PyPOP

Documentation: https://numericalalgorithmsgroup.github.io/pypop/doc.html

A python package for calculating POP metrics from application profiles, primarily designed for literate programming using Jupyter notebooks.

PyPOP currently consumes Extrae *.prv traces, but is designed with a view to adding support for additional formats.

Requirements

• Extrae (for trace capture)
• Paraver/Paramedir (for trace analysis)
• Dimemas optional (for ideal network analysis)
• Numpy
• Pandas

Paramedir and Dimemas must be available on the system PATH (Linux $PATH or Windows %PATH% variables) so that they can be found by PyPOP.

Installation

PyPOP itself can be installed using the pip package manager:

$ pip install [--user] NAG-PyPOP

The optional --user directive instructs pip to install to the users home directory instead of the system site package directory.

N.B: Some linux distributions are still shipping Python2 alongside Python3. PyPOP requires Python3 and so you may need to install using the pip3 command rather than the pip command.

Note that the Extrae tool is needed to collect application traces and the Paraver/Paramedir and Dimemas tools must be installed for trace Analysis.

Usage

Jupyter notebooks are intended to be the primary interface to PyPOP. This guide uses several example notebooks to demonstrate the core functionality of PyPOP for calculation of the POP Metrics as well as advanced analysis of trace files.

PyPOP comes with example notebooks. These are located in the examples directory, which can be found using the pypop.examples module:

$ python -m pypop.examples
/home/phil/repos/pypop/pypop/examples

Copy these to directory where you have read permissions, e.g.

  $ cp -vr $(python -m pypop.examples) $HOME/pypop_examples

These notebooks demonstrate usage of the main elements of the package.

See the quickstart guide and API documentation for more detail on usage.

Copyright (c) 2019, Numerical Algorithms Group Ltd.