]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 9ad5e60)
vulkan: Reuse conversion results in prealloc_y (#15410)
authorJeff Bolz <redacted>
Thu, 21 Aug 2025 14:55:00 +0000 (09:55 -0500)
committerGitHub <redacted>
Thu, 21 Aug 2025 14:55:00 +0000 (16:55 +0200)
commit96452a3fa426de83494fb0268a636aa1bfe557fe
tree43c9e6f01f59d799f4a5afabb7c654b761d2acactree
parent9ad5e60dba38a6718366b7ac43e7d8e8abdc36c9commit | diff
vulkan: Reuse conversion results in prealloc_y (#15410)

* vulkan: Reuse conversion results in prealloc_y

Cache the pipeline and tensor that were most recently used to fill prealloc_y,
and skip the conversion if the current pipeline/tensor match.

* don't use shared pointer for prealloc_y_last_pipeline_used
ggml/src/ggml-vulkan/ggml-vulkan.cpp diff | blob | history
tests/test-backend-ops.cpp diff | blob | history
Packaging of ggml-org/llama.cpp