[Bug tree-optimization/111916] New: wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs)

[Bug tree-optimization/111916] New: wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs)

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

            Bug ID: 111916
           Summary: wrong code at -O1 and above on x86_64-linux-gnu (the
                    generated code hangs)
           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.

Compiler Explorer: https://godbolt.org/z/aEEKbdY6E

[563] % 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/14.0.0/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 14.0.0 20231022 (experimental) (GCC) 
[564] % 
[564] % gcctk -O0 small.c; ./a.out
[565] % 
[565] % gcctk -O1 small.c
[566] % timeout -s 9 5 ./a.out
Killed
[567] % 
[567] % cat small.c
#pragma pack(1)
struct A {
  int b : 4;
  int c : 11;
  int d : 2;
  int e : 5;
} f;
int main() {
  struct A g = {1, 1, 1, 1};
  while (!g.b)
    f = g;
  return 0;
}

[Bug tree-optimization/111916] [14 Regression] wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs)

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|wrong code at -O1 and above |[14 Regression] wrong code
                   |on x86_64-linux-gnu (the    |at -O1 and above on
                   |generated code hangs)       |x86_64-linux-gnu (the
                   |                            |generated code hangs)
                 CC|                            |tkoenig at gcc dot gnu.org
           Keywords|                            |wrong-code
   Target Milestone|---                         |14.0

--- Comment #1 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
Also occurs on POWER, so likely target-independent.  Does not happen
for 13.2.

[Bug tree-optimization/111916] [14 Regression] wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs)

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

--- Comment #2 from Sam James <sjames at gcc dot gnu.org> ---
I'll bisect.

[Bug tree-optimization/111916] [14 Regression] wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
                 CC|                            |pinskia at gcc dot gnu.org,
                   |                            |rguenth at gcc dot gnu.org
   Last reconfirmed|                            |2023-10-22

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.


Either r14-4677-g3aaf704bca3e01c72d9fe6de01ae5416dba6b5db or
r14-4612-g6decda1a35be5764101987c210b5693a0d914e58 .

[Bug tree-optimization/111916] [14 Regression] wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs) since r14-4612-g6decda1a35be5764101987c210b5693a0d914e58

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[14 Regression] wrong code  |[14 Regression] wrong code
                   |at -O1 and above on         |at -O1 and above on
                   |x86_64-linux-gnu (the       |x86_64-linux-gnu (the
                   |generated code hangs)       |generated code hangs) since
                   |                            |r14-4612-g6decda1a35be57641
                   |                            |01987c210b5693a0d914e58

--- Comment #4 from Sam James <sjames at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #3)
> Confirmed.
> 
> 
> Either r14-4677-g3aaf704bca3e01c72d9fe6de01ae5416dba6b5db or
> r14-4612-g6decda1a35be5764101987c210b5693a0d914e58 .

yep, r14-4612-g6decda1a35be5764101987c210b5693a0d914e58

6decda1a35be5764101987c210b5693a0d914e58 is the first bad commit
commit 6decda1a35be5764101987c210b5693a0d914e58
Author: Richard Biener <rguenther@suse.de>
Date:   Thu Oct 12 11:34:57 2023 +0200

    tree-optimization/111779 - Handle some BIT_FIELD_REFs in SRA

[Bug tree-optimization/111916] [14 Regression] wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs) since r14-4612-g6decda1a35be5764101987c210b5693a0d914e58

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
             Status|NEW                         |ASSIGNED
            Version|unknown                     |14.0
           Priority|P3                          |P1

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
I will have a look.

[Bug tree-optimization/111916] [14 Regression] wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs) since r14-4612-g6decda1a35be5764101987c210b5693a0d914e58

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

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

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

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
ESRA does

 int main ()
 {
+  unsigned char g;
   struct A g;
   unsigned char _1;
   unsigned char _2;

   <bb 2> :
   g = *.LC0;
+  g_6 = BIT_FIELD_REF <*.LC0, 8, 0>;
   goto <bb 4>; [INV]

   <bb 3> :
+  BIT_FIELD_REF <g, 8, 0> = g_6;
   f = g;

   <bb 4> :
-  _1 = BIT_FIELD_REF <g, 8, 0>;
+  _1 = g_6;
   _2 = _1 & 15;
   if (_2 == 0)
     goto <bb 3>; [INV]

which looks OK, but then SRA does

+  unsigned char SR.5;
+  unsigned char g;
   unsigned char g;
   struct A g;
   unsigned char _1;
   unsigned char _11;

   <bb 2> [local count: 118111600]:
+  SR.5_10 = SR.5_2(D);
   g = *.LC0;
-  g_5 = BIT_FIELD_REF <*.LC0, 8, 0>;
+  g_14 = SR.5_10;
+  g_5 = SR.5_10;
   _11 = g_5 & 15;
   if (_11 == 0)
     goto <bb 3>; [0.00%]
@@ -19,6 +33,8 @@

   <bb 3> [local count: 955630224]:
   BIT_FIELD_REF <g, 8, 0> = g_5;
+  g_16 = MEM <unsigned char> [(struct A *)&g];
+  MEM <unsigned char> [(struct A *)&g] = g_16;
   f = g;
   _1 = g_5 & 15;
   goto <bb 3>; [100.00%]

which looks weird.  First initialize_constant_pool_replacements inserts

SR.5 = BIT_FIELD_REF <*.LC0, 8, 0>;

which loosk OK, but then sra_modify_assign wrecks this, replacing the RHS
with 'SR.5', yielding

SR.5 = SR.5

this happens before the

  /* Avoid modifying initializations of constant-pool replacements.  */
  if (racc && (racc->replacement_decl == lhs))
    return SRA_AM_NONE; 

code.

I can paper over with the following, but I wonder if sra_modify_expr needs
that very same check for constant-pool replacements?

diff --git a/gcc/tree-sra.cc b/gcc/tree-sra.cc
index f8dff8b27d7..efb52453250 100644
--- a/gcc/tree-sra.cc
+++ b/gcc/tree-sra.cc
@@ -4275,7 +4275,8 @@ sra_modify_assign (gimple *stmt, gimple_stmt_iterator
*gsi)

   if (TREE_CODE (rhs) == REALPART_EXPR || TREE_CODE (lhs) == REALPART_EXPR
       || TREE_CODE (rhs) == IMAGPART_EXPR || TREE_CODE (lhs) == IMAGPART_EXPR
-      || TREE_CODE (rhs) == BIT_FIELD_REF || TREE_CODE (lhs) == BIT_FIELD_REF)
+      || (TREE_CODE (rhs) == BIT_FIELD_REF && !sra_handled_bf_read_p (rhs))
+      || TREE_CODE (lhs) == BIT_FIELD_REF)
     {
       modify_this_stmt = sra_modify_expr (gimple_assign_rhs1_ptr (stmt),
                                          gsi, false);

I will test this now.

[Bug tree-optimization/111916] [14 Regression] wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs) since r14-4612-g6decda1a35be5764101987c210b5693a0d914e58

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

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

https://gcc.gnu.org/g:458db9b6149b2e9bef94ab76909eb914ed9f675a

commit r14-4866-g458db9b6149b2e9bef94ab76909eb914ed9f675a
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Oct 23 14:08:41 2023 +0200

    tree-optimization/111916 - SRA of BIT_FIELD_REF of constant pool entries

    The following adjusts a leftover BIT_FIELD_REF special-casing to only
    cover the cases general code doesn't handle.

            PR tree-optimization/111916
            * tree-sra.cc (sra_modify_assign): Do not lower all
            BIT_FIELD_REF reads that are sra_handled_bf_read_p.

            * gcc.dg/torture/pr111916.c: New testcase.

[Bug tree-optimization/111916] [14 Regression] wrong code at -O1 and above on x86_64-linux-gnu (the generated code hangs) since r14-4612-g6decda1a35be5764101987c210b5693a0d914e58

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111916

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

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

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.