from the README
MultiFloats.jl is the product of significant original research on high-precision computer arithmetic, culminating in the discovery of a novel class of fast branch-free algorithmsfor floating-point arithmetic beyond machine precision. These state-of-the-art algorithms are both faster and more accurate than all previous techniques.
https://purl.stanford.edu/gt930wy1453)
https://theory.stanford.edu/~aiken/publications/papers/cav25...
https://theory.stanford.edu/~aiken/publications/papers/sc25....
