Abstract
Processor self-scheduling is an effective distributed dynamic scheduling scheme for parallel nested loops in multiprocessor systems. Since self-scheduling is nonpreemptive and uses busy-waiting as the basic technique for task synchronization, deadlocks may possibly arise. Conditions that allow deadlock-free processor self-scheduling are identified. Control tokens and data tokens are used to model the precedence relation and data dependences. Control tokens, data tokens, and processors are regarded as three kinds of resources needed by tasks. Using this resource model, three possible self-scheduling schemes with different degrees of parallelism are studied, and a proposal for preventing deadlocks is given.
Original language | English (US) |
---|---|
Title of host publication | Proceedings of the International Conference on Parallel Processing |
Editors | Sartaj K. Sahni |
Publisher | Pennsylvania State Univ Press |
Pages | 11-18 |
Number of pages | 8 |
ISBN (Print) | 0271006080 |
State | Published - Dec 1 1987 |
Event | Proc Int Conf Parallel Process 1987 - Universal Park, PA, USA Duration: Aug 17 1987 → Aug 21 1987 |
Other
Other | Proc Int Conf Parallel Process 1987 |
---|---|
City | Universal Park, PA, USA |
Period | 8/17/87 → 8/21/87 |