64-BitBrainstorm_2026/AMSS-NCKU
8
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: port GPU code to CUDA 13 and enable GPU computation

Browse Source 
Major changes:
   - Update makefile.inc for CUDA 13.1 with sm_89 architecture (RTX 4050)
   - Replace deprecated cudaThreadSynchronize() with cudaDeviceSynchronize()
   - Add CUDA_SAFE_CALL macro for CUDA 13 compatibility
   - Fix duplicate function definitions (compare_result_gpu, SHStep)
   - Fix syntax error in bssn_step_gpu.C
   - Enable GPU calculation in AMSS_NCKU_Input.py
   - Successfully build ABEGPU executable
This commit is contained in:
CGH0S7
2026-01-13 18:15:49 +00:00
parent b27e071cde
commit 75be0968fc
6 changed files with 83 additions and 68 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
AMSS_NCKU_source/makefile.inc
View File

@@ -3,7 +3,7 @@ filein = -I/usr/include -I/usr/include/openmpi-x86_64 -I/usr/lib/gcc/x86_64-lin
##filein = -I/usr/include/ -I/usr/lib/x86_64-linux-gnu/openmpi/include/ -I/usr/lib/x86_64-linux-gnu/openmpi/lib/ -I/usr/lib/gcc/x86_64-linux-gnu/11/ -I/usr/include/c++/11/ -I/usr/lib/cuda/include
LDLIBS = -L/usr/lib64/openmpi/lib -Wl,-rpath,/usr/lib64/openmpi/lib -lmpi -lgfortran
LDLIBS = -L/usr/lib64/openmpi/lib -Wl,-rpath,/usr/lib64/openmpi/lib -lmpi -lgfortran -L/usr/local/cuda-13.1/lib64 -Wl,-rpath,/usr/local/cuda-13.1/lib64 -lcudart -lcuda
##LDLIBS = -L/usr/lib/x86_64-linux-gnu -L/usr/lib64 -L/usr/lib/gcc/x86_64-linux-gnu/11 -lgfortran -L/usr/lib/cuda/lib64 -lcudart -lmpi -lgfortran
CXXAPPFLAGS = -O3 -Wno-deprecated -Dfortran3 -Dnewc
@@ -15,7 +15,8 @@ CXX = g++
CC = gcc
CLINKER = mpic++
Cu = nvcc
CUDA_LIB_PATH = -L/opt/nvidia/hpc_sdk/Linux_x86_64/24.5/cuda/lib64 -I/usr/include -I/opt/nvidia/hpc_sdk/Linux_x86_64/24.5/cuda/include
Cu = /usr/local/cuda-13.1/bin/nvcc
CUDA_LIB_PATH = -L/usr/local/cuda-13.1/lib64 -I/usr/include -I/usr/local/cuda-13.1/include
#CUDA_APP_FLAGS = -c -g -O3 --ptxas-options=-v -arch compute_13 -code compute_13,sm_13 -Dfortran3 -Dnewc
CUDA_APP_FLAGS = -c -g -O3 --ptxas-options=-v -Dfortran3 -Dnewc
# RTX 4050 uses Ada Lovelace architecture (compute capability 8.9)
CUDA_APP_FLAGS = -c -g -O3 --ptxas-options=-v -arch=sm_89 -Dfortran3 -Dnewc