]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: f4d7e54)
Faster AVX2 dot product for IQ2_XS (#5187)
authorKawrakow <redacted>
Tue, 30 Jan 2024 13:15:07 +0000 (15:15 +0200)
committerGitHub <redacted>
Tue, 30 Jan 2024 13:15:07 +0000 (15:15 +0200)
commit8e14e3ddb3744566aef7bc0fa734180e47ae6bdf
treeb7a3c7488bdc3fe81749889bb1754c1aed54a568tree
parentf4d7e5497485ce6ce0e322533930b7da4657dd2dcommit | diff
Faster AVX2 dot product for IQ2_XS (#5187)

* iq2xs: faster AVX2 dot product

* iq2xs: small AVX2 imrovement

* Speed up computing sign bits in AVX2 iq2_xs dot product

---------

Co-authored-by: Iwan Kawrakow <redacted>
Co-authored-by: Peter Reid <redacted>
ggml-quants.c diff | blob | history
Packaging of ggml-org/llama.cpp