llama.cpp
vulkan: improve partial offloading performance on AMD
#19976

Merged

vulkan: improve partial offloading performance on AMD #19976

0cc4m merged 7 commits into master from 0cc4m/vulkan-partial-offload-fix

vulkan: fix and enable cpy_tensor_async function

6943f830

use transfer_queue for async transfers on AMD, synchronize with timel…

5abb7d55

update offload_op logic

ca3481f3

fix missing transfer submission

e72fb936

disable async transfer queue on AMD GCN

29955d39

revert op batch size change

32adb28b

0cc4m requested a review from

jeffbolznv 12 days ago

github-actions added Vulkan

github-actions added ggml

fix cpy_tensor_async checks

b2bc5eb1

jeffbolznv approved these changes on 2026-02-28

0cc4m merged 31914624 into master 10 days ago

0cc4m deleted the 0cc4m/vulkan-partial-offload-fix branch 10 days ago

Reviewers

jeffbolznv

Assignees

No one assigned

Labels

Vulkan ggml

Milestone

No milestone