Student Highlight – Diogo Telmo Neves – PhD Thesis: “Separation of Concerns in Advanced Computing”

2010 April 30

by Joana Ferreira

The focus of Diogo’s PhD research work is to raise the abstraction level of parallel programming by developing a new set of programming constructs that promotes a stronger separation of concerns in parallel computing. The idea is to separate the domain specific code from parallelization issues, that is, hidden from programmers (such as scientists) as far as possible the complexity of parallel programming, by identifying parallelization concerns that can be specified as separate modules and investigating the use of different programming paradigms and programming constructs for each concern (e.g., a data flow model to specify data distributions, algorithmic skeletons to address coordination of tasks, etc.).

A framework named YaSkel (figure 1) was already developed as result of Diogo’s PhD research activities. YaSkel is based on Skeletons and Software Patterns, it allows parallelizing legacy code seamlessly through the use of the Dependency Injection Pattern. This work was presented by Diogo and his Portuguese supervisor – Professor João Luís Sobral – at ISPDC 2009 (International Symposium on Parallel and Distributed Computing).

At this moment Diogo is spending a semester, from (past) February to July, at the University of Texas at Austin integrated on a research group, which leader is Dr. Keshav Pingali – Diogo’s supervisor at UT Austin. During this period Diogo will be focused on his research, besides other tasks he will be working on graph algorithms. One of those algorithms constructs Phylogenetic trees (figure 2) upon a given set of taxa, and was written by Tandy Warnow et al. Despite the main goals will be to apply separation of concerns and to achieve better performance from parallelization of the algorithm, another important goal is to do applied science.