Shared memory parallelism and low-rank approximation techniques applied to direct solvers in FEM simulation
Abstract
In this paper, the performance of a parallel sparse direct solver on a shared memory multicore system is presented. Large size test matrices arising from finite element simulation of induction heating industrial applications are used in order to evaluate the performance improvements due to low-rank representations and multicore parallelization
Keywords
Shared-Memory Parallelism
FEMs
Linear Systems
Sparse Matrices
Eddy Currents
Electromagnetic Heating
Memory Management
Parallel Sparse Direct Solver
Multicore Parallelization
Low-Rank Representations
Test Matrices
Large Size
FEM Simulation
Parallel Processing
Multiprocessing Systems
Matrix Algebra
Mathematics Computing
Induction Heating
Approximation Theory
Finite Element Analysis
Electrical Engineering Computing
Origin : Files produced by the author(s)