Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] RFC: Java MPI bindings
From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2012-02-07 15:47:34


On Feb 7, 2012, at 2:33 PM, George Bosilca wrote:

> This doesn't sound like a very good idea, despite a significant support from a lot of institutions. There is no standardization efforts in the targeted community, and championing a broader support in the Java world was not one of our main target.

This is a bit of a chicken-and-egg issue.

You can't standardize something until you know what the good thing is to standardize. We currently have no Java bindings -- there were several in the late 90s, but all of them have bit-rotted in one way or another. Java -- as a performant technology -- has come a long way since then.

Hence, this is an attempt by those in the MPI community to go get some real-world experience on what to standardize.

> OMPI does not include the Boost bindings, despite the fact that it was developed at IU.

Apples and oranges.

Boost.mpi = class library
Boost.mpi != bindings
Boost.mpi has its own, separate community.
Boost.mpi hasn't been developed in quite a while.

This effort = bindings (i.e., 1:1 mapping to the C MPI bindings)
This effort != class library
This effort has a group that is trying to join the MPI community

> OMPI does not include Python nor R bindings despite their large user community.

They've also:

- never asked to be part of Open MPI
- support more than just Open MPI

If all goes well, the Java bindings will someday support more than Open MPI. But for today, the enterprise players are choosing to put efforts and resources only into the enterprise-class MPI implementation: Open MPI.

> Why suddenly should we provide unstandardized Java bindings?

Because multiple members of the Open MPI developer community would like to go explore this space.

The way these bindings interact is basically additional stuff in configure (to find the java compiler and the like) and a new directory under ompi/mpi/java.

It's the moral equivalent of a new component.

> I think we should not tackle such inclusion before there is at least a beginning of a standardization effort in the targeted community.

Chicken-and-egg issue. You know as well as I do that the MPI Forum won't talk about Java bindings until a (sizeable) community can be identified who can demonstrate real-world use cases.

> They have to step up and address their needs (if they are real), instead of relying on us to take a decision.

Er... have you noticed Ralph's new employer? He's at Greenplum now. They're a hadoop company.

They basically hired him to enable MPI in a Hadoop world.

Sooo... I'd say that they *are* stepping up. :-)

-- 
Jeff Squyres
jsquyres_at_[hidden]
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/