examples/vp9_spatial_svc_encoder.c

   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 /*
  12  * This is an example demonstrating how to implement a multi-layer
  13  * VP9 encoding scheme based on spatial scalability for video applications
  14  * that benefit from a scalable bitstream.
  15  */
  16
  17 #include <math.h>
  18 #include <stdarg.h>
  19 #include <stdlib.h>
  20 #include <string.h>
  21 #include <time.h>
  22
  23
  24 #include "../args.h"
  25 #include "../tools_common.h"
  26 #include "../video_writer.h"
  27
  28 #include "vpx/svc_context.h"
  29 #include "vpx/vp8cx.h"
  30 #include "vpx/vpx_encoder.h"
  31 #include "../vpxstats.h"
  32 #define OUTPUT_RC_STATS 1
  33
  34 static const arg_def_t skip_frames_arg =
  35     ARG_DEF("s", "skip-frames", 1, "input frames to skip");
  36 static const arg_def_t frames_arg =
  37     ARG_DEF("f", "frames", 1, "number of frames to encode");
  38 static const arg_def_t threads_arg =
  39     ARG_DEF("th", "threads", 1, "number of threads to use");
  40 #if OUTPUT_RC_STATS
  41 static const arg_def_t output_rc_stats_arg =
  42     ARG_DEF("rcstat", "output_rc_stats", 1, "output rc stats");
  43 #endif
  44 static const arg_def_t width_arg = ARG_DEF("w", "width", 1, "source width");
  45 static const arg_def_t height_arg = ARG_DEF("h", "height", 1, "source height");
  46 static const arg_def_t timebase_arg =
  47     ARG_DEF("t", "timebase", 1, "timebase (num/den)");
  48 static const arg_def_t bitrate_arg = ARG_DEF(
  49     "b", "target-bitrate", 1, "encoding bitrate, in kilobits per second");
  50 static const arg_def_t spatial_layers_arg =
  51     ARG_DEF("sl", "spatial-layers", 1, "number of spatial SVC layers");
  52 static const arg_def_t temporal_layers_arg =
  53     ARG_DEF("tl", "temporal-layers", 1, "number of temporal SVC layers");
  54 static const arg_def_t temporal_layering_mode_arg =
  55     ARG_DEF("tlm", "temporal-layering-mode", 1, "temporal layering scheme."
  56         "VP9E_TEMPORAL_LAYERING_MODE");
  57 static const arg_def_t kf_dist_arg =
  58     ARG_DEF("k", "kf-dist", 1, "number of frames between keyframes");
  59 static const arg_def_t scale_factors_arg =
  60     ARG_DEF("r", "scale-factors", 1, "scale factors (lowest to highest layer)");
  61 static const arg_def_t passes_arg =
  62     ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
  63 static const arg_def_t pass_arg =
  64     ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
  65 static const arg_def_t fpf_name_arg =
  66     ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");
  67 static const arg_def_t min_q_arg =
  68     ARG_DEF(NULL, "min-q", 1, "Minimum quantizer");
  69 static const arg_def_t max_q_arg =
  70     ARG_DEF(NULL, "max-q", 1, "Maximum quantizer");
  71 static const arg_def_t min_bitrate_arg =
  72     ARG_DEF(NULL, "min-bitrate", 1, "Minimum bitrate");
  73 static const arg_def_t max_bitrate_arg =
  74     ARG_DEF(NULL, "max-bitrate", 1, "Maximum bitrate");
  75 static const arg_def_t lag_in_frame_arg =
  76     ARG_DEF(NULL, "lag-in-frames", 1, "Number of frame to input before "
  77         "generating any outputs");
  78 static const arg_def_t rc_end_usage_arg =
  79     ARG_DEF(NULL, "rc-end-usage", 1, "0 - 3: VBR, CBR, CQ, Q");
  80 static const arg_def_t speed_arg =
  81     ARG_DEF("sp", "speed", 1, "speed configuration");
  82
  83 #if CONFIG_VP9_HIGHBITDEPTH
  84 static const struct arg_enum_list bitdepth_enum[] = {
  85   {"8",  VPX_BITS_8},
  86   {"10", VPX_BITS_10},
  87   {"12", VPX_BITS_12},
  88   {NULL, 0}
  89 };
  90
  91 static const arg_def_t bitdepth_arg =
  92     ARG_DEF_ENUM("d", "bit-depth", 1, "Bit depth for codec 8, 10 or 12. ",
  93                  bitdepth_enum);
  94 #endif  // CONFIG_VP9_HIGHBITDEPTH
  95
  96
  97 static const arg_def_t *svc_args[] = {
  98   &frames_arg,        &width_arg,         &height_arg,
  99   &timebase_arg,      &bitrate_arg,       &skip_frames_arg, &spatial_layers_arg,
 100   &kf_dist_arg,       &scale_factors_arg, &passes_arg,      &pass_arg,
 101   &fpf_name_arg,      &min_q_arg,         &max_q_arg,       &min_bitrate_arg,
 102   &max_bitrate_arg,   &temporal_layers_arg, &temporal_layering_mode_arg,
 103   &lag_in_frame_arg,  &threads_arg,
 104 #if OUTPUT_RC_STATS
 105   &output_rc_stats_arg,
 106 #endif
 107
 108 #if CONFIG_VP9_HIGHBITDEPTH
 109   &bitdepth_arg,
 110 #endif
 111   &speed_arg,
 112   &rc_end_usage_arg,  NULL
 113 };
 114
 115 static const uint32_t default_frames_to_skip = 0;
 116 static const uint32_t default_frames_to_code = 60 * 60;
 117 static const uint32_t default_width = 1920;
 118 static const uint32_t default_height = 1080;
 119 static const uint32_t default_timebase_num = 1;
 120 static const uint32_t default_timebase_den = 60;
 121 static const uint32_t default_bitrate = 1000;
 122 static const uint32_t default_spatial_layers = 5;
 123 static const uint32_t default_temporal_layers = 1;
 124 static const uint32_t default_kf_dist = 100;
 125 static const uint32_t default_temporal_layering_mode = 0;
 126 static const uint32_t default_output_rc_stats = 0;
 127 static const int32_t default_speed = -1;  // -1 means use library default.
 128 static const uint32_t default_threads = 0;  // zero means use library default.
 129
 130 typedef struct {
 131   const char *input_filename;
 132   const char *output_filename;
 133   uint32_t frames_to_code;
 134   uint32_t frames_to_skip;
 135   struct VpxInputContext input_ctx;
 136   stats_io_t rc_stats;
 137   int passes;
 138   int pass;
 139 } AppInput;
 140
 141 static const char *exec_name;
 142
 143 void usage_exit(void) {
 144   fprintf(stderr, "Usage: %s <options> input_filename output_filename\n",
 145           exec_name);
 146   fprintf(stderr, "Options:\n");
 147   arg_show_usage(stderr, svc_args);
 148   exit(EXIT_FAILURE);
 149 }
 150
 151 static void parse_command_line(int argc, const char **argv_,
 152                                AppInput *app_input, SvcContext *svc_ctx,
 153                                vpx_codec_enc_cfg_t *enc_cfg) {
 154   struct arg arg = {0};
 155   char **argv = NULL;
 156   char **argi = NULL;
 157   char **argj = NULL;
 158   vpx_codec_err_t res;
 159   int passes = 0;
 160   int pass = 0;
 161   const char *fpf_file_name = NULL;
 162   unsigned int min_bitrate = 0;
 163   unsigned int max_bitrate = 0;
 164   char string_options[1024] = {0};
 165
 166   // initialize SvcContext with parameters that will be passed to vpx_svc_init
 167   svc_ctx->log_level = SVC_LOG_DEBUG;
 168   svc_ctx->spatial_layers = default_spatial_layers;
 169   svc_ctx->temporal_layers = default_temporal_layers;
 170   svc_ctx->temporal_layering_mode = default_temporal_layering_mode;
 171 #if OUTPUT_RC_STATS
 172   svc_ctx->output_rc_stat = default_output_rc_stats;
 173 #endif
 174   svc_ctx->speed = default_speed;
 175   svc_ctx->threads = default_threads;
 176
 177   // start with default encoder configuration
 178   res = vpx_codec_enc_config_default(vpx_codec_vp9_cx(), enc_cfg, 0);
 179   if (res) {
 180     die("Failed to get config: %s\n", vpx_codec_err_to_string(res));
 181   }
 182   // update enc_cfg with app default values
 183   enc_cfg->g_w = default_width;
 184   enc_cfg->g_h = default_height;
 185   enc_cfg->g_timebase.num = default_timebase_num;
 186   enc_cfg->g_timebase.den = default_timebase_den;
 187   enc_cfg->rc_target_bitrate = default_bitrate;
 188   enc_cfg->kf_min_dist = default_kf_dist;
 189   enc_cfg->kf_max_dist = default_kf_dist;
 190   enc_cfg->rc_end_usage = VPX_CQ;
 191
 192   // initialize AppInput with default values
 193   app_input->frames_to_code = default_frames_to_code;
 194   app_input->frames_to_skip = default_frames_to_skip;
 195
 196   // process command line options
 197   argv = argv_dup(argc - 1, argv_ + 1);
 198   for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
 199     arg.argv_step = 1;
 200
 201     if (arg_match(&arg, &frames_arg, argi)) {
 202       app_input->frames_to_code = arg_parse_uint(&arg);
 203     } else if (arg_match(&arg, &width_arg, argi)) {
 204       enc_cfg->g_w = arg_parse_uint(&arg);
 205     } else if (arg_match(&arg, &height_arg, argi)) {
 206       enc_cfg->g_h = arg_parse_uint(&arg);
 207     } else if (arg_match(&arg, &timebase_arg, argi)) {
 208       enc_cfg->g_timebase = arg_parse_rational(&arg);
 209     } else if (arg_match(&arg, &bitrate_arg, argi)) {
 210       enc_cfg->rc_target_bitrate = arg_parse_uint(&arg);
 211     } else if (arg_match(&arg, &skip_frames_arg, argi)) {
 212       app_input->frames_to_skip = arg_parse_uint(&arg);
 213     } else if (arg_match(&arg, &spatial_layers_arg, argi)) {
 214       svc_ctx->spatial_layers = arg_parse_uint(&arg);
 215     } else if (arg_match(&arg, &temporal_layers_arg, argi)) {
 216       svc_ctx->temporal_layers = arg_parse_uint(&arg);
 217 #if OUTPUT_RC_STATS
 218     } else if (arg_match(&arg, &output_rc_stats_arg, argi)) {
 219       svc_ctx->output_rc_stat = arg_parse_uint(&arg);
 220 #endif
 221     } else if (arg_match(&arg, &speed_arg, argi)) {
 222       svc_ctx->speed = arg_parse_uint(&arg);
 223     } else if (arg_match(&arg, &threads_arg, argi)) {
 224       svc_ctx->threads = arg_parse_uint(&arg);
 225     } else if (arg_match(&arg, &temporal_layering_mode_arg, argi)) {
 226       svc_ctx->temporal_layering_mode =
 227           enc_cfg->temporal_layering_mode = arg_parse_int(&arg);
 228       if (svc_ctx->temporal_layering_mode) {
 229         enc_cfg->g_error_resilient = 1;
 230       }
 231     } else if (arg_match(&arg, &kf_dist_arg, argi)) {
 232       enc_cfg->kf_min_dist = arg_parse_uint(&arg);
 233       enc_cfg->kf_max_dist = enc_cfg->kf_min_dist;
 234     } else if (arg_match(&arg, &scale_factors_arg, argi)) {
 235       snprintf(string_options, sizeof(string_options), "%s scale-factors=%s",
 236                string_options, arg.val);
 237     } else if (arg_match(&arg, &passes_arg, argi)) {
 238       passes = arg_parse_uint(&arg);
 239       if (passes < 1 || passes > 2) {
 240         die("Error: Invalid number of passes (%d)\n", passes);
 241       }
 242     } else if (arg_match(&arg, &pass_arg, argi)) {
 243       pass = arg_parse_uint(&arg);
 244       if (pass < 1 || pass > 2) {
 245         die("Error: Invalid pass selected (%d)\n", pass);
 246       }
 247     } else if (arg_match(&arg, &fpf_name_arg, argi)) {
 248       fpf_file_name = arg.val;
 249     } else if (arg_match(&arg, &min_q_arg, argi)) {
 250       snprintf(string_options, sizeof(string_options), "%s min-quantizers=%s",
 251                string_options, arg.val);
 252     } else if (arg_match(&arg, &max_q_arg, argi)) {
 253       snprintf(string_options, sizeof(string_options), "%s max-quantizers=%s",
 254                string_options, arg.val);
 255     } else if (arg_match(&arg, &min_bitrate_arg, argi)) {
 256       min_bitrate = arg_parse_uint(&arg);
 257     } else if (arg_match(&arg, &max_bitrate_arg, argi)) {
 258       max_bitrate = arg_parse_uint(&arg);
 259     } else if (arg_match(&arg, &lag_in_frame_arg, argi)) {
 260       enc_cfg->g_lag_in_frames = arg_parse_uint(&arg);
 261     } else if (arg_match(&arg, &rc_end_usage_arg, argi)) {
 262       enc_cfg->rc_end_usage = arg_parse_uint(&arg);
 263 #if CONFIG_VP9_HIGHBITDEPTH
 264     } else if (arg_match(&arg, &bitdepth_arg, argi)) {
 265       enc_cfg->g_bit_depth = arg_parse_enum_or_int(&arg);
 266       switch (enc_cfg->g_bit_depth) {
 267         case VPX_BITS_8:
 268           enc_cfg->g_input_bit_depth = 8;
 269           enc_cfg->g_profile = 0;
 270           break;
 271         case VPX_BITS_10:
 272           enc_cfg->g_input_bit_depth = 10;
 273           enc_cfg->g_profile = 2;
 274           break;
 275          case VPX_BITS_12:
 276           enc_cfg->g_input_bit_depth = 12;
 277           enc_cfg->g_profile = 2;
 278           break;
 279         default:
 280           die("Error: Invalid bit depth selected (%d)\n", enc_cfg->g_bit_depth);
 281           break;
 282       }
 283 #endif  // CONFIG_VP9_HIGHBITDEPTH
 284     } else {
 285       ++argj;
 286     }
 287   }
 288
 289   // There will be a space in front of the string options
 290   if (strlen(string_options) > 0)
 291     vpx_svc_set_options(svc_ctx, string_options + 1);
 292
 293   if (passes == 0 || passes == 1) {
 294     if (pass) {
 295       fprintf(stderr, "pass is ignored since there's only one pass\n");
 296     }
 297     enc_cfg->g_pass = VPX_RC_ONE_PASS;
 298   } else {
 299     if (pass == 0) {
 300       die("pass must be specified when passes is 2\n");
 301     }
 302
 303     if (fpf_file_name == NULL) {
 304       die("fpf must be specified when passes is 2\n");
 305     }
 306
 307     if (pass == 1) {
 308       enc_cfg->g_pass = VPX_RC_FIRST_PASS;
 309       if (!stats_open_file(&app_input->rc_stats, fpf_file_name, 0)) {
 310         fatal("Failed to open statistics store");
 311       }
 312     } else {
 313       enc_cfg->g_pass = VPX_RC_LAST_PASS;
 314       if (!stats_open_file(&app_input->rc_stats, fpf_file_name, 1)) {
 315         fatal("Failed to open statistics store");
 316       }
 317       enc_cfg->rc_twopass_stats_in = stats_get(&app_input->rc_stats);
 318     }
 319     app_input->passes = passes;
 320     app_input->pass = pass;
 321   }
 322
 323   if (enc_cfg->rc_target_bitrate > 0) {
 324     if (min_bitrate > 0) {
 325       enc_cfg->rc_2pass_vbr_minsection_pct =
 326           min_bitrate * 100 / enc_cfg->rc_target_bitrate;
 327     }
 328     if (max_bitrate > 0) {
 329       enc_cfg->rc_2pass_vbr_maxsection_pct =
 330           max_bitrate * 100 / enc_cfg->rc_target_bitrate;
 331     }
 332   }
 333
 334   // Check for unrecognized options
 335   for (argi = argv; *argi; ++argi)
 336     if (argi[0][0] == '-' && strlen(argi[0]) > 1)
 337       die("Error: Unrecognized option %s\n", *argi);
 338
 339   if (argv[0] == NULL || argv[1] == 0) {
 340     usage_exit();
 341   }
 342   app_input->input_filename = argv[0];
 343   app_input->output_filename = argv[1];
 344   free(argv);
 345
 346   if (enc_cfg->g_w < 16 || enc_cfg->g_w % 2 || enc_cfg->g_h < 16 ||
 347       enc_cfg->g_h % 2)
 348     die("Invalid resolution: %d x %d\n", enc_cfg->g_w, enc_cfg->g_h);
 349
 350   printf(
 351       "Codec %s\nframes: %d, skip: %d\n"
 352       "layers: %d\n"
 353       "width %d, height: %d,\n"
 354       "num: %d, den: %d, bitrate: %d,\n"
 355       "gop size: %d\n",
 356       vpx_codec_iface_name(vpx_codec_vp9_cx()), app_input->frames_to_code,
 357       app_input->frames_to_skip,
 358       svc_ctx->spatial_layers, enc_cfg->g_w, enc_cfg->g_h,
 359       enc_cfg->g_timebase.num, enc_cfg->g_timebase.den,
 360       enc_cfg->rc_target_bitrate, enc_cfg->kf_max_dist);
 361 }
 362
 363 #if OUTPUT_RC_STATS
 364 // For rate control encoding stats.
 365 struct RateControlStats {
 366   // Number of input frames per layer.
 367   int layer_input_frames[VPX_MAX_LAYERS];
 368   // Total (cumulative) number of encoded frames per layer.
 369   int layer_tot_enc_frames[VPX_MAX_LAYERS];
 370   // Number of encoded non-key frames per layer.
 371   int layer_enc_frames[VPX_MAX_LAYERS];
 372   // Framerate per layer (cumulative).
 373   double layer_framerate[VPX_MAX_LAYERS];
 374   // Target average frame size per layer (per-frame-bandwidth per layer).
 375   double layer_pfb[VPX_MAX_LAYERS];
 376   // Actual average frame size per layer.
 377   double layer_avg_frame_size[VPX_MAX_LAYERS];
 378   // Average rate mismatch per layer (|target - actual| / target).
 379   double layer_avg_rate_mismatch[VPX_MAX_LAYERS];
 380   // Actual encoding bitrate per layer (cumulative).
 381   double layer_encoding_bitrate[VPX_MAX_LAYERS];
 382   // Average of the short-time encoder actual bitrate.
 383   // TODO(marpan): Should we add these short-time stats for each layer?
 384   double avg_st_encoding_bitrate;
 385   // Variance of the short-time encoder actual bitrate.
 386   double variance_st_encoding_bitrate;
 387   // Window (number of frames) for computing short-time encoding bitrate.
 388   int window_size;
 389   // Number of window measurements.
 390   int window_count;
 391 };
 392
 393 // Note: these rate control stats assume only 1 key frame in the
 394 // sequence (i.e., first frame only).
 395 static void set_rate_control_stats(struct RateControlStats *rc,
 396                                      vpx_codec_enc_cfg_t *cfg) {
 397   unsigned int sl, tl;
 398   // Set the layer (cumulative) framerate and the target layer (non-cumulative)
 399   // per-frame-bandwidth, for the rate control encoding stats below.
 400   const double framerate = cfg->g_timebase.den / cfg->g_timebase.num;
 401
 402   for (sl = 0; sl < cfg->ss_number_layers; ++sl) {
 403     for (tl = 0; tl < cfg->ts_number_layers; ++tl) {
 404       const int layer = sl * cfg->ts_number_layers + tl;
 405       const int tlayer0 = sl * cfg->ts_number_layers;
 406       rc->layer_framerate[layer] =
 407           framerate / cfg->ts_rate_decimator[tl];
 408       if (tl > 0) {
 409         rc->layer_pfb[layer] = 1000.0 *
 410             (cfg->layer_target_bitrate[layer] -
 411                 cfg->layer_target_bitrate[layer - 1]) /
 412             (rc->layer_framerate[layer] -
 413                 rc->layer_framerate[layer - 1]);
 414       } else {
 415         rc->layer_pfb[tlayer0] = 1000.0 *
 416             cfg->layer_target_bitrate[tlayer0] /
 417             rc->layer_framerate[tlayer0];
 418       }
 419       rc->layer_input_frames[layer] = 0;
 420       rc->layer_enc_frames[layer] = 0;
 421       rc->layer_tot_enc_frames[layer] = 0;
 422       rc->layer_encoding_bitrate[layer] = 0.0;
 423       rc->layer_avg_frame_size[layer] = 0.0;
 424       rc->layer_avg_rate_mismatch[layer] = 0.0;
 425     }
 426   }
 427   rc->window_count = 0;
 428   rc->window_size = 15;
 429   rc->avg_st_encoding_bitrate = 0.0;
 430   rc->variance_st_encoding_bitrate = 0.0;
 431 }
 432
 433 static void printout_rate_control_summary(struct RateControlStats *rc,
 434                                           vpx_codec_enc_cfg_t *cfg,
 435                                           int frame_cnt) {
 436   unsigned int sl, tl;
 437   int tot_num_frames = 0;
 438   double perc_fluctuation = 0.0;
 439   printf("Total number of processed frames: %d\n\n", frame_cnt - 1);
 440   printf("Rate control layer stats for sl%d tl%d layer(s):\n\n",
 441       cfg->ss_number_layers, cfg->ts_number_layers);
 442   for (sl = 0; sl < cfg->ss_number_layers; ++sl) {
 443     for (tl = 0; tl < cfg->ts_number_layers; ++tl) {
 444       const int layer = sl * cfg->ts_number_layers + tl;
 445       const int num_dropped = (tl > 0) ?
 446           (rc->layer_input_frames[layer] - rc->layer_enc_frames[layer]) :
 447           (rc->layer_input_frames[layer] - rc->layer_enc_frames[layer] - 1);
 448       if (!sl)
 449         tot_num_frames += rc->layer_input_frames[layer];
 450       rc->layer_encoding_bitrate[layer] = 0.001 * rc->layer_framerate[layer] *
 451           rc->layer_encoding_bitrate[layer] / tot_num_frames;
 452       rc->layer_avg_frame_size[layer] = rc->layer_avg_frame_size[layer] /
 453           rc->layer_enc_frames[layer];
 454       rc->layer_avg_rate_mismatch[layer] =
 455           100.0 * rc->layer_avg_rate_mismatch[layer] /
 456           rc->layer_enc_frames[layer];
 457       printf("For layer#: sl%d tl%d \n", sl, tl);
 458       printf("Bitrate (target vs actual): %d %f.0 kbps\n",
 459              cfg->layer_target_bitrate[layer],
 460              rc->layer_encoding_bitrate[layer]);
 461       printf("Average frame size (target vs actual): %f %f bits\n",
 462              rc->layer_pfb[layer], rc->layer_avg_frame_size[layer]);
 463       printf("Average rate_mismatch: %f\n",
 464              rc->layer_avg_rate_mismatch[layer]);
 465       printf("Number of input frames, encoded (non-key) frames, "
 466           "and percent dropped frames: %d %d %f.0 \n",
 467           rc->layer_input_frames[layer], rc->layer_enc_frames[layer],
 468           100.0 * num_dropped / rc->layer_input_frames[layer]);
 469       printf("\n");
 470     }
 471   }
 472   rc->avg_st_encoding_bitrate = rc->avg_st_encoding_bitrate / rc->window_count;
 473   rc->variance_st_encoding_bitrate =
 474       rc->variance_st_encoding_bitrate / rc->window_count -
 475       (rc->avg_st_encoding_bitrate * rc->avg_st_encoding_bitrate);
 476   perc_fluctuation = 100.0 * sqrt(rc->variance_st_encoding_bitrate) /
 477       rc->avg_st_encoding_bitrate;
 478   printf("Short-time stats, for window of %d frames: \n", rc->window_size);
 479   printf("Average, rms-variance, and percent-fluct: %f %f %f \n",
 480          rc->avg_st_encoding_bitrate,
 481          sqrt(rc->variance_st_encoding_bitrate),
 482          perc_fluctuation);
 483   if (frame_cnt != tot_num_frames)
 484     die("Error: Number of input frames not equal to output encoded frames != "
 485         "%d tot_num_frames = %d\n", frame_cnt, tot_num_frames);
 486 }
 487
 488 vpx_codec_err_t parse_superframe_index(const uint8_t *data,
 489                                        size_t data_sz,
 490                                        uint32_t sizes[8], int *count) {
 491   // A chunk ending with a byte matching 0xc0 is an invalid chunk unless
 492   // it is a super frame index. If the last byte of real video compression
 493   // data is 0xc0 the encoder must add a 0 byte. If we have the marker but
 494   // not the associated matching marker byte at the front of the index we have
 495   // an invalid bitstream and need to return an error.
 496
 497   uint8_t marker;
 498
 499   marker = *(data + data_sz - 1);
 500   *count = 0;
 501
 502
 503   if ((marker & 0xe0) == 0xc0) {
 504     const uint32_t frames = (marker & 0x7) + 1;
 505     const uint32_t mag = ((marker >> 3) & 0x3) + 1;
 506     const size_t index_sz = 2 + mag * frames;
 507
 508     // This chunk is marked as having a superframe index but doesn't have
 509     // enough data for it, thus it's an invalid superframe index.
 510     if (data_sz < index_sz)
 511       return VPX_CODEC_CORRUPT_FRAME;
 512
 513     {
 514       const uint8_t marker2 = *(data + data_sz - index_sz);
 515
 516       // This chunk is marked as having a superframe index but doesn't have
 517       // the matching marker byte at the front of the index therefore it's an
 518       // invalid chunk.
 519       if (marker != marker2)
 520         return VPX_CODEC_CORRUPT_FRAME;
 521     }
 522
 523     {
 524       // Found a valid superframe index.
 525       uint32_t i, j;
 526       const uint8_t *x = &data[data_sz - index_sz + 1];
 527
 528       for (i = 0; i < frames; ++i) {
 529         uint32_t this_sz = 0;
 530
 531         for (j = 0; j < mag; ++j)
 532           this_sz |= (*x++) << (j * 8);
 533         sizes[i] = this_sz;
 534       }
 535       *count = frames;
 536     }
 537   }
 538   return VPX_CODEC_OK;
 539 }
 540 #endif
 541
 542 int main(int argc, const char **argv) {
 543   AppInput app_input = {0};
 544   VpxVideoWriter *writer = NULL;
 545   VpxVideoInfo info = {0};
 546   vpx_codec_ctx_t codec;
 547   vpx_codec_enc_cfg_t enc_cfg;
 548   SvcContext svc_ctx;
 549   uint32_t i;
 550   uint32_t frame_cnt = 0;
 551   vpx_image_t raw;
 552   vpx_codec_err_t res;
 553   int pts = 0;            /* PTS starts at 0 */
 554   int frame_duration = 1; /* 1 timebase tick per frame */
 555   FILE *infile = NULL;
 556   int end_of_stream = 0;
 557   int frames_received = 0;
 558 #if OUTPUT_RC_STATS
 559   VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = {NULL};
 560   struct RateControlStats rc;
 561   vpx_svc_layer_id_t layer_id;
 562   int sl, tl;
 563   double sum_bitrate = 0.0;
 564   double sum_bitrate2 = 0.0;
 565   double framerate  = 30.0;
 566 #endif
 567   memset(&svc_ctx, 0, sizeof(svc_ctx));
 568   svc_ctx.log_print = 1;
 569   exec_name = argv[0];
 570   parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg);
 571
 572   // Allocate image buffer
 573 #if CONFIG_VP9_HIGHBITDEPTH
 574   if (!vpx_img_alloc(&raw, enc_cfg.g_input_bit_depth == 8 ?
 575                          VPX_IMG_FMT_I420 : VPX_IMG_FMT_I42016,
 576                      enc_cfg.g_w, enc_cfg.g_h, 32)) {
 577     die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
 578   }
 579 #else
 580   if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, enc_cfg.g_w, enc_cfg.g_h, 32)) {
 581     die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
 582   }
 583 #endif  // CONFIG_VP9_HIGHBITDEPTH
 584
 585   if (!(infile = fopen(app_input.input_filename, "rb")))
 586     die("Failed to open %s for reading\n", app_input.input_filename);
 587
 588   // Initialize codec
 589   if (vpx_svc_init(&svc_ctx, &codec, vpx_codec_vp9_cx(), &enc_cfg) !=
 590       VPX_CODEC_OK)
 591     die("Failed to initialize encoder\n");
 592
 593 #if OUTPUT_RC_STATS
 594   if (svc_ctx.output_rc_stat) {
 595     set_rate_control_stats(&rc, &enc_cfg);
 596     framerate = enc_cfg.g_timebase.den / enc_cfg.g_timebase.num;
 597   }
 598 #endif
 599
 600   info.codec_fourcc = VP9_FOURCC;
 601   info.time_base.numerator = enc_cfg.g_timebase.num;
 602   info.time_base.denominator = enc_cfg.g_timebase.den;
 603
 604   if (!(app_input.passes == 2 && app_input.pass == 1)) {
 605     // We don't save the bitstream for the 1st pass on two pass rate control
 606     writer = vpx_video_writer_open(app_input.output_filename, kContainerIVF,
 607                                    &info);
 608     if (!writer)
 609       die("Failed to open %s for writing\n", app_input.output_filename);
 610   }
 611 #if OUTPUT_RC_STATS
 612   // For now, just write temporal layer streams.
 613   // TODO(wonkap): do spatial by re-writing superframe.
 614   if (svc_ctx.output_rc_stat) {
 615     for (tl = 0; tl < enc_cfg.ts_number_layers; ++tl) {
 616       char file_name[PATH_MAX];
 617
 618       snprintf(file_name, sizeof(file_name), "%s_t%d.ivf",
 619                app_input.output_filename, tl);
 620       outfile[tl] = vpx_video_writer_open(file_name, kContainerIVF, &info);
 621       if (!outfile[tl])
 622         die("Failed to open %s for writing", file_name);
 623     }
 624   }
 625 #endif
 626
 627   // skip initial frames
 628   for (i = 0; i < app_input.frames_to_skip; ++i)
 629     vpx_img_read(&raw, infile);
 630
 631   if (svc_ctx.speed != -1)
 632     vpx_codec_control(&codec, VP8E_SET_CPUUSED, svc_ctx.speed);
 633   if (svc_ctx.threads)
 634     vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, (svc_ctx.threads >> 1));
 635
 636   // Encode frames
 637   while (!end_of_stream) {
 638     vpx_codec_iter_t iter = NULL;
 639     const vpx_codec_cx_pkt_t *cx_pkt;
 640     if (frame_cnt >= app_input.frames_to_code || !vpx_img_read(&raw, infile)) {
 641       // We need one extra vpx_svc_encode call at end of stream to flush
 642       // encoder and get remaining data
 643       end_of_stream = 1;
 644     }
 645
 646     res = vpx_svc_encode(&svc_ctx, &codec, (end_of_stream ? NULL : &raw),
 647                          pts, frame_duration, svc_ctx.speed >= 5 ?
 648                          VPX_DL_REALTIME : VPX_DL_GOOD_QUALITY);
 649
 650     printf("%s", vpx_svc_get_message(&svc_ctx));
 651     if (res != VPX_CODEC_OK) {
 652       die_codec(&codec, "Failed to encode frame");
 653     }
 654
 655     while ((cx_pkt = vpx_codec_get_cx_data(&codec, &iter)) != NULL) {
 656       switch (cx_pkt->kind) {
 657         case VPX_CODEC_CX_FRAME_PKT: {
 658           if (cx_pkt->data.frame.sz > 0) {
 659 #if OUTPUT_RC_STATS
 660             uint32_t sizes[8];
 661             int count = 0;
 662 #endif
 663             vpx_video_writer_write_frame(writer,
 664                                          cx_pkt->data.frame.buf,
 665                                          cx_pkt->data.frame.sz,
 666                                          cx_pkt->data.frame.pts);
 667 #if OUTPUT_RC_STATS
 668             // TODO(marpan/wonkap): Put this (to line728) in separate function.
 669             if (svc_ctx.output_rc_stat) {
 670               vpx_codec_control(&codec, VP9E_GET_SVC_LAYER_ID, &layer_id);
 671               parse_superframe_index(cx_pkt->data.frame.buf,
 672                                      cx_pkt->data.frame.sz, sizes, &count);
 673               for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 674                 ++rc.layer_input_frames[sl * enc_cfg.ts_number_layers +
 675                                         layer_id.temporal_layer_id];
 676               }
 677               for (tl = layer_id.temporal_layer_id;
 678                   tl < enc_cfg.ts_number_layers; ++tl) {
 679                 vpx_video_writer_write_frame(outfile[tl],
 680                                              cx_pkt->data.frame.buf,
 681                                              cx_pkt->data.frame.sz,
 682                                              cx_pkt->data.frame.pts);
 683               }
 684
 685               for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 686                 for (tl = layer_id.temporal_layer_id;
 687                     tl < enc_cfg.ts_number_layers; ++tl) {
 688                   const int layer = sl * enc_cfg.ts_number_layers + tl;
 689                   ++rc.layer_tot_enc_frames[layer];
 690                   rc.layer_encoding_bitrate[layer] += 8.0 * sizes[sl];
 691                   // Keep count of rate control stats per layer, for non-key
 692                   // frames.
 693                   if (tl == layer_id.temporal_layer_id &&
 694                       !(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
 695                     rc.layer_avg_frame_size[layer] += 8.0 * sizes[sl];
 696                     rc.layer_avg_rate_mismatch[layer] +=
 697                         fabs(8.0 * sizes[sl] - rc.layer_pfb[layer]) /
 698                         rc.layer_pfb[layer];
 699                     ++rc.layer_enc_frames[layer];
 700                   }
 701                 }
 702               }
 703
 704               // Update for short-time encoding bitrate states, for moving
 705               // window of size rc->window, shifted by rc->window / 2.
 706               // Ignore first window segment, due to key frame.
 707               if (frame_cnt > rc.window_size) {
 708                 tl = layer_id.temporal_layer_id;
 709                 for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 710                   sum_bitrate += 0.001 * 8.0 * sizes[sl] * framerate;
 711                 }
 712                 if (frame_cnt % rc.window_size == 0) {
 713                   rc.window_count += 1;
 714                   rc.avg_st_encoding_bitrate += sum_bitrate / rc.window_size;
 715                   rc.variance_st_encoding_bitrate +=
 716                       (sum_bitrate / rc.window_size) *
 717                       (sum_bitrate / rc.window_size);
 718                   sum_bitrate = 0.0;
 719                 }
 720               }
 721
 722               // Second shifted window.
 723               if (frame_cnt > rc.window_size + rc.window_size / 2) {
 724                tl = layer_id.temporal_layer_id;
 725                for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 726                  sum_bitrate2 += 0.001 * 8.0 * sizes[sl] * framerate;
 727                }
 728
 729                if (frame_cnt > 2 * rc.window_size &&
 730                   frame_cnt % rc.window_size == 0) {
 731                  rc.window_count += 1;
 732                  rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
 733                  rc.variance_st_encoding_bitrate +=
 734                     (sum_bitrate2 / rc.window_size) *
 735                     (sum_bitrate2 / rc.window_size);
 736                  sum_bitrate2 = 0.0;
 737                }
 738               }
 739             }
 740 #endif
 741           }
 742
 743           printf("SVC frame: %d, kf: %d, size: %d, pts: %d\n", frames_received,
 744                  !!(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY),
 745                  (int)cx_pkt->data.frame.sz, (int)cx_pkt->data.frame.pts);
 746           ++frames_received;
 747           break;
 748         }
 749         case VPX_CODEC_STATS_PKT: {
 750           stats_write(&app_input.rc_stats,
 751                       cx_pkt->data.twopass_stats.buf,
 752                       cx_pkt->data.twopass_stats.sz);
 753           break;
 754         }
 755         default: {
 756           break;
 757         }
 758       }
 759     }
 760
 761     if (!end_of_stream) {
 762       ++frame_cnt;
 763       pts += frame_duration;
 764     }
 765   }
 766   printf("Processed %d frames\n", frame_cnt);
 767   fclose(infile);
 768 #if OUTPUT_RC_STATS
 769   if (svc_ctx.output_rc_stat) {
 770     printout_rate_control_summary(&rc, &enc_cfg, frame_cnt);
 771     printf("\n");
 772   }
 773 #endif
 774   if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
 775   if (app_input.passes == 2)
 776     stats_close(&app_input.rc_stats, 1);
 777   if (writer) {
 778     vpx_video_writer_close(writer);
 779   }
 780 #if OUTPUT_RC_STATS
 781   if (svc_ctx.output_rc_stat) {
 782     for (tl = 0; tl < enc_cfg.ts_number_layers; ++tl) {
 783       vpx_video_writer_close(outfile[tl]);
 784     }
 785   }
 786 #endif
 787   vpx_img_free(&raw);
 788   // display average size, psnr
 789   printf("%s", vpx_svc_dump_statistics(&svc_ctx));
 790   vpx_svc_release(&svc_ctx);
 791   return EXIT_SUCCESS;
 792 }