fl2sqrt, fl2sqr, fl2/, flsqrt/error accuraccy regression

Open samth opened this issue 3 years ago • 0 comments

PR #71 introduced a regression in the accuracy of some double-flonum functions: fl2sqrt, fl2sqr, fl2/, flsqrt/error. This was to avoid using the translated code from crlibm for licensing reasons. As far as I know, the algorithms used in crlibm are not published anywhere, and the currently-implemented algorithms are the best ones I could find. But I don't know this area and I'm sure more could be done.