Session 6: programming

Abstract

omputing and especially programming is always going through some kind of evolution, with each generation being smarter and more powerful than the last one. In today’s competitive world the quality of software is a key to economic success and stability. The design of programming environments, their implementation and their evaluation in different application environments become more and more important. In this session we are investigating some latest developments in this field. The first paper “Exploiting the Multilevel Parallelism and the Problem Structure in the Numerical Solution of Stiff ODEs” is by J.M. Mantas Ruiz, J.O. Lopera and J.A. Carrillo. The paper presents a component-based methodology to derive parallel stiff ordinary differential equation solvers for distributed parallel machines. The methodology allows the exploitation of the multilevel parallelism of this kind of numerical algorithms and the particular structure of ODE systems by using parallel linear algebra modules. The results show that the parallel solver performs specially well with these ODE systems and quite well with narrow banded systems. The second paper by G. Spezzano, G. Folino and C. Pizzuti is entitled “Improving Induction Decision Trees with Parallel Genetic Programming”. A parallel genetic programming approach to induce decision trees in large data sets is presented. It is a new approach to induce decision trees for the data mining task of classification. The method is able to deal with large data sets since it uses a parallel implementation of genetic programming through the grid model. The third paper “Programming Distributed Systems with Group_IO” is by F. Guerra, J. Miranda, J.M. Santos, E. Martel, L. Hernandez and E.Pulido. The paper describes Group_IO, a library written in Ada which facilitates the construction of distributed applications by means of the group paradigm. Group_IO offers a simple interface to the implementation of reliable, atomic, causal, and uniform multicast, and it allows client-server interactions where the client may be a group. The fourth paper “A Technique to Build Ada Preprocessors” by J. Miranda, F. Guerra, J. Martin and A. Gonzalez proposes a technique to add a pre-processing phase to the GNU NYU Ada Translator, a front-end and runtime system for Ada 95. “Towards the Design of an Automatically Tuned Linear Algebra Library” is the title of the fifth paper by J. Cuenca, D. Gimenez and J. Gonzalez. This paper analyses a possible architecture of an automatically tuned parallel linear algebra library. It is applied on different parallel platforms and with several basic libraries. The experiments performed show that this technique may be used successfully to develop automatically tuned parallel linear algebra libraries in the future. C