]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 77f9c6b)
vulkan: fix mul_mat_vec failure in backend tests (#12529)
authorJeff Bolz <redacted>
Mon, 24 Mar 2025 06:56:17 +0000 (01:56 -0500)
committerGitHub <redacted>
Mon, 24 Mar 2025 06:56:17 +0000 (07:56 +0100)
commit9b169a4d4e01af7bc07a6981b53b27c18c9470d8
tree1adaf5969f6fd7cbc392d1b9c9e8b8c9f47b22d9tree
parent77f9c6bbe55fccd9ea567794024cb80943947901commit | diff
vulkan: fix mul_mat_vec failure in backend tests (#12529)

The OOB calculation could be wrong if the last iteration was during one of
the unrolled loops. Adjust the unrolling counts to avoid this. Add a couple
new backend tests that hit this failure on NVIDIA GPUs.
ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp diff | blob | history
tests/test-backend-ops.cpp diff | blob | history
Packaging of ggml-org/llama.cpp