[Feature Tracking] MLA-FP8 Hopper kernel #990

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Open

yzh119 opened this issue Mar 31, 2025 · 0 comments

Collaborator

yzh119 commented Mar 31, 2025 •

edited

Loading

Pipeline design

Stages

LOAD-Q
LOAD-K
GEMM-QK
SMEM-TRANSPOSE-K
GEMM-PV-LEFT
GEMM-PV-RIGHT
WRITE-O

2WG Example

WG1: LOAD-K(1,1) -> LOAD-Q(q) -> LOAD-K(1, 2) -> SMEM-TRANSPOSE-K(1, 1) -> GEMM-PV-LEFT(1, 1) -> LOAD-K(1, 3) -> SMEM-TRANSPOSE-K(1, 2) -> GEMM-PV-LEFT(1, 2) -> ...
WG2: GEMM-QK(1, 1) -> GEMM-PV-RIGHT(1, 1) -> GEMM-QK(1, 2) -> GEMM-PV-RIGHT(1, 2) -> ...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment