[hatari-devel] Higher precision when generating samples -> use nanosleep

[David Savinkoff]

>
> But that's what my latest committed patch does : it really computes the 
> exact number of samples, adjusting it by +1 when needed to take care of 
> the rouding errors.
>
> Uncomment line 1147 in sound.c and you will see the number of generated 
> samples ; doing the sum on 1 second should give the same number of 
> samples as the current replay frequency.
>
> I understand what your patch does, and I agree with the method, but now 
> that my latest fix computes the right number of samples, we should only 
> have to modify Main_WaitOnVbl().
> For example, to emulate a nearly correct 60 Hz VBL, we would alternate 
> between 16.66 ms and 16.67 ms using nanosleep(). No need to modify 
> sound.c or audio.c.
>
>Nicolas
>

Yes. sound.c may not have to be touched by my code.

Integrate the following into main.c Main_WaitOnVbl():

	/* Increase emulation rate to maintain sound synchronization */
	if (nSamplesToGenerate + nGeneratedSamples < SoundBufferSize + nSamplesPerFrame)
		/* Shorten emulated VBL time by given number of ms/FRAME_TICK_FACTOR */
		nFrameTickSkew = -20;
	else
	{
		if (nSamplesToGenerate + nGeneratedSamples > SoundBufferSize + nSamplesPerFrame*2)
			/* Decrease emulation rate to maintain sound synchronization */
			nFrameTickSkew = 20;
		else
			nFrameTickSkew = 0;
	}

And remove it from sound.c Sound_SetSamplesPassed()

David