Evaluating Performance And Energy Consumption Of Multithreaded Applications In Cc-Numa Multicore Processors

This paper discusses the methodology used in designing and implementing a full-system cc-NUMA multicore architectural simulator based on the open source gem5 simulation framework. We analyzed the performance and power consumption of multithreaded applications in cc-NUMA multicore processors and the experimental results of the PARSEC benchmark suite demonstrated that fine-tuning cc-NUMA multicore architectural parameters such as shared cache geometry and replacement policies can lead to significant performance improvement and energy reduction of multithreaded applications.