Re: Re: [SynCE-WindowsMobile5] Re: Updated WM5 patch

Hi,

> On Thursday 20 April 2006 15:29, Ole André Vadla Ravnås wrote:
> > Oops, I forgot to mention one little, but important, detail: It is no
> > longer necessary to set the MTU, just stick with the default at 1500.
> > At least I've not had any issues with it with the latest patch when
> > testing against a Qtek 9100 (HTC Wizard) device.
> Great improvement! But unfortunately, i only have USB-1 device sticking at
> 12MB/s :-)  Though, that the default MTU is now working, is something for me!

My device is only capable of 10 Mbps, so USB-1 shouldn't be a limiting
factor. In my experience with 10 Mbps cdc ethernet based USB devices
the top speed I've ever reached has been ~500 kB/sec (+/-, don't
remember exactly), so the current rndis_host performance should be
close to the limit -- it'll be interesting to test it on Windows
though. :-)

> I will host the patch on sourceforge until it has entered the official kernel.

Great! Much appreciated!

Regards,
Ole André


> > On 4/20/06, Ole André Vadla Ravnås <oleavr@xxxxxxxxx> wrote:
> > > Hi,
> > >
> > > Here's an updated version of the original WM5 patch. It doesn't yet
> > > solve the compatibility problem with some WM5 devices, which is still
> > > under investigation.
> > >
> > > What's new is that performance has been improved from ~150 kB/sec
> > > (testing with pcp, with vdccm running without any debug output), to
> > > 450-460 kB/sec, which is a 3x performance improvement. This was
> > > achieved by fixing a bug in usbnet.c where it doesn't cope with
> > > partial frames, so that if one USB transfer contains say 300 bytes of
> > > the first RNDIS packet message, and the next transfer contains the
> > > rest of it plus say 5 more RNDIS packet messages, they would all just
> > > get dropped resulting in multiple re-transmits, and thus a big
> > > performance hit.
> > > I'm not sure if my fix is the "right" approach, as I'm still quite new
> > > to kernel development in general, but at least it's a step in the
> > > right direction.
> > >
> > > There's also much more debug output now, as rndis_host does far more
> > > querying when the device is connected, which could prove useful in
> > > pinpointing the differences between working and non-working devices.
> > >
> > > Regards,
> > > Ole André
> >
> > -------------------------------------------------------
> > Using Tomcat but need to do more? Need to support web services, security?
> > Get stuff done quickly with pre-integrated technology to make your job
> > easier Download IBM WebSphere Application Server v.1.0.1 based on Apache
> > Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid0709&bid&3057&dat1642
> > _______________________________________________
> > SynCE-WindowsMobile5 mailing list
> > SynCE-WindowsMobile5@xxxxxxxxxxxxxxxxxxxxx
> > https://lists.sourceforge.net/lists/listinfo/synce-windowsmobile5
>
>
> -------------------------------------------------------
> Using Tomcat but need to do more? Need to support web services, security?
> Get stuff done quickly with pre-integrated technology to make your job easier
> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmdlnk&kid0709&bid&3057&dat1642
> _______________________________________________
> Synce-devel mailing list
> Synce-devel@xxxxxxxxxxxxxxxxxxxxx
> https://lists.sourceforge.net/lists/listinfo/synce-devel
>


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid0709&bid&3057&dat1642