Computing Surface Integral Equation Matrices with Shared Memory Parallelization

An efficient shared memory parallelization strategy is presented for computing matrices resulting from surface integral equation based methods in time-harmonic electromagnetics (EM). In the numerical solution of EM problems with methods such as the method of moments (MoM) a significant portion of the solution time is dedicated to filling dense interaction matrices. Developing efficient parallelization strategies to take advantage of all of the available computational resources of multi-core central processing units (CPU) is thus critical in the fast solution of computationally demanding EM problems. The proposed scheme is applied to the electric field integral equation (EFIE)-based MoM and shown to demonstrate good parallel efficiency.