]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 8e5f498)
vulkan: support solve_tri with larger N/K values (#17781)
authorJeff Bolz <redacted>
Sat, 6 Dec 2025 07:56:45 +0000 (01:56 -0600)
committerGitHub <redacted>
Sat, 6 Dec 2025 07:56:45 +0000 (08:56 +0100)
commitc6c5e859798163c2e41d848d1157438467a2a34a
tree915613dd7a78dda2fe43d5dbd9c9bdc2d6c1288atree
parent8e5f4987b1c0f41ab80d4e95355e79eb6d169b8bcommit | diff
vulkan: support solve_tri with larger N/K values (#17781)

Split N into chunks to fit into shared memory.
If K > 128, use a larger workgroup with enough invocations.
Add perf tests matching qwen3next.
ggml/src/ggml-vulkan/ggml-vulkan.cpp diff | blob | history
ggml/src/ggml-vulkan/vulkan-shaders/solve_tri.comp diff | blob | history
tests/test-backend-ops.cpp diff | blob | history
Packaging of ggml-org/llama.cpp