[MLAS] Add an AVX-512 fused kernel for LinearAttention - #31674

Merged

mirounga merged 8 commits into

microsoft:mainfrom

mirounga:rs_buf72

Aug 12, 2026

Commits on Jul 27, 2026

Fix LinearAttention output shape inference for standard GQA (q > kv)
mirounga
committed
24f9b88

Commits on Jul 28, 2026

[Webgpu] Fix LinearAttention output shape for standard GQA (q > kv)
mirounga
committed
a6acfff

Commits on Aug 4, 2026

[MLAS] Move the LinearAttention recurrence into MLAS
mirounga
committed
188b06a
[MLAS] Add an AVX-512 fused kernel for LinearAttention
mirounga
committed
2482e83

Commits on Aug 5, 2026

Merge remote-tracking branch 'origin/main' into rs_buf72
mirounga
committed
3189073

Commits on Aug 6, 2026

[MLAS] Address review feedback on the LinearAttention AVX-512 kernel
mirounga
committed
b351482

Commits on Aug 10, 2026

Merge branch 'main' into rs_buf72
mirounga
committed
72a141b
[MLAS] Second round of review feedback on LinearAttention
mirounga
committed
39de2a4