Make the public gemm path work for the flavors it claims - #558

Merged

YangXu1990uiuc merged 4 commits into

NVIDIA:developfrom

YangXu1990uiuc:yanxu/gemm-reduction-public-path

Aug 13, 2026

Commits on Aug 12, 2026

Make the public gemm path work for the flavors it claims
YangXu1990uiuc
committed
f75e208
Seed a reduction output through the driver, on the kernel's stream
YangXu1990uiuc
committed
952bb20
Pack a reduction identity as the output's dtype, and seed on the MoE stream

YangXu1990uiuc
and
claude
committed
605e443

Commits on Aug 13, 2026

Seed a padded reduction output through the driver too

YangXu1990uiuc
and
claude
committed
99c4442