summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--DEPS4
-rw-r--r--media/ffmpeg/ffmpeg_common.cc5
-rw-r--r--media/filters/audio_renderer_impl.cc18
-rw-r--r--media/filters/ffmpeg_demuxer.cc13
-rw-r--r--media/filters/opus_audio_decoder.cc195
-rw-r--r--media/filters/opus_audio_decoder.h11
-rw-r--r--media/media.gyp4
7 files changed, 141 insertions, 109 deletions
diff --git a/DEPS b/DEPS
index 869c0f1..4c50d1b 100644
--- a/DEPS
+++ b/DEPS
@@ -36,8 +36,8 @@ vars = {
# These two FFmpeg variables must be updated together. One is used for SVN
# checkouts and the other for Git checkouts.
- "ffmpeg_revision": "240211",
- "ffmpeg_hash": "8dc45cbcad763762bb679de280bdc584f35aa22f",
+ "ffmpeg_revision": "240434",
+ "ffmpeg_hash": "c6c988923be6b8fd1f381522d478813e14505ce2",
"sfntly_revision": "228",
"lighttpd_revision": "33737",
diff --git a/media/ffmpeg/ffmpeg_common.cc b/media/ffmpeg/ffmpeg_common.cc
index 40696c6..6e7bd15 100644
--- a/media/ffmpeg/ffmpeg_common.cc
+++ b/media/ffmpeg/ffmpeg_common.cc
@@ -286,8 +286,9 @@ static void AVCodecContextToAudioDecoderConfig(
if (codec == kCodecOpus) {
// |codec_context->sample_fmt| is not set by FFmpeg because Opus decoding is
- // not enabled in FFmpeg, so we need to manually set the sample format.
- sample_format = kSampleFormatS16;
+ // not enabled in FFmpeg. It doesn't matter what value is set here, so long
+ // as it's valid, the true sample format is selected inside the decoder.
+ sample_format = kSampleFormatF32;
}
base::TimeDelta seek_preroll;
diff --git a/media/filters/audio_renderer_impl.cc b/media/filters/audio_renderer_impl.cc
index fdfb45d..fc5b7bf 100644
--- a/media/filters/audio_renderer_impl.cc
+++ b/media/filters/audio_renderer_impl.cc
@@ -399,8 +399,22 @@ bool AudioRendererImpl::HandleSplicerBuffer(
if (state_ == kUnderflow || state_ == kRebuffering)
ChangeState_Locked(kPlaying);
} else {
- if (state_ == kPrerolling && IsBeforePrerollTime(buffer))
- return true;
+ if (state_ == kPrerolling) {
+ if (IsBeforePrerollTime(buffer))
+ return true;
+
+ // Trim off any additional time before the preroll timestamp.
+ const base::TimeDelta trim_time =
+ preroll_timestamp_ - buffer->timestamp();
+ if (trim_time > base::TimeDelta()) {
+ buffer->TrimStart(buffer->frame_count() *
+ (static_cast<double>(trim_time.InMicroseconds()) /
+ buffer->duration().InMicroseconds()));
+ }
+ // If the entire buffer was trimmed, request a new one.
+ if (!buffer->frame_count())
+ return true;
+ }
if (state_ != kUninitialized && state_ != kStopped)
algorithm_->EnqueueBuffer(buffer);
diff --git a/media/filters/ffmpeg_demuxer.cc b/media/filters/ffmpeg_demuxer.cc
index e47c59e..6b80271 100644
--- a/media/filters/ffmpeg_demuxer.cc
+++ b/media/filters/ffmpeg_demuxer.cc
@@ -845,6 +845,19 @@ void FFmpegDemuxer::OnReadFrameDone(ScopedAVPacket packet, int result) {
packet.swap(new_packet);
}
+ // Special case for opus in ogg. FFmpeg is pre-trimming the codec delay
+ // from the packet timestamp. Chrome expects to handle this itself inside
+ // the decoder, so shift timestamps by the delay in this case.
+ // TODO(dalecurtis): Try to get fixed upstream. See http://crbug.com/328207
+ if (strcmp(glue_->format_context()->iformat->name, "ogg") == 0) {
+ const AVCodecContext* codec_context =
+ glue_->format_context()->streams[packet->stream_index]->codec;
+ if (codec_context->codec_id == AV_CODEC_ID_OPUS &&
+ codec_context->delay > 0) {
+ packet->pts += codec_context->delay;
+ }
+ }
+
FFmpegDemuxerStream* demuxer_stream = streams_[packet->stream_index];
demuxer_stream->EnqueuePacket(packet.Pass());
}
diff --git a/media/filters/opus_audio_decoder.cc b/media/filters/opus_audio_decoder.cc
index 37e1abd..e356720 100644
--- a/media/filters/opus_audio_decoder.cc
+++ b/media/filters/opus_audio_decoder.cc
@@ -25,7 +25,6 @@
namespace media {
static uint16 ReadLE16(const uint8* data, size_t data_size, int read_offset) {
- DCHECK(data);
uint16 value = 0;
DCHECK_LE(read_offset + sizeof(value), data_size);
memcpy(&value, data + read_offset, sizeof(value));
@@ -46,15 +45,8 @@ static int TimeDeltaToAudioFrames(base::TimeDelta time_delta,
// http://www.xiph.org/vorbis/doc/Vorbis_I_spec.html
static const int kMaxVorbisChannels = 8;
-// Opus allows for decode of S16 or float samples. OpusAudioDecoder always uses
-// S16 samples.
-static const int kBitsPerChannel = 16;
-static const int kBytesPerChannel = kBitsPerChannel / 8;
-
// Maximum packet size used in Xiph's opusdec and FFmpeg's libopusdec.
-static const int kMaxOpusOutputPacketSizeSamples = 960 * 6 * kMaxVorbisChannels;
-static const int kMaxOpusOutputPacketSizeBytes =
- kMaxOpusOutputPacketSizeSamples * kBytesPerChannel;
+static const int kMaxOpusOutputPacketSizeSamples = 960 * 6;
static void RemapOpusChannelLayout(const uint8* opus_mapping,
int num_channels,
@@ -206,13 +198,16 @@ struct OpusExtraData {
static bool ParseOpusExtraData(const uint8* data, int data_size,
const AudioDecoderConfig& config,
OpusExtraData* extra_data) {
- if (data_size < kOpusExtraDataSize)
+ if (data_size < kOpusExtraDataSize) {
+ DLOG(ERROR) << "Extra data size is too small:" << data_size;
return false;
+ }
extra_data->channels = *(data + kOpusExtraDataChannelsOffset);
if (extra_data->channels <= 0 || extra_data->channels > kMaxVorbisChannels) {
- DVLOG(0) << "invalid channel count in extra data: " << extra_data->channels;
+ DLOG(ERROR) << "invalid channel count in extra data: "
+ << extra_data->channels;
return false;
}
@@ -223,7 +218,7 @@ static bool ParseOpusExtraData(const uint8* data, int data_size,
if (!extra_data->channel_mapping) {
if (extra_data->channels > kMaxChannelsWithDefaultLayout) {
- DVLOG(0) << "Invalid extra data, missing stream map.";
+ DLOG(ERROR) << "Invalid extra data, missing stream map.";
return false;
}
@@ -234,8 +229,8 @@ static bool ParseOpusExtraData(const uint8* data, int data_size,
}
if (data_size < kOpusExtraDataStreamMapOffset + extra_data->channels) {
- DVLOG(0) << "Invalid stream map; insufficient data for current channel "
- << "count: " << extra_data->channels;
+ DLOG(ERROR) << "Invalid stream map; insufficient data for current channel "
+ << "count: " << extra_data->channels;
return false;
}
@@ -256,12 +251,14 @@ OpusAudioDecoder::OpusAudioDecoder(
weak_factory_(this),
demuxer_stream_(NULL),
opus_decoder_(NULL),
- bits_per_channel_(0),
channel_layout_(CHANNEL_LAYOUT_NONE),
samples_per_second_(0),
+ sample_format_(kSampleFormatF32),
+ bits_per_channel_(SampleFormatToBytesPerChannel(sample_format_) * 8),
last_input_timestamp_(kNoTimestamp()),
frames_to_discard_(0),
- frame_delay_at_start_(0) {
+ frame_delay_at_start_(0),
+ start_input_timestamp_(kNoTimestamp()) {
}
void OpusAudioDecoder::Initialize(
@@ -274,7 +271,7 @@ void OpusAudioDecoder::Initialize(
if (demuxer_stream_) {
// TODO(scherkus): initialization currently happens more than once in
// PipelineIntegrationTest.BasicPlayback.
- DVLOG(0) << "Initialize has already been called.";
+ DLOG(ERROR) << "Initialize has already been called.";
CHECK(false);
}
@@ -375,7 +372,7 @@ void OpusAudioDecoder::BufferReady(
// occurs with some damaged files.
if (input->timestamp() == kNoTimestamp() &&
output_timestamp_helper_->base_timestamp() == kNoTimestamp()) {
- DVLOG(1) << "Received a buffer without timestamps!";
+ DLOG(ERROR) << "Received a buffer without timestamps!";
base::ResetAndReturn(&read_cb_).Run(kDecodeError, NULL);
return;
}
@@ -384,13 +381,21 @@ void OpusAudioDecoder::BufferReady(
input->timestamp() != kNoTimestamp() &&
input->timestamp() < last_input_timestamp_) {
base::TimeDelta diff = input->timestamp() - last_input_timestamp_;
- DVLOG(1) << "Input timestamps are not monotonically increasing! "
- << " ts " << input->timestamp().InMicroseconds() << " us"
- << " diff " << diff.InMicroseconds() << " us";
+ DLOG(ERROR) << "Input timestamps are not monotonically increasing! "
+ << " ts " << input->timestamp().InMicroseconds() << " us"
+ << " diff " << diff.InMicroseconds() << " us";
base::ResetAndReturn(&read_cb_).Run(kDecodeError, NULL);
return;
}
+ // Apply the necessary codec delay.
+ if (start_input_timestamp_ == kNoTimestamp())
+ start_input_timestamp_ = input->timestamp();
+ if (last_input_timestamp_ == kNoTimestamp() &&
+ input->timestamp() == start_input_timestamp_) {
+ frames_to_discard_ = frame_delay_at_start_;
+ }
+
last_input_timestamp_ = input->timestamp();
scoped_refptr<AudioBuffer> output_buffer;
@@ -414,53 +419,41 @@ bool OpusAudioDecoder::ConfigureDecoder() {
const AudioDecoderConfig& config = demuxer_stream_->audio_decoder_config();
if (config.codec() != kCodecOpus) {
- DVLOG(0) << "codec must be kCodecOpus.";
+ DVLOG(1) << "Codec must be kCodecOpus.";
return false;
}
const int channel_count =
ChannelLayoutToChannelCount(config.channel_layout());
if (!config.IsValidConfig() || channel_count > kMaxVorbisChannels) {
- DVLOG(0) << "Invalid or unsupported audio stream -"
- << " codec: " << config.codec()
- << " channel count: " << channel_count
- << " channel layout: " << config.channel_layout()
- << " bits per channel: " << config.bits_per_channel()
- << " samples per second: " << config.samples_per_second();
- return false;
- }
-
- if (config.bits_per_channel() != kBitsPerChannel) {
- DVLOG(0) << "16 bit samples required.";
+ DLOG(ERROR) << "Invalid or unsupported audio stream -"
+ << " codec: " << config.codec()
+ << " channel count: " << channel_count
+ << " channel layout: " << config.channel_layout()
+ << " bits per channel: " << config.bits_per_channel()
+ << " samples per second: " << config.samples_per_second();
return false;
}
if (config.is_encrypted()) {
- DVLOG(0) << "Encrypted audio stream not supported.";
+ DLOG(ERROR) << "Encrypted audio stream not supported.";
return false;
}
if (opus_decoder_ &&
- (bits_per_channel_ != config.bits_per_channel() ||
- channel_layout_ != config.channel_layout() ||
+ (channel_layout_ != config.channel_layout() ||
samples_per_second_ != config.samples_per_second())) {
- DVLOG(1) << "Unsupported config change :";
- DVLOG(1) << "\tbits_per_channel : " << bits_per_channel_
- << " -> " << config.bits_per_channel();
- DVLOG(1) << "\tchannel_layout : " << channel_layout_
- << " -> " << config.channel_layout();
- DVLOG(1) << "\tsample_rate : " << samples_per_second_
- << " -> " << config.samples_per_second();
+ DLOG(ERROR) << "Unsupported config change -"
+ << ", channel_layout: " << channel_layout_
+ << " -> " << config.channel_layout()
+ << ", sample_rate: " << samples_per_second_
+ << " -> " << config.samples_per_second();
return false;
}
// Clean up existing decoder if necessary.
CloseDecoder();
- // Allocate the output buffer if necessary.
- if (!output_buffer_)
- output_buffer_.reset(new int16[kMaxOpusOutputPacketSizeSamples]);
-
// Parse the Opus Extra Data.
OpusExtraData opus_extra_data;
if (!ParseOpusExtraData(config.extra_data(), config.extra_data_size(),
@@ -468,24 +461,23 @@ bool OpusAudioDecoder::ConfigureDecoder() {
&opus_extra_data))
return false;
- if (!config.codec_delay().InMicroseconds())
- return false;
-
// Convert from seconds to samples.
timestamp_offset_ = config.codec_delay();
frame_delay_at_start_ = TimeDeltaToAudioFrames(config.codec_delay(),
config.samples_per_second());
- if (frame_delay_at_start_ < 0) {
- DVLOG(1) << "Invalid file. Incorrect value for codec delay.";
+ if (timestamp_offset_ <= base::TimeDelta() || frame_delay_at_start_ < 0) {
+ DLOG(ERROR) << "Invalid file. Incorrect value for codec delay: "
+ << config.codec_delay().InMicroseconds();
return false;
}
+
if (frame_delay_at_start_ != opus_extra_data.skip_samples) {
- DVLOG(1) << "Invalid file. Codec Delay in container does not match the "
- << "value in Opus Extra Data.";
+ DLOG(ERROR) << "Invalid file. Codec Delay in container does not match the "
+ << "value in Opus Extra Data.";
return false;
}
- uint8 channel_mapping[kMaxVorbisChannels];
+ uint8 channel_mapping[kMaxVorbisChannels] = {0};
memcpy(&channel_mapping,
kDefaultOpusChannelLayout,
kMaxChannelsWithDefaultLayout);
@@ -505,16 +497,16 @@ bool OpusAudioDecoder::ConfigureDecoder() {
channel_mapping,
&status);
if (!opus_decoder_ || status != OPUS_OK) {
- DVLOG(0) << "opus_multistream_decoder_create failed status="
- << opus_strerror(status);
+ DLOG(ERROR) << "opus_multistream_decoder_create failed status="
+ << opus_strerror(status);
return false;
}
- bits_per_channel_ = config.bits_per_channel();
channel_layout_ = config.channel_layout();
samples_per_second_ = config.samples_per_second();
output_timestamp_helper_.reset(
new AudioTimestampHelper(config.samples_per_second()));
+ start_input_timestamp_ = kNoTimestamp();
return true;
}
@@ -535,68 +527,68 @@ void OpusAudioDecoder::ResetTimestampState() {
bool OpusAudioDecoder::Decode(const scoped_refptr<DecoderBuffer>& input,
scoped_refptr<AudioBuffer>* output_buffer) {
- int frames_decoded = opus_multistream_decode(opus_decoder_,
- input->data(),
- input->data_size(),
- &output_buffer_[0],
- kMaxOpusOutputPacketSizeSamples,
- 0);
+ // Allocate a buffer for the output samples.
+ *output_buffer = AudioBuffer::CreateBuffer(
+ sample_format_,
+ ChannelLayoutToChannelCount(channel_layout_),
+ kMaxOpusOutputPacketSizeSamples);
+ const int buffer_size =
+ output_buffer->get()->channel_count() *
+ output_buffer->get()->frame_count() *
+ SampleFormatToBytesPerChannel(sample_format_);
+
+ float* float_output_buffer = reinterpret_cast<float*>(
+ output_buffer->get()->channel_data()[0]);
+ const int frames_decoded =
+ opus_multistream_decode_float(opus_decoder_,
+ input->data(),
+ input->data_size(),
+ float_output_buffer,
+ buffer_size,
+ 0);
+
if (frames_decoded < 0) {
- DVLOG(0) << "opus_multistream_decode failed for"
- << " timestamp: " << input->timestamp().InMicroseconds()
- << " us, duration: " << input->duration().InMicroseconds()
- << " us, packet size: " << input->data_size() << " bytes with"
- << " status: " << opus_strerror(frames_decoded);
+ DLOG(ERROR) << "opus_multistream_decode failed for"
+ << " timestamp: " << input->timestamp().InMicroseconds()
+ << " us, duration: " << input->duration().InMicroseconds()
+ << " us, packet size: " << input->data_size() << " bytes with"
+ << " status: " << opus_strerror(frames_decoded);
return false;
}
- uint8* decoded_audio_data = reinterpret_cast<uint8*>(&output_buffer_[0]);
- int bytes_decoded = frames_decoded *
- demuxer_stream_->audio_decoder_config().bytes_per_frame();
- DCHECK_LE(bytes_decoded, kMaxOpusOutputPacketSizeBytes);
-
if (output_timestamp_helper_->base_timestamp() == kNoTimestamp() &&
!input->end_of_stream()) {
DCHECK(input->timestamp() != kNoTimestamp());
output_timestamp_helper_->SetBaseTimestamp(input->timestamp());
}
- // Skip samples should be equal to codec delay when the file starts and when
- // there is a seek to zero.
- // TODO(vigneshv): This should be checked for start of stream rather than
- // input timestamp of zero to accomodate streams that don't start at zero.
- if (input->timestamp() == base::TimeDelta())
- frames_to_discard_ = frame_delay_at_start_;
+ // Trim off any extraneous allocation.
+ DCHECK_LE(frames_decoded, output_buffer->get()->frame_count());
+ const int trim_frames = output_buffer->get()->frame_count() - frames_decoded;
+ if (trim_frames > 0)
+ output_buffer->get()->TrimEnd(trim_frames);
- if (bytes_decoded > 0 && frames_decoded > frames_to_discard_) {
- // Copy the audio samples into an output buffer.
- uint8* data[] = { decoded_audio_data };
- *output_buffer = AudioBuffer::CopyFrom(
- kSampleFormatS16,
- ChannelLayoutToChannelCount(channel_layout_),
- frames_decoded,
- data,
- output_timestamp_helper_->GetTimestamp() - timestamp_offset_,
- output_timestamp_helper_->GetFrameDuration(frames_decoded));
- output_timestamp_helper_->AddFrames(frames_decoded);
+ // Handle frame discard and trimming.
+ int frames_to_output = frames_decoded;
+ if (frames_decoded > frames_to_discard_) {
if (frames_to_discard_ > 0) {
output_buffer->get()->TrimStart(frames_to_discard_);
- frames_decoded -= frames_to_discard_;
+ frames_to_output -= frames_to_discard_;
frames_to_discard_ = 0;
}
if (input->discard_padding().InMicroseconds() > 0) {
int discard_padding = TimeDeltaToAudioFrames(input->discard_padding(),
samples_per_second_);
- if (discard_padding < 0 || discard_padding > frames_decoded) {
+ if (discard_padding < 0 || discard_padding > frames_to_output) {
DVLOG(1) << "Invalid file. Incorrect discard padding value.";
return false;
}
output_buffer->get()->TrimEnd(discard_padding);
- frames_decoded -= discard_padding;
+ frames_to_output -= discard_padding;
}
- } else if (bytes_decoded > 0) {
- frames_to_discard_ -= frames_decoded;
- frames_decoded = 0;
+ } else {
+ frames_to_discard_ -= frames_to_output;
+ frames_to_output = 0;
}
// Decoding finished successfully, update statistics.
@@ -604,8 +596,15 @@ bool OpusAudioDecoder::Decode(const scoped_refptr<DecoderBuffer>& input,
statistics.audio_bytes_decoded = input->data_size();
statistics_cb_.Run(statistics);
+ // Assign timestamp and duration to the buffer.
+ output_buffer->get()->set_timestamp(
+ output_timestamp_helper_->GetTimestamp() - timestamp_offset_);
+ output_buffer->get()->set_duration(
+ output_timestamp_helper_->GetFrameDuration(frames_to_output));
+ output_timestamp_helper_->AddFrames(frames_decoded);
+
// Discard the buffer to indicate we need more data.
- if (!frames_decoded)
+ if (!frames_to_output)
*output_buffer = NULL;
return true;
diff --git a/media/filters/opus_audio_decoder.h b/media/filters/opus_audio_decoder.h
index 50ba069..982458b 100644
--- a/media/filters/opus_audio_decoder.h
+++ b/media/filters/opus_audio_decoder.h
@@ -10,6 +10,7 @@
#include "base/time/time.h"
#include "media/base/audio_decoder.h"
#include "media/base/demuxer_stream.h"
+#include "media/base/sample_format.h"
struct OpusMSDecoder;
@@ -62,9 +63,10 @@ class MEDIA_EXPORT OpusAudioDecoder : public AudioDecoder {
OpusMSDecoder* opus_decoder_;
// Decoded audio format.
- int bits_per_channel_;
ChannelLayout channel_layout_;
int samples_per_second_;
+ const SampleFormat sample_format_;
+ const int bits_per_channel_;
// Used for computing output timestamps.
scoped_ptr<AudioTimestampHelper> output_timestamp_helper_;
@@ -80,16 +82,15 @@ class MEDIA_EXPORT OpusAudioDecoder : public AudioDecoder {
int frames_to_discard_;
// Number of frames to be discarded at the start of the stream. This value
- // is typically the CodecDelay value from the container.
+ // is typically the CodecDelay value from the container. This value should
+ // only be applied when input timestamp is |start_input_timestamp_|.
int frame_delay_at_start_;
+ base::TimeDelta start_input_timestamp_;
// Timestamp to be subtracted from all the frames. This is typically computed
// from the CodecDelay value in the container.
base::TimeDelta timestamp_offset_;
- // Buffer for output from libopus.
- scoped_ptr<int16[]> output_buffer_;
-
DISALLOW_IMPLICIT_CONSTRUCTORS(OpusAudioDecoder);
};
diff --git a/media/media.gyp b/media/media.gyp
index fe3b797..97e2f5d 100644
--- a/media/media.gyp
+++ b/media/media.gyp
@@ -560,6 +560,10 @@
'base/media.h',
'base/media_stub.cc',
],
+ 'sources!': [
+ 'filters/opus_audio_decoder.cc',
+ 'filters/opus_audio_decoder.h',
+ ],
'conditions': [
['android_webview_build==0', {
'dependencies': [
generated by cgit v1.1 at 2025-01-09 22:21:08 +0000