Diverse Compiling for Microprocessor Fault Detection in Temporal Redundant Systems

As hardware components are expected to become ever more unreliable due to the technology scaling, hardware errors have become unavoidable. Dependable systems that rely on a correct functionality often use redundancy to detect such hardware faults during operation. However, to design costefficient reliable systems, it is crucial to effectively exploit the available redundancy. Thus, researchers have investigated ways to automatically add cost-efficient diversity to software to increase the efficiency of redundancy strategies. One of these automated software diversification methods is diverse compiling, which exploits the diversity introduced by different compilers and different optimization flags. However, there is a lack of statistics regarding the efficiency of the approach for detecting certain types of faults. In this paper, we contribute towards filling this gap by evaluating the diverse compiling approach regarding its ability to detect faults in the microprocessor. We experimentally quantify the efficiency of diverse compiling for three benchmarks regarding the detection of register and instruction decoder faults for an ARM9 processor. For these applications, our fault injection campaigns show that about 90% register faults and 70% instruction decoder faults can be detected with diverse compiling. These results indicate that this approach is quite promising for improving the efficiency of redundancy techniques.