Summary: The Amber System: Parallel Programming
on a
Network of Multiprocessors
Jeffrey S. Chase, Franz G. Amador, Edward D. Lazowska
Henry M. Levy and Richard J. Littlefield
Department of Computer Science
University of Washington
Seattle, WA 98195
Abstract
Microprocessor­based shared­memory multiprocessors are becoming widely available and
promise to provide cost­effective high­performance computing.
This paper describes a programming system called Amber which permits a single application
program to use a homogeneous network of multiprocessors in a uniform way, making the net­
work appear to the application as an integrated, non­uniform memory access, shared­memory
multiprocessor. This simplifies the development of applications and allows compute­intensive
parallel programs to effectively harness the potential of multiple nodes.
Amber programs are written using an object­oriented subset of the C++ programming lan­
guage, supplemented with primitives for managing concurrency and distribution. Amber pro­
vides a network­wide shared­object virtual memory in which coherence is provided by hardware
means for locally­executing threads, and by software means for remote accesses. Amber runs

Source: Anderson, Richard - Department of Computer Science and Engineering, University of Washington at Seattle

Collections: Computer Technologies and Information Sciences