In my research I have a full unsymmetric system of equations to solve. Just
to get an answer I've been using the ludcmp and lubksb routines from
Numerical Recipes.
This hasn't been a problem until I moved to several thousand degrees of
freedom, e.g., 5-10k. Now solutions times are unacceptable.
My questions are as follows:
- can someone recommend a Fortran public domain routine for solving a full
unsymmetric system of equations that might be faster?
- should I consider an iterative solver?
- I'm looking at ways to create a banded matrix (without too much loss of
accuracy). Can someone recommend a solver for unsymmetric banded matrices?
Dr. Harry R. Millwater Jr.
Assistant Professor of Mechanical Engineering
The University of Texas at San Antonio
6900 North Loop 1604 West
San Antonio, TX 78249-0670
(210) 458-4481 Voice
(210) 458-5589 Fax
|