[Bug middle-end/54921] New: [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[Bug middle-end/54921] New: [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54921

             Bug #: 54921
           Summary: [4.8 Regression] wrong code with -Os
                    -fno-omit-frame-pointer -fsched2-use-superblocks
                    -fstack-protector -ftree-slp-vectorize
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: zsojka@seznam.cz


Created attachment 28441
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28441
reduced testcase

Output:
$ gcc -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector
-ftree-slp-vectorize testcase.c
$ ./a.out 
Aborted

(gdb) p a
$1 = {a = 1, b = '\000' <repeats 31 times>}
(gdb) p b
$2 = {a = 2, b = '\000' <repeats 31 times>}

Tested revisions:
r192420 - fail
4.7 r191640 - OK

[Bug middle-end/54921] [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54921

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

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

[Bug middle-end/54921] [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[mpolacek at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54921

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-10-22
                 CC|                            |mpolacek at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #1 from Marek Polacek <mpolacek at gcc dot gnu.org> 2012-10-22 13:38:20 UTC ---
Confirmed.

[Bug middle-end/54921] [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54921

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

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

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-10-23 11:21:55 UTC ---
Looks to be cselib + find_base_term related.
We have:
49 bp=sp
50 sp=sp-0x30
...
46 r8=bp-0x30
47 r9=r8+0x4
...
43 di=r9
17 {cx=0;di=cx<<0x2+di;[di]=0;use ax;...}
18 [bp-0x2c]=si

The reason for insn 46 and 47 being separate is that r8 is used later on.

When sched-deps.c asks whether there is any output_dependence in between
the insn 17 (rep_stossi) and insn 18 stores, it returns that there is not, even
when there is.  This is because the bp-0x2c store is value:2 - 0x2c, where
find_base_term (value:2) is (address:DI -4), i.e. HFP base, while when calling
find_base_term on the value for di (== r9), it uses the loc value of r8 + 4
and value of r8 has sp register among its locs (in addition to value:2 - 0x30),
so find_base_term on the value of di returns (address:DI -1), i.e. sp based,
and
thus base_alias_check returns 0.

I'm afraid if alias.c wants to argue that sp based accesses can't alias with
hfp based accesses with frame_pointer_needed, perhaps cselib.c needs to do what
I've done so far in var-tracking.c, in particular right after the assignment to
hard frame pointer in the prologue do cselib_invalidate_rtx
(stack_pointer_rtx);
so that sp based VALUEs don't have hfp anywhere in locs and similarly hfp based
VALUEs don't have sp anywhere in locs.

[Bug middle-end/54921] [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54921

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |jakub at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-11-07 13:14:17 UTC ---
http://gcc.gnu.org/ml/gcc-patches/2012-10/msg02035.html

[Bug middle-end/54921] [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54921

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-11-19 18:07:33 UTC ---
Created attachment 28735
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28735
gcc48-pr54921.patch

Untested updated patch.

[Bug middle-end/54921] [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54921

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-11-20 08:34:51 UTC ---
Author: jakub
Date: Tue Nov 20 08:34:43 2012
New Revision: 193647

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=193647
Log:
    PR rtl-optimization/54921
    * cselib.h (fp_setter_insn): New prototype.
    * cselib.c (fp_setter_insn): New function.
    (cselib_process_insn): If frame_pointer_needed,
    call cselib_invalidate_rtx (stack_pointer_rtx) after
    processing a frame pointer setter.
    * var-tracking.c (fp_setter): Removed.
    (vt_initialize): Use fp_setter_insn instead of fp_setter.

    * gcc.dg/pr54921.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/pr54921.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/cselib.c
    trunk/gcc/cselib.h
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/var-tracking.c

[Bug middle-end/54921] [4.8 Regression] wrong code with -Os -fno-omit-frame-pointer -fsched2-use-superblocks -fstack-protector -ftree-slp-vectorize

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54921

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

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

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-11-20 08:39:58 UTC ---
Fixed.