A Study on Setting Processor or CPU Affinity in Multi-Core Architecture for Parallel Computing
Abstract: Early days when we are having Single-core systems, then lot of work is to done by the single processor like scheduling kernel, servicing the interrupts, run the tasks concurrently and so on. The performance was may degrades for the time-critical applications. There are some real-time applications like weather forecasting, car crash analysis, military application, defense applications and so on, where speed is important. To improve the speed, adding or having Multi-core architecture or multiple cores is not sufficient, it is important to make their use efficiently. In Multi-core architecture, the parallelism is required to make effective use of hardware present in our system. It is important for the application developers take care about the processor affinity which may affect the performance of their application. When using Multi-core system to run an Multi threaded application, the migration of process or thread from one core to another, there will be a probability of cache locality or memory locality lost. The main motivation behind setting the processor affinity is to resist the migration of process or thread in Multi-core systems.
Keywords: Cache Locality, Memory Locality, Multi-core Architecture, Migration, Ping-Pong effect, Processor Affinity
How to Cite?: Saroj A. Shambharkar, "A Study on Setting Processor or CPU Affinity in Multi-Core Architecture for Parallel Computing", Volume 4 Issue 5, May 2015, International Journal of Science and Research (IJSR), Pages: 1987-1990, https://www.ijsr.net/getabstract.php?paperid=SUB154652, DOI: https://dx.doi.org/10.21275/SUB154652