Dear Openmpi Users,
I am reserving several processors with SGE upon which I
want to run a number of openmpi jobs, all of which
individually (and combined) use less than the reserved number
of processors. The code I am using uses BLACS, and when
blacs_pinfo is called I get a seg fault. If the code doesn't
call blacs_pinfo it runs fine being submitted in this manner.
blacs_pinfo simply returns the number of available processors,
so I suspect this is an issue with SGE and openmpi and the
requested node number being different to that given to mpirun.
Can anyone explain why this would happen with openmpi jobs
using BLACS on the SGE? And suggest maybe a way around it?
Many thanks
Conn
example
submission script:
#!/bin/bash -f -l
#$ -V
#$ -N test
#$ -S /bin/bash
#$ -cwd
#$ -l vf=1800M
#$ -pe ib-ompi 12
#$ -q infiniband.q
BIN=~/bin/program
for i in XPOL,YPOL,ZPOL; do
mkdir ${TMPDIR}/4ZP;
mkdir ${TMPDIR}/4ZP/$i;
cp ./4ZP/$i/* ${TMPDIR}/4ZP/$i;
done
cd ${TMPDIR}/4ZP/XPOL;
mpirun -np 4 -machinefile ${TMPDIR}/machines $BIN > output &
cd ${TMPDIR}/4ZP/YPOL;
mpirun -np 4 -machinefile ${TMPDIR}/machines $BIN > output &
cd ${TMPDIR}/4ZP/ZPOL;
mpirun -np 4 -machinefile ${TMPDIR}/machines $BIN > output ;
for i in XPOL YPOL ZPOL ; do
cp ${TMPDIR}/4ZP/$i/* ${HOME}/4ZP/$i;
done
blacs_pinfo.c:
#include "Bdef.h"
#if (INTFACE == C_CALL)
void Cblacs_pinfo(int *mypnum, int *nprocs)
#else
F_VOID_FUNC blacs_pinfo_(int *mypnum, int *nprocs)
#endif
{
int ierr;
extern int BI_Iam, BI_Np;
/*
* If this is our first call, will need to set up some stuff
*/
if (BI_F77_MPI_COMM_WORLD == NULL)
{
/*
* The BLACS always call f77's mpi_init. If the user is using C, he should
* explicitly call MPI_Init . . .
*/
MPI_Initialized(nprocs);
#ifdef MainInF77
if (!(*nprocs)) bi_f77_init_();
#else
if (!(*nprocs))
BI_BlacsErr(-1, -1, __FILE__,
"Users with C main programs must explicitly call MPI_Init");
#endif
BI_F77_MPI_COMM_WORLD = (int *) malloc(sizeof(int));
#ifdef UseF77Mpi
BI_F77_MPI_CONSTANTS = (int *)
malloc(23*sizeof(int));
ierr = 1;
bi_f77_get_constants_(BI_F77_MPI_COMM_WORLD, &ierr, BI_F77_MPI_CONSTANTS);
#else
ierr = 0;
bi_f77_get_constants_(BI_F77_MPI_COMM_WORLD, &ierr, nprocs);
#endif
BI_MPI_Comm_size(BI_MPI_COMM_WORLD, &BI_Np, ierr);
BI_MPI_Comm_rank(BI_MPI_COMM_WORLD, &BI_Iam, ierr);
}
*mypnum = BI_Iam;
*nprocs = BI_Np;
}