G'day,

Just FYI to provide a closure to this thread for those that are in a similar 
situation:

I finally gave up and installed Suse 9.2 on a spare partition. Data bandwidth 
is now 32Mbyte/sec using the same hardware. The NetBSD ugen driver whilst 
much slower at 4Mbyte/sec also has problems in sustaining this speed even 
under low system load suffering from loss of data. Linux doesn't exhibit 
these problems under same conditions.

Not being a kernel hacker myself, I'm not sure how to improve the USB driver 
and hence I have no other choice but to run this application under Linux for 
now.

cheerio Berndt

On Mon, 14 Feb 2005 11:35 am, Lennart Augustsson wrote:
> The ugen driver isn't really made for good throughput.
> Maybe you want to rool your own driver.
>
>  -- Lennart
>
> Berndt Josef Wulf wrote:
> > On Mon, 14 Feb 2005 02:19 am, Jon Buller wrote:
> >>On 00 Jan 00 00:00:00,
> >>
> >>Berndt Josef Wulf wrote:
> >>>G'day,
> >>>
> >>>Where can I find benchtests performed for USB2 compliant controllers on
> >>>NetBSD-i386 that would give some indication on the expected maximum data
> >>>bandwidth? Currently, I'm only managing a maximum of 4MBytes/sec with a
> >>>Intel 82801DB USB EHCI Controller (rev. 0x01) connected to a Cypress
> >>>7C63013 device (FX2), which is a long way off from its theoratical
> >>>480Mbits/sec of a fullspeed USB2 interface.
> >>
> >>You meant high-speed USB interface, didn't you?  Full-speed is
> >>10Mbits/sec (or is it 12?)  High-speed is the fast one.  I (and
> >>everyone else I know) thinks the USB people were on crack when they
> >>thought up those names.  If you really are configuring it at full speed,
> >>there is some chance for improvement there (2x or 3x), but not 100x.
> >
> > The measured data bandwidth is 4MBytes/sec. This is only 10% of the
> > maximum bandwidth theoretically achievable. Linux developers report data
> > bandwidth of 32-40MBytes/sec under the same conditions.
> >
> > Is there generic support for asynchronous reading and writing?  That
> > is, multiple reads and writes outstanding, which notifies
> > their completion? (Signals, polling, etc).
> >
> > Is there a USB specific ioctl that allows submission of an USB "URB"
> > (USB Request Block) asynchronously?
> >
> > Is there a way of improving data throughput on this device or is this the
> > best we currently can do? Find below some kernel and userland messages on
> > system configuration and performance.
> >
> > ./program_and_start.sh
> > Using device /dev/ugen0 on bus /dev/usb3 vendor id 0xfffe product id
> > 0x0002 Read 102400000 bytes in 25 seconds and 1068 microseconds. Rate of
> > 4095825 bytes/second
> >
> >  ehci0 at pci0 dev 29 function 7: Intel 82801DB USB EHCI Controller (rev.
> > 0x01)
> >  ehci0: interrupting at irq 11
> >  ehci0: EHCI version 1.0
> >  ehci0: companion controllers, 2 ports each: uhci0 uhci1 uhci2
> >  usb3 at ehci0: USB revision 2.0
> >  uhub3 at usb3
> >  uhub3: Intel EHCI root hub, class 9/0, rev 2.00/1.00, addr 1
> >  uhub3: single transaction translator
> >  uhub3: 6 ports with 6 removable, self powered
> >
> > usbdevs -v
> > [,,,]
> > Controller /dev/usb3:
> > addr 1: high speed, self powered, config 1, EHCI root hub(0x0000),
> > Intel(0x8086), rev 1.00
> >  port 1 addr 2: high speed, self powered, config 1, USRP Rev 2(0x0002),
> > Free Software Folks(0xfffe), rev 1.02
> >  port 2 powered
> >  port 3 powered
> >  port 4 powered
> >  port 5 powered
> >  port 6 powered

-- 
Every man who says frankly and fully what he thinks is doing a public service.
[Leslie Stephen]