Towards Fault Simulation at Mixed Register-Transfer/Gate-Level Models

Abstract

Safety-critical designs used in automotive applications need to ensure reliable operations even under hostile operating conditions. As these designs grow in size and complexity, they are facing an increased risk of failure. Consequently, the methods applied to validate the reliability of designs require increasingly more compute resources (e.g., fault simulation time) and manual efforts. Rigorous and highly automated safety analysis methods are needed to cope with this rising complexity. In this paper, we propose a model-based safety analysis flow to enable fault injection at different abstraction levels of a design. The fault simulation is performed at register transfer level (RTL) of a design, in which parts of the design targeted for fault simulation are represented with gate-level granularity. This mixed representation of a design provides a significant rise in fault simulation performance while maintaining the same accuracy as a gate-level fault simulation. To demonstrate the applicability of the proposed approach, various RISC- V based CPU subsystems that are part of automotive SoCs are considered for fault simulation. The experimental results show an increase of 3.5x - 8.4x in the fault simulation performance with substantially less manual effort as all the design activities are automated utilizing a model-driven RTL generation flow.