P27: Parallelization of the Particle-In-Cell Monte Carlo Collision (PIC-MCC) Algorithm for Plasma Simulation on Intel MIC Xeon Phi Architecture
SessionPoster Reception
Event Type
ACM Student Research Competition
Poster
Reception
TimeTuesday, November 14th5:15pm - 7pm
LocationFour Seasons Ballroom
DescriptionThe implementation of 2D-3v (2D in space and 3D in velocity space) PIC-MCC (Particle-In-Cell Monte Carlo Collision) method involves the computational solution of Vlasov-Poisson equations. This provides the spatial and temporal evolution of the charged-particle velocity distribution functions in plasma under the effect of self-consistent electromagnetic fields and collisions. Stringent numerical constraints associated with the PIC code makes it computationally prohibitive on CPU.
In our work, parallelization and optimization techniques have been extended to this simulation, along with a novel approach that involves developing a ‘self-aware’ code that triggers sorting in order to maintain cache-coherence while reducing the total sorting time during iterations.
We present the effect of important numerical parameters on speed-up. Finally, we compare the scalability and performance of the parallelization and optimization strategies on Intel® Xeon™ E5-2630, Xeon Phi™ 5110p and Xeon Phi™ 7250 relative to a serial implementation on Intel® i5.
In our work, parallelization and optimization techniques have been extended to this simulation, along with a novel approach that involves developing a ‘self-aware’ code that triggers sorting in order to maintain cache-coherence while reducing the total sorting time during iterations.
We present the effect of important numerical parameters on speed-up. Finally, we compare the scalability and performance of the parallelization and optimization strategies on Intel® Xeon™ E5-2630, Xeon Phi™ 5110p and Xeon Phi™ 7250 relative to a serial implementation on Intel® i5.
Authors
