this is the kind of esoteric academic bullshit that's completely disconnected from reality. who the hell cares about approximating floating-point addition when we can't even build reliable software? https://www.reddit.com/user/self