Commits

Branch selector

 main

User selector

All users

All time

Commit History

Commits on Jun 11, 2026

• build: dropped NONE option for ONEDNN_GPU_VENDOR knob

  vpirogov

  committed
  8cde797

  Copy full SHA for 8cde797

• doc: fixup for ONEDNN_CODE_COVERAGE supported options

  vpirogov

  committed
  ddbbccc

  Copy full SHA for ddbbccc

• build: introduced functions to handle build knobs behavior

  vpirogov

  committed
  86978e1

  Copy full SHA for 86978e1

• tests: benchdnn: fix grouped binary po no ref memory path

  mzhukova

  committed
  7f10b68

  Copy full SHA for 7f10b68

• src: gpu: intel: matmul: make shim hearders readable for gemm PO

  Show description for 602b01b
  

  h-sadia

  and

  umar456

  committed
  602b01b

  Copy full SHA for 602b01b

• doc: primitives: matmul: add note for GPU PO support

  h-sadia

  committed
  aa98993

  Copy full SHA for aa98993

• src: gpu: intel: matmul: enable PO for reference grouped gemm

  h-sadia

  committed
  cafe18c

  Copy full SHA for cafe18c

• src: gpu: intel: matmul: enable PO for ggemm (bin_mul + elt_swish)

  h-sadia

  committed
  13e6484

  Copy full SHA for 13e6484

• cpu: aarch64: jit: use hardware sp for stack management

  Show description for d2e067a
  Sqvid

  committed
  d2e067a

  Copy full SHA for d2e067a

• cpu: aarch64: binary: Avoid parallel overhead on tiny workloads.

  renato-arantes

  committed
  5ac8ab8

  Copy full SHA for 5ac8ab8

• common: gated_mlp: enable binary post-ops in gmlp

  hidefromkgb

  committed
  fc8d627

  Copy full SHA for fc8d627

Commits on Jun 10, 2026

• doc: set ONEDNN_BUILD_DOC as OFF by default

  mzhukova

  committed
  9329a26

  Copy full SHA for 9329a26

• tests: gtests: persistent_cache_api: harden the test

  echeresh

  committed
  5428256

  Copy full SHA for 5428256

• xe: compute: device_info: fix serialization

  echeresh

  committed
  d15707f

  Copy full SHA for d15707f

• common: serialization: add std::string support

  echeresh

  committed
  1dfcc28

  Copy full SHA for 1dfcc28

• benchdnn: dnnl_memory: can't return size for runtime arguments

  dzarukin

  committed
  468722f

  Copy full SHA for 468722f

• graph: utils: support matching disconnected subgraphs

  

  ElaineBao

  authored and

  TaoLv

  committed
  8ff5338

  Copy full SHA for 8ff5338

Commits on Jun 9, 2026

• cpu: x64: matmul: amx_blocking: fix int overflow for large sizes

  dzarukin

  committed
  0ecf484

  Copy full SHA for 0ecf484

• benchdnn: enable the failed summary for other modes

  dzarukin

  committed
  fdfcf52

  Copy full SHA for fdfcf52

• common: memory_desc: update the check to fit for blocked tags

  dzarukin

  committed
  ad0f5cb

  Copy full SHA for ad0f5cb

• common: memory_desc: add int4 stride sanity check

  dzarukin

  committed
  41048d5

  Copy full SHA for 41048d5

• common: move memory_desc_strides_check to memory_desc.cpp

  dzarukin

  committed
  1ece732

  Copy full SHA for 1ece732

• benchdnn: inputs: remove invalid test cases post int4 strides check

  dzarukin

  committed
  258b90c

  Copy full SHA for 258b90c

• cpu: x64: matmul: rename m_and_k_contiguous to m_and_k_adjacent

  densamoilov

  committed
  a65a1d9

  Copy full SHA for a65a1d9

• benchdnn: inputs: matmul: update regression test

  Show description for 4c54884
  densamoilov

  committed
  4c54884

  Copy full SHA for 4c54884

• cpu: x64: matmul: remove incorrect C stride adjustment

  Show description for 359fdc2
  densamoilov

  committed
  359fdc2

  Copy full SHA for 359fdc2

• cpu: x64: matmul: remove dead code

  Show description for 7b3cb6a
  densamoilov

  committed
  7b3cb6a

  Copy full SHA for 7b3cb6a

• xe: copy plan: use elements for dependency tracking

  atkassen

  committed
  180dab0

  Copy full SHA for 180dab0

• ngen: fix formatting

  atkassen

  committed
  58421c4

  Copy full SHA for 58421c4

• ngen: fixup register regioning edge case

  atkassen

  committed
  4914be4

  Copy full SHA for 4914be4

• xe3p: copy plan: fixup `dnscl` instruction dump

  atkassen

  committed
  d1f82bd

  Copy full SHA for d1f82bd

• xe: codegen: reorder: use integer paths for like types

  atkassen

  committed
  57eacd2

  Copy full SHA for 57eacd2

• xe3p: copy plan: combine shift + align into single instruction

  atkassen

  committed
  b2c721a

  Copy full SHA for b2c721a

• xe3p: copy plan: pad `shfl` operands to a valid SIMD

  atkassen

  committed
  99215de

  Copy full SHA for 99215de

• xe: copy plan: correct temporary grf check for atomic writes

  atkassen

  committed
  9b4927f

  Copy full SHA for 9b4927f