Big Calculations (was [Wien] Allocation error 494)

[L. D. Marks]

I can only give a partial answer, and I have not had time to really get to
grips with it. I've never used mpi before so these might be trivial
problems which I should find the answer for. So far, Several issues:

a) mpirun not being found -- probably requres /opt/mpich/intel/bin (or gnu
version?) to be defined in .bashrc and/or .cshrc. There might be some
issues in exporting of variables since Wien is csh based and the available
mpirun scripts I have are bash based.

b) machines not being found. There are probably some configuration files
that need to be setup correctly and, at least in my first attempts, it was
defaulting to a p4 procgroup (-p4pg) rather than using machines. (Probably
an issue with how mpirun is called in lapwXpara.)

c) /opt/mpich/intel/share not being correctly propogated to all nodes (mpi
configuration issues).

d) A Sigsev on one node, source unknown. Probably related to incorrect mpi
configuration although it might be within lapw1_mpi.

If anyone has worked out configuring mpi on an Intel system it might be
useful to post some information to the list.

On Tue, 22 Feb 2005, Fred Nastos wrote:

>
> I heard back from the group I helped compile the code, and they
> too have problems with the mpi runs.  As far as I know the code
> compiles cleanly, but fails on execution (sorry, I have no more info).
>
> Can you tell us the problem you encountered?
>
> Thanks.
>
> > N.B., this would be for calculations which even with RKMAX=4.5 are taking
> > about 1.5G so an adequate calculation would be >> 2G.
> >
> > N.N.B., it might be necessary to modify lapw1 to stick some of the huge
> > arrays on disk, e.g. using mmap calls.
> >
> > On Tue, 22 Feb 2005, Michael Gurnett wrote:
> > > As far as I know, unless your running on a 64 bit machine, the most the
> > > operating system can allocate in a single chunk is 2GB.
> > >
> > > Michael
> > >
> > > ----- Original Message -----
> > > From: "Rebecca Nicholls" <rebecca.nicholls at st-catherines.oxford.ac.uk>
> > > To: <wien at zeus.theochem.tuwien.ac.at>
> > > Sent: Tuesday, February 22, 2005 1:59 PM
> > > Subject: [Wien] Allocation error 494
> > >
> > > > hello everyone
> > > >
> > > > I was wondering if anyone had ever had trouble with allocation
> > > > errors on a 4G machine.  I sucessfully got a large job to start
> > > > on a 2G machine, using 1.5G RAM and 1G swap.  When I try to run
> > > > this same job on a 4G machine, I get allocation errors.  It seems
> > > > I'm having trouble using more than 2G out of the 4G of ram, and
> > > > the program is not using any swap.
> > > >
> > > > I've changed the stacksize to unlimited:
> > > >
> > > > cputime         unlimited
> > > > filesize        unlimited
> > > > datasize        unlimited
> > > > stacksize       unlimited
> > > > coredumpsize    0 kbytes
> > > > memoryuse       unlimited
> > > > vmemoryuse      unlimited
> > > > descriptors     1024
> > > > memorylocked    unlimited
> > > > maxproc         7168
> > > >
> > > > The 2G machine is a dual-processor (although I was only running
> > > > the job on one processor) Xeon 2.2Gz running red hat linux (kernel
> > > > 2.4.18-3smp).
> > > >
> > > > The 4G machine is a pentium 4 running suse (kernal 2.4.21-260-smp4G).
> > > >
> > > > Thanks a lot
> > > > Best wishes
> > > >
> > > > Rebecca
> > > >
> > > > _______________________________________________
> > > > Wien mailing list
> > > > Wien at zeus.theochem.tuwien.ac.at
> > > > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> > >
> > > _______________________________________________
> > > Wien mailing list
> > > Wien at zeus.theochem.tuwien.ac.at
> > > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> >
> > -----------------------------------------------
> > Laurence Marks
> > Department of Materials Science and Engineering
> > MSE Rm 2036 Cook Hall
> > 2220 N Campus Drive
> > Northwestern University
> > Evanston, IL 60201, USA
> > Tel: (847) 491-3996 Fax: (847) 491-7820
> > email: L - marks @ northwestern . edu
> > http://www.numis.northwestern.edu
> > -----------------------------------------------
> >
> > _______________________________________________
> > Wien mailing list
> > Wien at zeus.theochem.tuwien.ac.at
> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>

-----------------------------------------------
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60201, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L - marks @ northwestern . edu
http://www.numis.northwestern.edu
-----------------------------------------------