Re: [PATCH] ALSA: usb: refine delay information with USB frame counter

[Clemens Ladisch <clemens@ladisch.de>]

Pierre-Louis Bossart wrote:
> > > +	if (frame_diff < 0)
> > > +		frame_diff += 1024; /* handle 10-bit wrap-around */
> > 
> > After reading through some HCD source files, I deduced the following
> > wrap-arounds:
> > 
> > EHCI: 8-10 bits (default 9)
> > FHCI: 11 bits
> > IMX21: 16 bits
> > ISP1760: 10 bits
> > OHCI: 16 bits
> > OXU210HP: 10 bits
> > R8A66597: 10 bits
> > SL811: 5 bits(!?)
> > UHCI: 32 bits
> > xHCI: ?
> > 
> > As long as there isn't an API that tells us about the valid range of the
> > frame counter, we should probably mask off all except the lowest few bits.
> 
> If we only used the 8 least-significant bits, it'd still leave us with up to
> 256 ms. That should work I guess. But the SL811 with 5 bits is a problem,
> that makes 32ms, this is probably too low.

Sorry, I read the code wrong; that driver returns a 16-bit frame counter, it
just doesn't allow to schedule more than 32 frame AFAICT.  And that driver
doesn't support USB audio anyway, so 8 bits should be fine.


Regards,
Clemens