/***************************************************************************
 *                                                                         *
 *   LinuxSampler - modular, streaming capable sampler                     *
 *                                                                         *
 *   Copyright (C) 2003,2004 by Benno Senoner and Christian Schoenebeck    *
 *   Copyright (C) 2005-2020 Christian Schoenebeck                         *
 *   Copyright (C) 2009-2012 Grigor Iliev                                  *
 *   Copyright (C) 2013-2017 Andreas Persson                               *
 *                                                                         *
 *   This program is free software; you can redistribute it and/or modify  *
 *   it under the terms of the GNU General Public License as published by  *
 *   the Free Software Foundation; either version 2 of the License, or     *
 *   (at your option) any later version.                                   *
 *                                                                         *
 *   This program is distributed in the hope that it will be useful,       *
 *   but WITHOUT ANY WARRANTY; without even the implied warranty of        *
 *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         *
 *   GNU General Public License for more details.                          *
 *                                                                         *
 *   You should have received a copy of the GNU General Public License     *
 *   along with this program; if not, write to the Free Software           *
 *   Foundation, Inc., 59 Temple Place, Suite 330, Boston,                 *
 *   MA  02111-1307  USA                                                   *
 ***************************************************************************/

#include "AbstractVoice.h"

namespace LinuxSampler {

    AbstractVoice::AbstractVoice(SignalUnitRack* pRack): pSignalUnitRack(pRack) {
        pEngineChannel = NULL;
        pLFO1 = new LFOClusterUnsigned(1.0f);  // amplitude LFO (0..1 range)
        pLFO2 = new LFOClusterUnsigned(1.0f);  // filter LFO (0..1 range)
        pLFO3 = new LFOClusterSigned(1200.0f); // pitch LFO (-1200..+1200 range)
        PlaybackState = playback_state_end;
        SynthesisMode = 0; // set all mode bits to 0 first
        // select synthesis implementation (asm core is not supported ATM)
        #if 0 // CONFIG_ASM && ARCH_X86
        SYNTHESIS_MODE_SET_IMPLEMENTATION(SynthesisMode, Features::supportsMMX() && Features::supportsSSE());
        #else
        SYNTHESIS_MODE_SET_IMPLEMENTATION(SynthesisMode, false);
        #endif
        SYNTHESIS_MODE_SET_PROFILING(SynthesisMode, gig::Profiler::isEnabled());

        finalSynthesisParameters.filterLeft.Reset();
        finalSynthesisParameters.filterRight.Reset();
        
        pEq          = NULL;
        bEqSupport   = false;
    }

    AbstractVoice::~AbstractVoice() {
        if (pLFO1) delete pLFO1;
        if (pLFO2) delete pLFO2;
        if (pLFO3) delete pLFO3;
        
        if(pEq != NULL) delete pEq;
    }
            
    void AbstractVoice::CreateEq() {
        if(!bEqSupport) return;
        if(pEq != NULL) delete pEq;
        pEq = new EqSupport;
        pEq->InitEffect(GetEngine()->pAudioOutputDevice);
    }

    /**
     *  Resets voice variables. Should only be called if rendering process is
     *  suspended / not running.
     */
    void AbstractVoice::Reset() {
        finalSynthesisParameters.filterLeft.Reset();
        finalSynthesisParameters.filterRight.Reset();
        DiskStreamRef.pStream = NULL;
        DiskStreamRef.hStream = 0;
        DiskStreamRef.State   = Stream::state_unused;
        DiskStreamRef.OrderID = 0;
        PlaybackState = playback_state_end;
        itTriggerEvent = Pool<Event>::Iterator();
        itKillEvent    = Pool<Event>::Iterator();
    }

    /**
     *  Initializes and triggers the voice, a disk stream will be launched if
     *  needed.
     *
     *  @param pEngineChannel - engine channel on which this voice was ordered
     *  @param itNoteOnEvent  - event that caused triggering of this voice
     *  @param PitchBend      - MIDI detune factor (-8192 ... +8191)
     *  @param pRegion- points to the region which provides sample wave(s) and articulation data
     *  @param VoiceType      - type of this voice
     *  @param iKeyGroup      - a value > 0 defines a key group in which this voice is member of
     *  @returns 0 on success, a value < 0 if the voice wasn't triggered
     *           (either due to an error or e.g. because no region is
     *           defined for the given key)
     */
    int AbstractVoice::Trigger (
        AbstractEngineChannel*  pEngineChannel,
        Pool<Event>::Iterator&  itNoteOnEvent,
        int                     PitchBend,
        type_t                  VoiceType,
        int                     iKeyGroup
    ) {
        this->pEngineChannel = pEngineChannel;
        Orphan = false;

        #if CONFIG_DEVMODE
        if (itNoteOnEvent->FragmentPos() > GetEngine()->MaxSamplesPerCycle) { // just a sanity check for debugging
            dmsg(1,("Voice::Trigger(): ERROR, TriggerDelay > Totalsamples\n"));
        }
        #endif // CONFIG_DEVMODE

        Type            = VoiceType;
        pNote           = pEngineChannel->pEngine->NoteByID( itNoteOnEvent->Param.Note.ID );
        PlaybackState   = playback_state_init; // mark voice as triggered, but no audio rendered yet
        Delay           = itNoteOnEvent->FragmentPos();
        itTriggerEvent  = itNoteOnEvent;
        itKillEvent     = Pool<Event>::Iterator();
        MidiKeyBase* pKeyInfo = GetMidiKeyInfo(MIDIKey());

        // when editing key groups with an instrument editor while sound was
        // already loaded, ActiveKeyGroups may not have the KeyGroup in question
        // so use find() here instead of array subscript operator[] to avoid an
        // implied creation of a NULL entry, to prevent a crash while editing
        // instruments
        {
            AbstractEngineChannel::ActiveKeyGroupMap::const_iterator it =
                pEngineChannel->ActiveKeyGroups.find(iKeyGroup);
            pGroupEvents =
                (iKeyGroup && it != pEngineChannel->ActiveKeyGroups.end()) ?
                    it->second : NULL;
        }

        SmplInfo   = GetSampleInfo();
        RgnInfo    = GetRegionInfo();
        InstrInfo  = GetInstrumentInfo();
        
        MIDIPan    = CalculatePan(pEngineChannel->iLastPanRequest);

        AboutToTrigger();

        // calculate volume
        const double velocityAttenuation = GetVelocityAttenuation(MIDIVelocity());
        float volume = CalculateVolume(velocityAttenuation) * pKeyInfo->Volume;
        if (volume <= 0) return -1;

        // select channel mode (mono or stereo)
        SYNTHESIS_MODE_SET_CHANNELS(SynthesisMode, SmplInfo.ChannelCount == 2);
        // select bit depth (16 or 24)
        SYNTHESIS_MODE_SET_BITDEPTH24(SynthesisMode, SmplInfo.BitDepth == 24);

        // get starting crossfade volume level
        float crossfadeVolume = CalculateCrossfadeVolume(MIDIVelocity());

        VolumeLeft  = volume * pKeyInfo->PanLeft;
        VolumeRight = volume * pKeyInfo->PanRight;

        // this rate is used for rather mellow volume fades
        const float subfragmentRate = GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE;
        // this rate is used for very fast volume fades
        const float quickRampRate = RTMath::Min(subfragmentRate, GetEngine()->SampleRate * 0.001f /* approx. 13ms */);
        CrossfadeSmoother.trigger(crossfadeVolume, subfragmentRate);

        VolumeSmoother.trigger(pEngineChannel->MidiVolume, subfragmentRate);
        NoteVolume.setCurveOnly(pNote ? pNote->Override.VolumeCurve : DEFAULT_FADE_CURVE);
        NoteVolume.setCurrentValue(pNote ? pNote->Override.Volume.Value : 1.f);
        NoteVolume.setDefaultDuration(pNote ? pNote->Override.VolumeTime : DEFAULT_NOTE_VOLUME_TIME_S);
        NoteVolume.setFinal(pNote ? pNote->Override.Volume.Final : false);

        // Check if the sample needs disk streaming or is too short for that
        long cachedsamples = GetSampleCacheSize() / SmplInfo.FrameSize;
        DiskVoice          = cachedsamples < SmplInfo.TotalFrameCount;

        SetSampleStartOffset();

        if (DiskVoice) { // voice to be streamed from disk
            if (cachedsamples > (GetEngine()->MaxSamplesPerCycle << CONFIG_MAX_PITCH)) {
                //TODO: this calculation is too pessimistic
                MaxRAMPos = cachedsamples - (GetEngine()->MaxSamplesPerCycle << CONFIG_MAX_PITCH);
            } else {
                // The cache is too small to fit a max sample buffer.
                // Setting MaxRAMPos to 0 will probably cause a click
                // in the audio, but it's better than not handling
                // this case at all, which would have caused the
                // unsigned MaxRAMPos to be set to a negative number.
                MaxRAMPos = 0;
            }

            // check if there's a loop defined which completely fits into the cached (RAM) part of the sample
            RAMLoop = (SmplInfo.HasLoops && (SmplInfo.LoopStart + SmplInfo.LoopLength) <= MaxRAMPos);

            if (OrderNewStream()) return -1;
            dmsg(4,("Disk voice launched (cached samples: %ld, total Samples: %d, MaxRAMPos: %lu, RAMLooping: %s)\n", cachedsamples, SmplInfo.TotalFrameCount, MaxRAMPos, (RAMLoop) ? "yes" : "no"));
        }
        else { // RAM only voice
            MaxRAMPos = cachedsamples;
            RAMLoop = (SmplInfo.HasLoops);
            dmsg(4,("RAM only voice launched (Looping: %s)\n", (RAMLoop) ? "yes" : "no"));
        }
        if (RAMLoop) {
            loop.uiTotalCycles = SmplInfo.LoopPlayCount;
            loop.uiCyclesLeft  = SmplInfo.LoopPlayCount;
            loop.uiStart       = SmplInfo.LoopStart;
            loop.uiEnd         = SmplInfo.LoopStart + SmplInfo.LoopLength;
            loop.uiSize        = SmplInfo.LoopLength;
        }

        Pitch = CalculatePitchInfo(PitchBend);
        NotePitch.setCurveOnly(pNote ? pNote->Override.PitchCurve : DEFAULT_FADE_CURVE);
        NotePitch.setCurrentValue(pNote ? pNote->Override.Pitch.Value : 1.0f);
        NotePitch.setFinal(pNote ? pNote->Override.Pitch.Final : false);
        NotePitch.setDefaultDuration(pNote ? pNote->Override.PitchTime : DEFAULT_NOTE_PITCH_TIME_S);
        NoteCutoff.Value = (pNote) ? pNote->Override.Cutoff.Value : 1.0f;
        NoteCutoff.Final = (pNote) ? pNote->Override.Cutoff.isFinal() : false;
        NoteResonance.Value = (pNote) ? pNote->Override.Resonance.Value : 1.0f;
        NoteResonance.Final = (pNote) ? pNote->Override.Resonance.Final : false;

        // the length of the decay and release curves are dependent on the velocity
        const double velrelease = 1 / GetVelocityRelease(MIDIVelocity());

        if (pSignalUnitRack == NULL) { // setup EG 1 (VCA EG)
            // get current value of EG1 controller
            double eg1controllervalue = GetEG1ControllerValue(MIDIVelocity());

            // calculate influence of EG1 controller on EG1's parameters
            EGInfo egInfo = CalculateEG1ControllerInfluence(eg1controllervalue);

            if (pNote) {
                pNote->Override.Attack.applyTo(egInfo.Attack);
                pNote->Override.Decay.applyTo(egInfo.Decay);
                pNote->Override.Release.applyTo(egInfo.Release);
            }

            TriggerEG1(egInfo, velrelease, velocityAttenuation, GetEngine()->SampleRate, MIDIVelocity());
        } else {
            pSignalUnitRack->Trigger();
        }

        const uint8_t pan = (pSignalUnitRack) ? pSignalUnitRack->GetEndpointUnit()->CalculatePan(MIDIPan) : MIDIPan;
        for (int c = 0; c < 2; ++c) {
            float value = (pNote) ? AbstractEngine::PanCurveValueNorm(pNote->Override.Pan.Value, c) : 1.f;
            NotePan[c].setCurveOnly(pNote ? pNote->Override.PanCurve : DEFAULT_FADE_CURVE);
            NotePan[c].setCurrentValue(value);
            NotePan[c].setFinal(pNote ? pNote->Override.Pan.Final : false);
            NotePan[c].setDefaultDuration(pNote ? pNote->Override.PanTime : DEFAULT_NOTE_PAN_TIME_S);
        }

        PanLeftSmoother.trigger(
            AbstractEngine::PanCurve[128 - pan],
            quickRampRate //NOTE: maybe we should have 2 separate pan smoothers, one for MIDI CC10 (with slow rate) and one for instrument script change_pan() calls (with fast rate)
        );
        PanRightSmoother.trigger(
            AbstractEngine::PanCurve[pan],
            quickRampRate //NOTE: maybe we should have 2 separate pan smoothers, one for MIDI CC10 (with slow rate) and one for instrument script change_pan() calls (with fast rate)
        );

#if CONFIG_INTERPOLATE_VOLUME
        // setup initial volume in synthesis parameters
    #if CONFIG_PROCESS_MUTED_CHANNELS
        if (pEngineChannel->GetMute()) {
            finalSynthesisParameters.fFinalVolumeLeft  = 0;
            finalSynthesisParameters.fFinalVolumeRight = 0;
        }
        else
    #else
        {
            float finalVolume = pEngineChannel->MidiVolume * crossfadeVolume;
            float fModVolume;
            if (pSignalUnitRack == NULL) {
                fModVolume = pEG1->getLevel();
            } else {
                fModVolume = pSignalUnitRack->GetEndpointUnit()->GetVolume();
            }
            NoteVolume.applyCurrentValueTo(fModVolume);
            finalVolume *= fModVolume;

            float panL = PanLeftSmoother.render();
            float panR = PanRightSmoother.render();
            NotePan[0].applyCurrentValueTo(panL);
            NotePan[1].applyCurrentValueTo(panR);

            finalSynthesisParameters.fFinalVolumeLeft  = finalVolume * VolumeLeft  * panL;
            finalSynthesisParameters.fFinalVolumeRight = finalVolume * VolumeRight * panR;
        }
    #endif
#endif

        if (pSignalUnitRack == NULL) {
            // setup EG 2 (VCF Cutoff EG)
            {
                // get current value of EG2 controller
                double eg2controllervalue = GetEG2ControllerValue(MIDIVelocity());

                // calculate influence of EG2 controller on EG2's parameters
                EGInfo egInfo = CalculateEG2ControllerInfluence(eg2controllervalue);

                if (pNote) {
                    pNote->Override.CutoffAttack.applyTo(egInfo.Attack);
                    pNote->Override.CutoffDecay.applyTo(egInfo.Decay);
                    pNote->Override.CutoffRelease.applyTo(egInfo.Release);
                }

                TriggerEG2(egInfo, velrelease, velocityAttenuation, GetEngine()->SampleRate, MIDIVelocity());
            }


            // setup EG 3 (VCO EG)
            {
                // if portamento mode is on, we dedicate EG3 purely for portamento, otherwise if portamento is off we do as told by the patch
                bool  bPortamento = pEngineChannel->PortamentoMode && pEngineChannel->PortamentoPos >= 0.0f;
                float eg3depth = (bPortamento)
                             ? RTMath::CentsToFreqRatio((pEngineChannel->PortamentoPos - (float) MIDIKey()) * 100)
                             : RTMath::CentsToFreqRatio(RgnInfo.EG3Depth);
                float eg3time = (bPortamento)
                            ? pEngineChannel->PortamentoTime
                            : RgnInfo.EG3Attack;
                EG3.trigger(eg3depth, eg3time, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                dmsg(5,("PortamentoPos=%f, depth=%f, time=%f\n", pEngineChannel->PortamentoPos, eg3depth, eg3time));
            }


            // setup LFO 1 (VCA LFO)
            InitLFO1();
            // setup LFO 2 (VCF Cutoff LFO)
            InitLFO2();
            // setup LFO 3 (VCO LFO)
            InitLFO3();
        }


        #if CONFIG_FORCE_FILTER
        const bool bUseFilter = true;
        #else // use filter only if instrument file told so
        const bool bUseFilter = RgnInfo.VCFEnabled;
        #endif // CONFIG_FORCE_FILTER
        SYNTHESIS_MODE_SET_FILTER(SynthesisMode, bUseFilter);
        if (bUseFilter) {
            #ifdef CONFIG_OVERRIDE_CUTOFF_CTRL
            VCFCutoffCtrl.controller = CONFIG_OVERRIDE_CUTOFF_CTRL;
            #else // use the one defined in the instrument file
            VCFCutoffCtrl.controller = GetVCFCutoffCtrl();
            #endif // CONFIG_OVERRIDE_CUTOFF_CTRL

            #ifdef CONFIG_OVERRIDE_RESONANCE_CTRL
            VCFResonanceCtrl.controller = CONFIG_OVERRIDE_RESONANCE_CTRL;
            #else // use the one defined in the instrument file
            VCFResonanceCtrl.controller = GetVCFResonanceCtrl();
            #endif // CONFIG_OVERRIDE_RESONANCE_CTRL

            #ifndef CONFIG_OVERRIDE_FILTER_TYPE
            finalSynthesisParameters.filterLeft.SetType(RgnInfo.VCFType);
            finalSynthesisParameters.filterRight.SetType(RgnInfo.VCFType);
            #else // override filter type
            finalSynthesisParameters.filterLeft.SetType(CONFIG_OVERRIDE_FILTER_TYPE);
            finalSynthesisParameters.filterRight.SetType(CONFIG_OVERRIDE_FILTER_TYPE);
            #endif // CONFIG_OVERRIDE_FILTER_TYPE

            VCFCutoffCtrl.value    = pEngineChannel->ControllerTable[VCFCutoffCtrl.controller];
            VCFResonanceCtrl.value = pEngineChannel->ControllerTable[VCFResonanceCtrl.controller];

            // calculate cutoff frequency
            CutoffBase = CalculateCutoffBase(MIDIVelocity());

            VCFCutoffCtrl.fvalue = CalculateFinalCutoff(CutoffBase);

            // calculate resonance
            float resonance = (float) (VCFResonanceCtrl.controller ? VCFResonanceCtrl.value : RgnInfo.VCFResonance);
            VCFResonanceCtrl.fvalue = resonance;
        } else {
            VCFCutoffCtrl.controller    = 0;
            VCFResonanceCtrl.controller = 0;
        }
        
        const bool bEq =
            pSignalUnitRack != NULL && pSignalUnitRack->HasEq() && pEq->HasSupport();

        if (bEq) {
            pEq->GetInChannelLeft()->Clear();
            pEq->GetInChannelRight()->Clear();
            pEq->RenderAudio(GetEngine()->pAudioOutputDevice->MaxSamplesPerCycle());
        }

        return 0; // success
    }
    
    void AbstractVoice::SetSampleStartOffset() {
        double pos = RgnInfo.SampleStartOffset; // offset where we should start playback of sample

        // if another sample playback start position was requested by instrument
        // script (built-in script function play_note())
        if (pNote && pNote->Override.SampleOffset >= 0) {
            double overridePos =
                double(SmplInfo.SampleRate) * double(pNote->Override.SampleOffset) / 1000000.0;
            if (overridePos < SmplInfo.TotalFrameCount)
                pos = overridePos;
        }

        finalSynthesisParameters.dPos = pos;
        Pos = pos;
    }

    /**
     *  Synthesizes the current audio fragment for this voice.
     *
     *  @param Samples - number of sample points to be rendered in this audio
     *                   fragment cycle
     *  @param pSrc    - pointer to input sample data
     *  @param Skip    - number of sample points to skip in output buffer
     */
    void AbstractVoice::Synthesize(uint Samples, sample_t* pSrc, uint Skip) {
        bool delay = false; // Whether the voice playback should be delayed for this call
        
        if (pSignalUnitRack != NULL) {
            uint delaySteps = pSignalUnitRack->GetEndpointUnit()->DelayTrigger();
            if (delaySteps > 0) { // delay on the endpoint unit means delay of the voice playback
                if (delaySteps >= Samples) {
                    pSignalUnitRack->GetEndpointUnit()->DecreaseDelay(Samples);
                    delay = true;
                } else {
                    pSignalUnitRack->GetEndpointUnit()->DecreaseDelay(delaySteps);
                    Samples -= delaySteps;
                    Skip += delaySteps;
                }
            }
        }
        
        AbstractEngineChannel* pChannel = pEngineChannel;
        MidiKeyBase* pMidiKeyInfo = GetMidiKeyInfo(MIDIKey());

        const bool bVoiceRequiresDedicatedRouting =
            pEngineChannel->GetFxSendCount() > 0 &&
            (pMidiKeyInfo->ReverbSend || pMidiKeyInfo->ChorusSend);
        
        const bool bEq =
            pSignalUnitRack != NULL && pSignalUnitRack->HasEq() && pEq->HasSupport();

        if (bEq) {
            pEq->GetInChannelLeft()->Clear();
            pEq->GetInChannelRight()->Clear();
            finalSynthesisParameters.pOutLeft  = &pEq->GetInChannelLeft()->Buffer()[Skip];
            finalSynthesisParameters.pOutRight = &pEq->GetInChannelRight()->Buffer()[Skip];
            pSignalUnitRack->UpdateEqSettings(pEq);
        } else if (bVoiceRequiresDedicatedRouting) {
            finalSynthesisParameters.pOutLeft  = &GetEngine()->pDedicatedVoiceChannelLeft->Buffer()[Skip];
            finalSynthesisParameters.pOutRight = &GetEngine()->pDedicatedVoiceChannelRight->Buffer()[Skip];
        } else {
            finalSynthesisParameters.pOutLeft  = &pChannel->pChannelLeft->Buffer()[Skip];
            finalSynthesisParameters.pOutRight = &pChannel->pChannelRight->Buffer()[Skip];
        }
        finalSynthesisParameters.pSrc = pSrc;

        RTList<Event>::Iterator itCCEvent = pChannel->pEvents->first();
        RTList<Event>::Iterator itNoteEvent;
        GetFirstEventOnKey(HostKey(), itNoteEvent);

        RTList<Event>::Iterator itGroupEvent;
        if (pGroupEvents && !Orphan) itGroupEvent = pGroupEvents->first();

        if (itTriggerEvent) { // skip events that happened before this voice was triggered
            while (itCCEvent && itCCEvent->FragmentPos() <= Skip) ++itCCEvent;
            while (itGroupEvent && itGroupEvent->FragmentPos() <= Skip) ++itGroupEvent;

            // we can't simply compare the timestamp here, because note events
            // might happen on the same time stamp, so we have to deal on the
            // actual sequence the note events arrived instead (see bug #112)
            for (; itNoteEvent; ++itNoteEvent) {
                if (itTriggerEvent == itNoteEvent) {
                    ++itNoteEvent;
                    break;
                }
            }
        }

        uint killPos = 0;
        if (itKillEvent) {
            int maxFadeOutPos = Samples - GetEngine()->GetMinFadeOutSamples();
            if (maxFadeOutPos < 0) {
                // There's not enough space in buffer to do a fade out
                // from max volume (this can only happen for audio
                // drivers that use Samples < MaxSamplesPerCycle).
                // End the EG1 here, at pos 0, with a shorter max fade
                // out time.
                if (pSignalUnitRack == NULL) {
                    pEG1->enterFadeOutStage(Samples / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                } else {
                    pSignalUnitRack->EnterFadeOutStage(Samples / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                }
                itKillEvent = Pool<Event>::Iterator();
            } else {
                killPos = RTMath::Min(itKillEvent->FragmentPos(), maxFadeOutPos);
            }
        }

        uint i = Skip;
        while (i < Samples) {
            int iSubFragmentEnd = RTMath::Min(i + CONFIG_DEFAULT_SUBFRAGMENT_SIZE, Samples);

            // initialize all final synthesis parameters
            fFinalCutoff    = VCFCutoffCtrl.fvalue;
            fFinalResonance = VCFResonanceCtrl.fvalue;

            // process MIDI control change, aftertouch and pitchbend events for this subfragment
            processCCEvents(itCCEvent, iSubFragmentEnd);
            uint8_t pan = MIDIPan;
            if (pSignalUnitRack != NULL) pan = pSignalUnitRack->GetEndpointUnit()->CalculatePan(MIDIPan);

            PanLeftSmoother.update(AbstractEngine::PanCurve[128 - pan]);
            PanRightSmoother.update(AbstractEngine::PanCurve[pan]);

            finalSynthesisParameters.fFinalPitch = Pitch.PitchBase * Pitch.PitchBend;

            float fFinalVolume = VolumeSmoother.render() * CrossfadeSmoother.render();
#if CONFIG_PROCESS_MUTED_CHANNELS
            if (pChannel->GetMute()) fFinalVolume = 0;
#endif

            // process transition events (note on, note off & sustain pedal)
            processTransitionEvents(itNoteEvent, iSubFragmentEnd);
            processGroupEvents(itGroupEvent, iSubFragmentEnd);

            float fModVolume = 1;
            float fModPitch  = 1;

            if (pSignalUnitRack == NULL) {
                // if the voice was killed in this subfragment, or if the
                // filter EG is finished, switch EG1 to fade out stage
                if ((itKillEvent && killPos <= iSubFragmentEnd) ||
                    (SYNTHESIS_MODE_GET_FILTER(SynthesisMode) &&
                    pEG2->getSegmentType() == EG::segment_end)) {
                    pEG1->enterFadeOutStage();
                    itKillEvent = Pool<Event>::Iterator();
                }

                // process envelope generators
                switch (pEG1->getSegmentType()) {
                    case EG::segment_lin:
                        fModVolume *= pEG1->processLin();
                        break;
                    case EG::segment_exp:
                        fModVolume *= pEG1->processExp();
                        break;
                    case EG::segment_end:
                        fModVolume *= pEG1->getLevel();
                        break; // noop
                    case EG::segment_pow:
                        fModVolume *= pEG1->processPow();
                        break;
                }
                switch (pEG2->getSegmentType()) {
                    case EG::segment_lin:
                        fFinalCutoff *= pEG2->processLin();
                        break;
                    case EG::segment_exp:
                        fFinalCutoff *= pEG2->processExp();
                        break;
                    case EG::segment_end:
                        fFinalCutoff *= pEG2->getLevel();
                        break; // noop
                    case EG::segment_pow:
                        fFinalCutoff *= pEG2->processPow();
                        break;
                }
                if (EG3.active()) fModPitch *= EG3.render();

                // process low frequency oscillators
                if (bLFO1Enabled) fModVolume   *= (1.0f - pLFO1->render());
                if (bLFO2Enabled) fFinalCutoff *= (1.0f - pLFO2->render());
                if (bLFO3Enabled) fModPitch *= RTMath::CentsToFreqRatio(pLFO3->render());
            } else {
                // if the voice was killed in this subfragment, enter fade out stage
                if (itKillEvent && killPos <= iSubFragmentEnd) {
                    pSignalUnitRack->EnterFadeOutStage();
                    itKillEvent = Pool<Event>::Iterator();
                }
                
                // if the filter EG is finished, switch EG1 to fade out stage
                /*if (SYNTHESIS_MODE_GET_FILTER(SynthesisMode) &&
                    pEG2->getSegmentType() == EG::segment_end) {
                    pEG1->enterFadeOutStage();
                    itKillEvent = Pool<Event>::Iterator();
                }*/
                // TODO: ^^^

                fFinalVolume   *= pSignalUnitRack->GetEndpointUnit()->GetVolume();
                fFinalCutoff    = pSignalUnitRack->GetEndpointUnit()->CalculateFilterCutoff(fFinalCutoff);
                fFinalResonance = pSignalUnitRack->GetEndpointUnit()->CalculateResonance(fFinalResonance);
                
                fModPitch = pSignalUnitRack->GetEndpointUnit()->CalculatePitch(fModPitch);
            }

            NoteVolume.renderApplyTo(fModVolume);
            NotePitch.renderApplyTo(fModPitch);
            NoteCutoff.applyTo(fFinalCutoff);
            NoteResonance.applyTo(fFinalResonance);

            fFinalVolume *= fModVolume;

            finalSynthesisParameters.fFinalPitch *= fModPitch;

            // limit the pitch so we don't read outside the buffer
            finalSynthesisParameters.fFinalPitch = RTMath::Min(finalSynthesisParameters.fFinalPitch, float(1 << CONFIG_MAX_PITCH));

            // if filter enabled then update filter coefficients
            if (SYNTHESIS_MODE_GET_FILTER(SynthesisMode)) {
                finalSynthesisParameters.filterLeft.SetParameters(fFinalCutoff, fFinalResonance, GetEngine()->SampleRate);
                finalSynthesisParameters.filterRight.SetParameters(fFinalCutoff, fFinalResonance, GetEngine()->SampleRate);
            }

            // do we need resampling?
            const float __PLUS_ONE_CENT  = 1.000577789506554859250142541782224725466f;
            const float __MINUS_ONE_CENT = 0.9994225441413807496009516495583113737666f;
            const bool bResamplingRequired = !(finalSynthesisParameters.fFinalPitch <= __PLUS_ONE_CENT &&
                                               finalSynthesisParameters.fFinalPitch >= __MINUS_ONE_CENT);
            SYNTHESIS_MODE_SET_INTERPOLATE(SynthesisMode, bResamplingRequired);

            // prepare final synthesis parameters structure
            finalSynthesisParameters.uiToGo            = iSubFragmentEnd - i;

            float panL = PanLeftSmoother.render();
            float panR = PanRightSmoother.render();
            NotePan[0].renderApplyTo(panL);
            NotePan[1].renderApplyTo(panR);

#if CONFIG_INTERPOLATE_VOLUME
            finalSynthesisParameters.fFinalVolumeDeltaLeft  =
                (fFinalVolume * VolumeLeft  * panL -
                 finalSynthesisParameters.fFinalVolumeLeft) / finalSynthesisParameters.uiToGo;
            finalSynthesisParameters.fFinalVolumeDeltaRight =
                (fFinalVolume * VolumeRight * panR -
                 finalSynthesisParameters.fFinalVolumeRight) / finalSynthesisParameters.uiToGo;
#else
            finalSynthesisParameters.fFinalVolumeLeft  =
                fFinalVolume * VolumeLeft  * panL;
            finalSynthesisParameters.fFinalVolumeRight =
                fFinalVolume * VolumeRight * panR;
#endif
            // render audio for one subfragment
            if (!delay) RunSynthesisFunction(SynthesisMode, &finalSynthesisParameters, &loop);

            if (pSignalUnitRack == NULL) {
                // stop the rendering if volume EG is finished
                if (pEG1->getSegmentType() == EG::segment_end) break;
            } else {
                // stop the rendering if the endpoint unit is not active
                if (!pSignalUnitRack->GetEndpointUnit()->Active()) break;
            }

            const double newPos = Pos + (iSubFragmentEnd - i) * finalSynthesisParameters.fFinalPitch;

            if (pSignalUnitRack == NULL) {
                // increment envelopes' positions
                if (pEG1->active()) {

                    // if sample has a loop and loop start has been reached in this subfragment, send a special event to EG1 to let it finish the attack hold stage
                    if (SmplInfo.HasLoops && Pos <= SmplInfo.LoopStart && SmplInfo.LoopStart < newPos) {
                        pEG1->update(EG::event_hold_end, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                    }

                    pEG1->increment(1);
                    if (!pEG1->toStageEndLeft()) pEG1->update(EG::event_stage_end, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                }
                if (pEG2->active()) {
                    pEG2->increment(1);
                    if (!pEG2->toStageEndLeft()) pEG2->update(EG::event_stage_end, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                }
                EG3.increment(1);
                if (!EG3.toEndLeft()) EG3.update(); // neutralize envelope coefficient if end reached
            } else {
                    // if sample has a loop and loop start has been reached in this subfragment, send a special event to EG1 to let it finish the attack hold stage
                    /*if (SmplInfo.HasLoops && Pos <= SmplInfo.LoopStart && SmplInfo.LoopStart < newPos) {
                        pEG1->update(EG::event_hold_end, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                    }*/
                // TODO: ^^^
                
                if (!delay) pSignalUnitRack->Increment();
            }

            Pos = newPos;
            i = iSubFragmentEnd;
        }
        
        if (delay) return;

        if (bVoiceRequiresDedicatedRouting) {
            if (bEq) {
                pEq->RenderAudio(Samples);
                pEq->GetOutChannelLeft()->CopyTo(GetEngine()->pDedicatedVoiceChannelLeft, Samples);
                pEq->GetOutChannelRight()->CopyTo(GetEngine()->pDedicatedVoiceChannelRight, Samples);
            }
            optional<float> effectSendLevels[2] = {
                pMidiKeyInfo->ReverbSend,
                pMidiKeyInfo->ChorusSend
            };
            GetEngine()->RouteDedicatedVoiceChannels(pEngineChannel, effectSendLevels, Samples);
        } else if (bEq) {
            pEq->RenderAudio(Samples);
            pEq->GetOutChannelLeft()->MixTo(pChannel->pChannelLeft, Samples);
            pEq->GetOutChannelRight()->MixTo(pChannel->pChannelRight, Samples);
        }
    }

    /**
     * Process given list of MIDI control change, aftertouch and pitch bend
     * events for the given time.
     *
     * @param itEvent - iterator pointing to the next event to be processed
     * @param End     - youngest time stamp where processing should be stopped
     */
    void AbstractVoice::processCCEvents(RTList<Event>::Iterator& itEvent, uint End) {
        for (; itEvent && itEvent->FragmentPos() <= End; ++itEvent) {
            if ((itEvent->Type == Event::type_control_change || itEvent->Type == Event::type_channel_pressure)
                && itEvent->Param.CC.Controller) // if (valid) MIDI control change event
            {
                if (itEvent->Param.CC.Controller == VCFCutoffCtrl.controller) {
                    ProcessCutoffEvent(itEvent);
                }
                if (itEvent->Param.CC.Controller == VCFResonanceCtrl.controller) {
                    processResonanceEvent(itEvent);
                }
                if (itEvent->Param.CC.Controller == CTRL_TABLE_IDX_AFTERTOUCH ||
                    itEvent->Type == Event::type_channel_pressure)
                {
                    ProcessChannelPressureEvent(itEvent);
                }
                if (pSignalUnitRack == NULL) {
                    if (itEvent->Param.CC.Controller == pLFO1->ExtController) {
                        pLFO1->updateByMIDICtrlValue(itEvent->Param.CC.Value);
                    }
                    if (itEvent->Param.CC.Controller == pLFO2->ExtController) {
                        pLFO2->updateByMIDICtrlValue(itEvent->Param.CC.Value);
                    }
                    if (itEvent->Param.CC.Controller == pLFO3->ExtController) {
                        pLFO3->updateByMIDICtrlValue(itEvent->Param.CC.Value);
                    }
                }
                if (itEvent->Param.CC.Controller == 7) { // volume
                    VolumeSmoother.update(AbstractEngine::VolumeCurve[itEvent->Param.CC.Value]);
                } else if (itEvent->Param.CC.Controller == 10) { // panpot
                    MIDIPan = CalculatePan(itEvent->Param.CC.Value);
                }
            } else if (itEvent->Type == Event::type_pitchbend) { // if pitch bend event
                processPitchEvent(itEvent);
            } else if (itEvent->Type == Event::type_note_pressure) {
                ProcessPolyphonicKeyPressureEvent(itEvent);
            }

            ProcessCCEvent(itEvent);
            if (pSignalUnitRack != NULL) {
                pSignalUnitRack->ProcessCCEvent(itEvent);
            }
        }
    }

    void AbstractVoice::processPitchEvent(RTList<Event>::Iterator& itEvent) {
        Pitch.PitchBend = RTMath::CentsToFreqRatio(itEvent->Param.Pitch.Pitch * Pitch.PitchBendRange);
    }

    void AbstractVoice::processResonanceEvent(RTList<Event>::Iterator& itEvent) {
        // convert absolute controller value to differential
        const int ctrldelta = itEvent->Param.CC.Value - VCFResonanceCtrl.value;
        VCFResonanceCtrl.value = itEvent->Param.CC.Value;
        const float resonancedelta = (float) ctrldelta;
        fFinalResonance += resonancedelta;
        // needed for initialization of parameter
        VCFResonanceCtrl.fvalue = itEvent->Param.CC.Value;
    }

    /**
     * Process given list of MIDI note on, note off, sustain pedal events and
     * note synthesis parameter events for the given time.
     *
     * @param itEvent - iterator pointing to the next event to be processed
     * @param End     - youngest time stamp where processing should be stopped
     */
    void AbstractVoice::processTransitionEvents(RTList<Event>::Iterator& itEvent, uint End) {
        for (; itEvent && itEvent->FragmentPos() <= End; ++itEvent) {
            // some voice types ignore note off 
            if (!(Type & (Voice::type_one_shot | Voice::type_release_trigger | Voice::type_controller_triggered))) {
                if (itEvent->Type == Event::type_release_key) {
                    EnterReleaseStage();
                } else if (itEvent->Type == Event::type_cancel_release_key) {
                    if (pSignalUnitRack == NULL) {
                        pEG1->update(EG::event_cancel_release, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        pEG2->update(EG::event_cancel_release, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                    } else {
                        pSignalUnitRack->CancelRelease();
                    }
                }
            }
            // process stop-note events (caused by built-in instrument script function note_off())
            if (itEvent->Type == Event::type_release_note && pNote &&
                pEngineChannel->pEngine->NoteByID( itEvent->Param.Note.ID ) == pNote)
            {
                EnterReleaseStage();
            }
            // process kill-note events (caused by built-in instrument script function fade_out())
            if (itEvent->Type == Event::type_kill_note && pNote &&
                pEngineChannel->pEngine->NoteByID( itEvent->Param.Note.ID ) == pNote)
            {
                Kill(itEvent);
            }
            // process synthesis parameter events (caused by built-in realt-time instrument script functions)
            if (itEvent->Type == Event::type_note_synth_param && pNote &&
                pEngineChannel->pEngine->NoteByID( itEvent->Param.NoteSynthParam.NoteID ) == pNote)
            {
                switch (itEvent->Param.NoteSynthParam.Type) {
                    case Event::synth_param_volume:
                        NoteVolume.fadeTo(itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        NoteVolume.setFinal(itEvent->Param.NoteSynthParam.isFinal());
                        break;
                    case Event::synth_param_volume_time:
                        NoteVolume.setDefaultDuration(itEvent->Param.NoteSynthParam.AbsValue);
                        break;
                    case Event::synth_param_volume_curve:
                        NoteVolume.setCurve((fade_curve_t)itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        break;
                    case Event::synth_param_pitch:
                        NotePitch.fadeTo(itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        NotePitch.setFinal(itEvent->Param.NoteSynthParam.isFinal());
                        break;
                    case Event::synth_param_pitch_time:
                        NotePitch.setDefaultDuration(itEvent->Param.NoteSynthParam.AbsValue);
                        break;
                    case Event::synth_param_pitch_curve:
                        NotePitch.setCurve((fade_curve_t)itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        break;
                    case Event::synth_param_pan:
                        NotePan[0].fadeTo(
                            AbstractEngine::PanCurveValueNorm(itEvent->Param.NoteSynthParam.AbsValue, 0 /*left*/),
                            GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE
                        );
                        NotePan[1].fadeTo(
                            AbstractEngine::PanCurveValueNorm(itEvent->Param.NoteSynthParam.AbsValue, 1 /*right*/),
                            GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE
                        );
                        NotePan[0].setFinal(itEvent->Param.NoteSynthParam.isFinal());
                        NotePan[1].setFinal(itEvent->Param.NoteSynthParam.isFinal());
                        break;
                    case Event::synth_param_pan_time:
                        NotePan[0].setDefaultDuration(itEvent->Param.NoteSynthParam.AbsValue);
                        NotePan[1].setDefaultDuration(itEvent->Param.NoteSynthParam.AbsValue);
                        break;
                    case Event::synth_param_pan_curve:
                        NotePan[0].setCurve((fade_curve_t)itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        NotePan[1].setCurve((fade_curve_t)itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        break;
                    case Event::synth_param_cutoff:
                        NoteCutoff.Value = itEvent->Param.NoteSynthParam.AbsValue;
                        NoteCutoff.Final = itEvent->Param.NoteSynthParam.isFinal();
                        break;
                    case Event::synth_param_resonance:
                        NoteResonance.Value = itEvent->Param.NoteSynthParam.AbsValue;
                        NoteResonance.Final = itEvent->Param.NoteSynthParam.isFinal();
                        break;
                    case Event::synth_param_amp_lfo_depth:
                        pLFO1->setScriptDepthFactor(
                            itEvent->Param.NoteSynthParam.AbsValue,
                            itEvent->Param.NoteSynthParam.isFinal()
                        );
                        break;
                    case Event::synth_param_amp_lfo_freq:
                        if (itEvent->Param.NoteSynthParam.isFinal())
                            pLFO1->setScriptFrequencyFinal(itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        else
                            pLFO1->setScriptFrequencyFactor(itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        break;
                    case Event::synth_param_cutoff_lfo_depth:
                        pLFO2->setScriptDepthFactor(
                            itEvent->Param.NoteSynthParam.AbsValue,
                            itEvent->Param.NoteSynthParam.isFinal()
                        );
                        break;
                    case Event::synth_param_cutoff_lfo_freq:
                        if (itEvent->Param.NoteSynthParam.isFinal())
                            pLFO2->setScriptFrequencyFinal(itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        else
                            pLFO2->setScriptFrequencyFactor(itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        break;
                    case Event::synth_param_pitch_lfo_depth:
                        pLFO3->setScriptDepthFactor(
                            itEvent->Param.NoteSynthParam.AbsValue,
                            itEvent->Param.NoteSynthParam.isFinal()
                        );
                        break;
                    case Event::synth_param_pitch_lfo_freq:
                        pLFO3->setScriptFrequencyFactor(itEvent->Param.NoteSynthParam.AbsValue, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
                        break;

                    case Event::synth_param_attack:
                    case Event::synth_param_decay:
                    case Event::synth_param_sustain:
                    case Event::synth_param_release:
                    case Event::synth_param_cutoff_attack:
                    case Event::synth_param_cutoff_decay:
                    case Event::synth_param_cutoff_sustain:
                    case Event::synth_param_cutoff_release:
                        break; // noop
                }
            }
        }
    }

    /**
     * Process given list of events aimed at all voices in a key group.
     *
     * @param itEvent - iterator pointing to the next event to be processed
     * @param End     - youngest time stamp where processing should be stopped
     */
    void AbstractVoice::processGroupEvents(RTList<Event>::Iterator& itEvent, uint End) {
        for (; itEvent && itEvent->FragmentPos() <= End; ++itEvent) {
            ProcessGroupEvent(itEvent);
        }
    }

    /** @brief Update current portamento position.
     *
     * Will be called when portamento mode is enabled to get the final
     * portamento position of this active voice from where the next voice(s)
     * might continue to slide on.
     *
     * @param itNoteOffEvent - event which causes this voice to die soon
     */
    void AbstractVoice::UpdatePortamentoPos(Pool<Event>::Iterator& itNoteOffEvent) {
        if (pSignalUnitRack == NULL) {
            const float fFinalEG3Level = EG3.level(itNoteOffEvent->FragmentPos());
            pEngineChannel->PortamentoPos = (float) MIDIKey() + RTMath::FreqRatioToCents(fFinalEG3Level) * 0.01f;
        } else {
            // TODO: 
        }
    }

    /**
     *  Kill the voice in regular sense. Let the voice render audio until
     *  the kill event actually occured and then fade down the volume level
     *  very quickly and let the voice die finally. Unlike a normal release
     *  of a voice, a kill process cannot be cancalled and is therefore
     *  usually used for voice stealing and key group conflicts.
     *
     *  @param itKillEvent - event which caused the voice to be killed
     */
    void AbstractVoice::Kill(Pool<Event>::Iterator& itKillEvent) {
        #if CONFIG_DEVMODE
        if (!itKillEvent) dmsg(1,("AbstractVoice::Kill(): ERROR, !itKillEvent !!!\n"));
        if (itKillEvent && !itKillEvent.isValid()) dmsg(1,("AbstractVoice::Kill(): ERROR, itKillEvent invalid !!!\n"));
        #endif // CONFIG_DEVMODE

        if (itTriggerEvent && itKillEvent->FragmentPos() <= itTriggerEvent->FragmentPos()) return;
        this->itKillEvent = itKillEvent;
    }

    Voice::PitchInfo AbstractVoice::CalculatePitchInfo(int PitchBend) {
        PitchInfo pitch;
        double pitchbasecents = InstrInfo.FineTune + RgnInfo.FineTune + GetEngine()->ScaleTuning[MIDIKey() % 12];

        // GSt behaviour: maximum transpose up is 40 semitones. If
        // MIDI key is more than 40 semitones above unity note,
        // the transpose is not done.
        //
        // Update: Removed this GSt misbehavior. I don't think that any stock
        // gig sound requires it to resemble its original sound.
        // -- Christian, 2017-07-09
        if (!SmplInfo.Unpitched /* && (MIDIKey() - (int) RgnInfo.UnityNote) < 40*/)
            pitchbasecents += (MIDIKey() - (int) RgnInfo.UnityNote) * 100;

        pitch.PitchBase = RTMath::CentsToFreqRatioUnlimited(pitchbasecents) * (double(SmplInfo.SampleRate) / double(GetEngine()->SampleRate));
        pitch.PitchBendRange = 1.0 / 8192.0 * 100.0 * InstrInfo.PitchbendRange;
        pitch.PitchBend = RTMath::CentsToFreqRatio(PitchBend * pitch.PitchBendRange);

        return pitch;
    }
    
    void AbstractVoice::onScaleTuningChanged() {
        PitchInfo pitch = this->Pitch;
        double pitchbasecents = InstrInfo.FineTune + RgnInfo.FineTune + GetEngine()->ScaleTuning[MIDIKey() % 12];
        
        // GSt behaviour: maximum transpose up is 40 semitones. If
        // MIDI key is more than 40 semitones above unity note,
        // the transpose is not done.
        //
        // Update: Removed this GSt misbehavior. I don't think that any stock
        // gig sound requires it to resemble its original sound.
        // -- Christian, 2017-07-09
        if (!SmplInfo.Unpitched /* && (MIDIKey() - (int) RgnInfo.UnityNote) < 40*/)
            pitchbasecents += (MIDIKey() - (int) RgnInfo.UnityNote) * 100;
        
        pitch.PitchBase = RTMath::CentsToFreqRatioUnlimited(pitchbasecents) * (double(SmplInfo.SampleRate) / double(GetEngine()->SampleRate));
        this->Pitch = pitch;
    }

    double AbstractVoice::CalculateVolume(double velocityAttenuation) {
        // For 16 bit samples, we downscale by 32768 to convert from
        // int16 value range to DSP value range (which is
        // -1.0..1.0). For 24 bit, we downscale from int32.
        float volume = velocityAttenuation / (SmplInfo.BitDepth == 16 ? 32768.0f : 32768.0f * 65536.0f);

        volume *= GetSampleAttenuation() * pEngineChannel->GlobalVolume * GLOBAL_VOLUME;

        // the volume of release triggered samples depends on note length
        if (Type & Voice::type_release_trigger) {
            float noteLength = float(GetEngine()->FrameTime + Delay -
                GetNoteOnTime(MIDIKey()) ) / GetEngine()->SampleRate;

            volume *= GetReleaseTriggerAttenuation(noteLength);
        }

        return volume;
    }

    float AbstractVoice::GetReleaseTriggerAttenuation(float noteLength) {
        return 1 - RgnInfo.ReleaseTriggerDecay * noteLength;
    }

    void AbstractVoice::EnterReleaseStage() {
        if (pSignalUnitRack == NULL) {
            pEG1->update(EG::event_release, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
            pEG2->update(EG::event_release, GetEngine()->SampleRate / CONFIG_DEFAULT_SUBFRAGMENT_SIZE);
        } else {
            pSignalUnitRack->EnterReleaseStage();
        }
    }

    bool AbstractVoice::EG1Finished() {
        if (pSignalUnitRack == NULL) {
            return pEG1->getSegmentType() == EG::segment_end;
        } else {
            return !pSignalUnitRack->GetEndpointUnit()->Active();
        }
    }

} // namespace LinuxSampler