browserperfrunner

This repository contains a set of script and tools that automate running benchmarks with different browsers.

The focus is on benchmarking the browser itself.

Most of the code from this repository is also available on the WebKit repository.

The idea of having this also on a separate repository is to allow running this scripts without having to clone the whole WebKit repository (which is huge)

This is specially useful for running benchmarks on embedded boards.

• If you are using Yocto, there is a recipe for installing this (and the required deps) in meta-webkit

Currently there are two main tools here:

• run-benchmark: allows to run benchmarks and saves the perf data to a json file
• browserperfdash-benchmark: its a wrapper over run-benchmark that allows to upload the perf data to a dashboard like https://github.com/Igalia/browserperfdash. The idea is to run this continously on a bot with different browser versions.

Dependencies

In the case of Linux, the scripts need the following dependencies installed on the system:

sudo apt-get install curl make patch perl libarchive-zip-perl procps psmisc ruby \
     subversion python2.7 python-psutil python-setuptools python-gobject gir1.2-gtk-3.0

Running browserperfdash-benchmark

• Assuming that the required dependencies (above) are installed, the scripts can be run directly from the git checkout.

• This is an example of how to run browserperfdash-benchmark

./browserperfdash-benchmark --plan sunspider \
	--config-file /etc/browserperfdash-benchmark-config.txt \
	--browser chrome --browser-version m67

Check the config-file-example.txt to see the format of the config file.

• To see the available plans:

./run-benchmark --list-plans

And to run all plans one after the other simply pass --allplans