Hi Jakub,
Bootstrapped/regtested on x86_64-linux and i686-linux (together with
Richard's patch), ok for trunk?
A couple of points:
with this patch, we will have an algorithm that will evaluate NORM2
in a different way than before, possibly leading to regressions
in some cases where we previously generated good results and
would return +/- Inf now.
Also, with this patch, we would use a very different algorithm
at run-time to compile-time, potentially leading to differences
far exceeding a few ulps. I am not sure I like that idea.
I'll look at the literature and generate some test cases, and
come back to you.
From curiosity, did you chose this algorithm from literature?
Regards
Thomas
