PDFflow: PDF interpolation for hardware accelerators

https://img.shields.io/badge/arXiv-hep--ph%2F2009.06635-%23B31B1B.svg https://zenodo.org/badge/DOI/10.5281/zenodo.3964190.svg

PDFflow is a parton distribution function interpolation library written in Python and based on the TensorFlow framework. It is developed with a focus on speed and efficiency, enabling researchers to perform very expensive calculation as quick and easy as possible.

How to obtain the code

Open Source

The pdfflow package is open source and available at https://github.com/N3PDF/pdfflow


The package can be installed with pip:

python3 -m pip install pdfflow

If you prefer a manual installation just use:

git clone https://github.com/N3PDF/pdfflow
cd pdfflow
python3 setup.py install

or if you are planning to extend or develop code just use:

python3 setup.py develop


PDFflow is developed within the Particle Physics group of the University of Milan. Theoretical calculations in particle physics are incredibly time consuming operations, sometimes taking months in big clusters all around the world.

These expensive calculations are driven by the high dimensional phase space that need to be integrated but also by a lack of expertise in new techniques on high performance computation. Indeed, while at the theoretical level these are some of the most complicated calculations performed by mankind; at the technical level most of these calculations are performed using very dated code and methodologies that are unable to make us of the available resources.

With PDFflow we aim to fill this gap between theoretical calculations and technical performance by providing a framework which can automatically make the best of the machine in which it runs. To that end PDFflow is based on two technologies that together will enable a new age of research.

How to cite pdfflow?

When using pdfflow in your research, please cite the following publications:

https://img.shields.io/badge/arXiv-hep--ph%2F2009.06635-%23B31B1B.svg https://zenodo.org/badge/DOI/10.5281/zenodo.3691926.svg


    author = "Carrazza, Stefano and Cruz-Martinez, Juan M. and Rossi, Marco",
    title = "{PDFFlow: parton distribution functions on GPU}",
    eprint = "2009.06635",
    archivePrefix = "arXiv",
    primaryClass = "hep-ph",
    month = "9",
    year = "2020"

    author       = {Juan Cruz-Martinez and
                    Marco Rossi and
                    Stefano Carrazza},
    title        = {N3PDF/pdfflow: PDFFlow 1.0},
    month        = sep,
    year         = 2020,
    publisher    = {Zenodo},
    version      = {v1.0},
    doi          = {10.5281/zenodo.3964190},
    url          = {https://doi.org/10.5281/zenodo.3964190}


Why the name pdfflow?

It is a combination of the names PDF and Tensorflow.

  • PDFs: Parton Distribution Functions (or PDFs) are at the core of LHC phenomenology by providing a description of the parton content of the proton.

  • TensorFlow: the framework developed by Google and made public in November of 2015 is a perfect combination between performance and usability. With a focus on Deep Learning, TensorFlow provides an algebra library able to easily run operations in many different devices: CPUs, GPUs, TPUs with little input by the developer. Write your code once.

PDF interpolation with tensorflow