[Bug lto/40903] New: LTO doesn't merge common sections properly

[Bug lto/40903] New: LTO doesn't merge common sections properly

[rguenth at gcc dot gnu dot org]

t1.c
double i;

t2.c
int i;
int main () { return i; }

should be accepted with -fcommon.

What needs to be done is that we have to keep a list of decls per symtab
entry that we might end up merging to and we need to return the proper
decl when looking up the canonical decl.

A warning is still appropriate here.

This happens in several SPEC 2000 and SPEC 2006 benchmarks.


-- 
           Summary: LTO doesn't merge common sections properly
           Product: gcc
           Version: lto
            Status: UNCONFIRMED
          Keywords: rejects-valid
          Severity: normal
          Priority: P3
         Component: lto
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: rguenth at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2009-07-29 13:39 -------
I'm working on this.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-07-29 13:39:41
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rth at gcc dot gnu dot org]

------- Comment #2 from rth at gcc dot gnu dot org  2009-07-29 17:55 -------
I believe a "proper" way to handle this, preserving the semantics that
the linker provides, is to transform this into

union {
  double _1;
  int _2;
} i;

and replace occurrences with i._1, i._2, etc.

Perhaps a more interesting example to look at, besides the arguable
coding errors in SPEC, is the usages

core_cia.c:} saved_config __attribute((common));
core_t2.c:} t2_saved_config __attribute((common));
core_titan.c:} saved_config[4] __attribute__((common));
core_tsunami.c:} saved_config[2] __attribute__((common));
sys_sio.c:} saved_config __attribute((common));

in linux/arch/alpha/kernel/.  Each usage is for a different hardware
type, and therefore mutually exclusive.

I believe there are Fortran common blocks that expect union-like
semantics as well.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rguenther at suse dot de]

------- Comment #3 from rguenther at suse dot de  2009-07-29 18:01 -------
Subject: Re:  LTO doesn't merge common sections properly

On Wed, 29 Jul 2009, rth at gcc dot gnu dot org wrote:

> ------- Comment #2 from rth at gcc dot gnu dot org  2009-07-29 17:55 -------
> I believe a "proper" way to handle this, preserving the semantics that
> the linker provides, is to transform this into
> 
> union {
>   double _1;
>   int _2;
> } i;
> 
> and replace occurrences with i._1, i._2, etc.

That's another possibility.  Though it would be only necessary if
you want to support punning between the entries.

Just keeping the decls and uses unmerged will get the exact same
effects as with non-LTO.  Apart from that if there _are_ overlapping
uses those would be disambiguated by the alias-oracle and you'd
eventually get different behavior than from non-LTO mode (though
arguably the behavior is just undefined in that case).

Keeping the decls unmerged is certainly the less invasive approach.

> Perhaps a more interesting example to look at, besides the arguable
> coding errors in SPEC, is the usages
> 
> core_cia.c:} saved_config __attribute((common));
> core_t2.c:} t2_saved_config __attribute((common));
> core_titan.c:} saved_config[4] __attribute__((common));
> core_tsunami.c:} saved_config[2] __attribute__((common));
> sys_sio.c:} saved_config __attribute((common));
> 
> in linux/arch/alpha/kernel/.  Each usage is for a different hardware
> type, and therefore mutually exclusive.

That should work indeed.

> I believe there are Fortran common blocks that expect union-like
> semantics as well.

Yes, they are marked as DECL_COMMON as well.

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rth at gcc dot gnu dot org]

------- Comment #4 from rth at gcc dot gnu dot org  2009-07-29 18:10 -------
So LTO still produces N output object files for N input files?

Cause you can't just output

  .comm i,4,4
  .comm i,8,8

in one object file, and I didn't see any changes to varasm.c...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rguenther at suse dot de]

------- Comment #5 from rguenther at suse dot de  2009-07-29 18:18 -------
Subject: Re:  LTO doesn't merge common sections properly

On Wed, 29 Jul 2009, rth at gcc dot gnu dot org wrote:

> ------- Comment #4 from rth at gcc dot gnu dot org  2009-07-29 18:10 -------
> So LTO still produces N output object files for N input files?
> 
> Cause you can't just output
> 
>   .comm i,4,4
>   .comm i,8,8
> 
> in one object file, and I didn't see any changes to varasm.c...

Hm, it seems to output just the first decl, so depending on the
order of files on the commandline I either get 

.comm i,4,4

or

.comm i,8,8

I have to check where that happens.

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rguenth at gcc dot gnu dot org]

------- Comment #6 from rguenth at gcc dot gnu dot org  2009-07-29 19:20 -------
Ah, it's because I don't push to lto_global_var_decls.  We can at
write-global-declarations time walk over all global vars and exchange the
written decl for
the largest in the chain.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rguenth at gcc dot gnu dot org]

------- Comment #7 from rguenth at gcc dot gnu dot org  2009-07-30 16:24 -------
Subject: Bug 40903

Author: rguenth
Date: Thu Jul 30 16:24:05 2009
New Revision: 150262

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=150262
Log:
2009-07-30  Richard Guenther  <rguenther@suse.de>

        PR lto/40903
        * lto-symtab.c (lto_symtab_compatible): Only warn for mismatched
        types.
        (lto_symtab_merge_decl): For decls we cannot merge chain them
        appropriately in the symtab entry.
        (lto_symtab_prevailing_decl): Return a matching entry from the
        symtab chain.
        * lto.c (read_cgraph_and_symbols): After fixing up decls choose
        the largest decl for output and free TREE_CHAIN for further
        use.

        * gcc.dg/lto/20090729_0.c: New testcase.
        * gcc.dg/lto/20090729_1.c: Likewise.

Added:
    branches/lto/gcc/testsuite/gcc.dg/lto/20090729_0.c
    branches/lto/gcc/testsuite/gcc.dg/lto/20090729_1.c
Modified:
    branches/lto/gcc/ChangeLog.lto
    branches/lto/gcc/lto-symtab.c
    branches/lto/gcc/lto/ChangeLog
    branches/lto/gcc/lto/lto.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rguenth at gcc dot gnu dot org]

------- Comment #9 from rguenth at gcc dot gnu dot org  2009-07-30 16:25 -------
,


-- 

rguenth at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rguenth at gcc dot gnu dot org]

------- Comment #8 from rguenth at gcc dot gnu dot org  2009-07-30 16:24 -------
Fixed.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903

[Bug lto/40903] LTO doesn't merge common sections properly

[rguenth at gcc dot gnu dot org]

------- Comment #10 from rguenth at gcc dot gnu dot org  2009-08-03 08:46 -------
Subject: Bug 40903

Author: rguenth
Date: Mon Aug  3 08:45:37 2009
New Revision: 150365

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=150365
Log:
2009-07-30  Richard Guenther  <rguenther@suse.de>

        PR lto/40903
        * gcc.dg/lto/20090729_0.c: New testcase.
        * gcc.dg/lto/20090729_1.c: Likewise.

Modified:
    branches/lto/gcc/testsuite/ChangeLog.lto


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40903