2 ms·

There’s a link at the bottom to a paper exploring that question.

• dahart 2y ago
  Good point, I hadn’t read Lomont’s paper and should have. I read the section in the Wikipedia article talking about it, and did try the constant that it suggests, however it depends on doing extra Newton iterations and I looked at the relative error of the initial guess without Newton. I can see in the paper he found something within 1 bit of what I found. I’m not certain mine’s better, but my python script claims it is.