CSGF Conference Notes

Just returned from 5 days in Washington, DC for the annual computational science graduate fellowship conference.  No live notes/tweets, so just attempting to synthesize some of the major ideas before I get back to paper writing. Most of the conference material, including workshop slides, are archived online.

The high performance computing workshop made two things clear to me – a lot of parallel computing can be done without lots of communication between nodes (i.e. Monte Carlo, simulations, and all the embarrassingly parallel stuff we now call massively serial stuff), and the next generation of supercomputers – exoscale computers – will be pretty much impossible to program or use, with lots of processors per chip but no memory, while the major limitation becomes (already is) power. It’s not the only was forward, as one fellow pointed me to the massive shared-memory machine at PSC, blacklight. The NERSC hands-on session was particularly good.

The student talks were as excellent as they were diverse.  I think the fellows this year did a particularly good job of keeping the talks entertaining to a wide audience, if completely irrelevant to my work.  See the agenda for details, but some of the astrophysics talks probably took the cake.  As always, the poster session was expertly organized into shifts and abundant good food.

I always enjoy hearing from the various agency directors.  We didn’t have a general this year to yell at us about how awesome science was, but we did here a lot aboutpredictive science and quantifying uncertainty, with some emphasis on ill-defined but policy relevant systems.  Hey, that’s what I do! (in the forecasting ecosystem collapse work.)  Only I think ecology isn’t what they have in mind. It was great to see familiar faces of fellows and alumni and meet the incoming students.  Celebrating 20 years, CSGF program has adopted the tag-line for this conference as “Building a Community of Leaders.”  With a diverse network of researchers in all kinds of interdisciplinary fields, I think it’s not too far off.