Co-design of a particle-in-cell plasma simulation code for Intel Xeon Phi: A first look at Knights landing
Paper i proceeding, 2016

Three dimensional particle-in-cell laser-plasma simulation is an important area of computational physics. Solving state-of-the-art problems requires large-scale simulation on a supercomputer using specialized codes. A growing demand in computational resources inspires research in improving efficiency and co-design for supercomputers based on many-core architectures. This paper presents first performance results of the particle-in-cell plasma simulation code PICADOR on the recently introduced Knights Landing generation of Intel Xeon Phi. A straightforward rebuilding of the code yields a 2.43 x speedup compared to the previous Knights Corner generation. Further code optimization results in an additional 1.89 x speedup. The optimization performed is beneficial not only for Knights Landing, but also for high-end CPUs and Knights Corner. The optimized version achieves 100 GFLOPS double precision performance on a Knights Landing device with the speedups of 2.35 x compared to a 14-core Haswell CPU and 3.47 x compared to a 61-core Knights Corner Xeon Phi.

Författare

I. Surmin

Lobachevsky University

S. Bastrakov

Lobachevsky University

Z. Matveev

Intel Technologies, Inc., Russia

E. Efimenko

Lobachevsky University

Russian Academy of Sciences

Arkady Gonoskov

Chalmers, Fysik, Teoretisk fysik

I. Meyerov

Lobachevsky University

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

03029743 (ISSN) 16113349 (eISSN)

Vol. 10049 319-329
9783319499550 (ISBN)

Ämneskategorier

Annan teknik

DOI

10.1007/978-3-319-49956-7_25

ISBN

9783319499550

Mer information

Skapat

2017-10-07