Abstract | ||
---|---|---|
Hardware redundancy is a common method for improving the reliability of a system. The disadvantage of this approach is the hardware overhead and the additional power consumption. This contribution proposes a strategy for implementing low-cost triple modular redundancy (TMR) on coarse-grained reconfigurable architectures (CGRAs). Low-cost TMR is achieved by utilizing unused functional units (FUs) for the redundant computation of results. This is realized by combining the FUs of three processing elements in a fault-tolerant data path. Experimental results show that the proposed approach reduces area in a 8-bit architecture by 12.8% and average power consumption is decreased between 1.6 and 18.6% when compared with a fault-tolerant CGRA implemented by conventional TMR. |
Year | DOI | Venue |
---|---|---|
2011 | 10.1109/ReConFig.2011.57 | Reconfigurable Computing and FPGAs |
Keywords | Field | DocType |
low-cost triple modular redundancy,average power consumption,hardware overhead,coarse-grained reconfigurable architectures,low-cost tmr,conventional tmr,fault-tolerant data path,additional power consumption,hardware redundancy,fault-tolerant cgra,fault tolerant,redundancy,computer architecture,fault tolerance,functional unit,fault tolerant system,error detection,reliability | Data path,Computer science,Parallel computing,Triple modular redundancy,Real-time computing,Error detection and correction,Fault tolerance,Redundancy (engineering),Computation,Power consumption,Hardware redundancy,Embedded system | Conference |
ISBN | Citations | PageRank |
978-1-4577-1734-5 | 11 | 0.65 |
References | Authors | |
8 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Thomas Schweizer | 1 | 74 | 10.73 |
Philipp Schlicker | 2 | 11 | 0.65 |
Sven Eisenhardt | 3 | 36 | 4.54 |
Tommy Kuhn | 4 | 51 | 6.75 |
Wolfgang Rosenstiel | 5 | 1462 | 212.32 |