Department of Computer Science, UNC Chapel Hill

LUGPU:

LUGPU versus LAPACK on theCPU

We have observed considerable speedups with LUGPU over sgetrf and sgetc2 running on the CPU. The following graphs show the timings for two test systems. In both cases, LUGPU runs faster than the LAPACK implementation. In the case of full pivoting, the speedups are order of magnitudes faster than the CPU implementation.

LUGPU vs LAPACK Full Pivoting LUGPU vs LAPACK Partial Pivoting

©2003 Department of Computer Science, UNC Chapel Hill