On Wed, 2007-10-31 at 18:48 +0530, Avadhoot Punde wrote: why do you believe you need to do this as a kernel module?

On Wed, 2007-10-31 at 18:48 +0530, Avadhoot Punde wrote: For an example of how to use ALSA, you may want to have a look at this kit: http://amirhirsch.com/ps3synth.tar $ tar -xf ps3synth.tar && cd ps3synth If you are on Intel hardware, you'll need to edit line 49 in play.c from: // Signed 16-bit big-endian format snd_pcm_hw_params_set_format(handle, params, SND_PCM_FORMAT_S16_BE); .. to: // Signed 16-bit little-endian format snd_pcm_hw_params_set_format(handle, params, SND_PCM_FORMAT_S16_LE); .. and then: $ make && ./square|./filter|./play Despite the name of the archive, this is all standard C. --

On Thu, 2007-11-01 at 15:09 +0100, Jens M Andreasen wrote: I don't know if Avadhoot is still around here? Anyway, although the ALSA part in this archive (play.c) looks great and should solve your initial question, at least sine.c is seriously broken. It will run for a few seconds and then start living a random life of its own ... An alternative approach could look like this: --8<------------------------------------ sine.c ------- #include <unistd.h> // 16 bit Taylor approximation of sin() float fastsin(float x) { float x2 = x*x; return x*(x2*(x2*(x2*(x2*(1.0/36288.0) - (1.0/5040.0)) + (1.0/120.0)) - (1.0/6.0)) + 1.0); } int main() { short out[256][2]; float x = 22000.0, // amplitude y = 0.0, a = 2.0 * fastsin(440 * 3.14159/44100); int i = 0; for(;;) { x -= a*y; y += a*x; if(++i == 256) { write(1, out, sizeof(short) * 256 * 2); i = 0; } out[i][0] = out[i][1] = x; } } --8<--------------------------------------- EOF -------- .. which will run forever as long as you keep the frequency below 1/6 of the samplerate. The beuty of this variant is that by letting sin and cos alternate - slightly out of phase with each other - the generated errors tend to take out each other, stabilizing the system. Well, at least this is my understanding of the dicussion over at http://www.musicdsp.org