Scalasca  (Scalasca 2.3.1, revision 14987)
Scalable Performance Analysis of Large-Scale Applications
Analysis report examination

The results of the runtime summarization and/or the automatic trace analysis are stored in one or more reports (i.e., CUBE4 files) in the measurement experiment directory. These reports can be postprocessed and examined using the scalasca -examine (or short square) command, providing an experiment directory name as argument:

  % scalasca -examine [options] <experiment_name>

Postprocessing is done the first time an experiment is examined, before launching the Cube analysis report browser. If the scalasca -examine command is provided with an already processed experiment directory, or with a CUBE4 file specified as argument, the viewer is launched immediately.

Instead of interactively examining the measurement analysis results, a textual score report can also be obtained using the -s option without launching the viewer:

  % scalasca -examine -s <experiment_name>

This score report comes from the scorep-score utility and provides a breakdown of the different types of regions included in the measurement and their estimated associated trace buffer capacity requirements, aggregate trace size and largest process trace buffer size (max_buf), which can be used to set up a filtering file and to determine an appropriate SCOREP_TOTAL_MEMORY setting for a subsequent trace measurement. See Section Optimizing the measurement configuration for more details.

The Cube viewer can also be directly used on an experiment archive – opening a dialog window to choose one of the contained CUBE4 files – or an individual CUBE4 file as shown below:

  % cube <experiment_name>
  % cube <file>.cubex

However, keep in mind that no postprocessing is performed in this case, so that only a subset of Scalasca's analyses and metrics may be shown.



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