Showing papers on "Source transformation published in 2010"
12 Sep 2010
TL;DR: The link between these two components in order to compute second derivatives is described, illustrated by a simplified parallel implementation of Burgers' equation in a second-order optimization setting, for example, Newton's method.
Abstract: A source transformation tool for algorithmic differentiation is introduced, capable of transforming MPI-enabled code into secondorder adjoint code. Our derivative code compiler (dcc) is used for the source transformation while a runtime library handles the adjoining of the MPI routines. This paper describes in detail the link between these two components in order to compute second derivatives. This process is illustrated by a simplified parallel implementation of Burgers' equation in a second-order optimization setting, for example, Newton's method.
7 citations
TL;DR: A code generator is utilized to create libraries that overload intrinsics for derivative computation, and approaches to improve the efficiency of the generated code are discussed.
2 citations
01 Jan 2010
TL;DR: Tapenade (2), a source transformation automatic dierentiation tool, is used to obtain the sensitivity of a CFD system, whereby the independent and dependent variables are user- dened.
Abstract: This paper presents some insights into constructing sensitivity algorithms, the tools required, what the constraints are and how to assemble the program. In this work Tapenade (2), a source transformation automatic dierentiation tool, is used to obtain the sensitivity of a CFD system, whereby the independent and dependent variables are user- dened. Points on how to prepare the source code are detailed, with the aim of enabling the programmer to submit safe transformable code to Tapenade whilst retaining the use of modern Fortran structures, making use of modules, derived data types, pointers, etc.
1 citations
11 Jul 2010
TL;DR: In this paper, the form-invariant nature of Maxwell's equations has led to the transformation electromagnetics/optics device design technique, where electromagnetic phenomena in the transformed coordinate system are interpreted as the result of modifications to material properties in the untransformed system.
Abstract: The form-invariant nature of Maxwell's equations has led to the transformation electromagnetics/optics device design technique [1], where electromagnetic phenomena in the transformed coordinate system are interpreted as the result of modifications to material properties in the untransformed system. Starting with the invisibility cloaks [2], the coordinate transformation approach has been used to design a variety of electromagnetic/optical devices such as a field rotator [3] and beam bends/expanders [4, 5]. Source transformation results if the original domain contains sources [6]; the source distribution in the transformed system is given in terms of the original source configuration and the coordinate transformation. The source transformation has been recently applied to arrays [7].