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

TL;DR

This paper compares the performance of GPU and CPU implementations of Fourier pseudospectral algorithms using different programming extensions, highlighting ease of porting and efficiency of GPU methods.

Contribution

It provides a comparative analysis of GPU and CPU implementations of pseudospectral algorithms using OpenACC, CUDA FORTRAN, and standard FORTRAN, demonstrating porting ease and performance.

Findings

GPU implementations outperform CPU in spectral algorithms.

OpenACC and CUDA FORTRAN enable efficient GPU porting.

GPU methods show ease of implementation with existing codes.

Abstract

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