[Bug c/101985] New: vec_cpsgn parameter order

[Bug c/101985] New: vec_cpsgn parameter order

[simb611alt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

            Bug ID: 101985
           Summary: vec_cpsgn parameter order
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: simb611alt at gmail dot com
  Target Milestone: ---

I have found an issue where the parameter order of the vec_cpsgn function on
POWER is reversed. Clang has recently fixed the same issue in their code which
leaves only GCC using the reversed order.
XLC always used the order clang currently uses.

Patch link for clang -
https://github.com/llvm/llvm-project/commit/1af037f643fc5499f83d92e5aec199950871d475

Link to POWER documentation -
http://cdn.openpowerfoundation.org/wp-content/uploads/resources/Intrinsics-Reference/content/VIPR.reference.vecfns.html

#include <stdio.h>
#include <altivec.h>

int main() {
    vector float a = {  1,  2, - 3, - 4};
    vector float b = {-10, 20, -30,  40};

    a = vec_cpsgn(a, b);
    printf("%f %f %f %f\n", a[0], a[1], a[2], a[3]);
    return 0;
}

Output for GCC 11.0.1:
-1.000000 2.000000 -3.000000 4.000000

Output for clang 12.0.1:
10.000000 20.000000 -30.000000 -40.000000

Output for xlc 16.1.1:
10.000000 20.000000 -30.000000 -40.000000

[Bug target/101985] vec_cpsgn parameter order

[wschmidt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

Bill Schmidt <wschmidt at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
           Assignee|unassigned at gcc dot gnu.org      |wschmidt at gcc dot gnu.org
   Last reconfirmed|                            |2021-09-23

--- Comment #1 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
Confirmed.  I'll take a look.

[Bug target/101985] vec_cpsgn parameter order

[wschmidt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

--- Comment #2 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
Patch posted at
https://gcc.gnu.org/pipermail/gcc-patches/2021-September/580235.html.

[Bug target/101985] vec_cpsgn parameter order

[wschmidt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

--- Comment #3 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
Kunwar, can you please tell us (if you don't mind) where the problem was
detected?  Since we're changing behavior of the intrinsic, we'll need to
document this, and knowing whether we have problematic code in the wild that
relies on present behavior would be helpful.

[Bug target/101985] vec_cpsgn parameter order

[simb611alt at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

--- Comment #4 from Kunwar Maheep Singh <simb611alt at gmail dot com> ---
(In reply to Bill Schmidt from comment #3)
> Kunwar, can you please tell us (if you don't mind) where the problem was
> detected?  Since we're changing behavior of the intrinsic, we'll need to
> document this, and knowing whether we have problematic code in the wild that
> relies on present behavior would be helpful.

I detected it in SIMDe (https://github.com/simd-everywhere/simde).
Relevant commit -
https://github.com/simd-everywhere/simde/commit/1ba15969fdc1b302c0e50084420412368a14cd96

[Bug target/101985] vec_cpsgn parameter order

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

--- Comment #5 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by William Schmidt <wschmidt@gcc.gnu.org>:

https://gcc.gnu.org/g:76ba473b99c30ddec6171840a76292d6d4b67e7c

commit r12-4361-g76ba473b99c30ddec6171840a76292d6d4b67e7c
Author: Bill Schmidt <wschmidt@linux.ibm.com>
Date:   Tue Oct 12 17:37:16 2021 -0500

    rs6000: Fix vec_cpsgn parameter order (PR101985)

    The vec_cpsgn built-in function API differs in argument order from the
    copysign<mode>3 convention.  Currently that pattern is incorrctly used to
    implement vec_cpsgn.  Fix that by reversing the operand order of the
    builtin while leaving the existing pattern in place to implement copysignf
    for vector modes.

    Part of the fix when using the new built-in support requires an adjustment
    to a pending patch that replaces much of altivec.h with an automatically
    generated file.

    Also fix a bug in the new built-in overload infrastructure where we were
    using the VSX form of the VEC_COPYSIGN built-in when we should default to
    the VMX form.

    2021-10-12  Bill Schmidt  <wschmidt@linux.ibm.com>

    gcc/
            PR target/101985
            * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
            * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
            avoid generating an automatic #define of vec_cpsgn.  Use the
            correct built-in for V4SFmode that doesn't depend on VSX.

    gcc/testsuite/
            PR target/101985
            * gcc.target/powerpc/pr101985-1.c: New.
            * gcc.target/powerpc/pr101985-2.c: New.

[Bug target/101985] vec_cpsgn parameter order

[wschmidt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

--- Comment #6 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
Currently fixed on trunk.  The parts of the fix that don't involve the new
builtin infrastructure should be backported to all releases after some burn-in
time.

[Bug target/101985] vec_cpsgn parameter order

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-11 branch has been updated by William Schmidt
<wschmidt@gcc.gnu.org>:

https://gcc.gnu.org/g:8fb9bf6667db0c9bac4145fcc02280b8d3133a27

commit r11-9272-g8fb9bf6667db0c9bac4145fcc02280b8d3133a27
Author: Bill Schmidt <wschmidt@linux.ibm.com>
Date:   Tue Nov 23 14:33:18 2021 -0600

    rs6000: Fix vec_cpsgn parameter order (PR101985)

    Backport r12-4361 to releases/gcc-11.

    2021-11-23  Bill Schmidt  <wschmidt@linux.ibm.com>

    gcc/
            PR target/101985
            * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.

    gcc/testsuite/
            PR target/101985
            * gcc.target/powerpc/pr101985-1.c: New.
            * gcc.target/powerpc/pr101985-2.c: New.

[Bug target/101985] vec_cpsgn parameter order

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by William Schmidt
<wschmidt@gcc.gnu.org>:

https://gcc.gnu.org/g:049aa656393299d2ff23f0292c6f49e5057e2909

commit r10-10289-g049aa656393299d2ff23f0292c6f49e5057e2909
Author: Bill Schmidt <wschmidt@linux.ibm.com>
Date:   Tue Nov 23 15:57:27 2021 -0600

    rs6000: Fix vec_cpsgn parameter order (PR101985)

    Backport r12-4361 to releases/gcc-10.

    2021-11-23  Bill Schmidt  <wschmidt@linux.ibm.com>

    gcc/
            PR target/101985
            * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.

    gcc/testsuite/
            PR target/101985
            * gcc.target/powerpc/pr101985-1.c: New.
            * gcc.target/powerpc/pr101985-2.c: New.

[Bug target/101985] vec_cpsgn parameter order

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-9 branch has been updated by William Schmidt
<wschmidt@gcc.gnu.org>:

https://gcc.gnu.org/g:4326a1cd664e56167ed23de8beecfe250f466fcb

commit r9-9839-g4326a1cd664e56167ed23de8beecfe250f466fcb
Author: Bill Schmidt <wschmidt@linux.ibm.com>
Date:   Tue Nov 23 16:14:01 2021 -0600

    rs6000: Fix vec_cpsgn parameter order (PR101985)

    Backport r12-4361 to releases/gcc-9.

    2021-11-23  Bill Schmidt  <wschmidt@linux.ibm.com>

    gcc/
            PR target/101985
            * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.

    gcc/testsuite/
            PR target/101985
            * gcc.target/powerpc/pr101985-1.c: New.
            * gcc.target/powerpc/pr101985-2.c: New.

[Bug target/101985] vec_cpsgn parameter order

[wschmidt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

Bill Schmidt <wschmidt at gcc dot gnu.org> changed:

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

--- Comment #10 from Bill Schmidt <wschmidt at gcc dot gnu.org> ---
Now fixed everywhere.

[Bug target/101985] vec_cpsgn parameter order

[mark.j.abraham at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101985

Mark Abraham <mark.j.abraham at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mark.j.abraham at gmail dot com

--- Comment #11 from Mark Abraham <mark.j.abraham at gmail dot com> ---
Also discovered in the SIMD compatibility layer for GROMACS
https://gitlab.com/gromacs/gromacs/-/issues/4661