Re: [pooma-dev] r1, MPI, and doof2d: What is Wrong?

[Scott Haney]

I'm hazy about the R1 options, but isn't it --commlib? Also, I don't recall whether it is mandatory to give the number of processes with --processes NP. It may be redundant since you specified the -np for mpirun, but I'm not sure.

Scott

On Tuesday, October 16, 2001, at 08:58 PM, Jeffrey Oldham wrote:

> Attached is the the r1 version of doof2d.  I have tried to run it
> using MPI, but it keeps falling over, probably because I do not know
> how to use MPI or I am too tired.
>
> Is the problem and a solution obvious?
>
> > mpirun -np 1 ./doof2d --comm MPI --nofieldcompression 400 1 > results_mpi_p1_400_1
> MPI: MPI_COMM_WORLD rank 0 has terminated without calling MPI_Finalize()
> MPI: aborting job
> mpirun -np 1 ./doof2d --comm MPI --nofieldcompression 400 1 > results_mpi_p1_400_1 died at ./r1-Doof2d-tester.pl line 28.
>
> > mpirun -np 4 ./doof2d --comm MPI --nofieldcompression 800 2 > results_mpi_p4_400_2
> MPI: MPI_COMM_WORLD rank 3 has terminated without calling MPI_Finalize()
> MPI: aborting job
> mpirun -np 4 ./doof2d --comm MPI --nofieldcompression 800 2 > results_mpi_p4_400_2 died at ./r1-Doof2d-tester.pl line 28.
>
> Pooma r1 was configured with './conf SGI64 KAI OPTIMIZE MPI'.
>
> Thanks,
> Jeffrey D. Oldham
> address@hidden
--
Scott W. Haney
Development Manager
Proximation LLC
2960 Rodeo Park Drive West
Santa Fe, NM 87505
Voice: 505-424-3809 x101
FAX: 505-438-4161