Simd v5.2.124

Algorithms

New features

• Base implementation, SSE4.1, AVX2, AVX-512BW, NEON optimizations of function BgraToYuv422pV2.
• NEON optimizations of function BgraToYuv444pV2.
• NEON optimizations of function BgraToYuv420pV2.
• NEON optimizations of function Float32ToBFloat16.
• NEON optimizations of function BFloat16ToFloat32.
• Base implementation, SSE4.1, AVX2, AVX-512BW, NEON optimizations of function BgraToYuva420pV2.
• Support of SynetUnaryOperation32fErf in Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of function SynetUnaryOperation32f.
• Base implementation, SSE4.1, AVX2 optimizations of function SynetGelu32f.

Improving

• Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of function SynetSoftmaxLayerForward.

Bug fixing

• Error in method View::ToOcv.

Removing

• Support of all formats besides NHWC and NCHW for function SynetAddBias.
• Support of all formats besides NHWC and NCHW for function SynetLrnLayerCrossChannels.
• Support of all formats besides NHWC and NCHW for function SynetPreluLayerForward.
• Support of all formats besides NHWC and NCHW for function SynetScaleLayerForward.
• Support of all formats besides NHWC and NCHW for function SynetFusedLayerForward0.
• Support of all formats besides NHWC and NCHW for function SynetFusedLayerForward1.
• Support of all formats besides NHWC and NCHW for function SynetFusedLayerForward2.
• Support of all formats besides NHWC and NCHW for function SynetFusedLayerForward3.
• Support of all formats besides NHWC and NCHW for function SynetFusedLayerForward4.
• Support of all formats besides NHWC and NCHW for function SynetFusedLayerForward8.
• Support of all formats besides NHWC and NCHW for function SynetFusedLayerForward9.
• Function SynetReorderFilter.
• Function SynetReorderImage.
• Function SynetTensorAlignment.
• Function SynetSpecifyTensorFormat.
• Support of all formats besides NHWC and NCHW for enumeration SimdTensorFormatType.

Renaming

• Function from SynetUnaryOperation32fLayerForward to SynetUnaryOperation32f.

Test framework

New features

• Test command line argument '-ts' to print statistics of time of tests execution.
• Tests for verifying functionality of function BgraToYuv422pV2.
• Tests for verifying functionality of function BgraToYuva420pV2.
• Improving header of performance report.
• Tests for verifying functionality of function SynetGelu32f.

Bug fixing

• Error in test SynetUnaryOperation32fLayerForward.

Simd v5.2.123

Algorithms

Bug fixing

• MSVS-2022 compiler errors in AmxBf16 project.
• Clang compiler error in method Array::Release.
• MSVS-2022 compiler warnings in file SimdSse41ResizerNearest.cpp.
• MSVS-2022 compiler warnings in file SimdAvx512bwResizerNearest.cpp.
• Error in SSE4.1, AVX, AVX2, AVX-512BW kernels of ConvolutionNhwcDirect_2 (fixed type).
• MSVS-2022 compiler warnings in file SimdAvx2RecursiveBilateralFilter.cpp.
• Error in file SimdFrame.hpp (function Simd::Convert).
• Error in AVX2 optimizations of class RecursiveBilateralFilterFast (x86 only).
• MSVS-2022 compiler error in file SimdInit.h (ARM64).
• MSVS-2022 compiler errors in file SimdLog.h (ARM64).
• MSVS-2022 compiler errors in file SimdConversion.h (ARM64).
• MSVS-2022 compiler errors in file SimdNeonYuvToHue.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonAbsDifferenceSum.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonDetection.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonHog.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonLaplace.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonMeanFilter3x3.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonNeural.cpp (ARM64).
• MSVS-2022 compiler warnings in file SimdNeonNeuralConvolution.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonSobel.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonSynetConversion.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonSynetConvolution8i.cpp (ARM64).
• MSVS-2022 compiler errors in file SimdNeonYuvToHue.cpp (ARM64).
• Wrong assert in AVX-512BW optimization of function BgraToYuv420pV2.
• MSVS-2017 compiler errors in AVX-512BW optimizations of function AlphaBlendingBgraToYuv420p.
• MSVS-2017 compiler errors in AVX-512BW optimizations of WarpAffine engine.
• MSVS-2015 compiler errors in file SimdFmadd.h (Win32).
• MSVS-2015 compiler errors in SSE4.1 and AVX2 optimizations of WarpAffine engine (Win32).
• Crash in AVX2 optimizations of function CosineDistance16f.
• Crash in SSE4.1, AVX2, AVX-512BW, NEON optimizations of function HogLiteResizeFeatures.
• Clang compiler warnings in file SimdBaseRecursiveBilateralFilter.cpp.
• Clang compiler warnings in file SimdSse41RecursiveBilateralFilter.cpp.
• Clang linker error in method Motion::Detector::GenerateSearchRegionScanlines.
• Crashes in AVX-512BW optimizations of WarpAffine engine (MSVS-2022, Release).
• Internal compiler error in file SimdAvx512bwSynetConvolution32f.cpp (MSVS-2017, Release).
• Error in SSE4.1, AVX2 optimizations of function BgraToYuv444pV2 (MSVS-2015, Release, Win32).
• Error in SSE4.1, AVX2 optimizations of function BgraToYuv420pV2 (MSVS-2015, Release, Win32).
• Error in AVX2 optimizations of function Yuva444pToBgraV2 (MSVS-2015, Release, Win32).
• Error in AVX2 optimizations of function Yuv444pToBgraV2 (MSVS-2015, Release, Win32).
• Error in AVX2 optimizations of function Yuv420pToBgraV2 (MSVS-2015, Release, Win32).
• Error in AVX2 optimizations of function AlphaBlendingBgraToYuv420p (MSVS-2015, Release, Win32).
• Error in AVX2 optimizations of class ResizerByteArea2x2 (MSVS-2015, Release, Win32).
• Error in AVX2 optimizations of function Uyvy422ToBgr (MSVS-2015, Release, Win32).

Test framework

New features

• Handling of Windows exceptions in AutoTest.

Bug fixing

• Error in test Nv12SaveAsJpegToMemoryAutoTest.
• Error in test SynetAdd8iAutoTest.
• Error in test SynetConvolution8iForwardAutoTest.
• Error in test SynetScale8iForwardAutoTest.
• Error in test WarpAffineAutoTest.
• Error in test ResizeBilinearAutoTest.
• Error in test SynetMergedConvolution8iForwardAutoTest.
• Error in function MakeAutoTests (multithreaded environment).
• Error in test Float32ToBFloat16AutoTest.
• Error in test SynetConvert32fTo8uAutoTest (MSVS-2015 and MSVS-2017, Release, Win32).
• Error in test SynetMergedConvolution32fForwardAutoTest (MSVS-2015 and MSVS-2017, Release, Win32).
• Error in test CosineDistancesMxNp16fAutoTest.
• Error in test VectorNormNp16fAutoTest.

Infrastructure

New features

• Script BuildAll.cmd to build MSVS solution for all configurations and platforms.
• Github actions script for CMake (build and test for x86_x64, Linux).
• Github actions script for CMake (cross platform build for ARM, Linux).
• Github actions script for Cmake (cross platform build for ARM64, Linux).
• Github actions script for CMake (cross platform build for PowerPC, Linux).
• Github actions script for CMake (build and test for clang, Linux).
• Github actions script for MSBuild (build and test for Visual Studio 2022, Windows).
• Script GetThreadCount.cmd.
• Github actions script for MSBuild (build and test for Visual Studio 2019, Windows).
• Github actions script for MSBuild (build and test for Visual Studio 2017, Windows).
• Github actions script for MSBuild (build and test for Visual Studio 2015, Windows).

Renaming

• Script TestVisualStudio.cmd to TestAll.cmd.

Documentation

Bug fixing

• Wrong description of Cmake parameters.

Simd v5.2.122

Algorithms

New features

• New API of function Avx512bw::TileZero (AMX emulation).
• New API of function Avx512bw::TileLoad (AMX emulation).
• New API of function Avx512bw::TileStore (AMX emulation).
• New API of function Avx512bw::TileMatMulBf16 (AMX emulation).
• New API of function Avx512bw::TileMatMul8u8i (AMX emulation).
• Function Avx512bw::TileMatMulFp16(AMX emulation).
• The mark of function SimdInterferenceIncrement as deprecated.
• The mark of function SimdInterferenceIncrementMasked as deprecated.
• The mark of function SimdInterferenceDecrement as deprecated.
• The mark of function SimdInterferenceDecrementMasked as deprecated.
• The mark of function SimdSynetReorderImage as deprecated.
• The mark of function SimdSynetReorderFilter as deprecated.
• SimdTensorData16f (16-bit floating point) tensor type.
• The mark of function SimdSynetSpecifyTensorFormat as deprecated.
• The mark of function SimdSynetTensorAlignment as deprecated.
• Base implementation, SSE4.1, AVX2, AVX-512BW, NEON optimizations of SynetPermute engine.
• NEON optimizations of function Yuva444pToBgraV2.
• NEON optimizations of function AlphaBlending2x.
• SSE4.1, AVX2, AVX-512BW optimizations of function BgraToYuv444pV2.
• SSE4.1, AVX2, AVX-512BW optimizations of function BgraToYuv420pV2.
• Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of function AlphaBlendingBgraToYuv420p.
• Parameter 'copy' to View::Capure.
• Method Array::Release.

Improving

• NEON optimizations of function TransformImage.

Bug fixing

• Clang compiler error in function Simd::WarpAffine.
• MSVS-2022 compiler warnings in file SimdBaseRecursiveBilateralFilter.cpp.
• MSVS-2022 compiler warnings in file SimdSse41RecursiveBilateralFilter.cpp.
• MSVS-2015 compiler error in file SimdAvx2RecursiveBilateralFilter.cpp.
• Error in method MergConvParam32f::Valid.
• Crash in constructor of Simd::TileConf.
• Crash in AVX and AVX2 optimizations of function SynetInnerProductLayerForward.
• MSVS-2022 compiler error in file SimdAvx2RecursiveBilateralFilter.cpp (Win32 target).
• GCC compiler error in file SimdParallel.hpp (for AVX2 optimizations).

Test framework

New features

• Tests for verifying functionality of SynetPermut engine.
• Tests for verifying functionality of function AlphaBlendingBgraToYuv420p.

Bug fixing

• Crash in test GaussianBlurAutoTest.

Infrastructure

New features

• Install target in Cmake.
• Uninstall target in Cmake.

Renaming

• Project Amx to AmxBf16.

Simd v5.2.121

Algorithms

New features

• SIMD_DEPRECATED macro.
• The mark of function SimdSvmSumLinear as deprecated.
• SSE4.1, AVX2, AVX-512BW optimizations of function SynetNormalizeLayerForward.
• Enumeration SimdWarpAffineFlags.
• Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of class WarpAffineNearest.
• Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of class WarpAffineBilinear.
• Multi-threaded optimizations of class WarpAffineNearest.
• Multi-threaded optimizations of class WarpAffineBilinear.
• Function Simd::WarpAffine.
• Function Simd::Mean.
• Function Simd::OtsuThreshold.
• Function Simd::RecursiveBilateralFilter.
• The mark of function SimdEdgeBackgroundGrowRangeSlow as deprecated.
• The mark of function SimdEdgeBackgroundGrowRangeFast as deprecated.
• The mark of function SimdEdgeBackgroundIncrementCount as deprecated.
• The mark of function SimdEdgeBackgroundAdjustRange as deprecated.
• The mark of function SimdEdgeBackgroundAdjustRangeMasked as deprecated.
• The mark of function SimdEdgeBackgroundShiftRange as deprecated.
• The mark of function SimdEdgeBackgroundShiftRangeMasked as deprecated.
• The mark of function Simd::EdgeBackgroundGrowRangeSlow as deprecated.
• The mark of function Simd::EdgeBackgroundGrowRangeFast as deprecated.
• The mark of function Simd::EdgeBackgroundIncrementCount as deprecated.
• The mark of function Simd::EdgeBackgroundAdjustRange as deprecated.
• The mark of function Simd::EdgeBackgroundAdjustRangeMasked as deprecated.
• The mark of function Simd::EdgeBackgroundShiftRange as deprecated.
• The mark of function Simd::EdgeBackgroundShiftRangeMasked as deprecated.

Bug fixing

• Wrong assert in AVX-512BW optimizations of function BgrToRgb.
• MSVS compiler bug (Windows, Arm64).
• Error in function Simd::DrawLine.

Test framework

New features

• Tests for verifying functionality of WarpAffine engine.
• Special tests for verifying functionality of WarpAffine engine.

Infrastructure

New features

• SIMD_OPENCV Cmake option to test Simd with OpenCV support.

Documentation

Improving

• Using example in description of function RecursiveBilateralFilterInit.

Simd v5.2.120

Algorithms

New features

• AVX2 optimizations of class RecursiveBilateralFilterFast.
• Base implementation of function SynetNormalizeLayerForward.

Bug fixing

• Error in SSE4.1 optimizations of function SynetSetInput.
• MSVS compiler warning in AMX optimizations of class SynetConvolution8iNhwcDirect.
• MSVS compiler warning in AMX optimizations of class SynetMergedConvolution8iCdc.
• MSVS compiler warning in AMX optimizations of class SynetMergedConvolution8iCd.
• MSVS compiler warning in AMX optimizations of class SynetMergedConvolution8iDc.
• Error in AVX and AVX2 optimizations of function SynetInnerProductLayerForward.
• Using of SIMD_CPP_2011_ENABLE macro outside of library.

Test framework

New features

• Tests for verifying functionality of function SynetNormalizeLayerForward.

Removing

• Data test for function Fill.
• Data test for function FillFrame.
• Data test for function FillBgra.
• Data test for function FillBgr.
• Data test for function FillPixel.
• Data test for function Float32ToFloat16.
• Data test for function Float16ToFloat32.
• Data test for function SquaredDifferenceSum16f.
• Data test for function CosineDistance16f.
• Data test for function Float32ToUint8.
• Data test for function Uint8ToFloat32.
• Data test for function MeanFilter3x3.
• Data test for function MedianFilterRhomb3x3.
• Data test for function MedianFilterRhomb5x5.
• Data test for function MedianFilterSquare3x3.
• Data test for function MedianFilterSquare5x5.
• Data test for function GaussianBlur3x3.
• Data test for function AbsGradientSaturatedSum.
• Data test for function LbpEstimate.
• Data test for function NormalizeHistogram.
• Data test for function SobelDx.
• Data test for function SobelDxAbs.
• Data test for function SobelDy.
• Data test for function SobelDyAbs.
• Data test for function ContourMetrics.
• Data test for function Laplace.
• Data test for function LaplaceAbs.
• Data test for function Histogram.
• Data test for function HistogramMasked.
• Data test for function HistogramConditional.
• Data test for function AbsSecondDerivativeHistogram.
• Data test for function ChangeColors.
• Data test for function HogDirectionHistograms.
• Data test for function HogExtractFeatures.
• Data test for function HogDeinterleave.
• Data test for function HogFilterSeparable.
• Data test for function HogLiteExtractFeatures.
• Data test for function HogLiteFilterFeatures.
• Data test for function HogLiteResizeFeatures.
• Data test for function HogLiteCompressFeatures.
• Data test for function HogLiteFilterSeparable.
• Data test for function HogLiteFindMax7x7.
• Data test for function HogLiteCreateMask.
• Data test for function Integral.
• Data test for function InterferenceIncrement.
• Data test for function InterferenceIncrementMasked.
• Data test for function InterferenceDecrement.
• Data test for function InterferenceDecrementMasked.
• Data test for function InterleaveUv.
• Data test for function InterleaveBgr.
• Data test for function InterleaveBgra.
• Data test for function NeuralConvert.
• Data test for function NeuralProductSum.
• Data test for function NeuralAddVectorMultipliedByValue.
• Data test for function NeuralAddVector.
• Data test for function NeuralAddValue.
• Data test for function NeuralRoughSigmoid.
• Data test for function NeuralRoughSigmoid2.
• Data test for function NeuralDerivativeSigmoid.
• Data test for function NeuralRoughTanh.
• Data test for function NeuralDerivativeTanh.
• Data test for function NeuralDerivativeRelu.
• Data test for function NeuralPow.
• Data test for function NeuralUpdateWeights.
• Data test for function NeuralAdaptiveGradientUpdate.
• Data test for function NeuralPooling1x1Max3x3.
• Data test for function NeuralPooling2x2Max2x2.
• Data test for function NeuralPooling2x2Max3x3.
• Data test for function NeuralAddConvolution2x2Forward.
• Data test for function NeuralAddConvolution3x3Forward.
• Data test for function NeuralAddConvolution4x4Forward.
• Data test for function NeuralAddConvolution5x5Forward.
• Data test for function NeuralAddConvolution2x2Backward.
• Data test for function NeuralAddConvolution3x3Backward.
• Data test for function NeuralAddConvolution4x4Backward.
• Data test for function NeuralAddConvolution5x5Backward.
• Data test for function NeuralAddConvolution2x2Sum.
• Data test for function NeuralAddConvolution3x3Sum.
• Data test for function NeuralAddConvolution4x4Sum.
• Data test for function NeuralAddConvolution5x5Sum.
• Data test for function NeuralConvolutionForward.
• Data test for function OperationBinary8u.
• Data test for function OperationBinary16i.
• Data test for function VectorProduct.
• Data test for function ReduceColor2x2.
• Data test for function ReduceGray2x2.
• Data test for function ReduceGray3x3.
• Data test for function ReduceGray4x4.
• Data test for function ReduceGray5x5.
• Data test for function Reorder16bit.
• Data test for function Reorder32bit.
• Data test for function Reorder64bit.
• Data test for function ResizeBilinear.
• Data test for function SegmentationShrinkRegion.
• Data test for function SegmentationFillSingleHoles.
• Data test for function SegmentationChangeIndex.
• Data test for function SegmentationPropagate2x2.
• Data test for function ShiftBilinear.
• Data test for function GetStatistic.
• Data test for function GetMoments.
• Data test for function GetRowSums.
• Data test for function GetColSums.
• Data test for function GetAbsDyRowSums.
• Data test for function GetAbsDxColSums.
• Data test for function ValueSum.
• Data test for function SquareSum.
• Data test for function SobelDxAbsSum.
• Data test for function SobelDyAbsSum.
• Data test for function LaplaceAbsSum.
• Data test for function ValueSquareSum.
• Data test for function CorrelationSum.
• Data test for function StretchGray2x2.
• Data test for function SvmSumLinear.
• Data test for function SynetEltwiseLayerForward.
• Data test for function TextureBoostedSaturatedGradient.
• Data test for function TextureBoostedUv.
• Data test for function TextureGetDifferenceSum.
• Data test for function TexturePerformCompensation.
• Data test for function Yuv444pToBgr.
• Data test for function Yuv422pToBgr.
• Data test for function Yuv420pToBgr.
• Data test for function Yuv444pToHsl.
• Data test for function Yuv444pToHsv.
• Data test for function Yuv444pToHue.
• Data test for function Yuv420pToHue.
• Data test for function Yuv444pToBgra.
• Data test for function Yuv422pToBgra.
• Data test for function Yuv420pToBgra.
• Data test infrastructure.

Infrastructure

New features

• SIMD_RUNTIME CMake build option.

Simd v5.1.119

Algorithms

New features

• AMX optimizations of class SynetConvolution8iNhwcDirect.
• AMX optimizations of class SynetMergedConvolution8iCdc.
• AMX optimizations of class SynetMergedConvolution8iCd.
• AMX optimizations of class SynetMergedConvolution8iDc.

Improving

• Optimization of using of memory buffer in class SynetConvolution8iNhwcDirect.

Bug fixing

• MSVS compiler bug (Windows, Arm64).

Removing

• AVX-512VNNI optimizations of function SetDepthwise in class SynetConvolution8iNhwcDirect (it is equal to AVX-512BW version).

Test framework

New features

Removing

• Data test for function BayerToBgr.
• Data test for function BayerToBgra.
• Data test for function Bgr48pToBgra32.
• Data test for function Binarization.
• Data test for function AveragingBinarization.
• Data test for function ConditionalCount8u.
• Data test for function ConditionalCount16i.
• Data test for function ConditionalSum.
• Data test for function ConditionalSquareSum.
• Data test for function ConditionalSquareGradientSum.
• Data test for function ConditionalFill.
• Data test for function Copy.
• Data test for function CopyFrame.
• Data test for function Crc32c.
• Data test for function DeinterleaveUv.
• Data test for function DeinterleaveBgr.
• Data test for function DeinterleaveBgra.
• Data test for function DetectionHaarDetect32fp.
• Data test for function DetectionHaarDetect32fi.
• Data test for function DetectionLbpDetect32fp.
• Data test for function DetectionLbpDetect32fi.
• Data test for function DetectionLbpDetect16ip.
• Data test for function DetectionLbpDetect16ii.
• Data test for function AbsDifferenceSum.
• Data test for function AbsDifferenceSumMasked.
• Data test for function AbsDifferenceSums3x3.
• Data test for function AbsDifferenceSums3x3Masked.
• Data test for function SquaredDifferenceSum.
• Data test for function SquaredDifferenceSumMasked.
• Data test for function SquaredDifferenceSum32f.
• Data test for function SquaredDifferenceKahanSum32f.
• Data test for function CosineDistance32f.
• Data test for function AlphaBlending.
• Data test for function AlphaFilling.
• Data test for function EdgeBackgroundGrowRangeSlow.
• Data test for function EdgeBackgroundGrowRangeFast.
• Data test for function EdgeBackgroundIncrementCount.
• Data test for function EdgeBackgroundAdjustRange.
• Data test for function EdgeBackgroundAdjustRangeMasked.
• Data test for function EdgeBackgroundShiftRange.
• Data test for function EdgeBackgroundShiftRangeMasked.

Simd v5.1.118

Algorithms

New features

• Base implementation, SSE4.1 optimizations of RecursiveBilateralFilter engine.
• Support of ARGB32 format in View.
• Support of ARGB32 format in function AlphaPremultiply.
• Support of ARGB32 format in function AlphaUnpremultiply.
• AVX-512BW optimizations of function TileMatMul8u8i (AMX tile emulation).

Improving

• Base implementation of class ImagePngLoader.

Bug fixing

• Build errors on MacOS Arm64.
• Clang compiler warning (-mfpu=neon -mfpu=neon-fp16).
• Compiler errors (C++-98 specific).

Test framework

New features

• Tests for verifying functionality of RecursiveBilateralFilter engine.

Removing

• Data test for function AbsDifference.
• Data test for function AddFeatureDifference.
• Data test for function BgraToBgr.
• Data test for function BgraToGray.
• Data test for function BgrToGray.
• Data test for function BgrToHsl.
• Data test for function BgrToHsv.
• Data test for function GrayToBgr.
• Data test for function Int16ToGray.
• Data test for function BgrToBayer.
• Data test for function BgraToBayer.
• Data test for function BgrToBgra.
• Data test for function GrayToBgra.
• Data test for function BgraToYuv420p.
• Data test for function BgraToYuv422p.
• Data test for function BgraToYuv444p.
• Data test for function BgrToYuv420p.
• Data test for function BgrToYuv422p.
• Data test for function BgrToYuv444p.
• Data test for function BackgroundGrowRangeSlow.
• Data test for function BackgroundGrowRangeFast.
• Data test for function BackgroundIncrementCount.
• Data test for function BackgroundAdjustRange.
• Data test for function BackgroundAdjustRangeMasked.
• Data test for function BackgroundShiftRange.
• Data test for function BackgroundShiftRangeMasked.
• Data test for function BackgroundInitMask.

Simd v5.1.117

Algorithms

New features

• Base implementation, SSE4.1, AVX2 optimizations of function AlphaBlending2x.

Bug fixing

• Buffer overrun in Base implementation of function SynetFusedLayerForward9.
• Buffer overrun in SSE4.1 optimization of class SynetScale8i.
• Buffer overrun in AVX-512BW optimizations of class ResizerNearest.
• Buffer overrun in AVX-512BW optimizations of class ResizerByteBilinear.
• Buffer overrun in AVX-512BW optimizations of class ResizerByteBicubic.
• Buffer overrun in AVX-512BW optimizations of class ResizerByteArea1x1.
• Buffer overrun in AVX-512BW optimizations of class ResizerByteArea2x2.
• Buffer overrun in AVX-512BW optimizations of function TransformImage.
• Error in AVX-512BW optimizations of function Yuv420pToUyvy422.
• Crash in std::unordered_map after calling of some Simd function (Simd does not clear MMX registers after using).
• Error (possible negative output values) in AVX-512BW optimizations of function CosineDistancesMxNp16f.
• Error (possible negative output values) in AVX-512BW optimizations of function CosineDistancesMxNa16f.
• Error in AVX-512BW optimizations of function TransformImage.
• Valgrind warning in OutputMemoryStream.
• Memory leak in Base implementation of class ImagePngLoader.

Replacing

• Replace SSE2 optimizations to SSE4.1 for function SegmentationChangeIndex.
• Replace SSE2 optimizations to SSE4.1 for function SegmentationFillSingleHoles.
• Replace SSE2 optimizations to SSE4.1 for function SegmentationPropagate2x2.
• Replace SSE2 optimizations to SSE4.1 for function ShiftBilinear.
• Replace SSE2 optimizations to SSE4.1 for function SobelDx.
• Replace SSE2 optimizations to SSE4.1 for function SobelDy.
• Replace SSE2 optimizations to SSE4.1 for function ContourAnchors.
• Replace SSE2 optimizations to SSE4.1 for function SquaredDifferenceSum.
• Replace SSE2 optimizations to SSE4.1 for function SquaredDifferenceSumMasked.
• Replace SSE2 optimizations to SSE4.1 for function SquaredDifferenceSum32f.
• Replace SSE2 optimizations to SSE4.1 for function SquaredDifferenceKahanSum32f.
• Replace SSE2 optimizations to SSE4.1 for function GetStatistic.
• Replace SSE2 optimizations to SSE4.1 for function GetMoments.
• Replace SSE2 optimizations to SSE4.1 for function GetObjectMoments.
• Replace SSE2 optimizations to SSE4.1 for function GetRowSums.
• Replace SSE2 optimizations to SSE4.1 for function GetColSums.
• Replace SSE2 optimizations to SSE4.1 for function GetAbsDyRowSums.
• Replace SSE2 optimizations to SSE4.1 for function GetAbsDxColSums.
• Replace SSE2 optimizations to SSE4.1 for function ValueSum.
• Replace SSE2 optimizations to SSE4.1 for function SquareSum.
• Replace SSE2 optimizations to SSE4.1 for function ValueSquareSum.
• Replace SSE2 optimizations to SSE4.1 for function CorrelationSum.
• Replace SSE2 optimizations to SSE4.1 for function StretchGray2x2.
• Replace SSE2 optimizations to SSE4.1 for function TextureBoostedSaturatedGradient.
• Replace SSE2 optimizations to SSE4.1 for function TextureBoostedUv.
• Replace SSE2 optimizations to SSE4.1 for function TextureGetDifferenceSum.
• Replace SSE2 optimizations to SSE4.1 for function TexturePerformCompensation.
• Replace SSE2 optimizations to SSE4.1 for function Yuv420pToHue.
• Replace SSE2 optimizations to SSE4.1 for function Yuv444pToHue.
• Replace SSE2 optimizations to SSE4.1 for function Yuva420pToBgra.
• Replace SSE2 optimizations to SSE4.1 for function Yuv420pToBgra.
• Replace SSE2 optimizations to SSE4.1 for function Yuv420pToBgraV2.
• Replace SSE2 optimizations to SSE4.1 for function Yuv422pToBgra.
• Replace SSE2 optimizations to SSE4.1 for function Yuv444pToBgra.
• Replace SSE2 optimizations to SSE4.1 for function Yuv444pToBgraV2.
• Replace SSE2 optimizations to SSE4.1 for function SynetPoolingAverage.
• Replace SSE2 optimizations to SSE4.1 for function SynetScaleLayerForward.
• Replace SSE2 optimizations to SSE4.1 for function SynetConvert32fTo8u.
• Replace SSE2 optimizations to SSE4.1 for function SynetReorderImage.
• Replace SSE2 optimizations to SSE4.1 for function SynetReorderFilter.
• Replace SSE2 optimizations to SSE4.1 for function SynetFusedLayerForward0.
• Replace SSE2 optimizations to SSE4.1 for function SynetFusedLayerForward1.
• Replace SSE2 optimizations to SSE4.1 for function SynetFusedLayerForward2.
• Replace SSE2 optimizations to SSE4.1 for function SynetFusedLayerForward3.
• Replace SSE2 optimizations to SSE4.1 for function SynetFusedLayerForward4.
• Replace SSE2 optimizations to SSE4.1 for function SynetFusedLayerForward8.
• Replace SSE2 optimizations to SSE4.1 for function SynetFusedLayerForward9.
• Replace SSE2 optimizations to SSE4.1 for function SynetDeconvolution32fInit.
• Replace SSE2 optimizations to SSE4.1 for class SynetDeconvolution32fGemmNN.
• Replace SSE2 optimizations to SSE4.1 for class SynetDeconvolution32fNhwcDirect2x2.
• Replace SSE2 optimizations to SSE4.1 for function SynetConvolution32fInit.
• Replace SSE2 optimizations to SSE4.1 for class SynetConvolution32fDepthwiseDotProduct.
• Replace SSE2 optimizations to SSE4.1 for class SynetConvolution32fWinograd.
• Replace SSE2 optimizations to SSE4.1 for class SynetConvolution32fDirectNchw.
• Replace SSE2 optimizations to SSE4.1 for class SynetConvolution32fNhwcDirect.
• Replace SSE2 optimizations to SSE4.1 for class SynetConvolution32fDirectNhwc.
• Replace SSE2 optimizations to SSE4.1 for class SynetConvolution32fGemmNN.
• Replace SSE2 optimizations to SSE4.1 for class SynetMergedConvolution32fCdc.
• Replace SSE2 optimizations to SSE4.1 for class SynetMergedConvolution32fCd.
• Replace SSE2 optimizations to SSE4.1 for class SynetMergedConvolution32fDc.
• Replace SSE2 optimizations to SSE4.1 for function SynetMergedConvolution32fInit.
• Replace SSE2 optimizations to SSE4.1 for function SynetAddBias.
• Replace SSE2 optimizations to SSE4.1 for function SynetEltwiseLayerForward.
• Replace SSE2 optimizations to SSE4.1 for function SynetInnerProductLayerForward.
• Replace SSE2 optimizations to SSE4.1 for function SynetLrnLayerCrossChannels.
• Replace SSE2 optimizations to SSE4.1 for function SynetShuffleLayerForward.
• Replace SSE2 optimizations to SSE4.1 for function SynetSoftmaxLayerForward.
• Replace SSE2 optimizations to SSE4.1 for function SynetElu32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetHardSigmoid32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetHswish32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetMish32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetPreluLayerForward.
• Replace SSE2 optimizations to SSE4.1 for function SynetRelu32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetRestrictRange32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetSigmoid32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetSoftplus32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetSwish32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetTanh32f.
• Replace SSE2 optimizations to SSE4.1 for function SynetGemm32fNN.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel1x3Block1x4SetFilter.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel1x3Block1x4SetInput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel1x3Block1x4SetOutput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel1x5Block1x4SetFilter.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel1x5Block1x4SetInput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel1x5Block1x4SetOutput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel2x2Block2x2SetFilter.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel2x2Block2x2SetInput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel2x2Block2x2SetOutput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel2x2Block4x4SetFilter.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel2x2Block4x4SetInput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel2x2Block4x4SetOutput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block2x2SetFilter.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block2x2SetInput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block2x2SetOutput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block3x3SetFilter.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block3x3SetInput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block3x3SetOutput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block4x4SetFilter.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block4x4SetInput.
• Replace SSE2 optimizations to SSE4.1 for function WinogradKernel3x3Block4x4SetOutput.

Test...

Simd v5.0.116

Algorithms

New features

• Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of function Yuva444pToBgraV2.
• Function SimdEmpty.
• Checking of no man's land watermarks in function SimdFree.

Improving

• AVX-512BW optimizations of AMX tile emulation.
• AMX optimizations of class SynetConvolution32fBf16Nhwc.
• AMX optimizations of class SynetMergedConvolution32fBf16Cdc.
• AMX optimizations of class SynetMergedConvolution32fBf16Cd.

Bug fixing

• GCC linker error when SIMD_AMX_EMULATE macro is switched on.
• Error in SSE4.1, AVX2, AVX-512BW, AMX optimizations of class SynetConvolution32fBf16Nhwc.
• Wrong assert in SSE4.1 and AVX-512BW optimizations of class ResizerNearest.
• Error in AVX optimizations of class SynetMergedConvolution32fCdc.
• Error in Base implementation, SSE4.1, AVX2, AVX-512BW, AVX-512BF16, AMX optimizations of class SynetMergedConvolution32fBf16Cdc.
• External buffer reading overflow in class SynetMergedConvolution32fBf16Cdc.
• External buffer reading overflow in class SynetMergedConvolution32fBf16Cd.
• External buffer reading overflow in class SynetMergedConvolution32fBf16Dc.
• FP32 overflow in SSE2, AVX2, AVX-512BW, NEON optimizations of function Tanh.
• Error in function Base::SynetConvolution32fGemmNN::ImgToCol.
• Error in SSE4.1, AVX2 optimizations of class ResizerByteArea2x2.
• Buffer overrun in SSE4.1, AVX2 optimizations of class ResizerNearest.
• Buffer overrun in SSE4.1, AVX2 optimizations of class ResizerByteBilinear.
• Buffer overrun in SSE4.1, AVX2 optimizations of class ResizerByteBicubic.
• Buffer overrun in SSE4.1, AVX2 optimizations of class ResizerByteArea1x1.
• Buffer overrun in SSE4.1, AVX2 optimizations of class ResizerByteArea2x2.

Replacing

• Replace SSE2 optimizations to SSE4.1 for function SvmSumLinear.
• Replace SSE2 optimizations to SSE4.1 for function AbsDifference.
• Replace SSE2 optimizations to SSE4.1 for function AbsDifferenceSum.
• Replace SSE2 optimizations to SSE4.1 for function AbsDifferenceSumMasked.
• Replace SSE2 optimizations to SSE4.1 for function AbsDifferenceSums3x3.
• Replace SSE2 optimizations to SSE4.1 for function AbsDifferenceSums3x3Masked.
• Replace SSE2 optimizations to SSE4.1 for function AbsGradientSaturatedSum.
• Replace SSE2 optimizations to SSE4.1 for function AddFeatureDifference.
• Replace SSE2 optimizations to SSE4.1 for function AlphaBlending.
• Replace SSE2 optimizations to SSE4.1 for function AlphaBlendingUniform.
• Replace SSE2 optimizations to SSE4.1 for function AlphaFilling.
• Replace SSE2 optimizations to SSE4.1 for function AlphaPremultiply.
• Replace SSE2 optimizations to SSE4.1 for function BackgroundGrowRangeSlow.
• Replace SSE2 optimizations to SSE4.1 for function BackgroundGrowRangeFast.
• Replace SSE2 optimizations to SSE4.1 for function BackgroundIncrementCount.
• Replace SSE2 optimizations to SSE4.1 for function BackgroundAdjustRange.
• Replace SSE2 optimizations to SSE4.1 for function BackgroundAdjustRangeMasked.
• Replace SSE2 optimizations to SSE4.1 for function BackgroundShiftRange.
• Replace SSE2 optimizations to SSE4.1 for function BackgroundShiftRangeMasked.
• Replace SSE2 optimizations to SSE4.1 for function BackgroundInitMask.
• Replace SSE2 optimizations to SSE4.1 for function EdgeBackgroundGrowRangeSlow.
• Replace SSE2 optimizations to SSE4.1 for function EdgeBackgroundGrowRangeFast.
• Replace SSE2 optimizations to SSE4.1 for function EdgeBackgroundIncrementCount.
• Replace SSE2 optimizations to SSE4.1 for function EdgeBackgroundAdjustRange.
• Replace SSE2 optimizations to SSE4.1 for function EdgeBackgroundAdjustRangeMasked.
• Replace SSE2 optimizations to SSE4.1 for function EdgeBackgroundShiftRangeMasked.
• Replace SSE2 optimizations to SSE4.1 for function BayerToBgra.
• Replace SSE2 optimizations to SSE4.1 for function BgraToGray.
• Replace SSE2 optimizations to SSE4.1 for function BgraToYuv420p.
• Replace SSE2 optimizations to SSE4.1 for function BgraToYuv422p.
• Replace SSE2 optimizations to SSE4.1 for function BgraToYuv444p.
• Replace SSE2 optimizations to SSE4.1 for function BgraToYuva420p.
• Replace SSE2 optimizations to SSE4.1 for function BgrToGray.
• Replace SSE2 optimizations to SSE4.1 for function RgbaToGray.
• Replace SSE2 optimizations to SSE4.1 for function Bgr48pToBgra32.
• Replace SSE2 optimizations to SSE4.1 for function Binarization.
• Replace SSE2 optimizations to SSE4.1 for function AveragingBinarization.
• Replace SSE2 optimizations to SSE4.1 for function ConditionalCount8u.
• Replace SSE2 optimizations to SSE4.1 for function ConditionalCount16i.
• Replace SSE2 optimizations to SSE4.1 for function ConditionalSum.
• Replace SSE2 optimizations to SSE4.1 for function ConditionalSquareSum.
• Replace SSE2 optimizations to SSE4.1 for function ConditionalSquareGradientSum.
• Replace SSE2 optimizations to SSE4.1 for function ConditionalFill.
• Replace SSE2 optimizations to SSE4.1 for function DeinterleaveUv.
• Replace SSE2 optimizations to SSE4.1 for function Fill32f.
• Replace SSE2 optimizations to SSE4.1 for function FillBgr.
• Replace SSE2 optimizations to SSE4.1 for function FillBgra.
• Replace SSE2 optimizations to SSE4.1 for function FillPixel.
• Replace SSE2 optimizations to SSE4.1 for function CosineDistance32f.
• Replace SSE2 optimizations to SSE4.1 for function Float32ToUint8.
• Replace SSE2 optimizations to SSE4.1 for function Uint8ToFloat32.
• Replace SSE2 optimizations to SSE4.1 for function GaussianBlur3x3.
• Replace SSE2 optimizations to SSE4.1 for function GrayToBgra.
• Replace SSE2 optimizations to SSE4.1 for function AbsSecondDerivativeHistogram.
• Replace SSE2 optimizations to SSE4.1 for function HistogramMasked.
• Replace SSE2 optimizations to SSE4.1 for function HistogramConditional.
• Replace SSE2 optimizations to SSE4.1 for function HogDirectionHistograms.
• Replace SSE2 optimizations to SSE4.1 for function HogDeinterleave.
• Replace SSE2 optimizations to SSE4.1 for function HogFilterSeparable.
• Replace SSE2 optimizations to SSE4.1 for function Int16ToGray.
• Replace SSE2 optimizations to SSE4.1 for function InterferenceIncrement.
• Replace SSE2 optimizations to SSE4.1 for function InterferenceIncrementMasked.
• Replace SSE2 optimizations to SSE4.1 for function InterferenceDecrement.
• Replace SSE2 optimizations to SSE4.1 for function InterferenceDecrementMasked.
• Replace SSE2 optimizations to SSE4.1 for function InterleaveUv.
• Replace SSE2 optimizations to SSE4.1 for function Laplace.
• Replace SSE2 optimizations to SSE4.1 for function LbpEstimate.
• Replace SSE2 optimizations to SSE4.1 for function MeanFilter3x3.
• Replace SSE2 optimizations to SSE4.1 for function MedianFilterRhomb3x3.
• Replace SSE2 optimizations to SSE4.1 for function MedianFilterRhomb5x5.
• Replace SSE2 optimizations to SSE4.1 for function MedianFilterSquare3x3.
• Replace SSE2 optimizations to SSE4.1 for function MedianFilterSquare5x5.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution2x2Forward.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution3x3Forward.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution4x4Forward.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution5x5Forward.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution2x2Backward.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution3x3Backward.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution4x4Backward.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution5x5Backward.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution2x2Sum.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution3x3Sum.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution4x4Sum.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddConvolution5x5Sum.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAdaptiveGradientUpdate.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddVectorMultipliedByValue.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddVector.
• Replace SSE2 optimizations to SSE4.1 for function NeuralAddValue.
• Replace SSE2 optimizations to SSE4.1 for function NeuralConvert.
• Replace SSE2 optimizations to SSE4.1 for function NeuralDerivativeRelu.
• Replace SSE2 optimizations to SSE4.1 for function NeuralDerivativeSigmoid.
• Replace SSE2 optimizations to SSE4.1 for function NeuralDerivativeTanh.
• Replace SSE2 optimizations to SSE4.1 for function NeuralPooling1x1Max3x3.
• Replace SSE2 optimizations to SSE4.1 for function NeuralPooling2x2Max2x2.
• Replace SSE2 optimizations to SSE4.1 for function NeuralPooling2x2Max3x3.
• Replace SSE2 optimizations to SSE4.1 for function NeuralPow.
• Replace SSE2 optimizations to SSE4.1 for function NeuralProdu...

Simd v5.0.115

Algorithms

New features

• Base implementation, SSE4.1, AVX2, AVX-512BW, AVX-512BF16, AMX optimizations of class SynetMergedConvolution32fBf16Cdc.
• Base implementation, SSE4.1, AVX2, AVX-512BW, AVX-512BF16, AMX optimizations of class SynetMergedConvolution32fBf16Cd.
• Base implementation, SSE4.1, AVX2, AVX-512BW, AVX-512BF16, AMX optimizations of class SynetMergedConvolution32fBf16Dc.
• AVX-512BF16 extension support.
• AVX-512BF16 optimizations of function Float32ToBFloat16.
• AVX-512BF16, AMX optimizations of class SynetConvolution32fBf16Nhwc.
• AMX extension support.
• Support of 3D pooling in Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of function SynetPoolingMax32f.

Improving

• AVX-512BW optimizations of function Fill32f.

Renaming

• Rename function SynetPoolingForwardAverage to SynetPoolingAverage.
• Rename function SynetPoolingForwardMax32f to SynetPoolingMax32f.
• Rename function SynetPoolingForwardMax8u to SynetPoolingMax8u.

Replacing

• Replace AVX-512F optimizations to AVX-512BW for function SvmSumLinear.
• Replace AVX-512F optimizations to AVX-512BW for function Fill32f.
• Replace AVX-512F optimizations to AVX-512BW for class ResizerNearest.
• Replace AVX-512F optimizations to AVX-512BW for class ResizerFloatBilinear.
• Replace AVX-512F optimizations to AVX-512BW for function SquaredDifferenceSum32f.
• Replace AVX-512F optimizations to AVX-512BW for function SquaredDifferenceKahanSum32f.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralConvolutionForward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution2x2Forward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution2x2Backward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution2x2Sum.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution3x3Forward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution3x3Backward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution3x3Sum.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution4x4Forward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution4x4Backward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution4x4Sum.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution5x5Forward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution5x5Backward.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddConvolution5x5Sum.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralProductSum.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAdaptiveGradientUpdate.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralPooling1x1Max3x3.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralPooling2x2Max2x2.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralPooling2x2Max3x3.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralUpdateWeights.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddValue.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddVector.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralAddVectorMultipliedByValue.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralRoughSigmoid.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralRoughSigmoid2.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralDerivativeSigmoid.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralRoughTanh.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralDerivativeTanh.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralDerivativeRelu.
• Replace AVX-512F optimizations to AVX-512BW for function NeuralPow.
• Replace AVX-512F optimizations to AVX-512BW for class SynetConvolution32fGemmNN.
• Replace AVX-512F optimizations to AVX-512BW for class SynetConvolution32fGemmNT.
• Replace AVX-512F optimizations to AVX-512BW for class SynetConvolution32fWinograd.
• Replace AVX-512F optimizations to AVX-512BW for class SynetDeconvolution32fGemmNN.
• Replace AVX-512F optimizations to AVX-512BW for class SynetDeconvolution32fNhwcDirect2x2.
• Replace AVX-512F optimizations to AVX-512BW for function SynetDeconvolution32fInit.
• Replace AVX-512F optimizations to AVX-512BW for class SynetInnerProduct32fGemm.
• Replace AVX-512F optimizations to AVX-512BW for class SynetInnerProduct32fProd.
• Replace AVX-512F optimizations to AVX-512BW for function SynetInnerProduct32fInit.
• Replace AVX-512F optimizations to AVX-512BW for function ConvolutionBiasAndActivation.
• Replace AVX-512F optimizations to AVX-512BW for function SynetReorderImage.
• Replace AVX-512F optimizations to AVX-512BW for function SynetReorderFilter.
• Replace AVX-512F optimizations to AVX-512BW for function Gemm32fNN.
• Replace AVX-512F optimizations to AVX-512BW for function Gemm32fNT.
• Replace AVX-512F optimizations to AVX-512BW for function SynetFusedLayerForward0.
• Replace AVX-512F optimizations to AVX-512BW for function SynetFusedLayerForward1.
• Replace AVX-512F optimizations to AVX-512BW for function SynetFusedLayerForward2.
• Replace AVX-512F optimizations to AVX-512BW for function SynetFusedLayerForward3.
• Replace AVX-512F optimizations to AVX-512BW for function SynetFusedLayerForward4.
• Replace AVX-512F optimizations to AVX-512BW for function SynetFusedLayerForward8.
• Replace AVX-512F optimizations to AVX-512BW for function SynetFusedLayerForward9.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel1x3Block1x4SetFilter.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel1x3Block1x4SetInput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel1x3Block1x4SetOutput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel1x5Block1x4SetFilter.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel1x5Block1x4SetInput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel1x5Block1x4SetOutput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel2x2Block2x2SetFilter.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel2x2Block2x2SetInput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel2x2Block2x2SetOutput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel2x2Block4x4SetFilter.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel2x2Block4x4SetInput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel2x2Block4x4SetOutput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block2x2SetFilter.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block2x2SetInput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block2x2SetOutput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block3x3SetFilter.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block3x3SetInput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block3x3SetOutput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block4x4SetFilter.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block4x4SetInput.
• Replace AVX-512F optimizations to AVX-512BW for function WinogradKernel3x3Block4x4SetOutput.
• Replace AVX-512F optimizations to AVX-512BW for function SynetElu32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetHardSigmoid32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetHswish32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetMish32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetPreluLayerForward.
• Replace AVX-512F optimizations to AVX-512BW for function SynetRelu32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetRestrictRange32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetSigmoid32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetSoftplus32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetSwish32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetTanh32f.
• Replace AVX-512F optimizations to AVX-512BW for function SynetScaleLayerForward.
• Replace AVX-512F optimizations to AVX-512BW for function SynetPoolingAverage.
• Replace AVX-512F optimizations to AVX-512BW for function SynetAddBias.
• Replace AVX-512F optimizations to AVX-512BW for function SynetEltwiseLayerForward.
• Replace AVX-512F optimizations to AVX-512BW for function SynetInnerProductLayerForward.
• Replace AVX-512F optimizations to AVX-512BW for function SynetLrnLayerCrossChannels.