llama.cpp
hip : substituted bpermute ops with swizzle ops (gfx906, maybe all AMD)
#16291

Open

hip : substituted bpermute ops with swizzle ops (gfx906, maybe all AMD) #16291

iacopPBK wants to merge 7 commits into ggml-org:master from iacopPBK:master

iacopPBK requested a review from

ggerganov 13 days ago

iacopPBK requested a review from

JohannesGaessler 13 days ago

iacopPBK requested a review from

slaren 13 days ago

github-actions added Nvidia GPU

github-actions added ggml

IMbackK assigned

IMbackK 12 days ago

Optimize AMD GFX906 flash attention with DS_SWIZZLE instrinsics Autho…

45d6cec3

iacopPBK force pushed from 6a2203f8 to 45d6cec3 11 days ago

Update README.md

50e47774

Update README.md

aad3f096

Update README.md

1e97e687

Update README.md

4513b24d

Update README.md

ec8ac2d8

Update README.md

229d7a74

IMbackK requested changes on 2025-09-30

JohannesGaessler commented on 2025-09-30

Reviewers

IMbackK

JohannesGaessler

ggerganov

slaren

Assignees

IMbackK

Labels

Nvidia GPU ggml

Milestone

No milestone