[Bug tree-optimization/115120] Bad interaction between ivcanon and early break vectorization

["rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115120

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Blocks|                            |53947
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2024-05-17

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Note this isn't really because of IVCANON but because the IV is live.  IVCANON
adds a downward counting IV historically to enable RTL doloop transforms.

One could argue this should nowadays be IVOPTs job (in the past IVOPTs was
done on RTL).

So what's really the issue is that IVOPTs does a bad job (I don't really
think it even tries) to replace one vector IV with another.

We could try to change IVCANON to avoid creating a canonical downward counting
IV if one already exists (I don't think it even avoids that case) and also
avoid creating a downward counting IV with step -1 when a upward counting
IV with step 1 already controls the exit and verify IVOPTs turns it into
a downward counting one when profitable.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53947
[Bug 53947] [meta-bug] vectorizer missed-optimizations