[AG-GEMM] Add warp specialized ag-gemm that supports KDA and MLA up projections - #41

Merged

MaoZiming merged 16 commits into

uccl-project:mainfrom

ShawnWeiChew:ag-gemm-kda-mla-pub

Sep 10, 2026

Commits on Sep 8, 2026

[ag-gemm-kda-mla] initial setup
ShawnWeiChew
committed
cca5d36
chore: revert old Makefile configurations
ShawnWeiChew
committed
277ef7a
chore: remove todo
ShawnWeiChew
committed
1e757a9
chore: generalize kernel name
ShawnWeiChew
committed
1de2386

Commits on Sep 9, 2026

refactor: merge blackwell and hopper ag-gemm-bench
ShawnWeiChew
committed
0242787
fix: update makefile testing script
ShawnWeiChew
committed
bb9bfb9
fix: makefile and sync to prevent hang
ShawnWeiChew
committed
59404bf
fix: add TK correctness check
ShawnWeiChew
committed
a939f44
feat: transpose B tile
ShawnWeiChew
committed
b2c4354
tune supergroup
ShawnWeiChew
committed
067d2bb
compile tuning results
ShawnWeiChew
committed
9ef951a
try chaning epilogue
ShawnWeiChew
committed
58bf6ef
chore: remove formatting
ShawnWeiChew
committed
c529bca
revert another unecessary change
ShawnWeiChew
committed
5ba431f
more formatting
ShawnWeiChew
committed
5325095

Commits on Sep 10, 2026

feat: add some minor optimizations
ShawnWeiChew
committed
da74306