aomenc.c 83.7 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_DECODERS
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

36 37
#if CONFIG_AV1_ENCODER
#include "aom/aomcx.h"
John Koleszar's avatar
John Koleszar committed
38
#endif
39 40
#if CONFIG_AV1_DECODER
#include "aom/aomdx.h"
John Koleszar's avatar
John Koleszar committed
41 42
#endif

43
#include "./aomstats.h"
44
#include "./rate_hist.h"
45
#include "./warnings.h"
46 47 48
#include "aom/aom_integer.h"
#include "aom_ports/aom_timer.h"
#include "aom_ports/mem_ops.h"
49
#if CONFIG_WEBM_IO
50
#include "./webmenc.h"
51
#endif
52
#include "./y4minput.h"
53

John Koleszar's avatar
John Koleszar committed
54
/* Swallow warnings about unused results of fread/fwrite */
clang-format's avatar
clang-format committed
55
static size_t wrap_fread(void *ptr, size_t size, size_t nmemb, FILE *stream) {
John Koleszar's avatar
John Koleszar committed
56
  return fread(ptr, size, nmemb, stream);
John Koleszar's avatar
John Koleszar committed
57 58 59 60
}
#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
61 62
                          FILE *stream) {
  return fwrite(ptr, size, nmemb, stream);
John Koleszar's avatar
John Koleszar committed
63 64 65
}
#define fwrite wrap_fwrite

John Koleszar's avatar
John Koleszar committed
66 67
static const char *exec_name;

68
static void warn_or_exit_on_errorv(aom_codec_ctx_t *ctx, int fatal,
69
                                   const char *s, va_list ap) {
John Koleszar's avatar
John Koleszar committed
70
  if (ctx->err) {
71
    const char *detail = aom_codec_error_detail(ctx);
John Koleszar's avatar
John Koleszar committed
72

John Koleszar's avatar
John Koleszar committed
73
    vfprintf(stderr, s, ap);
74
    fprintf(stderr, ": %s\n", aom_codec_error(ctx));
John Koleszar's avatar
John Koleszar committed
75

clang-format's avatar
clang-format committed
76
    if (detail) fprintf(stderr, "    %s\n", detail);
John Koleszar's avatar
John Koleszar committed
77

clang-format's avatar
clang-format committed
78
    if (fatal) exit(EXIT_FAILURE);
John Koleszar's avatar
John Koleszar committed
79
  }
John Koleszar's avatar
John Koleszar committed
80 81
}

82
static void ctx_exit_on_error(aom_codec_ctx_t *ctx, const char *s, ...) {
83 84 85 86 87 88 89
  va_list ap;

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

90
static void warn_or_exit_on_error(aom_codec_ctx_t *ctx, int fatal,
91 92 93 94 95 96 97 98
                                  const char *s, ...) {
  va_list ap;

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

99
static int read_frame(struct AvxInputContext *input_ctx, aom_image_t *img) {
100 101
  FILE *f = input_ctx->file;
  y4m_input *y4m = &input_ctx->y4m;
John Koleszar's avatar
John Koleszar committed
102 103
  int shortread = 0;

104
  if (input_ctx->file_type == FILE_TYPE_Y4M) {
clang-format's avatar
clang-format committed
105
    if (y4m_input_fetch_frame(y4m, f, img) < 1) return 0;
John Koleszar's avatar
John Koleszar committed
106
  } else {
107
    shortread = read_yuv_frame(input_ctx, img);
John Koleszar's avatar
John Koleszar committed
108
  }
John Koleszar's avatar
John Koleszar committed
109

John Koleszar's avatar
John Koleszar committed
110
  return !shortread;
John Koleszar's avatar
John Koleszar committed
111 112
}

113
static int file_is_y4m(const char detect[4]) {
John Koleszar's avatar
John Koleszar committed
114 115 116 117
  if (memcmp(detect, "YUV4", 4) == 0) {
    return 1;
  }
  return 0;
118 119
}

120
static int fourcc_is_ivf(const char detect[4]) {
121 122 123 124 125
  if (memcmp(detect, "DKIF", 4) == 0) {
    return 1;
  }
  return 0;
}
John Koleszar's avatar
John Koleszar committed
126

clang-format's avatar
clang-format committed
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
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");
148
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
149 150
static const arg_def_t fpmbf_name =
    ARG_DEF(NULL, "fpmbf", 1, "First pass block statistics file name");
151
#endif
clang-format's avatar
clang-format committed
152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
static const arg_def_t limit =
    ARG_DEF(NULL, "limit", 1, "Stop encoding after n input frames");
static const arg_def_t skip =
    ARG_DEF(NULL, "skip", 1, "Skip the first n input frames");
static const arg_def_t deadline =
    ARG_DEF("d", "deadline", 1, "Deadline per frame (usec)");
static const arg_def_t best_dl =
    ARG_DEF(NULL, "best", 0, "Use Best Quality Deadline");
static const arg_def_t good_dl =
    ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
static const arg_def_t rt_dl =
    ARG_DEF(NULL, "rt", 0, "Use Realtime Quality Deadline");
static const arg_def_t quietarg =
    ARG_DEF("q", "quiet", 0, "Do not print encode progress");
static const arg_def_t verbosearg =
    ARG_DEF("v", "verbose", 0, "Show encoder parameters");
static const arg_def_t psnrarg =
    ARG_DEF(NULL, "psnr", 0, "Show PSNR in status line");
170

171
static const struct arg_enum_list test_decode_enum[] = {
clang-format's avatar
clang-format committed
172 173 174 175
  { "off", TEST_DECODE_OFF },
  { "fatal", TEST_DECODE_FATAL },
  { "warn", TEST_DECODE_WARN },
  { NULL, 0 }
176
};
Deb Mukherjee's avatar
Deb Mukherjee committed
177 178
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
179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196
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.");
197

198
#if CONFIG_AOM_HIGHBITDEPTH
199 200
static const arg_def_t test16bitinternalarg = ARG_DEF(
    NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
201 202 203 204 205 206 207 208 209 210 211

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");
212
#endif
213

clang-format's avatar
clang-format committed
214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
static const arg_def_t *main_args[] = { &debugmode,
                                        &outputfile,
                                        &codecarg,
                                        &passes,
                                        &pass_arg,
                                        &fpf_name,
                                        &limit,
                                        &skip,
                                        &deadline,
                                        &best_dl,
                                        &good_dl,
                                        &rt_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
245 246
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");
247
#if CONFIG_WEBM_IO
248
static const struct arg_enum_list stereo_mode_enum[] = {
clang-format's avatar
clang-format committed
249 250 251 252 253 254
  { "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 }
255
};
Deb Mukherjee's avatar
Deb Mukherjee committed
256 257
static const arg_def_t stereo_mode = ARG_DEF_ENUM(
    NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
258
#endif
Deb Mukherjee's avatar
Deb Mukherjee committed
259 260
static const arg_def_t timebase = ARG_DEF(
    NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
clang-format's avatar
clang-format committed
261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
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");

static const arg_def_t *global_args[] = { &use_yv12,
                                          &use_i420,
                                          &use_i422,
                                          &use_i444,
                                          &use_i440,
                                          &usage,
                                          &threads,
                                          &profile,
                                          &width,
                                          &height,
276
#if CONFIG_WEBM_IO
clang-format's avatar
clang-format committed
277
                                          &stereo_mode,
278
#endif
clang-format's avatar
clang-format committed
279 280 281
                                          &timebase,
                                          &framerate,
                                          &error_resilient,
282
#if CONFIG_AOM_HIGHBITDEPTH
clang-format's avatar
clang-format committed
283
                                          &test16bitinternalarg,
284
                                          &bitdeptharg,
285
#endif
clang-format's avatar
clang-format committed
286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
                                          &lag_in_frames,
                                          NULL };

static const arg_def_t dropframe_thresh =
    ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
static const arg_def_t resize_allowed =
    ARG_DEF(NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
static const arg_def_t resize_width =
    ARG_DEF(NULL, "resize-width", 1, "Width of encoded frame");
static const arg_def_t resize_height =
    ARG_DEF(NULL, "resize-height", 1, "Height of encoded frame");
static const arg_def_t resize_up_thresh =
    ARG_DEF(NULL, "resize-up", 1, "Upscale threshold (buf %)");
static const arg_def_t resize_down_thresh =
    ARG_DEF(NULL, "resize-down", 1, "Downscale threshold (buf %)");
301 302 303 304
static const struct arg_enum_list end_usage_enum[] = { { "vbr", AOM_VBR },
                                                       { "cbr", AOM_CBR },
                                                       { "cq", AOM_CQ },
                                                       { "q", AOM_Q },
clang-format's avatar
clang-format committed
305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323
                                                       { NULL, 0 } };
static const arg_def_t end_usage =
    ARG_DEF_ENUM(NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
static const arg_def_t target_bitrate =
    ARG_DEF(NULL, "target-bitrate", 1, "Bitrate (kbps)");
static const arg_def_t min_quantizer =
    ARG_DEF(NULL, "min-q", 1, "Minimum (best) quantizer");
static const arg_def_t max_quantizer =
    ARG_DEF(NULL, "max-q", 1, "Maximum (worst) quantizer");
static const arg_def_t undershoot_pct =
    ARG_DEF(NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
static const arg_def_t overshoot_pct =
    ARG_DEF(NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
static const arg_def_t buf_sz =
    ARG_DEF(NULL, "buf-sz", 1, "Client buffer size (ms)");
static const arg_def_t buf_initial_sz =
    ARG_DEF(NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
static const arg_def_t buf_optimal_sz =
    ARG_DEF(NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
John Koleszar's avatar
John Koleszar committed
324
static const arg_def_t *rc_args[] = {
clang-format's avatar
clang-format committed
325 326 327 328
  &dropframe_thresh, &resize_allowed,     &resize_width,   &resize_height,
  &resize_up_thresh, &resize_down_thresh, &end_usage,      &target_bitrate,
  &min_quantizer,    &max_quantizer,      &undershoot_pct, &overshoot_pct,
  &buf_sz,           &buf_initial_sz,     &buf_optimal_sz, NULL
John Koleszar's avatar
John Koleszar committed
329 330
};

clang-format's avatar
clang-format committed
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
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)");
static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1, "AltRef type");
John Koleszar's avatar
John Koleszar committed
362
static const struct arg_enum_list tuning_enum[] = {
363
  { "psnr", AOM_TUNE_PSNR }, { "ssim", AOM_TUNE_SSIM }, { NULL, 0 }
John Koleszar's avatar
John Koleszar committed
364
};
clang-format's avatar
clang-format committed
365 366 367 368 369 370
static const arg_def_t tune_ssim =
    ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
static const arg_def_t cq_level =
    ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
static const arg_def_t max_intra_rate_pct =
    ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
371

372 373
#if CONFIG_AV1_ENCODER
static const arg_def_t cpu_used_av1 =
clang-format's avatar
clang-format committed
374 375 376 377 378 379 380 381
    ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-8..8)");
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)");
static const arg_def_t lossless =
    ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
382 383 384 385 386 387 388 389 390
#if CONFIG_AOM_QM
static const arg_def_t enable_qm =
    ARG_DEF(NULL, "enable_qm", 1,
            "Enable quantisation matrices (0: false (default), 1: true)");
static const arg_def_t qm_min = ARG_DEF(
    NULL, "qm_min", 1, "Min quant matrix flatness (0..15), default is 8");
static const arg_def_t qm_max = ARG_DEF(
    NULL, "qm_max", 1, "Max quant matrix flatness (0..15), default is 16");
#endif
391 392 393 394 395 396 397
#if CONFIG_TILE_GROUPS
static const arg_def_t num_tg =
    ARG_DEF(NULL, "num-tile-groups", 1, "Maximum number of tile groups, default is 1");
static const arg_def_t mtu_size =
    ARG_DEF(NULL, "mtu-size", 1,
            "MTU size for a tile group, default is 0 (no MTU targeting), overrides maximum number of tile groups");
#endif
Yaowu Xu's avatar
Yaowu Xu committed
398 399 400 401
static const arg_def_t frame_parallel_decoding =
    ARG_DEF(NULL, "frame-parallel", 1,
            "Enable frame parallel decodability features "
            "(0: false (default), 1: true)");
Thomas's avatar
Thomas committed
402
#if CONFIG_DELTA_Q
403 404 405
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
406
    "3: cyclic refresh, 4: delta quant)");
407
#else
Thomas's avatar
Thomas committed
408 409 410
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
411
    "3: cyclic refresh)");
412
#endif
clang-format's avatar
clang-format committed
413 414 415
static const arg_def_t frame_periodic_boost =
    ARG_DEF(NULL, "frame-boost", 1,
            "Enable frame periodic boost (0: off (default), 1: on)");
416 417
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
418 419
static const arg_def_t max_inter_rate_pct =
    ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
420 421 422 423 424 425
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)");
426

Yaowu Xu's avatar
Yaowu Xu committed
427
static const struct arg_enum_list color_space_enum[] = {
428 429 430 431 432 433 434 435
  { "unknown", AOM_CS_UNKNOWN },
  { "bt601", AOM_CS_BT_601 },
  { "bt709", AOM_CS_BT_709 },
  { "smpte170", AOM_CS_SMPTE_170 },
  { "smpte240", AOM_CS_SMPTE_240 },
  { "bt2020", AOM_CS_BT_2020 },
  { "reserved", AOM_CS_RESERVED },
  { "sRGB", AOM_CS_SRGB },
Yaowu Xu's avatar
Yaowu Xu committed
436 437 438
  { NULL, 0 }
};

clang-format's avatar
clang-format committed
439 440 441
static const arg_def_t input_color_space =
    ARG_DEF_ENUM(NULL, "color-space", 1, "The color space of input content:",
                 color_space_enum);
Yaowu Xu's avatar
Yaowu Xu committed
442

443
static const struct arg_enum_list tune_content_enum[] = {
444 445
  { "default", AOM_CONTENT_DEFAULT },
  { "screen", AOM_CONTENT_SCREEN },
clang-format's avatar
clang-format committed
446
  { NULL, 0 }
447 448 449 450
};

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
451
#endif
452

453
#if CONFIG_AV1_ENCODER
454 455
#if CONFIG_EXT_PARTITION
static const struct arg_enum_list superblock_size_enum[] = {
456 457 458
  { "dynamic", AOM_SUPERBLOCK_SIZE_DYNAMIC },
  { "64", AOM_SUPERBLOCK_SIZE_64X64 },
  { "128", AOM_SUPERBLOCK_SIZE_128X128 },
clang-format's avatar
clang-format committed
459
  { NULL, 0 }
460 461 462 463 464
};
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

465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
static const arg_def_t *av1_args[] = { &cpu_used_av1,
                                       &auto_altref,
                                       &sharpness,
                                       &static_thresh,
                                       &tile_cols,
                                       &tile_rows,
                                       &arnr_maxframes,
                                       &arnr_strength,
                                       &arnr_type,
                                       &tune_ssim,
                                       &cq_level,
                                       &max_intra_rate_pct,
                                       &max_inter_rate_pct,
                                       &gf_cbr_boost_pct,
                                       &lossless,
480 481 482 483 484
#if CONFIG_AOM_QM
                                       &enable_qm,
                                       &qm_min,
                                       &qm_max,
#endif
485 486 487 488 489 490 491 492
                                       &frame_parallel_decoding,
                                       &aq_mode,
                                       &frame_periodic_boost,
                                       &noise_sens,
                                       &tune_content,
                                       &input_color_space,
                                       &min_gf_interval,
                                       &max_gf_interval,
493
#if CONFIG_EXT_PARTITION
494
                                       &superblock_size,
495
#endif  // CONFIG_EXT_PARTITION
496 497 498 499
#if CONFIG_TILE_GROUPS
                                       &num_tg,
                                       &mtu_size,
#endif
500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519
#if CONFIG_AOM_HIGHBITDEPTH
                                       &bitdeptharg,
                                       &inbitdeptharg,
#endif  // CONFIG_AOM_HIGHBITDEPTH
                                       NULL };
static const int av1_arg_ctrl_map[] = { AOME_SET_CPUUSED,
                                        AOME_SET_ENABLEAUTOALTREF,
                                        AOME_SET_SHARPNESS,
                                        AOME_SET_STATIC_THRESHOLD,
                                        AV1E_SET_TILE_COLUMNS,
                                        AV1E_SET_TILE_ROWS,
                                        AOME_SET_ARNR_MAXFRAMES,
                                        AOME_SET_ARNR_STRENGTH,
                                        AOME_SET_ARNR_TYPE,
                                        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,
520 521 522 523 524
#if CONFIG_AOM_QM
                                        AV1E_SET_ENABLE_QM,
                                        AV1E_SET_QM_MIN,
                                        AV1E_SET_QM_MAX,
#endif
525 526 527 528 529 530 531 532
                                        AV1E_SET_FRAME_PARALLEL_DECODING,
                                        AV1E_SET_AQ_MODE,
                                        AV1E_SET_FRAME_PERIODIC_BOOST,
                                        AV1E_SET_NOISE_SENSITIVITY,
                                        AV1E_SET_TUNE_CONTENT,
                                        AV1E_SET_COLOR_SPACE,
                                        AV1E_SET_MIN_GF_INTERVAL,
                                        AV1E_SET_MAX_GF_INTERVAL,
533
#if CONFIG_EXT_PARTITION
534
                                        AV1E_SET_SUPERBLOCK_SIZE,
535
#endif  // CONFIG_EXT_PARTITION
536 537 538 539
#if CONFIG_TILE_GROUPS
                                        AV1E_SET_NUM_TG,
                                        AV1E_SET_MTU,
#endif
540
                                        0 };
hui su's avatar
hui su committed
541 542
#endif

John Koleszar's avatar
John Koleszar committed
543 544
static const arg_def_t *no_args[] = { NULL };

545
void usage_exit(void) {
John Koleszar's avatar
John Koleszar committed
546
  int i;
547
  const int num_encoder = get_aom_encoder_count();
John Koleszar's avatar
John Koleszar committed
548 549 550 551 552

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

  fprintf(stderr, "\nOptions:\n");
John Koleszar's avatar
John Koleszar committed
553
  arg_show_usage(stderr, main_args);
John Koleszar's avatar
John Koleszar committed
554
  fprintf(stderr, "\nEncoder Global Options:\n");
John Koleszar's avatar
John Koleszar committed
555
  arg_show_usage(stderr, global_args);
John Koleszar's avatar
John Koleszar committed
556
  fprintf(stderr, "\nRate Control Options:\n");
John Koleszar's avatar
John Koleszar committed
557
  arg_show_usage(stderr, rc_args);
John Koleszar's avatar
John Koleszar committed
558
  fprintf(stderr, "\nTwopass Rate Control Options:\n");
John Koleszar's avatar
John Koleszar committed
559
  arg_show_usage(stderr, rc_twopass_args);
John Koleszar's avatar
John Koleszar committed
560
  fprintf(stderr, "\nKeyframe Placement Options:\n");
John Koleszar's avatar
John Koleszar committed
561
  arg_show_usage(stderr, kf_args);
562 563 564
#if CONFIG_AV1_ENCODER
  fprintf(stderr, "\nAV1 Specific Options:\n");
  arg_show_usage(stderr, av1_args);
John Koleszar's avatar
John Koleszar committed
565
#endif
clang-format's avatar
clang-format committed
566 567
  fprintf(stderr,
          "\nStream timebase (--timebase):\n"
John Koleszar's avatar
John Koleszar committed
568 569
          "  The desired precision of timestamps in the output, expressed\n"
          "  in fractional seconds. Default is 1/1000.\n");
570
  fprintf(stderr, "\nIncluded encoders:\n\n");
John Koleszar's avatar
John Koleszar committed
571

572
  for (i = 0; i < num_encoder; ++i) {
573
    const AvxInterface *const encoder = get_aom_encoder_by_index(i);
clang-format's avatar
clang-format committed
574 575
    const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
    fprintf(stderr, "    %-6s - %s %s\n", encoder->name,
576
            aom_codec_iface_name(encoder->codec_interface()), defstr);
577
  }
578 579
  fprintf(stderr, "\n        ");
  fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
John Koleszar's avatar
John Koleszar committed
580 581

  exit(EXIT_FAILURE);
John Koleszar's avatar
John Koleszar committed
582 583
}

clang-format's avatar
clang-format committed
584
#define mmin(a, b) ((a) < (b) ? (a) : (b))
585

586 587 588
#if CONFIG_AOM_HIGHBITDEPTH
static void find_mismatch_high(const aom_image_t *const img1,
                               const aom_image_t *const img2, int yloc[4],
clang-format's avatar
clang-format committed
589
                               int uloc[4], int vloc[4]) {
590 591 592 593 594 595 596 597 598 599 600 601
  uint16_t *plane1, *plane2;
  uint32_t stride1, stride2;
  const uint32_t bsize = 64;
  const uint32_t bsizey = bsize >> img1->y_chroma_shift;
  const uint32_t bsizex = bsize >> img1->x_chroma_shift;
  const uint32_t c_w =
      (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
  const uint32_t c_h =
      (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
  int match = 1;
  uint32_t i, j;
  yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
602 603 604 605
  plane1 = (uint16_t *)img1->planes[AOM_PLANE_Y];
  plane2 = (uint16_t *)img2->planes[AOM_PLANE_Y];
  stride1 = img1->stride[AOM_PLANE_Y] / 2;
  stride2 = img2->stride[AOM_PLANE_Y] / 2;
606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627
  for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
    for (j = 0; match && j < img1->d_w; j += bsize) {
      int k, l;
      const int si = mmin(i + bsize, img1->d_h) - i;
      const int sj = mmin(j + bsize, img1->d_w) - j;
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
          if (*(plane1 + (i + k) * stride1 + j + l) !=
              *(plane2 + (i + k) * stride2 + j + l)) {
            yloc[0] = i + k;
            yloc[1] = j + l;
            yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
            yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
            match = 0;
            break;
          }
        }
      }
    }
  }

  uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
628 629 630 631
  plane1 = (uint16_t *)img1->planes[AOM_PLANE_U];
  plane2 = (uint16_t *)img2->planes[AOM_PLANE_U];
  stride1 = img1->stride[AOM_PLANE_U] / 2;
  stride2 = img2->stride[AOM_PLANE_U] / 2;
632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653
  for (i = 0, match = 1; match && i < c_h; i += bsizey) {
    for (j = 0; match && j < c_w; j += bsizex) {
      int k, l;
      const int si = mmin(i + bsizey, c_h - i);
      const int sj = mmin(j + bsizex, c_w - j);
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
          if (*(plane1 + (i + k) * stride1 + j + l) !=
              *(plane2 + (i + k) * stride2 + j + l)) {
            uloc[0] = i + k;
            uloc[1] = j + l;
            uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
            uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
            match = 0;
            break;
          }
        }
      }
    }
  }

  vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
654 655 656 657
  plane1 = (uint16_t *)img1->planes[AOM_PLANE_V];
  plane2 = (uint16_t *)img2->planes[AOM_PLANE_V];
  stride1 = img1->stride[AOM_PLANE_V] / 2;
  stride2 = img2->stride[AOM_PLANE_V] / 2;
658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680
  for (i = 0, match = 1; match && i < c_h; i += bsizey) {
    for (j = 0; match && j < c_w; j += bsizex) {
      int k, l;
      const int si = mmin(i + bsizey, c_h - i);
      const int sj = mmin(j + bsizex, c_w - j);
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
          if (*(plane1 + (i + k) * stride1 + j + l) !=
              *(plane2 + (i + k) * stride2 + j + l)) {
            vloc[0] = i + k;
            vloc[1] = j + l;
            vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
            vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
            match = 0;
            break;
          }
        }
      }
    }
  }
}
#endif

681 682
static void find_mismatch(const aom_image_t *const img1,
                          const aom_image_t *const img2, int yloc[4],
clang-format's avatar
clang-format committed
683
                          int uloc[4], int vloc[4]) {
684 685 686 687 688 689 690 691 692
  const uint32_t bsize = 64;
  const uint32_t bsizey = bsize >> img1->y_chroma_shift;
  const uint32_t bsizex = bsize >> img1->x_chroma_shift;
  const uint32_t c_w =
      (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
  const uint32_t c_h =
      (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
  int match = 1;
  uint32_t i, j;
693
  yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
694 695
  for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
    for (j = 0; match && j < img1->d_w; j += bsize) {
696
      int k, l;
James Zern's avatar
James Zern committed
697 698 699 700
      const int si = mmin(i + bsize, img1->d_h) - i;
      const int sj = mmin(j + bsize, img1->d_w) - j;
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
701 702 703 704
          if (*(img1->planes[AOM_PLANE_Y] +
                (i + k) * img1->stride[AOM_PLANE_Y] + j + l) !=
              *(img2->planes[AOM_PLANE_Y] +
                (i + k) * img2->stride[AOM_PLANE_Y] + j + l)) {
705 706
            yloc[0] = i + k;
            yloc[1] = j + l;
707 708 709 710
            yloc[2] = *(img1->planes[AOM_PLANE_Y] +
                        (i + k) * img1->stride[AOM_PLANE_Y] + j + l);
            yloc[3] = *(img2->planes[AOM_PLANE_Y] +
                        (i + k) * img2->stride[AOM_PLANE_Y] + j + l);
711
            match = 0;
712
            break;
713 714
          }
        }
James Zern's avatar
James Zern committed
715
      }
716
    }
717
  }
718

719
  uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
720
  for (i = 0, match = 1; match && i < c_h; i += bsizey) {
John Koleszar's avatar
John Koleszar committed
721
    for (j = 0; match && j < c_w; j += bsizex) {
722
      int k, l;
James Zern's avatar
James Zern committed
723 724 725 726
      const int si = mmin(i + bsizey, c_h - i);
      const int sj = mmin(j + bsizex, c_w - j);
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
727 728 729 730
          if (*(img1->planes[AOM_PLANE_U] +
                (i + k) * img1->stride[AOM_PLANE_U] + j + l) !=
              *(img2->planes[AOM_PLANE_U] +
                (i + k) * img2->stride[AOM_PLANE_U] + j + l)) {
731 732
            uloc[0] = i + k;
            uloc[1] = j + l;
733 734 735 736
            uloc[2] = *(img1->planes[AOM_PLANE_U] +
                        (i + k) * img1->stride[AOM_PLANE_U] + j + l);
            uloc[3] = *(img2->planes[AOM_PLANE_U] +
                        (i + k) * img2->stride[AOM_PLANE_U] + j + l);
737 738 739 740
            match = 0;
            break;
          }
        }
James Zern's avatar
James Zern committed
741
      }
742 743
    }
  }
744
  vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
745
  for (i = 0, match = 1; match && i < c_h; i += bsizey) {
John Koleszar's avatar
John Koleszar committed
746
    for (j = 0; match && j < c_w; j += bsizex) {
747
      int k, l;
James Zern's avatar
James Zern committed
748 749 750 751
      const int si = mmin(i + bsizey, c_h - i);
      const int sj = mmin(j + bsizex, c_w - j);
      for (k = 0; match && k < si; ++k) {
        for (l = 0; match && l < sj; ++l) {
752 753 754 755
          if (*(img1->planes[AOM_PLANE_V] +
                (i + k) * img1->stride[AOM_PLANE_V] + j + l) !=
              *(img2->planes[AOM_PLANE_V] +
                (i + k) * img2->stride[AOM_PLANE_V] + j + l)) {
756 757
            vloc[0] = i + k;
            vloc[1] = j + l;
758 759 760 761
            vloc[2] = *(img1->planes[AOM_PLANE_V] +
                        (i + k) * img1->stride[AOM_PLANE_V] + j + l);
            vloc[3] = *(img2->planes[AOM_PLANE_V] +
                        (i + k) * img2->stride[AOM_PLANE_V] + j + l);
762 763 764 765
            match = 0;
            break;
          }
        }
James Zern's avatar
James Zern committed
766
      }
767 768
    }
  }
769 770
}

771 772
static int compare_img(const aom_image_t *const img1,
                       const aom_image_t *const img2) {
773
  uint32_t l_w = img1->d_w;
clang-format's avatar
clang-format committed
774
  uint32_t c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
775 776 777
  const uint32_t c_h =
      (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
  uint32_t i;
John Koleszar's avatar
John Koleszar committed
778
  int match = 1;
779

John Koleszar's avatar
John Koleszar committed
780
  match &= (img1->fmt == img2->fmt);
781 782
  match &= (img1->d_w == img2->d_w);
  match &= (img1->d_h == img2->d_h);
783 784
#if CONFIG_AOM_HIGHBITDEPTH
  if (img1->fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
785 786 787 788
    l_w *= 2;
    c_w *= 2;
  }
#endif
789

James Zern's avatar
James Zern committed
790
  for (i = 0; i < img1->d_h; ++i)
791 792
    match &= (memcmp(img1->planes[AOM_PLANE_Y] + i * img1->stride[AOM_PLANE_Y],
                     img2->planes[AOM_PLANE_Y] + i * img2->stride[AOM_PLANE_Y],
793
                     l_w) == 0);
794

James Zern's avatar
James Zern committed
795
  for (i = 0; i < c_h; ++i)
796 797
    match &= (memcmp(img1->planes[AOM_PLANE_U] + i * img1->stride[AOM_PLANE_U],
                     img2->planes[AOM_PLANE_U] + i * img2->stride[AOM_PLANE_U],
798
                     c_w) == 0);
799

James Zern's avatar
James Zern committed
800
  for (i = 0; i < c_h; ++i)
801 802
    match &= (memcmp(img1->planes[AOM_PLANE_V] + i * img1->stride[AOM_PLANE_V],
                     img2->planes[AOM_PLANE_V] + i * img2->stride[AOM_PLANE_V],
803
                     c_w) == 0);
804

John Koleszar's avatar
John Koleszar committed
805
  return match;
806 807
}

clang-format's avatar
clang-format committed
808
#define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
809 810
#if CONFIG_AV1_ENCODER
#define ARG_CTRL_CNT_MAX NELEMENTS(av1_arg_ctrl_map)
811
#endif
John Koleszar's avatar
John Koleszar committed
812

813 814
#if !CONFIG_WEBM_IO
typedef int stereo_format_t;
clang-format's avatar
clang-format committed
815 816 817
struct WebmOutputContext {
  int debug;
};
818 819
#endif

820
/* Per-stream configuration */
John Koleszar's avatar
John Koleszar committed
821
struct stream_config {
822
  struct aom_codec_enc_cfg cfg;
clang-format's avatar
clang-format committed
823 824
  const char *out_fn;
  const char *stats_fn;
825
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
826
  const char *fpmb_stats_fn;
827
#endif
clang-format's avatar
clang-format committed
828 829 830 831
  stereo_format_t stereo_fmt;
  int arg_ctrls[ARG_CTRL_CNT_MAX][2];
  int arg_ctrl_cnt;
  int write_webm;
832
#if CONFIG_AOM_HIGHBITDEPTH
833
  // whether to use 16bit internal buffers
clang-format's avatar
clang-format committed
834
  int use_16bit_internal;
835
#endif
836 837
};

John Koleszar's avatar
John Koleszar committed
838
struct stream_state {
clang-format's avatar
clang-format committed
839 840 841 842 843 844 845 846 847 848 849
  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];
850
  aom_codec_ctx_t encoder;
clang-format's avatar
clang-format committed
851 852 853 854
  unsigned int frames_out;
  uint64_t cx_time;
  size_t nbytes;
  stats_io_t stats;
855
#if CONFIG_FP_MB_STATS
clang-format's avatar
clang-format committed
856
  stats_io_t fpmb_stats;
857
#endif
858 859
  struct aom_image *img;
  aom_codec_ctx_t decoder;
clang-format's avatar
clang-format committed
860
  int mismatch_seen;
861 862
};

clang-format's avatar
clang-format committed
863
static void validate_positive_rational(const char *msg,
864
                                       struct aom_rational *rat) {
John Koleszar's avatar
John Koleszar committed
865 866 867 868
  if (rat->den < 0) {
    rat->num *= -1;
    rat->den *= -1;
  }
869

clang-format's avatar
clang-format committed
870
  if (rat->num < 0) die("Error: %s must be positive\n", msg);
871

clang-format's avatar
clang-format committed
872
  if (!rat->den) die("Error: %s has zero denominator\n", msg);
873 874
}

875
static void parse_global_config(struct AvxEncoderConfig *global, char **argv) {
clang-format's avatar
clang-format committed
876 877
  char **argi, **argj;
  struct arg arg;
878
  const int num_encoder = get_aom_encoder_count();
879

clang-format's avatar
clang-format committed
880
  if (num_encoder < 1) die("Error: no valid encoder available\n");
John Koleszar's avatar
John Koleszar committed
881 882 883

  /* Initialize default parameters */
  memset(global, 0, sizeof(*global));
884
  global->codec = get_aom_encoder_by_index(num_encoder - 1);
885
  global->passes = 0;
886
  global->color_type = I420;
887
  /* Assign default deadline to good quality */
888
  global->deadline = AOM_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
889 890 891 892 893

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

    if (arg_match(&arg, &codecarg, argi)) {
894
      global->codec = get_aom_encoder_by_name(arg.val);
895 896
      if (!global->codec)
        die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
John Koleszar's avatar
John Koleszar committed
897
    } else if (arg_match(&arg, &passes, argi)) {
John Koleszar's avatar
John Koleszar committed
898
      global->passes = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
899

John Koleszar's avatar
John Koleszar committed
900 901
      if (global->passes < 1 || global->passes > 2)
        die("Error: Invalid number of passes (%d)\n", global->passes);
John Koleszar's avatar
John Koleszar committed
902
    } else if (arg_match(&arg, &pass_arg, argi)) {
John Koleszar's avatar
John Koleszar committed
903 904 905
      global->pass = arg_parse_uint(&arg);

      if (global->pass < 1 || global->pass > 2)
clang-format's avatar
clang-format committed
906
        die("Error: Invalid pass selected (%d)\n", global->pass);
John Koleszar's avatar
John Koleszar committed
907 908
    } else if (arg_match(&arg, &usage, argi))
      global->usage = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
909
    else if (arg_match(&arg, &deadline, argi))
John Koleszar's avatar
John Koleszar committed
910
      global->deadline = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
911
    else if (arg_match(&arg, &best_dl, argi))
912
      global->deadline = AOM_DL_BEST_QUALITY;
John Koleszar's avatar
John Koleszar committed
913
    else if (arg_match(&arg, &good_dl, argi))
914
      global->deadline = AOM_DL_GOOD_QUALITY;
John Koleszar's avatar
John Koleszar committed
915
    else if (arg_match(&arg, &rt_dl, argi))
916
      global->deadline = AOM_DL_REALTIME;
John Koleszar's avatar
John Koleszar committed
917
    else if (arg_match(&arg, &use_yv12, argi))
918
      global->color_type = YV12;
John Koleszar's avatar
John Koleszar committed
919
    else if (arg_match(&arg, &use_i420, argi))
920 921 922 923 924
      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
925 926
    else if (arg_match(&arg, &use_i440, argi))
      global->color_type = I440;
John Koleszar's avatar
John Koleszar committed
927 928 929 930
    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
931
    else if (arg_match(&arg, &limit, argi))
John Koleszar's avatar
John Koleszar committed
932
      global->limit = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
933
    else if (arg_match(&arg, &skip, argi))
John Koleszar's avatar
John Koleszar committed
934
      global->skip_frames = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
935
    else if (arg_match(&arg, &psnrarg, argi))
John Koleszar's avatar
John Koleszar committed
936
      global->show_psnr = 1;
John Koleszar's avatar
John Koleszar committed
937
    else if (arg_match(&arg, &recontest, argi))
938
      global->test_decode = arg_parse_enum_or_int(&arg);
John Koleszar's avatar
John Koleszar committed
939
    else if (arg_match(&arg, &framerate, argi)) {
John Koleszar's avatar
John Koleszar committed
940 941 942 943 944
      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
945
    else if (arg_match(&arg, &debugmode, argi))
John Koleszar's avatar
John Koleszar committed
946
      global->debug = 1;
John Koleszar's avatar
John Koleszar committed
947
    else if (arg_match(&arg, &q_hist_n, argi))
John Koleszar's avatar
John Koleszar committed
948
      global->show_q_hist_buckets = arg_parse_uint(&arg);
John Koleszar's avatar
John Koleszar committed
949
    else if (arg_match(&arg, &rate_hist_n, argi))
John Koleszar's avatar
John Koleszar committed
950
      global->show_rate_hist_buckets = arg_parse_uint(&arg);
951 952 953 954
    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
955 956 957
    else
      argj++;
  }
John Koleszar's avatar
John Koleszar committed
958

959 960 961
  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
962 963
      warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
           global->pass);
964 965 966
      global->passes = global->pass;
    }
  }
John Koleszar's avatar
John Koleszar committed
967
  /* Validate global config */
968
  if (global->passes == 0) {
969 970
#if CONFIG_AV1_ENCODER
    // Make default AV1 passes = 2 until there is a better quality 1-pass
971
    // encoder
972
    if (global->codec != NULL && global->codec->name != NULL)
973 974
      global->passes = (strcmp(global->codec->name, "av1") == 0 &&
                        global->deadline != AOM_DL_REALTIME)
clang-format's avatar
clang-format committed
975 976
                           ? 2
                           : 1;
James Zern's avatar
James Zern committed
977 978 979
#else
    global->passes = 1;
#endif
980
  }
981

982
  if (global->deadline == AOM_DL_REALTIME && global->passes > 1) {
983 984
    warn("Enforcing one-pass encoding in realtime mode\n");
    global->passes = 1;
John Koleszar's avatar
John Koleszar committed
985
  }
986 987
}

988
static void open_input_file(struct AvxInputContext *input) {
John Koleszar's avatar
John Koleszar committed
989
  /* Parse certain options from the input file, if possible */
clang-format's avatar
clang-format committed
990 991
  input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
                                             : set_binary_mode(stdin);
John Koleszar's avatar
John Koleszar committed
992

clang-format's avatar
clang-format committed
993
  if (!input->file) fatal("Failed to open input file");
John Koleszar's avatar
John Koleszar committed
994

995 996 997 998 999 1000 1001 1002
  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);
  }

1003 1004 1005 1006
  /* 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
1007 1008 1009 1010 1011 1012
  /* 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
1013
  if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
1014 1015
    if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
                       input->only_i420) >= 0) {
John Koleszar's avatar
John Koleszar committed
1016
      input->file_type = FILE_TYPE_Y4M;
1017 1018
      input->width = input->y4m.pic_w;
      input->height = input->y4m.pic_h;
1019 1020
      input->pixel_aspect_ratio.numerator = input->y4m.par_n;
      input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1021 1022
      input->framerate.numerator = input->y4m.fps_n;
      input->framerate.denominator = input->y4m.fps_d;
1023
      input->fmt = input->y4m.aom_fmt;
1024
      input->bit_depth = input->y4m.bit_depth;
John Koleszar's avatar
John Koleszar committed
1025 1026
    } else
      fatal("Unsupported Y4M stream.");
1027
  } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1028
    fatal("IVF is not supported as input.");
John Koleszar's avatar
John Koleszar committed
1029 1030
  } else {
    input->file_type = FILE_TYPE_RAW;
John Koleszar's avatar
John Koleszar committed
1031
  }
1032 1033
}

1034
static void close_input_file(struct AvxInputContext *input) {
John Koleszar's avatar
John Koleszar committed
1035
  fclose(input->file);
clang-format's avatar
clang-format committed
1036
  if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
1037 1038
}

1039
static struct stream_state *new_stream(struct AvxEncoderConfig *global,
1040
                                       struct stream_state *prev) {
John Koleszar's avatar
John Koleszar committed
1041 1042 1043
  struct stream_state *stream;

  stream = calloc(1, sizeof(*stream));
1044
  if (stream == NULL) {
John Koleszar's avatar
John Koleszar committed
1045
    fatal("Failed to allocate new stream.");
1046 1047
  }

John Koleszar's avatar
John Koleszar committed
1048 1049 1050 1051 1052
  if (prev) {
    memcpy(stream, prev, sizeof(*stream));
    stream->index++;
    prev->next = stream;
  } else {
1053
    aom_codec_err_t res;
John Koleszar's avatar
John Koleszar committed
1054 1055

    /* Populate encoder configuration */
1056
    res = aom_codec_enc_config_default(global->codec->codec_interface(),
clang-format's avatar
clang-format committed
1057
                                       &stream->config.cfg, global->usage);
1058
    if (res) fatal("Failed to get config: %s\n", aom_codec_err_to_string(res));
John Koleszar's avatar
John Koleszar committed
1059

John Koleszar's avatar
John Koleszar committed
1060 1061 1062 1063
    /* 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;
1064

John Koleszar's avatar
John Koleszar committed
1065 1066 1067 1068 1069
    /* 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
1070

John Koleszar's avatar
John Koleszar committed
1071 1072
    /* Initialize remaining stream parameters */
    stream->config.write_webm = 1;
1073
#if CONFIG_WEBM_IO
1074
    stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1075 1076 1077
    stream->webm_ctx.last_pts_ns = -1;
    stream->webm_ctx.writer = NULL;
    stream->webm_ctx.segment = NULL;
1078
#endif
John Koleszar's avatar
John Koleszar committed
1079

John Koleszar's avatar
John Koleszar committed
1080
    /* Allows removal of the application version from the EBML tags */
1081
    stream->webm_ctx.debug = global->debug;
1082 1083

    /* Default lag_in_frames is 0 in realtime mode */
Yaowu Xu's avatar