[Bug tree-optimization/115661] New: wrong code at -O{2,3} on x86_64-linux-gnu

[Bug tree-optimization/115661] New: wrong code at -O{2,3} on x86_64-linux-gnu

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

            Bug ID: 115661
           Summary: wrong code at -O{2,3} on x86_64-linux-gnu
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zhendong.su at inf dot ethz.ch
  Target Milestone: ---

This appears to be a recent regression as it doesn't reproduce with 14.1.

Compiler Explorer: https://godbolt.org/z/1dneGGoTE

[517] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/15.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--enable-checking=yes --prefix=/local/suz-local/software/local/gcc-trunk
--enable-sanitizers --enable-languages=c,c++ --disable-werror --enable-multilib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 15.0.0 20240626 (experimental) (GCC) 
[518] % 
[518] % gcctk -O0 small.c; ./a.out
[519] % gcctk -O2 small.c
[520] % ./a.out
0
[521] % cat small.c
int printf(const char *, ...);
int a, b[6], c = 2, d, e = 446984904, *g, h, *i = &b[5];
int *m() {
  int *f = &c;
  if (d < 6)
    return &b[5];
  return f;
}
void n() {
  int j;
  for (; a < 1; a++)
    for (j = 0; j < 1; j++) {
      g = m();
      int k = *g;
      if (k) {
        int l = (3504049682 + e) & 3504049682;
        h = ~(k * l) * 7455084268;
        if (h > -1436818931) {
          printf("0\n");
          continue;
        }
        return;
      }
    }
}
int main() {
  *i = 1;
  n();
  return 0;
}

[Bug tree-optimization/115661] wrong code at -O{2,3} on x86_64-linux-gnu

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

--- Comment #1 from Zhendong Su <zhendong.su at inf dot ethz.ch> ---
The following might be the same or related although it needs several flags to
trigger. 

Compiler Explorer: https://godbolt.org/z/Y9414PbTh

[567] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/15.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--enable-checking=yes --prefix=/local/suz-local/software/local/gcc-trunk
--enable-sanitizers --enable-languages=c,c++ --disable-werror --enable-multilib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 15.0.0 20240626 (experimental) (GCC) 
[568] % 
[568] % gcctk -O2 small.c; ./a.out
[569] % 
[569] % gcctk -O2 -fno-tree-pre -fschedule-insns -fno-forward-propagate
-fno-guess-branch-probability -fno-gcse small.c
[570] % ./a.out
0
[571] % 
[571] % cat small.c
int printf(const char *, ...);
struct {
  unsigned : 8;
  volatile unsigned b : 1;
  int c : 31;
  int d : 5;
} g, k, m, n, p;
unsigned e = 5;
volatile int f = 1;
int h, i, l, o, r, q;
long j, a;
int main() {
  if (0)
  L:
    for (; e; e++)
      ;
  for (; j < 5; j = 0)
    if (f) {
      int t = 7 - (((k.d | 7) ^ m.c) - a) - ~(n.b & -o), u, v = -t ^ q,
          w = e - h;
      u = k.d;
      if (i) {
        if (w > 6)
          goto L;
        printf("%d", l);
      }
      g.d = (g.d & (n.b - 5)) - (-((k.d - 7 - o) & t) ^ (v & q));
      n.b = t;
      m.c = ~(t - (7L & m.c));
      k.d = ~(a & (m.c | ~o) & ~u) & g.d;
      if ((unsigned) p.d >= e)
        printf("0\n");
      return 0;
    }
  return 0;
}

[Bug tree-optimization/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|wrong code at -O{2,3} on    |[15 Regression] wrong code
                   |x86_64-linux-gnu            |at -O{2,3} on
                   |                            |x86_64-linux-gnu
   Target Milestone|---                         |15.0
            Version|unknown                     |15.0
           Keywords|                            |needs-bisection, wrong-code

[Bug tree-optimization/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

Sam James <sjames at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |Evgeny.Karpov at microsoft dot com
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=115635

--- Comment #2 from Sam James <sjames at gcc dot gnu.org> ---
Bisect says either of the two which is also in the range I suspected for
PR115635:

63512c72df09 Extract ix86 dllimport implementation to mingw
104d06c028c6 (HEAD) Move mingw_* declarations to the mingw folder

[Bug tree-optimization/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

Sam James <sjames at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|needs-bisection             |
     Ever confirmed|0                           |1
            Summary|[15 Regression] wrong code  |[15 Regression] wrong code
                   |at -O{2,3} on               |at -O{2,3} on
                   |x86_64-linux-gnu            |x86_64-linux-gnu since
                   |                            |r15-1599-g63512c72df09b4
   Last reconfirmed|                            |2024-06-26
             Status|UNCONFIRMED                 |NEW

--- Comment #3 from Sam James <sjames at gcc dot gnu.org> ---
r15-1599-g63512c72df09b4, supposedly?

[Bug tree-optimization/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

--- Comment #4 from Sam James <sjames at gcc dot gnu.org> ---
This fixes it for me as a hack:
```
--- a/gcc/config/i386/i386-expand.cc
+++ b/gcc/config/i386/i386-expand.cc
@@ -412,20 +412,10 @@ ix86_expand_move (machine_mode mode, rtx operands[])
        }
       else
        {
-#if TARGET_PECOFF
-         tmp = legitimize_pe_coff_symbol (op1, addend != NULL_RTX);
-         if (tmp)
-           {
-             op1 = tmp;
-             if (!addend)
-               break;
-           }
-         else
            {
              op1 = operands[1];
              break;
            }
-#endif
        }

       if (addend)
```

Before r15-1599-g63512c72df09b4, legitimize_pe_coff_symbol would return
NULL_RTX for non-PECOFF targets, so the else branch would get taken.

[Bug target/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Note the ix86_GOT_alias_set change also looks wrong.  AFAICS it now sets
MEM_ALIAS_SET to -1 which is ALIAS_SET_MEMORY_BARRIER and a pessimization.

[Bug target/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[Evgeny.Karpov at microsoft dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

--- Comment #6 from Evgeny Karpov <Evgeny.Karpov at microsoft dot com> ---
Sorry for the regression.
I am working on the fix and will try to validate and submit it today.

[Bug target/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

Sam James <sjames at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |Evgeny.Karpov at microsoft dot com
             Status|NEW                         |ASSIGNED

[Bug target/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

--- Comment #7 from Sam James <sjames at gcc dot gnu.org> ---
To the reporter: I should say, thanks for finding this so quickly - it made it
a lot easier to debug.

[Bug target/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[Evgeny.Karpov at microsoft dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

--- Comment #8 from Evgeny Karpov <Evgeny.Karpov at microsoft dot com> ---
Thank you for reporting the issues and discussing the root causes. 
It helped in preparing the patch.

The patch is under review.
https://gcc.gnu.org/pipermail/gcc-patches/2024-June/655807.html

[Bug target/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

--- Comment #9 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Christophe Lyon <clyon@gcc.gnu.org>:

https://gcc.gnu.org/g:c4b7b62bfa10816c1e08f0f9597d857f11379688

commit r15-1708-gc4b7b62bfa10816c1e08f0f9597d857f11379688
Author: Evgeny Karpov <Evgeny.Karpov@microsoft.com>
Date:   Fri Jun 28 12:37:12 2024 +0000

    i386: Fix regression after refactoring legitimize_pe_coff_symbol,
ix86_GOT_alias_set and PE_COFF_LEGITIMIZE_EXTERN_DECL [PR115635]

    This patch fixes 3 bugs reported after merging the "Add DLL
    import/export implementation to AArch64" series.
    https://gcc.gnu.org/pipermail/gcc-patches/2024-June/653955.html The
    series refactors the i386 codebase to reuse it in AArch64, which
    triggers some bugs.

    Bug 115661 - [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu
    since r15-1599-g63512c72df09b4
    https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

    Bug 115635 - [15 regression] Bootstrap fails with failed self-test
    with the rust fe (diagnostic-path.cc:1153: test_empty_path: FAIL:
    ASSERT_FALSE ((path.interprocedural_p ()))) since
    r15-1599-g63512c72df09b4
    https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115635

    Issue 1. In some code, i386 has been relying on the
    legitimize_pe_coff_symbol call on all platforms and should return
    NULL_RTX if it is not supported.

    Fix: NULL_RTX handling has been added when the target does not support
    PECOFF.

    Issue 2. ix86_GOT_alias_set is used on all platforms and cannot be
    extracted to mingw.

    Fix: ix86_GOT_alias_set has been returned as it was and is used on all
    platforms for i386.

    Bug 115643 - [15 regression] aarch64-w64-mingw32 support today breaks
    x86_64-w64-mingw32 build cannot represent relocation type BFD_RELOC_64
    since r15-1602-ged20feebd9ea31
    https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115643

    Issue 3. PE_COFF_EXTERN_DECL_SHOULD_BE_LEGITIMIZED has been added and
    used with a negative operator for a complex expression without braces.

    Fix: Braces has been added, and
    PE_COFF_EXTERN_DECL_SHOULD_BE_LEGITIMIZED has been renamed to
    PE_COFF_LEGITIMIZE_EXTERN_DECL.

    2024-06-28  Evgeny Karpov <Evgeny.Karpov@microsoft.com>

            gcc/ChangeLog:
            PR bootstrap/115635
            PR target/115643
            PR target/115661
            * config/aarch64/cygming.h
            (PE_COFF_EXTERN_DECL_SHOULD_BE_LEGITIMIZED): Rename to
            PE_COFF_LEGITIMIZE_EXTERN_DECL.
            (PE_COFF_LEGITIMIZE_EXTERN_DECL): Likewise.
            * config/i386/cygming.h (GOT_ALIAS_SET): Remove the diffinition to
            reuse it from i386.h.
            (PE_COFF_EXTERN_DECL_SHOULD_BE_LEGITIMIZED): Rename to
            PE_COFF_LEGITIMIZE_EXTERN_DECL.
            (PE_COFF_LEGITIMIZE_EXTERN_DECL): Likewise.
            * config/i386/i386-expand.cc (ix86_expand_move): Return
            ix86_GOT_alias_set.
            * config/i386/i386-expand.h (ix86_GOT_alias_set): Likewise.
            * config/i386/i386.cc (ix86_GOT_alias_set): Likewise.
            * config/i386/i386.h (GOT_ALIAS_SET): Likewise.
            * config/mingw/winnt-dll.cc (get_dllimport_decl): Use
            GOT_ALIAS_SET.
            (legitimize_pe_coff_symbol): Rename to
            PE_COFF_LEGITIMIZE_EXTERN_DECL.
            * config/mingw/winnt-dll.h (ix86_GOT_alias_set): Declare
            ix86_GOT_alias_set.

[Bug target/115661] [15 Regression] wrong code at -O{2,3} on x86_64-linux-gnu since r15-1599-g63512c72df09b4

[clyon at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115661

Christophe Lyon <clyon at gcc dot gnu.org> changed:

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

--- Comment #10 from Christophe Lyon <clyon at gcc dot gnu.org> ---
Fixed.