The development of large computational codes such as those used to perform simulations in quantum mechanics to calculate properties of materials, or to predict the aerodynamics forces around airplanes, still typically requires several human-years. However the pace of research and education demands much more rapid software tool development. A goal of this project is to discover how programming languages can be adopted to particular domains in order to more effectively develop high performance large scale codes.
Our hypothesis is that these problems can, at least partially, be effectively addressed if developers had interactive development tools that can be extended and customized for their particular problem and application domain. We are focusing on materials science as a case study owing to our experience of code development in this area, but our aim is to reach a general audience in computational science.
We are thus developing an interactive extensible language framework for scientific software development specifically aimed at rapid software development. We are augmenting programming languages such as ANSI C with high level language constructs to allow developers to more directly specify problem solutions. We will also develop interactive frameworks for easily experimenting with expressions written in these extended languages. Such interactive frameworks, such as MATLAB, have proven helpful in writing and understanding solutions to computational problems. This approach of creating an interactive extensible language framework will provide a way to help speed development of large scale computer software.
Eric Van Wyk & Yousef Saad, University of Minnesota
Jim Chelikowsky, University of Texas at Austin