[Trilinos-Users] Rythmos 1D FEM example fails to run in parallel

[Peles, Slaven Export License Required - US UTRC]

Ross, thanks for your reply. I was actually looking at "1DefmInterface.C" and it seems as if there was at least an attempt to make this run in parallel. Anyhow, do you (or anybody else) know of a Rythmos example with adaptive BDF solver that is designed to run in parallel, and which can be freely shared? I'm trying to learn how to use Rythmos for parallel computations and having a working example would certainly help. 

Thanks,
Slaven


> 
> Slaven,
> 
> From looking at:
> 
>     packages/rythmos/example/epetra/1Dfem/CMakeLists.txt
> 
> And:
> 
>     http://trilinos.sandia.gov/cdash/viewTest.php?onlypassed&buildid=205481
> 
> it does not appear that this example was meant to run in parallel.
> 
> - Ross
> 
> 
> 
> >
> > Hi,
> >
> > I wonder if anyone successfully ran Rythmos 1D FEM example in parallel?
> > (packages/rythmos/example/epetra/1Dfem, version 10.0.4)
> >
> > I compiled this example on a Linux box with 8 cpu cores. The code runs
> > fine on one cpu; when I run it on 2-4 cpus the results come with a
> > large error compared to the "exact" solution and computation diverges
> > if I try to run it on a larger number of cpus. I tried adaptive BDF and
> > fixed step backward Euler methods, with and without NOX, and with
> > different tolerances, but the same problem occurs.
> >
> > If anyone has a suggestion how to troubleshoot this, please let me
> > know. Also, let me know if you want me to supply any additional
> > information.
> >
> > Thanks,
> > --
> > Slaven Peles
> >