[Bug rtl-optimization/53487] New: [4.8 Regression] Unrecognizable insn for conditional move

[Bug rtl-optimization/53487] New: [4.8 Regression] Unrecognizable insn for conditional move

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53487

             Bug #: 53487
           Summary: [4.8 Regression] Unrecognizable insn for conditional
                    move
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: wschmidt@gcc.gnu.org
                CC: bergner@gcc.gnu.org, meissner@gcc.gnu.org,
                    pinskia@gcc.gnu.org
              Host: powerpc64-linux
            Target: powerpc64-linux
             Build: powerpc64-linux


Created attachment 27496
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27496
Reduced testcase

The attached test (reduced from 456.hmmer) fails to compile beginning with
r187183, as shown:

wschmidt@bns:~/delta> gcc -c -o cluster.o  -m64 -O3 -mcpu=power7 -ffast-math
-misel cluster.i
cluster.i: In function ‘Cluster’:
cluster.i:30:1: error: unrecognizable insn:
 }
 ^
(insn 156 155 157 8 (set (reg:SI 289)
        (if_then_else:SI (ge (reg:SF 291)
                (const_double:SF 0.0 [0x0.0p+0]))
            (subreg/s:SI (reg/v:DI 258 [ i+-4 ]) 4)
            (subreg/s:SI (reg/v:DI 265 [ row+-4 ]) 4))) -1
     (nil))
cluster.i:30:1: internal compiler error: in extract_insn, at recog.c:2130
 }
 ^
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

This form of conditional move insn isn't recognized on powerpc64 because the
floating-point comparison is incompatible with a conditional move of integer
arguments.  Prior to this revision, code generation recognized the
incompatibility and did not generate the conditional move.

r187183 is:

2012-05-04  Andrew Pinski  <apinski@cavium.com>

    * expr.c (get_def_for_expr_class): New function.
    (convert_tree_comp_to_rtx): New function.
    (expand_cond_expr_using_cmove): New function.
    (expand_expr_real_2 <case COND_EXPR>): Call
    expand_cond_expr_using_cmove first and return if it succeeds.
    Remove the check for HAVE_conditional_move since we should have
    already converted it to a conditional move.
    * config/i386/i386.c (ix86_expand_int_movcc): Disallow comparison
    modes of DImode for 32bits and TImode.

[Bug rtl-optimization/53487] [4.8 Regression] Unrecognizable insn for conditional move

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53487

William J. Schmidt <wschmidt at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code
   Target Milestone|---                         |4.8.0

[Bug rtl-optimization/53487] [4.8 Regression] Unrecognizable insn for conditional move

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53487

--- Comment #1 from William J. Schmidt <wschmidt at gcc dot gnu.org> 2012-05-25 16:09:43 UTC ---
Dumps show that the ICE occurs in the first pass following expand.

[Bug rtl-optimization/53487] [4.8 Regression] Unrecognizable insn for conditional move

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53487

William J. Schmidt <wschmidt at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #27496|0                           |1
        is obsolete|                            |

--- Comment #2 from William J. Schmidt <wschmidt at gcc dot gnu.org> 2012-05-25 16:13:29 UTC ---
Created attachment 27498
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27498
Reduced testcase in text/plain

Replaced attachment with one in text/plain format.

[Bug target/53487] [4.8 Regression] Unrecognizable insn for conditional move

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53487

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|rtl-optimization            |target

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-05-25 17:53:00 UTC ---
The target needs to reject the cases in movsicc it cannot handle the
comparisons.  The x86 and arm backends have been fixed to do exactly this.

[Bug target/53487] [4.8 Regression] Unrecognizable insn for conditional move

[meissner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53487

--- Comment #4 from Michael Meissner <meissner at gcc dot gnu.org> 2012-06-05 19:40:39 UTC ---
Author: meissner
Date: Tue Jun  5 19:40:34 2012
New Revision: 188248

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=188248
Log:
[gcc]
2012-06-04  Michael Meissner  <meissner@linux.vnet.ibm.com>

    PR target/53487
    * config/rs6000/rs6000.c (rs6000_generate_compare): If we are
    doing an unsigned compare, make sure the second argument is not a
    negative constant.
    (rs6000_emit_cmove): Don't allow floating point comparisons when
    generating ISEL moves.

[gcc/testsuite]
2012-06-04  Michael Meissner  <meissner@linux.vnet.ibm.com>

    * gcc.target/powerpc/pr53487.c: New test.


Added:
    trunk/gcc/testsuite/gcc.target/powerpc/pr53487.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/rs6000/rs6000.c
    trunk/gcc/testsuite/ChangeLog

[Bug target/53487] [4.8 Regression] Unrecognizable insn for conditional move

[meissner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53487

Michael Meissner <meissner at gcc dot gnu.org> changed:

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

--- Comment #5 from Michael Meissner <meissner at gcc dot gnu.org> 2012-08-16 22:59:35 UTC ---
Fixed on June 5, 2012.