]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
overview / pkg / ggml / sources / ggml / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: f4eb726)
vulkan: improve partial offloading performance on AMD (llama/19976)
authorRuben Ortlam <redacted>
Sun, 1 Mar 2026 16:32:14 +0000 (17:32 +0100)
committerGeorgi Gerganov <redacted>
Sun, 15 Mar 2026 19:50:13 +0000 (21:50 +0200)
commit162fd7c136861a1ce1d97177b031526a6428edc4
tree5a13d292dff34d87813730bae85e93539cb30d64tree
parentf4eb7267dad77f0b38ea107d734cb20102b37656commit | diff
vulkan: improve partial offloading performance on AMD (llama/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
src/ggml-vulkan/ggml-vulkan.cpp diff | blob | history
Packaging of ggml-org/ggml