Identifying thread interference with performance variation

Understanding the performance of a multi-threaded appli- cation is difficult. The threads interfere when they access the same hardware resource, which slows down their execution. Unfortunately, current profiling tools are unable to identify the most problematic interference, because they cannot classify interference on different hardware resources. In this paper, we propose a holistic metric able to simultaneously classify inter- ference on different hardware resources. The metric considers performance variation as a universal indicator. We propose an interactive profiling tool to compute this metric. With an evaluation of 27 applications we show that our profiling tool successfully identify 10 performance bottlenecks caused by 6 different kinds of interferenc