Optimal Dependability and Fine Granular Error Resilience Methodology for Reconfigurable Systems

Adaptivity to changing demands is a key feature of many systems in computationally challenging applications. This feature achieved with reconfigurable hardware, providing a good balance between implementation efficiency and flexibility. However, for some areas the specification may also include a high level of system dependability. In particular, FPGA-based implementations may be used in systems that require long-time dependable operation over several years. This paper presents a new approach for online self-detecting and handling permanent faults in the field programmable gate array (FPGA) based soft processors with the lowest overhead in time and area. The method presented includes both the fine-grain diagnosis of faults in an FPGA and a technique for repair by re-configuration at the lowest cost. The results were achieved using Xilinx FPGAs based on the 28 nm technology and two soft processors.