Applications

We are interested in developing applications for reconfigurable supercomputers such as the Cray XD1, with dual-Opteron nodes, Xilinx Virtex 2 Pro FPGA, and a proprietary interconnection network. Our applications include Monte Carlo Radiative Heat Transfer (also available from FPL 2004, Springer Verlag, Publisher), Metropolitan Road Network Simulation (IEEE FCCM 2005), and simulation of lattice gas hydrodynamics.

Monte Carlo Radiative Heat Transfer

The RASR team is researching mapping floating point applications onto FPGAs. Newer, very large FPGAs can accommodate floating point's large operand size and excessive area use, so can potentially be used for implementing supercomputing applications. In addition, studies suggest that peak floating-point performance of FPGAs is growing significantly faster than peak floating point performance for a CPU. Monte Carlo radiative heat transfer is a floating point supercomputing application that traces a large number of photons emitted from the surfaces of a 2-D enclosure. We mapped the application's most compute-intensive inner loop onto Xilinx Virtex II and Virtex II Pro FPGAs. Using three single-precision floating point pipelines, we found that we could achieve a speed up of 10.37X versus running the application on a 3 GHz Pentium IV Xeon workstation.

Metropolitan Road Network Simulation

This work demonstrates that road traffic simulation of entire metropolitan areas is possible with reconfigurable supercomputing that combines 64-bit microprocessors and FPGAs in a high bandwidth, low latency interconnect. Previously, traffic simulation on FPGAs was limited to very short road segments or required a very large number of FPGAs. Our data streaming approach overcomes scaling issues associated with direct implementations and still allows for high-level parallelism by dividing the data sets between hardware and software across the reconfigurable supercomputer. Using one FPGA on the Cray XD1 supercomputer, we are able to achieve a 12.8X speed up over the AMD microprocessor. This result paves the way for accelerating other large infrastructure simulations.