mirror of
https://github.com/deepseek-ai/DeepGEMM.git
synced 2026-11-05 00:47:26 +00:00
[GH-ISSUE #59] [Question] Why not use TMA to load RHS scales? #38
Labels
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/DeepGEMM#38
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @ZJUGuoShuai on GitHub (Mar 12, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepGEMM/issues/59
The implementation uses TMA to load for LHS, LHS scaling factors, and RHS matrices. Why not use TMA to load RHS scales? Is it just for overlapping it with the TMA store in Math Warp-Groups?
@soundOfDestiny commented on GitHub (Mar 13, 2025):
It's okay to use TMA to load RHS scales.
@soundOfDestiny commented on GitHub (Mar 16, 2025):
example:
github.com/ademeure/DeeperGEMM@d1e54c9633Thanks @ademeure