This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).
Graphics Processors (GPUs) are potentially a cost effective and low power vehicle for science and engineering research that requires high performance computation. The primary challenge to the use of GPUs more broadly is the difficulty in programming. Dr. Walker and a team of colleagues representing five different scientific and engineering disciplines propose to pursue research topics in each of the disciplines. By selecting important research topics which require a fundamentally similar computational algorithm for a class of problems labelled "n-body problems", the project offers opportunity for meaningful interdisciplinary collaboration across scientific domains that are normally quite distinct. Since, solutions to this class of problem are particularly well suited to GPUs, there is likelihood of advances in multiple areas of scientific interest at a fraction of alternative costs and power. Therefore NSF's Office of Cyberinfrastructure (OCI) is supporting the acquisition of the instrument.
Atronomers understanding dark matter, biological chemists predicting drug interactions, cognitive neuroscientists exploring human behavior and nanotechnologists designing energy conversion devices are among the focus areas studied with the Vanderbilt GPU cluster acquired with this grant. The commonality among these important applications is the n-body nature of the simulations. That is, all involve extraordinary numbers of small easily desribed objects. When combined together with the laws of nature, they can describe rich natural phenomena leading scientific and engineering discoveries. The GPU cluster is perfectly designed to accelerate research of this type. The GPU cluster attracted 93 research groups and well over one hundred investigators exploring as many important scientific questions. In addition, the grant was used to help train hundreds of students in computational science using GPU processors. This next geenration of computer-oriented researchers will devise and attack new problems that were until now intractable. Among the findings include a new nanostructured device (VSSL or variably-spaced superlattice) that will improve energy conversion efficiency for direct thermal to electrical power. This technology can reduce society's dependence on non-renewable energy resources by producing additional power from otherwise wasted heat. In cognitive neuroscience, the results of the simulations performed at Vanderbilt have brought us closer to mimicing human brain activities. Understanding of these phenomena will advance the development of treatments for debilitating neural diseaes such as Alzheimers or depression. Protein folding studies have also been performed to help us understand to fundamental chemical reactions in the body that govern cell growth and pathology. These reactions are critical to developing new drugs and other treatments for ailments ranging from cancer to the common cold. Since n-body problems are not scale dependent, we used the cluster to study inhomogeneous distributions of matter across the universe. These studies show how our universe formed in the time close to the big bang and help answer recent questions involving the expanding universe and dark matter.
