llama.cpp
CUDA: fix thread/block count in quantized cpy kernel launches
#26731

Merged

CUDA: fix thread/block count in quantized cpy kernel launches #26731

ggerganov merged 2 commits into ggml-org:master from grafail:cuda-cpy-launch

CUDA: fix thread/block count in quantized cpy kernel launches

6862f133

grafail requested a review 10 days ago

github-actions added ggml

github-actions added CUDA

am17an assigned

am17an 10 days ago

grafail requested a review from

ggerganov 10 days ago

github-actions added testing

tests: add uneven block count cpy case

bf25a480

grafail force pushed from c0ef90ef to bf25a480 10 days ago

am17an approved these changes on 2026-08-07

am17an added merge ready

ggerganov merged 69bf6437 into master 9 days ago

Reviewers

am17an

ggerganov

Assignees

am17an

Labels

testing ggml merge ready CUDA

Milestone

No milestone