Laplace solver running on GPU using CUDA, with CPU version for comparison. In addition, the CPU version contains support for OpenMP and OpenACC.
Change the problem size by modifying NUM in laplace_cpu.c and laplace_gpu.cu. To make all the executables (CPU, OpenMP, OpenACC, and GPU) type make.
To run each version, simply type
$ ./laplace_cpu
$ ./laplace_gpu
$ ./laplace_omp
$ ./laplace_acc
The makefile is set up to use gcc to compile the CPU version, nvcc (part of CUDA) to make the GPU version, and 'pgcc' (part of the Portland Group's Accelerator compilers) to make the OpenMP and OpenACC versions. gcc may also be used to make the OpenMP version by changing the appropriate line in the makefile.
The most up-to-date version of laplace_gpu can be found at the GitHub repository on GitHub.
laplace_gpu is released under the modified BSD license, see LICENSE for details.
Created by Kyle Niemeyer. Email address: kyle.niemeyer@gmail.com