Ticket #7214 (closed: fixed)
Replace uses of boostmpi with mpi4py
| Reported by: | Russell Taylor | Owned by: | Russell Taylor |
|---|---|---|---|
| Priority: | critical | Milestone: | Release 2.6 |
| Component: | Python | Keywords: | MPI |
| Cc: | Blocked By: | ||
| Blocking: | Tester: | Vickie Lynch |
Description (last modified by Nick Draper) (diff)
The boostmpi python package depends on boost-python (it is python bindings to the Boost.MPI library generated - unsurprisingly - with boost-python) but these days Mantid builds a version of boost-python into our libraries, meaning that there is a conflict if something imports boostmpi (which SNSPowderReduction does).
Since we only use it to determine rank, and the package is no longer maintained, we should switch to using the mpi4py package (http://mpi4py.scipy.org/) instead.
Change History
comment:2 Changed 7 years ago by Russell Taylor
Re #7214. Switch SNSPowderReduction from using boostmpi to mpi4py.
Changeset: 7a5ae466eb8d0ae42fa59498cb8aaf9a93e71a6d
comment:3 Changed 7 years ago by Russell Taylor
Typed the wrong ticket number in this one:
Re #7213. Switch examples to us mpi4py in place of boostmpi.
Changeset: 7d3ba96e65808489d3eb05c06885cdcca5c6a905
comment:4 Changed 7 years ago by Russell Taylor
- Status changed from accepted to verify
- Resolution set to fixed
Branch is feature/7214_mpi4py_replaces_boostmpi
To test make sure in particular that SNSPowderReduction works with an MPI build of Mantid. Double-check that there are no remaining references to 'mpi.world'. You'll need to have mpi4py installed - see http://www.mantidproject.org/Parallel_Mantid_using_MPI for instructions on that.
comment:5 Changed 7 years ago by Vickie Lynch
- Status changed from verify to verifying
- Tester set to Vickie Lynch
comment:6 Changed 7 years ago by Vickie Lynch
- Status changed from verifying to closed
Ran parallel reduction of Nomad data on 8 cores and it worked with mpi4py
