]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
benchmark : add tool for timing q4_0 matrix multiplication (#653)
authorSebastianApel <redacted>
Thu, 13 Apr 2023 12:46:23 +0000 (14:46 +0200)
committerGitHub <redacted>
Thu, 13 Apr 2023 12:46:23 +0000 (15:46 +0300)
commit95ea26f6e92d620a5437f576b80868aee7f808d6
tree9256b919bf5939cbc18a91a6975f453f0fcb45a5
parent82d146df9b43cf677e0dbce20b03cf864958a0cc
benchmark : add tool for timing q4_0 matrix multiplication (#653)

* Initial version of q4_0 matrix multiplication benchmark

* Bugfix: Added dependency to ggml.o to benchmark

* Reviewer requests: added parameter for threads, switched to ggml_time_us()

* Reviewer input: removed rtsc, use epsilon for check

* Review comment: Removed set_locale

* Feature: Param for numer of iterations, Bugfix for use of parameter threads

* Reviewer suggestion: Moved to examples

* Reviewer feedback: Updated clean: and benchmark: sections

---------

Co-authored-by: Georgi Gerganov <redacted>
Makefile
examples/benchmark/benchmark-q4_0-matmult.c [new file with mode: 0644]