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

cuda sources (+1 squashed commits)

Browse source 
Squashed commits:

[d3aedc03] add source universally
This commit is contained in:
Concedo 2023-09-12 18:43:18 +08:00
parent 74384cfbb5
commit 78b2602844
1 changed files with 5 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

10
CMakeLists.txt
View file

@ -43,7 +43,7 @@ if (NOT MSVC)
endif()
# 3rd party libs
option(LLAMA_CUBLAS "llama: use CUDA" OFF)
option(LLAMA_CUBLAS "llama: use CUDA" ON)
set(LLAMA_CUDA_MMQ_Y "64" CACHE STRING "llama: y tile size for mmq CUDA kernels")
set(LLAMA_CUDA_DMMV_X "32" CACHE STRING "llama: x stride for dmmv CUDA kernels")
set(LLAMA_CUDA_DMMV_Y "1" CACHE STRING "llama: y block size for dmmv CUDA kernels")
@ -68,6 +68,10 @@ find_package(Threads REQUIRED)
add_compile_definitions(GGML_USE_K_QUANTS)
add_compile_definitions(LOG_DISABLE_LOGS)
set(GGML_SOURCES_CUDA ggml-cuda.cu ggml-cuda.h)
set(GGML_V2_CUDA_SOURCES otherarch/ggml_v2-cuda.cu otherarch/ggml_v2-cuda.h)
set(GGML_V2_LEGACY_CUDA_SOURCES otherarch/ggml_v2-cuda-legacy.cu otherarch/ggml_v2-cuda-legacy.h)
if (LLAMA_CUBLAS)
cmake_minimum_required(VERSION 3.17)
@ -77,10 +81,6 @@ if (LLAMA_CUBLAS)
enable_language(CUDA)
set(GGML_SOURCES_CUDA ggml-cuda.cu ggml-cuda.h)
set(GGML_V2_CUDA_SOURCES otherarch/ggml_v2-cuda.cu otherarch/ggml_v2-cuda.h)
set(GGML_V2_LEGACY_CUDA_SOURCES otherarch/ggml_v2-cuda-legacy.cu otherarch/ggml_v2-cuda-legacy.h)
add_compile_definitions(GGML_USE_CUBLAS)
#add_compile_definitions(GGML_CUDA_CUBLAS) #remove to not use cublas
add_compile_definitions(GGML_CUDA_MMQ_Y=${LLAMA_CUDA_MMQ_Y})