1

I've been following the tutorial I found at mpitutorial.com, and I've been able to get by on that so far. However, after setting up my AWS cluster, I have hit a small snag. The program is only running on one processor, even with a host file, and the world size is still 1. As a result, I cannot practice with the send/receive code properly.

This is the output I get from mpirun -host master,node001 -n 4 ./mpi_hello_world:

Hello world from processor master, rank 0 out of 1 processors Hello world from processor master, rank 0 out of 1 processors Hello world from processor node001, rank 0 out of 1 processors Hello world from processor node001, rank 0 out of 1 processors

Can anyone tell me why the -host flag isn't actually running the processes on the other instance I started up?

Inglonias
  • 468
  • 1
  • 5
  • 18
  • 1
    This is a duplicate of [Wrong mpi number of processors](http://stackoverflow.com/questions/9990947/wrong-mpi-number-of-processors), [MPI Unexpected output](http://stackoverflow.com/questions/15095121/mpi-unexpected-output), [MPI_Rank return same process number for all process](http://stackoverflow.com/questions/20287564/mpi-rank-return-same-process-number-for-all-process) and several other questions. It is the result of running (with `mpirun`) using a different flavour of MPI than you compiled with (`mpicc`, etc). – Jonathan Dursi Jun 16 '14 at 15:17
  • That's correct in my case, however that doesn't help me. I used Starcluster to configure the server, and I don't know how to fix this particular issue. Following the instructions about purging the various MPI installs failed, and I can't seem to reinstall a single MPI package. Any idea what I can do? – Inglonias Jun 16 '14 at 15:29
  • Use an absolute path to your `mpirun` executable. – Wesley Bland Jun 18 '14 at 13:40

1 Answers1

1

Using the mpich2 plugin from StarCluster fixes this issue. Thanks for your help, everybody!

Cashew
  • 1,466
  • 2
  • 16
  • 21
Inglonias
  • 468
  • 1
  • 5
  • 18