I should probably explain the reasoning behind this. I don't think it makes
much sense to the end user to specify the audio buffer length as a size in
bytes. It's hard to understand, and, if you specify it in bytes, then the
resulting performance is dependent on the other audio settings - the
frequency, the number of channels, the resolution. This makes it difficult to
provide sensible defaults that perform equally well independent of the audio
platform.
The default - 100 ms should provide acceptable results without too much CPU
usage. You should reduce this if you notice a lag between the event which
should produce a sound effect and hearing the sound effect. Decreasing the
latency will increase the CPU usage. Increasing the latency will reduce the
CPU usage.
