https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54192
--- Comment #14 from Richard Biener <rguenth at gcc dot gnu.org> --- I think we want to have a flag specifying the FP exception mechanism in effect with the default being IEEE exception flags only (no control flow effect) and "alternate" being offered, which may have a (subset of) cases trapping, and a "none" which would allow transforms as for -fno-trapping-math. This would allow for -ffp-exceptions=flags -fnon-call-exceptions to not consider all FP operations as EH sources. -ftrapping-math would be legacy and map to =alternate?
