CuPyBe documentation

CuPyBe is a collection of utilities that allow to convert the data contained in .cubex files produced by Score-P into data structures that can be interacted with through Python libraries (e.g., Pandas).

It is basically a bunch of functions that use cube_dump and parse its output (see the Cube Command line tools guide in html and its PDF version). For this reason, in order for CuPyBe to work it is necessary to have cubelib installed.

Main ideas

The two most important functions to use are possibly process_cubex and process_multi, which are used to parse a single (or many) .cubex files at once, yielding data that is mostly stored into pandas.DataFrames. In addition to that, the calltree is also stored in a recursive fashion using CubeTreeNode objects.

Notes

• The “system tree” is not dealt with at the moment. This means that data can be indexed only using Thread IDs instead of a somewhat more sophisticated way using nodes and MPI ranks. Only the “call tree” is implemented.

• The inclusive/exclusive conversions are not fully implemented. For those metrics that are “Inclusive convertible”, the conversion from exclusive to inclusive along the call tree have been implemented. Anyway it is possible to ask cube_dump for the inclusive metrics if one likes.

Table of content

• Modules
  • The merger module
  • The calltree module
  • Utilities for dealing with .cubex files
  • The metrics module
  • Tools for call tree conversions
  • Tools for index conversions
  • Tree parsing
• Examples
  • Bar Charts with Matplotlib
  • Interactive plots with Plotly

Indices and tables

• Index

• Module Index

• Search Page

Issues and bug reports

For issues and bug reports, please use the Issue feature on GitHub. In case this is inconvenient or not possible, write an email to michele.mesiti on the domain swansea.ac.uk.

License

CUPyBE is licensed according to the 3-clause license.