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

ggml backend interface wip

Browse source 
refactor ggml-cuda
This commit is contained in:
slaren 2023-07-10 17:32:06 +02:00
parent 6e7cca4047
commit 0d2b66c638
15 changed files with 4480 additions and 4782 deletions
Download patch file Download diff file Expand all files Collapse all files

10
CMakeLists.txt
View file

@ -67,7 +67,7 @@ endif()
option(LLAMA_ACCELERATE "llama: enable Accelerate framework" ON)
option(LLAMA_BLAS "llama: use BLAS" OFF)
set(LLAMA_BLAS_VENDOR "Generic" CACHE STRING "llama: BLAS library vendor")
option(LLAMA_CUBLAS "llama: use cuBLAS" OFF)
option(LLAMA_CUDA "llama: use CUDA" OFF)
option(LLAMA_CUDA_FORCE_DMMV "llama: use dmmv instead of mmvq CUDA kernels" OFF)
set(LLAMA_CUDA_DMMV_X "32" CACHE STRING "llama: x stride for dmmv CUDA kernels")
set(LLAMA_CUDA_MMV_Y "1" CACHE STRING "llama: y block size for mmv CUDA kernels")
@ -239,18 +239,18 @@ if (LLAMA_K_QUANTS)
endif()
endif()
if (LLAMA_CUBLAS)
if (LLAMA_CUDA)
cmake_minimum_required(VERSION 3.17)
find_package(CUDAToolkit)
if (CUDAToolkit_FOUND)
message(STATUS "cuBLAS found")
message(STATUS "CUDA found")
enable_language(CUDA)
set(GGML_SOURCES_CUDA ggml-cuda.cu ggml-cuda.h)
add_compile_definitions(GGML_USE_CUBLAS)
add_compile_definitions(GGML_USE_CUDA)
if (LLAMA_CUDA_FORCE_DMMV)
add_compile_definitions(GGML_CUDA_FORCE_DMMV)
endif()
@ -280,7 +280,7 @@ if (LLAMA_CUBLAS)
message(STATUS "Using CUDA architectures: ${CMAKE_CUDA_ARCHITECTURES}")
else()
message(WARNING "cuBLAS not found")
message(WARNING "CUDA not found")
endif()
endif()