On Fri, Dec 11, 2020 at 09:07:45AM +0100, Marcus Glocker wrote:
> 
> After doing some deeper analyzes in to asmc_wait() I agree to that.
> Something seems to go fundamental wrong there.  In every asmc_update()
> execution, I can see asmc_wait() timeout 9 times, always on the
> ASMC_ACCEPT check.  That means out of ~48ms to execute asmc_update(),
> we spend 45ms waiting for timeouts!
> 
> When this can be fixed, I guess we should be fine with the remaining
> ~3ms to execute asmc_update().
> 

Even 3ms seems the wrong order of magnitude; that's ~3000000 CPU
cycles on a 1GHz processor. This is 3 usb frames, so this will still
hurt audio in low-latency corner cases.

Reply via email to