[Bug tree-optimization/64081] New: [5 Regression] r217827 prevents RTL loop unroll

[Bug tree-optimization/64081] New: [5 Regression] r217827 prevents RTL loop unroll

[izamyatin at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

            Bug ID: 64081
           Summary: [5 Regression] r217827 prevents RTL loop unroll
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: izamyatin at gmail dot com
            Target: x86

Created attachment 34123
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34123&action=edit
reproducer

Noticed that for the attached test no RTL loop unroll started to happen.
It is because of changes in dom - namely, I see in dumps that dom2 complicates
loop structure. (probably because of changes in lookup_avail_expr?)
Looks like r217827 doesn't mean this :)

Options that should be used - just "-O2  -funroll-loops"

[Bug tree-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2014-11-26
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
   Target Milestone|---                         |5.0
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
I will have a look.

[Bug tree-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Well, the change allows DOM to CSE pos = 0 in

  <bb 2>:
  pos = 0;
  dir = 1;
  _45 = (long unsigned int) argc_9(D);
  if (_45 != 0)
    goto <bb 3>;
  else
    goto <bb 21>;

  <bb 3>:
  pretmp_75 = data;
  pretmp_77 = token;
  arr1.5_23 = arr1;
  arr2.7_27 = arr2;
  pos_lsm.17_22 = pos;

which in turn allows jump threading to do its work.

  Registering jump thread: (11, 12) incoming edge;  (12, 5) joiner;  (5, 23)
normal; (23, 21) nocopy;
...
  Threaded jump 11 --> 12 to 25

this changes the loop to have two latches (thus it becomes a loop nest), and
it adds one exit (now the loop has three).  Further down the road the extra
loop is no longer there but the three exits remain.

I don't see what is wrong with what DOM does here.

We do miss some interesting kind of optimization opportunities like
transforming

  if (prephitmp_87 == 1)
    goto <bb 9>;
  else
    goto <bb 10>;

  <bb 9>:
  _24 = arr1.5_23 + _62;
  pos.6_25 = *_24;
  goto <bb 11>;

  <bb 10>:
  _28 = arr2.7_27 + _62;
  pos.8_29 = *_28;

  <bb 11>:
  # prephitmp_89 = PHI <pos.6_25(9), pos.8_29(10)>

to

  if (prephitmp_87 == 1)
    goto <bb 9>;
  else
    goto <bb 11>;

  <bb 9>:
  goto <bb 11>;

  <bb 11>:
  # _24 = PHI <arr1.5_23, arr2.7_27>
  _28 = _24 + _62;
  prephitmp_89 = *_28;

sinking common computations through a PHI.

With the followup optimization in out-of-SSA to coalesce arr1.5_23 and
arr2.7_27 which means we can drop the conditional entirely.

Heh.  Fun idea.

Anyway - for this PR it is RTL unroll IV analysis that needs to be improved.

[Bug tree-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW
                 CC|                            |rguenth at gcc dot gnu.org
           Assignee|rguenth at gcc dot gnu.org         |unassigned at gcc dot gnu.org

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[ienkovich at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #3 from ienkovich at gcc dot gnu.org ---
Author: ienkovich
Date: Mon Jan 19 13:58:54 2015
New Revision: 219842

URL: https://gcc.gnu.org/viewcvs?rev=219842&root=gcc&view=rev
Log:
gcc/

    PR rtl-optimization/64081
    * loop-iv.c (def_pred_latch_p): New function.
    (latch_dominating_def): Allow specific cases with non-single
    definitions.
    (iv_get_reaching_def): Likewise.
    (check_complex_exit_p): New function.
    (check_simple_exit): Use check_complex_exit_p to allow certain cases
    with exits not executing on any iteration.

gcc/testsuite/

    PR rtl-optimization/64081
    * gcc.dg/pr64081.c: New test.


Added:
    trunk/gcc/testsuite/gcc.dg/pr64081.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/loop-iv.c
    trunk/gcc/testsuite/ChangeLog

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[ienkovich at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #4 from ienkovich at gcc dot gnu.org ---
Author: ienkovich
Date: Tue Jan 20 08:29:09 2015
New Revision: 219880

URL: https://gcc.gnu.org/viewcvs?rev=219880&root=gcc&view=rev
Log:
gcc/

    PR bootstrap/64676
    Revert:
    2015-01-19  Igor Zamyatin  <igor.zamyatin@intel.com>

    PR rtl-optimization/64081
    * loop-iv.c (def_pred_latch_p): New function.
    (latch_dominating_def): Allow specific cases with non-single
    definitions.
    (iv_get_reaching_def): Likewise.
    (check_complex_exit_p): New function.
    (check_simple_exit): Use check_complex_exit_p to allow certain cases
    with exits not executing on any iteration.

gcc/testsuite/

    PR bootstrap/64676
    Revert:
    2014-01-19  Igor Zamyatin  <igor.zamyatin@intel.com>

    PR rtl-optimization/64081
    * gcc.dg/pr64081.c: New test.


Removed:
    trunk/gcc/testsuite/gcc.dg/pr64081.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/loop-iv.c
    trunk/gcc/testsuite/ChangeLog

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081
Bug 64081 depends on bug 64676, which changed state.

Bug 64676 Summary: [5 Regression] SEGV in tree-ssa-structalias.c solve_constraint
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64676

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[law at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |law at redhat dot com
         Resolution|---                         |FIXED

--- Comment #5 from Jeffrey A. Law <law at redhat dot com> ---
Igor fixed this on the trunk.

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|FIXED                       |---
     Ever confirmed|1                           |0

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Jeffrey A. Law from comment #5)
> Igor fixed this on the trunk.

The patch was revert.

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[dje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

David Edelsohn <dje at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dje at gcc dot gnu.org

--- Comment #7 from David Edelsohn <dje at gcc dot gnu.org> ---
I ran the gcc and g++ testsuites on AIX using the stage 1 compiler with the
patch that breaks bootstrap by miscompiling the stage 2 compiler. 
Unfortunately, the testsuite results are identical with and without the patch,
meaning the testsuite does not evoke and demonstrate the miscompilation.

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[izamyatin at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #8 from Igor Zamyatin <izamyatin at gmail dot com> ---
Created attachment 34524
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34524&action=edit
patch to try AIX bootstrap

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[izamyatin at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #9 from Igor Zamyatin <izamyatin at gmail dot com> ---
David, could you please try attached patch?

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[dje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #10 from David Edelsohn <dje at gcc dot gnu.org> ---
Unfortunately the attached patch causes the same failure demonstrating a
miscompilation of the stage 2 compiler.

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[izamyatin at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #11 from Igor Zamyatin <izamyatin at gmail dot com> ---
Could you please provide details of your compiler configuration for me to try
to reproduce the problem?

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[dje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #12 from David Edelsohn <dje at gcc dot gnu.org> ---
GCC on AIX.  One can use gcc111 in the GCC Compiler Farm.

configure flags: --disable-werror --enable-languages=c,c++,fortran,objc
--with-gmp=/opt/cfarm --with-cloog=no --with-ppl=no
--with-libiconv-prefix=/opt/cfarm --disable-libstdcxx-pch --enable-__cxa_atexit

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[izamyatin at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #13 from Igor Zamyatin <izamyatin at gmail dot com> ---
(In reply to David Edelsohn from comment #12)
> GCC on AIX.  One can use gcc111 in the GCC Compiler Farm.
> 
Thanks! I've sent a request for an access to gcc111 but got no response so
far...

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #14 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Any progress on this?

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[izamyatin at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

--- Comment #15 from Igor Zamyatin <izamyatin at gmail dot com> ---
Got an access to AIX machine, planning to look at it next week

[Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll

[steven at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

Steven Bosscher <steven at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|2014-11-26 00:00:00         |2015-03-24
     Ever confirmed|0                           |1

[Bug rtl-optimization/64081] [5/6 Regression] r217827 prevents RTL loop unroll

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|5.0                         |5.2

--- Comment #16 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 5.1 has been released.

[Bug rtl-optimization/64081] [5/6 Regression] r217827 prevents RTL loop unroll

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|5.2                         |5.3

--- Comment #17 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 5.2 is being released, adjusting target milestone to 5.3.

[Bug rtl-optimization/64081] [8/9/10/11/12 Regression] r217828 prevents RTL loop unroll

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|11.0                        |11.2

--- Comment #65 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 11.1 has been released, retargeting bugs to GCC 11.2.

[Bug rtl-optimization/64081] [9/10/11/12 Regression] r217828 prevents RTL loop unroll

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|11.2                        |11.3

--- Comment #66 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 11.2 is being released, retargeting bugs to GCC 11.3

[Bug rtl-optimization/64081] [9/10/11/12 Regression] r217828 prevents RTL loop unroll

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|11.3                        |11.4

--- Comment #67 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 11.3 is being released, retargeting bugs to GCC 11.4.

[Bug rtl-optimization/64081] [10/11/12/13 Regression] r217828 prevents RTL loop unroll

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |needs-bisection

--- Comment #68 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note for the original testcase in comment #0, GCC 9 starts to unroll the code
again. There are gimple level changes that changed the IR again.

Would be interesting to find out what patch changed that.

[Bug rtl-optimization/64081] [10/11/12/13/14 Regression] r217828 prevents RTL loop unroll

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|11.4                        |11.5

--- Comment #69 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 11.4 is being released, retargeting bugs to GCC 11.5.

[Bug rtl-optimization/64081] [11/12 Regression] r217828 prevents RTL loop unroll

[law at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64081

Jeffrey A. Law <law at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[11/12/13/14 Regression]    |[11/12 Regression] r217828
                   |r217828 prevents RTL loop   |prevents RTL loop unroll
                   |unroll                      |
                 CC|                            |law at gcc dot gnu.org

--- Comment #70 from Jeffrey A. Law <law at gcc dot gnu.org> ---
Per c#68.  Given there's been ~6 years since any significant work was done on
this BZ, I'm not too keen on actually bisecting.  I'm just happy it's behaving
as expected.

Removing gcc-14 and gcc-13 regression markers.