This chapter provides a few guidelines for making best use of the performance benefits offered by Sun S3L. The range of topics covered include
Sun S3L functions that benefit from cyclic distribution
Sun S3L functions that benefit from distributing only the last axis
Using shared memory
Performance guidelines specific to FFT routines
Performance guidelines specific to dense SVD routines
Performance guidelines specific to dense linear system solvers
Performance guidelines specific to banded solvers
Performance guidelines specific to sparse linear systems solvers
Performance guidelines specific to dense matrix operations
Support for convolution, deconvolution, correlation, and autocorrelation