https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128737
--- Comment #2 from Thomas Koenig <tkoenig at gcc dot gnu.org> --- (In reply to Richard Biener from comment #1) > probably a duplicate, the GIMPLE level re-association pass doesn't consider > signed integers (due to overflow UB) Yep, replacing INTEGER with UNSIGNED gets the same code for the _v1 and _v2 versions. Under Fortran rules, it is permitted to reorder expressions as long as parentheses are honored. Is there a hint that Fortran can give the middle end that this sort of thing is OK?
