Re: Testing matrix solvers (fem)
- From: "Evgenii Rudnyi" <usenet@xxxxxxxxx>
- Date: 25 Feb 2007 09:47:39 -0800
I'm developing some sparse matrix solvers (iterative) for FEM
applications, and hence need some robust way to compare the accuracies
of different solver and preconditioner types. There must be a a free
collection of _unassembled_ elemental matrices with right-hand sides
and solutions. Preferably resulting in large and sparse _unsymmetric_
stiffness matrices. I checked the Harwell-Boeing collection, but there
were no suitable cases, only symmetric. Does anyone have experience
with such a thing?
You can have a look at Oberwolfach benchmark collection. Well, most
matrices are symmetric though.
http://www.imtek.uni-freiburg.de/simulation/benchmark/
Yet, you can find here really huge matrices from the bone models. If
you like challenges, try this
http://www.imtek.uni-freiburg.de/simulation/benchmark/wb/bone/
If you know someone with ANSYS nearby, ask him generate for you model
you like and then use MOR for ANSYS to dump matrices.
Best wishes,
Evgenii
http://ModelReduction.com/
.
- References:
- Testing matrix solvers (fem)
- From: maxsmith12345
- Testing matrix solvers (fem)
- Prev by Date: Re: Computing the Sine and Cosine Integrals
- Next by Date: NR: svdcmp
- Previous by thread: Re: Testing matrix solvers (fem)
- Next by thread: Call for Paper-Bulletin of Korean Mathematical Society
- Index(es):