/green-algorithms-tool

Primary LanguagePythonCreative Commons Attribution 4.0 InternationalCC-BY-4.0

Green Algorithms

www.green-algorithms.org

Generic badge

Maintenance Open Source? Yes!


Methods and data

The methodology behind the Green Algorithms project is described in our publication: https://onlinelibrary.wiley.com/doi/10.1002/advs.202100707

All the data used for the calculator are in the /data directory above.

Questions, issues, suggestions? Want to contribute?

Start by opening an issue here, and we will try to address it quickly: https://github.com/GreenAlgorithms/green-algorithms-tool/issues

You can also contact us at: green.algorithms@gmail.com

How to cite this work

Lannelongue, L., Grealey, J., Inouye, M., Green Algorithms: Quantifying the Carbon Footprint of Computation. Adv. Sci. 2021, 2100707. https://doi.org/10.1002/advs.202100707

FAQ

Ask Me Anything !

Should I include the number of processors, number of cores, or number of threads used?

For CPUs, the number of cores (CPUs usually have 4-12 cores per processor). For GPUs, the number of GPUs. If using multi-threading on CPUs (i.e. using more threads than cores), still input the number of cores, but be aware that your emissions might be underestimated.

What if my processor is not in the list?

You can select "Other" and find the TDP (Thermal Design Power) value on the manufacturer's website. Plus, add a comment on this issue so that we can add it to the list!

What if my country is not in the list?

Add a comment on this issue so that we can add it to the list!
(some countries are more secretive than others about their energy mix). You can use the world average, or a close proxy, for your estimations.

Can I compare algorithms impact independantly of the location?

Yes, simply use the "Energy needed" (in W) displayed next to the carbon emissions.

How do I find the usage factor of my processors?

It depends on your system. For example if you're using SLURM, seff <job_id> will give you the "CPU Efficiency". Similar commands exist for the different systems, and if you can't find it, you can just leave the default value of 1.

How do I estimate my PSF (Pragmatic Scaling Factor)?

Try to estimate how many times you need to run your full analysis to get results you're happy with. It can be trials and errors, parameters optimisations, memory issues etc.

What if I found a bug in the tool?

Open an issue on the GitHub so that we can look at it.

Credits

Licence

This work is licensed under a Creative Commons Attribution 4.0 International License.

CC BY 4.0

CC BY 4.0