Fixed opus encoding completion

This commit is contained in:
Ilya Laktyushin 2019-08-12 05:22:13 +03:00
parent 41c468c139
commit b4541f7495
3 changed files with 201 additions and 212 deletions

View File

@ -81,13 +81,13 @@ typedef enum {
NSDictionary *outputSettings = @
{
AVFormatIDKey: @(kAudioFormatLinearPCM),
AVSampleRateKey: @(TGBridgeAudioEncoderSampleRate),
AVNumberOfChannelsKey: @1,
AVLinearPCMBitDepthKey: @16,
AVLinearPCMIsFloatKey: @false,
AVLinearPCMIsBigEndianKey: @false,
AVLinearPCMIsNonInterleaved: @false
AVFormatIDKey: @(kAudioFormatLinearPCM),
AVSampleRateKey: @(TGBridgeAudioEncoderSampleRate),
AVNumberOfChannelsKey: @1,
AVLinearPCMBitDepthKey: @16,
AVLinearPCMIsFloatKey: @false,
AVLinearPCMIsBigEndianKey: @false,
AVLinearPCMIsNonInterleaved: @false
};
_readerOutput = [AVAssetReaderAudioMixOutput assetReaderAudioMixOutputWithAudioTracks:asset.tracks audioSettings:outputSettings];
@ -114,9 +114,9 @@ typedef enum {
static ATQueue *queue = nil;
static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^
{
queue = [[ATQueue alloc] initWithName:@"org.telegram.opusAudioEncoderQueue"];
});
{
queue = [[ATQueue alloc] initWithName:@"org.telegram.opusAudioEncoderQueue"];
});
return queue;
}
@ -126,74 +126,74 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
- (void)startWithCompletion:(void (^)(NSString *, int32_t))completion
{
[[TGBridgeAudioEncoder processingQueue] dispatch:^
{
_oggWriter = [[TGOggOpusWriter alloc] init];
if (![_oggWriter beginWithDataItem:_tempFileItem])
{
[self cleanup];
return;
}
[_assetReader startReading];
while (_assetReader.status != AVAssetReaderStatusCompleted)
{
if (_assetReader.status == AVAssetReaderStatusReading)
{
CMSampleBufferRef nextBuffer = [_readerOutput copyNextSampleBuffer];
if (nextBuffer)
{
AudioBufferList abl;
CMBlockBufferRef blockBuffer;
CMSampleBufferGetAudioBufferListWithRetainedBlockBuffer(nextBuffer, NULL, &abl, sizeof(abl), NULL, NULL, kCMSampleBufferFlag_AudioBufferList_Assure16ByteAlignment, &blockBuffer);
[[TGBridgeAudioEncoder processingQueue] dispatch:^
{
[self _processBuffer:&abl.mBuffers[0]];
CFRelease(nextBuffer);
CFRelease(blockBuffer);
}];
}
else
{
[[TGBridgeAudioEncoder processingQueue] dispatch:^
{
if (_tailLength > 0) {
[_oggWriter writeFrame:(uint8_t *)_audioBuffer.bytes frameByteCount:(NSUInteger)_tailLength];
}
}];
break;
}
}
}
[[TGBridgeAudioEncoder processingQueue] dispatch:^
{
TGFileDataItem *dataItemResult = nil;
NSTimeInterval durationResult = 0.0;
NSUInteger totalBytes = 0;
if (_assetReader.status == AVAssetReaderStatusCompleted)
{
NSLog(@"finished");
if (_oggWriter != nil)
{
dataItemResult = _tempFileItem;
durationResult = [_oggWriter encodedDuration];
totalBytes = [_oggWriter encodedBytes];
}
[self cleanup];
}
//TGLog(@"[TGBridgeAudioEncoder#%x convert time: %f ms]", self, (CFAbsoluteTimeGetCurrent() - startTime) * 1000.0);
if (completion != nil)
completion(dataItemResult.path, (int32_t)durationResult);
}];
}];
{
_oggWriter = [[TGOggOpusWriter alloc] init];
if (![_oggWriter beginWithDataItem:_tempFileItem])
{
[self cleanup];
return;
}
[_assetReader startReading];
while (_assetReader.status != AVAssetReaderStatusCompleted)
{
if (_assetReader.status == AVAssetReaderStatusReading)
{
CMSampleBufferRef nextBuffer = [_readerOutput copyNextSampleBuffer];
if (nextBuffer)
{
AudioBufferList abl;
CMBlockBufferRef blockBuffer;
CMSampleBufferGetAudioBufferListWithRetainedBlockBuffer(nextBuffer, NULL, &abl, sizeof(abl), NULL, NULL, kCMSampleBufferFlag_AudioBufferList_Assure16ByteAlignment, &blockBuffer);
[[TGBridgeAudioEncoder processingQueue] dispatch:^
{
[self _processBuffer:&abl.mBuffers[0]];
CFRelease(nextBuffer);
CFRelease(blockBuffer);
}];
}
else
{
[[TGBridgeAudioEncoder processingQueue] dispatch:^
{
if (_tailLength > 0) {
[_oggWriter writeFrame:(uint8_t *)_audioBuffer.bytes frameByteCount:(NSUInteger)_tailLength];
}
}];
break;
}
}
}
[[TGBridgeAudioEncoder processingQueue] dispatch:^
{
TGFileDataItem *dataItemResult = nil;
NSTimeInterval durationResult = 0.0;
NSUInteger totalBytes = 0;
if (_assetReader.status == AVAssetReaderStatusCompleted)
{
NSLog(@"finished");
if (_oggWriter != nil && [_oggWriter writeFrame:NULL frameByteCount:0])
{
dataItemResult = _tempFileItem;
durationResult = [_oggWriter encodedDuration];
totalBytes = [_oggWriter encodedBytes];
}
[self cleanup];
}
//TGLog(@"[TGBridgeAudioEncoder#%x convert time: %f ms]", self, (CFAbsoluteTimeGetCurrent() - startTime) * 1000.0);
if (completion != nil)
completion(dataItemResult.path, (int32_t)durationResult);
}];
}];
}
- (void)_processBuffer:(AudioBuffer const *)buffer
@ -306,11 +306,11 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
[_queue dispatch:^
{
_fileName = filePath;
_length = [[[NSFileManager defaultManager] attributesOfItemAtPath:_fileName error:nil][NSFileSize] unsignedIntegerValue];
_fileExists = [[NSFileManager defaultManager] fileExistsAtPath:_fileName];
}];
{
_fileName = filePath;
_length = [[[NSFileManager defaultManager] attributesOfItemAtPath:_fileName error:nil][NSFileSize] unsignedIntegerValue];
_fileExists = [[NSFileManager defaultManager] fileExistsAtPath:_fileName];
}];
}
return self;
}
@ -322,38 +322,38 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
- (void)moveToPath:(NSString *)path
{
[_queue dispatch:^
{
[[NSFileManager defaultManager] moveItemAtPath:_fileName toPath:path error:nil];
_fileName = path;
}];
{
[[NSFileManager defaultManager] moveItemAtPath:_fileName toPath:path error:nil];
_fileName = path;
}];
}
- (void)remove
{
[_queue dispatch:^
{
[[NSFileManager defaultManager] removeItemAtPath:_fileName error:nil];
}];
{
[[NSFileManager defaultManager] removeItemAtPath:_fileName error:nil];
}];
}
- (void)appendData:(NSData *)data
{
[_queue dispatch:^
{
if (!_fileExists)
{
[[NSFileManager defaultManager] createFileAtPath:_fileName contents:nil attributes:nil];
_fileExists = true;
}
NSFileHandle *file = [NSFileHandle fileHandleForUpdatingAtPath:_fileName];
[file seekToEndOfFile];
[file writeData:data];
[file synchronizeFile];
[file closeFile];
_length += data.length;
[_data appendData:data];
}];
{
if (!_fileExists)
{
[[NSFileManager defaultManager] createFileAtPath:_fileName contents:nil attributes:nil];
_fileExists = true;
}
NSFileHandle *file = [NSFileHandle fileHandleForUpdatingAtPath:_fileName];
[file seekToEndOfFile];
[file writeData:data];
[file synchronizeFile];
[file closeFile];
_length += data.length;
[_data appendData:data];
}];
}
- (NSData *)readDataAtOffset:(NSUInteger)offset length:(NSUInteger)length
@ -361,14 +361,14 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
__block NSData *data = nil;
[_queue dispatch:^
{
NSFileHandle *file = [NSFileHandle fileHandleForUpdatingAtPath:_fileName];
[file seekToFileOffset:(unsigned long long)offset];
data = [file readDataOfLength:length];
if (data.length != length)
//TGLog(@"Read data length mismatch");
[file closeFile];
} synchronous:true];
{
NSFileHandle *file = [NSFileHandle fileHandleForUpdatingAtPath:_fileName];
[file seekToFileOffset:(unsigned long long)offset];
data = [file readDataOfLength:length];
if (data.length != length)
//TGLog(@"Read data length mismatch");
[file closeFile];
} synchronous:true];
return data;
}
@ -377,9 +377,9 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
{
__block NSUInteger result = 0;
[_queue dispatch:^
{
result = _length;
} synchronous:true];
{
result = _length;
} synchronous:true];
return result;
}
@ -420,11 +420,11 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
static ATQueue *queue = nil;
static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^
{
queue = [[ATQueue alloc] init];
queue->_nativeQueue = dispatch_get_main_queue();
queue->_isMainQueue = true;
});
{
queue = [[ATQueue alloc] init];
queue->_nativeQueue = dispatch_get_main_queue();
queue->_isMainQueue = true;
});
return queue;
}
@ -434,9 +434,9 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
static ATQueue *queue = nil;
static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^
{
queue = [[ATQueue alloc] initWithNativeQueue:dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0)];
});
{
queue = [[ATQueue alloc] initWithNativeQueue:dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0)];
});
return queue;
}
@ -446,9 +446,9 @@ static const int encoderPacketSizeInBytes = TGBridgeAudioEncoderSampleRate / 100
static ATQueue *queue = nil;
static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^
{
queue = [[ATQueue alloc] initWithNativeQueue:dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_BACKGROUND, 0)];
});
{
queue = [[ATQueue alloc] initWithNativeQueue:dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_BACKGROUND, 0)];
});
return queue;
}

View File

@ -319,7 +319,7 @@ final class ManagedAudioRecorderContext {
self.idleTimerExtensionDisposable = (Signal<Void, NoError> { subscriber in
return pushIdleTimerExtension()
} |> delay(5.0, queue: queue)).start()
} |> delay(5.0, queue: queue)).start()
}
deinit {
@ -406,19 +406,19 @@ final class ManagedAudioRecorderContext {
strongSelf.audioSessionAcquired(headset: state.isHeadsetConnected)
}
}
}, deactivate: { [weak self] in
return Signal { subscriber in
queue.async {
if let strongSelf = self {
strongSelf.hasAudioSession = false
strongSelf.stop()
strongSelf.recordingState.set(.stopped)
subscriber.putCompletion()
}
}, deactivate: { [weak self] in
return Signal { subscriber in
queue.async {
if let strongSelf = self {
strongSelf.hasAudioSession = false
strongSelf.stop()
strongSelf.recordingState.set(.stopped)
subscriber.putCompletion()
}
return EmptyDisposable
}
return EmptyDisposable
}
})
}
}
@ -592,53 +592,57 @@ final class ManagedAudioRecorderContext {
}
func takeData() -> RecordedAudioData? {
var scaledSamplesMemory = malloc(100 * 2)!
var scaledSamples: UnsafeMutablePointer<Int16> = scaledSamplesMemory.assumingMemoryBound(to: Int16.self)
defer {
free(scaledSamplesMemory)
}
memset(scaledSamples, 0, 100 * 2);
var waveform: Data?
let count = self.compressedWaveformSamples.count / 2
self.compressedWaveformSamples.withUnsafeMutableBytes { (samples: UnsafeMutablePointer<Int16>) -> Void in
for i in 0 ..< count {
let sample = samples[i]
let index = i * 100 / count
if (scaledSamples[index] < sample) {
scaledSamples[index] = sample;
if self.oggWriter.writeFrame(nil, frameByteCount: 0) {
var scaledSamplesMemory = malloc(100 * 2)!
var scaledSamples: UnsafeMutablePointer<Int16> = scaledSamplesMemory.assumingMemoryBound(to: Int16.self)
defer {
free(scaledSamplesMemory)
}
memset(scaledSamples, 0, 100 * 2);
var waveform: Data?
let count = self.compressedWaveformSamples.count / 2
self.compressedWaveformSamples.withUnsafeMutableBytes { (samples: UnsafeMutablePointer<Int16>) -> Void in
for i in 0 ..< count {
let sample = samples[i]
let index = i * 100 / count
if (scaledSamples[index] < sample) {
scaledSamples[index] = sample;
}
}
}
var peak: Int16 = 0
var sumSamples: Int64 = 0
for i in 0 ..< 100 {
let sample = scaledSamples[i]
if peak < sample {
peak = sample
var peak: Int16 = 0
var sumSamples: Int64 = 0
for i in 0 ..< 100 {
let sample = scaledSamples[i]
if peak < sample {
peak = sample
}
sumSamples += Int64(sample)
}
sumSamples += Int64(sample)
}
var calculatedPeak: UInt16 = 0
calculatedPeak = UInt16((Double(sumSamples) * 1.8 / 100.0))
if calculatedPeak < 2500 {
calculatedPeak = 2500
var calculatedPeak: UInt16 = 0
calculatedPeak = UInt16((Double(sumSamples) * 1.8 / 100.0))
if calculatedPeak < 2500 {
calculatedPeak = 2500
}
for i in 0 ..< 100 {
let sample: UInt16 = UInt16(Int64(scaledSamples[i]))
let minPeak = min(Int64(sample), Int64(calculatedPeak))
let resultPeak = minPeak * 31 / Int64(calculatedPeak)
scaledSamples[i] = Int16(clamping: min(31, resultPeak))
}
let resultWaveform = AudioWaveform(samples: Data(bytes: scaledSamplesMemory, count: 100 * 2), peak: 31)
let bitstream = resultWaveform.makeBitstream()
waveform = AudioWaveform(bitstream: bitstream, bitsPerSample: 5).makeBitstream()
}
for i in 0 ..< 100 {
let sample: UInt16 = UInt16(Int64(scaledSamples[i]))
let minPeak = min(Int64(sample), Int64(calculatedPeak))
let resultPeak = minPeak * 31 / Int64(calculatedPeak)
scaledSamples[i] = Int16(clamping: min(31, resultPeak))
}
let resultWaveform = AudioWaveform(samples: Data(bytes: scaledSamplesMemory, count: 100 * 2), peak: 31)
let bitstream = resultWaveform.makeBitstream()
waveform = AudioWaveform(bitstream: bitstream, bitsPerSample: 5).makeBitstream()
return RecordedAudioData(compressedData: self.dataItem.data(), duration: self.oggWriter.encodedDuration(), waveform: waveform)
} else {
return nil
}
return RecordedAudioData(compressedData: self.dataItem.data(), duration: self.oggWriter.encodedDuration(), waveform: waveform)
}
}

View File

@ -426,39 +426,24 @@ static inline int writeOggPage(ogg_page *page, TGDataItem *fileItem)
pages_out++;
}
op.packet = (unsigned char *)_packet;
op.bytes = nbBytes;
op.b_o_s = 0;
op.granulepos = enc_granulepos;
if (op.e_o_s)
{
/* We compute the final GP as ceil(len*48k/input_rate). When a resampling
decoder does the matching floor(len*input/48k) conversion the length will
be exactly the same as the input.
*/
op.granulepos = ((total_samples * 48000 + rate - 1) / rate) + header.preskip;
}
op.packetno = 2 + _packetId;
ogg_stream_packetin(&os, &op);
last_segments += size_segments;
if (framePcmBytes != NULL) {
op.packet = (unsigned char *)_packet;
op.bytes = nbBytes;
op.b_o_s = 0;
op.granulepos = enc_granulepos;
/* The downside of early reading is if the input is an exact
multiple of the frame_size you'll get an extra frame that needs
to get cropped off. The downside of late reading is added delay.
If your ogg_delay is 120ms or less we'll assume you want the
low delay behavior.
*/
/*if ((!op.e_o_s) && max_ogg_delay > 5760)
{
nb_samples = inopt.read_samples(inopt.readdata, input, frame_size);
total_samples += nb_samples;
if (nb_samples < frame_size)
eos = 1;
if (nb_samples == 0)
op.e_o_s = 1;
if (op.e_o_s)
{
/* We compute the final GP as ceil(len*48k/input_rate). When a resampling
decoder does the matching floor(len*input/48k) conversion the length will
be exactly the same as the input.
*/
op.granulepos = ((total_samples * 48000 + rate - 1) / rate) + header.preskip;
}
op.packetno = 2 + _packetId;
ogg_stream_packetin(&os, &op);
last_segments += size_segments;
}
else
nb_samples = -1;*/
// If the stream is over or we're sure that the delayed flush will fire, go ahead and flush now to avoid adding delay
while ((op.e_o_s || (enc_granulepos + (frame_size * 48000 / coding_rate) - last_granulepos > max_ogg_delay) ||