The Rapsodia tool is the result of a collaboration between
Further contributions have been made by Rapsodia is intended for the computation of higher order derivative information of numerical models written in Fortran, C or C++ by automatic differentiation.  Rapsodia consists of  two parts:
  1. A Python-based code generator that produces a C++ or Fortran library for the propagation of univariate Taylor polynomials for a given derivative order and number of directions. The code generator relies on inlining and loop unrolling to aid subsequent compiler optimization.  
  2. Implementations of the algorithm that interpolates derivative tensor entries from Taylor polynomial coefficients provided for both, C++ and Fortran.
Please follow the links in the top frame to download the software, read the code documentation, the manual or refer to timing tests.
The last change sets were pushed to the source code repository on this server on 2014-03-05 .


[CharpentierRUm] I. Charpentier, J. Utke:
Rapsodia: User Manual

[Utke2013SAf] J. Utke, B.T. Rearden, R.A. Lefebvre:
Sensitivity Analysis for Mixed-Language Numerical Models

[Reed2012CAD] J. Reed, J. Utke, and H. Abdel-Khalik:
Combining Automatic Differentiation Methods for High-Dimensional Nonlinear Models

[Buntinas2010Egl] D. Buntinas, A. Malozemoff, J. Utke:
Efficient generated libraries for asynchronous derivative computation

[Buntinas2010Egl] D. Buntinas, A. Malozemoff, J. Utke:
Multithreaded derivative computation with generated libraries (extended journal version of the above conference paper)

[Charpentier2009Fho] I. Charpentier, J. Utke:
Fast higher-order derivative tensors with Rapsodia

[Charpentier2008EHO] I. Charpentier, C. Dal Cappello, J. Utke:
Efficient Higher-Order Derivatives of the Hypergeometric Function

web site last updated: Friday, 28-Mar-2014 08:58:43 CDT