[Bug c/50949] New: ICE on armhf with neon code

[Bug c/50949] New: ICE on armhf with neon code

[konstantinos.margaritis at linaro dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50949

             Bug #: 50949
           Summary: ICE on armhf with neon code
    Classification: Unclassified
           Product: gcc
           Version: 4.6.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: konstantinos.margaritis@linaro.org
              Host: arm-linux-gnueabihf
            Target: arm-linux-gnueabihf
             Build: arm-linux-gnueabihf


Created attachment 25680
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25680
minimal testcase produced using delta tool from source (propriatery driver
including NEON code)

(Note: this is a different ICE than #PR50948, but produced from the
same file as the previous, in fact it's produced from the same
function, albeit with a small modification).
When building neon code on armhf -which I cannot attach to its full,
due to licensing reason, unfortunately- I get an ICE. Using delta I
was able to get a minimal testcase (which is different than the one in
#PR50948), that produces the exact ICE. The testcase is attached. Here
is the message and the cmd line used:

$ gcc -O -mfpu=neon matrix2-min.i
matrix2-min.i: In function ‘matrixTranspose’:
matrix2-min.i:38:5: error: insn does not satisfy its constraints:
(insn 41 16 17 2 (set (mem/c/i:XI (pre_dec:SI (reg/f:SI 5 r5 [145]))
[0 S64 A64])
        (reg:XI 111 d24)) matrix2-min.i:21 753 {*neon_movxi}
     (expr_list:REG_INC (reg/f:SI 5 r5 [145])
        (nil)))
matrix2-min.i:38:5: internal compiler error: in
reload_cse_simplify_operands, at postreload.c:403
Please submit a full bug report,
with preprocessed source if appropriate.
See <file:///usr/share/doc/gcc-4.6/README.Bugs> for instructions.
Preprocessed source stored into /tmp/ccpGsI3S.out file, please attach
this to your bugreport.

This ICE is not reproducible on gcc-4.6 on oneiric/armel.

This bug has also been filed to the Debian BTS:

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=646163

Konstantinos

[Bug target/50949] ICE on armhf with neon code

[rearnsha at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50949

Richard Earnshaw <rearnsha at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2011-11-01
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Earnshaw <rearnsha at gcc dot gnu.org> 2011-11-01 20:44:42 UTC ---
arm-linux-gnueabihf isn't a configuration supported by FSF GCC; if this is
vendor distribution you need to contact your vendor.  Alternatively you need to
show how to produce this on the FSF sources.

[Bug target/50949] ICE on armhf with neon code

[konstantinos.margaritis at linaro dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50949

--- Comment #2 from Konstantinos Margaritis <konstantinos.margaritis at linaro dot org> 2011-11-01 21:05:42 UTC ---
This is the full cmd line used:

gcc -g -O -mfpu=neon -mfloat-abi=hard -march=armv7-a -mthumb -fpermissive -c
matrix.i

This is the actual gcc revision used in the Debian package: 20111028 (r180603).

[Bug target/50949] ICE on armhf with neon code

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=50949

ktkachov at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
                 CC|                            |ktkachov at gcc dot gnu.org
         Resolution|---                         |DUPLICATE

--- Comment #3 from ktkachov at gcc dot gnu.org ---
Dup.

*** This bug has been marked as a duplicate of bug 50948 ***