[ExecuTorch][WebGPU] Add 256-thread "steel" q4gsw prefill GEMM - #20815

Merged

JCNTH merged 11 commits into

gh/JCNTH/3/origfrom

gh/JCNTH/4/orig

Jul 9, 2026

Commits on Jul 9, 2026

[ExecuTorch][WebGPU] Add 256-thread "steel" q4gsw prefill GEMM
JCNTH
committed
efb7b5a
[ExecuTorch][WebGPU] Steel q4gsw prefill GEMM — test coverage
JCNTH
committed
9117995
[ExecuTorch][WebGPU] Add f16-multiply variant of the steel q4gsw prefill GEMM
JCNTH
committed
e2e9afe
[ExecuTorch][WebGPU] Test coverage for the f16-multiply steel q4gsw GEMM
JCNTH
committed
a888231
[ExecuTorch][WebGPU] Add opt-in native f16 KV cache (−280 MiB, default-OFF)
JCNTH
committed
364b335
[ExecuTorch][WebGPU] Test coverage for the f16 KV cache
JCNTH
committed
5fa85e8
[ExecuTorch][WebGPU] Reuse pool for single-op scratch (-86.7 MiB)
JCNTH
committed
43a053b
[ExecuTorch][WebGPU] Packed-word-dequant f16 steel q4gsw prefill GEMM
JCNTH
committed
4a4f0a8
[ExecuTorch][WebGPU] Test coverage for the packed-word-dequant f16 steel GEMM
JCNTH
committed
0690a89
[ExecuTorch][WebGPU] f16-accumulate (pwdqf16acc) steel q4gsw prefill GEMM
JCNTH
committed
45ea327
[ExecuTorch][WebGPU] Test coverage for the f16-accumulate steel GEMM
JCNTH
committed
32cb3c1