https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107283
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The link you've provided is about how to do it with severe limitations.
So, we could do it that way only if the ranger tells us the integral (or
floating point) values are in a certain range (+ -ffast-math or similar that
you don't care about infinities etc.).
