A Parallel Hierarchical Approach for Automatic Differentiation
by
Marco Mancini
Department of Electronics, Computer Science and Systems, University of Calabria, Italy

Efficient Automatic Differentiation (AD) algorithms of computer programs can be designed by exploiting the associativity of the chain rule of calculus, the program structure and high-performance computing techniques. The main aim of this work is to describe an AD algorithm that, based on a source transformation paradigm, generates efficiently first-order derivatives by using a hierarchical approach and that takes advantage of parallel technology in order to perform derivative computation concurrently. The proposed AD algorithm has been developed by considering a shared-memory paradigm as a parallel computational model. Its performance has been evaluated by considering some test problems of the MINPACK-2 collection. The computational experiments have been carried out on a SGI Origin 2000 and some encouraging numerical results are presented.

http://www.parcolab.unical.it/~mancini/ad2000_par.ps

Date received: February 11, 2000

Copyright © 2000 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cads-68.