[Bug testsuite/103264] New: [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[Bug testsuite/103264] New: [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[seurer at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103264

            Bug ID: 103264
           Summary: [12 regression] gcc.dg/tree-prof/merge_block.c fails
                    after r12-5236
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: testsuite
          Assignee: unassigned at gcc dot gnu.org
          Reporter: seurer at gcc dot gnu.org
  Target Milestone: ---

Tried g:5aa91072e24c1e16a5ec641b48b64c9c9f199f13, r12-5236
make  -k check-gcc RUNTESTFLAGS="tree-prof.exp=gcc.dg/tree-prof/merge_block.c"
FAIL: gcc.dg/tree-prof/merge_block.c scan-tree-dump-not optimized "Invalid sum"
# of expected passes            4
# of unexpected failures        1
# of unsupported tests          1

commit 5aa91072e24c1e16a5ec641b48b64c9c9f199f13 (HEAD, refs/bisect/bad)
Author: Jan Hubicka <jh@suse.cz>
Date:   Sat Nov 13 22:25:23 2021 +0100

    Implement DSE of dead functions calls storing memory.

[Bug testsuite/103264] [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[hubicka at kam dot mff.cuni.cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103264

--- Comment #1 from hubicka at kam dot mff.cuni.cz ---
What breaks in the testcase is updating profile after complete loop
unroling.  I suspect the unrolling is enabled by the extra DSE.

[Bug testsuite/103264] [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103264

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |marxin at gcc dot gnu.org
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-11-16
             Status|UNCONFIRMED                 |NEW

[Bug testsuite/103264] [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103264

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
   Target Milestone|---                         |12.0

[Bug testsuite/103264] [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103264

--- Comment #2 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
OK, the test wants to verify that main() became single basic block that has
correct frequency.  This is still true.  What we get wrong is that the function
t() gets mismatches:
int t ()
{
  int i;
  int _7;
  int _10;
  int _13;

;;   basic block 2, loop depth 0, count 312727306 (estimated locally, globally
0), probably never executed
;;    prev block 0, next block 3, flags: (NEW, REACHABLE, VISITED)
;;    pred:       ENTRY [always]  count:312727306 (estimated locally, globally
0) (FALLTHRU,EXECUTABLE)
  _7 = a[0];
  if (_7 != 0)
    goto <bb 6>; [5.50%]
  else
    goto <bb 3>; [94.50%]
;;    succ:       6 [5.5% (guessed)]  count:17200002 (estimated locally,
globally 0) (TRUE_VALUE,EXECUTABLE)
;;                3 [94.5% (guessed)]  count:295527304 (estimated locally,
globally 0) (FALSE_VALUE,EXECUTABLE)

;;   basic block 3, loop depth 0, count 299827304 (estimated locally, globally
0), probably never executed
;;   Invalid sum of incoming counts 295527304 (estimated locally, globally 0),
should be 299827304 (estimated locally, globally 0)
;;    prev block 2, next block 4, flags: (NEW, REACHABLE, VISITED)
;;    pred:       2 [94.5% (guessed)]  count:295527304 (estimated locally,
globally 0) (FALSE_VALUE,EXECUTABLE)
  _10 = a[1];
  if (_10 != 0)
    goto <bb 6>; [5.50%]
  else
    goto <bb 4>; [94.50%]
;;    succ:       6 [5.5% (guessed)]  count:16490502 (estimated locally,
globally 0) (TRUE_VALUE,EXECUTABLE)
;;                4 [94.5% (guessed)]  count:283336802 (estimated locally,
globally 0) (FALSE_VALUE,EXECUTABLE)

;;   basic block 4, loop depth 0, count 287459432 (estimated locally, globally
0), probably never executed
;;   Invalid sum of incoming counts 283336802 (estimated locally, globally 0),
should be 287459432 (estimated locally, globally 0)
;;    prev block 3, next block 5, flags: (NEW, REACHABLE, VISITED)
;;    pred:       3 [94.5% (guessed)]  count:283336802 (estimated locally,
globally 0) (FALSE_VALUE,EXECUTABLE)
  _13 = a[2];
  if (_13 != 0)
    goto <bb 6>; [5.50%]
  else
    goto <bb 5>; [94.50%]
;;    succ:       6 [5.5% (guessed)]  count:15810269 (estimated locally,
globally 0) (TRUE_VALUE,EXECUTABLE)
;;                5 [94.5% (guessed)]  count:271649163 (estimated locally,
globally 0) (FALSE_VALUE,EXECUTABLE)

;;   basic block 5, loop depth 0, count 271649163 (estimated locally, globally
0), probably never executed
;;    prev block 4, next block 6, flags: (NEW)
;;    pred:       4 [94.5% (guessed)]  count:271649163 (estimated locally,
globally 0) (FALSE_VALUE,EXECUTABLE)
;;    succ:       6 [always]  count:271649163 (estimated locally, globally 0)
(FALLTHRU)

;;   basic block 6, loop depth 0, count 312727306 (estimated locally, globally
0), probably never executed
;;   Invalid sum of incoming counts 321149936 (estimated locally, globally 0),
should be 312727306 (estimated locally, globally 0)
;;    prev block 5, next block 1, flags: (NEW, REACHABLE, VISITED)
;;    pred:       3 [5.5% (guessed)]  count:16490502 (estimated locally,
globally 0) (TRUE_VALUE,EXECUTABLE)
;;                5 [always]  count:271649163 (estimated locally, globally 0)
(FALLTHRU)
;;                4 [5.5% (guessed)]  count:15810269 (estimated locally,
globally 0) (TRUE_VALUE,EXECUTABLE)
;;                2 [5.5% (guessed)]  count:17200002 (estimated locally,
globally 0) (TRUE_VALUE,EXECUTABLE)
  # i_5 = PHI <1(3), 3(5), 2(4), 0(2)>
  return i_5;
;;    succ:       EXIT [always]  count:312727306 (estimated locally, globally
0)

}

which appears in cunrolli. I don't think this is related to the modref change.

Martin, can you please bisect it?

[Bug testsuite/103264] [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103264

--- Comment #3 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Aha, I see it.  The difference with IPA modref is that we now eliminate the
loop before profile instrumentation.  This leads to difference because the
function t() is never executed and thus we use guessed profile on it.

Now without modref we are not able to optimize out the loop and we actualy
profile it and then scale down the profile to zero.  This makes the cunrolli
updating bug to disappear and also shows weird behaviour of profile scaling -
when we get count to 0 we should maintain local profile.

So to get back to the original behaviour we want to add -fno-ipa-modref but
there are at least two other issues - the cunrolli misupading profile (which if
I recall may be partly intentional) and profile scalling killing local profile
when dropping count to 0.

[Bug testsuite/103264] [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103264

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jan Hubicka <hubicka@gcc.gnu.org>:

https://gcc.gnu.org/g:c8260767aa3b41017b075d8fde3a4065fa637db7

commit r12-5438-gc8260767aa3b41017b075d8fde3a4065fa637db7
Author: Jan Hubicka <jh@suse.cz>
Date:   Sun Nov 21 16:13:40 2021 +0100

    Fix failure merge_block.c testcase

    gcc/testsuite/ChangeLog:

    2021-11-21  Jan Hubicka  <hubicka@ucw.cz>

            PR ipa/103264
            * gcc.dg/tree-prof/merge_block.c: Add -fno-ipa-modref

[Bug testsuite/103264] [12 regression] gcc.dg/tree-prof/merge_block.c fails after r12-5236

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103264

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |jakub at gcc dot gnu.org
         Resolution|---                         |FIXED

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.