Repository navigation

[MLAS][KleidiAI] Integrate Depthwise 3x3 Conv Kernel - #29441

Open

Jonathan Clohessy (JonathanC-ARM) wants to merge 10 commits into

microsoft:mainfrom

JonathanC-ARM:jonclo01/dw_conv_integration

Open

[MLAS][KleidiAI] Integrate Depthwise 3x3 Conv Kernel#29441
Jonathan Clohessy (JonathanC-ARM) wants to merge 10 commits into
microsoft:mainfrom
JonathanC-ARM:jonclo01/dw_conv_integration

Commits on Nov 5, 2026
- Integrate DW Conv Kernel
  
  Jonathan Clohessy (JonathanC-ARM)
  authored and
  Melike Kaptan (melkap01-Arm)
  committed
  2dcfc42
- review points implemented
  Melike Kaptan (melkap01-Arm)
  committed
  2150dd9
- pre packed weights added, tests improved to cover larger channel counts, refactor cleaned
  Melike Kaptan (melkap01-Arm)
  committed
  fa579e7
- threading added into the DepthwiseConvKleidiAI
  Melike Kaptan (melkap01-Arm)
  committed
  eb628da
- co pilot review comments adressed
  Melike Kaptan (melkap01-Arm)
  committed
  5cae7e4
- codex review points on clamping addressed
  Melike Kaptan (melkap01-Arm)
  committed
  a0b21b2
- NaN bias case fallback provided
  Melike Kaptan (melkap01-Arm)
  committed
  28cd39f
- Preserve NaNs in KleidiAI depthwise convolution
  Melike Kaptan (melkap01-Arm)
  committed
  a43c69d
- lintrunner run
  Melike Kaptan (melkap01-Arm)
  committed
  32411d4
Commits on Nov 9, 2026
- Merge branch 'microsoft:main' into jonclo01/dw_conv_integration
  Melike Kaptan (melkap01-Arm)
  authored
  b46f439