https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125873
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> --- I cannot observe different behavior of fold_plusminus_mult_expr. Reverting the offending revision ontop of trunk does not fix the testcase. But confirmed that 16 works. Checking 4bcb92dcb23daa446^ vs 4bcb92dcb23daa446 now.
