| commit | 9c88775ceb5505318c88a94f91e3d8e033b908f7 | [log] [tgz] |
|---|---|---|
| author | Michele Di Giorgio <michele.digiorgio@arm.com> | Thu Mar 31 11:23:45 2016 +0100 |
| committer | Michele Di Giorgio <michele.digiorgio@arm.com> | Thu Mar 31 11:28:42 2016 +0100 |
| tree | 8d8fb47d2cb56dc3a78173820c394d127089a2fe | |
| parent | 8eb2939f8b5113cf4efe783543e5fc57bd38e032 [diff] |
README.md: improve structure of README file
The Behavioural Analysis and Regression Toolkit is based on TRAPpy. The primary goal is to assert behaviours using the FTrace output from the kernel.
The framework is designed to cater to a wide range of audience. Aiding developers as well as automating the testing of “difficult to test” behaviours.
Making sure that the code that you are writing is doing the right thing.
Plotting/Asserting performance behaviours between different revisions of the kernel.
Verifying behaviours when different components/patches are integrated.
$ sudo apt install trace-cmd kernelshark
$ sudo apt install python-pip python-dev
$ sudo apt install libfreetype6-dev libpng12-dev python-nose $ sudo pip install numpy matplotlib pandas ipython[all] $ sudo pip install --upgrade trappy
ipython[all] will install IPython Notebook, a web based interactive python programming interface. It is required if you plan to use interactive plotting in BART.
$ sudo pip install --upgrade bart-py
Instead of installing TRAPpy and BART using pip you should the repositories:
$ git clone git@github.com:ARM-software/bart.git $ git clone git@github.com:ARM-software/trappy.git
Add the directories to your PYTHONPATH
$ export PYTHONPATH=$BASE_DIR/bart:$BASE_DIR/trappy:$PYTHONPATH
BART also provides a generic Trace Analysis Language, which allows the user to construct complex relation statements on trace data and assert their expected behaviours. The usage of the Analyzer module can be seen for the thermal behaviours here
Enables assertion and the calculation of the following parameters:
The total time that the task spent on a CPU executing.
Assert that a task switched between CPUs/Clusters in a given window of time.
The ratio of the execution time to the total time.
The average difference between two switch-in or two switch-out events of a task.
The first CPU that a task ran on.
Calculate and assert the total residency of a task on a CPU or cluster.
The Scheduler assertions also use TRAPpy's EventPlot to provide a kernelshark like timeline for the tasks under consideration. (in IPython notebooks).
A notebook explaining the usage of the framework for asserting the deadline scheduler behaviours can be seen here.