https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125107
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> --- The series is expected to make some BB vectorizations profitable that were not previously. That doesn't mean we do vectorize only profitable cases, of course. Notably we still (IMO) vectorize too many two-lane patterns which when otherwise enough execution resources are avialable, only ties the CPU schedulers hands.
