- a463285 4x16 QS8 GEMM use 4 less registers, avoiding push/pop. by Frank Barchard · 3 years, 11 months ago
- ab8c4c8 Add basic unit tests for graph rewriting procedure by XNNPACK Team · 3 years, 11 months ago
- 46aadda Enable 1x16 QS8 assembly GEMM for Neon dotproduct by Frank Barchard · 3 years, 11 months ago
- 59df88b 4x16 QS8 GEMM defer params by Frank Barchard · 3 years, 11 months ago
- f1fd89e 1x16 QS8 GEMM AARCH64 assembly microkernel using dot product. by Frank Barchard · 3 years, 11 months ago
- a5237a5 Rename 4x16 GEMM dot product microkernel file name to allow for future variations. by Frank Barchard · 3 years, 11 months ago
- bc0c729 Enable GEMM 4x16 QS8 using dot product microkernels. by Frank Barchard · 3 years, 11 months ago
- d2750b0 Limit testing range for FP16 global-average-pooling to 1.0e-3f to 1 to avoid by Frank Barchard · 3 years, 11 months ago
- 31bb45b 4x16 QS8 GEMM AARCH64 assembly microkernel using dot product. by Frank Barchard · 3 years, 11 months ago
- a117ce7 Fix graph rewriting procedure for graph without NCHW2NHWC compatible nodes/ops by XNNPACK Team · 3 years, 11 months ago
- 3b26206 Renumber labels in assembly sequentially by Frank Barchard · 3 years, 11 months ago
- aed3278 extend hardswish test range to -4 to 4 for better coverage by Frank Barchard · 3 years, 11 months ago
- 87318a2 Refactor scalar and NEON Sigmoid microkernels by Marat Dukhan · 3 years, 11 months ago
- fe017b6 FP16 binary testers use 0.01 as min random value to avoid rdiv producing inf by Frank Barchard · 3 years, 11 months ago
- 6edeade Refactor SSE2/SSE4/AVX2/AVX512F/WAsm SIMD Sigmoid microkernels by Marat Dukhan · 3 years, 11 months ago
- 66ccf64 Rename QS8 generator templates by Marat Dukhan · 4 years ago
- c3001e1 Refactor Sigmoid evaluation stubs by Marat Dukhan · 4 years ago
- 9501298 Refactor SSE2/AVX/AVX2/AVX512 Sigmoid evaluation stubs by Marat Dukhan · 4 years ago
- 04fd73a Reoptimize LUT-based ExpMinus evaluation stubs by Marat Dukhan · 4 years ago
- 1f256fc Reoptimize LUT-based Sigmoid evaluation stubs by Marat Dukhan · 4 years ago
- 3cfe8f5 Refactor SSE2 Sigmoid evaluation stub by Marat Dukhan · 4 years ago
- a48848f 4x8, 6x8 and 8x16 Neon dot product GEMM microkernels by Frank Barchard · 4 years ago
- 2e4727b AVX versions of Sigmoid evaluation stubs by Marat Dukhan · 4 years ago
- d9ca7e6 AVX512F versions of Sigmoid microkernel by Marat Dukhan · 4 years ago
- ae24dfc Fix CMake build by Marat Dukhan · 4 years ago
- 6a34c5f Refactor AVX512F Sigmoid evaluation stubs by Marat Dukhan · 4 years ago
- 2fa1745 6x16 QS8 GEMM for Neon dot product by Frank Barchard · 4 years ago
- 80a8ac5 AVX512F LUT-based with SCALEF Exp evaluation stubs by Marat Dukhan · 4 years ago
- 9490613 Reoptimize LUT-based NEON Sigmoid evaluation stubs by Marat Dukhan · 4 years ago
- b32b018 Reoptimize LUT-based SSE2 Exp evaluation stub by Marat Dukhan · 4 years ago
- 272139e Reoptimize LUT-based NEONFMA Exp/ExpMinus evaluation stubs by Marat Dukhan · 4 years ago
- d08abcb AVX Exp evaluation stub by Marat Dukhan · 4 years ago
- 8e17294 LUT-based SSE2 Exp evaluation stub by Marat Dukhan · 4 years ago
- 47ffc41 AVX512F Sigmoid evaluation stubs by Marat Dukhan · 4 years ago
- 5876c83 Reoptimize SSE4 LUT-based Sigmoid microkernels by Marat Dukhan · 4 years ago
- 6dd7136 Use LUT-based Sigmoid microkernels on SSE2/SSE4 systems by Marat Dukhan · 4 years ago
- d243c1a LUT-based SSE Sigmoid microkernels by Marat Dukhan · 4 years ago
- f8475d6 SSE2 version of LUT64 Sigmoid evaluation stub by Marat Dukhan · 4 years ago
- 2b9d29b FP16 unittest tests expect less error values near zero. by Frank Barchard · 4 years ago
- 6844730 Fix segfault due to out-of-bounds reads of internally allocated values by Marat Dukhan · 4 years ago
- fc2844e FP16 hardswish operator benchmark by Frank Barchard · 4 years ago
- 7d2c1f2 FP16 tests allow for results that are near zero. by Frank Barchard · 4 years ago
- 66f3ccd End-to-end FP16 MobileNet v3 benchmarks by Marat Dukhan · 4 years ago
- d04e2dd Alphabetically order FP16 operators by Marat Dukhan · 4 years ago
- a96948e FP16 HardSwish operator by Frank Barchard · 4 years ago
- ef4ce31 Remove trailing whitespace by Marat Dukhan · 4 years ago
- d4c8303 Enable NEON DOT QS8 [I]GEMM microkernels on ARM64 by Marat Dukhan · 4 years ago
- 0ea6a77 FP16 binary multiply operator by Frank Barchard · 4 years ago
- e6dc0b6 AVX2 versions of QS8 VADD[C] microkernels by Marat Dukhan · 4 years ago
- bb9225e SSE4.1 and XOP versions of MUL32 VADD[C] microkernels by Marat Dukhan · 4 years ago
- 2ffc5e6 AVX512 versions of QS8 DWCONV microkernels by Marat Dukhan · 4 years ago
- 70a9618 End-to-end QS8 MobileNet v2 benchmark by Marat Dukhan · 4 years ago
- ff20948 QS8 version of ND ADD operator by Marat Dukhan · 4 years ago
- 5df27f8 WAsm SIMD versions of QS8 VADD[C] microkernels by Marat Dukhan · 4 years ago
- ba7b279 NEON variants of QS8 VADD[C] microkernels by Marat Dukhan · 4 years ago
- 95caee5 Test zero_point and scale parameters in QS8 VADD[C] microkernels by Marat Dukhan · 4 years ago
- 9c7308f vbinary microkernels unrolled to x8 for scalar and web assembly and x16 web assembly simd by Frank Barchard · 4 years ago
- a635a29 WASM and WASMSIMD test ifdef improved. by Frank Barchard · 4 years ago
- 37297a6 F32-RELU unrolled more for improved performance on Web Assembly by Frank Barchard · 4 years ago
- 4422f10 FP16 hswish fix cast fix for improved GCC compatability. by Frank Barchard · 4 years ago
- 12c5777 Optimization: 2x partial unroll to load 8 contiguous bytes. by Benoit Jacob · 4 years ago
- 919328c Improve compatibility with -fno-lax-vector-conversions by Marat Dukhan · 4 years ago
- 7359463 Fix incompatibilities with ARM GCC by Marat Dukhan · 4 years ago
- c2146cc Added support for specialized batch_size == 1 convolutions. by Fabio Riccardi · 4 years ago
- 62d752e Function macro for WASM assembly by Frank Barchard · 4 years ago
- 9bd894b Relu wasm assembly use if statements instead of block for readability. by Frank Barchard · 4 years ago
- bcedc08 Move WAsm assembly source into its own list of sources by Frank Barchard · 4 years ago
- 9f5ccb3 f32 relu microkernel WASM assembly x2 and x4 unrolled loops by Frank Barchard · 4 years ago
- 6fe0b53 add wasm32 to common build scripts to define macros for unittests and benchmarks by Frank Barchard · 4 years ago
- a05487f Add xnn_qs8_igemm_minmax_ukernel_${MR}x${NR}c4__neondot (ARMv8.2+dotprod). by Benoit Jacob · 4 years ago
- 799ac75 Use more verbose compiler flags for NEON DOT ISA on AArch32 by Marat Dukhan · 4 years ago
- a964473 Add xnn_qs8_gemm_minmax_ukernel_${MR}x${NR}c4__neondot (ARMv8.2+dotprod). by Benoit Jacob · 4 years ago
- 0af63ab Include polyfills for intrinsics in QS8 AVX512 GEMM/IGEMM microkernels by Marat Dukhan · 4 years ago
- d91624e f32 relu microkernel WASM assembly optimized implementation by Frank Barchard · 4 years ago
- 2a18f7e Polyfill _mm512_set_epi8 for older gcc by Marat Dukhan · 4 years ago
- 0270d9f QS8 VADDC microkernels in SSE2 and SSE4.1 implementations by Marat Dukhan · 4 years ago
- 8432486 Fix CMake build by Marat Dukhan · 4 years ago
- f28cddf Initialize QS8 microkernels in ARM/ARM64 builds by Marat Dukhan · 4 years ago
- 281262d NEON variant of QS8 GAVGPOOL microkernel by Marat Dukhan · 4 years ago
- 023bcf9 NEON variant of QS8 DWCONV microkernels by Marat Dukhan · 4 years ago
- d9f3ad4 QS8 ADD microkernels in SSE2 and SSE4.1 implementations by Marat Dukhan · 4 years ago
- bb00b1d AVX512 variants of QS8 GEMM and IGEMM microkernels by Marat Dukhan · 4 years ago
- facecc5 Fix fusion of Constant Pad with asymmetric padding by Marat Dukhan · 4 years ago
- f124e88 Polyfill _mm_loadu_si32 and _mm_storeu_si32 intrinsics by Marat Dukhan · 4 years ago
- 4de076d Fix xnn_update_qs8_avgpool_params on WAsm SIMD by Marat Dukhan · 4 years, 1 month ago
- 674778d Add binary op microkernels with RELU activation by Frank Barchard · 4 years, 1 month ago
- 75215d8 Enable XOP versions of GEMM/IGEMM microkernels by Marat Dukhan · 4 years, 1 month ago
- c15aa4e Remove XOP variants of QS8 DWCONV by Marat Dukhan · 4 years, 1 month ago
- 5b3af47 Re-generate QS8 and QU8 microkernels from templates by Marat Dukhan · 4 years, 1 month ago
- 453f4b8 Fix CMake build by Marat Dukhan · 4 years, 1 month ago
- c9c320e CMake build fix by Frank Barchard · 4 years, 1 month ago
- b33fc0e Add xnn_q{u,s}8_gemm_minmax_ukernel_MRxNRc4__scalar by Benoit Jacob · 4 years, 1 month ago
- 9e0b539 QS8 variant of NWC Global Average Pooling operator by Marat Dukhan · 4 years, 1 month ago
- 4013552 AVX2 versions of QS8 DWCONV microkernels using 16-bit multiplication by Marat Dukhan · 4 years, 1 month ago
- 54e95a0 Unify implementation of NWC Global Average Pooling across data types by Marat Dukhan · 4 years, 1 month ago
- b5e3d17 Multipass QS8 GAVGPOOL microkernel in WAsm SIMD implementation by Marat Dukhan · 4 years, 1 month ago
- ef45180 Unipass QS8 GAVGPOOL microkernel in WAsm SIMD implementation by Marat Dukhan · 4 years, 1 month ago
- 159688f Multipass QS8 GAVGPOOL microkernels in SSE2/SSSE3/SSE4.1 implementations by Marat Dukhan · 4 years, 1 month ago
- 4ed53f4 Unipass QS8 GAVGPOOL microkernels in SSE2/SSSE3/SSE4.1 implementations by Marat Dukhan · 4 years, 1 month ago
- 07e5040 Initialize QS8 microkernels for WAsm SIMD by Marat Dukhan · 4 years, 1 month ago