https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101639

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
If we "fix" pattern mask precision calculation by processing PHIs after stmts
(so we get the backedge mask precision computed) there's the issue that we
mishandle the reduction epilogue when facing with a mask reduction.

Reply via email to