arXiv subjects

Javad Moshfegh

Publications and source records attributed to Javad Moshfegh.

3 recordsLinked to original sources

arXiv2020

Direct Domain Decomposition Method (D3M) for Finite Element Electromagnetic Computations

An exact arithmetic, memory efficient direct solution method for finite element method (FEM) computations is outlined. Unlike conventional black-box or low-rank direct solvers that are opaque to the underlying physical problem, the proposed method leverages physical insights at every stage of the development through a new symmetric domain decomposition method (DDM) with one set of Lagrange multipliers. Comparisons with state-of-the-art exact direct solvers on electrically large problems suggest up to 10 times less memory and better run-time complexity while maintaining the same accuracy.

cs.CE↗

arXiv2020

Direct Solution of FEM Models: Are Sparse Direct Solvers the Best Strategy?

A brief summary of direct solution approaches for finite element methods (FEM) in computational electromagnetics (CEM) is given along with an alternative direct solution based on domain decomposition (DD). Unlike recent trends in approximate/low-rank solvers, this work focuses on `numerically exact' solution methods as they are more reliable for complex `real-life' models. Preliminary studies on general three dimensional geometries with unstructured FEM meshes suggest that the proposed direct DD methodology offers significant memory advantages over highly optimized, high-performance sparse direct solver libraries, while maintaining approximately comparable or slightly slower serial serial execution speed but with significantly better parallel and GPU processing prospects.

cs.CE↗

arXiv2020

Parallel Direct Domain Decomposition Methods (D3M) for Finite Elements

A parallel direct solution approach based on domain decomposition method (DDM) and directed acyclic graph (DAG) scheduling is outlined. Computations are represented as a sequence of small tasks that operate on domains of DDM or dense matrix blocks of a reduced matrix. These tasks can be statically scheduled for parallel execution using their DAG dependencies and weights that depend on estimates of computation and communication costs. Performance comparison with MUMPS 5.1.2 on electrically large problems suggest up to 20% better parallel efficiency, 30% less memory and slightly faster in run-time, while maintaining the same accuracy.

cs.DC↗