Commit b1c79bf
committed
vulkan: Use larger workgroups for mul_mat_vec when M is small
Also use subgroup instructions for (part of) the reduction when supported.
Without this, the more expensive reductions would eat into the benefits of
the larger workgroups.1 parent 19f4dec commit b1c79bf
File tree
3 files changed
+135
-54
lines changed- ggml/src/ggml-vulkan
- vulkan-shaders
3 files changed
+135
-54
lines changed
0 commit comments