PR #26264 cuda: unblock mmq for MoE on sm_60

cuda: unblock mmq for MoE on sm_60 #26264

am17an merged 3 commits into ggml-org:master from dfriehs:p100-moe-mmq

dfriehs requested a review 80 days ago

github-actions added ggml

github-actions added CUDA

am17an approved these changes on 2026-07-29

dfriehs marked this pull request as draft 79 days ago

dfriehs force pushed 78 days ago

dfriehs marked this pull request as ready for review 78 days ago

cuda: unblock mmq for MoE on sm_60

ddf75a2f

cuda: duplicate mmq-config-pascal for dp4a and older

70f323f0

cuda: reduce occupancy on non-dp4a pascal for Q2_K, Q4_K, Q5_K, Q6_K

1809286c

dfriehs force pushed to 1809286c 53 days ago

IMbackK approved these changes on 2026-08-26

am17an approved these changes on 2026-08-26

am17an merged fc35562b into master 52 days ago

dfriehs deleted the p100-moe-mmq branch 30 days ago

Reviewers

am17an

IMbackK

Assignees

No one assigned

Labels

ggml CUDA

Milestone

No milestone