All the code is available here : https://github.com/HugoJPMartin/SPLC2021
All results tables are available in tables.pdf
The "Automated Performance Specialization.ipynb" notebook contains the explaination through code of all 3 developped approaches for performance specialization.
The "Experiments.ipynb" notebook contains the scripts made for the experiments and the latex table generation. The "experiments.py" is an equivalent outside of a notebook.
The "Results.ipynb" notebook allows to recreate the table from the paper from raw results data.
The dataset for Linux is too big for Github, we made it available on Zenodo : https://zenodo.org/record/4943884
It is possible to directly download the dataset :
wget https://zenodo.org/api/files/6008ca9e-bf65-4c35-8b06-992dbd7a1bf8/Linux.csv
We also provide a Docker image to ensure the possibility to run the experiments witht the original packages.
It is available on Docker Hub, accessible with that command :
docker run -i -p 8888:8888 hmartinirisa/splc2021
This will run a Jupyter server that will allow to run the differents notebooks.
To access the notebooks, open the last link on the bottom of the image.
It is also possible to build the image locally with the content of this repository :
docker build -t splc2021 .
To run the local image :
docker run -i -p 8888:8888 splc2021