extension__({ \5939
+ simde__m128i ret; \5940
+ if ((imm8) <= 0) { \5941
+ ret = a; \5942
+ } else if ((imm8) > 31) { \5943
+ ret = simde_mm_setzero_si128(); \5944
+ } else { \5945
+ ret = simde__m128i_from_neon_i32(vshlq_n_s32( \5946
+ simde__m128i_to_neon_i32(a), ((imm8)&31))); \5947
+ } \5948
+ ret; \5949
+ }))5950
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)5951
+#define simde_mm_slli_epi32(a, imm8) \5952
+ ((imm8 < 32) \5953
+ ? wasm_i32x4_shl(simde__m128i_to_private(a).wasm_v128, imm8) \5954
+ : wasm_i32x4_const(0, 0, 0, 0))5955
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)5956
+#define simde_mm_slli_epi32(a, imm8) \5957
+ (__extension__({ \5958
+ simde__m128i ret; \5959
+ if ((imm8) <= 0) { \5960
+ ret = a; \5961
+ } else if ((imm8) > 31) { \5962
+ ret = simde_mm_setzero_si128(); \5963
+ } else { \5964
+ ret = simde__m128i_from_altivec_i32( \5965
+ vec_sl(simde__m128i_to_altivec_i32(a), \5966
+ vec_splats(HEDLEY_STATIC_CAST( \5967
+ unsigned int, (imm8)&31)))); \5968
+ } \5969
+ ret; \5970
+ }))5971
+#endif5972
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)5973
+#define _mm_slli_epi32(a, imm8) simde_mm_slli_epi32(a, imm8)5974
+#endif5975
+5976
+SIMDE_FUNCTION_ATTRIBUTES5977
+simde__m128i simde_mm_slli_epi64(simde__m128i a, const int imm8)5978
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)5979
+{5980
+ if (HEDLEY_UNLIKELY((imm8 > 63))) {5981
+ return simde_mm_setzero_si128();5982
+ }5983
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);5984
+5985
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)5986
+ r_.i64 = a_.i64 << imm8;5987
+#else5988
+ SIMDE_VECTORIZE5989
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {5990
+ r_.i64[i] = a_.i64[i] << (imm8 & 0xff);5991
+ }5992
+#endif5993
+5994
+ return simde__m128i_from_private(r_);5995
+}5996
+#if defined(SIMDE_X86_SSE2_NATIVE)5997
+#define simde_mm_slli_epi64(a, imm8) _mm_slli_epi64(a, imm8)5998
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)5999
+#define simde_mm_slli_epi64(a, imm8) \6000
+ (__extension__({ \6001
+ simde__m128i ret; \6002
+ if ((imm8) <= 0) { \6003
+ ret = a; \6004
+ } else if ((imm8) > 63) { \6005
+ ret = simde_mm_setzero_si128(); \6006
+ } else { \6007
+ ret = simde__m128i_from_neon_i64(vshlq_n_s64( \6008
+ simde__m128i_to_neon_i64(a), ((imm8)&63))); \6009
+ } \6010
+ ret; \6011
+ }))6012
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6013
+#define simde_mm_slli_epi64(a, imm8) \6014
+ ((imm8 < 64) \6015
+ ? wasm_i64x2_shl(simde__m128i_to_private(a).wasm_v128, imm8) \6016
+ : wasm_i64x2_const(0, 0))6017
+#endif6018
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6019
+#define _mm_slli_epi64(a, imm8) simde_mm_slli_epi64(a, imm8)6020
+#endif6021
+6022
+SIMDE_FUNCTION_ATTRIBUTES6023
+simde__m128i simde_mm_srli_epi16(simde__m128i a, const int imm8)6024
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)6025
+{6026
+ if (HEDLEY_UNLIKELY((imm8 > 15))) {6027
+ return simde_mm_setzero_si128();6028
+ }6029
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);6030
+6031
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)6032
+ r_.u16 = a_.u16 >> imm8;6033
+#else6034
+ SIMDE_VECTORIZE6035
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {6036
+ r_.u16[i] = a_.u16[i] >> (imm8 & 0xff);6037
+ }6038
+#endif6039
+6040
+ return simde__m128i_from_private(r_);6041
+}6042
+#if defined(SIMDE_X86_SSE2_NATIVE)6043
+#define simde_mm_srli_epi16(a, imm8) _mm_srli_epi16(a, imm8)6044
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)6045
+#define simde_mm_srli_epi16(a, imm8) \6046
+ (__extension__({ \6047
+ simde__m128i ret; \6048
+ if ((imm8) <= 0) { \6049
+ ret = a; \6050
+ } else if ((imm8) > 15) { \6051
+ ret = simde_mm_setzero_si128(); \6052
+ } else { \6053
+ ret = simde__m128i_from_neon_u16(vshrq_n_u16( \6054
+ simde__m128i_to_neon_u16(a), \6055
+ (((imm8)&15) | (((imm8)&15) == 0)))); \6056
+ } \6057
+ ret; \6058
+ }))6059
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6060
+#define simde_mm_srli_epi16(a, imm8) \6061
+ ((imm8 < 16) \6062
+ ? wasm_u16x8_shr(simde__m128i_to_private(a).wasm_v128, imm8) \6063
+ : wasm_i16x8_const(0, 0, 0, 0, 0, 0, 0, 0))6064
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)6065
+#define simde_mm_srli_epi16(a, imm8) \6066
+ ((imm8 & ~15) ? simde_mm_setzero_si128() \6067
+ : simde__m128i_from_altivec_i16( \6068
+ vec_sr(simde__m128i_to_altivec_i16(a), \6069
+ vec_splat_u16(HEDLEY_STATIC_CAST( \6070
+ unsigned short, imm8)))))6071
+#endif6072
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6073
+#define _mm_srli_epi16(a, imm8) simde_mm_srli_epi16(a, imm8)6074
+#endif6075
+6076
+SIMDE_FUNCTION_ATTRIBUTES6077
+simde__m128i simde_mm_srli_epi32(simde__m128i a, const int imm8)6078
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)6079
+{6080
+ if (HEDLEY_UNLIKELY((imm8 > 31))) {6081
+ return simde_mm_setzero_si128();6082
+ }6083
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);6084
+6085
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)6086
+ r_.u32 = a_.u32 >> (imm8 & 0xff);6087
+#else6088
+ SIMDE_VECTORIZE6089
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {6090
+ r_.u32[i] = a_.u32[i] >> (imm8 & 0xff);6091
+ }6092
+#endif6093
+6094
+ return simde__m128i_from_private(r_);6095
+}6096
+#if defined(SIMDE_X86_SSE2_NATIVE)6097
+#define simde_mm_srli_epi32(a, imm8) _mm_srli_epi32(a, imm8)6098
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)6099
+#define simde_mm_srli_epi32(a, imm8) \6100
+ (__extension__({ \6101
+ simde__m128i ret; \6102
+ if ((imm8) <= 0) { \6103
+ ret = a; \6104
+ } else if ((imm8) > 31) { \6105
+ ret = simde_mm_setzero_si128(); \6106
+ } else { \6107
+ ret = simde__m128i_from_neon_u32(vshrq_n_u32( \6108
+ simde__m128i_to_neon_u32(a), \6109
+ (((imm8)&31) | (((imm8)&31) == 0)))); \6110
+ } \6111
+ ret; \6112
+ }))6113
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6114
+#define simde_mm_srli_epi32(a, imm8) \6115
+ ((imm8 < 32) \6116
+ ? wasm_u32x4_shr(simde__m128i_to_private(a).wasm_v128, imm8) \6117
+ : wasm_i32x4_const(0, 0, 0, 0))6118
+#elif defined(SIMDE_POWER_ALTIVEC_P8_NATIVE)6119
+#define simde_mm_srli_epi32(a, imm8) \6120
+ (__extension__({ \6121
+ simde__m128i ret; \6122
+ if ((imm8) <= 0) { \6123
+ ret = a; \6124
+ } else if ((imm8) > 31) { \6125
+ ret = simde_mm_setzero_si128(); \6126
+ } else { \6127
+ ret = simde__m128i_from_altivec_i32( \6128
+ vec_sr(simde__m128i_to_altivec_i32(a), \6129
+ vec_splats(HEDLEY_STATIC_CAST( \6130
+ unsigned int, (imm8)&31)))); \6131
+ } \6132
+ ret; \6133
+ }))6134
+#endif6135
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6136
+#define _mm_srli_epi32(a, imm8) simde_mm_srli_epi32(a, imm8)6137
+#endif6138
+6139
+SIMDE_FUNCTION_ATTRIBUTES6140
+simde__m128i simde_mm_srli_epi64(simde__m128i a, const int imm8)6141
+ SIMDE_REQUIRE_CONSTANT_RANGE(imm8, 0, 255)6142
+{6143
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);6144
+6145
+ if (HEDLEY_UNLIKELY((imm8 & 63) != imm8))6146
+ return simde_mm_setzero_si128();6147
+6148
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6149
+ r_.neon_u64 = vshlq_u64(a_.neon_u64, vdupq_n_s64(-imm8));6150
+#else6151
+#if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && !defined(SIMDE_BUG_GCC_94488)6152
+ r_.u64 = a_.u64 >> imm8;6153
+#else6154
+ SIMDE_VECTORIZE6155
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {6156
+ r_.u64[i] = a_.u64[i] >> imm8;6157
+ }6158
+#endif6159
+#endif6160
+6161
+ return simde__m128i_from_private(r_);6162
+}6163
+#if defined(SIMDE_X86_SSE2_NATIVE)6164
+#define simde_mm_srli_epi64(a, imm8) _mm_srli_epi64(a, imm8)6165
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)6166
+#define simde_mm_srli_epi64(a, imm8) \6167
+ (__extension__({ \6168
+ simde__m128i ret; \6169
+ if ((imm8) <= 0) { \6170
+ ret = a; \6171
+ } else if ((imm8) > 63) { \6172
+ ret = simde_mm_setzero_si128(); \6173
+ } else { \6174
+ ret = simde__m128i_from_neon_u64(vshrq_n_u64( \6175
+ simde__m128i_to_neon_u64(a), \6176
+ (((imm8)&63) | (((imm8)&63) == 0)))); \6177
+ } \6178
+ ret; \6179
+ }))6180
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6181
+#define simde_mm_srli_epi64(a, imm8) \6182
+ ((imm8 < 64) \6183
+ ? wasm_u64x2_shr(simde__m128i_to_private(a).wasm_v128, imm8) \6184
+ : wasm_i64x2_const(0, 0))6185
+#endif6186
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6187
+#define _mm_srli_epi64(a, imm8) simde_mm_srli_epi64(a, imm8)6188
+#endif6189
+6190
+SIMDE_FUNCTION_ATTRIBUTES6191
+void simde_mm_store_pd(simde_float64 mem_addr[HEDLEY_ARRAY_PARAM(2)],6192
+ simde__m128d a)6193
+{6194
+#if defined(SIMDE_X86_SSE2_NATIVE)6195
+ _mm_store_pd(mem_addr, a);6196
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)6197
+ vst1q_f64(mem_addr, simde__m128d_to_private(a).neon_f64);6198
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)6199
+ vst1q_s64(HEDLEY_REINTERPRET_CAST(int64_t *, mem_addr),6200
+ simde__m128d_to_private(a).neon_i64);6201
+#else6202
+ simde_memcpy(SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m128d), &a,6203
+ sizeof(a));6204
+#endif6205
+}6206
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6207
+#define _mm_store_pd(mem_addr, a) \6208
+ simde_mm_store_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6209
+#endif6210
+6211
+SIMDE_FUNCTION_ATTRIBUTES6212
+void simde_mm_store1_pd(simde_float64 mem_addr[HEDLEY_ARRAY_PARAM(2)],6213
+ simde__m128d a)6214
+{6215
+#if defined(SIMDE_X86_SSE2_NATIVE)6216
+ _mm_store1_pd(mem_addr, a);6217
+#else6218
+ simde__m128d_private a_ = simde__m128d_to_private(a);6219
+6220
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)6221
+ vst1q_f64(mem_addr, vdupq_laneq_f64(a_.neon_f64, 0));6222
+#else6223
+ mem_addr[0] = a_.f64[0];6224
+ mem_addr[1] = a_.f64[0];6225
+#endif6226
+#endif6227
+}6228
+#define simde_mm_store_pd1(mem_addr, a) \6229
+ simde_mm_store1_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6230
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6231
+#define _mm_store1_pd(mem_addr, a) \6232
+ simde_mm_store1_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6233
+#define _mm_store_pd1(mem_addr, a) \6234
+ simde_mm_store_pd1(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6235
+#endif6236
+6237
+SIMDE_FUNCTION_ATTRIBUTES6238
+void simde_mm_store_sd(simde_float64 *mem_addr, simde__m128d a)6239
+{6240
+#if defined(SIMDE_X86_SSE2_NATIVE)6241
+ _mm_store_sd(mem_addr, a);6242
+#else6243
+ simde__m128d_private a_ = simde__m128d_to_private(a);6244
+6245
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)6246
+ const simde_float64 v = vgetq_lane_f64(a_.neon_f64, 0);6247
+ simde_memcpy(mem_addr, &v, sizeof(v));6248
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)6249
+ const int64_t v = vgetq_lane_s64(a_.neon_i64, 0);6250
+ simde_memcpy(HEDLEY_REINTERPRET_CAST(int64_t *, mem_addr), &v,6251
+ sizeof(v));6252
+#else6253
+ simde_float64 v = a_.f64[0];6254
+ simde_memcpy(mem_addr, &v, sizeof(simde_float64));6255
+#endif6256
+#endif6257
+}6258
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6259
+#define _mm_store_sd(mem_addr, a) \6260
+ simde_mm_store_sd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6261
+#endif6262
+6263
+SIMDE_FUNCTION_ATTRIBUTES6264
+void simde_mm_store_si128(simde__m128i *mem_addr, simde__m128i a)6265
+{6266
+#if defined(SIMDE_X86_SSE2_NATIVE)6267
+ _mm_store_si128(HEDLEY_STATIC_CAST(__m128i *, mem_addr), a);6268
+#else6269
+ simde__m128i_private a_ = simde__m128i_to_private(a);6270
+6271
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6272
+ vst1q_s32(HEDLEY_REINTERPRET_CAST(int32_t *, mem_addr), a_.neon_i32);6273
+#else6274
+ simde_memcpy(SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m128i), &a_,6275
+ sizeof(a_));6276
+#endif6277
+#endif6278
+}6279
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6280
+#define _mm_store_si128(mem_addr, a) simde_mm_store_si128(mem_addr, a)6281
+#endif6282
+6283
+SIMDE_FUNCTION_ATTRIBUTES6284
+void simde_mm_storeh_pd(simde_float64 *mem_addr, simde__m128d a)6285
+{6286
+#if defined(SIMDE_X86_SSE2_NATIVE)6287
+ _mm_storeh_pd(mem_addr, a);6288
+#else6289
+ simde__m128d_private a_ = simde__m128d_to_private(a);6290
+6291
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)6292
+ *mem_addr = vgetq_lane_f64(a_.neon_f64, 1);6293
+#else6294
+ *mem_addr = a_.f64[1];6295
+#endif6296
+#endif6297
+}6298
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6299
+#define _mm_storeh_pd(mem_addr, a) \6300
+ simde_mm_storeh_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6301
+#endif6302
+6303
+SIMDE_FUNCTION_ATTRIBUTES6304
+void simde_mm_storel_epi64(simde__m128i *mem_addr, simde__m128i a)6305
+{6306
+#if defined(SIMDE_X86_SSE2_NATIVE)6307
+ _mm_storel_epi64(HEDLEY_STATIC_CAST(__m128i *, mem_addr), a);6308
+#else6309
+ simde__m128i_private a_ = simde__m128i_to_private(a);6310
+ int64_t tmp;6311
+6312
+ /* memcpy to prevent aliasing, tmp because we can't take the6313
+ * address of a vector element. */6314
+6315
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6316
+ tmp = vgetq_lane_s64(a_.neon_i64, 0);6317
+#elif defined(SIMDE_POWER_ALTIVEC_P7_NATIVE)6318
+#if defined(SIMDE_BUG_GCC_95227)6319
+ (void)a_;6320
+#endif6321
+ tmp = vec_extract(a_.altivec_i64, 0);6322
+#else6323
+ tmp = a_.i64[0];6324
+#endif6325
+6326
+ simde_memcpy(mem_addr, &tmp, sizeof(tmp));6327
+#endif6328
+}6329
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6330
+#define _mm_storel_epi64(mem_addr, a) simde_mm_storel_epi64(mem_addr, a)6331
+#endif6332
+6333
+SIMDE_FUNCTION_ATTRIBUTES6334
+void simde_mm_storel_pd(simde_float64 *mem_addr, simde__m128d a)6335
+{6336
+#if defined(SIMDE_X86_SSE2_NATIVE)6337
+ _mm_storel_pd(mem_addr, a);6338
+#else6339
+ simde__m128d_private a_ = simde__m128d_to_private(a);6340
+6341
+ simde_float64 tmp;6342
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)6343
+ tmp = vgetq_lane_f64(a_.neon_f64, 0);6344
+#else6345
+ tmp = a_.f64[0];6346
+#endif6347
+ simde_memcpy(mem_addr, &tmp, sizeof(tmp));6348
+#endif6349
+}6350
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6351
+#define _mm_storel_pd(mem_addr, a) \6352
+ simde_mm_storel_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6353
+#endif6354
+6355
+SIMDE_FUNCTION_ATTRIBUTES6356
+void simde_mm_storer_pd(simde_float64 mem_addr[2], simde__m128d a)6357
+{6358
+#if defined(SIMDE_X86_SSE2_NATIVE)6359
+ _mm_storer_pd(mem_addr, a);6360
+#else6361
+ simde__m128d_private a_ = simde__m128d_to_private(a);6362
+6363
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6364
+ vst1q_s64(HEDLEY_REINTERPRET_CAST(int64_t *, mem_addr),6365
+ vextq_s64(a_.neon_i64, a_.neon_i64, 1));6366
+#elif defined(SIMDE_SHUFFLE_VECTOR_)6367
+ a_.f64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.f64, a_.f64, 1, 0);6368
+ simde_mm_store_pd(mem_addr, simde__m128d_from_private(a_));6369
+#else6370
+ mem_addr[0] = a_.f64[1];6371
+ mem_addr[1] = a_.f64[0];6372
+#endif6373
+#endif6374
+}6375
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6376
+#define _mm_storer_pd(mem_addr, a) \6377
+ simde_mm_storer_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6378
+#endif6379
+6380
+SIMDE_FUNCTION_ATTRIBUTES6381
+void simde_mm_storeu_pd(simde_float64 *mem_addr, simde__m128d a)6382
+{6383
+#if defined(SIMDE_X86_SSE2_NATIVE)6384
+ _mm_storeu_pd(mem_addr, a);6385
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)6386
+ vst1q_f64(mem_addr, simde__m128d_to_private(a).neon_f64);6387
+#else6388
+ simde_memcpy(mem_addr, &a, sizeof(a));6389
+#endif6390
+}6391
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6392
+#define _mm_storeu_pd(mem_addr, a) \6393
+ simde_mm_storeu_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6394
+#endif6395
+6396
+SIMDE_FUNCTION_ATTRIBUTES6397
+void simde_mm_storeu_si128(simde__m128i *mem_addr, simde__m128i a)6398
+{6399
+#if defined(SIMDE_X86_SSE2_NATIVE)6400
+ _mm_storeu_si128(HEDLEY_STATIC_CAST(__m128i *, mem_addr), a);6401
+#else6402
+ simde_memcpy(mem_addr, &a, sizeof(a));6403
+#endif6404
+}6405
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6406
+#define _mm_storeu_si128(mem_addr, a) simde_mm_storeu_si128(mem_addr, a)6407
+#endif6408
+6409
+SIMDE_FUNCTION_ATTRIBUTES6410
+void simde_mm_storeu_si16(void *mem_addr, simde__m128i a)6411
+{6412
+#if defined(SIMDE_X86_SSE2_NATIVE) && \6413
+ (SIMDE_DETECT_CLANG_VERSION_CHECK(8, 0, 0) || \6414
+ HEDLEY_GCC_VERSION_CHECK(11, 0, 0) || \6415
+ HEDLEY_INTEL_VERSION_CHECK(20, 21, 1))6416
+ _mm_storeu_si16(mem_addr, a);6417
+#else6418
+ int16_t val = simde_x_mm_cvtsi128_si16(a);6419
+ simde_memcpy(mem_addr, &val, sizeof(val));6420
+#endif6421
+}6422
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6423
+#define _mm_storeu_si16(mem_addr, a) simde_mm_storeu_si16(mem_addr, a)6424
+#endif6425
+6426
+SIMDE_FUNCTION_ATTRIBUTES6427
+void simde_mm_storeu_si32(void *mem_addr, simde__m128i a)6428
+{6429
+#if defined(SIMDE_X86_SSE2_NATIVE) && \6430
+ (SIMDE_DETECT_CLANG_VERSION_CHECK(8, 0, 0) || \6431
+ HEDLEY_GCC_VERSION_CHECK(11, 0, 0) || \6432
+ HEDLEY_INTEL_VERSION_CHECK(20, 21, 1))6433
+ _mm_storeu_si32(mem_addr, a);6434
+#else6435
+ int32_t val = simde_mm_cvtsi128_si32(a);6436
+ simde_memcpy(mem_addr, &val, sizeof(val));6437
+#endif6438
+}6439
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6440
+#define _mm_storeu_si32(mem_addr, a) simde_mm_storeu_si32(mem_addr, a)6441
+#endif6442
+6443
+SIMDE_FUNCTION_ATTRIBUTES6444
+void simde_mm_storeu_si64(void *mem_addr, simde__m128i a)6445
+{6446
+#if defined(SIMDE_X86_SSE2_NATIVE) && \6447
+ (SIMDE_DETECT_CLANG_VERSION_CHECK(8, 0, 0) || \6448
+ HEDLEY_GCC_VERSION_CHECK(11, 0, 0) || \6449
+ HEDLEY_INTEL_VERSION_CHECK(20, 21, 1))6450
+ _mm_storeu_si64(mem_addr, a);6451
+#else6452
+ int64_t val = simde_mm_cvtsi128_si64(a);6453
+ simde_memcpy(mem_addr, &val, sizeof(val));6454
+#endif6455
+}6456
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6457
+#define _mm_storeu_si64(mem_addr, a) simde_mm_storeu_si64(mem_addr, a)6458
+#endif6459
+6460
+SIMDE_FUNCTION_ATTRIBUTES6461
+void simde_mm_stream_pd(simde_float64 mem_addr[HEDLEY_ARRAY_PARAM(2)],6462
+ simde__m128d a)6463
+{6464
+#if defined(SIMDE_X86_SSE2_NATIVE)6465
+ _mm_stream_pd(mem_addr, a);6466
+#else6467
+ simde_memcpy(mem_addr, &a, sizeof(a));6468
+#endif6469
+}6470
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6471
+#define _mm_stream_pd(mem_addr, a) \6472
+ simde_mm_stream_pd(HEDLEY_REINTERPRET_CAST(double *, mem_addr), a)6473
+#endif6474
+6475
+SIMDE_FUNCTION_ATTRIBUTES6476
+void simde_mm_stream_si128(simde__m128i *mem_addr, simde__m128i a)6477
+{6478
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64)6479
+ _mm_stream_si128(HEDLEY_STATIC_CAST(__m128i *, mem_addr), a);6480
+#else6481
+ simde_memcpy(mem_addr, &a, sizeof(a));6482
+#endif6483
+}6484
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6485
+#define _mm_stream_si128(mem_addr, a) simde_mm_stream_si128(mem_addr, a)6486
+#endif6487
+6488
+SIMDE_FUNCTION_ATTRIBUTES6489
+void simde_mm_stream_si32(int32_t *mem_addr, int32_t a)6490
+{6491
+#if defined(SIMDE_X86_SSE2_NATIVE)6492
+ _mm_stream_si32(mem_addr, a);6493
+#else6494
+ *mem_addr = a;6495
+#endif6496
+}6497
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6498
+#define _mm_stream_si32(mem_addr, a) simde_mm_stream_si32(mem_addr, a)6499
+#endif6500
+6501
+SIMDE_FUNCTION_ATTRIBUTES6502
+void simde_mm_stream_si64(int64_t *mem_addr, int64_t a)6503
+{6504
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_ARCH_AMD64) && \6505
+ !defined(HEDLEY_MSVC_VERSION)6506
+ _mm_stream_si64(SIMDE_CHECKED_REINTERPRET_CAST(long long int *,6507
+ int64_t *, mem_addr),6508
+ a);6509
+#else6510
+ *mem_addr = a;6511
+#endif6512
+}6513
+#define simde_mm_stream_si64x(mem_addr, a) simde_mm_stream_si64(mem_addr, a)6514
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6515
+#define _mm_stream_si64(mem_addr, a) \6516
+ simde_mm_stream_si64(SIMDE_CHECKED_REINTERPRET_CAST( \6517
+ int64_t *, __int64 *, mem_addr), \6518
+ a)6519
+#define _mm_stream_si64x(mem_addr, a) \6520
+ simde_mm_stream_si64(SIMDE_CHECKED_REINTERPRET_CAST( \6521
+ int64_t *, __int64 *, mem_addr), \6522
+ a)6523
+#endif6524
+6525
+SIMDE_FUNCTION_ATTRIBUTES6526
+simde__m128i simde_mm_sub_epi8(simde__m128i a, simde__m128i b)6527
+{6528
+#if defined(SIMDE_X86_SSE2_NATIVE)6529
+ return _mm_sub_epi8(a, b);6530
+#else6531
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6532
+ b_ = simde__m128i_to_private(b);6533
+6534
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6535
+ r_.neon_i8 = vsubq_s8(a_.neon_i8, b_.neon_i8);6536
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6537
+ r_.i8 = a_.i8 - b_.i8;6538
+#else6539
+ SIMDE_VECTORIZE6540
+ for (size_t i = 0; i < (sizeof(r_.i8) / sizeof(r_.i8[0])); i++) {6541
+ r_.i8[i] = a_.i8[i] - b_.i8[i];6542
+ }6543
+#endif6544
+6545
+ return simde__m128i_from_private(r_);6546
+#endif6547
+}6548
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6549
+#define _mm_sub_epi8(a, b) simde_mm_sub_epi8(a, b)6550
+#endif6551
+6552
+SIMDE_FUNCTION_ATTRIBUTES6553
+simde__m128i simde_mm_sub_epi16(simde__m128i a, simde__m128i b)6554
+{6555
+#if defined(SIMDE_X86_SSE2_NATIVE)6556
+ return _mm_sub_epi16(a, b);6557
+#else6558
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6559
+ b_ = simde__m128i_to_private(b);6560
+6561
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6562
+ r_.neon_i16 = vsubq_s16(a_.neon_i16, b_.neon_i16);6563
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6564
+ r_.i16 = a_.i16 - b_.i16;6565
+#else6566
+ SIMDE_VECTORIZE6567
+ for (size_t i = 0; i < (sizeof(r_.i16) / sizeof(r_.i16[0])); i++) {6568
+ r_.i16[i] = a_.i16[i] - b_.i16[i];6569
+ }6570
+#endif6571
+6572
+ return simde__m128i_from_private(r_);6573
+#endif6574
+}6575
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6576
+#define _mm_sub_epi16(a, b) simde_mm_sub_epi16(a, b)6577
+#endif6578
+6579
+SIMDE_FUNCTION_ATTRIBUTES6580
+simde__m128i simde_mm_sub_epi32(simde__m128i a, simde__m128i b)6581
+{6582
+#if defined(SIMDE_X86_SSE2_NATIVE)6583
+ return _mm_sub_epi32(a, b);6584
+#else6585
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6586
+ b_ = simde__m128i_to_private(b);6587
+6588
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6589
+ r_.neon_i32 = vsubq_s32(a_.neon_i32, b_.neon_i32);6590
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6591
+ r_.i32 = a_.i32 - b_.i32;6592
+#else6593
+ SIMDE_VECTORIZE6594
+ for (size_t i = 0; i < (sizeof(r_.i32) / sizeof(r_.i32[0])); i++) {6595
+ r_.i32[i] = a_.i32[i] - b_.i32[i];6596
+ }6597
+#endif6598
+6599
+ return simde__m128i_from_private(r_);6600
+#endif6601
+}6602
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6603
+#define _mm_sub_epi32(a, b) simde_mm_sub_epi32(a, b)6604
+#endif6605
+6606
+SIMDE_FUNCTION_ATTRIBUTES6607
+simde__m128i simde_mm_sub_epi64(simde__m128i a, simde__m128i b)6608
+{6609
+#if defined(SIMDE_X86_SSE2_NATIVE)6610
+ return _mm_sub_epi64(a, b);6611
+#else6612
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6613
+ b_ = simde__m128i_to_private(b);6614
+6615
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6616
+ r_.neon_i64 = vsubq_s64(a_.neon_i64, b_.neon_i64);6617
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6618
+ r_.i64 = a_.i64 - b_.i64;6619
+#else6620
+ SIMDE_VECTORIZE6621
+ for (size_t i = 0; i < (sizeof(r_.i64) / sizeof(r_.i64[0])); i++) {6622
+ r_.i64[i] = a_.i64[i] - b_.i64[i];6623
+ }6624
+#endif6625
+6626
+ return simde__m128i_from_private(r_);6627
+#endif6628
+}6629
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6630
+#define _mm_sub_epi64(a, b) simde_mm_sub_epi64(a, b)6631
+#endif6632
+6633
+SIMDE_FUNCTION_ATTRIBUTES6634
+simde__m128i simde_x_mm_sub_epu32(simde__m128i a, simde__m128i b)6635
+{6636
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6637
+ b_ = simde__m128i_to_private(b);6638
+6639
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6640
+ r_.u32 = a_.u32 - b_.u32;6641
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)6642
+ r_.neon_u32 = vsubq_u32(a_.neon_u32, b_.neon_u32);6643
+#else6644
+ SIMDE_VECTORIZE6645
+ for (size_t i = 0; i < (sizeof(r_.u32) / sizeof(r_.u32[0])); i++) {6646
+ r_.u32[i] = a_.u32[i] - b_.u32[i];6647
+ }6648
+#endif6649
+6650
+ return simde__m128i_from_private(r_);6651
+}6652
+6653
+SIMDE_FUNCTION_ATTRIBUTES6654
+simde__m128d simde_mm_sub_pd(simde__m128d a, simde__m128d b)6655
+{6656
+#if defined(SIMDE_X86_SSE2_NATIVE)6657
+ return _mm_sub_pd(a, b);6658
+#else6659
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),6660
+ b_ = simde__m128d_to_private(b);6661
+6662
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6663
+ r_.f64 = a_.f64 - b_.f64;6664
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)6665
+ r_.neon_f64 = vsubq_f64(a_.neon_f64, b_.neon_f64);6666
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6667
+ r_.wasm_v128 = wasm_f64x2_sub(a_.wasm_v128, b_.wasm_v128);6668
+#else6669
+ SIMDE_VECTORIZE6670
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {6671
+ r_.f64[i] = a_.f64[i] - b_.f64[i];6672
+ }6673
+#endif6674
+6675
+ return simde__m128d_from_private(r_);6676
+#endif6677
+}6678
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6679
+#define _mm_sub_pd(a, b) simde_mm_sub_pd(a, b)6680
+#endif6681
+6682
+SIMDE_FUNCTION_ATTRIBUTES6683
+simde__m128d simde_mm_sub_sd(simde__m128d a, simde__m128d b)6684
+{6685
+#if defined(SIMDE_X86_SSE2_NATIVE)6686
+ return _mm_sub_sd(a, b);6687
+#elif (SIMDE_NATURAL_VECTOR_SIZE > 0)6688
+ return simde_mm_move_sd(a, simde_mm_sub_pd(a, b));6689
+#else6690
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),6691
+ b_ = simde__m128d_to_private(b);6692
+6693
+ r_.f64[0] = a_.f64[0] - b_.f64[0];6694
+ r_.f64[1] = a_.f64[1];6695
+6696
+ return simde__m128d_from_private(r_);6697
+#endif6698
+}6699
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6700
+#define _mm_sub_sd(a, b) simde_mm_sub_sd(a, b)6701
+#endif6702
+6703
+SIMDE_FUNCTION_ATTRIBUTES6704
+simde__m64 simde_mm_sub_si64(simde__m64 a, simde__m64 b)6705
+{6706
+#if defined(SIMDE_X86_SSE2_NATIVE) && defined(SIMDE_X86_MMX_NATIVE)6707
+ return _mm_sub_si64(a, b);6708
+#else6709
+ simde__m64_private r_, a_ = simde__m64_to_private(a),6710
+ b_ = simde__m64_to_private(b);6711
+6712
+#if defined(SIMDE_VECTOR_SUBSCRIPT_OPS)6713
+ r_.i64 = a_.i64 - b_.i64;6714
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)6715
+ r_.neon_i64 = vsub_s64(a_.neon_i64, b_.neon_i64);6716
+#else6717
+ r_.i64[0] = a_.i64[0] - b_.i64[0];6718
+#endif6719
+6720
+ return simde__m64_from_private(r_);6721
+#endif6722
+}6723
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6724
+#define _mm_sub_si64(a, b) simde_mm_sub_si64(a, b)6725
+#endif6726
+6727
+SIMDE_FUNCTION_ATTRIBUTES6728
+simde__m128i simde_mm_subs_epi8(simde__m128i a, simde__m128i b)6729
+{6730
+#if defined(SIMDE_X86_SSE2_NATIVE)6731
+ return _mm_subs_epi8(a, b);6732
+#else6733
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6734
+ b_ = simde__m128i_to_private(b);6735
+6736
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6737
+ r_.neon_i8 = vqsubq_s8(a_.neon_i8, b_.neon_i8);6738
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6739
+ r_.wasm_v128 = wasm_i8x16_sub_saturate(a_.wasm_v128, b_.wasm_v128);6740
+#else6741
+ SIMDE_VECTORIZE6742
+ for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i8[0])); i++) {6743
+ if (((b_.i8[i]) > 0 && (a_.i8[i]) < INT8_MIN + (b_.i8[i]))) {6744
+ r_.i8[i] = INT8_MIN;6745
+ } else if ((b_.i8[i]) < 0 &&6746
+ (a_.i8[i]) > INT8_MAX + (b_.i8[i])) {6747
+ r_.i8[i] = INT8_MAX;6748
+ } else {6749
+ r_.i8[i] = (a_.i8[i]) - (b_.i8[i]);6750
+ }6751
+ }6752
+#endif6753
+6754
+ return simde__m128i_from_private(r_);6755
+#endif6756
+}6757
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6758
+#define _mm_subs_epi8(a, b) simde_mm_subs_epi8(a, b)6759
+#endif6760
+6761
+SIMDE_FUNCTION_ATTRIBUTES6762
+simde__m128i simde_mm_subs_epi16(simde__m128i a, simde__m128i b)6763
+{6764
+#if defined(SIMDE_X86_SSE2_NATIVE)6765
+ return _mm_subs_epi16(a, b);6766
+#else6767
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6768
+ b_ = simde__m128i_to_private(b);6769
+6770
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6771
+ r_.neon_i16 = vqsubq_s16(a_.neon_i16, b_.neon_i16);6772
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6773
+ r_.wasm_v128 = wasm_i16x8_sub_saturate(a_.wasm_v128, b_.wasm_v128);6774
+#else6775
+ SIMDE_VECTORIZE6776
+ for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i16[0])); i++) {6777
+ if (((b_.i16[i]) > 0 &&6778
+ (a_.i16[i]) < INT16_MIN + (b_.i16[i]))) {6779
+ r_.i16[i] = INT16_MIN;6780
+ } else if ((b_.i16[i]) < 0 &&6781
+ (a_.i16[i]) > INT16_MAX + (b_.i16[i])) {6782
+ r_.i16[i] = INT16_MAX;6783
+ } else {6784
+ r_.i16[i] = (a_.i16[i]) - (b_.i16[i]);6785
+ }6786
+ }6787
+#endif6788
+6789
+ return simde__m128i_from_private(r_);6790
+#endif6791
+}6792
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6793
+#define _mm_subs_epi16(a, b) simde_mm_subs_epi16(a, b)6794
+#endif6795
+6796
+SIMDE_FUNCTION_ATTRIBUTES6797
+simde__m128i simde_mm_subs_epu8(simde__m128i a, simde__m128i b)6798
+{6799
+#if defined(SIMDE_X86_SSE2_NATIVE)6800
+ return _mm_subs_epu8(a, b);6801
+#else6802
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6803
+ b_ = simde__m128i_to_private(b);6804
+6805
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6806
+ r_.neon_u8 = vqsubq_u8(a_.neon_u8, b_.neon_u8);6807
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6808
+ r_.wasm_v128 = wasm_u8x16_sub_saturate(a_.wasm_v128, b_.wasm_v128);6809
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)6810
+ r_.altivec_u8 = vec_subs(a_.altivec_u8, b_.altivec_u8);6811
+#else6812
+ SIMDE_VECTORIZE6813
+ for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i8[0])); i++) {6814
+ const int32_t x = a_.u8[i] - b_.u8[i];6815
+ if (x < 0) {6816
+ r_.u8[i] = 0;6817
+ } else if (x > UINT8_MAX) {6818
+ r_.u8[i] = UINT8_MAX;6819
+ } else {6820
+ r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, x);6821
+ }6822
+ }6823
+#endif6824
+6825
+ return simde__m128i_from_private(r_);6826
+#endif6827
+}6828
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6829
+#define _mm_subs_epu8(a, b) simde_mm_subs_epu8(a, b)6830
+#endif6831
+6832
+SIMDE_FUNCTION_ATTRIBUTES6833
+simde__m128i simde_mm_subs_epu16(simde__m128i a, simde__m128i b)6834
+{6835
+#if defined(SIMDE_X86_SSE2_NATIVE)6836
+ return _mm_subs_epu16(a, b);6837
+#else6838
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),6839
+ b_ = simde__m128i_to_private(b);6840
+6841
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)6842
+ r_.neon_u16 = vqsubq_u16(a_.neon_u16, b_.neon_u16);6843
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6844
+ r_.wasm_v128 = wasm_u16x8_sub_saturate(a_.wasm_v128, b_.wasm_v128);6845
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)6846
+ r_.altivec_u16 = vec_subs(a_.altivec_u16, b_.altivec_u16);6847
+#else6848
+ SIMDE_VECTORIZE6849
+ for (size_t i = 0; i < (sizeof(r_) / sizeof(r_.i16[0])); i++) {6850
+ const int32_t x = a_.u16[i] - b_.u16[i];6851
+ if (x < 0) {6852
+ r_.u16[i] = 0;6853
+ } else if (x > UINT16_MAX) {6854
+ r_.u16[i] = UINT16_MAX;6855
+ } else {6856
+ r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, x);6857
+ }6858
+ }6859
+#endif6860
+6861
+ return simde__m128i_from_private(r_);6862
+#endif6863
+}6864
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6865
+#define _mm_subs_epu16(a, b) simde_mm_subs_epu16(a, b)6866
+#endif6867
+6868
+SIMDE_FUNCTION_ATTRIBUTES6869
+int simde_mm_ucomieq_sd(simde__m128d a, simde__m128d b)6870
+{6871
+#if defined(SIMDE_X86_SSE2_NATIVE)6872
+ return _mm_ucomieq_sd(a, b);6873
+#else6874
+ simde__m128d_private a_ = simde__m128d_to_private(a),6875
+ b_ = simde__m128d_to_private(b);6876
+ int r;6877
+6878
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)6879
+ uint64x2_t a_not_nan = vceqq_f64(a_.neon_f64, a_.neon_f64);6880
+ uint64x2_t b_not_nan = vceqq_f64(b_.neon_f64, b_.neon_f64);6881
+ uint64x2_t a_or_b_nan = vreinterpretq_u64_u32(vmvnq_u32(6882
+ vreinterpretq_u32_u64(vandq_u64(a_not_nan, b_not_nan))));6883
+ uint64x2_t a_eq_b = vceqq_f64(a_.neon_f64, b_.neon_f64);6884
+ r = !!(vgetq_lane_u64(vorrq_u64(a_or_b_nan, a_eq_b), 0) != 0);6885
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6886
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) ==6887
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);6888
+#elif defined(SIMDE_HAVE_FENV_H)6889
+ fenv_t envp;6890
+ int x = feholdexcept(&envp);6891
+ r = a_.f64[0] == b_.f64[0];6892
+ if (HEDLEY_LIKELY(x == 0))6893
+ fesetenv(&envp);6894
+#else6895
+ r = a_.f64[0] == b_.f64[0];6896
+#endif6897
+6898
+ return r;6899
+#endif6900
+}6901
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6902
+#define _mm_ucomieq_sd(a, b) simde_mm_ucomieq_sd(a, b)6903
+#endif6904
+6905
+SIMDE_FUNCTION_ATTRIBUTES6906
+int simde_mm_ucomige_sd(simde__m128d a, simde__m128d b)6907
+{6908
+#if defined(SIMDE_X86_SSE2_NATIVE)6909
+ return _mm_ucomige_sd(a, b);6910
+#else6911
+ simde__m128d_private a_ = simde__m128d_to_private(a),6912
+ b_ = simde__m128d_to_private(b);6913
+ int r;6914
+6915
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)6916
+ uint64x2_t a_not_nan = vceqq_f64(a_.neon_f64, a_.neon_f64);6917
+ uint64x2_t b_not_nan = vceqq_f64(b_.neon_f64, b_.neon_f64);6918
+ uint64x2_t a_and_b_not_nan = vandq_u64(a_not_nan, b_not_nan);6919
+ uint64x2_t a_ge_b = vcgeq_f64(a_.neon_f64, b_.neon_f64);6920
+ r = !!(vgetq_lane_u64(vandq_u64(a_and_b_not_nan, a_ge_b), 0) != 0);6921
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6922
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) >=6923
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);6924
+#elif defined(SIMDE_HAVE_FENV_H)6925
+ fenv_t envp;6926
+ int x = feholdexcept(&envp);6927
+ r = a_.f64[0] >= b_.f64[0];6928
+ if (HEDLEY_LIKELY(x == 0))6929
+ fesetenv(&envp);6930
+#else6931
+ r = a_.f64[0] >= b_.f64[0];6932
+#endif6933
+6934
+ return r;6935
+#endif6936
+}6937
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6938
+#define _mm_ucomige_sd(a, b) simde_mm_ucomige_sd(a, b)6939
+#endif6940
+6941
+SIMDE_FUNCTION_ATTRIBUTES6942
+int simde_mm_ucomigt_sd(simde__m128d a, simde__m128d b)6943
+{6944
+#if defined(SIMDE_X86_SSE2_NATIVE)6945
+ return _mm_ucomigt_sd(a, b);6946
+#else6947
+ simde__m128d_private a_ = simde__m128d_to_private(a),6948
+ b_ = simde__m128d_to_private(b);6949
+ int r;6950
+6951
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)6952
+ uint64x2_t a_not_nan = vceqq_f64(a_.neon_f64, a_.neon_f64);6953
+ uint64x2_t b_not_nan = vceqq_f64(b_.neon_f64, b_.neon_f64);6954
+ uint64x2_t a_and_b_not_nan = vandq_u64(a_not_nan, b_not_nan);6955
+ uint64x2_t a_gt_b = vcgtq_f64(a_.neon_f64, b_.neon_f64);6956
+ r = !!(vgetq_lane_u64(vandq_u64(a_and_b_not_nan, a_gt_b), 0) != 0);6957
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6958
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) >6959
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);6960
+#elif defined(SIMDE_HAVE_FENV_H)6961
+ fenv_t envp;6962
+ int x = feholdexcept(&envp);6963
+ r = a_.f64[0] > b_.f64[0];6964
+ if (HEDLEY_LIKELY(x == 0))6965
+ fesetenv(&envp);6966
+#else6967
+ r = a_.f64[0] > b_.f64[0];6968
+#endif6969
+6970
+ return r;6971
+#endif6972
+}6973
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)6974
+#define _mm_ucomigt_sd(a, b) simde_mm_ucomigt_sd(a, b)6975
+#endif6976
+6977
+SIMDE_FUNCTION_ATTRIBUTES6978
+int simde_mm_ucomile_sd(simde__m128d a, simde__m128d b)6979
+{6980
+#if defined(SIMDE_X86_SSE2_NATIVE)6981
+ return _mm_ucomile_sd(a, b);6982
+#else6983
+ simde__m128d_private a_ = simde__m128d_to_private(a),6984
+ b_ = simde__m128d_to_private(b);6985
+ int r;6986
+6987
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)6988
+ uint64x2_t a_not_nan = vceqq_f64(a_.neon_f64, a_.neon_f64);6989
+ uint64x2_t b_not_nan = vceqq_f64(b_.neon_f64, b_.neon_f64);6990
+ uint64x2_t a_or_b_nan = vreinterpretq_u64_u32(vmvnq_u32(6991
+ vreinterpretq_u32_u64(vandq_u64(a_not_nan, b_not_nan))));6992
+ uint64x2_t a_le_b = vcleq_f64(a_.neon_f64, b_.neon_f64);6993
+ r = !!(vgetq_lane_u64(vorrq_u64(a_or_b_nan, a_le_b), 0) != 0);6994
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)6995
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) <=6996
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);6997
+#elif defined(SIMDE_HAVE_FENV_H)6998
+ fenv_t envp;6999
+ int x = feholdexcept(&envp);7000
+ r = a_.f64[0] <= b_.f64[0];7001
+ if (HEDLEY_LIKELY(x == 0))7002
+ fesetenv(&envp);7003
+#else7004
+ r = a_.f64[0] <= b_.f64[0];7005
+#endif7006
+7007
+ return r;7008
+#endif7009
+}7010
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7011
+#define _mm_ucomile_sd(a, b) simde_mm_ucomile_sd(a, b)7012
+#endif7013
+7014
+SIMDE_FUNCTION_ATTRIBUTES7015
+int simde_mm_ucomilt_sd(simde__m128d a, simde__m128d b)7016
+{7017
+#if defined(SIMDE_X86_SSE2_NATIVE)7018
+ return _mm_ucomilt_sd(a, b);7019
+#else7020
+ simde__m128d_private a_ = simde__m128d_to_private(a),7021
+ b_ = simde__m128d_to_private(b);7022
+ int r;7023
+7024
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7025
+ uint64x2_t a_not_nan = vceqq_f64(a_.neon_f64, a_.neon_f64);7026
+ uint64x2_t b_not_nan = vceqq_f64(b_.neon_f64, b_.neon_f64);7027
+ uint64x2_t a_or_b_nan = vreinterpretq_u64_u32(vmvnq_u32(7028
+ vreinterpretq_u32_u64(vandq_u64(a_not_nan, b_not_nan))));7029
+ uint64x2_t a_lt_b = vcltq_f64(a_.neon_f64, b_.neon_f64);7030
+ r = !!(vgetq_lane_u64(vorrq_u64(a_or_b_nan, a_lt_b), 0) != 0);7031
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)7032
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) <7033
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);7034
+#elif defined(SIMDE_HAVE_FENV_H)7035
+ fenv_t envp;7036
+ int x = feholdexcept(&envp);7037
+ r = a_.f64[0] < b_.f64[0];7038
+ if (HEDLEY_LIKELY(x == 0))7039
+ fesetenv(&envp);7040
+#else7041
+ r = a_.f64[0] < b_.f64[0];7042
+#endif7043
+7044
+ return r;7045
+#endif7046
+}7047
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7048
+#define _mm_ucomilt_sd(a, b) simde_mm_ucomilt_sd(a, b)7049
+#endif7050
+7051
+SIMDE_FUNCTION_ATTRIBUTES7052
+int simde_mm_ucomineq_sd(simde__m128d a, simde__m128d b)7053
+{7054
+#if defined(SIMDE_X86_SSE2_NATIVE)7055
+ return _mm_ucomineq_sd(a, b);7056
+#else7057
+ simde__m128d_private a_ = simde__m128d_to_private(a),7058
+ b_ = simde__m128d_to_private(b);7059
+ int r;7060
+7061
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7062
+ uint64x2_t a_not_nan = vceqq_f64(a_.neon_f64, a_.neon_f64);7063
+ uint64x2_t b_not_nan = vceqq_f64(b_.neon_f64, b_.neon_f64);7064
+ uint64x2_t a_and_b_not_nan = vandq_u64(a_not_nan, b_not_nan);7065
+ uint64x2_t a_neq_b = vreinterpretq_u64_u32(vmvnq_u32(7066
+ vreinterpretq_u32_u64(vceqq_f64(a_.neon_f64, b_.neon_f64))));7067
+ r = !!(vgetq_lane_u64(vandq_u64(a_and_b_not_nan, a_neq_b), 0) != 0);7068
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)7069
+ return wasm_f64x2_extract_lane(a_.wasm_v128, 0) !=7070
+ wasm_f64x2_extract_lane(b_.wasm_v128, 0);7071
+#elif defined(SIMDE_HAVE_FENV_H)7072
+ fenv_t envp;7073
+ int x = feholdexcept(&envp);7074
+ r = a_.f64[0] != b_.f64[0];7075
+ if (HEDLEY_LIKELY(x == 0))7076
+ fesetenv(&envp);7077
+#else7078
+ r = a_.f64[0] != b_.f64[0];7079
+#endif7080
+7081
+ return r;7082
+#endif7083
+}7084
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7085
+#define _mm_ucomineq_sd(a, b) simde_mm_ucomineq_sd(a, b)7086
+#endif7087
+7088
+#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)7089
+HEDLEY_DIAGNOSTIC_PUSH7090
+SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_7091
+#endif7092
+7093
+#if defined(SIMDE_DIAGNOSTIC_DISABLE_UNINITIALIZED_)7094
+HEDLEY_DIAGNOSTIC_POP7095
+#endif7096
+7097
+SIMDE_FUNCTION_ATTRIBUTES7098
+void simde_mm_lfence(void)7099
+{7100
+#if defined(SIMDE_X86_SSE2_NATIVE)7101
+ _mm_lfence();7102
+#else7103
+ simde_mm_sfence();7104
+#endif7105
+}7106
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7107
+#define _mm_lfence() simde_mm_lfence()7108
+#endif7109
+7110
+SIMDE_FUNCTION_ATTRIBUTES7111
+void simde_mm_mfence(void)7112
+{7113
+#if defined(SIMDE_X86_SSE2_NATIVE)7114
+ _mm_mfence();7115
+#else7116
+ simde_mm_sfence();7117
+#endif7118
+}7119
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7120
+#define _mm_mfence() simde_mm_mfence()7121
+#endif7122
+7123
+SIMDE_FUNCTION_ATTRIBUTES7124
+simde__m128i simde_mm_unpackhi_epi8(simde__m128i a, simde__m128i b)7125
+{7126
+#if defined(SIMDE_X86_SSE2_NATIVE)7127
+ return _mm_unpackhi_epi8(a, b);7128
+#else7129
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7130
+ b_ = simde__m128i_to_private(b);7131
+7132
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7133
+ r_.neon_i8 = vzip2q_s8(a_.neon_i8, b_.neon_i8);7134
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)7135
+ int8x8_t a1 = vreinterpret_s8_s16(vget_high_s16(a_.neon_i16));7136
+ int8x8_t b1 = vreinterpret_s8_s16(vget_high_s16(b_.neon_i16));7137
+ int8x8x2_t result = vzip_s8(a1, b1);7138
+ r_.neon_i8 = vcombine_s8(result.val[0], result.val[1]);7139
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7140
+ r_.i8 = SIMDE_SHUFFLE_VECTOR_(8, 16, a_.i8, b_.i8, 8, 24, 9, 25, 10, 26,7141
+ 11, 27, 12, 28, 13, 29, 14, 30, 15, 31);7142
+#else7143
+ SIMDE_VECTORIZE7144
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i8[0])) / 2); i++) {7145
+ r_.i8[(i * 2)] =7146
+ a_.i8[i + ((sizeof(r_) / sizeof(r_.i8[0])) / 2)];7147
+ r_.i8[(i * 2) + 1] =7148
+ b_.i8[i + ((sizeof(r_) / sizeof(r_.i8[0])) / 2)];7149
+ }7150
+#endif7151
+7152
+ return simde__m128i_from_private(r_);7153
+#endif7154
+}7155
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7156
+#define _mm_unpackhi_epi8(a, b) simde_mm_unpackhi_epi8(a, b)7157
+#endif7158
+7159
+SIMDE_FUNCTION_ATTRIBUTES7160
+simde__m128i simde_mm_unpackhi_epi16(simde__m128i a, simde__m128i b)7161
+{7162
+#if defined(SIMDE_X86_SSE2_NATIVE)7163
+ return _mm_unpackhi_epi16(a, b);7164
+#else7165
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7166
+ b_ = simde__m128i_to_private(b);7167
+7168
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7169
+ r_.neon_i16 = vzip2q_s16(a_.neon_i16, b_.neon_i16);7170
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)7171
+ int16x4_t a1 = vget_high_s16(a_.neon_i16);7172
+ int16x4_t b1 = vget_high_s16(b_.neon_i16);7173
+ int16x4x2_t result = vzip_s16(a1, b1);7174
+ r_.neon_i16 = vcombine_s16(result.val[0], result.val[1]);7175
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7176
+ r_.i16 = SIMDE_SHUFFLE_VECTOR_(16, 16, a_.i16, b_.i16, 4, 12, 5, 13, 6,7177
+ 14, 7, 15);7178
+#else7179
+ SIMDE_VECTORIZE7180
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i16[0])) / 2); i++) {7181
+ r_.i16[(i * 2)] =7182
+ a_.i16[i + ((sizeof(r_) / sizeof(r_.i16[0])) / 2)];7183
+ r_.i16[(i * 2) + 1] =7184
+ b_.i16[i + ((sizeof(r_) / sizeof(r_.i16[0])) / 2)];7185
+ }7186
+#endif7187
+7188
+ return simde__m128i_from_private(r_);7189
+#endif7190
+}7191
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7192
+#define _mm_unpackhi_epi16(a, b) simde_mm_unpackhi_epi16(a, b)7193
+#endif7194
+7195
+SIMDE_FUNCTION_ATTRIBUTES7196
+simde__m128i simde_mm_unpackhi_epi32(simde__m128i a, simde__m128i b)7197
+{7198
+#if defined(SIMDE_X86_SSE2_NATIVE)7199
+ return _mm_unpackhi_epi32(a, b);7200
+#else7201
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7202
+ b_ = simde__m128i_to_private(b);7203
+7204
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7205
+ r_.neon_i32 = vzip2q_s32(a_.neon_i32, b_.neon_i32);7206
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)7207
+ int32x2_t a1 = vget_high_s32(a_.neon_i32);7208
+ int32x2_t b1 = vget_high_s32(b_.neon_i32);7209
+ int32x2x2_t result = vzip_s32(a1, b1);7210
+ r_.neon_i32 = vcombine_s32(result.val[0], result.val[1]);7211
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7212
+ r_.i32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.i32, b_.i32, 2, 6, 3, 7);7213
+#else7214
+ SIMDE_VECTORIZE7215
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i32[0])) / 2); i++) {7216
+ r_.i32[(i * 2)] =7217
+ a_.i32[i + ((sizeof(r_) / sizeof(r_.i32[0])) / 2)];7218
+ r_.i32[(i * 2) + 1] =7219
+ b_.i32[i + ((sizeof(r_) / sizeof(r_.i32[0])) / 2)];7220
+ }7221
+#endif7222
+7223
+ return simde__m128i_from_private(r_);7224
+#endif7225
+}7226
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7227
+#define _mm_unpackhi_epi32(a, b) simde_mm_unpackhi_epi32(a, b)7228
+#endif7229
+7230
+SIMDE_FUNCTION_ATTRIBUTES7231
+simde__m128i simde_mm_unpackhi_epi64(simde__m128i a, simde__m128i b)7232
+{7233
+#if defined(SIMDE_X86_SSE2_NATIVE)7234
+ return _mm_unpackhi_epi64(a, b);7235
+#else7236
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7237
+ b_ = simde__m128i_to_private(b);7238
+7239
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)7240
+ int64x1_t a_h = vget_high_s64(a_.neon_i64);7241
+ int64x1_t b_h = vget_high_s64(b_.neon_i64);7242
+ r_.neon_i64 = vcombine_s64(a_h, b_h);7243
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7244
+ r_.i64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.i64, b_.i64, 1, 3);7245
+#else7246
+ SIMDE_VECTORIZE7247
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i64[0])) / 2); i++) {7248
+ r_.i64[(i * 2)] =7249
+ a_.i64[i + ((sizeof(r_) / sizeof(r_.i64[0])) / 2)];7250
+ r_.i64[(i * 2) + 1] =7251
+ b_.i64[i + ((sizeof(r_) / sizeof(r_.i64[0])) / 2)];7252
+ }7253
+#endif7254
+7255
+ return simde__m128i_from_private(r_);7256
+#endif7257
+}7258
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7259
+#define _mm_unpackhi_epi64(a, b) simde_mm_unpackhi_epi64(a, b)7260
+#endif7261
+7262
+SIMDE_FUNCTION_ATTRIBUTES7263
+simde__m128d simde_mm_unpackhi_pd(simde__m128d a, simde__m128d b)7264
+{7265
+#if defined(SIMDE_X86_SSE2_NATIVE)7266
+ return _mm_unpackhi_pd(a, b);7267
+#else7268
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),7269
+ b_ = simde__m128d_to_private(b);7270
+7271
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7272
+ float64x1_t a_l = vget_high_f64(a_.f64);7273
+ float64x1_t b_l = vget_high_f64(b_.f64);7274
+ r_.neon_f64 = vcombine_f64(a_l, b_l);7275
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)7276
+ r_.wasm_v128 = wasm_v64x2_shuffle(a_.wasm_v128, b_.wasm_v128, 1, 3);7277
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7278
+ r_.f64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.f64, b_.f64, 1, 3);7279
+#else7280
+ SIMDE_VECTORIZE7281
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.f64[0])) / 2); i++) {7282
+ r_.f64[(i * 2)] =7283
+ a_.f64[i + ((sizeof(r_) / sizeof(r_.f64[0])) / 2)];7284
+ r_.f64[(i * 2) + 1] =7285
+ b_.f64[i + ((sizeof(r_) / sizeof(r_.f64[0])) / 2)];7286
+ }7287
+#endif7288
+7289
+ return simde__m128d_from_private(r_);7290
+#endif7291
+}7292
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7293
+#define _mm_unpackhi_pd(a, b) simde_mm_unpackhi_pd(a, b)7294
+#endif7295
+7296
+SIMDE_FUNCTION_ATTRIBUTES7297
+simde__m128i simde_mm_unpacklo_epi8(simde__m128i a, simde__m128i b)7298
+{7299
+#if defined(SIMDE_X86_SSE2_NATIVE)7300
+ return _mm_unpacklo_epi8(a, b);7301
+#else7302
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7303
+ b_ = simde__m128i_to_private(b);7304
+7305
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7306
+ r_.neon_i8 = vzip1q_s8(a_.neon_i8, b_.neon_i8);7307
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)7308
+ int8x8_t a1 = vreinterpret_s8_s16(vget_low_s16(a_.neon_i16));7309
+ int8x8_t b1 = vreinterpret_s8_s16(vget_low_s16(b_.neon_i16));7310
+ int8x8x2_t result = vzip_s8(a1, b1);7311
+ r_.neon_i8 = vcombine_s8(result.val[0], result.val[1]);7312
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7313
+ r_.i8 = SIMDE_SHUFFLE_VECTOR_(8, 16, a_.i8, b_.i8, 0, 16, 1, 17, 2, 18,7314
+ 3, 19, 4, 20, 5, 21, 6, 22, 7, 23);7315
+#else7316
+ SIMDE_VECTORIZE7317
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i8[0])) / 2); i++) {7318
+ r_.i8[(i * 2)] = a_.i8[i];7319
+ r_.i8[(i * 2) + 1] = b_.i8[i];7320
+ }7321
+#endif7322
+7323
+ return simde__m128i_from_private(r_);7324
+#endif7325
+}7326
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7327
+#define _mm_unpacklo_epi8(a, b) simde_mm_unpacklo_epi8(a, b)7328
+#endif7329
+7330
+SIMDE_FUNCTION_ATTRIBUTES7331
+simde__m128i simde_mm_unpacklo_epi16(simde__m128i a, simde__m128i b)7332
+{7333
+#if defined(SIMDE_X86_SSE2_NATIVE)7334
+ return _mm_unpacklo_epi16(a, b);7335
+#else7336
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7337
+ b_ = simde__m128i_to_private(b);7338
+7339
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7340
+ r_.neon_i16 = vzip1q_s16(a_.neon_i16, b_.neon_i16);7341
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)7342
+ int16x4_t a1 = vget_low_s16(a_.neon_i16);7343
+ int16x4_t b1 = vget_low_s16(b_.neon_i16);7344
+ int16x4x2_t result = vzip_s16(a1, b1);7345
+ r_.neon_i16 = vcombine_s16(result.val[0], result.val[1]);7346
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7347
+ r_.i16 = SIMDE_SHUFFLE_VECTOR_(16, 16, a_.i16, b_.i16, 0, 8, 1, 9, 2,7348
+ 10, 3, 11);7349
+#else7350
+ SIMDE_VECTORIZE7351
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i16[0])) / 2); i++) {7352
+ r_.i16[(i * 2)] = a_.i16[i];7353
+ r_.i16[(i * 2) + 1] = b_.i16[i];7354
+ }7355
+#endif7356
+7357
+ return simde__m128i_from_private(r_);7358
+#endif7359
+}7360
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7361
+#define _mm_unpacklo_epi16(a, b) simde_mm_unpacklo_epi16(a, b)7362
+#endif7363
+7364
+SIMDE_FUNCTION_ATTRIBUTES7365
+simde__m128i simde_mm_unpacklo_epi32(simde__m128i a, simde__m128i b)7366
+{7367
+#if defined(SIMDE_X86_SSE2_NATIVE)7368
+ return _mm_unpacklo_epi32(a, b);7369
+#else7370
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7371
+ b_ = simde__m128i_to_private(b);7372
+7373
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7374
+ r_.neon_i32 = vzip1q_s32(a_.neon_i32, b_.neon_i32);7375
+#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)7376
+ int32x2_t a1 = vget_low_s32(a_.neon_i32);7377
+ int32x2_t b1 = vget_low_s32(b_.neon_i32);7378
+ int32x2x2_t result = vzip_s32(a1, b1);7379
+ r_.neon_i32 = vcombine_s32(result.val[0], result.val[1]);7380
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7381
+ r_.i32 = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.i32, b_.i32, 0, 4, 1, 5);7382
+#else7383
+ SIMDE_VECTORIZE7384
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i32[0])) / 2); i++) {7385
+ r_.i32[(i * 2)] = a_.i32[i];7386
+ r_.i32[(i * 2) + 1] = b_.i32[i];7387
+ }7388
+#endif7389
+7390
+ return simde__m128i_from_private(r_);7391
+#endif7392
+}7393
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7394
+#define _mm_unpacklo_epi32(a, b) simde_mm_unpacklo_epi32(a, b)7395
+#endif7396
+7397
+SIMDE_FUNCTION_ATTRIBUTES7398
+simde__m128i simde_mm_unpacklo_epi64(simde__m128i a, simde__m128i b)7399
+{7400
+#if defined(SIMDE_X86_SSE2_NATIVE)7401
+ return _mm_unpacklo_epi64(a, b);7402
+#else7403
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7404
+ b_ = simde__m128i_to_private(b);7405
+7406
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)7407
+ int64x1_t a_l = vget_low_s64(a_.i64);7408
+ int64x1_t b_l = vget_low_s64(b_.i64);7409
+ r_.neon_i64 = vcombine_s64(a_l, b_l);7410
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7411
+ r_.i64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.i64, b_.i64, 0, 2);7412
+#else7413
+ SIMDE_VECTORIZE7414
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.i64[0])) / 2); i++) {7415
+ r_.i64[(i * 2)] = a_.i64[i];7416
+ r_.i64[(i * 2) + 1] = b_.i64[i];7417
+ }7418
+#endif7419
+7420
+ return simde__m128i_from_private(r_);7421
+#endif7422
+}7423
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7424
+#define _mm_unpacklo_epi64(a, b) simde_mm_unpacklo_epi64(a, b)7425
+#endif7426
+7427
+SIMDE_FUNCTION_ATTRIBUTES7428
+simde__m128d simde_mm_unpacklo_pd(simde__m128d a, simde__m128d b)7429
+{7430
+#if defined(SIMDE_X86_SSE2_NATIVE)7431
+ return _mm_unpacklo_pd(a, b);7432
+#else7433
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a),7434
+ b_ = simde__m128d_to_private(b);7435
+7436
+#if defined(SIMDE_ARM_NEON_A64V8_NATIVE)7437
+ float64x1_t a_l = vget_low_f64(a_.f64);7438
+ float64x1_t b_l = vget_low_f64(b_.f64);7439
+ r_.neon_f64 = vcombine_f64(a_l, b_l);7440
+#elif defined(SIMDE_SHUFFLE_VECTOR_)7441
+ r_.f64 = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.f64, b_.f64, 0, 2);7442
+#else7443
+ SIMDE_VECTORIZE7444
+ for (size_t i = 0; i < ((sizeof(r_) / sizeof(r_.f64[0])) / 2); i++) {7445
+ r_.f64[(i * 2)] = a_.f64[i];7446
+ r_.f64[(i * 2) + 1] = b_.f64[i];7447
+ }7448
+#endif7449
+7450
+ return simde__m128d_from_private(r_);7451
+#endif7452
+}7453
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7454
+#define _mm_unpacklo_pd(a, b) simde_mm_unpacklo_pd(a, b)7455
+#endif7456
+7457
+SIMDE_FUNCTION_ATTRIBUTES7458
+simde__m128d simde_x_mm_negate_pd(simde__m128d a)7459
+{7460
+#if defined(SIMDE_X86_SSE_NATIVE)7461
+ return simde_mm_xor_pd(a, _mm_set1_pd(SIMDE_FLOAT64_C(-0.0)));7462
+#else7463
+ simde__m128d_private r_, a_ = simde__m128d_to_private(a);7464
+7465
+#if defined(SIMDE_POWER_ALTIVEC_P8_NATIVE) && \7466
+ (!defined(HEDLEY_GCC_VERSION) || HEDLEY_GCC_VERSION_CHECK(8, 1, 0))7467
+ r_.altivec_f64 = vec_neg(a_.altivec_f64);7468
+#elif defined(SIMDE_ARM_NEON_A64V8_NATIVE)7469
+ r_.neon_f64 = vnegq_f64(a_.neon_f64);7470
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)7471
+ r_.wasm_v128 = wasm_f64x2_neg(a_.wasm_v128);7472
+#elif defined(SIMDE_VECTOR_NEGATE)7473
+ r_.f64 = -a_.f64;7474
+#else7475
+ SIMDE_VECTORIZE7476
+ for (size_t i = 0; i < (sizeof(r_.f64) / sizeof(r_.f64[0])); i++) {7477
+ r_.f64[i] = -a_.f64[i];7478
+ }7479
+#endif7480
+7481
+ return simde__m128d_from_private(r_);7482
+#endif7483
+}7484
+7485
+SIMDE_FUNCTION_ATTRIBUTES7486
+simde__m128i simde_mm_xor_si128(simde__m128i a, simde__m128i b)7487
+{7488
+#if defined(SIMDE_X86_SSE2_NATIVE)7489
+ return _mm_xor_si128(a, b);7490
+#else7491
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a),7492
+ b_ = simde__m128i_to_private(b);7493
+7494
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)7495
+ r_.neon_i32 = veorq_s32(a_.neon_i32, b_.neon_i32);7496
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)7497
+ r_.altivec_i32 = vec_xor(a_.altivec_i32, b_.altivec_i32);7498
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)7499
+ r_.i32f = a_.i32f ^ b_.i32f;7500
+#else7501
+ SIMDE_VECTORIZE7502
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {7503
+ r_.i32f[i] = a_.i32f[i] ^ b_.i32f[i];7504
+ }7505
+#endif7506
+7507
+ return simde__m128i_from_private(r_);7508
+#endif7509
+}7510
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7511
+#define _mm_xor_si128(a, b) simde_mm_xor_si128(a, b)7512
+#endif7513
+7514
+SIMDE_FUNCTION_ATTRIBUTES7515
+simde__m128i simde_x_mm_not_si128(simde__m128i a)7516
+{7517
+#if defined(SIMDE_X86_AVX512VL_NATIVE)7518
+ return _mm_ternarylogic_epi32(a, a, a, 0x55);7519
+#else7520
+ simde__m128i_private r_, a_ = simde__m128i_to_private(a);7521
+7522
+#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)7523
+ r_.neon_i32 = vmvnq_s32(a_.neon_i32);7524
+#elif defined(SIMDE_POWER_ALTIVEC_P6_NATIVE)7525
+ r_.altivec_i32 = vec_nor(a_.altivec_i32, a_.altivec_i32);7526
+#elif defined(SIMDE_WASM_SIMD128_NATIVE)7527
+ r_.wasm_v128 = wasm_v128_not(a_.wasm_v128);7528
+#elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)7529
+ r_.i32f = ~a_.i32f;7530
+#else7531
+ SIMDE_VECTORIZE7532
+ for (size_t i = 0; i < (sizeof(r_.i32f) / sizeof(r_.i32f[0])); i++) {7533
+ r_.i32f[i] = ~(a_.i32f[i]);7534
+ }7535
+#endif7536
+7537
+ return simde__m128i_from_private(r_);7538
+#endif7539
+}7540
+7541
+#define SIMDE_MM_SHUFFLE2(x, y) (((x) << 1) | (y))7542
+#if defined(SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES)7543
+#define _MM_SHUFFLE2(x, y) SIMDE_MM_SHUFFLE2(x, y)7544
+#endif7545
+7546
+SIMDE_END_DECLS_7547
+7548
+HEDLEY_DIAGNOSTIC_POP7549
+7550
+#endif /* !defined(SIMDE_X86_SSE2_H) */7551
obs-studio-26.1.0.tar.xz/libobs/util/sse-intrin.h -> obs-studio-26.1.1.tar.xz/libobs/util/sse-intrin.h
Changed
61
1
2
3
#pragma once4
5
-#if NEEDS_SIMDE6
-7
-#include "simde/sse2.h"8
-9
-#define __m128 simde__m12810
-#define _mm_setzero_ps simde_mm_setzero_ps11
-#define _mm_set_ps simde_mm_set_ps12
-#define _mm_add_ps simde_mm_add_ps13
-#define _mm_sub_ps simde_mm_sub_ps14
-#define _mm_mul_ps simde_mm_mul_ps15
-#define _mm_div_ps simde_mm_div_ps16
-#define _mm_set1_ps simde_mm_set1_ps17
-#define _mm_movehl_ps simde_mm_movehl_ps18
-#define _mm_shuffle_ps simde_mm_shuffle_ps19
-#define _mm_min_ps simde_mm_min_ps20
-#define _mm_max_ps simde_mm_max_ps21
-#define _mm_movelh_ps simde_mm_movelh_ps22
-#define _mm_unpacklo_ps simde_mm_unpacklo_ps23
-#define _mm_unpackhi_ps simde_mm_unpackhi_ps24
-#define _mm_load_ps simde_mm_load_ps25
-#define _mm_andnot_ps simde_mm_andnot_ps26
-#define _mm_storeu_ps simde_mm_storeu_ps27
-#define _mm_loadu_ps simde_mm_loadu_ps28
-29
-#define __m128i simde__m128i30
-#define _mm_set1_epi32 simde_mm_set1_epi3231
-#define _mm_set1_epi16 simde_mm_set1_epi1632
-#define _mm_load_si128 simde_mm_load_si12833
-#define _mm_packs_epi32 simde_mm_packs_epi3234
-#define _mm_srli_si128 simde_mm_srli_si12835
-#define _mm_and_si128 simde_mm_and_si12836
-#define _mm_packus_epi16 simde_mm_packus_epi1637
-#define _mm_add_epi64 simde_mm_add_epi6438
-#define _mm_shuffle_epi32 simde_mm_shuffle_epi3239
-#define _mm_srai_epi16 simde_mm_srai_epi1640
-#define _mm_shufflelo_epi16 simde_mm_shufflelo_epi1641
-#define _mm_storeu_si128 simde_mm_storeu_si12842
-43
-#define _MM_SHUFFLE SIMDE_MM_SHUFFLE44
-#define _MM_TRANSPOSE4_PS SIMDE_MM_TRANSPOSE4_PS45
-46
-#else47
-48
-#if defined(__aarch64__) || defined(__arm__)49
-#include <arm_neon.h>50
-#include "sse2neon.h"51
-#else52
-#include <xmmintrin.h>53
+#if defined(_MSC_VER)54
#include <emmintrin.h>55
-#endif56
-57
+#else58
+#define SIMDE_ENABLE_NATIVE_ALIASES59
+#include "simde/x86/sse2.h"60
#endif61
obs-studio-26.1.0.tar.xz/plugins/coreaudio-encoder/CMakeLists.txt -> obs-studio-26.1.1.tar.xz/plugins/coreaudio-encoder/CMakeLists.txt
Changed
25
1
2
encoder.cpp)3
4
if (WIN32)5
+ # Set compiler flag before adding resource file6
+ if (MINGW)7
+ set_source_files_properties(${coreaudio-encoder_SOURCES}8
+ PROPERTIES COMPILE_FLAGS "-Wno-multichar")9
+ endif()10
+11
set(MODULE_DESCRIPTION "OBS Core Audio encoder")12
configure_file(${CMAKE_SOURCE_DIR}/cmake/winrc/obs-module.rc.in coreaudio-encoder.rc)13
list(APPEND coreaudio-encoder_SOURCES14
coreaudio-encoder.rc)15
set(coreaudio-encoder_HEADERS windows-imports.h)16
set(coreaudio-encoder_LIBS )17
-18
- if (MINGW)19
- set_source_files_properties(${coreaudio-encoder_SOURCES}20
- PROPERTIES COMPILE_FLAGS "-Wno-multichar")21
- endif()22
else()23
find_library(COREFOUNDATION CoreFoundation)24
find_library(COREAUDIO CoreAudio)25
obs-studio-26.1.0.tar.xz/plugins/decklink/DecklinkInput.cpp -> obs-studio-26.1.1.tar.xz/plugins/decklink/DecklinkInput.cpp
Changed
10
1
2
return false;3
}4
5
- if (!instance->StartCapture(mode, bmdVideoConnection,6
+ if (!instance->StartCapture(mode, allow10Bit, bmdVideoConnection,7
bmdAudioConnection)) {8
instance = nullptr;9
return false;10
obs-studio-26.1.0.tar.xz/plugins/decklink/DecklinkInput.hpp -> obs-studio-26.1.1.tar.xz/plugins/decklink/DecklinkInput.hpp
Changed
9
1
2
std::string hash;3
long long id;4
bool swap = false;5
+ bool allow10Bit = false;6
BMDVideoConnection videoConnection;7
BMDAudioConnection audioConnection;8
};9
obs-studio-26.1.0.tar.xz/plugins/decklink/OBSVideoFrame.cpp -> obs-studio-26.1.1.tar.xz/plugins/decklink/OBSVideoFrame.cpp
Changed
19
1
2
#include "OBSVideoFrame.h"3
4
-OBSVideoFrame::OBSVideoFrame(long width, long height)5
+OBSVideoFrame::OBSVideoFrame(long width, long height,6
+ BMDPixelFormat pixelFormat)7
{8
+ int bpp = 2;9
this->width = width;10
this->height = height;11
- this->rowBytes = width * 2;12
- this->data = new unsigned char[width * height * 2 + 1];13
+ this->rowBytes = width * bpp;14
+ this->data = new unsigned char[width * height * bpp + 1];15
+ this->pixelFormat = pixelFormat;16
}17
18
HRESULT OBSVideoFrame::SetFlags(BMDFrameFlags newFlags)19
obs-studio-26.1.0.tar.xz/plugins/decklink/OBSVideoFrame.h -> obs-studio-26.1.1.tar.xz/plugins/decklink/OBSVideoFrame.h
Changed
10
1
2
unsigned char *data;3
4
public:5
- OBSVideoFrame(long width, long height);6
+ OBSVideoFrame(long width, long height, BMDPixelFormat pixelFormat);7
8
HRESULT STDMETHODCALLTYPE SetFlags(BMDFrameFlags newFlags) override;9
10
obs-studio-26.1.0.tar.xz/plugins/decklink/const.h -> obs-studio-26.1.1.tar.xz/plugins/decklink/const.h
Changed
14
1
2
#define AUTO_START "auto_start"3
#define KEYER "keyer"4
#define SWAP "swap"5
+#define ALLOW_10_BIT "allow_10_bit"6
7
#define TEXT_DEVICE obs_module_text("Device")8
#define TEXT_VIDEO_CONNECTION obs_module_text("VideoConnection")9
10
#define TEXT_ENABLE_KEYER obs_module_text("Keyer")11
#define TEXT_SWAP obs_module_text("SwapFC-LFE")12
#define TEXT_SWAP_TOOLTIP obs_module_text("SwapFC-LFE.Tooltip")13
+#define TEXT_ALLOW_10_BIT obs_module_text("Allow10Bit")14
obs-studio-26.1.0.tar.xz/plugins/decklink/data/locale/en-US.ini -> obs-studio-26.1.1.tar.xz/plugins/decklink/data/locale/en-US.ini
Changed
7
1
2
SwapFC-LFE.Tooltip="Swap Front Center Channel and LFE Channel"3
VideoConnection="Video Connection"4
AudioConnection="Audio Connection"5
+Allow10Bit="Allow 10 Bit (Required for SDI captions, may cause performance overhead)"6
\ No newline at end of file7
obs-studio-26.1.0.tar.xz/plugins/decklink/decklink-device-instance.cpp -> obs-studio-26.1.1.tar.xz/plugins/decklink/decklink-device-instance.cpp
Changed
138
1
2
return VIDEO_FORMAT_BGRX;3
4
default:5
- case bmdFormat8BitYUV:;6
+ case bmdFormat8BitYUV:7
+ case bmdFormat10BitYUV:;8
+ return VIDEO_FORMAT_UYVY;9
}10
-11
- return VIDEO_FORMAT_UYVY;12
}13
14
static inline int ConvertChannelFormat(speaker_layout format)15
16
packets->Release();17
}18
19
- IDeckLinkVideoConversion *frameConverter =20
- CreateVideoConversionInstance();21
+ IDeckLinkVideoFrame *frame;22
+ if (videoFrame->GetPixelFormat() != convertFrame->GetPixelFormat()) {23
+ IDeckLinkVideoConversion *frameConverter =24
+ CreateVideoConversionInstance();25
+26
+ frameConverter->ConvertFrame(videoFrame, convertFrame);27
28
- frameConverter->ConvertFrame(videoFrame, convertFrame);29
+ frame = convertFrame;30
+ } else {31
+ frame = videoFrame;32
+ }33
34
void *bytes;35
- if (convertFrame->GetBytes(&bytes) != S_OK) {36
+ if (frame->GetBytes(&bytes) != S_OK) {37
LOG(LOG_WARNING, "Failed to get video frame data");38
return;39
}40
41
currentFrame.data[0] = (uint8_t *)bytes;42
- currentFrame.linesize[0] = (uint32_t)convertFrame->GetRowBytes();43
- currentFrame.width = (uint32_t)convertFrame->GetWidth();44
- currentFrame.height = (uint32_t)convertFrame->GetHeight();45
+ currentFrame.linesize[0] = (uint32_t)frame->GetRowBytes();46
+ currentFrame.width = (uint32_t)frame->GetWidth();47
+ currentFrame.height = (uint32_t)frame->GetHeight();48
currentFrame.timestamp = timestamp;49
50
obs_source_output_video2(51
52
currentFrame.color_range_min,53
currentFrame.color_range_max);54
55
- if (convertFrame) {56
- delete convertFrame;57
+ delete convertFrame;58
+59
+ BMDPixelFormat convertFormat;60
+ switch (pixelFormat) {61
+ case bmdFormat8BitBGRA:62
+ convertFormat = bmdFormat8BitBGRA;63
+ break;64
+ default:65
+ case bmdFormat10BitYUV:66
+ case bmdFormat8BitYUV:;67
+ convertFormat = bmdFormat8BitYUV;68
+ break;69
}70
- convertFrame = new OBSVideoFrame(mode_->GetWidth(), mode_->GetHeight());71
+72
+ convertFrame = new OBSVideoFrame(mode_->GetWidth(), mode_->GetHeight(),73
+ convertFormat);74
75
#ifdef LOG_SETUP_VIDEO_FORMAT76
LOG(LOG_INFO, "Setup video format: %s, %s, %s",77
78
}79
80
bool DeckLinkDeviceInstance::StartCapture(DeckLinkDeviceMode *mode_,81
+ bool allow10Bit_,82
BMDVideoConnection bmdVideoConnection,83
BMDAudioConnection bmdAudioConnection)84
{85
86
bool isauto = mode_->GetName() == "Auto";87
if (isauto) {88
displayMode = bmdModeNTSC;89
- pixelFormat = bmdFormat10BitYUV;90
+ if (allow10Bit) {91
+ pixelFormat = bmdFormat10BitYUV;92
+ } else {93
+ pixelFormat = bmdFormat8BitYUV;94
+ }95
flags = bmdVideoInputEnableFormatDetection;96
} else {97
displayMode = mode_->GetDisplayMode();98
99
flags = bmdVideoInputFlagDefault;100
}101
102
+ allow10Bit = allow10Bit_;103
+104
const HRESULT videoResult =105
input->EnableVideoInput(displayMode, pixelFormat, flags);106
if (videoResult != S_OK) {107
108
{109
110
if (events & bmdVideoInputColorspaceChanged) {111
- switch (detectedSignalFlags) {112
- case bmdDetectedVideoInputRGB444:113
+ if (detectedSignalFlags & bmdDetectedVideoInputRGB444) {114
pixelFormat = bmdFormat8BitBGRA;115
- break;116
-117
- default:118
- case bmdDetectedVideoInputYCbCr422:119
- pixelFormat = bmdFormat10BitYUV;120
- break;121
+ }122
+ if (detectedSignalFlags & bmdDetectedVideoInputYCbCr422) {123
+ if (detectedSignalFlags &124
+ bmdDetectedVideoInput10BitDepth) {125
+ if (allow10Bit) {126
+ pixelFormat = bmdFormat10BitYUV;127
+ } else {128
+ pixelFormat = bmdFormat8BitYUV;129
+ }130
+ }131
+ if (detectedSignalFlags &132
+ bmdDetectedVideoInput8BitDepth) {133
+ pixelFormat = bmdFormat8BitYUV;134
+ }135
}136
}137
138
obs-studio-26.1.0.tar.xz/plugins/decklink/decklink-device-instance.hpp -> obs-studio-26.1.1.tar.xz/plugins/decklink/decklink-device-instance.hpp
Changed
18
1
2
AudioRepacker *audioRepacker = nullptr;3
speaker_layout channelFormat = SPEAKERS_STEREO;4
bool swap;5
+ bool allow10Bit;6
7
OBSVideoFrame *convertFrame = nullptr;8
IDeckLinkMutableVideoFrame *decklinkOutputFrame = nullptr;9
10
11
inline DeckLinkDeviceMode *GetMode() const { return mode; }12
13
- bool StartCapture(DeckLinkDeviceMode *mode,14
+ bool StartCapture(DeckLinkDeviceMode *mode, bool allow10Bit,15
BMDVideoConnection bmdVideoConnection,16
BMDAudioConnection bmdAudioConnection);17
bool StopCapture(void);18
obs-studio-26.1.0.tar.xz/plugins/decklink/decklink-source.cpp -> obs-studio-26.1.1.tar.xz/plugins/decklink/decklink-source.cpp
Changed
36
1
2
decklink->SetChannelFormat(channelFormat);3
decklink->hash = std::string(hash);4
decklink->swap = obs_data_get_bool(settings, SWAP);5
+ decklink->allow10Bit = obs_data_get_bool(settings, ALLOW_10_BIT);6
decklink->Activate(device, id, videoConnection, audioConnection);7
}8
9
10
list = obs_properties_get(props, PIXEL_FORMAT);11
obs_property_set_visible(list, id != MODE_ID_AUTO);12
13
+ auto allow10BitProp = obs_properties_get(props, ALLOW_10_BIT);14
+ obs_property_set_visible(allow10BitProp, id == MODE_ID_AUTO);15
+16
return true;17
}18
19
20
OBS_COMBO_FORMAT_INT);21
22
obs_property_list_add_int(list, "8-bit YUV", bmdFormat8BitYUV);23
+ obs_property_list_add_int(list, "10-bit YUV", bmdFormat10BitYUV);24
obs_property_list_add_int(list, "8-bit BGRA", bmdFormat8BitBGRA);25
26
list = obs_properties_add_list(props, COLOR_SPACE, TEXT_COLOR_SPACE,27
28
29
obs_properties_add_bool(props, DEACTIVATE_WNS, TEXT_DWNS);30
31
+ obs_properties_add_bool(props, ALLOW_10_BIT, TEXT_ALLOW_10_BIT);32
+33
UNUSED_PARAMETER(data);34
return props;35
}36
obs-studio-26.1.0.tar.xz/plugins/image-source/image-source.c -> obs-studio-26.1.1.tar.xz/plugins/image-source/image-source.c
Changed
17
1
2
}3
4
static const char *image_filter =5
- "All formats (*.bmp *.tga *.png *.jpeg *.jpg *.gif *.psd);;"6
+ "All formats (*.bmp *.tga *.png *.jpeg *.jpg *.gif *.psd *.webp);;"7
"BMP Files (*.bmp);;"8
"Targa Files (*.tga);;"9
"PNG Files (*.png);;"10
"JPEG Files (*.jpeg *.jpg);;"11
"GIF Files (*.gif);;"12
"PSD Files (*.psd);;"13
+ "WebP Files (*.webp);;"14
"All Files (*.*)";15
16
static obs_properties_t *image_source_properties(void *data)17
obs-studio-26.1.0.tar.xz/plugins/image-source/obs-slideshow.c -> obs-studio-26.1.1.tar.xz/plugins/image-source/obs-slideshow.c
Changed
22
1
2
if (!ss->transition || !ss->slide_time)3
return;4
5
- if (ss->restart_on_activate && !ss->randomize && ss->use_cut) {6
+ if (ss->restart_on_activate && ss->use_cut) {7
ss->elapsed = 0.0f;8
- ss->cur_item = 0;9
+ ss->cur_item = ss->randomize ? random_file(ss) : 0;10
do_transition(ss, false);11
ss->restart_on_activate = false;12
ss->use_cut = false;13
14
}15
16
static const char *file_filter =17
- "Image files (*.bmp *.tga *.png *.jpeg *.jpg *.gif)";18
+ "Image files (*.bmp *.tga *.png *.jpeg *.jpg *.gif *.webp)";19
20
static const char *aspects[] = {"16:9", "16:10", "4:3", "1:1"};21
22
obs-studio-26.1.0.tar.xz/plugins/linux-jack/jack-wrapper.c -> obs-studio-26.1.1.tar.xz/plugins/linux-jack/jack-wrapper.c
Changed
70
1
2
int jack_process_callback(jack_nframes_t nframes, void *arg)3
{4
struct jack_data *data = (struct jack_data *)arg;5
+ jack_nframes_t current_frames;6
+ jack_time_t current_usecs, next_usecs;7
+ float period_usecs;8
+9
+ uint64_t now = os_gettime_ns();10
+11
if (data == 0)12
return 0;13
14
- pthread_mutex_lock(&data->jack_mutex);15
-16
struct obs_source_audio out;17
out.speakers = jack_channels_to_obs_speakers(data->channels);18
out.samples_per_sec = jack_get_sample_rate(data->jack_client);19
20
}21
22
out.frames = nframes;23
- out.timestamp = os_gettime_ns() -24
- jack_frames_to_time(data->jack_client, nframes);25
+ if (!jack_get_cycle_times(data->jack_client, ¤t_frames,26
+ ¤t_usecs, &next_usecs, &period_usecs)) {27
+ out.timestamp = now - (int64_t)(period_usecs * 1000);28
+ } else {29
+ out.timestamp = now - util_mul_div64(nframes, 1000000000ULL,30
+ data->samples_per_sec);31
+ blog(LOG_WARNING,32
+ "jack_get_cycle_times error: guessing timestamp");33
+ }34
35
+ /* FIXME: this function is not realtime-safe, we should do something36
+ * about this */37
obs_source_output_audio(data->source, &out);38
- pthread_mutex_unlock(&data->jack_mutex);39
return 0;40
}41
42
43
44
data->jack_ports[i] = jack_port_register(45
data->jack_client, port_name, JACK_DEFAULT_AUDIO_TYPE,46
- JackPortIsInput, 0);47
+ JackPortIsInput | JackPortIsTerminal, 0);48
if (data->jack_ports[i] == NULL) {49
blog(LOG_ERROR,50
"jack_port_register Error:"51
52
pthread_mutex_lock(&data->jack_mutex);53
54
if (data->jack_client) {55
+ jack_client_close(data->jack_client);56
if (data->jack_ports != NULL) {57
- for (int i = 0; i < data->channels; ++i) {58
- if (data->jack_ports[i] != NULL)59
- jack_port_unregister(60
- data->jack_client,61
- data->jack_ports[i]);62
- }63
bfree(data->jack_ports);64
data->jack_ports = NULL;65
}66
- jack_client_close(data->jack_client);67
data->jack_client = NULL;68
}69
pthread_mutex_unlock(&data->jack_mutex);70
obs-studio-26.1.0.tar.xz/plugins/mac-virtualcam/src/dal-plugin/CMSampleBufferUtils.mm -> obs-studio-26.1.1.tar.xz/plugins/mac-virtualcam/src/dal-plugin/CMSampleBufferUtils.mm
Changed
11
1
2
3
static void releaseNSData(void *o, void *block, size_t size)4
{5
+ UNUSED_PARAMETER(block);6
+ UNUSED_PARAMETER(size);7
+8
NSData *data = (__bridge_transfer NSData *)o;9
data = nil; // Assuming ARC is enabled10
}11
obs-studio-26.1.0.tar.xz/plugins/mac-virtualcam/src/dal-plugin/Logging.h -> obs-studio-26.1.1.tar.xz/plugins/mac-virtualcam/src/dal-plugin/Logging.h
Changed
8
1
2
#define VLogFunc(fmt, ...)3
#define ELog(fmt, ...) DLog(fmt, ##__VA_ARGS__)4
5
+#define UNUSED_PARAMETER(param) (void)param6
+7
#endif /* Logging_h */8
obs-studio-26.1.0.tar.xz/plugins/mac-virtualcam/src/dal-plugin/OBSDALDevice.mm -> obs-studio-26.1.1.tar.xz/plugins/mac-virtualcam/src/dal-plugin/OBSDALDevice.mm
Changed
53
1
2
case kCMIODevicePropertyDeviceMaster:3
return sizeof(pid_t);4
default:5
- DLog(@"Device unhandled getPropertyDataSizeWithAddress for %@",6
- [OBSDALObjectStore7
- StringFromPropertySelector:address.mSelector]);8
+ break;9
};10
11
return 0;12
13
*dataUsed = sizeof(pid_t);14
break;15
default:16
- DLog(@"Device unhandled getPropertyDataWithAddress for %@",17
- [OBSDALObjectStore18
- StringFromPropertySelector:address.mSelector]);19
- *dataUsed = 0;20
break;21
};22
}23
24
case kCMIODevicePropertyLinkedCoreAudioDeviceUID:25
return false;26
default:27
- DLog(@"Device unhandled hasPropertyWithAddress for %@",28
- [OBSDALObjectStore29
- StringFromPropertySelector:address.mSelector]);30
return false;31
};32
}33
34
case kCMIODevicePropertyDeviceMaster:35
return true;36
default:37
- DLog(@"Device unhandled isPropertySettableWithAddress for %@",38
- [OBSDALObjectStore39
- StringFromPropertySelector:address.mSelector]);40
return false;41
};42
}43
44
self.masterPid = *static_cast<const pid_t *>(data);45
break;46
default:47
- DLog(@"Device unhandled setPropertyDataWithAddress for %@",48
- [OBSDALObjectStore49
- StringFromPropertySelector:address.mSelector]);50
break;51
};52
}53
obs-studio-26.1.0.tar.xz/plugins/mac-virtualcam/src/dal-plugin/OBSDALPlugInInterface.mm -> obs-studio-26.1.1.tar.xz/plugins/mac-virtualcam/src/dal-plugin/OBSDALPlugInInterface.mm
Changed
136
1
2
3
ULONG HardwarePlugIn_AddRef(CMIOHardwarePlugInRef self)4
{5
+ UNUSED_PARAMETER(self);6
+7
sRefCount += 1;8
DLogFunc(@"sRefCount now = %d", sRefCount);9
return sRefCount;10
11
12
ULONG HardwarePlugIn_Release(CMIOHardwarePlugInRef self)13
{14
+ UNUSED_PARAMETER(self);15
+16
sRefCount -= 1;17
DLogFunc(@"sRefCount now = %d", sRefCount);18
return sRefCount;19
20
HRESULT HardwarePlugIn_QueryInterface(CMIOHardwarePlugInRef self, REFIID uuid,21
LPVOID *interface)22
{23
+ UNUSED_PARAMETER(self);24
DLogFunc(@"");25
26
if (!interface) {27
28
void HardwarePlugIn_ObjectShow(CMIOHardwarePlugInRef self,29
CMIOObjectID objectID)30
{31
+ UNUSED_PARAMETER(objectID);32
DLogFunc(@"self=%p", self);33
}34
35
36
CMIOObjectID objectID,37
const CMIOObjectPropertyAddress *address)38
{39
+ UNUSED_PARAMETER(self);40
41
NSObject<CMIOObject> *object =42
[OBSDALObjectStore GetObjectWithId:objectID];43
44
const CMIOObjectPropertyAddress *address, UInt32 qualifierDataSize,45
const void *qualifierData, UInt32 *dataSize)46
{47
+ UNUSED_PARAMETER(self);48
49
NSObject<CMIOObject> *object =50
[OBSDALObjectStore GetObjectWithId:objectID];51
52
const void *qualifierData, UInt32 dataSize, UInt32 *dataUsed,53
void *data)54
{55
+ UNUSED_PARAMETER(self);56
57
NSObject<CMIOObject> *object =58
[OBSDALObjectStore GetObjectWithId:objectID];59
60
OSStatus HardwarePlugIn_DeviceSuspend(CMIOHardwarePlugInRef self,61
CMIODeviceID deviceID)62
{63
+ UNUSED_PARAMETER(deviceID);64
+65
DLogFunc(@"self=%p", self);66
return kCMIOHardwareNoError;67
}68
69
OSStatus HardwarePlugIn_DeviceResume(CMIOHardwarePlugInRef self,70
CMIODeviceID deviceID)71
{72
+ UNUSED_PARAMETER(deviceID);73
+74
DLogFunc(@"self=%p", self);75
return kCMIOHardwareNoError;76
}77
78
CMIODeviceID deviceID,79
CMIODeviceAVCCommand *ioAVCCommand)80
{81
+ UNUSED_PARAMETER(deviceID);82
+ UNUSED_PARAMETER(ioAVCCommand);83
+84
DLogFunc(@"self=%p", self);85
return kCMIOHardwareNoError;86
}87
88
CMIODeviceID deviceID,89
CMIODeviceRS422Command *ioRS422Command)90
{91
+ UNUSED_PARAMETER(deviceID);92
+ UNUSED_PARAMETER(ioRS422Command);93
+94
DLogFunc(@"self=%p", self);95
return kCMIOHardwareNoError;96
}97
98
OSStatus HardwarePlugIn_StreamDeckPlay(CMIOHardwarePlugInRef self,99
CMIOStreamID streamID)100
{101
+ UNUSED_PARAMETER(streamID);102
+103
DLogFunc(@"self=%p", self);104
return kCMIOHardwareIllegalOperationError;105
}106
107
OSStatus HardwarePlugIn_StreamDeckStop(CMIOHardwarePlugInRef self,108
CMIOStreamID streamID)109
{110
+ UNUSED_PARAMETER(streamID);111
+112
DLogFunc(@"self=%p", self);113
return kCMIOHardwareIllegalOperationError;114
}115
116
OSStatus HardwarePlugIn_StreamDeckJog(CMIOHardwarePlugInRef self,117
CMIOStreamID streamID, SInt32 speed)118
{119
+ UNUSED_PARAMETER(streamID);120
+ UNUSED_PARAMETER(speed);121
+122
DLogFunc(@"self=%p", self);123
return kCMIOHardwareIllegalOperationError;124
}125
126
Float64 requestedTimecode,127
Boolean playOnCue)128
{129
+ UNUSED_PARAMETER(streamID);130
+ UNUSED_PARAMETER(requestedTimecode);131
+ UNUSED_PARAMETER(playOnCue);132
+133
DLogFunc(@"self=%p", self);134
return kCMIOHardwareIllegalOperationError;135
}136
obs-studio-26.1.0.tar.xz/plugins/mac-virtualcam/src/dal-plugin/OBSDALPluginMain.mm -> obs-studio-26.1.1.tar.xz/plugins/mac-virtualcam/src/dal-plugin/OBSDALPluginMain.mm
Changed
10
1
2
extern "C" {3
void *PlugInMain(CFAllocatorRef allocator, CFUUIDRef requestedTypeUUID)4
{5
+ UNUSED_PARAMETER(allocator);6
+7
DLogFunc(@"version=%@", PLUGIN_VERSION);8
if (!CFEqual(requestedTypeUUID, kCMIOHardwarePlugInTypeID)) {9
return 0;10
obs-studio-26.1.0.tar.xz/plugins/mac-virtualcam/src/dal-plugin/OBSDALStream.mm -> obs-studio-26.1.1.tar.xz/plugins/mac-virtualcam/src/dal-plugin/OBSDALStream.mm
Changed
56
1
2
- (void)fillFrame3
{4
if (CMSimpleQueueGetFullness(self.queue) >= 1.0) {5
- DLog(@"Queue is full, bailing out");6
return;7
}8
9
10
case kCMIOStreamPropertyClock:11
return sizeof(CFTypeRef);12
default:13
- DLog(@"Stream unhandled getPropertyDataSizeWithAddress for %@",14
- [OBSDALObjectStore15
- StringFromPropertySelector:address.mSelector]);16
return 0;17
};18
}19
20
*dataUsed = sizeof(CFTypeRef);21
break;22
default:23
- DLog(@"Stream unhandled getPropertyDataWithAddress for %@",24
- [OBSDALObjectStore25
- StringFromPropertySelector:address.mSelector]);26
*dataUsed = 0;27
};28
}29
30
StringFromPropertySelector:address.mSelector]);31
return false;32
default:33
- DLog(@"Stream unhandled hasPropertyWithAddress for %@",34
- [OBSDALObjectStore35
- StringFromPropertySelector:address.mSelector]);36
return false;37
};38
}39
40
- (BOOL)isPropertySettableWithAddress:(CMIOObjectPropertyAddress)address41
{42
- DLog(@"Stream unhandled isPropertySettableWithAddress for %@",43
- [OBSDALObjectStore StringFromPropertySelector:address.mSelector]);44
return false;45
}46
47
48
dataSize:(UInt32)dataSize49
data:(nonnull const void *)data50
{51
- DLog(@"Stream unhandled setPropertyDataWithAddress for %@",52
- [OBSDALObjectStore StringFromPropertySelector:address.mSelector]);53
}54
55
@end56
obs-studio-26.1.0.tar.xz/plugins/mac-virtualcam/src/obs-plugin/plugin-main.mm -> obs-studio-26.1.1.tar.xz/plugins/mac-virtualcam/src/obs-plugin/plugin-main.mm
Changed
45
1
2
static void *virtualcam_output_create(obs_data_t *settings,3
obs_output_t *output)4
{5
+ UNUSED_PARAMETER(settings);6
+7
outputRef = output;8
9
blog(LOG_DEBUG, "output_create");10
11
12
static void virtualcam_output_destroy(void *data)13
{14
+ UNUSED_PARAMETER(data);15
blog(LOG_DEBUG, "output_destroy");16
sMachServer = nil;17
}18
19
static bool virtualcam_output_start(void *data)20
{21
+ UNUSED_PARAMETER(data);22
+23
bool hasDalPlugin = check_dal_plugin();24
25
if (!hasDalPlugin) {26
27
28
static void virtualcam_output_stop(void *data, uint64_t ts)29
{30
+ UNUSED_PARAMETER(data);31
+ UNUSED_PARAMETER(ts);32
+33
blog(LOG_DEBUG, "output_stop");34
obs_output_end_data_capture(outputRef);35
[sMachServer stop];36
37
38
static void virtualcam_output_raw_video(void *data, struct video_data *frame)39
{40
+ UNUSED_PARAMETER(data);41
+42
uint8_t *outData = frame->data[0];43
if (frame->linesize[0] != (videoInfo.output_width * 2)) {44
blog(LOG_ERROR,45
obs-studio-26.1.0.tar.xz/plugins/obs-browser/obs-browser-plugin.cpp -> obs-studio-26.1.1.tar.xz/plugins/obs-browser/obs-browser-plugin.cpp
Changed
14
1
2
}3
obs_data_release(private_data);4
#endif5
+6
+#if defined(__APPLE__) && CHROME_VERSION_BUILD < 41837
+ // Make sure CEF malloc hijacking happens early in the process8
+ obs_browser_initialize();9
+#endif10
+11
return true;12
}13
14
obs-studio-26.1.0.tar.xz/plugins/obs-ffmpeg/ffmpeg-mux/ffmpeg-mux.c -> obs-studio-26.1.1.tar.xz/plugins/obs-ffmpeg/ffmpeg-mux/ffmpeg-mux.c
Changed
10
1
2
}3
4
/* Treat "Invalid data found when processing input" and "Invalid argument" as non-fatal */5
- if (ret == AVERROR_INVALIDDATA || ret == EINVAL) {6
+ if (ret == AVERROR_INVALIDDATA || ret == -EINVAL) {7
return true;8
}9
10
obs-studio-26.1.0.tar.xz/plugins/rtmp-services/data/package.json -> obs-studio-26.1.1.tar.xz/plugins/rtmp-services/data/package.json
Changed
14
1
2
{3
"url": "https://obsproject.com/obs2_update/rtmp-services",4
- "version": 161,5
+ "version": 163,6
"files": [7
{8
"name": "services.json",9
- "version": 16110
+ "version": 16311
}12
]13
}14
obs-studio-26.1.0.tar.xz/plugins/rtmp-services/data/services.json -> obs-studio-26.1.1.tar.xz/plugins/rtmp-services/data/services.json
Changed
136
1
2
}3
},4
{5
- "name": "VIMM",6
+ "name": "Loola.tv",7
+ "common": false,8
"servers": [9
{10
- "name": "Europe: Frankfurt",11
- "url": "rtmp://eu.vimm.tv/live"12
+ "name": "US East: Virginia",13
+ "url": "rtmp://rtmp.loola.tv/push"14
},15
{16
- "name": "North America: Montreal",17
- "url": "rtmp://us.vimm.tv/live"18
+ "name": "EU Central: Germany",19
+ "url": "rtmp://rtmp-eu.loola.tv/push"20
+ },21
+ {22
+ "name": "South America: Brazil",23
+ "url": "rtmp://rtmp-sa.loola.tv/push"24
+ },25
+ {26
+ "name": "Asia/Pacific: Singapore",27
+ "url": "rtmp://rtmp-sg.loola.tv/push"28
+ },29
+ {30
+ "name": "Middle East: Bahrain",31
+ "url": "rtmp://rtmp-me.loola.tv/push"32
}33
],34
"recommended": {35
"keyint": 2,36
- "max video bitrate": 8000,37
- "max audio bitrate": 320,38
+ "profile": "high",39
+ "max video bitrate": 2500,40
+ "max audio bitrate": 160,41
+ "bframes": 2,42
"x264opts": "scenecut=0"43
}44
},45
{46
- "name": "Smashcast",47
+ "name": "VIMM",48
"servers": [49
{50
- "name": "Default",51
- "url": "rtmp://live.hitbox.tv/push"52
- },53
- {54
- "name": "EU-North: Amsterdam, Netherlands",55
- "url": "rtmp://live.ams.hitbox.tv/push"56
- },57
- {58
- "name": "EU-West: Paris, France",59
- "url": "rtmp://live.cdg.hitbox.tv/push"60
- },61
- {62
- "name": "EU-South: Milan, Italia",63
- "url": "rtmp://live.mxp.hitbox.tv/push"64
- },65
- {66
- "name": "Russia: Moscow",67
- "url": "rtmp://live.dme.hitbox.tv/push"68
- },69
- {70
- "name": "US-East: New York",71
- "url": "rtmp://live.jfk.hitbox.tv/push"72
- },73
- {74
- "name": "US-West: San Francisco",75
- "url": "rtmp://live.sfo.hitbox.tv/push"76
- },77
- {78
- "name": "US-West: Los Angeles",79
- "url": "rtmp://live.lax.hitbox.tv/push"80
- },81
- {82
- "name": "South America: Sao Paulo, Brazil",83
- "url": "rtmp://live.gru.hitbox.tv/push"84
- },85
- {86
- "name": "Asia: Singapore",87
- "url": "rtmp://live.sin.hitbox.tv/push"88
+ "name": "Europe: Frankfurt",89
+ "url": "rtmp://eu.vimm.tv/live"90
},91
{92
- "name": "Oceania: Sydney, Australia",93
- "url": "rtmp://live.syd.hitbox.tv/push"94
+ "name": "North America: Montreal",95
+ "url": "rtmp://us.vimm.tv/live"96
}97
],98
"recommended": {99
"keyint": 2,100
- "profile": "high",101
- "max video bitrate": 3500,102
- "max audio bitrate": 320103
+ "max video bitrate": 8000,104
+ "max audio bitrate": 320,105
+ "x264opts": "scenecut=0"106
}107
},108
{109
110
{111
"name": "US: New York, NY",112
"url": "rtmp://live-nyc.vaughnsoft.net/live"113
- }, 114
+ },115
{116
"name": "US: Miami, FL",117
"url": "rtmp://live-mia.vaughnsoft.net/live"118
119
{120
"name": "US: New York, NY",121
"url": "rtmp://live-nyc.vaughnsoft.net/live"122
- }, 123
+ },124
{125
"name": "US: Miami, FL",126
"url": "rtmp://live-mia.vaughnsoft.net/live"127
128
"max audio bitrate": 160,129
"x264opts": "tune=zerolatency"130
}131
- },132
+ },133
{134
"name": "Mux",135
"servers": [136
obs-studio-26.1.0.tar.xz/plugins/win-dshow/libdshowcapture/dshowcapture.hpp -> obs-studio-26.1.1.tar.xz/plugins/win-dshow/libdshowcapture/dshowcapture.hpp
Changed
10
1
2
3
#define DSHOWCAPTURE_VERSION_MAJOR 04
#define DSHOWCAPTURE_VERSION_MINOR 85
-#define DSHOWCAPTURE_VERSION_PATCH 56
+#define DSHOWCAPTURE_VERSION_PATCH 67
8
#define MAKE_DSHOWCAPTURE_VERSION(major, minor, patch) \9
((major << 24) | (minor << 16) | (patch))10
obs-studio-26.1.0.tar.xz/plugins/win-dshow/libdshowcapture/source/dshow-base.cpp -> obs-studio-26.1.1.tar.xz/plugins/win-dshow/libdshowcapture/source/dshow-base.cpp
Changed
198
1
2
return hr;3
}4
5
+static HRESULT GetFriendlyName(REFCLSID deviceClass, const wchar_t *devPath,6
+ wchar_t *name, int nameSize)7
+{8
+ /* Sanity checks */9
+ if (!devPath)10
+ return E_POINTER;11
+ if (!name)12
+ return E_POINTER;13
+14
+ /* Create device enumerator */15
+ ComPtr<ICreateDevEnum> createDevEnum;16
+ HRESULT hr = CoCreateInstance(CLSID_SystemDeviceEnum, NULL,17
+ CLSCTX_INPROC_SERVER, IID_ICreateDevEnum,18
+ (void **)&createDevEnum);19
+20
+ /* Enumerate filters */21
+ ComPtr<IEnumMoniker> enumMoniker;22
+ if (SUCCEEDED(hr)) {23
+ /* returns S_FALSE if no devices are installed */24
+ hr = createDevEnum->CreateClassEnumerator(deviceClass,25
+ &enumMoniker, 0);26
+ if (!enumMoniker)27
+ hr = E_FAIL;28
+ }29
+30
+ /* Cycle through the enumeration */31
+ if (SUCCEEDED(hr)) {32
+ ULONG fetched = 0;33
+ ComPtr<IMoniker> moniker;34
+35
+ enumMoniker->Reset();36
+37
+ while (enumMoniker->Next(1, &moniker, &fetched) == S_OK) {38
+39
+ /* Get device path from moniker */40
+ wchar_t monikerDevPath[512];41
+ hr = ReadProperty(moniker, L"DevicePath",42
+ monikerDevPath,43
+ _ARRAYSIZE(monikerDevPath));44
+45
+ /* Find desired filter */46
+ if (wcscmp(devPath, monikerDevPath) == 0) {47
+48
+ /* Get friendly name */49
+ hr = ReadProperty(moniker, L"FriendlyName",50
+ name, nameSize);51
+ return hr;52
+ }53
+ }54
+ }55
+56
+ return E_FAIL;57
+}58
+59
+static bool MatchFriendlyNames(const wchar_t *vidName, const wchar_t *audName)60
+{61
+ /* Sanity checks */62
+ if (!vidName)63
+ return false;64
+ if (!audName)65
+ return false;66
+67
+ /* Convert strings to lower case */68
+ wstring strVidName = vidName;69
+ for (wchar_t &c : strVidName)70
+ c = (wchar_t)tolower(c);71
+ wstring strAudName = audName;72
+ for (wchar_t &c : strAudName)73
+ c = (wchar_t)tolower(c);74
+75
+ /* Remove 'video' from friendly name */76
+ size_t posVid;77
+ wstring searchVid[] = {L"(video) ", L"(video)", L"video ", L"video"};78
+ for (int i = 0; i < _ARRAYSIZE(searchVid); i++) {79
+ wstring &search = searchVid[i];80
+ while ((posVid = strVidName.find(search)) !=81
+ std::string::npos) {82
+ strVidName.replace(posVid, search.length(), L"");83
+ }84
+ }85
+86
+ /* Remove 'audio' from friendly name */87
+ size_t posAud;88
+ wstring searchAud[] = {L"(audio) ", L"(audio)", L"audio ", L"audio"};89
+ for (int i = 0; i < _ARRAYSIZE(searchAud); i++) {90
+ wstring &search = searchAud[i];91
+ while ((posAud = strAudName.find(search)) !=92
+ std::string::npos) {93
+ strAudName.replace(posAud, search.length(), L"");94
+ }95
+ }96
+97
+ return strVidName == strAudName;98
+}99
+100
static bool GetDeviceAudioFilterInternal(REFCLSID deviceClass,101
const wchar_t *vidDevPath,102
- IBaseFilter **audioCaptureFilter)103
+ IBaseFilter **audioCaptureFilter,104
+ bool matchFilterName = false)105
{106
/* Get video device instance path */107
wchar_t vidDevInstPath[512];108
109
return false;110
#endif111
112
+ /* Get friendly name */113
+ wchar_t vidName[512];114
+ if (matchFilterName) {115
+ hr = GetFriendlyName(CLSID_VideoInputDeviceCategory, vidDevPath,116
+ vidName, _ARRAYSIZE(vidName));117
+ if (FAILED(hr))118
+ return false;119
+ }120
+121
/* Create device enumerator */122
ComPtr<ICreateDevEnum> createDevEnum;123
if (SUCCEEDED(hr))124
125
126
while (enumMoniker->Next(1, &moniker, &fetched) == S_OK) {127
bool samePath = false;128
-#if 0129
- /* Get friendly name (helpful for debugging) */130
- wchar_t friendlyName[512];131
- ReadProperty(moniker, L"FriendlyName", friendlyName,132
- _ARRAYSIZE(friendlyName));133
-#endif134
135
/* Get device path */136
wchar_t audDevPath[512];137
138
139
/* Get audio capture filter */140
if (samePath) {141
- hr = moniker->BindToObject(142
- 0, 0, IID_IBaseFilter,143
- (void **)audioCaptureFilter);144
- if (SUCCEEDED(hr))145
- return true;146
+ /* Match video and audio filter names */147
+ bool isSameFilterName = false;148
+ if (matchFilterName) {149
+ wchar_t audName[512];150
+ hr = ReadProperty(moniker,151
+ L"FriendlyName",152
+ audName,153
+ _ARRAYSIZE(audName));154
+ if (SUCCEEDED(hr)) {155
+ isSameFilterName =156
+ MatchFriendlyNames(157
+ vidName,158
+ audName);159
+ }160
+ }161
+162
+ if (!matchFilterName || isSameFilterName) {163
+ hr = moniker->BindToObject(164
+ 0, 0, IID_IBaseFilter,165
+ (void **)audioCaptureFilter);166
+ if (SUCCEEDED(hr))167
+ return true;168
+ }169
}170
}171
}172
173
bool GetDeviceAudioFilter(const wchar_t *vidDevPath,174
IBaseFilter **audioCaptureFilter)175
{176
- /* Search in "Audio capture sources" */177
+ /* Search in "Audio capture sources" and match filter name */178
bool success = GetDeviceAudioFilterInternal(179
- CLSID_AudioInputDeviceCategory, vidDevPath, audioCaptureFilter);180
+ CLSID_AudioInputDeviceCategory, vidDevPath, audioCaptureFilter,181
+ true);182
+183
+ /* Search in "WDM Streaming Capture Devices" and match filter name */184
+ if (!success)185
+ success = GetDeviceAudioFilterInternal(KSCATEGORY_CAPTURE,186
+ vidDevPath,187
+ audioCaptureFilter,188
+ true);189
+190
+ /* Search in "Audio capture sources" */191
+ if (!success)192
+ success = GetDeviceAudioFilterInternal(193
+ CLSID_AudioInputDeviceCategory, vidDevPath,194
+ audioCaptureFilter);195
196
/* Search in "WDM Streaming Capture Devices" */197
if (!success)198
obs-studio-26.1.0.tar.xz/plugins/win-dshow/libdshowcapture/source/output-filter.cpp -> obs-studio-26.1.1.tar.xz/plugins/win-dshow/libdshowcapture/source/output-filter.cpp
Changed
15
1
2
* USA3
*/4
5
-#include <strsafe.h>6
#include "output-filter.hpp"7
#include "dshow-formats.hpp"8
#include "log.hpp"9
10
+#include <strsafe.h>11
+12
namespace DShow {13
14
#if 015