Tested in Tsubame f-node
Used Matrix N=2048
4 processed for mpi
| Method | Time (s) | GFlops | Error |
|---|---|---|---|
| MPI(example) | 34.253257 | 0.501554 | 0.000364 |
| MPI+openmp | 5.137721 | 3.343870 | 0.000364 |
| MPI+openmp+SIMD | 0.859402 | 19.990485 | 0.000364 |
| MPI+openmp+SIMD+CacheBlocking | 0.049206 | 349.140059 | 0.000258 |
| Cuda | 0.052758 | 325.635400 | 0.000364 |
| MPI+Cuda | 1.605541 | 10.700359 | 0.000364 |
module load gcc
module load intel-mpi
module load gcc
module load cuda/11.2.146
module load openmpi
mpicxx example.cpp
mpirun -np 4 ./a.out
mpicxx mpi_openmp.cpp -fopenmp
mpirun -np 4 ./a.out
mpicxx mpi_openmp_simd.cpp -fopenmp -fopt-info-optimized -march=native -O3
mpirun -np 4 ./a.out
mpicxx mpi_openmp_simd_cacheBlocking.cpp -fopenmp -fopt-info-optimized -march=native -O3
mpirun -np 4 ./a.out
nvcc cuda.cu
./a.out
nvcc mpi_cuda.cu -lmpi
./a.out