[PATCH v4 2/2] x86_64: roundeven with sse4.1 support

H.J. Lu hjl.tools@gmail.com
Mon Jun 1 02:04:21 GMT 2020


On Sun, May 31, 2020 at 6:28 PM Cui, Lili <lili.cui@intel.com> wrote:
>
>
>
>
>
> From: H.J. Lu <hjl.tools@gmail.com>
> Sent: Friday, May 29, 2020 7:30 PM
> To: Cui, Lili <lili.cui@intel.com>
> Cc: Florian Weimer <fweimer@redhat.com>; H.J. Lu via Libc-alpha <libc-alpha@sourceware.org>; Shen-Ta Hsieh <ibmibmibm.tw@gmail.com>
> Subject: Re: [PATCH v4 2/2] x86_64: roundeven with sse4.1 support
>
>
>
> On Fri, May 29, 2020 at 1:48 AM Cui, Lili <lili.cui@intel.com> wrote:
>
>
>
>
>
> > -----Original Message-----
>
> > From: H.J. Lu <hjl.tools@gmail.com>
>
> > Sent: Thursday, May 28, 2020 8:32 PM
>
> > To: Florian Weimer <fweimer@redhat.com>; Cui, Lili <lili.cui@intel.com>
>
> > Cc: H.J. Lu via Libc-alpha <libc-alpha@sourceware.org>; Shen-Ta Hsieh
>
> > <ibmibmibm.tw@gmail.com>
>
> > Subject: Re: [PATCH v4 2/2] x86_64: roundeven with sse4.1 support
>
> >
>
> > On Thu, May 28, 2020 at 5:22 AM Florian Weimer <fweimer@redhat.com>
>
> > wrote:
>
> > >
>
> > > * H. J. Lu via Libc-alpha:
>
> > >
>
> > > >> Here is a benchmark result on my AMD Ryzen 9 3900X system:
>
> > > >
>
> > > > Since we don't know or may not care SSE4 machines without AVX,
>
> > > > should we make it to AVX only?
>
> > >
>
> > > What about Goldmont/Tremont?  Those are current CPUs which do not
>
> > > support AVX, but I think they have sufficient SSE4 support levels for
>
> > > this change.
>
> > >
>
> >
>
> > Good point.  Lili, please collect glibc micro benchmark roundeven/roundevenf
>
> > data before and after:
>
> >
>
> > https://sourceware.org/pipermail/libc-alpha/2020-May/113533.html
>
> >
>
> > on Tremont.
>
> >
>
> > --
>
> > H.J.
>
>
>
> Hi H.J,
>
>
>
> Result is here.
>
> benchmark result before this commit on Tremont
>
>
>
>
>
>
>
> benchmark result after this commit on Tremont
>
>
>
>
>
>
>
>
>
> Hi Lili,
>
>
>
> The results are empty.
>
>
>
> --
>
> H.J.
>
>
>
> Hi H.J,
>
>
>
> Sorry for that my format has some problems, data is here.
>
>
>
> benchmark result before this commit on Tremont
>
>
>
>    "roundeven":                                        "roundevenf":
>
>     "duration": 2.19422e+09,                    "duration": 2.19402e+09,
>
>     "iterations": 1.44514e+08,                  "iterations": 1.4184e+08,
>
>     "max": 43.258,                                       "max": 53.07,
>
>     "min": 11.052,                                        "min": 12.052,
>
>     "mean": 15.1835                                    "mean": 15.4683
>
>
>
> benchmark result after this commit on Tremont
>
>
>
>      "roundeven":                                   "roundevenf":
>
>      "duration": 2.19144e+09,             "duration": 2.19218e+09,
>
>      "iterations": 2.17075e+08,           "iterations": 1.97982e+08,
>
>      "max": 395.428,                             "max": 34.928,
>
>      "min": 10.044,                                "min": 11.02,
>
>      "mean": 10.0953                           "mean": 11.0726
>
>
>

Looks good.

Thanks.

-- 
H.J.


More information about the Libc-alpha mailing list