[Bug middle-end/50633] New: [4.7 Regression] New test failures

[Bug middle-end/50633] New: [4.7 Regression] New test failures

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

             Bug #: 50633
           Summary: [4.7 Regression] New test failures
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: hjl.tools@gmail.com


On Linux/x32, revision 179566 gave

FAIL: gcc.c-torture/execute/20030914-1.c execution,  -O1 
FAIL: gcc.c-torture/execute/20030914-1.c execution,  -O2 
FAIL: gcc.c-torture/execute/20030914-1.c execution,  -O2 -flto 
FAIL: gcc.c-torture/execute/20030914-1.c execution,  -O2 -flto
-flto-partition=none 
FAIL: gcc.dg/pr50132.c (internal compiler error)
FAIL: gcc.dg/pr50132.c (test for excess errors)
FAIL: gfortran.dg/g77/short.f  -O1  execution test
FAIL: gfortran.dg/g77/short.f  -O2  execution test
FAIL: gfortran.dg/g77/short.f  -Os  execution test
FAIL: gfortran.dg/reshape_rank7.f90  -O1  execution test
FAIL: gfortran.fortran-torture/execute/entry_4.f90 execution,  -O2 
FAIL: gfortran.fortran-torture/execute/entry_4.f90 execution,  -O2
-fbounds-check 
FAIL: gfortran.fortran-torture/execute/entry_4.f90 execution,  -O2
-fomit-frame-pointer -finline-functions 
FAIL: gfortran.fortran-torture/execute/entry_4.f90 execution,  -O2
-fomit-frame-pointer -finline-functions -funroll-loops 
FAIL: gfortran.fortran-torture/execute/entry_4.f90 execution,  -O3 -g 
FAIL: gfortran.fortran-torture/execute/entry_4.f90 execution, -O2
-ftree-vectorize -msse2 
FAIL: gfortran.fortran-torture/execute/stack_varsize.f90 execution,  -O3 -g 
FAIL: gfortran.fortran-torture/execute/stack_varsize.f90 execution, -O2
-ftree-vectorize -msse2 
FAIL: libgomp.c++/collapse-2.C  -O1  execution test
FAIL: libgomp.fortran/appendix-a/a.16.1.f90  -O1  execution test
FAIL: libgomp.fortran/omp_workshare2.f  -O  execution test

Revision 179528 is OK.

[Bug middle-end/50633] [4.7 Regression] New test failures

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

Rainer Orth <ro at gcc dot gnu.org> changed:

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

--- Comment #1 from Rainer Orth <ro at gcc dot gnu.org> 2011-10-06 16:05:18 UTC ---
I also see the entry_4.f90 failures on i386-pc-solaris2.10 (but only those).

  Rainer

[Bug middle-end/50633] [4.7 Regression] New test failures

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-10-06 16:23:37 UTC ---
> FAIL: gfortran.fortran-torture/execute/entry_4.f90 execution,  -O2 

It is pr50628.

> FAIL: gcc.dg/pr50132.c (internal compiler error)

The ICE is

/opt/gcc/work/gcc/testsuite/gcc.dg/pr50132.c: In function 'foo':
/opt/gcc/work/gcc/testsuite/gcc.dg/pr50132.c:10:1: internal compiler error: in
maybe_record_trace_start, at dwarf2cfi.c:2243

I don't see the other failures on x86_64-apple-darwin10.

[Bug middle-end/50633] [4.7 Regression] New test failures

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

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

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

--- Comment #3 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-06 17:39:24 UTC ---
-fshrink-wrap caused those failures:

FAIL: gcc.c-torture/execute/20030914-1.c execution,  -O1 
FAIL: gcc.c-torture/execute/20030914-1.c execution,  -O2 
FAIL: gcc.c-torture/execute/20030914-1.c execution,  -O2 -flto 
FAIL: gcc.c-torture/execute/20030914-1.c execution,  -O2 -flto
-flto-partition=none 
FAIL: gfortran.dg/g77/short.f  -O1  execution test
FAIL: gfortran.dg/g77/short.f  -O2  execution test
FAIL: gfortran.dg/g77/short.f  -Os  execution test
FAIL: gfortran.dg/reshape_rank7.f90  -O1  execution test
FAIL: gfortran.fortran-torture/execute/stack_varsize.f90 execution,  -O3 -g 
FAIL: gfortran.fortran-torture/execute/stack_varsize.f90 execution, -O2
-ftree-vectorize -msse2 
FAIL: libgomp.c++/collapse-2.C  -O1  execution test
FAIL: libgomp.fortran/appendix-a/a.16.1.f90  -O1  execution test
FAIL: libgomp.fortran/omp_workshare2.f  -O  execution test

[Bug middle-end/50633] [4.7 Regression] New test failures

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #4 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-06 17:55:48 UTC ---
Here is a small testcase. shrink-wrap screwed up stack adjustment
for local variables:

[hjl@gnu-mic-2 pr50633]$ cat x.i
struct s { int val[16]; };

extern double f (struct s pb, double pc);

int main ()
{
  struct s x;
  int i;

  for (i = 0; i < 16; i++)
    x.val[i] = i + 1;
  if (f (x, 10000.0L) != 10136.0L)
    __builtin_abort ();
  return 0;
}
[hjl@gnu-mic-2 pr50633]$ /export/build/gnu/gcc-x32/build-x86_64-linux/gcc/xgcc
-B/export/build/gnu/gcc-x32/build-x86_64-linux/gcc/ -mx32 -O -S x.i
-fno-shrink-wrap -o good.s
[hjl@gnu-mic-2 pr50633]$ /export/build/gnu/gcc-x32/build-x86_64-linux/gcc/xgcc
-B/export/build/gnu/gcc-x32/build-x86_64-linux/gcc/ -mx32 -O -S x.i -o bad.s
[hjl@gnu-mic-2 pr50633]$ cat good.s
    .file    "x.i"
    .text
    .globl    main
    .type    main, @function
main:
.LFB0:
    .cfi_startproc
    subq    $136, %rsp
    .cfi_def_cfa_offset 144
    movl    $0, %eax
    movl    %esp, %ecx
    addl    $60, %ecx
.L2:
    addl    $1, %eax
    leal    (%rcx,%rax,4), %edx
    movl    %eax, (%edx)
    cmpl    $16, %eax
    jne    .L2
    movq    64(%rsp), %rax
    movq    %rax, (%rsp)
    movq    72(%rsp), %rax
    movq    %rax, 8(%rsp)
    movq    80(%rsp), %rax
    movq    %rax, 16(%rsp)
    movq    88(%rsp), %rax
    movq    %rax, 24(%rsp)
    movq    96(%rsp), %rax
    movq    %rax, 32(%rsp)
    movq    104(%rsp), %rax
    movq    %rax, 40(%rsp)
    movq    112(%rsp), %rax
    movq    %rax, 48(%rsp)
    movq    120(%rsp), %rax
    movq    %rax, 56(%rsp)
    movsd    .LC0(%rip), %xmm0
    call    f
    ucomisd    .LC1(%rip), %xmm0
    jp    .L5
    je    .L7
.L5:
    call    abort
.L7:
    movl    $0, %eax
    addq    $136, %rsp
    .cfi_def_cfa_offset 8
    ret
    .cfi_endproc
.LFE0:
    .size    main, .-main
    .section    .rodata.cst8,"aM",@progbits,8
    .align 8
.LC0:
    .long    0
    .long    1086556160
    .align 8
.LC1:
    .long    0
    .long    1086573568
    .ident    "GCC: (GNU) 4.7.0 20111005 (experimental)"
    .section    .note.GNU-stack,"",@progbits
[hjl@gnu-mic-2 pr50633]$ cat bad.s
    .file    "x.i"
    .text
    .globl    main
    .type    main, @function
main:
.LFB0:
    .cfi_startproc
    movl    $0, %eax
    movl    %esp, %ecx
    addl    $60, %ecx
.L2:
    addl    $1, %eax
    leal    (%rcx,%rax,4), %edx
    movl    %eax, (%edx)
    cmpl    $16, %eax
    jne    .L2
    subq    $136, %rsp
    .cfi_def_cfa_offset 144
    movq    64(%rsp), %rax
    movq    %rax, (%rsp)
    movq    72(%rsp), %rax
    movq    %rax, 8(%rsp)
    movq    80(%rsp), %rax
    movq    %rax, 16(%rsp)
    movq    88(%rsp), %rax
    movq    %rax, 24(%rsp)
    movq    96(%rsp), %rax
    movq    %rax, 32(%rsp)
    movq    104(%rsp), %rax
    movq    %rax, 40(%rsp)
    movq    112(%rsp), %rax
    movq    %rax, 48(%rsp)
    movq    120(%rsp), %rax
    movq    %rax, 56(%rsp)
    movsd    .LC0(%rip), %xmm0
    call    f
    ucomisd    .LC1(%rip), %xmm0
    jp    .L5
    je    .L7
.L5:
    call    abort
.L7:
    movl    $0, %eax
    addq    $136, %rsp
    .cfi_def_cfa_offset 8
    ret
    .cfi_endproc
.LFE0:
    .size    main, .-main
    .section    .rodata.cst8,"aM",@progbits,8
    .align 8
.LC0:
    .long    0
    .long    1086556160
    .align 8
.LC1:
    .long    0
    .long    1086573568
    .ident    "GCC: (GNU) 4.7.0 20111005 (experimental)"
    .section    .note.GNU-stack,"",@progbits
[hjl@gnu-mic-2 pr50633]$ diff -up good.s bad.s
--- good.s    2011-10-06 10:52:45.410961190 -0700
+++ bad.s    2011-10-06 10:52:48.938896097 -0700
@@ -5,8 +5,6 @@
 main:
 .LFB0:
     .cfi_startproc
-    subq    $136, %rsp
-    .cfi_def_cfa_offset 144
     movl    $0, %eax
     movl    %esp, %ecx
     addl    $60, %ecx
@@ -16,6 +14,8 @@ main:
     movl    %eax, (%edx)
     cmpl    $16, %eax
     jne    .L2
+    subq    $136, %rsp
+    .cfi_def_cfa_offset 144
     movq    64(%rsp), %rax
     movq    %rax, (%rsp)
     movq    72(%rsp), %rax
[hjl@gnu-mic-2 pr50633]$

[Bug middle-end/50633] [4.7 Regression] New test failures

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #5 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-06 18:38:21 UTC ---
static int
frame_required_for_rtx (rtx *loc, void *data ATTRIBUTE_UNUSED)
{
  rtx x = *loc;
  if (x == stack_pointer_rtx || x == hard_frame_pointer_rtx
      || x == arg_pointer_rtx || x == pic_offset_table_rtx)
    return 1;
  return 0;
}

failed to handle

(gdb) call debug_rtx (insn)
(insn 99 34 36 2 (set (reg:SI 2 cx [106])
        (reg:SI 7 sp)) x.i:11 64 {*movsi_internal}
     (nil))
(gdb) 

X32 has "(reg/f:DI 7 sp)" as hardware stack pointer.  It also
should check regno.

[Bug middle-end/50633] [4.7 Regression] New test failures

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #6 from Bernd Schmidt <bernds at gcc dot gnu.org> 2011-10-06 18:43:21 UTC ---
How do you build this? What's the target triplet?

[Bug middle-end/50633] [4.7 Regression] New test failures

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #7 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-06 18:50:14 UTC ---
This patch seems to work:

diff --git a/gcc/function.c b/gcc/function.c
index 863f09d..0bc1dd9 100644
--- a/gcc/function.c
+++ b/gcc/function.c
@@ -5312,6 +5312,21 @@ frame_required_for_rtx (rtx *loc, void *data
ATTRIBUTE_UNUSED)
   if (x == stack_pointer_rtx || x == hard_frame_pointer_rtx
       || x == arg_pointer_rtx || x == pic_offset_table_rtx)
     return 1;
+  if (x != NULL
+      && REG_P (x)
+      && ((stack_pointer_rtx != NULL
+      && REG_P (stack_pointer_rtx)
+      && REGNO (x) == REGNO (stack_pointer_rtx))
+      || (hard_frame_pointer_rtx != NULL
+          && REG_P (hard_frame_pointer_rtx)
+          && REGNO (x) == REGNO (hard_frame_pointer_rtx))
+      || (arg_pointer_rtx != NULL
+          && REG_P (arg_pointer_rtx)
+          && REGNO (x) == REGNO (arg_pointer_rtx))
+      || (pic_offset_table_rtx != NULL
+          && REG_P (pic_offset_table_rtx)
+          && REGNO (x) == REGNO (pic_offset_table_rtx))))
+    return 1;
   return 0;
 }

[Bug middle-end/50633] [4.7 Regression] New test failures

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.0

--- Comment #8 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-06 18:51:45 UTC ---
(In reply to comment #6)
> How do you build this? What's the target triplet?

If you build GCC for Linux/x86-64, you will get x32 code generation
and you can see the problem from assembly output.

[Bug middle-end/50633] [4.7 Regression] New test failures

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #9 from Bernd Schmidt <bernds at gcc dot gnu.org> 2011-10-06 19:59:18 UTC ---
Created attachment 25434
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25434
Test patch

Can you give this a try? Seems to fix the testcase.

[Bug middle-end/50633] [4.7 Regression] New test failures

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #10 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-06 20:09:13 UTC ---
(In reply to comment #9)
> Created attachment 25434 [details]
> Test patch
> 
> Can you give this a try? Seems to fix the testcase.

It works. Thanks.

[Bug middle-end/50633] [4.7 Regression] New test failures

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #11 from Bernd Schmidt <bernds at gcc dot gnu.org> 2011-10-07 12:10:11 UTC ---
I see none of these failures in the x86_64 test run with my last patch applied.
Can you confirm that it's fixed?

[Bug middle-end/50633] [4.7 Regression] New test failures

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

--- Comment #12 from Bernd Schmidt <bernds at gcc dot gnu.org> 2011-10-07 12:20:57 UTC ---
Sorry, commit didn't go through earlier - it's in now.

[Bug middle-end/50633] [4.7 Regression] New test failures

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50633

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

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

--- Comment #13 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-07 16:20:59 UTC ---
Fixed as of revision 179662.