[gpaw-users] [gpaw-developers] Memory usage

[Nichols A. Romero]

Is there someone on a platform where the "Initial overhead" actually works?

On my Mac Book, I always get zero.

On BG/P, I always get O(infinity).


----- Original Message -----
> Thank you for your input & explanation. It is true that the "Initial
> overhead" is too high in AIX power 6 machine (few thousand of MiB) and
> is small in Cray XT5.
> 
> I think it is necessary to well predict memory usage in large scale
> simulation.
> I don't want to be in the situation where I ask for few hundred cores
> then the job hang up due to the lack of memory
> or in the situation where I ask too many cores unnecessarily. In both
> case, time and resources are wasted.
> 
> Best regards,
> Duy
> --------------------------------------------------
> Duy Le
> PhD Student
> Department of Physics
> University of Central Florida.
> 
> "Men don't need hand to do things"
> 
> 
> 
> On Thu, Sep 23, 2010 at 1:27 PM, Ask Hjorth Larsen
> <askhl at fysik.dtu.dk> wrote:
> > A bit more info
> >
> > On Thu, 23 Sep 2010, Ask Hjorth Larsen wrote:
> > (memory related ramblings)
> >
> > Here's a script which will test the previously mentioned system
> > calls:
> >
> > http://www.student.dtu.dk/~ashj/opendir/memtest.py
> >
> > The maxrss() function actually tries to call several other
> > functions, and
> > returns the first nonzero result. These functions are accessed
> > directly by
> > calling the _VmB function with a number of different 'keys'.
> >
> > Here's the output on my computer:
> >
> > [askhl at newton ~]$ python memtest.py
> > 9.44921875
> > 137.46875
> > VmHWM: 137.46875
> > VmRss: 0.0
> > VmPeak: 150.69140625
> > VmSize: 150.69140625
> >
> > Here's the output on the Niflheim login node:
> >
> > [askhl at thul ~]$ python memtest.py
> > 14.359375
> > 142.37890625
> > VmHWM: 142.37890625
> > VmRss: 0.0
> > VmPeak: 326.02734375
> > VmSize: 326.02734375
> >
> > maxrss returns the same as VmHWM in this case because that is the
> > first
> > nonzero result among the function calls.
> >
> > Regards
> > Ask
> >
> 
> _______________________________________________
> gpaw-users mailing list
> gpaw-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users

-- 
Nichols A. Romero, Ph.D.
Argonne Leadership Computing Facility
Argonne National Laboratory
Building 240 Room 2-127
9700 South Cass Avenue
Argonne, IL 60490
(630) 252-3441