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_ports/vpx_timer.h"
  29 #include "vpx/svc_context.h"
  30 #include "vpx/vp8cx.h"
  31 #include "vpx/vpx_encoder.h"
  32 #include "../vpxstats.h"
  33 #include "vp9/encoder/vp9_encoder.h"
  34 #define OUTPUT_RC_STATS 1
  35
  36 static const arg_def_t skip_frames_arg =
  37     ARG_DEF("s", "skip-frames", 1, "input frames to skip");
  38 static const arg_def_t frames_arg =
  39     ARG_DEF("f", "frames", 1, "number of frames to encode");
  40 static const arg_def_t threads_arg =
  41     ARG_DEF("th", "threads", 1, "number of threads to use");
  42 #if OUTPUT_RC_STATS
  43 static const arg_def_t output_rc_stats_arg =
  44     ARG_DEF("rcstat", "output_rc_stats", 1, "output rc stats");
  45 #endif
  46 static const arg_def_t width_arg = ARG_DEF("w", "width", 1, "source width");
  47 static const arg_def_t height_arg = ARG_DEF("h", "height", 1, "source height");
  48 static const arg_def_t timebase_arg =
  49     ARG_DEF("t", "timebase", 1, "timebase (num/den)");
  50 static const arg_def_t bitrate_arg = ARG_DEF(
  51     "b", "target-bitrate", 1, "encoding bitrate, in kilobits per second");
  52 static const arg_def_t spatial_layers_arg =
  53     ARG_DEF("sl", "spatial-layers", 1, "number of spatial SVC layers");
  54 static const arg_def_t temporal_layers_arg =
  55     ARG_DEF("tl", "temporal-layers", 1, "number of temporal SVC layers");
  56 static const arg_def_t temporal_layering_mode_arg =
  57     ARG_DEF("tlm", "temporal-layering-mode", 1, "temporal layering scheme."
  58         "VP9E_TEMPORAL_LAYERING_MODE");
  59 static const arg_def_t kf_dist_arg =
  60     ARG_DEF("k", "kf-dist", 1, "number of frames between keyframes");
  61 static const arg_def_t scale_factors_arg =
  62     ARG_DEF("r", "scale-factors", 1, "scale factors (lowest to highest layer)");
  63 static const arg_def_t passes_arg =
  64     ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
  65 static const arg_def_t pass_arg =
  66     ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
  67 static const arg_def_t fpf_name_arg =
  68     ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");
  69 static const arg_def_t min_q_arg =
  70     ARG_DEF(NULL, "min-q", 1, "Minimum quantizer");
  71 static const arg_def_t max_q_arg =
  72     ARG_DEF(NULL, "max-q", 1, "Maximum quantizer");
  73 static const arg_def_t min_bitrate_arg =
  74     ARG_DEF(NULL, "min-bitrate", 1, "Minimum bitrate");
  75 static const arg_def_t max_bitrate_arg =
  76     ARG_DEF(NULL, "max-bitrate", 1, "Maximum bitrate");
  77 static const arg_def_t lag_in_frame_arg =
  78     ARG_DEF(NULL, "lag-in-frames", 1, "Number of frame to input before "
  79         "generating any outputs");
  80 static const arg_def_t rc_end_usage_arg =
  81     ARG_DEF(NULL, "rc-end-usage", 1, "0 - 3: VBR, CBR, CQ, Q");
  82 static const arg_def_t speed_arg =
  83     ARG_DEF("sp", "speed", 1, "speed configuration");
  84 static const arg_def_t aqmode_arg =
  85     ARG_DEF("aq", "aqmode", 1, "aq-mode off/on");
  86
  87 #if CONFIG_VP9_HIGHBITDEPTH
  88 static const struct arg_enum_list bitdepth_enum[] = {
  89   {"8",  VPX_BITS_8},
  90   {"10", VPX_BITS_10},
  91   {"12", VPX_BITS_12},
  92   {NULL, 0}
  93 };
  94
  95 static const arg_def_t bitdepth_arg =
  96     ARG_DEF_ENUM("d", "bit-depth", 1, "Bit depth for codec 8, 10 or 12. ",
  97                  bitdepth_enum);
  98 #endif  // CONFIG_VP9_HIGHBITDEPTH
  99
 100
 101 static const arg_def_t *svc_args[] = {
 102   &frames_arg,        &width_arg,         &height_arg,
 103   &timebase_arg,      &bitrate_arg,       &skip_frames_arg, &spatial_layers_arg,
 104   &kf_dist_arg,       &scale_factors_arg, &passes_arg,      &pass_arg,
 105   &fpf_name_arg,      &min_q_arg,         &max_q_arg,       &min_bitrate_arg,
 106   &max_bitrate_arg,   &temporal_layers_arg, &temporal_layering_mode_arg,
 107   &lag_in_frame_arg,  &threads_arg,       &aqmode_arg,
 108 #if OUTPUT_RC_STATS
 109   &output_rc_stats_arg,
 110 #endif
 111
 112 #if CONFIG_VP9_HIGHBITDEPTH
 113   &bitdepth_arg,
 114 #endif
 115   &speed_arg,
 116   &rc_end_usage_arg,  NULL
 117 };
 118
 119 static const uint32_t default_frames_to_skip = 0;
 120 static const uint32_t default_frames_to_code = 60 * 60;
 121 static const uint32_t default_width = 1920;
 122 static const uint32_t default_height = 1080;
 123 static const uint32_t default_timebase_num = 1;
 124 static const uint32_t default_timebase_den = 60;
 125 static const uint32_t default_bitrate = 1000;
 126 static const uint32_t default_spatial_layers = 5;
 127 static const uint32_t default_temporal_layers = 1;
 128 static const uint32_t default_kf_dist = 100;
 129 static const uint32_t default_temporal_layering_mode = 0;
 130 static const uint32_t default_output_rc_stats = 0;
 131 static const int32_t default_speed = -1;  // -1 means use library default.
 132 static const uint32_t default_threads = 0;  // zero means use library default.
 133
 134 typedef struct {
 135   const char *input_filename;
 136   const char *output_filename;
 137   uint32_t frames_to_code;
 138   uint32_t frames_to_skip;
 139   struct VpxInputContext input_ctx;
 140   stats_io_t rc_stats;
 141   int passes;
 142   int pass;
 143 } AppInput;
 144
 145 static const char *exec_name;
 146
 147 void usage_exit(void) {
 148   fprintf(stderr, "Usage: %s <options> input_filename output_filename\n",
 149           exec_name);
 150   fprintf(stderr, "Options:\n");
 151   arg_show_usage(stderr, svc_args);
 152   exit(EXIT_FAILURE);
 153 }
 154
 155 static void parse_command_line(int argc, const char **argv_,
 156                                AppInput *app_input, SvcContext *svc_ctx,
 157                                vpx_codec_enc_cfg_t *enc_cfg) {
 158   struct arg arg = {0};
 159   char **argv = NULL;
 160   char **argi = NULL;
 161   char **argj = NULL;
 162   vpx_codec_err_t res;
 163   int passes = 0;
 164   int pass = 0;
 165   const char *fpf_file_name = NULL;
 166   unsigned int min_bitrate = 0;
 167   unsigned int max_bitrate = 0;
 168   char string_options[1024] = {0};
 169
 170   // initialize SvcContext with parameters that will be passed to vpx_svc_init
 171   svc_ctx->log_level = SVC_LOG_DEBUG;
 172   svc_ctx->spatial_layers = default_spatial_layers;
 173   svc_ctx->temporal_layers = default_temporal_layers;
 174   svc_ctx->temporal_layering_mode = default_temporal_layering_mode;
 175 #if OUTPUT_RC_STATS
 176   svc_ctx->output_rc_stat = default_output_rc_stats;
 177 #endif
 178   svc_ctx->speed = default_speed;
 179   svc_ctx->threads = default_threads;
 180
 181   // start with default encoder configuration
 182   res = vpx_codec_enc_config_default(vpx_codec_vp9_cx(), enc_cfg, 0);
 183   if (res) {
 184     die("Failed to get config: %s\n", vpx_codec_err_to_string(res));
 185   }
 186   // update enc_cfg with app default values
 187   enc_cfg->g_w = default_width;
 188   enc_cfg->g_h = default_height;
 189   enc_cfg->g_timebase.num = default_timebase_num;
 190   enc_cfg->g_timebase.den = default_timebase_den;
 191   enc_cfg->rc_target_bitrate = default_bitrate;
 192   enc_cfg->kf_min_dist = default_kf_dist;
 193   enc_cfg->kf_max_dist = default_kf_dist;
 194   enc_cfg->rc_end_usage = VPX_CQ;
 195
 196   // initialize AppInput with default values
 197   app_input->frames_to_code = default_frames_to_code;
 198   app_input->frames_to_skip = default_frames_to_skip;
 199
 200   // process command line options
 201   argv = argv_dup(argc - 1, argv_ + 1);
 202   for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
 203     arg.argv_step = 1;
 204
 205     if (arg_match(&arg, &frames_arg, argi)) {
 206       app_input->frames_to_code = arg_parse_uint(&arg);
 207     } else if (arg_match(&arg, &width_arg, argi)) {
 208       enc_cfg->g_w = arg_parse_uint(&arg);
 209     } else if (arg_match(&arg, &height_arg, argi)) {
 210       enc_cfg->g_h = arg_parse_uint(&arg);
 211     } else if (arg_match(&arg, &timebase_arg, argi)) {
 212       enc_cfg->g_timebase = arg_parse_rational(&arg);
 213     } else if (arg_match(&arg, &bitrate_arg, argi)) {
 214       enc_cfg->rc_target_bitrate = arg_parse_uint(&arg);
 215     } else if (arg_match(&arg, &skip_frames_arg, argi)) {
 216       app_input->frames_to_skip = arg_parse_uint(&arg);
 217     } else if (arg_match(&arg, &spatial_layers_arg, argi)) {
 218       svc_ctx->spatial_layers = arg_parse_uint(&arg);
 219     } else if (arg_match(&arg, &temporal_layers_arg, argi)) {
 220       svc_ctx->temporal_layers = arg_parse_uint(&arg);
 221 #if OUTPUT_RC_STATS
 222     } else if (arg_match(&arg, &output_rc_stats_arg, argi)) {
 223       svc_ctx->output_rc_stat = arg_parse_uint(&arg);
 224 #endif
 225     } else if (arg_match(&arg, &speed_arg, argi)) {
 226       svc_ctx->speed = arg_parse_uint(&arg);
 227     } else if (arg_match(&arg, &aqmode_arg, argi)) {
 228       svc_ctx->aqmode = arg_parse_uint(&arg);
 229     } else if (arg_match(&arg, &threads_arg, argi)) {
 230       svc_ctx->threads = arg_parse_uint(&arg);
 231     } else if (arg_match(&arg, &temporal_layering_mode_arg, argi)) {
 232       svc_ctx->temporal_layering_mode =
 233           enc_cfg->temporal_layering_mode = arg_parse_int(&arg);
 234       if (svc_ctx->temporal_layering_mode) {
 235         enc_cfg->g_error_resilient = 1;
 236       }
 237     } else if (arg_match(&arg, &kf_dist_arg, argi)) {
 238       enc_cfg->kf_min_dist = arg_parse_uint(&arg);
 239       enc_cfg->kf_max_dist = enc_cfg->kf_min_dist;
 240     } else if (arg_match(&arg, &scale_factors_arg, argi)) {
 241       snprintf(string_options, sizeof(string_options), "%s scale-factors=%s",
 242                string_options, arg.val);
 243     } else if (arg_match(&arg, &passes_arg, argi)) {
 244       passes = arg_parse_uint(&arg);
 245       if (passes < 1 || passes > 2) {
 246         die("Error: Invalid number of passes (%d)\n", passes);
 247       }
 248     } else if (arg_match(&arg, &pass_arg, argi)) {
 249       pass = arg_parse_uint(&arg);
 250       if (pass < 1 || pass > 2) {
 251         die("Error: Invalid pass selected (%d)\n", pass);
 252       }
 253     } else if (arg_match(&arg, &fpf_name_arg, argi)) {
 254       fpf_file_name = arg.val;
 255     } else if (arg_match(&arg, &min_q_arg, argi)) {
 256       snprintf(string_options, sizeof(string_options), "%s min-quantizers=%s",
 257                string_options, arg.val);
 258     } else if (arg_match(&arg, &max_q_arg, argi)) {
 259       snprintf(string_options, sizeof(string_options), "%s max-quantizers=%s",
 260                string_options, arg.val);
 261     } else if (arg_match(&arg, &min_bitrate_arg, argi)) {
 262       min_bitrate = arg_parse_uint(&arg);
 263     } else if (arg_match(&arg, &max_bitrate_arg, argi)) {
 264       max_bitrate = arg_parse_uint(&arg);
 265     } else if (arg_match(&arg, &lag_in_frame_arg, argi)) {
 266       enc_cfg->g_lag_in_frames = arg_parse_uint(&arg);
 267     } else if (arg_match(&arg, &rc_end_usage_arg, argi)) {
 268       enc_cfg->rc_end_usage = arg_parse_uint(&arg);
 269 #if CONFIG_VP9_HIGHBITDEPTH
 270     } else if (arg_match(&arg, &bitdepth_arg, argi)) {
 271       enc_cfg->g_bit_depth = arg_parse_enum_or_int(&arg);
 272       switch (enc_cfg->g_bit_depth) {
 273         case VPX_BITS_8:
 274           enc_cfg->g_input_bit_depth = 8;
 275           enc_cfg->g_profile = 0;
 276           break;
 277         case VPX_BITS_10:
 278           enc_cfg->g_input_bit_depth = 10;
 279           enc_cfg->g_profile = 2;
 280           break;
 281          case VPX_BITS_12:
 282           enc_cfg->g_input_bit_depth = 12;
 283           enc_cfg->g_profile = 2;
 284           break;
 285         default:
 286           die("Error: Invalid bit depth selected (%d)\n", enc_cfg->g_bit_depth);
 287           break;
 288       }
 289 #endif  // CONFIG_VP9_HIGHBITDEPTH
 290     } else {
 291       ++argj;
 292     }
 293   }
 294
 295   // There will be a space in front of the string options
 296   if (strlen(string_options) > 0)
 297     vpx_svc_set_options(svc_ctx, string_options + 1);
 298
 299   if (passes == 0 || passes == 1) {
 300     if (pass) {
 301       fprintf(stderr, "pass is ignored since there's only one pass\n");
 302     }
 303     enc_cfg->g_pass = VPX_RC_ONE_PASS;
 304   } else {
 305     if (pass == 0) {
 306       die("pass must be specified when passes is 2\n");
 307     }
 308
 309     if (fpf_file_name == NULL) {
 310       die("fpf must be specified when passes is 2\n");
 311     }
 312
 313     if (pass == 1) {
 314       enc_cfg->g_pass = VPX_RC_FIRST_PASS;
 315       if (!stats_open_file(&app_input->rc_stats, fpf_file_name, 0)) {
 316         fatal("Failed to open statistics store");
 317       }
 318     } else {
 319       enc_cfg->g_pass = VPX_RC_LAST_PASS;
 320       if (!stats_open_file(&app_input->rc_stats, fpf_file_name, 1)) {
 321         fatal("Failed to open statistics store");
 322       }
 323       enc_cfg->rc_twopass_stats_in = stats_get(&app_input->rc_stats);
 324     }
 325     app_input->passes = passes;
 326     app_input->pass = pass;
 327   }
 328
 329   if (enc_cfg->rc_target_bitrate > 0) {
 330     if (min_bitrate > 0) {
 331       enc_cfg->rc_2pass_vbr_minsection_pct =
 332           min_bitrate * 100 / enc_cfg->rc_target_bitrate;
 333     }
 334     if (max_bitrate > 0) {
 335       enc_cfg->rc_2pass_vbr_maxsection_pct =
 336           max_bitrate * 100 / enc_cfg->rc_target_bitrate;
 337     }
 338   }
 339
 340   // Check for unrecognized options
 341   for (argi = argv; *argi; ++argi)
 342     if (argi[0][0] == '-' && strlen(argi[0]) > 1)
 343       die("Error: Unrecognized option %s\n", *argi);
 344
 345   if (argv[0] == NULL || argv[1] == 0) {
 346     usage_exit();
 347   }
 348   app_input->input_filename = argv[0];
 349   app_input->output_filename = argv[1];
 350   free(argv);
 351
 352   if (enc_cfg->g_w < 16 || enc_cfg->g_w % 2 || enc_cfg->g_h < 16 ||
 353       enc_cfg->g_h % 2)
 354     die("Invalid resolution: %d x %d\n", enc_cfg->g_w, enc_cfg->g_h);
 355
 356   printf(
 357       "Codec %s\nframes: %d, skip: %d\n"
 358       "layers: %d\n"
 359       "width %d, height: %d,\n"
 360       "num: %d, den: %d, bitrate: %d,\n"
 361       "gop size: %d\n",
 362       vpx_codec_iface_name(vpx_codec_vp9_cx()), app_input->frames_to_code,
 363       app_input->frames_to_skip,
 364       svc_ctx->spatial_layers, enc_cfg->g_w, enc_cfg->g_h,
 365       enc_cfg->g_timebase.num, enc_cfg->g_timebase.den,
 366       enc_cfg->rc_target_bitrate, enc_cfg->kf_max_dist);
 367 }
 368
 369 #if OUTPUT_RC_STATS
 370 // For rate control encoding stats.
 371 struct RateControlStats {
 372   // Number of input frames per layer.
 373   int layer_input_frames[VPX_MAX_LAYERS];
 374   // Total (cumulative) number of encoded frames per layer.
 375   int layer_tot_enc_frames[VPX_MAX_LAYERS];
 376   // Number of encoded non-key frames per layer.
 377   int layer_enc_frames[VPX_MAX_LAYERS];
 378   // Framerate per layer (cumulative).
 379   double layer_framerate[VPX_MAX_LAYERS];
 380   // Target average frame size per layer (per-frame-bandwidth per layer).
 381   double layer_pfb[VPX_MAX_LAYERS];
 382   // Actual average frame size per layer.
 383   double layer_avg_frame_size[VPX_MAX_LAYERS];
 384   // Average rate mismatch per layer (|target - actual| / target).
 385   double layer_avg_rate_mismatch[VPX_MAX_LAYERS];
 386   // Actual encoding bitrate per layer (cumulative).
 387   double layer_encoding_bitrate[VPX_MAX_LAYERS];
 388   // Average of the short-time encoder actual bitrate.
 389   // TODO(marpan): Should we add these short-time stats for each layer?
 390   double avg_st_encoding_bitrate;
 391   // Variance of the short-time encoder actual bitrate.
 392   double variance_st_encoding_bitrate;
 393   // Window (number of frames) for computing short-time encoding bitrate.
 394   int window_size;
 395   // Number of window measurements.
 396   int window_count;
 397 };
 398
 399 // Note: these rate control stats assume only 1 key frame in the
 400 // sequence (i.e., first frame only).
 401 static void set_rate_control_stats(struct RateControlStats *rc,
 402                                      vpx_codec_enc_cfg_t *cfg) {
 403   unsigned int sl, tl;
 404   // Set the layer (cumulative) framerate and the target layer (non-cumulative)
 405   // per-frame-bandwidth, for the rate control encoding stats below.
 406   const double framerate = cfg->g_timebase.den / cfg->g_timebase.num;
 407
 408   for (sl = 0; sl < cfg->ss_number_layers; ++sl) {
 409     for (tl = 0; tl < cfg->ts_number_layers; ++tl) {
 410       const int layer = sl * cfg->ts_number_layers + tl;
 411       const int tlayer0 = sl * cfg->ts_number_layers;
 412       if (cfg->ts_number_layers == 1)
 413         rc->layer_framerate[layer] = framerate;
 414       else
 415         rc->layer_framerate[layer] =
 416           framerate / cfg->ts_rate_decimator[tl];
 417       if (tl > 0) {
 418         rc->layer_pfb[layer] = 1000.0 *
 419             (cfg->layer_target_bitrate[layer] -
 420                 cfg->layer_target_bitrate[layer - 1]) /
 421             (rc->layer_framerate[layer] -
 422                 rc->layer_framerate[layer - 1]);
 423       } else {
 424         rc->layer_pfb[tlayer0] = 1000.0 *
 425             cfg->layer_target_bitrate[tlayer0] /
 426             rc->layer_framerate[tlayer0];
 427       }
 428       rc->layer_input_frames[layer] = 0;
 429       rc->layer_enc_frames[layer] = 0;
 430       rc->layer_tot_enc_frames[layer] = 0;
 431       rc->layer_encoding_bitrate[layer] = 0.0;
 432       rc->layer_avg_frame_size[layer] = 0.0;
 433       rc->layer_avg_rate_mismatch[layer] = 0.0;
 434     }
 435   }
 436   rc->window_count = 0;
 437   rc->window_size = 15;
 438   rc->avg_st_encoding_bitrate = 0.0;
 439   rc->variance_st_encoding_bitrate = 0.0;
 440 }
 441
 442 static void printout_rate_control_summary(struct RateControlStats *rc,
 443                                           vpx_codec_enc_cfg_t *cfg,
 444                                           int frame_cnt) {
 445   unsigned int sl, tl;
 446   int tot_num_frames = 0;
 447   double perc_fluctuation = 0.0;
 448   printf("Total number of processed frames: %d\n\n", frame_cnt - 1);
 449   printf("Rate control layer stats for sl%d tl%d layer(s):\n\n",
 450       cfg->ss_number_layers, cfg->ts_number_layers);
 451   for (sl = 0; sl < cfg->ss_number_layers; ++sl) {
 452     for (tl = 0; tl < cfg->ts_number_layers; ++tl) {
 453       const int layer = sl * cfg->ts_number_layers + tl;
 454       const int num_dropped = (tl > 0) ?
 455           (rc->layer_input_frames[layer] - rc->layer_enc_frames[layer]) :
 456           (rc->layer_input_frames[layer] - rc->layer_enc_frames[layer] - 1);
 457       if (!sl)
 458         tot_num_frames += rc->layer_input_frames[layer];
 459       rc->layer_encoding_bitrate[layer] = 0.001 * rc->layer_framerate[layer] *
 460           rc->layer_encoding_bitrate[layer] / tot_num_frames;
 461       rc->layer_avg_frame_size[layer] = rc->layer_avg_frame_size[layer] /
 462           rc->layer_enc_frames[layer];
 463       rc->layer_avg_rate_mismatch[layer] =
 464           100.0 * rc->layer_avg_rate_mismatch[layer] /
 465           rc->layer_enc_frames[layer];
 466       printf("For layer#: sl%d tl%d \n", sl, tl);
 467       printf("Bitrate (target vs actual): %d %f.0 kbps\n",
 468              cfg->layer_target_bitrate[layer],
 469              rc->layer_encoding_bitrate[layer]);
 470       printf("Average frame size (target vs actual): %f %f bits\n",
 471              rc->layer_pfb[layer], rc->layer_avg_frame_size[layer]);
 472       printf("Average rate_mismatch: %f\n",
 473              rc->layer_avg_rate_mismatch[layer]);
 474       printf("Number of input frames, encoded (non-key) frames, "
 475           "and percent dropped frames: %d %d %f.0 \n",
 476           rc->layer_input_frames[layer], rc->layer_enc_frames[layer],
 477           100.0 * num_dropped / rc->layer_input_frames[layer]);
 478       printf("\n");
 479     }
 480   }
 481   rc->avg_st_encoding_bitrate = rc->avg_st_encoding_bitrate / rc->window_count;
 482   rc->variance_st_encoding_bitrate =
 483       rc->variance_st_encoding_bitrate / rc->window_count -
 484       (rc->avg_st_encoding_bitrate * rc->avg_st_encoding_bitrate);
 485   perc_fluctuation = 100.0 * sqrt(rc->variance_st_encoding_bitrate) /
 486       rc->avg_st_encoding_bitrate;
 487   printf("Short-time stats, for window of %d frames: \n", rc->window_size);
 488   printf("Average, rms-variance, and percent-fluct: %f %f %f \n",
 489          rc->avg_st_encoding_bitrate,
 490          sqrt(rc->variance_st_encoding_bitrate),
 491          perc_fluctuation);
 492   if (frame_cnt != tot_num_frames)
 493     die("Error: Number of input frames not equal to output encoded frames != "
 494         "%d tot_num_frames = %d\n", frame_cnt, tot_num_frames);
 495 }
 496
 497 vpx_codec_err_t parse_superframe_index(const uint8_t *data,
 498                                        size_t data_sz,
 499                                        uint32_t sizes[8], int *count) {
 500   // A chunk ending with a byte matching 0xc0 is an invalid chunk unless
 501   // it is a super frame index. If the last byte of real video compression
 502   // data is 0xc0 the encoder must add a 0 byte. If we have the marker but
 503   // not the associated matching marker byte at the front of the index we have
 504   // an invalid bitstream and need to return an error.
 505
 506   uint8_t marker;
 507
 508   marker = *(data + data_sz - 1);
 509   *count = 0;
 510
 511
 512   if ((marker & 0xe0) == 0xc0) {
 513     const uint32_t frames = (marker & 0x7) + 1;
 514     const uint32_t mag = ((marker >> 3) & 0x3) + 1;
 515     const size_t index_sz = 2 + mag * frames;
 516
 517     // This chunk is marked as having a superframe index but doesn't have
 518     // enough data for it, thus it's an invalid superframe index.
 519     if (data_sz < index_sz)
 520       return VPX_CODEC_CORRUPT_FRAME;
 521
 522     {
 523       const uint8_t marker2 = *(data + data_sz - index_sz);
 524
 525       // This chunk is marked as having a superframe index but doesn't have
 526       // the matching marker byte at the front of the index therefore it's an
 527       // invalid chunk.
 528       if (marker != marker2)
 529         return VPX_CODEC_CORRUPT_FRAME;
 530     }
 531
 532     {
 533       // Found a valid superframe index.
 534       uint32_t i, j;
 535       const uint8_t *x = &data[data_sz - index_sz + 1];
 536
 537       for (i = 0; i < frames; ++i) {
 538         uint32_t this_sz = 0;
 539
 540         for (j = 0; j < mag; ++j)
 541           this_sz |= (*x++) << (j * 8);
 542         sizes[i] = this_sz;
 543       }
 544       *count = frames;
 545     }
 546   }
 547   return VPX_CODEC_OK;
 548 }
 549 #endif
 550
 551 // Example pattern for spatial layers and 2 temporal layers used in the
 552 // bypass/flexible mode. The pattern corresponds to the pattern
 553 // VP9E_TEMPORAL_LAYERING_MODE_0101 (temporal_layering_mode == 2) used in
 554 // non-flexible mode.
 555 void set_frame_flags_bypass_mode(int sl, int tl, int num_spatial_layers,
 556                                  int is_key_frame,
 557                                  vpx_svc_ref_frame_config_t *ref_frame_config) {
 558   for (sl = 0; sl < num_spatial_layers; ++sl) {
 559     if (!tl) {
 560       if (!sl) {
 561         ref_frame_config->frame_flags[sl] = VP8_EFLAG_NO_REF_GF |
 562                                             VP8_EFLAG_NO_REF_ARF |
 563                                             VP8_EFLAG_NO_UPD_GF |
 564                                             VP8_EFLAG_NO_UPD_ARF;
 565       } else {
 566         if (is_key_frame) {
 567           ref_frame_config->frame_flags[sl] = VP8_EFLAG_NO_REF_LAST |
 568                                               VP8_EFLAG_NO_REF_ARF |
 569                                               VP8_EFLAG_NO_UPD_GF |
 570                                               VP8_EFLAG_NO_UPD_ARF;
 571         } else {
 572         ref_frame_config->frame_flags[sl] = VP8_EFLAG_NO_REF_ARF |
 573                                             VP8_EFLAG_NO_UPD_GF |
 574                                             VP8_EFLAG_NO_UPD_ARF;
 575         }
 576       }
 577     } else if (tl == 1) {
 578       if (!sl) {
 579         ref_frame_config->frame_flags[sl] = VP8_EFLAG_NO_REF_GF |
 580                                             VP8_EFLAG_NO_REF_ARF |
 581                                             VP8_EFLAG_NO_UPD_LAST |
 582                                             VP8_EFLAG_NO_UPD_GF;
 583       } else {
 584         ref_frame_config->frame_flags[sl] = VP8_EFLAG_NO_REF_ARF |
 585                                             VP8_EFLAG_NO_UPD_LAST |
 586                                             VP8_EFLAG_NO_UPD_GF;
 587       }
 588     }
 589     if (tl == 0) {
 590       ref_frame_config->lst_fb_idx[sl] = sl;
 591       if (sl)
 592         ref_frame_config->gld_fb_idx[sl] = sl - 1;
 593       else
 594         ref_frame_config->gld_fb_idx[sl] = 0;
 595       ref_frame_config->alt_fb_idx[sl] = 0;
 596     } else if (tl == 1) {
 597       ref_frame_config->lst_fb_idx[sl] = sl;
 598       ref_frame_config->gld_fb_idx[sl] = num_spatial_layers + sl - 1;
 599       ref_frame_config->alt_fb_idx[sl] = num_spatial_layers + sl;
 600     }
 601   }
 602 }
 603
 604 int main(int argc, const char **argv) {
 605   AppInput app_input = {0};
 606   VpxVideoWriter *writer = NULL;
 607   VpxVideoInfo info = {0};
 608   vpx_codec_ctx_t codec;
 609   vpx_codec_enc_cfg_t enc_cfg;
 610   SvcContext svc_ctx;
 611   uint32_t i;
 612   uint32_t frame_cnt = 0;
 613   vpx_image_t raw;
 614   vpx_codec_err_t res;
 615   int pts = 0;            /* PTS starts at 0 */
 616   int frame_duration = 1; /* 1 timebase tick per frame */
 617   FILE *infile = NULL;
 618   int end_of_stream = 0;
 619   int frames_received = 0;
 620 #if OUTPUT_RC_STATS
 621   VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = {NULL};
 622   struct RateControlStats rc;
 623   vpx_svc_layer_id_t layer_id;
 624   vpx_svc_ref_frame_config_t ref_frame_config;
 625   int sl, tl;
 626   double sum_bitrate = 0.0;
 627   double sum_bitrate2 = 0.0;
 628   double framerate  = 30.0;
 629 #endif
 630   struct vpx_usec_timer timer;
 631   int64_t cx_time = 0;
 632   memset(&svc_ctx, 0, sizeof(svc_ctx));
 633   svc_ctx.log_print = 1;
 634   exec_name = argv[0];
 635   parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg);
 636
 637   // Allocate image buffer
 638 #if CONFIG_VP9_HIGHBITDEPTH
 639   if (!vpx_img_alloc(&raw, enc_cfg.g_input_bit_depth == 8 ?
 640                          VPX_IMG_FMT_I420 : VPX_IMG_FMT_I42016,
 641                      enc_cfg.g_w, enc_cfg.g_h, 32)) {
 642     die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
 643   }
 644 #else
 645   if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, enc_cfg.g_w, enc_cfg.g_h, 32)) {
 646     die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
 647   }
 648 #endif  // CONFIG_VP9_HIGHBITDEPTH
 649
 650   if (!(infile = fopen(app_input.input_filename, "rb")))
 651     die("Failed to open %s for reading\n", app_input.input_filename);
 652
 653   // Initialize codec
 654   if (vpx_svc_init(&svc_ctx, &codec, vpx_codec_vp9_cx(), &enc_cfg) !=
 655       VPX_CODEC_OK)
 656     die("Failed to initialize encoder\n");
 657
 658 #if OUTPUT_RC_STATS
 659   if (svc_ctx.output_rc_stat) {
 660     set_rate_control_stats(&rc, &enc_cfg);
 661     framerate = enc_cfg.g_timebase.den / enc_cfg.g_timebase.num;
 662   }
 663 #endif
 664
 665   info.codec_fourcc = VP9_FOURCC;
 666   info.time_base.numerator = enc_cfg.g_timebase.num;
 667   info.time_base.denominator = enc_cfg.g_timebase.den;
 668
 669   if (!(app_input.passes == 2 && app_input.pass == 1)) {
 670     // We don't save the bitstream for the 1st pass on two pass rate control
 671     writer = vpx_video_writer_open(app_input.output_filename, kContainerIVF,
 672                                    &info);
 673     if (!writer)
 674       die("Failed to open %s for writing\n", app_input.output_filename);
 675   }
 676 #if OUTPUT_RC_STATS
 677   // For now, just write temporal layer streams.
 678   // TODO(wonkap): do spatial by re-writing superframe.
 679   if (svc_ctx.output_rc_stat) {
 680     for (tl = 0; tl < enc_cfg.ts_number_layers; ++tl) {
 681       char file_name[PATH_MAX];
 682
 683       snprintf(file_name, sizeof(file_name), "%s_t%d.ivf",
 684                app_input.output_filename, tl);
 685       outfile[tl] = vpx_video_writer_open(file_name, kContainerIVF, &info);
 686       if (!outfile[tl])
 687         die("Failed to open %s for writing", file_name);
 688     }
 689   }
 690 #endif
 691
 692   // skip initial frames
 693   for (i = 0; i < app_input.frames_to_skip; ++i)
 694     vpx_img_read(&raw, infile);
 695
 696   if (svc_ctx.speed != -1)
 697     vpx_codec_control(&codec, VP8E_SET_CPUUSED, svc_ctx.speed);
 698   if (svc_ctx.threads)
 699     vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, (svc_ctx.threads >> 1));
 700   if (svc_ctx.speed >= 5 && svc_ctx.aqmode == 1)
 701     vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
 702
 703
 704   // Encode frames
 705   while (!end_of_stream) {
 706     vpx_codec_iter_t iter = NULL;
 707     const vpx_codec_cx_pkt_t *cx_pkt;
 708     if (frame_cnt >= app_input.frames_to_code || !vpx_img_read(&raw, infile)) {
 709       // We need one extra vpx_svc_encode call at end of stream to flush
 710       // encoder and get remaining data
 711       end_of_stream = 1;
 712     }
 713
 714     // For BYPASS/FLEXIBLE mode, set the frame flags (reference and updates)
 715     // and the buffer indices for each spatial layer of the current
 716     // (super)frame to be encoded. The temporal layer_id for the current frame
 717     // also needs to be set.
 718     // TODO(marpan): Should rename the "VP9E_TEMPORAL_LAYERING_MODE_BYPASS"
 719     // mode to "VP9E_LAYERING_MODE_BYPASS".
 720     if (svc_ctx.temporal_layering_mode == VP9E_TEMPORAL_LAYERING_MODE_BYPASS) {
 721       layer_id.spatial_layer_id = 0;
 722       // Example for 2 temporal layers.
 723       if (frame_cnt % 2 == 0)
 724         layer_id.temporal_layer_id = 0;
 725       else
 726         layer_id.temporal_layer_id = 1;
 727       // Note that we only set the temporal layer_id, since we are calling
 728       // the encode for the whole superframe. The encoder will internally loop
 729       // over all the spatial layers for the current superframe.
 730       vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id);
 731       set_frame_flags_bypass_mode(sl, layer_id.temporal_layer_id,
 732                                   svc_ctx.spatial_layers,
 733                                   frame_cnt == 0,
 734                                   &ref_frame_config);
 735       vpx_codec_control(&codec, VP9E_SET_SVC_REF_FRAME_CONFIG,
 736                         &ref_frame_config);
 737       // Keep track of input frames, to account for frame drops in rate control
 738       // stats/metrics.
 739       for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 740         ++rc.layer_input_frames[sl * enc_cfg.ts_number_layers +
 741                                 layer_id.temporal_layer_id];
 742       }
 743     }
 744
 745     vpx_usec_timer_start(&timer);
 746     res = vpx_svc_encode(&svc_ctx, &codec, (end_of_stream ? NULL : &raw),
 747                          pts, frame_duration, svc_ctx.speed >= 5 ?
 748                          VPX_DL_REALTIME : VPX_DL_GOOD_QUALITY);
 749     vpx_usec_timer_mark(&timer);
 750     cx_time += vpx_usec_timer_elapsed(&timer);
 751
 752     printf("%s", vpx_svc_get_message(&svc_ctx));
 753     fflush(stdout);
 754     if (res != VPX_CODEC_OK) {
 755       die_codec(&codec, "Failed to encode frame");
 756     }
 757
 758     while ((cx_pkt = vpx_codec_get_cx_data(&codec, &iter)) != NULL) {
 759       switch (cx_pkt->kind) {
 760         case VPX_CODEC_CX_FRAME_PKT: {
 761           SvcInternal_t *const si = (SvcInternal_t *)svc_ctx.internal;
 762           if (cx_pkt->data.frame.sz > 0) {
 763 #if OUTPUT_RC_STATS
 764             uint32_t sizes[8];
 765             int count = 0;
 766 #endif
 767             vpx_video_writer_write_frame(writer,
 768                                          cx_pkt->data.frame.buf,
 769                                          cx_pkt->data.frame.sz,
 770                                          cx_pkt->data.frame.pts);
 771 #if OUTPUT_RC_STATS
 772             // TODO(marpan/wonkap): Put this (to line728) in separate function.
 773             if (svc_ctx.output_rc_stat) {
 774               vpx_codec_control(&codec, VP9E_GET_SVC_LAYER_ID, &layer_id);
 775               parse_superframe_index(cx_pkt->data.frame.buf,
 776                                      cx_pkt->data.frame.sz, sizes, &count);
 777               // Note computing input_layer_frames here won't account for frame
 778               // drops in rate control stats.
 779               // TODO(marpan): Fix this for non-bypass mode so we can get stats
 780               // for dropped frames.
 781               if (svc_ctx.temporal_layering_mode !=
 782                   VP9E_TEMPORAL_LAYERING_MODE_BYPASS) {
 783                 for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 784                   ++rc.layer_input_frames[sl * enc_cfg.ts_number_layers +
 785                                          layer_id.temporal_layer_id];
 786                 }
 787               }
 788               for (tl = layer_id.temporal_layer_id;
 789                   tl < enc_cfg.ts_number_layers; ++tl) {
 790                 vpx_video_writer_write_frame(outfile[tl],
 791                                              cx_pkt->data.frame.buf,
 792                                              cx_pkt->data.frame.sz,
 793                                              cx_pkt->data.frame.pts);
 794               }
 795
 796               for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 797                 for (tl = layer_id.temporal_layer_id;
 798                     tl < enc_cfg.ts_number_layers; ++tl) {
 799                   const int layer = sl * enc_cfg.ts_number_layers + tl;
 800                   ++rc.layer_tot_enc_frames[layer];
 801                   rc.layer_encoding_bitrate[layer] += 8.0 * sizes[sl];
 802                   // Keep count of rate control stats per layer, for non-key
 803                   // frames.
 804                   if (tl == layer_id.temporal_layer_id &&
 805                       !(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
 806                     rc.layer_avg_frame_size[layer] += 8.0 * sizes[sl];
 807                     rc.layer_avg_rate_mismatch[layer] +=
 808                         fabs(8.0 * sizes[sl] - rc.layer_pfb[layer]) /
 809                         rc.layer_pfb[layer];
 810                     ++rc.layer_enc_frames[layer];
 811                   }
 812                 }
 813               }
 814
 815               // Update for short-time encoding bitrate states, for moving
 816               // window of size rc->window, shifted by rc->window / 2.
 817               // Ignore first window segment, due to key frame.
 818               if (frame_cnt > rc.window_size) {
 819                 tl = layer_id.temporal_layer_id;
 820                 for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 821                   sum_bitrate += 0.001 * 8.0 * sizes[sl] * framerate;
 822                 }
 823                 if (frame_cnt % rc.window_size == 0) {
 824                   rc.window_count += 1;
 825                   rc.avg_st_encoding_bitrate += sum_bitrate / rc.window_size;
 826                   rc.variance_st_encoding_bitrate +=
 827                       (sum_bitrate / rc.window_size) *
 828                       (sum_bitrate / rc.window_size);
 829                   sum_bitrate = 0.0;
 830                 }
 831               }
 832
 833               // Second shifted window.
 834               if (frame_cnt > rc.window_size + rc.window_size / 2) {
 835                tl = layer_id.temporal_layer_id;
 836                for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 837                  sum_bitrate2 += 0.001 * 8.0 * sizes[sl] * framerate;
 838                }
 839
 840                if (frame_cnt > 2 * rc.window_size &&
 841                   frame_cnt % rc.window_size == 0) {
 842                  rc.window_count += 1;
 843                  rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
 844                  rc.variance_st_encoding_bitrate +=
 845                     (sum_bitrate2 / rc.window_size) *
 846                     (sum_bitrate2 / rc.window_size);
 847                  sum_bitrate2 = 0.0;
 848                }
 849               }
 850             }
 851 #endif
 852           }
 853
 854           printf("SVC frame: %d, kf: %d, size: %d, pts: %d\n", frames_received,
 855                  !!(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY),
 856                  (int)cx_pkt->data.frame.sz, (int)cx_pkt->data.frame.pts);
 857           if (enc_cfg.ss_number_layers == 1 && enc_cfg.ts_number_layers == 1)
 858             si->bytes_sum[0] += (int)cx_pkt->data.frame.sz;
 859           ++frames_received;
 860           break;
 861         }
 862         case VPX_CODEC_STATS_PKT: {
 863           stats_write(&app_input.rc_stats,
 864                       cx_pkt->data.twopass_stats.buf,
 865                       cx_pkt->data.twopass_stats.sz);
 866           break;
 867         }
 868         default: {
 869           break;
 870         }
 871       }
 872     }
 873
 874     if (!end_of_stream) {
 875       ++frame_cnt;
 876       pts += frame_duration;
 877     }
 878   }
 879
 880   // Compensate for the extra frame count for the bypass mode.
 881   if (svc_ctx.temporal_layering_mode == VP9E_TEMPORAL_LAYERING_MODE_BYPASS) {
 882     for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
 883       const int layer = sl * enc_cfg.ts_number_layers +
 884           layer_id.temporal_layer_id;
 885       --rc.layer_input_frames[layer];
 886     }
 887   }
 888
 889   printf("Processed %d frames\n", frame_cnt);
 890   fclose(infile);
 891 #if OUTPUT_RC_STATS
 892   if (svc_ctx.output_rc_stat) {
 893     printout_rate_control_summary(&rc, &enc_cfg, frame_cnt);
 894     printf("\n");
 895   }
 896 #endif
 897   if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
 898   if (app_input.passes == 2)
 899     stats_close(&app_input.rc_stats, 1);
 900   if (writer) {
 901     vpx_video_writer_close(writer);
 902   }
 903 #if OUTPUT_RC_STATS
 904   if (svc_ctx.output_rc_stat) {
 905     for (tl = 0; tl < enc_cfg.ts_number_layers; ++tl) {
 906       vpx_video_writer_close(outfile[tl]);
 907     }
 908   }
 909 #endif
 910   printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
 911          frame_cnt,
 912          1000 * (float)cx_time / (double)(frame_cnt * 1000000),
 913          1000000 * (double)frame_cnt / (double)cx_time);
 914   vpx_img_free(&raw);
 915   // display average size, psnr
 916   printf("%s", vpx_svc_dump_statistics(&svc_ctx));
 917   vpx_svc_release(&svc_ctx);
 918   return EXIT_SUCCESS;
 919 }