https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66799
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|WAITING |ASSIGNED
Component|tree-optimization |testsuite
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Ok, sorry for responding so late. The testcase has aligned accesses only
and we now might be able to propagate simply via IPA alignment propagation.
Not sure why it started to "work" in the revision range but the vectorizer
indeed analyzes all accesses to be aligned.
I think fixing the testcase is best (not align the source array) as only
that triggers a runtime error if we get things wrong (and miss the
versioning check). The testcase was made to align that array with r104445
but I don't see a good reason for this today.
