[Bug tree-optimization/46312] New: gcc.dg/vec-scal-opt2.c fails for ARM targets.

[Bug tree-optimization/46312] New: gcc.dg/vec-scal-opt2.c fails for ARM targets.

[ramana at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

           Summary: gcc.dg/vec-scal-opt2.c fails for ARM targets.
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ramana@gcc.gnu.org
                CC: rguenth@gcc.gnu.org
              Host: x86_64-unknown-linux-gnu
            Target: arm-linux-gnueabi
             Build: x86_64-unknown-linux-gnu


Hi, 

This test fails for ARM targets configured in the following manner -


/home/ramana/cross-build/src/trunk/configure --target=arm-none-linux-gnueabi
--with-cpu=cortex-a9 --with-fpu=neon --with-float=softfp 

The dump file doesn't contain << 2 and hence the test fails. 

This fails right from : r166303 for ARM


Author: rguenth
Date: Thu Nov  4 10:54:33 2010
New Revision: 166303

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=166303
Log:
2010-11-04  Artjoms Sinkarovs <artyom.shinakroff@gmail.com>
    Richard Guenther  <rguenther@suse.de>

    * passes.c (init_optimization_passes): Move veclower after SSA.
    * tree-vect-generic.c (uniform_vector_p): New function.
    (expand_vector_operations_1): Handle conversion.
    (pass_lower_vector): Update SSA form.

    testsuite/
    * gcc.dg/vec-scal-opt.c: New testcase.
    * gcc.dg/vec-scal-opt1.c: New testcase.
    * gcc.dg/vec-scal-opt2.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/vec-scal-opt.c
    trunk/gcc/testsuite/gcc.dg/vec-scal-opt1.c
    trunk/gcc/testsuite/gcc.dg/vec-scal-opt2.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/passes.c
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-vect-generic.c


cheers
Ramana

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2010.11.05 11:29:41
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-05 11:29:41 UTC ---
Hm, of course.  Does adding

/* { dg-require-effective-target vect_shift } */

make it unsupported?  (I see that would be true for
check_effective_target_arm32,
so, does gcc.dg/vect/vect-shift-1.c also fail?

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

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> 2010-11-05 11:39:19 UTC ---
It fails even on i686-linux:
Executing on host: /usr/src/gcc/obj333/gcc/xgcc -B/usr/src/gcc/obj333/gcc/
/usr/src/gcc/gcc/testsuite/gcc.dg/vec-scal-opt.c   -O -fdump-tree-veclower -S 
-o vec-scal-opt.s    (timeout = 300)
PASS: gcc.dg/vec-scal-opt.c (test for excess errors)
FAIL: gcc.dg/vec-scal-opt.c scan-tree-dump-times veclower ">> k.[0-9_]*" 1
Executing on host: /usr/src/gcc/obj333/gcc/xgcc -B/usr/src/gcc/obj333/gcc/
/usr/src/gcc/gcc/testsuite/gcc.dg/vec-scal-opt1.c   -O -fdump-tree-veclower -S 
-o vec-scal-opt1.s    (timeout = 300)
PASS: gcc.dg/vec-scal-opt1.c (test for excess errors)
FAIL: gcc.dg/vec-scal-opt1.c scan-tree-dump-times veclower ">> 2" 1
Executing on host: /usr/src/gcc/obj333/gcc/xgcc -B/usr/src/gcc/obj333/gcc/
/usr/src/gcc/gcc/testsuite/gcc.dg/vec-scal-opt2.c   -O -fdump-tree-veclower -S 
-o vec-scal-opt2.s    (timeout = 300)
PASS: gcc.dg/vec-scal-opt2.c (test for excess errors)
FAIL: gcc.dg/vec-scal-opt2.c scan-tree-dump-times veclower ">> 2" 1
Executing on host: /usr/src/gcc/obj333/gcc/xgcc -B/usr/src/gcc/obj333/gcc/
/usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c    -S  -o vector-1.s    (timeout =
300)
/usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c: In function 'f':
/usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c:13:7: error: wrong type argument
to bit-complement
compiler exited with status 1
output is:
/usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c: In function 'f':
/usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c:13:7: error: wrong type argument
to bit-complement

I guess the bug is that the testcase should be put into gcc.dg/vect/ rather
than gcc.dg/ directly.

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[ramana.r at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

--- Comment #3 from Ramana Radhakrishnan <ramana.r at gmail dot com> 2010-11-05 11:41:40 UTC ---
> --- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-05 11:29:41 UTC ---
> Hm, of course.  Does adding
>
> /* { dg-require-effective-target vect_shift } */
>
> make it unsupported?  (I see that would be true for
> check_effective_target_arm32,

Isn't that true for arm32 and hence making that check would be a no-op
won't it ?  We do have vector shifts for Neon.


> so, does gcc.dg/vect/vect-shift-1.c also fail?

I begin to see an execute failure with gcc.dg/vect/vect-shift-1.c

gcc.dg/vect-scalop-2.c and gcc.dg/vect-scalop.c also appear to fail similarly .

Ramana

From the latest test results gcc.dg/vect/vect-shift-1.c also appears
to have an execution failure.







>
> --
> Configure bugmail: http://gcc.gnu.org/bugzilla/userprefs.cgi?tab=email
> ------- You are receiving this mail because: -------
> You reported the bug.
>

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[rguenther at suse dot de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

--- Comment #4 from rguenther at suse dot de <rguenther at suse dot de> 2010-11-05 11:43:48 UTC ---
On Fri, 5 Nov 2010, jakub at gcc dot gnu.org wrote:

> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312
> 
> 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> 2010-11-05 11:39:19 UTC ---
> It fails even on i686-linux:

Hm, it tested ok on x86_64 with -m32.

> Executing on host: /usr/src/gcc/obj333/gcc/xgcc -B/usr/src/gcc/obj333/gcc/
> /usr/src/gcc/gcc/testsuite/gcc.dg/vec-scal-opt.c   -O -fdump-tree-veclower -S 
> -o vec-scal-opt.s    (timeout = 300)
> PASS: gcc.dg/vec-scal-opt.c (test for excess errors)
> FAIL: gcc.dg/vec-scal-opt.c scan-tree-dump-times veclower ">> k.[0-9_]*" 1
> Executing on host: /usr/src/gcc/obj333/gcc/xgcc -B/usr/src/gcc/obj333/gcc/
> /usr/src/gcc/gcc/testsuite/gcc.dg/vec-scal-opt1.c   -O -fdump-tree-veclower -S 
> -o vec-scal-opt1.s    (timeout = 300)
> PASS: gcc.dg/vec-scal-opt1.c (test for excess errors)
> FAIL: gcc.dg/vec-scal-opt1.c scan-tree-dump-times veclower ">> 2" 1
> Executing on host: /usr/src/gcc/obj333/gcc/xgcc -B/usr/src/gcc/obj333/gcc/
> /usr/src/gcc/gcc/testsuite/gcc.dg/vec-scal-opt2.c   -O -fdump-tree-veclower -S 
> -o vec-scal-opt2.s    (timeout = 300)
> PASS: gcc.dg/vec-scal-opt2.c (test for excess errors)
> FAIL: gcc.dg/vec-scal-opt2.c scan-tree-dump-times veclower ">> 2" 1
> Executing on host: /usr/src/gcc/obj333/gcc/xgcc -B/usr/src/gcc/obj333/gcc/
> /usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c    -S  -o vector-1.s    (timeout =
> 300)
> /usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c: In function 'f':
> /usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c:13:7: error: wrong type argument
> to bit-complement
> compiler exited with status 1
> output is:
> /usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c: In function 'f':
> /usr/src/gcc/gcc/testsuite/gcc.dg/vector-1.c:13:7: error: wrong type argument
> to bit-complement
> 
> I guess the bug is that the testcase should be put into gcc.dg/vect/ rather
> than gcc.dg/ directly.

But that enables -ftree-vectorize, but ok - I guess that doesn't matter
too much.

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

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

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

--- Comment #5 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-09 10:41:28 UTC ---
*** Bug 46380 has been marked as a duplicate of this bug. ***

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

--- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-09 13:45:54 UTC ---
Author: rguenth
Date: Tue Nov  9 13:45:45 2010
New Revision: 166489

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=166489
Log:
2010-11-09  Richard Guenther  <rguenther@suse.de>

    PR testsuite/46312
    * gcc.dg/vec-scal-opt.c: Move ...
    * gcc.dg/vect/vect-scal-opt.c: ... here.
    * gcc.dg/vec-scal-opt1.c: Move ...
    * gcc.dg/vect/vect-scal-opt1.c: ... here.
    * gcc.dg/vec-scal-opt2.c: Move ...
    * gcc.dg/vect/vect-scal-opt2.c: ... here.

Added:
    trunk/gcc/testsuite/gcc.dg/vect/vect-scal-opt.c
      - copied, changed from r166486, trunk/gcc/testsuite/gcc.dg/vec-scal-opt.c
    trunk/gcc/testsuite/gcc.dg/vect/vect-scal-opt1.c
      - copied, changed from r166486,
trunk/gcc/testsuite/gcc.dg/vec-scal-opt1.c
    trunk/gcc/testsuite/gcc.dg/vect/vect-scal-opt2.c
      - copied, changed from r166486,
trunk/gcc/testsuite/gcc.dg/vec-scal-opt2.c
Removed:
    trunk/gcc/testsuite/gcc.dg/vec-scal-opt.c
    trunk/gcc/testsuite/gcc.dg/vec-scal-opt1.c
    trunk/gcc/testsuite/gcc.dg/vec-scal-opt2.c
Modified:
    trunk/gcc/testsuite/ChangeLog

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.6.0

--- Comment #7 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-09 13:46:40 UTC ---
Fixed.

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[ramana at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

Ramana Radhakrishnan <ramana at gcc dot gnu.org> changed:

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

--- Comment #8 from Ramana Radhakrishnan <ramana at gcc dot gnu.org> 2010-11-12 10:51:06 UTC ---
So this gets fixed by moving the testcase to a different directory for x86_32 ? 

FAIL: gcc.dg/vect/vect-scal-opt.c scan-tree-dump-times veclower ">> k.[0-9_]*"
1
FAIL: gcc.dg/vect/vect-scal-opt1.c scan-tree-dump-times veclower ">> 2" 1
FAIL: gcc.dg/vect/vect-scal-opt2.c scan-tree-dump-times veclower ">> 2" 1

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[ramana at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

--- Comment #9 from Ramana Radhakrishnan <ramana at gcc dot gnu.org> 2010-11-12 10:51:58 UTC ---
(In reply to comment #8)
> So this gets fixed by moving the testcase to a different directory for x86_32 ? 
> 
> FAIL: gcc.dg/vect/vect-scal-opt.c scan-tree-dump-times veclower ">> k.[0-9_]*"
> 1
> FAIL: gcc.dg/vect/vect-scal-opt1.c scan-tree-dump-times veclower ">> 2" 1
> FAIL: gcc.dg/vect/vect-scal-opt2.c scan-tree-dump-times veclower ">> 2" 1

Oh and those are the latest results on ARM.

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-12 11:50:53 UTC ---
No, it isn't unfortunately fixed anywhere.

vect.exp unfortunately only puts the correct options as default options (if no
dg-options is provided).

I guess either the testcase should be limited to x86_64-linux (or other targets
where vectorization is always possible without special flags), or needs a list
of special dg-options for each of the interesting targets, or we need some
*.exp magic somewhere.

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[irar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

--- Comment #11 from irar at gcc dot gnu.org 2010-11-17 10:00:34 UTC ---
Author: irar
Date: Wed Nov 17 10:00:27 2010
New Revision: 166860

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=166860
Log:

        PR tree-optimization/46312
        * gcc.dg/vect/vect-scal-opt2.c: Rename to...
        * gcc.dg/vect/vec-scal-opt2.c: ... this. Expect to pass only
        on targets that support vector shifts with scalar shift
        argument.
        Remove dg-options.
        * gcc.dg/vect/vect-scal-opt.c: Rename to...
        * gcc.dg/vect/vec-scal-opt.c: ... this, and change likewise.
        * gcc.dg/vect/vect-scal-opt1.c: Rename to...
        * gcc.dg/vect/vec-scal-opt1.c: ... this, and change likewise.
        * gcc.dg/vect/vect.exp: Run vec-scal-* tests with relevant
        flags.
        * gcc.dg/vect/vect-114.c: Expect to pass only on targets that
        support misaligned data accesses.
        * gcc.dg/vect/pr43432.c, gcc.dg/vect/vect-15.c: Likewise.
        * gcc.dg/vect/pr43430-2.c: Remove dg-options.
        * lib/target-supports.exp
        (check_effective_target_vect_shift_scalar): New.


Added:
    trunk/gcc/testsuite/gcc.dg/vect/vec-scal-opt.c
    trunk/gcc/testsuite/gcc.dg/vect/vec-scal-opt1.c
    trunk/gcc/testsuite/gcc.dg/vect/vec-scal-opt2.c
Removed:
    trunk/gcc/testsuite/gcc.dg/vect/vect-scal-opt.c
    trunk/gcc/testsuite/gcc.dg/vect/vect-scal-opt1.c
    trunk/gcc/testsuite/gcc.dg/vect/vect-scal-opt2.c
Modified:
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/gcc.dg/vect/pr43430-2.c
    trunk/gcc/testsuite/gcc.dg/vect/pr43432.c
    trunk/gcc/testsuite/gcc.dg/vect/vect-114.c
    trunk/gcc/testsuite/gcc.dg/vect/vect-15.c
    trunk/gcc/testsuite/gcc.dg/vect/vect.exp
    trunk/gcc/testsuite/lib/target-supports.exp

[Bug tree-optimization/46312] gcc.dg/vec-scal-opt2.c fails for ARM targets.

[irar at il dot ibm.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46312

Ira Rosen <irar at il dot ibm.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |RESOLVED
                 CC|                            |irar at il dot ibm.com
         Resolution|                            |FIXED

--- Comment #12 from Ira Rosen <irar at il dot ibm.com> 2010-11-21 08:01:08 UTC ---
Fixed.