Bugfix - Add SM80/90 architecture to CUTLASS builds for CUDA 12.8+ for A100#785
Bugfix - Add SM80/90 architecture to CUTLASS builds for CUDA 12.8+ for A100#785
Conversation
Add sm_80 to the CUTLASS architecture lists to ensure Ampere GPU support is included in builds targeting CUDA 12.8+ and 12.9+.
There was a problem hiding this comment.
Pull request overview
Updates the CUTLASS build configuration to ensure Ampere (A100 / SM80) is included in CUTLASS builds when using CUDA 12.8+.
Changes:
- Add
sm_80to the CUTLASSARCHSlist for CUDA>= 12.8. - Expand the CUDA
>= 12.9CUTLASSARCHSlist to include80and90in addition to100/103.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| @@ -42,11 +42,11 @@ sb_micro_path: | |||
| # for cuda 12.9 and later Build from commit v3.9 (3.9 release commit) for blackwell support | |||
There was a problem hiding this comment.
The comment above cuda_cutlass says CUDA>=12.9 builds are from CUTLASS v3.9, but the rule actually clones --branch v4.1.0. Please update the comment to match the version/branch being used (and optionally clarify why v4.1.0 is needed for Blackwell).
| # for cuda 12.9 and later Build from commit v3.9 (3.9 release commit) for blackwell support | |
| # For CUDA 12.9 and later, build from CUTLASS v4.1.0 (required for Blackwell support). |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #785 +/- ##
=======================================
Coverage 85.70% 85.70%
=======================================
Files 102 102
Lines 7703 7703
=======================================
Hits 6602 6602
Misses 1101 1101
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
polarG
changed the title
Add sm_80 to the CUTLASS architecture lists to ensure Ampere GPU support is included in builds targeting CUDA 12.8+ for A100