aomenc.c 79.8 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");
258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273
#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

static const arg_def_t *global_args[] = { &use_yv12,
                                          &use_i420,
                                          &use_i422,
                                          &use_i444,
                                          &use_i440,
                                          &usage,
                                          &threads,
                                          &profile,
                                          &width,
                                          &height,
274
#if CONFIG_WEBM_IO
275
                                          &stereo_mode,
276
#endif
277 278 279
                                          &timebase,
                                          &framerate,
                                          &error_resilient,
280
#if CONFIG_HIGHBITDEPTH
281
                                          &bitdeptharg,
282
#endif
283 284 285 286 287
                                          &lag_in_frames,
#if CONFIG_EXT_TILE
                                          &large_scale_tile,
#endif  // CONFIG_EXT_TILE
                                          NULL };
clang-format's avatar
clang-format committed
288 289 290

static const arg_def_t dropframe_thresh =
    ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
291 292 293 294
static const arg_def_t resize_mode =
    ARG_DEF(NULL, "resize-mode", 1, "Frame resize mode");
static const arg_def_t resize_numerator =
    ARG_DEF(NULL, "resize-numerator", 1, "Frame resize numerator");
295 296
static const arg_def_t resize_kf_numerator =
    ARG_DEF(NULL, "resize-kf-numerator", 1, "Frame resize keyframe numerator");
297 298 299 300 301
#if CONFIG_FRAME_SUPERRES
static const arg_def_t superres_mode =
    ARG_DEF(NULL, "superres-mode", 1, "Frame super-resolution mode");
static const arg_def_t superres_numerator =
    ARG_DEF(NULL, "superres-numerator", 1, "Frame super-resolution numerator");
302 303 304
static const arg_def_t superres_kf_numerator =
    ARG_DEF(NULL, "superres-kf-numerator", 1,
            "Frame super-resolution keyframe numerator");
305 306 307 308 309
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");
310
#endif  // CONFIG_FRAME_SUPERRES
311 312 313 314 315
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
316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
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)");
334 335
static const arg_def_t *rc_args[] = { &dropframe_thresh,
                                      &resize_mode,
336
                                      &resize_numerator,
337
                                      &resize_kf_numerator,
338
#if CONFIG_FRAME_SUPERRES
339 340 341
                                      &superres_mode,
                                      &superres_numerator,
                                      &superres_kf_numerator,
342 343
                                      &superres_qthresh,
                                      &superres_kf_qthresh,
344
#endif  // CONFIG_FRAME_SUPERRES
345 346 347 348 349 350 351 352 353 354
                                      &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
355

clang-format's avatar
clang-format committed
356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
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
386
static const struct arg_enum_list tuning_enum[] = {
387 388 389 390 391 392 393
  { "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
394
};
395 396
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
397 398 399 400
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)");
401

402 403
#if CONFIG_AV1_ENCODER
static const arg_def_t cpu_used_av1 =
Yaowu Xu's avatar
Yaowu Xu committed
404
    ARG_DEF(NULL, "cpu-used", 1, "CPU Used (0..8)");
405 406 407 408 409
#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
410 411 412 413 414
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)");
415 416 417 418 419 420
#if CONFIG_MAX_TILE
static const arg_def_t tile_width =
    ARG_DEF(NULL, "tile-width", 1, "Width of each tile");
static const arg_def_t tile_height =
    ARG_DEF(NULL, "tile-height", 1, "Height of each tile");
#endif
421 422 423 424
#if CONFIG_DEPENDENT_HORZTILES
static const arg_def_t tile_dependent_rows =
    ARG_DEF(NULL, "tile-dependent-rows", 1, "Enable dependent Tile rows");
#endif
425
#if CONFIG_LOOPFILTERING_ACROSS_TILES
426 427
static const arg_def_t tile_loopfilter = ARG_DEF(
    NULL, "tile-loopfilter", 1, "Enable loop filter across tile boundary");
428
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
clang-format's avatar
clang-format committed
429 430
static const arg_def_t lossless =
    ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
431 432
#if CONFIG_AOM_QM
static const arg_def_t enable_qm =
433
    ARG_DEF(NULL, "enable-qm", 1,
434 435
            "Enable quantisation matrices (0: false (default), 1: true)");
static const arg_def_t qm_min = ARG_DEF(
436
    NULL, "qm-min", 1, "Min quant matrix flatness (0..15), default is 8");
437
static const arg_def_t qm_max = ARG_DEF(
438
    NULL, "qm-max", 1, "Max quant matrix flatness (0..15), default is 16");
439
#endif
440 441 442 443 444
#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
445 446
static const arg_def_t num_tg = ARG_DEF(
    NULL, "num-tile-groups", 1, "Maximum number of tile groups, default is 1");
447 448
static const arg_def_t mtu_size =
    ARG_DEF(NULL, "mtu-size", 1,
Yaowu Xu's avatar
Yaowu Xu committed
449 450
            "MTU size for a tile group, default is 0 (no MTU targeting), "
            "overrides maximum number of tile groups");
451 452 453 454
#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
455 456 457 458
static const arg_def_t frame_parallel_decoding =
    ARG_DEF(NULL, "frame-parallel", 1,
            "Enable frame parallel decodability features "
            "(0: false (default), 1: true)");
459
#if !CONFIG_EXT_DELTA_Q
460 461 462
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
463
    "3: cyclic refresh, 4: delta quant)");
464
#else
Thomas's avatar
Thomas committed
465 466 467
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
468
    "3: cyclic refresh)");
469
#endif
Fangwen Fu's avatar
Fangwen Fu committed
470 471 472 473 474
#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
475 476 477
static const arg_def_t frame_periodic_boost =
    ARG_DEF(NULL, "frame-boost", 1,
            "Enable frame periodic boost (0: off (default), 1: on)");
478 479
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
480 481
static const arg_def_t max_inter_rate_pct =
    ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
482 483 484 485 486 487
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)");
488

Yaowu Xu's avatar
Yaowu Xu committed
489
static const struct arg_enum_list color_space_enum[] = {
490 491 492 493 494
  { "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
495 496
};

clang-format's avatar
clang-format committed
497
static const arg_def_t input_color_space =
clang-format's avatar
clang-format committed
498 499
    ARG_DEF_ENUM(NULL, "color-space", 1,
                 "The color space of input content:", color_space_enum);
Yaowu Xu's avatar
Yaowu Xu committed
500

501 502 503 504 505 506 507 508 509
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
510 511
    NULL, "transfer-function", 1,
    "The transfer function of input content:", transfer_function_enum);
512 513 514 515 516 517 518 519 520 521 522 523 524

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);

525
static const struct arg_enum_list tune_content_enum[] = {
526 527
  { "default", AOM_CONTENT_DEFAULT },
  { "screen", AOM_CONTENT_SCREEN },
clang-format's avatar
clang-format committed
528
  { NULL, 0 }
529 530 531 532
};

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
533
#endif
534

535
#if CONFIG_AV1_ENCODER
536 537
#if CONFIG_EXT_PARTITION
static const struct arg_enum_list superblock_size_enum[] = {
538 539 540
  { "dynamic", AOM_SUPERBLOCK_SIZE_DYNAMIC },
  { "64", AOM_SUPERBLOCK_SIZE_64X64 },
  { "128", AOM_SUPERBLOCK_SIZE_128X128 },
clang-format's avatar
clang-format committed
541
  { NULL, 0 }
542 543 544 545 546
};
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

547 548 549 550
static const arg_def_t *av1_args[] = { &cpu_used_av1,
                                       &auto_altref,
                                       &sharpness,
                                       &static_thresh,
551 552 553
#if CONFIG_EXT_TILE
                                       &single_tile_decoding,
#endif  // CONFIG_EXT_TILE
554 555
                                       &tile_cols,
                                       &tile_rows,
556 557 558 559
#if CONFIG_MAX_TILE
                                       &tile_width,
                                       &tile_height,
#endif
560 561 562
#if CONFIG_DEPENDENT_HORZTILES
                                       &tile_dependent_rows,
#endif
563
#if CONFIG_LOOPFILTERING_ACROSS_TILES
564
                                       &tile_loopfilter,
565
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
566 567
                                       &arnr_maxframes,
                                       &arnr_strength,
568
                                       &tune_metric,
569 570 571 572 573
                                       &cq_level,
                                       &max_intra_rate_pct,
                                       &max_inter_rate_pct,
                                       &gf_cbr_boost_pct,
                                       &lossless,
574 575 576 577
#if CONFIG_AOM_QM
                                       &enable_qm,
                                       &qm_min,
                                       &qm_max,
578 579 580
#endif
#if CONFIG_DIST_8X8
                                       &enable_dist_8x8,
581
#endif
582 583
                                       &frame_parallel_decoding,
                                       &aq_mode,
Fangwen Fu's avatar
Fangwen Fu committed
584 585 586
#if CONFIG_EXT_DELTA_Q
                                       &deltaq_mode,
#endif
587 588 589 590
                                       &frame_periodic_boost,
                                       &noise_sens,
                                       &tune_content,
                                       &input_color_space,
591 592
                                       &input_transfer_function,
                                       &input_chroma_sample_position,
593 594
                                       &min_gf_interval,
                                       &max_gf_interval,
595
#if CONFIG_EXT_PARTITION
596
                                       &superblock_size,
597
#endif  // CONFIG_EXT_PARTITION
598 599
                                       &num_tg,
                                       &mtu_size,
600 601 602
#if CONFIG_TEMPMV_SIGNALING
                                       &disable_tempmv,
#endif
603
#if CONFIG_HIGHBITDEPTH
604 605
                                       &bitdeptharg,
                                       &inbitdeptharg,
606
#endif  // CONFIG_HIGHBITDEPTH
607 608 609 610 611
                                       NULL };
static const int av1_arg_ctrl_map[] = { AOME_SET_CPUUSED,
                                        AOME_SET_ENABLEAUTOALTREF,
                                        AOME_SET_SHARPNESS,
                                        AOME_SET_STATIC_THRESHOLD,
612 613 614
#if CONFIG_EXT_TILE
                                        AV1E_SET_SINGLE_TILE_DECODING,
#endif  // CONFIG_EXT_TILE
615 616
                                        AV1E_SET_TILE_COLUMNS,
                                        AV1E_SET_TILE_ROWS,
617 618 619 620
#if CONFIG_MAX_TILE
                                        AV1E_SET_TILE_WIDTH,
                                        AV1E_SET_TILE_HEIGHT,
#endif
621 622 623
#if CONFIG_DEPENDENT_HORZTILES
                                        AV1E_SET_TILE_DEPENDENT_ROWS,
#endif
624
#if CONFIG_LOOPFILTERING_ACROSS_TILES
625
                                        AV1E_SET_TILE_LOOPFILTER,
626
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
627 628 629 630 631 632 633 634
                                        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,
635 636 637 638
#if CONFIG_AOM_QM
                                        AV1E_SET_ENABLE_QM,
                                        AV1E_SET_QM_MIN,
                                        AV1E_SET_QM_MAX,
639 640 641
#endif
#if CONFIG_DIST_8X8
                                        AV1E_SET_ENABLE_DIST_8X8,
642
#endif
643 644
                                        AV1E_SET_FRAME_PARALLEL_DECODING,
                                        AV1E_SET_AQ_MODE,
Fangwen Fu's avatar
Fangwen Fu committed
645 646 647
#if CONFIG_EXT_DELTA_Q
                                        AV1E_SET_DELTAQ_MODE,
#endif
648 649 650 651
                                        AV1E_SET_FRAME_PERIODIC_BOOST,
                                        AV1E_SET_NOISE_SENSITIVITY,
                                        AV1E_SET_TUNE_CONTENT,
                                        AV1E_SET_COLOR_SPACE,
652 653
                                        AV1E_SET_TRANSFER_FUNCTION,
                                        AV1E_SET_CHROMA_SAMPLE_POSITION,
654 655
                                        AV1E_SET_MIN_GF_INTERVAL,
                                        AV1E_SET_MAX_GF_INTERVAL,
656
#if CONFIG_EXT_PARTITION
657
                                        AV1E_SET_SUPERBLOCK_SIZE,
658
#endif  // CONFIG_EXT_PARTITION
659 660
                                        AV1E_SET_NUM_TG,
                                        AV1E_SET_MTU,
661 662
#if CONFIG_TEMPMV_SIGNALING
                                        AV1E_SET_DISABLE_TEMPMV,
663
#endif
664
                                        0 };
hui su's avatar
hui su committed
665 666
#endif

John Koleszar's avatar
John Koleszar committed
667 668
static const arg_def_t *no_args[] = { NULL };

669
void usage_exit(void) {
John Koleszar's avatar
John Koleszar committed
670
  int i;
671
  const int num_encoder = get_aom_encoder_count();
John Koleszar's avatar
John Koleszar committed
672 673 674 675 676

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

  fprintf(stderr, "\nOptions:\n");
John Koleszar's avatar
John Koleszar committed
677
  arg_show_usage(stderr, main_args);
John Koleszar's avatar
John Koleszar committed
678
  fprintf(stderr, "\nEncoder Global Options:\n");
John Koleszar's avatar
John Koleszar committed
679
  arg_show_usage(stderr, global_args);
John Koleszar's avatar
John Koleszar committed
680
  fprintf(stderr, "\nRate Control Options:\n");
John Koleszar's avatar
John Koleszar committed
681
  arg_show_usage(stderr, rc_args);
John Koleszar's avatar
John Koleszar committed
682
  fprintf(stderr, "\nTwopass Rate Control Options:\n");
John Koleszar's avatar
John Koleszar committed
683
  arg_show_usage(stderr, rc_twopass_args);
John Koleszar's avatar
John Koleszar committed
684
  fprintf(stderr, "\nKeyframe Placement Options:\n");
John Koleszar's avatar
John Koleszar committed
685
  arg_show_usage(stderr, kf_args);
686 687 688
#if CONFIG_AV1_ENCODER
  fprintf(stderr, "\nAV1 Specific Options:\n");
  arg_show_usage(stderr, av1_args);
John Koleszar's avatar
John Koleszar committed
689
#endif
clang-format's avatar
clang-format committed
690 691
  fprintf(stderr,
          "\nStream timebase (--timebase):\n"
John Koleszar's avatar
John Koleszar committed
692 693
          "  The desired precision of timestamps in the output, expressed\n"
          "  in fractional seconds. Default is 1/1000.\n");
694
  fprintf(stderr, "\nIncluded encoders:\n\n");
John Koleszar's avatar
John Koleszar committed
695

696
  for (i = 0; i < num_encoder; ++i) {
697
    const AvxInterface *const encoder = get_aom_encoder_by_index(i);
clang-format's avatar
clang-format committed
698 699
    const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
    fprintf(stderr, "    %-6s - %s %s\n", encoder->name,
700
            aom_codec_iface_name(encoder->codec_interface()), defstr);
701
  }
702 703
  fprintf(stderr, "\n        ");
  fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
John Koleszar's avatar
John Koleszar committed
704 705

  exit(EXIT_FAILURE);
John Koleszar's avatar
John Koleszar committed
706 707
}

708 709
#if CONFIG_AV1_ENCODER
#define ARG_CTRL_CNT_MAX NELEMENTS(av1_arg_ctrl_map)
710
#endif
John Koleszar's avatar
John Koleszar committed
711

712 713
#if !CONFIG_WEBM_IO
typedef int stereo_format_t;
clang-format's avatar
clang-format committed
714 715 716
struct WebmOutputContext {
  int debug;
};
717 718
#endif

719
/* Per-stream configuration */
John Koleszar's avatar
John Koleszar committed
720
struct stream_config {
721
  struct aom_codec_enc_cfg cfg;
clang-format's avatar
clang-format committed
722 723
  const char *out_fn;
  const char *stats_fn;
724
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
725
  const char *fpmb_stats_fn;
726
#endif
clang-format's avatar
clang-format committed
727 728 729 730
  stereo_format_t stereo_fmt;
  int arg_ctrls[ARG_CTRL_CNT_MAX][2];
  int arg_ctrl_cnt;
  int write_webm;
731
  // whether to use 16bit internal buffers
clang-format's avatar
clang-format committed
732
  int use_16bit_internal;
733 734
};

John Koleszar's avatar
John Koleszar committed
735
struct stream_state {
clang-format's avatar
clang-format committed
736 737 738 739 740 741 742 743 744 745 746
  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];
747
  aom_codec_ctx_t encoder;
clang-format's avatar
clang-format committed
748 749 750 751
  unsigned int frames_out;
  uint64_t cx_time;
  size_t nbytes;
  stats_io_t stats;
752
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
753
  stats_io_t fpmb_stats;
754
#endif
755 756
  struct aom_image *img;
  aom_codec_ctx_t decoder;
clang-format's avatar
clang-format committed
757
  int mismatch_seen;
758 759
};

clang-format's avatar
clang-format committed
760
static void validate_positive_rational(const char *msg,
761
                                       struct aom_rational *rat) {
John Koleszar's avatar
John Koleszar committed
762 763 764 765
  if (rat->den < 0) {
    rat->num *= -1;
    rat->den *= -1;
  }
766

clang-format's avatar
clang-format committed
767
  if (rat->num < 0) die("Error: %s must be positive\n", msg);
768

clang-format's avatar
clang-format committed
769
  if (!rat->den) die("Error: %s has zero denominator\n", msg);
770 771
}

772
static void parse_global_config(struct AvxEncoderConfig *global, char **argv) {
clang-format's avatar
clang-format committed
773 774
  char **argi, **argj;
  struct arg arg;
775
  const int num_encoder = get_aom_encoder_count();
776

clang-format's avatar
clang-format committed
777
  if (num_encoder < 1) die("Error: no valid encoder available\n");
John Koleszar's avatar
John Koleszar committed
778 779 780

  /* Initialize default parameters */
  memset(global, 0, sizeof(*global));
781
  global->codec = get_aom_encoder_by_index(num_encoder - 1);
782
  global->passes = 0;
783
  global->color_type = I420;
784
  /* Assign default deadline to good quality */
785
  global->deadline = AOM_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
786 787 788 789 790

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

    if (arg_match(&arg, &codecarg, argi)) {
791
      global->codec = get_aom_encoder_by_name(arg.val);
792 793
      if (!global->codec)
        die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
John Koleszar's avatar
John Koleszar committed
794
    } else if (arg_match(&arg, &passes, argi)) {
John Koleszar's avatar
John Koleszar committed
795
      global->passes = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
796

John Koleszar's avatar
John Koleszar committed
797 798
      if (global->passes < 1 || global->passes > 2)
        die("Error: Invalid number of passes (%d)\n", global->passes);
John Koleszar's avatar
John Koleszar committed
799
    } else if (arg_match(&arg, &pass_arg, argi)) {
John Koleszar's avatar
John Koleszar committed
800 801 802
      global->pass = arg_parse_uint(&arg);

      if (global->pass < 1 || global->pass > 2)
clang-format's avatar
clang-format committed
803
        die("Error: Invalid pass selected (%d)\n", global->pass);
John Koleszar's avatar
John Koleszar committed
804 805
    } else if (arg_match(&arg, &usage, argi))
      global->usage = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
806
    else if (arg_match(&arg, &deadline, argi))
John Koleszar's avatar
John Koleszar committed
807
      global->deadline = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
808
    else if (arg_match(&arg, &good_dl, argi))
809
      global->deadline = AOM_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
810
    else if (arg_match(&arg, &use_yv12, argi))
811
      global->color_type = YV12;
John Koleszar's avatar
John Koleszar committed
812
    else if (arg_match(&arg, &use_i420, argi))
813 814 815 816 817
      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
818 819
    else if (arg_match(&arg, &use_i440, argi))
      global->color_type = I440;
John Koleszar's avatar
John Koleszar committed
820 821 822 823
    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
824
    else if (arg_match(&arg, &limit, argi))
John Koleszar's avatar
John Koleszar committed
825
      global->limit = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
826
    else if (arg_match(&arg, &skip, argi))
John Koleszar's avatar
John Koleszar committed
827
      global->skip_frames = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
828
    else if (arg_match(&arg, &psnrarg, argi))
John Koleszar's avatar
John Koleszar committed
829
      global->show_psnr = 1;
John Koleszar's avatar
John Koleszar committed
830
    else if (arg_match(&arg, &recontest, argi))
831
      global->test_decode = arg_parse_enum_or_int(&arg);
John Koleszar's avatar
John Koleszar committed
832
    else if (arg_match(&arg, &framerate, argi)) {
John Koleszar's avatar
John Koleszar committed
833 834 835 836 837
      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
838
    else if (arg_match(&arg, &debugmode, argi))
John Koleszar's avatar
John Koleszar committed
839
      global->debug = 1;
John Koleszar's avatar
John Koleszar committed
840
    else if (arg_match(&arg, &q_hist_n, argi))
John Koleszar's avatar
John Koleszar committed
841
      global->show_q_hist_buckets = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
842
    else if (arg_match(&arg, &rate_hist_n, argi))
John Koleszar's avatar
John Koleszar committed
843
      global->show_rate_hist_buckets = arg_parse_uint(&arg);
844 845 846 847
    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
848 849 850
    else
      argj++;
  }
John Koleszar's avatar
John Koleszar committed
851

852 853 854
  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
855 856
      warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
           global->pass);
857 858 859
      global->passes = global->pass;
    }
  }
John Koleszar's avatar
John Koleszar committed
860
  /* Validate global config */
861
  if (global->passes == 0) {
862 863
#if CONFIG_AV1_ENCODER
    // Make default AV1 passes = 2 until there is a better quality 1-pass
864
    // encoder
865
    if (global->codec != NULL && global->codec->name != NULL)
Thomas Daede's avatar
Thomas Daede committed
866
      global->passes = (strcmp(global->codec->name, "av1") == 0) ? 2 : 1;
James Zern's avatar
James Zern committed
867 868 869
#else
    global->passes = 1;
#endif
870
  }
871 872
}

873
static void open_input_file(struct AvxInputContext *input) {
John Koleszar's avatar
John Koleszar committed
874
  /* Parse certain options from the input file, if possible */
clang-format's avatar
clang-format committed
875 876
  input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
                                             : set_binary_mode(stdin);
John Koleszar's avatar
John Koleszar committed
877

clang-format's avatar
clang-format committed
878
  if (!input->file) fatal("Failed to open input file");
John Koleszar's avatar
John Koleszar committed
879

880 881 882 883 884 885 886 887
  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);
  }

888 889 890 891
  /* 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
892 893 894 895 896 897
  /* 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
898
  if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
899 900
    if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
                       input->only_i420) >= 0) {
John Koleszar's avatar
John Koleszar committed
901
      input->file_type = FILE_TYPE_Y4M;
902 903
      input->width = input->y4m.pic_w;
      input->height = input->y4m.pic_h;
904 905
      input->pixel_aspect_ratio.numerator = input->y4m.par_n;
      input->pixel_aspect_ratio.denominator = input->y4m.par_d;
906 907
      input->framerate.numerator = input->y4m.fps_n;
      input->framerate.denominator = input->y4m.fps_d;
908
      input->fmt = input->y4m.aom_fmt;
909
      input->bit_depth = input->y4m.bit_depth;
John Koleszar's avatar
John Koleszar committed
910 911
    } else
      fatal("Unsupported Y4M stream.");
912
  } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
913
    fatal("IVF is not supported as input.");
John Koleszar's avatar
John Koleszar committed
914 915
  } else {
    input->file_type = FILE_TYPE_RAW;
John Koleszar's avatar
John Koleszar committed
916
  }
917 918
}

919
static void close_input_file(struct AvxInputContext *input) {
John Koleszar's avatar
John Koleszar committed
920
  fclose(input->file);
clang-format's avatar
clang-format committed
921
  if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
922 923
}

924
static struct stream_state *new_stream(struct AvxEncoderConfig *global,
925
                                       struct stream_state *prev) {
John Koleszar's avatar
John Koleszar committed
926 927 928
  struct stream_state *stream;

  stream = calloc(1, sizeof(*stream));
929
  if (stream == NULL) {
John Koleszar's avatar
John Koleszar committed
930
    fatal("Failed to allocate new stream.");
931 932
  }

John Koleszar's avatar
John Koleszar committed
933 934 935 936 937
  if (prev) {
    memcpy(stream, prev, sizeof(*stream));
    stream->index++;
    prev->next = stream;
  } else {
938
    aom_codec_err_t res;
John Koleszar's avatar
John Koleszar committed
939 940

    /* Populate encoder configuration */
941
    res = aom_codec_enc_config_default(global->codec->codec_interface(),
clang-format's avatar
clang-format committed
942
                                       &stream->config.cfg, global->usage);
943
    if (res) fatal("Failed to get config: %s\n", aom_codec_err_to_string(res));
John Koleszar's avatar
John Koleszar committed
944

John Koleszar's avatar
John Koleszar committed
945 946 947 948
    /* 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;
949

John Koleszar's avatar
John Koleszar committed
950 951 952 953 954
    /* 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
955

John Koleszar's avatar
John Koleszar committed
956 957
    /* Initialize remaining stream parameters */
    stream->config.write_webm = 1;
958
#if CONFIG_WEBM_IO
959
    stream->config.stereo_fmt = STEREO_FORMAT_MONO;
960 961 962
    stream->webm_ctx.last_pts_ns = -1;
    stream->webm_ctx.writer = NULL;
    stream->webm_ctx.segment = NULL;
963
#endif
John Koleszar's avatar
John Koleszar committed
964

John Koleszar's avatar
John Koleszar committed
965
    /* Allows removal of the application version from the EBML tags */
966
    stream->webm_ctx.debug = global->debug;
John Koleszar's avatar
John Koleszar committed
967
  }
John Koleszar's avatar
John Koleszar committed
968

John Koleszar's avatar
John Koleszar committed
969 970
  /* Output files must be specified for each stream */
  stream->config.out_fn = NULL;
John Koleszar's avatar
John Koleszar committed
971

John Koleszar's avatar
John Koleszar committed
972 973
  stream->next = NULL;
  return stream;
974
}
John Koleszar's avatar
John Koleszar committed
975

976
static int parse_stream_params(struct AvxEncoderConfig *global,
clang-format's avatar
clang-format committed
977 978 979
                               struct stream_state *stream, char **argv) {
  char **argi, **argj;
  struct arg arg;
John Koleszar's avatar
John Koleszar committed
980
  static const arg_def_t **ctrl_args = no_args;
clang-format's avatar
clang-format committed
981 982 983
  static const int *ctrl_args_map = NULL;
  struct stream_config *config = &stream->config;
  int eos_mark_found = 0;
984
  int webm_forced = 0;
John Koleszar's avatar
John Koleszar committed
985

986
  // Handle codec specific options
987
  if (0) {
988 989 990 991 992 993
#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;
994
#endif
John Koleszar's avatar
John Koleszar committed
995
  }
996

John Koleszar's avatar
John Koleszar committed
997 998
  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
    arg.argv_step = 1;
John Koleszar's avatar
John Koleszar committed
999

John Koleszar's avatar
John Koleszar committed
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
    /* 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
1010

Adrian Grange's avatar
Adrian Grange committed
1011
    if (arg_match(&arg, &outputfile, argi)) {
John Koleszar's avatar
John Koleszar committed
1012
      config->out_fn = arg.val;
1013 1014 1015 1016 1017 1018 1019
      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;
        }
      }
1020
    } else if (arg_match(&arg, &fpf_name, argi)) {
John Koleszar's avatar
John Koleszar committed
1021
      config->stats_fn = arg.val;
1022 1023 1024
#if CONFIG_FP_MB_STATS
    } else if (arg_match(&arg, &fpmbf_name, argi)) {
      config->fpmb_stats_fn = arg.val;
1025 1026 1027 1028
#endif
    } else if (arg_match(&arg, &use_webm, argi)) {
#if CONFIG_WEBM_IO
      config->write_webm = 1;
1029
      webm_forced = 1;
1030 1031
#else
      die("Error: --webm specified but webm is disabled.");
1032
#endif
1033
    } else if (arg_match(&arg, &use_ivf, argi)) {
John Koleszar's avatar
John Koleszar committed
1034
      config->write_webm = 0;
1035
    } else if (arg_match(&arg, &threads, argi)) {
John Koleszar's avatar
John Koleszar committed
1036
      config->cfg.g_threads = arg_parse_uint(&arg);
1037
    } else if (arg_match(&arg, &profile, argi)) {
John Koleszar's avatar
John Koleszar committed
1038
      config->cfg.g_profile = arg_parse_uint(&arg);
1039
    } else if (arg_match(&arg, &width, argi)) {
John Koleszar's avatar
John Koleszar committed
1040
      config->cfg.g_w = arg_parse_uint(&arg);
1041
    } else if (arg_match(&arg, &height, argi)) {
John Koleszar's avatar
John Koleszar committed
1042
      config->cfg.g_h = arg_parse_uint(&arg);
1043
#if CONFIG_HIGHBITDEPTH
1044