[Bug other/50283] New: [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[Bug other/50283] New: [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

             Bug #: 50283
           Summary: [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution
                    test
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: other
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: danglin@gcc.gnu.org
                CC: rth@gcc.gnu.org
              Host: hppa2.0w-hp-hpux11.11
            Target: hppa2.0w-hp-hpux11.11
             Build: hppa2.0w-hp-hpux11.11


Executing on host: /test/gnu/gcc/objdir/gcc/testsuite/g++/../../g++
-B/test/gnu/gcc/objdir/gcc/testsuite/g++/../../
/test/gnu/gcc/gcc/gcc/testsuite/g++.dg/eh/simd-1.C  -nostdinc++
-I/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/libstdc++-v3/include/hppa2.0w-hp-hpux11.11
-I/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/libstdc++-v3/include
-I/test/gnu/gcc/gcc/libstdc++-v3/libsupc++
-I/test/gnu/gcc/gcc/libstdc++-v3/include/backward
-I/test/gnu/gcc/gcc/libstdc++-v3/testsuite/util -fmessage-length=0  -O -Wno-abi
   -L/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/./libstdc++-v3/src/.libs 
-B/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/./libstdc++-v3/src/.libs 
-L/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/./libstdc++-v3/src/.libs  -lm   -o
./simd-1.exe    (timeout = 300)
PASS: g++.dg/eh/simd-1.C (test for excess errors)
Setting LD_LIBRARY_PATH to
.:/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/./libstdc++-v3/src/.libs:/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/./libstdc++-v3/src/.libs:/test/gnu/gcc/objdir/gcc:.:/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/./libstdc++-v3/src/.libs:/test/gnu/gcc/objdir/hppa2.0w-hp-hpux11.11/./libstdc++-v3/src/.libs:/test/gnu/gcc/objdir/gcc
WARNING: program timed out.
FAIL: g++.dg/eh/simd-1.C execution test

The throw in f3 results in an infinite unwind loop.

Thought at first that this might be caused by stack allocation instruction
being placed in call delay slot, but fail also occurs at -O0.

Similar fails are:
FAIL: g++.dg/eh/simd-2.C execution test
FAIL: g++.dg/opt/eh2.C execution test

[Bug other/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #2 from John David Anglin <danglin at gcc dot gnu.org> 2011-09-03 18:21:24 UTC ---
Created attachment 25187
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25187
.s file

[Bug other/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #1 from John David Anglin <danglin at gcc dot gnu.org> 2011-09-03 18:20:58 UTC ---
(In reply to comment #0)

> Thought at first that this might be caused by stack allocation instruction
> being placed in call delay slot, but fail also occurs at -O0.

This is wrong.  The test doesn't fail at -O0 (I ran the wrong executable).
At -O, the CFI labels appear incorrectly placed

[Bug other/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

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

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

[Bug other/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #3 from John David Anglin <danglin at gcc dot gnu.org> 2011-10-03 19:43:40 UTC ---
Created attachment 25405
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25405
.final dump

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |EH
           Priority|P3                          |P1
          Component|other                       |middle-end

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #4 from John David Anglin <danglin at gcc dot gnu.org> 2011-10-10 14:33:17 UTC ---
The bug was introduced in revision 177218:

2011-08-02  Richard Henderson  <rth@redhat.com>

        PR target/49864
        * reg-notes.def (REG_ARGS_SIZE): New.
        * calls.c (emit_call_1): Emit REG_ARGS_SIZE for call_pop.
        (expand_call): Add REG_ARGS_SIZE to emit_stack_restore.
        * cfgcleanup.c (old_insns_match_p): Don't allow cross-jumping to
        different stack levels.
        * combine-stack-adj.c (adjust_frame_related_expr): Remove.
        (maybe_move_args_size_note): New.
        (combine_stack_adjustments_for_block): Use it.
        * combine.c (distribute_notes): Place REG_ARGS_SIZE.
        * dwarf2cfi.c (dw_cfi_row_struct): Remove args_size member.
        (dw_trace_info): Add beg_true_args_size, end_true_args_size,
        beg_delay_args_size, end_delay_args_size, eh_head, args_size_undefined.
        (cur_cfa): New.
        (queued_args_size): Remove.
        (add_cfi_args_size): Assert size is non-negative.
        (stack_adjust_offset, dwarf2out_args_size): Remove.
        (dwarf2out_stack_adjust, dwarf2out_notice_stack_adjust): Remove.
        (notice_args_size, notice_eh_throw): New.
        (dwarf2out_frame_debug_def_cfa): Use cur_cfa.
        (dwarf2out_frame_debug_adjust_cfa): Likewise.
        (dwarf2out_frame_debug_cfa_offset): Likewise.
        (dwarf2out_frame_debug_expr): Likewise.  Don't stack_adjust_offset.
        (dwarf2out_frame_debug): Don't handle non-frame-related-p insns.
        (change_cfi_row): Don't emit args_size.
        (maybe_record_trace_start_abnormal): Split out from ...
        (maybe_record_trace_start): Here.  Set args_size_undefined.
        (create_trace_edges): Update to match.
        (scan_trace): Handle REG_ARGS_SIZE.
        (connect_traces): Connect args_size between EH insns.
        * emit-rtl.c (try_split): Handle REG_ARGS_SIZE.
        * explow.c (suppress_reg_args_size): New.
        (adjust_stack_1): Split out from ...
        (adjust_stack): ... here.
        (anti_adjust_stack): Use it.
        (allocate_dynamic_stack_space): Suppress REG_ARGS_SIZE.
        * expr.c (mem_autoinc_base): New.
        (fixup_args_size_notes): New.
        (emit_single_push_insn_1): Rename from emit_single_push_insn.
        (emit_single_push_insn): New.  Generate REG_ARGS_SIZE.
        * recog.c (peep2_attempt): Handle REG_ARGS_SIZE.
        * reload1.c (reload_as_needed): Likewise.
        * rtl.h (fixup_args_size_notes): Declare.

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #5 from John David Anglin <danglin at gcc dot gnu.org> 2011-11-21 17:04:20 UTC ---
Regarding label placement of frame related insns, we have the following
rtl for f2:

(insn 14 2 15 (sequence [
            (call_insn 5 2 11 (parallel [
                        (call (mem:SI (symbol_ref/v:SI ("@_Z2f3v") [flags
0x203]  <function_decl 7afc7b80 f3>) [0 f3 S4 A32])
                            (const_int 16 [0x10]))
                        (clobber (reg:SI 1 %r1))
                        (clobber (reg:SI 2 %r2))
                        (use (const_int 0 [0]))
                    ]) /test/gnu/gcc/gcc/gcc/testsuite/g++.dg/eh/simd-1.C:30
199 {call_symref}
                 (expr_list:REG_NORETURN (const_int 0 [0])
                    (nil))
                (nil))
            (insn/f 11 5 15 (set (reg/f:SI 30 %r30)
                    (plus:SI (reg/f:SI 30 %r30)
                        (const_int 64 [0x40]))) 111 {addsi3}
                 (nil))
        ]) /test/gnu/gcc/gcc/gcc/testsuite/g++.dg/eh/simd-1.C:30 -1
     (nil))

The assembler for f2 is:

    .SPACE $TEXT$
    .NSUBSPA $CODE$
    .align 4
    .EXPORT _Z2f2v,ENTRY,PRIV_LEV=3
L$FB0002:
_Z2f2v:
    .PROC
    .CALLINFO FRAME=64,CALLS,SAVE_RP
    .ENTRY
    stw %r2,-20(%r30)
L$CFI0001:
    .CALL 
    bl _Z2f3v,%r2
    ldo 64(%r30),%r30
L$CFI0002:
    nop
    .EXIT
    .PROCEND
L$FE0002:

L$CFI0002 is now after the call sequence whereas it used to be
before the sequence.

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #6 from John David Anglin <danglin at gcc dot gnu.org> 2011-11-27 18:35:52 UTC ---
Fails have changed somewhat but are still present:

WARNING: program timed out.FAIL: g++.dg/eh/simd-1.C -std=gnu++98 execution test
WARNING: program timed out.FAIL: g++.dg/eh/simd-1.C -std=gnu++11 execution test
FAIL: g++.dg/eh/simd-2.C -std=gnu++98 execution testWARNING: program timed out.
FAIL: g++.dg/eh/simd-2.C -std=gnu++11 execution test
WARNING: program timed out.FAIL: g++.dg/opt/eh2.C -std=gnu++98 execution test
WARNING: program timed out.FAIL: g++.dg/opt/eh2.C -std=gnu++11 execution test

Tested first with -fno-delay-branch and it doesn't fail.

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

John David Anglin <danglin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #25187|0                           |1
        is obsolete|                            |

--- Comment #7 from John David Anglin <danglin at gcc dot gnu.org> 2011-11-27 18:46:17 UTC ---
Created attachment 25920
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25920
.s file

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #8 from John David Anglin <danglin at gcc dot gnu.org> 2011-11-27 18:54:10 UTC ---
This is hack, but test doesn't fail if I interchange ldo and
.cfi_def_cfa_offset
lines:

    ldo 64(%r30),%r30
    .cfi_def_cfa_offset -64

How should a stack adjust in the final instruction of a delayed branch
be described?

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #9 from Richard Henderson <rth at gcc dot gnu.org> 2011-11-27 18:59:40 UTC ---
You don't.  We're supposed to prevent frame-related insns
from appearing in branch delay slots.

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[dave.anglin at bell dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #10 from dave.anglin at bell dot net 2011-11-27 19:07:46 UTC ---
On 27-Nov-11, at 1:59 PM, rth at gcc dot gnu.org wrote:

> You don't.  We're supposed to prevent frame-related insns
> from appearing in branch delay slots.

Is this a reorg issue?

--
John David Anglin    dave.anglin@bell.net

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #11 from John David Anglin <danglin at gcc dot gnu.org> 2011-12-01 02:08:20 UTC ---
Author: danglin
Date: Thu Dec  1 02:08:10 2011
New Revision: 181868

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=181868
Log:
    PR middle-end/50283
    * config/pa/pa.md (in_branch_delay): Disallow frame related insns.
    (in_nullified_branch_delay): Likewise.
    (in_call_delay): Likewise.


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

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

John David Anglin <danglin at gcc dot gnu.org> changed:

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

--- Comment #12 from John David Anglin <danglin at gcc dot gnu.org> 2011-12-01 02:17:03 UTC ---
Fixed.

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #13 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-12-16 02:16:00 UTC ---
(In reply to comment #10)
> On 27-Nov-11, at 1:59 PM, rth at gcc dot gnu.org wrote:
> 
> > You don't.  We're supposed to prevent frame-related insns
> > from appearing in branch delay slots.
> 
> Is this a reorg issue?

I am creating more generic patch for this for all targets that use branch delay
slots.

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[dave.anglin at bell dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #14 from dave.anglin at bell dot net 2011-12-17 04:12:48 UTC ---
On 15-Dec-11, at 9:16 PM, pinskia at gcc dot gnu.org wrote:

> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283
>
> --- Comment #13 from Andrew Pinski <pinskia at gcc dot gnu.org>  
> 2011-12-16 02:16:00 UTC ---
> (In reply to comment #10)
>> On 27-Nov-11, at 1:59 PM, rth at gcc dot gnu.org wrote:
>>
>>> You don't.  We're supposed to prevent frame-related insns
>>> from appearing in branch delay slots.
>>
>> Is this a reorg issue?
>
> I am creating more generic patch for this for all targets that use  
> branch delay
> slots.


My question may have been a bit obtuse but I got the impression that  
this should
be treated as a target specific problem.  Would you please revert my  
target patch
with your generic patch?

Thanks,
Dave
--
John David Anglin    dave.anglin@bell.net

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

vries at gcc dot gnu.org changed:

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

--- Comment #15 from vries at gcc dot gnu.org 2012-01-10 11:31:55 UTC ---
(In reply to comment #9)
> You don't.  We're supposed to prevent frame-related insns
> from appearing in branch delay slots.

Richard,

do you mean this in general, or just for calls?

Either way, we should be able to formulate an assert in scan_trace that checks
this condition. Do you think that would be useful?

Thanks,
- Tom

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #16 from Richard Henderson <rth at gcc dot gnu.org> 2012-01-10 21:40:22 UTC ---
Calls are especially problematic, yes.  You've just fixed a bug
for branches; hopefully that's the last of them.

Though if it were a matter of preferences, I would expect that
choosing a non-frame-related insn for the delay slot over a
frame-related insn of otherwise equal priority would produce
less heartburn in consumers even if it's possibly to handle it
within gcc.

You absolutely correct that we ought to have a check for this.

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #17 from vries at gcc dot gnu.org 2012-01-11 09:45:02 UTC ---
>> Either way, we should be able to formulate an assert in scan_trace that
>> checks this condition. Do you think that would be useful?

> You absolutely correct that we ought to have a check for this.

Is this check good, or is there another component (f.i.
target/configure-specific)? :
...
Index: gcc/dwarf2cfi.c
===================================================================
--- gcc/dwarf2cfi.c (revision 183070)
+++ gcc/dwarf2cfi.c (working copy)
@@ -2474,6 +2474,7 @@ scan_trace (dw_trace_info *trace)
       for (i = 1; i < n; ++i)
         {
           elt = XVECEXP (pat, 0, i);
+          gcc_assert (!(CALL_P (control) && RTX_FRAME_RELATED_P (elt)));
           scan_insn_after (elt);
         }

...

A mips64el-linux-gnu build already fails this assert at libgcc2.c, on an insn:
...
(insn 92 73 29 (sequence [
            (call_insn/j 27 73 74 (set (reg:DI 2 $2)
                    (call (mem:SI (reg/f:SI 25 $25 [205]) [0 __fixunssfdi S4
A32])
                        (unspec [
                                (const_int 0 [0])
                                (symbol_ref:SI ("__fixunssfdi") [flags 0x41]
<function_decl 0xf7dbe400 __fixunssfdi>)
                            ] UNSPEC_CALL_ATTR))) libgcc2.c:1393 571
{sibcall_value_internal}
                 (expr_list:REG_DEAD (reg:SF 44 $f12)
                    (expr_list:REG_DEAD (reg/f:SI 25 $25 [205])
                        (nil)))
                (expr_list:REG_DEP_TRUE (use (reg:SI 79 $fakec))
                    (expr_list:REG_NORETURN (use (reg:SF 44 $f12))
                        (nil))))
            (insn/f 74 27 86 (set (reg/f:SI 29 $sp)
                    (plus:SI (reg/f:SI 29 $sp)
                        (const_int 16 [0x10]))) 10 {*addsi3}
                 (expr_list:REG_CFA_DEF_CFA (reg/f:SI 29 $sp)
                    (expr_list:REG_CFA_RESTORE (reg:DI 28 $28)
                        (expr_list:REG_CFA_RESTORE (reg:DI 31 $31)
                            (nil)))))
        ]) libgcc2.c:1393 -1
     (nil))
...

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #18 from vries at gcc dot gnu.org 2012-01-28 21:00:14 UTC ---
Submitted generic patch:
http://gcc.gnu.org/ml/gcc-patches/2012-01/msg01540.html

[Bug middle-end/50283] [4.7 Regression] FAIL: g++.dg/eh/simd-1.C execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50283

--- Comment #19 from John David Anglin <danglin at gcc dot gnu.org> 2012-12-06 02:19:55 UTC ---
Author: danglin
Date: Thu Dec  6 02:19:48 2012
New Revision: 194238

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=194238
Log:
    Backport for mainline:
    2011-11-30  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>

    PR middle-end/50283
    * config/pa/pa.md (in_branch_delay): Disallow frame related insns.
    (in_nullified_branch_delay): Likewise.
    (in_call_delay): Likewise.

    2012-11-12  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>

    PR target/55195
    * config/pa/pa.md (type): Add sibcall and sh_func_adrs insn types.
    (in_branch_delay): Don't allow sibcall or sh_func_adrs insns.
    (in_nullified_branch_delay): Likewise.
    (in_call_delay): Likewise.
    Define delay for sibcall insns.  Adjust Z3 and Z4 insn reservations for
    new types.  Add opaque cond to mark all calls, sibcalls, dyncalls and
    the $$sh_func_adrs call as variable.  Update type of sibcalls and
    $$sh_func_adrs call.
    * config/pa/pa.c (pa_adjust_insn_length): Revise to return updated
    length instead of adjustment.  Handle negative and undefined call
    adjustments for insn_default_length.  Remove adjustment for millicode
    insn with unfilled delay slot.
    (pa_output_millicode_call): Update for revised millicode length.
    * config/pa/pa.h (ADJUST_INSN_LENGTH): Revise to set LENGTH.

    2012-12-05  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>

    * config/pa/pa.md: Use "const_int 0" instead of match_test to simplify
    opaque cond in all call insns.


Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/config/pa/pa.c
    branches/gcc-4_6-branch/gcc/config/pa/pa.h
    branches/gcc-4_6-branch/gcc/config/pa/pa.md