[Bug fortran/110725] New: internal compiler error: in expand_expr_real_1, at expr.cc:10897

[Bug fortran/110725] New: internal compiler error: in expand_expr_real_1, at expr.cc:10897

[tonycurtis32 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

            Bug ID: 110725
           Summary: internal compiler error: in expand_expr_real_1, at
                    expr.cc:10897
           Product: gcc
           Version: 13.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: tonycurtis32 at gmail dot com
  Target Milestone: ---

Created attachment 55573
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55573&action=edit
preprocessed source

arch = a64fx (arm 8.2 + sve)

==

Using built-in specs.
COLLECT_GCC=gfortran
COLLECT_LTO_WRAPPER=/lustre/software/gcc/13.1.0/libexec/gcc/aarch64-unknown-linux-gnu/13.1.0/lto-wrapper
Target: aarch64-unknown-linux-gnu
Configured with: ../gcc-13.1.0/configure --with-mpfr-lib=/lib64
--with-gmp-lib=/lib64 --with-gmp-include=/usr/include
--with-mpfr-include=/usr/include --with-mpc-lib=/lib64
--with-mpc-include=/usr/include --prefix=/lustre/software/gcc/13.1.0
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 13.1.0 (GCC)

==

command-line = gfortran -ggdb -fopenmp -mcmodel=large -mcpu=a64fx -O3 -o swim
swim.F
(-O0 compiles ok, gcc 12.2 works fine too with optimization)

==

swim.F:107:72:

  107 |       DO 50 I=1,MP1
      |                                                                       
1
Warning: Fortran 2018 deleted feature: Shared DO termination label 50 at (1)
swim.F:129:72:

  129 |       DO 60 I=1,M
      |                                                                       
1
Warning: Fortran 2018 deleted feature: Shared DO termination label 60 at (1)
swim.F:173:72:

  173 |       DO 86 I=1,MP1
      |                                                                       
1
Warning: Fortran 2018 deleted feature: Shared DO termination label 86 at (1)
swim.F:206:72:

  206 |       DO 100 I=1,M
      |                                                                       
1
Warning: Fortran 2018 deleted feature: Shared DO termination label 100 at (1)
swim.F:278:72:

  278 |       DO 200 I=1,M
      |                                                                       
1
Warning: Fortran 2018 deleted feature: Shared DO termination label 200 at (1)
swim.F:340:72:

  340 |       DO 400 I=1,MP1
      |                                                                       
1
Warning: Fortran 2018 deleted feature: Shared DO termination label 400 at (1)
swim.F:376:72:

  376 |       DO 300 I=1,M
      |                                                                       
1
Warning: Fortran 2018 deleted feature: Shared DO termination label 300 at (1)
during RTL pass: expand
swim.F:404:72:

  404 | !$omp target
      |                                                                       
^
internal compiler error: in expand_expr_real_1, at expr.cc:10897
0x9d1067 expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:10897
0x9cce9b expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier,
rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:9000
0x9cce9b expand_expr(tree_node*, rtx_def*, machine_mode, expand_modifier)
        ../../gcc-13.1.0/gcc/expr.h:310
0x9cce9b expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:11234
0x9ce88f expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier,
rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:9000
0x9ce88f expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:11463
0x9cff3b expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier,
rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:9000
0x9cff3b expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:10805
0x9cce9b expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier,
rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:9000
0x9cce9b expand_expr(tree_node*, rtx_def*, machine_mode, expand_modifier)
        ../../gcc-13.1.0/gcc/expr.h:310
0x9cce9b expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:11234
0x9cff3b expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier,
rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:9000
0x9cff3b expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
expand_modifier, rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:10805
0x9da10f expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier,
rtx_def**, bool)
        ../../gcc-13.1.0/gcc/expr.cc:9000
0x9da10f store_expr(tree_node*, rtx_def*, int, bool, bool)
        ../../gcc-13.1.0/gcc/expr.cc:6330
0x9db73b expand_assignment(tree_node*, tree_node*, bool)
        ../../gcc-13.1.0/gcc/expr.cc:6048
0x8ca803 expand_gimple_stmt_1
        ../../gcc-13.1.0/gcc/cfgexpand.cc:3946
0x8ca803 expand_gimple_stmt
        ../../gcc-13.1.0/gcc/cfgexpand.cc:4044
0x8cf367 expand_gimple_basic_block
        ../../gcc-13.1.0/gcc/cfgexpand.cc:6106
0x8d0e3f execute
        ../../gcc-13.1.0/gcc/cfgexpand.cc:6841
Please submit a full bug report, with preprocessed source.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

anlauf at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
            Summary|internal compiler error: in |[13/14 Regression] internal
                   |expand_expr_real_1, at      |compiler error: in
                   |expr.cc:10897               |expand_expr_real_1, at
                   |                            |expr.cc:10897
   Target Milestone|---                         |13.2
           Keywords|                            |ice-on-valid-code, openmp
   Last reconfirmed|                            |2023-07-18
           Priority|P3                          |P4

--- Comment #1 from anlauf at gcc dot gnu.org ---
The code is accepted by Nvidia.

No ICE at -O0, but ICE at -Og, -O1, and higher.

ICE confirmed also on x86_64, so likely not target-specific.

[Bug fortran/110725] [13/14 Regression,openmp] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

kargl at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[13/14 Regression] internal |[13/14 Regression,openmp]
                   |compiler error: in          |internal compiler error: in
                   |expand_expr_real_1, at      |expand_expr_real_1, at
                   |expr.cc:10897               |expr.cc:10897
                 CC|                            |kargl at gcc dot gnu.org

--- Comment #2 from kargl at gcc dot gnu.org ---
Reduced testcase.  The '!$omp end teams' line in subroutine initial appears to
be out-of-place.


      module swim_mod

      INTEGER, PARAMETER :: N1=7702, N2=7702

      DOUBLE PRECISION, ALLOCATABLE, DIMENSION(:,:) :: U, V

      INTEGER :: M, N, MP1, NP1

!$omp declare target(U, V)
!$omp declare target(M,N,MP1,NP1)

      CONTAINS

      SUBROUTINE ALLOC
         IMPLICIT NONE
!$omp target update to(M,N,MP1,NP1)
!$omp&
         ALLOCATE(U(NP1,MP1), V(NP1,MP1))
      END SUBROUTINE

      SUBROUTINE INITAL
      INTEGER I
!$omp target
!$omp teams
!$omp distribute parallel do simd
      DO 75 I=1,M
         U(I+1,N+1) = U(I+1,1)
         V(I,1) = V(I,N+1)
   75 CONTINUE
!$omp end teams
      U(1,N+1) = U(M+1,1)
      V(M+1,1) = V(1,N+1)
!$omp end target
      END SUBROUTINE

      end module swim_mod

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |accepts-invalid
           Priority|P4                          |P3
            Summary|[13/14 Regression,openmp]   |[13/14 Regression] internal
                   |internal compiler error: in |compiler error: in
                   |expand_expr_real_1, at      |expand_expr_real_1, at
                   |expr.cc:10897               |expr.cc:10897

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
That would make it accepts-invalid then?

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

kargl at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |14.0
      Known to work|                            |12.2.0

--- Comment #4 from kargl at gcc dot gnu.org ---
(In reply to Richard Biener from comment #3)
> That would make it accepts-invalid then?

I'm not an expert with OpenMP, so cannot make a decision here.
Note, the original code compiles with 12.2.0.  I don't have 
13.x installed for testing.

While reducing the code to something much smaller, I notice the
idiom of

!$omp target
!$omp teams
!$omp distribute parallel do simd
      DO 75 I=1,M
        <Fortran code here>
   75 CONTINUE
!$omp end teams
!$omp end target

everywhere except for the one occurrence in the reduce code of

!$omp target
!$omp teams
!$omp distribute parallel do simd
      DO 75 I=1,M
         U(I+1,N+1) = U(I+1,1)
         V(I,1) = V(I,N+1)
   75 CONTINUE
!$omp end teams
      U(1,N+1) = U(M+1,1)      <-- This seems out-of-place when compared
      V(M+1,1) = V(1,N+1)      <-- to all other code.
!$omp end target
      END SUBROUTINE

If I move the '!$omp end teams' down two lines, the code compiles.  If I remove
the two lines of Fortran code, the code compiles.

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

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

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

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
This is certainly accepts-invalid if it is accepted.
E.g. OpenMP 5.0 says on this in 2.20, 5.1 in
https://www.openmp.org/spec-html/5.1/openmpse30.html#x155-1880002.22
"If a teams construct is nested within a target construct, that target
construct must contain no statements, declarations or directives outside of the
teams construct."
Though, diagnosing this is hard, because e.g. parsing of teams construct's
clauses can result in some code in between the target and teams nested in it,
and that is valid.
I mean say
int foo (void) { return 42; }
#pragma omp declare target (foo)

int
main ()
{
#pragma omp target
#pragma omp teams num_teams (foo ())
;
}

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Tobias Burnus <burnus@gcc.gnu.org>:

https://gcc.gnu.org/g:2e31fe431b08b0302e1fa8a1c18ee51adafd41df

commit r14-2754-g2e31fe431b08b0302e1fa8a1c18ee51adafd41df
Author: Tobias Burnus <tobias@codesourcery.com>
Date:   Mon Jul 24 22:57:07 2023 +0200

    OpenMP/Fortran: Reject not strictly nested target -> teams [PR110725,
PR71065]

    OpenMP requires: "If a teams region is nested inside a target region, the
    corresponding target construct must not contain any statements,
declarations
    or directives outside of the corresponding teams construct."

    This commit checks now for this restriction.

            PR fortran/110725
            PR middle-end/71065

    gcc/fortran/ChangeLog:

            * gfortran.h (gfc_omp_clauses): Add contains_teams_construct.
            * openmp.cc (resolve_omp_target): New; check for teams nesting.
            (gfc_resolve_omp_directive): Call it.
            * parse.cc (decode_omp_directive): Set contains_teams_construct
            on enclosing ST_OMP_TARGET.

    gcc/testsuite/ChangeLog:

            * gfortran.dg/gomp/pr99226.f90: Update dg-error.
            * gfortran.dg/gomp/teams-5.f90: New test.

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

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

https://gcc.gnu.org/g:09dda270380fe13e7b4722305cb1122df1f779a0

commit r14-2761-g09dda270380fe13e7b4722305cb1122df1f779a0
Author: Tobias Burnus <tobias@codesourcery.com>
Date:   Tue Jul 25 15:43:58 2023 +0200

    OpenMP/Fortran: Reject declarations between target + teams

    While commit r14-2754-g2e31fe431b08b0302e1fa8a1c18ee51adafd41df
    detected executable statements, declarations do not show up as
    executable statements.  Hence, we now check whether the first
    statement after TARGET is TEAMS - such that we can detect data
    statements like type or variable declarations.  Fortran semantics
    ensures that only executable directives/statemens can come after
    '!$omp end teams' such that those can be detected with the
    previous check.

    Note that statements returning ST_NONE such as 'omp nothing' or
    'omp error at(compilation)' will still slip through.

            PR fortran/110725
            PR middle-end/71065

    gcc/fortran/ChangeLog:

            * gfortran.h (gfc_omp_clauses): Add target_first_st_is_teams.
            * parse.cc (parse_omp_structured_block): Set it if the first
            statement in the structured block of a TARGET is TEAMS or
            a combined/composite starting with TEAMS.
            * openmp.cc (resolve_omp_target): Also show an error for
            contains_teams_construct without target_first_st_is_teams.

    gcc/testsuite/ChangeLog:

            * gfortran.dg/gomp/teams-6.f90: New test.

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|13.2                        |13.3

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 13.2 is being released, retargeting bugs to GCC 13.3.

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Tobias Burnus <burnus@gcc.gnu.org>:

https://gcc.gnu.org/g:081e25d3cfd86c4094999ded0bbe99b91762013c

commit r14-2826-g081e25d3cfd86c4094999ded0bbe99b91762013c
Author: Tobias Burnus <tobias@codesourcery.com>
Date:   Thu Jul 27 18:14:11 2023 +0200

    OpenMP/Fortran: Extend reject code between target + teams [PR71065,
PR110725]

    The previous version failed to diagnose when the 'teams' was nested
    more deeply inside the target region, e.g. inside a DO or some
    block or structured block.

                PR fortran/110725
                PR middle-end/71065

    gcc/fortran/ChangeLog:

            * openmp.cc (resolve_omp_target): Minor cleanup.
            * parse.cc (decode_omp_directive): Find TARGET statement
            also higher in the stack.

    gcc/testsuite/ChangeLog:

            * gfortran.dg/gomp/teams-6.f90: Extend.

[Bug fortran/110725] [13/14 Regression] internal compiler error: in expand_expr_real_1, at expr.cc:10897

[law at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110725

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

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

--- Comment #10 from Jeffrey A. Law <law at gcc dot gnu.org> ---
Fixed on the trunk.