media/cast/sender/vp8_encoder.cc

   1 // Copyright 2014 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/cast/sender/vp8_encoder.h"
   6
   7 #include "base/logging.h"
   8 #include "media/base/video_frame.h"
   9 #include "media/cast/cast_defines.h"
  10 #include "media/cast/net/cast_transport_config.h"
  11 #include "third_party/libvpx/source/libvpx/vpx/vp8cx.h"
  12
  13 namespace media {
  14 namespace cast {
  15
  16 namespace {
  17
  18 // After a pause in the video stream, what is the maximum duration amount to
  19 // pass to the encoder for the next frame (in terms of 1/max_fps sized periods)?
  20 // This essentially controls the encoded size of the first frame that follows a
  21 // pause in the video stream.
  22 const int kRestartFramePeriods = 3;
  23
  24 }  // namespace
  25
  26 Vp8Encoder::Vp8Encoder(const VideoSenderConfig& video_config)
  27     : cast_config_(video_config),
  28       use_multiple_video_buffers_(
  29           cast_config_.max_number_of_video_buffers_used ==
  30           kNumberOfVp8VideoBuffers),
  31       key_frame_requested_(true),
  32       bitrate_kbit_(cast_config_.start_bitrate / 1000),
  33       last_encoded_frame_id_(kStartFrameId),
  34       last_acked_frame_id_(kStartFrameId),
  35       undroppable_frames_(0) {
  36   config_.g_timebase.den = 0;  // Not initialized.
  37
  38   for (int i = 0; i < kNumberOfVp8VideoBuffers; ++i) {
  39     buffer_state_[i].frame_id = last_encoded_frame_id_;
  40     buffer_state_[i].state = kBufferStartState;
  41   }
  42
  43   // VP8 have 3 buffers available for prediction, with
  44   // max_number_of_video_buffers_used set to 1 we maximize the coding efficiency
  45   // however in this mode we can not skip frames in the receiver to catch up
  46   // after a temporary network outage; with max_number_of_video_buffers_used
  47   // set to 3 we allow 2 frames to be skipped by the receiver without error
  48   // propagation.
  49   DCHECK(cast_config_.max_number_of_video_buffers_used == 1 ||
  50          cast_config_.max_number_of_video_buffers_used ==
  51              kNumberOfVp8VideoBuffers)
  52       << "Invalid argument";
  53
  54   thread_checker_.DetachFromThread();
  55 }
  56
  57 Vp8Encoder::~Vp8Encoder() {
  58   DCHECK(thread_checker_.CalledOnValidThread());
  59   if (is_initialized())
  60     vpx_codec_destroy(&encoder_);
  61 }
  62
  63 void Vp8Encoder::Initialize() {
  64   DCHECK(thread_checker_.CalledOnValidThread());
  65   DCHECK(!is_initialized());
  66   // The encoder will be created/configured when the first frame encode is
  67   // requested.
  68 }
  69
  70 void Vp8Encoder::ConfigureForNewFrameSize(const gfx::Size& frame_size) {
  71   if (is_initialized()) {
  72     // Workaround for VP8 bug: If the new size is strictly less-than-or-equal to
  73     // the old size, in terms of area, the existing encoder instance can
  74     // continue.  Otherwise, completely tear-down and re-create a new encoder to
  75     // avoid a shutdown crash.
  76     if (frame_size.GetArea() <= gfx::Size(config_.g_w, config_.g_h).GetArea() &&
  77         !use_multiple_video_buffers_) {
  78       DVLOG(1) << "Continuing to use existing encoder at smaller frame size: "
  79                << gfx::Size(config_.g_w, config_.g_h).ToString() << " --> "
  80                << frame_size.ToString();
  81       config_.g_w = frame_size.width();
  82       config_.g_h = frame_size.height();
  83       CHECK_EQ(vpx_codec_enc_config_set(&encoder_, &config_), VPX_CODEC_OK)
  84           << "Failed to update frame size in encoder config.";
  85       return;
  86     }
  87
  88     DVLOG(1) << "Destroying/Re-Creating encoder for larger frame size: "
  89              << gfx::Size(config_.g_w, config_.g_h).ToString() << " --> "
  90              << frame_size.ToString();
  91     vpx_codec_destroy(&encoder_);
  92   } else {
  93     DVLOG(1) << "Creating encoder for the first frame; size: "
  94              << frame_size.ToString();
  95   }
  96
  97   // Reset multi-buffer mode state.
  98   last_acked_frame_id_ = last_encoded_frame_id_;
  99   undroppable_frames_ = 0;
 100   for (int i = 0; i < kNumberOfVp8VideoBuffers; ++i) {
 101     buffer_state_[i].frame_id = last_encoded_frame_id_;
 102     buffer_state_[i].state = kBufferStartState;
 103   }
 104
 105   // Populate encoder configuration with default values.
 106   CHECK_EQ(vpx_codec_enc_config_default(vpx_codec_vp8_cx(), &config_, 0),
 107            VPX_CODEC_OK);
 108
 109   config_.g_threads = cast_config_.number_of_encode_threads;
 110   config_.g_w = frame_size.width();
 111   config_.g_h = frame_size.height();
 112   // Set the timebase to match that of base::TimeDelta.
 113   config_.g_timebase.num = 1;
 114   config_.g_timebase.den = base::Time::kMicrosecondsPerSecond;
 115   if (use_multiple_video_buffers_) {
 116     // We must enable error resilience when we use multiple buffers, due to
 117     // codec requirements.
 118     config_.g_error_resilient = 1;
 119   }
 120   // |g_pass| and |g_lag_in_frames| must be "one pass" and zero, respectively,
 121   // in order for VP8 to support changing frame sizes during encoding:
 122   config_.g_pass = VPX_RC_ONE_PASS;
 123   config_.g_lag_in_frames = 0;  // Immediate data output for each frame.
 124
 125   // Rate control settings.
 126   config_.rc_dropframe_thresh = 0;  // The encoder may not drop any frames.
 127   config_.rc_resize_allowed = 0;  // TODO(miu): Why not?  Investigate this.
 128   config_.rc_end_usage = VPX_CBR;
 129   config_.rc_target_bitrate = bitrate_kbit_;
 130   config_.rc_min_quantizer = cast_config_.min_qp;
 131   config_.rc_max_quantizer = cast_config_.max_qp;
 132   // TODO(miu): Revisit these now that the encoder is being successfully
 133   // micro-managed.
 134   config_.rc_undershoot_pct = 100;
 135   config_.rc_overshoot_pct = 15;
 136   // TODO(miu): Document why these rc_buf_*_sz values were chosen and/or
 137   // research for better values.  Should they be computed from the target
 138   // playout delay?
 139   config_.rc_buf_initial_sz = 500;
 140   config_.rc_buf_optimal_sz = 600;
 141   config_.rc_buf_sz = 1000;
 142
 143   config_.kf_mode = VPX_KF_DISABLED;
 144
 145   vpx_codec_flags_t flags = 0;
 146   CHECK_EQ(vpx_codec_enc_init(&encoder_, vpx_codec_vp8_cx(), &config_, flags),
 147            VPX_CODEC_OK);
 148
 149   // Raise the threshold for considering macroblocks as static.  The default is
 150   // zero, so this setting makes the encoder less sensitive to motion.  This
 151   // lowers the probability of needing to utilize more CPU to search for motion
 152   // vectors.
 153   CHECK_EQ(vpx_codec_control(&encoder_, VP8E_SET_STATIC_THRESHOLD, 1),
 154            VPX_CODEC_OK);
 155
 156   // Improve quality by enabling sets of codec features that utilize more CPU.
 157   // The default is zero, with increasingly more CPU to be used as the value is
 158   // more negative.
 159   // TODO(miu): Document why this value was chosen and expected behaviors.
 160   // Should this be dynamic w.r.t. hardware performance?
 161   CHECK_EQ(vpx_codec_control(&encoder_, VP8E_SET_CPUUSED, -6), VPX_CODEC_OK);
 162 }
 163
 164 void Vp8Encoder::Encode(const scoped_refptr<media::VideoFrame>& video_frame,
 165                         const base::TimeTicks& reference_time,
 166                         EncodedFrame* encoded_frame) {
 167   DCHECK(thread_checker_.CalledOnValidThread());
 168   DCHECK(encoded_frame);
 169
 170   // Initialize on-demand.  Later, if the video frame size has changed, update
 171   // the encoder configuration.
 172   const gfx::Size frame_size = video_frame->visible_rect().size();
 173   if (!is_initialized() || gfx::Size(config_.g_w, config_.g_h) != frame_size)
 174     ConfigureForNewFrameSize(frame_size);
 175
 176   uint32 latest_frame_id_to_reference;
 177   Vp8Buffers buffer_to_update;
 178   vpx_codec_flags_t flags = 0;
 179   if (key_frame_requested_) {
 180     flags = VPX_EFLAG_FORCE_KF;
 181     // Self reference.
 182     latest_frame_id_to_reference = last_encoded_frame_id_ + 1;
 183     // We can pick any buffer as buffer_to_update since we update
 184     // them all.
 185     buffer_to_update = kLastBuffer;
 186   } else {
 187     // Reference all acked frames (buffers).
 188     latest_frame_id_to_reference = GetCodecReferenceFlags(&flags);
 189     buffer_to_update = GetNextBufferToUpdate();
 190     GetCodecUpdateFlags(buffer_to_update, &flags);
 191   }
 192
 193   // Wrapper for vpx_codec_encode() to access the YUV data in the |video_frame|.
 194   // Only the VISIBLE rectangle within |video_frame| is exposed to the codec.
 195   vpx_image_t vpx_image;
 196   vpx_image_t* const result = vpx_img_wrap(
 197       &vpx_image,
 198       VPX_IMG_FMT_I420,
 199       frame_size.width(),
 200       frame_size.height(),
 201       1,
 202       video_frame->data(VideoFrame::kYPlane));
 203   DCHECK_EQ(result, &vpx_image);
 204   vpx_image.planes[VPX_PLANE_Y] =
 205       video_frame->visible_data(VideoFrame::kYPlane);
 206   vpx_image.planes[VPX_PLANE_U] =
 207       video_frame->visible_data(VideoFrame::kUPlane);
 208   vpx_image.planes[VPX_PLANE_V] =
 209       video_frame->visible_data(VideoFrame::kVPlane);
 210   vpx_image.stride[VPX_PLANE_Y] = video_frame->stride(VideoFrame::kYPlane);
 211   vpx_image.stride[VPX_PLANE_U] = video_frame->stride(VideoFrame::kUPlane);
 212   vpx_image.stride[VPX_PLANE_V] = video_frame->stride(VideoFrame::kVPlane);
 213
 214   // The frame duration given to the VP8 codec affects a number of important
 215   // behaviors, including: per-frame bandwidth, CPU time spent encoding,
 216   // temporal quality trade-offs, and key/golden/alt-ref frame generation
 217   // intervals.  Use the actual amount of time between the current and previous
 218   // frames as a prediction for the next frame's duration, but bound the
 219   // prediction to account for the fact that the frame rate can be highly
 220   // variable, including long pauses in the video stream.
 221   const base::TimeDelta minimum_frame_duration =
 222       base::TimeDelta::FromSecondsD(1.0 / cast_config_.max_frame_rate);
 223   const base::TimeDelta maximum_frame_duration =
 224       base::TimeDelta::FromSecondsD(static_cast<double>(kRestartFramePeriods) /
 225                                         cast_config_.max_frame_rate);
 226   const base::TimeDelta last_frame_duration =
 227       video_frame->timestamp() - last_frame_timestamp_;
 228   const base::TimeDelta predicted_frame_duration =
 229       std::max(minimum_frame_duration,
 230                std::min(maximum_frame_duration, last_frame_duration));
 231   last_frame_timestamp_ = video_frame->timestamp();
 232
 233   // Encode the frame.  The presentation time stamp argument here is fixed to
 234   // zero to force the encoder to base its single-frame bandwidth calculations
 235   // entirely on |predicted_frame_duration| and the target bitrate setting being
 236   // micro-managed via calls to UpdateRates().
 237   CHECK_EQ(vpx_codec_encode(&encoder_,
 238                             &vpx_image,
 239                             0,
 240                             predicted_frame_duration.InMicroseconds(),
 241                             flags,
 242                             VPX_DL_REALTIME),
 243            VPX_CODEC_OK)
 244       << "BUG: Invalid arguments passed to vpx_codec_encode().";
 245
 246   // Pull data from the encoder, populating a new EncodedFrame.
 247   encoded_frame->frame_id = ++last_encoded_frame_id_;
 248   const vpx_codec_cx_pkt_t* pkt = NULL;
 249   vpx_codec_iter_t iter = NULL;
 250   while ((pkt = vpx_codec_get_cx_data(&encoder_, &iter)) != NULL) {
 251     if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
 252       continue;
 253     if (pkt->data.frame.flags & VPX_FRAME_IS_KEY) {
 254       // TODO(hubbe): Replace "dependency" with a "bool is_key_frame".
 255       encoded_frame->dependency = EncodedFrame::KEY;
 256       encoded_frame->referenced_frame_id = encoded_frame->frame_id;
 257     } else {
 258       encoded_frame->dependency = EncodedFrame::DEPENDENT;
 259       // Frame dependencies could theoretically be relaxed by looking for the
 260       // VPX_FRAME_IS_DROPPABLE flag, but in recent testing (Oct 2014), this
 261       // flag never seems to be set.
 262       encoded_frame->referenced_frame_id = latest_frame_id_to_reference;
 263     }
 264     encoded_frame->rtp_timestamp =
 265         TimeDeltaToRtpDelta(video_frame->timestamp(), kVideoFrequency);
 266     encoded_frame->reference_time = reference_time;
 267     encoded_frame->data.assign(
 268         static_cast<const uint8*>(pkt->data.frame.buf),
 269         static_cast<const uint8*>(pkt->data.frame.buf) + pkt->data.frame.sz);
 270     break;  // Done, since all data is provided in one CX_FRAME_PKT packet.
 271   }
 272   DCHECK(!encoded_frame->data.empty())
 273       << "BUG: Encoder must provide data since lagged encoding is disabled.";
 274
 275   DVLOG(2) << "VP8 encoded frame_id " << encoded_frame->frame_id
 276            << ", sized:" << encoded_frame->data.size();
 277
 278   if (encoded_frame->dependency == EncodedFrame::KEY) {
 279     key_frame_requested_ = false;
 280
 281     for (int i = 0; i < kNumberOfVp8VideoBuffers; ++i) {
 282       buffer_state_[i].state = kBufferSent;
 283       buffer_state_[i].frame_id = encoded_frame->frame_id;
 284     }
 285   } else {
 286     if (buffer_to_update != kNoBuffer) {
 287       buffer_state_[buffer_to_update].state = kBufferSent;
 288       buffer_state_[buffer_to_update].frame_id = encoded_frame->frame_id;
 289     }
 290   }
 291 }
 292
 293 uint32 Vp8Encoder::GetCodecReferenceFlags(vpx_codec_flags_t* flags) {
 294   if (!use_multiple_video_buffers_)
 295     return last_encoded_frame_id_;
 296
 297   const uint32 kMagicFrameOffset = 512;
 298   // We set latest_frame_to_reference to an old frame so that
 299   // IsNewerFrameId will work correctly.
 300   uint32 latest_frame_to_reference =
 301       last_encoded_frame_id_ - kMagicFrameOffset;
 302
 303   // Reference all acked frames.
 304   // TODO(hubbe): We may also want to allow references to the
 305   // last encoded frame, if that frame was assigned to a buffer.
 306   for (int i = 0; i < kNumberOfVp8VideoBuffers; ++i) {
 307     if (buffer_state_[i].state == kBufferAcked) {
 308       if (IsNewerFrameId(buffer_state_[i].frame_id,
 309                          latest_frame_to_reference)) {
 310         latest_frame_to_reference = buffer_state_[i].frame_id;
 311       }
 312     } else {
 313       switch (i) {
 314         case kAltRefBuffer:
 315           *flags |= VP8_EFLAG_NO_REF_ARF;
 316           break;
 317         case kGoldenBuffer:
 318           *flags |= VP8_EFLAG_NO_REF_GF;
 319           break;
 320         case kLastBuffer:
 321           *flags |= VP8_EFLAG_NO_REF_LAST;
 322           break;
 323       }
 324     }
 325   }
 326
 327   if (latest_frame_to_reference ==
 328       last_encoded_frame_id_ - kMagicFrameOffset) {
 329     // We have nothing to reference, it's kind of like a key frame,
 330     // but doesn't reset buffers.
 331     latest_frame_to_reference = last_encoded_frame_id_ + 1;
 332   }
 333
 334   return latest_frame_to_reference;
 335 }
 336
 337 Vp8Encoder::Vp8Buffers Vp8Encoder::GetNextBufferToUpdate() {
 338   if (!use_multiple_video_buffers_)
 339     return kNoBuffer;
 340
 341   // The goal here is to make sure that we always keep one ACKed
 342   // buffer while trying to get an ACK for a newer buffer as we go.
 343   // Here are the rules for which buffer to select for update:
 344   // 1. If there is a buffer in state kStartState, use it.
 345   // 2. If there is a buffer other than the oldest buffer
 346   //    which is Acked, use the oldest buffer.
 347   // 3. If there are Sent buffers which are older than
 348   //    latest_acked_frame_, use the oldest one.
 349   // 4. If all else fails, just overwrite the newest buffer,
 350   //    but no more than 3 times in a row.
 351   //    TODO(hubbe): Figure out if 3 is optimal.
 352   // Note, rule 1-3 describe cases where there is a "free" buffer
 353   // that we can use. Rule 4 describes what happens when there is
 354   // no free buffer available.
 355
 356   // Buffers, sorted from oldest frame to newest.
 357   Vp8Encoder::Vp8Buffers buffers[kNumberOfVp8VideoBuffers];
 358
 359   for (int i = 0; i < kNumberOfVp8VideoBuffers; ++i) {
 360     Vp8Encoder::Vp8Buffers buffer = static_cast<Vp8Encoder::Vp8Buffers>(i);
 361
 362     // Rule 1
 363     if (buffer_state_[buffer].state == kBufferStartState) {
 364       undroppable_frames_ = 0;
 365       return buffer;
 366     }
 367     buffers[buffer] = buffer;
 368   }
 369
 370   // Sorting three elements with selection sort.
 371   for (int i = 0; i < kNumberOfVp8VideoBuffers - 1; i++) {
 372     for (int j = i + 1; j < kNumberOfVp8VideoBuffers; j++) {
 373       if (IsOlderFrameId(buffer_state_[buffers[j]].frame_id,
 374                          buffer_state_[buffers[i]].frame_id)) {
 375         std::swap(buffers[i], buffers[j]);
 376       }
 377     }
 378   }
 379
 380   // Rule 2
 381   if (buffer_state_[buffers[1]].state == kBufferAcked ||
 382       buffer_state_[buffers[2]].state == kBufferAcked) {
 383     undroppable_frames_ = 0;
 384     return buffers[0];
 385   }
 386
 387   // Rule 3
 388   for (int i = 0; i < kNumberOfVp8VideoBuffers; i++) {
 389     if (buffer_state_[buffers[i]].state == kBufferSent &&
 390         IsOlderFrameId(buffer_state_[buffers[i]].frame_id,
 391                        last_acked_frame_id_)) {
 392       undroppable_frames_ = 0;
 393       return buffers[i];
 394     }
 395   }
 396
 397   // Rule 4
 398   if (undroppable_frames_ >= 3) {
 399     undroppable_frames_ = 0;
 400     return kNoBuffer;
 401   } else {
 402     undroppable_frames_++;
 403     return buffers[kNumberOfVp8VideoBuffers - 1];
 404   }
 405 }
 406
 407 void Vp8Encoder::GetCodecUpdateFlags(Vp8Buffers buffer_to_update,
 408                                      vpx_codec_flags_t* flags) {
 409   if (!use_multiple_video_buffers_)
 410     return;
 411
 412   // Update at most one buffer, except for key-frames.
 413   switch (buffer_to_update) {
 414     case kAltRefBuffer:
 415       *flags |= VP8_EFLAG_NO_UPD_GF;
 416       *flags |= VP8_EFLAG_NO_UPD_LAST;
 417       break;
 418     case kLastBuffer:
 419       *flags |= VP8_EFLAG_NO_UPD_GF;
 420       *flags |= VP8_EFLAG_NO_UPD_ARF;
 421       break;
 422     case kGoldenBuffer:
 423       *flags |= VP8_EFLAG_NO_UPD_ARF;
 424       *flags |= VP8_EFLAG_NO_UPD_LAST;
 425       break;
 426     case kNoBuffer:
 427       *flags |= VP8_EFLAG_NO_UPD_ARF;
 428       *flags |= VP8_EFLAG_NO_UPD_GF;
 429       *flags |= VP8_EFLAG_NO_UPD_LAST;
 430       *flags |= VP8_EFLAG_NO_UPD_ENTROPY;
 431       break;
 432   }
 433 }
 434
 435 void Vp8Encoder::UpdateRates(uint32 new_bitrate) {
 436   DCHECK(thread_checker_.CalledOnValidThread());
 437
 438   if (!is_initialized())
 439     return;
 440
 441   uint32 new_bitrate_kbit = new_bitrate / 1000;
 442   if (config_.rc_target_bitrate == new_bitrate_kbit)
 443     return;
 444
 445   config_.rc_target_bitrate = bitrate_kbit_ = new_bitrate_kbit;
 446
 447   // Update encoder context.
 448   if (vpx_codec_enc_config_set(&encoder_, &config_)) {
 449     NOTREACHED() << "Invalid return value";
 450   }
 451
 452   VLOG(1) << "VP8 new rc_target_bitrate: " << new_bitrate_kbit << " kbps";
 453 }
 454
 455 void Vp8Encoder::LatestFrameIdToReference(uint32 frame_id) {
 456   DCHECK(thread_checker_.CalledOnValidThread());
 457   if (!use_multiple_video_buffers_)
 458     return;
 459
 460   VLOG(2) << "VP8 ok to reference frame:" << static_cast<int>(frame_id);
 461   for (int i = 0; i < kNumberOfVp8VideoBuffers; ++i) {
 462     if (frame_id == buffer_state_[i].frame_id) {
 463       buffer_state_[i].state = kBufferAcked;
 464       break;
 465     }
 466   }
 467   if (IsOlderFrameId(last_acked_frame_id_, frame_id)) {
 468     last_acked_frame_id_ = frame_id;
 469   }
 470 }
 471
 472 void Vp8Encoder::GenerateKeyFrame() {
 473   DCHECK(thread_checker_.CalledOnValidThread());
 474   key_frame_requested_ = true;
 475 }
 476
 477 }  // namespace cast
 478 }  // namespace media