[Wien] reduce memory usage

Pavel Ondracka pavel.ondracka at email.cz
Mon Feb 9 11:58:07 CET 2015


On Mon, 2015-02-09 at 11:33 +0100, Pavel Ondracka wrote:
> On Mon, 2015-02-09 at 01:36 -0800, Yundi Quan wrote:
> > Hi, 
> > Is there a way of reducing the memory usage? I have a case with 72
> > atom per unit cell. And my cluster has 8 GB member per node and each
> > node has 8 cores. I submitted the job and got the error message saying
> > that insufficient virtual memory. Can someone give me any suggestions?
> 
> Hi,
> 
> I don't know about any way to reduce the total memory required, you can
> however get around it by smarter parallelisation, e.g. if you use more
> nodes with mpi the memory requirement per process (and hence per core
> and node) decreases, so just try to tweak you .machines file until the
> calculation fits.
> 
> Also if you have hyperthreading on and you can disable it, do it.
> 
> Best regards
> Pavel
> > 
Forgot to mention most important thing, the RKMAX value is mainly
accountable for memory requirements, but always do convergence tests
when you change it (this actually may not save you from the need to do
one calculation with the higher RKMAX, unless you can check the
convergence of the parameter you are interested in in some smaller
similar system).

BTW if you set correct NMATMAX value during compilation, Wien reduces
RKMAX automatically so the calculation always fits and prints a warning
in log.
Too high NMATMAX is not so much a problem when your calculation require
more memory than total virtual memory because the process just crashes
and notices you. However when you need more memory than physical memory
but the calculation still fits in the swap space, you get no warning,
just really slow speed.

Best regards
Pavel

> > 
> > Yundi
> > _______________________________________________
> > Wien mailing list
> > Wien at zeus.theochem.tuwien.ac.at
> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> > SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
> 
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html




More information about the Wien mailing list