VI-HPS

Introduction

This page introduces the performance tools developed by the VI-HPS Institute (Virtual Institute for High Productivity Supercomputing). The following tools are installed on CSCS systems and can assist developers of simulation codes to address their needs in performance analysis:

  • Score-P is a highly scalable instrumentation and measurement infrastructure for profiling, event tracing, and online analysis. It supports a wide range of HPC platforms (CRAY, IBM, Linux) and programming models (MPI, OpenMP and OpenACC/CUDA). Score-P provides core measurement services for a range of specialized analysis tools, such as Vampir, Scalasca, TAU, or Periscope.
  • Scalasca supports the performance optimization of parallel programs with a collection of scalable trace-based tools for in-depth analyses of concurrent behavior. The analysis identifies potential performance bottlenecks - in particular those concerning communication and synchronization - and offers guidance in exploring their causes.
  • Vampir is a performance visualizer that allows to quickly study a program's runtime behavior at a fine level of details. This includes the display of detailed performance event recordings over time in timelines and aggregated profiles. Interactive navigation and zooming are the key features of the tool, which help to quickly identify inefficient or faulty parts of a program.

  • All these tools are available in a single modulefile:

    •     module load Score-P/3.0-CrayCCE-2016.11
    •     module load Score-P/3.0-CrayGNU-2016.11
    •     module load Score-P/3.0-CrayIntel-2016.11
    •     module load Score-P/3.0-CrayPGI-2016.11

    and

    •     module load Scalasca/2.3.1-CrayGNU-2016.11
    •     module load Scalasca/2.3.1-CrayPGI-2016.11
    •     module load Scalasca/2.3.1-CrayIntel-2016.11

CSCS training material

In October 2014, CSCS organised a course focusing on "Practical Performance Analysis of Parallel Applications". The goal of this course is to let the users familiarise and practice the tools described above, analyse their parallel application codes and learn performance analysis techniques that can improve their execution performance and scalability.

The course starts with basic application instrumentation and measurement to generate execution profiles, then improves measurement quality via customization capabilities, and ends with interactive and automated analyses of execution traces.

The material (presentations and hands-on practical exercises on Piz Daint Cray XC30 system) is available here:

  1. CSCS YouTube channel
  2. Workshop outline
  3. Intro-VIHPS.pdf
  4. Engineering.pdf
  5. BTMZ-PizDaint.pdf
  6. Score-P.pdf
  7. Cube.pdf
  8. Vampir.pdf
  9. Scalasca.pdf
  10. jacobi_HandsOn.pdf
  11. EnergyCounters_OpenACC.pdf
  12. CASITA.pdf

Troubleshooting

For further information please check:

If everything fails, please contact help(at)cscs.ch.