Fascinating! At some point I read one of the previous explanations cited in this paper. I’m delighted to see Robertson went one step further and algebraically derived the constant optimal post-Newton iteration, producing optimal constants for 64 and 128 bit floats.
it turns out there’s a
Phdbachelor thesis on this topic.Edited to correct error per child post
Looks like a bachelor’s thesis, not a Ph.D thesis.
Pretty big difference there.
You’re right, my mistake.
Fascinating! At some point I read one of the previous explanations cited in this paper. I’m delighted to see Robertson went one step further and algebraically derived the constant optimal post-Newton iteration, producing optimal constants for 64 and 128 bit floats.
Thank you! I was actually just arguing with a friend about this a couple days ago and digging up the original source was on my to-do list.