Pardiso size limitation?

Pardiso size limitation?

Does Pardiso have a size limitation? My number of equations is about 240,000, and Pardiso returned an error code -2 (not enough memory). The the packed sparse matrix size is only about 68MB. My machine is a single processor P4 with 2GB RAM and 100GB disk space. Also I don't have a problem to solve a model with 180,000 equations.

Is there anyway I can overcome the problem?


4 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.


I need some more information and thenI will call in the big guns to answer. You don't say what operating system you are using (32-bit or 64-bit), the processor or the version of MKL you are using.

Please add this information and we will try to respond.



The question was also posted on the intel premier support (issue # 337721), which included detail configuration information (W32, MKL8.0.017 etc). There are a couple of e-mail exchanges on the issue, but the problem is still under investigation. We have done some further experiments ourselves, and found out the Pardiso solver is not working too well with ATL application. For example, the Pardiso solver works fine for the problem large matrix in a console application, even with an extra 500MB overhead array being allocated in the application. However, it would fail in an ATL application without any overhead array. Given the fact that ATL may preserve up to 200MB space, we are still puzzled on what happened on the other 300MB space. If you want to replicate the problem, you can download the sample projects following the instruction of one of the e-mails in issue# 337721.

We have also found 2 problems with MKL8.0.1.003, and they are related to ATL applications too.

  1. Link error with VS.Net 2005 (issue# 340092)
  2. Pardiso solver crashes with VS.Net 2003 (issue 340075)

Thanks for your attention! The solver is great, it just needs some cleanups.



I faced with similar problem in PARDISO. I described it in the issue #331593.

I hope that you will improve this "features" :)


Leave a Comment

Please sign in to add a comment. Not a member? Join today