[CPMD-list] Memory Requirement Problem

[Axel Kohlmeyer]

On Fri, 6 Jul 2007, Maurice de Koning wrote:

MK> Dear all,

dear maurice,

MK> I'm trying to run a TDDFT electronic spectra calculation using the
MK> parallel version (compiled with ifort 9 and mkl 8.0)  of CPMD on an
MK> SGI-Altix cluster with 12 Itanium2 CPU's and 24 Gb of memory. After
MK> the initial Lanczos diagonalization, which runs fine, I get the
MK> following type of message for all CPU's:
MK> 
MK> **************************************************************************************
MK>  CHEMICAL POTENTIAL =                             2.3082842040 EV
MK>  ================================================================
MK>  ==                END OF STATE INITIALIZATION                 ==
MK>  ================================================================
MK> 
MK>  ***   SPECTRA| SIZE OF THE PROGRAM IS  406432/48887376 kBYTES ***
MK>  ==                      SINGLET STATES                        ==
MK>  NUMBER OF STATES TO BE INITIALIZED                             5
MK>  TOTAL NUMBER OF TEST VECTORS                                  10
MK>   MEMORY| MEMORY REQUIRED:   303161702  WORDS
MK> 
MK> 
MK>  PROGRAM STOPS IN SUBROUTINE MEMORY| TOO BIG VALUE [PROC=   3]
MK> MPI: On host topazio, Program /home/esmeralda/dekoning/dekoning/cpmd/CPMD-3.11.1/cpmd.x, Rank 3, Process 10585 called MPI_Abort(<communicator>, 999)
MK> 
MK> MPI: --------stack traceback-------
MK>  ****************************************************************************************
MK> 
MK> Evidently, this is a memory problem. Is there any way I can work around this?

the malloc function on itanium/ifort only support allocations of up to 
2GB, your job is asking for 2.3GB. if you are only using a few of the 
12 CPUs you may get away with doubling the number of cpus. if you are
already using all 12, you have to find a larger machine to run the job
as you won't fit the full job into 24GB (minus the requirements of 
the OS etc.) anyways.

cheers,
   axel.

MK> 
MK> Thanks,
MK> 
MK> Maurice
MK> _______________________________________________
MK> CPMD-list mailing list
MK> CPMD-list at cpmd.org
MK> http://cpmd.org/mailman/listinfo/cpmd-list
MK> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.