> > Trying to optimize following sequence:
> > - send command to bulk ep 4. I call this the 'master-urb'
> > - recieve data from bulk ep 2. I call this the 'slave-urb'
>
>Is there a reason you can't submit the read urb to ep 2 before sending
>the command to ep 4?
Good idea...
Well, what can i say. mount'ing the 64MB SmartMedia now takes 8.5 sec (vs
16.6) !
But a 'cat /mnt/sm/* >/dev/null' still freezes the machine after 10-20 sec.
I've tried it with both uhci + usb-uhci (2.4.4). Basically this is the same
behavior as what i described under "a)" - except that they don't instantly
freeze.
Possibly a SMP issue ? (all frozen machines were 2xSMP).. I'll check that.
> Or, how about having the completion handler for ep
>4 submit the read urb to ep 2? Then you _know_ the write command
>completed before trying to read (you can't rely on the status !=
>EINPROGRESS as I've been reminded at times :)
That's exactly what i tired to describe as "c)". Works perfectly for hours
- but with the usual performache penalty.
So I'm still puzzling about why the machine freezes when queueing both urb's ?
And also where this turn-time penalty comes from; especially why is it
still there, when the slave urb is submit()'d in the completion handler ?
- sda
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
http://lists.sourceforge.net/lists/listinfo/linux-usb-devel
