[Bug target/58158] New: internal compiler error: in extract_insn, at recog.c:2150 while compiling ImageMagick on mipsel

[Bug target/58158] New: internal compiler error: in extract_insn, at recog.c:2150 while compiling ImageMagick on mipsel

[aaro.koskinen at iki dot fi]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

            Bug ID: 58158
           Summary: internal compiler error: in extract_insn, at
                    recog.c:2150 while compiling ImageMagick on mipsel
           Product: gcc
           Version: 4.8.1
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: aaro.koskinen at iki dot fi

Created attachment 30657
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30657&action=edit
Source files (original & pre-processed).

Compiler details:

Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/mipsel-linux-gnu/4.8.1/lto-wrapper
Target: mipsel-linux-gnu
Configured with: /home/aaro/los/work/shared/gcc-4.8.1/configure --build=none
--host=mipsel-linux-gnu --target=mipsel-linux-gnu --prefix=/usr --disable-nls
--disable-bootstrap --enable-languages=c,c++ --with-system-zlib --enable-shared
--disable-static --with-arch=loongson2f --with-abi=32 --enable-targets=all
Thread model: posix
gcc version 4.8.1 (GCC)

Steps to reproduce:

$ gcc -O2 mpeg.i
coders/mpeg.c: In function 'WriteMPEGImage':
coders/mpeg.c:625:1: error: unrecognizable insn:
 }
 ^
(insn 223 222 224 26 (set (reg:SI 351 [ D.11743 ])
        (if_then_else:SI (ne:CC (reg:CC 67 $fcc0)
                (const_int 0 [0]))
            (reg:SI 351 [ D.11743 ])
            (reg:SI 354))) -1
     (nil))
coders/mpeg.c:625:1: internal compiler error: in extract_insn, at recog.c:2150
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

Source file is attached.

[Bug target/58158] internal compiler error: in extract_insn, at recog.c:2150 while compiling ImageMagick on mipsel

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

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

[Bug target/58158] internal compiler error: in extract_insn, at recog.c:2150 while compiling ImageMagick on mipsel

[aaro.koskinen at iki dot fi]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #1 from Aaro Koskinen <aaro.koskinen at iki dot fi> ---
A simplified reproducer:

$ cat bar.c
extern void foo(void);

static inline double max(const double x, const double y)
{
        return (x > y) ? x : y;
}

void bar(int d)
{
        int i;

        for (i = 0; i < (int)max(100.0 * d, 1.0); i++) {
                foo();
        }
}
$ gcc -c -O2 bar.c
bar.c: In function 'bar':
bar.c:15:1: error: unrecognizable insn:
 }
 ^
(insn 23 22 24 2 (set (reg:SI 203 [ D.1511 ])
        (if_then_else:SI (ne:CC (reg:CC 67 $fcc0)
                (const_int 0 [0]))
            (reg:SI 203 [ D.1511 ])
            (reg:SI 206))) -1
     (nil))
bar.c:15:1: internal compiler error: in extract_insn, at recog.c:2150
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

[Bug target/58158] internal compiler error: in extract_insn, at recog.c:2150 while compiling ImageMagick on mipsel

[aaro.koskinen at iki dot fi]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

Aaro Koskinen <aaro.koskinen at iki dot fi> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |4.7.3
      Known to fail|                            |4.8.0, 4.8.1

--- Comment #2 from Aaro Koskinen <aaro.koskinen at iki dot fi> ---
Reproducible also with 4.8.0 but not with 4.7.3.

[Bug target/58158] internal compiler error: in extract_insn, at recog.c:2150 while compiling ImageMagick on mipsel

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

Mikael Pettersson <mikpe at it dot uu.se> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikpe at it dot uu.se,
                   |                            |pinskia at gcc dot gnu.org

--- Comment #3 from Mikael Pettersson <mikpe at it dot uu.se> ---
Started with the "Improve COND_EXPR expansion" patch in r187183.  Still occurs
with recent trunk and head of 4.8 branch.  Author CC:d.

[Bug target/58158] internal compiler error: in extract_insn, at recog.c:2150 while compiling ImageMagick on mipsel

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code
             Target|mipsel-linux-gnu            |mipsel-linux-gnu
                   |                            |(loongson2f)

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This is a loongson2f specific bug in the back-end.  It works correctly for
-march=octeon/-march=mips32/-march=mips64 .

[Bug target/58158] internal compiler error: in extract_insn, at recog.c:2150 while compiling ImageMagick on mipsel

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-08-22
               Host|mipsel-linux-gnu            |
     Ever confirmed|0                           |1
              Build|i486-linux-gnu (gcc (Debian |
                   |4.8.1-6) 4.8.1)             |

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The backend says we have conditional moves on the floating point CC modes and
expands it that way but really the target says we don't and it causes the ICE.

Since it is only loongson2f only bug I am not going to fix it.  I will let the
loongson maintainer fix it since I don't know what loongson2f really has when
it comes to conditional moves.

Confirmed.

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.8.2
            Summary|internal compiler error: in |[4.8/4.9 Regression] ICE
                   |extract_insn, at            |with conditional moves on
                   |recog.c:2150 while          |GPRs with a floating point
                   |compiling ImageMagick on    |conditional on mipsel with
                   |mipsel                      |loongson2f

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[aaro.koskinen at iki dot fi]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #7 from Aaro Koskinen <aaro.koskinen at iki dot fi> ---
Loongson 2F is MIPS III + movz/movn only. I am currently running my compiler
with the below workaround, and it seems to help:

--- gcc-4.8.1.orig/gcc/config/mips/mips.md      2013-02-25 15:53:16.000000000
+0200
+++ gcc-4.8.1/gcc/config/mips/mips.md   2013-08-23 23:09:13.000000000 +0300
@@ -6708,20 +6708,23 @@
 ;; These are the main define_expand's used to make conditional moves.

 (define_expand "mov<mode>cc"
   [(set (match_dup 4) (match_operand 1 "comparison_operator"))
    (set (match_operand:GPR 0 "register_operand")
        (if_then_else:GPR (match_dup 5)
                          (match_operand:GPR 2 "reg_or_0_operand")
                          (match_operand:GPR 3 "reg_or_0_operand")))]
   "ISA_HAS_CONDMOVE"
 {
+  if (!ISA_HAS_FP_CONDMOVE &&
+      GET_MODE_CLASS (GET_MODE (XEXP (operands[1], 0))) != MODE_INT)
+    FAIL;
   mips_expand_conditional_move (operands);
   DONE;
 })

 (define_expand "mov<mode>cc"
   [(set (match_dup 4) (match_operand 1 "comparison_operator"))
    (set (match_operand:SCALARF 0 "register_operand")
        (if_then_else:SCALARF (match_dup 5)
                              (match_operand:SCALARF 2 "register_operand")
                              (match_operand:SCALARF 3 "register_operand")))]

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.2                       |4.8.3

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 4.8.2 has been released.

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |biergaizi2009 at gmail dot com

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 59620 has been marked as a duplicate of this bug. ***

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[biergaizi2009 at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #11 from Tom Li <biergaizi2009 at gmail dot com> ---
My Gentoo system is nearly broken because of this bug :(

I'm going to report it to Loongson's mailing list.

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[biergaizi2009 at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #12 from Tom Li <biergaizi2009 at gmail dot com> ---
And I just applied Aaro Koskinen's patch on gcc 4.8.2, recompile gcc and
resolve this issue.

Patch for 4.8.2:

--- gcc-4.8.2/gcc/config/mips/mips.md   2013-02-25 21:53:16.000000000 +0800
+++ gcc-4.8.2-fixed/gcc/config/mips/mips.md     2013-12-31 20:23:26.021484375
+0800
@@ -6715,6 +6715,9 @@
                          (match_operand:GPR 3 "reg_or_0_operand")))]
   "ISA_HAS_CONDMOVE"
 {
+  if (!ISA_HAS_FP_CONDMOVE &&
+      GET_MODE_CLASS (GET_MODE (XEXP (operands[1], 0))) != MODE_INT)
+    FAIL;
   mips_expand_conditional_move (operands);
   DONE;
 })

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

Richard Henderson <rth at gcc dot gnu.org> changed:

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

--- Comment #13 from Richard Henderson <rth at gcc dot gnu.org> ---
(In reply to Tom Li from comment #12)
>  {
> +  if (!ISA_HAS_FP_CONDMOVE &&
> +      GET_MODE_CLASS (GET_MODE (XEXP (operands[1], 0))) != MODE_INT)
> +    FAIL;

The patch is clearly wrong.  It's attempting to look through
a subreg around operands[1], but of course that subreg will
not always exist.

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #14 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Richard Henderson from comment #13)
> (In reply to Tom Li from comment #12)
> >  {
> > +  if (!ISA_HAS_FP_CONDMOVE &&
> > +      GET_MODE_CLASS (GET_MODE (XEXP (operands[1], 0))) != MODE_INT)
> > +    FAIL;
> 
> The patch is clearly wrong.  It's attempting to look through
> a subreg around operands[1], but of course that subreg will
> not always exist.

Actually it is correct as operands[1] will be an comparison_operator which
always have two operands itself.

[Bug target/58158] [4.8/4.9/4.10 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.3                       |4.8.4

--- Comment #15 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 4.8.3 is being released, adjusting target milestone.

[Bug target/58158] [4.8/4.9/5 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.4                       |4.8.5

--- Comment #16 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 4.8.4 has been released.

[Bug target/58158] [4.8/4.9/5 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[sje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

Steve Ellcey <sje at gcc dot gnu.org> changed:

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

--- Comment #17 from Steve Ellcey <sje at gcc dot gnu.org> ---
Updated patch posted to gcc-patches.

https://gcc.gnu.org/ml/gcc-patches/2015-03/msg00106.html

[Bug target/58158] [4.8/4.9/5 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[aaro.koskinen at iki dot fi]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #18 from Aaro Koskinen <aaro.koskinen at iki dot fi> ---
(In reply to Steve Ellcey from comment #17)
> Updated patch posted to gcc-patches.
> 
> https://gcc.gnu.org/ml/gcc-patches/2015-03/msg00106.html

Thanks - always good to get rid of out-of-tree patches. By review, your patch
looks good, but if you like I could also test it, if you can point out some
tree
or commit against which I should test it against. Thanks.

[Bug target/58158] [4.8/4.9/5 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[sje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #19 from Steve Ellcey <sje at gcc dot gnu.org> ---
Author: sje
Date: Tue Mar  3 00:07:19 2015
New Revision: 221128

URL: https://gcc.gnu.org/viewcvs?rev=221128&root=gcc&view=rev
Log:
2015-03-02  Steve Ellcey  <sellcey@imgtec.com>

    PR target/58158
    * config/mips/mips.md (mov<mode>cc): Change ISA_HAS_SEL check to
    !ISA_HAS_FP_CONDMOVE.

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

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[biergaizi2009 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #20 from Tom Li <biergaizi2009 at gmail dot com> ---
Great job, I'll test it.

[Bug target/58158] [4.8/4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.5                       |4.9.3

--- Comment #21 from Richard Biener <rguenth at gcc dot gnu.org> ---
The gcc-4_8-branch is being closed, re-targeting regressions to 4.9.3.

[Bug target/58158] [4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

--- Comment #22 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 4.9.3 has been released.

[Bug target/58158] [4.9 Regression] ICE with conditional moves on GPRs with a floating point conditional on mipsel with loongson2f

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58158

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.9.3                       |4.9.4