On 3/31/23 16:34, Mark Brown wrote: > On Fri, Mar 31, 2023 at 02:26:33PM +0000, Emil Abildgaard Svendsen wrote: >> On 3/31/23 04:55, Shengjiu Wang wrote: > >>> There are different requirements for this slot width. Some need physical >>> width, >>> Some need format width. We need to be careful about change here. > >> I might be wrong but shouldn't physical width always correspond to what >> can be physically measured. If you need the slot width, the same as >> format width you use a format with matching widths like >> SNDRV_PCM_FORMAT_S24_3LE? > > The point is more that things are likely to be relying on the > current behaviour, for example CODECs that don't actually support > 24 bit audio due to a power of two requirement but cope fine when > you play 24 bit audio in a 32 bit timeslot. This creates issues > changing things even if the new behaviour is in some sense > better. This is one of the unfortunate consequences of DT.
I guess if you want to do it runtime you have to create an audio card that will do it for you. I would have preferred it to be handled as close to hardware as possible. But that's a good enough compromise for not breaking current behavior. Thanks! Kind regards, Emil Svendsen
