Files
cantata/replaygain/ffmpeginput.cpp
2022-01-31 14:07:26 +11:00

689 lines
22 KiB
C++

/*
* Cantata
*
* Copyright (c) 2011-2022 Craig Drummond <craig.p.drummond@gmail.com>
*
*/
/* This class is a C++/Qt version of input_ffmpeg.c from libebur128
*/
/* See LICENSE file for copyright and license details. */
#ifdef __cplusplus
#define __STDC_CONSTANT_MACROS
#ifdef _STDINT_H
#undef _STDINT_H
#endif
#include <stdint.h>
#endif
#ifdef __cplusplus
extern "C" {
#endif
#include <libavcodec/avcodec.h>
#include <libavformat/avformat.h>
#if LIBAVFORMAT_VERSION_MAJOR >= 54
#include <libavutil/channel_layout.h>
#endif
#ifdef __cplusplus
}
#endif
#include <QMutex>
#include <QFile>
#include <QString>
#include <QList>
#include <QByteArray>
#include "ebur128/ebur128.h"
#include "ffmpeginput.h"
static QMutex mutex;
#ifndef AVCODEC_MAX_AUDIO_FRAME_SIZE
#define AVCODEC_MAX_AUDIO_FRAME_SIZE 192000 // 1 second of 48khz 32bit audio
#endif
#if defined FF_INPUT_BUFFER_PADDING_SIZE
#define FFMPEG_INPUT_BUFFER_PADDING_SIZE FF_INPUT_BUFFER_PADDING_SIZE
#elif defined AV_INPUT_BUFFER_PADDING_SIZE
#define FFMPEG_INPUT_BUFFER_PADDING_SIZE AV_INPUT_BUFFER_PADDING_SIZE
#else
#define FFMPEG_INPUT_BUFFER_PADDING_SIZE 32
#endif
#define BUFFER_SIZE ((((AVCODEC_MAX_AUDIO_FRAME_SIZE * 3) / 2) * sizeof(int16_t)) + FFMPEG_INPUT_BUFFER_PADDING_SIZE)
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(57, 89, 100) // Not 100% of version here!
#define GET_CODEC_TYPE(A) A->codecpar->codec_type
#else
#define GET_CODEC_TYPE(A) A->codec->codec_type
#endif
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(57, 24, 0)
#define AV_FREE(PKT) av_packet_unref(PKT)
#else
#define AV_FREE(PKT) av_free_packet(PKT)
#endif
struct FfmpegInput::Handle {
Handle()
: formatContext(0)
, codecContext(0)
, codec(0)
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(53, 35, 0)
, frame(0)
, gotFrame(0)
, packetLeft(false)
, flushing(false)
#endif
, audioStream(0)
, currentBytes(0) {
packet = av_packet_alloc();
origPacket = NULL;
audioBuffer = (int16_t*)av_malloc(BUFFER_SIZE);
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(53, 35, 0)
#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(55, 39, 101)
frame = avcodec_alloc_frame();
#else
frame = av_frame_alloc();
#endif
#endif
}
~Handle() {
if(origPacket != NULL) {
av_packet_free(&origPacket);
}
if(packet != NULL) {
av_packet_free(&packet);
}
if (audioBuffer) {
av_free(audioBuffer);
}
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(53, 35, 0)
if (frame) {
#if LIBAVCODEC_VERSION_INT <= AV_VERSION_INT(54, 23, 100)
av_free(&frame);
#elif LIBAVCODEC_VERSION_INT < AV_VERSION_INT(55, 39, 101)
avcodec_free_frame(&frame);
#else
av_frame_free(&frame);
#endif
}
#endif
}
AVFormatContext *formatContext;
AVCodecContext *codecContext;
const AVCodec *codec;
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(53, 35, 0)
AVFrame *frame;
int gotFrame;
bool packetLeft;
bool flushing;
AVPacket* origPacket;
#endif
AVPacket* packet;
int audioStream;
int16_t *audioBuffer;
float buffer[BUFFER_SIZE / 2 + 1];
QList<QByteArray> bufferList;
size_t currentBytes;
};
void FfmpegInput::init()
{
static int i=false;
if (!i) {
#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(58, 18, 100)
av_register_all();
#endif
av_log_set_level(AV_LOG_ERROR);
i=true;
}
}
FfmpegInput::FfmpegInput(const QString &fileName)
{
mutex.lock();
handle=new Handle;
bool ok=true;
#if LIBAVFORMAT_VERSION_MAJOR >= 54 || \
(LIBAVFORMAT_VERSION_MAJOR == 53 && LIBAVFORMAT_VERSION_INT >= AV_VERSION_INT(53, 2, 0)) || \
(LIBAVFORMAT_VERSION_MAJOR == 52 && LIBAVFORMAT_VERSION_INT >= AV_VERSION_INT(52, 110, 0))
if (avformat_open_input(&handle->formatContext, QFile::encodeName(fileName).constData(), NULL, NULL) != 0)
#else
if (av_open_input_file(&handle->formatContext, QFile::encodeName(fileName).constData(), NULL, 0, NULL) != 0)
#endif
{
mutex.unlock();
delete handle;
handle=0;
return;
}
#if LIBAVFORMAT_VERSION_INT >= AV_VERSION_INT(53, 21, 0)
if (ok && avformat_find_stream_info(handle->formatContext, 0) < 0) {
ok=false;
}
#else
if (ok && av_find_stream_info(handle->formatContext) < 0) {
ok=false;
}
#endif
// Find the first audio stream
if (ok) {
handle->audioStream = -1;
for (size_t j = 0; j < handle->formatContext->nb_streams; ++j) {
if (GET_CODEC_TYPE(handle->formatContext->streams[j]) ==
#if LIBAVCODEC_VERSION_MAJOR >= 53 || \
(LIBAVCODEC_VERSION_MAJOR == 52 && LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(52, 64, 0))
AVMEDIA_TYPE_AUDIO
#else
CODEC_TYPE_AUDIO
#endif
) {
handle->audioStream = (int) j;
break;
}
}
if (-1==handle->audioStream) {
ok=false;
}
}
if (ok) {
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(57, 89, 100) // Not 100% of version here!
handle->codec=avcodec_find_decoder(handle->formatContext->streams[handle->audioStream]->codecpar->codec_id);
handle->codecContext=avcodec_alloc_context3(handle->codec);
if (avcodec_parameters_to_context(handle->codecContext, handle->formatContext->streams[handle->audioStream]->codecpar)<0) {
ok = false;
}
#else
// Get a pointer to the codec context for the audio stream
handle->codecContext = handle->formatContext->streams[handle->audioStream]->codec;
// request float output if supported
#if LIBAVCODEC_VERSION_MAJOR >= 54
handle->codecContext->request_sample_fmt = AV_SAMPLE_FMT_FLT;
#elif(LIBAVCODEC_VERSION_MAJOR == 53 && LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(53, 4, 0))
handle->codecContext->request_sample_fmt = SAMPLE_FMT_FLT;
#endif
// Find the decoder for the video stream, and open codec...
handle->codec = avcodec_find_decoder(handle->codecContext->codec_id);
#endif
if (ok) {
QString floatCodec=QLatin1String(handle->codec->name)+QLatin1String("float");
const AVCodec *possibleFloatCodec = avcodec_find_decoder_by_name(floatCodec.toLatin1().constData());
if (possibleFloatCodec) {
handle->codec = possibleFloatCodec;
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(57, 89, 100) // Not 100% of version here!
avcodec_free_context(&handle->codecContext);
handle->codecContext=avcodec_alloc_context3(handle->codec);
if (avcodec_parameters_to_context(handle->codecContext, handle->formatContext->streams[handle->audioStream]->codecpar)<0) {
ok = false;
}
#endif
}
if (!handle->codec ||
#if LIBAVCODEC_VERSION_MAJOR >= 53
avcodec_open2(handle->codecContext, handle->codec, NULL) < 0)
#else
avcodec_open(handle->codecContext, handle->codec) < 0)
#endif
{
ok=false;
}
}
}
if (ok) {
mutex.unlock();
} else {
#if LIBAVFORMAT_VERSION_INT >= AV_VERSION_INT(53, 21, 0)
avformat_close_input(&handle->formatContext);
#else
av_close_input_file(handle->formatContext);
#endif
mutex.unlock();
delete handle;
handle=0;
}
}
FfmpegInput::~FfmpegInput()
{
if (handle) {
mutex.lock();
avcodec_close(handle->codecContext);
#if LIBAVFORMAT_VERSION_INT >= AV_VERSION_INT(53, 21, 0)
avformat_close_input(&handle->formatContext);
#else
av_close_input_file(handle->formatContext);
#endif
mutex.unlock();
delete handle;
handle=0;
}
}
size_t FfmpegInput::totalFrames() const
{
if (!handle) {
return 0;
}
double tmp = (double) handle->formatContext->streams[handle->audioStream]->duration
* (double) handle->formatContext->streams[handle->audioStream]->time_base.num
/ (double) handle->formatContext->streams[handle->audioStream]->time_base.den
* (double) handle->codecContext->sample_rate;
return tmp<=0.0 ? 0 : (size_t) (tmp + 0.5);
}
unsigned int FfmpegInput::channels() const
{
return handle ? handle->codecContext->channels : 0;
}
unsigned long FfmpegInput::sampleRate() const
{
return handle ? handle->codecContext->sample_rate : 0;
}
float * FfmpegInput::buffer() const
{
return handle ? handle->buffer : 0;
}
bool FfmpegInput::setChannelMap(int *st) const
{
if (handle && handle->codecContext->channel_layout) {
unsigned int mapIndex = 0;
int bitCounter = 0;
while (mapIndex < (unsigned) handle->codecContext->channels) {
if (handle->codecContext->channel_layout & (1 << bitCounter)) {
switch (1 << bitCounter) {
#if LIBAVFORMAT_VERSION_MAJOR >= 54
case AV_CH_FRONT_LEFT:
#else
case CH_FRONT_LEFT:
#endif
st[mapIndex] = EBUR128_LEFT;
break;
#if LIBAVFORMAT_VERSION_MAJOR >= 54
case AV_CH_FRONT_RIGHT:
#else
case CH_FRONT_RIGHT:
#endif
st[mapIndex] = EBUR128_RIGHT;
break;
#if LIBAVFORMAT_VERSION_MAJOR >= 54
case AV_CH_FRONT_CENTER:
#else
case CH_FRONT_CENTER:
#endif
st[mapIndex] = EBUR128_CENTER;
break;
#if LIBAVFORMAT_VERSION_MAJOR >= 54
case AV_CH_BACK_LEFT:
#else
case CH_BACK_LEFT:
#endif
st[mapIndex] = EBUR128_LEFT_SURROUND;
break;
#if LIBAVFORMAT_VERSION_MAJOR >= 54
case AV_CH_BACK_RIGHT:
#else
case CH_BACK_RIGHT:
#endif
st[mapIndex] = EBUR128_RIGHT_SURROUND;
break;
default:
st[mapIndex] = EBUR128_UNUSED;
break;
}
++mapIndex;
}
++bitCounter;
}
return true;
}
return false;
}
size_t FfmpegInput::readFrames()
{
if (!handle || !channels()) {
return 0;
}
size_t bufferPosition=0, numberRead=0;
while (handle->currentBytes < BUFFER_SIZE) {
numberRead = readOnePacket();
if (!numberRead) {
break;
}
size_t bufferSize=numberRead * channels() * sizeof(float);
handle->bufferList.append(QByteArray((const char *)(handle->buffer), bufferSize));
handle->currentBytes += bufferSize;
}
while (handle->bufferList.count() && handle->bufferList.first().size() + bufferPosition <= BUFFER_SIZE) {
QByteArray b=handle->bufferList.takeAt(0);
memcpy((char *) handle->buffer + bufferPosition, b.constData(), b.size());
bufferPosition += b.size();
handle->currentBytes -= b.size();
}
return bufferPosition / sizeof(float) / channels();
}
#if LIBAVCODEC_VERSION_MAJOR >= 54
static int decodePacket(FfmpegInput::Handle *handle)
{
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(57, 89, 100) // Not 100% of version here!
int ret = avcodec_receive_frame(handle->codecContext, handle->frame);
if (0 == ret) {
handle->gotFrame = true;
}
if (AVERROR(EAGAIN) == ret) {
ret = 0;
}
if (0 == ret) {
ret = avcodec_send_packet(handle->codecContext, handle->packet);
}
if (AVERROR(EAGAIN) == ret) {
ret = 0;
}
return ret<0 ? ret : handle->packet->size;
#else
int ret = avcodec_decode_audio4(handle->codecContext, handle->frame, &handle->gotFrame, handle->packet);
if (ret < 0) {
return ret;
}
return FFMIN(ret, handle->packet->size);
#endif
}
size_t FfmpegInput::readOnePacket()
{
if (!handle) {
return 0;
}
start:
if (handle->flushing) {
handle->packet->data = NULL;
handle->packet->size = 0;
decodePacket(handle);
if (!handle->gotFrame) {
return 0;
} else {
goto write_to_buffer;
}
}
if (handle->packetLeft) {
goto packetLeft;
}
//next_frame:
for (;;) {
if (av_read_frame(handle->formatContext, handle->packet) < 0) {
handle->flushing = true;
goto start;
}
if (handle->packet->stream_index != handle->audioStream) {
AV_FREE(handle->packet);
continue;
} else {
break;
}
}
int ret;
if(handle->origPacket != NULL) {
AV_FREE(handle->origPacket);
}
handle->origPacket = av_packet_clone(handle->packet);
packetLeft:
ret = decodePacket(handle);
if (ret < 0) {
goto free_packet;
}
handle->packet->data += ret;
handle->packet->size -= ret;
if (handle->packet->size > 0) {
handle->packetLeft = true;
} else {
free_packet:
AV_FREE(handle->origPacket);
handle->origPacket = NULL;
handle->packetLeft = false;
}
if (!handle->gotFrame) {
goto start;
}
write_to_buffer: ;
size_t numberRead=handle->frame->nb_samples;
/* TODO: fix this */
int numChannels = handle->codecContext->channels;
// channels = handle->frame->channels;
if (handle->frame->nb_samples * numChannels > (int)sizeof handle->buffer) {
return 0;
}
switch (handle->codecContext->sample_fmt) {
case AV_SAMPLE_FMT_S16: {
int16_t *dataShort = (int16_t*)handle->frame->extended_data[0];
for (int i = 0; i < handle->frame->nb_samples * numChannels; ++i) {
handle->buffer[i] = ((float) dataShort[i])/qMax(-(float) SHRT_MIN, (float) SHRT_MAX);
}
break;
}
case AV_SAMPLE_FMT_S32: {
int32_t *dataInt = (int32_t*)handle->frame->extended_data[0];
for (int i = 0; i < handle->frame->nb_samples * numChannels; ++i) {
handle->buffer[i] = ((float) dataInt[i])/qMax(-(float) INT_MIN, (float) INT_MAX);
}
break;
}
case AV_SAMPLE_FMT_FLT: {
float *dataFloat = (float*)handle->frame->extended_data[0];
for (int i = 0; i < handle->frame->nb_samples * numChannels; ++i) {
handle->buffer[i] = dataFloat[i];
}
break;
}
case AV_SAMPLE_FMT_DBL: {
double *dataDouble = (double*)handle->frame->extended_data[0];
for (int i = 0; i < handle->frame->nb_samples * numChannels; ++i) {
handle->buffer[i] = (float)dataDouble[i];
}
break;
}
case AV_SAMPLE_FMT_S16P: {
uint8_t **ed = handle->frame->extended_data;
for (int i = 0; i < handle->frame->nb_samples * numChannels; ++i) {
int currentChannel = i / handle->frame->nb_samples;
int currentSample = i % handle->frame->nb_samples;
handle->buffer[currentSample * numChannels + currentChannel] = ((float)((int16_t*) ed[currentChannel])[currentSample])/qMax(-(float)SHRT_MIN, (float)SHRT_MAX);
}
break;
}
case AV_SAMPLE_FMT_S32P: {
uint8_t **ed = handle->frame->extended_data;
for (int i = 0; i < handle->frame->nb_samples * numChannels; ++i) {
int currentChannel = i / handle->frame->nb_samples;
int currentSample = i % handle->frame->nb_samples;
handle->buffer[currentSample * numChannels + currentChannel] = ((float) ((int32_t*) ed[currentChannel])[currentSample])/qMax(-(float)INT_MIN, (float)INT_MAX);
}
break;
}
case AV_SAMPLE_FMT_FLTP: {
uint8_t **ed = handle->frame->extended_data;
for (int i = 0; i < handle->frame->nb_samples * numChannels; ++i) {
int currentChannel = i / handle->frame->nb_samples;
int currentSample = i % handle->frame->nb_samples;
handle->buffer[currentSample * numChannels + currentChannel] = ((float*) ed[currentChannel])[currentSample];
}
break;
}
case AV_SAMPLE_FMT_DBLP:{
uint8_t **ed = handle->frame->extended_data;
for (int i = 0; i < handle->frame->nb_samples * numChannels; ++i) {
int currentChannel = i / handle->frame->nb_samples;
int currentSample = i % handle->frame->nb_samples;
handle->buffer[currentSample * numChannels + currentChannel] = ((double*) ed[currentChannel])[currentSample];
}
break;
}
case AV_SAMPLE_FMT_U8:
case AV_SAMPLE_FMT_NONE:
case AV_SAMPLE_FMT_NB:
default:
numberRead=0;
goto out;
}
out:
return numberRead;
}
#else
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(53, 35, 0)
static int decodeAudio(FfmpegInput::Handle *handle, int *frame_size_ptr)
{
int ret, got_frame = 0;
if (!handle->frame) {
return AVERROR(ENOMEM);
}
#if LIBAVFORMAT_VERSION_INT < AV_VERSION_INT(55, 0, 0)
if (handle->codecContext->get_buffer != avcodec_default_get_buffer) {
handle->codecContext->get_buffer = avcodec_default_get_buffer;
handle->codecContext->release_buffer = avcodec_default_release_buffer;
}
#endif
ret = avcodec_decode_audio4(handle->codecContext, handle->frame, &got_frame, handle->packet);
if (ret >= 0 && got_frame) {
int ch, plane_size;
int planar = av_sample_fmt_is_planar(handle->codecContext->sample_fmt);
int data_size = av_samples_get_buffer_size(&plane_size, handle->codecContext->channels,
handle->frame->nb_samples, handle->codecContext->sample_fmt, 1);
if (*frame_size_ptr < data_size) {
return AVERROR(EINVAL);
}
memcpy(handle->audioBuffer, handle->frame->extended_data[0], plane_size);
if (planar && handle->codecContext->channels > 1) {
uint8_t *out = ((uint8_t *)(handle->audioBuffer)) + plane_size;
for (ch = 1; ch < handle->codecContext->channels; ch++) {
memcpy(out, handle->frame->extended_data[ch], plane_size);
out += plane_size;
}
}
*frame_size_ptr = data_size;
} else {
*frame_size_ptr = 0;
}
return ret;
}
#endif
size_t FfmpegInput::readOnePacket()
{
if (!handle) {
return 0;
}
next_frame:
for (;;) {
if (av_read_frame(handle->formatContext, handle->packet) < 0) {
return 0;
}
if (handle->packet->stream_index == handle->audioStream) {
break;
}
AV_FREE(handle->packet);
}
size_t numberRead=0;
int dataSize=BUFFER_SIZE;
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(53, 35, 0)
int len = decodeAudio(handle, &dataSize);
#elif LIBAVCODEC_VERSION_MAJOR >= 53 || \
(LIBAVCODEC_VERSION_MAJOR == 52 && LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(52, 23, 0))
int len = avcodec_decode_audio3(handle->codecContext, handle->audioBuffer, &dataSize, handle->packet);
#else
int len = avcodec_decode_audio2(handle->codecContext, handle->audioBuffer, &dataSize, handle->packet->data, handle->packet->size);
#endif
if (len < 0) {
goto out;
}
/* No data used, (happens with metadata frames for example) */
if (len <= 0) {
AV_FREE(handle->packet);
goto next_frame;
}
switch (handle->codecContext->sample_fmt) {
case SAMPLE_FMT_S16: {
int16_t *dataShort = (int16_t*) handle->audioBuffer;
numberRead = (size_t) dataSize / sizeof(int16_t) / (size_t) handle->codecContext->channels;
for (unsigned int i = 0; i < (size_t) dataSize / sizeof(int16_t); ++i) {
handle->buffer[i] = ((float) dataShort[i]) / qMax(-(float) SHRT_MIN, (float) SHRT_MAX);
}
break;
}
case SAMPLE_FMT_S32: {
int32_t *dataInt = (int32_t*) handle->audioBuffer;
numberRead = (size_t) dataSize / sizeof(int32_t) / (size_t) handle->codecContext->channels;
for (unsigned int i = 0; i < (size_t) dataSize / sizeof(int32_t); ++i) {
handle->buffer[i] = ((float) dataInt[i]) / qMax(-(float) INT_MIN, (float) INT_MAX);
}
break;
}
case SAMPLE_FMT_FLT: {
float *dataFloat = (float*) handle->audioBuffer;
numberRead = (size_t) dataSize / sizeof(float) / (size_t) handle->codecContext->channels;
for (unsigned int i = 0; i < (size_t) dataSize / sizeof(float); ++i) {
handle->buffer[i] = dataFloat[i];
}
break;
}
case SAMPLE_FMT_DBL: {
double *dataDouble = (double*) handle->audioBuffer;
numberRead = (size_t) dataSize / sizeof(double) / (size_t) handle->codecContext->channels;
for (unsigned int i = 0; i < (size_t) dataSize / sizeof(double); ++i) {
handle->buffer[i] = (float) dataDouble[i];
}
break;
}
case SAMPLE_FMT_U8:
case SAMPLE_FMT_NONE:
case SAMPLE_FMT_NB:
default:
numberRead=0;
goto out;
}
out:
AV_FREE(handle->packet);
return numberRead;
}
#endif
bool FfmpegInput::isFloatCodec() const
{
return handle && handle->codec && QString(QLatin1String(handle->codec->name)).endsWith(QLatin1String("float"));
}