[Bug middle-end/49671] New: volatile goes missing after inlining

[Bug middle-end/49671] New: volatile goes missing after inlining

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49671

           Summary: volatile goes missing after inlining
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: pinskia@gcc.gnu.org


Take:
volatile int t;
static inline int cvmx_atomic_get32(volatile int *ptr)
{
    return *ptr;
}
void f(void)
{
  while (!cvmx_atomic_get32(&t))
    ;
}

Nothing fancy but after inlining, we have:
  D.2699_5 = t;
Which is wrong as t is volatile

[Bug middle-end/49671] [4.6/4.7 Regression] volatile goes missing after inlining

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49671

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
      Known to work|                            |4.5.1
   Target Milestone|---                         |4.6.2
            Summary|volatile goes missing after |[4.6/4.7 Regression]
                   |inlining                    |volatile goes missing after
                   |                            |inlining
      Known to fail|                            |4.6.1, 4.7.0

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-07-07 21:41:35 UTC ---
It was broken since at least "GCC: (GNU) 4.6.0 20110210 (experimental) [trunk
revision 170025]"

[Bug middle-end/49671] [4.6/4.7 Regression] volatile goes missing after inlining

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49671

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2011.07.07 22:48:48
         AssignedTo|unassigned at gcc dot       |pinskia at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-07-07 22:48:48 UTC ---
Note the following is also broken:
int t;
static inline int cvmx_atomic_get32(int *ptr)
{
    return *(volatile int*)ptr;
}
void f(void)
{
  while (!cvmx_atomic_get32(&t))
    ;
}
--- CUT ----
But that is caused by IPA-SRA.
I have a fix for both issues.

[Bug middle-end/49671] [4.6/4.7 Regression] volatile goes missing after inlining

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49671

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-07-07 23:07:02 UTC ---
Created attachment 24708
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24708
Fix which I am going to test

[Bug middle-end/49671] [4.6/4.7 Regression] volatile goes missing after inlining

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49671

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-07-08 10:30:32 UTC ---
Doesn't the outermost component also need TREE_THIS_VOLATILE?  So I suppose
sink the

  TREE_THIS_VOLATILE (*tp) = TREE_THIS_VOLATILE (old);

from the else branch as well.

[Bug middle-end/49671] [4.6/4.7 Regression] volatile goes missing after inlining

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49671

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-07-26 05:28:50 UTC ---
Author: pinskia
Date: Tue Jul 26 05:28:45 2011
New Revision: 176782

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=176782
Log:
2011-07-25  Andrew Pinski  <apinski@cavium.com>

    PR tree-opt/49671
    * tree-inline.c (remap_gimple_op_r): Copy TREE_THIS_VOLATILE and
    TREE_THIS_NOTRAP into the inner most MEM_REF.
    Always copy TREE_THIS_VOLATILE.
    * tree-sra.c (ptr_parm_has_direct_uses): Check that the lhs, rhs and
    arguments are not volatile references.

2011-07-25  Andrew Pinski  <apinski@cavium.com>

    PR tree-opt/49671
    * gcc.dg/tree-ssa/pr49671-1.c: New testcase.
    * gcc.dg/tree-ssa/pr49671-2.c: New testcase.


Added:
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr49671-1.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr49671-2.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-inline.c
    trunk/gcc/tree-sra.c

[Bug middle-end/49671] [4.6/4.7 Regression] volatile goes missing after inlining

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49671

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-07-26 07:52:27 UTC ---
Author: pinskia
Date: Tue Jul 26 07:52:24 2011
New Revision: 176785

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=176785
Log:
2011-07-25  Andrew Pinski  <apinski@cavium.com>

    PR tree-opt/49671
    * tree-inline.c (remap_gimple_op_r): Copy TREE_THIS_VOLATILE and
    TREE_THIS_NOTRAP into the inner most MEM_REF.
    Always copy TREE_THIS_VOLATILE.
    * tree-sra.c (ptr_parm_has_direct_uses): Check that the lhs, rhs and
    arguments are not volatile references.

2011-07-25  Andrew Pinski  <apinski@cavium.com>

    PR tree-opt/49671
    * gcc.dg/tree-ssa/pr49671-1.c: New testcase.
    * gcc.dg/tree-ssa/pr49671-2.c: New testcase.


Added:
    branches/gcc-4_6-branch/gcc/testsuite/gcc.dg/tree-ssa/pr49671-1.c
      - copied unchanged from r176782,
trunk/gcc/testsuite/gcc.dg/tree-ssa/pr49671-1.c
    branches/gcc-4_6-branch/gcc/testsuite/gcc.dg/tree-ssa/pr49671-2.c
      - copied unchanged from r176782,
trunk/gcc/testsuite/gcc.dg/tree-ssa/pr49671-2.c
Modified:
    branches/gcc-4_6-branch/   (props changed)
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/config/rs6000/rs6000.c   (props changed)
    branches/gcc-4_6-branch/gcc/config/rs6000/rs6000.h   (props changed)
    branches/gcc-4_6-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_6-branch/gcc/tree-inline.c
    branches/gcc-4_6-branch/gcc/tree-sra.c

Propchange: branches/gcc-4_6-branch/
            ('svn:mergeinfo' modified)

Propchange: branches/gcc-4_6-branch/gcc/config/rs6000/rs6000.c
            ('svn:mergeinfo' modified)

Propchange: branches/gcc-4_6-branch/gcc/config/rs6000/rs6000.h
            ('svn:mergeinfo' modified)

[Bug middle-end/49671] [4.6/4.7 Regression] volatile goes missing after inlining

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49671

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

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

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-07-26 07:53:25 UTC ---
Fixed.