On 6/11/21 6:53 AM, Giovanni Mascellani wrote:
Hi,
Il 10/06/21 21:25, Derek Lesho ha scritto:
Yeah that would be really useful to see. My impression is that, since on Windows the stream has a separate request queue, the simplest implementation would just stop processing requests on the stream queue while the stream is paused. This wouldn't cause problems with the source receiving the start request, as it has its own queue and then upon starting again the stream would run through its saved requests.
In the end, I didn't end up implementing this strategy. I agree that it is conceptually simpler and probably it is what happens on Windows, but I believe my implementation (v2) should be essentially equivalent and it requires less locking and coordination between different queues (because there is just one).
Yep, fully agreed. I believe my original message was unclear, I more meant "given that Windows decided to use the (admittedly more complex) multi-queue approach, the simplest implementation on their end would probably just stop processing requests on pause, and therefore it would probably be more accurate to queue requests than samples".
Giovanni.