On Tue, Aug 12, 2014 at 08:40:14PM +0530, Vinod Koul wrote: > On Mon, Aug 11, 2014 at 01:19:12PM +0100, Mark Brown wrote:
> > That doesn't mean it's the best way to go in the end though - long term > > I expect to see us move away from that to something more data driven and > > DAPM integrated. Requiring drivers to open code things rather than > > factoring out the code doesn't seem like the right way forwards. > Okay if that is the direction then this sound good to me. > Only one last nitpick on how do we set different format value apart from > sample rate, bits etc here. We have devices talking on same port and each one > is > different format(one use i2s, one uses tdm ...), so how do we handle that > part? Doing that sort of thing automatically is probably best handled by having some actual concrete idea of what's on the link and how it's sharing the resources which we don't really have in detail right now - we loose track of TDM when things hit the link and can't do things like notice that only two timeslots are active even though the link could have eight and so end up powering more than we need to. What I'd like to be able to do is to trace the individual per-channel and per-stream audio paths over the link, I think multiple configurations would fall out of being able to do that though I don't have a terribly concrete idea for a design right now (I sort of have one but it's got a lot of handwaving in it and isn't really concrete enough to write down).
signature.asc
Description: Digital signature
