Always call aom_codec_encode for every spatial layer in the libaom AV1 encoder wrapper.
Bug: none
Change-Id: I8556c4ba14393b958f4012fe9942af5523aae356
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/236341
Reviewed-by: Marco Paniconi <marpan@google.com>
Reviewed-by: Jerome Jiang <jianj@google.com>
Commit-Queue: Philip Eliasson <philipel@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35280}
diff --git a/modules/video_coding/codecs/av1/libaom_av1_encoder.cc b/modules/video_coding/codecs/av1/libaom_av1_encoder.cc
index 9c7cc89..5990772 100644
--- a/modules/video_coding/codecs/av1/libaom_av1_encoder.cc
+++ b/modules/video_coding/codecs/av1/libaom_av1_encoder.cc
@@ -616,20 +616,38 @@
const uint32_t duration =
kRtpTicksPerSecond / static_cast<float>(encoder_settings_.maxFramerate);
- for (size_t i = 0; i < layer_frames.size(); ++i) {
- ScalableVideoController::LayerFrameConfig& layer_frame = layer_frames[i];
- const bool end_of_picture = i == layer_frames.size() - 1;
+ const size_t num_spatial_layers =
+ svc_params_ ? svc_params_->number_spatial_layers : 1;
+ size_t next_layer_frame_index = 0;
+ for (size_t i = 0; i < num_spatial_layers; ++i) {
+ // The libaom AV1 encoder requires that `aom_codec_encode` is called for
+ // every spatial layer, even if the configured bitrate for that layer is
+ // zero. For zero bitrate spatial layers no frames will be produced.
+ absl::optional<ScalableVideoController::LayerFrameConfig>
+ non_encoded_layer_frame;
+ ScalableVideoController::LayerFrameConfig* layer_frame;
+ if (next_layer_frame_index < layer_frames.size() &&
+ layer_frames[next_layer_frame_index].SpatialId() ==
+ static_cast<int>(i)) {
+ layer_frame = &layer_frames[next_layer_frame_index];
+ next_layer_frame_index++;
+ } else {
+ // For layers that are not encoded only the spatial id matters.
+ non_encoded_layer_frame.emplace().S(i);
+ layer_frame = &*non_encoded_layer_frame;
+ }
+ const bool end_of_picture = (next_layer_frame_index == layer_frames.size());
aom_enc_frame_flags_t flags =
- layer_frame.IsKeyframe() ? AOM_EFLAG_FORCE_KF : 0;
+ layer_frame->IsKeyframe() ? AOM_EFLAG_FORCE_KF : 0;
if (SvcEnabled()) {
- SetSvcLayerId(layer_frame);
- SetSvcRefFrameConfig(layer_frame);
+ SetSvcLayerId(*layer_frame);
+ SetSvcRefFrameConfig(*layer_frame);
aom_codec_err_t ret =
aom_codec_control(&ctx_, AV1E_SET_ERROR_RESILIENT_MODE,
- layer_frame.TemporalId() > 0 ? 1 : 0);
+ layer_frame->TemporalId() > 0 ? 1 : 0);
if (ret != AOM_CODEC_OK) {
RTC_LOG(LS_WARNING) << "LibaomAv1Encoder::Encode returned " << ret
<< " on control AV1E_SET_ERROR_RESILIENT_MODE.";
@@ -646,6 +664,10 @@
return WEBRTC_VIDEO_CODEC_ERROR;
}
+ if (non_encoded_layer_frame) {
+ continue;
+ }
+
// Get encoded image data.
EncodedImage encoded_image;
aom_codec_iter_t iter = nullptr;
@@ -663,9 +685,10 @@
/*size=*/pkt->data.frame.sz));
if ((pkt->data.frame.flags & AOM_EFLAG_FORCE_KF) != 0) {
- layer_frame.Keyframe();
+ layer_frame->Keyframe();
}
- encoded_image._frameType = layer_frame.IsKeyframe()
+
+ encoded_image._frameType = layer_frame->IsKeyframe()
? VideoFrameType::kVideoFrameKey
: VideoFrameType::kVideoFrameDelta;
encoded_image.SetTimestamp(frame.timestamp());
@@ -675,8 +698,8 @@
// If encoded image width/height info are added to aom_codec_cx_pkt_t,
// use those values in lieu of the values in frame.
if (svc_params_) {
- int n = svc_params_->scaling_factor_num[layer_frame.SpatialId()];
- int d = svc_params_->scaling_factor_den[layer_frame.SpatialId()];
+ int n = svc_params_->scaling_factor_num[layer_frame->SpatialId()];
+ int d = svc_params_->scaling_factor_den[layer_frame->SpatialId()];
encoded_image._encodedWidth = cfg_.g_w * n / d;
encoded_image._encodedHeight = cfg_.g_h * n / d;
} else {
@@ -702,9 +725,9 @@
CodecSpecificInfo codec_specific_info;
codec_specific_info.codecType = kVideoCodecAV1;
codec_specific_info.end_of_picture = end_of_picture;
- bool is_keyframe = layer_frame.IsKeyframe();
+ bool is_keyframe = layer_frame->IsKeyframe();
codec_specific_info.generic_frame_info =
- svc_controller_->OnEncodeDone(std::move(layer_frame));
+ svc_controller_->OnEncodeDone(std::move(*layer_frame));
if (is_keyframe && codec_specific_info.generic_frame_info) {
codec_specific_info.template_structure =
svc_controller_->DependencyStructure();
@@ -747,6 +770,7 @@
svc_controller_->OnRatesUpdated(parameters.bitrate);
cfg_.rc_target_bitrate = parameters.bitrate.get_sum_kbps();
+ aom_codec_err_t error_code = aom_codec_enc_config_set(&ctx_, &cfg_);
if (SvcEnabled()) {
for (int sid = 0; sid < svc_params_->number_spatial_layers; ++sid) {
@@ -769,9 +793,6 @@
// Set frame rate to closest integer value.
encoder_settings_.maxFramerate =
static_cast<uint32_t>(parameters.framerate_fps + 0.5);
-
- // Update encoder context.
- aom_codec_err_t error_code = aom_codec_enc_config_set(&ctx_, &cfg_);
if (error_code != AOM_CODEC_OK) {
RTC_LOG(LS_WARNING) << "Error configuring encoder, error code: "
<< error_code;
