llama.cpp
HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and shapes
#14949

Merged

HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and shapes #14949

IMbackK merged 2 commits into ggml-org:master from IMbackK:mi100mfma

IMbackK requested a review from

JohannesGaessler 225 days ago

github-actions added Nvidia GPU

github-actions added ggml

deepsek commented on 2025-07-29

HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and sh…

0e7995d4

IMbackK force pushed to 0e7995d4 224 days ago

JohannesGaessler approved these changes on 2025-07-30

Remove parentheses

7875fdfe

IMbackK merged ad4a7001 into master 224 days ago

Reviewers

JohannesGaessler

deepsek

Assignees

No one assigned

Labels

Nvidia GPU ggml

Milestone

No milestone