The Second Workshop on Manufacturable and Dependable Multicore Architectures at Nanoscale (MEDIAN’13) Quantitative Analysis of Soft Error Propagation at RTL

Radiation-induced soft error is one of the main issues of system reliability with the continuous technology scaling. Soft error analysis at early design phase is essential for applying appropriate mitigation techniques to meet the reliability requirements. This paper proposes a novel approach to investigate the soft error propagation properties at behavioral register transfer level (RTL), especially for the control paths of the design. Modeling the control path as state transition system, the formal method probabilistic model checking is adopted to quantify the error sensitivities of all signals in the RTL design over multiple time steps. The experimental results reveal that error sensitivities across different variables show significant variance, which provides great potential for cost-efficient selective protection against soft error at higher abstraction levels.