A flexible instruction generation framework for extending embedded processors

Abstract

Modern platform-based design involves the domain-specific extension of embedded processors to fit customer requirements. To accomplish this task, the possibilities offered by recent extensible processors for their instruction set and microarchitectural customization have to be exploited. In this paper, a design approach that encapsulates automated workload characterization and highly-controllable instruction and application-specific functional unit (AFU) generation is utilized for fast extension space exploration of embedded processors. It is proved that a relatively small number of unique AFUs is needed in order to support embedded applications from the MiBench and Powerstone suites. It is possible to achieve 1.8times to 6.8times performance improvements although further possibilities such as subword parallelization are not currently regarded