Numerically, the last solution step during the calculation is to solve a linear system of equations. You can choose between different solvers. Generally one can distinguish direct and iterative solvers. Direct solvers often need more memory and are not as fast as iterative solvers. However, they are quite robust and may be the better choice if the iterative solvers suffer from convergence problems.
- Spooles is a direct equation solver allowing parallel calculations and being pretty robust. Therefore, if uncertain, better choose this equation solver. However, Spooles generally needs more memory than iterative solvers.
- Multfront is a direct solver of the multifrontal type. It is easy to set up and behaves well for most problems.
- MUMPS is a general purpose direct solver of the multifrontal type. It provides a lot of parameter settings to allow the best fitting to your problems needs.
- LDLT is a direct solver which uses a Gaussian Algortihm. It is comparatively slow for big problems.
- Iterative Scaling is often fast but less robust compared to Spooles. Until present, parallel calculations are not possible with Iterative Scaling.
- Iterative Cholesky is often fast but less robust compared to Spooles. Until present, parallel calculations are not possible with Iterative Cholesky.
- PETSC is an iterative solver speacially designed to deal with large systems. It scales very effectively in parallel and is the best choice for large problems.
- GCPC is an iterative solver of the pre-conditioned conjugate gradient type. It scales well in parallel and is also usable for large problems.