[Bug tree-optimization/58956] New: wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[Bug tree-optimization/58956] New: wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[su at cs dot ucdavis.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

            Bug ID: 58956
           Summary: wrong code at -O1 and above (affecting gcc 4.6 to
                    trunk)
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: su at cs dot ucdavis.edu

The current gcc trunk (as well as gcc 4.6.x, 4.7.x, and 4.8.x) produces wrong
code for the following testcase on x86_64-linux-gnu when compiled at -O1 and
above in 64-bit mode. 

$ gcc-trunk -v
Using built-in specs.
COLLECT_GCC=gcc-trunk
COLLECT_LTO_WRAPPER=/usr/local/gcc-trunk/libexec/gcc/x86_64-unknown-linux-gnu/4.9.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../gcc-trunk/configure
--enable-languages=c,c++,objc,obj-c++,fortran,lto --disable-werror
--enable-checking=none --with-gmp=/usr/local/gcc-trunk
--with-mpfr=/usr/local/gcc-trunk --with-mpc=/usr/local/gcc-trunk
--with-cloog=/usr/local/gcc-trunk --prefix=/usr/local/gcc-trunk
Thread model: posix
gcc version 4.9.0 20131031 (experimental) [trunk revision 204248] (GCC) 
$ 
$ gcc-trunk -O0 small.c; a.out
1
$ gcc-trunk -O1 small.c; a.out
0
$ gcc-4.8 -O1 small.c; a.out
0
$ gcc-4.7 -O1 small.c; a.out
0
$ gcc-4.6 -O1 small.c; a.out
0
$ 
$ clang-trunk -O1 small.c; a.out
1
$ icc -O1 small.c; a.out
1
$ 


---------------------------------------


int printf (const char *, ...);

struct S
{
  int f0;
} a = {1}, b, g, *c = &b, **f = &c;

int *d, **e = &d, h;

void __assert_fail ();

struct S
foo ()
{
  *e = &h;
  if (!d) 
    __assert_fail ();
  *f = &g;
  return a;
}

int
main ()
{
  struct S *i = c;
  *i = foo ();
  printf ("%d\n", b.f0);
  return 0;
}

[Bug tree-optimization/58956] [4.7/4.8/4.9 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[mpolacek at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-11-01
                 CC|                            |mpolacek at gcc dot gnu.org
            Summary|wrong code at -O1 and above |[4.7/4.8/4.9 Regression]
                   |(affecting gcc 4.6 to       |wrong code at -O1 and above
                   |trunk)                      |(affecting gcc 4.6 to
                   |                            |trunk)
     Ever confirmed|0                           |1

--- Comment #1 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Interesting.  Confirmed.

[Bug tree-optimization/58956] [4.7/4.8/4.9 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
   Target Milestone|---                         |4.7.4

[Bug tree-optimization/58956] [4.7/4.8/4.9 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[mikpelinux at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

Mikael Pettersson <mikpelinux at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikpelinux at gmail dot com

--- Comment #2 from Mikael Pettersson <mikpelinux at gmail dot com> ---
gcc 4.5.4 and 4.4.7 are also broken, but 4.3.6 is Ok.  Started with r142396.

The effect of that revision is that

{
  struct S *i = c;
  *i = foo ();

is compiled as if it had been written

{
  struct S tmp = foo ();
  *c = tmp;

which is wrong since foo may (and in this case does) change c.

[Bug middle-end/58956] [4.7/4.8/4.9 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2
             Status|NEW                         |ASSIGNED
          Component|tree-optimization           |middle-end
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Mine.

[Bug middle-end/58956] [4.7/4.8/4.9 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Well, on GIMPLE all is ok

  <bb 2>:
  i_2 = c;
  *i_2 = foo ();

but we expand it like

;; *i_2 = foo ();   --- note, TER applies!

(insn 5 4 6 (set (reg:QI 0 ax)
        (const_int 0 [0])) t.i:26 -1
     (nil))

(call_insn 6 5 7 (set (reg:SI 0 ax)
        (call (mem:QI (symbol_ref:DI ("foo") [flags 0x3]  <function_decl
0x7f2ae1c46200 foo>) [0 foo S1 A8])
            (const_int 0 [0]))) t.i:26 -1
     (expr_list:REG_EH_REGION (const_int 0 [0])
        (nil))
    (expr_list:REG_DEP_TRUE (use (reg:QI 0 ax))
        (nil)))

(insn 7 6 8 (set (reg:SI 90)
        (reg:SI 0 ax)) t.i:26 -1
     (nil))

(insn 8 7 9 (set (reg:SI 91)
        (reg:SI 90)) t.i:26 -1

(insn 9 8 10 (set (reg/f:DI 92)
        (mem/f/c:DI (symbol_ref:DI ("c") [flags 0x2]  <var_decl 0x7f2ae1c48260
c>) [0 c+0 S8 A64])) t.i:26 -1
     (nil))

(insn 10 9 0 (set (mem:SI (reg/f:DI 92) [0 *i_2+0 S4 A32])
        (reg:SI 91)) t.i:26 -1
     (nil))

So indeed expand sees wrong IL - *c = foo () here.  I have a patch.

[Bug middle-end/58956] [4.7/4.8/4.9 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Author: rguenth
Date: Tue Nov 19 13:28:35 2013
New Revision: 205026

URL: http://gcc.gnu.org/viewcvs?rev=205026&root=gcc&view=rev
Log:
2013-11-19  Richard Biener  <rguenther@suse.de>

    PR middle-end/58956
    * tree-ssa-ter.c (find_replaceable_in_bb): Avoid forwarding
    loads into stmts that may clobber it.

    * gcc.dg/torture/pr58956.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/torture/pr58956.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-ssa-ter.c

[Bug middle-end/58956] [4.7/4.8 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |4.9.0
            Summary|[4.7/4.8/4.9 Regression]    |[4.7/4.8 Regression] wrong
                   |wrong code at -O1 and above |code at -O1 and above
                   |(affecting gcc 4.6 to       |(affecting gcc 4.6 to
                   |trunk)                      |trunk)
      Known to fail|                            |4.8.2

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed on trunk sofar.

[Bug middle-end/58956] [4.7/4.8 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Author: rguenth
Date: Thu Dec  5 13:10:20 2013
New Revision: 205709

URL: http://gcc.gnu.org/viewcvs?rev=205709&root=gcc&view=rev
Log:
2013-12-05  Richard Biener  <rguenther@suse.de>

    Backport from mainline
    2013-11-19  Richard Biener  <rguenther@suse.de>

    PR middle-end/58956
    * tree-ssa-ter.c (find_replaceable_in_bb): Avoid forwarding
    loads into stmts that may clobber it.

    * gcc.dg/torture/pr58956.c: New testcase.

Added:
    branches/gcc-4_8-branch/gcc/testsuite/gcc.dg/torture/pr58956.c
Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_8-branch/gcc/tree-ssa-ter.c

[Bug middle-end/58956] [4.7 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Mon Dec 16 08:09:05 2013
New Revision: 206009

URL: http://gcc.gnu.org/viewcvs?rev=206009&root=gcc&view=rev
Log:
    PR middle-end/58956
    PR middle-end/59470
    * gimple-walk.h (walk_stmt_load_store_addr_fn): New typedef.
    (walk_stmt_load_store_addr_ops, walk_stmt_load_store_ops): Use it
    for callback params.
    * gimple-walk.c (walk_stmt_load_store_ops): Likewise.
    (walk_stmt_load_store_addr_ops): Likewise.  Adjust all callback
    calls to supply the gimple operand containing the base tree
    as an extra argument.
    * tree-ssa-ter.c: Include gimple-walk.h.
    (find_ssaname, find_ssaname_in_store): New helper functions.
    (find_replaceable_in_bb): For calls or GIMPLE_ASM, only set
    same_root_var if USE is used somewhere in the stores of the stmt.
    * ipa-prop.c (visit_ref_for_mod_analysis): Remove name of the stmt
    argument and ATTRIBUTE_UNUSED, add another unnamed tree argument.
    * ipa-pure-const.c (check_load, check_store, check_ipa_load,
    check_ipa_store): Likewise.
    * gimple.c (gimple_ior_addresses_taken_1, check_loadstore): Likewise.
    * ipa-split.c (test_nonssa_use, mark_nonssa_use): Likewise.
    (verify_non_ssa_vars, visit_bb): Adjust their callers.
    * cfgexpand.c (add_scope_conflicts_1): Use
    walk_stmt_load_store_addr_fn type for visit variable.
    (visit_op, visit_conflict): Remove name of the stmt
    argument and ATTRIBUTE_UNUSED, add another unnamed tree argument.
    * tree-sra.c (asm_visit_addr): Likewise.  Remove name of the data
    argument and ATTRIBUTE_UNUSED.
    * cgraphbuild.c (mark_address, mark_load, mark_store): Add another
    unnamed tree argument.
    * gimple-ssa-isolate-paths.c (check_loadstore): Likewise.  Remove
    ATTRIBUTE_UNUSED from stmt parameter.

    * gcc.target/i386/pr59470.c: New test.

Added:
    trunk/gcc/testsuite/gcc.target/i386/pr59470.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/cfgexpand.c
    trunk/gcc/cgraphbuild.c
    trunk/gcc/gimple-ssa-isolate-paths.c
    trunk/gcc/gimple-walk.c
    trunk/gcc/gimple-walk.h
    trunk/gcc/gimple.c
    trunk/gcc/ipa-prop.c
    trunk/gcc/ipa-pure-const.c
    trunk/gcc/ipa-split.c
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-sra.c
    trunk/gcc/tree-ssa-ter.c

[Bug middle-end/58956] [4.7 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Tue Jan  7 16:49:22 2014
New Revision: 206396

URL: http://gcc.gnu.org/viewcvs?rev=206396&root=gcc&view=rev
Log:
    Backported from mainline
    2013-12-16  Jakub Jelinek  <jakub@redhat.com>

    PR middle-end/58956
    PR middle-end/59470
    * gimple.h (walk_stmt_load_store_addr_fn): New typedef.
    (walk_stmt_load_store_addr_ops, walk_stmt_load_store_ops): Use it
    for callback params.
    * gimple.c (walk_stmt_load_store_ops): Likewise.
    (walk_stmt_load_store_addr_ops): Likewise.  Adjust all callback
    calls to supply the gimple operand containing the base tree
    as an extra argument.
    * tree-ssa-ter.c (find_ssaname, find_ssaname_in_store): New helper
    functions.
    (find_replaceable_in_bb): For calls or GIMPLE_ASM, only set
    same_root_var if USE is used somewhere in the stores of the stmt.
    * ipa-prop.c (visit_ref_for_mod_analysis): Remove name of the stmt
    argument and ATTRIBUTE_UNUSED, add another unnamed tree argument.
    * ipa-pure-const.c (check_load, check_store, check_ipa_load,
    check_ipa_store): Likewise.
    * gimple.c (gimple_ior_addresses_taken_1): Likewise.
    * ipa-split.c (test_nonssa_use, mark_nonssa_use): Likewise.
    (verify_non_ssa_vars, visit_bb): Adjust their callers.
    * cfgexpand.c (add_scope_conflicts_1): Use
    walk_stmt_load_store_addr_fn type for visit variable.
    (visit_op, visit_conflict): Remove name of the stmt
    argument and ATTRIBUTE_UNUSED, add another unnamed tree argument.
    * tree-sra.c (asm_visit_addr): Likewise.  Remove name of the data
    argument and ATTRIBUTE_UNUSED.
    * cgraphbuild.c (mark_address, mark_load, mark_store): Add another
    unnamed tree argument.

    * gcc.target/i386/pr59470.c: New test.

Added:
    branches/gcc-4_8-branch/gcc/testsuite/gcc.target/i386/pr59470.c
Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/cfgexpand.c
    branches/gcc-4_8-branch/gcc/cgraphbuild.c
    branches/gcc-4_8-branch/gcc/gimple.c
    branches/gcc-4_8-branch/gcc/gimple.h
    branches/gcc-4_8-branch/gcc/ipa-prop.c
    branches/gcc-4_8-branch/gcc/ipa-pure-const.c
    branches/gcc-4_8-branch/gcc/ipa-split.c
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_8-branch/gcc/tree-sra.c
    branches/gcc-4_8-branch/gcc/tree-ssa-ter.c

[Bug middle-end/58956] [4.7 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:

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

--- Comment #10 from Hans-Peter Nilsson <hp at gcc dot gnu.org> ---
Is this fixed but still open?  See also PR59584; the commit and backport just
above this comment caused it.

[Bug middle-end/58956] [4.7 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
It's still not fixed on the 4.7 branch.

[Bug middle-end/58956] [4.7 Regression] wrong code at -O1 and above (affecting gcc 4.6 to trunk)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58956

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED
   Target Milestone|4.7.4                       |4.8.3
      Known to fail|                            |4.7.4

--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed for 4.8.3.