[Bug tree-optimization/94524] [8/9/10 Regression] wrong code with vector modulo operation since r0-117580

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94524

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|wrong code with vector      |[8/9/10 Regression] wrong
                   |modulo operation            |code with vector modulo
                   |                            |operation since r0-117580
                 CC|                            |jakub at gcc dot gnu.org
   Target Milestone|---                         |8.5
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2020-04-07

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with my r0-117580-g4ee4c52c64cc1eeda53aae6e221b5b1bd9bd7421 so I
believe this should work correctly in 4.7 (and 4.6 didn't accept it).
The vect pattern code tried to follow what the expander does, but in that case
there is match.pd rule that canonicalizes x % -C into x % C.  Will need to dig
up if something in the expansion tweaks the sign of the divisor or if it has
the same bug as the vect pattern.