Extra-P

Download Build requirements Documentation ChangeLog

 

Extra-P is an automatic performance-modeling tool that supports the user in the identification of scalability bugs. A scalability bug is a part of the program whose scaling behavior is unintentionally poor, that is, much worse than expected. A performance model is a formula that expresses a performance metric of interest such as execution time or energy consumption as a function of one or more execution parameters such as the size of the input problem or the number of processors.

Extra-P uses measurements of various performance metrics at different execution configurations as input to generate performance models of code regions (including their calling context) as a function of the execution parameters. All it takes to search for scalability issues even in full-blown codes is to run a manageable number of small-scale performance experiments, launch Extra-P, and compare the asymptotic or extrapolated performance of the worst instances to the expectations.

Extra-P generates not only a list of potential scalability bugs but also human-readable models for all performance metrics available such as floating-point operations or bytes sent by MPI calls that can be further analyzed and compared to identify the root causes of scalability issues.

Extra-P is developed by TU Darmstadt – in collaboration with ETH Zurich. For questions regarding Extra-P please send a message to the Extra-P team.

This software is free but copyrighted

© 2020 Technical University of Darmstadt, Darmstadt, Germany

By downloading and using this software you automatically agree to comply with the regulations as described in the Extra-P license agreement.

VersionDateDescription
Latest

Latest Release 

The latest release can always be found on GitHub.

 

source (GitHub)

Python package
Extra-P 4.0 packages for download
4.0 12-Nov-2020

This release is completely rewritten in Python; it can be found on GitHub.

  • Sparse modeling support
  • Support for more input files
  • A python-based version of Extra-P
  • Improved Python-based command-line tool
  • Minor improvements and fixes

source (GitHub)

Python package

Supported Platforms

Extra-P has been successfully tested on the following platforms:

  • Linux (x86_64)
  • Windows 10 (x86_64)
  • MacOS (x86_64, ARM64)

Please report success/failure on other platforms to the Extra-P development team.

 

Older Versions

This software is free but copyrighted

© 2015-2018 Technische Universität Darmstadt, Darmstadt, Germany
© 2014-2015 Lawrence Livermore National Laboratory, Livermore, California, USA
© 2009-2015 German Research School for Simulation Sciences GmbH, Jülich/Aachen, Germany
© 1998-2015 Forschungszentrum Jülich GmbH, Germany

By downloading and using this software you automatically agree to comply with the regulations as described in the Extra-P license agreement.

VersionDateDescription
3.0 07-Feb-2018

New requirements:

Includes:

Command line tools

  • Added the command line tool extrap-modeler to create a Extra-P experiment file with input data and models on a shell.
  • Refactored extrap-print to produce a console output of an Extra-P experiment file. Removed the 95% confidence interval from the extrap-print output. Added additional options to get a list of definitions.

Multi-parameter support

  • Added possibility to open set of CUBE files with multiple parameters. The number of parameters can be specified before loading the CUBE files. After a change to the number of parameters, it tries to guess the structure of the measurement directories.
  • Added a model generator for multi-parameter models. Depending whether the experiment contains only one parameter or multiple parameters, the single-parameter model generators or the multi-parameter model generator can be selected for model generation.
  • Several new displays for two-parameter models:
    • A display that plots selected models next to each other in distinct surface plots
    • A display that plots all selected models in the same surface plot
    • A scatter plot that shows for each parameter only the dominating function
    • The dominating model as the surface plot
    • The dominating model as a heat map
    • Contour plots
    • Interpolated contour plots
    • All selected models in the same surface plot with transparent surface and measurement points
  • Added projection possibilities. Thus, when more parameters are available than the display can handle, it shows a orthogonal projection. The user can choose which parameter is shown on which axis and the value for the remaining parameters.

Minor improvements and fixes

  • Fixed Opening of experiment files which contain a model generated with the refining model generator.
  • Allow to pass an Extra-P experiment file to the GUI via command line argument. The file is then opened on startup of the GUI.
Extra-P 3.0 packages for download
source
MD5sum : 4a5e6996be73aff870d2bde52f4aacd6
Extra-P 2.0 packages for download
source
MD5sum : 6ec38ae6fb15575e7c05a2a0e8b0c993
Features of previous releases can be found in the changelog.

Supported Platforms

Extra-P has been successfully tested on the following platforms:

  • Linux (x86_64)
  • Mac OS X (x86_64) (10.8 and younger)

Please report success/failure on other platforms to the Extra-P development team.