[Kernel][Perf] Add Hopper (SM90) tuned config for batch-invariant persistent matmul - #53446

Open

yuvalluria wants to merge 7 commits into

vllm-project:mainfrom

yuvalluria:add-hopper-matmul-config

Commits on Sep 6, 2026

[Kernel][Perf] Tune Hopper (SM90) batch-invariant matmul configs via H100 NVL sweep

yuvalluria
and
claude
committed
ffea982
perf(hopper): remove benchmark methodology comment from hopper configs

yuvalluria
and
claude
committed
3a72715
perf(hopper): scope NVL configs to H100 NVL only via GPU name check

yuvalluria
and
claude
committed
b213ad3

Commits on Sep 9, 2026

Commits on Sep 16, 2026