AppleWin/source/Speaker.cpp

/*
AppleWin : An Apple //e emulator for Windows

Copyright (C) 1994-1996, Michael O'Brien
Copyright (C) 1999-2001, Oliver Schmidt
Copyright (C) 2002-2005, Tom Charlesworth
Copyright (C) 2006-2007, Tom Charlesworth, Michael Pohoreski

AppleWin 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.

AppleWin 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 AppleWin; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
*/

/* Description: Speaker emulation
 *
 * Author: Various
 */

#include "StdAfx.h"

#include "Speaker.h"
#include "Core.h"
#include "CPU.h"
#include "Interface.h"
#include "Log.h"
#include "Memory.h"
#include "SoundCore.h"
#include "YamlHelper.h"
#include "Riff.h"

#include "Debugger/Debug.h"	// For DWORD extbench

// Notes:
//
// [OLD: 23.191 Apple CLKs == 44100Hz (CLK_6502/44100)]
// 23 Apple CLKS per PC sample (played back at 44.1KHz)
// 
//
// The speaker's wave output drives how much 6502 emulation is done in real-time, eg:
// If the speaker's wave buffer is running out of sample-data, then more 6502 cycles
// need to be executed to top-up the wave buffer.
// This is in contrast to the AY8910 voices, which can simply generate more data if
// their buffers are running low.
//

// NB. Setting g_nSPKR_NumChannels=1 still works (ie. mono).
// . Retain it for a while in case there are regressions with the new 2-channel code, then remove it.
static const unsigned short g_nSPKR_NumChannels = 2;
static const DWORD g_dwDSSpkrBufferSize = MAX_SAMPLES * sizeof(short) * g_nSPKR_NumChannels;

//-------------------------------------

static short*	g_pSpeakerBuffer = NULL;

// Globals (SOUND_WAVE)
const short		SPKR_DATA_INIT = (short)0x8000;

short		g_nSpeakerData	= SPKR_DATA_INIT;
static UINT		g_nBufferIdx	= 0;		// Sample index

static short*	g_pRemainderBuffer = NULL;
static UINT		g_nRemainderBufferSize;		// Setup in SpkrInitialize()
static UINT		g_nRemainderBufferIdx;		// Setup in SpkrInitialize()

// Application-wide globals:
SoundType_e		soundtype		= SOUND_WAVE;
double		    g_fClksPerSpkrSample;		// Setup in SetClksPerSpkrSample()

// Allow temporary quietening of speaker (8 bit DAC)
bool			g_bQuieterSpeaker = false;

// Globals
static unsigned __int64	g_nSpkrQuietCycleCount = 0;
static unsigned __int64 g_nSpkrLastCycle = 0;
static bool g_bSpkrToggleFlag = false;
static VOICE SpeakerVoice;
static bool g_bSpkrAvailable = false;

//-----------------------------------------------------------------------------

// Forward refs:
static ULONG   Spkr_SubmitWaveBuffer_FullSpeed(short* pSpeakerBuffer, ULONG nNumSamples);
static ULONG   Spkr_SubmitWaveBuffer(short* pSpeakerBuffer, ULONG nNumSamples);
static void    Spkr_SetActive(bool bActive);
static void    Spkr_DSUninit();

//-----------------------------------------------------------------------------

static bool g_bSpkrOutputToRiff = false;

void Spkr_OutputToRiff(void)
{
	g_bSpkrOutputToRiff = true;
}

UINT Spkr_GetNumChannels(void)
{
	return g_nSPKR_NumChannels;
}

//=============================================================================

static void DisplayBenchmarkResults ()
{
  DWORD totaltime = GetTickCount()-extbench;
  GetFrame().VideoRedrawScreen();
  std::string strText = StrFormat("This benchmark took %u.%02u seconds.",
								  (unsigned)(totaltime / 1000),
								  (unsigned)((totaltime / 10) % 100));
  GetFrame().FrameMessageBox(strText.c_str(),
							 "Benchmark Results",
							 MB_ICONINFORMATION | MB_SETFOREGROUND);
}

//=============================================================================
//
// DC filtering V2  (Riccardo Macri May 2015) (GH#275)
//
// To prevent loud clicks on Window's sound buffer underruns and constant DC
// being sent out to amplifiers (some soundcards are DC coupled) which is
// not good for them, an attenuator slowly drops the speaker output
// to 0 after the speaker (or 8 bit DAC) has been idle for a couple hundred
// milliseconds.
//
// The approach works as follows:
// - SpkrToggle() is called when the speaker state is flipped by accessing $C030
// - This brings audio up to date then calls ResetDCFilter()
// - ResetDCFilter() sets a counter to a high value
// - every audio sample is processed by DCFilter() as follows:
//   - if the counter is >= 32768, the speaker has been recently toggled
//     and the samples are unaffected
//   - if the counter is < 32768 but > 0, it is used to scale the
//     sample to reduce +ve or -ve speaker states towards zero
//   - In the two cases above, the counter is decremented 
//   - if the counter is zero, the speaker has been silent for a while
//     and the output is 0 regardless of the speaker state.
//
// - the initial "high value" is chosen so 10000/44100 = about a 
//   quarter of a second of speaker inactivity is needed before attenuation
//   begins.
//
//   NOTE: The attenuation is not ever reducing the level of audio, just
//         the DC offset at which the speaker has been left.  
//
//  This approach has zero impact on any speaker tones including PWM
//  due to the samples being unchanged for at least 0.25 seconds after
//  any speaker activity.
// 

static UINT g_uDCFilterState = 0;

inline void ResetDCFilter(void)
{
	// reset the attenuator with an additional 250ms of full gain
	// (10000 samples) before it starts attenuating
	g_uDCFilterState = 32768 + 10000;
}

inline short DCFilter(short sample_in)
{
	if (g_uDCFilterState == 0)		// no sound for a while, stay 0
		return 0;

	if (g_uDCFilterState >= 32768)	// full gain after recent sound
	{
		g_uDCFilterState--;
		return sample_in;
	}

	return (((int)sample_in) * (g_uDCFilterState--)) / 32768;	// scale & divide by 32768 (NB. Don't ">>15" as undefined behaviour)
}

//=============================================================================

static void SetClksPerSpkrSample()
{
//	// 23.191 clks for 44.1Khz (when 6502 CLK=1.0Mhz)
//	g_fClksPerSpkrSample = g_fCurrentCLK6502 / (double)SPKR_SAMPLE_RATE;

	// Use integer value: Better for MJ Mahon's RT.SYNTH.DSK (integer multiples of 1.023MHz Clk)
	// . 23 clks @ 1.023MHz
	g_fClksPerSpkrSample = (double) (UINT) (g_fCurrentCLK6502 / (double)SPKR_SAMPLE_RATE);
}

//=============================================================================

static void InitRemainderBuffer()
{
	delete [] g_pRemainderBuffer;

	SetClksPerSpkrSample();

	g_nRemainderBufferSize = (UINT) g_fClksPerSpkrSample;
	if ((double)g_nRemainderBufferSize < g_fClksPerSpkrSample)
		g_nRemainderBufferSize++;

	g_pRemainderBuffer = new short [g_nRemainderBufferSize];
	memset(g_pRemainderBuffer, 0, g_nRemainderBufferSize);

	g_nRemainderBufferIdx = 0;
}

//
// ----- ALL GLOBALLY ACCESSIBLE FUNCTIONS ARE BELOW THIS LINE -----
//

//=============================================================================

void SpkrDestroy ()
{
	Spkr_DSUninit();

	//

	if(soundtype == SOUND_WAVE)
	{
		delete [] g_pSpeakerBuffer;
		delete [] g_pRemainderBuffer;
		
		g_pSpeakerBuffer = NULL;
		g_pRemainderBuffer = NULL;
	}
}

//=============================================================================

void SpkrInitialize ()
{
	if(g_fh)
	{
		fprintf(g_fh, "Spkr Config: soundtype = %d ", (int) soundtype);
		switch(soundtype)
		{
			case SOUND_NONE:   fprintf(g_fh, "(NONE)\n"); break;
			case SOUND_WAVE:   fprintf(g_fh, "(WAVE)\n"); break;
			default:           fprintf(g_fh, "(UNDEFINED!)\n"); break;
		}
	}

	if(g_bDisableDirectSound)
	{
		SpeakerVoice.bMute = true;
		LogFileOutput("SpkrInitialize: g_bDisableDirectSound=1... SpeakerVoice.bMute=true\n");
	}
	else
	{
		g_bSpkrAvailable = Spkr_DSInit();
		LogFileOutput("Spkr_DSInit(), res=%d\n", g_bSpkrAvailable ? 1 : 0);
		if (!g_bSpkrAvailable)
		{
			GetFrame().FrameMessageBox(
				TEXT("The emulator is unable to initialize a waveform ")
				TEXT("output device.  Make sure you have a sound card ")
				TEXT("and a driver installed and that Windows is ")
				TEXT("correctly configured to use the driver.  Also ")
				TEXT("ensure that no other program is currently using ")
				TEXT("the device."),
				TEXT("Configuration"),
				MB_ICONEXCLAMATION | MB_SETFOREGROUND);
		}
	}

	//

	if (soundtype == SOUND_WAVE)
	{
		InitRemainderBuffer();

		g_pSpeakerBuffer = new short [SPKR_SAMPLE_RATE * g_nSPKR_NumChannels];	// Buffer can hold a max of 1 seconds worth of samples
	}
}

//=============================================================================

// NB. Called when /g_fCurrentCLK6502/ changes
void SpkrReinitialize ()
{
	if (soundtype == SOUND_WAVE)
	{
		InitRemainderBuffer();
	}
}

//=============================================================================

void SpkrReset()
{
	g_nBufferIdx = 0;
	g_nSpkrQuietCycleCount = 0;
	g_bSpkrToggleFlag = false;

	InitRemainderBuffer();
	Spkr_SubmitWaveBuffer(NULL, 0);
	Spkr_SetActive(false);
	Spkr_Unmute();
}

//=============================================================================

void SpkrSetEmulationType (SoundType_e newtype)
{
  SpkrDestroy();	// GH#295: Destroy for all types (even SOUND_NONE)

  soundtype = newtype;
  if (soundtype != SOUND_NONE)
    SpkrInitialize();
}

//=============================================================================

static void ReinitRemainderBuffer(UINT nCyclesRemaining)
{
	if(nCyclesRemaining == 0)
		return;

	for(g_nRemainderBufferIdx=0; g_nRemainderBufferIdx<nCyclesRemaining; g_nRemainderBufferIdx++)
		g_pRemainderBuffer[g_nRemainderBufferIdx] = g_nSpeakerData;

	_ASSERT(g_nRemainderBufferIdx < g_nRemainderBufferSize);
}

static void UpdateRemainderBuffer(ULONG* pnCycleDiff)
{
	if(g_nRemainderBufferIdx)
	{
		while((g_nRemainderBufferIdx < g_nRemainderBufferSize) && *pnCycleDiff)
		{
			g_pRemainderBuffer[g_nRemainderBufferIdx] = g_nSpeakerData;
			g_nRemainderBufferIdx++;
			(*pnCycleDiff)--;
		}

		if(g_nRemainderBufferIdx == g_nRemainderBufferSize)
		{
			g_nRemainderBufferIdx = 0;
			signed long nSampleMean = 0;
			for(UINT i=0; i<g_nRemainderBufferSize; i++)
				nSampleMean += (signed long) g_pRemainderBuffer[i];
			nSampleMean /= (signed long) g_nRemainderBufferSize;

			if (g_nBufferIdx < SPKR_SAMPLE_RATE - 1)
			{
				if (g_nSPKR_NumChannels == 1)
				{
					g_pSpeakerBuffer[g_nBufferIdx] = DCFilter((short)nSampleMean);
				}
				else
				{
					const short sample = DCFilter((short)nSampleMean);
					g_pSpeakerBuffer[g_nBufferIdx * 2 + 0] = sample;
					g_pSpeakerBuffer[g_nBufferIdx * 2 + 1] = sample;
				}
				g_nBufferIdx++;
			}
		}
	}
}

static void UpdateSpkr()
{
  if(!g_bFullSpeed || SoundCore_GetTimerState())
  {
	  ULONG nCycleDiff = (ULONG) (g_nCumulativeCycles - g_nSpkrLastCycle);

	  UpdateRemainderBuffer(&nCycleDiff);

	  ULONG nNumSamples = (ULONG) ((double)nCycleDiff / g_fClksPerSpkrSample);

	  ULONG nCyclesRemaining = (ULONG) ((double)nCycleDiff - (double)nNumSamples * g_fClksPerSpkrSample);

	  while ((nNumSamples--) && (g_nBufferIdx < SPKR_SAMPLE_RATE - 1))
	  {
		  if (g_nSPKR_NumChannels == 1)
		  {
			  g_pSpeakerBuffer[g_nBufferIdx] = DCFilter(g_nSpeakerData);
		  }
		  else
		  {
			  const short sample = DCFilter(g_nSpeakerData);
			  g_pSpeakerBuffer[g_nBufferIdx * 2 + 0] = sample;
			  g_pSpeakerBuffer[g_nBufferIdx * 2 + 1] = sample;
		  }
		  g_nBufferIdx++;
	  }

	  ReinitRemainderBuffer(nCyclesRemaining);	// Partially fill 1Mhz sample buffer
  }

  g_nSpkrLastCycle = g_nCumulativeCycles;
}

//=============================================================================

// Called by emulation code when Speaker I/O reg is accessed
//

BYTE __stdcall SpkrToggle (WORD, WORD, BYTE, BYTE, ULONG nExecutedCycles)
{
  g_bSpkrToggleFlag = true;

  if(!g_bFullSpeed)
	Spkr_SetActive(true);

  //


  if (extbench)
  {
    DisplayBenchmarkResults();
    extbench = 0;
  }

  if (soundtype == SOUND_WAVE)
  {
	  CpuCalcCycles(nExecutedCycles);

	  UpdateSpkr();

      short speakerDriveLevel = SPKR_DATA_INIT;
      if (g_bQuieterSpeaker)	// quieten the speaker if 8 bit DAC in use
        speakerDriveLevel /= 4;	// NB. Don't shift -ve number right: undefined behaviour (MSDN says: implementation-dependent)

      // When full-speed: Don't ResetDCFilter(), otherwise get occasional clicks when speaker toggled
      if (!g_bFullSpeed)
        ResetDCFilter();

      if (g_nSpeakerData == speakerDriveLevel)
        g_nSpeakerData = ~speakerDriveLevel;
      else
        g_nSpeakerData = speakerDriveLevel;
  }

  return MemReadFloatingBus(nExecutedCycles);
}

//=============================================================================

// Called by ContinueExecution()
void SpkrUpdate (DWORD totalcycles)
{
#ifdef LOG_PERF_TIMINGS
	extern UINT64 g_timeSpeaker;
	PerfMarker perfMarker(g_timeSpeaker);
#endif

  if(!g_bSpkrToggleFlag)
  {
	  if(!g_nSpkrQuietCycleCount)
	  {
		  g_nSpkrQuietCycleCount = g_nCumulativeCycles;
	  }
	  else if(g_nCumulativeCycles - g_nSpkrQuietCycleCount > (unsigned __int64)g_fCurrentCLK6502/5)
	  {
		  // After 0.2 sec of Apple time, deactivate spkr voice
		  // . This allows emulator to auto-switch to full-speed g_nAppMode for fast disk access
		  Spkr_SetActive(false);
	  }
  }
  else
  {
      g_nSpkrQuietCycleCount = 0;
      g_bSpkrToggleFlag = false;
  }

  //

  if (soundtype == SOUND_WAVE)
  {
	  UpdateSpkr();
	  ULONG nSamplesUsed;

	  if(g_bFullSpeed)
		  nSamplesUsed = Spkr_SubmitWaveBuffer_FullSpeed(g_pSpeakerBuffer, g_nBufferIdx);
	  else
		  nSamplesUsed = Spkr_SubmitWaveBuffer(g_pSpeakerBuffer, g_nBufferIdx);

	  _ASSERT(nSamplesUsed <= g_nBufferIdx);
	  memmove(g_pSpeakerBuffer, &g_pSpeakerBuffer[nSamplesUsed], (g_nBufferIdx - nSamplesUsed) * sizeof(short) * g_nSPKR_NumChannels);
	  g_nBufferIdx -= nSamplesUsed;
  }
}

// Called from SoundCore_TimerFunc() for FADE_OUT
void SpkrUpdate_Timer()
{
	if (soundtype == SOUND_WAVE)
	{
		UpdateSpkr();
		ULONG nSamplesUsed;

		nSamplesUsed = Spkr_SubmitWaveBuffer_FullSpeed(g_pSpeakerBuffer, g_nBufferIdx);

		_ASSERT(nSamplesUsed <=	g_nBufferIdx);
		memmove(g_pSpeakerBuffer, &g_pSpeakerBuffer[nSamplesUsed], (g_nBufferIdx - nSamplesUsed) * sizeof(short) * g_nSPKR_NumChannels);
		g_nBufferIdx -=	nSamplesUsed;
	}
}

//=============================================================================

static DWORD dwByteOffset = (DWORD)-1;
static int nNumSamplesError = 0;
static int nDbgSpkrCnt = 0;

// FullSpeed g_nAppMode, 2 cases:
// i) Short burst of full-speed, so PlayCursor doesn't complete sound from previous fixed-speed session.
// ii) Long burst of full-speed, so PlayCursor completes sound from previous fixed-speed session.

// Try to:
// 1) Output remaining samples from SpeakerBuffer (from previous fixed-speed session)
// 2) Output pad samples to keep the VoiceBuffer topped-up

// If nNumSamples>0 then these are from previous fixed-speed session.
// - Output these before outputting zero-pad samples.

static ULONG Spkr_SubmitWaveBuffer_FullSpeed(short* pSpeakerBuffer, ULONG nNumSamples)
{
	nDbgSpkrCnt++;

	if(!SpeakerVoice.bActive)
		return nNumSamples;

	// pSpeakerBuffer can't be NULL, as reset clears g_bFullSpeed, so 1st SpkrUpdate() never calls here
	_ASSERT(pSpeakerBuffer != NULL);

	//

	DWORD dwDSLockedBufferSize0, dwDSLockedBufferSize1;
	SHORT *pDSLockedBuffer0, *pDSLockedBuffer1;
	//bool bBufferError = false;

	DWORD dwCurrentPlayCursor, dwCurrentWriteCursor;
	HRESULT hr = SpeakerVoice.lpDSBvoice->GetCurrentPosition(&dwCurrentPlayCursor, &dwCurrentWriteCursor);
	if(FAILED(hr))
		return nNumSamples;

	if(dwByteOffset == (DWORD)-1)
	{
		// First time in this func (probably after re-init (Spkr_SubmitWaveBuffer()))

		dwByteOffset = dwCurrentPlayCursor + (g_dwDSSpkrBufferSize/8)*3;	// Ideal: 0.375 is between 0.25 & 0.50 full
		dwByteOffset %= g_dwDSSpkrBufferSize;
		//LogOutput("[Submit_FS] PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X [REINIT]\n", dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamples);
	}
	else
	{
		// Check that our offset isn't between Play & Write positions

		if(dwCurrentWriteCursor > dwCurrentPlayCursor)
		{
			// |-----PxxxxxW-----|
			if((dwByteOffset > dwCurrentPlayCursor) && (dwByteOffset < dwCurrentWriteCursor))
			{
				//LogOutput("[Submit_FS] PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X xxx\n", dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamples);
				dwByteOffset = dwCurrentWriteCursor;
			}
		}
		else
		{
			// |xxW----------Pxxx|
			if((dwByteOffset > dwCurrentPlayCursor) || (dwByteOffset < dwCurrentWriteCursor))
			{
				//LogOutput("[Submit_FS] PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X XXX\n", dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamples);
				dwByteOffset = dwCurrentWriteCursor;
			}
		}
	}

	// Calc bytes remaining to be played
	int nBytesRemaining = dwByteOffset - dwCurrentPlayCursor;
	if(nBytesRemaining < 0)
		nBytesRemaining += g_dwDSSpkrBufferSize;
	if((nBytesRemaining == 0) && (dwCurrentPlayCursor != dwCurrentWriteCursor))
		nBytesRemaining = g_dwDSSpkrBufferSize;		// Case when complete buffer is to be played

	//

	UINT nNumPadSamples = 0;

	if(nBytesRemaining < g_dwDSSpkrBufferSize / 4)
	{
		// < 1/4 of play-buffer remaining (need *more* data)
		nNumPadSamples = ((g_dwDSSpkrBufferSize / 4) - nBytesRemaining) / (sizeof(short) * g_nSPKR_NumChannels);

		if(nNumPadSamples > nNumSamples)
			nNumPadSamples -= nNumSamples;
		else
			nNumPadSamples = 0;

		// NB. If nNumPadSamples>0 then all nNumSamples are to be used
	}

	//

	UINT nBytesFree = g_dwDSSpkrBufferSize - nBytesRemaining;	// Calc free buffer space
	ULONG nNumSamplesToUse = nNumSamples + nNumPadSamples;

	if (nNumSamplesToUse * sizeof(short) * g_nSPKR_NumChannels > nBytesFree)
		nNumSamplesToUse = nBytesFree / (sizeof(short) * g_nSPKR_NumChannels);

	//

	if(nNumSamplesToUse >= 128)	// Limit the buffer unlock/locking to a minimum
	{
		hr = DSGetLock(SpeakerVoice.lpDSBvoice,
			dwByteOffset, (DWORD)nNumSamplesToUse * sizeof(short) * g_nSPKR_NumChannels,
			&pDSLockedBuffer0, &dwDSLockedBufferSize0,
			&pDSLockedBuffer1, &dwDSLockedBufferSize1);
		if (FAILED(hr))
			return nNumSamples;

		//

		DWORD dwBufferSize0 = 0;
		DWORD dwBufferSize1 = 0;

		if(nNumSamples)
		{
			//LogOutput("[Submit_FS] C=%08X, PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X ***\n", nDbgSpkrCnt, dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamples);

			if (nNumSamples * sizeof(short) * g_nSPKR_NumChannels <= dwDSLockedBufferSize0)
			{
				dwBufferSize0 = nNumSamples * sizeof(short) * g_nSPKR_NumChannels;
				dwBufferSize1 = 0;
			}
			else
			{
				dwBufferSize0 = dwDSLockedBufferSize0;
				dwBufferSize1 = nNumSamples * sizeof(short) * g_nSPKR_NumChannels - dwDSLockedBufferSize0;

				if(dwBufferSize1 > dwDSLockedBufferSize1)
					dwBufferSize1 = dwDSLockedBufferSize1;
			}
			
			memcpy(pDSLockedBuffer0, &pSpeakerBuffer[0], dwBufferSize0);
			if (g_bSpkrOutputToRiff)
				RiffPutSamples(pDSLockedBuffer0, dwBufferSize0 / (sizeof(short) * g_nSPKR_NumChannels));
			nNumSamples = dwBufferSize0 / (sizeof(short) * g_nSPKR_NumChannels);

			if(pDSLockedBuffer1 && dwBufferSize1)
			{
				memcpy(pDSLockedBuffer1, &pSpeakerBuffer[dwDSLockedBufferSize0/sizeof(short)], dwBufferSize1);
				if (g_bSpkrOutputToRiff)
					RiffPutSamples(pDSLockedBuffer1, dwBufferSize1 / (sizeof(short) * g_nSPKR_NumChannels));
				nNumSamples += dwBufferSize1 / (sizeof(short) * g_nSPKR_NumChannels);
			}
		}

		if(nNumPadSamples)
		{
			//LogOutput("[Submit_FS] C=%08X, PC=%08X, WC=%08X, Diff=%08X, Off=%08X, PS=%08X, Data=%04X\n", nDbgSpkrCnt, dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumPadSamples, g_nSpeakerData);

			dwBufferSize0 = dwDSLockedBufferSize0 - dwBufferSize0;
			dwBufferSize1 = dwDSLockedBufferSize1 - dwBufferSize1;

			if(dwBufferSize0)
			{
				const UINT numSamples = dwBufferSize0 / (sizeof(short) * g_nSPKR_NumChannels);
				if (g_nSPKR_NumChannels == 1)
				{
					std::fill_n(pDSLockedBuffer0, numSamples, DCFilter(g_nSpeakerData));
				}
				else
				{
					for (UINT i = 0; i < numSamples; i++)
					{
						const short sample = DCFilter(g_nSpeakerData);
						pDSLockedBuffer0[i * 2 + 0] = sample;
						pDSLockedBuffer0[i * 2 + 1] = sample;
					}
				}

				if (g_bSpkrOutputToRiff)
					RiffPutSamples(pDSLockedBuffer0, numSamples);
			}

			if(pDSLockedBuffer1)
			{
				const UINT numSamples = dwBufferSize0 / (sizeof(short) * g_nSPKR_NumChannels);
				if (g_nSPKR_NumChannels == 1)
				{
					std::fill_n(pDSLockedBuffer1, numSamples, DCFilter(g_nSpeakerData));
				}
				else
				{
					for (UINT i = 0; i < numSamples; i++)
					{
						const short sample = DCFilter(g_nSpeakerData);
						pDSLockedBuffer1[i * 2 + 0] = sample;
						pDSLockedBuffer1[i * 2 + 1] = sample;
					}
				}

				if (g_bSpkrOutputToRiff)
					RiffPutSamples(pDSLockedBuffer1, numSamples);
			}
		}

		// Commit sound buffer
		hr = SpeakerVoice.lpDSBvoice->Unlock((void*)pDSLockedBuffer0, dwDSLockedBufferSize0,
											(void*)pDSLockedBuffer1, dwDSLockedBufferSize1);
		if(FAILED(hr))
			return nNumSamples;

		dwByteOffset = (dwByteOffset + (DWORD)nNumSamplesToUse*sizeof(short)*g_nSPKR_NumChannels) % g_dwDSSpkrBufferSize;
	}

	return nNumSamples;
}

//-----------------------------------------------------------------------------

static ULONG Spkr_SubmitWaveBuffer(short* pSpeakerBuffer, ULONG nNumSamples)
{
	nDbgSpkrCnt++;

	if(!SpeakerVoice.bActive)
		return nNumSamples;

	if(pSpeakerBuffer == NULL)
	{
		// Re-init from SpkrReset()
		dwByteOffset = (DWORD)-1;
		nNumSamplesError = 0;

		// Don't call DSZeroVoiceBuffer() - get noise with "VIA AC'97 Enhanced Audio Controller"
		// . I guess SpeakerVoice.Stop() isn't really working and the new zero buffer causes noise corruption when submitted.
		bool res = DSZeroVoiceWritableBuffer(&SpeakerVoice, g_dwDSSpkrBufferSize);
		LogFileOutput("Spkr_SubmitWaveBuffer: DSZeroVoiceWritableBuffer, res=%d\n", res ? 1 : 0);

		return 0;
	}

	//

	DWORD dwDSLockedBufferSize0, dwDSLockedBufferSize1;
	SHORT *pDSLockedBuffer0, *pDSLockedBuffer1;
	bool bBufferError = false;

	DWORD dwCurrentPlayCursor, dwCurrentWriteCursor;
	HRESULT hr = SpeakerVoice.lpDSBvoice->GetCurrentPosition(&dwCurrentPlayCursor, &dwCurrentWriteCursor);
	if (FAILED(hr))
	{
		LogFileOutput("Spkr_SubmitWaveBuffer: GetCurrentPosition failed (%08X)\n", hr);
		return nNumSamples;
	}

	if(dwByteOffset == (DWORD)-1)
	{
		// First time in this func (probably after re-init (above))

		dwByteOffset = dwCurrentPlayCursor + (g_dwDSSpkrBufferSize/8)*3;	// Ideal: 0.375 is between 0.25 & 0.50 full
		dwByteOffset %= g_dwDSSpkrBufferSize;
		//LogOutput("[Submit]   PC=%08X, WC=%08X, Diff=%08X, Off=%08X [REINIT]\n", dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset);
	}
	else
	{
		// Check that our offset isn't between Play & Write positions

		if(dwCurrentWriteCursor > dwCurrentPlayCursor)
		{
			// |-----PxxxxxW-----|
			if((dwByteOffset > dwCurrentPlayCursor) && (dwByteOffset < dwCurrentWriteCursor))
			{
				double fTicksSecs = (double)GetTickCount() / 1000.0;
				//LogOutput("%010.3f: [Submit]    PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X xxx\n", fTicksSecs, dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamples);
				//LogFileOutput("%010.3f: [Submit]    PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X xxx\n", fTicksSecs, dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamples);

				dwByteOffset = dwCurrentWriteCursor;
				nNumSamplesError = 0;
				bBufferError = true;
			}
		}
		else
		{
			// |xxW----------Pxxx|
			if((dwByteOffset > dwCurrentPlayCursor) || (dwByteOffset < dwCurrentWriteCursor))
			{
				double fTicksSecs = (double)GetTickCount() / 1000.0;
				//LogOutput("%010.3f: [Submit]    PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X XXX\n", fTicksSecs, dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamples);
				//LogFileOutput("%010.3f: [Submit]    PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X XXX\n", fTicksSecs, dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamples);

				dwByteOffset = dwCurrentWriteCursor;
				nNumSamplesError = 0;
				bBufferError = true;
			}
		}
	}

	// Calc bytes remaining to be played
	int nBytesRemaining = dwByteOffset - dwCurrentPlayCursor;
	if(nBytesRemaining < 0)
		nBytesRemaining += g_dwDSSpkrBufferSize;
	if((nBytesRemaining == 0) && (dwCurrentPlayCursor != dwCurrentWriteCursor))
		nBytesRemaining = g_dwDSSpkrBufferSize;		// Case when complete buffer is to be played

	// Calc correction factor so that play-buffer doesn't under/overflow
	const int nErrorInc = SoundCore_GetErrorInc();
	if(nBytesRemaining < g_dwDSSpkrBufferSize / 4)
		nNumSamplesError += nErrorInc;				// < 1/4 of play-buffer remaining (need *more* data)
	else if(nBytesRemaining > g_dwDSSpkrBufferSize / 2)
		nNumSamplesError -= nErrorInc;				// > 1/2 of play-buffer remaining (need *less* data)
	else
		nNumSamplesError = 0;						// Acceptable amount of data in buffer

	const int nErrorMax = SoundCore_GetErrorMax();				// Cap feedback to +/-nMaxError units
	if(nNumSamplesError < -nErrorMax) nNumSamplesError = -nErrorMax;
	if(nNumSamplesError >  nErrorMax) nNumSamplesError =  nErrorMax;
	g_nCpuCyclesFeedback = (int) ((double)nNumSamplesError * g_fClksPerSpkrSample);

	//

	UINT nBytesFree = g_dwDSSpkrBufferSize - nBytesRemaining;	// Calc free buffer space
	ULONG nNumSamplesToUse = nNumSamples;

	if(nNumSamplesToUse * sizeof(short) > nBytesFree)
		nNumSamplesToUse = nBytesFree / sizeof(short);

	if(bBufferError)
		pSpeakerBuffer = &pSpeakerBuffer[nNumSamples - nNumSamplesToUse];

	//

	if(nNumSamplesToUse)
	{
		//LogOutput("[Submit]    C=%08X, PC=%08X, WC=%08X, Diff=%08X, Off=%08X, NS=%08X +++\n", nDbgSpkrCnt, dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor, dwByteOffset, nNumSamplesToUse);

		hr = DSGetLock(SpeakerVoice.lpDSBvoice,
			dwByteOffset, (DWORD)nNumSamplesToUse * sizeof(short) * g_nSPKR_NumChannels,
			&pDSLockedBuffer0, &dwDSLockedBufferSize0,
			&pDSLockedBuffer1, &dwDSLockedBufferSize1);
		if (FAILED(hr))
		{
			LogFileOutput("Spkr_SubmitWaveBuffer: DSGetLock failed\n");
			return nNumSamples;
		}

		memcpy(pDSLockedBuffer0, &pSpeakerBuffer[0], dwDSLockedBufferSize0);
		if (g_bSpkrOutputToRiff)
			RiffPutSamples(pDSLockedBuffer0, dwDSLockedBufferSize0 / (sizeof(short) * g_nSPKR_NumChannels));

		if(pDSLockedBuffer1)
		{
			memcpy(pDSLockedBuffer1, &pSpeakerBuffer[dwDSLockedBufferSize0/sizeof(short)], dwDSLockedBufferSize1);
			if (g_bSpkrOutputToRiff)
				RiffPutSamples(pDSLockedBuffer1, dwDSLockedBufferSize1 / (sizeof(short) * g_nSPKR_NumChannels));
		}

		// Commit sound buffer
		hr = SpeakerVoice.lpDSBvoice->Unlock((void*)pDSLockedBuffer0, dwDSLockedBufferSize0,
											(void*)pDSLockedBuffer1, dwDSLockedBufferSize1);
		if (FAILED(hr))
		{
			LogFileOutput("Spkr_SubmitWaveBuffer: Unlock failed (%08X)\n", hr);
			return nNumSamples;
		}

		dwByteOffset = (dwByteOffset + (DWORD)nNumSamplesToUse*sizeof(short)*g_nSPKR_NumChannels) % g_dwDSSpkrBufferSize;
	}

	return bBufferError ? nNumSamples : nNumSamplesToUse;
}

//-----------------------------------------------------------------------------

// NB. Not currently used
void Spkr_Mute()
{
	if(SpeakerVoice.bActive && !SpeakerVoice.bMute)
	{
		HRESULT hr = SpeakerVoice.lpDSBvoice->SetVolume(DSBVOLUME_MIN);
		LogFileOutput("Spkr_Mute: SetVolume(%d) res = %08X\n", DSBVOLUME_MIN, hr);
		SpeakerVoice.bMute = true;
	}
}

// NB. Only called by SpkrReset()
void Spkr_Unmute()
{
	if(SpeakerVoice.bActive && SpeakerVoice.bMute)
	{
		HRESULT hr = SpeakerVoice.lpDSBvoice->SetVolume(SpeakerVoice.nVolume);
		LogFileOutput("Spkr_Unmute: SetVolume(%d) res = %08X\n", SpeakerVoice.nVolume, hr);
		SpeakerVoice.bMute = false;
	}
}

//-----------------------------------------------------------------------------

static bool g_bSpkrRecentlyActive = false;

static void Spkr_SetActive(bool bActive)
{
	if(!SpeakerVoice.bActive)
		return;

	if(bActive)
	{
		// Called by SpkrToggle() or SpkrReset()
		g_bSpkrRecentlyActive = true;
		SpeakerVoice.bRecentlyActive = true;
	}
	else
	{
		// Called by SpkrUpdate() after 0.2s of speaker inactivity
		g_bSpkrRecentlyActive = false;
		SpeakerVoice.bRecentlyActive = false;
		g_bQuieterSpeaker = 0;	// undo any muting (for 8 bit DAC)
	}
}

bool Spkr_IsActive()
{
	return g_bSpkrRecentlyActive;
}

//-----------------------------------------------------------------------------

DWORD SpkrGetVolume()
{
	return SpeakerVoice.dwUserVolume;
}

void SpkrSetVolume(DWORD dwVolume, DWORD dwVolumeMax)
{
	SpeakerVoice.dwUserVolume = dwVolume;

	SpeakerVoice.nVolume = NewVolume(dwVolume, dwVolumeMax);

	if (SpeakerVoice.bActive && !SpeakerVoice.bMute)
	{
		HRESULT hr = SpeakerVoice.lpDSBvoice->SetVolume(SpeakerVoice.nVolume);
		LogFileOutput("SpkrSetVolume: SetVolume(%d) res = %08X\n", SpeakerVoice.nVolume, hr);
	}
}

//=============================================================================

bool Spkr_DSInit()
{
	//
	// Create single Apple speaker voice
	//

	if (!g_bDSAvailable)
	{
		LogFileOutput("Spkr_DSInit: g_bDSAvailable=0\n");
		return false;
	}

	SpeakerVoice.bIsSpeaker = true;

	HRESULT hr = DSGetSoundBuffer(&SpeakerVoice, DSBCAPS_CTRLVOLUME, g_dwDSSpkrBufferSize, SPKR_SAMPLE_RATE, g_nSPKR_NumChannels, "Spkr");
	if (FAILED(hr))
	{
		LogFileOutput("Spkr_DSInit: DSGetSoundBuffer failed (%08X)\n", hr);
		return false;
	}

	if (!DSZeroVoiceBuffer(&SpeakerVoice, g_dwDSSpkrBufferSize))
	{
		LogFileOutput("Spkr_DSInit: DSZeroVoiceBuffer failed\n");
		return false;
	}

	SpeakerVoice.bActive = true;

	// Volume might've been setup from value in Registry
	if(!SpeakerVoice.nVolume)
		SpeakerVoice.nVolume = DSBVOLUME_MAX;

	hr = SpeakerVoice.lpDSBvoice->SetVolume(SpeakerVoice.nVolume);
	LogFileOutput("Spkr_DSInit: SetVolume(%d) res = %08X\n", SpeakerVoice.nVolume, hr);

	//

	DWORD dwCurrentPlayCursor, dwCurrentWriteCursor;
	hr = SpeakerVoice.lpDSBvoice->GetCurrentPosition(&dwCurrentPlayCursor, &dwCurrentWriteCursor);
	if (FAILED(hr))
		LogFileOutput("Spkr_DSInit: GetCurrentPosition failed (%08X)\n", hr);
	if (SUCCEEDED(hr) && (dwCurrentPlayCursor == dwCurrentWriteCursor))
	{
		// KLUDGE: For my WinXP PC with "VIA AC'97 Enhanced Audio Controller"
		// . Not required for my Win98SE/WinXP PC with PCI "Soundblaster Live!"
		Sleep(200);

		hr = SpeakerVoice.lpDSBvoice->GetCurrentPosition(&dwCurrentPlayCursor, &dwCurrentWriteCursor);
		LogFileOutput("Spkr_DSInit: GetCurrentPosition kludge (%08X)\n", hr);
		LogOutput("[DSInit] PC=%08X, WC=%08X, Diff=%08X\n", dwCurrentPlayCursor, dwCurrentWriteCursor, dwCurrentWriteCursor-dwCurrentPlayCursor);
	}

	return true;
}

static void Spkr_DSUninit()
{
	if(SpeakerVoice.lpDSBvoice && SpeakerVoice.bActive)
		DSVoiceStop(&SpeakerVoice);

	DSReleaseSoundBuffer(&SpeakerVoice);
}

//=============================================================================

#define SS_YAML_KEY_LASTCYCLE "Last Cycle"

static const std::string& SpkrGetSnapshotStructName(void)
{
	static const std::string name("Speaker");
	return name;
}

void SpkrSaveSnapshot(YamlSaveHelper& yamlSaveHelper)
{
	YamlSaveHelper::Label state(yamlSaveHelper, "%s:\n", SpkrGetSnapshotStructName().c_str());
	yamlSaveHelper.SaveHexUint64(SS_YAML_KEY_LASTCYCLE, g_nSpkrLastCycle);
}

void SpkrLoadSnapshot(YamlLoadHelper& yamlLoadHelper)
{
	if (!yamlLoadHelper.GetSubMap(SpkrGetSnapshotStructName()))
		return;

	g_nSpkrLastCycle = yamlLoadHelper.LoadUint64(SS_YAML_KEY_LASTCYCLE);

	yamlLoadHelper.PopMap();
}