Home

Awesome

SCOOP logo

SCOOP (Scalable COncurrent Operations in Python) is a distributed task module allowing concurrent parallel programming on various environments, from heterogeneous grids to supercomputers. Its documentation is available on http://scoop.readthedocs.org/ .

Philosophy

SCOOP was designed from the following ideas:

These tenets are translated concretely in a minimum number of functions allowing maximum parallel efficiency while keeping at minimum the inner knowledge required to use them. It is implemented with Python 3 in mind while being compatible with Python 2.6+ to allow fast prototyping without sacrificing efficiency and speed.

Some comments we received on SCOOP:

Features

SCOOP features and advantages over futures, multiprocessing and similar modules are as follows:

Anatomy of a SCOOPed program

SCOOP can handle multiple diversified multi-layered tasks. With it, you can submit your different functions and data simultaneously and effortlessly while the framework executes them locally or remotely. Contrarily to most multiprocessing frameworks, it allows to launch subtasks within tasks.

Intro tree

Through SCOOP, you can execute simultaneously tasks that are different by nature, shown by the task color, or different by complexity, shown by the task radius. The module will handle the physical considerations of parallelization, such as task distribution over your resources (load balancing), communications, etc.

Applications

The common applications of SCOOP consist but is not limited to:

Citing SCOOP

Authors of scientific papers including results generated using SCOOP are encouraged to cite the following paper.

{{{ @inproceedings{SCOOP_XSEDE2014, title={Once you SCOOP, no need to fork}, author={Hold-Geoffroy, Yannick and Gagnon, Olivier and Parizeau, Marc}, booktitle={Proceedings of the 2014 Annual Conference on Extreme Science and Engineering Discovery Environment}, pages={60}, year={2014}, organization={ACM} } }}}

Useful links

You can download the latest stable version, check the project documentation, post to the mailing list or submit an issue if you've found one.