The University of British Columbia

This paper proposes a novel technique for preventing a wide range of data errors from corrupting the execution of applications. The proposed technique enables automated derivation of fine-grained, application-specific error detectors. An algorithm based on dynamic traces of application execution is developed for extracting the set of error detector classes, parameters, and locations in order to maximize the error detection coverage for a target application. The paper also presents an automatic framework for synthesizing the set of detectors in hardware to enable low-overhead runtime checking of the application execution. Coverage (evaluated using fault injection) of the error detectors derived using the proposed methodology, the additional hardware resources needed, and performance overhead for several benchmark programs are also reported