Skip to content
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

[Tracking Issue] Performance with CUDA 12.8 #780

Open
6 tasks
yzh119 opened this issue Feb 4, 2025 · 0 comments
Open
6 tasks

[Tracking Issue] Performance with CUDA 12.8 #780

yzh119 opened this issue Feb 4, 2025 · 0 comments

Comments

@yzh119
Copy link
Collaborator

yzh119 commented Feb 4, 2025

CUDA 12.8 was released recently, this issue tracks the performance of flashinfer on cu128:

  • Performance Degradation: hopper-based vector sparse attention kernel got worse performance on cu128 than on cu124
    • Check PTX/SASS code
    • Try setting different number of registers for producers & consumers.
  • Performance Improvement for BatchPrefillWithRaggedKVCache (the causal attention performance for HEAD_DIM=256 is bad)
    • Change the UseSchedulerBarrier setting.
    • Tweak the pipeline design.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@yzh119