Developing a runtime for running SVP programs across networks of multi-core machines

From Master Projects
Jump to: navigation, search

has title::Runtime System for Running SVP Programs Across Networks of Multi-core Machines
status: ongoing
Master: project within::Parallel and Distributed Computer Systems
Student name: student name::Andrei MATEI
Start start date:=2010/02/01
End end date:=2010/08/28
Supervisor: Chris Jesshope (UvA)
Second reader: has second reader::Andrew S. Tanenbaum
Poster: has poster::Media:Media:Posternaam.pdf

Signature supervisor



SVP is a parallel programming paradigm developed by the Computer Systems Architecture group at the UvA, lead by Prof. Chris Jesshope. Contrasting with other concurrency models, SVP has a number of attractive properties: it allows one to express all the concurrency available in an algorithm, it captures locality in communication and the concurrency and synchronization specified always allow a well-defined sequential schedule.

The objective of this project is multifold: - design an SVP runtime that permits distribution of work over a network, implementing a distributed shared memory system - design a mapping engine - a server that makes delegation decisions for computations - devising techniques for iteratively adapting the mapping decisions according to the run-time behavior of the software components