llama.cpp
≈65% speedup of the AVX-512 implementation of `ggml_vec_dot_q4_0()`
#933

Merged

≈65% speedup of the AVX-512 implementation of `ggml_vec_dot_q4_0()` #933

dfyz merged 1 commit into ggml-org:master from dfyz:master

dfyz closed this 2 years ago

dfyz force pushed to 0e07e6a8 2 years ago

dfyz reopened this 2 years ago

ggerganov added performance

ggerganov added high priority

ggerganov approved these changes on 2023-04-15

dfyz force pushed from d787348b 2 years ago

dfyz force pushed to 6a4fa4d9 2 years ago

dfyz merged f266259a into master 2 years ago

Speedup the AVX-512 implementation of ggml_vec_dot_q4_0()

6a4fa4d9

Reviewers

ggerganov

Assignees

No one assigned

Labels

performance high priority

Milestone

No milestone