Files
palemoon27/dom/media/platforms/ffmpeg/FFmpegAudioDecoder.cpp
T
roytam1 0b4ed75d7e import changes from `dev' branch of rmottola/Arctic-Fox:
- formatting (49c876691)
- adapted  Bug 1163227: Part5: Increase verbosity content for debugging purposes. r=cpearce (cfca0b7bf)
- Bug 1184634 - Rename MediaPromise to MozPromise. r=gerald (e9617bf70)
- space (c5bafde74)
- Bug 1158293 - Fix ResumeStateMachine()'s seek handling r=cpearce (67dbe7dfc)
- reapply part of Bug 1158448 Part 2 (a953e6c6b)
- Bug 1184634 - Rename "TaskQueue()" accessor to "OwnerThread()". r=gerald (118965bae)
- Bug 1145411: Reduce Apple audio decoder verbosity. r=mattwoodrow (a9f2c4092)
- Bug 1174055 - Remove some now uncalled WMF functions. r=jya (8cddc9d0e)
- Bug 1179094: Use TimeUnit in PlatformDecoderModule. r=cpearce (3d8fd9cd7)
- align warnings to Bug 1154133 - Remove sync dispatches in EMEDecryptor, and mark MediaTaskQueue::SyncDispatch as deprecated. (d93e2f431)
- align some permissions and comments (54eb9e910)
- Bug 1155268 - Add checked methods to convert from an AbstractThread to a concrete type. r=cpearce (5deff5f04)
- align comment (c508b1aaf)
- Bug 857895 - Only enforce tail dispatch if both source and dest support it. r=bholley (2e6d03096)
- Bug 1184634 - Rename MediaTaskQueue to TaskQueue. r=gerald (5f18c2616)
- Bug 1127656 - Remove class inheritance hierarchy from gonk PlatformDeocdeModule. r=ajones (eebf0eb20)
- Bug 1184634 - Move various includes into the mozilla namespace. r=gerald (a9b516578)
- Bug 1190636 - Replace AutoStringVector with Rooted usage; r=njn (5d287ee81)
- Bug 1190911 - Replace AutoIdValueVector with normal Rooted usage; r=jonco (b6c8ce668)
- Bug 1149343 - Followup to fix bustage on a CLOSED TREE (f20f9f1a2)
- Bug 1172841. Set number of WMF decoder threads based on CPU count. r=cpearce (af74b7654)
- Bug 1166760 - Query audio sampling rate & channel number from codec. r=sotaro (ee5fb569a)
-  Bug 1175071 - Remove those codes which handles ADTS. r=jya. (00b1d3dad)
- Bug 1178682 - Reject the decoded sample if timestamp is revert. r=sotaro (4cd029e8c)
2021-07-27 16:39:15 +08:00

214 lines
6.7 KiB
C++

/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim:set ts=2 sw=2 sts=2 et cindent: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "mozilla/TaskQueue.h"
#include "FFmpegRuntimeLinker.h"
#include "FFmpegAudioDecoder.h"
#include "TimeUnits.h"
#define MAX_CHANNELS 16
namespace mozilla
{
static int (*avcodec_decode_audio4)(AVCodecContext*,AVFrame*,
int*,const AVPacket*) = nullptr;
static void (*av_init_packet1)(AVPacket*) = nullptr;
FFmpegAudioDecoder<LIBAV_VER>::FFmpegAudioDecoder(
FlushableTaskQueue* aTaskQueue, MediaDataDecoderCallback* aCallback,
const AudioInfo& aConfig)
: FFmpegDataDecoder(aTaskQueue, GetCodecId(aConfig.mMimeType))
, mCallback(aCallback)
{
MOZ_COUNT_CTOR(FFmpegAudioDecoder);
// Use a new MediaByteBuffer as the object will be modified during initialization.
mExtraData = new MediaByteBuffer;
mExtraData->AppendElements(*aConfig.mCodecSpecificConfig);
}
nsresult
FFmpegAudioDecoder<LIBAV_VER>::Init()
{
nsresult rv = FFmpegDataDecoder::Init();
NS_ENSURE_SUCCESS(rv, rv);
avcodec_decode_audio4 = (decltype(avcodec_decode_audio4))FFmpegRuntimeLinker::avc_ptr[_decode_audio4];
av_init_packet1 = (decltype(av_init_packet1))FFmpegRuntimeLinker::avc_ptr[_init_packet];
return NS_OK;
}
void
FFmpegAudioDecoder<LIBAV_VER>::InitCodecContext()
{
MOZ_ASSERT(mCodecContext);
// We do not want to set this value to 0 as FFmpeg by default will
// use the number of cores, which with our mozlibavutil get_cpu_count
// isn't implemented.
mCodecContext->thread_count = 1;
// FFmpeg takes this as a suggestion for what format to use for audio samples.
uint32_t major, minor, micro;
FFmpegRuntimeLinker::GetVersion(major, minor, micro);
// LibAV 0.8 produces rubbish float interleaved samples, request 16 bits audio.
mCodecContext->request_sample_fmt =
(major == 53) ? AV_SAMPLE_FMT_S16 : AV_SAMPLE_FMT_FLT;
}
static AudioDataValue*
CopyAndPackAudio(AVFrame* aFrame, uint32_t aNumChannels, uint32_t aNumAFrames)
{
MOZ_ASSERT(aNumChannels <= MAX_CHANNELS);
nsAutoArrayPtr<AudioDataValue> audio(
new AudioDataValue[aNumChannels * aNumAFrames]);
if (aFrame->format == AV_SAMPLE_FMT_FLT) {
// Audio data already packed. No need to do anything other than copy it
// into a buffer we own.
memcpy(audio, aFrame->data[0],
aNumChannels * aNumAFrames * sizeof(AudioDataValue));
} else if (aFrame->format == AV_SAMPLE_FMT_FLTP) {
// Planar audio data. Pack it into something we can understand.
AudioDataValue* tmp = audio;
AudioDataValue** data = reinterpret_cast<AudioDataValue**>(aFrame->data);
for (uint32_t frame = 0; frame < aNumAFrames; frame++) {
for (uint32_t channel = 0; channel < aNumChannels; channel++) {
*tmp++ = data[channel][frame];
}
}
} else if (aFrame->format == AV_SAMPLE_FMT_S16) {
// Audio data already packed. Need to convert from S16 to 32 bits Float
AudioDataValue* tmp = audio;
int16_t* data = reinterpret_cast<int16_t**>(aFrame->data)[0];
for (uint32_t frame = 0; frame < aNumAFrames; frame++) {
for (uint32_t channel = 0; channel < aNumChannels; channel++) {
*tmp++ = AudioSampleToFloat(*data++);
}
}
} else if (aFrame->format == AV_SAMPLE_FMT_S16P) {
// Planar audio data. Convert it from S16 to 32 bits float
// and pack it into something we can understand.
AudioDataValue* tmp = audio;
int16_t** data = reinterpret_cast<int16_t**>(aFrame->data);
for (uint32_t frame = 0; frame < aNumAFrames; frame++) {
for (uint32_t channel = 0; channel < aNumChannels; channel++) {
*tmp++ = AudioSampleToFloat(data[channel][frame]);
}
}
}
return audio.forget();
}
void
FFmpegAudioDecoder<LIBAV_VER>::DecodePacket(MediaRawData* aSample)
{
AVPacket packet;
av_init_packet1(&packet);
packet.data = const_cast<uint8_t*>(aSample->Data());
packet.size = aSample->Size();
if (!PrepareFrame()) {
NS_WARNING("FFmpeg audio decoder failed to allocate frame.");
mCallback->Error();
return;
}
int64_t samplePosition = aSample->mOffset;
media::TimeUnit pts = media::TimeUnit::FromMicroseconds(aSample->mTime);
while (packet.size > 0) {
int decoded;
int bytesConsumed =
avcodec_decode_audio4(mCodecContext, mFrame, &decoded, &packet);
if (bytesConsumed < 0) {
NS_WARNING("FFmpeg audio decoder error.");
mCallback->Error();
return;
}
if (decoded) {
uint32_t numChannels = mCodecContext->channels;
uint32_t samplingRate = mCodecContext->sample_rate;
nsAutoArrayPtr<AudioDataValue> audio(
CopyAndPackAudio(mFrame, numChannels, mFrame->nb_samples));
media::TimeUnit duration =
FramesToTimeUnit(mFrame->nb_samples, samplingRate);
if (!duration.IsValid()) {
NS_WARNING("Invalid count of accumulated audio samples");
mCallback->Error();
return;
}
nsRefPtr<AudioData> data = new AudioData(samplePosition,
pts.ToMicroseconds(),
duration.ToMicroseconds(),
mFrame->nb_samples,
audio.forget(),
numChannels,
samplingRate);
mCallback->Output(data);
pts += duration;
if (!pts.IsValid()) {
NS_WARNING("Invalid count of accumulated audio samples");
mCallback->Error();
return;
}
}
packet.data += bytesConsumed;
packet.size -= bytesConsumed;
samplePosition += bytesConsumed;
}
if (mTaskQueue->IsEmpty()) {
mCallback->InputExhausted();
}
}
nsresult
FFmpegAudioDecoder<LIBAV_VER>::Input(MediaRawData* aSample)
{
nsCOMPtr<nsIRunnable> runnable(NS_NewRunnableMethodWithArg<nsRefPtr<MediaRawData>>(
this, &FFmpegAudioDecoder::DecodePacket, nsRefPtr<MediaRawData>(aSample)));
mTaskQueue->Dispatch(runnable.forget());
return NS_OK;
}
nsresult
FFmpegAudioDecoder<LIBAV_VER>::Drain()
{
mTaskQueue->AwaitIdle();
mCallback->DrainComplete();
return Flush();
}
AVCodecID
FFmpegAudioDecoder<LIBAV_VER>::GetCodecId(const nsACString& aMimeType)
{
if (aMimeType.EqualsLiteral("audio/mpeg")) {
return AV_CODEC_ID_MP3;
}
if (aMimeType.EqualsLiteral("audio/mp4a-latm")) {
return AV_CODEC_ID_AAC;
}
return AV_CODEC_ID_NONE;
}
FFmpegAudioDecoder<LIBAV_VER>::~FFmpegAudioDecoder()
{
MOZ_COUNT_DTOR(FFmpegAudioDecoder);
}
} // namespace mozilla