Skip to main content


GPU computing research

Research Achievements

GPU computing research

As his minor project, IGERT trainee Kipton Barros worked with Brower and collaborators to develop optimized code for lattice QCD on graphics processing units (GPUs). The effort resulted in order-of-magnitude speedups over traditional computer architectures. Barros has since taken a postdoctoral position at Northwestern University, where he is leveraging the experience he gained in GPU computing. This project has spurred a major initiative to use GPUs in the U.S. LQCD community, with the initial proof-of-concept code having grown into the "QUDA" library for LQCD. A lead developer of this library is Ronald Babich, a former IGERT trainee and now postdoctoral fellow at BU Center for Computational Science. Partly as a result of this work, Boston University has become a hub for interdisciplinary research in GPU computing. In particular, a GPU cluster was recently installed with support from an NSF EAGER award (Brower, PI), to be used by researchers in LQCD, CFD, and other fields.