Home | History | Annotate | Download | only in examples
      1 /*
      2  *  Copyright (c) 2012 The WebM project authors. All Rights Reserved.
      3  *
      4  *  Use of this source code is governed by a BSD-style license
      5  *  that can be found in the LICENSE file in the root of the source
      6  *  tree. An additional intellectual property rights grant can be found
      7  *  in the file PATENTS.  All contributing project authors may
      8  *  be found in the AUTHORS file in the root of the source tree.
      9  */
     10 
     11 //  This is an example demonstrating how to implement a multi-layer VP9
     12 //  encoding scheme based on temporal scalability for video applications
     13 //  that benefit from a scalable bitstream.
     14 
     15 #include <math.h>
     16 #include <stdio.h>
     17 #include <stdlib.h>
     18 #include <string.h>
     19 
     20 #define VPX_CODEC_DISABLE_COMPAT 1
     21 #include "./vpx_config.h"
     22 #include "vpx_ports/vpx_timer.h"
     23 #include "vpx/vp8cx.h"
     24 #include "vpx/vpx_encoder.h"
     25 
     26 #include "./tools_common.h"
     27 #include "./video_writer.h"
     28 
     29 static const char *exec_name;
     30 
     31 void usage_exit() {
     32   exit(EXIT_FAILURE);
     33 }
     34 
     35 static int mode_to_num_layers[12] = {1, 2, 2, 3, 3, 3, 3, 5, 2, 3, 3, 3};
     36 
     37 // For rate control encoding stats.
     38 struct RateControlMetrics {
     39   // Number of input frames per layer.
     40   int layer_input_frames[VPX_TS_MAX_LAYERS];
     41   // Total (cumulative) number of encoded frames per layer.
     42   int layer_tot_enc_frames[VPX_TS_MAX_LAYERS];
     43   // Number of encoded non-key frames per layer.
     44   int layer_enc_frames[VPX_TS_MAX_LAYERS];
     45   // Framerate per layer layer (cumulative).
     46   double layer_framerate[VPX_TS_MAX_LAYERS];
     47   // Target average frame size per layer (per-frame-bandwidth per layer).
     48   double layer_pfb[VPX_TS_MAX_LAYERS];
     49   // Actual average frame size per layer.
     50   double layer_avg_frame_size[VPX_TS_MAX_LAYERS];
     51   // Average rate mismatch per layer (|target - actual| / target).
     52   double layer_avg_rate_mismatch[VPX_TS_MAX_LAYERS];
     53   // Actual encoding bitrate per layer (cumulative).
     54   double layer_encoding_bitrate[VPX_TS_MAX_LAYERS];
     55 };
     56 
     57 // Note: these rate control metrics assume only 1 key frame in the
     58 // sequence (i.e., first frame only). So for temporal pattern# 7
     59 // (which has key frame for every frame on base layer), the metrics
     60 // computation will be off/wrong.
     61 // TODO(marpan): Update these metrics to account for multiple key frames
     62 // in the stream.
     63 static void set_rate_control_metrics(struct RateControlMetrics *rc,
     64                                      vpx_codec_enc_cfg_t *cfg) {
     65   unsigned int i = 0;
     66   // Set the layer (cumulative) framerate and the target layer (non-cumulative)
     67   // per-frame-bandwidth, for the rate control encoding stats below.
     68   const double framerate = cfg->g_timebase.den / cfg->g_timebase.num;
     69   rc->layer_framerate[0] = framerate / cfg->ts_rate_decimator[0];
     70   rc->layer_pfb[0] = 1000.0 * cfg->ts_target_bitrate[0] /
     71       rc->layer_framerate[0];
     72   for (i = 0; i < cfg->ts_number_layers; ++i) {
     73     if (i > 0) {
     74       rc->layer_framerate[i] = framerate / cfg->ts_rate_decimator[i];
     75       rc->layer_pfb[i] = 1000.0 *
     76           (cfg->ts_target_bitrate[i] - cfg->ts_target_bitrate[i - 1]) /
     77           (rc->layer_framerate[i] - rc->layer_framerate[i - 1]);
     78     }
     79     rc->layer_input_frames[i] = 0;
     80     rc->layer_enc_frames[i] = 0;
     81     rc->layer_tot_enc_frames[i] = 0;
     82     rc->layer_encoding_bitrate[i] = 0.0;
     83     rc->layer_avg_frame_size[i] = 0.0;
     84     rc->layer_avg_rate_mismatch[i] = 0.0;
     85   }
     86 }
     87 
     88 static void printout_rate_control_summary(struct RateControlMetrics *rc,
     89                                           vpx_codec_enc_cfg_t *cfg,
     90                                           int frame_cnt) {
     91   unsigned int i = 0;
     92   int tot_num_frames = 0;
     93   printf("Total number of processed frames: %d\n\n", frame_cnt -1);
     94   printf("Rate control layer stats for %d layer(s):\n\n",
     95       cfg->ts_number_layers);
     96   for (i = 0; i < cfg->ts_number_layers; ++i) {
     97     const int num_dropped = (i > 0) ?
     98         (rc->layer_input_frames[i] - rc->layer_enc_frames[i]) :
     99         (rc->layer_input_frames[i] - rc->layer_enc_frames[i] - 1);
    100     tot_num_frames += rc->layer_input_frames[i];
    101     rc->layer_encoding_bitrate[i] = 0.001 * rc->layer_framerate[i] *
    102         rc->layer_encoding_bitrate[i] / tot_num_frames;
    103     rc->layer_avg_frame_size[i] = rc->layer_avg_frame_size[i] /
    104         rc->layer_enc_frames[i];
    105     rc->layer_avg_rate_mismatch[i] = 100.0 * rc->layer_avg_rate_mismatch[i] /
    106         rc->layer_enc_frames[i];
    107     printf("For layer#: %d \n", i);
    108     printf("Bitrate (target vs actual): %d %f \n", cfg->ts_target_bitrate[i],
    109            rc->layer_encoding_bitrate[i]);
    110     printf("Average frame size (target vs actual): %f %f \n", rc->layer_pfb[i],
    111            rc->layer_avg_frame_size[i]);
    112     printf("Average rate_mismatch: %f \n", rc->layer_avg_rate_mismatch[i]);
    113     printf("Number of input frames, encoded (non-key) frames, "
    114         "and perc dropped frames: %d %d %f \n", rc->layer_input_frames[i],
    115         rc->layer_enc_frames[i],
    116         100.0 * num_dropped / rc->layer_input_frames[i]);
    117     printf("\n");
    118   }
    119   if ((frame_cnt - 1) != tot_num_frames)
    120     die("Error: Number of input frames not equal to output! \n");
    121 }
    122 
    123 // Temporal scaling parameters:
    124 // NOTE: The 3 prediction frames cannot be used interchangeably due to
    125 // differences in the way they are handled throughout the code. The
    126 // frames should be allocated to layers in the order LAST, GF, ARF.
    127 // Other combinations work, but may produce slightly inferior results.
    128 static void set_temporal_layer_pattern(int layering_mode,
    129                                        vpx_codec_enc_cfg_t *cfg,
    130                                        int *layer_flags,
    131                                        int *flag_periodicity) {
    132   switch (layering_mode) {
    133     case 0: {
    134       // 1-layer.
    135       int ids[1] = {0};
    136       cfg->ts_periodicity = 1;
    137       *flag_periodicity = 1;
    138       cfg->ts_number_layers = 1;
    139       cfg->ts_rate_decimator[0] = 1;
    140       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    141       // Update L only.
    142       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_UPD_GF |
    143           VP8_EFLAG_NO_UPD_ARF;
    144       break;
    145     }
    146     case 1: {
    147       // 2-layers, 2-frame period.
    148       int ids[2] = {0, 1};
    149       cfg->ts_periodicity = 2;
    150       *flag_periodicity = 2;
    151       cfg->ts_number_layers = 2;
    152       cfg->ts_rate_decimator[0] = 2;
    153       cfg->ts_rate_decimator[1] = 1;
    154       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    155 #if 1
    156       // 0=L, 1=GF, Intra-layer prediction enabled.
    157       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_UPD_GF |
    158           VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
    159       layer_flags[1] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
    160           VP8_EFLAG_NO_REF_ARF;
    161 #else
    162        // 0=L, 1=GF, Intra-layer prediction disabled.
    163       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_UPD_GF |
    164           VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
    165       layer_flags[1] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
    166           VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_REF_LAST;
    167 #endif
    168       break;
    169     }
    170     case 2: {
    171       // 2-layers, 3-frame period.
    172       int ids[3] = {0, 1, 1};
    173       cfg->ts_periodicity = 3;
    174       *flag_periodicity = 3;
    175       cfg->ts_number_layers = 2;
    176       cfg->ts_rate_decimator[0] = 3;
    177       cfg->ts_rate_decimator[1] = 1;
    178       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    179       // 0=L, 1=GF, Intra-layer prediction enabled.
    180       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_REF_GF |
    181           VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    182       layer_flags[1] =
    183       layer_flags[2] = VP8_EFLAG_NO_REF_GF  | VP8_EFLAG_NO_REF_ARF |
    184           VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
    185       break;
    186     }
    187     case 3: {
    188       // 3-layers, 6-frame period.
    189       int ids[6] = {0, 2, 2, 1, 2, 2};
    190       cfg->ts_periodicity = 6;
    191       *flag_periodicity = 6;
    192       cfg->ts_number_layers = 3;
    193       cfg->ts_rate_decimator[0] = 6;
    194       cfg->ts_rate_decimator[1] = 3;
    195       cfg->ts_rate_decimator[2] = 1;
    196       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    197       // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
    198       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_REF_GF |
    199           VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    200       layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_ARF |
    201           VP8_EFLAG_NO_UPD_LAST;
    202       layer_flags[1] =
    203       layer_flags[2] =
    204       layer_flags[4] =
    205       layer_flags[5] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_LAST;
    206       break;
    207     }
    208     case 4: {
    209       // 3-layers, 4-frame period.
    210       int ids[4] = {0, 2, 1, 2};
    211       cfg->ts_periodicity = 4;
    212       *flag_periodicity = 4;
    213       cfg->ts_number_layers = 3;
    214       cfg->ts_rate_decimator[0] = 4;
    215       cfg->ts_rate_decimator[1] = 2;
    216       cfg->ts_rate_decimator[2] = 1;
    217       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    218       // 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
    219       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_REF_GF |
    220           VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    221       layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
    222           VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
    223       layer_flags[1] =
    224       layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
    225           VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    226       break;
    227     }
    228     case 5: {
    229       // 3-layers, 4-frame period.
    230       int ids[4] = {0, 2, 1, 2};
    231       cfg->ts_periodicity = 4;
    232       *flag_periodicity = 4;
    233       cfg->ts_number_layers     = 3;
    234       cfg->ts_rate_decimator[0] = 4;
    235       cfg->ts_rate_decimator[1] = 2;
    236       cfg->ts_rate_decimator[2] = 1;
    237       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    238       // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled in layer 1, disabled
    239       // in layer 2.
    240       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_REF_GF |
    241           VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    242       layer_flags[2] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
    243           VP8_EFLAG_NO_UPD_ARF;
    244       layer_flags[1] =
    245       layer_flags[3] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
    246           VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    247       break;
    248     }
    249     case 6: {
    250       // 3-layers, 4-frame period.
    251       int ids[4] = {0, 2, 1, 2};
    252       cfg->ts_periodicity = 4;
    253       *flag_periodicity = 4;
    254       cfg->ts_number_layers = 3;
    255       cfg->ts_rate_decimator[0] = 4;
    256       cfg->ts_rate_decimator[1] = 2;
    257       cfg->ts_rate_decimator[2] = 1;
    258       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    259       // 0=L, 1=GF, 2=ARF, Intra-layer prediction enabled.
    260       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_REF_GF |
    261           VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    262       layer_flags[2] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
    263           VP8_EFLAG_NO_UPD_ARF;
    264       layer_flags[1] =
    265       layer_flags[3] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
    266       break;
    267     }
    268     case 7: {
    269       // NOTE: Probably of academic interest only.
    270       // 5-layers, 16-frame period.
    271       int ids[16] = {0, 4, 3, 4, 2, 4, 3, 4, 1, 4, 3, 4, 2, 4, 3, 4};
    272       cfg->ts_periodicity = 16;
    273       *flag_periodicity = 16;
    274       cfg->ts_number_layers = 5;
    275       cfg->ts_rate_decimator[0] = 16;
    276       cfg->ts_rate_decimator[1] = 8;
    277       cfg->ts_rate_decimator[2] = 4;
    278       cfg->ts_rate_decimator[3] = 2;
    279       cfg->ts_rate_decimator[4] = 1;
    280       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    281       layer_flags[0]  = VPX_EFLAG_FORCE_KF;
    282       layer_flags[1]  =
    283       layer_flags[3]  =
    284       layer_flags[5]  =
    285       layer_flags[7]  =
    286       layer_flags[9]  =
    287       layer_flags[11] =
    288       layer_flags[13] =
    289       layer_flags[15] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
    290           VP8_EFLAG_NO_UPD_ARF;
    291       layer_flags[2]  =
    292       layer_flags[6]  =
    293       layer_flags[10] =
    294       layer_flags[14] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_GF;
    295       layer_flags[4] =
    296       layer_flags[12] = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_UPD_ARF;
    297       layer_flags[8]  = VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_GF;
    298       break;
    299     }
    300     case 8: {
    301       // 2-layers, with sync point at first frame of layer 1.
    302       int ids[2] = {0, 1};
    303       cfg->ts_periodicity = 2;
    304       *flag_periodicity = 8;
    305       cfg->ts_number_layers = 2;
    306       cfg->ts_rate_decimator[0] = 2;
    307       cfg->ts_rate_decimator[1] = 1;
    308       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    309       // 0=L, 1=GF.
    310       // ARF is used as predictor for all frames, and is only updated on
    311       // key frame. Sync point every 8 frames.
    312 
    313       // Layer 0: predict from L and ARF, update L and G.
    314       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_REF_GF |
    315           VP8_EFLAG_NO_UPD_ARF;
    316       // Layer 1: sync point: predict from L and ARF, and update G.
    317       layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_LAST |
    318           VP8_EFLAG_NO_UPD_ARF;
    319       // Layer 0, predict from L and ARF, update L.
    320       layer_flags[2] = VP8_EFLAG_NO_REF_GF  | VP8_EFLAG_NO_UPD_GF |
    321           VP8_EFLAG_NO_UPD_ARF;
    322       // Layer 1: predict from L, G and ARF, and update G.
    323       layer_flags[3] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
    324           VP8_EFLAG_NO_UPD_ENTROPY;
    325       // Layer 0.
    326       layer_flags[4] = layer_flags[2];
    327       // Layer 1.
    328       layer_flags[5] = layer_flags[3];
    329       // Layer 0.
    330       layer_flags[6] = layer_flags[4];
    331       // Layer 1.
    332       layer_flags[7] = layer_flags[5];
    333      break;
    334     }
    335     case 9: {
    336       // 3-layers: Sync points for layer 1 and 2 every 8 frames.
    337       int ids[4] = {0, 2, 1, 2};
    338       cfg->ts_periodicity = 4;
    339       *flag_periodicity = 8;
    340       cfg->ts_number_layers = 3;
    341       cfg->ts_rate_decimator[0] = 4;
    342       cfg->ts_rate_decimator[1] = 2;
    343       cfg->ts_rate_decimator[2] = 1;
    344       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    345       // 0=L, 1=GF, 2=ARF.
    346       layer_flags[0] = VPX_EFLAG_FORCE_KF  | VP8_EFLAG_NO_REF_GF |
    347           VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    348       layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
    349           VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
    350       layer_flags[2] = VP8_EFLAG_NO_REF_GF   | VP8_EFLAG_NO_REF_ARF |
    351           VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ARF;
    352       layer_flags[3] =
    353       layer_flags[5] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
    354       layer_flags[4] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
    355           VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
    356       layer_flags[6] = VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_UPD_LAST |
    357           VP8_EFLAG_NO_UPD_ARF;
    358       layer_flags[7] = VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
    359           VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_ENTROPY;
    360       break;
    361     }
    362     case 10: {
    363       // 3-layers structure where ARF is used as predictor for all frames,
    364       // and is only updated on key frame.
    365       // Sync points for layer 1 and 2 every 8 frames.
    366 
    367       int ids[4] = {0, 2, 1, 2};
    368       cfg->ts_periodicity = 4;
    369       *flag_periodicity = 8;
    370       cfg->ts_number_layers = 3;
    371       cfg->ts_rate_decimator[0] = 4;
    372       cfg->ts_rate_decimator[1] = 2;
    373       cfg->ts_rate_decimator[2] = 1;
    374       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    375       // 0=L, 1=GF, 2=ARF.
    376       // Layer 0: predict from L and ARF; update L and G.
    377       layer_flags[0] = VPX_EFLAG_FORCE_KF | VP8_EFLAG_NO_UPD_ARF |
    378           VP8_EFLAG_NO_REF_GF;
    379       // Layer 2: sync point: predict from L and ARF; update none.
    380       layer_flags[1] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_GF |
    381           VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST |
    382           VP8_EFLAG_NO_UPD_ENTROPY;
    383       // Layer 1: sync point: predict from L and ARF; update G.
    384       layer_flags[2] = VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_UPD_ARF |
    385           VP8_EFLAG_NO_UPD_LAST;
    386       // Layer 2: predict from L, G, ARF; update none.
    387       layer_flags[3] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
    388           VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ENTROPY;
    389       // Layer 0: predict from L and ARF; update L.
    390       layer_flags[4] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
    391           VP8_EFLAG_NO_REF_GF;
    392       // Layer 2: predict from L, G, ARF; update none.
    393       layer_flags[5] = layer_flags[3];
    394       // Layer 1: predict from L, G, ARF; update G.
    395       layer_flags[6] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
    396       // Layer 2: predict from L, G, ARF; update none.
    397       layer_flags[7] = layer_flags[3];
    398       break;
    399     }
    400     case 11:
    401     default: {
    402       // 3-layers structure as in case 10, but no sync/refresh points for
    403       // layer 1 and 2.
    404       int ids[4] = {0, 2, 1, 2};
    405       cfg->ts_periodicity = 4;
    406       *flag_periodicity = 8;
    407       cfg->ts_number_layers = 3;
    408       cfg->ts_rate_decimator[0] = 4;
    409       cfg->ts_rate_decimator[1] = 2;
    410       cfg->ts_rate_decimator[2] = 1;
    411       memcpy(cfg->ts_layer_id, ids, sizeof(ids));
    412       // 0=L, 1=GF, 2=ARF.
    413       // Layer 0: predict from L and ARF; update L.
    414       layer_flags[0] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
    415           VP8_EFLAG_NO_REF_GF;
    416       layer_flags[4] = layer_flags[0];
    417       // Layer 1: predict from L, G, ARF; update G.
    418       layer_flags[2] = VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
    419       layer_flags[6] = layer_flags[2];
    420       // Layer 2: predict from L, G, ARF; update none.
    421       layer_flags[1] = VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
    422           VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_ENTROPY;
    423       layer_flags[3] = layer_flags[1];
    424       layer_flags[5] = layer_flags[1];
    425       layer_flags[7] = layer_flags[1];
    426       break;
    427     }
    428   }
    429 }
    430 
    431 int main(int argc, char **argv) {
    432   VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS];
    433   vpx_codec_ctx_t codec;
    434   vpx_codec_enc_cfg_t cfg;
    435   int frame_cnt = 0;
    436   vpx_image_t raw;
    437   vpx_codec_err_t res;
    438   unsigned int width;
    439   unsigned int height;
    440   int speed;
    441   int frame_avail;
    442   int got_data;
    443   int flags = 0;
    444   unsigned int i;
    445   int pts = 0;  // PTS starts at 0.
    446   int frame_duration = 1;  // 1 timebase tick per frame.
    447   int layering_mode = 0;
    448   int layer_flags[VPX_TS_MAX_PERIODICITY] = {0};
    449   int flag_periodicity = 1;
    450   int max_intra_size_pct;
    451   vpx_svc_layer_id_t layer_id = {0, 0};
    452   const VpxInterface *encoder = NULL;
    453   FILE *infile = NULL;
    454   struct RateControlMetrics rc;
    455   int64_t cx_time = 0;
    456 
    457   exec_name = argv[0];
    458   // Check usage and arguments.
    459   if (argc < 11) {
    460     die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
    461         "<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
    462         "<Rate_0> ... <Rate_nlayers-1> \n", argv[0]);
    463   }
    464 
    465   encoder = get_vpx_encoder_by_name(argv[3]);
    466   if (!encoder)
    467     die("Unsupported codec.");
    468 
    469   printf("Using %s\n", vpx_codec_iface_name(encoder->interface()));
    470 
    471   width = strtol(argv[4], NULL, 0);
    472   height = strtol(argv[5], NULL, 0);
    473   if (width < 16 || width % 2 || height < 16 || height % 2) {
    474     die("Invalid resolution: %d x %d", width, height);
    475   }
    476 
    477   layering_mode = strtol(argv[10], NULL, 0);
    478   if (layering_mode < 0 || layering_mode > 12) {
    479     die("Invalid layering mode (0..12) %s", argv[10]);
    480   }
    481 
    482   if (argc != 11 + mode_to_num_layers[layering_mode]) {
    483     die("Invalid number of arguments");
    484   }
    485 
    486   if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {
    487     die("Failed to allocate image", width, height);
    488   }
    489 
    490   // Populate encoder configuration.
    491   res = vpx_codec_enc_config_default(encoder->interface(), &cfg, 0);
    492   if (res) {
    493     printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));
    494     return EXIT_FAILURE;
    495   }
    496 
    497   // Update the default configuration with our settings.
    498   cfg.g_w = width;
    499   cfg.g_h = height;
    500 
    501   // Timebase format e.g. 30fps: numerator=1, demoninator = 30.
    502   cfg.g_timebase.num = strtol(argv[6], NULL, 0);
    503   cfg.g_timebase.den = strtol(argv[7], NULL, 0);
    504 
    505   speed = strtol(argv[8], NULL, 0);
    506   if (speed < 0) {
    507     die("Invalid speed setting: must be positive");
    508   }
    509 
    510   for (i = 11; (int)i < 11 + mode_to_num_layers[layering_mode]; ++i) {
    511     cfg.ts_target_bitrate[i - 11] = strtol(argv[i], NULL, 0);
    512   }
    513 
    514   // Real time parameters.
    515   cfg.rc_dropframe_thresh = strtol(argv[9], NULL, 0);
    516   cfg.rc_end_usage = VPX_CBR;
    517   cfg.rc_resize_allowed = 0;
    518   cfg.rc_min_quantizer = 2;
    519   cfg.rc_max_quantizer = 56;
    520   cfg.rc_undershoot_pct = 50;
    521   cfg.rc_overshoot_pct = 50;
    522   cfg.rc_buf_initial_sz = 500;
    523   cfg.rc_buf_optimal_sz = 600;
    524   cfg.rc_buf_sz = 1000;
    525 
    526   // Enable error resilient mode.
    527   cfg.g_error_resilient = 1;
    528   cfg.g_lag_in_frames   = 0;
    529   cfg.kf_mode = VPX_KF_DISABLED;
    530 
    531   // Disable automatic keyframe placement.
    532   cfg.kf_min_dist = cfg.kf_max_dist = 3000;
    533 
    534   set_temporal_layer_pattern(layering_mode,
    535                              &cfg,
    536                              layer_flags,
    537                              &flag_periodicity);
    538 
    539   set_rate_control_metrics(&rc, &cfg);
    540 
    541   // Target bandwidth for the whole stream.
    542   // Set to ts_target_bitrate for highest layer (total bitrate).
    543   cfg.rc_target_bitrate = cfg.ts_target_bitrate[cfg.ts_number_layers - 1];
    544 
    545   // Open input file.
    546   if (!(infile = fopen(argv[1], "rb"))) {
    547     die("Failed to open %s for reading", argv[1]);
    548   }
    549 
    550   // Open an output file for each stream.
    551   for (i = 0; i < cfg.ts_number_layers; ++i) {
    552     char file_name[PATH_MAX];
    553     VpxVideoInfo info;
    554     info.codec_fourcc = encoder->fourcc;
    555     info.frame_width = cfg.g_w;
    556     info.frame_height = cfg.g_h;
    557     info.time_base.numerator = cfg.g_timebase.num;
    558     info.time_base.denominator = cfg.g_timebase.den;
    559 
    560     snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i);
    561     outfile[i] = vpx_video_writer_open(file_name, kContainerIVF, &info);
    562     if (!outfile[i])
    563       die("Failed to open %s for writing", file_name);
    564   }
    565   // No spatial layers in this encoder.
    566   cfg.ss_number_layers = 1;
    567 
    568   // Initialize codec.
    569   if (vpx_codec_enc_init(&codec, encoder->interface(), &cfg, 0))
    570     die_codec(&codec, "Failed to initialize encoder");
    571 
    572   if (strncmp(encoder->name, "vp8", 3) == 0) {
    573     vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed);
    574     vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, 1);
    575   } else if (strncmp(encoder->name, "vp9", 3) == 0) {
    576       vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed);
    577       vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
    578       vpx_codec_control(&codec, VP9E_SET_FRAME_PERIODIC_BOOST, 0);
    579       vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, 0);
    580       if (vpx_codec_control(&codec, VP9E_SET_SVC, 1)) {
    581         die_codec(&codec, "Failed to set SVC");
    582     }
    583   }
    584   vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
    585   vpx_codec_control(&codec, VP8E_SET_TOKEN_PARTITIONS, 1);
    586   // This controls the maximum target size of the key frame.
    587   // For generating smaller key frames, use a smaller max_intra_size_pct
    588   // value, like 100 or 200.
    589   max_intra_size_pct = (int) (((double)cfg.rc_buf_optimal_sz * 0.5)
    590       * ((double) cfg.g_timebase.den / cfg.g_timebase.num) / 10.0);
    591   // For low-quality key frame.
    592   max_intra_size_pct = 200;
    593   vpx_codec_control(&codec, VP8E_SET_MAX_INTRA_BITRATE_PCT, max_intra_size_pct);
    594 
    595   frame_avail = 1;
    596   while (frame_avail || got_data) {
    597     struct vpx_usec_timer timer;
    598     vpx_codec_iter_t iter = NULL;
    599     const vpx_codec_cx_pkt_t *pkt;
    600     // Update the temporal layer_id. No spatial layers in this test.
    601     layer_id.spatial_layer_id = 0;
    602     layer_id.temporal_layer_id =
    603         cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
    604     if (strncmp(encoder->name, "vp9", 3) == 0) {
    605       vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id);
    606     }
    607     flags = layer_flags[frame_cnt % flag_periodicity];
    608     frame_avail = vpx_img_read(&raw, infile);
    609     if (frame_avail)
    610       ++rc.layer_input_frames[layer_id.temporal_layer_id];
    611     vpx_usec_timer_start(&timer);
    612     if (vpx_codec_encode(&codec, frame_avail? &raw : NULL, pts, 1, flags,
    613         VPX_DL_REALTIME)) {
    614       die_codec(&codec, "Failed to encode frame");
    615     }
    616     vpx_usec_timer_mark(&timer);
    617     cx_time += vpx_usec_timer_elapsed(&timer);
    618     // Reset KF flag.
    619     if (layering_mode != 7) {
    620       layer_flags[0] &= ~VPX_EFLAG_FORCE_KF;
    621     }
    622     got_data = 0;
    623     while ( (pkt = vpx_codec_get_cx_data(&codec, &iter)) ) {
    624       got_data = 1;
    625       switch (pkt->kind) {
    626         case VPX_CODEC_CX_FRAME_PKT:
    627           for (i = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
    628               i < cfg.ts_number_layers; ++i) {
    629             vpx_video_writer_write_frame(outfile[i], pkt->data.frame.buf,
    630                                          pkt->data.frame.sz, pts);
    631             ++rc.layer_tot_enc_frames[i];
    632             rc.layer_encoding_bitrate[i] += 8.0 * pkt->data.frame.sz;
    633             // Keep count of rate control stats per layer (for non-key frames).
    634             if (i == cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity] &&
    635                 !(pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
    636               rc.layer_avg_frame_size[i] += 8.0 * pkt->data.frame.sz;
    637               rc.layer_avg_rate_mismatch[i] +=
    638                   fabs(8.0 * pkt->data.frame.sz - rc.layer_pfb[i]) /
    639                   rc.layer_pfb[i];
    640               ++rc.layer_enc_frames[i];
    641             }
    642           }
    643           break;
    644           default:
    645             break;
    646       }
    647     }
    648     ++frame_cnt;
    649     pts += frame_duration;
    650   }
    651   fclose(infile);
    652   printout_rate_control_summary(&rc, &cfg, frame_cnt);
    653   printf("\n");
    654   printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
    655           frame_cnt,
    656           1000 * (float)cx_time / (double)(frame_cnt * 1000000),
    657           1000000 * (double)frame_cnt / (double)cx_time);
    658 
    659   if (vpx_codec_destroy(&codec))
    660     die_codec(&codec, "Failed to destroy codec");
    661 
    662   // Try to rewrite the output file headers with the actual frame count.
    663   for (i = 0; i < cfg.ts_number_layers; ++i)
    664     vpx_video_writer_close(outfile[i]);
    665 
    666   return EXIT_SUCCESS;
    667 }
    668