Re: 128+ proccessor jobs on 64bit Widows hang upon completion of the simulation.

From: Eric Bohm (ebohm_at_cs.uiuc.edu)
Date: Mon Jul 30 2007 - 11:13:09 CDT

Yes.

You may also need to revise your arch (and possible Make.charm) to
point to the new charm target as the new charm will be in
net-win64-tcp and the old in net-win64.

 

Ilya Chorny wrote:
> Will I have to rebuild NAMD as well?
>
>
>
> On 7/30/07, *Eric Bohm* <ebohm_at_uiuc.edu <mailto:ebohm_at_uiuc.edu>> wrote:
>
> add tcp as a parameter after the machine specification.
>
> ./build charm net-win64 tcp
>
> Assuming net-win64 was your target before.
>
>
> Ilya Chorny wrote:
> > How do I do that?
> >
> > Thanks,
> >
> > Ilya
> >
> >
> > On 7/30/07, *Eric Bohm* <ebohm_at_uiuc.edu <mailto:ebohm_at_uiuc.edu>
> <mailto:ebohm_at_uiuc.edu <mailto:ebohm_at_uiuc.edu>>> wrote:
> >
> > You could try using tcp in your charm build instead of udp
> (the
> > default).
> >
> > Ilya Chorny wrote:
> > > I ran the apoa1 benchmark on my 64 bit Windows cluster
> running NAMD
> > > with 2,4,8, 16, 32,and 64 processors and the jobs exited
> at the
> > end of
> > > the simulation, however 128, and 256 processor jobs just
> hang at the
> > > end of the simulation. All the output files were correctly
> written.
> > > The network I am running is GigE. Any ideas?
> > >
> > > Thanks!!!!
> > >
> > > Ilya
> > >
> > > --
> > > Ilya Chorny Ph.D.
> > > Stroud Lab
> > > UC San Francisco
> >
> >
> >
> >
> > --
> > Ilya Chorny Ph.D.
>
>
>
>
> --
> Ilya Chorny Ph.D.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:45:01 CST