Thinking Parallel

It’s been a short while since I came back from IWOMP but I would still like to share some of the news and experiences I have brought back from there. It’s been a really great experience for me, both at the conference and in the city of Beijing – which is always worth a trip, if you ask me. So here they go:

• Tasking is the next big thing in OpenMP 3.0. And it really is big. I had the opportunity to talk and think this through in multiple sessions with many of the people involved in designing that feature and even play with a reference implementation for a bit (thanks to the UPC-people at Barcelona for that) and it really does take OpenMP to a whole new level. New classes of applications are able to be parallelized with OpenMP without much of a problem (e.g. recursive algorithms or other, more irregular ones). The last details of tasking are being finalized at this time but this is the one feature that will most definitely make it into OpenMP 3.0. Very exciting :grin:!
• The target date for a release of OpenMP 3.0 is the time of the Supercomputing conference this year, which is November 10th-16th. Of course, this release date is not yet confirmed but the schedule has been set. Will be very interesting to see what other features make it in until that date.
• A couple of new members have joined the OpenMP ARB recently (the ARB is the Architecture Review Board, which basically controls the fate of the everything related to the specification), the most prominent ones being Microsoft, AMD and Cray. This pushes the list of member organizations up to 16 – which are in case you did not know (in no particular order):
  • ASCI
  • cOMPunity
  • EPCC
  • NASA
  • STMicroelectronics / Portland Group
  • Fujitsu
  • HP
  • IBM
  • Intel
  • NEC
  • SGI
  • Sun
  • RWTH Aachen University
  • Microsoft
  • AMD
  • Cray

  Quite a long list of supporters now, don’t you think?

• Two talks at this conference have used the JOMP-implementation as basis for their research. This is an implementation of OpenMP on top of Java originally done by Mark Bull from EPCC, but has been unmaintained for some time. As can be seen by these talks, there is obviously some demand for OpenMP on Java, though.
• GCC 4.2. has recently been released, including OpenMP-support for the very first time. This should help mainstream adaption, as it adds another compiler to the growing list of compilers that can handle OpenMP.
• I have also been able to do some interviews with a lot of very interesting people, more on that as soon as I have managed to type them up *sigh*.

I have probably forgotten to put down many other things I should be saying here, but I am sure I will remember them sooner or later and just post them then. As you can see, although it may not look like it from the outside (e.g. by judging from the sorry state of the OpenMP-homepage), the community is busy and there are exciting developments ahead.