]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
CUDA: fix thread/block count in quantized cpy kernel launches (#26731)
authorRafail Giavrimis <redacted>
Sat, 8 Aug 2026 04:40:04 +0000 (05:40 +0100)
committerGitHub <redacted>
Sat, 8 Aug 2026 04:40:04 +0000 (07:40 +0300)
commit69bf6437914596fbbc4caf09a7ac16f2acdd1a94
treeb0ce255996a1bc91c6a64ea2d301b3a7a635fbcb
parent3653e6d6d547ec763317d9ecd0ace334a7e21359
CUDA: fix thread/block count in quantized cpy kernel launches (#26731)

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

* tests: add uneven block count cpy case
ggml/src/ggml-cuda/cpy.cu
tests/test-backend-ops.cpp