Date: Sun, 19 Aug 2007 09:57:50 +0200 From: Pawel Jakub Dawidek <pjd@FreeBSD.org> To: Jeff Roberson <jroberson@chesapeake.net> Cc: freebsd-arch@freebsd.org Subject: Re: Lockless uidinfo. Message-ID: <20070819075750.GB11792@garage.freebsd.pl> In-Reply-To: <20070818163503.T568@10.0.0.1> References: <20070818120056.GA6498@garage.freebsd.pl> <20070818220756.GH6498@garage.freebsd.pl> <20070818230917.GI6498@garage.freebsd.pl> <20070818163503.T568@10.0.0.1>
next in thread | previous in thread | raw e-mail | index | archive | help
--0eh6TmSyL6TZE2Uz
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Sat, Aug 18, 2007 at 04:35:42PM -0700, Jeff Roberson wrote:
> On Sun, 19 Aug 2007, Pawel Jakub Dawidek wrote:
> >Ok, after implementing atomic_fetchadd_long() on amd64, we get additional
> >6% of performance improvement:
> >
> >x ./uidinfo_lockfree.txt (atomic_cmpset_long loop)
> >+ ./uidinfo_waitfree.txt (atomic_fetchadd_long)
> >+-----------------------------------------------------------------------=
-------+
> >| =
=20
> >+|
> >| =
=20
> >+|
> >|x xx xx =
=20
> >+ ++|
> >| |__MA___| =
=20
> >|AM|
> >+-----------------------------------------------------------------------=
-------+
> > N Min Max Median Avg Stdd=
ev
> >x 5 1561566 1575987 1568964 1569767 5853.1=
399
> >+ 5 1662362 1665936 1665810 1664881.8 1541.2=
693
> >Difference at 95.0% confidence
> > 95114.8 +/- 6241.96
> > 6.05917% +/- 0.397636%
> > (Student's t, pooled s =3D 4279.88)
>=20
> How does this effect the single-threaded performance? Do you attribute=
=20
> this to atomic fetchadd being cheaper than atomic cmpset? What is your=
=20
> processor?
CPU: Intel(R) Xeon(R) CPU E5310 @ 1.60GHz (1597.65-MHz
K8-class CPU)
Origin =3D "GenuineIntel" Id =3D 0x6f7 Stepping =3D 7
Features=3D0xbfebfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PG=
E,MCA,CMOV,PAT,PSE36,CLFLUSH,DTS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE>
Features2=3D0x4e33d<SSE3,RSVD2,MON,DS_CPL,VMX,TM2,SSSE3,CX16,xTPR,PDCM,DC=
A>
AMD Features=3D0x20100800<SYSCALL,NX,LM>
AMD Features2=3D0x1<LAHF>
Cores per package: 4
Ok, I changed the code to something like this:
long old;
int diff, loops;
atomic_add_int(&uidinfo_cnt1, 1);
if (diff > 0) {
loops =3D 0;
do {
loops++;
old =3D uip->ui_sbsize;
if (old + diff > max)
return (0);
} while (atomic_cmpset_long(&uip->ui_sbsize, old, old + diff) =3D=3D 0);
if (loops > 1)
atomic_add_int(&uidinfo_cnt2, loops);
} else {
atomic_add_long(&uip->ui_sbsize, (long)diff);
}
This allows me to see how many additional loops I do, because with
lock-free version we still can have contention and loop, that's why
wait-free version is superior.
Actually I was a bit surprised with the results:
debug.uidinfo.cnt1: 88746008
debug.uidinfo.cnt2: 31296304
(Running 8 processes.)
Which means, because of contention, we do 31296304 additional atomic
operations, which is about 30% more.
--=20
Pawel Jakub Dawidek http://www.wheel.pl
pjd@FreeBSD.org http://www.FreeBSD.org
FreeBSD committer Am I Evil? Yes, I Am!
--0eh6TmSyL6TZE2Uz
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4 (FreeBSD)
iD8DBQFGx/f9ForvXbEpPzQRAoIwAKCL/fLfk/Wow6njyNFLyOXjKky5RwCfUKoX
7ZGZAv/M+5w9Xu5RFPFoJRE=
=yhzP
-----END PGP SIGNATURE-----
--0eh6TmSyL6TZE2Uz--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070819075750.GB11792>