I am currently trying to run a very simple ScalaPACK test program on a
LAM/MPI system, but keep running into a problem with MPI error
messages. If anyone could help me on this, it would be appreciated.
The code fails in BLACS_GRIDINFO, but manages BLACS_PINFO and BLACS_GET fine.
The code, which is simplified from the Scalapack website after the
original failed at the same stage:
PROGRAM HELLO
* -- BLACS example code --
* Written by Clint Whaley 7/26/94
* Performs a simple check-in type hello world
* ..
* .. External Functions ..
INTEGER BLACS_PNUM
EXTERNAL BLACS_PNUM
* ..
* .. Variable Declaration ..
INTEGER CONTXT, IAM, NPROCS, NPROW, NPCOL, MYPROW, MYPCOL
INTEGER ICALLER, I, J, HISROW, HISCOL, ERROR
*
* Determine my process number and the number of processes in
* machine
*
* CALL MPI_INIT (ERROR)
CALL BLACS_PINFO(IAM, NPROCS)
WRITE (*,*) "Process", IAM+1, "out of", NPROCS
*
* If in PVM, create virtual machine if it doesn't exist
*
IF (NPROCS .LT. 1) THEN
IF (IAM .EQ. 0) THEN
WRITE(*, 1000)
READ(*, I) NPROCS
END IF
CALL BLACS_SETUP(IAM, NPROCS)
END IF
*
* Set up process grid that is as close to square as possible
*
NPROW = INT( SQRT( REAL(NPROCS) ) )
NPCOL = NPROCS / NPROW
*
* Get default system context, and define grid
*
CALL BLACS_GET(0, 0, CONTXT)
WRITE (*,*) "Process", IAM+1, "context", CONTXT
CALL BLACS_GRIDINIT(CONTXT, 'Row', NPROW, NPCOL)
WRITE (*,*) "Context after gridinit", CONTXT
CALL BLACS_GRIDINFO(CONTXT, NPROW, NPCOL, MYPROW, MYPCOL)
WRITE (*,*) "Gridinfo", MYPROW, MYPCOL
* CALL MPI_FINALIZE(error)
1000 FORMAT ('How many processes in the machine?')
STOP
END
The output is:
Process 1 out of 2
Process 1 context 0
MPI_Comm_group: invalid communicator: Invalid argument (rank 0, MPI_COMM_WORLD)
Rank (0, MPI_COMM_WORLD): Call stack within LAM:
Rank (0, MPI_COMM_WORLD): - MPI_Comm_group()
Rank (0, MPI_COMM_WORLD): - main()
Process 2 out of 2
Process 2 context 0
MPI_Comm_group: invalid communicator: Invalid argument (rank 1, MPI_COMM_WORLD)
Rank (1, MPI_COMM_WORLD): Call stack within LAM:
Rank (1, MPI_COMM_WORLD): - MPI_Comm_group()
Rank (1, MPI_COMM_WORLD): - main()
The code is compiled with:
mpif77 -o hello hello.f libscalapack.a blacsF77init_MPI-LINUX-0.a
blacs_MPI-LINUX-0.a blacsF77init_MPI-LINUX-0.a
/home/lindkvis/Applications/Atlas/lib/libf77blas.a
/home/lindkvis/Applications/Atlas/lib/libatlas.a
On this system:
LAM/MPI: 7.0.6 Prefix: /usr
Architecture: x86_64-redhat-linux-gnu
Configured by: bhcompile
Configured on: Tue Nov 2 16:57:23 EST 2004
Configure host: dolly.build.redhat.com
C bindings: yes
C++ bindings: yes
Fortran bindings: yes
C profiling: yes
C++ profiling: yes
Fortran profiling: yes
ROMIO support: yes
IMPI support: no
Debug support: no
Purify clean: no
SSI boot: globus (Module v0.5)
SSI boot: rsh (Module v1.0)
SSI coll: lam_basic (Module v7.0)
SSI coll: smp (Module v1.0)
SSI rpi: crtcp (Module v1.0.1)
SSI rpi: lamd (Module v7.0)
SSI rpi: sysv (Module v7.0)
SSI rpi: tcp (Module v7.0)
SSI rpi: usysv (Module v7.0)
And run with:
mpirun -np 2 hello
Using the following lamhost (dual core AMD Opteron test system):
localhost cpu=2
I tried adding a MPI_INIT and subsequent MPI_FINALIZE without any
effect. Regular MPI c++ code runs fine on this system.
It uses Atlas for lapack and blas and Scalapack compiled using Atlas.
I apologise if this is not the best mailing list to post to, but I
thought there is probably considerable scalapack experience on this
mailing list.
--
Gaute Lindkvist
|
