llama.cpp
f01a7549 - CUDA: pick MMQ tile size against ncols_opt set on the host side

Commit

1 day ago

CUDA: pick MMQ tile size against ncols_opt set on the host side Assisted-by: Claude Fable 5.1 Claude-Session: https://claude.ai/code/session_011SYPfRhKoUpU3gMsGxq6go

References

pr-24546-recreated

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

Author

pwilkin

Committer

pwilkin

Parents

79800853

llama.cpp f01a7549 - CUDA: pick MMQ tile size against ncols_opt set on the host side

llama.cpp
f01a7549 - CUDA: pick MMQ tile size against ncols_opt set on the host side