Skip to content

build(cuda): enable GGML_CUDA_GRAPHS on CUDA builds #26

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

Merged
mudler merged 1 commit into from
Jun 12, 2026
Merged

build(cuda): enable GGML_CUDA_GRAPHS on CUDA builds #26

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
8 changes: 8 additions & 0 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,14 @@ set(GGML_METAL ${PARAKEET_GGML_METAL} CACHE BOOL "" FORCE)
set(GGML_VULKAN ${PARAKEET_GGML_VULKAN} CACHE BOOL "" FORCE)
set(GGML_HIP ${PARAKEET_GGML_HIP} CACHE BOOL "" FORCE)

# CUDA graphs: ggml defaults this OFF, but on CUDA builds it captures/replays the
# compute graph for a small free speedup (~1% measured on GB10, never negative).
# Turn it on whenever we forward CUDA so every CUDA build (docker/release/local)
# gets it. The runtime kill-switch GGML_CUDA_DISABLE_GRAPHS=1 still works.
if(PARAKEET_GGML_CUDA)
set(GGML_CUDA_GRAPHS ON CACHE BOOL "" FORCE)
endif()

# Performance: -march=native (GGML_NATIVE) and tinyBLAS SGEMM (GGML_LLAMAFILE)
# give meaningful free speedups (~30% and ~25% per the rt-detr.cpp benchmarks).
# Force them on unless the caller explicitly overrides.
Expand Down
Loading