Scaling Analysis of a Hierarchical Parallelization of Large Inverse Multiple-Scattering Solutions

We propose a hierarchical parallelization strategy to improve the scalability of inverse multiple-scattering solutions. The inverse solver parallelizes the independent forward solutions corresponding to different illuminations. For further scaling out on large numbers of computing nodes, each forward solver parallelizes the dense and large matrix-vector multiplications accelerated by the multilevel fast multipole algorithm. Numerical results on up to 1,024 of CPU nodes show that the former and latter parallelizations have 95% and 73% strong-scaling efficiencies, respectively. ACM Reference format: Mert Hidayetoğlu, Carl Pearson, Izzat El Hajj, Weng Cho Chew, Levent Gürel, and Wen-Mei Hwu. 2017. Scaling Analysis of a Hierarchical Parallelization of Large Inverse Multiple-Scattering Solutions. In Proceedings of Supercomputing, Denver, CO, November 2017 (SC17), 2 pages. https://doi.org/10.1145/nnnnnnn.nnnnnnn

[1]  Weng Cho Chew,et al.  Scalable parallel DBIM solutions of inverse-scattering problems , 2017, 2017 Computing and Electromagnetics International Workshop (CEM).

[2]  Weng Cho Chew,et al.  A frequency-domain formulation of the Fréchet derivative to exploit the inherent parallelism of the distorted Born iterative method , 2006 .

[3]  W. Chew,et al.  Fast inverse scattering solutions using the distorted Born iterative method and the multilevel fast multipole algorithm. , 2010, The Journal of the Acoustical Society of America.

[4]  Weng Cho Chew,et al.  Large inverse-scattering solutions with DBIM on GPU-enabled supercomputers , 2017, 2017 International Applied Computational Electromagnetics Society Symposium - Italy (ACES).