]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 66d65ec)
vulkan: improve partial offloading performance on AMD (#19976)
authorRuben Ortlam <redacted>
Sun, 1 Mar 2026 16:32:14 +0000 (17:32 +0100)
committerGitHub <redacted>
Sun, 1 Mar 2026 16:32:14 +0000 (17:32 +0100)
commit319146247e643695f94a558e8ae686277dd4f8da
tree44e7fd415366709b1b29f054446ea4b40446b8c2tree
parent66d65ec29ba7c440cbc31b6f63b74a17b536ba65commit | diff
vulkan: improve partial offloading performance on AMD (#19976)

* vulkan: fix and enable cpy_tensor_async function

* use transfer_queue for async transfers on AMD, synchronize with timeline semaphore

* update offload_op logic

* fix missing transfer submission

* disable async transfer queue on AMD GCN

* revert op batch size change

* fix cpy_tensor_async checks
ggml/src/ggml-vulkan/ggml-vulkan.cpp diff | blob | history
Packaging of ggml-org/llama.cpp