Virtual Institute — High Productivity Supercomputing

SC'13 tutorial: Hands-on Practical Hybrid Parallel Application Performance Engineering (Denver, CO, USA)

Date

Sunday 17th November 2013

Presenters

  • Markus Geimer, Jülich Supercomputing Centre
  • Sameer Shende, University of Oregon
  • Bert Wesarg, Technische Universität Dresden
  • Brian Wylie, Jülich Supercomputing Centre

Location

The tutorial takes place as part of the SC'13 conference in room 301 of the Colorado Convention Center, Denver, CO, USA. In preparation for the hands-on exercises, participants are encouraged to download the latest VI-HPS Linux ISO for execution within VirtualBox on their notebook computer.

Abtract

This tutorial presents state-of-the-art performance tools for leading-edge HPC systems founded on the Score-P community instrumentation and measurement infrastructure, demonstrating how they can be used for performance engineering of effective scientific applications based on standard MPI, OpenMP, hybrid MPI+OpenMP, and increasingly common usage of accelerators. Parallel performance evaluation tools from the VI-HPS (Virtual Institute High Productivity Supercomputing) are introduced and featured in hands-on exercises with Scalasca, Vampir and TAU. We present the complete workflow of performance engineering, including instrumentation, measurement (profiling and tracing, timing and PAPI hardware counters), data storage, analysis, and visualization. Emphasis is placed on how tools are used in combination for identifying performance problems and investigating optimization alternatives. Using their own notebook computers with a provided Linux Live-ISO image containing the tools (booted from DVD/USB or within a virtual machine) will help to prepare participants to locate and diagnose performance bottlenecks in their own parallel programs.

Programme

08:30 Introduction
  • Introduction to VI-HPS & parallel application engineering
  • VI-HPS Linux Live-ISO and MPI+OpenMP example
  • Instrumentation & measurement of applications with Score-P
  • 10:00 (break)
    10:30 Profiling
  • Profile examination with CUBE
  • Configuration & customization of Score-P measurements
  • Specialized Score-P measurements & analyses
  • 12:00
    (lunch)
    13:30 Tracing
  • Automated trace analysis with Scalasca
  • Interactive trace analysis with Vampir
  • 15:00 (break)
    15:30 Further steps
  • Profile examination with TAU ParaProf
  • Performance data management & mining with TAU PerfExplorer
  • Finding typical parallel performance bottlenecks
  • Review & conclusion
  • 17:00 (adjourn)

    Registration

    Registration via the SC'13 conference website (or on-site) is possible with or without including the conference technical program, exhibition and workshops.

    Contact

    Brian Wylie
    Jülich Supercomputing Centre
    Phone: +49 2461 61-6589
    Email: b.wylie@fz-juelich.de