[Bug c++/114409] New: ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[Bug c++/114409] New: ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

            Bug ID: 114409
           Summary: ICE after adding novector pragmas (internal compiler
                    error: in tsubst_expr, at cp/pt.cc:21794)
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: sjames at gcc dot gnu.org
  Target Milestone: ---

Created attachment 57751
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57751&action=edit
EarlyCSE.cpp.ii.xz

While working on PR114403, I was adding novector pragmas and ended up hitting
an ICE.

```
$ g++ -c
/home/sam/data/build/llvm-project-test/lib/Transforms/Scalar/CMakeFiles/LLVMScalarOpts.dir/EarlyCSE.cpp.ii
In file included from
/home/sam/git/llvm-project/llvm/lib/Transforms/Scalar/EarlyCSE.cpp:18:
/home/sam/git/llvm-project/llvm/include/llvm/ADT/ScopedHashTable.h: In
instantiation of ‘llvm::ScopedHashTableScope<K, V, KInfo,
AllocatorTy>::~ScopedHashTableScope() [with K = {anonymous}::SimpleValue; V =
llvm::Value*; KInfo = llvm::DenseMapInfo<{anonymous}::SimpleValue>; AllocatorTy
= llvm::RecyclingAllocator<llvm::BumpPtrAllocatorImpl<>,
llvm::ScopedHashTableVal<{anonymous}::SimpleValue, llvm::Value*> >]’:
/home/sam/git/llvm-project/llvm/lib/Transforms/Scalar/EarlyCSE.cpp:671:9:  
required from here
  671 |       : Scope(AvailableValues), LoadScope(AvailableLoads),
      |         ^~~~~~~~~~~~~~~~~~~~~~
/home/sam/git/llvm-project/llvm/include/llvm/ADT/ScopedHashTable.h:241:36:
internal compiler error: in tsubst_expr, at cp/pt.cc:21794
  241 |   while (ScopedHashTableVal<K, V> *ThisEntry = LastValInScope) {
      |                                    ^~~~~~~~~
0x55963b55a154 tsubst_expr(tree_node*, tree_node*, int, tree_node*)
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:21794
0x55963cdda158 tsubst_expr(tree_node*, tree_node*, int, tree_node*)
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:20637
0x55963b55a4ff tsubst_expr(tree_node*, tree_node*, int, tree_node*)
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:21776
0x55963d07cfa1 tsubst_stmt
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:19438
0x55963d07dbe9 tsubst_stmt
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:18676
0x55963d07cfe3 tsubst_stmt
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:18398
0x55963d07cea6 tsubst_stmt
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:18771
0x55963d07cfe3 tsubst_stmt
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:18398
0x55963d07cea6 tsubst_stmt
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:18771
0x55963d3347e8 instantiate_body
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:27064
0x55963cf46498 instantiate_decl(tree_node*, bool, bool)
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:27349
0x55963ca8a8c1 instantiate_pending_templates(int)
        /usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/pt.cc:27425
0x55963ca7f13a c_parse_final_cleanups()
       
/usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/cp/decl2.cc:5151
0x55963d25b647 c_common_parse_file()
       
/usr/src/debug/sys-devel/gcc-14.0.9999/gcc-14.0.9999/gcc/c-family/c-opts.cc:1329
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See <https://bugs.gentoo.org/> for instructions.
```

[Bug c++/114409] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #1 from Sam James <sjames at gcc dot gnu.org> ---
If you can think of a workaround, please let me know, as unfortunately this
loop is the one where the assert in the other bug gets hit :(

[Bug c++/114409] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Just an FYI, using unroll instead of novect also ICEs. Let me reduce this and
see if it is a regression.

[Bug c++/114409] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Sam James from comment #1)
> If you can think of a workaround, please let me know, as unfortunately this
> loop is the one where the assert in the other bug gets hit :(

An easy workaround is to add:
asm("":::"memory");

Right after the opening `{`.
But that might be too much of a hammer in some cases ...

[Bug c++/114409] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #3)
> (In reply to Sam James from comment #1)
> > If you can think of a workaround, please let me know, as unfortunately this
> > loop is the one where the assert in the other bug gets hit :(
> 
> An easy workaround is to add:
> asm("":::"memory");
> 
> Right after the opening `{`.
> But that might be too much of a hammer in some cases ...

Even a simple `__builtin_printf("");` added to the loop might avoid the
vectorization to figure out where the wrong code happens.

[Bug c++/114409] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #5 from Sam James <sjames at gcc dot gnu.org> ---
Thanks, I'll play with those...

[Bug c++/114409] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
The alternative is "bisceting" with -fdbg-cnt=vect_loop (IIRC there were some
ICEs reported when using that, so YMMV)

[Bug c++/114409] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

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

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

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Reduced testcase:
template <typename T>
void
foo (T x)
{
  #pragma GCC novector
  while (T y = x)
    {
      ++y;
    }
}

void
bar ()
{
  foo (-42);
}

[Bug c++/114409] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Perhaps better testcase:

template <typename T>
T
foo (T)
{
  static T t;
  return 42 - ++t;
}

template <typename T>
void
bar (T x)
{
  #pragma GCC novector
  while (T y = foo (x))
    ;
}

void
baz ()
{
  bar (0);
}

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0
                 CC|                            |jason at gcc dot gnu.org,
                   |                            |ppalka at gcc dot gnu.org
           Priority|P3                          |P1
            Summary|ICE after adding novector   |[14 Regression] ICE after
                   |pragmas (internal compiler  |adding novector pragmas
                   |error: in tsubst_expr, at   |(internal compiler error:
                   |cp/pt.cc:21794)             |in tsubst_expr, at
                   |                            |cp/pt.cc:21794)

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r14-4229-g9c62af101e11e1cce573c2b3d2e18b403412dbc8
The ANNOTATE_EXPR has a COMPOUND_EXPR with DECL_EXPR inside of it and
tsubst_copy_and_build doesn't handle that.
Wonder if the DECL_EXPR shouldn't be before the ANNOTATE_EXPR instead, at least
without templates I see in original dump
        int y;
    <<cleanup_point <<< Unknown tree: expr_stmt
      (void) (y = foo<int> (x)) >>>>>;
    if (!ANNOTATE_EXPR <y != 0, no-vector>) goto <D.2804>;

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2024-03-21
             Status|UNCONFIRMED                 |NEW

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Note, when not in a template, such as
void
qux (int x)
{
  #pragma GCC novector
  while (int y = foo (x))
    ;
}
with the above foo template, we ignore the annotation because the middle-end
doesn't find it at the right spot:
warning: ignoring loop annotation
C doesn't allow defining a variable in while loop's condition, so can't
cross-check what it does instead.

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 57758
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57758&action=edit
gcc14-pr114409.patch

Ah, I have a fix for the bogus warning, ANNOTATE_EXPR really needs to wrap the
whole condition rather than just part of it.
But, I have no idea what to do with the templates case.

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Reduced to use `GCC unroll` instead:
```
template <int> int t() {
#pragma GCC unroll 4
    while (int ThisEntry = 0) { }
}
int tt = t<1>();
```

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794) since r14-4229-g9c62af101e11e1cce573c2b3d2e18b403412dbc8

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #13 from Richard Biener <rguenth at gcc dot gnu.org> ---
The question is what handles the COMPOUND_EXPR with DECL_EXPR when the
ANNOTATE_EXPR isn't around.  tsubst_expr doesn't handle DECL_EXPRs.

It's built even w/o the #pragma via finish_cond and finish_while_stmt_cond
and in that case it's tsubst_stmt that handles the COMPOUND_EXPR and
that function does handle DECL_EXPRs.  Maybe ANNOTATE_EXPR should be handled
at the "stmt" level [as well]?  Or the COMPOUND_EXPR handling in
tsubst_expr should recurse to tsubst_stmt?

That said, ( DECL_EXPR <X>, .. use of X ) is phishy?

Jason?

The following fixes it:

diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
index bf4b89d8413..dae423a751f 100644
--- a/gcc/cp/pt.cc
+++ b/gcc/cp/pt.cc
@@ -20635,8 +20635,11 @@ tsubst_expr (tree t, tree args, tsubst_flags_t
complain, tree in_decl)

     case COMPOUND_EXPR:
       {
-       tree op0 = tsubst_expr (TREE_OPERAND (t, 0), args,
+       tree op0 = tsubst_stmt (TREE_OPERAND (t, 0), args,
                                complain & ~tf_decltype, in_decl);
+       if (op0 == NULL_TREE)
+         /* If the first operand was a statement, we're done with it.  */
+         RETURN (RECUR (TREE_OPERAND (t, 1)));
        RETURN (build_x_compound_expr (EXPR_LOCATION (t),
                                       op0,
                                       RECUR (TREE_OPERAND (t, 1)),

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794) since r14-4229-g9c62af101e11e1cce573c2b3d2e18b403412dbc8

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #14 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I had another look at this.
The reason why this works without the pragmas (i.e. without ANNOTATE_EXPRs) is
that
the WHILE_COND/FOR_COND are handled not by tsubst_expr but tsubst_stmt:
    case FOR_STMT:
      stmt = begin_for_stmt (NULL_TREE, NULL_TREE);
      RECUR (FOR_INIT_STMT (t));
      finish_init_stmt (stmt);
      tmp = RECUR (FOR_COND (t));
      finish_for_cond (tmp, stmt, false, 0, false);
and
    case WHILE_STMT:
      stmt = begin_while_stmt ();
      tmp = RECUR (WHILE_COND (t));
      finish_while_stmt_cond (tmp, stmt, false, 0, false);
in tsubst_stmt, where RECUR in that case means tsubst_stmt, not tsubst_expr.
So, one possible fix is to use tsubst_stmt rather than RECUR aka tsubst_expr
on the first operand of ANNOTATE_EXPR.
And another would be to move the ANNOTATE_EXPR handling from tsubst_expr to
tsubst_stmt.

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794) since r14-4229-g9c62af101e11e1cce573c2b3d2e18b403412dbc8

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

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

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

--- Comment #15 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 57917
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57917&action=edit
gcc14-pr114409-1.patch

One possible untested fix.

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794) since r14-4229-g9c62af101e11e1cce573c2b3d2e18b403412dbc8

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #16 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 57918
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57918&action=edit
gcc14-pr114409-2.patch

Another possible untested fix.

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794) since r14-4229-g9c62af101e11e1cce573c2b3d2e18b403412dbc8

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #17 from GCC 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:cb46aca0a07355abf2f0b04f52087bca8f848524

commit r14-9910-gcb46aca0a07355abf2f0b04f52087bca8f848524
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Thu Apr 11 09:46:00 2024 +0200

    c++: Fix ANNOTATE_EXPR instantiation [PR114409]

    The following testcase ICEs starting with the r14-4229 PR111529
    change which moved ANNOTATE_EXPR handling from tsubst_expr to
    tsubst_copy_and_build.
    ANNOTATE_EXPR is only allowed in the IL to wrap a loop condition,
    and the loop condition of while/for loops can be a COMPOUND_EXPR
    with DECL_EXPR in the first operand and the corresponding VAR_DECL
    in the second, as created by finish_cond
          else if (!empty_expr_stmt_p (cond))
            expr = build2 (COMPOUND_EXPR, TREE_TYPE (expr), cond, expr);
    Since then Patrick reworked the instantiation, so that we have now
    tsubst_stmt and tsubst_expr and ANNOTATE_EXPR ended up in the latter,
    while only tsubst_stmt can handle DECL_EXPR.

    Now, the reason why the while/for loops with variable declaration
    in the condition works in templates without the pragmas (i.e. without
    ANNOTATE_EXPR) is that both the FOR_STMT and WHILE_STMT handling uses
    RECUR aka tsubst_stmt in handling of the *_COND operand:
        case FOR_STMT:
          stmt = begin_for_stmt (NULL_TREE, NULL_TREE);
          RECUR (FOR_INIT_STMT (t));
          finish_init_stmt (stmt);
          tmp = RECUR (FOR_COND (t));
          finish_for_cond (tmp, stmt, false, 0, false);
    and
        case WHILE_STMT:
          stmt = begin_while_stmt ();
          tmp = RECUR (WHILE_COND (t));
          finish_while_stmt_cond (tmp, stmt, false, 0, false);
    Therefore, it will handle DECL_EXPR embedded in COMPOUND_EXPR of the
    {WHILE,FOR}_COND just fine.
    But if that COMPOUND_EXPR with DECL_EXPR is wrapped with one or more
    ANNOTATE_EXPRs, because ANNOTATE_EXPR is now done solely in tsubst_expr
    and uses RECUR there (i.e. tsubst_expr), it will ICE on DECL_EXPR in there.

    This could be fixed by keeping ANNOTATE_EXPR handling in tsubst_expr but
    using tsubst_stmt for the first operand, but this patch instead
    moves ANNOTATE_EXPR handling to tsubst_stmt (and uses tsubst_expr for the
    second/third operand).

    2024-04-11  Jakub Jelinek  <jakub@redhat.com>

            PR c++/114409
            * pt.cc (tsubst_expr) <case ANNOTATE_EXPR>: Move to ...
            (tsubst_stmt) <case ANNOTATE_EXPR>: ... here.  Use tsubst_expr
            instead of RECUR for the last 2 arguments.

            * g++.dg/ext/pr114409-2.C: New test.

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794) since r14-4229-g9c62af101e11e1cce573c2b3d2e18b403412dbc8

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

--- Comment #18 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The P1 bug has been fixed, I've moved the rest to PR114691.

[Bug c++/114409] [14 Regression] ICE after adding novector pragmas (internal compiler error: in tsubst_expr, at cp/pt.cc:21794) since r14-4229-g9c62af101e11e1cce573c2b3d2e18b403412dbc8

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114409

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

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

--- Comment #19 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
.