[Bug c/34825] New: ICE (SIGSEGV) in compiling valid code with -funsafe-math-optimizations

[Bug c/34825] New: ICE (SIGSEGV) in compiling valid code with -funsafe-math-optimizations

[t-gcc-bugzilla at snowelm dot com]

gcc reports ICE when compiling SVMlight software package
(http://svmlight.joachims.org/) on x86_64 with -funsafe-math-optimizations.
The following is the reduced code from the svm_hideo.c in SVMlight 6.01.

Occurs on: gcc-4.2.2 on x86_64, gcc-4.3.0 on x86_64,  
           gcc-4.2.3 on i386, gcc-4.3.0 on i386
Doesn't occur on: gcc-4.1.3 on x86_64, gcc-4.1.3 on i386

$ cat >test.i <<__END__
double optimize_hildreth_despo(long n,long b1,long b2, double *g, double *g0,
double *ce, double *ce0, double *low, double *up, double *init)
{
        double t;

        if(ce[b1] != ce[b2]) {
                t = 0;
        }
        else if(((g[b1*n+b1]>0) || (g[b2*n+b2]>0))) {
               
t=((ce[b2]/ce[b1])*g0[b1]-g0[b2]+ce0[0]*(g[b1*n+b1]*ce[b2]/ce[b1]-g[b1*n+b2]/ce[b1]))/((ce[b2]*ce[b2]/(ce[b1]*ce[b1]))*g[b1*n+b1]+g[b2*n+b2]-2*(g[b1*n+b2]*ce[b2]/ce[b1]))-init[b2];
                if((up[b2]-init[b2]) < t) {
                        t=up[b2]-init[b2];
                }
                if((init[b2]-low[b2]) < -t) {
                        t=-(init[b2]-low[b2]);
                }
                if((up[b1]-init[b1]) < -t) {
                        t=-(up[b1]-init[b1]);
                }
                if((init[b1]-low[b1]) < t) {
                        t=init[b1]-low[b1];
                }
        }
        return t;
}
__END__
$ gcc-4.3 -funsafe-math-optimizations -O -c test.i
test.i: In function 'optimize_hildreth_despo':
test.i:2: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
$ gcc-4.3 -v
Using built-in specs.
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Debian 4.3-20080104-1'
--with-bugurl=file:///usr/share/doc/gcc-4.3/README.Bugs
--enable-languages=c,c++,fortran,objc,obj-c++,treelang --prefix=/usr
--enable-shared --with-system-zlib --libexecdir=/usr/lib
--without-included-gettext --enable-threads=posix --enable-nls
--with-gxx-include-dir=/usr/include/c++/4.3 --program-suffix=-4.3
--enable-clocale=gnu --enable-libstdcxx-debug --enable-mpfr
--enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu
--target=x86_64-linux-gnu
Thread model: posix
gcc version 4.3.0 20080104 (experimental) [trunk revision 131316] (Debian
4.3-20080104-1) 
$ uname -a
Linux rainbow 2.6.22-3-amd64 #1 SMP Thu Oct 11 15:23:23 UTC 2007 x86_64
GNU/Linux


-- 
           Summary: ICE (SIGSEGV) in compiling valid code with -funsafe-
                    math-optimizations
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: t-gcc-bugzilla at snowelm dot com
 GCC build triplet: x86_64-linux-gnu
  GCC host triplet: x86_64-linux-gnu
GCC target triplet: x86_64-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2/4.3 regression] ICE with -funsafe-math-optimizations

[reichelt at gcc dot gnu dot org]

------- Comment #1 from reichelt at gcc dot gnu dot org  2008-01-17 09:32 -------
Confirmed.
Reduced testcase:

==================================
double foo(double x, double y)
{
  return x == y ? x/y*x/y : 0;
}
==================================


-- 

reichelt at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |reichelt at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
          Component|middle-end                  |tree-optimization
     Ever Confirmed|0                           |1
           Keywords|                            |ice-on-valid-code
   Last reconfirmed|0000-00-00 00:00:00         |2008-01-17 09:32:48
               date|                            |
            Summary|ICE (SIGSEGV) in compiling  |[4.2/4.3 regression] ICE
                   |valid code with -funsafe-   |with -funsafe-math-
                   |math-optimizations          |optimizations
   Target Milestone|---                         |4.2.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2/4.3 regression] ICE with -funsafe-math-optimizations

[ubizjak at gmail dot com]

------- Comment #2 from ubizjak at gmail dot com  2008-01-17 10:09 -------
(gdb) set args -O -funsafe-math-optimizations pr34825.c
(gdb) break fancy_abort
Breakpoint 1 at 0x81697e0: file ../../gcc-svn/trunk/gcc/diagnostic.c, line 659.
(gdb) run
Starting program: /home/uros/gcc-build/gcc/cc1 -O -funsafe-math-optimizations
pr34825.c
 foo
Analyzing compilation unit
Performing interprocedural optimizations
 <visibility> <early_local_cleanups> <inline> <static-var>
<pure-const>Assembling functions:
 foo
Program received signal SIGSEGV, Segmentation fault.
execute_cse_reciprocals_1 (def_bsi=0x0, def=0xb7cf4ed4)
    at ../../gcc-svn/trunk/gcc/tree-flow-inline.h:1534
1534          imm->next_imm_name = imm->imm_use->next;
(gdb) bt
#0  execute_cse_reciprocals_1 (def_bsi=0x0, def=0xb7cf4ed4)
    at ../../gcc-svn/trunk/gcc/tree-flow-inline.h:1534
#1  0x0845bf0a in execute_cse_reciprocals () at
../../gcc-svn/trunk/gcc/tree-ssa-math-opts.c:473
#2  0x082c50aa in execute_one_pass (pass=0x8903440) at
../../gcc-svn/trunk/gcc/passes.c:1118
#3  0x082c529f in execute_pass_list (pass=0x8903440) at
../../gcc-svn/trunk/gcc/passes.c:1171
#4  0x082c52b2 in execute_pass_list (pass=0x8902840) at
../../gcc-svn/trunk/gcc/passes.c:1172
#5  0x083b173b in tree_rest_of_compilation (fndecl=0xb7cef9a0)
    at ../../gcc-svn/trunk/gcc/tree-optimize.c:404
#6  0x085790a5 in cgraph_expand_function (node=0xb7c8c280)
    at ../../gcc-svn/trunk/gcc/cgraphunit.c:1152
#7  0x0857bda7 in cgraph_optimize () at
../../gcc-svn/trunk/gcc/cgraphunit.c:1215
#8  0x0805ecd6 in c_write_global_declarations () at
../../gcc-svn/trunk/gcc/c-decl.c:8077
#9  0x0834ba83 in toplev_main (argc=4, argv=0xbfb9b594) at
../../gcc-svn/trunk/gcc/toplev.c:1055
#10 0x080e1aef in main (argc=0, argv=0x19) at ../../gcc-svn/trunk/gcc/main.c:35


-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bonzini at gnu dot org
  GCC build triplet|x86_64-linux-gnu            |x86-linux-gnu
   GCC host triplet|x86_64-linux-gnu            |x86-linux-gnu
 GCC target triplet|x86_64-linux-gnu            |x86-linux-gnu
   Last reconfirmed|2008-01-17 09:32:48         |2008-01-17 10:09:08
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2/4.3 regression] ICE with -funsafe-math-optimizations

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2008-01-17 10:15 -------
Err yeah, this pass doesn't keep immediate uses (and SSA form) up-to-date.

Should be easy to fix.


-- 

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|NEW                         |ASSIGNED
   Last reconfirmed|2008-01-17 10:09:08         |2008-01-17 10:15:50
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2/4.3 regression] ICE with -funsafe-math-optimizations

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2008-01-17 10:45 -------
So, we enter the recip pass with

<bb 2>:
  if (x_2(D) == y_3(D))
    goto <bb 3>;
  else
    goto <bb 4>;

<bb 3>:
  D.1547_4 = x_2(D) / x_2(D);
  D.1548_5 = D.1547_4 * x_2(D);
  D.1545_6 = D.1548_5 / x_2(D);

<bb 4>:
  # D.1545_1 = PHI <D.1545_6(3), 0.0(2)>
  return D.1545_1;

that is, we compute x/x*x/x.  Recip now 'optimizes' this to

  reciptmp.6_10 = 1.0e+0 / x_2(D);
  D.1547_4 = __builtin_pow (reciptmp.6_10, 2.0e+0);
  D.1548_5 = D.1547_4 * x_2(D);
  D.1545_6 = D.1548_5 * reciptmp.6_10;

which is bogus anyway.  But the ICE is gone :P

Well, it registers the division x_2(D)/x_2(D) twice, which also makes
it a wrong-code regression.

The easiest thing is probably to just not do reciprocal optimization
for x / x.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
           Priority|P3                          |P1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2/4.3 regression] ICE with -funsafe-math-optimizations

[ubizjak at gmail dot com]

------- Comment #5 from ubizjak at gmail dot com  2008-01-17 11:09 -------
(In reply to comment #4)

> The easiest thing is probably to just not do reciprocal optimization
> for x / x.

This will be later converted to 1.0 anyway.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2/4.3 regression] ICE with -funsafe-math-optimizations

[rguenther at suse dot de]

------- Comment #6 from rguenther at suse dot de  2008-01-17 12:13 -------
Subject: Re:  [4.2/4.3 regression] ICE with
 -funsafe-math-optimizations

On Thu, 17 Jan 2008, ubizjak at gmail dot com wrote:

> ------- Comment #5 from ubizjak at gmail dot com  2008-01-17 11:09 -------
> (In reply to comment #4)
> 
> > The easiest thing is probably to just not do reciprocal optimization
> > for x / x.
> 
> This will be later converted to 1.0 anyway.

Only for -ffinite-math-only and -fno-trapping-math.

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2/4.3 regression] ICE with -funsafe-math-optimizations

[rguenth at gcc dot gnu dot org]

------- Comment #7 from rguenth at gcc dot gnu dot org  2008-01-17 12:18 -------
Subject: Bug 34825

Author: rguenth
Date: Thu Jan 17 12:17:24 2008
New Revision: 131595

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=131595
Log:
2008-01-17  Richard Guenther  <rguenther@suse.de>

        PR tree-optimization/34825
        * tree-ssa-math-opts.c (is_division_by): Do not recognize
        x / x as division to handle.

        * gcc.dg/pr34825.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/pr34825.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-ssa-math-opts.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2 regression] ICE with -funsafe-math-optimizations

[rguenth at gcc dot gnu dot org]

------- Comment #8 from rguenth at gcc dot gnu dot org  2008-01-17 12:29 -------
Fixed on the trunk.  Unassigning.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|rguenth at gcc dot gnu dot  |unassigned at gcc dot gnu
                   |org                         |dot org
             Status|ASSIGNED                    |NEW
      Known to fail|                            |4.2.3
      Known to work|                            |4.3.0
            Summary|[4.2/4.3 regression] ICE    |[4.2 regression] ICE with -
                   |with -funsafe-math-         |funsafe-math-optimizations
                   |optimizations               |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2 regression] ICE with -funsafe-math-optimizations

[jsm28 at gcc dot gnu dot org]

------- Comment #9 from jsm28 at gcc dot gnu dot org  2008-02-01 16:55 -------
4.2.3 is being released now, changing milestones of open bugs to 4.2.4.


-- 

jsm28 at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.2.3                       |4.2.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2 regression] ICE with -funsafe-math-optimizations

[rguenth at gcc dot gnu dot org]

------- Comment #10 from rguenth at gcc dot gnu dot org  2008-02-05 12:31 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825

[Bug tree-optimization/34825] [4.2 regression] ICE with -funsafe-math-optimizations

[rguenth at gcc dot gnu dot org]

------- Comment #11 from rguenth at gcc dot gnu dot org  2008-02-05 12:32 -------
Subject: Bug 34825

Author: rguenth
Date: Tue Feb  5 12:31:50 2008
New Revision: 132119

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=132119
Log:
2008-05-02  Richard Guenther  <rguenther@suse.de>

        Backport from mainline:
        2008-01-17  Richard Guenther  <rguenther@suse.de>

        PR tree-optimization/34825
        * tree-ssa-math-opts.c (is_division_by): Do not recognize
        x / x as division to handle.

        * gcc.dg/pr34825.c: New testcase.

Added:
    branches/gcc-4_2-branch/gcc/testsuite/gcc.dg/pr34825.c
      - copied unchanged from r131595, trunk/gcc/testsuite/gcc.dg/pr34825.c
Modified:
    branches/gcc-4_2-branch/gcc/ChangeLog
    branches/gcc-4_2-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_2-branch/gcc/tree-ssa-math-opts.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34825