path: root/audio.c

/*
 * rpihddevice - VDR HD output device for Raspberry Pi
 * Copyright (C) 2014, 2015, 2016 Thomas Reufer
 *
 * 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "audio.h"
#include "setup.h"
#include "omx.h"

#include <vdr/tools.h>
#include <vdr/remux.h>

extern "C" {
#include <libavcodec/avcodec.h>
#include <libavutil/log.h>
#include <libavutil/opt.h>

#ifdef ENABLE_AAC_LATM
#warning "experimental AAC-LATM frame parser enabled, only 2ch/48kHz supported!"
#endif

// ffmpeg's resampling
#ifdef HAVE_LIBSWRESAMPLE
#  include <libswresample/swresample.h>
#  define DO_RESAMPLE
#endif

// libav's resampling
#ifdef HAVE_LIBAVRESAMPLE
#  include <libavresample/avresample.h>
#  include <libavutil/samplefmt.h>
#  define DO_RESAMPLE
#  define SwrContext AVAudioResampleContext
#  define swr_alloc  avresample_alloc_context
#  define swr_init   avresample_open
#  define swr_free   avresample_free
#  define swr_convert(ctx, dst, out_cnt, src, in_cnt) \
		avresample_convert(ctx, dst, 0, out_cnt, (uint8_t**)src, 0, in_cnt)
#endif

// legacy libavcodec
#if LIBAVCODEC_VERSION_MAJOR < 55
#  define av_frame_alloc       avcodec_alloc_frame
#  define av_frame_free        avcodec_free_frame
#  define av_frame_unref       avcodec_get_frame_defaults
#  define AV_CODEC_ID_MP3      CODEC_ID_MP3
#  define AV_CODEC_ID_AC3      CODEC_ID_AC3
#  define AV_CODEC_ID_EAC3     CODEC_ID_EAC3
#  define AV_CODEC_ID_AAC      CODEC_ID_AAC
#  define AV_CODEC_ID_AAC_LATM CODEC_ID_AAC_LATM
#  define AV_CODEC_ID_DTS      CODEC_ID_DTS
#endif

#if LIBAVCODEC_VERSION_MAJOR < 54
#  define avcodec_free_frame av_free
#endif

// prevent depreciated warnings for >ffmpeg-1.2.x and >libav-9.x
#if LIBAVCODEC_VERSION_MAJOR > 54
#  undef FF_API_REQUEST_CHANNELS
#endif
}

#include <queue>
#include <string.h>

#define AVPKT_BUFFER_SIZE (KILOBYTE(256))

class cRpiAudioDecoder::cParser
{

public:

	cParser() :
		m_mutex(new cMutex()),
		m_codec(cAudioCodec::eInvalid),
		m_channels(0),
		m_samplingRate(0),
		m_size(0),
		m_parsed(true)
	{
	}

	~cParser()
	{
		delete m_mutex;
	}

	AVPacket* Packet(void)
	{
		return &m_packet;
	}

	cAudioCodec::eCodec GetCodec(void)
	{
		if (!m_parsed)
			Parse();
		return m_codec;
	}

	unsigned int GetChannels(void)
	{
		if (!m_parsed)
			Parse();
		return m_channels;
	}

	unsigned int GetSamplingRate(void)
	{
		if (!m_parsed)
			Parse();
		return m_samplingRate;
	}

	unsigned int GetFrameSize(void)
	{
		if (!m_parsed)
			Parse();
		return m_packet.size;
	}

	int64_t GetPts(void)
	{
		int64_t pts = OMX_INVALID_PTS;
		m_mutex->Lock();

		if (!m_ptsQueue.empty())
			pts = m_ptsQueue.front()->pts;

		m_mutex->Unlock();
		return pts;
	}

	unsigned int GetFreeSpace(void)
	{
		return AVPKT_BUFFER_SIZE - m_size - FF_INPUT_BUFFER_PADDING_SIZE;
	}

	bool Empty(void)
	{
		if (!m_parsed)
			Parse();
		return m_packet.size == 0;
	}

	int Init(void)
	{
		if (!av_new_packet(&m_packet, AVPKT_BUFFER_SIZE))
		{
			Reset();
			return 0;
		}
		return -1;
	}

	int DeInit(void)
	{
		av_free_packet(&m_packet);
		return 0;
	}

	void Reset(void)
	{
		m_mutex->Lock();
		m_codec = cAudioCodec::eInvalid;
		m_channels = 0;
		m_samplingRate = 0;
		m_packet.size = 0;
		m_size = 0;
		m_parsed = true; // parser is empty, no need for parsing
		memset(m_packet.data, 0, FF_INPUT_BUFFER_PADDING_SIZE);

		while (!m_ptsQueue.empty())
		{
			delete m_ptsQueue.front();
			m_ptsQueue.pop();
		}
		m_mutex->Unlock();
	}

	bool Append(const unsigned char *data, int64_t pts, unsigned int length)
	{
		m_mutex->Lock();
		bool ret = true;

		if (m_size + length + FF_INPUT_BUFFER_PADDING_SIZE > AVPKT_BUFFER_SIZE)
			ret = false;
		else
		{
			memcpy(m_packet.data + m_size, data, length);
			m_size += length;
			memset(m_packet.data + m_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);

			Pts* entry = new Pts(pts, length);
			m_ptsQueue.push(entry);

			m_parsed = false;
		}
		m_mutex->Unlock();
		return ret;
	}

	void Shrink(unsigned int length, bool retainPts = false)
	{
		m_mutex->Lock();

		if (length < m_size)
		{
			memmove(m_packet.data, m_packet.data + length, m_size - length);
			m_size -= length;
			memset(m_packet.data + m_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);

			while (!m_ptsQueue.empty() && length)
			{
				if (m_ptsQueue.front()->length <= length)
				{
					length -= m_ptsQueue.front()->length;
					delete m_ptsQueue.front();
					m_ptsQueue.pop();
				}
				else
				{
					// clear current PTS since it's not valid anymore after
					// shrinking the packet
					if (!retainPts)
						m_ptsQueue.front()->pts = OMX_INVALID_PTS;

					m_ptsQueue.front()->length -= length;
					length = 0;
				}
			}

			m_parsed = false;
		}
		else
			Reset();

		m_mutex->Unlock();
	}
	
private:

	cParser(const cParser&);
	cParser& operator= (const cParser&);

	// Check format of first audio packet in buffer. If format has been
	// guessed, but packet is not yet complete, codec is set with a length
	// of 0. Once the buffer contains either the exact amount of expected
	// data or another valid packet start after the first frame, packet
	// size is set to the first frame length.
	// Valid packets are always moved to the buffer start, if no valid
	// audio frame has been found, packet gets cleared.

	void Parse()
	{
		m_mutex->Lock();

		cAudioCodec::eCodec codec = cAudioCodec::eInvalid;
		unsigned int channels = 0;
		unsigned int offset = 0;
		unsigned int frameSize = 0;
		unsigned int samplingRate = 0;

		while (m_size - offset >= 4)
		{
			// 0xFFE...      MPEG audio
			// 0x0B77...     (E)AC-3 audio
			// 0xFFF...      AAC audio
			// 0x7FFE8001... DTS audio
			// PCM audio can't be found

			const uint8_t *p = m_packet.data + offset;
			unsigned int n = m_size - offset;

			switch (FastCheck(p))
			{
			case cAudioCodec::eMPG:
				if (MpegCheck(p, n, frameSize, channels, samplingRate))
					codec = cAudioCodec::eMPG;
				break;

			case cAudioCodec::eAC3:
				if (Ac3Check(p, n, frameSize, channels, samplingRate))
				{
					codec = cAudioCodec::eAC3;
					if (n > 5 && p[5] > (10 << 3))
						codec = cAudioCodec::eEAC3;
				}
				break;

			case cAudioCodec::eAAC:
				if (AdtsCheck(p, n, frameSize, channels, samplingRate))
					codec = cAudioCodec::eAAC;
				break;

#ifdef ENABLE_AAC_LATM
			case cAudioCodec::eAAC_LATM:
				if (LatmCheck(p, n, frameSize, channels, samplingRate))
					codec = cAudioCodec::eAAC_LATM;
				break;
#endif

			case cAudioCodec::eDTS:
				if (DtsCheck(p, n, frameSize, channels, samplingRate))
					codec = cAudioCodec::eDTS;
				break;

			default:
				break;
			}

			if (codec != cAudioCodec::eInvalid)
			{
				// if there is enough data in buffer, check if predicted next
				// frame start is valid
				if (n < frameSize + 4 ||
						FastCheck(p + frameSize) != cAudioCodec::eInvalid)
				{
					// if codec has been detected but buffer does not yet
					// contains a complete frame, set size to zero to prevent
					// frame from being decoded
					if (frameSize > n)
						frameSize = 0;

					break;
				}
			}

			++offset;
		}

		if (offset)
		{
			DBG("audio parser skipped %u of %u bytes", offset, m_size);
			Shrink(offset, true);
		}

		if (codec != cAudioCodec::eInvalid)
		{
			m_codec = codec;
			m_channels = channels;
			m_samplingRate = samplingRate;
			m_packet.size = frameSize;
		}
		else
			m_packet.size = 0;

		m_parsed = true;
		m_mutex->Unlock();
	}

	struct Pts
	{
		Pts(int64_t _pts, unsigned int _length)
			: pts(_pts), length(_length) { };

		int64_t 		pts;
		unsigned int 	length;
	};

	cMutex*				m_mutex;
	AVPacket 			m_packet;
	cAudioCodec::eCodec m_codec;
	unsigned int		m_channels;
	unsigned int		m_samplingRate;
	unsigned int		m_size;
	std::queue<Pts*> 	m_ptsQueue;
	bool				m_parsed;

	/* ---------------------------------------------------------------------- */
	/*     audio codec parser helper functions, based on vdr-softhddevice     */
	/* ---------------------------------------------------------------------- */

	static const uint16_t BitRateTable[2][3][16];
	static const uint16_t MpegSampleRateTable[4];
	static const uint32_t Mpeg4SampleRateTable[16];
	static const uint16_t Ac3SampleRateTable[4];
	static const uint16_t Ac3FrameSizeTable[38][3];
	static const uint32_t DtsSampleRateTable[16];

	static cAudioCodec::eCodec FastCheck(const uint8_t *p)
	{
		return 	FastMpegCheck(p)  ? cAudioCodec::eMPG      :
				FastAc3Check (p)  ? cAudioCodec::eAC3      :
				FastAdtsCheck(p)  ? cAudioCodec::eAAC      :
#ifdef ENABLE_AAC_LATM
				FastLatmCheck(p)  ? cAudioCodec::eAAC_LATM :
#endif
				FastDtsCheck (p)  ? cAudioCodec::eDTS      :
									cAudioCodec::eInvalid;
	}

	///
	///	Fast check for MPEG audio.
	///
	///	0xFFE... MPEG audio
	///
	static bool FastMpegCheck(const uint8_t *p)
	{
		if (p[0] != 0xFF)			// 11bit frame sync
			return false;
		if ((p[1] & 0xE0) != 0xE0)
			return false;
		if ((p[1] & 0x18) == 0x08)	// version ID - 01 reserved
			return false;
		if (!(p[1] & 0x06))			// layer description - 00 reserved
			return false;
		if ((p[2] & 0xF0) == 0xF0)	// bit rate index - 1111 reserved
			return false;
		if ((p[2] & 0x0C) == 0x0C)	// sampling rate index - 11 reserved
			return false;
		return true;
	}

	///	Check for MPEG audio.
	///
	///	0xFFEx already checked.
	///
	///	From: http://www.mpgedit.org/mpgedit/mpeg_format/mpeghdr.htm
	///
	///	AAAAAAAA AAABBCCD EEEEFFGH IIJJKLMM
	///
	///	o a 11x Frame sync
	///	o b 2x	MPEG audio version (2.5, reserved, 2, 1)
	///	o c 2x	Layer (reserved, III, II, I)
	///	o e 2x	BitRate index
	///	o f 2x	SampleRate index (41000, 48000, 32000, 0)
	///	o g 1x	Padding bit
	/// o h 1x  Private bit
	/// o i 2x  Channel mode
	///	o ..	Doesn't care
	///
	///	frame length:
	///	Layer I:
	///		FrameLengthInBytes = (12 * BitRate / SampleRate + Padding) * 4
	///	Layer II & III:
	///		FrameLengthInBytes = 144 * BitRate / SampleRate + Padding
	///
	static bool MpegCheck(const uint8_t *p, unsigned int size,
			unsigned int &frameSize, unsigned int &channels,
			unsigned int &samplingRate)
	{
		frameSize = size;
		if (size < 4)
			return true;

		int cmode = (p[3] >> 6) & 0x03;
		int mpeg2 = !(p[1] & 0x08) && (p[1] & 0x10);
		int mpeg25 = !(p[1] & 0x08) && !(p[1] & 0x10);
		int layer = 4 - ((p[1] >> 1) & 0x03);
		int padding = (p[2] >> 1) & 0x01;

		// channel mode = [ stereo, joint stereo, dual channel, mono]
		channels = cmode == 0x03 ? 1 : 2;

		samplingRate = MpegSampleRateTable[(p[2] >> 2) & 0x03];
		if (!samplingRate)
			return false;

		samplingRate >>= mpeg2;		// MPEG 2 half rate
		samplingRate >>= mpeg25;	// MPEG 2.5 quarter rate

		int bit_rate =
				BitRateTable[mpeg2 | mpeg25][layer - 1][(p[2] >> 4) & 0x0F];
		if (!bit_rate)
			return false;

		switch (layer)
		{
		case 1:
			frameSize = (12000 * bit_rate) / samplingRate;
			frameSize = (frameSize + padding) * 4;
			break;
		case 2:
		case 3:
		default:
			frameSize = (144000 * bit_rate) / samplingRate;
			frameSize = frameSize + padding;
			break;
		}
		return true;
	}

	///
	///	Fast check for (E-)AC-3 audio.
	///
	///	0x0B77... AC-3 audio
	///
	static bool FastAc3Check(const uint8_t *p)
	{
		if (p[0] != 0x0B)			// 16bit sync
			return false;
		if (p[1] != 0x77)
			return false;
		return true;
	}

	///
	///	Check for (E-)AC-3 audio.
	///
	///	0x0B77xxxxxx already checked.
	///
	///	o AC-3 Header
	///	AAAAAAAA AAAAAAAA BBBBBBBB BBBBBBBB CCDDDDDD EEEEEFFF GGGxxxxx
	///
	///	o a 16x Frame sync, always 0x0B77
	///	o b 16x CRC 16
	///	o c 2x	Sample rate ( 48000, 44100, 32000, reserved )
	///	o d 6x	Frame size code
	///	o e 5x	Bit stream ID
	///	o f 3x	Bit stream mode
	/// o g 3x  Audio coding mode
	///
	///	o E-AC-3 Header
	///	AAAAAAAA AAAAAAAA BBCCCDDD DDDDDDDD EEFFGGGH IIIII...
	///
	///	o a 16x Frame sync, always 0x0B77
	///	o b 2x	Frame type
	///	o c 3x	Sub stream ID
	///	o d 11x Frame size - 1 in words
	///	o e 2x	Frame size code
	///	o f 2x	Frame size code 2
	/// o g 3x  Channel mode
	/// 0 h 1x  LFE on
	///
	static bool Ac3Check(const uint8_t *p, unsigned int size,
			unsigned int &frameSize, unsigned int &channels,
			unsigned int &samplingRate)
	{
		frameSize = size;
		if (size < 7)
			return true;

		int acmod;
		bool lfe;
		int fscod = (p[4] & 0xC0) >> 6;

		samplingRate = Ac3SampleRateTable[fscod];

		if (p[5] > (10 << 3))		// E-AC-3
		{
			if (fscod == 0x03)
			{
				int fscod2 = (p[4] & 0x30) >> 4;
				if (fscod2 == 0x03)
					return false;		// invalid fscod & fscod2

				samplingRate = Ac3SampleRateTable[fscod2] / 2;
			}

			acmod = (p[4] & 0x0E) >> 1;	// number of channels, LFE excluded
			lfe = p[4] & 0x01;

			frameSize = ((p[2] & 0x07) << 8) + p[3] + 1;
			frameSize *= 2;
		}
		else						// AC-3
		{
			if (fscod == 0x03)		// invalid sample rate
				return false;

			int frmsizcod = p[4] & 0x3F;
			if (frmsizcod > 37)		// invalid frame size
				return false;

			acmod = p[6] >> 5;		// number of channels, LFE excluded

			int lfe_bptr = 51;		// position of LFE bit in header for 2.0
			if ((acmod & 0x01) && (acmod != 0x01))
				lfe_bptr += 2;		// skip center mix level
			if (acmod & 0x04)
				lfe_bptr += 2;		// skip surround mix level
			if (acmod == 0x02)
				lfe_bptr += 2;		// skip surround mode
			lfe = (p[lfe_bptr / 8] & (1 << (7 - (lfe_bptr % 8))));

			// invalid is checked above
			frameSize = Ac3FrameSizeTable[frmsizcod][fscod] * 2;
		}

		channels =
			acmod == 0x00 ? 2 : 	// Ch1, Ch2
			acmod == 0x01 ? 1 : 	// C
			acmod == 0x02 ? 2 : 	// L, R
			acmod == 0x03 ? 3 : 	// L, C, R
			acmod == 0x04 ? 3 : 	// L, R, S
			acmod == 0x05 ? 4 : 	// L, C, R, S
			acmod == 0x06 ? 4 : 	// L, R, RL, RR
			acmod == 0x07 ? 5 : 0;	// L, C, R, RL, RR

		if (lfe) channels++;
		return true;
	}

#ifdef ENABLE_AAC_LATM
	///
	///	Fast check for AAC LATM audio.
	///
	///	0x56E... AAC LATM audio
	///
	static bool FastLatmCheck(const uint8_t *p)
	{
		if (p[0] != 0x56)			// 11bit sync
			return false;
		if ((p[1] & 0xE0) != 0xE0)
			return false;
		return true;
	}

	///
	///	Check for AAC LATM audio.
	///
	///	0x56Exxx already checked.
	///
	static bool LatmCheck(const uint8_t *p, unsigned int size,
			unsigned int &frameSize, unsigned int &channels,
			unsigned int &samplingRate)
	{
		frameSize = size;
		if (size < 3)
			return true;

		// to do: determine channels
		channels = 2;

		// to do: determine sampling rate
		samplingRate = 48000;

		// 13 bit frame size without header
		frameSize = ((p[1] & 0x1F) << 8) + p[2];
		frameSize += 3;
		return true;
	}
#endif
	
	///
	///	Fast check for ADTS Audio Data Transport Stream.
	///
	///	0xFFF...  ADTS audio
	///
	static bool FastAdtsCheck(const uint8_t *p)
	{
		if (p[0] != 0xFF)			// 12bit sync
			return false;
		if ((p[1] & 0xF6) != 0xF0)	// sync + layer must be 0
			return false;
		if ((p[2] & 0x3C) == 0x3C)	// sampling frequency index != 15
			return false;
		return true;
	}

	///
	///	Check for ADTS Audio Data Transport Stream.
	///
	///	0xFFF already checked.
	///
	///	AAAAAAAA AAAABCCD EEFFFFGH HHIJKLMM MMMMMMMM MMMOOOOO OOOOOOPP
	///	(QQQQQQQQ QQQQQQQ)
	///
	///	o A*12	sync word 0xFFF
	///	o B*1	MPEG Version: 0 for MPEG-4, 1 for MPEG-2
	///	o C*2	layer: always 0
	///	o ..
	///	o F*4	sampling frequency index (15 is invalid)
	///	o ..
	/// o H*3	MPEG-4 channel configuration
	/// o ...
	///	o M*13	frame length
	///
	static bool AdtsCheck(const uint8_t *p, unsigned int size,
			unsigned int &frameSize, unsigned int &channels,
			unsigned int &samplingRate)
	{
		frameSize = size;
		if (size < 6)
			return true;

		samplingRate = Mpeg4SampleRateTable[(p[2] >> 2) & 0x0F];

		frameSize = (p[3] & 0x03) << 11;
		frameSize |= (p[4] & 0xFF) << 3;
		frameSize |= (p[5] & 0xE0) >> 5;

	    int cConf = (p[2] & 0x01) << 7;
	    cConf |= (p[3] & 0xC0) >> 6;
	    channels =
	    	cConf == 0x00 ? 0 : // defined in AOT specific config
			cConf == 0x01 ? 1 : // C
	    	cConf == 0x02 ? 2 : // L, R
	    	cConf == 0x03 ? 3 : // C, L, R
	    	cConf == 0x04 ? 4 : // C, L, R, RC
	    	cConf == 0x05 ? 5 : // C, L, R, RL, RR
	    	cConf == 0x06 ? 6 : // C, L, R, RL, RR, LFE
	    	cConf == 0x07 ? 8 : // C, L, R, SL, SR, RL, RR, LFE
				0;

		if (!samplingRate || !channels)
			return false;

	    return true;
	}

	///
	///	Fast check for DTS Audio Data Transport Stream.
	///
	///	0x7FFE8001....  DTS audio
	///
	static bool FastDtsCheck(const uint8_t *p)
	{
		if (p[0] != 0x7F)			// 32bit sync
			return false;
		if (p[1] != 0xFE)
			return false;
		if (p[2] != 0x80)
			return false;
		if (p[3] != 0x01)
			return false;
		return true;
	}

	///
	///	Check for DTS Audio Data Transport Stream.
	///
	///	0x7FFE8001 already checked.
	///
	///	AAAAAAAA AAAAAAAA AAAAAAAA AAAAAAAA BCCCCCDE EEEEEEFF FFFFFFFF FFFFGGGG
	/// GGHHHHII IIIJKLMN OOOPQRRS TTTTTTTT TTTTTTTT UVVVVWWX XXYZaaaa
	///
	///	o A*32	sync word 0x7FFE8001
	///	o B*1   frame type
	///	o C*5   deficit sample count
	///	o D*1   CRC present flag
	///	o E*7   number of PCM sample blocks
	///	o F*14  primary frame size
	///	o G*6   audio channel arrangement
	///	o H*4   core audio sampling frequency
	///	o I*5   transmission bit rate
	///	o J*1   embedded downmix enabled
	///	o K*1   embedded dynamic range flag
	///	o L*1   embedded time stamp flag
	///	o M*1   auxiliary data flag
	///	o N*1   HDCD
	///	o O*3   extension audio descriptor flag
	///	o P*1   extended coding flag
	///	o Q*1   audio sync word insertion flag
	///	o R*2   low frequency effects flag
	///	o S*1   predictor history flag
	///	o T*16  header CRC check (if CRC present flag set)
	///	o U*1   multi rate interpolator switch
	///	o V*4   encoder software revision
	///	o W*2   copy history
	///	o X*3   source PCM resolution
	///	o Y*1   front sum/difference flag
	///	o Z*1   surrounds sum/difference flag
	///	o a*4   dialog normalization parameter
	///
	static bool DtsCheck(const uint8_t *p, unsigned int size,
			unsigned int &frameSize, unsigned int &channels,
			unsigned int &samplingRate)
	{
		frameSize = size;
		if (size < 11)
			return true;

		frameSize = ((p[5] & 0x03) << 12) + (p[6] << 4) + ((p[7] & 0xF0) >> 4);
		frameSize++;

		samplingRate = DtsSampleRateTable[(p[8] & 0x3C) >> 2];

		int amode = ((p[7] & 0x0F) << 2) + ((p[8] & 0xC0) >> 6);
		channels =
			amode == 0x00 ? 1 : 	// mono
			amode == 0x02 ? 2 : 	// L, R
			amode == 0x03 ? 2 : 	// (L + R), (L - R)
			amode == 0x04 ? 2 : 	// LT, RT
			amode == 0x05 ? 3 : 	// L, R, C
			amode == 0x06 ? 3 : 	// L, R, S
			amode == 0x08 ? 4 : 	// L, R, RL, RR
			amode == 0x09 ? 5 : 0;	// L, C, R, RL, RR

		if (!samplingRate || !channels)
			return false;

		if (p[10] & 0x06) channels++;
		return true;
	}
};

///
///	MPEG bit rate table.
///
///	BitRateTable[Version][Layer][Index]
///
const uint16_t cRpiAudioDecoder::cParser::BitRateTable[2][3][16] =
{
	{	// MPEG Version 1
		{0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, 0},
		{0, 32, 48, 56,  64,  80,  96, 112, 128, 160, 192, 224, 256, 320, 384, 0},
		{0, 32, 40, 48,  56,  64,  80,  96, 112, 128, 160, 192, 224, 256, 320, 0}
	},
	{	// MPEG Version 2 & 2.5
		{0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256, 0},
		{0,  8, 16, 24, 32, 40, 48,  56,  64,  80,  96, 112, 128, 144, 160, 0},
		{0,  8, 16, 24, 32, 40, 48,  56,  64,  80,  96, 112, 128, 144, 160, 0}
	}
};

///
///	MPEG sample rate table.
///
const uint16_t cRpiAudioDecoder::cParser::MpegSampleRateTable[4] =
	{ 44100, 48000, 32000, 0 };

///
///	MPEG-4 sample rate table.
///
const uint32_t cRpiAudioDecoder::cParser::Mpeg4SampleRateTable[16] = {
		96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050,
		16000, 12000, 11025,  8000,  7350,     0,     0,     0
};

///
///	AC-3 sample rate table.
///
const uint16_t cRpiAudioDecoder::cParser::Ac3SampleRateTable[4] =
	{ 48000, 44100, 32000, 0 };

///
///	Possible AC-3 frame sizes.
///
///	from ATSC A/52 table 5.18 frame size code table.
///
const uint16_t cRpiAudioDecoder::cParser::Ac3FrameSizeTable[38][3] =
{
	{  64,   69,   96}, {  64,   70,   96}, {  80,   87,  120}, { 80,  88,  120},
	{  96,  104,  144}, {  96,  105,  144}, { 112,  121,  168}, {112, 122,  168},
	{ 128,  139,  192}, { 128,  140,  192}, { 160,  174,  240}, {160, 175,  240},
	{ 192,  208,  288}, { 192,  209,  288}, { 224,  243,  336}, {224, 244,  336},
	{ 256,  278,  384}, { 256,  279,  384}, { 320,  348,  480}, {320, 349,  480},
	{ 384,  417,  576}, { 384,  418,  576}, { 448,  487,  672}, {448, 488,  672},
	{ 512,  557,  768}, { 512,  558,  768}, { 640,  696,  960}, {640, 697,  960},
	{ 768,  835, 1152}, { 768,  836, 1152}, { 896,  975, 1344}, {896, 976, 1344},
	{1024, 1114, 1536}, {1024, 1115, 1536}, {1152, 1253, 1728},
	{1152, 1254, 1728}, {1280, 1393, 1920}, {1280, 1394, 1920},
};

///
///	DTS sample rate table.
///
const uint32_t cRpiAudioDecoder::cParser::DtsSampleRateTable[16] =
	{ 0,  8000, 16000, 32000, 64000,
	  0, 11025, 22050, 44100, 88200,
	  0, 12000, 24000, 48000, 96000, 0 };

/* ------------------------------------------------------------------------- */

#define AV_CH_LAYOUT(ch) ( \
		ch == 1 ? AV_CH_LAYOUT_MONO    : \
		ch == 2 ? AV_CH_LAYOUT_STEREO  : \
		ch == 3 ? AV_CH_LAYOUT_2POINT1 : \
		ch == 6 ? AV_CH_LAYOUT_5POINT1 : 0)

#define AV_SAMPLE_STR(fmt) ( \
		fmt == AV_SAMPLE_FMT_U8   ? "U8"             : \
		fmt == AV_SAMPLE_FMT_S16  ? "S16"            : \
		fmt == AV_SAMPLE_FMT_S32  ? "S32"            : \
		fmt == AV_SAMPLE_FMT_FLT  ? "float"          : \
		fmt == AV_SAMPLE_FMT_DBL  ? "double"         : \
		fmt == AV_SAMPLE_FMT_U8P  ? "U8, planar"     : \
		fmt == AV_SAMPLE_FMT_S16P ? "S16, planar"    : \
		fmt == AV_SAMPLE_FMT_S32P ? "S32, planar"    : \
		fmt == AV_SAMPLE_FMT_FLTP ? "float, planar"  : \
		fmt == AV_SAMPLE_FMT_DBLP ? "double, planar" : "unknown")

/* ------------------------------------------------------------------------- */

class cRpiAudioRender
{

public:

	cRpiAudioRender(cOmx *omx) :
		m_mutex(new cMutex()),
		m_omx(omx),
		m_port(cRpiAudioPort::eLocal),
		m_codec(cAudioCodec::eInvalid),
		m_inChannels(0),
		m_outChannels(0),
		m_samplingRate(0),
		m_frameSize(0),
		m_configured(false),
		m_running(false),
#ifdef DO_RESAMPLE
		m_resample(0),
		m_resamplerConfigured(false),
#endif
		m_pcmSampleFormat(AV_SAMPLE_FMT_NONE),
		m_pts(0)
	{
	}

	~cRpiAudioRender()
	{
		Flush();
#ifdef DO_RESAMPLE
		swr_free(&m_resample);
#endif
		delete m_mutex;
	}

	int WriteSamples(uint8_t** data, int samples, int64_t pts,
			AVSampleFormat sampleFormat = AV_SAMPLE_FMT_NONE)
	{
		if (!Ready())
			return 0;

		m_mutex->Lock();
		int copied = 0;

		if (sampleFormat == AV_SAMPLE_FMT_NONE)
		{
			// pass through
			while (samples > copied)
			{
				OMX_BUFFERHEADERTYPE *buf = m_omx->GetAudioBuffer(pts);
				if (!buf)
					break;

				unsigned int len = samples - copied;
				if (len > buf->nAllocLen)
					len = buf->nAllocLen;

				memcpy(buf->pBuffer, *data + copied, len);
				buf->nFilledLen = len;

				if (!m_omx->EmptyAudioBuffer(buf))
					break;

				copied += len;
				pts = 0;
			}
		}
		else
		{
#ifdef DO_RESAMPLE
			// local decode, do resampling
			if (!m_resamplerConfigured || m_pcmSampleFormat != sampleFormat)
			{
				m_pcmSampleFormat = sampleFormat;
				ApplyResamplerSettings();
			}
			if (m_resample)
			{
				m_pts = pts ? pts : m_pts;
				OMX_BUFFERHEADERTYPE *buf = m_omx->GetAudioBuffer(m_pts);
				if (buf)
				{
					if (buf->nAllocLen >= (samples * m_outChannels *
						av_get_bytes_per_sample(AV_SAMPLE_FMT_S16)))
					{
						uint8_t *dst[] = { buf->pBuffer };
						int copiedSamples = swr_convert(m_resample,
							dst, samples, (const uint8_t **)data, samples);

						buf->nFilledLen = av_samples_get_buffer_size(NULL,
							m_outChannels, copiedSamples, AV_SAMPLE_FMT_S16, 1);

						m_pts += copiedSamples * 90000 / m_samplingRate;
					}
					copied = m_omx->EmptyAudioBuffer(buf) ? samples : 0;
				}
			}
#else
			// local decode, no resampling
			m_pts = pts ? pts : m_pts;
			OMX_BUFFERHEADERTYPE *buf = m_omx->GetAudioBuffer(m_pts);
			if (buf)
			{
				unsigned int size = samples * m_outChannels *
						av_get_bytes_per_sample(AV_SAMPLE_FMT_S16);
				if (buf->nAllocLen >= size)
				{
					memcpy(buf->pBuffer, *data, size);
					buf->nFilledLen = size;
					m_pts += samples * 90000 / m_samplingRate;
				}
				copied = m_omx->EmptyAudioBuffer(buf) ? samples : 0;
			}
#endif
		}
		m_mutex->Unlock();
		return copied;
	}

	void Flush(void)
	{
		m_mutex->Lock();
		if (m_running)
			m_omx->StopAudio();
		m_configured = false;
		m_running = false;
		m_pts = 0;
		m_mutex->Unlock();
	}

	void SetCodec(cAudioCodec::eCodec codec, unsigned int channels,
			unsigned int samplingRate, unsigned int frameSize)
	{
		m_mutex->Lock();
		if (codec != cAudioCodec::eInvalid && channels > 0)
		{
			m_inChannels = channels;
			cRpiAudioPort::ePort newPort = cRpiSetup::GetAudioPort();
			cAudioCodec::eCodec newCodec = cAudioCodec::ePCM;

			DLOG("new audio codec: %dch %s", channels, cAudioCodec::Str(codec));

			if (newPort == cRpiAudioPort::eHDMI)
			{
				// check if pass through is possible
				if (cRpiSetup::IsAudioFormatSupported(codec, channels,
							samplingRate))
					newCodec = codec;

				// check for multi channel PCM, stereo downmix if not supported
				else if (!cRpiSetup::IsAudioFormatSupported(cAudioCodec::ePCM,
						channels, samplingRate))
					channels = 2;
			}
			else
				channels = 2;

			// if the user changes the port, this should change immediately
			if (newPort != m_port)
				Flush();

			// save new settings to be applied when render is ready
			if (newPort != m_port || m_codec != newCodec ||
					m_outChannels != channels || m_samplingRate != samplingRate)
			{
				m_configured = false;
				m_port = newPort;
				m_codec = newCodec;
				m_outChannels = channels;
				m_samplingRate = samplingRate;
				m_frameSize = frameSize;
			}
#ifdef DO_RESAMPLE
			m_resamplerConfigured = false;
#endif
		}
		m_mutex->Unlock();
	}

	bool IsPassthrough(void)
	{
		return m_codec != cAudioCodec::ePCM;
	}

	int GetChannels(void)
	{
		return m_outChannels;
	}

	bool Ready(void)
	{
		if (!m_configured)
		{
			// wait until render is ready before applying new settings
			if (m_running && m_omx->GetAudioLatency())
				return false;

			ApplyRenderSettings();
		}
		return true;
	}

private:

	cRpiAudioRender(const cRpiAudioRender&);
	cRpiAudioRender& operator= (const cRpiAudioRender&);

	void ApplyRenderSettings(void)
	{
		if (m_running)
			m_omx->StopAudio();

		if (m_codec != cAudioCodec::eInvalid)
		{
			if (m_port == cRpiAudioPort::eHDMI)
				cRpiSetup::SetHDMIChannelMapping(m_codec != cAudioCodec::ePCM,
						m_outChannels);

			m_omx->SetupAudioRender(m_codec, m_outChannels, m_port,
					m_samplingRate, m_frameSize);

			DLOG("set %s audio output format to %dch %s, %d.%dkHz%s",
					cRpiAudioPort::Str(m_port), m_outChannels,
					cAudioCodec::Str(m_codec),
					m_samplingRate / 1000, (m_samplingRate % 1000) / 100,
					m_codec != cAudioCodec::ePCM ? " (pass-through)" : "");
		}
		m_running = m_codec != cAudioCodec::eInvalid;
		m_configured = true;
	}

#ifdef DO_RESAMPLE
	void ApplyResamplerSettings(void)
	{
		swr_free(&m_resample);
		m_resample = swr_alloc();
		if (m_resample)
		{
			av_opt_set_int(m_resample, "in_sample_rate", m_samplingRate, 0);
			av_opt_set_int(m_resample, "in_sample_fmt", m_pcmSampleFormat, 0);
			av_opt_set_int(m_resample, "in_channel_count", m_inChannels, 0);
			av_opt_set_int(m_resample, "in_channel_layout",
					AV_CH_LAYOUT(m_inChannels), 0);

			av_opt_set_int(m_resample, "out_sample_rate", m_samplingRate, 0);
			av_opt_set_int(m_resample, "out_sample_fmt", AV_SAMPLE_FMT_S16, 0);
			av_opt_set_int(m_resample, "out_channel_count", m_outChannels, 0);
			av_opt_set_int(m_resample, "out_channel_layout",
					AV_CH_LAYOUT(m_outChannels), 0);

			swr_init(m_resample);
			m_resamplerConfigured = true;
		}
		else
			ELOG("failed to allocate resampling context!");
	}
#endif

	cMutex		        *m_mutex;
	cOmx		        *m_omx;

	cRpiAudioPort::ePort m_port;
	cAudioCodec::eCodec  m_codec;
	unsigned int         m_inChannels;
	unsigned int         m_outChannels;
	unsigned int         m_samplingRate;
	unsigned int         m_frameSize;
	bool                 m_configured;
	bool                 m_running;

#ifdef DO_RESAMPLE
	SwrContext          *m_resample;
	bool                 m_resamplerConfigured;
#endif

	AVSampleFormat       m_pcmSampleFormat;
	int64_t              m_pts;
};

/* ------------------------------------------------------------------------- */

cRpiAudioDecoder::cRpiAudioDecoder(cOmx *omx) :
	cThread("audio decoder"),
	m_passthrough(false),
	m_reset(false),
	m_setupChanged(true),
	m_wait(new cCondWait()),
	m_parser(new cParser()),
	m_render(new cRpiAudioRender(omx))
{
	memset(m_codecs, 0, sizeof(m_codecs));
}

cRpiAudioDecoder::~cRpiAudioDecoder()
{
	if (Active())
		Reset();

	delete m_render;
	delete m_parser;
	delete m_wait;
}

extern int SysLogLevel;

int cRpiAudioDecoder::Init(void)
{
	int ret = m_parser->Init();
	if (ret)
		return ret;

	avcodec_register_all();

	av_log_set_level(
			SysLogLevel > 2 ? AV_LOG_VERBOSE :
			SysLogLevel > 1 ? AV_LOG_INFO : AV_LOG_ERROR);
	av_log_set_callback(&Log);

	m_codecs[cAudioCodec::ePCM     ].codec = NULL;
	m_codecs[cAudioCodec::eMPG     ].codec = avcodec_find_decoder(AV_CODEC_ID_MP3);
	m_codecs[cAudioCodec::eAC3     ].codec = avcodec_find_decoder(AV_CODEC_ID_AC3);
	m_codecs[cAudioCodec::eEAC3    ].codec = avcodec_find_decoder(AV_CODEC_ID_EAC3);
	m_codecs[cAudioCodec::eAAC     ].codec = avcodec_find_decoder(AV_CODEC_ID_AAC);
#ifdef ENABLE_AAC_LATM
	m_codecs[cAudioCodec::eAAC_LATM].codec = avcodec_find_decoder(AV_CODEC_ID_AAC_LATM);
#endif
	m_codecs[cAudioCodec::eDTS     ].codec = avcodec_find_decoder(AV_CODEC_ID_DTS);

	for (int i = 0; i < cAudioCodec::eNumCodecs; i++)
	{
		cAudioCodec::eCodec codec = static_cast<cAudioCodec::eCodec>(i);
		if (m_codecs[codec].codec)
		{
			m_codecs[codec].context = avcodec_alloc_context3(m_codecs[codec].codec);
			if (!m_codecs[codec].context)
			{
				ELOG("failed to allocate %s context!", cAudioCodec::Str(codec));
				ret = -1;
				break;
			}
			if (avcodec_open2(m_codecs[codec].context, m_codecs[codec].codec, NULL) < 0)
			{
				ELOG("failed to open %s decoder!", cAudioCodec::Str(codec));
				ret = -1;
				break;
			}
		}
	}

	if (!ret)
	{
		cRpiSetup::SetAudioSetupChangedCallback(&OnAudioSetupChanged, this);
		Start();
	}
	else
		DeInit();

	return ret;
}

int cRpiAudioDecoder::DeInit(void)
{
	Lock();

	Reset();
	Cancel(-1);
	m_wait->Signal();

	while (Active())
		cCondWait::SleepMs(5);

	m_render->Flush();
	cRpiSetup::SetAudioSetupChangedCallback(0);

	for (int i = 0; i < cAudioCodec::eNumCodecs; i++)
	{
		cAudioCodec::eCodec codec = static_cast<cAudioCodec::eCodec>(i);
		if (m_codecs[codec].codec)
			avcodec_close(m_codecs[codec].context);
	}

	av_log_set_callback(&av_log_default_callback);
	m_parser->DeInit();

	Unlock();
	return 0;
}

bool cRpiAudioDecoder::WriteData(const unsigned char *buf, unsigned int length,
		int64_t pts)
{
	Lock();

	bool ret = m_parser->Append(buf, pts, length);
	if (ret)
		m_wait->Signal();

	Unlock();
	return ret;
}

void cRpiAudioDecoder::Reset(void)
{
	Lock();
	m_reset = true;
	m_wait->Signal();
	while (m_reset)
		cCondWait::SleepMs(5);
	Unlock();
}

bool cRpiAudioDecoder::Poll(void)
{
	return m_parser->GetFreeSpace() > KILOBYTE(16);
}

void cRpiAudioDecoder::HandleAudioSetupChanged()
{
	DBG("HandleAudioSetupChanged()");
	m_setupChanged = true;
}

void cRpiAudioDecoder::Action(void)
{
	SetPriority(-15);
	DLOG("cAudioDecoder() thread started");

	unsigned int channels = 0;
	unsigned int samplingRate = 0;
	cAudioCodec::eCodec codec = cAudioCodec::eInvalid;

	AVFrame *frame = av_frame_alloc();
	if (!frame)
	{
		ELOG("failed to allocate audio frame!");
		return;
	}

	while (Running())
	{
		if (m_reset)
		{
			m_parser->Reset();
			m_render->Flush();
			av_frame_unref(frame);
			m_reset = false;
		}

		// test for codec change if there is data in parser and no left over
		if (!m_parser->Empty() && !frame->nb_samples)
			m_setupChanged |= codec != m_parser->GetCodec() ||
				channels != m_parser->GetChannels() ||
				samplingRate != m_parser->GetSamplingRate();

		// if necessary, set up audio codec
		if (!m_parser->Empty() && m_setupChanged)
		{
			if (codec != m_parser->GetCodec() && codec != cAudioCodec::eInvalid)
				avcodec_flush_buffers(m_codecs[codec].context);

			codec = m_parser->GetCodec();
			channels = m_parser->GetChannels();
			samplingRate = m_parser->GetSamplingRate();

			// validate channel layout and apply new audio parameters
			if (AV_CH_LAYOUT(channels))
			{
				m_setupChanged = false;
				m_render->SetCodec(codec, channels, samplingRate,
						m_parser->GetFrameSize());

#ifndef DO_RESAMPLE
#if FF_API_REQUEST_CHANNELS
				// if there's no libswresample, let decoder do the down mix
				m_codecs[codec].context->request_channels =
						m_render->GetChannels();
#endif
				m_codecs[codec].context->request_channel_layout =
						AV_CH_LAYOUT(m_render->GetChannels());
#endif
			}
			m_reset = m_setupChanged;
			continue;
		}

		// if there's audio data available...
		if (!m_parser->Empty())
		{
			// ... either pass through if render is ready
			if (m_render->IsPassthrough())
			{
				if (m_render->Ready())
				{
					int len = m_render->WriteSamples(&m_parser->Packet()->data,
							m_parser->Packet()->size, m_parser->GetPts());
					if (len)
					{
						m_parser->Shrink(len);
						continue;
					}
				}
			}
			// ... or decode if there's no leftover
			else if (!frame->nb_samples)
			{
				int gotFrame = 0;
				int len = avcodec_decode_audio4(m_codecs[codec].context,
						frame, &gotFrame, m_parser->Packet());

				if (len > 0 && gotFrame)
				{
					frame->pts = m_parser->GetPts();
					m_parser->Shrink(len);
				}
				else
				{
					ELOG("failed to decode audio frame!");
					m_parser->Reset();
					av_frame_unref(frame);
					continue;
				}
			}
		}
		// if there's leftover, pass decoded audio data to render when ready
		if (frame->nb_samples && m_render->Ready())
		{
			int len = m_render->WriteSamples(frame->extended_data,
					frame->nb_samples, frame->pts,
					(AVSampleFormat)frame->format);
			if (len)
			{
				av_frame_unref(frame);
				continue;
			}
		}
		// nothing to be done...
		m_wait->Wait(50);
	}

	av_frame_free(&frame);
	DLOG("cAudioDecoder() thread ended");
}

void cRpiAudioDecoder::Log(void* ptr, int level, const char* fmt, va_list vl)
{
	if (level == AV_LOG_QUIET)
		return;

	char line[128];
	vsnprintf(line, sizeof(line), fmt, vl);

	if (level <= AV_LOG_ERROR)
		ELOG("[libav] %s", line);
	else if (level <= AV_LOG_INFO)
		ILOG("[libav] %s", line);
	else if (level <= AV_LOG_VERBOSE)
		DLOG("[libav] %s", line);
}