https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126334
Robin Dapp <rdapp at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Version|17.0 |14.0
--- Comment #2 from Robin Dapp <rdapp at gcc dot gnu.org> ---
Thinking about it some more, stride 0 might just be a special case of the
broader "stride < element size" problem. For a byte stride 1 and 64-bit
elements we still overlap and have the same problem. So the vectorizer
versioning condition should be if (stride >= effective element_size).
