Paper Info
Title
Run-time Task Overlapping on Multiprocessor Platforms
Abstract
Today's embedded applications often consist of multiple concurrent tasks. These tasks are decomposed into sub- tasks which are in turn assigned and scheduled on mul- tiple different processors to achieve the optimal perfor- mance/energy combination. Previous work introduced sys- tematical approaches to make performance-energy trade- offs explorations for each individual task and used the exploration results at run-time to fulll system-level con- straints. However, they did not exploit the fact that the concurrent tasks can be executed in an overlapped fash- ion. In this paper, we propose a simple yet powerful on-line technique that performs task overlapping by run-time sub- task re-scheduling. By doing so, a multiprocessor system with concurrent tasks can achieve better performance with- out extra energy consumption. We have applied our algo- rithm to a set of randomly-generated task graphs, obtain- ing encouraging improvements over non-overlapped task, and also having less overall energy consumption than a previous DVS method for real-time tasks. Then, we have demonstrated the algorithm on real-life video- and image- processing applications implemented on a dual-processor TI TMS320C6202 board: We have achieved a reduction of 22-29% in the application execution time, while the im- pact of run-time scheduling overhead proved to be negligi- ble (1.55%).
Year
DOI
Venue
2010
10.1007/s11265-008-0302-7
Journal of Signal Processing Systems
Keywords
Field
DocType
Task scheduling,On-line optimization,Multiprocessor platforms
Graph theory,Concurrency control,Scheduling (computing),Computer science,Parallel computing,Multiprocessing,Real-time computing,Exploit,Execution time,Computer multitasking,Energy consumption,Distributed computing
Journal
Volume
Issue
ISSN
60
2
1939-8018
Citations 
PageRank 
References 
2
0.40
17
Authors
3
Name
Order
Citations
PageRank
Zhe Ma1153.19
Daniele Paolo Scarpazza2987.43
Francky Catthoor321.75