Past and Future Directions for Concurrent Task Scheduling

• Published in: Concurrent Objects and Beyond pp. 167 - 195
• Main Author: Halstead, Robert H.
• Format: Book Chapter
• Language: English
• Published: Berlin, Heidelberg Springer Berlin Heidelberg 2014
• Series: Lecture Notes in Computer Science
• Subjects: Fair Schedule; Future Object; Multicore Processor; Speculative Computing; Task Schedule
• ISBN: 9783662444702; 3662444704
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-662-44471-9_8

A wave of parallel processing research in the 1970s and 1980s developed various techniques for concurrent task scheduling, including work-stealing scheduling and lazy task creation, and various ideas for supporting speculative computing, including the sponsor model, but these ideas did not see large-scale use as long as uniprocessor clock speeds continued to increase rapidly from year to year. Now that the increase in clock speeds has slowed dramatically and multicore processors have become the answer for increasing the computing throughput of processor chips, increasing the performance of everyday applications on multicore processors by using parallelism has taken on greater importance, so concurrent task scheduling techniques are getting a second look. Work stealing and lazy task creation have now been incorporated into a wide range of systems capable of “industrial strength” application execution, but support for speculative computing still lags behind. This paper traces these techniques from their origins to their use in present-day systems and suggests some directions for further investigation and development in the speculative computing area.