Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 10 Jun 2024 16:31:02 -0700
From:      Ryan Libby <rlibby@gmail.com>
To:        Ed Maste <emaste@freebsd.org>, kargl@freebsd.org
Cc:        src-committers@freebsd.org, dev-commits-src-all@freebsd.org,  dev-commits-src-main@freebsd.org
Subject:   Re: git: 888796ade284 - main - libm: fma: correct zero sign with small inputs
Message-ID:  <CAHgpiFyTM-cWgyStKa4Ln7NNv84DtxjJw0sAP_-VynPvCe-=Ag@mail.gmail.com>
In-Reply-To: <202406081556.458FuJYN022360@gitrepo.freebsd.org>
References:  <202406081556.458FuJYN022360@gitrepo.freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Jun 8, 2024 at 8:56=E2=80=AFAM Ed Maste <emaste@freebsd.org> wrote:
>
> The branch main has been updated by emaste:
>
> URL: https://cgit.FreeBSD.org/src/commit/?id=3D888796ade2842486d3167067e8=
034254c38aadd3
>
> commit 888796ade2842486d3167067e8034254c38aadd3
> Author:     Ed Maste <emaste@FreeBSD.org>
> AuthorDate: 2024-03-19 14:31:39 +0000
> Commit:     Ed Maste <emaste@FreeBSD.org>
> CommitDate: 2024-06-08 15:55:36 +0000
>
>     libm: fma: correct zero sign with small inputs
>
>     PR:             277783
>     Reported by:    Victor Stinner
>     Submitted by:   kargl
>     MFC after:      1 week
>     Differential Revision: https://reviews.freebsd.org/D44433
> ---
>  lib/msun/src/s_fma.c  | 4 +++-
>  lib/msun/src/s_fmal.c | 4 +++-
>  2 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/lib/msun/src/s_fma.c b/lib/msun/src/s_fma.c
> index b8a342646d85..4d08b40cc71a 100644
> --- a/lib/msun/src/s_fma.c
> +++ b/lib/msun/src/s_fma.c
> @@ -267,7 +267,9 @@ fma(double x, double y, double z)
>                  */
>                 fesetround(oround);
>                 volatile double vzs =3D zs; /* XXX gcc CSE bug workaround=
 */
> -               return (xy.hi + vzs + ldexp(xy.lo, spread));
> +               xs =3D ldexp(xy.lo, spread);
> +               xy.hi +=3D vzs;
> +               return (xy.hi =3D=3D 0 ? xs : xy.hi + xs);
>         }
>
>         if (oround !=3D FE_TONEAREST) {
> diff --git a/lib/msun/src/s_fmal.c b/lib/msun/src/s_fmal.c
> index 3d333632127c..12f9c364670b 100644
> --- a/lib/msun/src/s_fmal.c
> +++ b/lib/msun/src/s_fmal.c
> @@ -248,7 +248,9 @@ fmal(long double x, long double y, long double z)
>                  */
>                 fesetround(oround);
>                 volatile long double vzs =3D zs; /* XXX gcc CSE bug worka=
round */
> -               return (xy.hi + vzs + ldexpl(xy.lo, spread));
> +               xs =3D ldexpl(xy.lo, spread);
> +               xy.hi +=3D vzs;
> +               return (xy.hi =3D=3D 0 ? xs : xy.hi + xs);
>         }
>
>         if (oround !=3D FE_TONEAREST) {

This seems to have caused the lib/msun/fma_tests:zeroes test to fail in
the FE_ROUNDDOWN mode on amd64, now finding 0 while expecting -0.  I
don't know if the test is wrong or too strict, or if the new result is
wrong.

https://ci.freebsd.org/job/FreeBSD-main-amd64-test/25249/testReport/junit/l=
ib.msun/fma_test/zeroes/

Reproduces with
kyua debug -k /usr/tests/Kyuafile lib/msun/fma_test:zeroes

Ryan



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAHgpiFyTM-cWgyStKa4Ln7NNv84DtxjJw0sAP_-VynPvCe-=Ag>