[Bug middle-end/50561] New: [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[Bug middle-end/50561] New: [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[matt at use dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

             Bug #: 50561
           Summary: [4.7 regression] ICE when compiling zlib with -O2
                    -floop-flatten -floop-strip-mine
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: matt@use.net


getting this with trunk r179143, on Ubuntu 11.10/x86-64:

/usr/lib/gcc-snapshot/bin/x86_64-linux-gnu-gcc  -O2 -floop-flatten
-floop-strip-mine -c ~/src/trees.i trees.c: In function 'init_block':
trees.c:414:13: internal compiler error: in psct_dynamic_dim, at
graphite-poly.h:659

It doesn't happen with -O1, or if you remove either of the loop options. This
doesn't happen with 4.6.1.

Pre-processed source is attached, but it quick to reproduce with zlib 1.2.3.4
sources.

[Bug middle-end/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[matt at use dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

--- Comment #1 from Matt Hargett <matt at use dot net> 2011-09-28 20:59:07 UTC ---
Created attachment 25378
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25378
pre-processed source of the file that triggers the ICE

[Bug middle-end/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[matt at use dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

Matt Hargett <matt at use dot net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |major

[Bug middle-end/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.0

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011-10-27
                 CC|                            |spop at gcc dot gnu.org
          Component|middle-end                  |tree-optimization
     Ever Confirmed|0                           |1
           Severity|major                       |normal

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-10-27 10:32:06 UTC ---
Confirmed.

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[matt at use dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

--- Comment #3 from Matt Hargett <matt at use dot net> 2011-12-10 01:38:00 UTC ---
Just re-verified that this is still a problem with trunk as of 2011-12-09.

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-01-09 14:41:04 UTC ---
Re-confirmed.  Reduced testcase:

void f (unsigned *s)
{
  int n;
  for (n = 0; n < 256; n++)
    s[n] = 0;
}

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #5 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-01-09 15:28:01 UTC ---
The asserts in

static inline ppl_dimension_type
psct_dynamic_dim (poly_bb_p pbb, graphite_dim_t level)
{
  graphite_dim_t result = 1 + 2 * level;

  gcc_assert (result < pbb_nb_scattering_transform (pbb));
  return result;
}

static inline ppl_dimension_type
psct_static_dim (poly_bb_p pbb, graphite_dim_t level)
{
  graphite_dim_t result = 2 * level;

  gcc_assert (result < pbb_nb_scattering_transform (pbb));
  return result;
}

are not consistent in case both a dynamic and static dimension should
always exist at the same time.  Changing psct_dynamic_dim to use <=
fixes the issue.

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW
                 CC|                            |grosser at gcc dot gnu.org
         AssignedTo|rguenth at gcc dot gnu.org  |unassigned at gcc dot
                   |                            |gnu.org

--- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-02-08 13:43:24 UTC ---
Not mine, my fix was bogus.  Tobias?

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[grosser at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

--- Comment #7 from Tobias Grosser <grosser at gcc dot gnu.org> 2012-02-08 15:38:19 UTC ---
Created attachment 26614
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26614
A possible fix

Here a possible fix. The commit message is:

    Removing old scattering dimensions after loop flattening is incorrect.

    As long as we use PPL we are not able to remove the old scattering
dimensions.
    The reason is that these dimensions are not entirely unused.  They are not
    necessary as part of the scheduling vector, as the earlier dimensions
already
    unambiguously define the execution time, however they may still be needed
to
    carry modulo constraints as introduced e.g. by strip mining.  The correct
    solution would be to project these dimensions out of the scattering
polyhedra.
    In case they are still required to carry modulo constraints they should be
kept
    internally as existentially quantified dimensions.  PPL does only support
    projection of rational polyhedra, however in this case we need an integer
    projection. With isl this will be trivial to implement.  For now we just
leave
    the dimensions. This is a little ugly, but should be correct.

Could you verify it it works for you?

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

--- Comment #8 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-02-08 16:02:18 UTC ---
Sure, I will check.

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

--- Comment #9 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2012-02-08 17:02:16 UTC ---
I saw a similar ICE for the polyhedron test air.f90. The patch in comment #7
fixes the ICE. However it produces a wrong code for -O2 and above, unless I use
-fwhole-program.

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

--- Comment #10 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-02-15 12:20:23 UTC ---
Author: rguenth
Date: Wed Feb 15 12:20:10 2012
New Revision: 184265

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=184265
Log:
2012-02-15  Tobias Grosser <grosser@fim.uni-passau.de>

    PR tree-optimization/50561
    * graphite-flattening.c (lst_project_loop): Do not
    remove old scattering dimensions after flattening.
    (lst_do_flatten): Likewise.

    * gcc.dg/graphite/pr50561.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/graphite/pr50561.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/graphite-flattening.c
    trunk/gcc/testsuite/ChangeLog

[Bug tree-optimization/50561] [4.7 regression] ICE when compiling zlib with -O2 -floop-flatten -floop-strip-mine

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50561

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

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

--- Comment #11 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-02-15 12:21:18 UTC ---
Fixed.