Confused about FloatVectorOperations::multiply compared to regular multiply

General JUCE discussion
1

Here is three pieces of code, which I thought would produce the exact same results, but does not.

Note for all three examples, in between reading and writing samples, I am doing some “wavefolding”, that is processing signals that exceeds -1 or 1, so overflow wraps down into valid range.

  1. First one is what I need;
while (voiceSamples < numSamples)
{
	// Read sample
	signal = TGVoiceBuffer.getSample (0, voiceSamples) * multiplier;

	// Wrap signal here
	
	// Write sample
	TGVoiceBuffer.setSample (0, voiceSamples, signal);
	
	++voiceSamples;
}

TGVoiceBuffer.applyGain (multiplier);
while (voiceSamples < numSamples)
{
	// Read sample
	signal = TGVoiceBuffer.getSample (0, voiceSamples);

	// Wrap signal here
	
	// Write sample
	TGVoiceBuffer.setSample (0, voiceSamples, signal);
	
	++voiceSamples;
}

FloatVectorOperations::multiply (TGVoiceBuffer.getWritePointer (0, 0), multiplier, numSamples);
while (voiceSamples < numSamples)
{
	// Read sample
	signal = TGVoiceBuffer.getSample (0, voiceSamples);

	// Wrap signal here
	
	// Write sample
	TGVoiceBuffer.setSample (0, voiceSamples, signal);
	
	++voiceSamples;
}

The samples produces by gain on method 2 and multiplication on method 3, is different from method one, which is what I need.

How, if possible, do I use FloatVectorOperations, so it does exactly as in first example?

2

Are you sure the values are actually different? Have you debugged the code and inspected the values?

3

I have not debugged, but I can easily see it in the synth I am coding as it has a continuous graph of the waveform (samples), kind of an oscilloscope, and I can hear it.

4

Ok I need to take the day off programming! I made a mistake - Please move on, nothing to see here. Sorry for wasting your time.

5

If you are using the Juce Synthesiser classes, you should not operate with the Synthesiser voice buffers with the 2 and 3 methods, only method 1 will work. The reason for that is that the buffer you get into the voice’s renderNextBlock already contains the audio from the other voices. The synthesiser class uses a shared audio buffer for the voices. The voices should only add/mix their audio into the buffer, not do anything else with the audio. Considering that, also your implementation for method 1 seems wrong, you should not use setSample on the buffer, instead addSample should be used.

6

Nah it was me overseeing a bad mistake in the code. And those buffers are not the one’s “coming” from the Processor, they are local to each voice.

7

OK, nice you got it working.

8

Two weeks straight of coding from morning to night, time for a break.