CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-29 08:41:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: correct the lowest Maxwell supported by CUDA 12 (#11984)

Browse Source 
* CUDA: correct the lowest Maxwell supported by CUDA 12

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit is contained in:
PureJourney
2025-02-21 19:21:05 +08:00
committed by GitHub
parent 0b3863ff95
commit ecc8e3aeff
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-cuda/CMakeLists.txt
View File

@@ -7,7 +7,7 @@ if (CUDAToolkit_FOUND)
if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES) if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
# native == GPUs available at build time # native == GPUs available at build time
# 52 == Maxwell, lowest CUDA 12 standard # 50 == Maxwell, lowest CUDA 12 standard
# 60 == P100, FP16 CUDA intrinsics # 60 == P100, FP16 CUDA intrinsics
# 61 == Pascal, __dp4a instruction (per-byte integer dot product) # 61 == Pascal, __dp4a instruction (per-byte integer dot product)
# 70 == V100, FP16 tensor cores # 70 == V100, FP16 tensor cores
@@ -17,7 +17,7 @@ if (CUDAToolkit_FOUND)
elseif(GGML_CUDA_F16 OR GGML_CUDA_DMMV_F16) elseif(GGML_CUDA_F16 OR GGML_CUDA_DMMV_F16)
set(CMAKE_CUDA_ARCHITECTURES "60;61;70;75;80") set(CMAKE_CUDA_ARCHITECTURES "60;61;70;75;80")
else() else()
set(CMAKE_CUDA_ARCHITECTURES "52;61;70;75;80") set(CMAKE_CUDA_ARCHITECTURES "50;61;70;75;80")
endif() endif()
endif() endif()
message(STATUS "Using CUDA architectures: ${CMAKE_CUDA_ARCHITECTURES}") message(STATUS "Using CUDA architectures: ${CMAKE_CUDA_ARCHITECTURES}")