Abstract
Algorithmic adaptions to use next-generation computers closer to their potential are underway. Instead of squeezing out flops – the traditional goal of algorithmic optimality, which once served as a reasonable proxy for all associated costs – algorithms must now squeeze synchronizations, memory, and data transfers, while extra flops on locally cached data represent only small costs in time and energy. After decades of programming model stability with bulk synchronous processing, new programming models and new algorithmic capabilities (to make forays into, e.g., data assimilation, inverse problems, and uncertainty qualification) must be co-designed with the hardware. The speaker briefly recaps the architectural constraints and application opportunities. He then concentrates on two types of tasks each occupies a large portion of all scientific computing cycles: large dense symmetric/Hermitian systems (covariance, Hamiltonians, Hessians, Schur complements) and large sparse Poisson/Helmholtz systems (solids, fluids, electromagnetism, radiation diffusion, gravitation). He examines progress in porting solvers for these tasks to the hybrid distributed-shared programming environment, including the graphics processing unit (GPU) and the Many Integrated Core (MIC) architectures that make up the cores of the top scientific systems “on the floor” and “on the books”.
About the speaker
Prof David E Keyes got his BSE in Aerospace and Mechanical Engineering from Princeton University in 1978 and his PhD in Applied Mathematics from Harvard University in 1984. He started his academic career at Yale University, where he taught for eight years, prior to joining Old Dominion University and the Institute for Computer Applications in Science & Engineering at the NASA Langley Research Center in 1993. Then he moved to Columbia University where he was the Fu Foundation Professor of Applied Mathematics until 2011. From 2009 to 2012, he served as the inaugural Dean of Mathematical and Computer Sciences and Engineering at the King Abdullah University of Science and Technology (KAUST), where he is currently the Professor of Applied Mathematics and Computer Science and Director of Extreme Computing Research Center.
Prof Keyes works at the algorithmic interface between parallel computing and the numerical analysis of partial differential equations. He has been the author and editor of more than a dozen US federal agency reports and member of several federal advisory committees on computational science and engineering and high performance computing. His representative publications include KBLAS: An Optimized Library for Dense Matrix-Vector Multiplication on GPU Accelerators, Multicore-optimized Wavefront Diamond Blocking for Optimizing Stencil Updates, etc.
Prof Keyes has been recognized as fellows of the American Mathematical Society (AMS) (2012) and the Society for Industrial and Applied Mathematics (SIAM) (2011). He is also honored with the Sidney Fernbach Award of the IEEE Computer Society (2007) and the Gordon Bell Prize of the Association for Computer Machinery (ACM) (1999).
|
