GMICRO/500 microprocessor: pipeline structure of superscalar architecture

Abstract

The GMICRO/500 pipelined instruction execution mechanism is described. The 5-stage dual-pipeline superscalar architecture is examined and examples of basic instruction execution timing are analyzed, illustrating the effect of a pipe bypass mechanism and dedicated resident branch instruction caches. The benefit of microprogram-controlled instruction execution for high-speed execution of high-level language instructions is shown. Overall GMICRO/500 performance is evaluated in Dhrystones.<>