llama.cpp
vulkan: add tiled-transpose concat shader for Qwen models
#30591
Merged

vulkan: add tiled-transpose concat shader for Qwen models #30591

ggerganov merged 6 commits into master from 0cc4m/vulkan-qwen-concat-fix
0cc4m
0cc4m 0cc4m requested a review 26 days ago
github-actions github-actions added Vulkan
github-actions github-actions added ggml
jeffbolznv
0cc4m
0cc4m
jeffbolznv
jeffbolznv commented on 2026-10-15
0cc4m
jeffbolznv
0cc4m 0cc4m requested a review from ggerganov ggerganov 22 days ago
github-actions github-actions added testing
tczank
0cc4m 0cc4m force pushed from 52e41a44 to cc28c8df 9 days ago
oscarmherrera
Nathanw1014 vulkan: tiled transpose for dim-0 concat with a transposed source
5795fcf3
0cc4m enable by default, add smaller shader sizes
c5bef907
0cc4m clean up comments
af6b215f
0cc4m improve shader
ad5bb319
0cc4m Add non-contiguous concat tests
429bf415
0cc4m handle x or y axis overflow
90b7a327
0cc4m 0cc4m force pushed from cc28c8df to 90b7a327 1 day ago
0cc4m
jeffbolznv
jeffbolznv approved these changes on 2026-11-09
nerun
0cc4m 0cc4m added merge ready
0cc4m
ggerganov ggerganov merged ae4893a0 into master 1 day ago
0cc4m 0cc4m deleted the 0cc4m/vulkan-qwen-concat-fix branch 1 day ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone