]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
vulkan: subgroup size tuning (#12087)
authorDaniele <redacted>
Mon, 17 Mar 2025 11:42:33 +0000 (12:42 +0100)
committerGitHub <redacted>
Mon, 17 Mar 2025 11:42:33 +0000 (12:42 +0100)
commitcf2270e4d3685ac46f4a166d8718997ba7cbc45a
treec02d33f35e075e015ce6d690253eaf8cfa4f6e13
parentf07690c930f74d82d4f108e567c7092544847f77
vulkan: subgroup size tuning (#12087)

* vulkan: subgroup size test

* Vulkan: Add device architecture enum and logic to recognize AMD generations

* vulkan: use new architecture logic to specify subgroup size

* Initial vulkan subgroup size tuning for RDNA3

* vulkan: commonize RDNA subgroup tuning

* vulkan: override subgroup size if required_subgroup_size = 0

* vulkan: disable warp 32 for RDNA3

* vulkan: fine tuned RDNA1 subgroup sizes

* vulkan: adjusted subgroup size map

* vulkan: fixed RDNA2 subgroup map

---------

Co-authored-by: 0cc4m <redacted>
ggml/src/ggml-vulkan/ggml-vulkan.cpp