vpxenc.c 77.2 KB
Newer Older
John Koleszar's avatar
John Koleszar committed
1
/*
2
 *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
John Koleszar's avatar
John Koleszar committed
3
 *
4
 *  Use of this source code is governed by a BSD-style license
5 6
 *  that can be found in the LICENSE file in the root of the source
 *  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.
John Koleszar's avatar
John Koleszar committed
9 10
 */

11
#include "./vpxenc.h"
12
#include "./vpx_config.h"
John Koleszar's avatar
John Koleszar committed
13

14 15
#include <assert.h>
#include <limits.h>
16
#include <math.h>
17
#include <stdarg.h>
John Koleszar's avatar
John Koleszar committed
18 19 20
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
21

22 23 24 25
#if CONFIG_LIBYUV
#include "third_party/libyuv/include/libyuv/scale.h"
#endif

26
#include "vpx/vpx_encoder.h"
27
#if CONFIG_DECODERS
28
#include "vpx/vpx_decoder.h"
29
#endif
30

31 32
#include "./args.h"
#include "./ivfenc.h"
33
#include "./tools_common.h"
34

35
#if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
36
#include "vpx/vp8cx.h"
37
#endif
38
#if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
39
#include "vpx/vp8dx.h"
40 41
#endif

42
#include "vpx/vpx_integer.h"
John Koleszar's avatar
John Koleszar committed
43 44
#include "vpx_ports/mem_ops.h"
#include "vpx_ports/vpx_timer.h"
45
#include "./rate_hist.h"
46
#include "./vpxstats.h"
47
#include "./warnings.h"
48
#if CONFIG_WEBM_IO
49
#include "./webmenc.h"
50
#endif
51
#include "./y4minput.h"
52

53 54
/* Swallow warnings about unused results of fread/fwrite */
static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
55 56
                         FILE *stream) {
  return fread(ptr, size, nmemb, stream);
57 58 59 60
}
#define fread wrap_fread

static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
61 62
                          FILE *stream) {
  return fwrite(ptr, size, nmemb, stream);
63 64 65 66
}
#define fwrite wrap_fwrite


John Koleszar's avatar
John Koleszar committed
67 68
static const char *exec_name;

69 70
static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
                                   const char *s, va_list ap) {
John Koleszar's avatar
John Koleszar committed
71 72
  if (ctx->err) {
    const char *detail = vpx_codec_error_detail(ctx);
John Koleszar's avatar
John Koleszar committed
73

74 75
    vfprintf(stderr, s, ap);
    fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
John Koleszar's avatar
John Koleszar committed
76

John Koleszar's avatar
John Koleszar committed
77 78
    if (detail)
      fprintf(stderr, "    %s\n", detail);
John Koleszar's avatar
John Koleszar committed
79

80 81
    if (fatal)
      exit(EXIT_FAILURE);
John Koleszar's avatar
John Koleszar committed
82
  }
John Koleszar's avatar
John Koleszar committed
83 84
}

85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101
static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
  va_list ap;

  va_start(ap, s);
  warn_or_exit_on_errorv(ctx, 1, s, ap);
  va_end(ap);
}

static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
                                  const char *s, ...) {
  va_list ap;

  va_start(ap, s);
  warn_or_exit_on_errorv(ctx, fatal, s, ap);
  va_end(ap);
}

102
static int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
103 104
  FILE *f = input_ctx->file;
  y4m_input *y4m = &input_ctx->y4m;
John Koleszar's avatar
John Koleszar committed
105 106
  int shortread = 0;

107
  if (input_ctx->file_type == FILE_TYPE_Y4M) {
John Koleszar's avatar
John Koleszar committed
108 109 110
    if (y4m_input_fetch_frame(y4m, f, img) < 1)
      return 0;
  } else {
111
    shortread = read_yuv_frame(input_ctx, img);
John Koleszar's avatar
John Koleszar committed
112
  }
John Koleszar's avatar
John Koleszar committed
113

John Koleszar's avatar
John Koleszar committed
114
  return !shortread;
John Koleszar's avatar
John Koleszar committed
115 116
}

117
static int file_is_y4m(const char detect[4]) {
John Koleszar's avatar
John Koleszar committed
118 119 120 121
  if (memcmp(detect, "YUV4", 4) == 0) {
    return 1;
  }
  return 0;
122 123
}

124
static int fourcc_is_ivf(const char detect[4]) {
125 126 127 128 129
  if (memcmp(detect, "DKIF", 4) == 0) {
    return 1;
  }
  return 0;
}
130

131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152
static const arg_def_t debugmode = ARG_DEF(
    "D", "debug", 0, "Debug mode (makes output deterministic)");
static const arg_def_t outputfile = ARG_DEF(
    "o", "output", 1, "Output filename");
static const arg_def_t use_yv12 = ARG_DEF(
    NULL, "yv12", 0, "Input file is YV12 ");
static const arg_def_t use_i420 = ARG_DEF(
    NULL, "i420", 0, "Input file is I420 (default)");
static const arg_def_t use_i422 = ARG_DEF(
    NULL, "i422", 0, "Input file is I422");
static const arg_def_t use_i444 = ARG_DEF(
    NULL, "i444", 0, "Input file is I444");
static const arg_def_t use_i440 = ARG_DEF(
    NULL, "i440", 0, "Input file is I440");
static const arg_def_t codecarg = ARG_DEF(
    NULL, "codec", 1, "Codec to use");
static const arg_def_t passes = ARG_DEF(
    "p", "passes", 1, "Number of passes (1/2)");
static const arg_def_t pass_arg = ARG_DEF(
    NULL, "pass", 1, "Pass to execute (1/2)");
static const arg_def_t fpf_name = ARG_DEF(
    NULL, "fpf", 1, "First pass statistics file name");
153
#if CONFIG_FP_MB_STATS
154 155
static const arg_def_t fpmbf_name = ARG_DEF(
    NULL, "fpmbf", 1, "First pass block statistics file name");
156
#endif
157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
static const arg_def_t limit = ARG_DEF(
    NULL, "limit", 1, "Stop encoding after n input frames");
static const arg_def_t skip = ARG_DEF(
    NULL, "skip", 1, "Skip the first n input frames");
static const arg_def_t deadline = ARG_DEF(
    "d", "deadline", 1, "Deadline per frame (usec)");
static const arg_def_t best_dl = ARG_DEF(
    NULL, "best", 0, "Use Best Quality Deadline");
static const arg_def_t good_dl = ARG_DEF(
    NULL, "good", 0, "Use Good Quality Deadline");
static const arg_def_t rt_dl = ARG_DEF(
    NULL, "rt", 0, "Use Realtime Quality Deadline");
static const arg_def_t quietarg = ARG_DEF(
    "q", "quiet", 0, "Do not print encode progress");
static const arg_def_t verbosearg = ARG_DEF(
    "v", "verbose", 0, "Show encoder parameters");
static const arg_def_t psnrarg = ARG_DEF(
    NULL, "psnr", 0, "Show PSNR in status line");
175

176 177 178 179 180 181
static const struct arg_enum_list test_decode_enum[] = {
  {"off",   TEST_DECODE_OFF},
  {"fatal", TEST_DECODE_FATAL},
  {"warn",  TEST_DECODE_WARN},
  {NULL, 0}
};
182 183 184 185
static const arg_def_t recontest = ARG_DEF_ENUM(
    NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
static const arg_def_t framerate = ARG_DEF(
    NULL, "fps", 1, "Stream frame rate (rate/scale)");
186 187
static const arg_def_t use_webm = ARG_DEF(
    NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
188
static const arg_def_t use_ivf = ARG_DEF(
189
    NULL, "ivf", 0, "Output IVF");
190 191 192 193 194 195 196 197 198 199 200 201 202
static const arg_def_t out_part = ARG_DEF(
    "P", "output-partitions", 0,
    "Makes encoder output partitions. Requires IVF output!");
static const arg_def_t q_hist_n = ARG_DEF(
    NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
static const arg_def_t rate_hist_n = ARG_DEF(
    NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
static const arg_def_t disable_warnings = ARG_DEF(
    NULL, "disable-warnings", 0,
    "Disable warnings about potentially incorrect encode settings.");
static const arg_def_t disable_warning_prompt = ARG_DEF(
    "y", "disable-warning-prompt", 0,
    "Display warnings, but do not prompt user to continue.");
203

204 205 206 207
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
static const arg_def_t test16bitinternalarg = ARG_DEF(
    NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
#endif
208

John Koleszar's avatar
John Koleszar committed
209 210 211
static const arg_def_t *main_args[] = {
  &debugmode,
  &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
212
  &deadline, &best_dl, &good_dl, &rt_dl,
213
  &quietarg, &verbosearg, &psnrarg, &use_webm, &use_ivf, &out_part, &q_hist_n,
214
  &rate_hist_n, &disable_warnings, &disable_warning_prompt,
John Koleszar's avatar
John Koleszar committed
215
  NULL
John Koleszar's avatar
John Koleszar committed
216 217
};

218 219 220 221 222 223 224 225
static const arg_def_t usage = ARG_DEF(
    "u", "usage", 1, "Usage profile number to use");
static const arg_def_t threads = ARG_DEF(
    "t", "threads", 1, "Max number of threads to use");
static const arg_def_t profile = ARG_DEF(
    NULL, "profile", 1, "Bitstream profile number to use");
static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
226
#if CONFIG_WEBM_IO
227
static const struct arg_enum_list stereo_mode_enum[] = {
John Koleszar's avatar
John Koleszar committed
228 229 230 231 232 233
  {"mono", STEREO_FORMAT_MONO},
  {"left-right", STEREO_FORMAT_LEFT_RIGHT},
  {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
  {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
  {"right-left", STEREO_FORMAT_RIGHT_LEFT},
  {NULL, 0}
234
};
235 236
static const arg_def_t stereo_mode = ARG_DEF_ENUM(
    NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
237
#endif
238 239 240 241 242 243
static const arg_def_t timebase = ARG_DEF(
    NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
static const arg_def_t error_resilient = ARG_DEF(
    NULL, "error-resilient", 1, "Enable error resiliency features");
static const arg_def_t lag_in_frames = ARG_DEF(
    NULL, "lag-in-frames", 1, "Max number of frames to lag");
John Koleszar's avatar
John Koleszar committed
244

John Koleszar's avatar
John Koleszar committed
245
static const arg_def_t *global_args[] = {
246
  &use_yv12, &use_i420, &use_i422, &use_i444, &use_i440,
247
  &usage, &threads, &profile,
248 249 250 251 252
  &width, &height,
#if CONFIG_WEBM_IO
  &stereo_mode,
#endif
  &timebase, &framerate,
253
  &error_resilient,
254 255 256
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
  &test16bitinternalarg,
#endif
John Koleszar's avatar
John Koleszar committed
257
  &lag_in_frames, NULL
John Koleszar's avatar
John Koleszar committed
258 259
};

260 261 262 263 264 265 266 267 268 269 270 271
static const arg_def_t dropframe_thresh = ARG_DEF(
    NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
static const arg_def_t resize_allowed = ARG_DEF(
    NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
static const arg_def_t resize_width = ARG_DEF(
    NULL, "resize-width", 1, "Width of encoded frame");
static const arg_def_t resize_height = ARG_DEF(
    NULL, "resize-height", 1, "Height of encoded frame");
static const arg_def_t resize_up_thresh = ARG_DEF(
    NULL, "resize-up", 1, "Upscale threshold (buf %)");
static const arg_def_t resize_down_thresh = ARG_DEF(
    NULL, "resize-down", 1, "Downscale threshold (buf %)");
272
static const struct arg_enum_list end_usage_enum[] = {
John Koleszar's avatar
John Koleszar committed
273 274 275
  {"vbr", VPX_VBR},
  {"cbr", VPX_CBR},
  {"cq",  VPX_CQ},
276
  {"q",   VPX_Q},
John Koleszar's avatar
John Koleszar committed
277
  {NULL, 0}
278
};
279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
static const arg_def_t end_usage = ARG_DEF_ENUM(
    NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
static const arg_def_t target_bitrate = ARG_DEF(
    NULL, "target-bitrate", 1, "Bitrate (kbps)");
static const arg_def_t min_quantizer = ARG_DEF(
    NULL, "min-q", 1, "Minimum (best) quantizer");
static const arg_def_t max_quantizer = ARG_DEF(
    NULL, "max-q", 1, "Maximum (worst) quantizer");
static const arg_def_t undershoot_pct = ARG_DEF(
    NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
static const arg_def_t overshoot_pct = ARG_DEF(
    NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
static const arg_def_t buf_sz = ARG_DEF(
    NULL, "buf-sz", 1, "Client buffer size (ms)");
static const arg_def_t buf_initial_sz = ARG_DEF(
    NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
static const arg_def_t buf_optimal_sz = ARG_DEF(
    NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
John Koleszar's avatar
John Koleszar committed
297
static const arg_def_t *rc_args[] = {
298 299 300 301
  &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
  &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
  &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
  &buf_initial_sz, &buf_optimal_sz, NULL
John Koleszar's avatar
John Koleszar committed
302 303 304
};


305 306 307 308 309 310
static const arg_def_t bias_pct = ARG_DEF(
    NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
static const arg_def_t minsection_pct = ARG_DEF(
    NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
static const arg_def_t maxsection_pct = ARG_DEF(
    NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
John Koleszar's avatar
John Koleszar committed
311 312
static const arg_def_t *rc_twopass_args[] = {
  &bias_pct, &minsection_pct, &maxsection_pct, NULL
John Koleszar's avatar
John Koleszar committed
313 314 315
};


316 317 318 319 320 321
static const arg_def_t kf_min_dist = ARG_DEF(
    NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
static const arg_def_t kf_max_dist = ARG_DEF(
    NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
static const arg_def_t kf_disabled = ARG_DEF(
    NULL, "disable-kf", 0, "Disable keyframe placement");
John Koleszar's avatar
John Koleszar committed
322 323
static const arg_def_t *kf_args[] = {
  &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
John Koleszar's avatar
John Koleszar committed
324 325 326
};


327 328 329 330 331 332
static const arg_def_t noise_sens = ARG_DEF(
    NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
static const arg_def_t sharpness = ARG_DEF(
    NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
static const arg_def_t static_thresh = ARG_DEF(
    NULL, "static-thresh", 1, "Motion detection threshold");
333
static const arg_def_t cpu_used_vp8 = ARG_DEF(
334
    NULL, "cpu-used", 1, "CPU Used (-16..16)");
335 336
static const arg_def_t cpu_used_vp9 = ARG_DEF(
    NULL, "cpu-used", 1, "CPU Used (-8..8)");
337 338 339 340 341 342 343 344
static const arg_def_t auto_altref = ARG_DEF(
    NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
static const arg_def_t arnr_maxframes = ARG_DEF(
    NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
static const arg_def_t arnr_strength = ARG_DEF(
    NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
static const arg_def_t arnr_type = ARG_DEF(
    NULL, "arnr-type", 1, "AltRef type");
345
static const struct arg_enum_list tuning_enum[] = {
John Koleszar's avatar
John Koleszar committed
346 347 348
  {"psnr", VP8_TUNE_PSNR},
  {"ssim", VP8_TUNE_SSIM},
  {NULL, 0}
349
};
350 351 352 353 354 355
static const arg_def_t tune_ssim = ARG_DEF_ENUM(
    NULL, "tune", 1, "Material to favor", tuning_enum);
static const arg_def_t cq_level = ARG_DEF(
    NULL, "cq-level", 1, "Constant/Constrained Quality level");
static const arg_def_t max_intra_rate_pct = ARG_DEF(
    NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
356 357
static const arg_def_t max_inter_rate_pct = ARG_DEF(
    NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
358 359
static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
    NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
John Koleszar's avatar
John Koleszar committed
360

Marco's avatar
Marco committed
361 362 363
static const arg_def_t screen_content_mode = ARG_DEF(NULL, "screen-content-mode", 1,
                                                     "Screen content mode");

364
#if CONFIG_VP8_ENCODER
365 366
static const arg_def_t token_parts = ARG_DEF(
    NULL, "token-parts", 1, "Number of token partitions to use, log2");
John Koleszar's avatar
John Koleszar committed
367
static const arg_def_t *vp8_args[] = {
368
  &cpu_used_vp8, &auto_altref, &noise_sens, &sharpness, &static_thresh,
John Koleszar's avatar
John Koleszar committed
369
  &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
Marco's avatar
Marco committed
370
  &tune_ssim, &cq_level, &max_intra_rate_pct, &screen_content_mode,
371
  NULL
John Koleszar's avatar
John Koleszar committed
372
};
373 374 375 376 377 378
static const int vp8_arg_ctrl_map[] = {
  VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
  VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
  VP8E_SET_TOKEN_PARTITIONS,
  VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
  VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
Marco's avatar
Marco committed
379
  VP8E_SET_SCREEN_CONTENT_MODE,
380 381 382 383 384
  0
};
#endif

#if CONFIG_VP9_ENCODER
385 386 387 388 389 390
static const arg_def_t tile_cols = ARG_DEF(
    NULL, "tile-columns", 1, "Number of tile columns to use, log2");
static const arg_def_t tile_rows = ARG_DEF(
    NULL, "tile-rows", 1, "Number of tile rows to use, log2");
static const arg_def_t lossless = ARG_DEF(
    NULL, "lossless", 1, "Lossless mode");
391 392 393 394
static const arg_def_t frame_parallel_decoding = ARG_DEF(
    NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
395
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
396 397
    "3: cyclic refresh)");
static const arg_def_t frame_periodic_boost = ARG_DEF(
398
    NULL, "frame-boost", 1,
399
    "Enable frame periodic boost (0: off (default), 1: on)");
400

Yaowu Xu's avatar
Yaowu Xu committed
401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416
static const struct arg_enum_list color_space_enum[] = {
  { "unknown", VPX_CS_UNKNOWN },
  { "bt601", VPX_CS_BT_601 },
  { "bt709", VPX_CS_BT_709 },
  { "smpte170", VPX_CS_SMPTE_170 },
  { "smpte240", VPX_CS_SMPTE_240 },
  { "bt2020", VPX_CS_BT_2020 },
  { "reserved", VPX_CS_RESERVED },
  { "sRGB", VPX_CS_SRGB },
  { NULL, 0 }
};

static const arg_def_t input_color_space = ARG_DEF_ENUM(
    NULL, "color-space", 1,
    "The color space of input content:", color_space_enum);

417 418 419 420 421 422 423 424
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
static const struct arg_enum_list bitdepth_enum[] = {
  {"8",  VPX_BITS_8},
  {"10", VPX_BITS_10},
  {"12", VPX_BITS_12},
  {NULL, 0}
};

425 426 427 428 429 430
static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
    "b", "bit-depth", 1,
    "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
    bitdepth_enum);
static const arg_def_t inbitdeptharg = ARG_DEF(
    NULL, "input-bit-depth", 1, "Bit depth of input");
431 432
#endif

433 434 435 436 437 438 439 440 441
static const struct arg_enum_list tune_content_enum[] = {
  {"default", VP9E_CONTENT_DEFAULT},
  {"screen", VP9E_CONTENT_SCREEN},
  {NULL, 0}
};

static const arg_def_t tune_content = ARG_DEF_ENUM(
    NULL, "tune-content", 1, "Tune content type", tune_content_enum);

442
static const arg_def_t *vp9_args[] = {
443
  &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
444
  &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
445 446
  &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
  &gf_cbr_boost_pct, &lossless,
447
  &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
Yaowu Xu's avatar
Yaowu Xu committed
448
  &noise_sens, &tune_content, &input_color_space,
449 450 451
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
  &bitdeptharg, &inbitdeptharg,
#endif
John Koleszar's avatar
John Koleszar committed
452
  NULL
John Koleszar's avatar
John Koleszar committed
453
};
454
static const int vp9_arg_ctrl_map[] = {
John Koleszar's avatar
John Koleszar committed
455
  VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
456
  VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
457
  VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
458
  VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
John Koleszar's avatar
John Koleszar committed
459
  VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
460
  VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
461
  VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
462
  VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
463
  VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
John Koleszar's avatar
John Koleszar committed
464
  0
John Koleszar's avatar
John Koleszar committed
465 466 467 468 469
};
#endif

static const arg_def_t *no_args[] = { NULL };

470
void usage_exit(void) {
John Koleszar's avatar
John Koleszar committed
471
  int i;
472
  const int num_encoder = get_vpx_encoder_count();
John Koleszar's avatar
John Koleszar committed
473 474 475 476 477

  fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
          exec_name);

  fprintf(stderr, "\nOptions:\n");
478
  arg_show_usage(stderr, main_args);
John Koleszar's avatar
John Koleszar committed
479
  fprintf(stderr, "\nEncoder Global Options:\n");
480
  arg_show_usage(stderr, global_args);
John Koleszar's avatar
John Koleszar committed
481
  fprintf(stderr, "\nRate Control Options:\n");
482
  arg_show_usage(stderr, rc_args);
John Koleszar's avatar
John Koleszar committed
483
  fprintf(stderr, "\nTwopass Rate Control Options:\n");
484
  arg_show_usage(stderr, rc_twopass_args);
John Koleszar's avatar
John Koleszar committed
485
  fprintf(stderr, "\nKeyframe Placement Options:\n");
486
  arg_show_usage(stderr, kf_args);
John Koleszar's avatar
John Koleszar committed
487
#if CONFIG_VP8_ENCODER
John Koleszar's avatar
John Koleszar committed
488
  fprintf(stderr, "\nVP8 Specific Options:\n");
489
  arg_show_usage(stderr, vp8_args);
John Koleszar's avatar
John Koleszar committed
490
#endif
491 492
#if CONFIG_VP9_ENCODER
  fprintf(stderr, "\nVP9 Specific Options:\n");
493
  arg_show_usage(stderr, vp9_args);
John Koleszar's avatar
John Koleszar committed
494
#endif
John Koleszar's avatar
John Koleszar committed
495 496 497
  fprintf(stderr, "\nStream timebase (--timebase):\n"
          "  The desired precision of timestamps in the output, expressed\n"
          "  in fractional seconds. Default is 1/1000.\n");
498
  fprintf(stderr, "\nIncluded encoders:\n\n");
John Koleszar's avatar
John Koleszar committed
499

500
  for (i = 0; i < num_encoder; ++i) {
501
    const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
502 503 504 505
    const char* defstr = (i == (num_encoder - 1)) ? "(default)" : "";
      fprintf(stderr, "    %-6s - %s %s\n",
              encoder->name, vpx_codec_iface_name(encoder->codec_interface()),
              defstr);
506
  }
507 508
  fprintf(stderr, "\n        ");
  fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
John Koleszar's avatar
John Koleszar committed
509 510

  exit(EXIT_FAILURE);
John Koleszar's avatar
John Koleszar committed
511 512
}

513
#define mmin(a, b)  ((a) < (b) ? (a) : (b))
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609

#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
static void find_mismatch_high(const vpx_image_t *const img1,
                               const vpx_image_t *const img2,
                               int yloc[4], int uloc[4], int vloc[4]) {
  uint16_t *plane1, *plane2;
  uint32_t stride1, stride2;
  const uint32_t bsize = 64;
  const uint32_t bsizey = bsize >> img1->y_chroma_shift;
  const uint32_t bsizex = bsize >> img1->x_chroma_shift;
  const uint32_t c_w =
      (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
  const uint32_t c_h =
      (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
  int match = 1;
  uint32_t i, j;
  yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
  plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
  plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
  stride1 = img1->stride[VPX_PLANE_Y]/2;
  stride2 = img2->stride[VPX_PLANE_Y]/2;
  for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
    for (j = 0; match && j < img1->d_w; j += bsize) {
      int k, l;
      const int si = mmin(i + bsize, img1->d_h) - i;
      const int sj = mmin(j + bsize, img1->d_w) - j;
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
          if (*(plane1 + (i + k) * stride1 + j + l) !=
              *(plane2 + (i + k) * stride2 + j + l)) {
            yloc[0] = i + k;
            yloc[1] = j + l;
            yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
            yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
            match = 0;
            break;
          }
        }
      }
    }
  }

  uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
  plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
  plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
  stride1 = img1->stride[VPX_PLANE_U]/2;
  stride2 = img2->stride[VPX_PLANE_U]/2;
  for (i = 0, match = 1; match && i < c_h; i += bsizey) {
    for (j = 0; match && j < c_w; j += bsizex) {
      int k, l;
      const int si = mmin(i + bsizey, c_h - i);
      const int sj = mmin(j + bsizex, c_w - j);
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
          if (*(plane1 + (i + k) * stride1 + j + l) !=
              *(plane2 + (i + k) * stride2 + j + l)) {
            uloc[0] = i + k;
            uloc[1] = j + l;
            uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
            uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
            match = 0;
            break;
          }
        }
      }
    }
  }

  vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
  plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
  plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
  stride1 = img1->stride[VPX_PLANE_V]/2;
  stride2 = img2->stride[VPX_PLANE_V]/2;
  for (i = 0, match = 1; match && i < c_h; i += bsizey) {
    for (j = 0; match && j < c_w; j += bsizex) {
      int k, l;
      const int si = mmin(i + bsizey, c_h - i);
      const int sj = mmin(j + bsizex, c_w - j);
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
          if (*(plane1 + (i + k) * stride1 + j + l) !=
              *(plane2 + (i + k) * stride2 + j + l)) {
            vloc[0] = i + k;
            vloc[1] = j + l;
            vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
            vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
            match = 0;
            break;
          }
        }
      }
    }
  }
}
#endif

James Zern's avatar
James Zern committed
610 611
static void find_mismatch(const vpx_image_t *const img1,
                          const vpx_image_t *const img2,
612
                          int yloc[4], int uloc[4], int vloc[4]) {
613 614 615 616 617 618 619 620 621
  const uint32_t bsize = 64;
  const uint32_t bsizey = bsize >> img1->y_chroma_shift;
  const uint32_t bsizex = bsize >> img1->x_chroma_shift;
  const uint32_t c_w =
      (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
  const uint32_t c_h =
      (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
  int match = 1;
  uint32_t i, j;
622
  yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
623 624
  for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
    for (j = 0; match && j < img1->d_w; j += bsize) {
625
      int k, l;
James Zern's avatar
James Zern committed
626 627 628 629
      const int si = mmin(i + bsize, img1->d_h) - i;
      const int sj = mmin(j + bsize, img1->d_w) - j;
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
630 631 632 633 634 635
          if (*(img1->planes[VPX_PLANE_Y] +
                (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
              *(img2->planes[VPX_PLANE_Y] +
                (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
            yloc[0] = i + k;
            yloc[1] = j + l;
636 637 638 639
            yloc[2] = *(img1->planes[VPX_PLANE_Y] +
                        (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
            yloc[3] = *(img2->planes[VPX_PLANE_Y] +
                        (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
640
            match = 0;
641
            break;
642 643
          }
        }
James Zern's avatar
James Zern committed
644
      }
645
    }
646
  }
647

648
  uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
649
  for (i = 0, match = 1; match && i < c_h; i += bsizey) {
650
    for (j = 0; match && j < c_w; j += bsizex) {
651
      int k, l;
James Zern's avatar
James Zern committed
652 653 654 655
      const int si = mmin(i + bsizey, c_h - i);
      const int sj = mmin(j + bsizex, c_w - j);
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
656 657 658 659 660 661
          if (*(img1->planes[VPX_PLANE_U] +
                (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
              *(img2->planes[VPX_PLANE_U] +
                (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
            uloc[0] = i + k;
            uloc[1] = j + l;
662 663 664
            uloc[2] = *(img1->planes[VPX_PLANE_U] +
                        (i + k) * img1->stride[VPX_PLANE_U] + j + l);
            uloc[3] = *(img2->planes[VPX_PLANE_U] +
James Zern's avatar
James Zern committed
665
                        (i + k) * img2->stride[VPX_PLANE_U] + j + l);
666 667 668 669
            match = 0;
            break;
          }
        }
James Zern's avatar
James Zern committed
670
      }
671 672
    }
  }
673
  vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
674
  for (i = 0, match = 1; match && i < c_h; i += bsizey) {
675
    for (j = 0; match && j < c_w; j += bsizex) {
676
      int k, l;
James Zern's avatar
James Zern committed
677 678 679 680
      const int si = mmin(i + bsizey, c_h - i);
      const int sj = mmin(j + bsizex, c_w - j);
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
681 682 683 684 685 686
          if (*(img1->planes[VPX_PLANE_V] +
                (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
              *(img2->planes[VPX_PLANE_V] +
                (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
            vloc[0] = i + k;
            vloc[1] = j + l;
687 688 689 690
            vloc[2] = *(img1->planes[VPX_PLANE_V] +
                        (i + k) * img1->stride[VPX_PLANE_V] + j + l);
            vloc[3] = *(img2->planes[VPX_PLANE_V] +
                        (i + k) * img2->stride[VPX_PLANE_V] + j + l);
691 692 693 694
            match = 0;
            break;
          }
        }
James Zern's avatar
James Zern committed
695
      }
696 697
    }
  }
698 699
}

James Zern's avatar
James Zern committed
700 701
static int compare_img(const vpx_image_t *const img1,
                       const vpx_image_t *const img2) {
702 703
  uint32_t l_w = img1->d_w;
  uint32_t c_w =
704 705 706 707
      (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
  const uint32_t c_h =
      (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
  uint32_t i;
John Koleszar's avatar
John Koleszar committed
708
  int match = 1;
709

John Koleszar's avatar
John Koleszar committed
710
  match &= (img1->fmt == img2->fmt);
711 712
  match &= (img1->d_w == img2->d_w);
  match &= (img1->d_h == img2->d_h);
713 714 715 716 717 718
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
  if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
    l_w *= 2;
    c_w *= 2;
  }
#endif
719

James Zern's avatar
James Zern committed
720 721 722
  for (i = 0; i < img1->d_h; ++i)
    match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
                     img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
723
                     l_w) == 0);
724

James Zern's avatar
James Zern committed
725 726 727
  for (i = 0; i < c_h; ++i)
    match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
                     img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
728
                     c_w) == 0);
729

James Zern's avatar
James Zern committed
730 731 732
  for (i = 0; i < c_h; ++i)
    match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
                     img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
733
                     c_w) == 0);
734

John Koleszar's avatar
John Koleszar committed
735
  return match;
736 737
}

738

739
#define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
740 741
#define MAX(x,y) ((x)>(y)?(x):(y))
#if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
742
#define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
743 744 745 746 747 748
#elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
#define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
#else
#define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
                             NELEMENTS(vp9_arg_ctrl_map))
#endif
John Koleszar's avatar
John Koleszar committed
749

750 751 752 753 754
#if !CONFIG_WEBM_IO
typedef int stereo_format_t;
struct EbmlGlobal { int debug; };
#endif

755
/* Per-stream configuration */
756 757 758 759
struct stream_config {
  struct vpx_codec_enc_cfg  cfg;
  const char               *out_fn;
  const char               *stats_fn;
760 761 762
#if CONFIG_FP_MB_STATS
  const char               *fpmb_stats_fn;
#endif
763 764 765 766 767
  stereo_format_t           stereo_fmt;
  int                       arg_ctrls[ARG_CTRL_CNT_MAX][2];
  int                       arg_ctrl_cnt;
  int                       write_webm;
  int                       have_kf_max_dist;
768 769 770 771
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
  // whether to use 16bit internal buffers
  int                       use_16bit_internal;
#endif
772 773 774
};


775 776 777 778 779
struct stream_state {
  int                       index;
  struct stream_state      *next;
  struct stream_config      config;
  FILE                     *file;
780
  struct rate_hist         *rate_hist;
781
  struct EbmlGlobal         ebml;
782 783 784 785 786 787 788 789 790 791
  uint64_t                  psnr_sse_total;
  uint64_t                  psnr_samples_total;
  double                    psnr_totals[4];
  int                       psnr_count;
  int                       counts[64];
  vpx_codec_ctx_t           encoder;
  unsigned int              frames_out;
  uint64_t                  cx_time;
  size_t                    nbytes;
  stats_io_t                stats;
792 793 794
#if CONFIG_FP_MB_STATS
  stats_io_t                fpmb_stats;
#endif
795
  struct vpx_image         *img;
796 797
  vpx_codec_ctx_t           decoder;
  int                       mismatch_seen;
798 799 800
};


801 802
static void validate_positive_rational(const char          *msg,
                                       struct vpx_rational *rat) {
803 804 805 806
  if (rat->den < 0) {
    rat->num *= -1;
    rat->den *= -1;
  }
807

808 809
  if (rat->num < 0)
    die("Error: %s must be positive\n", msg);
810

811 812
  if (!rat->den)
    die("Error: %s has zero denominator\n", msg);
813 814 815
}


816
static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
817 818
  char       **argi, **argj;
  struct arg   arg;
819 820 821 822
  const int num_encoder = get_vpx_encoder_count();

  if (num_encoder < 1)
    die("Error: no valid encoder available\n");
823 824 825

  /* Initialize default parameters */
  memset(global, 0, sizeof(*global));
826
  global->codec = get_vpx_encoder_by_index(num_encoder - 1);
827
  global->passes = 0;
828
  global->color_type = I420;
829 830
  /* Assign default deadline to good quality */
  global->deadline = VPX_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
831 832 833 834 835

  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
    arg.argv_step = 1;

    if (arg_match(&arg, &codecarg, argi)) {
836 837 838
      global->codec = get_vpx_encoder_by_name(arg.val);
      if (!global->codec)
        die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
John Koleszar's avatar
John Koleszar committed
839
    } else if (arg_match(&arg, &passes, argi)) {
840
      global->passes = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
841

842 843
      if (global->passes < 1 || global->passes > 2)
        die("Error: Invalid number of passes (%d)\n", global->passes);
John Koleszar's avatar
John Koleszar committed
844
    } else if (arg_match(&arg, &pass_arg, argi)) {
845 846 847 848 849 850 851
      global->pass = arg_parse_uint(&arg);

      if (global->pass < 1 || global->pass > 2)
        die("Error: Invalid pass selected (%d)\n",
            global->pass);
    } else if (arg_match(&arg, &usage, argi))
      global->usage = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
852
    else if (arg_match(&arg, &deadline, argi))
853
      global->deadline = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
854
    else if (arg_match(&arg, &best_dl, argi))
855
      global->deadline = VPX_DL_BEST_QUALITY;
John Koleszar's avatar
John Koleszar committed
856
    else if (arg_match(&arg, &good_dl, argi))
857
      global->deadline = VPX_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
858
    else if (arg_match(&arg, &rt_dl, argi))
859 860
      global->deadline = VPX_DL_REALTIME;
    else if (arg_match(&arg, &use_yv12, argi))
861
      global->color_type = YV12;
862
    else if (arg_match(&arg, &use_i420, argi))
863 864 865 866 867
      global->color_type = I420;
    else if (arg_match(&arg, &use_i422, argi))
      global->color_type = I422;
    else if (arg_match(&arg, &use_i444, argi))
      global->color_type = I444;
868 869
    else if (arg_match(&arg, &use_i440, argi))
      global->color_type = I440;
870 871 872 873
    else if (arg_match(&arg, &quietarg, argi))
      global->quiet = 1;
    else if (arg_match(&arg, &verbosearg, argi))
      global->verbose = 1;
John Koleszar's avatar
John Koleszar committed
874
    else if (arg_match(&arg, &limit, argi))
875
      global->limit = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
876
    else if (arg_match(&arg, &skip, argi))
877
      global->skip_frames = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
878
    else if (arg_match(&arg, &psnrarg, argi))
879
      global->show_psnr = 1;
John Koleszar's avatar
John Koleszar committed
880
    else if (arg_match(&arg, &recontest, argi))
881
      global->test_decode = arg_parse_enum_or_int(&arg);
John Koleszar's avatar
John Koleszar committed
882
    else if (arg_match(&arg, &framerate, argi)) {
883 884 885 886 887
      global->framerate = arg_parse_rational(&arg);
      validate_positive_rational(arg.name, &global->framerate);
      global->have_framerate = 1;
    } else if (arg_match(&arg, &out_part, argi))
      global->out_part = 1;
John Koleszar's avatar
John Koleszar committed
888
    else if (arg_match(&arg, &debugmode, argi))
889
      global->debug = 1;
John Koleszar's avatar
John Koleszar committed
890
    else if (arg_match(&arg, &q_hist_n, argi))
891
      global->show_q_hist_buckets = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
892
    else if (arg_match(&arg, &rate_hist_n, argi))
893
      global->show_rate_hist_buckets = arg_parse_uint(&arg);
894 895 896 897
    else if (arg_match(&arg, &disable_warnings, argi))
      global->disable_warnings = 1;
    else if (arg_match(&arg, &disable_warning_prompt, argi))
      global->disable_warning_prompt = 1;
John Koleszar's avatar
John Koleszar committed
898 899 900
    else
      argj++;
  }
John Koleszar's avatar
John Koleszar committed
901

902 903 904 905 906 907 908 909
  if (global->pass) {
    /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
    if (global->pass > global->passes) {
      warn("Assuming --pass=%d implies --passes=%d\n",
           global->pass, global->pass);
      global->passes = global->pass;
    }
  }
910
  /* Validate global config */
911
  if (global->passes == 0) {
James Zern's avatar
James Zern committed
912
#if CONFIG_VP9_ENCODER
913 914
    // Make default VP9 passes = 2 until there is a better quality 1-pass
    // encoder
915 916 917
    if (global->codec != NULL && global->codec->name != NULL)
      global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
                        global->deadline != VPX_DL_REALTIME) ? 2 : 1;
James Zern's avatar
James Zern committed
918 919 920
#else
    global->passes = 1;
#endif
921
  }
922

923 924 925 926
  if (global->deadline == VPX_DL_REALTIME &&
      global->passes > 1) {
    warn("Enforcing one-pass encoding in realtime mode\n");
    global->passes = 1;
927
  }
928 929 930
}


931
static void open_input_file(struct VpxInputContext *input) {
932
  /* Parse certain options from the input file, if possible */
933 934
  input->file = strcmp(input->filename, "-")
      ? fopen(input->filename, "rb") : set_binary_mode(stdin);
935 936 937

  if (!input->file)
    fatal("Failed to open input file");
John Koleszar's avatar
John Koleszar committed
938

939 940 941 942 943 944 945 946
  if (!fseeko(input->file, 0, SEEK_END)) {
    /* Input file is seekable. Figure out how long it is, so we can get
     * progress info.
     */
    input->length = ftello(input->file);
    rewind(input->file);
  }

947 948 949 950 951 952 953
  /* For RAW input sources, these bytes will applied on the first frame
   *  in read_frame().
   */
  input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
  input->detect.position = 0;

  if (input->detect.buf_read == 4
954
      && file_is_y4m(input->detect.buf)) {
955 956
    if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
                       input->only_i420) >= 0) {
957
      input->file_type = FILE_TYPE_Y4M;
958 959 960 961
      input->width = input->y4m.pic_w;
      input->height = input->y4m.pic_h;
      input->framerate.numerator = input->y4m.fps_n;
      input->framerate.denominator = input->y4m.fps_d;
962
      input->fmt = input->y4m.vpx_fmt;
963
      input->bit_depth = input->y4m.bit_depth;
964 965
    } else
      fatal("Unsupported Y4M stream.");
966
  } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
967
    fatal("IVF is not supported as input.");
968 969
  } else {
    input->file_type = FILE_TYPE_RAW;
John Koleszar's avatar
John Koleszar committed
970
  }
971 972 973
}


974
static void close_input_file(struct VpxInputContext *input) {
975 976 977
  fclose(input->file);
  if (input->file_type == FILE_TYPE_Y4M)
    y4m_input_close(&input->y4m);
978 979
}

980
static struct stream_state *new_stream(struct VpxEncoderConfig *global,
981
                                       struct stream_state *prev) {
982 983 984
  struct stream_state *stream;

  stream = calloc(1, sizeof(*stream));
985
  if (stream == NULL) {
986
    fatal("Failed to allocate new stream.");
987 988
  }

989 990 991 992 993 994 995 996
  if (prev) {
    memcpy(stream, prev, sizeof(*stream));
    stream->index++;
    prev->next = stream;
  } else {
    vpx_codec_err_t  res;

    /* Populate encoder configuration */
997
    res = vpx_codec_enc_config_default(global->codec->codec_interface(),
998 999 1000 1001
                                       &stream->config.cfg,
                                       global->usage);
    if (res)
      fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
John Koleszar's avatar
John Koleszar committed
1002

1003 1004 1005 1006
    /* Change the default timebase to a high enough value so that the
     * encoder will always create strictly increasing timestamps.
     */
    stream->config.cfg.g_timebase.den = 1000;
1007

1008 1009 1010 1011 1012
    /* Never use the library's default resolution, require it be parsed
     * from the file or set on the command line.
     */
    stream->config.cfg.g_w = 0;
    stream->config.cfg.g_h = 0;
John Koleszar's avatar
John Koleszar committed
1013

1014 1015
    /* Initialize remaining stream parameters */
    stream->config.write_webm = 1;
1016
#if CONFIG_WEBM_IO
1017
    stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1018 1019 1020
    stream->ebml.last_pts_ns = -1;
    stream->ebml.writer = NULL;
    stream->ebml.segment = NULL;
1021
#endif
John Koleszar's avatar
John Koleszar committed
1022

1023 1024
    /* Allows removal of the application version from the EBML tags */
    stream->ebml.debug = global->debug;
1025 1026 1027 1028

    /* Default lag_in_frames is 0 in realtime mode */
    if (global->deadline == VPX_DL_REALTIME)
      stream->config.cfg.g_lag_in_frames = 0;
John Koleszar's avatar
John Koleszar committed
1029
  }
John Koleszar's avatar
John Koleszar committed
1030

1031 1032
  /* Output files must be specified for each stream */
  stream->config.out_fn = NULL;
John Koleszar's avatar
John Koleszar committed
1033

1034 1035
  stream->next = NULL;
  return stream;
1036
}
John Koleszar's avatar
John Koleszar committed
1037

1038

1039
static int parse_stream_params(struct VpxEncoderConfig *global,
1040
                               struct stream_state  *stream,
1041 1042 1043 1044 1045 1046 1047
                               char **argv) {
  char                   **argi, **argj;
  struct arg               arg;
  static const arg_def_t **ctrl_args = no_args;
  static const int        *ctrl_args_map = NULL;
  struct stream_config    *config = &stream->config;
  int                      eos_mark_found = 0;
1048 1049 1050
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
  int                      test_16bit_internal = 0;
#endif
1051

1052
  // Handle codec specific options
1053 1054
  if (0) {
#if CONFIG_VP8_ENCODER
1055
  } else if (strcmp(global->codec->name, "vp8") == 0) {
1056 1057
    ctrl_args = vp8_args;
    ctrl_args_map = vp8_arg_ctrl_map;
1058 1059
#endif
#if CONFIG_VP9_ENCODER
1060
  } else if (strcmp(global->codec->name, "vp9") == 0) {
1061 1062 1063
    ctrl_args = vp9_args;
    ctrl_args_map = vp9_arg_ctrl_map;
#endif
1064
  }
1065

John Koleszar's avatar
John Koleszar committed
1066 1067
  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
    arg.argv_step = 1;
John Koleszar's avatar
John Koleszar committed
1068

1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
    /* Once we've found an end-of-stream marker (--) we want to continue
     * shifting arguments but not consuming them.
     */
    if (eos_mark_found) {
      argj++;
      continue;
    } else if (!strcmp(*argj, "--")) {
      eos_mark_found = 1;
      continue;
    }
John Koleszar's avatar
John Koleszar committed
1079

Adrian Grange's avatar
Adrian Grange committed
1080
    if (arg_match(&arg, &outputfile, argi)) {
1081
      config->out_fn = arg.val;
1082
    } else if (arg_match(&arg, &fpf_name, argi)) {
1083
      config->stats_fn = arg.val;
1084 1085 1086
#if CONFIG_FP_MB_STATS
    } else if (arg_match(&arg, &fpmbf_name, argi)) {
      config->fpmb_stats_fn = arg.val;
1087 1088 1089 1090 1091 1092
#endif
    } else if (arg_match(&arg, &use_webm, argi)) {
#if CONFIG_WEBM_IO
      config->write_webm = 1;
#else
      die("Error: --webm specified but webm is disabled.");
1093
#endif
1094
    } else if (arg_match(&arg, &use_ivf, argi)) {
1095
      config->write_webm = 0;
1096
    } else if (arg_match(&arg, &threads, argi)) {
1097
      config->cfg.g_threads = arg_parse_uint(&arg);
1098
    } else if (arg_match(&arg, &profile, argi)) {
1099
      config->cfg.g_profile = arg_parse_uint(&arg);
1100
    } else if (arg_match(&arg, &width, argi)) {
1101
      config->cfg.g_w = arg_parse_uint(&arg);
1102
    } else if (arg_match(&arg, &height, argi)) {
1103
      config->cfg.g_h = arg_parse_uint(&arg);
1104 1105 1106 1107 1108 1109
#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
    } else if (arg_match(&arg, &bitdeptharg, argi)) {
      config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
    } else if (arg_match(&arg, &inbitdeptharg, argi)) {
      config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
#endif
1110
#if CONFIG_WEBM_IO
1111
    } else if (arg_match(&arg, &stereo_mode, argi)) {
1112
      config->stereo_fmt = arg_parse_enum_or_int(&arg);
1113
#endif
1114
    } else if (arg_match(&arg, &timebase, argi)) {
1115 1116
      config->cfg.g_timebase = arg_parse_rational(&arg);
      validate_positive_rational(arg.name, &config->cfg.g_timebase);
1117
    } else if (arg_match(&arg, &error_resilient, argi)) {
1118
      config->cfg.g_error_resilient = arg_parse_uint(&arg);
1119
    } else if (arg_match(&arg, &lag_in_frames, argi)) {
1120
      config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1121 1122 1123 1124 1125 1126
      if (global->deadline == VPX_DL_REALTIME &&
          config->cfg.g_lag_in_frames != 0) {
        warn("non-zero %s option ignored in realtime mode.\n", arg.name);
        config->cfg.g_lag_in_frames = 0;
      }
    } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1127
      config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1128
    } else if (arg_match(&arg, &resize_allowed, argi)) {
1129
      config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1130 1131 1132 1133
    } else if (arg_match(&arg, &resize_width, argi)) {
      config->cfg.rc_scaled_width = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &resize_height, argi)) {
      config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1134
    } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1135
      config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1136
    } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1137
      config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1138
    } else if (arg_match(&arg, &end_usage, argi)) {
1139
      config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1140
    } else if (arg_match(&arg, &target_bitrate, argi)) {
1141
      config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1142
    } else if (arg_match(&arg, &min_quantizer, argi)) {
1143
      config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1144
    } else if (arg_match(&arg, &max_quantizer, argi)) {
1145
      config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1146
    } else if (arg_match(&arg, &undershoot_pct, argi)) {
1147
      config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1148
    } else if (arg_match(&arg, &overshoot_pct, argi)) {
1149
      config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1150
    } else if (arg_match(&arg, &buf_sz, argi)) {
1151
      config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1152
    } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1153
      config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1154
    } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1155
      config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1156 1157
    } else if (arg_match(&arg, &bias_pct, argi)) {
        config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1158 1159
      if (global->passes < 2)
        warn("option %s ignored in one-pass mode.\n", arg.name);
John Koleszar's avatar
John Koleszar committed
1160
    } else if (arg_match(&arg, &minsection_pct, argi)) {
1161
      config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
1162

1163 1164
      if (global->passes < 2)
        warn("option %s ignored in one-pass mode.\n", arg.name);
John Koleszar's avatar
John Koleszar committed
1165
    } else if (arg_match(&arg, &maxsection_pct, argi)) {
1166
      config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
1167

1168 1169
      if (global->passes < 2)
        warn("option %s ignored in one-pass mode.\n", arg.name);
Deb Mukherjee's avatar