llama.cpp
vulkan: Dynamic subgroup size support for Q6_K mat_vec
#10536
Merged

vulkan: Dynamic subgroup size support for Q6_K mat_vec #10536

0cc4m merged 6 commits into ggml-org:master from vulkan
netrunnereve
netrunnereve subgroup 64 version with subgroup add. 15% faster
0aa5fd08
github-actions github-actions added Vulkan
github-actions github-actions added ggml
jeffbolznv
jeffbolznv
jeffbolznv commented on 2024-11-27
netrunnereve check for subgroup multiple of 16 and greater than 16
7c313b5f
jeffbolznv
jeffbolznv commented on 2024-11-28
sorasoras
0cc4m
sorasoras
0cc4m
sorasoras
0cc4m
netrunnereve Merge https://github.com/ggerganov/llama.cpp into vulkan
31a1d8af
netrunnereve subgroup sizes are always a power of 2 (https://github.com/KhronosGro…
97e0c686
netrunnereve force 16 sequential threads per block
2bca8122
jeffbolznv
jeffbolznv approved these changes on 2024-11-29
netrunnereve
netrunnereve netrunnereve marked this pull request as ready for review 1 year ago
0cc4m
0cc4m
0cc4m approved these changes on 2024-11-29
netrunnereve make 16 subgroup size a constant
b65961bf
0cc4m
0cc4m approved these changes on 2024-11-30
0cc4m 0cc4m merged 0533e7fb into master 1 year ago
netrunnereve netrunnereve deleted the vulkan branch 1 year ago
netrunnereve
0cc4m

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone