http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57371
--- Comment #2 from Marc Glisse <glisse at gcc dot gnu.org> --- Thank you for this very detailed analysis :-) (In reply to jos...@codesourcery.com from comment #1) > * If not all values of the integer type can be represented in the > floating-point type exactly, then the change to an integer comparison > would lose an "inexact" exception, so not be appropriate with > -ftrapping-math (strictly interpreted; GCC may not follow that > consistently). Indeed. Note that I believe splitting -ftrapping-math=(no|normal|strict), or maybe even further (I know some people using overflow, but noone who uses the inexact exception, although there must be a few) would help convince people to add those checks (I abandoned PR 53805 and PR 53806 in part because it wasn't clear which one was wanted), and so would making -fno-trapping-math the default (PR 54192), so we don't fear regressing code speed for default users who don't care.
