Scalasca  (Scalasca 2.3.1, revision 14987)
Scalable Performance Analysis of Large-Scale Applications
Getting started

This chapter provides an introduction to the use of the Scalasca Trace Tools on the basis of the analysis of an example application. The most prominent features are addressed, and at times a reference to later chapters with more in-depth information on the corresponding topic is given.

Use of the Scalasca Trace Tools involves three phases: instrumentation of the target application, execution measurement collection and analysis, and examination of the analysis report. For instrumentation and measurement, the Scalasca Trace Tools 2.x release series leverages the Score-P infrastructure, while the Cube graphical user interface is used for analysis report examination. Scalasca complements the functionality provided by Cube and Score-P with scalable automatic trace-analysis components, as well as convenience commands for controlling execution measurement collection and analysis, and analysis report postprocessing.

Most of Scalasca's functionality can be accessed through the scalasca command, which provides action options that in turn invoke the corresponding underlying commands scorep, scan and square. These actions are:

  1. scalasca -instrument

    (or short skin) familiar to users of the Scalasca 1.x series is deprecated and only provided for backward compatibility. It tries to map the command-line options of the Scalasca 1.x instrumenter onto corresponding options of Score-P's instrumenter command scorep – as far as this is possible. However, to take full advantage of the improved functionality provided by Score-P, users are strongly encouraged to use the scorep instrumenter command directly. Please refer to the Score-P User Manual [13] for details. To assist in transitioning existing measurement configurations to Score-P, the Scalasca instrumentation wrapper prints the converted command that is actually executed to standard output.

  2. scalasca -analyze

    (or short scan) is used to control the Score-P measurement environment during the execution of the target application (supporting both runtime summarization and/or event trace collection, optionally including hardware-counter information), and to automatically initiate Scalasca's trace analysis after measurement completion if tracing was requested.

  3. scalasca -examine

    (or short square) is used to postprocess the analysis report generated by a Score-P profiling measurement and/or Scalasca's automatic post-mortem trace analysis, and to start the analysis report examination browser Cube.

To get a brief usage summary, call the scalasca command without arguments, or use scalasca --quickref to open the Scalasca Quick Reference (with a suitable PDF viewer).

Under the hood, the Scalasca convenience commands leverage a number of other commands provided by Score-P and Cube. Therefore, it is generally advisable to include the executable directories of appropriate installations of all three components in the shell search path (PATH).

The following three sections provide a quick overview of each of these actions and how to use them during the corresponding step of the performance analysis, before a tutorial-style full workflow example is presented in Section A full workflow example.


Scalasca    Copyright © 1998–2016 Forschungszentrum Jülich GmbH, Jülich Supercomputing Centre
Copyright © 2009–2015 German Research School for Simulation Sciences GmbH, Laboratory for Parallel Programming