HPCC_optimization

feature

more than 9x faster than baseline(Using CUDA lib)

requirements

• FTW3.3.8
• CUDA 9 or 10
• Intel MKL or cblas lib
• MPI lib (e.g., Mpich, openMPI)

compile FFTW3

./configure --prefix=/home/ringbo/hpcc/fftw-3.3.8/install --enable-mpi --enable-avx512 --enable-sse2 --enable-generic-simd256 --enable-single