TY - JOUR
T1 - A New Decomposition Solver for Complex Electromagnetic Problems [EM Programmer's Notebook]
AU - Zhao, Xunwang
AU - Chen, Yan
AU - Zhang, Huanhuan
AU - Zhang, Yu
AU - Sarkar, Tapan K.
N1 - Funding Information:
This work is supported by grants from the International S&T Coop eration program (2016YFE0121600), the National Natural Science Foun dation of China (NSFC) (61301069), the New Century Excellent Talents of the University of China program (NCET-13-0949), the 863 Program (2012AA01A308), the Fundamental Research Funds for the Central Uni versities research project(JB160218), and the Special Program for Applied Research on Super Computation of the NSFC–Guangdong Joint Fund (the second phase).
Publisher Copyright:
© 2017 IEEE.
PY - 2017/6
Y1 - 2017/6
N2 - A parallel higher-order method of moments (HOMoM) with a newly developed reduced-communication, lower-upper (RCLU) decomposition solver is proposed in this article. The method uses 201,600 central processing unit (CPU) cores on a supercomputer located in Guangzhou, China. Our code achieves an extremely high parallel efficiency when simulating a large aircraft that has been discretized in the method-of-moments (MoM) context, using a higher-order quadrilateral patch basis, into approximately 1.06 million unknowns for the surface-current distribution. Remarkably, its solution using the classical lower-upper (LU) solver only takes roughly half an hour. In addition, a review of the in-core and out-of-core algorithms of an HOMoM is presented, with a focus on their parallel implementation. The parallel performance of the methodology is also demonstrated on some challenging applications.
AB - A parallel higher-order method of moments (HOMoM) with a newly developed reduced-communication, lower-upper (RCLU) decomposition solver is proposed in this article. The method uses 201,600 central processing unit (CPU) cores on a supercomputer located in Guangzhou, China. Our code achieves an extremely high parallel efficiency when simulating a large aircraft that has been discretized in the method-of-moments (MoM) context, using a higher-order quadrilateral patch basis, into approximately 1.06 million unknowns for the surface-current distribution. Remarkably, its solution using the classical lower-upper (LU) solver only takes roughly half an hour. In addition, a review of the in-core and out-of-core algorithms of an HOMoM is presented, with a focus on their parallel implementation. The parallel performance of the methodology is also demonstrated on some challenging applications.
UR - http://www.scopus.com/inward/record.url?scp=85027369913&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85027369913&partnerID=8YFLogxK
U2 - 10.1109/MAP.2017.2687119
DO - 10.1109/MAP.2017.2687119
M3 - Article
AN - SCOPUS:85027369913
SN - 1045-9243
VL - 59
SP - 131
EP - 140
JO - IEEE Antennas and Propagation Magazine
JF - IEEE Antennas and Propagation Magazine
IS - 3
M1 - 7938824
ER -