Research Exhibit
This is the home page for our 13th major research exhibit at the IEEE/ACM Supercomputing conference. The exhibit is again under the title Aggregate.Org / University of Kentucky, the informal research consortium led by our KAOS (Compilers, Hardware Architectures, and Operating Systems) group here at the University of Kentucky's Department of Electrical and Computer Engineering.
This year, the buzz at SC was largely about various types of attached processors: FPGAs, Clearspeed, Cell, and GPUs. We've been fans of attached processors for quite some years, but we were very disappointed to see that FPGA-based systems in particular seem to be priced very non-competitively... around $15K for an entry-level system. GPUs might not be the easiest architectural models to deal with, but at least they don't have the same pricing problem. Our work with GPUs was quite well recieved.
Although we had a very wide range of topics to discuss, the work emphasized in our exhibit is summarized in the following white papers, which were distributed from our research exhibit, booth 232, at SC06:
It is difficult to know how people will react to the exhibit until after the show begins. This year, of the many things presented, three seemed to be most interesting to the most people:
Our research exhibit was positioned nicely at one end of the aisle that spanned the entire show floor, and we pushed our backlit signs still higher this year, with the result that our booth was visible from quite far away. We estimate about 600 visitors came to our booth, again with a pretty high average "dwell time." The strange part was that Broadcom was next to us -- the mixing of industry and research exhibits hasn't worked out very well for the research folks because all industry booths get to pick locations before any research booth, no matter what the "points" for the organization. Actually, the problem is much worse for SC07, where all the good positions were taken by industry before any research exhibitor got a selection... we'll see how that works out.
From far away, the most striking things were the two backlit signs. Under the one in the left front of the booth was a 5-panel cluster video wall using our AFN hardware. The top display was actually kept separate, showing a simple slide show. The four bottom displays were used to show dynamic bar graphs of the predicted and measured power and execution time for a large set of benchmarks.
Under the sign in the rear right side of the booth were a display of the current fisheye views of the booth and a 4-machine laptop video wall using an AFN. This year, we had two cameras capturing fisheye images of the booth and posting them live at Aggregate.Org:
Up close, the core of our booth was again our pentagonal rear-projection display, which continues to have the magical power to cause lots of folks to peek under the leather curtains to see what the display technology is. The PC in front houses an ATI FireStream GPU, and our little MMX "bunnyman" sat upon it advertising the fact that both are variations on the SWAR execution model compiler technology that we've played a significant role in developing.
To be posted shortly....
The only thing set in stone is our name.