Dense Singular Value Decomposition (SVD)

Sun S3L includes routines for performing the singular value decomposition of real single- or double-precision arrays. These routines include options for computing only the singular values or the right and/or left singular eigenvectors.

For the S3L_gen_svd function, optimal performance is achieved when block-cyclic distribution is used on the S3L array operand. S3L_gen_svd works best on large arrays. Performance is also better when S3L_gen_svd executes on a small number of processes.