https://github.com/bccp/nbodykit

Analysis kit for large-scale structure datasets, the massively parallel way
https://github.com/bccp/nbodykit

astrophysics clustering cosmology data-analysis large-scale-structure mpi mpi4py parallel-computing python

Last synced: about 2 months ago
JSON representation

Analysis kit for large-scale structure datasets, the massively parallel way

• Host: GitHub
• URL: https://github.com/bccp/nbodykit
• Owner: bccp
• License: gpl-3.0
• Created: 2015-04-21T19:43:37.000Z (almost 10 years ago)
• Default Branch: master
• Last Pushed: 2022-09-06T17:17:00.000Z (over 2 years ago)
• Last Synced: 2024-10-07T08:12:17.409Z (4 months ago)
• Topics: astrophysics, clustering, cosmology, data-analysis, large-scale-structure, mpi, mpi4py, parallel-computing, python
• Language: Python
• Homepage: http://nbodykit.rtfd.io
• Size: 20.9 MB
• Stars: 111
• Watchers: 21
• Forks: 60
• Open Issues: 93
• Metadata Files:
  • Readme: README.SGI-MPT
  • Changelog: CHANGES.rst
  • Contributing: CONTRIBUTING.rst
  • License: LICENSE

Awesome Lists containing this project

README

        
This file describes using nbodykit (and related parallel MPI modules) with SGI-MPT.

SGI-MPT is more standard compliant than openmpi or mpich. As a result:

1. we need collective object destruction. This can be mimiced in Python by

setting.

.. code::

    export PYTHONHASHSEED=0

2. we need to enable MPI_SHEPHERD mode. It is not well documented, but otherwise

mpiexec will fail.

.. code::

    export MPI_SHEPHERD=true

3. runtests needs an extra argument to tell it to use mpiexec instead of the popular

but non-standard mpirun.

.. code::

    python runtests.py --mpirun="mpiexec -n 4"