TY - GEN
T1 - Parallel particle-in-cell performance optimization
T2 - 31st IEEE International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2017
AU - Narayanan, Ramachandran Kodanganallur
AU - Madduri, Kamesh
N1 - Publisher Copyright:
© 2017 IEEE.
PY - 2017/6/30
Y1 - 2017/6/30
N2 - The particle-in-cell (PIC) numerical technique is frequently used in physics and engineering simulations. In this work, we describe ES-PIC, a new shared-memory parallel implementation of the PIC technique for electrospray simulations. Electrospray simulations are used in aerospace applications, and the goal of an electrospray simulation is to understand behavior of an electrospray thruster or a colloid thruster. We discuss performance optimizations for various steps of a PICbased electrospray simulation. One of the main steps in this simulation is solving the Poisson partial differential equation, and this step can be in turn converted to solving a system of linear equations. We develop a parallel implementation of the Multigrid method for this step. We demonstrate that ESPIC is significantly faster than other parallel PIC electrospray simulation implementations on Intel Xeon multicore platforms. Further, ES-PIC can serve as a real-world scientific computing benchmark for analyzing parallel system performance.
AB - The particle-in-cell (PIC) numerical technique is frequently used in physics and engineering simulations. In this work, we describe ES-PIC, a new shared-memory parallel implementation of the PIC technique for electrospray simulations. Electrospray simulations are used in aerospace applications, and the goal of an electrospray simulation is to understand behavior of an electrospray thruster or a colloid thruster. We discuss performance optimizations for various steps of a PICbased electrospray simulation. One of the main steps in this simulation is solving the Poisson partial differential equation, and this step can be in turn converted to solving a system of linear equations. We develop a parallel implementation of the Multigrid method for this step. We demonstrate that ESPIC is significantly faster than other parallel PIC electrospray simulation implementations on Intel Xeon multicore platforms. Further, ES-PIC can serve as a real-world scientific computing benchmark for analyzing parallel system performance.
UR - http://www.scopus.com/inward/record.url?scp=85028046359&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85028046359&partnerID=8YFLogxK
U2 - 10.1109/IPDPSW.2017.160
DO - 10.1109/IPDPSW.2017.160
M3 - Conference contribution
AN - SCOPUS:85028046359
T3 - Proceedings - 2017 IEEE 31st International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2017
SP - 1158
EP - 1167
BT - Proceedings - 2017 IEEE 31st International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2017
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 29 May 2017 through 2 June 2017
ER -