[Bug tree-optimization/104759] [12 Regression] gcc.dg/vect/vect-multitypes-12.c FAILs

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104759

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |testsuite-fail
   Last reconfirmed|                            |2022-03-09
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
                 CC|                            |joel.hutton at arm dot com
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Since sparc doesn't have vect_unpack we probably expect the init loop in main()
to be vectorized (which is artificial) where it now fails to vectorize the
induction.  The likely culprit here is

2021-12-10  Joel Hutton  <joel.hutton@arm.com>

        PR tree-optimization/103523
        * tree-vect-loop.c (vectorizable_induction): Check for
        PLUS_EXPR/MINUS_EXPR support.

where SPARC here doesn't handle character vector add/subtract but then for
the testcase in question we'd have handled this with generic vectors eventually
before this rev.  One would need to check the generated code before this
to decide whether we should consider this a regression.

For gcc.dg/vect/vect-multitypes-12.c I'm simply going to exclude this loop
since it's not the important one to check for.