|
|
ba13db2 |
diff -Nrup -U 8 a/media/webrtc/trunk/webrtc/modules/video_coding/codecs/vp9/vp9_impl.cc b/media/webrtc/trunk/webrtc/modules/video_coding/codecs/vp9/vp9_impl.cc
|
|
|
ba13db2 |
--- a/media/webrtc/trunk/webrtc/modules/video_coding/codecs/vp9/vp9_impl.cc 2020-02-18 02:37:58.000000000 +0300
|
|
|
ba13db2 |
+++ b/media/webrtc/trunk/webrtc/modules/video_coding/codecs/vp9/vp9_impl.cc 2020-03-02 00:34:19.000000000 +0300
|
|
|
ba13db2 |
@@ -79,17 +79,19 @@ VP9EncoderImpl::VP9EncoderImpl()
|
|
|
ba13db2 |
num_temporal_layers_(0),
|
|
|
ba13db2 |
num_spatial_layers_(0),
|
|
|
ba13db2 |
num_cores_(0),
|
|
|
ba13db2 |
is_flexible_mode_(false),
|
|
|
ba13db2 |
frames_encoded_(0),
|
|
|
ba13db2 |
// Use two spatial when screensharing with flexible mode.
|
|
|
ba13db2 |
spatial_layer_(new ScreenshareLayersVP9(2)) {
|
|
|
ba13db2 |
memset(&codec_, 0, sizeof(codec_));
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
memset(&svc_params_, 0, sizeof(vpx_svc_extra_cfg_t));
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
uint32_t seed = rtc::Time32();
|
|
|
ba13db2 |
srand(seed);
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
VP9EncoderImpl::~VP9EncoderImpl() {
|
|
|
ba13db2 |
Release();
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
@@ -115,55 +117,66 @@ int VP9EncoderImpl::Release() {
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
inited_ = false;
|
|
|
ba13db2 |
return WEBRTC_VIDEO_CODEC_OK;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
bool VP9EncoderImpl::ExplicitlyConfiguredSpatialLayers() const {
|
|
|
ba13db2 |
// We check target_bitrate_bps of the 0th layer to see if the spatial layers
|
|
|
ba13db2 |
// (i.e. bitrates) were explicitly configured.
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
return num_spatial_layers_ > 1 &&
|
|
|
ba13db2 |
codec_.spatialLayers[0].target_bitrate_bps > 0;
|
|
|
ba13db2 |
+#else
|
|
|
ba13db2 |
+ return false;
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
bool VP9EncoderImpl::SetSvcRates() {
|
|
|
ba13db2 |
uint8_t i = 0;
|
|
|
ba13db2 |
|
|
|
ba13db2 |
if (ExplicitlyConfiguredSpatialLayers()) {
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
if (num_temporal_layers_ > 1) {
|
|
|
ba13db2 |
LOG(LS_ERROR) << "Multiple temporal layers when manually specifying "
|
|
|
ba13db2 |
"spatial layers not implemented yet!";
|
|
|
ba13db2 |
return false;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
int total_bitrate_bps = 0;
|
|
|
ba13db2 |
for (i = 0; i < num_spatial_layers_; ++i)
|
|
|
ba13db2 |
total_bitrate_bps += codec_.spatialLayers[i].target_bitrate_bps;
|
|
|
ba13db2 |
// If total bitrate differs now from what has been specified at the
|
|
|
ba13db2 |
// beginning, update the bitrates in the same ratio as before.
|
|
|
ba13db2 |
for (i = 0; i < num_spatial_layers_; ++i) {
|
|
|
ba13db2 |
config_->ss_target_bitrate[i] = config_->layer_target_bitrate[i] =
|
|
|
ba13db2 |
static_cast<int>(static_cast<int64_t>(config_->rc_target_bitrate) *
|
|
|
ba13db2 |
codec_.spatialLayers[i].target_bitrate_bps /
|
|
|
ba13db2 |
total_bitrate_bps);
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
} else {
|
|
|
ba13db2 |
float rate_ratio[VPX_MAX_LAYERS] = {0};
|
|
|
ba13db2 |
float total = 0;
|
|
|
ba13db2 |
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
for (i = 0; i < num_spatial_layers_; ++i) {
|
|
|
ba13db2 |
if (svc_params_.scaling_factor_num[i] <= 0 ||
|
|
|
ba13db2 |
svc_params_.scaling_factor_den[i] <= 0) {
|
|
|
ba13db2 |
LOG(LS_ERROR) << "Scaling factors not specified!";
|
|
|
ba13db2 |
return false;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
rate_ratio[i] =
|
|
|
ba13db2 |
static_cast<float>(svc_params_.scaling_factor_num[i]) /
|
|
|
ba13db2 |
svc_params_.scaling_factor_den[i];
|
|
|
ba13db2 |
total += rate_ratio[i];
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
+#else
|
|
|
ba13db2 |
+ rate_ratio[0] = 1;
|
|
|
ba13db2 |
+ total = 1;
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
|
|
|
ba13db2 |
for (i = 0; i < num_spatial_layers_; ++i) {
|
|
|
ba13db2 |
config_->ss_target_bitrate[i] = static_cast<unsigned int>(
|
|
|
ba13db2 |
config_->rc_target_bitrate * rate_ratio[i] / total);
|
|
|
ba13db2 |
if (num_temporal_layers_ == 1) {
|
|
|
ba13db2 |
config_->layer_target_bitrate[i] = config_->ss_target_bitrate[i];
|
|
|
ba13db2 |
} else if (num_temporal_layers_ == 2) {
|
|
|
ba13db2 |
config_->layer_target_bitrate[i * num_temporal_layers_] =
|
|
|
ba13db2 |
@@ -390,16 +403,17 @@ int VP9EncoderImpl::NumberOfThreads(int
|
|
|
ba13db2 |
return 2;
|
|
|
ba13db2 |
} else {
|
|
|
ba13db2 |
// 1 thread less than VGA.
|
|
|
ba13db2 |
return 1;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
int VP9EncoderImpl::InitAndSetControlSettings(const VideoCodec* inst) {
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
// Set QP-min/max per spatial and temporal layer.
|
|
|
ba13db2 |
int tot_num_layers = num_spatial_layers_ * num_temporal_layers_;
|
|
|
ba13db2 |
for (int i = 0; i < tot_num_layers; ++i) {
|
|
|
ba13db2 |
svc_params_.max_quantizers[i] = config_->rc_max_quantizer;
|
|
|
ba13db2 |
svc_params_.min_quantizers[i] = config_->rc_min_quantizer;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
config_->ss_number_layers = num_spatial_layers_;
|
|
|
ba13db2 |
if (ExplicitlyConfiguredSpatialLayers()) {
|
|
|
ba13db2 |
@@ -413,37 +427,41 @@ int VP9EncoderImpl::InitAndSetControlSet
|
|
|
ba13db2 |
for (int i = num_spatial_layers_ - 1; i >= 0; --i) {
|
|
|
ba13db2 |
// 1:2 scaling in each dimension.
|
|
|
ba13db2 |
svc_params_.scaling_factor_num[i] = scaling_factor_num;
|
|
|
ba13db2 |
svc_params_.scaling_factor_den[i] = 256;
|
|
|
ba13db2 |
if (codec_.mode != kScreensharing)
|
|
|
ba13db2 |
scaling_factor_num /= 2;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
|
|
|
ba13db2 |
if (!SetSvcRates()) {
|
|
|
ba13db2 |
return WEBRTC_VIDEO_CODEC_ERR_PARAMETER;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
if (vpx_codec_enc_init(encoder_, vpx_codec_vp9_cx(), config_, 0)) {
|
|
|
ba13db2 |
return WEBRTC_VIDEO_CODEC_UNINITIALIZED;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
vpx_codec_control(encoder_, VP8E_SET_CPUUSED, cpu_speed_);
|
|
|
ba13db2 |
vpx_codec_control(encoder_, VP8E_SET_MAX_INTRA_BITRATE_PCT,
|
|
|
ba13db2 |
rc_max_intra_target_);
|
|
|
ba13db2 |
vpx_codec_control(encoder_, VP9E_SET_AQ_MODE,
|
|
|
ba13db2 |
inst->VP9().adaptiveQpMode ? 3 : 0);
|
|
|
ba13db2 |
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
vpx_codec_control(
|
|
|
ba13db2 |
encoder_, VP9E_SET_SVC,
|
|
|
ba13db2 |
(num_temporal_layers_ > 1 || num_spatial_layers_ > 1) ? 1 : 0);
|
|
|
ba13db2 |
if (num_temporal_layers_ > 1 || num_spatial_layers_ > 1) {
|
|
|
ba13db2 |
vpx_codec_control(encoder_, VP9E_SET_SVC_PARAMETERS,
|
|
|
ba13db2 |
&svc_params_);
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
+
|
|
|
ba13db2 |
// Register callback for getting each spatial layer.
|
|
|
ba13db2 |
vpx_codec_priv_output_cx_pkt_cb_pair_t cbp = {
|
|
|
ba13db2 |
VP9EncoderImpl::EncoderOutputCodedPacketCallback,
|
|
|
ba13db2 |
reinterpret_cast<void*>(this)};
|
|
|
ba13db2 |
vpx_codec_control(encoder_, VP9E_REGISTER_CX_CALLBACK,
|
|
|
ba13db2 |
reinterpret_cast<void*>(&cbp));
|
|
|
ba13db2 |
|
|
|
ba13db2 |
// Control function to set the number of column tiles in encoding a frame, in
|
|
|
ba13db2 |
@@ -527,16 +545,17 @@ int VP9EncoderImpl::Encode(const VideoFr
|
|
|
ba13db2 |
|
|
|
ba13db2 |
vpx_enc_frame_flags_t flags = 0;
|
|
|
ba13db2 |
bool send_keyframe = (frame_type == kVideoFrameKey);
|
|
|
ba13db2 |
if (send_keyframe) {
|
|
|
ba13db2 |
// Key frame request from caller.
|
|
|
ba13db2 |
flags = VPX_EFLAG_FORCE_KF;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
if (is_flexible_mode_) {
|
|
|
ba13db2 |
SuperFrameRefSettings settings;
|
|
|
ba13db2 |
|
|
|
ba13db2 |
// These structs are copied when calling vpx_codec_control,
|
|
|
ba13db2 |
// therefore it is ok for them to go out of scope.
|
|
|
ba13db2 |
vpx_svc_ref_frame_config enc_layer_conf;
|
|
|
ba13db2 |
vpx_svc_layer_id layer_id;
|
|
|
ba13db2 |
|
|
|
ba13db2 |
@@ -548,16 +567,17 @@ int VP9EncoderImpl::Encode(const VideoFr
|
|
|
ba13db2 |
send_keyframe);
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
enc_layer_conf = GenerateRefsAndFlags(settings);
|
|
|
ba13db2 |
layer_id.temporal_layer_id = 0;
|
|
|
ba13db2 |
layer_id.spatial_layer_id = settings.start_layer;
|
|
|
ba13db2 |
vpx_codec_control(encoder_, VP9E_SET_SVC_LAYER_ID, &layer_id);
|
|
|
ba13db2 |
vpx_codec_control(encoder_, VP9E_SET_SVC_REF_FRAME_CONFIG, &enc_layer_conf);
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
|
|
|
ba13db2 |
assert(codec_.maxFramerate > 0);
|
|
|
ba13db2 |
uint32_t duration = 90000 / codec_.maxFramerate;
|
|
|
ba13db2 |
if (vpx_codec_encode(encoder_, raw_, timestamp_, duration, flags,
|
|
|
ba13db2 |
VPX_DL_REALTIME)) {
|
|
|
ba13db2 |
return WEBRTC_VIDEO_CODEC_ERROR;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
timestamp_ += duration;
|
|
|
ba13db2 |
@@ -686,29 +706,31 @@ void VP9EncoderImpl::PopulateCodecSpecif
|
|
|
ba13db2 |
vp9_info->p_diff[i] = p_diff_[layer_id.spatial_layer_id][i];
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
} else {
|
|
|
ba13db2 |
vp9_info->gof_idx =
|
|
|
ba13db2 |
static_cast<uint8_t>(frames_since_kf_ % gof_.num_frames_in_gof);
|
|
|
ba13db2 |
vp9_info->temporal_up_switch = gof_.temporal_up_switch[vp9_info->gof_idx];
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
if (vp9_info->ss_data_available) {
|
|
|
ba13db2 |
vp9_info->spatial_layer_resolution_present = true;
|
|
|
ba13db2 |
for (size_t i = 0; i < vp9_info->num_spatial_layers; ++i) {
|
|
|
ba13db2 |
vp9_info->width[i] = codec_.width *
|
|
|
ba13db2 |
svc_params_.scaling_factor_num[i] /
|
|
|
ba13db2 |
svc_params_.scaling_factor_den[i];
|
|
|
ba13db2 |
vp9_info->height[i] = codec_.height *
|
|
|
ba13db2 |
svc_params_.scaling_factor_num[i] /
|
|
|
ba13db2 |
svc_params_.scaling_factor_den[i];
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
- if (!vp9_info->flexible_mode) {
|
|
|
ba13db2 |
- vp9_info->gof.CopyGofInfoVP9(gof_);
|
|
|
ba13db2 |
- }
|
|
|
ba13db2 |
+ }
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
+ if (!vp9_info->flexible_mode) {
|
|
|
ba13db2 |
+ vp9_info->gof.CopyGofInfoVP9(gof_);
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
int VP9EncoderImpl::GetEncodedLayerFrame(const vpx_codec_cx_pkt* pkt) {
|
|
|
ba13db2 |
RTC_DCHECK_EQ(pkt->kind, VPX_CODEC_CX_FRAME_PKT);
|
|
|
ba13db2 |
|
|
|
ba13db2 |
if (pkt->data.frame.sz > encoded_image_._size) {
|
|
|
ba13db2 |
delete[] encoded_image_._buffer;
|
|
|
ba13db2 |
@@ -756,16 +778,17 @@ int VP9EncoderImpl::GetEncodedLayerFrame
|
|
|
ba13db2 |
vpx_codec_control(encoder_, VP8E_GET_LAST_QUANTIZER, &qp);
|
|
|
ba13db2 |
encoded_image_.qp_ = qp;
|
|
|
ba13db2 |
encoded_complete_callback_->OnEncodedImage(encoded_image_, &codec_specific,
|
|
|
ba13db2 |
&frag_info);
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
return WEBRTC_VIDEO_CODEC_OK;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
vpx_svc_ref_frame_config VP9EncoderImpl::GenerateRefsAndFlags(
|
|
|
ba13db2 |
const SuperFrameRefSettings& settings) {
|
|
|
ba13db2 |
static const vpx_enc_frame_flags_t kAllFlags =
|
|
|
ba13db2 |
VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_LAST |
|
|
|
ba13db2 |
VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_GF;
|
|
|
ba13db2 |
vpx_svc_ref_frame_config sf_conf = {};
|
|
|
ba13db2 |
if (settings.is_keyframe) {
|
|
|
ba13db2 |
// Used later on to make sure we don't make any invalid references.
|
|
|
ba13db2 |
@@ -857,16 +880,17 @@ vpx_svc_ref_frame_config VP9EncoderImpl:
|
|
|
ba13db2 |
buffer_updated_at_frame_[updated_buffer] = frames_encoded_;
|
|
|
ba13db2 |
sf_conf.frame_flags[layer_idx] = layer_flags;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
++frames_encoded_;
|
|
|
ba13db2 |
return sf_conf;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
|
|
|
ba13db2 |
int VP9EncoderImpl::SetChannelParameters(uint32_t packet_loss, int64_t rtt) {
|
|
|
ba13db2 |
return WEBRTC_VIDEO_CODEC_OK;
|
|
|
ba13db2 |
}
|
|
|
ba13db2 |
|
|
|
ba13db2 |
int VP9EncoderImpl::RegisterEncodeCompleteCallback(
|
|
|
ba13db2 |
EncodedImageCallback* callback) {
|
|
|
ba13db2 |
encoded_complete_callback_ = callback;
|
|
|
ba13db2 |
diff -Nrup -U 8 a/media/webrtc/trunk/webrtc/modules/video_coding/codecs/vp9/vp9_impl.h b/media/webrtc/trunk/webrtc/modules/video_coding/codecs/vp9/vp9_impl.h
|
|
|
ba13db2 |
--- a/media/webrtc/trunk/webrtc/modules/video_coding/codecs/vp9/vp9_impl.h 2020-02-18 02:37:58.000000000 +0300
|
|
|
ba13db2 |
+++ b/media/webrtc/trunk/webrtc/modules/video_coding/codecs/vp9/vp9_impl.h 2020-03-02 00:32:05.000000000 +0300
|
|
|
ba13db2 |
@@ -77,24 +77,26 @@ class VP9EncoderImpl : public VP9Encoder
|
|
|
ba13db2 |
|
|
|
ba13db2 |
void PopulateCodecSpecific(CodecSpecificInfo* codec_specific,
|
|
|
ba13db2 |
const vpx_codec_cx_pkt& pkt,
|
|
|
ba13db2 |
uint32_t timestamp);
|
|
|
ba13db2 |
|
|
|
ba13db2 |
bool ExplicitlyConfiguredSpatialLayers() const;
|
|
|
ba13db2 |
bool SetSvcRates();
|
|
|
ba13db2 |
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
// Used for flexible mode to set the flags and buffer references used
|
|
|
ba13db2 |
// by the encoder. Also calculates the references used by the RTP
|
|
|
ba13db2 |
// packetizer.
|
|
|
ba13db2 |
//
|
|
|
ba13db2 |
// Has to be called for every frame (keyframes included) to update the
|
|
|
ba13db2 |
// state used to calculate references.
|
|
|
ba13db2 |
vpx_svc_ref_frame_config GenerateRefsAndFlags(
|
|
|
ba13db2 |
const SuperFrameRefSettings& settings);
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
|
|
|
ba13db2 |
virtual int GetEncodedLayerFrame(const vpx_codec_cx_pkt* pkt);
|
|
|
ba13db2 |
|
|
|
ba13db2 |
// Callback function for outputting packets per spatial layer.
|
|
|
ba13db2 |
static void EncoderOutputCodedPacketCallback(vpx_codec_cx_pkt* pkt,
|
|
|
ba13db2 |
void* user_data);
|
|
|
ba13db2 |
|
|
|
ba13db2 |
// Determine maximum target for Intra frames
|
|
|
ba13db2 |
@@ -111,17 +113,19 @@ class VP9EncoderImpl : public VP9Encoder
|
|
|
ba13db2 |
bool inited_;
|
|
|
ba13db2 |
int64_t timestamp_;
|
|
|
ba13db2 |
uint16_t picture_id_;
|
|
|
ba13db2 |
int cpu_speed_;
|
|
|
ba13db2 |
uint32_t rc_max_intra_target_;
|
|
|
ba13db2 |
vpx_codec_ctx_t* encoder_;
|
|
|
ba13db2 |
vpx_codec_enc_cfg_t* config_;
|
|
|
ba13db2 |
vpx_image_t* raw_;
|
|
|
ba13db2 |
+#ifdef LIBVPX_SVC
|
|
|
ba13db2 |
vpx_svc_extra_cfg_t svc_params_;
|
|
|
ba13db2 |
+#endif
|
|
|
ba13db2 |
const VideoFrame* input_image_;
|
|
|
ba13db2 |
GofInfoVP9 gof_; // Contains each frame's temporal information for
|
|
|
ba13db2 |
// non-flexible mode.
|
|
|
ba13db2 |
uint8_t tl0_pic_idx_; // Only used in non-flexible mode.
|
|
|
ba13db2 |
size_t frames_since_kf_;
|
|
|
ba13db2 |
uint8_t num_temporal_layers_;
|
|
|
ba13db2 |
uint8_t num_spatial_layers_;
|
|
|
ba13db2 |
uint8_t num_cores_;
|