On Thu, 13.12.07 20:51, Denis Leroy (denis(a)poolshark.org) wrote:
>>> 4) Sound-using apps use ALSA/ESD as usual.
>>> - It's all actually going through pulseaudio, though.
>> Wait, wait. Since pulseaudio does not, actually, talk to the hardware,
>> applications using ALSA will be rerouted through pulseaudio, which in
>> talks to ALSA to produce sound, is that right?
> Sounds like pretty standard wrapping technique, if it is.
How much extra latency are we getting from this ? I've noticed the extra
lag in some of our games.
The latency that is introduced through the wrapping of APIs is most
likely not measurable. It's not that we do additional buffering or
anything. And it is certainly not hearable with human ears.
Of course, it would be better if we had native PA support in every
audio library we ship. I am working on that, bear with me.
If you can hear the lag in those games, than it is most likely caused
because we currently route all SDL output through the ESD
protocol. The ESD protocol doesn't offer any reasonable control of the
buffer or the latency, so that's not easy to change. All ESD data is
passed through a TCP socket, which needs to be fully filled up by the
client, and which thus adds major buffer and thus major latency to the
In F9 we'll hoepfully ship with SDL directly on top of PA. Stay tuned.
Lennart Poettering Red Hat, Inc.
lennart [at] poettering [dot] net ICQ# 11060553