17d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)// Copyright 2013 The Chromium Authors. All rights reserved.
27d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)// Use of this source code is governed by a BSD-style license that can be
37d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)// found in the LICENSE file.
47d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
57d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#ifndef MEDIA_BASE_AUDIO_BUFFER_H_
67d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#define MEDIA_BASE_AUDIO_BUFFER_H_
77d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
87d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#include <vector>
97d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
107d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#include "base/memory/aligned_memory.h"
117d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#include "base/memory/ref_counted.h"
127d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#include "base/memory/scoped_ptr.h"
13eb525c5499e34cc9c4b825d6d9e75bb07cc06aceBen Murdoch#include "base/time/time.h"
1423730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)#include "media/base/channel_layout.h"
157d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#include "media/base/media_export.h"
167d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#include "media/base/sample_format.h"
177d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
187d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)namespace media {
197d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)class AudioBus;
207d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
217d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)// An audio buffer that takes a copy of the data passed to it, holds it, and
227d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)// copies it into an AudioBus when needed. Also supports an end of stream
237d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)// marker.
247d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)class MEDIA_EXPORT AudioBuffer
257d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)    : public base::RefCountedThreadSafe<AudioBuffer> {
267d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles) public:
273551c9c881056c480085172ff9840cab31610854Torne (Richard Coles)  // Alignment of each channel's data; this must match what ffmpeg expects
283551c9c881056c480085172ff9840cab31610854Torne (Richard Coles)  // (which may be 0, 16, or 32, depending on the processor). Selecting 32 in
293551c9c881056c480085172ff9840cab31610854Torne (Richard Coles)  // order to work on all processors.
303551c9c881056c480085172ff9840cab31610854Torne (Richard Coles)  enum { kChannelAlignment = 32 };
313551c9c881056c480085172ff9840cab31610854Torne (Richard Coles)
327d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // Create an AudioBuffer whose channel data is copied from |data|. For
337d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // interleaved data, only the first buffer is used. For planar data, the
347d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // number of buffers must be equal to |channel_count|. |frame_count| is the
357d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // number of frames in each buffer. |data| must not be null and |frame_count|
367d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // must be >= 0.
377d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  static scoped_refptr<AudioBuffer> CopyFrom(SampleFormat sample_format,
3823730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)                                             ChannelLayout channel_layout,
39effb81e5f8246d0db0270817048dc992db66e9fbBen Murdoch                                             int channel_count,
4023730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)                                             int sample_rate,
417d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)                                             int frame_count,
427d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)                                             const uint8* const* data,
43010d83a9304c5a91596085d917d248abff47903aTorne (Richard Coles)                                             const base::TimeDelta timestamp);
447d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
45a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // Create an AudioBuffer with |frame_count| frames. Buffer is allocated, but
46a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // not initialized. Timestamp and duration are set to kNoTimestamp().
47a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  static scoped_refptr<AudioBuffer> CreateBuffer(SampleFormat sample_format,
4823730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)                                                 ChannelLayout channel_layout,
49effb81e5f8246d0db0270817048dc992db66e9fbBen Murdoch                                                 int channel_count,
5023730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)                                                 int sample_rate,
51a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)                                                 int frame_count);
52a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)
537dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  // Create an empty AudioBuffer with |frame_count| frames.
547dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  static scoped_refptr<AudioBuffer> CreateEmptyBuffer(
5523730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)      ChannelLayout channel_layout,
56effb81e5f8246d0db0270817048dc992db66e9fbBen Murdoch      int channel_count,
5723730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)      int sample_rate,
587dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch      int frame_count,
59010d83a9304c5a91596085d917d248abff47903aTorne (Richard Coles)      const base::TimeDelta timestamp);
607dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch
617d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // Create a AudioBuffer indicating we've reached end of stream.
627d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // Calling any method other than end_of_stream() on the resulting buffer
637d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // is disallowed.
647d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  static scoped_refptr<AudioBuffer> CreateEOSBuffer();
657d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
667d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // Copy frames into |dest|. |frames_to_copy| is the number of frames to copy.
677dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  // |source_frame_offset| specifies how many frames in the buffer to skip
687d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // first. |dest_frame_offset| is the frame offset in |dest|. The frames are
697d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // converted from their source format into planar float32 data (which is all
707d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // that AudioBus handles).
717d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  void ReadFrames(int frames_to_copy,
727d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)                  int source_frame_offset,
737d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)                  int dest_frame_offset,
747d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)                  AudioBus* dest);
757d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
767dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  // Trim an AudioBuffer by removing |frames_to_trim| frames from the start.
77a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // Timestamp and duration are adjusted to reflect the fewer frames.
787dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  // Note that repeated calls to TrimStart() may result in timestamp() and
797dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  // duration() being off by a few microseconds due to rounding issues.
807dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  void TrimStart(int frames_to_trim);
817dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch
82a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // Trim an AudioBuffer by removing |frames_to_trim| frames from the end.
83a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // Duration is adjusted to reflect the fewer frames.
84a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  void TrimEnd(int frames_to_trim);
85a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)
86010d83a9304c5a91596085d917d248abff47903aTorne (Richard Coles)  // Trim an AudioBuffer by removing |end - start| frames from [|start|, |end|).
87010d83a9304c5a91596085d917d248abff47903aTorne (Richard Coles)  // Even if |start| is zero, timestamp() is not adjusted, only duration().
88010d83a9304c5a91596085d917d248abff47903aTorne (Richard Coles)  void TrimRange(int start, int end);
89010d83a9304c5a91596085d917d248abff47903aTorne (Richard Coles)
907dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  // Return the number of channels.
917dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  int channel_count() const { return channel_count_; }
927dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch
937d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // Return the number of frames held.
947dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  int frame_count() const { return adjusted_frame_count_; }
957d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
9623730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)  // Return the sample rate.
9723730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)  int sample_rate() const { return sample_rate_; }
9823730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)
9923730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)  // Return the channel layout.
10023730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)  ChannelLayout channel_layout() const { return channel_layout_; }
10123730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)
1027d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  base::TimeDelta timestamp() const { return timestamp_; }
1037d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  base::TimeDelta duration() const { return duration_; }
1047dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  void set_timestamp(base::TimeDelta timestamp) { timestamp_ = timestamp; }
1057dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch
1067d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // If there's no data in this buffer, it represents end of stream.
1077dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  bool end_of_stream() const { return end_of_stream_; }
1087d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
109a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // Access to the raw buffer for ffmpeg to write directly to. Data for planar
1103551c9c881056c480085172ff9840cab31610854Torne (Richard Coles)  // data is grouped by channel. There is only 1 entry for interleaved formats.
1113551c9c881056c480085172ff9840cab31610854Torne (Richard Coles)  const std::vector<uint8*>& channel_data() const { return channel_data_; }
112a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)
1137d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles) private:
1147d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  friend class base::RefCountedThreadSafe<AudioBuffer>;
1157d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
1167d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // Allocates aligned contiguous buffer to hold all channel data (1 block for
1177d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // interleaved data, |channel_count| blocks for planar data), copies
118a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // [data,data+data_size) to the allocated buffer(s). If |data| is null, no
119a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // data is copied. If |create_buffer| is false, no data buffer is created (or
120a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)  // copied to).
1217d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  AudioBuffer(SampleFormat sample_format,
12223730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)              ChannelLayout channel_layout,
123effb81e5f8246d0db0270817048dc992db66e9fbBen Murdoch              int channel_count,
12423730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)              int sample_rate,
1257d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)              int frame_count,
126a36e5920737c6adbddd3e43b760e5de8431db6e0Torne (Richard Coles)              bool create_buffer,
1277d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)              const uint8* const* data,
128010d83a9304c5a91596085d917d248abff47903aTorne (Richard Coles)              const base::TimeDelta timestamp);
1297d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
1307d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  virtual ~AudioBuffer();
1317d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
1327dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  const SampleFormat sample_format_;
13323730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)  const ChannelLayout channel_layout_;
1347dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  const int channel_count_;
13523730a6e56a168d1879203e4b3819bb36e3d8f1fTorne (Richard Coles)  const int sample_rate_;
1367dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  int adjusted_frame_count_;
1377dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  int trim_start_;
1387dbb3d5cf0c15f500944d211057644d6a2f37371Ben Murdoch  const bool end_of_stream_;
1397d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  base::TimeDelta timestamp_;
1407d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  base::TimeDelta duration_;
1417d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
1427d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // Contiguous block of channel data.
1435d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  scoped_ptr<uint8, base::AlignedFreeDeleter> data_;
1447d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
1457d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  // For planar data, points to each channels data.
1467d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  std::vector<uint8*> channel_data_;
1477d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
1487d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)  DISALLOW_IMPLICIT_CONSTRUCTORS(AudioBuffer);
1497d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)};
1507d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
1517d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)}  // namespace media
1527d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)
1537d4cd473f85ac64c3747c96c277f9e506a0d2246Torne (Richard Coles)#endif  // MEDIA_BASE_AUDIO_BUFFER_H_
154