Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 31, 2025

• Fix KBLOCK size for qc2w GEMM scalar and neondot microkernel tests

  Show description for 4536794
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  4536794

  Copy full SHA for 4536794

• Fix KBLOCK size for neondot qc2w microkernel tests

  Show description for cafdb1e
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  cafdb1e

  Copy full SHA for cafdb1e

• Benchmarks for Apple macos / ios detect cpu frequency

  Show description for 4a0e2db
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  4a0e2db

  Copy full SHA for 4a0e2db

Commits on Dec 26, 2025

• Use tags to filter out tests for YNNPACK

  Show description for 25b42df
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  25b42df

  Copy full SHA for 25b42df

• Fix nc param to gemm/igemm headers

  Show description for bf97a38
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  bf97a38

  Copy full SHA for bf97a38

• QC2W NEONDOT defer kernel_zero_point compute to after loop

  Show description for 742fd64
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  742fd64

  Copy full SHA for 742fd64

• Remove unnecessary assert

  Show description for 9b2d3e8
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  9b2d3e8

  Copy full SHA for 9b2d3e8

• Fix `tile_k > 1` case for rewriting `transpose_a(stencil_copy(x))` -> `stencil_copy(transpose_a(x))`

  Show description for 2fff7b8
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  2fff7b8

  Copy full SHA for 2fff7b8

Commits on Dec 23, 2025

• Try to store output of stencil_copy at the innermost loop level.

  Show description for 4f8026d
  

  vksnk

  authored and

  xnnpack-bot

  committed
  4f8026d

  Copy full SHA for 4f8026d

• Fix stencil_copy bounds if multiple stencils use the same original dimension

  Show description for e1e7194
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  e1e7194

  Copy full SHA for e1e7194

• Define Hexagon type for intrinsics compatible with clang

  Show description for 911c20f
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  911c20f

  Copy full SHA for 911c20f

• 1. Adjust AVX reductions to match AVX512

  Show description for 20821fb
  

  alexander-shaposhnikov

  authored and

  xnnpack-bot

  committed
  20821fb

  Copy full SHA for 20821fb

• Fix stencil output extent calculation.

  Show description for da578ff
  

  vksnk

  authored and

  xnnpack-bot

  committed
  da578ff

  Copy full SHA for da578ff

Commits on Dec 22, 2025

• Remove unused Hexagon intrinsics from intrinsics-polyfill.h

  Show description for 549d5d2
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  549d5d2

  Copy full SHA for 549d5d2

• Include bias in initializer of dot for f32.

  Show description for c299a16
  

  mariecwhite

  authored and

  xnnpack-bot

  committed
  c299a16

  Copy full SHA for c299a16

Commits on Dec 20, 2025

• Fix bug in avx512 int32 helpers

  Show description for 7d0408c
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  7d0408c

  Copy full SHA for 7d0408c

• Don't use `EnumerateIndices` unnecessariily

  Show description for fc1d7b1
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  fc1d7b1

  Copy full SHA for fc1d7b1

Commits on Dec 19, 2025

• Change simd::vec to be recursively defined

  Show description for 82d56fb
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  82d56fb

  Copy full SHA for 82d56fb

• Fixed full byte requirement in 2-bit Fully Connected test for qs8 version.

  Show description for f52fbff
  

  Aelphy

  authored and

  xnnpack-bot

  committed
  f52fbff

  Copy full SHA for f52fbff

• Don't try to test int2 fully connected with YNNPACK

  Show description for ecdf584
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  ecdf584

  Copy full SHA for ecdf584

Commits on Dec 18, 2025

• Added qs8_qc2w FullyConnected.

  Show description for 91bc5c2
  

  Aelphy

  authored and

  xnnpack-bot

  committed
  91bc5c2

  Copy full SHA for 91bc5c2

• Support tile_k > 1 in transpose `stencil_copy` rewrite.

  Show description for e31a359
  

  mariecwhite

  authored and

  xnnpack-bot

  committed
  e31a359

  Copy full SHA for e31a359

• Fix `multi_vec`'s partial `load`/`store`

  Show description for 183297d
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  183297d

  Copy full SHA for 183297d

• Rewrite transpose_a(stencil_copy(x)) to stencil_copy(transpose_a(x))

  Show description for ff23b55
  xnnpack-bot

  committed
  ff23b55

  Copy full SHA for ff23b55

• Minor fixes to stencil_copy

  Show description for 0723f38
  xnnpack-bot

  committed
  0723f38

  Copy full SHA for 0723f38

• Enable transpose_a kernels for num_k_dims > 1

  Show description for 944d44d
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  944d44d

  Copy full SHA for 944d44d

Commits on Dec 17, 2025

• Tweak K1 = 1 reduction case

  Show description for 51dd4c8
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  51dd4c8

  Copy full SHA for 51dd4c8

• Merge pull request #9316 from salmanmkc:upgrade-github-actions-node24

  Show description for 996234f
  xnnpack-bot

  committed
  996234f

  Copy full SHA for 996234f

• Pass down the original kernel and bias buffers when creating a convolution operation.

  Show description for e317c78
  

  qukhan

  authored and

  xnnpack-bot

  committed
  e317c78

  Copy full SHA for e317c78

• Improved k1=1 reductions. Most of the cases have significant speed-up except scalar kernels, sum bf16_f32 for f16c and sse2, and x8_s32 for avx2.

  Show description for 103c631
  

  Aelphy

  authored and

  xnnpack-bot

  committed
  103c631

  Copy full SHA for 103c631

• Allowed not to pass channel wise zero point to accommodate asymmetric (zero point is assumed 0) 2-bit case.

  Show description for adf70ec
  

  Aelphy

  authored and

  xnnpack-bot

  committed
  adf70ec

  Copy full SHA for adf70ec

• Enable weight caching for quantized depthwise convolutions.

  Show description for 2c1a512
  

  qukhan

  authored and

  xnnpack-bot

  committed
  2c1a512

  Copy full SHA for 2c1a512

• Rewrite SIMD tests to make a new test for each architecture

  Show description for 4574c4d
  

  dsharletg

  authored and

  xnnpack-bot

  committed
  4574c4d

  Copy full SHA for 4574c4d

Commits on Dec 16, 2025

• Fixed n=0 case that left uninitialized vectors.

  Show description for 30cd310
  

  Aelphy

  authored and

  xnnpack-bot

  committed
  30cd310

  Copy full SHA for 30cd310

• Upgrade GitHub Actions for Node 24 compatibility

  salmanmkc

  committed
  7926389

  Copy full SHA for 7926389

Pagination

Previous

Next