]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
vulkan: Support asymmetric FA in coopmat2 path (#21753)
authorJeff Bolz <redacted>
Fri, 1 May 2026 13:28:32 +0000 (15:28 +0200)
committerGitHub <redacted>
Fri, 1 May 2026 13:28:32 +0000 (15:28 +0200)
commit05e141a6b34a1535096e2bb7828418df431f7efe
tree4e3aa35e637e4e18d1202c0e9ed1476c8fc12b84
parentaab68217b7bd8907135dd41fbb5bcb85fca06045
vulkan: Support asymmetric FA in coopmat2 path (#21753)

* vulkan: Support asymmetric FA in coopmat2 path

There has been some recent interest/experimentation with mixed quantization
types for FA. I had originally designed the cm2 FA shader with this in mind
(because I didn't realize it wasn't supported at the time!), this change
adds the missing pieces and enables it.

Also support Q1_0 since people have been trying that out (seems crazy, but
who knows).

We should be able to do similar things in the coopmat1/scalar path, but
there's another change open against the scalar path and I don't want to
conflict.

* reorder cases
ggml/src/ggml-vulkan/ggml-vulkan.cpp
ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.glsl
ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp
ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp
tests/test-backend-ops.cpp