From mboxrd@z Thu Jan  1 00:00:00 1970
MIME-Version: 1.0
In-Reply-To: <64cc61c082febe489ac3244473bb4ca4@brasstown.quanstro.net>
References: <64cc61c082febe489ac3244473bb4ca4@brasstown.quanstro.net>
Date: Sun, 22 Sep 2013 15:48:39 +1000
Message-ID: <CAJQxxw=KpdBcoC1suKHDsLbCX0mnm-pR+QNS-b11VEM2w-bOcQ@mail.gmail.com>
From: Bruce Ellis <bruce.ellis@gmail.com>
To: Fans of the OS Plan 9 from Bell Labs <9fans@9fans.net>
Content-Type: multipart/alternative; boundary=001a11c3851e9bda2004e6f2747c
Subject: Re: [9fans] userspace semlocks
Topicbox-Message-UUID: 80dcaa58-ead8-11e9-9d60-3106f5b1d025

--001a11c3851e9bda2004e6f2747c
Content-Type: text/plain; charset=UTF-8

you are missing the reason. read the paper.

brucee


On 22 September 2013 12:55, erik quanstrom <quanstro@quanstro.net> wrote:

> when i measure chan send performance with the attached program with
> the semaphore locks that have been made the default for sources and
> with the old locks, the old locks surprisingly outperform the new ones
> by a large margin.
>
> the test is let O be the number of buffers in the channel, and M be
> the number of sending procs, then cycles is the number of machine
> cycles required to send 1<<21 messages per proc, and receive them
> on a single listener.
>
> on my machine, i get the following raw numbers (averaged over a few tries):
>
>         new     1.84e9 cycles   O=10    M=1
>         old     1.10e9
>
>         new     4.61e9          O=0     M=1
>         old     4.38e9
>
>         new     1.55e10         O=10    M=8
>         old     2.74e10
>
>         new     3.64e10         O=0     M=8
>         old     5.14e10
>
> am i doing something fundamental wrong, or are the new locks substantially
> slower than the old ones?
>
> - erik

--001a11c3851e9bda2004e6f2747c
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">you are missing the reason. read the paper.<div><br></div>=
<div>brucee</div></div><div class=3D"gmail_extra"><br><br><div class=3D"gma=
il_quote">On 22 September 2013 12:55, erik quanstrom <span dir=3D"ltr">&lt;=
<a href=3D"mailto:quanstro@quanstro.net" target=3D"_blank">quanstro@quanstr=
o.net</a>&gt;</span> wrote:<br>
<blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1p=
x #ccc solid;padding-left:1ex">when i measure chan send performance with th=
e attached program with<br>
the semaphore locks that have been made the default for sources and<br>
with the old locks, the old locks surprisingly outperform the new ones<br>
by a large margin.<br>
<br>
the test is let O be the number of buffers in the channel, and M be<br>
the number of sending procs, then cycles is the number of machine<br>
cycles required to send 1&lt;&lt;21 messages per proc, and receive them<br>
on a single listener.<br>
<br>
on my machine, i get the following raw numbers (averaged over a few tries):=
<br>
<br>
=C2=A0 =C2=A0 =C2=A0 =C2=A0 new =C2=A0 =C2=A0 1.84e9 cycles =C2=A0 O=3D10 =
=C2=A0 =C2=A0M=3D1<br>
=C2=A0 =C2=A0 =C2=A0 =C2=A0 old =C2=A0 =C2=A0 1.10e9<br>
<br>
=C2=A0 =C2=A0 =C2=A0 =C2=A0 new =C2=A0 =C2=A0 4.61e9 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0O=3D0 =C2=A0 =C2=A0 M=3D1<br>
=C2=A0 =C2=A0 =C2=A0 =C2=A0 old =C2=A0 =C2=A0 4.38e9<br>
<br>
=C2=A0 =C2=A0 =C2=A0 =C2=A0 new =C2=A0 =C2=A0 1.55e10 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 O=3D10 =C2=A0 =C2=A0M=3D8<br>
=C2=A0 =C2=A0 =C2=A0 =C2=A0 old =C2=A0 =C2=A0 2.74e10<br>
<br>
=C2=A0 =C2=A0 =C2=A0 =C2=A0 new =C2=A0 =C2=A0 3.64e10 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 O=3D0 =C2=A0 =C2=A0 M=3D8<br>
=C2=A0 =C2=A0 =C2=A0 =C2=A0 old =C2=A0 =C2=A0 5.14e10<br>
<br>
am i doing something fundamental wrong, or are the new locks substantially<=
br>
slower than the old ones?<br>
<span class=3D"HOEnZb"><font color=3D"#888888"><br>
- erik</font></span></blockquote></div><br></div>

--001a11c3851e9bda2004e6f2747c--