[CP2K:794] Re: parallel distribution of data

[Nichols A. Romero]

Matt et al.,

I don't think its a bug. I think that the optimization algorithm is simply
running out of memory.
See line 570 in distribution_methods.F

There is an integer array called:
cost_ele_pair(nele,nele)

nele is equal to the number of atoms, for 8192 water molecules this is quite
large. When I work out the math,
the integer array size is 2.25 GB per core. This is a lot memory just for
optimization.

Could the 2D_MOLECULAR_DISTRIBUTION TRUE be a workaround? Do I define the
molecule through
some connectivity file?



On Tue, Mar 11, 2008 at 5:18 PM, Matt W <MattWa... at gmail.com> wrote:

>
> No, thankfully it's not related.  Looks like another overflow
> though...thanks for pushing things!
>
> Matt
>
> On Mar 11, 9:03 pm, "Nichols A. Romero" <naro... at gmail.com> wrote:
> > Guys,
> >
> > This might be related.
> >
> > I just tried H2O-8192.inp on 512 & 1024 cores and I received the error
> > below.
> >
> > BTW, I don't have this problem with the H2O-4096.inp
> >
> > *
> >  *** ERROR in distribute_molecules_2d (MODULE distribution_methods)  ***
> >  *
> >
> >  *** The memory allocation for the data object <cost_ele_pair> failed.
> The
> > ***
> >  *** requested memory size is -1835008 Kbytes
> > ***
> >
> >  *** Program stopped at line number 570 of MODULE distribution_methods
>  ***
> >
> >  *
> >  *** ERROR in distribute_molecules_2d (MODULE distribution_methods)  ***
> >  *
> >
> >  *** The memory allocation for the data object <cost_ele_pair> failed.
> The
> > ***
> >  *** requested memory size is -1835008 Kbytes
> > ***
> >
> > On Tue, Mar 11, 2008 at 10:14 AM, Nichols A. Romero <naro... at gmail.com>
> > wrote:
> >
> >
> >
> > > So can I. I am running that and some larger systems.
> >
> > > Many thanks to Matt!
> >
> > > On 3/11/08, Teodoro Laino <teodor... at gmail.com> wrote:
> >
> > > > I can confirm that with this bug-fix H2O-2048 is running nicely.
> >
> > > > teo
> >
> > > >  On 11 Mar 2008, at 12:09, Matt W wrote:
> >
> > > >  (not well tested though).
> >
> > > --
> > > Nichols A. Romero, Ph.D.
> > > DoD User Productivity Enhancement and Technology Transfer (PET) Group
> > > High Performance Technologies, Inc.
> > > Reston, VA
> > > 443-567-8328 (C)
> > > 410-278-2692 (O)
> >
> > --
> > Nichols A. Romero, Ph.D.
> > DoD User Productivity Enhancement and Technology Transfer (PET) Group
> > High Performance Technologies, Inc.
> > Reston, VA
> > 443-567-8328 (C)
> > 410-278-2692 (O)
> >
>


-- 
Nichols A. Romero, Ph.D.
DoD User Productivity Enhancement and Technology Transfer (PET) Group
High Performance Technologies, Inc.
Reston, VA
443-567-8328 (C)
410-278-2692 (O)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20080331/312fcb31/attachment.htm>