https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84077
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
Resolution|--- |INVALID
--- Comment #14 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Besides inline asm barriers another option is volatile {float,double} vars in
certain spots (to ensure rounding to {float,double} where required). And
again, you don't need to do anything if FLT_EVAL_METHOD is 0, or for double
computations also if FLT_EVAL_METHOD is 1.
