OpenSeesSP potential parallel computing problem

[Ozgun]

Hi,

I am using Manzari-Dafalias model with parallel processor. I have a very simple 1 column of soil on which I am doing gravity loading and dynamic time history analysis and it works in sequential computing without any problem. When I run it in parallel computing using SuperLU solver, I am getting the following warning in the beginning of the analysis.

"WARNING: MaterialStageParameter::setDomain() - no effect with material tag 1". And this warning appears for all of the material tags (in total 20).

I think that there is a problem with domain decomposition, or there may be problem with sending and receiving information between the cores. Does anybody have clearer explanation?

Also, while trying to do parallel computing, "NDMaterial::getTangent -- subclass responsibility" flows on the screen. Does anybody know what is the meaning of this?

Edit: In addition to the above stated issues, i found out that in case of using SuperLU with 4 cores for the case I have, the memory distribution is not balanced. The memory use of 1 core is 200 Mb, whereas the other 3 are using only 4 Mb of memory. In case of using Mumps solver, the memory distribution is equal but Mumps brings other bunch of problems in terms of convergence during gravity analysis (reported by others in too) therefore I am trying to avoid Mumps as much as possible.

In case of using Mumps, the error i get is following, (Note: the same analysis is working fine in sequential)

WARNING MumpsParallelSolver::solve(void)- Error -10 returned in substitution dmumps()
cause: Matrix is Singular Numerically
WARNING NewtonRaphson::solveCurrentStep() -the LinearSysOfEqn failed in solve()
DirectIntegrationAnalysis::analyze() - the Algorithm failed at time 5000
OpenSees > analyze failed, returned: -3 error flag
Gravity Elastic Execution time: 0 hours and 0 minutes and 1 seconds.
First run done.

Thank you,

Ozgun