[Bug tree-optimization/60026] New: ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[Bug tree-optimization/60026] New: ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[su at cs dot ucdavis.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

            Bug ID: 60026
           Summary: ICE at -O3 on valid code (with the optimize pragma) on
                    x86_64-linux-gnu
           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 following code using the optimize pragma causes an ICE when compiled with
the current gcc trunk and 4.8.x at -O3 on x86_64-linux-gnu (in both 32-bit and
64-bit modes). 

It is a regression from 4.7.x.

$ 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 --prefix=/usr/local/gcc-trunk
--enable-languages=c,c++ --disable-werror --enable-multilib
Thread model: posix
gcc version 4.9.0 20140131 (experimental) [trunk revision 207368] (GCC) 
$ 
$ gcc-trunk -O2 small.c; a.out
$ gcc-4.7.3 -O3 small.c; a.out
$ 
$ gcc-trunk -O3 small.c
small.c: In function ‘foo.constprop’:
small.c:8:10: internal compiler error: in parm_preserved_before_stmt_p, at
ipa-prop.c:760
 struct S foo (int p1, struct S p2)
          ^
0x84d137 parm_preserved_before_stmt_p
    ../../gcc-trunk/gcc/ipa-prop.c:760
0x84ecb4 ipa_load_from_parm_agg_1
    ../../gcc-trunk/gcc/ipa-prop.c:885
0x8531fa ipcp_transform_function(cgraph_node*)
    ../../gcc-trunk/gcc/ipa-prop.c:4776
0x915260 execute_one_ipa_transform_pass
    ../../gcc-trunk/gcc/passes.c:2066
0x915260 execute_all_ipa_transforms()
    ../../gcc-trunk/gcc/passes.c:2107
0x6a022a expand_function
    ../../gcc-trunk/gcc/cgraphunit.c:1767
0x6a2116 expand_all_functions
    ../../gcc-trunk/gcc/cgraphunit.c:1908
0x6a2116 compile()
    ../../gcc-trunk/gcc/cgraphunit.c:2252
0x6a2994 finalize_compilation_unit()
    ../../gcc-trunk/gcc/cgraphunit.c:2329
0x55c5e2 c_write_global_declarations()
    ../../gcc-trunk/gcc/c/c-decl.c:10401
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.
$ 


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


struct S
{
  int f;
} a;

#pragma GCC push_options
#pragma GCC optimize ("-O0")
struct S foo (int p1, struct S p2)
{
  int b = p2.f;
  return a; 
}
#pragma GCC pop_options

void
bar ()
{
  for (;a.f;)
    {
      struct S c = {0};
      foo (0, c);
    }
}

int
main ()
{
  bar(); 
  return 0;
}
>From gcc-bugs-return-442301-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Sun Feb 02 02:58:40 2014
Return-Path: <gcc-bugs-return-442301-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 22597 invoked by alias); 2 Feb 2014 02:58:39 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 22558 invoked by uid 48); 2 Feb 2014 02:58:33 -0000
From: "potswa at mac dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug libstdc++/57350] std::align missing
Date: Sun, 02 Feb 2014 02:58:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: libstdc++
X-Bugzilla-Version: 4.8.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: potswa at mac dot com
X-Bugzilla-Status: UNCONFIRMED
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-57350-4-la5jNUlIyz@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-57350-4@http.gcc.gnu.org/bugzilla/>
References: <bug-57350-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2014-02/txt/msg00058.txt.bz2
Content-length: 943

http://gcc.gnu.org/bugzilla/show_bug.cgi?idW350

--- Comment #4 from David Krauss <potswa at mac dot com> ---
Hmm, I recall preparing to submit a patch but not being able to decide which
header to modify.

Here's the aforementioned MIT-licensed code. The MIT license only requires
attribution which is satisfied by the changelog; anyway I don't care. I am
already a GNU contributor with FSF waiver back in 2009.

This code should probably be reviewed. I wrote it a long time ago and seldom
used it. Cannot recall whether it was intended to be 100% compliant.


inline void *align( std::size_t alignment, std::size_t size, void *&ptr,
std::size_t &space ) {
    auto pn = reinterpret_cast< std::size_t >( ptr );
    auto aligned = ( pn + alignment - 1 ) & - alignment;
    auto new_space = space - ( aligned - pn );
    if ( new_space < size ) return nullptr;
    space = new_space;
    return ptr = reinterpret_cast< void * >( aligned );
}

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |ice-checking
   Last reconfirmed|                            |2014-02-03
          Component|tree-optimization           |ipa
                 CC|                            |jamborm at gcc dot gnu.org
     Ever confirmed|0                           |1
            Summary|ICE at -O3 on valid code    |[4.8/4.9 Regression] ICE at
                   |(with the optimize pragma)  |-O3 on valid code (with the
                   |on x86_64-linux-gnu         |optimize pragma) on
                   |                            |x86_64-linux-gnu
   Target Milestone|---                         |4.8.3

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.

[Bug ipa/60026] [4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
   Target Milestone|4.8.3                       |4.9.0
            Summary|[4.8/4.9 Regression] ICE at |[4.9 Regression] ICE at -O3
                   |-O3 on valid code (with the |on valid code (with the
                   |optimize pragma) on         |optimize pragma) on
                   |x86_64-linux-gnu            |x86_64-linux-gnu

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Can't reproduce it on the 4.8 branch though.

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org
            Summary|[4.9 Regression] ICE at -O3 |[4.8/4.9 Regression] ICE at
                   |on valid code (with the     |-O3 on valid code (with the
                   |optimize pragma) on         |optimize pragma) on
                   |x86_64-linux-gnu            |x86_64-linux-gnu

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r193298.

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 32030
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32030&action=edit
gcc49-pr60026.patch

The problem is that for -O0 we don't create vdef/vuse at all, but I'd say we
shouldn't be cloning -O0 functions either, even when they don't have noclone
argument.  Thus perhaps something like attached patch?

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #4)
> Created attachment 32030 [details]
> gcc49-pr60026.patch
> 
> The problem is that for -O0 we don't create vdef/vuse at all, but I'd say we
> shouldn't be cloning -O0 functions either, even when they don't have noclone
> argument.  Thus perhaps something like attached patch?

While the idea of the patch is of course good, it seems that we only
set cannot_be_copied.* in copy_forbidden in tree-inline.c, so perhaps
it would be better to put the optimized test there as well so that we
never have any ordering issues?

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

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

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

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Martin Jambor from comment #5)
> (In reply to Jakub Jelinek from comment #4)
> > Created attachment 32030 [details]
> > gcc49-pr60026.patch
> > 
> > The problem is that for -O0 we don't create vdef/vuse at all, but I'd say we
> > shouldn't be cloning -O0 functions either, even when they don't have noclone
> > argument.  Thus perhaps something like attached patch?
> 
> While the idea of the patch is of course good, it seems that we only
> set cannot_be_copied.* in copy_forbidden in tree-inline.c, so perhaps
> it would be better to put the optimized test there as well so that we
> never have any ordering issues?

Not so easily, the problem is that copy_forbidden is called without fun being
actually cfun, and set_cfun is very expensive.  Do we ever call copy_forbidden
on functions that haven't been rewritten into SSA form yet?  If yes, it could
be even set e.g. during gimplification (if (!optimize)).

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #7 from Martin Jambor <jamborm at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #6)
> Not so easily, the problem is that copy_forbidden is called without fun
> being actually cfun, and set_cfun is very expensive. 

I think we can do the test without setting cfun by something similar
to ipa_func_spec_opts_forbid_analysis_p in ipa-prop.c.

> Do we ever call
> copy_forbidden on functions that haven't been rewritten into SSA form yet? 
> If yes, it could be even set e.g. during gimplification (if (!optimize)).

No, I don't think so.  It's really a matter of cleanness and a fairly
minor point at that.

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 32032
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32032&action=edit
gcc49-pr60026.patch

If you prefer say this, sure, it can be done instead (and fixes the testcase
too).

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #9 from Martin Jambor <jamborm at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #8)
> Created attachment 32032 [details]
> gcc49-pr60026.patch
> 
> If you prefer say this, sure, it can be done instead (and fixes the testcase
> too).

Yes, I do.  Thanks!

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
You could also check gimple_vop (fn) == NULL_TREE (not in virtual SSA form,
we probably should have a better way to query that).  But indeed guarding
this for -O0 is better.

[Bug ipa/60026] [4.8/4.9 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Tue Feb  4 13:04:37 2014
New Revision: 207463

URL: http://gcc.gnu.org/viewcvs?rev=207463&root=gcc&view=rev
Log:
    PR ipa/60026
    * tree-inline.c (copy_forbidden): Fail for
    __attribute__((optimize (0))) functions.

    * c-c++-common/torture/pr60026.c: New test.

Added:
    trunk/gcc/testsuite/c-c++-common/torture/pr60026.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-inline.c

[Bug ipa/60026] [4.8 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P1                          |P2
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org
   Target Milestone|4.9.0                       |4.8.3
            Summary|[4.8/4.9 Regression] ICE at |[4.8 Regression] ICE at -O3
                   |-O3 on valid code (with the |on valid code (with the
                   |optimize pragma) on         |optimize pragma) on
                   |x86_64-linux-gnu            |x86_64-linux-gnu

--- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed on the trunk so far.

[Bug ipa/60026] [4.8 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #13 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Fri Feb  7 23:54:17 2014
New Revision: 207621

URL: http://gcc.gnu.org/viewcvs?rev=207621&root=gcc&view=rev
Log:
    PR ipa/60026
    * ipa-cp.c (determine_versionability): Fail at -O0
    or __attribute__((optimize (0))) or -fno-ipa-cp functions.
    * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.

    Revert:
    2014-02-04  Jakub Jelinek  <jakub@redhat.com>

    PR ipa/60026
    * tree-inline.c (copy_forbidden): Fail for
    __attribute__((optimize (0))) functions.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/ipa-cp.c
    trunk/gcc/tree-inline.c
    trunk/gcc/tree-sra.c

[Bug ipa/60026] [4.8 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

--- Comment #14 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Thu Mar  6 07:57:12 2014
New Revision: 208364

URL: http://gcc.gnu.org/viewcvs?rev=208364&root=gcc&view=rev
Log:
    Backport from mainline
    2014-02-08  Jakub Jelinek  <jakub@redhat.com>

    PR ipa/60026
    * ipa-cp.c (determine_versionability): Fail at -O0
    or __attribute__((optimize (0))) or -fno-ipa-cp functions.
    * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.

    2014-02-06  Jakub Jelinek  <jakub@redhat.com>

    PR target/60062
    * tree.h (opts_for_fn): New inline function.
    (opt_for_fn): Define.
    * config/i386/i386.c (ix86_function_regparm): Use
    opt_for_fn (decl, optimize) instead of optimize.

    * gcc.c-torture/execute/pr60062.c: New test.
    * gcc.c-torture/execute/pr60072.c: New test.

    2014-02-04  Jakub Jelinek  <jakub@redhat.com>

    PR ipa/60026
    * c-c++-common/torture/pr60026.c: New test.

Added:
    branches/gcc-4_8-branch/gcc/testsuite/c-c++-common/torture/pr60026.c
    branches/gcc-4_8-branch/gcc/testsuite/gcc.c-torture/execute/pr60062.c
    branches/gcc-4_8-branch/gcc/testsuite/gcc.c-torture/execute/pr60072.c
Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/config/i386/i386.c
    branches/gcc-4_8-branch/gcc/ipa-cp.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.h

[Bug ipa/60026] [4.8 Regression] ICE at -O3 on valid code (with the optimize pragma) on x86_64-linux-gnu

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60026

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

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

--- Comment #15 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Hopefully fixed now.