llama.cpp
CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 (recreated)
#28552

Merged

CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 (recreated) #28552

IMbackK merged 2 commits into master from pr-24546-recreated

CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 …

79800853

CUDA: pick MMQ tile size against ncols_opt set on the host side

f01a7549

pwilkin requested a review 54 days ago

JohannesGaessler commented on 2026-09-07

JohannesGaessler assigned

JohannesGaessler 54 days ago

github-actions added ggml

github-actions added CUDA

JohannesGaessler approved these changes on 2026-09-08

IMbackK approved these changes on 2026-09-09

IMbackK merged d4abd573 into master 52 days ago

Reviewers

IMbackK

JohannesGaessler

Assignees

JohannesGaessler

Labels

ggml CUDA

Milestone

No milestone