Publication: Modeling the Performance of Interface Contraction
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

Modeling the Performance of Interface Contraction

- Article in a journal -
 

Author(s)
H. Martin Bücker , A. Rasch

Published in
ACM Transactions on Mathematical Software

Year
2003

Abstract
Automatic differentiation is a technique used to transform a given computer code implementing some mathematical function into another program capable of evaluating the function and its derivatives. Compared to numerical differentiation, the derivatives obtained from applying automatic differentiation are free from truncation error and their computation requires less time under a wide range of circumstances. To increase the efficiency of a black box approach of automatic differentiation further, a technique called interface contraction may be used. The idea behind interface contraction is to exploit the local structure of a given code to temporarily reduce the global number of derivatives propagated through the code. A performance model is proposed to predict the potential improvement in the execution time of a program making use of interface contraction compared to a program generated by a black box approach of automatic differentiation. The performance model is validated by numerical experiments from different application areas on a SUN UltraSPARC-III processor.

AD Theory and Techniques
Interface Contraction

BibTeX
@ARTICLE{
         Bucker2003MtP,
       title = "Modeling the Performance of Interface Contraction",
       journal = "{ACM} Transactions on Mathematical Software",
       pages = "440--457",
       abstract = "Automatic differentiation is a technique used to transform a given computer code
         implementing some mathematical function into another program capable of evaluating the function and
         its derivatives. Compared to numerical differentiation, the derivatives obtained from applying
         automatic differentiation are free from truncation error and their computation requires less time
         under a wide range of circumstances. To increase the efficiency of a black box approach of automatic
         differentiation further, a technique called interface contraction may be used. The idea behind
         interface contraction is to exploit the local structure of a given code to temporarily reduce the
         global number of derivatives propagated through the code. A performance model is proposed to predict
         the potential improvement in the execution time of a program making use of interface contraction
         compared to a program generated by a black box approach of automatic differentiation. The
         performance model is validated by numerical experiments from different application areas on a SUN
         UltraSPARC-III processor.",
       year = "2003",
       volume = "29",
       number = "4",
       ad_theotech = "Interface Contraction",
       author = "H. Martin B{\"u}cker and A. Rasch",
       doi = "http://doi.acm.org/10.1145/962437.962442"
}


back
  

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