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

make : add optional CUDA_NATIVE_ARCH (#2482)

Browse source 
Use the environment variable `CUDA_NATIVE_ARCH` if present to set NVCC arch. Otherwise, use `native`.
This commit is contained in:
Alex 2023-10-22 15:56:53 -04:00 committed by GitHub
parent 438c2ca830
commit 96981f37b1
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 4 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
Makefile
View file

@ -391,9 +391,12 @@ else
endif #LLAMA_CUDA_NVCC
ifdef CUDA_DOCKER_ARCH
NVCCFLAGS += -Wno-deprecated-gpu-targets -arch=$(CUDA_DOCKER_ARCH)
endif # CUDA_DOCKER_ARCH
ifdef CUDA_NATIVE_ARCH
NVCCFLAGS += -arch=$(CUDA_NATIVE_ARCH)
else
NVCCFLAGS += -arch=native
endif # CUDA_DOCKER_ARCH
endif # CUDA_NATIVE_ARCH
ifdef LLAMA_CUDA_FORCE_DMMV
NVCCFLAGS += -DGGML_CUDA_FORCE_DMMV
endif # LLAMA_CUDA_FORCE_DMMV