media/filters/ffmpeg_video_decoder.cc

   1 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "media/filters/ffmpeg_video_decoder.h"
   6
   7 #include <algorithm>
   8 #include <string>
   9
  10 #include "base/bind.h"
  11 #include "base/callback_helpers.h"
  12 #include "base/command_line.h"
  13 #include "base/location.h"
  14 #include "base/single_thread_task_runner.h"
  15 #include "base/strings/string_number_conversions.h"
  16 #include "media/base/bind_to_current_loop.h"
  17 #include "media/base/decoder_buffer.h"
  18 #include "media/base/limits.h"
  19 #include "media/base/media_switches.h"
  20 #include "media/base/pipeline.h"
  21 #include "media/base/video_decoder_config.h"
  22 #include "media/base/video_frame.h"
  23 #include "media/base/video_util.h"
  24 #include "media/ffmpeg/ffmpeg_common.h"
  25 #include "media/filters/ffmpeg_glue.h"
  26
  27 namespace media {
  28
  29 // Always try to use three threads for video decoding.  There is little reason
  30 // not to since current day CPUs tend to be multi-core and we measured
  31 // performance benefits on older machines such as P4s with hyperthreading.
  32 //
  33 // Handling decoding on separate threads also frees up the pipeline thread to
  34 // continue processing. Although it'd be nice to have the option of a single
  35 // decoding thread, FFmpeg treats having one thread the same as having zero
  36 // threads (i.e., avcodec_decode_video() will execute on the calling thread).
  37 // Yet another reason for having two threads :)
  38 static const int kDecodeThreads = 2;
  39 static const int kMaxDecodeThreads = 16;
  40
  41 // Returns the number of threads given the FFmpeg CodecID. Also inspects the
  42 // command line for a valid --video-threads flag.
  43 static int GetThreadCount(AVCodecID codec_id) {
  44   // Refer to http://crbug.com/93932 for tsan suppressions on decoding.
  45   int decode_threads = kDecodeThreads;
  46
  47   const base::CommandLine* cmd_line = base::CommandLine::ForCurrentProcess();
  48   std::string threads(cmd_line->GetSwitchValueASCII(switches::kVideoThreads));
  49   if (threads.empty() || !base::StringToInt(threads, &decode_threads))
  50     return decode_threads;
  51
  52   decode_threads = std::max(decode_threads, 0);
  53   decode_threads = std::min(decode_threads, kMaxDecodeThreads);
  54   return decode_threads;
  55 }
  56
  57 static int GetVideoBufferImpl(struct AVCodecContext* s,
  58                               AVFrame* frame,
  59                               int flags) {
  60   FFmpegVideoDecoder* decoder = static_cast<FFmpegVideoDecoder*>(s->opaque);
  61   return decoder->GetVideoBuffer(s, frame, flags);
  62 }
  63
  64 static void ReleaseVideoBufferImpl(void* opaque, uint8* data) {
  65   scoped_refptr<VideoFrame> video_frame;
  66   video_frame.swap(reinterpret_cast<VideoFrame**>(&opaque));
  67 }
  68
  69 static size_t RoundUp(size_t value, size_t alignment) {
  70   // Check that |alignment| is a power of 2.
  71   DCHECK((alignment + (alignment - 1)) == (alignment | (alignment - 1)));
  72   return ((value + (alignment - 1)) & ~(alignment - 1));
  73 }
  74
  75 FFmpegVideoDecoder::FFmpegVideoDecoder(
  76     const scoped_refptr<base::SingleThreadTaskRunner>& task_runner)
  77     : task_runner_(task_runner), state_(kUninitialized),
  78       decode_nalus_(false) {}
  79
  80 int FFmpegVideoDecoder::GetVideoBuffer(struct AVCodecContext* codec_context,
  81                                        AVFrame* frame,
  82                                        int flags) {
  83   // Don't use |codec_context_| here! With threaded decoding,
  84   // it will contain unsynchronized width/height/pix_fmt values,
  85   // whereas |codec_context| contains the current threads's
  86   // updated width/height/pix_fmt, which can change for adaptive
  87   // content.
  88   VideoFrame::Format format = PixelFormatToVideoFormat(codec_context->pix_fmt);
  89   if (format == VideoFrame::YV12 &&
  90       codec_context->colorspace == AVCOL_SPC_BT709) {
  91     format = VideoFrame::YV12HD;
  92   }
  93
  94   if (format == VideoFrame::UNKNOWN)
  95     return AVERROR(EINVAL);
  96   DCHECK(format == VideoFrame::YV12 || format == VideoFrame::YV16 ||
  97          format == VideoFrame::YV12J || format == VideoFrame::YV24 ||
  98          format == VideoFrame::YV12HD);
  99
 100   gfx::Size size(codec_context->width, codec_context->height);
 101   const int ret = av_image_check_size(size.width(), size.height(), 0, NULL);
 102   if (ret < 0)
 103     return ret;
 104
 105   gfx::Size natural_size;
 106   if (codec_context->sample_aspect_ratio.num > 0) {
 107     natural_size = GetNaturalSize(size,
 108                                   codec_context->sample_aspect_ratio.num,
 109                                   codec_context->sample_aspect_ratio.den);
 110   } else {
 111     natural_size = config_.natural_size();
 112   }
 113
 114   // FFmpeg has specific requirements on the allocation size of the frame.  The
 115   // following logic replicates FFmpeg's allocation strategy to ensure buffers
 116   // are not overread / overwritten.  See ff_init_buffer_info() for details.
 117   //
 118   // When lowres is non-zero, dimensions should be divided by 2^(lowres), but
 119   // since we don't use this, just DCHECK that it's zero.
 120   //
 121   // Always round up to a multiple of two to match VideoFrame restrictions on
 122   // frame alignment.
 123   DCHECK_EQ(codec_context->lowres, 0);
 124   gfx::Size coded_size(
 125       RoundUp(std::max(size.width(), codec_context->coded_width), 2),
 126       RoundUp(std::max(size.height(), codec_context->coded_height), 2));
 127
 128   if (!VideoFrame::IsValidConfig(
 129           format, coded_size, gfx::Rect(size), natural_size))
 130     return AVERROR(EINVAL);
 131
 132   scoped_refptr<VideoFrame> video_frame = frame_pool_.CreateFrame(
 133       format, coded_size, gfx::Rect(size), natural_size, kNoTimestamp());
 134
 135   for (int i = 0; i < 3; i++) {
 136     frame->data[i] = video_frame->data(i);
 137     frame->linesize[i] = video_frame->stride(i);
 138   }
 139
 140   frame->width = coded_size.width();
 141   frame->height = coded_size.height();
 142   frame->format = codec_context->pix_fmt;
 143   frame->reordered_opaque = codec_context->reordered_opaque;
 144
 145   // Now create an AVBufferRef for the data just allocated. It will own the
 146   // reference to the VideoFrame object.
 147   void* opaque = NULL;
 148   video_frame.swap(reinterpret_cast<VideoFrame**>(&opaque));
 149   frame->buf[0] =
 150       av_buffer_create(frame->data[0],
 151                        VideoFrame::AllocationSize(format, coded_size),
 152                        ReleaseVideoBufferImpl,
 153                        opaque,
 154                        0);
 155   return 0;
 156 }
 157
 158 std::string FFmpegVideoDecoder::GetDisplayName() const {
 159   return "FFmpegVideoDecoder";
 160 }
 161
 162 void FFmpegVideoDecoder::Initialize(const VideoDecoderConfig& config,
 163                                     bool low_delay,
 164                                     const PipelineStatusCB& status_cb,
 165                                     const OutputCB& output_cb) {
 166   DCHECK(task_runner_->BelongsToCurrentThread());
 167   DCHECK(!config.is_encrypted());
 168   DCHECK(!output_cb.is_null());
 169
 170   FFmpegGlue::InitializeFFmpeg();
 171
 172   config_ = config;
 173   PipelineStatusCB initialize_cb = BindToCurrentLoop(status_cb);
 174
 175   if (!config.IsValidConfig() || !ConfigureDecoder(low_delay)) {
 176     initialize_cb.Run(DECODER_ERROR_NOT_SUPPORTED);
 177     return;
 178   }
 179
 180   output_cb_ = BindToCurrentLoop(output_cb);
 181
 182   // Success!
 183   state_ = kNormal;
 184   initialize_cb.Run(PIPELINE_OK);
 185 }
 186
 187 void FFmpegVideoDecoder::Decode(const scoped_refptr<DecoderBuffer>& buffer,
 188                                 const DecodeCB& decode_cb) {
 189   DCHECK(task_runner_->BelongsToCurrentThread());
 190   DCHECK(buffer.get());
 191   DCHECK(!decode_cb.is_null());
 192   CHECK_NE(state_, kUninitialized);
 193
 194   DecodeCB decode_cb_bound = BindToCurrentLoop(decode_cb);
 195
 196   if (state_ == kError) {
 197     decode_cb_bound.Run(kDecodeError);
 198     return;
 199   }
 200
 201   if (state_ == kDecodeFinished) {
 202     decode_cb_bound.Run(kOk);
 203     return;
 204   }
 205
 206   DCHECK_EQ(state_, kNormal);
 207
 208   // During decode, because reads are issued asynchronously, it is possible to
 209   // receive multiple end of stream buffers since each decode is acked. When the
 210   // first end of stream buffer is read, FFmpeg may still have frames queued
 211   // up in the decoder so we need to go through the decode loop until it stops
 212   // giving sensible data.  After that, the decoder should output empty
 213   // frames.  There are three states the decoder can be in:
 214   //
 215   //   kNormal: This is the starting state. Buffers are decoded. Decode errors
 216   //            are discarded.
 217   //   kDecodeFinished: All calls return empty frames.
 218   //   kError: Unexpected error happened.
 219   //
 220   // These are the possible state transitions.
 221   //
 222   // kNormal -> kDecodeFinished:
 223   //     When EOS buffer is received and the codec has been flushed.
 224   // kNormal -> kError:
 225   //     A decoding error occurs and decoding needs to stop.
 226   // (any state) -> kNormal:
 227   //     Any time Reset() is called.
 228
 229   bool has_produced_frame;
 230   do {
 231     has_produced_frame = false;
 232     if (!FFmpegDecode(buffer, &has_produced_frame)) {
 233       state_ = kError;
 234       decode_cb_bound.Run(kDecodeError);
 235       return;
 236     }
 237     // Repeat to flush the decoder after receiving EOS buffer.
 238   } while (buffer->end_of_stream() && has_produced_frame);
 239
 240   if (buffer->end_of_stream())
 241     state_ = kDecodeFinished;
 242
 243   // VideoDecoderShim expects that |decode_cb| is called only after
 244   // |output_cb_|.
 245   decode_cb_bound.Run(kOk);
 246 }
 247
 248 void FFmpegVideoDecoder::Reset(const base::Closure& closure) {
 249   DCHECK(task_runner_->BelongsToCurrentThread());
 250
 251   avcodec_flush_buffers(codec_context_.get());
 252   state_ = kNormal;
 253   task_runner_->PostTask(FROM_HERE, closure);
 254 }
 255
 256 FFmpegVideoDecoder::~FFmpegVideoDecoder() {
 257   DCHECK(task_runner_->BelongsToCurrentThread());
 258
 259   if (state_ != kUninitialized)
 260     ReleaseFFmpegResources();
 261 }
 262
 263 bool FFmpegVideoDecoder::FFmpegDecode(
 264     const scoped_refptr<DecoderBuffer>& buffer,
 265     bool* has_produced_frame) {
 266   DCHECK(!*has_produced_frame);
 267
 268   // Create a packet for input data.
 269   // Due to FFmpeg API changes we no longer have const read-only pointers.
 270   AVPacket packet;
 271   av_init_packet(&packet);
 272   if (buffer->end_of_stream()) {
 273     packet.data = NULL;
 274     packet.size = 0;
 275   } else {
 276     packet.data = const_cast<uint8*>(buffer->data());
 277     packet.size = buffer->data_size();
 278
 279     // Let FFmpeg handle presentation timestamp reordering.
 280     codec_context_->reordered_opaque = buffer->timestamp().InMicroseconds();
 281   }
 282
 283   int frame_decoded = 0;
 284   int result = avcodec_decode_video2(codec_context_.get(),
 285                                      av_frame_.get(),
 286                                      &frame_decoded,
 287                                      &packet);
 288   // Log the problem if we can't decode a video frame and exit early.
 289   if (result < 0) {
 290     LOG(ERROR) << "Error decoding video: " << buffer->AsHumanReadableString();
 291     return false;
 292   }
 293
 294   // FFmpeg says some codecs might have multiple frames per packet.  Previous
 295   // discussions with rbultje@ indicate this shouldn't be true for the codecs
 296   // we use.
 297   DCHECK_EQ(result, packet.size);
 298
 299   // If no frame was produced then signal that more data is required to
 300   // produce more frames. This can happen under two circumstances:
 301   //   1) Decoder was recently initialized/flushed
 302   //   2) End of stream was reached and all internal frames have been output
 303   if (frame_decoded == 0) {
 304     return true;
 305   }
 306
 307   // TODO(fbarchard): Work around for FFmpeg http://crbug.com/27675
 308   // The decoder is in a bad state and not decoding correctly.
 309   // Checking for NULL avoids a crash in CopyPlane().
 310   if (!av_frame_->data[VideoFrame::kYPlane] ||
 311       !av_frame_->data[VideoFrame::kUPlane] ||
 312       !av_frame_->data[VideoFrame::kVPlane]) {
 313     LOG(ERROR) << "Video frame was produced yet has invalid frame data.";
 314     av_frame_unref(av_frame_.get());
 315     return false;
 316   }
 317
 318   scoped_refptr<VideoFrame> frame =
 319       reinterpret_cast<VideoFrame*>(av_buffer_get_opaque(av_frame_->buf[0]));
 320   frame->set_timestamp(
 321       base::TimeDelta::FromMicroseconds(av_frame_->reordered_opaque));
 322   *has_produced_frame = true;
 323   output_cb_.Run(frame);
 324
 325   av_frame_unref(av_frame_.get());
 326   return true;
 327 }
 328
 329 void FFmpegVideoDecoder::ReleaseFFmpegResources() {
 330   codec_context_.reset();
 331   av_frame_.reset();
 332 }
 333
 334 bool FFmpegVideoDecoder::ConfigureDecoder(bool low_delay) {
 335   // Release existing decoder resources if necessary.
 336   ReleaseFFmpegResources();
 337
 338   // Initialize AVCodecContext structure.
 339   codec_context_.reset(avcodec_alloc_context3(NULL));
 340   VideoDecoderConfigToAVCodecContext(config_, codec_context_.get());
 341
 342   codec_context_->thread_count = GetThreadCount(codec_context_->codec_id);
 343   codec_context_->thread_type = low_delay ? FF_THREAD_SLICE : FF_THREAD_FRAME;
 344   codec_context_->opaque = this;
 345   codec_context_->flags |= CODEC_FLAG_EMU_EDGE;
 346   codec_context_->get_buffer2 = GetVideoBufferImpl;
 347   codec_context_->refcounted_frames = 1;
 348
 349   if (decode_nalus_)
 350     codec_context_->flags2 |= CODEC_FLAG2_CHUNKS;
 351
 352   AVCodec* codec = avcodec_find_decoder(codec_context_->codec_id);
 353   if (!codec || avcodec_open2(codec_context_.get(), codec, NULL) < 0) {
 354     ReleaseFFmpegResources();
 355     return false;
 356   }
 357
 358   av_frame_.reset(av_frame_alloc());
 359   return true;
 360 }
 361
 362 }  // namespace media