ADIC - AN EXTENSIBLE AUTOMATIC DIFFERENTIATION TOOL FOR ANSI-C

In scientific computing, we often require the derivatives partial deri vative f/partial derivative x of a function f expressed as a program w ith respect to some input parameter(s) x, say, Automatic Differentiati on (AD) techniques augment the program with derivative computation by applying the chain rule of calculus to elementary operations in an aut omated fashion, This article introduces ADIC (Automatic Differentiatio n of C), a new AD tool for ANSI-C programs, ADIC is currently the only tool for ANSI-C that employs a source-to-source program transformatio n approach; that is, it takes a C code and produces a new C code that computes the original results as well as the derivatives, We first pre sent ADIC 'by example' to illustrate the functionality and ease of use of ADIC and then describe in detail the architecture of ADIC. ADIC in corporates a modular design that provides a foundation for both rapid prototyping of better AD algorithms and their sharing across AD tools for different languages, A component architecture called AIF (Automati c Differentiation Intermediate Form) separates core AD concepts from t heir language-specific implementation and allows the development of ge neric AD modules that can be reused directly in other AIF-based AD too ls, The language-specific ADIC front-end and back-end canonicalize C: programs to make them fit for semantic augmentation and manage, for ex ample, the association of a program variable with its derivative objec t, We also report on applications of ADIC to a semiconductor device si mulator, 3-D CFD grid generator, vehicle simulator, and neural network code. (C) 1997 by John Wiley & Sons, Ltd.