A Fuzzy Cuckoo-Search Driven Methodology for Design Space Exploration of Distributed Multiprocessor Embedded Systems
This chapter presents a methodology for conducting a Design Space Exploration (DSE) for Distributed Multi-Processor Embedded systems (DMPE). We introduce the notion of a Q-node to include quality-scaled tasks in the application model. A fuzzy rule-based r
- PDF / 705,490 Bytes
- 20 Pages / 439.37 x 666.142 pts Page_size
- 32 Downloads / 196 Views
Abstract This chapter presents a methodology for conducting a Design Space Exploration (DSE) for Distributed Multi-Processor Embedded systems (DMPE). We introduce the notion of a Q-node to include quality-scaled tasks in the application model. A fuzzy rule-based requirements elicitation framework allows the user to visualize and express the availability requirements in a flexible manner. We employ Cuckoo Search (CS), a metaheuristic that mimics the cuckoo birds’ breeding behavior, to explore the multi-objective design space. A fuzzy engine blends together multiple system objectives viz. Performance, Qualitative Availability and Cost-Effectiveness to calculate the overall fitness function. Experimental results illustrate the efficacy of the DSE tool in yielding high quality architectures in shorter run times and with lesser parameter tuning as compared with genetic algorithm. The fuzzy rules approach for fitness evaluation yields solutions with 24 % higher availability and 14 % higher performance as compared with a conventional approach using prefixed weights.
1 Introduction The Electronic Design Automation (EDA) industry has ushered in the era of pervasive computing where digital devices are indispensible for executing every aspect of modern civilization. Several EDA tools such as Vista from Mentor Graphics [1], EDA360 from Cadence [2], Platform Architect from Synopsys [3] and Simulink from Mathworks [4] are available for carrying out high level as well as low level system S. Chakraverty (B) Netaji Subhas Institute of Technology, Dwarka, Sector 3, New Delhi 110078, India e-mail: [email protected] A. Kumar Samsung Research Institute, Noida, India e-mail: [email protected]
M. A. Khan et al. (eds.), Embedded and Real Time System Development: A Software Engineering Perspective, Studies in Computational Intelligence 520, DOI: 10.1007/978-3-642-40888-5_5, © Springer-Verlag Berlin Heidelberg 2014
131
132
S. Chakraverty and A. Kumar
design. The field of EDA has been richly researched upon and utilized extensively by both academia and industry. Nevertheless, there are certain critical issues that specifically relate to the design of multi-objective DMPE systems. These issues need to be addressed more systematically and in a manner that supports active user participation in quantifying tradeoffs between conflicting design objectives. Multi-processor design optimization falls under the category of NP-Complete problems. Not only does it confront a very large search space but it has also to deal with several design objectives simultaneously. The system must be designed so as to achieve the desired real time performance levels, ensure a high degree of availability and accuracy at its service points and possess the ability to reconfigure in the presence of faults. All these deliverables must be achieved in a cost effective manner. Another level of complexity arises from the sheer diversity of implementation platforms that are available for executing the functional tasks. They include software implementations on a range of Instr
Data Loading...
