This award will support the acquisition of Nehalem nodes, each with twelve cores and more than 24 GB of memory, Infiniband switch and network, and an NVIDIA S20 series Tesla with Fermi GPUs. The proposed hardware is well suited to advance research in pure and applied mathematics at the University of Michigan. The hardware will be used for research on algebraic curves of high genus in positive characteristic, grain boundaries of polycrystalline metals and ceramics, Navier-Stokes solvers and exact solutions, very fast Fourier transforms for sparse data, tree codes and vortex methods, formal Laurent series, and quantum dots in materials science.
The SCREMS hardware will allow us to enrich the content of a range of courses, encompassing the arithmetic of dynamical systems, compressive sensing, image processing, and scientific computing. In addition, the hardware will allow us and our students and postdocs, some of whom belong to under-represented groups, to acquire expertise in parallel programming, high performance computing, and GPU computing. Thus the SCREMS hardware will be a valuable resource for training and professional development.
. The grant was used to purchase modern computer hardware consisting of Intel Xeon 5650 nodes, Infiniband network, an NVIDIA M2070 GPU system, and Intel SandyBridge machines equipped with Phi co-processors. Easy access to the latest computing technologies supported instruction and a variety of research. New research advanced, enabled, or accomplished using this grant ranges from pure mathematics to image processing, molecular dynamics, and fluid mechanics. A new algorithm for the automatic detection of anomalous patterns in large data sets captured from overhead cameras has been derived. One of the largest and most detailed mathematical models for Circadian rhythms was developed. A boundary integral method for the Poisson-Boltzmann equation was developed. Cohomology of cluster algebras is under investigation. Kinetic Monte-Carlo simulations of nano-scale phenomena such as liquid drop epitaxy were carried out. A new algorithm for the simulation of fully developed turbulence from the Navier-Stokes equations, with promise of reaching higher Reynolds numbers, has been developed. The grant provided access to modern computing hardware for several researchers and students. It was used to develop material for a class on scientific programming and computer architecture that is currently in its fourth iteration.
