Computing Systems for Data-Driven Science
Robert D. Mawhinney is a Professor of Physics at Columbia University. He received his B.S. in Physics from the University of South Florida in 1980 and his Ph.D. in Physics from Harvard University in 1987. His research is in the area of theoretical particle physics, with a particular focus on the behavior of quarks and the theory which describes their interactions, Quantum Chromodynamics (QCD). QCD is a highly non-linear theory, giving a first principles description of the strong interactions between quarks and many of the consequences of this theory are only available through large scale simulations. QCD is also an ideal numerical problem, with only a handful of input parameters and a very large number of output predictions, which are vital to interpreting and assimilating the results of many particle physics experiments.
To satisfy the large computational demands of QCD, Dr. Mawhinney and his colleagues at Columbia and other institutions have designed and built a number of massively parallel computers for their QCD simulations. The Quantum Chromodynamics on Digital Signal Processors (QCDSP) computers were built in 1997-1998, achieved a peak speed of 1 TFlop/s and won a Gordon Bell Prize at SC 98 for the best price to performance ratio. This was followed by the Quantum Chromodynamics on a Chip (QCDOC) computers in 2005, with 3, 10 TFlop/s machines built and run at the University of Edinburgh and Brookhaven National Laboratory. QCDOC was implemented using IBM’s ASIC technology and Dr. Alan Gara, a QCDSP and QCDOC designer, continued the development of this style of computing in IBM’s BlueGene series of Supercomputers.