On performance analysis of a multithreaded application parallelized by different programming models using intel VTune

Abstract

Multi-core processors are ubiquitous. Extracting the desired performance from them requires efficient techniques for partitioning a single piece of work into multiple fine-grained units of work in order to process them simultaneously. Understanding the performance behavior of a parallel system requires a close familiarity with the underlying architecture and the hardware counters. We present a performance analysis study of a multi-core system by a state-of-the-art parallel performance analyzer tool, the Intel VTune Performance Analyzer. We chose as a test-case a classic nested-loop application that exhibits unexpected performance gains using two different programming models on the same multi-core system. Our expectations were to be able to reason about the performance results by exploring the application behavior using the parallel analyzer tool. We found that it is very difficult to explain high-level performance measurements of multi-core systems by low-level hardware diagnosis. © 2011 Springer-Verlag Berlin Heidelberg.