It doesn't use any floats, and it doesn't use taylor series etc. Just 32 or 64 or 96 or however many iterations, and you get that many bits of the respective inverse function. This is the first one I worked out using this method, arctan:
It doesn't use any floats, and it doesn't use taylor series etc. Just 32 or 64 or 96 or however many iterations, and you get that many bits of the respective inverse function. This is the first one I worked out using this method, arctan: