Adifor 2.0: Automatic Differentiation of Fortran 77 Programs

  • Authors:
  • Christian Bischof;Peyvand Khademi;Andrew Mauer;Alan Carle

  • Affiliations:
  • -;-;-;-

  • Venue:
  • IEEE Computational Science & Engineering
  • Year:
  • 1996

Quantified Score

Hi-index 0.03

Visualization

Abstract

Many problems in scientific computing--optimization, sensitivity analysis, and others--require the calculation of derivatives. However, application codes are usually not constructed with derivatives in mind. The technique of automatic differentiation builds on the familiar rules of calculus to compute accurate derivatives from such codes. It propagates values, not expressions, avoiding the "expression swell" familiar to some users of symbolic math software. The Adifor 2.0 system is an advance in automatic differentiation that won the 1995 Wilkinson Prize for Numerical Software at the Third International Congress on Industrial and Applied Mathematics in Hamburg, Germany. Given an arbitrary Fortran 77 source code, this software generates an augmented code that computes the partial derivatives of all of the specified dependent variables with respect to all specified independent variables, in addition to the original result. IEEE CS&E is pleased to publish the first major article on how Adifor 2.0 works.