It took us longer than we intended, but we finally released Open MPI v1.7.4.
Woo hoo! (we got nice coverage from El Reg, too)
This is a monster release; it represents hundreds (thousands? millions?) of person-hours of work. Consider this a ginormous “thank you!” to the entire Open MPI community! Special thanks goes to Paul Hargrove of Lawrence Berkeley Labs for his hero efforts in regression testing throughout December and January.
v1.7.4 is the latest in Open MPI’s feature release series; here’s some of the notable new / improved features:
- Added much more MPI-3 support:
- Added support for remaining non-blocking collectives
- Added support for neighborhood collectives
- Updated C bindings with const and []
- Added MPI_COMM_IDUP
- Added MPI_COMM_CREATE_GROUP
- Support new MPI-3 mpi_f08 errata
- We’re now only missing the new one-sided functionality for MPI-3 completeness
- Added native support for the Intel Phi
- Added more CUDA support
- Added new MPI Java bindings
- Upgrade the embedded version of ROMIO to the latest available
And of course, lots and lots of little bug fixes and small improvements that aren’t worth calling out in detail (see the NEWS file for more information)
I have a java app to parallelize this year. Perfect for C++/MPI, but I wanted to keep it in Java. Crazy? Advice?
I’d say that this sounds like the exact case that we created the MPI Java bindings for. Give it a whirl with the Java bindings and send us your feedback — let us know what’s good and what’s bad.
Oscar Gisbert has been pushing in some further performance improvements after 1.7.4 was released, too — they’ll all be included in 1.7.5, which should be out sometime soon (!) (we had lots of new stuff queued up while 1.7.4 was undergoing its release process, so 1.7.5 isn’t far behind).
I encourage you to use the new Java bindings.
There is still little information… but you have the Java docs at this location: ompi-install/share/doc/openmpi/javadoc
Also you can write to the ompi mailing lists.