Ok I've uploaded wine 1.5.7 with pulse17 and a new version of pulse18.
Changes since v17 from changelog:
Code:
* Winepulse v18 patch
- Remove clock_pulse interpolation. It sadly couldn't work.
- Allow 2 * MinimumPeriod for shared buffers
- Fix all compiler warnings when compiling with 64-bits
- Dynamically select low latency mode if less than 2 default periods
are requested. This requires the rtkit patch to be useful.
* rtkit patch
- Uses rtkit to get realtime privileges in a safe manner.
This allows for more accurate timing and less chance of underruns.
* dsound patches to work nicer with winepulse
- Rework that uses event-based handling for rendering. In testing
with jackd, down to 3.3ms latency was possible. It will still
underrun once every few seconds, so low latency is not on by
default. 80 ms will be default latency.
* mmdevapi patch to reactivate midi with winepulse
- winmm will use winealsa for midi only
* Add patch so default valgrind on precise won't crash
- Don't bump fd limit with valgrind enabled.
In human form: I use rtkit's hook for wine that never got upstream to add reliability to winepulse. dsound rendering has been reworked slightly. If you are insane and into pro audio winepulse can be used to go arbitrarily low in latency. Only for WASAPI mode though. Hopefully this will make winepulse even more stable than before, and you no longer need to enable winealsa for midi.
If latency interests you, the new default latency for dsound mode is 80 ms, which is still less than before. I can help you patch dsound to go lower than 80 ms if you want, but you should really use WASAPI mode if your audio software supports it since it will work by default then. I had 3.3ms from dsound with underruns every second piping pulseaudio to jackd. One has to find the lower bound right? Because of obvious reasons I never enabled this by default. This will never be possible with winealsa as far as I can tell..
~Maarten
Bookmarks