AMBER Archive (2003)

Subject: Re: Increase lastist in the &cntrl namelist

From: Ioana Cozmuta (ioana_at_nas.nasa.gov)
Date: Tue May 13 2003 - 15:26:48 CDT


Hi,

I've run amber here on the NASA clusters with the compilation
modifications I've mentioned. These are Origin 2000 or 3000 clusters with
500-1000 CPU's but for amber I've used only up to 128.

Let me know if you need more info,
Ioana

On Tue, 13 May 2003, Dodi Heryadi wrote:

> Hi,
>
> Just out of curiosity (especially for Guille): did you run amber on one of
> the NCSA's machines? I'd like to know this since it might be useful for
> other Amber users at NCSA. We have 64 bit version of Amber on both SGI
> Origin and IBM p690 machines.
>
> thanks,
> --Dodi
> http://www.ncsa.uiuc.edu/~dodi
>
> At 11:58 AM 5/13/2003 -0700, Ioana Cozmuta wrote:
> >Hi,
> >
> >You say that you have to deal with a big system so I think that only
> >increasing the value of lastist or MAX_ISTACK will not help. For system
> >the size you mention, the allocate command is normally trying to allocate
> >few GBytes of memory that is simply too much for the 32 bit addressing
> >space with which the amber7 package was built. [-n32 compiler option vs.
> >-64; normally 32 bit address space is plenty of address space for small
> >to medium problems]. Additionally, the MPI library runs out of 32 bit
> >memory space at about 100 or so MPI processes. So, if you want to do
> >problems with largish numbers of processors, you'll need to take some more
> >action.
> >
> >So you can do the following:
> >1. Change in Machines/Machine.sgi_mpi all the -n32 compile
> > options with -64.
> >
> >2. Modify the mm_pbsa/Makefile to add a default compile line
> > for C modules. Otherwise, it used the default C compile which used
> > -n32.
> >
> >The LEaP X11 utility is still built with -n32. Since this
> >is a GUI setup utility, I didn't switch it to use -64.
> >Then
> >cd amber7/src
> >rm -fr ../exe
> >make clean |& tee clean.out
> >make install |& tee install.out
> >
> >THis is what I did in amber7 on an IRIX system.
> >
> >Hope it is of help,
> >Ioana
> >
> >On Tue, 13 May 2003, GUILLERMINA L ESTIU wrote:
> >
> > >
> > > Hi,
> > > I am trying to run sander (minimization) (amber7) for a sistem of 50000
> > > residues (180000 atoms)
> > > In the first run I am minimizing only the solvent water molecules.
> > >
> > > The minimization stops with a message "Exceeding lastist in get_istack"
> > > The message belongs to istak.f subruitine. But it is not clear to me,
> > if I have
> > > fo Increase MAX_ISTACK in sizes.h and recompile, or there is a variable
> > > "lastist" which I can handle from the input.
> > >
> > > Thanks
> > > Guille
> > > Dr Guillermina Estiu
> > > Chemistry Department
> > > Pennsylvania State University
> > > PA 16802
> > >
> > >
>
>