The following technical report is available from
http://aib.informatik.rwth-aachen.de:
Toward Adjoint OpenFOAM
Niloofar Safiran, Uwe Naumann
AIB 2011-16
OpenFOAM is a package which simplifies the implementation of physical
models by mimicking the form of partial differential equations in
software. Use of OpenFOAM is rapidly expanding in the research
community and among industrial users, covering a wide range of
continuum models. AD provides accurate derivative values and better
runtime performance of the adjoint model compared with finite
differences. This paper presents a transformation of the OpenFOAM
code that can be used for calculating derivatives with adjoint mode
of AD. An example based on ODE is considered here.
The following technical report is available from
http://aib.informatik.rwth-aachen.de:
Solving Muller Games via Safety Games
Daniel Neider, Roman Rabinovich, Martin Zimmermann
AIB 2011-14
We show how to transform a Muller game with n vertices into
a safety game with (n!)^3 vertices whose solution allows to
determine the winning regions of the Muller game and a
winning strategy for one player.
The following technical report is available from
http://aib.informatik.rwth-aachen.de:
An Operational Semantics for Activity Diagrams using SMV
Shahar Maoz, Jan Oliver Ringert, Bernhard Rumpe
AIB 2011-07
This document defines an operational semantics for activity diagrams
(ADs) using a translation to SMV. The translation is inspired by the
work of Eshuis [Esh06] and extends it with support for data. Each
execution step of the SMV module obtained from an AD represents an
executed action of this AD with interleaved execution of concurrent
branches.
An implementation of the given translation was used in the context
of semantic differencing for ADs [MRR11]. We define the translation
and give two examples, showing ADs and their complete representation
in SMV.
The following technical report is available from
http://aib.informatik.rwth-aachen.de:
Toward Adjoint OpenMP
Michael Förster, Uwe Naumann, Jean Utke
AIB 2011-13
Shared-memory multiprocessing is becoming increasingly important in
high-performance scientific computing. Algorithmic differentiation
provides accurate derivative values and better runtime performance
of the adjoint model compared with finite differences. This paper
presents a source-to-source transformation of OpenMP augmented code
that can be used in source transformation tools for creating the
adjoint code of a given input code. Only some directives of the
OpenMP standard are considered here, namely, directives to
parallelize loops.
The following technical report is available from
http://aib.informatik.rwth-aachen.de:
A Linear Operational Semantics for Termination and Complexity Analysis
of ISO Prolog
Thomas Ströder, Fabian Emmes, Peter Schneider-Kamp, Jürgen Giesl,
Carsten Fuhs
AIB 2011-08
We present a new operational semantics for Prolog which covers all
constructs in the corresponding ISO standard (including "non-logical"
concepts like cuts, meta-programming, "all solution" predicates,
dynamic predicates, and exception handling). In contrast to the
classical operational semantics for logic programming, our semantics
is linear and not based on search trees. This has the advantage that
it is particularly suitable for automated program analyses such as
termination and complexity analysis. We prove that our new semantics
is equivalent to the ISO Prolog semantics, i.e., it computes the same
answer substitutions and the derivations in both semantics have
essentially the same length.