ref: 6be2b750b8968a0b19180a6474864c0b6a41db13
parent: 37e5325b9f985e7ce1cd8a128b9bebafded60f40
author: Marco Paniconi <marpan@google.com>
date: Sun Feb 2 09:30:43 EST 2014
Layer based rate control for CBR mode. This patch adds a buffer-based rate control for temporal layers, under CBR mode. Added vpx_temporal_scalable_patters.c encoder for testing temporal layers, for both vp9 and vp8 (replaces the old vp8_scalable_patterns). Updated datarate unittest with tests for temporal layer rate-targeting. Change-Id: I9cb6cce2494390ae6096ee17774af7fb9308bde7
--- a/examples.mk
+++ b/examples.mk
@@ -64,6 +64,11 @@
vp9_spatial_scalable_encoder.SRCS += tools_common.c tools_common.h
vp9_spatial_scalable_encoder.GUID = 4A38598D-627D-4505-9C7B-D4020C84100D
vp9_spatial_scalable_encoder.DESCRIPTION = Spatial Scalable Encoder
+UTILS-$(CONFIG_ENCODERS) += vpx_temporal_scalable_patterns.c
+vpx_temporal_scalable_patterns.SRCS += ivfenc.c ivfenc.h
+vpx_temporal_scalable_patterns.SRCS += tools_common.c tools_common.h
+vpx_temporal_scalable_patterns.GUID = B18C08F2-A439-4502-A78E-849BE3D60947
+vpx_temporal_scalable_patterns.DESCRIPTION = Temporal Scalability Encoder
ifeq ($(CONFIG_SHARED),no)
UTILS-$(CONFIG_VP9_ENCODER) += resize_util.c
--- a/test/datarate_test.cc
+++ b/test/datarate_test.cc
@@ -200,21 +200,102 @@
frame_number_ = 0;
first_drop_ = 0;
num_drops_ = 0;
- bits_total_ = 0;
- duration_ = 0.0;
+ // For testing up to 3 layers.
+ for (int i = 0; i < 3; ++i) {+ bits_total_[i] = 0;
+ }
}
+ //
+ // Frame flags and layer id for temporal layers.
+ //
+
+ // For two layers, test pattern is:
+ // 1 3
+ // 0 2 .....
+ // For three layers, test pattern is:
+ // 1 3 5 7
+ // 2 6
+ // 0 4 ....
+ // LAST is always update on base/layer 0, GOLDEN is updated on layer 1.
+ // For this 3 layer example, the 2rd enhancement layer (layer 2) does not
+ // update any reference frames.
+ int SetFrameFlags(int frame_num, int num_temp_layers) {+ int frame_flags = 0;
+ if (num_temp_layers == 2) {+ if (frame_num % 2 == 0) {+ // Layer 0: predict from L and ARF, update L.
+ frame_flags = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
+ VP8_EFLAG_NO_UPD_ARF;
+ } else {+ // Layer 1: predict from L, G and ARF, and update G.
+ frame_flags = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_ENTROPY;
+ }
+ } else if (num_temp_layers == 3) {+ if (frame_num % 4 == 0) {+ // Layer 0: predict from L and ARF; update L.
+ frame_flags = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_REF_GF;
+ } else if ((frame_num - 2) % 4 == 0) {+ // Layer 1: predict from L, G, ARF; update G.
+ frame_flags = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
+ } else if ((frame_num - 1) % 2 == 0) {+ // Layer 2: predict from L, G, ARF; update none.
+ frame_flags = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_UPD_LAST;
+ }
+ }
+ return frame_flags;
+ }
+
+ int SetLayerId(int frame_num, int num_temp_layers) {+ int layer_id = 0;
+ if (num_temp_layers == 2) {+ if (frame_num % 2 == 0) {+ layer_id = 0;
+ } else {+ layer_id = 1;
+ }
+ } else if (num_temp_layers == 3) {+ if (frame_num % 4 == 0) {+ layer_id = 0;
+ } else if ((frame_num - 2) % 4 == 0) {+ layer_id = 1;
+ } else if ((frame_num - 1) % 2 == 0) {+ layer_id = 2;
+ }
+ }
+ return layer_id;
+ }
+
virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
- ::libvpx_test::Encoder *encoder) {+ ::libvpx_test::Encoder *encoder) { if (video->frame() == 1) {encoder->Control(VP8E_SET_CPUUSED, set_cpu_used_);
}
+ if (cfg_.ts_number_layers > 1) {+ if (video->frame() == 1) {+ encoder->Control(VP9E_SET_SVC, 1);
+ }
+ vpx_svc_layer_id_t layer_id = {0, 0};+ layer_id.spatial_layer_id = 0;
+ frame_flags_ = SetFrameFlags(video->frame(), cfg_.ts_number_layers);
+ layer_id.temporal_layer_id = SetLayerId(video->frame(),
+ cfg_.ts_number_layers);
+ if (video->frame() > 0) {+ encoder->Control(VP9E_SET_SVC_LAYER_ID, &layer_id);
+ }
+ }
const vpx_rational_t tb = video->timebase();
timebase_ = static_cast<double>(tb.num) / tb.den;
duration_ = 0;
}
+
virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {+ int layer = SetLayerId(frame_number_, cfg_.ts_number_layers);
+
// Time since last timestamp = duration.
vpx_codec_pts_t duration = pkt->data.frame.pts - last_pts_;
@@ -227,8 +308,13 @@
<< pkt->data.frame.pts;
const size_t frame_size_in_bits = pkt->data.frame.sz * 8;
- bits_total_ += frame_size_in_bits;
+ // Update the total encoded bits. For temporal layers, update the cumulative
+ // encoded bits per layer.
+ for (int i = layer; i < static_cast<int>(cfg_.ts_number_layers); ++i) {+ bits_total_[i] += frame_size_in_bits;
+ }
+
// If first drop not set and we have a drop set it to this time.
if (!first_drop_ && duration > 1)
first_drop_ = last_pts_ + 1;
@@ -244,10 +330,13 @@
}
virtual void EndPassHook(void) {- if (bits_total_) {+ for (int layer = 0; layer < static_cast<int>(cfg_.ts_number_layers);
+ layer++) {duration_ = (last_pts_ + 1) * timebase_;
- // Effective file datarate:
- effective_datarate_ = ((bits_total_) / 1000.0) / duration_;
+ if (bits_total_[layer]) {+ // Effective file datarate:
+ effective_datarate_[layer] = (bits_total_[layer] / 1000.0) / duration_;
+ }
}
}
@@ -254,9 +343,9 @@
vpx_codec_pts_t last_pts_;
double timebase_;
int frame_number_;
- int64_t bits_total_;
+ int64_t bits_total_[3];
double duration_;
- double effective_datarate_;
+ double effective_datarate_[3];
int set_cpu_used_;
int64_t bits_in_buffer_model_;
vpx_codec_pts_t first_drop_;
@@ -272,6 +361,7 @@
cfg_.rc_min_quantizer = 0;
cfg_.rc_max_quantizer = 63;
cfg_.rc_end_usage = VPX_CBR;
+ cfg_.g_lag_in_frames = 0;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,30, 1, 0, 140);
@@ -279,12 +369,10 @@
cfg_.rc_target_bitrate = i;
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
- ASSERT_GE(static_cast<double>(cfg_.rc_target_bitrate),
- effective_datarate_ * 0.85)
- << " The datarate for the file exceeds the target by too much!";
- ASSERT_LE(static_cast<double>(cfg_.rc_target_bitrate),
- effective_datarate_ * 1.15)
- << " The datarate for the file missed the target!";
+ ASSERT_GE(effective_datarate_[0], cfg_.rc_target_bitrate * 0.85)
+ << " The datarate for the file is lower than target by too much!";
+ ASSERT_LE(effective_datarate_[0], cfg_.rc_target_bitrate * 1.15)
+ << " The datarate for the file is greater than target by too much!";
}
}
@@ -309,10 +397,10 @@
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(static_cast<double>(cfg_.rc_target_bitrate),
- effective_datarate_ * 0.85)
+ effective_datarate_[0] * 0.85)
<< " The datarate for the file exceeds the target by too much!";
ASSERT_LE(static_cast<double>(cfg_.rc_target_bitrate),
- effective_datarate_ * 1.15)
+ effective_datarate_[0] * 1.15)
<< " The datarate for the file missed the target!"
<< cfg_.rc_target_bitrate << " "<< effective_datarate_;
}
@@ -334,6 +422,7 @@
cfg_.rc_max_quantizer = 50;
cfg_.rc_end_usage = VPX_CBR;
cfg_.rc_target_bitrate = 200;
+ cfg_.g_lag_in_frames = 0;
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,30, 1, 0, 140);
@@ -345,10 +434,10 @@
cfg_.rc_dropframe_thresh = i;
ResetModel();
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
- ASSERT_GE(effective_datarate_, cfg_.rc_target_bitrate * 0.85)
- << " The datarate for the file is lower than target by too much!";
- ASSERT_LE(effective_datarate_, cfg_.rc_target_bitrate * 1.15)
- << " The datarate for the file is greater than target by too much!";
+ ASSERT_GE(effective_datarate_[0], cfg_.rc_target_bitrate * 0.85)
+ << " The datarate for the file is lower than target by too much!";
+ ASSERT_LE(effective_datarate_[0], cfg_.rc_target_bitrate * 1.15)
+ << " The datarate for the file is greater than target by too much!";
ASSERT_LE(first_drop_, last_drop)
<< " The first dropped frame for drop_thresh " << i
<< " > first dropped frame for drop_thresh "
@@ -362,6 +451,81 @@
}
}
+// Check basic rate targeting for 2 temporal layers.
+TEST_P(DatarateTestVP9, BasicRateTargeting2TemporalLayers) {+ cfg_.rc_buf_initial_sz = 500;
+ cfg_.rc_buf_optimal_sz = 500;
+ cfg_.rc_buf_sz = 1000;
+ cfg_.rc_dropframe_thresh = 1;
+ cfg_.rc_min_quantizer = 0;
+ cfg_.rc_max_quantizer = 63;
+ cfg_.rc_end_usage = VPX_CBR;
+ cfg_.g_lag_in_frames = 0;
+
+ // 2 Temporal layers, no spatial layers: Framerate decimation (2, 1).
+ cfg_.ss_number_layers = 1;
+ cfg_.ts_number_layers = 2;
+ cfg_.ts_rate_decimator[0] = 2;
+ cfg_.ts_rate_decimator[1] = 1;
+
+ ::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,+ 30, 1, 0, 200);
+ for (int i = 200; i <= 800; i += 200) {+ cfg_.rc_target_bitrate = i;
+ ResetModel();
+ // 60-40 bitrate allocation for 2 temporal layers.
+ cfg_.ts_target_bitrate[0] = 60 * cfg_.rc_target_bitrate / 100;
+ cfg_.ts_target_bitrate[1] = cfg_.rc_target_bitrate;
+ ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+ for (int j = 0; j < static_cast<int>(cfg_.ts_number_layers); ++j) {+ ASSERT_GE(effective_datarate_[j], cfg_.ts_target_bitrate[j] * 0.85)
+ << " The datarate for the file is lower than target by too much, "
+ "for layer: " << j;
+ ASSERT_LE(effective_datarate_[j], cfg_.ts_target_bitrate[j] * 1.15)
+ << " The datarate for the file is greater than target by too much, "
+ "for layer: " << j;
+ }
+ }
+}
+
+// Check basic rate targeting for 3 temporal layers.
+TEST_P(DatarateTestVP9, BasicRateTargeting3TemporalLayers) {+ cfg_.rc_buf_initial_sz = 500;
+ cfg_.rc_buf_optimal_sz = 500;
+ cfg_.rc_buf_sz = 1000;
+ cfg_.rc_dropframe_thresh = 1;
+ cfg_.rc_min_quantizer = 0;
+ cfg_.rc_max_quantizer = 63;
+ cfg_.rc_end_usage = VPX_CBR;
+ cfg_.g_lag_in_frames = 0;
+
+ // 3 Temporal layers, no spatial layers: Framerate decimation (4, 2, 1).
+ cfg_.ss_number_layers = 1;
+ cfg_.ts_number_layers = 3;
+ cfg_.ts_rate_decimator[0] = 4;
+ cfg_.ts_rate_decimator[1] = 2;
+ cfg_.ts_rate_decimator[2] = 1;
+
+ ::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,+ 30, 1, 0, 200);
+ for (int i = 200; i <= 800; i += 200) {+ cfg_.rc_target_bitrate = i;
+ ResetModel();
+ // 40-20-40 bitrate allocation for 3 temporal layers.
+ cfg_.ts_target_bitrate[0] = 40 * cfg_.rc_target_bitrate / 100;
+ cfg_.ts_target_bitrate[1] = 60 * cfg_.rc_target_bitrate / 100;
+ cfg_.ts_target_bitrate[2] = cfg_.rc_target_bitrate;
+ ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+ for (int j = 0; j < static_cast<int>(cfg_.ts_number_layers); ++j) {+ ASSERT_GE(effective_datarate_[j], cfg_.ts_target_bitrate[j] * 0.85)
+ << " The datarate for the file is lower than target by too much, "
+ "for layer: " << j;
+ ASSERT_LE(effective_datarate_[j], cfg_.ts_target_bitrate[j] * 1.15)
+ << " The datarate for the file is greater than target by too much, "
+ "for layer: " << j;
+ }
+ }
+}
VP8_INSTANTIATE_TEST_CASE(DatarateTest, ALL_TEST_MODES);
VP9_INSTANTIATE_TEST_CASE(DatarateTestVP9,
::testing::Values(::libvpx_test::kOnePassGood),
--- a/test/encode_test_driver.h
+++ b/test/encode_test_driver.h
@@ -123,6 +123,11 @@
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
+ void Control(int ctrl_id, struct vpx_svc_layer_id *arg) {+ const vpx_codec_err_t res = vpx_codec_control_(&encoder_, ctrl_id, arg);
+ ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
+ }
+
void set_deadline(unsigned long deadline) {deadline_ = deadline;
}
--- a/vp9/common/vp9_onyx.h
+++ b/vp9/common/vp9_onyx.h
@@ -147,8 +147,12 @@
// END DATARATE CONTROL OPTIONS
// ----------------------------------------------------------------
- // Spatial scalability
- int ss_number_layers;
+ // Spatial and temporal scalability.
+ int ss_number_layers; // Number of spatial layers.
+ unsigned int ts_number_layers; // Number of temporal layers.
+ // Bitrate allocation (CBR mode) and framerate factor, for temporal layers.
+ unsigned int ts_target_bitrate[VPX_TS_MAX_LAYERS];
+ unsigned int ts_rate_decimator[VPX_TS_MAX_LAYERS];
// these parameters aren't to be used in final build don't use!!!
int play_alternate;
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -1158,6 +1158,107 @@
return (llval * llnum / llden);
}
+// Initialize layer content data from init_config().
+static void init_layer_context(VP9_COMP *const cpi) {+ int temporal_layer = 0;
+ cpi->svc.spatial_layer_id = 0;
+ cpi->svc.temporal_layer_id = 0;
+ for (temporal_layer = 0; temporal_layer < cpi->svc.number_temporal_layers;
+ ++temporal_layer) {+ LAYER_CONTEXT *lc = &cpi->svc.layer_context[temporal_layer];
+ lc->rc.active_worst_quality = q_trans[cpi->oxcf.worst_allowed_q];
+ lc->rc.avg_frame_qindex[INTER_FRAME] = q_trans[cpi->oxcf.worst_allowed_q];
+ lc->rc.last_q[INTER_FRAME] = q_trans[cpi->oxcf.worst_allowed_q];
+ lc->rc.ni_av_qi = lc->rc.active_worst_quality;
+ lc->rc.total_actual_bits = 0;
+ lc->rc.total_target_vs_actual = 0;
+ lc->rc.ni_tot_qi = 0;
+ lc->rc.tot_q = 0.0;
+ lc->rc.ni_frames = 0;
+ lc->rc.rate_correction_factor = 1.0;
+ lc->rc.key_frame_rate_correction_factor = 1.0;
+ lc->target_bandwidth = cpi->oxcf.ts_target_bitrate[temporal_layer] *
+ 1000;
+ lc->rc.buffer_level = rescale((int)(cpi->oxcf.starting_buffer_level),
+ lc->target_bandwidth, 1000);
+ lc->rc.bits_off_target = lc->rc.buffer_level;
+ }
+}
+
+// Update the layer context from a change_config() call.
+static void update_layer_context_change_config(VP9_COMP *const cpi,
+ const int target_bandwidth) {+ int temporal_layer = 0;
+ float bitrate_alloc = 1.0;
+ for (temporal_layer = 0; temporal_layer < cpi->svc.number_temporal_layers;
+ ++temporal_layer) {+ LAYER_CONTEXT *lc = &cpi->svc.layer_context[temporal_layer];
+ lc->target_bandwidth = cpi->oxcf.ts_target_bitrate[temporal_layer] * 1000;
+ bitrate_alloc = (float)lc->target_bandwidth / (float)target_bandwidth;
+ // Update buffer-related quantities.
+ lc->starting_buffer_level = cpi->oxcf.starting_buffer_level * bitrate_alloc;
+ lc->optimal_buffer_level = cpi->oxcf.optimal_buffer_level * bitrate_alloc;
+ lc->maximum_buffer_size = cpi->oxcf.maximum_buffer_size * bitrate_alloc;
+ lc->rc.bits_off_target = MIN(lc->rc.bits_off_target,
+ lc->maximum_buffer_size);
+ lc->rc.buffer_level = MIN(lc->rc.buffer_level, lc->maximum_buffer_size);
+ // Update framerate-related quantities.
+ lc->framerate = cpi->oxcf.framerate /
+ cpi->oxcf.ts_rate_decimator[temporal_layer];
+ lc->rc.av_per_frame_bandwidth = (int)(lc->target_bandwidth / lc->framerate);
+ lc->rc.per_frame_bandwidth = lc->rc.av_per_frame_bandwidth;
+ lc->rc.max_frame_bandwidth = cpi->rc.max_frame_bandwidth;
+ // Update qp-related quantities.
+ lc->rc.worst_quality = cpi->rc.worst_quality;
+ lc->rc.best_quality = cpi->rc.best_quality;
+ lc->rc.active_worst_quality = cpi->rc.active_worst_quality;
+ }
+}
+
+// Prior to encoding the frame, update framerate-related quantities
+// for the current layer.
+static void update_layer_framerate(VP9_COMP *const cpi) {+ int temporal_layer = cpi->svc.temporal_layer_id;
+ LAYER_CONTEXT *lc = &cpi->svc.layer_context[temporal_layer];
+ lc->framerate = cpi->oxcf.framerate /
+ cpi->oxcf.ts_rate_decimator[temporal_layer];
+ lc->rc.av_per_frame_bandwidth = (int)(lc->target_bandwidth /
+ lc->framerate);
+ lc->rc.per_frame_bandwidth = lc->rc.av_per_frame_bandwidth;
+ lc->rc.max_frame_bandwidth = cpi->rc.max_frame_bandwidth;
+}
+
+// Prior to encoding the frame, set the layer context, for the current layer
+// to be encoded, to the cpi struct.
+static void restore_layer_context(VP9_COMP *const cpi) {+ int temporal_layer = cpi->svc.temporal_layer_id;
+ LAYER_CONTEXT *lc = &cpi->svc.layer_context[temporal_layer];
+ int frame_since_key = cpi->rc.frames_since_key;
+ int frame_to_key = cpi->rc.frames_to_key;
+ memcpy(&cpi->rc, &lc->rc, sizeof(RATE_CONTROL));
+ cpi->target_bandwidth = lc->target_bandwidth;
+ cpi->oxcf.starting_buffer_level = lc->starting_buffer_level;
+ cpi->oxcf.optimal_buffer_level = lc->optimal_buffer_level;
+ cpi->oxcf.maximum_buffer_size = lc->maximum_buffer_size;
+ cpi->output_framerate = lc->framerate;
+ // Reset the frames_since_key and frames_to_key counters to their values
+ // before the layer restore. Keep these defined for the stream (not layer).
+ cpi->rc.frames_since_key = frame_since_key;
+ cpi->rc.frames_to_key = frame_to_key;
+}
+
+// Save the layer context after encoding the frame.
+static void save_layer_context(VP9_COMP *const cpi) {+ int temporal_layer = cpi->svc.temporal_layer_id;
+ LAYER_CONTEXT *lc = &cpi->svc.layer_context[temporal_layer];
+ memcpy(&lc->rc, &cpi->rc, sizeof(RATE_CONTROL));
+ lc->target_bandwidth = cpi->target_bandwidth;
+ lc->starting_buffer_level = cpi->oxcf.starting_buffer_level;
+ lc->optimal_buffer_level = cpi->oxcf.optimal_buffer_level;
+ lc->maximum_buffer_size = cpi->oxcf.maximum_buffer_size;
+ lc->framerate = cpi->output_framerate;
+}
+
static void set_tile_limits(VP9_COMP *cpi) {VP9_COMMON *const cm = &cpi->common;
@@ -1184,6 +1285,16 @@
cm->subsampling_y = 0;
vp9_alloc_compressor_data(cpi);
+ // Spatial scalability.
+ cpi->svc.number_spatial_layers = oxcf->ss_number_layers;
+ // Temporal scalability.
+ cpi->svc.number_temporal_layers = oxcf->ts_number_layers;
+
+ if (cpi->svc.number_temporal_layers > 1 &&
+ cpi->oxcf.end_usage == USAGE_STREAM_FROM_SERVER) {+ init_layer_context(cpi);
+ }
+
// change includes all joint functionality
vp9_change_config(ptr, oxcf);
@@ -1224,9 +1335,6 @@
cpi->gld_fb_idx = 1;
cpi->alt_fb_idx = 2;
- cpi->current_layer = 0;
- cpi->use_svc = 0;
-
set_tile_limits(cpi);
cpi->fixed_divide[0] = 0;
@@ -1234,7 +1342,6 @@
cpi->fixed_divide[i] = 0x80000 / i;
}
-
void vp9_change_config(VP9_PTR ptr, VP9_CONFIG *oxcf) {VP9_COMP *cpi = (VP9_COMP *)(ptr);
VP9_COMMON *const cm = &cpi->common;
@@ -1326,10 +1433,10 @@
cpi->oxcf.target_bandwidth, 1000);
// Under a configuration change, where maximum_buffer_size may change,
// keep buffer level clipped to the maximum allowed buffer size.
- if (cpi->rc.bits_off_target > cpi->oxcf.maximum_buffer_size) {- cpi->rc.bits_off_target = cpi->oxcf.maximum_buffer_size;
- cpi->rc.buffer_level = cpi->rc.bits_off_target;
- }
+ cpi->rc.bits_off_target = MIN(cpi->rc.bits_off_target,
+ cpi->oxcf.maximum_buffer_size);
+ cpi->rc.buffer_level = MIN(cpi->rc.buffer_level,
+ cpi->oxcf.maximum_buffer_size);
// Set up frame rate and related parameters rate control values.
vp9_new_framerate(cpi, cpi->oxcf.framerate);
@@ -1366,6 +1473,11 @@
}
update_frame_size(cpi);
+ if (cpi->svc.number_temporal_layers > 1 &&
+ cpi->oxcf.end_usage == USAGE_STREAM_FROM_SERVER) {+ update_layer_context_change_config(cpi, cpi->oxcf.target_bandwidth);
+ }
+
cpi->speed = cpi->oxcf.cpu_used;
if (cpi->oxcf.lag_in_frames == 0) {@@ -1589,6 +1701,8 @@
vp9_create_common(cm);
+ cpi->use_svc = 0;
+
init_config((VP9_PTR)cpi, oxcf);
init_pick_mode_context(cpi);
@@ -1604,9 +1718,6 @@
cpi->alt_is_last = 0;
cpi->gold_is_alt = 0;
- // Spatial scalability
- cpi->number_spatial_layers = oxcf->ss_number_layers;
-
// Create the encoder segmentation map and set all entries to 0
CHECK_MEM_ERROR(cm, cpi->segmentation_map,
vpx_calloc(cm->mi_rows * cm->mi_cols, 1));
@@ -3558,6 +3669,12 @@
adjust_frame_rate(cpi);
}
+ if (cpi->svc.number_temporal_layers > 1 &&
+ cpi->oxcf.end_usage == USAGE_STREAM_FROM_SERVER) {+ update_layer_framerate(cpi);
+ restore_layer_context(cpi);
+ }
+
// start with a 0 size frame
*size = 0;
@@ -3631,6 +3748,12 @@
if (*size > 0) {cpi->droppable = !frame_is_reference(cpi);
+ }
+
+ // Save layer specific state.
+ if (cpi->svc.number_temporal_layers > 1 &&
+ cpi->oxcf.end_usage == USAGE_STREAM_FROM_SERVER) {+ save_layer_context(cpi);
}
vpx_usec_timer_mark(&cmptimer);
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -421,6 +421,15 @@
int super_fast_rtc;
} SPEED_FEATURES;
+typedef struct {+ RATE_CONTROL rc;
+ int target_bandwidth;
+ int64_t starting_buffer_level;
+ int64_t optimal_buffer_level;
+ int64_t maximum_buffer_size;
+ double framerate;
+} LAYER_CONTEXT;
+
typedef struct VP9_COMP {DECLARE_ALIGNED(16, int16_t, y_quant[QINDEX_RANGE][8]);
DECLARE_ALIGNED(16, int16_t, y_quant_shift[QINDEX_RANGE][8]);
@@ -465,9 +474,6 @@
int gld_fb_idx;
int alt_fb_idx;
- int current_layer;
- int use_svc;
-
#if CONFIG_MULTIPLE_ARF
int alt_ref_fb_idx[REF_FRAMES - 3];
#endif
@@ -684,7 +690,18 @@
int initial_width;
int initial_height;
- int number_spatial_layers;
+ int use_svc;
+
+ struct svc {+ int spatial_layer_id;
+ int temporal_layer_id;
+ int number_spatial_layers;
+ int number_temporal_layers;
+ // Layer context used for rate control in CBR mode, only defined for
+ // temporal layers for now.
+ LAYER_CONTEXT layer_context[VPX_TS_MAX_LAYERS];
+ } svc;
+
int enable_encode_breakout; // Default value is 1. From first pass stats,
// encode_breakout may be disabled.
--- a/vp9/encoder/vp9_ratectrl.c
+++ b/vp9/encoder/vp9_ratectrl.c
@@ -252,6 +252,26 @@
rc->this_frame_target = target;
}
+
+// Update the buffer level for higher layers, given the encoded current layer.
+static void update_layer_buffer_level(VP9_COMP *const cpi,
+ int encoded_frame_size) {+ int temporal_layer = 0;
+ int current_temporal_layer = cpi->svc.temporal_layer_id;
+ for (temporal_layer = current_temporal_layer + 1;
+ temporal_layer < cpi->svc.number_temporal_layers; ++temporal_layer) {+ LAYER_CONTEXT *lc = &cpi->svc.layer_context[temporal_layer];
+ int bits_off_for_this_layer = (int)(lc->target_bandwidth / lc->framerate -
+ encoded_frame_size);
+ lc->rc.bits_off_target += bits_off_for_this_layer;
+
+ // Clip buffer level to maximum buffer size for the layer.
+ lc->rc.bits_off_target = MIN(lc->rc.bits_off_target,
+ lc->maximum_buffer_size);
+ lc->rc.buffer_level = lc->rc.bits_off_target;
+ }
+}
+
// Update the buffer level: leaky bucket model.
void vp9_update_buffer_level(VP9_COMP *const cpi, int encoded_frame_size) {const VP9_COMMON *const cm = &cpi->common;
@@ -266,7 +286,12 @@
}
// Clip the buffer level to the maximum specified buffer size.
- rc->buffer_level = MIN(rc->bits_off_target, oxcf->maximum_buffer_size);
+ rc->bits_off_target = MIN(rc->bits_off_target, oxcf->maximum_buffer_size);
+ rc->buffer_level = rc->bits_off_target;
+
+ if (cpi->use_svc && cpi->oxcf.end_usage == USAGE_STREAM_FROM_SERVER) {+ update_layer_buffer_level(cpi, encoded_frame_size);
+ }
}
int vp9_drop_frame(VP9_COMP *const cpi) {@@ -273,7 +298,6 @@
const VP9_CONFIG *oxcf = &cpi->oxcf;
RATE_CONTROL *const rc = &cpi->rc;
-
if (!oxcf->drop_frames_water_mark) {return 0;
} else {@@ -284,7 +308,7 @@
// If buffer is below drop_mark, for now just drop every other frame
// (starting with the next frame) until it increases back over drop_mark.
int drop_mark = (int)(oxcf->drop_frames_water_mark *
- oxcf->optimal_buffer_level / 100);
+ oxcf->optimal_buffer_level / 100);
if ((rc->buffer_level > drop_mark) &&
(rc->decimation_factor > 0)) {--rc->decimation_factor;
@@ -368,7 +392,6 @@
const int pct_high = MIN(-diff / one_pct_bits, oxcf->over_shoot_pct);
target += (target * pct_high) / 200;
}
-
return target;
}
@@ -428,7 +451,8 @@
if (cpi->common.frame_type == KEY_FRAME) {return cpi->rc.key_frame_rate_correction_factor;
} else {- if (cpi->refresh_alt_ref_frame || cpi->refresh_golden_frame)
+ if ((cpi->refresh_alt_ref_frame || cpi->refresh_golden_frame) &&
+ !(cpi->use_svc && cpi->oxcf.end_usage == USAGE_STREAM_FROM_SERVER))
return cpi->rc.gf_rate_correction_factor;
else
return cpi->rc.rate_correction_factor;
@@ -439,7 +463,8 @@
if (cpi->common.frame_type == KEY_FRAME) {cpi->rc.key_frame_rate_correction_factor = factor;
} else {- if (cpi->refresh_alt_ref_frame || cpi->refresh_golden_frame)
+ if ((cpi->refresh_alt_ref_frame || cpi->refresh_golden_frame) &&
+ !(cpi->use_svc && cpi->oxcf.end_usage == USAGE_STREAM_FROM_SERVER))
cpi->rc.gf_rate_correction_factor = factor;
else
cpi->rc.rate_correction_factor = factor;
--- a/vp9/vp9_cx_iface.c
+++ b/vp9/vp9_cx_iface.c
@@ -175,6 +175,23 @@
RANGE_CHECK(cfg, ss_number_layers, 1,
VPX_SS_MAX_LAYERS); /*Spatial layers max */
+
+ RANGE_CHECK(cfg, ts_number_layers, 1, VPX_TS_MAX_LAYERS);
+ if (cfg->ts_number_layers > 1) {+ unsigned int i;
+ for (i = 1; i <cfg->ts_number_layers; i++) {+ if (cfg->ts_target_bitrate[i] < cfg->ts_target_bitrate[i-1]) {+ ERROR("ts_target_bitrate entries are not increasing");+ }
+ }
+ RANGE_CHECK(cfg, ts_rate_decimator[cfg->ts_number_layers-1], 1, 1);
+ for (i = cfg->ts_number_layers-2; i > 0; i--) {+ if (cfg->ts_rate_decimator[i-1] != 2*cfg->ts_rate_decimator[i]) {+ ERROR("ts_rate_decimator factors are not powers of 2");+ }
+ }
+ }
+
/* VP8 does not support a lower bound on the keyframe interval in
* automatic keyframe placement mode.
*/
@@ -345,6 +362,19 @@
oxcf->aq_mode = vp8_cfg.aq_mode;
oxcf->ss_number_layers = cfg.ss_number_layers;
+
+ oxcf->ts_number_layers = cfg.ts_number_layers;
+
+ if (oxcf->ts_number_layers > 1) {+ memcpy(oxcf->ts_target_bitrate, cfg.ts_target_bitrate,
+ sizeof(cfg.ts_target_bitrate));
+ memcpy(oxcf->ts_rate_decimator, cfg.ts_rate_decimator,
+ sizeof(cfg.ts_rate_decimator));
+ } else if (oxcf->ts_number_layers == 1) {+ oxcf->ts_target_bitrate[0] = oxcf->target_bandwidth;
+ oxcf->ts_rate_decimator[0] = 1;
+ }
+
/*
printf("Current VP9 Settings: \n"); printf("target_bandwidth: %d\n", oxcf->target_bandwidth);@@ -1016,9 +1046,35 @@
va_list args) {int data = va_arg(args, int);
vp9_set_svc(ctx->cpi, data);
+ // CBR mode for SVC with both temporal and spatial layers not yet supported.
+ if (data == 1 &&
+ ctx->cfg.rc_end_usage == VPX_CBR &&
+ ctx->cfg.ss_number_layers > 1 &&
+ ctx->cfg.ts_number_layers > 1) {+ return VPX_CODEC_INVALID_PARAM;
+ }
return VPX_CODEC_OK;
}
+static vpx_codec_err_t vp9e_set_svc_layer_id(vpx_codec_alg_priv_t *ctx,
+ int ctr_id,
+ va_list args) {+ vpx_svc_layer_id_t *data = va_arg(args, vpx_svc_layer_id_t *);
+ VP9_COMP *cpi = (VP9_COMP *)ctx->cpi;
+ cpi->svc.spatial_layer_id = data->spatial_layer_id;
+ cpi->svc.temporal_layer_id = data->temporal_layer_id;
+ // Checks on valid layer_id input.
+ if (cpi->svc.temporal_layer_id < 0 ||
+ cpi->svc.temporal_layer_id >= ctx->cfg.ts_number_layers) {+ return VPX_CODEC_INVALID_PARAM;
+ }
+ if (cpi->svc.spatial_layer_id < 0 ||
+ cpi->svc.spatial_layer_id >= ctx->cfg.ss_number_layers) {+ return VPX_CODEC_INVALID_PARAM;
+ }
+ return VPX_CODEC_OK;
+}
+
static vpx_codec_err_t vp9e_set_svc_parameters(vpx_codec_alg_priv_t *ctx,
int ctr_id, va_list args) {vpx_svc_parameters_t *data = va_arg(args, vpx_svc_parameters_t *);
@@ -1031,7 +1087,9 @@
params = *(vpx_svc_parameters_t *)data;
- cpi->current_layer = params.layer;
+ cpi->svc.spatial_layer_id = params.spatial_layer;
+ cpi->svc.temporal_layer_id = params.temporal_layer;
+
cpi->lst_fb_idx = params.lst_fb_idx;
cpi->gld_fb_idx = params.gld_fb_idx;
cpi->alt_fb_idx = params.alt_fb_idx;
@@ -1080,6 +1138,7 @@
{VP9_GET_REFERENCE, get_reference}, {VP9E_SET_SVC, vp9e_set_svc}, {VP9E_SET_SVC_PARAMETERS, vp9e_set_svc_parameters},+ {VP9E_SET_SVC_LAYER_ID, vp9e_set_svc_layer_id}, { -1, NULL},};
@@ -1130,7 +1189,11 @@
9999, /* kf_max_dist */
VPX_SS_DEFAULT_LAYERS, /* ss_number_layers */
-
+ 1, /* ts_number_layers */
+ {0}, /* ts_target_bitrate */+ {0}, /* ts_rate_decimator */+ 0, /* ts_periodicity */
+ {0}, /* ts_layer_id */#if VPX_ENCODER_ABI_VERSION == (1 + VPX_CODEC_ABI_VERSION)
1, /* g_delete_first_pass_file */
"vp8.fpf" /* first pass filename */
--- a/vpx/src/svc_encodeframe.c
+++ b/vpx/src/svc_encodeframe.c
@@ -499,6 +499,7 @@
// modify encoder configuration
enc_cfg->ss_number_layers = si->layers;
+ enc_cfg->ts_number_layers = 1; // Temporal layers not used in this encoder.
enc_cfg->kf_mode = VPX_KF_DISABLED;
enc_cfg->g_pass = VPX_RC_ONE_PASS;
// Lag in frames not currently supported
@@ -691,7 +692,8 @@
SvcInternal *const si = get_svc_internal(svc_ctx);
memset(&svc_params, 0, sizeof(svc_params));
- svc_params.layer = si->layer;
+ svc_params.temporal_layer = 0;
+ svc_params.spatial_layer = si->layer;
svc_params.flags = si->enc_frame_flags;
layer = si->layer;
--- a/vpx/vp8cx.h
+++ b/vpx/vp8cx.h
@@ -194,7 +194,8 @@
VP9E_SET_AQ_MODE,
VP9E_SET_SVC,
- VP9E_SET_SVC_PARAMETERS
+ VP9E_SET_SVC_PARAMETERS,
+ VP9E_SET_SVC_LAYER_ID
};
/*!\brief vpx 1-D scaling mode
@@ -285,7 +286,8 @@
typedef struct vpx_svc_parameters {unsigned int width; /**< width of current spatial layer */
unsigned int height; /**< height of current spatial layer */
- int layer; /**< current layer number - 0 = base */
+ int spatial_layer; /**< current spatial layer number - 0 = base */
+ int temporal_layer; /**< current temporal layer number - 0 = base */
int flags; /**< encode frame flags */
int max_quantizer; /**< max quantizer for current layer */
int min_quantizer; /**< min quantizer for current layer */
@@ -295,6 +297,11 @@
int alt_fb_idx; /**< alt reference frame frame buffer index */
} vpx_svc_parameters_t;
+typedef struct vpx_svc_layer_id {+ int spatial_layer_id;
+ int temporal_layer_id;
+} vpx_svc_layer_id_t;
+
/*!\brief VP8 encoder control function parameter type
*
* Defines the data types that VP8E control functions take. Note that
@@ -316,6 +323,7 @@
VPX_CTRL_USE_TYPE(VP9E_SET_SVC, int)
VPX_CTRL_USE_TYPE(VP9E_SET_SVC_PARAMETERS, vpx_svc_parameters_t *)
+VPX_CTRL_USE_TYPE(VP9E_SET_SVC_LAYER_ID, vpx_svc_layer_id_t *)
VPX_CTRL_USE_TYPE(VP8E_SET_CPUUSED, int)
VPX_CTRL_USE_TYPE(VP8E_SET_ENABLEAUTOALTREF, unsigned int)
--- a/vpx/vpx_encoder.h
+++ b/vpx/vpx_encoder.h
@@ -604,47 +604,48 @@
* Spatial scalability settings (ss)
*/
- /*!\brief Number of coding layers (spatial)
+ /*!\brief Number of spatial coding layers.
*
- * This value specifies the number of coding layers to be used.
+ * This value specifies the number of spatial coding layers to be used.
*/
unsigned int ss_number_layers;
- /*!\brief Number of coding layers
+ /*!\brief Number of temporal coding layers.
*
- * This value specifies the number of coding layers to be used.
+ * This value specifies the number of temporal layers to be used.
*/
unsigned int ts_number_layers;
- /*!\brief Target bitrate for each layer
+ /*!\brief Target bitrate for each temporal layer.
*
- * These values specify the target coding bitrate for each coding layer.
+ * These values specify the target coding bitrate to be used for each
+ * temporal layer.
*/
unsigned int ts_target_bitrate[VPX_TS_MAX_LAYERS];
- /*!\brief Frame rate decimation factor for each layer
+ /*!\brief Frame rate decimation factor for each temporal layer.
*
* These values specify the frame rate decimation factors to apply
- * to each layer.
+ * to each temporal layer.
*/
unsigned int ts_rate_decimator[VPX_TS_MAX_LAYERS];
- /*!\brief Length of the sequence defining frame layer membership
- *
- * This value specifies the length of the sequence that defines the
- * membership of frames to layers. For example, if ts_periodicity=8 then
- * frames are assigned to coding layers with a repeated sequence of
- * length 8.
- */
+ /*!\brief Length of the sequence defining frame temporal layer membership.
+ *
+ * This value specifies the length of the sequence that defines the
+ * membership of frames to temporal layers. For example, if ts_periodicity=8
+ * then frames are assigned to coding layers with a repeated sequence of
+ * length 8.
+ */
unsigned int ts_periodicity;
- /*!\brief Template defining the membership of frames to coding layers
- *
- * This array defines the membership of frames to coding layers. For a
- * 2-layer encoding that assigns even numbered frames to one layer (0)
- * and odd numbered frames to a second layer (1) with ts_periodicity=8,
- * then ts_layer_id = (0,1,0,1,0,1,0,1).
- */
+ /*!\brief Template defining the membership of frames to temporal layers.
+ *
+ * This array defines the membership of frames to temporal coding layers.
+ * For a 2-layer encoding that assigns even numbered frames to one temporal
+ * layer (0) and odd numbered frames to a second temporal layer (1) with
+ * ts_periodicity=8, then ts_layer_id = (0,1,0,1,0,1,0,1).
+ */
unsigned int ts_layer_id[VPX_TS_MAX_PERIODICITY];
} vpx_codec_enc_cfg_t; /**< alias for struct vpx_codec_enc_cfg */
--- /dev/null
+++ b/vpx_temporal_scalable_patterns.c
@@ -1,0 +1,537 @@
+/*
+ * Copyright (c) 2012 The WebM 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.
+ */
+
+// This is an example demonstrating how to implement a multi-layer VP9
+// encoding scheme based on temporal scalability for video applications
+// that benefit from a scalable bitstream.
+
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#define VPX_CODEC_DISABLE_COMPAT 1
+#include "./ivfenc.h"
+#include "./tools_common.h"
+#include "./vpx_config.h"
+#include "vpx/vp8cx.h"
+#include "vpx/vpx_encoder.h"
+
+static const char *exec_name;
+
+void usage_exit() {+ exit(EXIT_FAILURE);
+}
+
+static int mode_to_num_layers[12] = {1, 2, 2, 3, 3, 3, 3, 5, 2, 3, 3, 3};+
+// Temporal scaling parameters:
+// NOTE: The 3 prediction frames cannot be used interchangeably due to
+// differences in the way they are handled throughout the code. The
+// frames should be allocated to layers in the order LAST, GF, ARF.
+// Other combinations work, but may produce slightly inferior results.
+static void set_temporal_layer_pattern(int layering_mode,
+ vpx_codec_enc_cfg_t *cfg,
+ int *layer_flags,
+ int *flag_periodicity) {+ switch (layering_mode) {+ case 0: {+ // 1-layer.
+ int ids[1] = {0};+ cfg->ts_periodicity = 1;
+ *flag_periodicity = 1;
+ cfg->ts_number_layers = 1;
+ cfg->ts_rate_decimator[0] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // Update L only.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
+ VP8_EFLAG_NO_UPD_ARF;
+ break;
+ }
+ case 1: {+ // 2-layers, 2-frame period.
+ int ids[2] = {0, 1};+ cfg->ts_periodicity = 2;
+ *flag_periodicity = 2;
+ cfg->ts_number_layers = 2;
+ cfg->ts_rate_decimator[0] = 2;
+ cfg->ts_rate_decimator[1] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+#if 1
+ // 0=L, 1=GF, Intra-layer prediction enabled.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
+ VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
+ layer_flags[1] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_REF_ARF;
+#else
+ // 0=L, 1=GF, Intra-layer prediction disabled.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_GF |
+ VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
+ layer_flags[1] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_REF_LAST;
+#endif
+ break;
+ }
+ case 2: {+ // 2-layers, 3-frame period.
+ int ids[3] = {0, 1, 1};+ cfg->ts_periodicity = 3;
+ *flag_periodicity = 3;
+ cfg->ts_number_layers = 2;
+ cfg->ts_rate_decimator[0] = 3;
+ cfg->ts_rate_decimator[1] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF, Intra-layer prediction enabled.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
+ VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[1] =
+ layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
+ VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
+ break;
+ }
+ case 3: {+ // 3-layers, 6-frame period.
+ int ids[6] = {0, 2, 2, 1, 2, 2};+ cfg->ts_periodicity = 6;
+ *flag_periodicity = 6;
+ cfg->ts_number_layers = 3;
+ cfg->ts_rate_decimator[0] = 6;
+ cfg->ts_rate_decimator[1] = 3;
+ cfg->ts_rate_decimator[2] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
+ VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_UPD_LAST;
+ layer_flags[1] =
+ layer_flags[2] =
+ layer_flags[4] =
+ layer_flags[5] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_LAST;
+ break;
+ }
+ case 4: {+ // 3-layers, 4-frame period.
+ int ids[4] = {0, 2, 1, 2};+ cfg->ts_periodicity = 4;
+ *flag_periodicity = 4;
+ cfg->ts_number_layers = 3;
+ cfg->ts_rate_decimator[0] = 4;
+ cfg->ts_rate_decimator[1] = 2;
+ cfg->ts_rate_decimator[2] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
+ VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
+ VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
+ layer_flags[1] =
+ layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ break;
+ }
+ case 5: {+ // 3-layers, 4-frame period.
+ int ids[4] = {0, 2, 1, 2};+ cfg->ts_periodicity = 4;
+ *flag_periodicity = 4;
+ cfg->ts_number_layers = 3;
+ cfg->ts_rate_decimator[0] = 4;
+ cfg->ts_rate_decimator[1] = 2;
+ cfg->ts_rate_decimator[2] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled in layer 1, disabled
+ // in layer 2.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
+ VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[2] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[1] =
+ layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ break;
+ }
+ case 6: {+ // 3-layers, 4-frame period.
+ int ids[4] = {0, 2, 1, 2};+ cfg->ts_periodicity = 4;
+ *flag_periodicity = 4;
+ cfg->ts_number_layers = 3;
+ cfg->ts_rate_decimator[0] = 4;
+ cfg->ts_rate_decimator[1] = 2;
+ cfg->ts_rate_decimator[2] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
+ VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[2] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[1] =
+ layer_flags[3] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
+ break;
+ }
+ case 7: {+ // NOTE: Probably of academic interest only.
+ // 5-layers, 16-frame period.
+ int ids[16] = {0, 4, 3, 4, 2, 4, 3, 4, 1, 4, 3, 4, 2, 4, 3, 4};+ cfg->ts_periodicity = 16;
+ *flag_periodicity = 16;
+ cfg->ts_number_layers = 5;
+ cfg->ts_rate_decimator[0] = 16;
+ cfg->ts_rate_decimator[1] = 8;
+ cfg->ts_rate_decimator[2] = 4;
+ cfg->ts_rate_decimator[3] = 2;
+ cfg->ts_rate_decimator[4] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ layer_flags[0] = VPX_EFLAG_FORCE_KF;
+ layer_flags[1] =
+ layer_flags[3] =
+ layer_flags[5] =
+ layer_flags[7] =
+ layer_flags[9] =
+ layer_flags[11] =
+ layer_flags[13] =
+ layer_flags[15] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
+ VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[2] =
+ layer_flags[6] =
+ layer_flags[10] =
+ layer_flags[14] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_GF;
+ layer_flags[4] =
+ layer_flags[12] = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[8] = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_GF;
+ break;
+ }
+ case 8: {+ // 2-layers, with sync point at first frame of layer 1.
+ int ids[2] = {0, 1};+ cfg->ts_periodicity = 2;
+ *flag_periodicity = 8;
+ cfg->ts_number_layers = 2;
+ cfg->ts_rate_decimator[0] = 2;
+ cfg->ts_rate_decimator[1] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF.
+ // ARF is used as predictor for all frames, and is only updated on
+ // key frame. Sync point every 8 frames.
+
+ // Layer 0: predict from L and ARF, update L and G.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
+ VP8_EFLAG_NO_UPD_ARF;
+ // Layer 1: sync point: predict from L and ARF, and update G.
+ layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_ARF;
+ // Layer 0, predict from L and ARF, update L.
+ layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
+ VP8_EFLAG_NO_UPD_ARF;
+ // Layer 1: predict from L, G and ARF, and update G.
+ layer_flags[3] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_ENTROPY;
+ // Layer 0.
+ layer_flags[4] = layer_flags[2];
+ // Layer 1.
+ layer_flags[5] = layer_flags[3];
+ // Layer 0.
+ layer_flags[6] = layer_flags[4];
+ // Layer 1.
+ layer_flags[7] = layer_flags[5];
+ break;
+ }
+ case 9: {+ // 3-layers: Sync points for layer 1 and 2 every 8 frames.
+ int ids[4] = {0, 2, 1, 2};+ cfg->ts_periodicity = 4;
+ *flag_periodicity = 8;
+ cfg->ts_number_layers = 3;
+ cfg->ts_rate_decimator[0] = 4;
+ cfg->ts_rate_decimator[1] = 2;
+ cfg->ts_rate_decimator[2] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF, 2=ARF.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_REF_GF |
+ VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
+ VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
+ layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
+ VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[3] =
+ layer_flags[5] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
+ layer_flags[4] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
+ VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[6] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_ARF;
+ layer_flags[7] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
+ VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_ENTROPY;
+ break;
+ }
+ case 10: {+ // 3-layers structure where ARF is used as predictor for all frames,
+ // and is only updated on key frame.
+ // Sync points for layer 1 and 2 every 8 frames.
+
+ int ids[4] = {0, 2, 1, 2};+ cfg->ts_periodicity = 4;
+ *flag_periodicity = 8;
+ cfg->ts_number_layers = 3;
+ cfg->ts_rate_decimator[0] = 4;
+ cfg->ts_rate_decimator[1] = 2;
+ cfg->ts_rate_decimator[2] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF, 2=ARF.
+ // Layer 0: predict from L and ARF; update L and G.
+ layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_REF_GF;
+ // Layer 2: sync point: predict from L and ARF; update none.
+ layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
+ VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
+ VP8_EFLAG_NO_UPD_ENTROPY;
+ // Layer 1: sync point: predict from L and ARF; update G.
+ layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_UPD_LAST;
+ // Layer 2: predict from L, G, ARF; update none.
+ layer_flags[3] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ENTROPY;
+ // Layer 0: predict from L and ARF; update L.
+ layer_flags[4] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_REF_GF;
+ // Layer 2: predict from L, G, ARF; update none.
+ layer_flags[5] = layer_flags[3];
+ // Layer 1: predict from L, G, ARF; update G.
+ layer_flags[6] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
+ // Layer 2: predict from L, G, ARF; update none.
+ layer_flags[7] = layer_flags[3];
+ break;
+ }
+ case 11:
+ default: {+ // 3-layers structure as in case 10, but no sync/refresh points for
+ // layer 1 and 2.
+ int ids[4] = {0, 2, 1, 2};+ cfg->ts_periodicity = 4;
+ *flag_periodicity = 8;
+ cfg->ts_number_layers = 3;
+ cfg->ts_rate_decimator[0] = 4;
+ cfg->ts_rate_decimator[1] = 2;
+ cfg->ts_rate_decimator[2] = 1;
+ memcpy(cfg->ts_layer_id, ids, sizeof(ids));
+ // 0=L, 1=GF, 2=ARF.
+ // Layer 0: predict from L and ARF; update L.
+ layer_flags[0] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_REF_GF;
+ layer_flags[4] = layer_flags[0];
+ // Layer 1: predict from L, G, ARF; update G.
+ layer_flags[2] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
+ layer_flags[6] = layer_flags[2];
+ // Layer 2: predict from L, G, ARF; update none.
+ layer_flags[1] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
+ VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ENTROPY;
+ layer_flags[3] = layer_flags[1];
+ layer_flags[5] = layer_flags[1];
+ layer_flags[7] = layer_flags[1];
+ break;
+ }
+ }
+}
+
+int main(int argc, char **argv) {+ FILE *outfile[VPX_TS_MAX_LAYERS];
+ vpx_codec_ctx_t codec;
+ vpx_codec_enc_cfg_t cfg;
+ int frame_cnt = 0;
+ vpx_image_t raw;
+ vpx_codec_err_t res;
+ unsigned int width;
+ unsigned int height;
+ int frame_avail;
+ int got_data;
+ int flags = 0;
+ int i;
+ int pts = 0; // PTS starts at 0.
+ int frame_duration = 1; // 1 timebase tick per frame.
+ int layering_mode = 0;
+ int frames_in_layer[VPX_TS_MAX_LAYERS] = {0};+ int layer_flags[VPX_TS_MAX_PERIODICITY] = {0};+ int flag_periodicity = 1;
+ int max_intra_size_pct;
+ vpx_svc_layer_id_t layer_id = {0, 0};+ char *codec_type;
+ // Default is VP8 codec.
+ const vpx_codec_iface_t *(*interface)(void) = vpx_codec_vp8_cx;
+ unsigned int fourcc = 0x30385056;
+ struct VpxInputContext input_ctx = {0};+
+ exec_name = argv[0];
+ // Check usage and arguments.
+ if (argc < 10) {+ die("Usage: %s <infile> <outfile> <width> <height> <rate_num> "+ " <rate_den> <mode> <Rate_0> ... <Rate_nlayers-1> "
+ " <codec_type(vp8/vp9)> \n", argv[0]);
+ }
+ width = strtol(argv[3], NULL, 0);
+ height = strtol(argv[4], NULL, 0);
+ if (width < 16 || width%2 || height <16 || height%2) {+ die("Invalid resolution: %d x %d", width, height);+ }
+
+ layering_mode = strtol(argv[7], NULL, 0);
+ if (layering_mode < 0 || layering_mode > 11) {+ die("Invalid mode (0..11) %s", argv[7]);+ }
+
+ if (argc != 8 + mode_to_num_layers[layering_mode] + 1) {+ die("Invalid number of arguments");+ }
+
+ if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {+ die("Failed to allocate image", width, height);+ }
+
+ codec_type = argv[8 + mode_to_num_layers[layering_mode]];
+ if (strncmp(codec_type, "vp9", 3) == 0) {+ interface = vpx_codec_vp9_cx;
+ fourcc = 0x30395056;
+ }
+ printf("Using %s\n", vpx_codec_iface_name(interface()));+
+ // Populate encoder configuration.
+ res = vpx_codec_enc_config_default(interface(), &cfg, 0);
+ if (res) {+ printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));+ return EXIT_FAILURE;
+ }
+
+ // Update the default configuration with our settings.
+ cfg.g_w = width;
+ cfg.g_h = height;
+
+ // Timebase format e.g. 30fps: numerator=1, demoninator = 30.
+ cfg.g_timebase.num = strtol(argv[5], NULL, 0);
+ cfg.g_timebase.den = strtol(argv[6], NULL, 0);
+
+ for (i = 8; i < 8 + mode_to_num_layers[layering_mode]; ++i) {+ cfg.ts_target_bitrate[i-8] = strtol(argv[i], NULL, 0);
+ }
+
+ // Real time parameters.
+ cfg.rc_dropframe_thresh = 0;
+ cfg.rc_end_usage = VPX_CBR;
+ cfg.rc_resize_allowed = 0;
+ cfg.rc_min_quantizer = 2;
+ cfg.rc_max_quantizer = 56;
+ cfg.rc_undershoot_pct = 100;
+ cfg.rc_overshoot_pct = 15;
+ cfg.rc_buf_initial_sz = 500;
+ cfg.rc_buf_optimal_sz = 600;
+ cfg.rc_buf_sz = 1000;
+
+ // Enable error resilient mode.
+ cfg.g_error_resilient = 1;
+ cfg.g_lag_in_frames = 0;
+ cfg.kf_mode = VPX_KF_DISABLED;
+
+ // Disable automatic keyframe placement.
+ cfg.kf_min_dist = cfg.kf_max_dist = 3000;
+
+ // Default setting for bitrate: used in special case of 1 layer (case 0).
+ cfg.rc_target_bitrate = cfg.ts_target_bitrate[0];
+
+ set_temporal_layer_pattern(layering_mode,
+ &cfg,
+ layer_flags,
+ &flag_periodicity);
+
+ // Open input file.
+ input_ctx.filename = argv[1];
+ if (!(input_ctx.file = fopen(input_ctx.filename, "rb"))) {+ die("Failed to open %s for reading", argv[1]);+ }
+
+ // Open an output file for each stream.
+ for (i = 0; i < cfg.ts_number_layers; ++i) {+ char file_name[512];
+ snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i);
+ if (!(outfile[i] = fopen(file_name, "wb")))
+ die("Failed to open %s for writing", file_name);+ ivf_write_file_header(outfile[i], &cfg, fourcc, 0);
+ }
+ // No spatial layers in this encoder.
+ cfg.ss_number_layers = 1;
+
+ // Initialize codec.
+ if (vpx_codec_enc_init(&codec, interface(), &cfg, 0))
+ die_codec(&codec, "Failed to initialize encoder");
+
+ vpx_codec_control(&codec, VP8E_SET_CPUUSED, -6);
+ vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, 1);
+ if (strncmp(codec_type, "vp9", 3) == 0) {+ vpx_codec_control(&codec, VP8E_SET_CPUUSED, 3);
+ vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, 0);
+ if (vpx_codec_control(&codec, VP9E_SET_SVC, 1)) {+ die_codec(&codec, "Failed to set SVC");
+ }
+ }
+ vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
+ vpx_codec_control(&codec, VP8E_SET_TOKEN_PARTITIONS, 1);
+ max_intra_size_pct = (int) (((double)cfg.rc_buf_optimal_sz * 0.5)
+ * ((double) cfg.g_timebase.den / cfg.g_timebase.num) / 10.0);
+ vpx_codec_control(&codec, VP8E_SET_MAX_INTRA_BITRATE_PCT, max_intra_size_pct);
+
+ frame_avail = 1;
+ while (frame_avail || got_data) {+ vpx_codec_iter_t iter = NULL;
+ const vpx_codec_cx_pkt_t *pkt;
+ // Update the temporal layer_id. No spatial layers in this test.
+ layer_id.spatial_layer_id = 0;
+ layer_id.temporal_layer_id =
+ cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
+ vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id);
+ flags = layer_flags[frame_cnt % flag_periodicity];
+ frame_avail = !read_yuv_frame(&input_ctx, &raw);
+ if (vpx_codec_encode(&codec, frame_avail? &raw : NULL, pts, 1, flags,
+ VPX_DL_REALTIME)) {+ die_codec(&codec, "Failed to encode frame");
+ }
+ // Reset KF flag.
+ if (layering_mode != 7) {+ layer_flags[0] &= ~VPX_EFLAG_FORCE_KF;
+ }
+ got_data = 0;
+ while ( (pkt = vpx_codec_get_cx_data(&codec, &iter)) ) {+ got_data = 1;
+ switch (pkt->kind) {+ case VPX_CODEC_CX_FRAME_PKT:
+ for (i = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
+ i < cfg.ts_number_layers; ++i) {+ ivf_write_frame_header(outfile[i], pts, pkt->data.frame.sz);
+ (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
+ outfile[i]);
+ ++frames_in_layer[i];
+ }
+ break;
+ default:
+ break;
+ }
+ }
+ ++frame_cnt;
+ pts += frame_duration;
+ }
+ fclose(input_ctx.file);
+ printf("Processed %d frames: \n", frame_cnt-1);+ if (vpx_codec_destroy(&codec)) {+ die_codec(&codec, "Failed to destroy codec");
+ }
+ // Try to rewrite the output file headers with the actual frame count.
+ for (i = 0; i < cfg.ts_number_layers; ++i) {+ if (!fseek(outfile[i], 0, SEEK_SET))
+ ivf_write_file_header(outfile[i], &cfg, fourcc, frame_cnt);
+ fclose(outfile[i]);
+ }
+ return EXIT_SUCCESS;
+}
--
⑨