Unique Problem with OpenseesMP

[shawnfa]

Hi everyone,

I've been trying to use OpenseesMP to analyze two different time history records simultaneously on one model.

It appears to run correctly, and I get an "Analysis (1) completed" and "Analysis(2) completed", but when I look at the output, both analyses have the exact same output. This output is not equal to either model if I run them sequentially. Also, if the time histories have a different number of steps OpenseesMP completely stops when the first completes. So clearly the analyses are not being solved in parallel and something is going very wrong.

I've tried using the -par command as well as using the additional commands and a "foreach" loop as described in the "Using the OpenSees Interpreter on Parallel Computers" document by Frank McKenna and Gregory L. Fenves, and neither work. I have also never heard of anyone having a problem like this. MPICH2 appears to works fine, as does OpenseesSP.

Any help or ideas would be greatly appreciated.

[fmk]

what solver and numberer are you using.

[shawnfa]

I'm using the Mumps solver and ParallelRCM numberer.

However I have tried others with no luck.

[fmk]

you don't want a parallel solver and numberer .. it treats the 2 seperate models as one model .. you want to keep the analysis and models seperate .. use the same system and solver that you would use in the case of a sequential script.

[shawnfa]

That was the problem.

Thank you very much!