A Statistical Characterization Approach to Estimate Software Execution Time in Multiprocessor Systems

International audienceAs multiprocessor systems become increasinglypopular, the timing performance must be carefully evaluatedto enable sound design decisions before final implementation.Nonetheless, shared resources can cause timing behavior ofsoftware hardly predictable. This paper introduces ameasurement based on performance capture method coupledwith statistical inference to infer the detailed statisticalcharacteristics. It supports simulating various work scene andcompanion. A set of experiments are presented to validate theintroduced methods and different possible results are shown.Based on the inference result, the paper provides severalsolutions to representing the execution time in multi-coresystem