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
iacopPBK iacopPBK requested a review from ggerganov ggerganov 13 days ago
iacopPBK iacopPBK requested a review from JohannesGaessler JohannesGaessler 13 days ago
iacopPBK iacopPBK requested a review from slaren slaren 13 days ago
github-actions github-actions added Nvidia GPU
github-actions github-actions added ggml
JohannesGaessler
iacopPBK
IMbackK
IMbackK IMbackK assigned IMbackK IMbackK 12 days ago
maximumbusdatatype
iacopPBK
JohannesGaessler
iacopPBK Optimize AMD GFX906 flash attention with DS_SWIZZLE instrinsics Autho…
45d6cec3
iacopPBK iacopPBK force pushed from 6a2203f8 to 45d6cec3 11 days ago
iacopPBK Update README.md
50e47774
iacopPBK Update README.md
aad3f096
iacopPBK Update README.md
1e97e687
iacopPBK Update README.md
4513b24d
iacopPBK Update README.md
ec8ac2d8
iacopPBK Update README.md
229d7a74
iacopPBK
IMbackK
IMbackK requested changes on 2025-09-30
JohannesGaessler
JohannesGaessler commented on 2025-09-30
maximumbusdatatype

Login to write a write a comment.

Login via GitHub

Assignees
Labels
Milestone