On Fri, Jun 5, 2009 at 10:53 AM, David Chisnall
<address@hidden> wrote:
I didn't reply to this earlier, but I think NSSound should be very wary of making any assumptions about the native sound format for the hardware for several reasons:
1) Expensive hardware now uses floating-point samples internally, and this is likely to become the case for cheap hardware in the future.
2) Handheld devices and telephony systems often use lower quality sound (even 8-bit or 22KHz) for speaker output.
3) A lot of sound hardware can do format conversion in the DSP, which uses less power than doing it on the CPU and doesn't take CPU time from other programs.
4) Any half-decent OS hides these details from you.
David
Thanks David, this information is definitely useful.
Eventhough I don't like doing so, I think I'm going to want to start over... this time applying all of what we've discussed about here. One thing you mentioned before, which I liked the idea (but wouldn't know where to start) was pushing the NSSound decoding and playing stuff to GNUstep-back (I know you said only output, but if we want to support streaming the decoding is going to have to be done just-in-time). Like I mentioned before, I'm doubting my reasoning behind picking OpenAL as the audio output library.
At this point, I think I am a lot better off than where I was 2-3 weeks ago when I started this.
Stefan