Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Multiple pipeline scheduling in vector supercomputers

Conference ·
OSTI ID:5213744
A parallel task scheduling model is proposed for multipipeline vector processors. This model can be applied to explore maximal concurrency in vector computers, such as the Cray-1, Cyber-205, Star-100, TI-ASC, and IBM 3838. The optimisation problem of simultaneously scheduling multiple pipelines with vector tasks is shown to be np-complete. Thus, several heuristic scheduling algorithms, which can be easily implemented in vector processors with low system overhead and high throughput performance, have been developed. 17 references.
OSTI ID:
5213744
Report Number(s):
CONF-820829-
Country of Publication:
United States
Language:
English