[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[doko at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

--- Comment #1 from Matthias Klose <doko at gcc dot gnu.org> ---
seen when building the dnaclust package, also seen in the qtbase-opensource-src
package. currently reducing

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |5.0

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

ktkachov at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-04-08
                 CC|                            |ktkachov at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #2 from ktkachov at gcc dot gnu.org ---
Confirmed with r221914.
A gcc from 27Mar works though for me.

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[doko at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

--- Comment #3 from Matthias Klose <doko at gcc dot gnu.org> ---
Created attachment 35259
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35259&action=edit
reduced file

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

ktkachov at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |ktkachov at gcc dot gnu.org

--- Comment #4 from ktkachov at gcc dot gnu.org ---
This looks similar to PR 64600.
The problem seems to be arm_canonicalize_comparison that
canonicalizes a comparison with 2147483647 (0x7fffffff)
into a comparison with that +1, that is 0x80000000 without properly
sign-extending it into 0xffffffff80000000 (which would make it -2147483648).

I have a fix, but it needs more testing

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

--- Comment #5 from ktkachov at gcc dot gnu.org ---
(In reply to ktkachov from comment #4)
> This looks similar to PR 64600.
> The problem seems to be arm_canonicalize_comparison that
> canonicalizes a comparison with 2147483647 (0x7fffffff)
> into a comparison with that +1, that is 0x80000000 without properly
> sign-extending it into 0xffffffff80000000 (which would make it -2147483648).

That's not right logic, need to think a bit harder on what to do here...

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

--- Comment #6 from ktkachov at gcc dot gnu.org ---
(In reply to ktkachov from comment #5)
> (In reply to ktkachov from comment #4)
> > This looks similar to PR 64600.
> > The problem seems to be arm_canonicalize_comparison that
> > canonicalizes a comparison with 2147483647 (0x7fffffff)
> > into a comparison with that +1, that is 0x80000000 without properly
> > sign-extending it into 0xffffffff80000000 (which would make it -2147483648).
> 
> That's not right logic, need to think a bit harder on what to do here...

err... Ignore this. I think doing the sign-extend here is the right thing to
do.
All these numbers close to INT_MIN/MAX are very fiddly :(

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to ktkachov from comment #6)
> (In reply to ktkachov from comment #5)
> > (In reply to ktkachov from comment #4)
> > > This looks similar to PR 64600.
> > > The problem seems to be arm_canonicalize_comparison that
> > > canonicalizes a comparison with 2147483647 (0x7fffffff)
> > > into a comparison with that +1, that is 0x80000000 without properly
> > > sign-extending it into 0xffffffff80000000 (which would make it -2147483648).
> > 
> > That's not right logic, need to think a bit harder on what to do here...
> 
> err... Ignore this. I think doing the sign-extend here is the right thing to
> do.
> All these numbers close to INT_MIN/MAX are very fiddly :(

Any progress on this?  We'd really like to build RC1 soon.

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

--- Comment #8 from ktkachov at gcc dot gnu.org ---
(In reply to Jakub Jelinek from comment #7)
> Any progress on this?  We'd really like to build RC1 soon.

Was just doing some testing, patch posted at:
https://gcc.gnu.org/ml/gcc-patches/2015-04/msg00437.html

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

--- Comment #9 from ktkachov at gcc dot gnu.org ---
Author: ktkachov
Date: Fri Apr 10 16:20:35 2015
New Revision: 221976

URL: https://gcc.gnu.org/viewcvs?rev=221976&root=gcc&view=rev
Log:
[ARM] PR 65694: Properly sign-extend large numbers before passing to GEN_INT in
arm_canonicalize_comparison

    PR target/65694
    * config/arm/arm.c (arm_canonicalize_comparison): Use ARM_SIGN_EXTEND
    when creating +1 values for SImode.

    PR target/65694
    * g++.dg/torture/pr65694.C: New test.

Added:
    trunk/gcc/testsuite/g++.dg/torture/pr65694.C
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/arm/arm.c
    trunk/gcc/testsuite/ChangeLog

[Bug target/65694] [5 Regression] ICE (in decompose, at rtl.h:2007) on arm-linux-gnueabihf

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65694

ktkachov at gcc dot gnu.org changed:

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

--- Comment #10 from ktkachov at gcc dot gnu.org ---
Fixed on trunk.