)]}' { "commit": "8d440c181c1bdc884c2d2d6d23a773ae51d178a1", "tree": "58778ec98faece087eb78461c40361f0d5b92285", "parents": [ "63e54f7374680a03beb3fb595ea756354dabaffd" ], "author": { "name": "Frank Barchard", "email": "fbarchard@google.com", "time": "Wed Oct 28 13:27:48 2026 -0700" }, "committer": { "name": "libyuv-scoped@luci-project-accounts.iam.gserviceaccount.com", "email": "libyuv-scoped@luci-project-accounts.iam.gserviceaccount.com", "time": "Thu Oct 29 12:46:05 2026 -0700" }, "message": "[libyuv] Upgrade x86 SIMD to AVX2/AVX-512 with 32-bit \u0026 row_win parity\n\nUpgrade all remaining SSE2 and SSSE3 max-ISA kernels across row, scale,\nrotate, and compare to AVX2 while removing superseded pre-AVX2 versions.\nImplement AVX2 for ARGBDetectRow, HalfRow, HalfRow_16,\nHalfWidthRow_16To8_Odd, AYUVTo{Y,UV,VU}Row, AR30To{ARGB,ABGR,AB30}Row,\nand TransposeWx8_16, and port all remaining Arm NEON (or higher) kernels\nto AVX2 (TransposeWxH_32, ScaleRowDown2*_16, ScaleUVRowDown2*,\nScaleUVRowDownEven, ScaleColsUp2, GaussCol*, Scale*Samples, and\nUnpackMT2T).\n\nPort 32-bit x86 (-m32 -fPIC) inline assembly across row_gcc.cc,\nrotate_gcc.cc, and compare_gcc.cc (including 5-GPR PIC support for\nI422ToAR30Row and I444/I422/I210/I410AlphaToARGBRow) so x86_32 matches\nx64 SIMD coverage across all 187 supported row/rotate/compare kernels.\nHandle 1..7 remainder elements directly in inline assembly without\ncalling *_C in ARGBDetectRow_AVX2, ARGBSepiaRow_AVX2,\nARGBColorMatrixRow_AVX2, ARGBQuantizeRow_AVX2, ARGBShadeRow_AVX2,\nARGBAffineRow_AVX2, GaussRow*_AVX2, GaussCol*_AVX2, and\n*ColorTableRow_AVX2. Fix unsigned 16-bit packing in\nScaleRowUp2_{Linear,Bilinear}_16_SSE2.\n\nAdd 64-bit AVX-512 (AVX512BW, AVX512VBMI, and AVX512VNNI, including\nSumSquareError_AVX512VNNI) implementations in *_gcc.cc and *_win.cc\nwherever Arm64 exceeds NEON, and handle arbitrary tail widths via\nopmasks in Mirror*_AVX512BW without _Any_ wrappers. Bring row_win.cc,\nscale_win.cc, rotate_win.cc, and compare_win.cc C++ intrinsics (and\n64-bit VISUALC_HAS_AVX2) to 100% parity with *_gcc.cc for MSVC and\nMemorySanitizer.\n\nWire YUY2ToUVRow_HVX and UYVYToUVRow_HVX in YUY2ToI420 and UYVYToI420,\nimplement all 13 missing ScaleUV*_16_C kernels so UVScale_16 supports\ndownscaling and general upscaling, make Ssim8x8, TransposeWxH,\nTransposeUVWxH, and TransposeWxH_16 static internal helpers, clean up\nHAS_* guards, enable direct row unittests so all 279 *_C functions have\nactive test coverage, and add Intel AMX-TF32 (kCpuHasAMXTF32, CPUID\nleaf 7 subleaf 1 EAX bit 4, LIBYUV_DISABLE_AMXTF32) detection.\n\nMark 16-bit RGB formats (RGB565, ARGB1555, and ARGB4444) and legacy\nfixed-matrix 16-bit RGB-to-YUV row functions as deprecated in headers\nand GEMINI.md, using two-step ARGB wrapper row functions in\nrow_common.cc.\n\nAMD Zen 6 Benchmark (1280 width, 10000 iterations):\n----------------------------------------------------------------\nFunction Before (ms) After (ms) Improvement\n----------------------------------------------------------------\nARGB1555ToARGBRow 0.83 0.49 1.69x\nARGB1555ToUVRow 16.00 3.21 4.99x\nARGB1555ToYRow 7.82 1.38 5.66x\nARGB4444ToUVRow 15.04 2.58 5.83x\nARGB4444ToYRow 7.42 1.07 6.94x\nARGBAffineRow 3.51 2.75 1.28x\nARGBColorMatrixRow 4.09 1.13 3.63x\nARGBColorTableRow 11.21 8.93 1.26x\nARGBLumaColorTableRow 10.01 10.21 0.98x\nARGBQuantizeRow 2.23 1.21 1.84x\nARGBSepiaRow 3.29 0.81 4.04x\nARGBSetRow 0.21 0.21 1.02x\nARGBShadeRow 1.66 0.87 1.91x\nARGBToRGB565DitherRow 1.22 0.50 2.43x\nARGBToRGB565Row 1.09 0.44 2.47x\nByteToFloatRow 0.88 0.41 2.14x\nConvert8To8Row 0.21 0.11 1.88x\nCopyRow_16 0.11 0.11 0.97x\nDetileRow 0.21 0.12 1.75x\nDetileRow_16 0.23 0.14 1.67x\nDetileSplitUVRow 0.29 0.21 1.38x\nDivideRow_16 0.14 0.11 1.30x\nGaussRow 3.29 1.27 2.59x\nGaussRow_F32 1.93 0.89 2.17x\nHalfFloatRow 1.16 0.74 1.57x\nI210AlphaToARGBRow 1.83 0.92 2.00x\nI210ToAR30Row 2.21 0.96 2.30x\nI210ToARGBRow 1.75 0.87 2.00x\nI212ToAR30Row 2.21 0.96 2.31x\nI212ToARGBRow 1.75 0.87 2.01x\nI410AlphaToARGBRow 1.73 0.86 2.00x\nI410ToAR30Row 2.04 0.89 2.28x\nI410ToARGBRow 1.64 0.82 2.00x\nI422ToARGB1555Row 2.94 2.27 1.30x\nI422ToARGB4444Row 2.94 2.27 1.29x\nI422ToRGB565Row 2.63 1.27 2.07x\nI422ToRGBARow 1.51 0.82 1.83x\nI444AlphaToARGBRow 1.52 0.87 1.73x\nI444ToARGBRow 1.48 0.81 1.82x\nI444ToRGB24Row 2.13 0.72 2.98x\nMergeUVRow_16 0.87 0.32 2.74x\nP210ToAR30Row 1.77 0.70 2.54x\nP210ToARGBRow 1.38 0.62 2.24x\nP410ToAR30Row 1.94 0.85 2.28x\nP410ToARGBRow 1.45 0.77 1.90x\nRGB565ToUVRow 15.84 2.88 5.49x\nRGB565ToYRow 7.82 1.22 6.42x\nRGBColorTableRow 8.42 6.95 1.21x\nSetRow 0.11 0.11 1.04x\nSobelRow 0.80 0.43 1.87x\nSobelToPlaneRow 0.27 0.12 2.25x\nSobelXRow 1.58 0.85 1.85x\nSobelXYRow 0.80 0.55 1.44x\nSobelYRow 1.58 0.85 1.85x\nUYVYToARGBRow 1.28 0.65 1.96x\nUYVYToUVRow 0.38 0.21 1.78x\nUYVYToYRow 0.21 0.11 1.87x\nYUY2ToARGBRow 1.28 0.65 1.96x\nYUY2ToUVRow 0.40 0.21 1.87x\nYUY2ToYRow 0.21 0.11 1.88x\n----------------------------------------------------------------\n\nRow/Rotate/Compare SIMD ISA per *_C (* \u003d uncalled in library):\n------------------------------------------------------------------------\nC Function arm64 x86_32 x64 row_win\n------------------------------------------------------------------------\nAB64ToARGBRow_C NEON AVX2 AVX2 AVX2\nABGRToAR30Row_C NEON AVX2 AVX512BW AVX512BW\nAR30ToAB30Row_C NEON AVX2 AVX512BW AVX512BW\nAR30ToABGRRow_C NEON AVX2 AVX512BW AVX512BW\nAR30ToARGBRow_C NEON AVX2 AVX512BW AVX512BW\nAR64ShuffleRow_C NEON AVX2 AVX512BW AVX512BW\nAR64ToAB64Row_C NEON AVX2 AVX512BW AVX512BW\nAR64ToARGBRow_C NEON AVX2 AVX2 AVX2\nARGB1555ToARGBRow_C SVE2 AVX2 AVX512BW AVX512BW\nARGB1555ToUVMatrixRow_C NEON AVX2 AVX2 AVX2\nARGB1555ToUVRow_C * NEON AVX2 AVX2 AVX2\nARGB1555ToYMatrixRow_C NEON AVX2 AVX2 AVX2\nARGB1555ToYRow_C * NEON AVX2 AVX2 AVX2\nARGB4444ToARGBRow_C NEON AVX2 AVX2 AVX2\nARGB4444ToUVMatrixRow_C NEON AVX2 AVX2 AVX2\nARGB4444ToUVRow_C * NEON AVX2 AVX2 AVX2\nARGB4444ToYMatrixRow_C NEON AVX2 AVX2 AVX2\nARGB4444ToYRow_C * NEON AVX2 AVX2 AVX2\nARGBAddRow_C NEON AVX2 AVX2 AVX2\nARGBAffineRow_C * NEON AVX2 AVX2 AVX2\nARGBAlphaBlendRow_C SME AVX2 AVX512BW AVX512BW\nARGBAttenuateRow_C NEON AVX2 AVX512BW AVX512BW\nARGBBlendRow_C NEON AVX2 AVX512BW AVX512BW\nARGBColorMatrixRow_C I8MM AVX2 AVX512VNNI AVX512VNNI\nARGBColorTableRow_C NEON AVX2 AVX2 AVX2\nARGBCopyAlphaRow_C NEON AVX2 AVX2 AVX2\nARGBCopyYToAlphaRow_C NEON AVX2 AVX512BW AVX512BW\nARGBDetectRow_C - AVX2 AVX2 AVX2\nARGBExtractAlphaRow_C NEON AVX2 AVX512BW AVX512BW\nARGBGrayRow_C DOTPROD AVX2 AVX512VNNI AVX512VNNI\nARGBLumaColorTableRow_C NEON AVX2 AVX2 AVX2\nARGBMirrorRow_C NEON AVX2 AVX512BW AVX512BW\nARGBMultiplyRow_C SME AVX2 AVX512BW AVX512BW\nARGBOrRow_C SME AVX2 AVX512BW AVX512BW\nARGBPolynomialRow_C NEON AVX2 AVX2 AVX2\nARGBQuantizeRow_C NEON AVX2 AVX2 AVX2\nARGBSepiaRow_C DOTPROD AVX2 AVX512VNNI AVX512VNNI\nARGBSetRow_C NEON AVX2 AVX2 AVX2\nARGBShadeRow_C NEON AVX2 AVX2 AVX2\nARGBShuffleRow_C NEON AVX2 AVX512BW AVX512BW\nARGBSubtractRow_C NEON AVX2 AVX2 AVX2\nARGBToAB64Row_C NEON AVX2 AVX2 AVX2\nARGBToABGRRow_C NEON AVX2 AVX512BW AVX512BW\nARGBToAR30Row_C NEON AVX2 AVX512BW AVX512BW\nARGBToAR64Row_C NEON AVX2 AVX2 AVX2\nARGBToARGB1555Row_C NEON AVX2 AVX2 AVX2\nARGBToARGB4444Row_C NEON AVX2 AVX2 AVX2\nARGBToBGRARow_C NEON AVX2 AVX512BW AVX512BW\nARGBToRAWRow_C SVE2 AVX2 AVX512BW AVX512BW\nARGBToRGB24Row_C SVE2 AVX2 AVX512BW AVX512BW\nARGBToRGB565DitherRow_C SVE2 AVX2 AVX512BW AVX512BW\nARGBToRGB565Row_C SVE2 AVX2 AVX512BW AVX512BW\nARGBToRGBARow_C NEON AVX2 AVX512BW AVX512BW\nARGBToUV444MatrixRow_C I8MM AVX2 AVX512VNNI AVX512VNNI\nARGBToUVMatrixRow_C SME AVX2 AVX512VNNI AVX512VNNI\nARGBToYMatrixRow_C SME AVX2 AVX512VNNI AVX512VNNI\nARGBUnattenuateRow_C NEON AVX2 AVX2 AVX2\nAYUVToUVRow_C SVE2 AVX2 AVX512VBMI AVX512VBMI\nAYUVToVURow_C SVE2 AVX2 AVX512VBMI AVX512VBMI\nAYUVToYRow_C NEON AVX2 AVX512VBMI AVX512VBMI\nBlendPlaneRow_C NEON AVX2 AVX512BW AVX512BW\nByteToFloatRow_C NEON AVX2 AVX2 AVX2\nComputeCumulativeSumRow_C NEON AVX2 AVX2 AVX2\nConvert16To8Row_C SME AVX2 AVX512BW AVX512BW\nConvert8To16Row_C SME AVX2 AVX512BW AVX512BW\nConvert8To8Row_C SME AVX2 AVX512BW AVX512BW\nCopyRow_16_C NEON AVX2 AVX2 AVX2\nCopyRow_C SME AVX AVX512BW AVX512BW\nCumulativeSumToAverageRow_C NEON AVX2 AVX2 AVX2\nDetileRow_16_C NEON AVX2 AVX2 AVX2\nDetileRow_C NEON AVX2 AVX2 AVX2\nDetileSplitUVRow_C NEON AVX2 AVX2 AVX2\nDetileToYUY2_C NEON AVX2 AVX2 AVX2\nDivideRow_16_C SVE2 AVX2 AVX512BW AVX512BW\nGaussCol_C * NEON AVX2 AVX2 AVX2\nGaussCol_F32_C NEON AVX2 AVX2 AVX2\nGaussRow_C * NEON AVX2 AVX2 AVX2\nGaussRow_F32_C NEON AVX2 AVX2 AVX2\nHalfFloatRow_C SVE2 AVX2 AVX512BW AVX512BW\nHalfMergeUVRow_C SVE2 AVX2 AVX512BW AVX512BW\nHalfRow_16To8_C SME AVX2 AVX512BW AVX512BW\nHalfRow_16_C - AVX2 AVX2 AVX2\nHalfRow_C - AVX2 AVX2 AVX2\nHalfWidthRow_16To8_C SME AVX2 AVX512BW AVX512BW\nHalfWidthRow_16To8_Odd_C - AVX2 AVX2 AVX2\nHammingDistance_C DOTPROD AVX2 AVX512BW AVX512BW\nHashDjb2_C NEON AVX2 AVX512BW AVX512BW\nI210AlphaToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI210ToAR30Row_C SME AVX2 AVX512BW AVX512BW\nI210ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI212ToAR30Row_C SME AVX2 AVX512BW AVX512BW\nI212ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI400ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI410AlphaToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI410ToAR30Row_C SME AVX2 AVX512BW AVX512BW\nI410ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI422AlphaToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI422ToAR30Row_C SME AVX2 AVX512BW AVX512BW\nI422ToARGB1555Row_C SME AVX2 AVX512BW AVX512BW\nI422ToARGB4444Row_C SME AVX2 AVX512BW AVX512BW\nI422ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI422ToRGB24Row_C SME AVX2 AVX512VBMI AVX512VBMI\nI422ToRGB565Row_C SME AVX2 AVX512BW AVX512BW\nI422ToRGBARow_C SME AVX2 AVX512BW AVX512BW\nI422ToUYVYRow_C NEON AVX2 AVX2 AVX2\nI422ToYUY2Row_C NEON AVX2 AVX2 AVX2\nI444AlphaToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI444ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nI444ToRGB24Row_C SME AVX2 AVX512VBMI AVX512VBMI\nInterpolateRow_16_C SME AVX2 AVX512BW AVX512BW\nInterpolateRow_C SME AVX2 AVX512BW AVX512BW\nJ400ToARGBRow_C NEON AVX2 AVX512BW AVX512BW\nMergeAR64Row_C NEON AVX2 AVX2 AVX2\nMergeARGB16To8Row_C NEON AVX2 AVX2 AVX2\nMergeARGBRow_C NEON AVX2 AVX2 AVX2\nMergeRGBRow_C NEON AVX2 AVX512VBMI AVX512VBMI\nMergeUVRow_16_C SME AVX2 AVX512BW AVX512BW\nMergeUVRow_C SME AVX2 AVX512BW AVX512BW\nMergeXR30Row_C NEON AVX2 AVX512BW AVX512BW\nMergeXR64Row_C NEON AVX2 AVX2 AVX2\nMergeXRGB16To8Row_C NEON AVX2 AVX2 AVX2\nMergeXRGBRow_C NEON AVX2 AVX2 AVX2\nMirrorRow_16_C NEON AVX2 AVX512BW AVX512BW\nMirrorRow_C NEON AVX2 AVX512BW AVX512BW\nMirrorSplitUVRow_C NEON AVX2 AVX512BW AVX512BW\nMirrorUVRow_C NEON AVX2 AVX512BW AVX512BW\nMultiplyRow_16_C SME AVX2 AVX512BW AVX512BW\nNV12ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nNV12ToRGB24Row_C SME AVX2 AVX512VBMI AVX512VBMI\nNV12ToRGB565Row_C NEON AVX2 AVX2 AVX2\nNV21ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nNV21ToRGB24Row_C SME AVX2 AVX512VBMI AVX512VBMI\nNV21ToYUV24Row_C NEON AVX2 AVX2 AVX2\nP210ToAR30Row_C SME AVX2 AVX512BW AVX512BW\nP210ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nP410ToAR30Row_C SME AVX2 AVX512BW AVX512BW\nP410ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nRAWToARGBRow_C SVE2 AVX2 AVX512BW AVX512BW\nRAWToRGB24Row_C SVE2 AVX2 AVX512VBMI AVX512VBMI\nRAWToRGBARow_C SVE2 AVX2 AVX512VBMI AVX512VBMI\nRGB24MirrorRow_C NEON AVX2 AVX512VBMI AVX512VBMI\nRGB24ToARGBRow_C SVE2 AVX2 AVX512BW AVX512BW\nRGB565ToARGBRow_C NEON AVX2 AVX2 AVX2\nRGB565ToUVMatrixRow_C NEON AVX2 AVX2 AVX2\nRGB565ToUVRow_C * NEON AVX2 AVX2 AVX2\nRGB565ToYMatrixRow_C NEON AVX2 AVX2 AVX2\nRGB565ToYRow_C * NEON AVX2 AVX2 AVX2\nRGBAToARGBRow_C NEON AVX2 AVX512BW AVX512BW\nRGBColorTableRow_C NEON AVX2 AVX2 AVX2\nRGBToUV444MatrixRow_C NEON AVX2 AVX512BW AVX512BW\nRGBToUVMatrixRow_C NEON AVX2 AVX512BW AVX512BW\nRGBToYMatrixRow_C DOTPROD AVX2 AVX512BW AVX512BW\nScaleMaxSamples_C * NEON AVX2 AVX2 AVX2\nScaleSamples_C * NEON AVX2 AVX2 AVX2\nScaleSumSamples_C * NEON AVX2 AVX2 AVX2\nSetRow_C NEON AVX2 AVX2 AVX2\nSobelRow_C NEON AVX2 AVX2 AVX2\nSobelToPlaneRow_C NEON AVX2 AVX2 AVX2\nSobelXRow_C NEON AVX2 AVX2 AVX2\nSobelXYRow_C NEON AVX2 AVX2 AVX2\nSobelYRow_C NEON AVX2 AVX2 AVX2\nSplitARGBRow_C NEON AVX2 AVX2 AVX2\nSplitRGBRow_C NEON AVX2 AVX2 AVX2\nSplitUVRow_16_C NEON AVX2 AVX2 AVX2\nSplitUVRow_C SVE2 AVX2 AVX512BW AVX512BW\nSplitXRGBRow_C NEON AVX2 AVX2 AVX2\nSsim4x4Row_C SME AVX2 AVX512BW AVX512BW\nSsimWindowRow_C SME AVX2 AVX512BW AVX512BW\nSumSquareError_C DOTPROD AVX2 AVX512VNNI AVX512VNNI\nSwapUVRow_C NEON AVX2 AVX512BW AVX512BW\nTransposeUVWx16_C - AVX2 AVX512BW AVX512BW\nTransposeUVWx8_C NEON AVX2 AVX2 AVX2\nTransposeWx16_C NEON AVX2 AVX512BW AVX512BW\nTransposeWx4_32_C * SVE2 AVX2 AVX512BW AVX512BW\nTransposeWx8_16_C SVE2 AVX2 AVX512BW AVX512BW\nTransposeWx8_C - AVX2 AVX2 AVX2\nTransposeWxH_32_C NEON AVX2 AVX2 AVX2\nUYVYToARGBRow_C SME AVX2 AVX512BW AVX512BW\nUYVYToUV422Row_C NEON AVX2 AVX2 AVX2\nUYVYToUVRow_C SVE2 AVX2 AVX512BW AVX512BW\nUYVYToYRow_C SVE2 AVX2 AVX512BW AVX512BW\nUnpackMT2T_C NEON AVX2 AVX2 AVX2\nYUY2ToARGBRow_C SME AVX2 AVX512BW AVX512BW\nYUY2ToNVUVRow_C NEON AVX2 AVX2 AVX2\nYUY2ToUV422Row_C NEON AVX2 AVX2 AVX2\nYUY2ToUVRow_C SVE2 AVX2 AVX512BW AVX512BW\nYUY2ToYRow_C SVE2 AVX2 AVX512BW AVX512BW\n------------------------------------------------------------------------\nSIMD (187 row/rot/cmp) 181 97% 187 100% 187 100% 187 100%\n------------------------------------------------------------------------\n\nScale SIMD ISA supported per *_C function:\n------------------------------------------------------------------------\nC Function arm64 x86_32 x64 row_win\n------------------------------------------------------------------------\nFixedDiv1_C - X86 X86 X86\nFixedDiv_C - X86 X86 X86\nScaleARGBCols64_C - - - -\nScaleARGBColsUp2_C - AVX2 AVX2 AVX2\nScaleARGBCols_C NEON AVX2 AVX2 AVX2\nScaleARGBFilterCols64_C - - - -\nScaleARGBFilterCols_C NEON AVX2 AVX2 AVX2\nScaleARGBRowDown2Box_C SME AVX2 AVX2 AVX2\nScaleARGBRowDown2Linear_C SME AVX2 AVX2 AVX2\nScaleARGBRowDown2_C SME AVX2 AVX2 AVX2\nScaleARGBRowDownEvenBox_C NEON AVX2 AVX2 AVX2\nScaleARGBRowDownEven_C NEON AVX2 AVX2 AVX2\nScaleAddCols0_C - - - -\nScaleAddCols1_16_C - - - -\nScaleAddCols1_C - - AVX512BW AVX512BW\nScaleAddCols2_16_C - - - -\nScaleAddCols2_C - - AVX512BW AVX512BW\nScaleAddRow_16_C - - - -\nScaleAddRow_C NEON AVX2 AVX512BW AVX512BW\nScaleColsUp2_16_C - - - -\nScaleColsUp2_C - AVX2 AVX2 AVX2\nScaleCols_16_C - - - -\nScaleCols_C - - - -\nScaleFilterCols64_16_C - - - -\nScaleFilterCols64_C - - - -\nScaleFilterCols_16_C - - - -\nScaleFilterCols_C NEON AVX2 AVX2 AVX2\nScaleRowDown2Box_16_C SME AVX2 AVX2 AVX2\nScaleRowDown2Box_C SME AVX2 AVX512BW AVX512BW\nScaleRowDown2Box_Odd_C - - AVX512BW AVX512BW\nScaleRowDown2Linear_16_C SME AVX2 AVX2 AVX2\nScaleRowDown2Linear_C SME AVX2 AVX512BW AVX512BW\nScaleRowDown2_16_C SME AVX2 AVX2 AVX2\nScaleRowDown2_C SME AVX2 AVX512BW AVX512BW\nScaleRowDown34_0_Box_16_C - - - -\nScaleRowDown34_0_Box_C NEON AVX2 AVX2 AVX2\nScaleRowDown34_16_C - - - -\nScaleRowDown34_1_Box_16_C - - - -\nScaleRowDown34_1_Box_C NEON AVX2 AVX2 AVX2\nScaleRowDown34_C NEON AVX2 AVX2 AVX2\nScaleRowDown38_16_C - - - -\nScaleRowDown38_2_Box_16_C - - - -\nScaleRowDown38_2_Box_C NEON AVX2 AVX2 AVX2\nScaleRowDown38_3_Box_16_C - - - -\nScaleRowDown38_3_Box_C NEON AVX2 AVX2 AVX2\nScaleRowDown38_C NEON AVX2 AVX2 AVX2\nScaleRowDown4Box_16_C - - - -\nScaleRowDown4Box_C NEON AVX2 AVX2 AVX2\nScaleRowDown4_16_C - - - -\nScaleRowDown4_C NEON AVX2 AVX2 AVX2\nScaleRowUp2_Bilinear_16_Any_C NEON AVX2 AVX2 AVX2\nScaleRowUp2_Bilinear_16_C NEON AVX2 AVX2 AVX2\nScaleRowUp2_Bilinear_Any_C NEON AVX2 AVX2 AVX2\nScaleRowUp2_Bilinear_C NEON AVX2 AVX2 AVX2\nScaleRowUp2_Linear_16_Any_C NEON AVX2 AVX2 AVX2\nScaleRowUp2_Linear_16_C NEON AVX2 AVX2 AVX2\nScaleRowUp2_Linear_Any_C NEON AVX2 AVX2 AVX2\nScaleRowUp2_Linear_C NEON AVX2 AVX2 AVX2\nScaleUVAddCols1_16_C - - - -\nScaleUVAddCols1_C - - - -\nScaleUVAddCols2_16_C - - - -\nScaleUVAddCols2_C - - - -\nScaleUVAddCols3_16_C - - - -\nScaleUVAddCols3_C SME AVX2 AVX512VBMI AVX512VBMI\nScaleUVCols64_16_C - - - -\nScaleUVCols64_C - - - -\nScaleUVColsUp2_16_C - - - -\nScaleUVColsUp2_C - - - -\nScaleUVCols_16_C - - - -\nScaleUVCols_C - - - -\nScaleUVFilterCols64_16_C - - - -\nScaleUVFilterCols64_C - - - -\nScaleUVFilterCols_16_C - - - -\nScaleUVFilterCols_C - - - -\nScaleUVRowDown2Box_16_C - - - -\nScaleUVRowDown2Box_C SME AVX2 AVX2 AVX2\nScaleUVRowDown2Linear_16_C - - - -\nScaleUVRowDown2Linear_C SME AVX2 AVX2 AVX2\nScaleUVRowDown2_16_C - - - -\nScaleUVRowDown2_C SME AVX2 AVX2 AVX2\nScaleUVRowDownEvenBox_16_C - - - -\nScaleUVRowDownEvenBox_C - - - -\nScaleUVRowDownEven_16_C - - - -\nScaleUVRowDownEven_C NEON AVX2 AVX2 AVX2\nScaleUVRowUp2_Bilinear_16_Any_C NEON AVX2 AVX2 AVX2\nScaleUVRowUp2_Bilinear_16_C NEON AVX2 AVX2 AVX2\nScaleUVRowUp2_Bilinear_Any_C NEON AVX2 AVX2 AVX2\nScaleUVRowUp2_Bilinear_C NEON AVX2 AVX2 AVX2\nScaleUVRowUp2_Linear_16_Any_C NEON AVX2 AVX2 AVX2\nScaleUVRowUp2_Linear_16_C NEON AVX2 AVX2 AVX2\nScaleUVRowUp2_Linear_Any_C NEON AVX2 AVX2 AVX2\nScaleUVRowUp2_Linear_C NEON AVX2 AVX2 AVX2\n------------------------------------------------------------------------\nSIMD (92 scale *_C) 44 48% 48 52% 51 55% 51 55%\n------------------------------------------------------------------------\n\nTest: libyuv_unittest\nBug: none\nChange-Id: I2b4a832f2a7fe1fe93487b453240cd15c1f5f940\nReviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/8676400\nCommit-Queue: Frank Barchard \u003cfbarchard@google.com\u003e\nReviewed-by: richard winterton \u003crrwinterton@gmail.com\u003e\n", "tree_diff": [ { "type": "modify", "old_id": "a82327f5b2e9b963c73c7a35dce7a716e2c9a814", "old_mode": 33188, "old_path": "GEMINI.md", "new_id": "923077777938d453916658c6238a3e2f9c79c56f", "new_mode": 33188, "new_path": "GEMINI.md" }, { "type": "modify", "old_id": "6dfc0d43dc7a545db5935f9e7bfbcb38421bbf92", "old_mode": 33188, "old_path": "README.chromium", "new_id": "0bac66a43c133168a119959866e0003bb4c9c2f2", "new_mode": 33188, "new_path": "README.chromium" }, { "type": "modify", "old_id": "a8c7aef6bc6fa9fe2f64b3fcbb14e5961d10853f", "old_mode": 33188, "old_path": "docs/environment_variables.md", "new_id": "a46744e45bb742e8873cc8c3b8d7d994bc4d16ad", "new_mode": 33188, "new_path": "docs/environment_variables.md" }, { "type": "modify", "old_id": "301a103e968f6b72d0176ad88bb087cae4eb5af5", "old_mode": 33188, "old_path": "include/libyuv/compare_row.h", "new_id": "03caa5c011d75460b82a95c5cd7a1e8411acd6c1", "new_mode": 33188, "new_path": "include/libyuv/compare_row.h" }, { "type": "modify", "old_id": "adf35df7f41baf31877ea497d6caaaa94cbfd1b0", "old_mode": 33188, "old_path": "include/libyuv/convert.h", "new_id": "fdaec5eaf541c0444f1e810a012cad00b5075b86", "new_mode": 33188, "new_path": "include/libyuv/convert.h" }, { "type": "modify", "old_id": "bcb2099d64bea81a1cfbcc9ac8820d270e556a85", "old_mode": 33188, "old_path": "include/libyuv/convert_argb.h", "new_id": "59b690f3a8fa2a75a2ece285174f62e044abdc69", "new_mode": 33188, "new_path": "include/libyuv/convert_argb.h" }, { "type": "modify", "old_id": "29d656d0aca27c6303478bf07d5518826607aee9", "old_mode": 33188, "old_path": "include/libyuv/convert_from_argb.h", "new_id": "817df6f9eae3b9b7e8da56f284e56bdcad2c42ec", "new_mode": 33188, "new_path": "include/libyuv/convert_from_argb.h" }, { "type": "modify", "old_id": "917df1cb40886429d4d4155adb3e048ecc9bafa7", "old_mode": 33188, "old_path": "include/libyuv/cpu_id.h", "new_id": "54512b766d14c159552db1d9f3d5cf4551ad47fb", "new_mode": 33188, "new_path": "include/libyuv/cpu_id.h" }, { "type": "modify", "old_id": "adb50c42d3cdceaf14df327d76b2055c739d7396", "old_mode": 33188, "old_path": "include/libyuv/cpu_support.h", "new_id": "227bf3bd117f5114d1452222368cdef41760ae45", "new_mode": 33188, "new_path": "include/libyuv/cpu_support.h" }, { "type": "modify", "old_id": "4897256dbc329112cf903c3a0bd1ea94535d69a9", "old_mode": 33188, "old_path": "include/libyuv/rotate_row.h", "new_id": "4ac43d388c3ef91dfe99e9b32988caddd113aef2", "new_mode": 33188, "new_path": "include/libyuv/rotate_row.h" }, { "type": "modify", "old_id": "f3c46bb2a2ee192b1342578707c77b628b01451d", "old_mode": 33188, "old_path": "include/libyuv/row.h", "new_id": "3fe8341b9b5394977d799ce9dfe4590460c7e8a2", "new_mode": 33188, "new_path": "include/libyuv/row.h" }, { "type": "modify", "old_id": "e58962e7ec4da21ee34c60979f2dd3b25fc0848e", "old_mode": 33188, "old_path": "include/libyuv/scale_row.h", "new_id": "a4d34c515059b0cde7a9131352caa0cf5101c268", "new_mode": 33188, "new_path": "include/libyuv/scale_row.h" }, { "type": "modify", "old_id": "f6cc8e40fb8affc6bcebbd134e00766c0fb9d41a", "old_mode": 33188, "old_path": "include/libyuv/scale_uv.h", "new_id": "168c5d94950b0964aa9c50360bc8b7c409e21876", "new_mode": 33188, "new_path": "include/libyuv/scale_uv.h" }, { "type": "modify", "old_id": "caf8590aa47c0120b6d66450cde82c09b4f1512f", "old_mode": 33188, "old_path": "include/libyuv/version.h", "new_id": "5ba507c6d8209cb3177a9898d495ffd27673844a", "new_mode": 33188, "new_path": "include/libyuv/version.h" }, { "type": "modify", "old_id": "e60707ae5d57663b3de72d16407e682a171a5372", "old_mode": 33188, "old_path": "include/libyuv/video_common.h", "new_id": "ee989f7869c205be555de5bf25511068947b5f10", "new_mode": 33188, "new_path": "include/libyuv/video_common.h" }, { "type": "modify", "old_id": "352cedb1b23ec465533350f2100b3531e303683c", "old_mode": 33188, "old_path": "source/compare.cc", "new_id": "c5ef006994e97931646777202e750d87fb593846", "new_mode": 33188, "new_path": "source/compare.cc" }, { "type": "modify", "old_id": "cd667fdd25f12f0378a7ad533be54d7b2bce4bdb", "old_mode": 33188, "old_path": "source/compare_common.cc", "new_id": "f50bc4708245f1d327f09727e8a981fa57ee026a", "new_mode": 33188, "new_path": "source/compare_common.cc" }, { "type": "modify", "old_id": "123d92fea13ddf70a2f48d8fcad36670a35fb794", "old_mode": 33188, "old_path": "source/compare_gcc.cc", "new_id": "4ddaa12f93174a5556ebb5d5ab66b27110cee306", "new_mode": 33188, "new_path": "source/compare_gcc.cc" }, { "type": "modify", "old_id": "d1166ad4665cc9a04c96a1e464d0051363a3aa1c", "old_mode": 33188, "old_path": "source/compare_win.cc", "new_id": "b969a30910ff0f767322c62126f597b7fd8de13e", "new_mode": 33188, "new_path": "source/compare_win.cc" }, { "type": "modify", "old_id": "67385d9ffb9c20308e559ce590ea102d554a00fb", "old_mode": 33188, "old_path": "source/convert.cc", "new_id": "50ea994844edd8ec3cbc59aeead41e132aca7a29", "new_mode": 33188, "new_path": "source/convert.cc" }, { "type": "modify", "old_id": "e9d8b9c1c8ac7da932567a66afbdebcb431f9259", "old_mode": 33188, "old_path": "source/convert_argb.cc", "new_id": "e79dcb8ad499e689c36fab788bb14555afb7dd52", "new_mode": 33188, "new_path": "source/convert_argb.cc" }, { "type": "modify", "old_id": "4bca7602fe0612beb9f833b7cbc8d5fc800aa07f", "old_mode": 33188, "old_path": "source/convert_from.cc", "new_id": "c4f35cfea38ea39c01eae0bc06f9305f948fed73", "new_mode": 33188, "new_path": "source/convert_from.cc" }, { "type": "modify", "old_id": "2337d54f7c0a88ddbaf36eef10257155999f552e", "old_mode": 33188, "old_path": "source/convert_from_argb.cc", "new_id": "cc7691314344a13b1658e9da83c61f410d511929", "new_mode": 33188, "new_path": "source/convert_from_argb.cc" }, { "type": "modify", "old_id": "bc699b647e485fad8f32b56b2f34028a447e9369", "old_mode": 33188, "old_path": "source/convert_to_argb.cc", "new_id": "d3b55687a006723c7490ce6219f9e71b037c8a3b", "new_mode": 33188, "new_path": "source/convert_to_argb.cc" }, { "type": "modify", "old_id": "67f78b170df069c253e820c7d57230fad9188450", "old_mode": 33188, "old_path": "source/convert_to_i420.cc", "new_id": "0d80d0c0d65cfba4b28e0928d0e4f91d73d92d3e", "new_mode": 33188, "new_path": "source/convert_to_i420.cc" }, { "type": "modify", "old_id": "7bfb66de2e5a72ebcb31301d2e01e3ebf8f3d0c0", "old_mode": 33188, "old_path": "source/cpu_id.cc", "new_id": "f78283a97254416dfd045236eab25b58521da2ec", "new_mode": 33188, "new_path": "source/cpu_id.cc" }, { "type": "modify", "old_id": "1521e3f055daa6e9929c6a0c556be99fd4f78778", "old_mode": 33188, "old_path": "source/planar_functions.cc", "new_id": "0c8a42542bd8c29b52f72cd04a6ad8aa12a7e4fa", "new_mode": 33188, "new_path": "source/planar_functions.cc" }, { "type": "modify", "old_id": "59bf008240242966b92259e886d7185a6622100d", "old_mode": 33188, "old_path": "source/rotate.cc", "new_id": "7b27333f1abb917ffb65930bd3b97556e37fe8c2", "new_mode": 33188, "new_path": "source/rotate.cc" }, { "type": "modify", "old_id": "499047cb40b094f0be06fc10e07b653ac33ab463", "old_mode": 33188, "old_path": "source/rotate_any.cc", "new_id": "bc203655a9e60de16e9a27f28127832bf2ac313d", "new_mode": 33188, "new_path": "source/rotate_any.cc" }, { "type": "modify", "old_id": "00900d1757a3b79d42a14946f431f5dae9cdeba2", "old_mode": 33188, "old_path": "source/rotate_argb.cc", "new_id": "1b0bf7417dea820c71ec70c6bfdffb7c06e14683", "new_mode": 33188, "new_path": "source/rotate_argb.cc" }, { "type": "modify", "old_id": "0abab4f7e040b067b4bcaedbb95a3b7b6605bc4c", "old_mode": 33188, "old_path": "source/rotate_common.cc", "new_id": "e4aab102f57f18aeb1e5fc1ab90ad5354a4aa918", "new_mode": 33188, "new_path": "source/rotate_common.cc" }, { "type": "modify", "old_id": "d9456a8421ef1016904d1fc87e2b11f87a845f36", "old_mode": 33188, "old_path": "source/rotate_gcc.cc", "new_id": "251054f2a61c0ff5e655bfbe5db39e0204218c5e", "new_mode": 33188, "new_path": "source/rotate_gcc.cc" }, { "type": "modify", "old_id": "707286a0f66d8287f0fcef43f00c1be553e86c5a", "old_mode": 33188, "old_path": "source/rotate_win.cc", "new_id": "d1d48dfea4e3df3f0b9cd59e90622cf063cb7d1b", "new_mode": 33188, "new_path": "source/rotate_win.cc" }, { "type": "modify", "old_id": "bb59eaf3efa7b76a62920316eb134825a24e62e3", "old_mode": 33188, "old_path": "source/row_any.cc", "new_id": "3ac640620e6c150522463e2b4f0834f3270c1cfd", "new_mode": 33188, "new_path": "source/row_any.cc" }, { "type": "modify", "old_id": "94de60ab33db11f801ac8e7bd8f20397697d245e", "old_mode": 33188, "old_path": "source/row_common.cc", "new_id": "951d3c92739e8006f4f47f6390895b0c8e462d91", "new_mode": 33188, "new_path": "source/row_common.cc" }, { "type": "modify", "old_id": "40f990dc00bf3231e89697e5a9c2560364c4be6f", "old_mode": 33188, "old_path": "source/row_gcc.cc", "new_id": "876fdd274fb7d1e7afa9013417f025139e5ee421", "new_mode": 33188, "new_path": "source/row_gcc.cc" }, { "type": "modify", "old_id": "da55bd12e0107e98106d1bb7e63e0a1941adbd24", "old_mode": 33188, "old_path": "source/row_win.cc", "new_id": "a986d79ef4dfbadf0e0824c99bfe7872124d17e8", "new_mode": 33188, "new_path": "source/row_win.cc" }, { "type": "modify", "old_id": "a9fdf79c868acc73e20ba3dfd952f63761c1d2df", "old_mode": 33188, "old_path": "source/scale.cc", "new_id": "f2157409cb3b9eb016ab89478a620c7baa02d82d", "new_mode": 33188, "new_path": "source/scale.cc" }, { "type": "modify", "old_id": "713f3f1866bf4b463ec608c662d956dd86432c3e", "old_mode": 33188, "old_path": "source/scale_any.cc", "new_id": "ec903768e1f0a049997b3eddb829c95f68941395", "new_mode": 33188, "new_path": "source/scale_any.cc" }, { "type": "modify", "old_id": "dd9df3612b34de4dbd0650cc53ebe763a6167ff1", "old_mode": 33188, "old_path": "source/scale_argb.cc", "new_id": "53280ea1662a8b27f35c677ad57852e1992df916", "new_mode": 33188, "new_path": "source/scale_argb.cc" }, { "type": "modify", "old_id": "0f1e7c845aaff046ace8f31e2eeb9fd439708070", "old_mode": 33188, "old_path": "source/scale_common.cc", "new_id": "a07a2183eba1439b7f240c63996b613f85b5c426", "new_mode": 33188, "new_path": "source/scale_common.cc" }, { "type": "modify", "old_id": "b018af795da3c34c9a223d73f8d06de9afa3976b", "old_mode": 33188, "old_path": "source/scale_gcc.cc", "new_id": "dc0383c37a08b8138c01000dc276c858578d70b9", "new_mode": 33188, "new_path": "source/scale_gcc.cc" }, { "type": "modify", "old_id": "324822a0e42564a07871d7463baacc4e1db5b0db", "old_mode": 33188, "old_path": "source/scale_uv.cc", "new_id": "706f27a80a6cbbc1809bdb2d932790f025fa36ee", "new_mode": 33188, "new_path": "source/scale_uv.cc" }, { "type": "modify", "old_id": "f2ad7f0784e99dcd366f3fc522d2d317e9067714", "old_mode": 33188, "old_path": "source/scale_win.cc", "new_id": "d638770832f2ad45436c5babff0149628d012b91", "new_mode": 33188, "new_path": "source/scale_win.cc" }, { "type": "modify", "old_id": "2556bedaaf108fc9554acdc2ef3d5f84941c5761", "old_mode": 33188, "old_path": "unit_test/compare_test.cc", "new_id": "ba0bd272c756d23fb76eacd6ac306a628dad70fd", "new_mode": 33188, "new_path": "unit_test/compare_test.cc" }, { "type": "modify", "old_id": "1a08f609b2bcf69de5cc390cbfdfa100184b7724", "old_mode": 33188, "old_path": "unit_test/cpu_test.cc", "new_id": "3829f8dd2342f9ba48bc593f2b29b088788e909a", "new_mode": 33188, "new_path": "unit_test/cpu_test.cc" }, { "type": "modify", "old_id": "01312ff60cb09a2c006d52876e1d1f9a8fb411f3", "old_mode": 33188, "old_path": "unit_test/planar_test.cc", "new_id": "5f7cf116e389b34664fa2e7f0ed2e6b533b109a4", "new_mode": 33188, "new_path": "unit_test/planar_test.cc" }, { "type": "modify", "old_id": "b9ece940966561aeeeb7ffd05612983d1d21f65e", "old_mode": 33188, "old_path": "unit_test/rotate_test.cc", "new_id": "5ddf899cbe51a006e8f19f16ad2c3c1e20d53b33", "new_mode": 33188, "new_path": "unit_test/rotate_test.cc" }, { "type": "modify", "old_id": "bf2ec6b8a76241c403a6939297656621d1e114f6", "old_mode": 33188, "old_path": "unit_test/scale_plane_test.cc", "new_id": "012377d5cd9c548b038f253e3b88703af4ee7655", "new_mode": 33188, "new_path": "unit_test/scale_plane_test.cc" }, { "type": "modify", "old_id": "5a8d160a96d14cb4489c59e06209ca440fda64d1", "old_mode": 33188, "old_path": "unit_test/scale_uv_test.cc", "new_id": "bbb6bbc680551d731e705fe71707f9dc8f1fdc67", "new_mode": 33188, "new_path": "unit_test/scale_uv_test.cc" }, { "type": "modify", "old_id": "ee1f63e6471e4bac5986145aa0203222b683fb1a", "old_mode": 33188, "old_path": "unit_test/unit_test.cc", "new_id": "c51b184e4ef2c4e0504636552d050f730cb567bd", "new_mode": 33188, "new_path": "unit_test/unit_test.cc" }, { "type": "modify", "old_id": "33cd4201a0b7e2d9d1c9332ccc5e5a400ab66dec", "old_mode": 33188, "old_path": "util/cpuid.c", "new_id": "d85595297046759c369db7ccc5af2765d329a5a2", "new_mode": 33188, "new_path": "util/cpuid.c" } ] }