| /* |
| * Copyright (c) 2013 The WebRTC project authors. All Rights Reserved. |
| * |
| * Use of this source code is governed by a BSD-style license |
| * that can be found in the LICENSE file in the root of the source |
| * tree. An additional intellectual property rights grant can be found |
| * in the file PATENTS. All contributing project authors may |
| * be found in the AUTHORS file in the root of the source tree. |
| */ |
| |
| #include "test/fake_encoder.h" |
| |
| #include <string.h> |
| |
| #include <algorithm> |
| #include <memory> |
| |
| #include "api/video_codecs/vp8_temporal_layers.h" |
| #include "common_types.h" // NOLINT(build/include) |
| #include "modules/video_coding/include/video_codec_interface.h" |
| #include "rtc_base/checks.h" |
| #include "system_wrappers/include/sleep.h" |
| |
| namespace webrtc { |
| namespace test { |
| |
| const int kKeyframeSizeFactor = 5; |
| |
| // Inverse of proportion of frames assigned to each temporal layer for all |
| // possible temporal layers numbers. |
| const int kTemporalLayerRateFactor[4][4] = { |
| {1, 0, 0, 0}, // 1/1 |
| {2, 2, 0, 0}, // 1/2 + 1/2 |
| {4, 4, 2, 0}, // 1/4 + 1/4 + 1/2 |
| {8, 8, 4, 2}, // 1/8 + 1/8 + 1/4 + 1/2 |
| }; |
| |
| FakeEncoder::FakeEncoder(Clock* clock) |
| : clock_(clock), |
| callback_(nullptr), |
| configured_input_framerate_(-1), |
| max_target_bitrate_kbps_(-1), |
| pending_keyframe_(true), |
| debt_bytes_(0) { |
| // Generate some arbitrary not-all-zero data |
| for (size_t i = 0; i < sizeof(encoded_buffer_); ++i) { |
| encoded_buffer_[i] = static_cast<uint8_t>(i); |
| } |
| for (bool& used : used_layers_) { |
| used = false; |
| } |
| } |
| |
| void FakeEncoder::SetMaxBitrate(int max_kbps) { |
| RTC_DCHECK_GE(max_kbps, -1); // max_kbps == -1 disables it. |
| rtc::CritScope cs(&crit_sect_); |
| max_target_bitrate_kbps_ = max_kbps; |
| } |
| |
| int32_t FakeEncoder::InitEncode(const VideoCodec* config, |
| int32_t number_of_cores, |
| size_t max_payload_size) { |
| rtc::CritScope cs(&crit_sect_); |
| config_ = *config; |
| target_bitrate_.SetBitrate(0, 0, config_.startBitrate * 1000); |
| configured_input_framerate_ = config_.maxFramerate; |
| pending_keyframe_ = true; |
| last_frame_info_ = FrameInfo(); |
| return 0; |
| } |
| |
| int32_t FakeEncoder::Encode(const VideoFrame& input_image, |
| const CodecSpecificInfo* codec_specific_info, |
| const std::vector<FrameType>* frame_types) { |
| unsigned char max_framerate; |
| unsigned char num_simulcast_streams; |
| SimulcastStream simulcast_streams[kMaxSimulcastStreams]; |
| EncodedImageCallback* callback; |
| VideoBitrateAllocation target_bitrate; |
| int framerate; |
| VideoCodecMode mode; |
| bool keyframe; |
| { |
| rtc::CritScope cs(&crit_sect_); |
| max_framerate = config_.maxFramerate; |
| num_simulcast_streams = config_.numberOfSimulcastStreams; |
| for (int i = 0; i < num_simulcast_streams; ++i) { |
| simulcast_streams[i] = config_.simulcastStream[i]; |
| } |
| callback = callback_; |
| target_bitrate = target_bitrate_; |
| mode = config_.mode; |
| if (configured_input_framerate_ > 0) { |
| framerate = configured_input_framerate_; |
| } else { |
| framerate = max_framerate; |
| } |
| keyframe = pending_keyframe_; |
| pending_keyframe_ = false; |
| } |
| |
| FrameInfo frame_info = |
| NextFrame(frame_types, keyframe, num_simulcast_streams, target_bitrate, |
| simulcast_streams, framerate); |
| for (uint8_t i = 0; i < frame_info.layers.size(); ++i) { |
| constexpr int kMinPayLoadLength = 10; |
| if (frame_info.layers[i].size < kMinPayLoadLength) { |
| // Drop this temporal layer. |
| continue; |
| } |
| |
| CodecSpecificInfo specifics; |
| memset(&specifics, 0, sizeof(specifics)); |
| specifics.codecType = kVideoCodecGeneric; |
| std::unique_ptr<uint8_t[]> encoded_buffer( |
| new uint8_t[frame_info.layers[i].size]); |
| memcpy(encoded_buffer.get(), encoded_buffer_, frame_info.layers[i].size); |
| EncodedImage encoded(encoded_buffer.get(), frame_info.layers[i].size, |
| sizeof(encoded_buffer_)); |
| encoded.SetTimestamp(input_image.timestamp()); |
| encoded.capture_time_ms_ = input_image.render_time_ms(); |
| encoded._frameType = |
| frame_info.keyframe ? kVideoFrameKey : kVideoFrameDelta; |
| encoded._encodedWidth = simulcast_streams[i].width; |
| encoded._encodedHeight = simulcast_streams[i].height; |
| encoded.rotation_ = input_image.rotation(); |
| encoded.content_type_ = (mode == VideoCodecMode::kScreensharing) |
| ? VideoContentType::SCREENSHARE |
| : VideoContentType::UNSPECIFIED; |
| encoded.SetSpatialIndex(i); |
| if (callback->OnEncodedImage(encoded, &specifics, nullptr).error != |
| EncodedImageCallback::Result::OK) { |
| return -1; |
| } |
| } |
| return 0; |
| } |
| |
| FakeEncoder::FrameInfo FakeEncoder::NextFrame( |
| const std::vector<FrameType>* frame_types, |
| bool keyframe, |
| uint8_t num_simulcast_streams, |
| const VideoBitrateAllocation& target_bitrate, |
| SimulcastStream simulcast_streams[kMaxSimulcastStreams], |
| int framerate) { |
| FrameInfo frame_info; |
| frame_info.keyframe = keyframe; |
| |
| if (frame_types) { |
| for (FrameType frame_type : *frame_types) { |
| if (frame_type == kVideoFrameKey) { |
| frame_info.keyframe = true; |
| break; |
| } |
| } |
| } |
| |
| rtc::CritScope cs(&crit_sect_); |
| for (uint8_t i = 0; i < num_simulcast_streams; ++i) { |
| if (target_bitrate.GetBitrate(i, 0) > 0) { |
| int temporal_id = last_frame_info_.layers.size() > i |
| ? ++last_frame_info_.layers[i].temporal_id % |
| simulcast_streams[i].numberOfTemporalLayers |
| : 0; |
| frame_info.layers.emplace_back(0, temporal_id); |
| } |
| } |
| |
| if (last_frame_info_.layers.size() < frame_info.layers.size()) { |
| // A new keyframe is needed since a new layer will be added. |
| frame_info.keyframe = true; |
| } |
| |
| for (uint8_t i = 0; i < frame_info.layers.size(); ++i) { |
| FrameInfo::SpatialLayer& layer_info = frame_info.layers[i]; |
| if (frame_info.keyframe) { |
| layer_info.temporal_id = 0; |
| size_t avg_frame_size = |
| (target_bitrate.GetBitrate(i, 0) + 7) * |
| kTemporalLayerRateFactor[frame_info.layers.size() - 1][i] / |
| (8 * framerate); |
| |
| // The first frame is a key frame and should be larger. |
| // Store the overshoot bytes and distribute them over the coming frames, |
| // so that we on average meet the bitrate target. |
| debt_bytes_ += (kKeyframeSizeFactor - 1) * avg_frame_size; |
| layer_info.size = kKeyframeSizeFactor * avg_frame_size; |
| } else { |
| size_t avg_frame_size = |
| (target_bitrate.GetBitrate(i, layer_info.temporal_id) + 7) * |
| kTemporalLayerRateFactor[frame_info.layers.size() - 1][i] / |
| (8 * framerate); |
| layer_info.size = avg_frame_size; |
| if (debt_bytes_ > 0) { |
| // Pay at most half of the frame size for old debts. |
| size_t payment_size = std::min(avg_frame_size / 2, debt_bytes_); |
| debt_bytes_ -= payment_size; |
| layer_info.size -= payment_size; |
| } |
| } |
| } |
| last_frame_info_ = frame_info; |
| return frame_info; |
| } |
| |
| int32_t FakeEncoder::RegisterEncodeCompleteCallback( |
| EncodedImageCallback* callback) { |
| rtc::CritScope cs(&crit_sect_); |
| callback_ = callback; |
| return 0; |
| } |
| |
| int32_t FakeEncoder::Release() { |
| return 0; |
| } |
| |
| int32_t FakeEncoder::SetRateAllocation( |
| const VideoBitrateAllocation& rate_allocation, |
| uint32_t framerate) { |
| rtc::CritScope cs(&crit_sect_); |
| target_bitrate_ = rate_allocation; |
| int allocated_bitrate_kbps = target_bitrate_.get_sum_kbps(); |
| |
| // Scale bitrate allocation to not exceed the given max target bitrate. |
| if (max_target_bitrate_kbps_ > 0 && |
| allocated_bitrate_kbps > max_target_bitrate_kbps_) { |
| for (uint8_t spatial_idx = 0; spatial_idx < kMaxSpatialLayers; |
| ++spatial_idx) { |
| for (uint8_t temporal_idx = 0; temporal_idx < kMaxTemporalStreams; |
| ++temporal_idx) { |
| if (target_bitrate_.HasBitrate(spatial_idx, temporal_idx)) { |
| uint32_t bitrate = |
| target_bitrate_.GetBitrate(spatial_idx, temporal_idx); |
| bitrate = static_cast<uint32_t>( |
| (bitrate * int64_t{max_target_bitrate_kbps_}) / |
| allocated_bitrate_kbps); |
| target_bitrate_.SetBitrate(spatial_idx, temporal_idx, bitrate); |
| } |
| } |
| } |
| } |
| |
| configured_input_framerate_ = framerate; |
| return 0; |
| } |
| |
| const char* FakeEncoder::kImplementationName = "fake_encoder"; |
| const char* FakeEncoder::ImplementationName() const { |
| return kImplementationName; |
| } |
| |
| int FakeEncoder::GetConfiguredInputFramerate() const { |
| rtc::CritScope cs(&crit_sect_); |
| return configured_input_framerate_; |
| } |
| |
| FakeH264Encoder::FakeH264Encoder(Clock* clock) |
| : FakeEncoder(clock), callback_(nullptr), idr_counter_(0) { |
| FakeEncoder::RegisterEncodeCompleteCallback(this); |
| } |
| |
| int32_t FakeH264Encoder::RegisterEncodeCompleteCallback( |
| EncodedImageCallback* callback) { |
| rtc::CritScope cs(&local_crit_sect_); |
| callback_ = callback; |
| return 0; |
| } |
| |
| EncodedImageCallback::Result FakeH264Encoder::OnEncodedImage( |
| const EncodedImage& encoded_image, |
| const CodecSpecificInfo* codec_specific_info, |
| const RTPFragmentationHeader* fragments) { |
| const size_t kSpsSize = 8; |
| const size_t kPpsSize = 11; |
| const int kIdrFrequency = 10; |
| EncodedImageCallback* callback; |
| int current_idr_counter; |
| { |
| rtc::CritScope cs(&local_crit_sect_); |
| callback = callback_; |
| current_idr_counter = idr_counter_; |
| ++idr_counter_; |
| } |
| RTPFragmentationHeader fragmentation; |
| if (current_idr_counter % kIdrFrequency == 0 && |
| encoded_image._length > kSpsSize + kPpsSize + 1) { |
| const size_t kNumSlices = 3; |
| fragmentation.VerifyAndAllocateFragmentationHeader(kNumSlices); |
| fragmentation.fragmentationOffset[0] = 0; |
| fragmentation.fragmentationLength[0] = kSpsSize; |
| fragmentation.fragmentationOffset[1] = kSpsSize; |
| fragmentation.fragmentationLength[1] = kPpsSize; |
| fragmentation.fragmentationOffset[2] = kSpsSize + kPpsSize; |
| fragmentation.fragmentationLength[2] = |
| encoded_image._length - (kSpsSize + kPpsSize); |
| const size_t kSpsNalHeader = 0x67; |
| const size_t kPpsNalHeader = 0x68; |
| const size_t kIdrNalHeader = 0x65; |
| encoded_image._buffer[fragmentation.fragmentationOffset[0]] = kSpsNalHeader; |
| encoded_image._buffer[fragmentation.fragmentationOffset[1]] = kPpsNalHeader; |
| encoded_image._buffer[fragmentation.fragmentationOffset[2]] = kIdrNalHeader; |
| } else { |
| const size_t kNumSlices = 1; |
| fragmentation.VerifyAndAllocateFragmentationHeader(kNumSlices); |
| fragmentation.fragmentationOffset[0] = 0; |
| fragmentation.fragmentationLength[0] = encoded_image._length; |
| const size_t kNalHeader = 0x41; |
| encoded_image._buffer[fragmentation.fragmentationOffset[0]] = kNalHeader; |
| } |
| uint8_t value = 0; |
| int fragment_counter = 0; |
| for (size_t i = 0; i < encoded_image._length; ++i) { |
| if (fragment_counter == fragmentation.fragmentationVectorSize || |
| i != fragmentation.fragmentationOffset[fragment_counter]) { |
| encoded_image._buffer[i] = value++; |
| } else { |
| ++fragment_counter; |
| } |
| } |
| CodecSpecificInfo specifics; |
| memset(&specifics, 0, sizeof(specifics)); |
| specifics.codecType = kVideoCodecH264; |
| specifics.codecSpecific.H264.packetization_mode = |
| H264PacketizationMode::NonInterleaved; |
| RTC_DCHECK(callback); |
| return callback->OnEncodedImage(encoded_image, &specifics, &fragmentation); |
| } |
| |
| DelayedEncoder::DelayedEncoder(Clock* clock, int delay_ms) |
| : test::FakeEncoder(clock), delay_ms_(delay_ms) { |
| // The encoder could be created on a different thread than |
| // it is being used on. |
| sequence_checker_.Detach(); |
| } |
| |
| void DelayedEncoder::SetDelay(int delay_ms) { |
| RTC_DCHECK_CALLED_SEQUENTIALLY(&sequence_checker_); |
| delay_ms_ = delay_ms; |
| } |
| |
| int32_t DelayedEncoder::Encode(const VideoFrame& input_image, |
| const CodecSpecificInfo* codec_specific_info, |
| const std::vector<FrameType>* frame_types) { |
| RTC_DCHECK_CALLED_SEQUENTIALLY(&sequence_checker_); |
| |
| SleepMs(delay_ms_); |
| |
| return FakeEncoder::Encode(input_image, codec_specific_info, frame_types); |
| } |
| |
| MultithreadedFakeH264Encoder::MultithreadedFakeH264Encoder(Clock* clock) |
| : test::FakeH264Encoder(clock), |
| current_queue_(0), |
| queue1_(nullptr), |
| queue2_(nullptr) { |
| // The encoder could be created on a different thread than |
| // it is being used on. |
| sequence_checker_.Detach(); |
| } |
| |
| int32_t MultithreadedFakeH264Encoder::InitEncode(const VideoCodec* config, |
| int32_t number_of_cores, |
| size_t max_payload_size) { |
| RTC_DCHECK_CALLED_SEQUENTIALLY(&sequence_checker_); |
| |
| queue1_.reset(new rtc::TaskQueue("Queue 1")); |
| queue2_.reset(new rtc::TaskQueue("Queue 2")); |
| |
| return FakeH264Encoder::InitEncode(config, number_of_cores, max_payload_size); |
| } |
| |
| class MultithreadedFakeH264Encoder::EncodeTask : public rtc::QueuedTask { |
| public: |
| EncodeTask(MultithreadedFakeH264Encoder* encoder, |
| const VideoFrame& input_image, |
| const CodecSpecificInfo* codec_specific_info, |
| const std::vector<FrameType>* frame_types) |
| : encoder_(encoder), |
| input_image_(input_image), |
| codec_specific_info_(), |
| frame_types_(*frame_types) { |
| if (codec_specific_info) |
| codec_specific_info_ = *codec_specific_info; |
| } |
| |
| private: |
| bool Run() override { |
| encoder_->EncodeCallback(input_image_, &codec_specific_info_, |
| &frame_types_); |
| return true; |
| } |
| |
| MultithreadedFakeH264Encoder* const encoder_; |
| VideoFrame input_image_; |
| CodecSpecificInfo codec_specific_info_; |
| std::vector<FrameType> frame_types_; |
| }; |
| |
| int32_t MultithreadedFakeH264Encoder::Encode( |
| const VideoFrame& input_image, |
| const CodecSpecificInfo* codec_specific_info, |
| const std::vector<FrameType>* frame_types) { |
| RTC_DCHECK_CALLED_SEQUENTIALLY(&sequence_checker_); |
| |
| std::unique_ptr<rtc::TaskQueue>& queue = |
| (current_queue_++ % 2 == 0) ? queue1_ : queue2_; |
| |
| if (!queue) { |
| return WEBRTC_VIDEO_CODEC_UNINITIALIZED; |
| } |
| |
| queue->PostTask(std::unique_ptr<rtc::QueuedTask>( |
| new EncodeTask(this, input_image, codec_specific_info, frame_types))); |
| |
| return WEBRTC_VIDEO_CODEC_OK; |
| } |
| |
| int32_t MultithreadedFakeH264Encoder::EncodeCallback( |
| const VideoFrame& input_image, |
| const CodecSpecificInfo* codec_specific_info, |
| const std::vector<FrameType>* frame_types) { |
| return FakeH264Encoder::Encode(input_image, codec_specific_info, frame_types); |
| } |
| |
| int32_t MultithreadedFakeH264Encoder::Release() { |
| RTC_DCHECK_CALLED_SEQUENTIALLY(&sequence_checker_); |
| |
| queue1_.reset(); |
| queue2_.reset(); |
| |
| return FakeH264Encoder::Release(); |
| } |
| |
| } // namespace test |
| } // namespace webrtc |