FEAP User Forum
FEAP => Parallel FEAP => Topic started by: Markus on July 12, 2012, 10:56:08 PM
-
Hello,
I am trying to calculate a coupled system of four degress of freedom in a parallel FEAP/PETSc run. The job converged very well while running in serial and using the PARTition command to apply a staggered scheme. How can I use this command in a parallel run? To simply write the PARTition table in the solve.file seems not to work. Is it in principle possible or do I have to look for a similar option of the solver I am using?
Thank you,
Markus
-
Markus,
I do not believe that partitions were ever implemented in parFEAP. I think you may have to do a little bit of programming to get this to work properly in parallel. The basic issue is that each equation partition (to be differentiated from domain decomposition partition) needs to have its own parallel stiffness matrix and RHS. This can probably be programmed without too much effort but before doing so, you should have a careful look at the source to see how PARTitions are handled.
-
You also may want to look at petsc field split preconditioner PCFIELDSPLIT
http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/PC/PCFIELDSPLIT.html
This allows you to partiton the Jacobian and solve each block separately, I think you could be able to use it to do what PARTition does.
-
Thanks a lot, this seems to be exactly what I have been looking for.