An Elastic Architecture Adaptable to Various Application Scenarios

摘要: The quantity of computer applications is increasing dramatically as the industry prospers. Meanwhile, even for one application, it has different requirements performance and power in scenarios. Although various processors with architectures emerge to fit scenarios, impossible design a dedicated processor meet all requirements. Furthermore, dealing uncertain significantly aggravates burden programmers system integrators achieve specific performance/power. In this paper, we propose elastic architecture (EA) provide uniform computing platform high elasticity, i.e., ratio worst-case best-case performance/power/performance-power trade-off, which can applications. It achieved by dynamically adjusting parameters (instruction set, branch predictor, data path, memory hierarchy, concurrency, status & control, so on) on demand. elasticity our prototype implementation EA, Sim-EA, ranges from 3.31 14.34, 5.41 arithmetic average, SPEC CPU2000 benchmark suites, provides great flexibility fulfill Moreover, Sim-EA reduce EDP (energy-delay product) 31.14 % average compared baseline fixed architecture. Besides, some subsequent experiments indicate negative correlation between application intervals’ lengths their elasticities.