/* ========================================
 *  Chamber2 - Chamber2.h
 *  Copyright (c) 2016 airwindows, Airwindows uses the MIT license
 * ======================================== */

#ifndef __Chamber2_H
#include "Chamber2.h"
#endif

void Chamber2::processReplacing(float **inputs, float **outputs, VstInt32 sampleFrames) 
{
    float* in1  =  inputs[0];
    float* in2  =  inputs[1];
    float* out1 = outputs[0];
    float* out2 = outputs[1];

	double overallscale = 1.0;
	overallscale /= 44100.0;
	overallscale *= getSampleRate();
	int cycleEnd = floor(overallscale);
	if (cycleEnd < 1) cycleEnd = 1;
	if (cycleEnd > 4) cycleEnd = 4;
	//this is going to be 2 for 88.1 or 96k, 3 for silly people, 4 for 176 or 192k
	if (cycle > cycleEnd-1) cycle = cycleEnd-1; //sanity check
	
	double size = (A*0.9)+0.1;
	double regen = (1.0-(pow(1.0-B,2)))*0.123;
	double echoScale = 1.0-C;
	double echo = 0.618033988749894848204586+((1.0-0.618033988749894848204586)*echoScale);
	double interpolate = (1.0-echo)*0.381966011250105;
	//this now goes from Chamber, to all the reverb delays being exactly the same
	//much larger usage of RAM due to the larger reverb delays everywhere, but
	//ability to go to an unusual variation on blurred delay.
	double wet = D*2.0;
	double dry = 2.0 - wet;
	if (wet > 1.0) wet = 1.0;
	if (wet < 0.0) wet = 0.0;
	if (dry > 1.0) dry = 1.0;
	if (dry < 0.0) dry = 0.0;
	//this reverb makes 50% full dry AND full wet, not crossfaded.
	//that's so it can be on submixes without cutting back dry channel when adjusted:
	//unless you go super heavy, you are only adjusting the added verb loudness.
	
	delayM = sqrt(9900*size);
	delayE = 9900*size;
	delayF = delayE*echo; 
	delayG = delayF*echo;
	delayH = delayG*echo;
	delayA = delayH*echo;
	delayB = delayA*echo;
	delayC = delayB*echo;
	delayD = delayC*echo;
	delayI = delayD*echo;
	delayJ = delayI*echo;
	delayK = delayJ*echo;
	delayL = delayK*echo;
	//initially designed around the Fibonnaci series, Chamber uses
	//delay coefficients that are all related to the Golden Ratio,
	//Turns out that as you continue to sustain them, it turns from a
	//chunky slapback effect into a smoother reverb tail that can
	//sustain infinitely.	
    
    while (--sampleFrames >= 0)
    {
		double inputSampleL = *in1;
		double inputSampleR = *in2;
		if (fabs(inputSampleL)<1.18e-23) inputSampleL = fpdL * 1.18e-17;
		if (fabs(inputSampleR)<1.18e-23) inputSampleR = fpdR * 1.18e-17;
		double drySampleL = inputSampleL;
		double drySampleR = inputSampleR;
		
		cycle++;
		if (cycle == cycleEnd) { //hit the end point and we do a reverb sample
			
			aML[countM] = inputSampleL;
			aMR[countM] = inputSampleR;
			countM++; if (countM < 0 || countM > delayM) countM = 0;
			inputSampleL = aML[countM-((countM > delayM)?delayM+1:0)];
			inputSampleR = aMR[countM-((countM > delayM)?delayM+1:0)];
			//predelay to make the first echo still be an echo even when blurred
			
			feedbackAL = (feedbackAL*(1.0-interpolate))+(previousAL*interpolate); previousAL = feedbackAL;
			feedbackBL = (feedbackBL*(1.0-interpolate))+(previousBL*interpolate); previousBL = feedbackBL;
			feedbackCL = (feedbackCL*(1.0-interpolate))+(previousCL*interpolate); previousCL = feedbackCL;
			feedbackDL = (feedbackDL*(1.0-interpolate))+(previousDL*interpolate); previousDL = feedbackDL;
			feedbackAR = (feedbackAR*(1.0-interpolate))+(previousAR*interpolate); previousAR = feedbackAR;
			feedbackBR = (feedbackBR*(1.0-interpolate))+(previousBR*interpolate); previousBR = feedbackBR;
			feedbackCR = (feedbackCR*(1.0-interpolate))+(previousCR*interpolate); previousCR = feedbackCR;
			feedbackDR = (feedbackDR*(1.0-interpolate))+(previousDR*interpolate); previousDR = feedbackDR;
			
			aIL[countI] = inputSampleL + (feedbackAL * regen);
			aJL[countJ] = inputSampleL + (feedbackBL * regen);
			aKL[countK] = inputSampleL + (feedbackCL * regen);
			aLL[countL] = inputSampleL + (feedbackDL * regen);
			aIR[countI] = inputSampleR + (feedbackAR * regen);
			aJR[countJ] = inputSampleR + (feedbackBR * regen);
			aKR[countK] = inputSampleR + (feedbackCR * regen);
			aLR[countL] = inputSampleR + (feedbackDR * regen);
			
			countI++; if (countI < 0 || countI > delayI) countI = 0;
			countJ++; if (countJ < 0 || countJ > delayJ) countJ = 0;
			countK++; if (countK < 0 || countK > delayK) countK = 0;
			countL++; if (countL < 0 || countL > delayL) countL = 0;
			
			double outIL = aIL[countI-((countI > delayI)?delayI+1:0)];
			double outJL = aJL[countJ-((countJ > delayJ)?delayJ+1:0)];
			double outKL = aKL[countK-((countK > delayK)?delayK+1:0)];
			double outLL = aLL[countL-((countL > delayL)?delayL+1:0)];
			double outIR = aIR[countI-((countI > delayI)?delayI+1:0)];
			double outJR = aJR[countJ-((countJ > delayJ)?delayJ+1:0)];
			double outKR = aKR[countK-((countK > delayK)?delayK+1:0)];
			double outLR = aLR[countL-((countL > delayL)?delayL+1:0)];
			//first block: now we have four outputs
			
			aAL[countA] = (outIL - (outJL + outKL + outLL));
			aBL[countB] = (outJL - (outIL + outKL + outLL));
			aCL[countC] = (outKL - (outIL + outJL + outLL));
			aDL[countD] = (outLL - (outIL + outJL + outKL));
			aAR[countA] = (outIR - (outJR + outKR + outLR));
			aBR[countB] = (outJR - (outIR + outKR + outLR));
			aCR[countC] = (outKR - (outIR + outJR + outLR));
			aDR[countD] = (outLR - (outIR + outJR + outKR));
			
			countA++; if (countA < 0 || countA > delayA) countA = 0;
			countB++; if (countB < 0 || countB > delayB) countB = 0;
			countC++; if (countC < 0 || countC > delayC) countC = 0;
			countD++; if (countD < 0 || countD > delayD) countD = 0;
			
			double outAL = aAL[countA-((countA > delayA)?delayA+1:0)];
			double outBL = aBL[countB-((countB > delayB)?delayB+1:0)];
			double outCL = aCL[countC-((countC > delayC)?delayC+1:0)];
			double outDL = aDL[countD-((countD > delayD)?delayD+1:0)];
			double outAR = aAR[countA-((countA > delayA)?delayA+1:0)];
			double outBR = aBR[countB-((countB > delayB)?delayB+1:0)];
			double outCR = aCR[countC-((countC > delayC)?delayC+1:0)];
			double outDR = aDR[countD-((countD > delayD)?delayD+1:0)];
			//second block: four more outputs
			
			aEL[countE] = (outAL - (outBL + outCL + outDL));
			aFL[countF] = (outBL - (outAL + outCL + outDL));
			aGL[countG] = (outCL - (outAL + outBL + outDL));
			aHL[countH] = (outDL - (outAL + outBL + outCL));
			aER[countE] = (outAR - (outBR + outCR + outDR));
			aFR[countF] = (outBR - (outAR + outCR + outDR));
			aGR[countG] = (outCR - (outAR + outBR + outDR));
			aHR[countH] = (outDR - (outAR + outBR + outCR));
			
			countE++; if (countE < 0 || countE > delayE) countE = 0;
			countF++; if (countF < 0 || countF > delayF) countF = 0;
			countG++; if (countG < 0 || countG > delayG) countG = 0;
			countH++; if (countH < 0 || countH > delayH) countH = 0;
			
			double outEL = aEL[countE-((countE > delayE)?delayE+1:0)];
			double outFL = aFL[countF-((countF > delayF)?delayF+1:0)];
			double outGL = aGL[countG-((countG > delayG)?delayG+1:0)];
			double outHL = aHL[countH-((countH > delayH)?delayH+1:0)];
			double outER = aER[countE-((countE > delayE)?delayE+1:0)];
			double outFR = aFR[countF-((countF > delayF)?delayF+1:0)];
			double outGR = aGR[countG-((countG > delayG)?delayG+1:0)];
			double outHR = aHR[countH-((countH > delayH)?delayH+1:0)];
			//third block: final outputs
			
			feedbackAR = (outEL - (outFL + outGL + outHL));
			feedbackBL = (outFL - (outEL + outGL + outHL));
			feedbackCR = (outGL - (outEL + outFL + outHL));
			feedbackDL = (outHL - (outEL + outFL + outGL));
			feedbackAL = (outER - (outFR + outGR + outHR));
			feedbackBR = (outFR - (outER + outGR + outHR));
			feedbackCL = (outGR - (outER + outFR + outHR));
			feedbackDR = (outHR - (outER + outFR + outGR));
			//which we need to feed back into the input again, a bit
			
			inputSampleL = (outEL + outFL + outGL + outHL)/8.0;
			inputSampleR = (outER + outFR + outGR + outHR)/8.0;
			//and take the final combined sum of outputs
			if (cycleEnd == 4) {
				lastRefL[0] = lastRefL[4]; //start from previous last
				lastRefL[2] = (lastRefL[0] + inputSampleL)/2; //half
				lastRefL[1] = (lastRefL[0] + lastRefL[2])/2; //one quarter
				lastRefL[3] = (lastRefL[2] + inputSampleL)/2; //three quarters
				lastRefL[4] = inputSampleL; //full
				lastRefR[0] = lastRefR[4]; //start from previous last
				lastRefR[2] = (lastRefR[0] + inputSampleR)/2; //half
				lastRefR[1] = (lastRefR[0] + lastRefR[2])/2; //one quarter
				lastRefR[3] = (lastRefR[2] + inputSampleR)/2; //three quarters
				lastRefR[4] = inputSampleR; //full
			}
			if (cycleEnd == 3) {
				lastRefL[0] = lastRefL[3]; //start from previous last
				lastRefL[2] = (lastRefL[0]+lastRefL[0]+inputSampleL)/3; //third
				lastRefL[1] = (lastRefL[0]+inputSampleL+inputSampleL)/3; //two thirds
				lastRefL[3] = inputSampleL; //full
				lastRefR[0] = lastRefR[3]; //start from previous last
				lastRefR[2] = (lastRefR[0]+lastRefR[0]+inputSampleR)/3; //third
				lastRefR[1] = (lastRefR[0]+inputSampleR+inputSampleR)/3; //two thirds
				lastRefR[3] = inputSampleR; //full
			}
			if (cycleEnd == 2) {
				lastRefL[0] = lastRefL[2]; //start from previous last
				lastRefL[1] = (lastRefL[0] + inputSampleL)/2; //half
				lastRefL[2] = inputSampleL; //full
				lastRefR[0] = lastRefR[2]; //start from previous last
				lastRefR[1] = (lastRefR[0] + inputSampleR)/2; //half
				lastRefR[2] = inputSampleR; //full
			}
			if (cycleEnd == 1) {
				lastRefL[0] = inputSampleL;
				lastRefR[0] = inputSampleR;
			}
			cycle = 0; //reset
			inputSampleL = lastRefL[cycle];
			inputSampleR = lastRefR[cycle];
		} else {
			inputSampleL = lastRefL[cycle];
			inputSampleR = lastRefR[cycle];
			//we are going through our references now
		}
		
		switch (cycleEnd) //multi-pole average using lastRef[] variables
		{
			case 4:
				lastRefL[8] = inputSampleL; inputSampleL = (inputSampleL+lastRefL[7])*0.5;
				lastRefL[7] = lastRefL[8]; //continue, do not break
				lastRefR[8] = inputSampleR; inputSampleR = (inputSampleR+lastRefR[7])*0.5;
				lastRefR[7] = lastRefR[8]; //continue, do not break
			case 3:
				lastRefL[8] = inputSampleL; inputSampleL = (inputSampleL+lastRefL[6])*0.5;
				lastRefL[6] = lastRefL[8]; //continue, do not break
				lastRefR[8] = inputSampleR; inputSampleR = (inputSampleR+lastRefR[6])*0.5;
				lastRefR[6] = lastRefR[8]; //continue, do not break
			case 2:
				lastRefL[8] = inputSampleL; inputSampleL = (inputSampleL+lastRefL[5])*0.5;
				lastRefL[5] = lastRefL[8]; //continue, do not break
				lastRefR[8] = inputSampleR; inputSampleR = (inputSampleR+lastRefR[5])*0.5;
				lastRefR[5] = lastRefR[8]; //continue, do not break
			case 1:
				break; //no further averaging
		}
				
		if (wet < 1.0) {inputSampleL *= wet; inputSampleR *= wet;}
		if (dry < 1.0) {drySampleL *= dry; drySampleR *= dry;}
		inputSampleL += drySampleL;
		inputSampleR += drySampleR;
		//this is our submix verb dry/wet: 0.5 is BOTH at FULL VOLUME
		//purpose is that, if you're adding verb, you're not altering other balances
		
		//begin 32 bit stereo floating point dither
		int expon; frexpf((float)inputSampleL, &expon);
		fpdL ^= fpdL << 13; fpdL ^= fpdL >> 17; fpdL ^= fpdL << 5;
		inputSampleL += ((double(fpdL)-uint32_t(0x7fffffff)) * 5.5e-36l * pow(2,expon+62));
		frexpf((float)inputSampleR, &expon);
		fpdR ^= fpdR << 13; fpdR ^= fpdR >> 17; fpdR ^= fpdR << 5;
		inputSampleR += ((double(fpdR)-uint32_t(0x7fffffff)) * 5.5e-36l * pow(2,expon+62));
		//end 32 bit stereo floating point dither
		
		*out1 = inputSampleL;
		*out2 = inputSampleR;

		in1++;
		in2++;
		out1++;
		out2++;
    }
}

void Chamber2::processDoubleReplacing(double **inputs, double **outputs, VstInt32 sampleFrames) 
{
    double* in1  =  inputs[0];
    double* in2  =  inputs[1];
    double* out1 = outputs[0];
    double* out2 = outputs[1];

	double overallscale = 1.0;
	overallscale /= 44100.0;
	overallscale *= getSampleRate();
	int cycleEnd = floor(overallscale);
	if (cycleEnd < 1) cycleEnd = 1;
	if (cycleEnd > 4) cycleEnd = 4;
	//this is going to be 2 for 88.1 or 96k, 3 for silly people, 4 for 176 or 192k
	if (cycle > cycleEnd-1) cycle = cycleEnd-1; //sanity check
	
	double size = (A*0.9)+0.1;
	double regen = (1.0-(pow(1.0-B,2)))*0.123;
	double echoScale = 1.0-C;
	double echo = 0.618033988749894848204586+((1.0-0.618033988749894848204586)*echoScale);
	double interpolate = (1.0-echo)*0.381966011250105;
	//this now goes from Chamber, to all the reverb delays being exactly the same
	//much larger usage of RAM due to the larger reverb delays everywhere, but
	//ability to go to an unusual variation on blurred delay.
	double wet = D*2.0;
	double dry = 2.0 - wet;
	if (wet > 1.0) wet = 1.0;
	if (wet < 0.0) wet = 0.0;
	if (dry > 1.0) dry = 1.0;
	if (dry < 0.0) dry = 0.0;
	//this reverb makes 50% full dry AND full wet, not crossfaded.
	//that's so it can be on submixes without cutting back dry channel when adjusted:
	//unless you go super heavy, you are only adjusting the added verb loudness.
	
	delayM = sqrt(9900*size);
	delayE = 9900*size;
	delayF = delayE*echo; 
	delayG = delayF*echo;
	delayH = delayG*echo;
	delayA = delayH*echo;
	delayB = delayA*echo;
	delayC = delayB*echo;
	delayD = delayC*echo;
	delayI = delayD*echo;
	delayJ = delayI*echo;
	delayK = delayJ*echo;
	delayL = delayK*echo;
	//initially designed around the Fibonnaci series, Chamber uses
	//delay coefficients that are all related to the Golden Ratio,
	//Turns out that as you continue to sustain them, it turns from a
	//chunky slapback effect into a smoother reverb tail that can
	//sustain infinitely.	
    
    while (--sampleFrames >= 0)
    {
		double inputSampleL = *in1;
		double inputSampleR = *in2;
		if (fabs(inputSampleL)<1.18e-23) inputSampleL = fpdL * 1.18e-17;
		if (fabs(inputSampleR)<1.18e-23) inputSampleR = fpdR * 1.18e-17;
		double drySampleL = inputSampleL;
		double drySampleR = inputSampleR;
		
		cycle++;
		if (cycle == cycleEnd) { //hit the end point and we do a reverb sample
			
			aML[countM] = inputSampleL;
			aMR[countM] = inputSampleR;
			countM++; if (countM < 0 || countM > delayM) countM = 0;
			inputSampleL = aML[countM-((countM > delayM)?delayM+1:0)];
			inputSampleR = aMR[countM-((countM > delayM)?delayM+1:0)];
			//predelay to make the first echo still be an echo even when blurred
			
			feedbackAL = (feedbackAL*(1.0-interpolate))+(previousAL*interpolate); previousAL = feedbackAL;
			feedbackBL = (feedbackBL*(1.0-interpolate))+(previousBL*interpolate); previousBL = feedbackBL;
			feedbackCL = (feedbackCL*(1.0-interpolate))+(previousCL*interpolate); previousCL = feedbackCL;
			feedbackDL = (feedbackDL*(1.0-interpolate))+(previousDL*interpolate); previousDL = feedbackDL;
			feedbackAR = (feedbackAR*(1.0-interpolate))+(previousAR*interpolate); previousAR = feedbackAR;
			feedbackBR = (feedbackBR*(1.0-interpolate))+(previousBR*interpolate); previousBR = feedbackBR;
			feedbackCR = (feedbackCR*(1.0-interpolate))+(previousCR*interpolate); previousCR = feedbackCR;
			feedbackDR = (feedbackDR*(1.0-interpolate))+(previousDR*interpolate); previousDR = feedbackDR;
			
			aIL[countI] = inputSampleL + (feedbackAL * regen);
			aJL[countJ] = inputSampleL + (feedbackBL * regen);
			aKL[countK] = inputSampleL + (feedbackCL * regen);
			aLL[countL] = inputSampleL + (feedbackDL * regen);
			aIR[countI] = inputSampleR + (feedbackAR * regen);
			aJR[countJ] = inputSampleR + (feedbackBR * regen);
			aKR[countK] = inputSampleR + (feedbackCR * regen);
			aLR[countL] = inputSampleR + (feedbackDR * regen);
			
			countI++; if (countI < 0 || countI > delayI) countI = 0;
			countJ++; if (countJ < 0 || countJ > delayJ) countJ = 0;
			countK++; if (countK < 0 || countK > delayK) countK = 0;
			countL++; if (countL < 0 || countL > delayL) countL = 0;
			
			double outIL = aIL[countI-((countI > delayI)?delayI+1:0)];
			double outJL = aJL[countJ-((countJ > delayJ)?delayJ+1:0)];
			double outKL = aKL[countK-((countK > delayK)?delayK+1:0)];
			double outLL = aLL[countL-((countL > delayL)?delayL+1:0)];
			double outIR = aIR[countI-((countI > delayI)?delayI+1:0)];
			double outJR = aJR[countJ-((countJ > delayJ)?delayJ+1:0)];
			double outKR = aKR[countK-((countK > delayK)?delayK+1:0)];
			double outLR = aLR[countL-((countL > delayL)?delayL+1:0)];
			//first block: now we have four outputs
			
			aAL[countA] = (outIL - (outJL + outKL + outLL));
			aBL[countB] = (outJL - (outIL + outKL + outLL));
			aCL[countC] = (outKL - (outIL + outJL + outLL));
			aDL[countD] = (outLL - (outIL + outJL + outKL));
			aAR[countA] = (outIR - (outJR + outKR + outLR));
			aBR[countB] = (outJR - (outIR + outKR + outLR));
			aCR[countC] = (outKR - (outIR + outJR + outLR));
			aDR[countD] = (outLR - (outIR + outJR + outKR));
			
			countA++; if (countA < 0 || countA > delayA) countA = 0;
			countB++; if (countB < 0 || countB > delayB) countB = 0;
			countC++; if (countC < 0 || countC > delayC) countC = 0;
			countD++; if (countD < 0 || countD > delayD) countD = 0;
			
			double outAL = aAL[countA-((countA > delayA)?delayA+1:0)];
			double outBL = aBL[countB-((countB > delayB)?delayB+1:0)];
			double outCL = aCL[countC-((countC > delayC)?delayC+1:0)];
			double outDL = aDL[countD-((countD > delayD)?delayD+1:0)];
			double outAR = aAR[countA-((countA > delayA)?delayA+1:0)];
			double outBR = aBR[countB-((countB > delayB)?delayB+1:0)];
			double outCR = aCR[countC-((countC > delayC)?delayC+1:0)];
			double outDR = aDR[countD-((countD > delayD)?delayD+1:0)];
			//second block: four more outputs
			
			aEL[countE] = (outAL - (outBL + outCL + outDL));
			aFL[countF] = (outBL - (outAL + outCL + outDL));
			aGL[countG] = (outCL - (outAL + outBL + outDL));
			aHL[countH] = (outDL - (outAL + outBL + outCL));
			aER[countE] = (outAR - (outBR + outCR + outDR));
			aFR[countF] = (outBR - (outAR + outCR + outDR));
			aGR[countG] = (outCR - (outAR + outBR + outDR));
			aHR[countH] = (outDR - (outAR + outBR + outCR));
			
			countE++; if (countE < 0 || countE > delayE) countE = 0;
			countF++; if (countF < 0 || countF > delayF) countF = 0;
			countG++; if (countG < 0 || countG > delayG) countG = 0;
			countH++; if (countH < 0 || countH > delayH) countH = 0;
			
			double outEL = aEL[countE-((countE > delayE)?delayE+1:0)];
			double outFL = aFL[countF-((countF > delayF)?delayF+1:0)];
			double outGL = aGL[countG-((countG > delayG)?delayG+1:0)];
			double outHL = aHL[countH-((countH > delayH)?delayH+1:0)];
			double outER = aER[countE-((countE > delayE)?delayE+1:0)];
			double outFR = aFR[countF-((countF > delayF)?delayF+1:0)];
			double outGR = aGR[countG-((countG > delayG)?delayG+1:0)];
			double outHR = aHR[countH-((countH > delayH)?delayH+1:0)];
			//third block: final outputs
			
			feedbackAR = (outEL - (outFL + outGL + outHL));
			feedbackBL = (outFL - (outEL + outGL + outHL));
			feedbackCR = (outGL - (outEL + outFL + outHL));
			feedbackDL = (outHL - (outEL + outFL + outGL));
			feedbackAL = (outER - (outFR + outGR + outHR));
			feedbackBR = (outFR - (outER + outGR + outHR));
			feedbackCL = (outGR - (outER + outFR + outHR));
			feedbackDR = (outHR - (outER + outFR + outGR));
			//which we need to feed back into the input again, a bit
			
			inputSampleL = (outEL + outFL + outGL + outHL)/8.0;
			inputSampleR = (outER + outFR + outGR + outHR)/8.0;
			//and take the final combined sum of outputs
			if (cycleEnd == 4) {
				lastRefL[0] = lastRefL[4]; //start from previous last
				lastRefL[2] = (lastRefL[0] + inputSampleL)/2; //half
				lastRefL[1] = (lastRefL[0] + lastRefL[2])/2; //one quarter
				lastRefL[3] = (lastRefL[2] + inputSampleL)/2; //three quarters
				lastRefL[4] = inputSampleL; //full
				lastRefR[0] = lastRefR[4]; //start from previous last
				lastRefR[2] = (lastRefR[0] + inputSampleR)/2; //half
				lastRefR[1] = (lastRefR[0] + lastRefR[2])/2; //one quarter
				lastRefR[3] = (lastRefR[2] + inputSampleR)/2; //three quarters
				lastRefR[4] = inputSampleR; //full
			}
			if (cycleEnd == 3) {
				lastRefL[0] = lastRefL[3]; //start from previous last
				lastRefL[2] = (lastRefL[0]+lastRefL[0]+inputSampleL)/3; //third
				lastRefL[1] = (lastRefL[0]+inputSampleL+inputSampleL)/3; //two thirds
				lastRefL[3] = inputSampleL; //full
				lastRefR[0] = lastRefR[3]; //start from previous last
				lastRefR[2] = (lastRefR[0]+lastRefR[0]+inputSampleR)/3; //third
				lastRefR[1] = (lastRefR[0]+inputSampleR+inputSampleR)/3; //two thirds
				lastRefR[3] = inputSampleR; //full
			}
			if (cycleEnd == 2) {
				lastRefL[0] = lastRefL[2]; //start from previous last
				lastRefL[1] = (lastRefL[0] + inputSampleL)/2; //half
				lastRefL[2] = inputSampleL; //full
				lastRefR[0] = lastRefR[2]; //start from previous last
				lastRefR[1] = (lastRefR[0] + inputSampleR)/2; //half
				lastRefR[2] = inputSampleR; //full
			}
			if (cycleEnd == 1) {
				lastRefL[0] = inputSampleL;
				lastRefR[0] = inputSampleR;
			}
			cycle = 0; //reset
			inputSampleL = lastRefL[cycle];
			inputSampleR = lastRefR[cycle];
		} else {
			inputSampleL = lastRefL[cycle];
			inputSampleR = lastRefR[cycle];
			//we are going through our references now
		}
		
		switch (cycleEnd) //multi-pole average using lastRef[] variables
		{
			case 4:
				lastRefL[8] = inputSampleL; inputSampleL = (inputSampleL+lastRefL[7])*0.5;
				lastRefL[7] = lastRefL[8]; //continue, do not break
				lastRefR[8] = inputSampleR; inputSampleR = (inputSampleR+lastRefR[7])*0.5;
				lastRefR[7] = lastRefR[8]; //continue, do not break
			case 3:
				lastRefL[8] = inputSampleL; inputSampleL = (inputSampleL+lastRefL[6])*0.5;
				lastRefL[6] = lastRefL[8]; //continue, do not break
				lastRefR[8] = inputSampleR; inputSampleR = (inputSampleR+lastRefR[6])*0.5;
				lastRefR[6] = lastRefR[8]; //continue, do not break
			case 2:
				lastRefL[8] = inputSampleL; inputSampleL = (inputSampleL+lastRefL[5])*0.5;
				lastRefL[5] = lastRefL[8]; //continue, do not break
				lastRefR[8] = inputSampleR; inputSampleR = (inputSampleR+lastRefR[5])*0.5;
				lastRefR[5] = lastRefR[8]; //continue, do not break
			case 1:
				break; //no further averaging
		}
		
		if (wet < 1.0) {inputSampleL *= wet; inputSampleR *= wet;}
		if (dry < 1.0) {drySampleL *= dry; drySampleR *= dry;}
		inputSampleL += drySampleL;
		inputSampleR += drySampleR;
		//this is our submix verb dry/wet: 0.5 is BOTH at FULL VOLUME
		//purpose is that, if you're adding verb, you're not altering other balances
		
		//begin 64 bit stereo floating point dither
		//int expon; frexp((double)inputSampleL, &expon);
		fpdL ^= fpdL << 13; fpdL ^= fpdL >> 17; fpdL ^= fpdL << 5;
		//inputSampleL += ((double(fpdL)-uint32_t(0x7fffffff)) * 1.1e-44l * pow(2,expon+62));
		//frexp((double)inputSampleR, &expon);
		fpdR ^= fpdR << 13; fpdR ^= fpdR >> 17; fpdR ^= fpdR << 5;
		//inputSampleR += ((double(fpdR)-uint32_t(0x7fffffff)) * 1.1e-44l * pow(2,expon+62));
		//end 64 bit stereo floating point dither
		
		*out1 = inputSampleL;
		*out2 = inputSampleR;

		in1++;
		in2++;
		out1++;
		out2++;
    }
}