Top > Seminars & Events > Seminars > Parallel-in-time integration for massively para...


Parallel-in-time integration for massively parallel computer simulations

Hold Date
2016-01-07 15:00〜2016-01-07 16:30
IMI Conference Room(W1-D-414), West Zone 1, Ito campus, Kyushu University
Object person
Daniel Ruprecht (University of Leeds)

I will introduce two different numerical algorithms, Parareal and PFASST, that allow to exploit concurrency along the time dimension for large-scale computer simulations solving partial differential equations. Performance results from parallel computations for both methods on different architectures and for different applications will be presented. The impact of using shared or distributed memory parallelisation on the performance of Parareal in terms of runtime, memory and energy will be analysed. Also, recent results showing how parallel-in-time integration can provide algorithm-based tolerance against hardware faults will be discussed.