Virtual semi-concurrent self-checking for heterogeneous MPSoC architectures

Heterogeneous MPSoCs designed for non-safety-critical embedded applications (e.g, streaming applications) may allow a (limited) presence of errors in the application results before detection and confinement are effected. In turn, this allows to reduce the massive redundancy required to achieve fully concurrent error detection, reaching a balance between time elapsed before errors are detected and introduced area/power overhead. We envision in this paper the adoption of a modified semi-concurrent online checking approach that uses deferred time-multiplexed software emulation of redundant tasks on the available hardware platform rather than straightforward hardware redundancy. Our approach reduces the amount of redundancy requested by previous self-checking approaches and allows greater flexibility of operation; rigidly synchronous operation of units concurring in a checking action at any given time is avoided, introducing a delayed checking solution.