[PETSC #3487] Re: PETSc survey
bsmith at mcs.anl.gov
Fri Jun 23 16:14:55 CEST 2000
Thanks for your response.
Note that we do now, in 2.0.28, provide Saad's ILU for the SeqAIJ matrices;
it is controlled by the usual -pc_type ilu with -pc_ilu_use_drop_tolerance
On Fri, 23 Jun 2000, Christophe Geuzaine wrote:
> William Gropp wrote:
> > Let us know what you have done with PETSc!
> First of all, thanks for the great job you did with PETSc. At the
> computational electromagnetics group of the University of Liege
> (Belgium), we use it as an alternative solver for large problems
> (Sparskit is used for smaller, sequential problems) in our code, GetDP
> (http://www.geuz.org/getdp/). Our main research concerns the
> mathematical and numerical modeling of electromagnetic phenomena in
> nonlinear three-dimensional systems taking mechanical and thermal
> effects into account.
> As you can see on http://www.geuz.org/getdp/ only sequential versions of
> the code are distributed for the moment. We are still in the testing
> phase for the PETSc-linked version (which already runs quite nicely on
> an SGI Origin as well as on a DEC Alpha cluster).
> > Please send us a short description of how you have used PETSc. If you have
> > a paper, web page, or other document describing your project and how you
> > used PETSc, please send that. Also let us know what you liked most and
> > least, and what features you would like to see in future versions of
> > PETSc.
> + good performance and scalability
> + source code
> + documentation
> - (too?) big (time stepping, non linear stuff, X interface, ...).
> - lack of ILUTP and some other newer ILU precond (cf. works from Saad et
> - what about OpenMP?
> - (too?) short documentation about partitioning (section 3.5 in the
> Christophe Geuzaine
> Tel: +32-(0)4-366.37.10 mailto:Christophe.Geuzaine at ulg.ac.be
> Fax: +32-(0)4-366.29.10 http://www.geuz.org