[Bug tree-optimization/66502] SCCVN can't handle PHIs optimistically optimally

[Bug tree-optimization/66502] SCCVN can't handle PHIs optimistically optimally

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66502

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2015-06-11
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
https://gcc.gnu.org/ml/gcc-patches/2015-06/msg00760.html was posted that
changes the current preference to handle the first testcase (but no longer
the second).  The first opportunity is frequently introduced by the vectorizer,
together with redundant loads, that are not eliminated in any way currently.

IVOPTs can remove redundant IVs but the issue is usually that it can't prove
the initial values are the same nor does it remove the redundant loads
leading to cost model issues (DOM might later remove the redundant loads).

The above mentioned patch passed bootstrap & regtest on
x86_64-unknown-linux-gnu.

Enlarging the SCCs to cover all candidates is a bit tricky in the current
framework, I still like to experiment with that idea.

[Bug tree-optimization/66502] SCCVN can't handle PHIs optimistically optimally

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66502

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Author: rguenth
Date: Thu Aug 13 07:06:10 2015
New Revision: 226850

URL: https://gcc.gnu.org/viewcvs?rev=226850&root=gcc&view=rev
Log:
2015-08-13  Richard Biener  <rguenther@suse.de>

        PR tree-optimization/66502
        PR tree-optimization/67167
        * tree-ssa-sccvn.c (vn_phi_compute_hash): Do not include
        backedge arguments.
        (vn_phi_lookup): Adjust.
        (vn_phi_insert): Likewise.
        (visit_phi): Prefer to value-number to another PHI node
        over value-numbering to a PHI argument.
        (init_scc_vn): Mark DFS back edges.

        * gcc.dg/tree-ssa/ssa-fre-46.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-46.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-ssa-sccvn.c

[Bug tree-optimization/66502] SCCVN can't handle PHIs optimistically optimally

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66502

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
SCCVN preference changed.  Missed optimization remains (now the other variant).

[Bug tree-optimization/66502] SCCVN can't handle PHIs optimistically optimally

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66502

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #4)
> SCCVN preference changed.  Missed optimization remains (now the other
> variant).

GCC 10 looks like it can handle both now. I don't know if that means this can
be marked as fixed though.

[Bug tree-optimization/66502] SCCVN can't handle PHIs optimistically optimally

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66502

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
      Known to work|                            |10.3.1
             Status|ASSIGNED                    |RESOLVED

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #5)
> (In reply to Richard Biener from comment #4)
> > SCCVN preference changed.  Missed optimization remains (now the other
> > variant).
> 
> GCC 10 looks like it can handle both now. I don't know if that means this
> can be marked as fixed though.

So it looks like we now first value-number j to a (backedge not executable)
and in the second iteration (backedge executable) we value-number it to i.
I guess the special-casing of a single executable edge makes this work now
and also ensure we don't oscillate between both choices.

[Bug tree-optimization/66502] SCCVN can't handle PHIs optimistically optimally

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66502

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Richard Biener <rguenth@gcc.gnu.org>:

https://gcc.gnu.org/g:cd83fd98a7e769ac7ded28377984f68238812a26

commit r12-6182-gcd83fd98a7e769ac7ded28377984f68238812a26
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Jan 3 10:48:57 2022 +0100

    tree-optimization/66502 - add testcase

    This adds the second testcase which we now also handle eliminating
    a redundant PHI node.

    2022-01-03  Richard Biener  <rguenther@suse.de>

            PR tree-optimization/66502
            * gcc.dg/tree-ssa/ssa-fre-98.c: New testcase.

[Bug tree-optimization/66502] SCCVN can't handle PHIs optimistically optimally

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66502

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |10.0