License status?

[maddyscientist]

Thank you for this work, it is a great demonstration of the reproducible summation algorithms.

The license of this source code doesn't seem to be specified. Would the author consider making this work available under a BSD / MIT / Apache license?

Thanks

[maddyscientist]

Any chance of an update here? I'd love to include look into porting this these algorithms to a GPU, and this looks like a great starting point. Any derived work would remain open source, if it matters.

[r-barnes]

@maddyscientist - I've moved this onto the MIT License. Note that some code is drawn from https://github.com/peterahrens/ReproBLAS, which uses a University of California license (details).

Sorry for the delay. I'd be interested to know how this works out for you :-)

[maddyscientist]

@r-barnes thank you very much. I'll update you if I make if and when I make any progress on this front.

[maddyscientist]

Just to note that I've mostly completed a port to CUDA. I've essentially ported ahrens2020.cpp to CUDA (in ahrens2020_cuda.cu). Still some optimization left to be done, but it's performing reasonably (get's up to 450 GB/s on a GPU that can get around 650 GB/s on a naive summation algorithm).

Up until now, I've kept fairly close to the style of the original code, but I may diverge / extend it to include other summation algorithms, e.g., higher-level floating-point expansions, fp128, double-double, etc.

[maddyscientist]

My branch is here: https://github.com/maddyscientist/reproducible_floating_sums/tree/feature/cuda

[r-barnes]

@maddyscientist - Very cool! Thanks for the link.