[Bug target/50655] New: Many of the new VIS2/VIS3 tests FAIL on Solaris

[Bug target/50655] New: Many of the new VIS2/VIS3 tests FAIL on Solaris

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50655

             Bug #: 50655
           Summary: Many of the new VIS2/VIS3 tests FAIL on Solaris
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ro@gcc.gnu.org
                CC: davem@davemloft.net, ebotcazou@gcc.gnu.org
              Host: sparc-sun-solaris2.1[01]
            Target: sparc-sun-solaris2.1[01]
             Build: sparc-sun-solaris2.1[01]


As of rev 179598, many of the new VIS2/VIS3 tests FAIL on Solaris 10 and 11,
both with Sun as and gas 2.21.1:

FAIL: gcc.target/sparc/cmask.c scan-assembler cmask8\\t%
FAIL: gcc.target/sparc/cmask.c scan-assembler cmask16\\t%
FAIL: gcc.target/sparc/cmask.c scan-assembler cmask32\\t%
FAIL: gcc.target/sparc/combined-1.c scan-assembler-times fandnot1\\t% 4
FAIL: gcc.target/sparc/combined-1.c scan-assembler-times for\\t% 4
FAIL: gcc.target/sparc/combined-1.c scan-assembler-times fxor\\t% 4
FAIL: gcc.target/sparc/fhalve.c scan-assembler fhadds\\t%
FAIL: gcc.target/sparc/fhalve.c scan-assembler fhaddd\\t%
FAIL: gcc.target/sparc/fhalve.c scan-assembler fhsubs\\t%
FAIL: gcc.target/sparc/fhalve.c scan-assembler fhsubd\\t%
FAIL: gcc.target/sparc/fhalve.c scan-assembler fnhadds\\t%
FAIL: gcc.target/sparc/fhalve.c scan-assembler fnhaddd\\t%
FAIL: gcc.target/sparc/fnegop.c scan-assembler fnadds\\t%
FAIL: gcc.target/sparc/fnegop.c scan-assembler fnaddd\\t%
FAIL: gcc.target/sparc/fnegop.c scan-assembler fnmuls\\t%
FAIL: gcc.target/sparc/fnegop.c scan-assembler fnmuld\\t%
FAIL: gcc.target/sparc/fnegop.c scan-assembler fnsmuld\\t%
FAIL: gcc.target/sparc/fpadds.c (test for excess errors)
UNRESOLVED: gcc.target/sparc/fpadds.c scan-assembler fpadds16\\t%
UNRESOLVED: gcc.target/sparc/fpadds.c scan-assembler fpadds16s\\t%
UNRESOLVED: gcc.target/sparc/fpadds.c scan-assembler fpsubs16\\t%
UNRESOLVED: gcc.target/sparc/fpadds.c scan-assembler fpsubs16s\\t%
UNRESOLVED: gcc.target/sparc/fpadds.c scan-assembler fpadds32\\t%
UNRESOLVED: gcc.target/sparc/fpadds.c scan-assembler fpadds32s\\t%
UNRESOLVED: gcc.target/sparc/fpadds.c scan-assembler fpsubs32\\t%
UNRESOLVED: gcc.target/sparc/fpadds.c scan-assembler fpsubs32s\\t%
FAIL: gcc.target/sparc/fshift.c (test for excess errors)
UNRESOLVED: gcc.target/sparc/fshift.c scan-assembler fsll16\\t%
UNRESOLVED: gcc.target/sparc/fshift.c scan-assembler fslas16\\t%
UNRESOLVED: gcc.target/sparc/fshift.c scan-assembler fsrl16\\t%
UNRESOLVED: gcc.target/sparc/fshift.c scan-assembler fsra16\\t%
UNRESOLVED: gcc.target/sparc/fshift.c scan-assembler fsll32\\t%
UNRESOLVED: gcc.target/sparc/fshift.c scan-assembler fslas32\\t%
UNRESOLVED: gcc.target/sparc/fshift.c scan-assembler fsrl32\\t%
UNRESOLVED: gcc.target/sparc/fshift.c scan-assembler fsra32\\t%
FAIL: gcc.target/sparc/fucmp.c scan-assembler fucmple8\\t%
FAIL: gcc.target/sparc/fucmp.c scan-assembler fucmpne8\\t%
FAIL: gcc.target/sparc/fucmp.c scan-assembler fucmpgt8\\t%
FAIL: gcc.target/sparc/fucmp.c scan-assembler fucmpeq8\\t%
FAIL: gcc.target/sparc/lzd.c scan-assembler-times lzd\\t% 3
FAIL: gcc.target/sparc/vis3misc.c (test for excess errors)
UNRESOLVED: gcc.target/sparc/vis3misc.c scan-assembler fchksm16\\t%
UNRESOLVED: gcc.target/sparc/vis3misc.c scan-assembler pdistn\\t%
UNRESOLVED: gcc.target/sparc/vis3misc.c scan-assembler fmean16\\t%
UNRESOLVED: gcc.target/sparc/vis3misc.c scan-assembler fpadd64\\t%
UNRESOLVED: gcc.target/sparc/vis3misc.c scan-assembler fpsub64\\t%
FAIL: gcc.target/sparc/xmul.c scan-assembler umulxhi\\t%
FAIL: gcc.target/sparc/xmul.c scan-assembler xmulx\\t%
FAIL: gcc.target/sparc/xmul.c scan-assembler xmulxhi\\t%

The problem seems to be the same everywhere:

$ /var/gcc/regression/trunk/11-gcc/build/gcc/xgcc
-B/var/gcc/regression/trunk/11-gcc/build/gcc/
/vol/gcc/src/hg/trunk/local/gcc/testsuite/gcc.target/sparc/cmask.c
-mcpu=niagara3 -mvis -S -o cmask.s
$ grep cmask cmask.s 
        .file   "cmask.c"
        call    __builtin_vis_cmask8, 0
        call    __builtin_vis_cmask16, 0
        call    __builtin_vis_cmask32, 0

  Rainer

[Bug target/50655] Many of the new VIS2/VIS3 tests FAIL on Solaris

[davem at davemloft dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50655

--- Comment #1 from David S. Miller <davem at davemloft dot net> 2011-10-07 15:01:55 UTC ---
Please try to figure out why the configure test is not detecting VIS3
instruction capabilities in your assembler.  That's why the VIS3 tests are
failing.

The combined-1.c test is not a VIS3 test, and is failing because of a known
register allocator issue that Eric has asked Vlad to look into.

[Bug target/50655] Many of the new VIS2/VIS3 tests FAIL on Solaris

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50655

--- Comment #2 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2011-10-07 15:16:16 UTC ---
> --- Comment #1 from David S. Miller <davem at davemloft dot net> 2011-10-07 15:01:55 UTC ---
> Please try to figure out why the configure test is not detecting VIS3
> instruction capabilities in your assembler.  That's why the VIS3 tests are
> failing.

On both Solaris 10 and 11, as complains:

configure:24069: /usr/ccs/bin/as   -xarch=v9d -o conftest.o conftest.s >&5
/usr/ccs/bin/as: "conftest.s", line 4: error: detect global register use not
cov
ered .register pseudo-op
/usr/ccs/bin/as: "conftest.s", line 4: error: detect global register use not
cov
ered .register pseudo-op
configure:24072: $? = 1

Adding

.register %g2, #scratch
.register %g3, #scratch

to the testcase allows the test to be assembled without error.

The Solaris 8 and 9 as errors out with

as: error: unknown option 'xarch=v9d'
usage: as [-V] [-Q{y,n}] [-f[O][if]#] [-q] [-s]
          [-S] [-K {pic,PIC}] [-o objfile] [-L] [-T]
          [-P [[-Ipath] [-Dname] [-Dname=def] [-Uname]]...]
          [-m [-Ym,path]] [-n] [-ul] [-xF] [-xarch=v7] [-xarch=v8] [-xarch=v8a]
[-xarch=v8plus] [-xarch=v8plusa] [-xarch=v8plusb]  [-xarch=v9] [-xarch=v9a]
[-xarch=v9b]  [-xcode={pic13,pic32}] file.s...

With gas 2.21.1, I get this failure instead:

configure:24069: /vol/gcc/bin/gas-2.21.1   -xarch=v9d -o conftest.o conftest.s
>
&5
Assembler messages:
Error: invalid architecture -xarch=v9d
Error: unrecognized option -\300v9d
configure:24072: $? = 1

Given that even the latest released gas doesn't support those insns, I
think the testsuite needs to accomodate the possibility that the
assembler used doesn't understand them.

> The combined-1.c test is not a VIS3 test, and is failing because of a known
> register allocator issue that Eric has asked Vlad to look into.

Ok, thanks.

    Rainer

[Bug target/50655] Many of the new VIS2/VIS3 tests FAIL on Solaris

[davem at davemloft dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50655

--- Comment #3 from David S. Miller <davem at davemloft dot net> 2011-10-07 16:45:52 UTC ---
Thanks, I'll add the necessary register directives and work on making the
testcases conditional on assembler support.

[Bug target/50655] Many of the new VIS2/VIS3 tests FAIL on Solaris

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50655

--- Comment #4 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2011-10-07 16:51:36 UTC ---
> --- Comment #3 from David S. Miller <davem at davemloft dot net> 2011-10-07 16:45:52 UTC ---
> Thanks, I'll add the necessary register directives and work on making the
> testcases conditional on assembler support.

Great, thanks.

       Rainer

[Bug target/50655] Many of the new VIS2/VIS3 tests FAIL on Solaris

[davem at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50655

--- Comment #5 from davem at gcc dot gnu.org 2011-10-07 17:23:53 UTC ---
Author: davem
Date: Fri Oct  7 17:23:47 2011
New Revision: 179667

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=179667
Log:
Fix VIS3 assembler check and conditionalize testsuite on VIS3 support.

gcc/

    PR 50655
    * configure.ac: Add .register directives to VIS3 test.
    * configure: Regenerate.

gcc/testsuite/

    PR 50655
    * gcc.target/sparc/sparc.exp: Add vis3 target test.
    * gcc.target/sparc/cmask.c: Use it.
    * gcc.target/sparc/fhalve.c: Likewise.
    * gcc.target/sparc/fnegop.c: Likewise.
    * gcc.target/sparc/fpadds.c: Likewise.
    * gcc.target/sparc/fshift.c: Likewise.
    * gcc.target/sparc/fucmp.c: Likewise.
    * gcc.target/sparc/lzd.c: Likewise.
    * gcc.target/sparc/vis3misc.c: Likewise.
    * gcc.target/sparc/xmul.c: Likewise.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/configure
    trunk/gcc/configure.ac
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/gcc.target/sparc/cmask.c
    trunk/gcc/testsuite/gcc.target/sparc/fhalve.c
    trunk/gcc/testsuite/gcc.target/sparc/fnegop.c
    trunk/gcc/testsuite/gcc.target/sparc/fpadds.c
    trunk/gcc/testsuite/gcc.target/sparc/fshift.c
    trunk/gcc/testsuite/gcc.target/sparc/fucmp.c
    trunk/gcc/testsuite/gcc.target/sparc/lzd.c
    trunk/gcc/testsuite/gcc.target/sparc/sparc.exp
    trunk/gcc/testsuite/gcc.target/sparc/vis3misc.c
    trunk/gcc/testsuite/gcc.target/sparc/xmul.c

[Bug target/50655] Many of the new VIS2/VIS3 tests FAIL on Solaris

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50655

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
                URL|                            |http://gcc.gnu.org/ml/gcc-p
                   |                            |atches/2011-10/msg01045.htm
                   |                            |l
         Resolution|                            |FIXED

--- Comment #6 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-10-27 13:24:46 UTC ---
.