aomenc.c 87.9 KB
Newer Older
John Koleszar's avatar
John Koleszar committed
1
/*
2
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
John Koleszar's avatar
John Koleszar committed
3
 *
4 5 6 7 8 9
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
John Koleszar's avatar
John Koleszar committed
10 11
 */

12 13
#include "./aomenc.h"
#include "./aom_config.h"
John Koleszar's avatar
John Koleszar committed
14

15 16
#include <assert.h>
#include <limits.h>
17
#include <math.h>
18
#include <stdarg.h>
John Koleszar's avatar
John Koleszar committed
19 20 21
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
22

23 24 25 26
#if CONFIG_LIBYUV
#include "third_party/libyuv/include/libyuv/scale.h"
#endif

27
#include "aom/aom_encoder.h"
28
#if CONFIG_AV1_DECODER
29
#include "aom/aom_decoder.h"
30
#endif
John Koleszar's avatar
John Koleszar committed
31

32 33
#include "./args.h"
#include "./ivfenc.h"
Tom Finegan's avatar
Tom Finegan committed
34
#include "./tools_common.h"
35
#include "examples/encoder_util.h"
36

37 38
#if CONFIG_AV1_ENCODER
#include "aom/aomcx.h"
John Koleszar's avatar
John Koleszar committed
39
#endif
40 41
#if CONFIG_AV1_DECODER
#include "aom/aomdx.h"
John Koleszar's avatar
John Koleszar committed
42 43
#endif

44
#include "./aomstats.h"
45
#include "./rate_hist.h"
46
#include "./warnings.h"
47
#include "aom/aom_integer.h"
48
#include "aom_dsp/aom_dsp_common.h"
49 50
#include "aom_ports/aom_timer.h"
#include "aom_ports/mem_ops.h"
51
#if CONFIG_WEBM_IO
52
#include "./webmenc.h"
53
#endif
54
#include "./y4minput.h"
55

John Koleszar's avatar
John Koleszar committed
56
/* Swallow warnings about unused results of fread/fwrite */
clang-format's avatar
clang-format committed
57
static size_t wrap_fread(void *ptr, size_t size, size_t nmemb, FILE *stream) {
John Koleszar's avatar
John Koleszar committed
58
  return fread(ptr, size, nmemb, stream);
John Koleszar's avatar
John Koleszar committed
59 60 61 62
}
#define fread wrap_fread

static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
John Koleszar's avatar
John Koleszar committed
63 64
                          FILE *stream) {
  return fwrite(ptr, size, nmemb, stream);
John Koleszar's avatar
John Koleszar committed
65 66 67
}
#define fwrite wrap_fwrite

John Koleszar's avatar
John Koleszar committed
68 69
static const char *exec_name;

70
static void warn_or_exit_on_errorv(aom_codec_ctx_t *ctx, int fatal,
71
                                   const char *s, va_list ap) {
John Koleszar's avatar
John Koleszar committed
72
  if (ctx->err) {
73
    const char *detail = aom_codec_error_detail(ctx);
John Koleszar's avatar
John Koleszar committed
74

John Koleszar's avatar
John Koleszar committed
75
    vfprintf(stderr, s, ap);
76
    fprintf(stderr, ": %s\n", aom_codec_error(ctx));
John Koleszar's avatar
John Koleszar committed
77

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

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

84
static void ctx_exit_on_error(aom_codec_ctx_t *ctx, const char *s, ...) {
85 86 87 88 89 90 91
  va_list ap;

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

92
static void warn_or_exit_on_error(aom_codec_ctx_t *ctx, int fatal,
93 94 95 96 97 98 99 100
                                  const char *s, ...) {
  va_list ap;

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

101
static int read_frame(struct AvxInputContext *input_ctx, aom_image_t *img) {
102 103
  FILE *f = input_ctx->file;
  y4m_input *y4m = &input_ctx->y4m;
John Koleszar's avatar
John Koleszar committed
104 105
  int shortread = 0;

106
  if (input_ctx->file_type == FILE_TYPE_Y4M) {
clang-format's avatar
clang-format committed
107
    if (y4m_input_fetch_frame(y4m, f, img) < 1) return 0;
John Koleszar's avatar
John Koleszar committed
108
  } else {
109
    shortread = read_yuv_frame(input_ctx, img);
John Koleszar's avatar
John Koleszar committed
110
  }
John Koleszar's avatar
John Koleszar committed
111

John Koleszar's avatar
John Koleszar committed
112
  return !shortread;
John Koleszar's avatar
John Koleszar committed
113 114
}

115
static int file_is_y4m(const char detect[4]) {
John Koleszar's avatar
John Koleszar committed
116 117 118 119
  if (memcmp(detect, "YUV4", 4) == 0) {
    return 1;
  }
  return 0;
120 121
}

122
static int fourcc_is_ivf(const char detect[4]) {
123 124 125 126 127
  if (memcmp(detect, "DKIF", 4) == 0) {
    return 1;
  }
  return 0;
}
John Koleszar's avatar
John Koleszar committed
128

James Zern's avatar
James Zern committed
129 130
static const arg_def_t help =
    ARG_DEF(NULL, "help", 0, "Show usage options and exit");
clang-format's avatar
clang-format committed
131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
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");
152
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
153 154
static const arg_def_t fpmbf_name =
    ARG_DEF(NULL, "fpmbf", 1, "First pass block statistics file name");
155
#endif
clang-format's avatar
clang-format committed
156 157 158 159
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");
160 161
static const arg_def_t good_dl =
    ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
clang-format's avatar
clang-format committed
162 163 164 165 166 167
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");
168

169
static const struct arg_enum_list test_decode_enum[] = {
clang-format's avatar
clang-format committed
170 171 172 173
  { "off", TEST_DECODE_OFF },
  { "fatal", TEST_DECODE_FATAL },
  { "warn", TEST_DECODE_WARN },
  { NULL, 0 }
174
};
Deb Mukherjee's avatar
Deb Mukherjee committed
175 176
static const arg_def_t recontest = ARG_DEF_ENUM(
    NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
clang-format's avatar
clang-format committed
177 178 179 180 181
static const arg_def_t framerate =
    ARG_DEF(NULL, "fps", 1, "Stream frame rate (rate/scale)");
static const arg_def_t use_webm =
    ARG_DEF(NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, "Output IVF");
182 183 184
#if CONFIG_OBU_NO_IVF
static const arg_def_t use_obu = ARG_DEF(NULL, "obu", 0, "Output OBU");
#endif
clang-format's avatar
clang-format committed
185 186 187 188 189 190 191 192 193 194 195 196 197
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.");
198 199 200 201 202 203 204 205 206 207
static const struct arg_enum_list bitdepth_enum[] = {
  { "8", AOM_BITS_8 }, { "10", AOM_BITS_10 }, { "12", AOM_BITS_12 }, { NULL, 0 }
};

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");
James Zern's avatar
James Zern committed
208 209
static const arg_def_t *main_args[] = { &help,
                                        &debugmode,
clang-format's avatar
clang-format committed
210 211 212 213 214 215 216
                                        &outputfile,
                                        &codecarg,
                                        &passes,
                                        &pass_arg,
                                        &fpf_name,
                                        &limit,
                                        &skip,
217
                                        &good_dl,
clang-format's avatar
clang-format committed
218 219 220 221 222
                                        &quietarg,
                                        &verbosearg,
                                        &psnrarg,
                                        &use_webm,
                                        &use_ivf,
223 224 225
#if CONFIG_OBU_NO_IVF
                                        &use_obu,
#endif
clang-format's avatar
clang-format committed
226 227 228 229 230 231 232 233 234 235 236 237 238 239
                                        &out_part,
                                        &q_hist_n,
                                        &rate_hist_n,
                                        &disable_warnings,
                                        &disable_warning_prompt,
                                        &recontest,
                                        NULL };

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");
Deb Mukherjee's avatar
Deb Mukherjee committed
240 241
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");
242
#if CONFIG_WEBM_IO
243
static const struct arg_enum_list stereo_mode_enum[] = {
clang-format's avatar
clang-format committed
244 245 246 247 248 249
  { "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 }
250
};
Deb Mukherjee's avatar
Deb Mukherjee committed
251 252
static const arg_def_t stereo_mode = ARG_DEF_ENUM(
    NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
253
#endif
Deb Mukherjee's avatar
Deb Mukherjee committed
254 255
static const arg_def_t timebase = ARG_DEF(
    NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
clang-format's avatar
clang-format committed
256 257 258 259
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");
260 261 262 263 264
#if CONFIG_EXT_TILE
static const arg_def_t large_scale_tile =
    ARG_DEF(NULL, "large-scale-tile", 1,
            "Large scale tile coding (0: off (default), 1: on)");
#endif  // CONFIG_EXT_TILE
265 266 267 268
#if CONFIG_MONO_VIDEO
static const arg_def_t monochrome =
    ARG_DEF(NULL, "monochrome", 0, "Monochrome video (no chroma planes)");
#endif  // CONFIG_MONO_VIDEO
269 270 271 272 273 274 275 276 277 278 279

static const arg_def_t *global_args[] = { &use_yv12,
                                          &use_i420,
                                          &use_i422,
                                          &use_i444,
                                          &use_i440,
                                          &usage,
                                          &threads,
                                          &profile,
                                          &width,
                                          &height,
280
#if CONFIG_WEBM_IO
281
                                          &stereo_mode,
282
#endif
283 284 285 286 287 288 289 290
                                          &timebase,
                                          &framerate,
                                          &error_resilient,
                                          &bitdeptharg,
                                          &lag_in_frames,
#if CONFIG_EXT_TILE
                                          &large_scale_tile,
#endif  // CONFIG_EXT_TILE
291 292 293
#if CONFIG_MONO_VIDEO
                                          &monochrome,
#endif  // CONFIG_MONO_VIDEO
294
                                          NULL };
clang-format's avatar
clang-format committed
295 296 297

static const arg_def_t dropframe_thresh =
    ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
298 299
static const arg_def_t resize_mode =
    ARG_DEF(NULL, "resize-mode", 1, "Frame resize mode");
300 301 302 303
static const arg_def_t resize_denominator =
    ARG_DEF(NULL, "resize-denominator", 1, "Frame resize denominator");
static const arg_def_t resize_kf_denominator = ARG_DEF(
    NULL, "resize-kf-denominator", 1, "Frame resize keyframe denominator");
304
#if CONFIG_HORZONLY_FRAME_SUPERRES
305 306
static const arg_def_t superres_mode =
    ARG_DEF(NULL, "superres-mode", 1, "Frame super-resolution mode");
307 308 309 310 311
static const arg_def_t superres_denominator = ARG_DEF(
    NULL, "superres-denominator", 1, "Frame super-resolution denominator");
static const arg_def_t superres_kf_denominator =
    ARG_DEF(NULL, "superres-kf-denominator", 1,
            "Frame super-resolution keyframe denominator");
312 313 314 315 316
static const arg_def_t superres_qthresh = ARG_DEF(
    NULL, "superres-qthresh", 1, "Frame super-resolution qindex threshold");
static const arg_def_t superres_kf_qthresh =
    ARG_DEF(NULL, "superres-kf-qthresh", 1,
            "Frame super-resolution keyframe qindex threshold");
317
#endif  // CONFIG_HORZONLY_FRAME_SUPERRES
318 319 320 321 322
static const struct arg_enum_list end_usage_enum[] = { { "vbr", AOM_VBR },
                                                       { "cbr", AOM_CBR },
                                                       { "cq", AOM_CQ },
                                                       { "q", AOM_Q },
                                                       { NULL, 0 } };
clang-format's avatar
clang-format committed
323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340
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)");
341 342
static const arg_def_t *rc_args[] = { &dropframe_thresh,
                                      &resize_mode,
343 344
                                      &resize_denominator,
                                      &resize_kf_denominator,
345
#if CONFIG_HORZONLY_FRAME_SUPERRES
346
                                      &superres_mode,
347 348
                                      &superres_denominator,
                                      &superres_kf_denominator,
349 350
                                      &superres_qthresh,
                                      &superres_kf_qthresh,
351
#endif  // CONFIG_HORZONLY_FRAME_SUPERRES
352 353 354 355 356 357 358 359 360 361
                                      &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
362

clang-format's avatar
clang-format committed
363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
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)");
static const arg_def_t *rc_twopass_args[] = { &bias_pct, &minsection_pct,
                                              &maxsection_pct, NULL };

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");
static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
                                      NULL };

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");
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)");
John Koleszar's avatar
John Koleszar committed
393
static const struct arg_enum_list tuning_enum[] = {
394 395 396 397 398 399 400
  { "psnr", AOM_TUNE_PSNR },
  { "ssim", AOM_TUNE_SSIM },
#ifdef CONFIG_DIST_8X8
  { "cdef-dist", AOM_TUNE_CDEF_DIST },
  { "daala-dist", AOM_TUNE_DAALA_DIST },
#endif
  { NULL, 0 }
John Koleszar's avatar
John Koleszar committed
401
};
402 403
static const arg_def_t tune_metric =
    ARG_DEF_ENUM(NULL, "tune", 1, "Distortion metric tuned with", tuning_enum);
clang-format's avatar
clang-format committed
404 405 406 407
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)");
408

409 410
#if CONFIG_AV1_ENCODER
static const arg_def_t cpu_used_av1 =
Yaowu Xu's avatar
Yaowu Xu committed
411
    ARG_DEF(NULL, "cpu-used", 1, "CPU Used (0..8)");
412 413
static const arg_def_t dev_sf_av1 =
    ARG_DEF(NULL, "dev-sf", 1, "Dev Speed (0..255)");
414 415 416 417 418
#if CONFIG_EXT_TILE
static const arg_def_t single_tile_decoding =
    ARG_DEF(NULL, "single-tile-decoding", 1,
            "Single tile decoding (0: off (default), 1: on)");
#endif  // CONFIG_EXT_TILE
clang-format's avatar
clang-format committed
419 420 421 422 423
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 (set to 0 while threads > 1)");
424 425
#if CONFIG_MAX_TILE
static const arg_def_t tile_width =
426
    ARG_DEF(NULL, "tile-width", 1, "Tile widths (comma separated)");
427
static const arg_def_t tile_height =
428
    ARG_DEF(NULL, "tile-height", 1, "Tile heights (command separated)");
429
#endif
430 431 432 433
#if CONFIG_DEPENDENT_HORZTILES
static const arg_def_t tile_dependent_rows =
    ARG_DEF(NULL, "tile-dependent-rows", 1, "Enable dependent Tile rows");
#endif
434
#if CONFIG_LOOPFILTERING_ACROSS_TILES
435 436 437 438 439 440 441 442
#if CONFIG_LOOPFILTERING_ACROSS_TILES_EXT
static const arg_def_t tile_loopfilter_v =
    ARG_DEF(NULL, "tile-loopfilter-v", 1,
            "Enable loop filter across vertical tile boundary");
static const arg_def_t tile_loopfilter_h =
    ARG_DEF(NULL, "tile-loopfilter-h", 1,
            "Enable loop filter across horizontal tile boundary");
#else
443 444
static const arg_def_t tile_loopfilter = ARG_DEF(
    NULL, "tile-loopfilter", 1, "Enable loop filter across tile boundary");
445
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES_EXT
446
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
clang-format's avatar
clang-format committed
447 448
static const arg_def_t lossless =
    ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
449 450 451 452
static const arg_def_t enable_cdef =
    ARG_DEF(NULL, "enable-cdef", 1,
            "Enable the constrained directional enhancement filter (0: false, "
            "1: true (default))");
453 454
#if CONFIG_AOM_QM
static const arg_def_t enable_qm =
455
    ARG_DEF(NULL, "enable-qm", 1,
456 457
            "Enable quantisation matrices (0: false (default), 1: true)");
static const arg_def_t qm_min = ARG_DEF(
458
    NULL, "qm-min", 1, "Min quant matrix flatness (0..15), default is 8");
459
static const arg_def_t qm_max = ARG_DEF(
Yaowu Xu's avatar
Yaowu Xu committed
460
    NULL, "qm-max", 1, "Max quant matrix flatness (0..15), default is 15");
461
#endif
462 463 464 465 466
#if CONFIG_DIST_8X8
static const arg_def_t enable_dist_8x8 =
    ARG_DEF(NULL, "enable-dist-8x8", 1,
            "Enable dist-8x8 (0: false (default), 1: true)");
#endif  // CONFIG_DIST_8X8
Yaowu Xu's avatar
Yaowu Xu committed
467 468
static const arg_def_t num_tg = ARG_DEF(
    NULL, "num-tile-groups", 1, "Maximum number of tile groups, default is 1");
469 470
static const arg_def_t mtu_size =
    ARG_DEF(NULL, "mtu-size", 1,
Yaowu Xu's avatar
Yaowu Xu committed
471 472
            "MTU size for a tile group, default is 0 (no MTU targeting), "
            "overrides maximum number of tile groups");
473 474 475 476 477 478 479 480 481 482
#if CONFIG_TIMING_INFO_IN_SEQ_HEADERS
static const struct arg_enum_list timing_info_enum[] = {
  { "unspecified", AOM_TIMING_UNSPECIFIED },
  { "constant", AOM_TIMING_EQUAL },
  { NULL, 0 }
};
static const arg_def_t timing_info =
    ARG_DEF_ENUM(NULL, "timing-info", 1,
                 "Signal timing info in the bitstream:", timing_info_enum);
#endif
483 484 485 486 487 488
#if CONFIG_FILM_GRAIN
static const arg_def_t film_grain_test =
    ARG_DEF(NULL, "film-grain-test", 1,
            "Film grain test vectors (0: none (default), 1: test-1  2: test-2, "
            "... 16: test-16)");
#endif
489 490 491 492
#if CONFIG_TEMPMV_SIGNALING
static const arg_def_t disable_tempmv = ARG_DEF(
    NULL, "disable-tempmv", 1, "Disable temporal mv prediction (default is 0)");
#endif
Yaowu Xu's avatar
Yaowu Xu committed
493 494 495 496
static const arg_def_t frame_parallel_decoding =
    ARG_DEF(NULL, "frame-parallel", 1,
            "Enable frame parallel decodability features "
            "(0: false (default), 1: true)");
497
#if !CONFIG_EXT_DELTA_Q
498 499 500
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
501
    "3: cyclic refresh, 4: delta quant)");
502
#else
Thomas's avatar
Thomas committed
503 504 505
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
506
    "3: cyclic refresh)");
507
#endif
Fangwen Fu's avatar
Fangwen Fu committed
508 509 510 511 512
#if CONFIG_EXT_DELTA_Q
static const arg_def_t deltaq_mode = ARG_DEF(
    NULL, "deltaq-mode", 1,
    "Delta qindex mode (0: off (default), 1: deltaq 2: deltaq + deltalf)");
#endif
clang-format's avatar
clang-format committed
513 514 515
static const arg_def_t frame_periodic_boost =
    ARG_DEF(NULL, "frame-boost", 1,
            "Enable frame periodic boost (0: off (default), 1: on)");
516 517
static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
    NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
clang-format's avatar
clang-format committed
518 519
static const arg_def_t max_inter_rate_pct =
    ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
520 521 522 523 524 525
static const arg_def_t min_gf_interval = ARG_DEF(
    NULL, "min-gf-interval", 1,
    "min gf/arf frame interval (default 0, indicating in-built behavior)");
static const arg_def_t max_gf_interval = ARG_DEF(
    NULL, "max-gf-interval", 1,
    "max gf/arf frame interval (default 0, indicating in-built behavior)");
526

Andrey Norkin's avatar
Andrey Norkin committed
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
#if CONFIG_CICP
static const struct arg_enum_list color_primaries_enum[] = {
  { "bt709", AOM_CICP_CP_BT_709 },
  { "unspecified", AOM_CICP_CP_UNSPECIFIED },
  { "bt601", AOM_CICP_CP_BT_601 },
  { "bt470m", AOM_CICP_CP_BT_470_M },
  { "bt470bg", AOM_CICP_CP_BT_470_B_G },
  { "smpte240", AOM_CICP_CP_SMPTE_240 },
  { "film", AOM_CICP_CP_GENERIC_FILM },
  { "bt2020", AOM_CICP_CP_BT_2020 },
  { "xyz", AOM_CICP_CP_XYZ },
  { "smpte431", AOM_CICP_CP_SMPTE_431 },
  { "smpte432", AOM_CICP_CP_SMPTE_432 },
  { "ebu3213", AOM_CICP_CP_EBU_3213 },
  { NULL, 0 }
};

static const arg_def_t input_color_primaries = ARG_DEF_ENUM(
    NULL, "color-primaries", 1,
    "Color primaries (CICP) of input content:", color_primaries_enum);

static const struct arg_enum_list transfer_characteristics_enum[] = {
  { "unspecified", AOM_CICP_CP_UNSPECIFIED },
  { "bt709", AOM_CICP_TC_BT_709 },
  { "bt470m", AOM_CICP_TC_BT_470_M },
  { "bt470bg", AOM_CICP_TC_BT_470_B_G },
  { "bt601", AOM_CICP_TC_BT_601 },
  { "smpte240", AOM_CICP_TC_SMPTE_240 },
  { "lin", AOM_CICP_TC_LINEAR },
  { "log100", AOM_CICP_TC_LOG_100 },
  { "log100sq10", AOM_CICP_TC_LOG_100_SQRT10 },
  { "iec61966", AOM_CICP_TC_IEC_61966 },
  { "bt1361", AOM_CICP_TC_BT_1361 },
  { "srgb", AOM_CICP_TC_SRGB },
  { "bt2020-10bit", AOM_CICP_TC_BT_2020_10_BIT },
  { "bt2020-12bit", AOM_CICP_TC_BT_2020_12_BIT },
  { "smpte2084", AOM_CICP_TC_SMPTE_2084 },
  { "hlg", AOM_CICP_TC_HLG },
  { "smpte428", AOM_CICP_TC_SMPTE_428 },
  { NULL, 0 }
};

static const arg_def_t input_transfer_characteristics =
    ARG_DEF_ENUM(NULL, "transfer-characteristics", 1,
                 "Transfer characteristics (CICP) of input content:",
                 transfer_characteristics_enum);

static const struct arg_enum_list matrix_coefficients_enum[] = {
  { "identity", AOM_CICP_MC_IDENTITY },
  { "bt709", AOM_CICP_MC_BT_709 },
  { "unspecified", AOM_CICP_MC_UNSPECIFIED },
  { "fcc73", AOM_CICP_MC_FCC },
  { "bt470bg", AOM_CICP_MC_BT_470_B_G },
  { "bt601", AOM_CICP_MC_BT_601 },
  { "smpte240", AOM_CICP_CP_SMPTE_240 },
  { "ycgco", AOM_CICP_MC_SMPTE_YCGCO },
  { "bt2020ncl", AOM_CICP_MC_BT_2020_NCL },
  { "bt2020cl", AOM_CICP_MC_BT_2020_CL },
  { "smpte2085", AOM_CICP_MC_SMPTE_2085 },
  { "chromncl", AOM_CICP_MC_CHROMAT_NCL },
  { "chromcl", AOM_CICP_MC_CHROMAT_CL },
  { "ictcp", AOM_CICP_MC_ICTCP },
  { NULL, 0 }
};

static const arg_def_t input_matrix_coefficients = ARG_DEF_ENUM(
    NULL, "matrix-coefficients", 1,
    "Matrix coefficients (CICP) of input content:", matrix_coefficients_enum);

#else
Yaowu Xu's avatar
Yaowu Xu committed
597
static const struct arg_enum_list color_space_enum[] = {
598 599 600 601 602
  { "unspecified", AOM_CS_UNKNOWN }, { "bt601", AOM_CS_BT_601 },
  { "bt709", AOM_CS_BT_709 },        { "smpte170", AOM_CS_SMPTE_170 },
  { "smpte240", AOM_CS_SMPTE_240 },  { "bt2020ncl", AOM_CS_BT_2020_NCL },
  { "bt2020cl", AOM_CS_BT_2020_CL }, { "sRGB", AOM_CS_SRGB },
  { "ictcp", AOM_CS_ICTCP },         { NULL, 0 }
Yaowu Xu's avatar
Yaowu Xu committed
603 604
};

clang-format's avatar
clang-format committed
605
static const arg_def_t input_color_space =
clang-format's avatar
clang-format committed
606 607
    ARG_DEF_ENUM(NULL, "color-space", 1,
                 "The color space of input content:", color_space_enum);
Yaowu Xu's avatar
Yaowu Xu committed
608

609 610 611 612 613 614 615 616 617
static const struct arg_enum_list transfer_function_enum[] = {
  { "unknown", AOM_TF_UNKNOWN },
  { "bt709", AOM_TF_BT_709 },
  { "pq", AOM_TF_PQ },
  { "hlg", AOM_TF_HLG },
  { NULL, 0 }
};

static const arg_def_t input_transfer_function = ARG_DEF_ENUM(
clang-format's avatar
clang-format committed
618 619
    NULL, "transfer-function", 1,
    "The transfer function of input content:", transfer_function_enum);
620

Andrey Norkin's avatar
Andrey Norkin committed
621
#endif
622 623 624 625 626 627 628 629 630 631 632 633
static const struct arg_enum_list chroma_sample_position_enum[] = {
  { "unknown", AOM_CSP_UNKNOWN },
  { "vertical", AOM_CSP_VERTICAL },
  { "colocated", AOM_CSP_COLOCATED },
  { NULL, 0 }
};

static const arg_def_t input_chroma_sample_position =
    ARG_DEF_ENUM(NULL, "chroma-sample-position", 1,
                 "The chroma sample position when chroma 4:2:0 is signaled:",
                 chroma_sample_position_enum);

634
static const struct arg_enum_list tune_content_enum[] = {
635 636
  { "default", AOM_CONTENT_DEFAULT },
  { "screen", AOM_CONTENT_SCREEN },
clang-format's avatar
clang-format committed
637
  { NULL, 0 }
638 639 640 641
};

static const arg_def_t tune_content = ARG_DEF_ENUM(
    NULL, "tune-content", 1, "Tune content type", tune_content_enum);
hui su's avatar
hui su committed
642
#endif
643

644
#if CONFIG_AV1_ENCODER
645 646
#if CONFIG_EXT_PARTITION
static const struct arg_enum_list superblock_size_enum[] = {
647 648 649
  { "dynamic", AOM_SUPERBLOCK_SIZE_DYNAMIC },
  { "64", AOM_SUPERBLOCK_SIZE_64X64 },
  { "128", AOM_SUPERBLOCK_SIZE_128X128 },
clang-format's avatar
clang-format committed
650
  { NULL, 0 }
651 652 653 654 655
};
static const arg_def_t superblock_size = ARG_DEF_ENUM(
    NULL, "sb-size", 1, "Superblock size to use", superblock_size_enum);
#endif  // CONFIG_EXT_PARTITION

656
static const arg_def_t *av1_args[] = { &cpu_used_av1,
657
                                       &dev_sf_av1,
658 659 660
                                       &auto_altref,
                                       &sharpness,
                                       &static_thresh,
661 662 663
#if CONFIG_EXT_TILE
                                       &single_tile_decoding,
#endif  // CONFIG_EXT_TILE
664 665
                                       &tile_cols,
                                       &tile_rows,
666 667 668
#if CONFIG_DEPENDENT_HORZTILES
                                       &tile_dependent_rows,
#endif
669
#if CONFIG_LOOPFILTERING_ACROSS_TILES
670 671 672 673
#if CONFIG_LOOPFILTERING_ACROSS_TILES_EXT
                                       &tile_loopfilter_v,
                                       &tile_loopfilter_h,
#else
674
                                       &tile_loopfilter,
675
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES_EXT
676
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
677 678
                                       &arnr_maxframes,
                                       &arnr_strength,
679
                                       &tune_metric,
680 681 682 683 684
                                       &cq_level,
                                       &max_intra_rate_pct,
                                       &max_inter_rate_pct,
                                       &gf_cbr_boost_pct,
                                       &lossless,
685
                                       &enable_cdef,
686 687 688 689
#if CONFIG_AOM_QM
                                       &enable_qm,
                                       &qm_min,
                                       &qm_max,
690 691 692
#endif
#if CONFIG_DIST_8X8
                                       &enable_dist_8x8,
693
#endif
694 695
                                       &frame_parallel_decoding,
                                       &aq_mode,
Fangwen Fu's avatar
Fangwen Fu committed
696 697 698
#if CONFIG_EXT_DELTA_Q
                                       &deltaq_mode,
#endif
699 700 701
                                       &frame_periodic_boost,
                                       &noise_sens,
                                       &tune_content,
Andrey Norkin's avatar
Andrey Norkin committed
702 703 704 705 706
#if CONFIG_CICP
                                       &input_color_primaries,
                                       &input_transfer_characteristics,
                                       &input_matrix_coefficients,
#else
707
                                       &input_color_space,
708
                                       &input_transfer_function,
Andrey Norkin's avatar
Andrey Norkin committed
709
#endif
710
                                       &input_chroma_sample_position,
711 712
                                       &min_gf_interval,
                                       &max_gf_interval,
713
#if CONFIG_EXT_PARTITION
714
                                       &superblock_size,
715
#endif  // CONFIG_EXT_PARTITION
716 717
                                       &num_tg,
                                       &mtu_size,
718 719 720
#if CONFIG_TIMING_INFO_IN_SEQ_HEADERS
                                       &timing_info,
#endif
721 722 723
#if CONFIG_FILM_GRAIN
                                       &film_grain_test,
#endif
724 725 726
#if CONFIG_TEMPMV_SIGNALING
                                       &disable_tempmv,
#endif
727 728 729 730
                                       &bitdeptharg,
                                       &inbitdeptharg,
                                       NULL };
static const int av1_arg_ctrl_map[] = { AOME_SET_CPUUSED,
731
                                        AOME_SET_DEVSF,
732 733 734
                                        AOME_SET_ENABLEAUTOALTREF,
                                        AOME_SET_SHARPNESS,
                                        AOME_SET_STATIC_THRESHOLD,
735 736 737
#if CONFIG_EXT_TILE
                                        AV1E_SET_SINGLE_TILE_DECODING,
#endif  // CONFIG_EXT_TILE
738 739
                                        AV1E_SET_TILE_COLUMNS,
                                        AV1E_SET_TILE_ROWS,
740 741 742
#if CONFIG_DEPENDENT_HORZTILES
                                        AV1E_SET_TILE_DEPENDENT_ROWS,
#endif
743
#if CONFIG_LOOPFILTERING_ACROSS_TILES
744 745 746 747
#if CONFIG_LOOPFILTERING_ACROSS_TILES_EXT
                                        AV1E_SET_TILE_LOOPFILTER_V,
                                        AV1E_SET_TILE_LOOPFILTER_H,
#else
748
                                        AV1E_SET_TILE_LOOPFILTER,
749
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES_EXT
750
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
751 752 753 754 755 756 757 758
                                        AOME_SET_ARNR_MAXFRAMES,
                                        AOME_SET_ARNR_STRENGTH,
                                        AOME_SET_TUNING,
                                        AOME_SET_CQ_LEVEL,
                                        AOME_SET_MAX_INTRA_BITRATE_PCT,
                                        AV1E_SET_MAX_INTER_BITRATE_PCT,
                                        AV1E_SET_GF_CBR_BOOST_PCT,
                                        AV1E_SET_LOSSLESS,
759
                                        AV1E_SET_ENABLE_CDEF,
760 761 762 763
#if CONFIG_AOM_QM
                                        AV1E_SET_ENABLE_QM,
                                        AV1E_SET_QM_MIN,
                                        AV1E_SET_QM_MAX,
764 765 766
#endif
#if CONFIG_DIST_8X8
                                        AV1E_SET_ENABLE_DIST_8X8,
767
#endif
768 769
                                        AV1E_SET_FRAME_PARALLEL_DECODING,
                                        AV1E_SET_AQ_MODE,
Fangwen Fu's avatar
Fangwen Fu committed
770 771 772
#if CONFIG_EXT_DELTA_Q
                                        AV1E_SET_DELTAQ_MODE,
#endif
773 774 775
                                        AV1E_SET_FRAME_PERIODIC_BOOST,
                                        AV1E_SET_NOISE_SENSITIVITY,
                                        AV1E_SET_TUNE_CONTENT,
Andrey Norkin's avatar
Andrey Norkin committed
776 777 778 779 780
#if CONFIG_CICP
                                        AV1E_SET_COLOR_PRIMARIES,
                                        AV1E_SET_TRANSFER_CHARACTERISTICS,
                                        AV1E_SET_MATRIX_COEFFICIENTS,
#else
781
                                        AV1E_SET_COLOR_SPACE,
782
                                        AV1E_SET_TRANSFER_FUNCTION,
Andrey Norkin's avatar
Andrey Norkin committed
783
#endif
784
                                        AV1E_SET_CHROMA_SAMPLE_POSITION,
785 786
                                        AV1E_SET_MIN_GF_INTERVAL,
                                        AV1E_SET_MAX_GF_INTERVAL,
787
#if CONFIG_EXT_PARTITION
788
                                        AV1E_SET_SUPERBLOCK_SIZE,
789
#endif  // CONFIG_EXT_PARTITION
790 791
                                        AV1E_SET_NUM_TG,
                                        AV1E_SET_MTU,
792 793 794
#if CONFIG_TIMING_INFO_IN_SEQ_HEADERS
                                        AV1E_SET_TIMING_INFO,
#endif
795 796 797
#if CONFIG_FILM_GRAIN
                                        AV1E_SET_FILM_GRAIN_TEST_VECTOR,
#endif
798 799
#if CONFIG_TEMPMV_SIGNALING
                                        AV1E_SET_DISABLE_TEMPMV,
800
#endif
801
                                        0 };
hui su's avatar
hui su committed
802 803
#endif

John Koleszar's avatar
John Koleszar committed
804 805
static const arg_def_t *no_args[] = { NULL };

James Zern's avatar
James Zern committed
806 807
void show_help(FILE *fout, int shorthelp) {
  fprintf(fout, "Usage: %s <options> -o dst_filename src_filename \n",
John Koleszar's avatar
John Koleszar committed
808 809
          exec_name);

James Zern's avatar
James Zern committed
810 811 812 813 814 815 816 817 818 819 820 821 822 823 824
  if (shorthelp) {
    fprintf(fout, "Use --help to see the full list of options.\n");
    return;
  }

  fprintf(fout, "\nOptions:\n");
  arg_show_usage(fout, main_args);
  fprintf(fout, "\nEncoder Global Options:\n");
  arg_show_usage(fout, global_args);
  fprintf(fout, "\nRate Control Options:\n");
  arg_show_usage(fout, rc_args);
  fprintf(fout, "\nTwopass Rate Control Options:\n");
  arg_show_usage(fout, rc_twopass_args);
  fprintf(fout, "\nKeyframe Placement Options:\n");
  arg_show_usage(fout, kf_args);
825
#if CONFIG_AV1_ENCODER
James Zern's avatar
James Zern committed
826 827
  fprintf(fout, "\nAV1 Specific Options:\n");
  arg_show_usage(fout, av1_args);
John Koleszar's avatar
John Koleszar committed
828
#endif
James Zern's avatar
James Zern committed
829
  fprintf(fout,
clang-format's avatar
clang-format committed
830
          "\nStream timebase (--timebase):\n"
John Koleszar's avatar
John Koleszar committed
831 832
          "  The desired precision of timestamps in the output, expressed\n"
          "  in fractional seconds. Default is 1/1000.\n");
James Zern's avatar
James Zern committed
833
  fprintf(fout, "\nIncluded encoders:\n\n");
John Koleszar's avatar
John Koleszar committed
834

James Zern's avatar
James Zern committed
835 836
  const int num_encoder = get_aom_encoder_count();
  for (int i = 0; i < num_encoder; ++i) {
837
    const AvxInterface *const encoder = get_aom_encoder_by_index(i);
clang-format's avatar
clang-format committed
838
    const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
James Zern's avatar
James Zern committed
839
    fprintf(fout, "    %-6s - %s %s\n", encoder->name,
840
            aom_codec_iface_name(encoder->codec_interface()), defstr);
841
  }
James Zern's avatar
James Zern committed
842 843 844
  fprintf(fout, "\n        ");
  fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
}
John Koleszar's avatar
John Koleszar committed
845

James Zern's avatar
James Zern committed
846 847
void usage_exit(void) {
  show_help(stderr, 1);
John Koleszar's avatar
John Koleszar committed
848
  exit(EXIT_FAILURE);
John Koleszar's avatar
John Koleszar committed
849 850
}

851 852
#if CONFIG_AV1_ENCODER
#define ARG_CTRL_CNT_MAX NELEMENTS(av1_arg_ctrl_map)
853
#endif
John Koleszar's avatar
John Koleszar committed
854

855 856
#if !CONFIG_WEBM_IO
typedef int stereo_format_t;
clang-format's avatar
clang-format committed
857 858 859
struct WebmOutputContext {
  int debug;
};
860 861
#endif

862
/* Per-stream configuration */
John Koleszar's avatar
John Koleszar committed
863
struct stream_config {
864
  struct aom_codec_enc_cfg cfg;
clang-format's avatar
clang-format committed
865 866
  const char *out_fn;
  const char *stats_fn;
867
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
868
  const char *fpmb_stats_fn;
869
#endif
clang-format's avatar
clang-format committed
870 871 872 873
  stereo_format_t stereo_fmt;
  int arg_ctrls[ARG_CTRL_CNT_MAX][2];
  int arg_ctrl_cnt;
  int write_webm;
874 875 876
#if CONFIG_OBU_NO_IVF
  int write_ivf;
#endif
877
  // whether to use 16bit internal buffers
clang-format's avatar
clang-format committed
878
  int use_16bit_internal;
879 880
};

John Koleszar's avatar
John Koleszar committed
881
struct stream_state {
clang-format's avatar
clang-format committed
882 883 884 885 886 887 888 889 890 891 892
  int index;
  struct stream_state *next;
  struct stream_config config;
  FILE *file;
  struct rate_hist *rate_hist;
  struct WebmOutputContext webm_ctx;
  uint64_t psnr_sse_total;
  uint64_t psnr_samples_total;
  double psnr_totals[4];
  int psnr_count;
  int counts[64];
893
  aom_codec_ctx_t encoder;
clang-format's avatar
clang-format committed
894 895 896 897
  unsigned int frames_out;
  uint64_t cx_time;
  size_t nbytes;
  stats_io_t stats;
898
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
899
  stats_io_t fpmb_stats;
900
#endif
901 902
  struct aom_image *img;
  aom_codec_ctx_t decoder;
clang-format's avatar
clang-format committed
903
  int mismatch_seen;
904 905
};

clang-format's avatar
clang-format committed
906
static void validate_positive_rational(const char *msg,
907
                                       struct aom_rational *rat) {
John Koleszar's avatar
John Koleszar committed
908 909 910 911
  if (rat->den < 0) {
    rat->num *= -1;
    rat->den *= -1;
  }
912

clang-format's avatar
clang-format committed
913
  if (rat->num < 0) die("Error: %s must be positive\n", msg);
914

clang-format's avatar
clang-format committed
915
  if (!rat->den) die("Error: %s has zero denominator\n", msg);
916 917
}

918
static void parse_global_config(struct AvxEncoderConfig *global, char **argv) {
clang-format's avatar
clang-format committed
919 920
  char **argi, **argj;
  struct arg arg;
921
  const int num_encoder = get_aom_encoder_count();
922

clang-format's avatar
clang-format committed
923
  if (num_encoder < 1) die("Error: no valid encoder available\n");
John Koleszar's avatar
John Koleszar committed
924 925 926

  /* Initialize default parameters */
  memset(global, 0, sizeof(*global));
927
  global->codec = get_aom_encoder_by_index(num_encoder - 1);
928
  global->passes = 0;
929
  global->color_type = I420;
John Koleszar's avatar
John Koleszar committed
930 931 932 933

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

James Zern's avatar
James Zern committed
934 935 936 937
    if (arg_match(&arg, &help, argi)) {
      show_help(stdout, 0);
      exit(EXIT_SUCCESS);
    } else if (arg_match(&arg, &codecarg, argi)) {
938
      global->codec = get_aom_encoder_by_name(arg.val);
939 940
      if (!global->codec)
        die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
John Koleszar's avatar
John Koleszar committed
941
    } else if (arg_match(&arg, &passes, argi)) {
John Koleszar's avatar
John Koleszar committed
942
      global->passes = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
943

John Koleszar's avatar
John Koleszar committed
944 945
      if (global->passes < 1 || global->passes > 2)
        die("Error: Invalid number of passes (%d)\n", global->passes);
John Koleszar's avatar
John Koleszar committed
946
    } else if (arg_match(&arg, &pass_arg, argi)) {
John Koleszar's avatar
John Koleszar committed
947 948 949
      global->pass = arg_parse_uint(&arg);

      if (global->pass < 1 || global->pass > 2)
clang-format's avatar
clang-format committed
950
        die("Error: Invalid pass selected (%d)\n", global->pass);
John Koleszar's avatar
John Koleszar committed
951 952
    } else if (arg_match(&arg, &usage, argi))
      global->usage = arg_parse_uint(&arg);
953 954
    else if (arg_match(&arg, &good_dl, argi))
      warn("Deprecated --good option! Ignoring\n");
John Koleszar's avatar
John Koleszar committed
955
    else if (arg_match(&arg, &use_yv12, argi))
956
      global->color_type = YV12;
John Koleszar's avatar
John Koleszar committed
957
    else if (arg_match(&arg, &use_i420, argi))
958 959 960 961 962
      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;
Deb Mukherjee's avatar
Deb Mukherjee committed
963 964
    else if (arg_match(&arg, &use_i440, argi))
      global->color_type = I440;
John Koleszar's avatar
John Koleszar committed
965 966 967 968
    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
969
    else if (arg_match(&arg, &limit, argi))
John Koleszar's avatar
John Koleszar committed
970
      global->limit = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
971
    else if (arg_match(&arg, &skip, argi))
John Koleszar's avatar
John Koleszar committed
972
      global->skip_frames = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
973
    else if (arg_match(&arg, &psnrarg, argi))
John Koleszar's avatar
John Koleszar committed
974
      global->show_psnr = 1;
John Koleszar's avatar
John Koleszar committed
975
    else if (arg_match(&arg, &recontest, argi))
976
      global->test_decode = arg_parse_enum_or_int(&arg);
John Koleszar's avatar
John Koleszar committed
977
    else if (arg_match(&arg, &framerate, argi)) {
John Koleszar's avatar
John Koleszar committed
978 979 980 981 982
      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
983
    else if (arg_match(&arg, &debugmode, argi))
John Koleszar's avatar
John Koleszar committed
984
      global->debug = 1;
John Koleszar's avatar
John Koleszar committed
985
    else if (arg_match(&arg, &q_hist_n, argi))
John Koleszar's avatar
John Koleszar committed
986
      global->show_q_hist_buckets = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
987
    else if (arg_match(&arg, &rate_hist_n, argi))
John Koleszar's avatar
John Koleszar committed
988
      global->show_rate_hist_buckets = arg_parse_uint(&arg);
989 990 991 992
    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
993 994 995
    else
      argj++;
  }
John Koleszar's avatar
John Koleszar committed
996

997 998 999
  if (global->pass) {
    /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
    if (global->pass > global->passes) {
clang-format's avatar
clang-format committed
1000 1001
      warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
           global->pass);
1002 1003 1004
      global->passes = global->pass;
    }
  }
John Koleszar's avatar
John Koleszar committed
1005
  /* Validate global config */
1006
  if (global->passes == 0) {
1007 1008
#if CONFIG_AV1_ENCODER
    // Make default AV1 passes = 2 until there is a better quality 1-pass
1009
    // encoder
1010
    if (global->codec != NULL && global->codec->name != NULL)
Thomas Daede's avatar
Thomas Daede committed
1011
      global->passes = (strcmp(global->codec->name, "av1") == 0) ? 2 : 1;
James Zern's avatar
James Zern committed
1012 1013 1014
#else
    global->passes = 1;
#endif
1015
  }
1016 1017
}

1018
static void open_input_file(struct AvxInputContext *input) {
John Koleszar's avatar
John Koleszar committed
1019
  /* Parse certain options from the input file, if possible */
clang-format's avatar
clang-format committed
1020 1021
  input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
                                             : set_binary_mode(stdin);
John Koleszar's avatar
John Koleszar committed
1022

clang-format's avatar
clang-format committed
1023
  if (!input->file) fatal("Failed to open input file");
John Koleszar's avatar
John Koleszar committed
1024

1025 1026 1027 1028 1029 1030 1031 1032
  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);
  }

1033 1034 1035 1036
  /* Default to 1:1 pixel aspect ratio. */
  input->pixel_aspect_ratio.numerator = 1;
  input->pixel_aspect_ratio.denominator = 1;

John Koleszar's avatar
John Koleszar committed
1037 1038 1039 1040 1041 1042
  /* 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;

clang-format's avatar
clang-format committed
1043
  if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
1044 1045
    if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
                       input->only_i420) >= 0) {
John Koleszar's avatar
John Koleszar committed
1046
      input->file_type = FILE_TYPE_Y4M;
1047 1048
      input->width = input->y4m.pic_w;
      input->height = input->y4m.pic_h;
1049 1050
      input->pixel_aspect_ratio.numerator = input->y4m.par_n;
      input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1051 1052
      input->framerate.numerator = input->y4m.fps_n;
      input->framerate.denominator = input->y4m.fps_d;
1053
      input->fmt = input->y4m.aom_fmt;
1054
      input->bit_depth = input->y4m.bit_depth;
John Koleszar's avatar
John Koleszar committed
1055 1056
    } else
      fatal("Unsupported Y4M stream.");
1057
  } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1058
    fatal("IVF is not supported as input.");
John Koleszar's avatar
John Koleszar committed
1059 1060
  } else {
    input->file_type = FILE_TYPE_RAW;
John Koleszar's avatar
John Koleszar committed
1061
  }
1062 1063
}

1064
static void close_input_file(struct AvxInputContext *input) {
John Koleszar's avatar
John Koleszar committed
1065
  fclose(input->file);
clang-format's avatar
clang-format committed
1066
  if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
1067 1068
}

1069
static struct stream_state *new_stream(struct AvxEncoderConfig *global,
1070
                                       struct stream_state *prev) {
John Koleszar's avatar
John Koleszar committed
1071 1072 1073
  struct stream_state *stream;

  stream = calloc(1, sizeof(*stream));
1074
  if (stream == NULL) {
John Koleszar's avatar
John Koleszar committed
1075
    fatal("Failed to allocate new stream.");
1076 1077
  }

John Koleszar's avatar
John Koleszar committed
1078 1079 1080 1081 1082
  if (prev) {
    memcpy(stream, prev, sizeof(*stream));
    stream->index++;
    prev->next = stream;
  } else {
1083
    aom_codec_err_t res;
John Koleszar's avatar
John Koleszar committed
1084 1085

    /* Populate encoder configuration */
1086
    res = aom_codec_enc_config_default(global->codec->codec_interface(),
clang-format's avatar
clang-format committed
1087
                                       &stream->config.cfg, global->usage);
1088
    if (res) fatal("Failed to get config: %s\n", aom_codec_err_to_string(res));
John Koleszar's avatar
John Koleszar committed
1089

John Koleszar's avatar
John Koleszar committed
1090 1091 1092 1093
    /* 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;
1094

John Koleszar's avatar
John Koleszar committed
1095 1096 1097 1098 1099
    /* 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
1100

John Koleszar's avatar
John Koleszar committed
1101 1102
    /* Initialize remaining stream parameters */
    stream->config.write_webm = 1;
1103 1104 1105
#if CONFIG_OBU_NO_IVF
    stream->config.write_ivf = 0;
#endif
1106
#if CONFIG_WEBM_IO
1107
    stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1108 1109 1110
    stream->webm_ctx.last_pts_ns = -1;
    stream->webm_ctx.writer = NULL;
    stream->webm_ctx.segment = NULL;
1111
#endif
John Koleszar's avatar
John Koleszar committed
1112

John Koleszar's avatar
John Koleszar committed
1113
    /* Allows removal of the application version from the EBML tags */
1114
    stream->webm_ctx<