OpenSees Crashes on Linux

jwaugh · Post by **jwaugh** » Tue Mar 29, 2005 1:51 pm

Frank, I built OpenSees on Linux and then when a friend was using it, it ran for a few hours and said "Sound server fatal error: CPU overload" It was coming from artsd. Why does this happen?

If it means anything when I built OpenSees I used 64 bit for as much as I could since my computer uses Opteron processors. Is this the source of my trouble and I need to rebuild the program using the 32 bit peices everywhere. Thanks

Post by **fmk** » Wed Mar 30, 2005 6:49 am

yes he sent the file ... just trying to get the next release out so i have not yet looked at it .. can't see how OpenSees is responsible for this one .. have
a look at the this thread on a possible fix.

http://lists.debian.org/debian-qt-kde/2 ... 00151.html

jwaugh · Post by **jwaugh** » Thu Mar 31, 2005 4:00 pm

Frank, is there a memory leak anywhere in OpenSees, because I am watching OpenSees run and it is just slowly eating up my memory, first the physical and then the swap file. Thanks for your assistance.

Post by **fmk** » Fri Apr 01, 2005 7:21 am

there may be .. just fixed a serious one with the new version of SuperLU
that we started to use .. if this is not it send me the i/p files .. better yet
use bugzilla to report a bug, then do a follow up & attach the scripts.

jwaugh · Post by **jwaugh** » Fri Apr 01, 2005 9:35 am

Ok, I put it into Bugzilla, it is bug #18. Sorry I can't give you a idea of where the problem is. I also attached the files where we observed it. I also believe the problem exists in Windows versions, however, it is so much slower to run that I won't be sure for a while.

jwaugh · Post by **jwaugh** » Fri May 13, 2005 9:27 am

Frank, I rebuilt OpenSees using the source from the CVS, as of 12:05 CST today. The build went fine, but when I tried to run the files I posted in the Bug report 18, I got a seg fault. It echoed to the screen:

Nodes Created
Soild Models Created
Soil Elements Created
Element Self Weight Applied
Segmentation Fault

This file hasn't changed since I ran it on on my build of the 1.6.1, so I don't think it is an error in the input. Thanks for all your help Frank.

Post by **fmk** » Fri May 13, 2005 11:39 am

the memory problem was fixed .. the i/p files are on my machine at home .. i was doing a profile of the example to send to boris .. send me a tarball or zip archive if you want me to look today, otherwise it's going to be sometime when i am at home.

jwaugh · Post by **jwaugh** » Mon May 16, 2005 11:26 am

So what is causing the seg fault if the memory problem has been fixed?Do you still want me to email you the files, I didn't see the reply until today. Thanks

Post by **fmk** » Mon May 16, 2005 3:58 pm

i don't know .. send the files.

jwaugh · Post by **jwaugh** » Fri May 20, 2005 8:50 am

Frank,

after a full clean rebuild of OpenSees the program doens't seg fault. But after it has run a few steps in the push over analysis it quits saying

"Root must be bracketed in ZBRENTstress"

What does this mean? I am not understanding that at all. THanks for the assistance.

Boris · Post by **Boris** » Mon May 23, 2005 10:02 am

Hello There,

This function zbrent is used in some of our algorithms so I suspect you are using something we have done...

Can you be more specific about what are you doing and what models you use, ...

BOris

Liam · Post by **Liam** » Fri May 27, 2005 9:07 am

Hi Boris

Liam here, Jon has been running my OpenSees model that we have conversed about previously. It is a pushover test of a pile embedded in soil and loaded at its top. It is a 3 dimensional model using brick elements for the soil and beam column elements for the pile. I am using the Template Elastic Plastic material models with Von Mises yield and potential surfaces. At each depth a single soil node is slaved to a single pile node and the top of the pile is loaded horizontally.

Similar to some of the the work you have been doing, and what we have discussed before

Hope this helps

Liam

Boris · Post by **Boris** » Wed Jul 06, 2005 9:20 am

Hello Liam,

Sorry about delayed answer, just got back from a month long trip to Europe...

OK, so you are pushing a pile, try first with elastic material just to make sure you have all the elements properly set...

Then go woth some simple elastic-plastic, vonMises would do for both yield and potential surfaces, make it perfectly plastic...

To tell you the truth, last time I had problems with ZBRENT was in '93/'94, so this might take some doing to debug...

Boris

Liam · Post by **Liam** » Wed Jul 06, 2005 1:51 pm

Hi Boris

Ran with elastic and with elastic perfectly plastic soil materials and seems to be OK, doesnt crash straight away. Must then be something to do with the use of the Armstrong Fredericks hardening rule. Maybe i have the syntax incorrect as it has changed before.

Maybe you could post the syntax that i should use and then i could run it again and see if that is anything to do with the problem

Thanks
Liam

Boris · Post by **Boris** » Fri Jul 08, 2005 2:04 pm

Hello Liam,

look at my web site, not much changed in last half year or so...

http://sokocalo.engr.ucdavis.edu/~jerem ... index.html

There is a line with OpenSes Commands (in item 1. subitem 1).

try to extract one single element and test with this model.

In the mean time check this web site:
http://cml00.engr.ucdavis.edu/~guanzhou/NEES_Project/

for models that are similar to what you do...

Boris

The OpenSees Community

OpenSees Crashes on Linux

OpenSees Crashes on Linux

zbrent

Model Information

OK, but I need more details

Armstron-Frederick...