On Wed, Aug 10, 2016 at 04:59:03PM -0500, Pierre-Louis Bossart wrote:
On 8/10/16 12:52 PM, Mark Brown wrote:
I'm not just hinting at that, I've openly stated it quite a few times now! :P For the simpler CODECs it's kind of marginal if you need to bother but for anything more complex (even things with PLLs) it seems like the way forwards.
interesting, thanks for the precision. I must admit I missed this concept completely and I didn't see any codec vendors work in this direction so far. Ironically the x86 part may be the most straightforward...
Architectures like x86 that are unable or unwilling to enable the use of the clock API are a blocker here, drivers can't make use of the API if it's not available. There are more architectures than x86 but it's the only one with any active ASoC use that I'm aware of, unfortunately the maintainers didn't react positively to attempts to enable the clock API in the past. I just sent a patch to try to fix it from the clock API side so let's hope that's a bit more helpful.