[Prism54-devel] Prism Nitro Support

[Luis R. Rodriguez]

--bPrm2PuLP7ysUh6c
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable


I'll reply to myself. Let's see... Most of these are educational
guesses.

On Thu, Mar 04, 2004 at 06:39:11PM -0500, Luis R. Rodriguez wrote:
> On Wed, Mar 03, 2004 at 04:51:19PM +0800, Rodel Miguel wrote:
> > Hi!
> >=20
> > I would like to ask if my solution would suffice prism nitro support wh=
ile frameburst is not yet integrated with wireless tools.  I just inserted =
the following line in isl_ioctl.c specifically in  prism54_mib_init functio=
n:
> >=20
> > 	/* set prism nitro support - rodel */
> >         MGT_SET_REQUEST_U32_NOCACHE(priv, DOT11_OID_MAXFRAMEBURST, init=
_burst);
> >=20
> >         /* set card profile - rodel */
> >         MGT_SET_REQUEST_U32_NOCACHE(priv, DOT11_OID_PROFILES, init_prof=
ile);
>=20
> If that's how you add "Prism Nitro" support then yes, that
> would be a proper place to put it to start it @ init time.=20
>=20
> Questions:
>=20
> 0. what is "prism nitro" ?

Marketing term for Prism firmware enhancements which allow you to
maximize the # packets sent out of the physical device / unit of time frame?

> 1. what's initial value are you using for init_burst and why?

Default value on all my cards is 0. I see an SMC Driver document (for
users) which suggests:

* Mixed-mode environment: 650
* 11g Only environments: 5000

I also noticed, default value is 0.

> 2. what is the units of DOT11_OID_MAXFRAMEBURST?

Not sure yet.

> 3. why are you setting the profile afterwards?

Doesn't seem necessary according to my tests.

> 4. was this a feature available for all versions of the firmware?

# Note: using nttcp -T -t -l16384 moth-wlan-gw

It doesn't seem like it. I tested my STA with 0.8.0.0.arm with both max
frameburst values to 0 and 5000 and I actually saw better results with
max frameburst set to 0 (19Mbit/s after 3 runs, it steadily stayed there).
With maxframeburst set to 5000 I get 14Mbit/s results.=20

> 5. is frameburst just an 802.11g feature? Ie, if set would a STA ASSOC we=
ll
> with 802.11b APs, and if set on the AP would it allow for non "frame
> burst" capable STAs to ASSOC?

It seems as a hack that works decently (or at least suggested) up to
values of 650 in mixed modes (B/G) and on G you can blast it off up to
5k.

I'm not sure but I'd guess 650 is an acceptable max framburst value you
can set your AP to if you're dealing in mixed mode and with cards not
capable of modifying this rate. More tests should prove/disprove this.

> > would the card be set to the specified frameburst whenever the driver i=
s loaded?   =20
>=20
> Yes.

If we were to set a default value I'd shoot then for 650 for now. We
*could* add an ioctl to max this to 5k... We are running out of ioctls
so we'll have to work conservatively now.

Answers/corrections to my answers/educational&experimental guesses are welc=
omed.

	Luis

--=20
GnuPG Key fingerprint =3D 113F B290 C6D2 0251 4D84  A34A 6ADD 4937 E20A 525E

--bPrm2PuLP7ysUh6c
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFASC8Kat1JN+IKUl4RAqqFAKCrNnS8KhTCfXU6MQJCXO3V7fmm2wCggah/
ytfLnqjfcDyi1JgVtBkqCD8=
=+bsF
-----END PGP SIGNATURE-----

--bPrm2PuLP7ysUh6c--