Error detection using dynamic dataflow verification

Abstract

A significant fraction of the circuitry in a modern processor is dedicated to converting the linear instruction stream into a representation that allows the execution of instructions in data dependence order, rather than program order, to extract instruction level parallelism. All errors caused by hardware faults in this circuitry - which includes the fetch and decode stages, renaming and scheduling logic, as well as the commit stage - will manifest themselves as incorrectly constructed dataflow graphs. Dynamic Dataflow Verification (DDFV) compares the dynamically constructed and executed dataflow graph to the expected dataflow graph of the static program binary, represented by a signature embedded in the instruction stream. The signature comparison enables comprehensive detection of transient errors, permanent errors, and design bugs in the dataflow circuitry. We show that DDFV detects errors with high probability, at a low hardware and performance cost. © 2007 IEEE.