Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Apr 10, 2021

• XOP versions of QS8 DWCONV MUL32 microkernels

  Show description for 3fd4e27
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  3fd4e27

  Copy full SHA for 3fd4e27

• Optimize QS8 GEMM/IGEMM microkernel selection for AVX

  Show description for 4181f94
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  4181f94

  Copy full SHA for 4181f94

• 2x4c2/3x4c2 microkernels for SSE2/SSSE3/SSE4.1/AVX/XOP

  Show description for 2e42787
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  2e42787

  Copy full SHA for 2e42787

Commits on Apr 8, 2021

• QS8 E2E GEMM benchmarks for C2 AVX microkernels

  Show description for be3d8fd
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  be3d8fd

  Copy full SHA for be3d8fd

• Add AVX microkernels to QS8 E2E GEMM benchmark

  Show description for d5cc508
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  d5cc508

  Copy full SHA for d5cc508

Commits on Apr 7, 2021

• Make xnn_initialize thread-safe

  Show description for 496389f
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  496389f

  Copy full SHA for 496389f

• QS8 move loads to end of loop, 1 every 2 neon instructions.

  Show description for e696c3f
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  e696c3f

  Copy full SHA for e696c3f

• Polyfill _mm_loadu_si32 in MUL32 QS8 DWCONV SSE4.1/AVX microkernels

  Show description for 60fc613
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  60fc613

  Copy full SHA for 60fc613

Commits on Apr 3, 2021

• MUL32 versions of SSE4.1 & AVX QS8 DWCONV microkernels

  Show description for 07feec8
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  07feec8

  Copy full SHA for 07feec8

• AVX versions of QS8 DWCONV microkernels

  Show description for fa0ab85
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  fa0ab85

  Copy full SHA for fa0ab85

Commits on Apr 2, 2021

• AVX versions of QS8 VADD/VADDC microkernels

  Show description for e9c4b96
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  e9c4b96

  Copy full SHA for e9c4b96

• AVX versions of QS8 GEMM/IGEMM microkernels

  Show description for a3c1633
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  a3c1633

  Copy full SHA for a3c1633

• Refactor code-generation templates for XOP microkernels

  Show description for b8ad46a
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  b8ad46a

  Copy full SHA for b8ad46a

• QS8 C8 GEMM/IGEMM use load a/b last technique for Cortex A75 performance.

  Show description for ae5082e
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  ae5082e

  Copy full SHA for ae5082e

• Include XOP headers in clang-cl compatible way. Fix google#1382.

  Show description for c409471
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  c409471

  Copy full SHA for c409471

Commits on Apr 1, 2021

• Fully Connected operator for QS8 datatype

  Show description for d23cb6e
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  d23cb6e

  Copy full SHA for d23cb6e

Commits on Mar 31, 2021

• Implement bilinear upsampling for SSE architecture

  Show description for b3ffd58
  

  ablavatski

  authored and

  xnnpack-bot

  committed
  b3ffd58

  Copy full SHA for b3ffd58

• Support quantized inference in Subgraph API with xnn_enable_qs8=true

  Show description for 1f5099e
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  1f5099e

  Copy full SHA for 1f5099e

• Rename xnnpack_enable_memopt Bazel option to xnn_enable_memopt

  Show description for 03f4621
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  03f4621

  Copy full SHA for 03f4621

• Bazel flag --xnn_enable_qs8=true to include QS8 operators in :xnnpack_for_tflite

  Show description for b939cdb
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  b939cdb

  Copy full SHA for b939cdb

• Remove unused :xnnpack_f32 target

  Show description for 96d95e0
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  96d95e0

  Copy full SHA for 96d95e0

Commits on Mar 30, 2021

• Remove Android NDK r17 support

  Show description for d09ca26
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  d09ca26

  Copy full SHA for d09ca26

• Rename TFLite- and TF.js- optimized targets

  Show description for f0cb70a
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  f0cb70a

  Copy full SHA for f0cb70a

• Validate static tensors in Subgraph API

  Show description for 09c0591
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  09c0591

  Copy full SHA for 09c0591

• Clarify documentation of xnn_define_quantized_tensor_value

  Show description for 3075719
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  3075719

  Copy full SHA for 3075719

• Extend Subgraph API to support quantized tensors

  Show description for 43ebc05
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  43ebc05

  Copy full SHA for 43ebc05

Commits on Mar 29, 2021

• Validate tensor data types in Subgraph API

  Show description for ccd3a1d
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  ccd3a1d

  Copy full SHA for ccd3a1d

• Update emscripten config settings to conform to the official emscripten toolchain.

  Show description for 3bfbdaf
  xnnpack-bot

  committed
  3bfbdaf

  Copy full SHA for 3bfbdaf

Commits on Mar 25, 2021

• Merge pull request google#1367 from JerryShih:upgrade-fp16

  Show description for f7baa13
  xnnpack-bot

  committed
  f7baa13

  Copy full SHA for f7baa13

• QS8 1X8C8 IGEMM microkernel

  Show description for 6e35de5
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  6e35de5

  Copy full SHA for 6e35de5

• QS8 1X8C8 GEMM microkernel

  Show description for b876263
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  b876263

  Copy full SHA for b876263

• Update FP16 dependency.

  Show description for accc49a
  JerryShih

  committed
  accc49a

  Copy full SHA for accc49a

Commits on Mar 24, 2021

• Validate tensor types in Subgraph API

  Show description for a0f9bdc
  

  Maratyszcza

  authored and

  xnnpack-bot

  committed
  a0f9bdc

  Copy full SHA for a0f9bdc

Commits on Mar 22, 2021

• Limit range of test values for f16 binary minmax ops.

  Show description for 967712d
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  967712d

  Copy full SHA for 967712d

• MOV 16b instead of 4s for GCC compatability. Fix google#1360

  Show description for 2c525e5
  

  fbarchard

  authored and

  xnnpack-bot

  committed
  2c525e5

  Copy full SHA for 2c525e5

Pagination

Previous

Next