Publication: Semi-automatic Parallelization of Direct and Inverse Problems for Geothermal Simulation
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

Semi-automatic Parallelization of Direct and Inverse Problems for Geothermal Simulation

- Part of a collection -
 

Area
Geophysics

Author(s)
H. M. Bücker , A. Rasch , V. Rath , A. Wolf

Published in
Proceedings of the 24th ACM Symposium on Applied Computing, Honolulu, Hawaii, USA, March 8--12, 2009

Year
2009

Publisher
ACM Press

Abstract
We describe a strategy for parallelizing a geothermal simulation package using the shared-memory programming model OpenMP. During the code development OpenMP is employed for the direct problem in such a way that, in a subsequent step, the OpenMP-parallelized code can be transformed via automatic differentiation into an OpenMP-paral-lelized code capable of computing derivatives for the inverse problem. Performance results on a Sun Fire X4600 using up to 16 threads are reported demonstrating that, for the derivative computation, an approach using nested parallelism is more scalable than a single level of parallelism.

AD Theory and Techniques
Parallelism

BibTeX
@INPROCEEDINGS{
         Bucker2009SaP,
       author = "H. M. B{\"u}cker and A. Rasch and V. Rath and A. Wolf",
       title = "Semi-automatic Parallelization of Direct and Inverse Problems for Geothermal
         Simulation",
       booktitle = "Proceedings of the 24th ACM Symposium on Applied Computing, Honolulu, Hawaii, USA,
         March~8--12, 2009",
       publisher = "ACM Press",
       pages = "971--975",
       doi = "10.1145/1529282.1529495",
       address = "New York",
       abstract = "We describe a strategy for parallelizing a geothermal simulation package using the
         shared-memory programming model OpenMP. During the code development OpenMP is employed for the
         direct problem in such a way that, in a subsequent step, the OpenMP-parallelized code can be
         transformed via automatic differentiation into an OpenMP-paral\-lelized code capable of
         computing derivatives for the inverse problem. Performance results on a Sun Fire X4600 using up to
         16 threads are reported demonstrating that, for the derivative computation, an approach using nested
         parallelism is more scalable than a single level of parallelism.",
       year = "2009",
       volume = "2",
       ad_area = "Geophysics",
       ad_theotech = "Parallelism"
}


back
  

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