]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
vulkan: support solve_tri with larger N/K values (llama/17781)
authorJeff Bolz <redacted>
Sat, 6 Dec 2025 07:56:45 +0000 (01:56 -0600)
committerGeorgi Gerganov <redacted>
Thu, 11 Dec 2025 13:32:57 +0000 (15:32 +0200)
commit0124b66315ed3875e48c9071dac4f0289b27b85b
treeed274eb442183daf01b3a370fcdd819b36029662
parent56d5a8a696686dcf0bda7786443b2828d0672c00
vulkan: support solve_tri with larger N/K values (llama/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.
src/ggml-vulkan/ggml-vulkan.cpp
src/ggml-vulkan/vulkan-shaders/solve_tri.comp
tests/test-backend-ops.cpp