How algorithm re-thinking can open the way to exascale

Recent years have seen a dramatic change in core technology. The end of Dennard scaling has now placed energy as a primary design principle in parwith performance, all the way from the hardware to the application software. To continue the exponential overall improvements, technologists have turned into multi-core chips and parallelism at all scales. This trend opens up new challenges in application development: how to program more complex machines and how to scale applications to the O(10^6) cores characteristic of leadership class supercomputers, while maintaining the pace with the very fast increase in power requirements. This talk will present several examples where a careful algorithm re-engineering has opened the way to fast, scalable, and energy-efficient solutions, for a range of applications from solution of complex PDEs to big data analytics.

contact: christian.vergara@polimi.it