Messages in this thread | | | Date | Thu, 14 Oct 1999 22:23:21 +0200 | From | Martin Dalecki <> | Subject | Re: Kernel OSS vs ALSA |
| |
Thomas Sailer wrote: > > Michael Poole wrote: > > > > Thomas Sailer <sailer@ife.ee.ethz.ch> writes: > > > > > > * simultaneous multiple streams, with mixing as needed > > > > > > User space IMO > > > > Bad idea IMO. For highly interactive applications (eg games, but also more > > serious applications of VR) you don't want to take (a minimum of) a full > > context switch before the sound card even has a prayer of seeing a new > > audio sample. It's just not reasonable to restrict access to the audio hw > > Having format converters and sample rate converters in the kernel is > scary. > Straight mixing is doable, but that means that every app accessing sound > has to be using the same format, likely the best the sound hardware can > support, > i.e. 48kHz 16bit 2channels for most current hardware. While this > may not necessarily break the OSS API, it will break many current > OSS apps. And yes, no mmap
I can just *emphasize* on this aspect of the whole: Doing sample rate conversion is an NOT AT ALL trivial task! Any trivial interpolation will have at best laughable results. There where times I where coding signal processing stuff, and writing proper adaptive filters itsn't something that should be done inside the kernel and it can be done (almost) poperly in a generic way. Or is there somebody out there who would wish to have the whole LINPACK inside of the kernel????!!!!
Just have a look at the posix RT calls and thing which use could be made of the for a user land daemon. If the context switch latency is a concern in some gaming, then, well really better tell the user to get a bigger processor.
--Marcin
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |