[Trilinos-Users] Sparse Matrix Memory Requirements Estimation

Williams, Alan B william at sandia.gov
Mon Jun 9 08:19:03 MDT 2008

I'll look into this. Perhaps the temporary memory that FECrsMatrix uses to hold the overlapping matrix contributions isn't being cleared. This could explain some overhead, but probably not that much.
How large is your executable?
Certainly it is worth running with valgrind or purify to check for memory leaks.

From: trilinos-users-bounces at software.sandia.gov [mailto:trilinos-users-bounces at software.sandia.gov] On Behalf Of Davood Ansari
Sent: Saturday, June 07, 2008 10:42 AM
To: trilinos-users at software.sandia.gov
Subject: [Trilinos-Users] Sparse Matrix Memory Requirements Estimation

Hi all

I am assembling a kind of FECrs_Matrix in Epetra.
I wonder if it is occupying too much memory.
After the assembling process is finished and all other memory resident arrays are deleted
and right before quitting main(), I use a getchar() to get the program hanging before I allow it to quit.
At this time, it occupies 1.3GB of my ram while it reports the number of non-zeros
(using NumGlobalNonzeros()) to be 32992643 entries.
My estimate is that this should only occupy about 32992643*(8 + 4) + some overhead
of bytes which equals to 395 Mega Bytes !!!!

Am I right and should I look for any memory leakage or some thing there ?
The mentioned code is not parallel.

Please comment

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://software.sandia.gov/mailman/private/trilinos-users/attachments/20080609/a167b978/attachment.html 

More information about the Trilinos-Users mailing list