vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml : add GGML_CUDA_USE_GRAPHS option, restore GGML_CUDA_FORCE_CUBLAS (cmake) (#8140)

Browse source
This commit is contained in:
slaren 2024-06-26 21:34:14 +02:00 • committed by GitHub
parent c7ab7b612c
commit 31ec3993f6
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 7 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

5
ggml/src/CMakeLists.txt
View file

@ -295,12 +295,15 @@ if (GGML_CUDA)
list(APPEND GGML_CDEF_PUBLIC GGML_USE_CUDA)
add_compile_definitions(GGML_CUDA_USE_GRAPHS)
add_compile_definitions(GGML_CUDA_DMMV_X=${GGML_CUDA_DMMV_X})
add_compile_definitions(GGML_CUDA_MMV_Y=${GGML_CUDA_MMV_Y})
add_compile_definitions(K_QUANTS_PER_ITERATION=${GGML_CUDA_KQUANTS_ITER})
add_compile_definitions(GGML_CUDA_PEER_MAX_BATCH_SIZE=${GGML_CUDA_PEER_MAX_BATCH_SIZE})
if (GGML_CUDA_USE_GRAPHS)
add_compile_definitions(GGML_CUDA_USE_GRAPHS)
endif()
if (GGML_CUDA_FORCE_DMMV)
add_compile_definitions(GGML_CUDA_FORCE_DMMV)
endif()