http://bugs.winehq.org/show_bug.cgi?id=9358
--- Comment #27 from Raymond superquad.vortex2@gmail.com 2010-05-05 21:24:37 --- (In reply to comment #21)
They are, sort of. Unfortunately alsa dmix uses power or 2 period sizes, and since nobody has hardware mixing I used that function to find out how big a period should be.
The current driver model doesn't allow me to know what the period size is, so I use a rough approximation to keep it near 10 ms if the primary rate is 44100 or 48000 which it is by default. The real problem is that alsa dmix sucks, so I have to work around its limitations. I really cannot when the expected latency is <100ms, so Im stuck with an imperfect solution that works for most people.
I don't understand why "The current driver model doesn't allow me to know what the period size" ?
The current implementation is setting buffer time and period time but dsound_test always fail since the driver allocate different buffer size when sampling rate is different