Conférence de Vivek Sarkar organisée par le département d'informatique.
In this talk, we claim that structured parallel programming greatly simplifies the task of writing correct and efficient parallel programs, while also capturing the flexibility of expressing asynchronous computations. We support this claim by introducing a family of structured parallel programming primitives developed in the Habanero Extreme Scale Software Research project at Rice University, and discussing semantic guarantees (e.g., deadlock freedom, datarace freedom, determinism, serial elision) for different subsets of parallel programs that can be constructed with these primitives. We will then summarize results for compiler optimizations, runtime systems , datarace detection and repair for these subsets. If time permits, we will briefly discuss future directions for data-driven repair and synthesis of parallel programs in the new Pliny project.
Voir aussi
|
Cursus :
Vivek Sarkar est professeur d'informatique à la Rice University. Il y mène notamment des recherches sur les langages de programmation et des programmes d'analyse afin d'améliorer les performances des logiciels informatiques.
Cliquer ICI pour fermer
Dernière mise à jour : 07/10/2015