[Bug target/47719] New: ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[Bug target/47719] New: ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[mans at mansr dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

           Summary: ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: mans@mansr.com


Created attachment 23326
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=23326
Preprocessed source

Using unpatched gcc from SVN built with portage (USE=vanilla):

$ arm-unknown-linux-gnueabi-gcc-4.6.0 --version
arm-unknown-linux-gnueabi-gcc-4.6.0 (Gentoo SVN) 4.6.0-pre9999 20110213
(experimental) rev. 170108

$ arm-unknown-linux-gnueabi-gcc-4.6.0 -I. -I"/home/mru/src/ffmpeg.dev"
-D_ISOC99_SOURCE -D_POSIX_C_SOURCE=200112 -D_FILE_OFFSET_BITS=64
-D_LARGEFILE_SOURCE -DHAVE_AV_CONFIG_H -mfloat-abi=softfp -mfpu=neon
-mcpu=cortex-a8 -std=c99 -fomit-frame-pointer -marm -pthread -g
-Wdeclaration-after-statement -Wall -Wno-parentheses -Wno-switch
-Wdisabled-optimization -Wpointer-arith -Wredundant-decls -Wno-pointer-sign
-Wcast-qual -Wwrite-strings -Wtype-limits -Wundef -Wmissing-prototypes -O3
-fno-math-errno -fno-signed-zeros -fno-tree-vectorize
-Werror=implicit-function-declaration -Werror=missing-prototypes          -MMD
-MF libavcodec/adxdec.d -MT libavcodec/adxdec.o -c -o libavcodec/adxdec.o
/home/mru/src/ffmpeg.dev/libavcodec/adxdec.c
/home/mru/src/ffmpeg.dev/libavcodec/adxdec.c: In function 'adx_decode_frame':
/home/mru/src/ffmpeg.dev/libavcodec/adxdec.c:167:1: internal compiler error: in
push_minipool_fix, at config/arm/arm.c:12084
Please submit a full bug report,
with preprocessed source if appropriate.

Preprocessed source attached.

Compilation succeeds if -fno-inline-functions is added to the flags.

[Bug target/47719] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[cltang at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

Chung-Lin Tang <cltang at gcc dot gnu.org> changed:

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

--- Comment #1 from Chung-Lin Tang <cltang at gcc dot gnu.org> 2011-02-17 08:29:38 UTC ---
This should be fixed by this patch, I encountered it when building ffmpeg too.
http://gcc.gnu.org/ml/gcc-patches/2011-02/msg00381.html

[Bug target/47719] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[mans at mansr dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

--- Comment #2 from Mans Rullgard <mans at mansr dot com> 2011-02-17 12:46:24 UTC ---
I can confirm this patch makes the file build without error.  I have not tested
the compiled code for correctness.

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[ramana at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

Ramana Radhakrishnan <ramana at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|arm                         |arm-none-linux-gnueabi
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |ice-on-valid-code
   Last reconfirmed|                            |2011.02.27 23:12:50
                 CC|                            |ramana at gcc dot gnu.org
     Ever Confirmed|0                           |1
            Summary|ICE compiling               |[4.6 regression] ICE
                   |libavcodec/adxdec.c         |compiling
                   |(FFmpeg) for ARM            |libavcodec/adxdec.c
                   |                            |(FFmpeg) for ARM
   Target Milestone|---                         |4.6.0

--- Comment #3 from Ramana Radhakrishnan <ramana at gcc dot gnu.org> 2011-02-27 23:12:50 UTC ---
Chung-Lin has a patch which is under review. This is a 4.6 regression compared
to all earlier versions.

Ramana

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-03-03 11:19:10 UTC ---
Marking as P1; ARM people, can this be resolved soon, so that it doesn't block
4.6 release for too long?
As the comments in postreload.c say, it might turn a movhi into
zero_extendhisi2 or
movqi into extendqisi2 for TARGET_THUMB, for arm_arch4 just movqi into
zero_extendqisi2 and for BYTES_BIG_ENDIAN movhi into extendhisi2.  Not sure for
which of the myriad of arm subtargets minipool is applicable though, or if
minipool is even applicable for movqi.  But if you emit some assembly insn
which in one case needs minipool attributes and from another zero/sign extend
insn
you emit exactly the same assembly, but don't have the minipool attributes on
it, it is a bug.

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[rearnsha at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

--- Comment #5 from Richard Earnshaw <rearnsha at gcc dot gnu.org> 2011-03-05 14:01:10 UTC ---
Author: rearnsha
Date: Sat Mar  5 14:01:07 2011
New Revision: 170698

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=170698
Log:
    PR target/47719
    * arm.md (movhi_insn_arch4):  Accept any immediate constant.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/arm/arm.md

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[rearnsha at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |rearnsha at gcc dot gnu.org
         Resolution|                            |FIXED

--- Comment #6 from Richard Earnshaw <rearnsha at gcc dot gnu.org> 2011-03-05 14:09:38 UTC ---
Fixed.

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[nitin.a.kamble at intel dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

--- Comment #7 from Nitin Kamble <nitin.a.kamble at intel dot com> 2011-05-25 22:42:44 UTC ---
Created attachment 24358
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24358
preprocessed output

This is the failing case.

$ ccache arm-poky-linux-gnueabi-gcc -march=armv7-a -mtune=cortex-a8 -mfpu=neon
-mfloat-abi=softfp -fno-tree-vectorize
--sysroot=/build_disk/poky_build/build0/tmp/sysroots/beagleboard -c -I.
-I../../../../src/gallium/include -I../../../../src/gallium/auxiliary
-I../../../../src/gallium/drivers
-I../../../../src/gallium/drivers/svga/include -O2 -pipe -g
-feliminate-unused-debug-types -Wall -Wmissing-prototypes -std=c99 -ffast-math
-fvisibility=hidden -fno-strict-aliasing -fPIC -D_GNU_SOURCE -DPTHREADS
-DHAVE_POSIX_MEMALIGN -DUSE_XSHM -std=gnu99 -fvisibility=hidden -DHAVE_STDINT_H
-DHAVE_SYS_TYPES_H svga_tgsi_insn.c -o svga_tgsi_insn.o
svga_tgsi_insn.c: In function 'svga_shader_emit_instructions':
svga_tgsi_insn.c:2969:1: internal compiler error: in push_minipool_fix, at
config/arm/arm.c:12084
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
make[4]: *** [svga_tgsi_insn.o] Error 1
make[4]: Leaving directory
`/build_disk/poky_build/build0/tmp/work/armv7a-poky-linux-gnueabi/mesa-xlib-1_7.10.2-r0/Mesa-7.10.2/src/gallium/drivers/svga'
make[3]: *** [default] Error 1
make[3]: Leaving directory
`/build_disk/poky_build/build0/tmp/work/armv7a-poky-linux-gnueabi/mesa-xlib-1_7.10.2-r0/Mesa-7.10.2/src/gallium/drivers'
make[2]: *** [default] Error 1
make[2]: Leaving directory
`/build_disk/poky_build/build0/tmp/work/armv7a-poky-linux-gnueabi/mesa-xlib-1_7.10.2-r0/Mesa-7.10.2/src/gallium'
make[1]: *** [subdirs] Error 1
make[1]: Leaving directory
`/build_disk/poky_build/build0/tmp/work/armv7a-poky-linux-gnueabi/mesa-xlib-1_7.10.2-r0/Mesa-7.10.2/src'
make: *** [default] Error 1
ERROR: oe_runmake failed

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[nitin.a.kamble at intel dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

Nitin Kamble <nitin.a.kamble at intel dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |nitin.a.kamble at intel dot
                   |                            |com

--- Comment #8 from Nitin Kamble <nitin.a.kamble at intel dot com> 2011-05-25 22:43:25 UTC ---
I am still seeing the failure with the fix applied to gcc 4.6.0

$ ccache arm-poky-linux-gnueabi-gcc -march=armv7-a -mtune=cortex-a8 -mfpu=neon
-mfloat-abi=softfp -fno-tree-vectorize
--sysroot=/build_disk/poky_build/build0/tmp/sysroots/beagleboard -c -I.
-I../../../../src/gallium/include -I../../../../src/gallium/auxiliary
-I../../../../src/gallium/drivers
-I../../../../src/gallium/drivers/svga/include -O2 -pipe -g
-feliminate-unused-debug-types -Wall -Wmissing-prototypes -std=c99 -ffast-math
-fvisibility=hidden -fno-strict-aliasing -fPIC -D_GNU_SOURCE -DPTHREADS
-DHAVE_POSIX_MEMALIGN -DUSE_XSHM -std=gnu99 -fvisibility=hidden -DHAVE_STDINT_H
-DHAVE_SYS_TYPES_H svga_tgsi_insn.c -o svga_tgsi_insn.o
svga_tgsi_insn.c: In function 'svga_shader_emit_instructions':
svga_tgsi_insn.c:2969:1: internal compiler error: in push_minipool_fix, at
config/arm/arm.c:12084
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
make[4]: *** [svga_tgsi_insn.o] Error 1
make[4]: Leaving directory
`/build_disk/poky_build/build0/tmp/work/armv7a-poky-linux-gnueabi/mesa-xlib-1_7.10.2-r0/Mesa-7.10.2/src/gallium/drivers/svga'
make[3]: *** [default] Error 1
make[3]: Leaving directory
`/build_disk/poky_build/build0/tmp/work/armv7a-poky-linux-gnueabi/mesa-xlib-1_7.10.2-r0/Mesa-7.10.2/src/gallium/drivers'
make[2]: *** [default] Error 1
make[2]: Leaving directory
`/build_disk/poky_build/build0/tmp/work/armv7a-poky-linux-gnueabi/mesa-xlib-1_7.10.2-r0/Mesa-7.10.2/src/gallium'
make[1]: *** [subdirs] Error 1
make[1]: Leaving directory
`/build_disk/poky_build/build0/tmp/work/armv7a-poky-linux-gnueabi/mesa-xlib-1_7.10.2-r0/Mesa-7.10.2/src'
make: *** [default] Error 1
ERROR: oe_runmake failed

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[nitin.a.kamble at intel dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

--- Comment #9 from Nitin Kamble <nitin.a.kamble at intel dot com> 2011-05-25 23:01:25 UTC ---
The ICE is on the gcc_assert line (config/arm/arm.c:12084) in the code bellow:

static void
push_minipool_fix (rtx insn, HOST_WIDE_INT address, rtx *loc,
                   enum machine_mode mode, rtx value)
{
  Mfix * fix = (Mfix *) obstack_alloc (&minipool_obstack, sizeof (* fix));

  fix->insn = insn;
  fix->address = address;
  fix->loc = loc;
  fix->mode = mode;
  fix->fix_size = MINIPOOL_FIX_SIZE (mode);
  fix->value = value;
  fix->forwards = get_attr_pool_range (insn);
  fix->backwards = get_attr_neg_pool_range (insn);
  fix->minipool = NULL;

  /* If an insn doesn't have a range defined for it, then it isn't
     expecting to be reworked by this code.  Better to stop now than
     to generate duff assembly code.  */
  gcc_assert (fix->forwards || fix->backwards);

  /* If an entry requires 8-byte alignment then assume all constant pools
     require 4 bytes of padding.  Trying to do this later on a per-pool
     basis is awkward because existing pool entries have to be modified.  */
  if (ARM_DOUBLEWORD_ALIGN && fix->fix_size >= 8)
    minipool_pad = 4;

  if (dump_file)

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[nitin.a.kamble at intel dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

--- Comment #10 from Nitin Kamble <nitin.a.kamble at intel dot com> 2011-05-25 23:05:33 UTC ---
Also adding  -fno-inline-functions parameter to gcc command line does not help.

[Bug target/47719] [4.6 regression] ICE compiling libavcodec/adxdec.c (FFmpeg) for ARM

[nitin.a.kamble at intel dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47719

--- Comment #11 from Nitin Kamble <nitin.a.kamble at intel dot com> 2011-05-26 16:26:45 UTC ---
I am not able to reopen this bug, can somebody reopen the bug?

Thanks,
Nitin