Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Feb 16, 2024

• Refactor `vunary` tests for `fp32` and `fp16` using generic functions templated on the kernel param types, and functions for the reference and tolerance values.

  Show description for 86bea1c
  

  gonnet

  authored and

  xnnpack-bot

  committed
  86bea1c

  Copy full SHA for 86bea1c

• Fix static slice test

  Show description for 768f49c
  

  Linkun Chen

  authored and

  xnnpack-bot

  committed
  768f49c

  Copy full SHA for 768f49c

Commits on Feb 15, 2024

• Update Slice to accept dynamic dimension

  Show description for d8f6274
  

  Linkun Chen

  authored and

  xnnpack-bot

  committed
  d8f6274

  Copy full SHA for d8f6274

• Reword the logging from xnn_reshape_external_value() to be more precise

  Show description for 39e6516
  

  Linkun Chen

  authored and

  xnnpack-bot

  committed
  39e6516

  Copy full SHA for 39e6516

• Disable avxvnni for 'MOBILE'

  Show description for cc4f729
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  cc4f729

  Copy full SHA for cc4f729

Commits on Feb 14, 2024

• QD8 F16 AVX2 GEMM template, test and benchmark

  Show description for 76cf722
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  76cf722

  Copy full SHA for 76cf722

Commits on Feb 13, 2024

• QD8_F16 AVXVNNI GEMM microkernels

  Show description for aa4499f
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  aa4499f

  Copy full SHA for aa4499f

• Fix static-reshape test

  Show description for 525740b
  

  Linkun Chen

  authored and

  xnnpack-bot

  committed
  525740b

  Copy full SHA for 525740b

Commits on Feb 12, 2024

• Update Reshape to accept dynamic dimension

  Show description for c06211a
  

  Linkun Chen

  authored and

  xnnpack-bot

  committed
  c06211a

  Copy full SHA for c06211a

Commits on Feb 11, 2024

• QD8 AVX2 GEMM/IGEMM microkernels support masked store on AVX512VL (AVX10)

  Show description for d5d373f
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  d5d373f

  Copy full SHA for d5d373f

Commits on Feb 9, 2024

• Create the `reshape-2d` subgraph node.

  Show description for 6db74a6
  

  gonnet

  authored and

  xnnpack-bot

  committed
  6db74a6

  Copy full SHA for 6db74a6

• Fix xnn_define_static_mean comment

  Show description for 2c58fe5
  

  Linkun Chen

  authored and

  xnnpack-bot

  committed
  2c58fe5

  Copy full SHA for 2c58fe5

• Disable RVV for config=android_riscv64

  Show description for 086b3ab
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  086b3ab

  Copy full SHA for 086b3ab

Commits on Feb 8, 2024

• Add xnn_reshape_runtime to trigger dynamic shape propagation

  Show description for b734670
  

  alankelly

  authored and

  xnnpack-bot

  committed
  b734670

  Copy full SHA for b734670

• Reshape output of static slice

  Show description for 24ea31c
  

  alankelly

  authored and

  xnnpack-bot

  committed
  24ea31c

  Copy full SHA for 24ea31c

• Add reshape output for static-mean

  Show description for 23f386b
  

  alankelly

  authored and

  xnnpack-bot

  committed
  23f386b

  Copy full SHA for 23f386b

• Only reallocate if new workspace size is greater than old. Not the opposite. That would just be silly.

  Show description for c0fcf0f
  

  alankelly

  authored and

  xnnpack-bot

  committed
  c0fcf0f

  Copy full SHA for c0fcf0f

• Set output num dims correctly for unary and binary elementwise ops

  Show description for 26e30ba
  

  alankelly

  authored and

  xnnpack-bot

  committed
  26e30ba

  Copy full SHA for 26e30ba

• Move non-constant shape parameters from create to reshape for global average pooling and global sum pooling

  Show description for ea9e1c6
  

  alankelly

  authored and

  xnnpack-bot

  committed
  ea9e1c6

  Copy full SHA for ea9e1c6

Commits on Feb 7, 2024

• Fix enable for QD8_F16_QD8W GEMM/IGEMM avx512 microkernel

  Show description for 784853f
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  784853f

  Copy full SHA for 784853f

Commits on Feb 6, 2024

• Enable QS8/QU8 GEMM/IGEMM 4x8C8 AVX512VL microkernels

  Show description for 6ef9832
  

  fbarchard

  authored and

  gonnet

  committed
  6ef9832

  Copy full SHA for 6ef9832

• Add -mf16c build flags to avx512 configs to allow FP16 output

  Show description for 0f233d4
  

  fbarchard

  authored and

  gonnet

  committed
  0f233d4

  Copy full SHA for 0f233d4

• Merge pull request #5974 from mcr229:reshape_softmax

  Show description for 49fff8e
  gonnet

  committed
  49fff8e

  Copy full SHA for 49fff8e

• Copybara import of the project:

  Show description for 0a8d146
  

  digantdesai

  authored and

  gonnet

  committed
  0a8d146

  Copy full SHA for 0a8d146

• Merge pull request #5975 from mcr229:reshape_prelu

  Show description for 8e54b63
  gonnet

  committed
  8e54b63

  Copy full SHA for 8e54b63

• Enable QD8_F16_QC4W/QC8W GEMM/IGEMM 5x8C8 AVX512VL microkernels

  Show description for 44f3431
  

  fbarchard

  authored and

  gonnet

  committed
  44f3431

  Copy full SHA for 44f3431

• Reshape sparse fully connected output.

  Show description for db94350
  

  alankelly

  authored and

  gonnet

  committed
  db94350

  Copy full SHA for db94350

• Convolution returns an error for invalid kernel zero point

  Show description for 9061d2b
  

  alankelly

  authored and

  gonnet

  committed
  9061d2b

  Copy full SHA for 9061d2b

• Reshape depthwise convolution 2d output

  Show description for 533625d
  

  alankelly

  authored and

  gonnet

  committed
  533625d

  Copy full SHA for 533625d

• Reshape deconvolution 2d output

  Show description for 211a55d
  

  alankelly

  authored and

  gonnet

  committed
  211a55d

  Copy full SHA for 211a55d

• Reshape argmax pooling 2d output

  Show description for f195077
  

  alankelly

  authored and

  gonnet

  committed
  f195077

  Copy full SHA for f195077

• Move channels, input_stride & output_stride from create to reshape for argmax pooling 2d

  Show description for 6b1e9e3
  

  alankelly

  authored and

  gonnet

  committed
  6b1e9e3

  Copy full SHA for 6b1e9e3

• Reshape static resize bilinear output

  Show description for 2ba43a4
  

  alankelly

  authored and

  gonnet

  committed
  2ba43a4

  Copy full SHA for 2ba43a4

• Move all constant parameters to create and non-constant ones to reshape for static resize bilinear

  Show description for 73595bf
  

  alankelly

  authored and

  gonnet

  committed
  73595bf

  Copy full SHA for 73595bf

• AVX512VL build of AVX2 MRx8C8 GEMM microkernels

  Show description for f072e81
  

  fbarchard

  authored and

  gonnet

  committed
  f072e81

  Copy full SHA for f072e81

Pagination

Previous

Next