Skip to content

benchmarks: build ggml with the tinyBLAS GEMM path (GGML_LLAMAFILE) llama.cpp enables by default #46

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
codspeed-hq wants to merge 1 commit into
base: master
Choose a base branch
from
Open

benchmarks: build ggml with the tinyBLAS GEMM path (GGML_LLAMAFILE) llama.cpp enables by default #46

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 11 additions & 0 deletions benchmarks/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,6 +58,17 @@ if (CODSPEED_MODE STREQUAL "simulation")
set(GGML_F16C ON CACHE BOOL "" FORCE)
endif()

# Match the ggml configuration that llama.cpp itself builds with. The root
# llama.cpp CMakeLists.txt changes the default of GGML_LLAMAFILE to ON (see
# "change the default for these ggml options"), which enables the tiled
# tinyBLAS matmul kernels in ggml-cpu. This project adds ggml directly as a
# subdirectory and therefore never sees that override, so without this the
# benchmarks measure the unoptimized per-row ggml_vec_dot() fallback instead of
# the GEMM path real llama.cpp inference takes.
if (NOT DEFINED GGML_LLAMAFILE)
set(GGML_LLAMAFILE_DEFAULT ON)
endif()

# Build the ggml compute core shipped in this repository.
add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/../ggml ${CMAKE_BINARY_DIR}/ggml)

Expand Down