[Bug libgomp/45838] New: FAIL: libgomp.c/pr34513.c execution test

[Bug libgomp/45838] New: FAIL: libgomp.c/pr34513.c execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

           Summary: FAIL: libgomp.c/pr34513.c execution test
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libgomp
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: danglin@gcc.gnu.org
              Host: hppa-unknown-linux-gnu
            Target: hppa-unknown-linux-gnu
             Build: hppa-unknown-linux-gnu


New libgomp fails:

FAIL: libgomp.c/pr34513.c execution test
FAIL: libgomp.c++/pr34513.C  -O1  execution test
FAIL: libgomp.c++/pr34513.C  -O2  execution test
FAIL: libgomp.c++/pr34513.C  -O3 -fomit-frame-pointer  execution test
FAIL: libgomp.c++/pr34513.C  -O3 -fomit-frame-pointer -funroll-loops  execution
test
FAIL: libgomp.c++/pr34513.C  -O3 -fomit-frame-pointer -funroll-all-loops
-finline-functions  execution test
FAIL: libgomp.c++/pr34513.C  -O3 -g  execution test
FAIL: libgomp.c++/pr34513.C  -Os  execution test

First seen in revision 164706.  164639 was ok.

[Bug libgomp/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|hppa-unknown-linux-gnu      |hppa*-*-*
               Host|hppa-unknown-linux-gnu      |hppa*-*-*
            Summary|FAIL: libgomp.c/pr34513.c   |[4.6 Regression] FAIL:
                   |execution test              |libgomp.c/pr34513.c
                   |                            |execution test
              Build|hppa-unknown-linux-gnu      |hppa*-*-*

--- Comment #1 from John David Anglin <danglin at gcc dot gnu.org> 2010-10-04 22:07:52 UTC ---
(gdb) r
Starting program:
/home/dave/gnu/gcc-4.6/objdir/hppa-linux/libgomp/testsuite/pr34513.xg 
[Thread debugging using libthread_db enabled]
[New Thread 0x41153440 (LWP 14854)]
[New Thread 0x41953440 (LWP 14855)]
[New Thread 0x42153440 (LWP 14856)]

Program received signal SIGABRT, Aborted.
0x4076487c in raise (sig=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:67
67    ../nptl/sysdeps/unix/sysv/linux/raise.c: No such file or directory.
    in ../nptl/sysdeps/unix/sysv/linux/raise.c
(gdb) bt
#0  0x4076487c in raise (sig=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:67
#1  0x40769228 in abort () at abort.c:92
#2  0x00010788 in main ()
    at /home/dave/gnu/gcc/gcc/libgomp/testsuite/libgomp.c/pr34513.c:31

errors = 3

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|libgomp                     |middle-end

--- Comment #2 from John David Anglin <danglin at gcc dot gnu.org> 2010-10-05 13:29:02 UTC ---
Introduced in revision 164689.

2010-09-28  Jan Hubicka  <jh@suse.cz>

        * builtin-attrs.def (ATTR_LEAF): New attribute.
        (ATTR_NOVOPS_LEAF_LIST, ATTR_LEAF_LIST, ATTR_NOTHROW_LEAF_LIST,
        ATTR_CONST_NOTHROW_LEAF_LIST, ATTR_PURE_NOTHROW_LEAF_LIST,
        ATTR_PURE_NOTHROW_NOVOPS_LEAF_LIST, ATTR_NORETURN_NOTHROW_LEAF_LIST,
        ATTR_MALLOC_NOTHROW_LEAF_LIST, ATTR_SENTINEL_NOTHROW_LEAF_LIST,
        ATTR_NOTHROW_NONNULL_LEAF, ATTR_CONST_NOTHROW_NONNULL_LEAF,
        ATTR_CONST_NOTHROW_TYPEGENERIC_LEAF, ATTR_PURE_NOTHROW_NONNULL_LEAF,
        ATTR_MALLOC_NOTHROW_NONNULL_LEAF): New attribute lists.
        * sync-builtins.def: Annotate all builtins by leaf.
        * omp-builtins.def: Annotate all builtins by leaf.
        * builtins.def: Annotate relevant builtins with leaf attribute.
        (ATTR_MATHFN_ERRNO, ATTR_MATHFN_FPROUNDING,
        ATTR_MATHFN_FPROUNDING_ERRNO, ATTR_MATHFN_FPROUNDING_STORE): Make
        leaf.

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[dave at hiauly1 dot hia.nrc.ca]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

--- Comment #4 from dave at hiauly1 dot hia.nrc.ca 2010-10-05 15:19:09 UTC ---
> Untested patch.  Several of the GOMP_* builtins are certainly not leaf.

Another issue may be that the sync builtins on hppa-unknown-linux-gnu are
implemented in a library.  This is also the case for some arm targets.
There are no sync builtins on hppa-hpux.

Will try patch.

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[dave at hiauly1 dot hia.nrc.ca]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

--- Comment #5 from dave at hiauly1 dot hia.nrc.ca 2010-10-05 18:14:57 UTC ---
> Untested patch.  Several of the GOMP_* builtins are certainly not leaf.

I did a quick test.  Unfortunately, libgomp.c/pr34513.c still fails.

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-10-06 07:47:32 UTC ---
Author: jakub
Date: Wed Oct  6 07:47:28 2010
New Revision: 165012

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=165012
Log:
    PR middle-end/45838
    * omp-builtins.def (BUILT_IN_GOMP_PARALLEL_LOOP_STATIC_START,
    BUILT_IN_GOMP_PARALLEL_LOOP_DYNAMIC_START,
    BUILT_IN_GOMP_PARALLEL_LOOP_GUIDED_START,
    BUILT_IN_GOMP_PARALLEL_LOOP_RUNTIME_START,
    BUILT_IN_GOMP_PARALLEL_START, BUILT_IN_GOMP_TASK,
    BUILT_IN_GOMP_PARALLEL_SECTIONS_START): Use
    Revert back to ATTR_NOTHROW_LIST instead of ATTR_NOTHROW_LEAF_LIST.

    * f95-lang.c (ATTR_NOTHROW_LEAF_LIST, ATTR_CONST_NOTHROW_LEAF_LIST,
    ATTR_NOTHROW_LIST, ATTR_CONST_NOTHROW_LIST): Define.
    (gfc_define_builtin): Change last argument to int bitmask from bool,
    control addition of TREE_NOTHROW and leaf attribute as well.
    (DO_DEFINE_MATH_BUILTIN): Adjust callers.
    (gfc_init_builtin_functions): Likewise.  Remove
    ATTR_{,CONST_}NOTHROW_LIST enum.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/f95-lang.c
    trunk/gcc/omp-builtins.def

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.6.0

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2010.11.03 15:47:43
     Ever Confirmed|0                           |1

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-03 15:47:43 UTC ---
That surprises me.  Can you try say -O1 -fopenmp testcase build before the
commit and after it (with the follow-up fixes backported), look at
-fdump-tree-optimized
dumps and see the differences in between (or in some late RTL dump?

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |ASSIGNED

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-16 12:13:53 UTC ---
Ah, hppa uses GOMP_atomic_{start,end}, and the leaf attribute makes fre think
it is ok to cache static var values for non-addressable vars across calls.  In
OpenMP, as they can be accessed by multiple threads, they can't be cached
across barrier points.  So I'll jus remove the leaf attribute from most of the
GOMP builtins that are barriers (and conservatively the starts too).  I've left
out just the omp_* functions and *_nowait.

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-16 12:16:52 UTC ---
Created attachment 22419
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22419
gcc46-pr45838.patch

Untested fix.

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #22419|0                           |1
        is obsolete|                            |

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-16 12:47:31 UTC ---
Created attachment 22420
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22420
gcc46-pr45838.patch

Actually, that's just too conservative, the *start function except
*parallel*start* and single_copy_start aren't barriers of any kind nor can call
functions from the current TU, so those should be okay.

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-16 13:37:28 UTC ---
Created attachment 22421
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22421
gcc46-pr45838.patch

Another, alternate, untested fix.

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

--- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-17 11:15:59 UTC ---
Author: jakub
Date: Wed Nov 17 11:15:55 2010
New Revision: 166863

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=166863
Log:
    PR middle-end/45838
    * tree-ssa-alias.c (ref_maybe_used_by_call_p_1,
    call_may_clobber_ref_p_1): Return true for __sync_* and some
    OpenMP builtins that act as threading barriers.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-ssa-alias.c

[Bug middle-end/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

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

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

--- Comment #13 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-17 11:18:18 UTC ---
Fixed.

[Bug libgomp/45838] [4.6 Regression] FAIL: libgomp.c/pr34513.c execution test

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45838

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2010-11-03 15:47:43         |
          Component|middle-end                  |libgomp
         AssignedTo|jakub at gcc dot gnu.org    |unassigned at gcc dot
                   |                            |gnu.org
   Target Milestone|4.6.0                       |---

--- Comment #14 from John David Anglin <danglin at gcc dot gnu.org> 2010-11-23 23:47:56 UTC ---
Resolved.