Optimal Microarchitectural Design Configuration Selection for Processor Hard- Error Reliability

Ying Zhang,  Lide Duan,  Bin Li,  Lu Peng
Louisiana State University


Abstract

Traditional design space exploration mainly focuses on performance and power consumption. However, as one of the first-class constraints for modern processor design, hard-error reliability has not been well studied. In this paper, we investigate the relationship between microarchitectural configurations and hard-error reliability by exploring a large processor design space. We employ a rule search strategy,i.e. Patient Rule Induction Method, to generate a set of rules which choose optimal configurations for processor hard-error reliability and its tradeoff with performance and power consumption.