A package for solving systems of sparse linear systems Ax=b, where A is
sparse and can be unsymmetric. There are options for choosing a good
pivot order, factorizing a subsequent matrix with the same pivot order and
nonzero pattern as a previously factorized matrix, and solving systems of
linear equations with the factors (with A, L, or U; or with their
transposes in the single/double precision versions). Iterative
refinement, with sparse backward error estimates, can be performed.
Single and double precision, complex, and complex double precision
(complex*16) routines are available.
WWW: http://www.cise.ufl.edu/research/sparse/umfpack/
This package also includes AMD, a set of routines for pre-ordering sparse
matrices prior to Cholesky or LU factorization using the aproximate
minimum degree ordering algorithm:
http://www.cise.ufl.edu/research/sparse/amd/
_________
Note: By default this port builds the C-callable interface using the CBLAS
interface from ATLAS, but other versions of BLAS could be used and there
is also a FORTRAN interface available.
Mark this port as DEPRECATED with and EXPIRATION_DATE of 2007-11-30, as it's an
outdated, unmaintained subset of math/suitesparse.
PR: 116965
Submitted by: bf <bf2006a@yahoo.com>
Use BLAS instead of ATLAS by default.
I used to recommend the use of ATLAS over the standard fortran BLAS.
Unfortunately for some time we are not packaging ATLAS anymore, and very
recently the source code of GotoBLAS has been made available for
research purposes. GotoBLAS is said to be faster than ATLAS and the
shared library is exchangeable with the standard BLAS. Curiously the
standard BLAS is known to beat both GotoBLAS and ATLAS when the problems
are not too complex, so I expect it's better for beginners.
I am hereto recommending the use of shared blas by default but leaving a
WITH_ATLAS knob in case someone doesn't have access to GotoBLAS and
wants to try to squeeze some more performance.
PR: ports/98070
Submitted by: Pedro F. Giffuni <giffunip (at) asme.org>
- Let the cblas library get overriden from the ports makefile;
- Add a test: target;
- Define -fPIC on amd64.
PR: 93005
Submitted by: Pedro F. Giffuni <giffunip (at) asme.org>