Hai Xiang Lin - TU Delft
‘Designing Parallel Numerical Algorithms for Exascale Computing’
The number of CPUs and cores in a parallel computer keeps on increasing and supercomputers with millions of cores have been already bulit. Within just a few years these hardware systems are expected to provide a peak performance of Exaflops, however, the incerased scale also brings a number of challenges to computational scientists and engineers. It is far from trivial to make efficient use of such a huge parallel computer. Increased scale also leads to increasing communication and synchronization overhead. In this we will review the techniques for reducing these overheads for parallel execution of numerical algorithms.
Hai Xiang Lin is with the group of Mathematical Physics, Delft Institute of Applied Mathematics (DIAM). His research interests include high performance computing, parallel and distributed numerical algorithms, simulation of large scale models, and big data. He has been working with applications such as 3-D shallow water transport models of the coastal areas, electrical power grids, volcanic ash distribution, and future transportation systems. He is also as a part time professor in Environmental Modelling and Data Analytics at Leiden University.