Repository navigation

Add SM80 (Ampere/A100) dense MLA decode kernel - #183

Open

bzantium wants to merge 2 commits into

deepseek-ai:mainfrom

bzantium:sm80-dense-decode-port

Open

Add SM80 (Ampere/A100) dense MLA decode kernel#183
bzantium wants to merge 2 commits into
deepseek-ai:mainfrom
bzantium:sm80-dense-decode-port

Commits on May 1, 2026
- Add SM80 (Ampere/A100) dense MLA decode kernel
  bzantium
  committed
  4f5ac5b
Commits on Aug 14, 2026
- Raise SM80 decode occupancy by shrinking the K staging tile
  bzantium
  committed
  d679661