[Bug tree-optimization/120704] [14/15 regression] Wrong code at -O3

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120704

--- Comment #14 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I think it went away with r16-6900-ge48704694d686de, but our bisect seed only
has for some strange reason r16-6899 and then r16-6918 (the former still emits
the 3 ubsan messages, the latter doesn't, and the first change in dumps between
those two is the lim2 pass).