]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 32c5411)
Metal: faster Q4_0 and Q4_1 matrix x vector kernels (#2212)
authorKawrakow <redacted>
Fri, 14 Jul 2023 09:46:21 +0000 (12:46 +0300)
committerGitHub <redacted>
Fri, 14 Jul 2023 09:46:21 +0000 (11:46 +0200)
commit27ad57a69b85bf12420a27e9945e580cc280be57
treef73b384b82088c94526a80a0eef1544eee2b1df7tree
parent32c54116318929c90fd7ae814cf9b5232cd44c36commit | diff
Metal: faster Q4_0 and Q4_1 matrix x vector kernels (#2212)

* 3-5% faster Q4_0 on Metal

* 7-25% faster Q4_1 on Metal

* Oops, forgot to delete the original Q4_1 kernel

---------

Co-authored-by: Iwan Kawrakow <redacted>
ggml-metal.m diff | blob | history
ggml-metal.metal diff | blob | history
Packaging of ggml-org/llama.cpp