Summary: Architectural Extensions to Support Efficient Communication Using Message
Prediction
Ahmad Afsahi
Department of Electrical and Computer Engineering
Queen's University
Kingston, Canada K7L 3N6
ahmad@ee.queensu.ca
Nikitas J. Dimopoulos
Department of Electrical and Computer Engineering
University of Victoria
and Computer Engineering Laboratory, ITS,
Delft University of Technology
nikitas@ece.uvic.ca
Abstract
With increasing uniprocessor and SMP computation
power, workstation clusters are becoming viable alterna-
tives to high performance computing systems.
Communication overhead affects the performance of
parallel computers significantly. A significant portion of
the software communication overhead is attributable to

Source: Afsahi, Ahmad - Department of Electrical and Computer Engineering, Queen's University (Kingston)

Collections: Computer Technologies and Information Sciences