Provided by: Democritus University of Thrace
Date Added: Jan 2013
Fault-tolerance is a crucial challenge for a number of application domains. Existing solutions to this problem are applied uniformly at the entire design imposing among others mentionable delay and power overheads. In this paper, the authors introduce a software-supported framework based on genetic algorithm for supporting fast application placement under fault-tolerant constraints. Rather than relevant placers that are executed sequentially, the proposed one exhibits inherent parallelisms, which can bene t from multi-core processors. Experimental results with various benchmarks showed that their framework leads to an average reduction of execution run-time by 6.3 with an increase in maximum operating frequency by 12% (on average).