Resource Management for Parallel Computers

Praun, Christoph von; Fineman, Jeremy T.; Leiserson, Charles E.; Gallopoulos, Efstratios; Snir, Marc; Heath, Michael R.; Grice, Don; White, Andrew G.; López, Pedro; Meister, Benoit; Vasilache, Nicolas; Wohlford, David; Baskaran, Muthu Manikandan; Leung, Allen; Lethin, Richard; Saltz, Joel; Das, Raja

doi:10.1007/978-0-387-09766-4_2262

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Note that the arrays χ domain mn and χ core mn have the same size. Additionally, each version of χ core mn is assigned to only a single CPU core in the first parallel do-loop to prevent problematic read/write interactions, i.e., race-conditions [34]. The first parallel do-loop in Algorithm 1 divides the cross-sections within the entire simulation domain over all N cores cores.…”

Section: Parallel Algorithmmentioning

confidence: 99%

A Parallel 3D Spatial Spectral Volume Integral Equation Method for Electromagnetic Scattering from Finite Scatterers

Eijsvogel,

Dilz,

van Beurden

2023

PIER B

View full text Add to dashboard Cite

Parallel computing for the three-dimensional spatial spectral volume integral equation method is presented for the computation of electromagnetic scattering by finite dielectric scatterers in a layered medium. The first part exploits the Gabor-frame expansion to compute the Gabor coefficients of scatterers in a parellel manner. The second part concerns the decomposition and restructuring of the matrix-vector product of this spatial spectral volume integral equation into (partially) independent components to enable parallel computing. Both capitalize on the hardware to reduce the computation time by shared-memory parallelism. Numerical experiments in the form of solving electrically large scattering problems, namely volumes up to 1300 cubic wavelengths, in combination with a large number of finite scatterers show a significant reduction in wall-clock time owing to parallel computing, while maintaining accuracy.

show abstract

Section: Parallel Algorithmmentioning

confidence: 99%