Skip to content

Simd v7.1.161

Choose a tag to compare

View all tags
@ermig1979 ermig1979 released this 04 May 11:53
v7.1.161
24611ee

Algorithms

New features
  • SSE4.1, AVX2, AVX-512BW, NEON optimizations of function BgrToHsv.
  • SSE4.1, AVX2, AVX-512BW, NEON optimizations of function Yuv44pToHsl.
  • SSE4.1, AVX2, AVX-512BW, NEON optimizations of function Yuv44pToHsv.
  • Base implementation, SSE4.1, AVX2, AVX-512BW optimizations of function SynetPoolingMax16b.
  • AVX-512BW optimizations of function AlphaBlending2x.
  • AVX2 optimizations of function BgrToBayer.
  • AVX2 optimizations of function BgraToBayer.
  • Base implementation, AMX-BF16 optimizations of class SynetConvolution16bNhwcGemmV2.
  • NEON optimizations of function BgrToHsl.
  • NEON optimizations of function BgrToLab.
  • NEON optimizations of function GrayToY.
  • NEON optimizations of function YToGray.
  • NEON optimizations of function Yuv444pToRgbaV2.
  • NEON optimizations of function SynetTiledScale2D32f.
  • NEON optimizations of function SynetNormalizeLayerForward.
  • NEON optimizations of function SynetNormalizeLayerForwardV2.
  • NEON optimizations of function SynetNormalizeLayerForwardV3.
  • NEON optimizations of function SynetNormalizeLayerForwardV4.
  • NEON optimizations of function SynetAdd8i.
  • NEON optimizations of function SynetConvert8uTo32f.
  • NEON optimizations of function SynetDequantizeLinear.
  • NEON optimizations of function SynetQuantizeLinear.
  • NEON optimizations of class SynetQuantizedAdd.
  • NEON optimizations of function SynetQuantizedConcatLayerForward.
  • NEON optimizations of function SynetQuantizedPreluLayerForward.
  • NEON optimizations of function SynetQuantizedScaleLayerForward.
  • NEON optimizations of function SynetQuantizedShuffleLayerForward.
  • NEON optimizations of class SynetScale8i.
  • SSE4.1 optimizations of class SynetScale16b.
  • Support of SVE extension (ARM/ARM64 platform).
  • SVE optimizations of function AbsDifference.
  • SimdCpuInfoSveSize member in SimdCpuInfoType enumeration.
  • SVE optimizations of function OperationBinary8u.
  • SVE optimizations of function OperationBinary16i.
  • SVE optimizations of function AbsGradientSaturatedSum.
  • SVE optimizations of function DeinterleaveUv.
  • SVE optimizations of function InterleaveUv.
  • SVE optimizations of function BgrToRgb.
Improving
  • SSE4.1, AVX2, AVX-512BW optimizations of class SynetScale16b.
Bug fixing
  • Error in NEON optimization of function TransformImage (BGR, T0 transform).
  • Error in AVX-512BW optimizations of function SynetSoftmax16b (MSVS, Win32).
  • Possible aligned load of unaligned memory in AVX2 optimizations of function AbsDifferenceSums3 (Windows 7, x64, gcc 8.1.0).
  • Possible aligned store in file SimdExtract.h for SSE4.1, AVX2, AVX-512BW.
  • Wrong assert condition in function Simd::Uyvy422ToBgr.
  • Wrong assert condition in function Simd::Uyvy422ToYuv420p.
  • Wrong assert condition in function Simd::Yuv420pToUyvy422.
  • Error in function Base::CpuCacheSize on ARM64 platform.
  • Error in NEON optimizations of function DescrInt::Encode32f.
  • Error in SSE4.1, AVX2, AVX-512BW, NEON optimizations of function BgraToRgba (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).
  • Error in SSE4.1, AVX2, NEON optimizations of function RgbaToGray (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).
  • Error in SSE4.1, AVX2, NEON optimizations of function BgrToRgb (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).
  • Error in SSE4.1, NEON optimizations of function Yuv444pToBgrV2 (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).
  • Error in SSE4.1, NEON optimizations of function Yuv444pToBgraV2 (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).
  • Error in SSE4.1, NEON optimizations of function Yuv444pToRgbV2 (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).
  • Error in SSE4.1, NEON optimizations of function Yuv444pToRgbaV2 (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).
  • Error in SSE4.1, AVX2, NEON optimizations of function BgraToBgr (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).
  • Error in SSE4.1, AVX2, NEON optimizations of function BgraToRgb (small width, internal call from ImageSaveToMemory, ImageLoadFromMemory).

Test framework

New features
  • Github action 'Test Python' step in build_and_test_gcc_new in cmake.yml.
  • Tests for verifying functionality of function SynetPoolingMax16b.
Bug fixing
  • Error in Python test ShiftDetectorFunctionsTest.
  • Error in Python test ShiftDetectorClassTest.
  • Error in test SquaredDifferenceSum32f.
  • Error in test NeuralAddConvolution2x2ForwardAutoTest.
  • Error in parsing of 'testThreads' command line options.
  • Error in performance report (Dupling of the same columns).

Infrastructure

New features
  • Job build_and_test_arm64 in Github actions script for CMake.
  • test.yml github action to test dev branch

Documentation

Bug fixing
  • Syntax and lexical errors in description of Python wrapper.
  • Wrong description of build process for Microsoft Visual Studio.
Assets 2
Loading