Webinar: How to identify causes of poor OpenMP parallel performance using the Intel VTune Amplifier
Often, after writing OpenMP code, our plots of speed-up and efficiency demonstrate suboptimal performance, and we want to understand why this is and where to focus our optimisation efforts. Intel’s VTune is a powerful software performance analysis tool, and can provide all the data we need for this. However, for the beginner it can be hard to know where to start with VTune, given the various options in terms of analysis type and data views, and the volume of data presented.
This talk is aimed at anyone who wants an introduction to using VTune to understand the causes of OpenMP underperformance. The talk describes a systematic way of using Intel’s VTune Amplifier to identify the sources of parallel inefficiency in OpenMP code, e.g. load imbalance, serial execution, OpenMP overheads and slowdown in processor throughput.
The webinar will include:
- Live VTune demo
- Advice on computing a range of useful performance metrics
- Examples of performance metrics from actual OpenMP code analysis
- How to use the instrumentation and tracing technology (ITT) API
About the presenter
Jonathan Boyle has several decades experience of working at a research software engineer, working primarily with Fortran, C, C++ and Python code, including MPI and OpenMP parallelisation. Recently Jonathan has been working at NAG with the Performance Optimisation and Productivity Centre of Excellence, offering performance optimisation services for parallel code.