llama.cpp
vulkan: prefer host-visible memory buffers on UMA devices
#22930

Merged

vulkan: prefer host-visible memory buffers on UMA devices #22930

0cc4m merged 2 commits into ggml-org:master from winstonma:vulkan-uma-allocation

winstonma requested a review 64 days ago

github-actions added Vulkan

github-actions added ggml

winstonma force pushed from 95fdd413 to 08319b9f 63 days ago

winstonma changed the title ~~UMA buffers get host-visible memory at allocation time~~ UMA buffers prefer host-visible memory 63 days ago

0cc4m commented on 2026-05-13

implement UMA host-visible memory

cd59801e

winstonma force pushed from dddb22e9 to cd59801e 31 days ago

update based on 0cc4m's suggestion

2a7a794e

0cc4m approved these changes on 2026-06-16

0cc4m changed the title ~~UMA buffers prefer host-visible memory~~ vulkan: prefer host-visible memory buffers on UMA devices 28 days ago

taronaeo approved these changes on 2026-06-16

0cc4m merged 32120c10 into master 28 days ago

winstonma deleted the vulkan-uma-allocation branch 28 days ago

Reviewers

taronaeo

0cc4m

Assignees

No one assigned

Labels

Vulkan ggml

Milestone

No milestone