Performance of FORTRAN and C GPU Extensions for a Benchmark Suite of Fourier Pseudospectral Algorithms

摘要

A comparison of PGI OpenACC, FORTRAN CUDA, and Nvidia CUDA pseudospectralmethods on a single GPU and GCC FORTRAN on single and multiple CPU cores isreported. The GPU implementations use CuFFT and the CPU implementations useFFTW. Porting pre-existing FORTRAN codes to utilize a GPUs is efficient andeasy to implement with OpenACC and CUDA FORTRAN. Example programs are provided.