Inexpensive Throughput Enhancement in Small-Scale Embedded Microprocessors with Differential Multithreading: Extended Results

Abstract

This paper examines differential multithreading (dMT) as an attractive organization for increasing throughput in simple, small-scale, pipelined processors like those used in embedded environments. dMT copes with pipeline stalls due to hazards and data- and instruction-cache misses by using duplicated pipeline registers to run instructions from an alternate thread. Results show that dMT boosts throughput substantially and can in fact replace dynamic branch prediction or can be used to reduce the sizes of the instruction and data caches. This report expands upon [1] by presenting extended results for additional dMT configurations. Note: Abstract extracted from PDF text