作者: Christian Bienia , Kai Li
DOI:
关键词:
摘要: Benchmarking has become one of the most important methods for quantitative performance evaluation processor and computer system designs. modern multiprocessors such as chip is challenging because their application domain, scalability parallelism requirements. In my thesis, I have developed a methodology to design effective benchmark suites demonstrated its effectiveness by developing deploying suite evaluating multiprocessors. More specifically, this thesis includes several contributions. First, shows that new needed behavior parallel programs significantly different from those represented SPLASH-2, popular over ten years ago. Second, quantitatively describes requirements characteristics set multithreaded underlying technology trends. Third, presents systematic approach scale select inputs with goal optimizing benchmarking accuracy subject constrained execution or simulation time. Finally, called PARSEC shared-memory Since initial release, been adopted many architecture groups in both research industry.