Different ncpus: convergence issue in OpenSeesSP

This forum is for issues related to parallel processing
and OpenSees using the new interpreters OpenSeesSP and OpenSeesMP

Moderator: selimgunay

Post Reply
zanakarimi
Posts: 27
Joined: Mon May 14, 2012 11:07 am
Location: University of Colorado Boulder
Contact:

Different ncpus: convergence issue in OpenSeesSP

Post by zanakarimi »

Hi everybody,
I have a question on parallel processing using OpenSeesSP. I created a very big Soil-Foundation-Structure model which has more than 200,000 DOFs. It works fine with 1, 2, and 6 ncpus on my PC, but it stops converging at some time steps when I use different numbers of processors. I am guessing the problem is with automatic domain decomposition. I need to speed up my analysis since it takes more than 48 hrs with ncpus=6; I can run my analysis on NEEShub HPCs with up to 48 processors, if I can solve this issue. In my model, the total number of elements is even. I have both SSPbrick and brickUP elements. I also tied lots of elements using Equal DOF command (e. g., between 3 and 4 DOFs elements).
Does any body have any idea what the problem can be?

Thanks in advance,
Zana,
mo_zarrin
Posts: 8
Joined: Mon Nov 16, 2009 4:50 am
Location: Kntu University of Technology

Re: Different ncpus: convergence issue in OpenSeesSP

Post by mo_zarrin »

Hi, could you please tell me which materials you have used?
Don't you have problems with update material stage command?
When the OpenSeesSP shows a problem like this, it seems that it will not work with the model. sorry to say this.
Just frank can help to these problems to be solved, but to some reasons he is not that much active to solve problems associated with openseesSp
mo_zarrin
Posts: 8
Joined: Mon Nov 16, 2009 4:50 am
Location: Kntu University of Technology

Re: Different ncpus: convergence issue in OpenSeesSP

Post by mo_zarrin »

Can you say why you have noted that you have even number of elements? Does this matter?
zanakarimi
Posts: 27
Joined: Mon May 14, 2012 11:07 am
Location: University of Colorado Boulder
Contact:

Re: Different ncpus: convergence issue in OpenSeesSP

Post by zanakarimi »

mo_zarrin,
I am using PDMY02 for the soil material. You are right; there was a problem with update material stage due to a missing part of the code in OpenSeesSP. But, Frank solved the problem long time ago and updated the .exe file. I suggest you to download the latest version of .exe. There should be no problem with it.
mo_zarrin
Posts: 8
Joined: Mon Nov 16, 2009 4:50 am
Location: Kntu University of Technology

Re: Different ncpus: convergence issue in OpenSeesSP

Post by mo_zarrin »

I am using the latest version.
I am going to use PDMY and PIMY material models. At the moment, I have a model which has only PIMY material and it shows update material stage error in large models.
Could you please for instance change your material model to PIMY and repeat your analysis and report the results to me, or sending an example large model for me to this address: mo_zarrin@yahoo.com
also, please check this link:
http://opensees.berkeley.edu/community/ ... =7&t=62049
Post Reply