vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-08-25 08:14:24 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Revert "try remove volta as a dedicated target b (+1 squashed commits)"

Browse source 
This reverts commit ddba580f00.
This commit is contained in:
Concedo 2025-12-06 21:31:44 +08:00
parent ddba580f00
commit d27949f22a
2 changed files with 4 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
CMakeLists.txt
View file

@ -139,10 +139,10 @@ if (LLAMA_CUBLAS)
elseif(CUDAToolkit_VERSION VERSION_GREATER 12)
add_compile_definitions(GGML_CUDA_USE_GRAPHS) #try enable cuda graphs on cu12 build
add_compile_definitions(KCPP_LIMIT_CUDA_MAX_ARCH=800)
set(CMAKE_CUDA_ARCHITECTURES "50-virtual;61-virtual;75-virtual;80-virtual") # lowest CUDA 12 standard + lowest for integer intrinsics
set(CMAKE_CUDA_ARCHITECTURES "50-virtual;61-virtual;70-virtual;75-virtual;80-virtual") # lowest CUDA 12 standard + lowest for integer intrinsics
else()
add_compile_definitions(KCPP_LIMIT_CUDA_MAX_ARCH=750) #will cause issues with ggml_cuda_highest_compiled_arch if removed
set(CMAKE_CUDA_ARCHITECTURES "35-virtual;50-virtual;61-virtual;75-virtual") # lowest CUDA 12 standard + lowest for integer intrinsics
set(CMAKE_CUDA_ARCHITECTURES "35-virtual;50-virtual;61-virtual;70-virtual;75-virtual") # lowest CUDA 12 standard + lowest for integer intrinsics
endif()
endif()
message(STATUS "Using CUDA architectures: ${CMAKE_CUDA_ARCHITECTURES}")

2
Makefile
View file

@ -226,6 +226,7 @@ NVCCFLAGS += -Wno-deprecated-gpu-targets \
-gencode arch=compute_35,code=compute_35 \
-gencode arch=compute_50,code=compute_50 \
-gencode arch=compute_61,code=compute_61 \
-gencode arch=compute_70,code=compute_70 \
-gencode arch=compute_75,code=compute_75 \
-DKCPP_LIMIT_CUDA_MAX_ARCH=750
@ -233,6 +234,7 @@ else ifdef LLAMA_ARCHES_CU12
NVCCFLAGS += -Wno-deprecated-gpu-targets \
-gencode arch=compute_50,code=compute_50 \
-gencode arch=compute_61,code=compute_61 \
-gencode arch=compute_70,code=compute_70 \
-gencode arch=compute_75,code=compute_75 \
-gencode arch=compute_80,code=compute_80 \
-DKCPP_LIMIT_CUDA_MAX_ARCH=800