Download Latest Version Simd v7.2.165 source code.zip (7.7 MB)
Email in envelope

Get an email when there's a new version of Simd Library

Home / v7.2.164
Name Modified Size InfoDownloads / Week
Parent folder
README.md 2026-08-03 15.6 kB
0
Simd v7.2.164 source code.tar.gz 2026-08-03 5.3 MB
0
Simd v7.2.164 source code.zip 2026-08-03 7.3 MB
0
Totals: 3 Items   12.7 MB 0

Algorithms

New features
  • SVE2 optimizations of function AbsDifference.
  • SVE2 optimizations of function AbsDifferenceSum.
  • SVE2 optimizations of function AbsDifferenceSumMasked.
  • SVE2 optimizations of function AbsDifferenceSums3x3.
  • SVE2 optimizations of function AbsDifferenceSums3x3Masked.
  • SVE2 optimizations of function AbsGradientSaturatedSum.
  • SVE2 optimizations of function Int16ToGray.
  • SVE2 optimizations of function OperationBinary8u.
  • SVE2 optimizations of function OperationBinary16i.
  • SVE2 optimizations of function DeinterleaveUv.
  • SVE2 optimizations of function DeinterleaveBgr.
  • SVE2 optimizations of function DeinterleaveBgra.
  • SVE2 optimizations of function InterleaveBgr.
  • SVE2 optimizations of function InterleaveBgra.
  • SVE2 optimizations of function GetStatistic.
  • NEON, SVE2 optimizations of function GetColSums.
  • SVE2 optimizations of class ResizerByteArea1x1.
  • SVE2 optimizations of class ResizerByteArea2x2.
  • SVE2 optimizations of class ResizerByteBilinearOpenCv.
  • SVE2 optimizations of class ResizerByteBilinear.
  • SVE2 optimizations of class ResizerByteBicubic.
  • SVE2 optimizations of class ResizerBf16Bilinear.
  • SVE2 optimizations of class ResizerFloatBilinear.
  • SVE2 optimizations of class ResizerNearest.
  • NEON, SVE2 optimizations of class RecursiveBilateralFilterFast.
  • NEON, SVE2 optimizations of class RecursiveBilateralFilterPrecize.
  • SVE2 optimizations of function StretchGray2x2.
  • SVE2 optimizations of function ShiftBilinear.
  • NEON, SVE2 optimizations of function SynetAdd16b.
  • SVE2 optimizations of function SynetAdd8i.
  • SVE2 optimizations of function SynetAddBias.
  • NEON, SVE2 optimizations of function SynetChannelSum16b.
  • SVE2 optimizations of function SynetConvert32fTo8u.
  • SVE2 optimizations of function SynetConvert8uTo32f.
  • SVE2 optimizations of class SynetConvolution8iNhwcDirect.
  • SVE2 optimizations of classes SynetDeconvolution32fGemmNN and SynetDeconvolution32fNhwcDirect2x2.
  • SVE2 optimizations of function SynetDequantizeLinear.
  • SVE2 optimizations of function SynetEltwiseLayerForward.
  • SVE2 optimizations of function SynetElu32f.
  • SVE2 optimizations of function SynetGelu32f.
  • SVE2 optimizations of function SynetHardSigmoid32f.
  • SVE2 optimizations of function SynetHswish32f.
  • SVE2 optimizations of classes SynetMergedConvolution32fCdc, SynetMergedConvolution32fCd and SynetMergedConvolution32fDc.
  • NEON, SVE2 optimizations of classes SynetMergedConvolution8iCdc, SynetMergedConvolution8iCd and SynetMergedConvolution8iDc.
  • SVE2 optimizations of function SynetMish32f.
  • SVE2 optimizations of function SynetPreluLayerForward.
  • SVE2 optimizations of function SynetRelu32f.
  • NEON, SVE2 optimizations of function SynetRelu16b.
  • SVE2 optimizations of function SynetRestrictRange32f.
  • SVE2 optimizations of function SynetSigmoid32f.
  • SVE2 optimizations of function SynetNormalizeLayerForward.
  • SVE2 optimizations of function SynetNormalizeLayerForwardV2.
  • NEON, SVE2 optimizations of function SynetNormalizeLayerForward16bV2.
  • SVE2 optimizations of function SynetNormalizeLayerForwardV3.
  • SVE2 optimizations of function SynetNormalizeLayerForwardV4.
  • SVE2 optimizations of function SynetPoolingAverage.
  • SVE2 optimizations of function SynetPoolingMax32f.
  • NEON, SVE2 optimizations of function SynetPoolingMax16b.
  • SVE2 optimizations of function SynetPoolingMax8u.
  • SVE2 optimizations of function SynetPermuteInit.
  • SVE2 optimizations of class SynetQuantizedAddUniform.
  • SVE2 optimizations of function SynetQuantizedScaleLayerForward.
  • SVE2 optimizations of function SynetScaleLayerForward.
  • SVE2 optimizations of class SynetScale8i.
  • NEON, SVE2 optimizations of class SynetScale16b.
  • SVE2 optimizations of function SynetShuffleLayerForward.
  • SVE2 optimizations of function SynetSoftmax32f.
  • NEON, SVE2 optimizations of function SynetSoftmax16b.
  • SVE2 optimizations of function SynetSoftplus32f.
  • SVE2 optimizations of function SynetSetInput.
  • SVE2 optimizations of function SynetSwish32f.
  • SVE2 optimizations of function SynetTanh32f.
  • SVE2 optimizations of function SynetTiledScale2D32f.
  • SVE2 optimizations of function SynetUnaryOperation32f.
  • NEON, SVE2 optimizations of function SynetInnerProduct8i.
  • SVE2 optimizations of function SynetInnerProductLayerForward.
  • SVE2 optimizations of function SynetLrnLayerCrossChannels.
  • NEON, SVE2 optimizations of class SynetGridSample2d32fBlZ.
  • SVE2 optimizations of function TextureBoostedSaturatedGradient.
  • SVE2 optimizations of function TextureBoostedUv.
  • NEON, SVE2 optimizations of class WarpAffineNearest.
  • NEON, SVE2 optimizations of class WarpAffineByteBilinear.
  • SVE2 optimizations of function WinogradKernel1x3Block1x4SetFilter.
  • SVE2 optimizations of function WinogradKernel1x3Block1x4SetInput.
  • SVE2 optimizations of function WinogradKernel1x3Block1x4SetOutput.
  • SVE2 optimizations of function WinogradKernel1x5Block1x4SetFilter.
  • SVE2 optimizations of function WinogradKernel1x5Block1x4SetInput.
  • SVE2 optimizations of function WinogradKernel1x5Block1x4SetOutput.
  • SVE2 optimizations of function WinogradKernel2x2Block2x2SetFilter.
  • SVE2 optimizations of function WinogradKernel2x2Block2x2SetInput.
  • SVE2 optimizations of function WinogradKernel2x2Block2x2SetOutput.
  • SVE2 optimizations of function WinogradKernel2x2Block4x4SetFilter.
  • SVE2 optimizations of function WinogradKernel2x2Block4x4SetInput.
  • SVE2 optimizations of function WinogradKernel2x2Block4x4SetOutput.
  • SVE2 optimizations of function WinogradKernel3x3Block2x2SetFilter.
  • SVE2 optimizations of function WinogradKernel3x3Block2x2SetInput.
  • SVE2 optimizations of function WinogradKernel3x3Block2x2SetOutput.
  • SVE2 optimizations of function WinogradKernel3x3Block3x3SetFilter.
  • SVE2 optimizations of function WinogradKernel3x3Block3x3SetInput.
  • SVE2 optimizations of function WinogradKernel3x3Block3x3SetOutput.
  • SVE2 optimizations of function WinogradKernel3x3Block4x4SetFilter.
  • SVE2 optimizations of function WinogradKernel3x3Block4x4SetInput.
  • SVE2 optimizations of function WinogradKernel3x3Block4x4SetOutput.
  • SVE2 optimizations of function TextureGetDifferenceSum.
  • SVE2 optimizations of function TexturePerformCompensation.
  • SVE2 optimizations of function TransformImage.
  • SVE2 optimizations of function Uyvy422ToBgr.
  • SVE2 optimizations of function Uyvy422ToYuv420p.
  • SVE2 optimizations of function Uint8ToFloat32.
  • SVE2 optimizations of function VectorProduct.
  • SVE2 optimizations of function VectorNormNa16f.
  • SVE2 optimizations of function VectorNormNp16f.
  • SVE2 optimizations of function Yuv420pToBgrV2.
  • SVE2 optimizations of function Yuv420pToRgbV2.
  • SVE2 optimizations of function Yuv420pToBgraV2.
  • SVE2 optimizations of function Yuva420pToBgraV2.
  • SVE2 optimizations of function Yuv420pToHue.
  • SVE2 optimizations of function Yuv420pToUyvy422.
  • SVE2 optimizations of function Yuv422pToBgrV2.
  • SVE2 optimizations of function Yuv422pToRgbV2.
  • SVE2 optimizations of function Yuv422pToBgraV2.
  • SVE2 optimizations of function Yuva422pToBgraV2.
  • SVE2 optimizations of function Yuva444pToBgraV2.
  • SVE2 optimizations of function Yuv444pToBgrV2.
  • SVE2 optimizations of function Yuv444pToRgbV2.
  • SVE2 optimizations of function Yuv444pToBgraV2.
  • SVE2 optimizations of function Yuv444pToHue.
  • SVE2 optimizations of function Yuv444pToHsl.
  • SVE2 optimizations of function Yuv444pToHsv.
  • SVE2 optimizations of function Yuv444pToRgbaV2.
  • SVE2 optimizations of function YToGray.
  • SVE2 optimizations of function SynetQuantizeLinear.
  • SVE2 optimizations of function SynetQuantizedConcatLayerForward.
  • SVE2 optimizations of function SynetQuantizedShuffleLayerForward.
  • SVE2 optimizations of function SynetQuantizedPreluLayerForward.
  • Base implementation, SSE4.1, AVX2, AVX-512BW, NEON, SVE2 optimizations of function SynetQuantizedHswish.
  • Base implementation, SSE4.1, AVX2, AVX-512BW, NEON, SVE2 optimizations of class SynetQuantizedMulUniversal.
  • Base implementation, SSE4.1, AVX2, AVX-512BW, NEON, SVE2 optimizations of function SynetQuantizedHardSigmoid.
  • Base implementation, SSE4.1, AVX2, AVX-512BW, NEON, SVE2 optimizations of function SynetQuantizedPoolingAverage.
  • Base implementation, SSE4.1, AVX2, AVX-512BW, NEON, SVE2 optimizations of function SynetSetInput (for case channels == 4).
Improving
  • SVE2 optimizations of function GrayToY.
  • SVE2 optimizations of function YToGray.
  • SSE4.1, AVX2, AVX-512BW, NEON, SVE2 optimizations of function GetColSums.
  • SVE2 optimizations of class SynetConvolution8iNhwcDirect.
  • SVE2 optimizations of function InterleaveBgra.
  • SVE2 optimizations of function InterleaveBgr.
  • SVE2 optimizations of function AlphaUnpremultiply.
  • SVE2 optimizations of function Base64Decode.
  • SVE2 optimizations of function BgrToBgra.
  • SVE2 optimizations of function RgbToBgra.
  • SVE2 optimizations of function DetectionLbpDetect16ii.
  • SVE2 optimizations of function GaussianBlur3x3.
Removing
  • SVE optimization of function AbsDifference.
  • SVE optimization of function AbsDifferenceSum.
  • SVE optimization of function AbsDifferenceSumMasked.
  • SVE optimization of function AbsDifferenceSums3x3.
  • SVE optimization of function AbsDifferenceSums3x3Masked.
  • SVE optimization of function AbsGradientSaturatedSum.
  • SVE optimization of function InterleaveUv.
  • SVE optimization of function InterleaveBgra.
  • SVE optimization of function InterleaveBgr.
  • SVE optimization of function DeinterleaveUv.
  • SVE optimization of function DeinterleaveBgra.
  • SVE optimization of function DeinterleaveBgr.
  • SVE optimization of function OperationBinary16i.
  • SVE optimization of function OperationBinary8u.
  • SVE optimization of function GetStatistic.
Bug fixing
  • Error in NEON optimization of function CosineDistance16f.
  • Error in SVE2 optimizations of class ResizerByteBilinear (buffer overrun in Debug for some destination widths).
Deprecation
  • Set 'deprecated' warning for function SimdHogDirectionHistograms.
  • Set 'deprecated' warning for function SimdHogExtractFeatures.
  • Set 'deprecated' warning for function SimdHogDeinterleave.
  • Set 'deprecated' warning for function SimdHogFilterSeparable.
  • Set 'deprecated' warning for function Simd::HogDirectionHistograms.
  • Set 'deprecated' warning for function SimdNeuralConvert.
  • Set 'deprecated' warning for function SimdNeuralProductSum.
  • Set 'deprecated' warning for function SimdNeuralAddVectorMultipliedByValue.
  • Set 'deprecated' warning for function SimdNeuralAddVector.
  • Set 'deprecated' warning for function SimdNeuralAddValue.
  • Set 'deprecated' warning for function SimdNeuralDerivativeSigmoid.
  • Set 'deprecated' warning for function SimdNeuralDerivativeTanh.
  • Set 'deprecated' warning for function SimdNeuralDerivativeRelu.
  • Set 'deprecated' warning for function SimdNeuralPow.
  • Set 'deprecated' warning for function SimdNeuralUpdateWeights.
  • Set 'deprecated' warning for function SimdNeuralAdaptiveGradientUpdate.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution2x2Forward.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution3x3Forward.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution4x4Forward.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution5x5Forward.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution2x2Backward.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution3x3Backward.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution4x4Backward.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution5x5Backward.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution2x2Sum.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution3x3Sum.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution4x4Sum.
  • Set 'deprecated' warning for function SimdNeuralAddConvolution5x5Sum.
  • Set 'deprecated' warning for function SimdNeuralConvolutionForward.
  • Set 'deprecated' warning for function SimdNeuralPooling1x1Max3x3.
  • Set 'deprecated' warning for function SimdNeuralPooling2x2Max2x2.
  • Set 'deprecated' warning for function SimdNeuralPooling2x2Max3x3.
  • Set 'deprecated' warning for function Simd::NeuralConvert.
  • Set 'deprecated' warning for function SimdVectorProduct.
  • Set 'deprecated' warning for function Simd::VectorProduct.

Test framework

New features
  • Tests for verifying functionality of function SynetQuantizedHardSigmoid.
  • Tests for verifying functionality of function SynetQuantizedHswish.
  • Tests for verifying functionality of function SynetQuantizedMulInit.
  • Tests for verifying functionality of function SynetQuantizedPoolingAverage.
Bug fixing
  • Logical error in DifferenceSum16fAutoTest.

Documentation

Improving
  • Improved description of function Simd::PrintInfo.
  • Improved description of function Simd::AbsDifference.
  • Improved description of function Simd::AbsDifferenceSum.
  • Improved description of function Simd::AbsDifferenceSums3x3.
  • Improved description of function Simd::AbsGradientSaturatedSum.
  • Improved description of function Simd::AddFeatureDifference.
  • Improved description of function Simd::AlphaBlending.
  • Improved description of function Simd::AlphaBlending2x.
  • Improved description of function Simd::AlphaBlendingBgraToYuv420p.
  • Improved description of function Simd::AlphaFilling.
  • Improved description of function Simd::AlphaPremultiply.
  • Improved description of function Simd::AlphaUnpremultiply.
    • Source: README.md, updated 2026-08-03