Power-efficient clustering via incomplete bypassing

Abstract

Researchers have proposed clustered microarchitectures for performance and energy efficiency. Typically, clustered microarchitectures offer fast, local bypassing between instructions within clusters but global bypasses are slower. Traditional clustered microarchitectures (TCM) are implemented by partitioning the register file and associated functional units to clusters. This paper demonstrates an alternate implementation - Incomplete bypass-based clustered microarchitecture (IBCM). IBCM reduces the length of bypass wires by 42.4% resulting in an 8.9% reduction of "Execute" stage delay. This delay reduction in the critical EX stage enables voltage scaling that results in significantly lower average power consumption (between 11.7% and 19.5% lower) while achieving identical performance.