The International Conference for High Performance Computing, Networking, Storage and Analysis
Efficient Datarace Detection in a Structured Programming Language.
Student: Trudy Lynn Firestone (University of Utah)
Supervisor: Ganesh Gopalakrishnan (University of Utah)
Abstract: Datarace detection is increasingly important in Computer Science as non-specialized computer scientists are required to write more parallel programs for better performance. Be- cause dataraces are an extremely common bug, detection of these conflicts is a necessary debugging tool. Unfortunately discovering dataraces is computationally expensive. How- ever, semantics guarantees provided by structured paral- lelism make it possible to create an efficient solution. To this end, we look to expand the race detector in the structured language, Habanero Java (HJ). Building upon HJ’s current race detector, which supports async/finish constructs and isolated regions, this research theoretically extends HJ’s al- gorithm to include the phaser construct.