Bottleneck identification and scheduling in multithreaded applications

Performance of multithreaded applications is limited by a vari-ety of bottlenecks, e.g. critical sections, barriers and slow pipeline stages. These bottlenecks serialize execution, waste valuable exe-cution cycles, and limit scalability of applications. This paper pro-poses Bottleneck Identification and Scheduling (BIS), a cooperative software-hardware mechanism to identify and accelerate the most critical bottlenecks. BIS identifies which bottlenecks are likely to reduce performance by measuring the number of cycles threads have to wait for each bottleneck, and accelerates those bottle-necks using one or more fast cores on an Asymmetric Chip Multi-Processor (ACMP). Unlike previous work that targets specific bot-tlenecks, BIS can identify and accelerate bottlenecks regardless of their type. We compare BIS to four previous approaches and show that it outperforms the best of them by 15 % on average. BIS ’ per-formance improvement increases as the number of cores and the number of fast cores in the system increase