> I anticipated this and came up with about 5ms worst case:
>
> 1 Submit packet to USB drivers
> 2 data to UART
> 3 data through loopback
> 4 data from UART
> 5 data from USB drivers
At least for OHCI, an idle host controller "should" be able to handle
(1) and (2) in less than a frame on average, assuming only one or two
USB transfers per packet ... but if there's more work queued, then it
can take more time. Same thing for (4) and (5), with the additional
variable of interrupt latencies.
> ...I'm using stock Intel UHCI.
The "uhci" driver seems to have higher latencies than the
"usb-uhci" driver (by almost another millisecond), which
in turn had higher latencies than "usb-ohci".
My own measurements were made by "ping" through the
"usbnet" driver, so they had USB and interrupt latencies
on both sides as well as network stack latencies. Usually
I see typical 1.5ms roundtrips with OHCI, and another
few ms with either UHCI driver. If you measure 15ms
using UHCI, that's 3-5 times slower than I'd expect.
- Dave
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
http://lists.sourceforge.net/lists/listinfo/linux-usb-devel
