hrev49858 adds 1 changeset to branch 'master'
old head: 235725ef91b6694691811f5d6e31aa56d43b1b5e
new head: 856cc59e58bb83de4a69daa680748ba9b52592d8
overview:
http://cgit.haiku-os.org/haiku/log/?qt=range&q=856cc59e58bb+%5E235725ef91b6
----------------------------------------------------------------------------
856cc59e58bb: AVCodecDecoder: use swresample to interleave audio channels.
Some codecs will always output audio in planar mode no matter what we
request. This is the case for example with AAC used for youtube. We now
use swresample to convert from planar to packed format.
Note that since swresample does its own buffering, we could probably do
away with some of the code that handled buffering before, making the
audio pipeline simpler and faster.
Fixes audio in youtube, but now the video plays at 2x speed. It seems
something is wrong with the timestamps. Possible things to investigate:
* why do we use the packet dts instead of the pts from the frames anyway?
* the pts and pkt_dts are in "stream time_base units". We seem to assume
microseconds for audio but this is probably not the case. Or did I
miss where the conversion is done?
[ Adrien Destugues <pulkomandy@xxxxxxxxx> ]
----------------------------------------------------------------------------
Revision: hrev49858
Commit: 856cc59e58bb83de4a69daa680748ba9b52592d8
URL: http://cgit.haiku-os.org/haiku/commit/?id=856cc59e58bb
Author: Adrien Destugues <pulkomandy@xxxxxxxxx>
Date: Wed Nov 18 21:43:29 2015 UTC
----------------------------------------------------------------------------
3 files changed, 35 insertions(+), 41 deletions(-)
.../media/plugins/ffmpeg/AVCodecDecoder.cpp | 65 ++++++++------------
.../media/plugins/ffmpeg/AVCodecDecoder.h | 10 ++-
src/add-ons/media/plugins/ffmpeg/Jamfile | 1 +
----------------------------------------------------------------------------
diff --git a/src/add-ons/media/plugins/ffmpeg/AVCodecDecoder.cpp
b/src/add-ons/media/plugins/ffmpeg/AVCodecDecoder.cpp
index bca32a3..b0b6b91 100644
--- a/src/add-ons/media/plugins/ffmpeg/AVCodecDecoder.cpp
+++ b/src/add-ons/media/plugins/ffmpeg/AVCodecDecoder.cpp
@@ -5,12 +5,14 @@
* Copyright (C) 2004 Marcus Overhagen
* Copyright (C) 2009 Stephan Amßus <superstippi@xxxxxx>
* Copyright (C) 2014 Colin Günther <coling@xxxxxx>
+ * Copyright (C) 2015 Adrien Destugues <pulkomandy@xxxxxxxxxxxxx>
*
* All rights reserved. Distributed under the terms of the MIT License.
*/
//! libavcodec based decoder for Haiku
+
#include "AVCodecDecoder.h"
#include <new>
@@ -95,6 +97,7 @@ AVCodecDecoder::AVCodecDecoder()
fIsAudio(false),
fCodec(NULL),
fContext(avcodec_alloc_context3(NULL)),
+ fResampleContext(NULL),
fDecodedData(NULL),
fDecodedDataSizeInBytes(0),
fPostProcessedDecodedPicture(avcodec_alloc_frame()),
@@ -123,7 +126,6 @@ AVCodecDecoder::AVCodecDecoder()
fAudioDecodeError(false),
fDecodedDataBuffer(avcodec_alloc_frame()),
- fDecodedDataBufferOffset(0),
fDecodedDataBufferSize(0)
{
TRACE("AVCodecDecoder::AVCodecDecoder()\n");
@@ -152,6 +154,7 @@ AVCodecDecoder::~AVCodecDecoder()
if (fCodecInitDone)
avcodec_close(fContext);
+ swr_free(&fResampleContext);
free(fChunkBuffer);
free(fDecodedData);
@@ -237,8 +240,7 @@ AVCodecDecoder::Setup(media_format* ioEncodedFormat, const
void* infoBuffer,
} else {
if (fIsAudio) {
fBlockAlign
- =
ioEncodedFormat->u.encoded_audio.output
- .buffer_size;
+ =
ioEncodedFormat->u.encoded_audio.output.buffer_size;
TRACE(" using buffer_size as block align:
%d\n",
fBlockAlign);
}
@@ -278,7 +280,6 @@ AVCodecDecoder::SeekedTo(int64 frame, bigtime_t time)
free(fChunkBuffer);
fChunkBuffer = NULL;
fChunkBufferSize = 0;
- fDecodedDataBufferOffset = 0;
fDecodedDataBufferSize = 0;
fDecodedDataSizeInBytes = 0;
@@ -366,7 +367,6 @@ AVCodecDecoder::_NegotiateAudioOutputFormat(media_format*
inOutFormat)
fChunkBuffer = NULL;
fChunkBufferSize = 0;
fAudioDecodeError = false;
- fDecodedDataBufferOffset = 0;
fDecodedDataBufferSize = 0;
_ResetTempPacket();
@@ -413,6 +413,13 @@ AVCodecDecoder::_NegotiateAudioOutputFormat(media_format*
inOutFormat)
if (fRawDecodedAudio->opaque == NULL)
return B_NO_MEMORY;
+ fResampleContext = swr_alloc_set_opts(NULL,
+ fContext->channel_layout, fContext->request_sample_fmt,
+ fContext->sample_rate,
+ fContext->channel_layout, fContext->sample_fmt,
fContext->sample_rate,
+ 0, NULL);
+ swr_init(fResampleContext);
+
TRACE(" bit_rate = %d, sample_rate = %d, channels = %d, "
"output frame size: %d, count: %ld, rate: %.2f\n",
fContext->bit_rate, fContext->sample_rate, fContext->channels,
@@ -902,24 +909,20 @@
AVCodecDecoder::_MoveAudioFramesToRawDecodedAudioAndUpdateStartTimes()
assert(fRawDecodedAudio->nb_samples < fOutputFrameCount);
assert(fOutputFrameRate > 0);
- int32 frames = min_c(fOutputFrameCount - fRawDecodedAudio->nb_samples,
- fDecodedDataBufferSize / fOutputFrameSize);
- if (frames == 0)
- debugger("fDecodedDataBufferSize not multiple of frame size!");
-
- size_t remainingSize = frames * fOutputFrameSize;
-
-#if 0
// Some decoders do not support format conversion on themselves, or use
// "planar" audio (each channel separated instead of interleaved
samples).
- // If this is a problem we will need to use swresample to convert the
data
- // here, instead of directly copying it.
- swr_convert(fResampleContext, fRawDecodedAudio->data,
- fDecodedDataBuffer->data + fDecodedDataBufferOffset, frames);
-#else
- memcpy(fRawDecodedAudio->data[0], fDecodedDataBuffer->data[0]
- + fDecodedDataBufferOffset, remainingSize);
-#endif
+ // In that case, we use swresample to convert the data (and it is
+ // smart enough to do just a copy, when possible)
+ int32 frames = swr_convert(fResampleContext, fRawDecodedAudio->data,
+ fOutputFrameCount - fRawDecodedAudio->nb_samples,
+ (const uint8_t**)fDecodedDataBuffer->data,
+ fDecodedDataBuffer->nb_samples);
+ if (frames < 0)
+ debugger("resampling failed");
+ size_t remainingSize = frames * fOutputFrameSize;
+
+ // libswresample handles all the buffering for us, how nice of them!
+ fDecodedDataBufferSize = 0;
bool firstAudioFramesCopiedToRawDecodedAudio
= fRawDecodedAudio->data[0] != fDecodedData;
@@ -936,19 +939,6 @@
AVCodecDecoder::_MoveAudioFramesToRawDecodedAudioAndUpdateStartTimes()
fRawDecodedAudio->data[0] += remainingSize;
fRawDecodedAudio->linesize[0] += remainingSize;
fRawDecodedAudio->nb_samples += frames;
-
- fDecodedDataBufferOffset += remainingSize;
- fDecodedDataBufferSize -= remainingSize;
-
- // Update start times accordingly
- bigtime_t framesTimeInterval = static_cast<bigtime_t>(
- (1000000LL * frames) / fOutputFrameRate);
- fDecodedDataBuffer->pkt_dts += framesTimeInterval;
- // Start time of buffer is updated in case that it contains
- // more audio frames to move.
- fTempPacket.dts += framesTimeInterval;
- // Start time of fTempPacket is updated in case the fTempPacket
- // contains more audio frames to decode.
}
@@ -970,8 +960,7 @@
AVCodecDecoder::_MoveAudioFramesToRawDecodedAudioAndUpdateStartTimes()
After this function returns successfully the caller can safely make the
following assumptions:
1. fDecodedDataBufferSize is greater than zero.
- 2. fDecodedDataBufferOffset is set to zero.
- 3. fDecodedDataBuffer contains audio frames.
+ 2. fDecodedDataBuffer contains audio frames.
\returns B_OK on successfully decoding one audio frame chunk.
\returns B_LAST_BUFFER_ERROR No more audio frame chunks available. From
@@ -983,7 +972,7 @@ AVCodecDecoder::_DecodeNextAudioFrameChunk()
{
assert(fDecodedDataBufferSize == 0);
- while(fDecodedDataBufferSize == 0) {
+ while (fDecodedDataBufferSize == 0) {
status_t loadingChunkStatus
= _LoadNextChunkIfNeededAndAssignStartTime();
if (loadingChunkStatus != B_OK)
@@ -1030,7 +1019,6 @@ AVCodecDecoder::_DecodeNextAudioFrameChunk()
Also see "Note" below.
2. fTempPacket was updated to exclude the data chunk that was
consumed
by avcodec_decode_audio4().
- 3. fDecodedDataBufferOffset is set to zero.
When this function failed to decode at least one audio frame due to a
decoding error the caller can safely make the following assumptions:
@@ -1057,7 +1045,6 @@
AVCodecDecoder::_DecodeSomeAudioFramesIntoEmptyDecodedDataBuffer()
assert(fTempPacket.size > 0);
avcodec_get_frame_defaults(fDecodedDataBuffer);
- fDecodedDataBufferOffset = 0;
int gotAudioFrame = 0;
int encodedDataSizeInBytes = avcodec_decode_audio4(fContext,
diff --git a/src/add-ons/media/plugins/ffmpeg/AVCodecDecoder.h
b/src/add-ons/media/plugins/ffmpeg/AVCodecDecoder.h
index a0545b8..1ccf494 100644
--- a/src/add-ons/media/plugins/ffmpeg/AVCodecDecoder.h
+++ b/src/add-ons/media/plugins/ffmpeg/AVCodecDecoder.h
@@ -4,6 +4,7 @@
* Copyright (C) 2001 Axel Dörfler.
* Copyright (C) 2004 Marcus Overhagen.
* Copyright (C) 2009 Stephan Aßmus <superstippi@xxxxxx>.
+ * Copyright (C) 2015 Adrien Destugues <pulkomandy@xxxxxxxxxxxxx>.
*
* All rights reserved. Distributed under the terms of the MIT License.
*/
@@ -12,13 +13,16 @@
//! libavcodec based decoder for Haiku
+
#include <MediaFormats.h>
+
extern "C" {
#include "avcodec.h"
- #include "swscale.h"
+ #include "swresample.h"
}
+
#include "DecoderPlugin.h"
#include "ReaderPlugin.h"
@@ -93,6 +97,7 @@ private:
// FFmpeg related members
AVCodec* fCodec;
AVCodecContext* fContext;
+ SwrContext* fResampleContext;
uint8_t* fDecodedData;
size_t fDecodedDataSizeInBytes;
AVFrame*
fPostProcessedDecodedPicture;
@@ -102,7 +107,9 @@ private:
bool fCodecInitDone;
gfx_convert_func fFormatConversionFunc;
+#if USE_SWS_FOR_COLOR_SPACE_CONVERSION
SwsContext* fSwsContext;
+#endif
char* fExtraData;
int fExtraDataSize;
@@ -119,7 +126,6 @@ private:
bool fAudioDecodeError;
AVFrame* fDecodedDataBuffer;
- int32
fDecodedDataBufferOffset;
int32 fDecodedDataBufferSize;
AVPacket fTempPacket;
diff --git a/src/add-ons/media/plugins/ffmpeg/Jamfile
b/src/add-ons/media/plugins/ffmpeg/Jamfile
index c3efdd0..9c3eeaa 100644
--- a/src/add-ons/media/plugins/ffmpeg/Jamfile
+++ b/src/add-ons/media/plugins/ffmpeg/Jamfile
@@ -47,6 +47,7 @@ for architectureObject in [ MultiArchSubDirSetup ] {
UseHeaders [ FDirName $(ffmpegHeaders) libavformat ] ;
UseHeaders [ FDirName $(ffmpegHeaders) libavutil ] ;
UseHeaders [ FDirName $(ffmpegHeaders) libswscale ] ;
+ UseHeaders [ FDirName $(ffmpegHeaders) libswresample ] ;
Addon [ MultiArchDefaultGristFiles ffmpeg ] :
$(sources)
