Publication: Explicit Loop Scheduling in OpenMP for Parallel Automatic Differentiation
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

Explicit Loop Scheduling in OpenMP for Parallel Automatic Differentiation

- Part of a collection -
 

Author(s)
H. M. Bücker , B. Lang , A. Rasch , C. H. Bischof , D. an Mey

Published in
Proceedings of the 16th Annual International Symposium on High Performance Computing Systems and Applications, Moncton, NB, Canada, June 16--19, 2002

Editor(s)
J. N. Almhana, V. C. Bhavsar

Year
2002

Publisher
IEEE Computer Society Press

Abstract
Derivatives of almost arbitrary functions can be evaluated efficiently by automatic differentiation whenever the functions are given in the form of computer programs in a high-level programming language such as Fortran, C, or C++. In contrast to numerical differentiation, where derivatives are only approximated, automatic differentiation generates derivatives that are accurate up to machine precision. Sophisticated software tools implementing the technology of automatic differentiation are capable of automatically generating code for the product of the Jacobian matrix and a so-called seed matrix. It is shown how these tools can benefit from concepts of shared memory programming to parallelize, in a completely mechanical fashion, the gradient operations associated with each statement of the given code. The feasibility of our approach is demonstrated by numerical experiments. They were performed with a code that was generated automatically by the ADIFOR system and augmented with OpenMP directives.

AD Theory and Techniques
Parallelism

BibTeX
@INPROCEEDINGS{
         Bucker2002ELS,
       author = "H. M. B{\"u}cker and B. Lang and A. Rasch and C. H. Bischof and D.~an~Mey",
       title = "Explicit Loop Scheduling in OpenMP for Parallel Automatic Differentiation",
       booktitle = "Proceedings of the 16th Annual International Symposium on High Performance
         Computing Systems and Applications, Moncton, NB, Canada, June~16--19, 2002",
       editor = "J. N. Almhana and V. C. Bhavsar",
       pages = "121--126",
       address = "Los Alamitos, CA",
       publisher = "IEEE Computer Society Press",
       doi = "10.1109/HPCSA.2002.1019144",
       url = "http://doi.ieeecomputersociety.org/10.1109/HPCSA.2002.1019144",
       abstract = "Derivatives of almost arbitrary functions can be evaluated efficiently by automatic
         differentiation whenever the functions are given in the form of computer programs in a high-level
         programming language such as Fortran, C, or C++. In contrast to numerical differentiation, where
         derivatives are only approximated, automatic differentiation generates derivatives that are accurate
         up to machine precision. Sophisticated software tools implementing the technology of automatic
         differentiation are capable of automatically generating code for the product of the Jacobian matrix
         and a so-called seed matrix. It is shown how these tools can benefit from concepts of shared memory
         programming to parallelize, in a completely mechanical fashion, the gradient operations associated
         with each statement of the given code. The feasibility of our approach is demonstrated by numerical
         experiments. They were performed with a code that was generated automatically by the Adifor system
         and augmented with OpenMP directives.",
       year = "2002",
       ad_theotech = "Parallelism"
}


back
  

Contact:
autodiff.org
Username:
Password:
(lost password)