[Bug target/55672] New: [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[Bug target/55672] New: [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

             Bug #: 55672
           Summary: [4.8 Regression] -fstack-check=generic ICEs in
                    print_reg, at config/i386/i386.c:13868
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: rguenth@gcc.gnu.org
            Target: i?86-*-*


int main ()
{
  int x[8];
  if (x[0] != 4)
    __builtin_abort ();
  return 0;
}

./cc1 -quiet run-id-pr47653.i -O -fstack-check=generic -m32
run-id-pr47653.i: In function 'main':
run-id-pr47653.i:7:1: internal compiler error: in print_reg, at
config/i386/i386.c:13868

This is the

FAIL: gcc.dg/graphite/run-id-pr47653.c (internal compiler error)
FAIL: gcc.dg/graphite/run-id-pr47653.c (test for excess errors)
WARNING: gcc.dg/graphite/run-id-pr47653.c compilation failed to produce
executable

FAIL on x86_64 32-bit testing.

[Bug target/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.8.0

[Bug target/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

Uros Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-12-13
     Ever Confirmed|0                           |1

--- Comment #1 from Uros Bizjak <ubizjak at gmail dot com> 2012-12-13 16:43:23 UTC ---
Confirmed, we tried to print "frame" reg from:

(insn 6 3 7 2 (set (reg:CCZ 17 flags)
        (compare:CCZ (mem/j/c:SI (plus:SI (reg/f:SI 20 frame)
                    (const_int -32 [0xffffffffffffffe0])) [0 x+0 S4 A32])
            (const_int 4 [0x4]))) pr55672.c:4 7 {*cmpsi_1}
     (nil))

[Bug target/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

Uros Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ra
                 CC|                            |vmakarov at gcc dot gnu.org

--- Comment #2 from Uros Bizjak <ubizjak at gmail dot com> 2012-12-13 16:49:45 UTC ---
LRA doesn't substitute frame reg with bp. In _.209r.reload, we still have:

(insn 6 3 7 2 (set (reg:CCZ 17 flags)
        (compare:CCZ (mem/j/c:SI (plus:SI (reg/f:SI 20 frame)
                    (const_int -32 [0xffffffffffffffe0])) [0 x+0 S4 A32])
            (const_int 4 [0x4]))) pr55672.c:4 7 {*cmpsi_1}
     (nil))

Adding CC.

[Bug target/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

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

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

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-12-17 10:51:45 UTC ---
Indeed, with LRA forced off reload replaces the 
(insn 6 3 7 2 (set (reg:CCZ 17 flags)
        (compare:CCZ (mem/c:SI (plus:SI (reg/f:SI 20 frame)
                    (const_int -32 [0xffffffffffffffe0])) [2 x+0 S4 A32])
            (const_int 4 [0x4]))) pr55672.c:4 7 {*cmpsi_1}
     (nil))
insn (only reference to frame register) with
(insn 6 3 7 2 (set (reg:CCZ 17 flags)
        (compare:CCZ (mem/c:SI (reg/f:SI 7 sp) [2 x+0 S4 A32])
            (const_int 4 [0x4]))) pr55672.c:4 7 {*cmpsi_1}
     (nil))
while LRA doesn't eliminate it.

[Bug target/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[vmakarov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

--- Comment #4 from Vladimir Makarov <vmakarov at gcc dot gnu.org> 2012-12-17 21:39:28 UTC ---
If stack_realign_p is true, frame_pointer_needed is also true.   So we can use
fp to eliminate frame but i386.c::x86_can_eliminate prohibits it.  The code
looks strange:


 if (stack_realign_fp)
    return ((from == ARG_POINTER_REGNUM
             && to == HARD_FRAME_POINTER_REGNUM)
            || (from == FRAME_POINTER_REGNUM
                && to == STACK_POINTER_REGNUM));

So we permit to change argument pointer but not frame pointer to FP which again
is strange IMHO.   Changing the code to

 if (stack_realign_fp)
    return ((from == ARG_POINTER_REGNUM
             && to == HARD_FRAME_POINTER_REGNUM)
            || (from == FRAME_POINTER_REGNUM
                && to == STACK_POINTER_REGNUM)
            || (from == FRAME_POINTER_REGNUM
                && to == HARD_FRAME_POINTER_REGNUM));

solves the problem.

[Bug target/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

Uros Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hjl.tools at gmail dot com

--- Comment #5 from Uros Bizjak <ubizjak at gmail dot com> 2012-12-18 09:54:17 UTC ---
CC author of stack realignment functionality.

[Bug target/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

--- Comment #6 from H.J. Lu <hjl.tools at gmail dot com> 2012-12-18 15:11:38 UTC ---
This is what we changed in reload for stack realignment:

diff --git a/gcc/reload1.c b/gcc/reload1.c
index f28b01c..9b81062 100644
--- a/gcc/reload1.c
+++ b/gcc/reload1.c
@@ -3663,8 +3663,11 @@ update_eliminables (HARD_REG_SET *pset)
   frame_pointer_needed = 1;
   for (ep = reg_eliminate; ep < &reg_eliminate[NUM_ELIMINABLE_REGS]; ep++)
     {
-      if (ep->can_eliminate && ep->from == FRAME_POINTER_REGNUM
-         && ep->to != HARD_FRAME_POINTER_REGNUM)
+      if (ep->can_eliminate
+         && ep->from == FRAME_POINTER_REGNUM
+         && ep->to != HARD_FRAME_POINTER_REGNUM
+         && (! SUPPORTS_STACK_ALIGNMENT
+             || ! crtl->stack_realign_needed))
        frame_pointer_needed = 0;

       if (! ep->can_eliminate && ep->can_eliminate_previous)
@@ -3720,7 +3723,10 @@ init_elim_table (void)
       ep->to = ep1->to;
       ep->can_eliminate = ep->can_eliminate_previous
        = (CAN_ELIMINATE (ep->from, ep->to)
-          && ! (ep->to == STACK_POINTER_REGNUM && frame_pointer_needed));
+          && ! (ep->to == STACK_POINTER_REGNUM
+                && frame_pointer_needed 
+                && (! SUPPORTS_STACK_ALIGNMENT
+                    || ! stack_realign_fp)));
     }
 #else
   reg_eliminate[0].from = reg_eliminate_1[0].from;


LRA should have similar codes.

[Bug target/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

--- Comment #7 from H.J. Lu <hjl.tools at gmail dot com> 2012-12-18 19:42:59 UTC ---
LRA doesn't handle HARD_FRAME_POINTER_IS_FRAME_POINTER at all.

[Bug rtl-optimization/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

H.J. Lu <hjl.tools at gmail dot com> changed:

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

--- Comment #8 from H.J. Lu <hjl.tools at gmail dot com> 2012-12-18 20:35:17 UTC ---
(In reply to comment #4)
> If stack_realign_p is true, frame_pointer_needed is also true.   So we can use
> fp to eliminate frame but i386.c::x86_can_eliminate prohibits it.  The code
> looks strange:
> 
> 
>  if (stack_realign_fp)
>     return ((from == ARG_POINTER_REGNUM
>              && to == HARD_FRAME_POINTER_REGNUM)
>             || (from == FRAME_POINTER_REGNUM
>                 && to == STACK_POINTER_REGNUM));
> 
> So we permit to change argument pointer but not frame pointer to FP which again
> is strange IMHO.   Changing the code to
> 
>  if (stack_realign_fp)
>     return ((from == ARG_POINTER_REGNUM
>              && to == HARD_FRAME_POINTER_REGNUM)
>             || (from == FRAME_POINTER_REGNUM
>                 && to == STACK_POINTER_REGNUM)
>             || (from == FRAME_POINTER_REGNUM
>                 && to == HARD_FRAME_POINTER_REGNUM));
> 
> solves the problem.

It fixes ICE, but generates questionable code:

main:
.LFB0:
    .cfi_startproc
    pushl    %ebp
    .cfi_def_cfa_offset 8
    .cfi_offset 5, -8
    movl    %esp, %ebp
    .cfi_def_cfa_register 5
    andl    $-16, %esp
    subl    $8236, %esp
    orl    $0, (%esp)
    addl    $8204, %esp
    cmpl    $4, -40(%ebp)
    je    .L2
    call    abort
.L2:
    movl    $0, %eax
    leave
    .cfi_restore 5
    .cfi_def_cfa 4, 4
    ret

Without LRA, we got

main:
.LFB0:
    .cfi_startproc
    pushl    %ebp
    .cfi_def_cfa_offset 8
    .cfi_offset 5, -8
    movl    %esp, %ebp
    .cfi_def_cfa_register 5
    andl    $-16, %esp
    subl    $8236, %esp
    orl    $0, (%esp)
    addl    $8204, %esp
    cmpl    $4, (%esp)
    je    .L2
    call    abort
.L2:
    movl    $0, %eax
    leave
    .cfi_restore 5
    .cfi_def_cfa 4, 4
    ret

The difference is

--- x.s    2012-12-18 12:24:17.072888139 -0800
+++ no-lra.s    2012-12-18 12:30:11.419157548 -0800
@@ -14,7 +14,7 @@ main:
     subl    $8236, %esp
     orl    $0, (%esp)
     addl    $8204, %esp
-    cmpl    $4, -40(%ebp)
+    cmpl    $4, (%esp)
     je    .L2
     call    abort
 .L2:

I think LRA generated code is wrong.  The reason we don't allow
converting software frame pointer to hardware frame pointer is
when stack alignment is needed, hardware frame pointer is used
to save stack pointer.  We can no longer use it for software
frame pointer.

[Bug rtl-optimization/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[vmakarov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

--- Comment #9 from Vladimir Makarov <vmakarov at gcc dot gnu.org> 2012-12-21 21:28:51 UTC ---
(In reply to comment #8)
> 
> The difference is
> 
> --- x.s    2012-12-18 12:24:17.072888139 -0800
> +++ no-lra.s    2012-12-18 12:30:11.419157548 -0800
> @@ -14,7 +14,7 @@ main:
>      subl    $8236, %esp
>      orl    $0, (%esp)
>      addl    $8204, %esp
> -    cmpl    $4, -40(%ebp)
> +    cmpl    $4, (%esp)
>      je    .L2
>      call    abort
>  .L2:
> 
> I think LRA generated code is wrong.  The reason we don't allow
> converting software frame pointer to hardware frame pointer is
> when stack alignment is needed, hardware frame pointer is used
> to save stack pointer.  We can no longer use it for software
> frame pointer.

H.J., thanks for the explanation.  Now I know what to do.  I'll start to work
on the solution after my vacation in 2 weeks.

[Bug rtl-optimization/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug rtl-optimization/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

--- Comment #10 from Uros Bizjak <ubizjak at gmail dot com> 2013-01-10 20:23:49 UTC ---
Author: vmakarov
Date: Thu Jan 10 20:07:55 2013
New Revision: 195095

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=195095
Log:
2013-01-10  Vladimir Makarov  <vmakarov@redhat.com>

    PR rtl-optimization/pr55672
    * lra-eliminations.c (mark_not_elimnable): Permit addition with
    const to be elimnable.

2013-01-10  Vladimir Makarov  <vmakarov@redhat.com>

    PR rtl-optimization/pr55672
    * gcc.target/i386/pr55672.c: New.


Added:
    trunk/gcc/testsuite/gcc.target/i386/pr55672.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/lra-eliminations.c
    trunk/gcc/testsuite/ChangeLog

[Bug rtl-optimization/55672] [4.8 Regression] -fstack-check=generic ICEs in print_reg, at config/i386/i386.c:13868

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55672

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

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

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-01-11 10:44:34 UTC ---
Assuming fixed.