explanation

What is GenBenchSite ▶️

GenBenchSite is a platform designed to compare the performance of different libraries and frameworks. It provides a detailed report of each library's speed, precision, and other parameters, making it easier for developers to choose the best library for their project.

explanation

GenBenchSite Brief explained 📰

GenBenchSite is designed to automate the process of comparing and testing different libraries. To achieve this, the test are written in configuration files, then given to GenBenchSite. The tests are executed in a controlled environment, and the results are then analyzed and compiled into easy-to-read reports. These reports are output as HTML files structured and then published on dedicated GitHub pages, where users can access them and see how the different libraries perform in a variety of scenarios. By automating the testing process, the website enables developers to save time and effort when evaluating libraries, and helps them make informed decisions.

graph explaning the structure of the directory needed to create a benchmark

Tutorial

First steps 👣

First of all, you'll need to get the project on your computer. To do so, you can either download the project directly from GitHub, clone it using the following command:

git clone https://github.com/White-On/BenchSite

or download it using the pip command:

pip install genbenchsite

Once you have the project on your computer, you can start creating your own benchmark. To do so, you'll need to create a new directory with a specific structure. The directory should contain 3 subdirectories: targets, themes, and site. The targets directory contains the configuration files for the libraries you want to test. The themes directory contains the configuration files for the tests you want to run. The site directory contains the configuration files for the website.

For more information on the structure of the directory graph explaning the structure of the directory needed to create a benchmark

If you are confused about the structure of the directory, you can use the following command to create a directory with the correct structure:

gbs init [name of your benchmark]

Or refer to the Matrix Computation Benchmark for an example of a benchmark.

tutorials

Setup and Launch 🚀

With the different libraries you are going to experiment on, we recommend to use a virtual environment as it will allow you to install the libraries you need without affecting your global python installation.

Create a virtual environment 📦

On Windows

Create a virtual environment using the following command:

python -m venv [name of your virtual environment]

Activate your virtual environment using the following command:

[name of your virtual environment]\Scripts\activate.bat

Deactivate your virtual environment using the following command:

[name of your virtual environment]\Scripts\deactivate.bat

On Linux

Create a virtual environment using the following command:

python -m venv [name of your virtual environment]

Activate your virtual environment using the following command:

source [name of your virtual environment]/bin/activate

Deactivate your virtual environment using the following command:

deactivate

More informations

You're now ready to launch your benchmark. Depending on where your benchmark is located, either locally or online, on a github repository, or if you want to publish the results on a github page, you'll need to run a different command. To see available commands run:

gbs --help

explanation

How we compare the targets 🤔

For the time being, we decided to compare results base on the Lexicographic Maximal Ordering Algorithm (LexMax). Each ranking is based on the number of wins, ties, and losses of each library. The target with the highest number of wins is ranked first, followed by the library with the second-highest number of wins, and so on. In the case of a tie, both libraries are ranked equally. The algorithm does not take into account the magnitude of the wins or losses, only the number of them.

We use it to compare all the data generated by the benchmarking process. For example, we run a task on a set of libraries, and we get the results. Each result is compared to the other result with the same argument, and we get a score for each argument. On the entire task, we get a vector of score for each library. We use the LexMax algorithm to compare the vector of score for each library and we get a ranking of the libraries for that task. We do this for each task and repeat it for the theme and the global ranking.

links

How to contribute ✍️

The benchmark website is an open-source project, and contributions from the community are welcome. To contribute, users can fork the project on GitHub, make changes to the code, and submit a pull request. Users can also contribute by reporting bugs, suggesting improvements, or sharing their benchmarking results.

GitHub repository