[Bug c/12662] New: ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[Bug c/12662] New: ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[lucier at math dot purdue dot edu]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12662

           Summary: ICE in cprop_into_phis, at tree-ssa-dom.c:1557
           Product: gcc
           Version: tree-ssa
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: lucier at math dot purdue dot edu
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: sparc-sun-solaris2.8
  GCC host triplet: sparc-sun-solaris2.8
GCC target triplet: sparc-sun-solaris2.8

with this compiler

banach-14% gcc -v
Reading specs from
/export/mdd0/users/lucier/local/gcc-ssa/bin/../lib/gcc/sparc-sun-solaris2.8/3.5-tree-ssa/specs
Configured with: ../configure --enable-languages=c
--prefix=/home/c/lucier/local/gcc-ssa
Thread model: posix
gcc version 3.5-tree-ssa 20031017 (merged 20031005)

on this source file
gcc -I../include -I. -Wall -W -Wno-unused -O1 -fno-math-errno -fschedule-insns2
-fno-trapping-math -fno-strict-aliasing -fomit-frame-pointer -fPIC -fno-common
-save-temps -DHAVE_CONFIG_H -D___PRIMAL -D___LIBRARY
-D___GAMBCDIR=\"/usr/local/\" -c _num.c
_num.c: In function `___H__20___num':

_num.c:5691: internal compiler error: in cprop_into_phis, at tree-ssa-dom.c:1557

http://www.math.purdue.edu/~lucier/bugzilla/2/_num.i.gz

we get this ICE

[Bug optimization/12662] [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12662


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c                           |optimization
           Keywords|                            |ice-on-valid-code
            Summary|ICE in cprop_into_phis, at  |[tree-ssa] ICE in
                   |tree-ssa-dom.c:1557         |cprop_into_phis, at tree-
                   |                            |ssa-dom.c:1557
   Target Milestone|---                         |tree-ssa

[Bug optimization/12662] [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[amacleod at redhat dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12662



------- Additional Comments From amacleod at redhat dot com  2003-10-17 17:21 -------

It looks like there is something left over from PRE in the phi nodes for a block:

This is the first PHI node for the block:

p print_generic_stmt (stderr, $6, 64)
 EUSE () [class:-1 phiop:1 bb:7685 ]


If you compile with -fno-tree-pre  the program finishes compilation.

Andrew

Re: [Bug optimization/12662] [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[Daniel Berlin]

On Oct 17, 2003, at 1:21 PM, amacleod at redhat dot com wrote:
>
>
> It looks like there is something left over from PRE in the phi nodes 
> for a block:
>
> This is the first PHI node for the block:
>
> p print_generic_stmt (stderr, $6, 64)
>  EUSE () [class:-1 phiop:1 bb:7685 ]
>


Err, PRE doesn't put anything in the phi_nodes chain, so this is 
certainly strange.
I'll see what's up.
:P

>
> If you compile with -fno-tree-pre  the program finishes compilation.
>
> Andrew

[Bug optimization/12662] [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[dberlin at dberlin dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12662



------- Additional Comments From dberlin at dberlin dot org  2003-10-17 17:28 -------
Subject: Re:  [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557


On Oct 17, 2003, at 1:21 PM, amacleod at redhat dot com wrote:
>
>
> It looks like there is something left over from PRE in the phi nodes 
> for a block:
>
> This is the first PHI node for the block:
>
> p print_generic_stmt (stderr, $6, 64)
>  EUSE () [class:-1 phiop:1 bb:7685 ]
>


Err, PRE doesn't put anything in the phi_nodes chain, so this is 
certainly strange.
I'll see what's up.
:P

>
> If you compile with -fno-tree-pre  the program finishes compilation.
>
> Andrew

[Bug optimization/12662] [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[dberlin at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12662



------- Additional Comments From dberlin at gcc dot gnu dot org  2003-10-17 20:25 -------
Well this is interesting:
<after PRE of the expression>
(gdb) p debug_generic_stmt (bb_ann (basic_block_info.data.bb[8157])->phi_nodes)
#   pretmp.1992_99947 = PHI <T.140_39112(8071), T.140_39231(8085)>;
pretmp.1992_99947 = PHI <T.140_39112(8071), T.140_39231(8085)>;

<after PRE is done>

(gdb) p debug_generic_stmt (bb_ann (basic_block_info.data.bb[8157])->phi_nodes)
 EUSE () [class:-1 phiop:1 bb:236 ]

[Bug optimization/12662] [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[dberlin at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12662


dberlin at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2003-10-17 23:29:36
               date|                            |


------- Additional Comments From dberlin at gcc dot gnu dot org  2003-10-17 23:29 -------
This is a Garbage Collection problem, because the BB annotations aren't marked, and thus, the phi 
nodes don't get marked. 

Because we push context in PRE (and never collect anywhere else in tree-ssa), only phi nodes PRE 
creates get collected.

Zdenek has patches to do proper marking, which should fix this.

[Bug optimization/12662] [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[lucier at math dot purdue dot edu]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12662



------- Additional Comments From lucier at math dot purdue dot edu  2003-10-19 19:51 -------
Subject: Re:  [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

As an aside, with -fno-tree-pre the compilation finishes, but the 
program is miscompiled somewhere.

[Bug optimization/12662] [tree-ssa] ICE in cprop_into_phis, at tree-ssa-dom.c:1557

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-04 09:04 -------
The orignal problem has been fixed now, can you submit a new bug for the miscompiling if it still 
happens.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12662