Design and analysis of efficient synthesis algorithms for EDAC functions in FPGAs

Abstract

Error detection and correction (EDAC) functions have been widely used for protecting memories from single event upsets (SEU), which occur in environments with high levels of radiation or in deep submicron manufacturing technologies. This paper presents three novel synthesis algorithms that obtain area-efficient implementations for a given EDAC function, with the ultimate aim of reducing the number of sensitive configuration bits in SRAM-based field-programmable gate arrays (FPGAs). Having less sensitive bits results in a lower chance of suffering an SEU in the EDAC circuitry, thus improving the overall reliability of the whole system. Besides minimizing area, the proposed algorithms also focus on improving other figures of merit like circuit speed and power consumption. The executed benchmarks show that, when compared with other modern synthesis tools, the proposed algorithms can reduce the number of utilized look-up tables (LUTs) up to a 34.48%. Such large reductions in area usage ultimately result in reliability improvements over 10% for the implemented EDAC cores, measured as MTBF (mean time between failures). On the other hand, maximum path delays and power consumptions can be reduced up to a 17.72% and 34.37%, respectively, on the placed and routed designs.