OSUOSL/Nero are experiencing Internet connectivity problems. This affects us as we're hosted with OSUOSL. We apologize for the inconvenience.

aomenc.c 80.7 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

clang-format's avatar
clang-format committed
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149
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");
150
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
151 152
static const arg_def_t fpmbf_name =
    ARG_DEF(NULL, "fpmbf", 1, "First pass block statistics file name");
153
#endif
clang-format's avatar
clang-format committed
154 155 156 157 158 159 160 161 162 163 164 165 166 167
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 good_dl =
    ARG_DEF(NULL, "good", 0, "Use Good 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");
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
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
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");
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.");
195

196
#if CONFIG_HIGHBITDEPTH
197 198 199 200 201 202 203 204 205 206
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");
207
#endif
208

clang-format's avatar
clang-format committed
209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
static const arg_def_t *main_args[] = { &debugmode,
                                        &outputfile,
                                        &codecarg,
                                        &passes,
                                        &pass_arg,
                                        &fpf_name,
                                        &limit,
                                        &skip,
                                        &deadline,
                                        &good_dl,
                                        &quietarg,
                                        &verbosearg,
                                        &psnrarg,
                                        &use_webm,
                                        &use_ivf,
                                        &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
238 239
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");
240
#if CONFIG_WEBM_IO
241
static const struct arg_enum_list stereo_mode_enum[] = {
clang-format's avatar
clang-format committed
242 243 244 245 246 247
  { "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 }
248
};
Deb Mukherjee's avatar
Deb Mukherjee committed
249 250
static const arg_def_t stereo_mode = ARG_DEF_ENUM(
    NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
251
#endif
Deb Mukherjee's avatar
Deb Mukherjee committed
252 253
static const arg_def_t timebase = ARG_DEF(
    NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
clang-format's avatar
clang-format committed
254 255 256 257
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");
258 259 260 261 262
#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
263 264 265 266
#if CONFIG_MONO_VIDEO
static const arg_def_t monochrome =
    ARG_DEF(NULL, "monochrome", 0, "Monochrome video (no chroma planes)");
#endif  // MONOCHROME
267 268 269 270 271 272 273 274 275 276 277

static const arg_def_t *global_args[] = { &use_yv12,
                                          &use_i420,
                                          &use_i422,
                                          &use_i444,
                                          &use_i440,
                                          &usage,
                                          &threads,
                                          &profile,
                                          &width,
                                          &height,
278
#if CONFIG_WEBM_IO
279
                                          &stereo_mode,
280
#endif
281 282 283
                                          &timebase,
                                          &framerate,
                                          &error_resilient,
284
#if CONFIG_HIGHBITDEPTH
285
                                          &bitdeptharg,
286
#endif
287 288 289 290
                                          &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");
304 305 306
#if CONFIG_FRAME_SUPERRES
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_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_FRAME_SUPERRES
346
                                      &superres_mode,
347 348
                                      &superres_denominator,
                                      &superres_kf_denominator,
349 350
                                      &superres_qthresh,
                                      &superres_kf_qthresh,
351
#endif  // CONFIG_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)");
412 413 414 415 416
#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
417 418 419 420 421
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)");
422 423
#if CONFIG_MAX_TILE
static const arg_def_t tile_width =
424
    ARG_DEF(NULL, "tile-width", 1, "Tile widths (comma separated)");
425
static const arg_def_t tile_height =
426
    ARG_DEF(NULL, "tile-height", 1, "Tile heights (command separated)");
427
#endif
428 429 430 431
#if CONFIG_DEPENDENT_HORZTILES
static const arg_def_t tile_dependent_rows =
    ARG_DEF(NULL, "tile-dependent-rows", 1, "Enable dependent Tile rows");
#endif
432
#if CONFIG_LOOPFILTERING_ACROSS_TILES
433 434
static const arg_def_t tile_loopfilter = ARG_DEF(
    NULL, "tile-loopfilter", 1, "Enable loop filter across tile boundary");
435
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
clang-format's avatar
clang-format committed
436 437
static const arg_def_t lossless =
    ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
438 439 440 441
static const arg_def_t enable_cdef =
    ARG_DEF(NULL, "enable-cdef", 1,
            "Enable the constrained directional enhancement filter (0: false, "
            "1: true (default))");
442 443
#if CONFIG_AOM_QM
static const arg_def_t enable_qm =
444
    ARG_DEF(NULL, "enable-qm", 1,
445 446
            "Enable quantisation matrices (0: false (default), 1: true)");
static const arg_def_t qm_min = ARG_DEF(
447
    NULL, "qm-min", 1, "Min quant matrix flatness (0..15), default is 8");
448
static const arg_def_t qm_max = ARG_DEF(
449
    NULL, "qm-max", 1, "Max quant matrix flatness (0..15), default is 16");
450
#endif
451 452 453 454 455
#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
456 457
static const arg_def_t num_tg = ARG_DEF(
    NULL, "num-tile-groups", 1, "Maximum number of tile groups, default is 1");
458 459
static const arg_def_t mtu_size =
    ARG_DEF(NULL, "mtu-size", 1,
Yaowu Xu's avatar
Yaowu Xu committed
460 461
            "MTU size for a tile group, default is 0 (no MTU targeting), "
            "overrides maximum number of tile groups");
462 463 464 465
#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
466 467 468 469
static const arg_def_t frame_parallel_decoding =
    ARG_DEF(NULL, "frame-parallel", 1,
            "Enable frame parallel decodability features "
            "(0: false (default), 1: true)");
470
#if !CONFIG_EXT_DELTA_Q
471 472 473
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
474
    "3: cyclic refresh, 4: delta quant)");
475
#else
Thomas's avatar
Thomas committed
476 477 478
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
479
    "3: cyclic refresh)");
480
#endif
Fangwen Fu's avatar
Fangwen Fu committed
481 482 483 484 485
#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
486 487 488
static const arg_def_t frame_periodic_boost =
    ARG_DEF(NULL, "frame-boost", 1,
            "Enable frame periodic boost (0: off (default), 1: on)");
489 490
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
491 492
static const arg_def_t max_inter_rate_pct =
    ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
493 494 495 496 497 498
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)");
499

Yaowu Xu's avatar
Yaowu Xu committed
500
static const struct arg_enum_list color_space_enum[] = {
501 502 503 504 505
  { "unknown", 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
506 507
};

clang-format's avatar
clang-format committed
508
static const arg_def_t input_color_space =
clang-format's avatar
clang-format committed
509 510
    ARG_DEF_ENUM(NULL, "color-space", 1,
                 "The color space of input content:", color_space_enum);
Yaowu Xu's avatar
Yaowu Xu committed
511

512 513 514 515 516 517 518 519 520
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
521 522
    NULL, "transfer-function", 1,
    "The transfer function of input content:", transfer_function_enum);
523 524 525 526 527 528 529 530 531 532 533 534 535

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);

536
static const struct arg_enum_list tune_content_enum[] = {
537 538
  { "default", AOM_CONTENT_DEFAULT },
  { "screen", AOM_CONTENT_SCREEN },
clang-format's avatar
clang-format committed
539
  { NULL, 0 }
540 541 542 543
};

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
544
#endif
545

546
#if CONFIG_AV1_ENCODER
547 548
#if CONFIG_EXT_PARTITION
static const struct arg_enum_list superblock_size_enum[] = {
549 550 551
  { "dynamic", AOM_SUPERBLOCK_SIZE_DYNAMIC },
  { "64", AOM_SUPERBLOCK_SIZE_64X64 },
  { "128", AOM_SUPERBLOCK_SIZE_128X128 },
clang-format's avatar
clang-format committed
552
  { NULL, 0 }
553 554 555 556 557
};
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

558 559 560 561
static const arg_def_t *av1_args[] = { &cpu_used_av1,
                                       &auto_altref,
                                       &sharpness,
                                       &static_thresh,
562 563 564
#if CONFIG_EXT_TILE
                                       &single_tile_decoding,
#endif  // CONFIG_EXT_TILE
565 566
                                       &tile_cols,
                                       &tile_rows,
567 568 569
#if CONFIG_DEPENDENT_HORZTILES
                                       &tile_dependent_rows,
#endif
570
#if CONFIG_LOOPFILTERING_ACROSS_TILES
571
                                       &tile_loopfilter,
572
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
573 574
                                       &arnr_maxframes,
                                       &arnr_strength,
575
                                       &tune_metric,
576 577 578 579 580
                                       &cq_level,
                                       &max_intra_rate_pct,
                                       &max_inter_rate_pct,
                                       &gf_cbr_boost_pct,
                                       &lossless,
581
                                       &enable_cdef,
582 583 584 585
#if CONFIG_AOM_QM
                                       &enable_qm,
                                       &qm_min,
                                       &qm_max,
586 587 588
#endif
#if CONFIG_DIST_8X8
                                       &enable_dist_8x8,
589
#endif
590 591
                                       &frame_parallel_decoding,
                                       &aq_mode,
Fangwen Fu's avatar
Fangwen Fu committed
592 593 594
#if CONFIG_EXT_DELTA_Q
                                       &deltaq_mode,
#endif
595 596 597 598
                                       &frame_periodic_boost,
                                       &noise_sens,
                                       &tune_content,
                                       &input_color_space,
599 600
                                       &input_transfer_function,
                                       &input_chroma_sample_position,
601 602
                                       &min_gf_interval,
                                       &max_gf_interval,
603
#if CONFIG_EXT_PARTITION
604
                                       &superblock_size,
605
#endif  // CONFIG_EXT_PARTITION
606 607
                                       &num_tg,
                                       &mtu_size,
608 609 610
#if CONFIG_TEMPMV_SIGNALING
                                       &disable_tempmv,
#endif
611
#if CONFIG_HIGHBITDEPTH
612 613
                                       &bitdeptharg,
                                       &inbitdeptharg,
614
#endif  // CONFIG_HIGHBITDEPTH
615 616 617 618 619
                                       NULL };
static const int av1_arg_ctrl_map[] = { AOME_SET_CPUUSED,
                                        AOME_SET_ENABLEAUTOALTREF,
                                        AOME_SET_SHARPNESS,
                                        AOME_SET_STATIC_THRESHOLD,
620 621 622
#if CONFIG_EXT_TILE
                                        AV1E_SET_SINGLE_TILE_DECODING,
#endif  // CONFIG_EXT_TILE
623 624
                                        AV1E_SET_TILE_COLUMNS,
                                        AV1E_SET_TILE_ROWS,
625 626 627
#if CONFIG_DEPENDENT_HORZTILES
                                        AV1E_SET_TILE_DEPENDENT_ROWS,
#endif
628
#if CONFIG_LOOPFILTERING_ACROSS_TILES
629
                                        AV1E_SET_TILE_LOOPFILTER,
630
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
631 632 633 634 635 636 637 638
                                        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,
639
                                        AV1E_SET_ENABLE_CDEF,
640 641 642 643
#if CONFIG_AOM_QM
                                        AV1E_SET_ENABLE_QM,
                                        AV1E_SET_QM_MIN,
                                        AV1E_SET_QM_MAX,
644 645 646
#endif
#if CONFIG_DIST_8X8
                                        AV1E_SET_ENABLE_DIST_8X8,
647
#endif
648 649
                                        AV1E_SET_FRAME_PARALLEL_DECODING,
                                        AV1E_SET_AQ_MODE,
Fangwen Fu's avatar
Fangwen Fu committed
650 651 652
#if CONFIG_EXT_DELTA_Q
                                        AV1E_SET_DELTAQ_MODE,
#endif
653 654 655 656
                                        AV1E_SET_FRAME_PERIODIC_BOOST,
                                        AV1E_SET_NOISE_SENSITIVITY,
                                        AV1E_SET_TUNE_CONTENT,
                                        AV1E_SET_COLOR_SPACE,
657 658
                                        AV1E_SET_TRANSFER_FUNCTION,
                                        AV1E_SET_CHROMA_SAMPLE_POSITION,
659 660
                                        AV1E_SET_MIN_GF_INTERVAL,
                                        AV1E_SET_MAX_GF_INTERVAL,
661
#if CONFIG_EXT_PARTITION
662
                                        AV1E_SET_SUPERBLOCK_SIZE,
663
#endif  // CONFIG_EXT_PARTITION
664 665
                                        AV1E_SET_NUM_TG,
                                        AV1E_SET_MTU,
666 667
#if CONFIG_TEMPMV_SIGNALING
                                        AV1E_SET_DISABLE_TEMPMV,
668
#endif
669
                                        0 };
hui su's avatar
hui su committed
670 671
#endif

John Koleszar's avatar
John Koleszar committed
672 673
static const arg_def_t *no_args[] = { NULL };

674
void usage_exit(void) {
John Koleszar's avatar
John Koleszar committed
675
  int i;
676
  const int num_encoder = get_aom_encoder_count();
John Koleszar's avatar
John Koleszar committed
677 678 679 680 681

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

  fprintf(stderr, "\nOptions:\n");
John Koleszar's avatar
John Koleszar committed
682
  arg_show_usage(stderr, main_args);
John Koleszar's avatar
John Koleszar committed
683
  fprintf(stderr, "\nEncoder Global Options:\n");
John Koleszar's avatar
John Koleszar committed
684
  arg_show_usage(stderr, global_args);
John Koleszar's avatar
John Koleszar committed
685
  fprintf(stderr, "\nRate Control Options:\n");
John Koleszar's avatar
John Koleszar committed
686
  arg_show_usage(stderr, rc_args);
John Koleszar's avatar
John Koleszar committed
687
  fprintf(stderr, "\nTwopass Rate Control Options:\n");
John Koleszar's avatar
John Koleszar committed
688
  arg_show_usage(stderr, rc_twopass_args);
John Koleszar's avatar
John Koleszar committed
689
  fprintf(stderr, "\nKeyframe Placement Options:\n");
John Koleszar's avatar
John Koleszar committed
690
  arg_show_usage(stderr, kf_args);
691 692 693
#if CONFIG_AV1_ENCODER
  fprintf(stderr, "\nAV1 Specific Options:\n");
  arg_show_usage(stderr, av1_args);
John Koleszar's avatar
John Koleszar committed
694
#endif
clang-format's avatar
clang-format committed
695 696
  fprintf(stderr,
          "\nStream timebase (--timebase):\n"
John Koleszar's avatar
John Koleszar committed
697 698
          "  The desired precision of timestamps in the output, expressed\n"
          "  in fractional seconds. Default is 1/1000.\n");
699
  fprintf(stderr, "\nIncluded encoders:\n\n");
John Koleszar's avatar
John Koleszar committed
700

701
  for (i = 0; i < num_encoder; ++i) {
702
    const AvxInterface *const encoder = get_aom_encoder_by_index(i);
clang-format's avatar
clang-format committed
703 704
    const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
    fprintf(stderr, "    %-6s - %s %s\n", encoder->name,
705
            aom_codec_iface_name(encoder->codec_interface()), defstr);
706
  }
707 708
  fprintf(stderr, "\n        ");
  fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
John Koleszar's avatar
John Koleszar committed
709 710

  exit(EXIT_FAILURE);
John Koleszar's avatar
John Koleszar committed
711 712
}

713 714
#if CONFIG_AV1_ENCODER
#define ARG_CTRL_CNT_MAX NELEMENTS(av1_arg_ctrl_map)
715
#endif
John Koleszar's avatar
John Koleszar committed
716

717 718
#if !CONFIG_WEBM_IO
typedef int stereo_format_t;
clang-format's avatar
clang-format committed
719 720 721
struct WebmOutputContext {
  int debug;
};
722 723
#endif

724
/* Per-stream configuration */
John Koleszar's avatar
John Koleszar committed
725
struct stream_config {
726
  struct aom_codec_enc_cfg cfg;
clang-format's avatar
clang-format committed
727 728
  const char *out_fn;
  const char *stats_fn;
729
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
730
  const char *fpmb_stats_fn;
731
#endif
clang-format's avatar
clang-format committed
732 733 734 735
  stereo_format_t stereo_fmt;
  int arg_ctrls[ARG_CTRL_CNT_MAX][2];
  int arg_ctrl_cnt;
  int write_webm;
736
  // whether to use 16bit internal buffers
clang-format's avatar
clang-format committed
737
  int use_16bit_internal;
738 739
};

John Koleszar's avatar
John Koleszar committed
740
struct stream_state {
clang-format's avatar
clang-format committed
741 742 743 744 745 746 747 748 749 750 751
  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];
752
  aom_codec_ctx_t encoder;
clang-format's avatar
clang-format committed
753 754 755 756
  unsigned int frames_out;
  uint64_t cx_time;
  size_t nbytes;
  stats_io_t stats;
757
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
758
  stats_io_t fpmb_stats;
759
#endif
760 761
  struct aom_image *img;
  aom_codec_ctx_t decoder;
clang-format's avatar
clang-format committed
762
  int mismatch_seen;
763 764
};

clang-format's avatar
clang-format committed
765
static void validate_positive_rational(const char *msg,
766
                                       struct aom_rational *rat) {
John Koleszar's avatar
John Koleszar committed
767 768 769 770
  if (rat->den < 0) {
    rat->num *= -1;
    rat->den *= -1;
  }
771

clang-format's avatar
clang-format committed
772
  if (rat->num < 0) die("Error: %s must be positive\n", msg);
773

clang-format's avatar
clang-format committed
774
  if (!rat->den) die("Error: %s has zero denominator\n", msg);
775 776
}

777
static void parse_global_config(struct AvxEncoderConfig *global, char **argv) {
clang-format's avatar
clang-format committed
778 779
  char **argi, **argj;
  struct arg arg;
780
  const int num_encoder = get_aom_encoder_count();
781

clang-format's avatar
clang-format committed
782
  if (num_encoder < 1) die("Error: no valid encoder available\n");
John Koleszar's avatar
John Koleszar committed
783 784 785

  /* Initialize default parameters */
  memset(global, 0, sizeof(*global));
786
  global->codec = get_aom_encoder_by_index(num_encoder - 1);
787
  global->passes = 0;
788
  global->color_type = I420;
789
  /* Assign default deadline to good quality */
790
  global->deadline = AOM_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
791 792 793 794 795

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

    if (arg_match(&arg, &codecarg, argi)) {
796
      global->codec = get_aom_encoder_by_name(arg.val);
797 798
      if (!global->codec)
        die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
John Koleszar's avatar
John Koleszar committed
799
    } else if (arg_match(&arg, &passes, argi)) {
John Koleszar's avatar
John Koleszar committed
800
      global->passes = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
801

John Koleszar's avatar
John Koleszar committed
802 803
      if (global->passes < 1 || global->passes > 2)
        die("Error: Invalid number of passes (%d)\n", global->passes);
John Koleszar's avatar
John Koleszar committed
804
    } else if (arg_match(&arg, &pass_arg, argi)) {
John Koleszar's avatar
John Koleszar committed
805 806 807
      global->pass = arg_parse_uint(&arg);

      if (global->pass < 1 || global->pass > 2)
clang-format's avatar
clang-format committed
808
        die("Error: Invalid pass selected (%d)\n", global->pass);
John Koleszar's avatar
John Koleszar committed
809 810
    } else if (arg_match(&arg, &usage, argi))
      global->usage = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
811
    else if (arg_match(&arg, &deadline, argi))
John Koleszar's avatar
John Koleszar committed
812
      global->deadline = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
813
    else if (arg_match(&arg, &good_dl, argi))
814
      global->deadline = AOM_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
815
    else if (arg_match(&arg, &use_yv12, argi))
816
      global->color_type = YV12;
John Koleszar's avatar
John Koleszar committed
817
    else if (arg_match(&arg, &use_i420, argi))
818 819 820 821 822
      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
823 824
    else if (arg_match(&arg, &use_i440, argi))
      global->color_type = I440;
John Koleszar's avatar
John Koleszar committed
825 826 827 828
    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
829
    else if (arg_match(&arg, &limit, argi))
John Koleszar's avatar
John Koleszar committed
830
      global->limit = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
831
    else if (arg_match(&arg, &skip, argi))
John Koleszar's avatar
John Koleszar committed
832
      global->skip_frames = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
833
    else if (arg_match(&arg, &psnrarg, argi))
John Koleszar's avatar
John Koleszar committed
834
      global->show_psnr = 1;
John Koleszar's avatar
John Koleszar committed
835
    else if (arg_match(&arg, &recontest, argi))
836
      global->test_decode = arg_parse_enum_or_int(&arg);
John Koleszar's avatar
John Koleszar committed
837
    else if (arg_match(&arg, &framerate, argi)) {
John Koleszar's avatar
John Koleszar committed
838 839 840 841 842
      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
843
    else if (arg_match(&arg, &debugmode, argi))
John Koleszar's avatar
John Koleszar committed
844
      global->debug = 1;
John Koleszar's avatar
John Koleszar committed
845
    else if (arg_match(&arg, &q_hist_n, argi))
John Koleszar's avatar
John Koleszar committed
846
      global->show_q_hist_buckets = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
847
    else if (arg_match(&arg, &rate_hist_n, argi))
John Koleszar's avatar
John Koleszar committed
848
      global->show_rate_hist_buckets = arg_parse_uint(&arg);
849 850 851 852
    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
853 854 855
    else
      argj++;
  }
John Koleszar's avatar
John Koleszar committed
856

857 858 859
  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
860 861
      warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
           global->pass);
862 863 864
      global->passes = global->pass;
    }
  }
John Koleszar's avatar
John Koleszar committed
865
  /* Validate global config */
866
  if (global->passes == 0) {
867 868
#if CONFIG_AV1_ENCODER
    // Make default AV1 passes = 2 until there is a better quality 1-pass
869
    // encoder
870
    if (global->codec != NULL && global->codec->name != NULL)
Thomas Daede's avatar
Thomas Daede committed
871
      global->passes = (strcmp(global->codec->name, "av1") == 0) ? 2 : 1;
James Zern's avatar
James Zern committed
872 873 874
#else
    global->passes = 1;
#endif
875
  }
876 877
}

878
static void open_input_file(struct AvxInputContext *input) {
John Koleszar's avatar
John Koleszar committed
879
  /* Parse certain options from the input file, if possible */
clang-format's avatar
clang-format committed
880 881
  input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
                                             : set_binary_mode(stdin);
John Koleszar's avatar
John Koleszar committed
882

clang-format's avatar
clang-format committed
883
  if (!input->file) fatal("Failed to open input file");
John Koleszar's avatar
John Koleszar committed
884

885 886 887 888 889 890 891 892
  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);
  }

893 894 895 896
  /* 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
897 898 899 900 901 902
  /* 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
903
  if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
904 905
    if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
                       input->only_i420) >= 0) {
John Koleszar's avatar
John Koleszar committed
906
      input->file_type = FILE_TYPE_Y4M;
907 908
      input->width = input->y4m.pic_w;
      input->height = input->y4m.pic_h;
909 910
      input->pixel_aspect_ratio.numerator = input->y4m.par_n;
      input->pixel_aspect_ratio.denominator = input->y4m.par_d;
911 912
      input->framerate.numerator = input->y4m.fps_n;
      input->framerate.denominator = input->y4m.fps_d;
913
      input->fmt = input->y4m.aom_fmt;
914
      input->bit_depth = input->y4m.bit_depth;
John Koleszar's avatar
John Koleszar committed
915 916
    } else
      fatal("Unsupported Y4M stream.");
917
  } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
918
    fatal("IVF is not supported as input.");
John Koleszar's avatar
John Koleszar committed
919 920
  } else {
    input->file_type = FILE_TYPE_RAW;
John Koleszar's avatar
John Koleszar committed
921
  }
922 923
}

924
static void close_input_file(struct AvxInputContext *input) {
John Koleszar's avatar
John Koleszar committed
925
  fclose(input->file);
clang-format's avatar
clang-format committed
926
  if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
927 928
}

929
static struct stream_state *new_stream(struct AvxEncoderConfig *global,
930
                                       struct stream_state *prev) {
John Koleszar's avatar
John Koleszar committed
931 932 933
  struct stream_state *stream;

  stream = calloc(1, sizeof(*stream));
934
  if (stream == NULL) {
John Koleszar's avatar
John Koleszar committed
935
    fatal("Failed to allocate new stream.");
936 937
  }

John Koleszar's avatar
John Koleszar committed
938 939 940 941 942
  if (prev) {
    memcpy(stream, prev, sizeof(*stream));
    stream->index++;
    prev->next = stream;
  } else {
943
    aom_codec_err_t res;
John Koleszar's avatar
John Koleszar committed
944 945

    /* Populate encoder configuration */
946
    res = aom_codec_enc_config_default(global->codec->codec_interface(),
clang-format's avatar
clang-format committed
947
                                       &stream->config.cfg, global->usage);
948
    if (res) fatal("Failed to get config: %s\n", aom_codec_err_to_string(res));
John Koleszar's avatar
John Koleszar committed
949

John Koleszar's avatar
John Koleszar committed
950 951 952 953
    /* 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;
954

John Koleszar's avatar
John Koleszar committed
955 956 957 958 959
    /* 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
960

John Koleszar's avatar
John Koleszar committed
961 962
    /* Initialize remaining stream parameters */
    stream->config.write_webm = 1;
963
#if CONFIG_WEBM_IO
964
    stream->config.stereo_fmt = STEREO_FORMAT_MONO;
965 966 967
    stream->webm_ctx.last_pts_ns = -1;
    stream->webm_ctx.writer = NULL;
    stream->webm_ctx.segment = NULL;
968
#endif
John Koleszar's avatar
John Koleszar committed
969

John Koleszar's avatar
John Koleszar committed
970
    /* Allows removal of the application version from the EBML tags */
971
    stream->webm_ctx.debug = global->debug;
John Koleszar's avatar
John Koleszar committed
972
  }
John Koleszar's avatar
John Koleszar committed
973

John Koleszar's avatar
John Koleszar committed
974 975
  /* Output files must be specified for each stream */
  stream->config.out_fn = NULL;
John Koleszar's avatar
John Koleszar committed
976

John Koleszar's avatar
John Koleszar committed
977 978
  stream->next = NULL;
  return stream;
979
}
John Koleszar's avatar
John Koleszar committed
980

981
static int parse_stream_params(struct AvxEncoderConfig *global,
clang-format's avatar
clang-format committed
982 983 984
                               struct stream_state *stream, char **argv) {
  char **argi, **argj;
  struct arg arg;
John Koleszar's avatar
John Koleszar committed
985
  static const arg_def_t **ctrl_args = no_args;
clang-format's avatar
clang-format committed
986 987 988
  static const int *ctrl_args_map = NULL;
  struct stream_config *config = &stream->config;
  int eos_mark_found = 0;
989
  int webm_forced = 0;
John Koleszar's avatar
John Koleszar committed
990

991
  // Handle codec specific options
992
  if (0) {
993 994 995 996 997 998
#if CONFIG_AV1_ENCODER
  } else if (strcmp(global->codec->name, "av1") == 0) {
    // TODO(jingning): Reuse AV1 specific encoder configuration parameters.
    // Consider to expand this set for AV1 encoder control.
    ctrl_args = av1_args;
    ctrl_args_map = av1_arg_ctrl_map;
999
#endif
John Koleszar's avatar
John Koleszar committed
1000
  }
1001

John Koleszar's avatar
John Koleszar committed
1002 1003
  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
    arg.argv_step = 1;
John Koleszar's avatar
John Koleszar committed
1004

John Koleszar's avatar
John Koleszar committed
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
    /* 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
1015

Adrian Grange's avatar
Adrian Grange committed
1016
    if (arg_match(&arg, &outputfile, argi)) {
John Koleszar's avatar
John Koleszar committed
1017
      config->out_fn = arg.val;
1018 1019 1020 1021 1022 1023 1024
      if (!webm_forced) {
        const size_t out_fn_len = strlen(config->out_fn);
        if (out_fn_len >= 4 &&
            !strcmp(config->out_fn + out_fn_len - 4, ".ivf")) {
          config->write_webm = 0;
        }
      }
1025
    } else if (arg_match(&arg, &fpf_name, argi)) {
John Koleszar's avatar
John Koleszar committed
1026
      config->stats_fn = arg.val;
1027 1028 1029
#if CONFIG_FP_MB_STATS
    } else if (arg_match(&arg, &fpmbf_name, argi)) {
      config->fpmb_stats_fn = arg.val;
1030 1031 1032 1033
#endif
    } else if (arg_match(&arg, &use_webm, argi)) {
#if CONFIG_WEBM_IO
      config->write_webm = 1;
1034
      webm_forced = 1;
1035 1036
#else
      die("Error: --webm specified but webm is disabled.");
1037
#endif
1038
    } else if (arg_match(&arg, &use_ivf, argi)) {
John Koleszar's avatar
John Koleszar committed
1039
      config->write_webm = 0;
1040
    } else if (arg_match(&arg, &threads, argi)) {
John Koleszar's avatar
John Koleszar committed