https://github.com/jbornschein/mpi4py-examples

mpi4py examples
https://github.com/jbornschein/mpi4py-examples

examples mpi4py python

Last synced: 6 days ago
JSON representation

mpi4py examples

• Host: GitHub
• URL: https://github.com/jbornschein/mpi4py-examples
• Owner: jbornschein
• License: mit
• Created: 2010-07-31T09:51:02.000Z (over 14 years ago)
• Default Branch: master
• Last Pushed: 2024-03-06T23:44:31.000Z (8 months ago)
• Last Synced: 2024-08-06T21:22:59.390Z (3 months ago)
• Topics: examples, mpi4py, python
• Language: Python
• Homepage:
• Size: 21.5 KB
• Stars: 367
• Watchers: 18
• Forks: 111
• Open Issues: 5
• Metadata Files:
  • Readme: README.txt
  • License: LICENSE

Awesome Lists containing this project

README

        

=== Dependencies ===

These programs depend on mpi4py (>= Version 1.0)

The mpi4py documentation and installation instructions 

can be found at:

   http://mpi4py.scipy.org/

=== How to run on a single (multi-core) host ===

Run it with 

 mpirun -np 4 ./some-program

where the number after "-np " is the number of parallel MPI 

processes to be started.

=== How to run on multiple hosts ===

If you want to run the program distributed over multiple hosts, 

you have to create a  which looks like:

-- hostfile --

host1   slots=4

host2   slots=4

host3   slots=4

--------------

Where "slots=" specifies the number of parallel processes that should be

started on that host.

Run it with

  mpirun --hostfile  ./some-program

=== Run on a cluster with the Torque Job scheduling system ===

There are two possibilities:

a) Run interactively:

Request an interactive session and allocate a number of processors/nodes for 

your session:

 $ qsub -I X -l nodes=4:ppn=4

Where "-I" means you want to work interactively, "-X" requests grapical

(X-Window) I/O -- (you can run arbitrary programs that open windows).  The

option "-l " specifies the resources you want to allocate.  "-l nodes=4:ppn=4"

requests four compute nodes with each having four processor cores 

[ppn =^ ProcessorsPerNode].  So in total you allocate 16 CPU cores. 

[The scheduler is free to run your job on two nodes having 8 CPU cores each]

Once your interactive session is ready, you run 

 $ mpirun ./your-program

  

mpirun automatically knowns how many parallel processes have to be started and

where they have to be started.

b) Submit as non-interactive batch-job:

Use 

 $ qsub -l nodes=4:ppn=4 ./your-jobfile.job

to submit jour job-file. Similar to the interactive case, "-l" again is used  

to request resources from the scheduling system. The job file usually is a 

simple shell script which specifies the commands to be run once your job 

starts. In addition, the jobfile can contain "#PBS " statements

which are used to specify additional options which could have been specified 

in the "qsub" commandline. Please see "man qsub" for details.