]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
vulkan: subgroup size tuning (llama/12087)
authorDaniele <redacted>
Mon, 17 Mar 2025 11:42:33 +0000 (12:42 +0100)
committerGeorgi Gerganov <redacted>
Thu, 27 Mar 2025 09:06:03 +0000 (11:06 +0200)
commit13eeebb1b28e54fe5f3604d305132bcaf7238e9a
treed264ff1c1e1beb522e3beb1b7650424a9f2f4245
parent905b834af1367d0166d1c3cec61ca0cc3dd35782
vulkan: subgroup size tuning (llama/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