Title :
Static micro-scheduling: Resource contention relief in multithreaded programs
Author :
Yuanfang Chen ; Xiaoming Li
Author_Institution :
Electr. & Comput. Eng. Dept., Univ. of Delaware, Newark, DE, USA
Abstract :
Parallelism helps performance but at the same time stresses computer resources that are shared among threads. In this paper, we propose a low-overhead hardware counter based profiling method to accurately identify time-relevant contention locations in the program, then these contentions are mitigated so that performance of multithreading tasks can be boosted by the reduction of unnecessary contention cycles. In our preliminary experiment using NAS Parallel Benchmark (NPB), the contention searching algorithm is able to find an severe memory contention loop in FT code. After contention mitigation, more than 10% of the total cycles is eliminated, and the execution time of FT is reduced by 3% at the same time.
Keywords :
multi-threading; resource allocation; scheduling; NAS Parallel Benchmark; computer resource; contention cycle; contention mitigation; contention searching algorithm; low-overhead hardware counter based profiling method; memory contention loop; multithreaded program; multithreading task; parallelism; resource contention; static microscheduling; time-relevant contention location; Benchmark testing; Hardware; Instruction sets; Message systems; Multicore processing; Radiation detectors; USA Councils; compiler; contention; hardware counter; optimization; profiling;
Conference_Titel :
Performance Computing and Communications Conference (IPCCC), 2012 IEEE 31st International
Conference_Location :
Austin, TX
Print_ISBN :
978-1-4673-4881-2
DOI :
10.1109/PCCC.2012.6407691