llama.cpp
CUDA: skip empty MoE tiles in MMQ stream-k fixup
#32067

Merged

CUDA: skip empty MoE tiles in MMQ stream-k fixup #32067

am17an merged 2 commits into ggml-org:master from lifeiteng:cuda-mmq-fixup-empty-tile

CUDA: skip empty MoE tiles in MMQ stream-k fixup

79ace80b

github-actions added testing

github-actions added ggml

github-actions added CUDA

lifeiteng marked this pull request as ready for review 3 days ago

lifeiteng requested a review from

ggerganov 3 days ago

lifeiteng requested a review 3 days ago

am17an approved these changes on 2026-11-11

JohannesGaessler commented on 2026-11-11

JohannesGaessler approved these changes on 2026-11-11

CUDA: drop the extra MUL_MAT_ID test case and the explanatory comment…

cf67da11

am17an approved these changes on 2026-11-12

am17an merged 5d4c8eeb into master 2 days ago

Reviewers

am17an

JohannesGaessler

ggerganov

Assignees

No one assigned

Labels

testing ggml CUDA

Milestone

No milestone