论文信息 - Applying TAF to generate efficient derivative code of Fortran 77‐95 programs

Applying TAF to generate efficient derivative code of Fortran 77‐95 programs

This paper features FastOpt's automatic differentiation (AD) tool Transformation of Algorithms in Fortran (TAF), a source to source translator for programs written in Fortran 77-95. TAF and its predecessor TAMC have a long record of successful large-scale applications. Here we give an overview on how TAF approaches typical challenges of AD such as handling of badly written program code, of large memory/disk requirements, of iterative solvers or of black box routines. We also point out, where the user is required to prepare his program code prior to invoking TAF.

Thomas Kaminski | Ralf Giering | R. Giering | T. Kaminski

[1] Andreas Griewank,et al. Achieving logarithmic growth of temporal and spatial complexity in reverse automatic differentiation , 1992 .

[2] Bruce Christianson,et al. Sharing storage using dirty vectors , 1996 .

[3] Andreas Griewank,et al. Automatic Differentiation of Algorithms: From Simulation to Optimization , 2000, Springer New York.

[4] Thomas Kaminski,et al. Recipes for adjoint code construction , 1998, TOMS.

[5] B. Christianson. Reverse aumulation and imploicit functions , 1998 .

[6] Andreas Griewank,et al. Evaluating derivatives - principles and techniques of algorithmic differentiation, Second Edition , 2000, Frontiers in applied mathematics.