Repository navigation
HIP/CUDA: Fix reduction in mmf getting tile widths not devisiable by 64 in mul_mat_id on wave64 devices - #30836
Merged
Merged
Conversation
…64 in mul_mat_id on wave64 devices
github-actions
Bot
added
ggml
thelittlefireman
approved these changes
Oct 19, 2026
IMbackK
added
the
merge ready
3 of 6 tasks
edwardyoon
pushed a commit
to edwardyoon/focus-llama
that referenced
this pull request
Oct 24, 2026
MuFengMuXue
pushed a commit
to MuFengMuXue/llama.cpp-gfx906
that referenced
this pull request
Oct 26, 2026
auRose94
pushed a commit
to auRose94/llama.cpp
that referenced
this pull request
Oct 28, 2026
1 task done
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Mmf reduces over ggml_cuda_get_physical_warp_size blocks, but the checks in mul_mat_id use WARP_SIZE instead, meaning that wave64 devices can end up reducing over a half wave sized block.
Overview
Noticed in test-llama-archs by laguna failing on mi100
Requirements