Since 1986 - Covering the Fastest Computers in the World and the People Who Run Them

Language Flags
November 13, 2009

SC09 Ready Or Not

Michael Feldman

It’s hard to believe that SC09 is going to kick off in just a few days in Portland, Oregon. As I get prepared for my great trek to the Northwest, I’m trying to anticipate what the big stories of this year’s show will be. The official focus areas of SC09 are bio-computing, the 3D Internet, and environmental sustainability, but as I peruse the conference schedule I see a few other themes emerging.

For example, this looks like the first year a broad swath of the community is giving serious thought to exascale computing. There are at least eight presentations on the schedule this year that focus on this topic. Of course, exascale systems are several years away, but companies like Cray, IBM, Intel, NVIDIA and others are already mapping out a strategy to get us there, and of course the academic community is all atwitter about it. If you’re still at the show on Friday, I’d recommend sitting in on The Road to Exascale panel to get the exa-scoop from some of the big names in academia and industry.

Speaking of Friday panels, there is another good one, which unfortunately runs concurrently with the exascale one mentioned above. But if you’re not ready for exaflops yet, try to catch Preparing the World for Ubiquitous Parallelism. As the title suggests, it is all about the trials and tribulations of parallel programming. Representatives from AMD, NVIDIA, Intel, Adobe Systems and DreamWorks will all be there. Tom Murphy from Contra Costa College will also be a panelist to offer his take on teaching parallelism in the classroom.

By the way, Tom along with his cohort, Paul Steinberg of the Intel Academic Community, host a “Teach Parallel” broadcast twice a month, where they talk about parallel programming with various computer science digerati. During the Supercomputing Conference, the Teach Parallel broadcasts will also be augmented with a live feed of all the SC09 plenary sessions, with the exception of Al Gore’s keynote.

Another big topic sure to permeate this year’s show is GPGPU. As my contacts from the world of NVIDIA pointed out, SC09 is going to be swimming in GPUs (especially theirs). According to them, more than 10 percent of the papers presented at the event reference their CUDA-GPU platform, while 17-plus system providers will be showing Tesla GPUs on the exhibit floor. The company also has a dozen or so software vendors at the show that are currently building products atop CUDA. According to NVIDIA’s SC09 Web page, the company will be demonstrating its first Fermi-equipped Tesla hardware at its booth, which, if I’m not mistaken, will be the first time a Fermi product is observed in the wild.

All the GPU activity in HPC seems to have spurred the FPGA lobby into action. A number of announcements from reconfigurable computing (RC) vendors are on tap for SC09. Also, on Sunday, before opening day festivities, there will be a Workshop on High-Performance Reconfigurable Computing Technology and Applications (HPRCTA’09). Dr. Alan George of CHREC (Center for High-Performance Reconfigurable Computing) will deliver a keynote address there. For those of you who can’t make the workshop, try to catch our interview with Dr. George during our conference coverage, as he explains why he thinks RC is poised for HPC stardom.

Oh, and yours truly is scheduled to be on a BoF panel (Communicating Virtual Science) on Wednesday night. The BoF will be a discussion about the inner workings of science and technology journalism. So if you’re interested in how the media magically turns vendor press releases and academic research papers into Pulitzer-prize winning prose, stop by and we’ll tell you our secrets.

Finally, as is always the case, there will be some big HPC vendor announcements during the show, but I can’t tell you about those quite yet.

So what did I leave out? Al Gore’s keynote, the TOP500, and the other 90 percent of the conference. Don’t worry, we’ve got it covered.