Publication
Title
Automatic differentiation for solving nonlinear partial differential equations : an efficient operator overloading approach
Author
Abstract
By resorting to Automatic Differentiation (AD) users of nonlinear PDE solvers can be relieved from the extra work of linearising a nonlinear PDE system and at the same time improve on the computational efficiency. This paper describes the main AD techniques and discusses how the operator overloading approach of AD can be extended to eliminate the overhead generally incurred with operator overloading. A recent AD system FastDer++, specially designed for this purpose, is integrated into a Least Squares solver. The necessary modifications to the general FEM algorithms. Code fragments and timing results demonstrate that (1) integrating AD with nonlinear PDE solvers leads to highly flexible code with a close resemblance to the mathematical expression of the problem, (2) coding and debugging efforts are greatly reduced, and (3) the computational efficiency is improved.
Language
English
Source (journal)
Numerical algorithms. - Basel
Publication
Basel : 2002
ISSN
1017-1398
Volume/pages
30:3-4(2002), p. 259-301
ISI
000177838300003
Full text (Publisher's DOI)
Full text (publisher's version - intranet only)
UAntwerpen
Faculty/Department
Publication type
Subject
External links
Web of Science
Record
Identification
Creation 24.02.2014
Last edited 12.10.2017