[Bug target/104307] New: [11/12 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[Bug target/104307] New: [11/12 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[asolokha at gmx dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

            Bug ID: 104307
           Summary: [11/12 Regression] '-fcompare-debug' failure (length)
                    w/ -mavx512f -O2
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: compare-debug-failure
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: asolokha at gmx dot com
  Target Milestone: ---
            Target: x86_64-unknown-linux-gnu

Created attachment 52319
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=52319&action=edit
gkd diff

gcc 12.0.1 20220123 snapshot (g:2da90ad39bf8fa9ee287e040d1f4411cb7a2e7ed) fails
-fcompare-debug check when compiling gcc/testsuite/gcc.target/i386/pr78669.c w/
-Os -funroll-loops:

% x86_64-unknown-linux-gnu-gcc-12.0.1 -mavx512f -O2 -fcompare-debug
-gno-statement-frontiers -c gcc/testsuite/gcc.target/i386/pr78669.c
x86_64-unknown-linux-gnu-gcc-12.0.1: error:
gcc/testsuite/gcc.target/i386/pr78669.c: '-fcompare-debug' failure (length)

gkd diff attached.

-fcompare-debug check also fails w/o -mavx512f, though the diff is different.

If I get the dumps right, this issue is not related to PR104054, as things
already start looking interesting since veclower21 tree dump.

[Bug target/104307] [11/12 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[asolokha at gmx dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

--- Comment #1 from Arseny Solokha <asolokha at gmx dot com> ---
(In reply to Arseny Solokha from comment #0)
> w/ -Os -funroll-loops:

It's an obvious copy-paste error, of course. It fails the check w/ -mavx512f
-O2.

[Bug target/104307] [11/12 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2022-01-31
           Priority|P3                          |P2
   Target Milestone|---                         |11.3
                 CC|                            |jakub at gcc dot gnu.org

[Bug middle-end/104307] [11/12 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r11-2577-g229752afe3156a3990dacaedb94c76846cebf132 but I think the
real bug is caused by r11-2362-g55c9695cbe91c406805ac0cd342949f32f13b779

[Bug middle-end/104307] [11/12 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

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

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

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 52320
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=52320&action=edit
gcc12-pr104307.patch

Untested fix.
Just the first part of the hunk is needed for this bug, I'm actually not 100%
sure about the rhs2/rhs3 case, but it feels it isn't right (it would e.g. try
to expand_vector_condition the same stmt multiple times).

I must say I don't understand the pushing of all uses into a vector, wouldn't
it make more sense to push there just the VEC_COND_EXPRs with such lhs?
Then we e.g. wouldn't need to do any expand_vector_condition calls if we have
1000 VEC_COND_EXPR uses and then a different use.

[Bug tree-optimization/104307] [11/12 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

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

https://gcc.gnu.org/g:e9bf6d6b0e1d67df6dcee042fbe37ab72c3a38d7

commit r12-6962-ge9bf6d6b0e1d67df6dcee042fbe37ab72c3a38d7
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Feb 1 16:02:54 2022 +0100

    veclower: Fix up -fcompare-debug issue in expand_vector_comparison
[PR104307]

    The following testcase fails -fcompare-debug, because
expand_vector_comparison
    since r11-1786-g1ac9258cca8030745d3c0b8f63186f0adf0ebc27 sets
    vec_cond_expr_only when it sees some use other than VEC_COND_EXPR that uses
    the lhs in its condition.
    Obviously we should ignore debug stmts when doing so, e.g. by not pushing
    them to uses.
    That would be a 2 liner change, but while looking at it, I'm also worried
    about VEC_COND_EXPRs that would use the lhs in more than one operand,
    like VEC_COND_EXPR <lhs, lhs, something> or VEC_COND_EXPR <lhs, something,
lhs>
    (sure, they ought to be folded, but what if they weren't).  Because if
    something like that happens, then FOR_EACH_IMM_USE_FAST would push the same
    stmt multiple times and expand_vector_condition can return true even when
    it modifies it (for vector bool masking).
    And lastly, it seems quite wasteful to safe_push statements that will just
    cause vec_cond_expr_only = false; and break; in the second loop, both for
    cases like 1000 immediate non-VEC_COND_EXPR uses and for cases like
    999 VEC_COND_EXPRs with lhs in cond followed by a single non-VEC_COND_EXPR
    use.  So this patch only pushes VEC_COND_EXPRs there.

    2022-02-01  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/104307
            * tree-vect-generic.cc (expand_vector_comparison): Don't push debug
            stmts to uses vector, just set vec_cond_expr_only to false for
            non-VEC_COND_EXPRs instead of pushing them into uses.  Treat
            VEC_COND_EXPRs that use lhs not just in rhs1, but rhs2 or rhs3 too
            like non-VEC_COND_EXPRs.

            * gcc.target/i386/pr104307.c: New test.

[Bug tree-optimization/104307] [11 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[11/12 Regression]          |[11 Regression]
                   |'-fcompare-debug' failure   |'-fcompare-debug' failure
                   |(length) w/ -mavx512f -O2   |(length) w/ -mavx512f -O2

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed on the trunk so far.

[Bug tree-optimization/104307] [11 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-11 branch has been updated by Jakub Jelinek
<jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:c81d1a061e5cfab6bd276c5be4cd6161b11e6d44

commit r11-9596-gc81d1a061e5cfab6bd276c5be4cd6161b11e6d44
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Tue Feb 1 16:02:54 2022 +0100

    veclower: Fix up -fcompare-debug issue in expand_vector_comparison
[PR104307]

    The following testcase fails -fcompare-debug, because
expand_vector_comparison
    since r11-1786-g1ac9258cca8030745d3c0b8f63186f0adf0ebc27 sets
    vec_cond_expr_only when it sees some use other than VEC_COND_EXPR that uses
    the lhs in its condition.
    Obviously we should ignore debug stmts when doing so, e.g. by not pushing
    them to uses.
    That would be a 2 liner change, but while looking at it, I'm also worried
    about VEC_COND_EXPRs that would use the lhs in more than one operand,
    like VEC_COND_EXPR <lhs, lhs, something> or VEC_COND_EXPR <lhs, something,
lhs>
    (sure, they ought to be folded, but what if they weren't).  Because if
    something like that happens, then FOR_EACH_IMM_USE_FAST would push the same
    stmt multiple times and expand_vector_condition can return true even when
    it modifies it (for vector bool masking).
    And lastly, it seems quite wasteful to safe_push statements that will just
    cause vec_cond_expr_only = false; and break; in the second loop, both for
    cases like 1000 immediate non-VEC_COND_EXPR uses and for cases like
    999 VEC_COND_EXPRs with lhs in cond followed by a single non-VEC_COND_EXPR
    use.  So this patch only pushes VEC_COND_EXPRs there.

    2022-02-01  Jakub Jelinek  <jakub@redhat.com>

            PR middle-end/104307
            * tree-vect-generic.c (expand_vector_comparison): Don't push debug
            stmts to uses vector, just set vec_cond_expr_only to false for
            non-VEC_COND_EXPRs instead of pushing them into uses.  Treat
            VEC_COND_EXPRs that use lhs not just in rhs1, but rhs2 or rhs3 too
            like non-VEC_COND_EXPRs.

            * gcc.target/i386/pr104307.c: New test.

    (cherry picked from commit e9bf6d6b0e1d67df6dcee042fbe37ab72c3a38d7)

[Bug tree-optimization/104307] [11 Regression] '-fcompare-debug' failure (length) w/ -mavx512f -O2

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104307

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

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

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed for 11.3+ too.