For multiprocessor embedded systems, the dynamic voltagescaling (DVS) technique can be applied to scheduledapplications for energy reduction. DVS utilizes slack in theschedule to slow down processes and save energy. Therefore,it is generally believed that the maximal energy savingis achieved on a schedule with the minimum makespan(maximal slack). Most current approaches treat task assignment,scheduling, and DVS separately. In this paper,we present a framework called CASPER (Combined Assignment,Scheduling, and PowER-management) that challengesthis common belief by integrating task schedulingand DVS under a single iterative optimization loop via geneticalgorithm. We have conducted extensive experimentsto validate the energy efficiency of CASPER. For homogeneousmultiprocessor systems (in which all processors areof the same type), we consider a recently proposed slack distributionalgorithm (PDP-SPM) [3]: applying PDP-SPMon the schedule with the minimal makespan gives an averageof 53.8% energy saving; CASPER finds schedules withslightly larger makespan but a 57.3% energy saving, a 7.8%improvement. For heterogeneous systems, we consider thepower variation DVS (PV-DVS) algorithm [13], CASPERimproves its energy efficiency by 8.2%. Finally, our resultsalso show that the proposed single loop CASPER frameworksaves 23.3% more energy over GMA+EE-GLSA [12],the only other known integrated approach with a nestedloop that combines scheduling and power management inthe inner loop but leaves assignment in the outer loop.
展开▼
