cuda: reject Blackwell in the hopper wgmma gate - #47
Merged
Conversation
The path's kernels are sm_90a-only fatbins (wgmma does not exist on Blackwell); cc >= 900 alone admits cc 1200+ where the launch fails. Blackwell support is a separate tcgen05 path.
khosravipasha
approved these changes
Jun 11, 2026
bri-prism
added a commit
that referenced
this pull request
Jun 12, 2026
The path's kernels are sm_90a-only fatbins (wgmma does not exist on Blackwell); cc >= 900 alone admits cc 1200+ where the launch fails. Blackwell support is a separate tcgen05 path.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
One-line correctness gate fix: the opt-in Hopper path's kernels are sm_90a-only fatbins, but
cc < 900alone admits Blackwell (cc 1200+), where the wgmma kernels do not exist and the launch fails. Rejectcc >= GGML_CUDA_CC_BLACKWELL. Blackwell gets its own tcgen05-based path (investigation underway); until then it falls through to stock MMQ, which already has native MXFP4 handling there.