[Prism54-devel] benchamark
Luis R. Rodriguez
mcgrof@ruslug.rutgers.edu
Fri, 5 Mar 2004 02:18:27 -0500
--Mit9XoPEfICDqq/V
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Thu, Mar 04, 2004 at 09:20:28AM +0100, Feyd wrote:
> Hi,
>=20
> I compiled the driver staticaly into the kernel to see if it affects perf=
ormace
> and the result is pleasant (I was getting about 10-12 Mbit with the drive=
r as
> module on those slow machines):
>=20
> notes:~# nttcp -T -t -l16384 gw
> Bytes Real s CPU s Real-MBit/s CPU-MBit/s Calls Real-C/s CP=
U-C/s
> l 33554432 11.17 0.31 24.0360 855.0180 2048 183.38 6=
523.3
> 1 33554432 11.19 1.04 23.9794 257.1612 22540 2013.50 21=
593.3
>=20
> after setting frameburst to 5000 (maximum allowed by windows driver) on b=
oth
> machines (./setoid eth0 0x1b000008 136 19 0 0):
>=20
> notes:~# nttcp -T -t -l16384 gw
> Bytes Real s CPU s Real-MBit/s CPU-MBit/s Calls Real-C/s CP=
U-C/s
> l 33554432 8.58 0.51 31.3002 524.3689 2048 238.80 4=
000.6
> 1 33554432 8.59 0.90 31.2364 298.6386 16113 1874.98 17=
926.0
These test are only performed with modules on systems running prism54
cards.=20
STA is PII 233Mhz / 56MB RAM
AP is PMMX 200Mhz / 60MB RAM
---------------------------------------------------------------------------=
------
My results with frameburst set to default (0) is:
wasp[~/prism54/prism54-cvs-latest]$ nttcp -T -t -l16384 moth-wlan-gw
Bytes Real s CPU s Real-MBit/s CPU-MBit/s Calls Real-C/s CPU-=
C/s
l 33554432 15.69 0.32 17.1102 844.2641 2048 130.54 644=
1.2
1 33554432 15.73 2.59 17.0688 103.6430 23171 1473.36 894=
6.3
---------------------------------------------------------------------------=
-------
Now, setting frameburst to a max of 5000 on both STA an AP I get:
wasp[~/prism54/prism54-cvs-latest]$ nttcp -T -t -l16384 moth-wlan-gw
Bytes Real s CPU s Real-MBit/s CPU-MBit/s Calls Real-C/s CPU-=
C/s
l 33554432 9.53 0.41 28.1540 656.4209 2048 214.80 500=
8.1
1 33554432 9.55 3.36 28.1147 79.8915 13341 1397.28 397=
0.5
---------------------------------------------------------------------------=
--------
Your list shows better results. I'd have to test building the driver
into the kernel, I'm not sure how it'd help speed things.
One thing is for sure, maxing out the frameburst is *good*. :)=20
--=20
GnuPG Key fingerprint =3D 113F B290 C6D2 0251 4D84 A34A 6ADD 4937 E20A 525E
--Mit9XoPEfICDqq/V
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
iD8DBQFASCnDat1JN+IKUl4RAipvAJ0bDPJT5FBTF0aqP6Q2DsA4dvVZiQCfW99d
qZd5ImQAOEbC1lqaJXnPaGA=
=nz0V
-----END PGP SIGNATURE-----
--Mit9XoPEfICDqq/V--