AOMedia AV1 Codec
aomenc
1/*
2 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3 *
4 * This source code is subject to the terms of the BSD 2 Clause License and
5 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6 * was not distributed with this source code in the LICENSE file, you can
7 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8 * Media Patent License 1.0 was not distributed with this source code in the
9 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 */
11
12#include "apps/aomenc.h"
13
14#include "config/aom_config.h"
15
16#include <assert.h>
17#include <limits.h>
18#include <math.h>
19#include <stdarg.h>
20#include <stdio.h>
21#include <stdlib.h>
22#include <string.h>
23
24#if CONFIG_AV1_DECODER
25#include "aom/aom_decoder.h"
26#include "aom/aomdx.h"
27#endif
28
29#include "aom/aom_encoder.h"
30#include "aom/aom_integer.h"
31#include "aom/aomcx.h"
32#include "aom_dsp/aom_dsp_common.h"
33#include "aom_ports/aom_timer.h"
34#include "aom_ports/mem_ops.h"
35#include "common/args.h"
36#include "common/ivfenc.h"
37#include "common/tools_common.h"
38#include "common/warnings.h"
39
40#if CONFIG_WEBM_IO
41#include "common/webmenc.h"
42#endif
43
44#include "common/y4minput.h"
45#include "examples/encoder_util.h"
46#include "stats/aomstats.h"
47#include "stats/rate_hist.h"
48
49#if CONFIG_LIBYUV
50#include <libyuv/scale.h>
51#endif
52
53/* Swallow warnings about unused results of fread/fwrite */
54static size_t wrap_fread(void *ptr, size_t size, size_t nmemb, FILE *stream) {
55 return fread(ptr, size, nmemb, stream);
56}
57#define fread wrap_fread
58
59static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
60 FILE *stream) {
61 return fwrite(ptr, size, nmemb, stream);
62}
63#define fwrite wrap_fwrite
64
65static const char *exec_name;
66
67static AOM_TOOLS_FORMAT_PRINTF(3, 0) void warn_or_exit_on_errorv(
68 aom_codec_ctx_t *ctx, int fatal, const char *s, va_list ap) {
69 if (ctx->err) {
70 const char *detail = aom_codec_error_detail(ctx);
71
72 vfprintf(stderr, s, ap);
73 fprintf(stderr, ": %s\n", aom_codec_error(ctx));
74
75 if (detail) fprintf(stderr, " %s\n", detail);
76
77 if (fatal) exit(EXIT_FAILURE);
78 }
79}
80
81static AOM_TOOLS_FORMAT_PRINTF(2,
82 3) void ctx_exit_on_error(aom_codec_ctx_t *ctx,
83 const char *s, ...) {
84 va_list ap;
85
86 va_start(ap, s);
87 warn_or_exit_on_errorv(ctx, 1, s, ap);
88 va_end(ap);
89}
90
91static AOM_TOOLS_FORMAT_PRINTF(3, 4) void warn_or_exit_on_error(
92 aom_codec_ctx_t *ctx, int fatal, const char *s, ...) {
93 va_list ap;
94
95 va_start(ap, s);
96 warn_or_exit_on_errorv(ctx, fatal, s, ap);
97 va_end(ap);
98}
99
100static int read_frame(struct AvxInputContext *input_ctx, aom_image_t *img) {
101 FILE *f = input_ctx->file;
102 y4m_input *y4m = &input_ctx->y4m;
103 int shortread = 0;
104
105 if (input_ctx->file_type == FILE_TYPE_Y4M) {
106 if (y4m_input_fetch_frame(y4m, f, img) < 1) return 0;
107 } else {
108 shortread = read_yuv_frame(input_ctx, img);
109 }
110
111 return !shortread;
112}
113
114static int file_is_y4m(const char detect[4]) {
115 if (memcmp(detect, "YUV4", 4) == 0) {
116 return 1;
117 }
118 return 0;
119}
120
121static int fourcc_is_ivf(const char detect[4]) {
122 if (memcmp(detect, "DKIF", 4) == 0) {
123 return 1;
124 }
125 return 0;
126}
127
128static const int av1_arg_ctrl_map[] = { AOME_SET_CPUUSED,
217#if CONFIG_DENOISE
220 AV1E_SET_ENABLE_DNL_DENOISING,
221#endif // CONFIG_DENOISE
231#if CONFIG_TUNE_VMAF
233#endif
240 0 };
241
242const arg_def_t *main_args[] = { &g_av1_codec_arg_defs.help,
243 &g_av1_codec_arg_defs.use_cfg,
244 &g_av1_codec_arg_defs.debugmode,
245 &g_av1_codec_arg_defs.outputfile,
246 &g_av1_codec_arg_defs.codecarg,
247 &g_av1_codec_arg_defs.passes,
248 &g_av1_codec_arg_defs.pass_arg,
249 &g_av1_codec_arg_defs.fpf_name,
250 &g_av1_codec_arg_defs.limit,
251 &g_av1_codec_arg_defs.skip,
252 &g_av1_codec_arg_defs.good_dl,
253 &g_av1_codec_arg_defs.rt_dl,
254 &g_av1_codec_arg_defs.ai_dl,
255 &g_av1_codec_arg_defs.quietarg,
256 &g_av1_codec_arg_defs.verbosearg,
257 &g_av1_codec_arg_defs.psnrarg,
258 &g_av1_codec_arg_defs.use_webm,
259 &g_av1_codec_arg_defs.use_ivf,
260 &g_av1_codec_arg_defs.use_obu,
261 &g_av1_codec_arg_defs.q_hist_n,
262 &g_av1_codec_arg_defs.rate_hist_n,
263 &g_av1_codec_arg_defs.disable_warnings,
264 &g_av1_codec_arg_defs.disable_warning_prompt,
265 &g_av1_codec_arg_defs.recontest,
266 NULL };
267
268const arg_def_t *global_args[] = {
269 &g_av1_codec_arg_defs.use_nv12,
270 &g_av1_codec_arg_defs.use_yv12,
271 &g_av1_codec_arg_defs.use_i420,
272 &g_av1_codec_arg_defs.use_i422,
273 &g_av1_codec_arg_defs.use_i444,
274 &g_av1_codec_arg_defs.usage,
275 &g_av1_codec_arg_defs.threads,
276 &g_av1_codec_arg_defs.profile,
277 &g_av1_codec_arg_defs.width,
278 &g_av1_codec_arg_defs.height,
279 &g_av1_codec_arg_defs.forced_max_frame_width,
280 &g_av1_codec_arg_defs.forced_max_frame_height,
281#if CONFIG_WEBM_IO
282 &g_av1_codec_arg_defs.stereo_mode,
283#endif
284 &g_av1_codec_arg_defs.timebase,
285 &g_av1_codec_arg_defs.framerate,
286 &g_av1_codec_arg_defs.global_error_resilient,
287 &g_av1_codec_arg_defs.bitdeptharg,
288 &g_av1_codec_arg_defs.inbitdeptharg,
289 &g_av1_codec_arg_defs.lag_in_frames,
290 &g_av1_codec_arg_defs.large_scale_tile,
291 &g_av1_codec_arg_defs.monochrome,
292 &g_av1_codec_arg_defs.full_still_picture_hdr,
293 &g_av1_codec_arg_defs.use_16bit_internal,
294 &g_av1_codec_arg_defs.save_as_annexb,
295 NULL
296};
297
298const arg_def_t *rc_args[] = { &g_av1_codec_arg_defs.dropframe_thresh,
299 &g_av1_codec_arg_defs.resize_mode,
300 &g_av1_codec_arg_defs.resize_denominator,
301 &g_av1_codec_arg_defs.resize_kf_denominator,
302 &g_av1_codec_arg_defs.superres_mode,
303 &g_av1_codec_arg_defs.superres_denominator,
304 &g_av1_codec_arg_defs.superres_kf_denominator,
305 &g_av1_codec_arg_defs.superres_qthresh,
306 &g_av1_codec_arg_defs.superres_kf_qthresh,
307 &g_av1_codec_arg_defs.end_usage,
308 &g_av1_codec_arg_defs.target_bitrate,
309 &g_av1_codec_arg_defs.min_quantizer,
310 &g_av1_codec_arg_defs.max_quantizer,
311 &g_av1_codec_arg_defs.undershoot_pct,
312 &g_av1_codec_arg_defs.overshoot_pct,
313 &g_av1_codec_arg_defs.buf_sz,
314 &g_av1_codec_arg_defs.buf_initial_sz,
315 &g_av1_codec_arg_defs.buf_optimal_sz,
316 &g_av1_codec_arg_defs.bias_pct,
317 &g_av1_codec_arg_defs.minsection_pct,
318 &g_av1_codec_arg_defs.maxsection_pct,
319 NULL };
320
321const arg_def_t *kf_args[] = { &g_av1_codec_arg_defs.fwd_kf_enabled,
322 &g_av1_codec_arg_defs.kf_min_dist,
323 &g_av1_codec_arg_defs.kf_max_dist,
324 &g_av1_codec_arg_defs.kf_disabled,
325 &g_av1_codec_arg_defs.sframe_dist,
326 &g_av1_codec_arg_defs.sframe_mode,
327 NULL };
328
329// TODO(bohanli): Currently all options are supported by the key & value API.
330// Consider removing the control ID usages?
331const arg_def_t *av1_ctrl_args[] = {
332 &g_av1_codec_arg_defs.cpu_used_av1,
333 &g_av1_codec_arg_defs.auto_altref,
334 &g_av1_codec_arg_defs.sharpness,
335 &g_av1_codec_arg_defs.static_thresh,
336 &g_av1_codec_arg_defs.rowmtarg,
337 &g_av1_codec_arg_defs.fpmtarg,
338 &g_av1_codec_arg_defs.tile_cols,
339 &g_av1_codec_arg_defs.tile_rows,
340 &g_av1_codec_arg_defs.enable_tpl_model,
341 &g_av1_codec_arg_defs.enable_keyframe_filtering,
342 &g_av1_codec_arg_defs.arnr_maxframes,
343 &g_av1_codec_arg_defs.arnr_strength,
344 &g_av1_codec_arg_defs.tune_metric,
345 &g_av1_codec_arg_defs.cq_level,
346 &g_av1_codec_arg_defs.max_intra_rate_pct,
347 &g_av1_codec_arg_defs.max_inter_rate_pct,
348 &g_av1_codec_arg_defs.gf_cbr_boost_pct,
349 &g_av1_codec_arg_defs.lossless,
350 &g_av1_codec_arg_defs.enable_cdef,
351 &g_av1_codec_arg_defs.enable_restoration,
352 &g_av1_codec_arg_defs.enable_rect_partitions,
353 &g_av1_codec_arg_defs.enable_ab_partitions,
354 &g_av1_codec_arg_defs.enable_1to4_partitions,
355 &g_av1_codec_arg_defs.min_partition_size,
356 &g_av1_codec_arg_defs.max_partition_size,
357 &g_av1_codec_arg_defs.enable_dual_filter,
358 &g_av1_codec_arg_defs.enable_chroma_deltaq,
359 &g_av1_codec_arg_defs.enable_intra_edge_filter,
360 &g_av1_codec_arg_defs.enable_order_hint,
361 &g_av1_codec_arg_defs.enable_tx64,
362 &g_av1_codec_arg_defs.enable_flip_idtx,
363 &g_av1_codec_arg_defs.enable_rect_tx,
364 &g_av1_codec_arg_defs.enable_dist_wtd_comp,
365 &g_av1_codec_arg_defs.enable_masked_comp,
366 &g_av1_codec_arg_defs.enable_onesided_comp,
367 &g_av1_codec_arg_defs.enable_interintra_comp,
368 &g_av1_codec_arg_defs.enable_smooth_interintra,
369 &g_av1_codec_arg_defs.enable_diff_wtd_comp,
370 &g_av1_codec_arg_defs.enable_interinter_wedge,
371 &g_av1_codec_arg_defs.enable_interintra_wedge,
372 &g_av1_codec_arg_defs.enable_global_motion,
373 &g_av1_codec_arg_defs.enable_warped_motion,
374 &g_av1_codec_arg_defs.enable_filter_intra,
375 &g_av1_codec_arg_defs.enable_smooth_intra,
376 &g_av1_codec_arg_defs.enable_paeth_intra,
377 &g_av1_codec_arg_defs.enable_cfl_intra,
378 &g_av1_codec_arg_defs.enable_diagonal_intra,
379 &g_av1_codec_arg_defs.force_video_mode,
380 &g_av1_codec_arg_defs.enable_obmc,
381 &g_av1_codec_arg_defs.enable_overlay,
382 &g_av1_codec_arg_defs.enable_palette,
383 &g_av1_codec_arg_defs.enable_intrabc,
384 &g_av1_codec_arg_defs.enable_angle_delta,
385 &g_av1_codec_arg_defs.disable_trellis_quant,
386 &g_av1_codec_arg_defs.enable_qm,
387 &g_av1_codec_arg_defs.qm_min,
388 &g_av1_codec_arg_defs.qm_max,
389 &g_av1_codec_arg_defs.reduced_tx_type_set,
390 &g_av1_codec_arg_defs.use_intra_dct_only,
391 &g_av1_codec_arg_defs.use_inter_dct_only,
392 &g_av1_codec_arg_defs.use_intra_default_tx_only,
393 &g_av1_codec_arg_defs.quant_b_adapt,
394 &g_av1_codec_arg_defs.coeff_cost_upd_freq,
395 &g_av1_codec_arg_defs.mode_cost_upd_freq,
396 &g_av1_codec_arg_defs.mv_cost_upd_freq,
397 &g_av1_codec_arg_defs.frame_parallel_decoding,
398 &g_av1_codec_arg_defs.error_resilient_mode,
399 &g_av1_codec_arg_defs.aq_mode,
400 &g_av1_codec_arg_defs.deltaq_mode,
401 &g_av1_codec_arg_defs.deltaq_strength,
402 &g_av1_codec_arg_defs.deltalf_mode,
403 &g_av1_codec_arg_defs.frame_periodic_boost,
404 &g_av1_codec_arg_defs.noise_sens,
405 &g_av1_codec_arg_defs.tune_content,
406 &g_av1_codec_arg_defs.cdf_update_mode,
407 &g_av1_codec_arg_defs.input_color_primaries,
408 &g_av1_codec_arg_defs.input_transfer_characteristics,
409 &g_av1_codec_arg_defs.input_matrix_coefficients,
410 &g_av1_codec_arg_defs.input_chroma_sample_position,
411 &g_av1_codec_arg_defs.min_gf_interval,
412 &g_av1_codec_arg_defs.max_gf_interval,
413 &g_av1_codec_arg_defs.gf_min_pyr_height,
414 &g_av1_codec_arg_defs.gf_max_pyr_height,
415 &g_av1_codec_arg_defs.superblock_size,
416 &g_av1_codec_arg_defs.num_tg,
417 &g_av1_codec_arg_defs.mtu_size,
418 &g_av1_codec_arg_defs.timing_info,
419 &g_av1_codec_arg_defs.film_grain_test,
420 &g_av1_codec_arg_defs.film_grain_table,
421#if CONFIG_DENOISE
422 &g_av1_codec_arg_defs.denoise_noise_level,
423 &g_av1_codec_arg_defs.denoise_block_size,
424 &g_av1_codec_arg_defs.enable_dnl_denoising,
425#endif // CONFIG_DENOISE
426 &g_av1_codec_arg_defs.max_reference_frames,
427 &g_av1_codec_arg_defs.reduced_reference_set,
428 &g_av1_codec_arg_defs.enable_ref_frame_mvs,
429 &g_av1_codec_arg_defs.target_seq_level_idx,
430 &g_av1_codec_arg_defs.set_tier_mask,
431 &g_av1_codec_arg_defs.set_min_cr,
432 &g_av1_codec_arg_defs.vbr_corpus_complexity_lap,
433 &g_av1_codec_arg_defs.input_chroma_subsampling_x,
434 &g_av1_codec_arg_defs.input_chroma_subsampling_y,
435#if CONFIG_TUNE_VMAF
436 &g_av1_codec_arg_defs.vmaf_model_path,
437#endif
438 &g_av1_codec_arg_defs.dv_cost_upd_freq,
439 &g_av1_codec_arg_defs.partition_info_path,
440 &g_av1_codec_arg_defs.enable_directional_intra,
441 &g_av1_codec_arg_defs.enable_tx_size_search,
442 &g_av1_codec_arg_defs.loopfilter_control,
443 &g_av1_codec_arg_defs.auto_intra_tools_off,
444 NULL,
445};
446
447const arg_def_t *av1_key_val_args[] = {
448 &g_av1_codec_arg_defs.passes,
449 &g_av1_codec_arg_defs.two_pass_output,
450 &g_av1_codec_arg_defs.second_pass_log,
451 &g_av1_codec_arg_defs.fwd_kf_dist,
452 &g_av1_codec_arg_defs.strict_level_conformance,
453 &g_av1_codec_arg_defs.dist_metric,
454 NULL,
455};
456
457static const arg_def_t *no_args[] = { NULL };
458
459static void show_help(FILE *fout, int shorthelp) {
460 fprintf(fout, "Usage: %s <options> -o dst_filename src_filename\n",
461 exec_name);
462
463 if (shorthelp) {
464 fprintf(fout, "Use --help to see the full list of options.\n");
465 return;
466 }
467
468 fprintf(fout, "\nOptions:\n");
469 arg_show_usage(fout, main_args);
470 fprintf(fout, "\nEncoder Global Options:\n");
471 arg_show_usage(fout, global_args);
472 fprintf(fout, "\nRate Control Options:\n");
473 arg_show_usage(fout, rc_args);
474 fprintf(fout, "\nKeyframe Placement Options:\n");
475 arg_show_usage(fout, kf_args);
476#if CONFIG_AV1_ENCODER
477 fprintf(fout, "\nAV1 Specific Options:\n");
478 arg_show_usage(fout, av1_ctrl_args);
479 arg_show_usage(fout, av1_key_val_args);
480#endif
481 fprintf(fout,
482 "\nStream timebase (--timebase):\n"
483 " The desired precision of timestamps in the output, expressed\n"
484 " in fractional seconds. Default is 1/1000.\n");
485 fprintf(fout, "\nIncluded encoders:\n\n");
486
487 const int num_encoder = get_aom_encoder_count();
488 for (int i = 0; i < num_encoder; ++i) {
489 aom_codec_iface_t *encoder = get_aom_encoder_by_index(i);
490 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
491 fprintf(fout, " %-6s - %s %s\n", get_short_name_by_aom_encoder(encoder),
492 aom_codec_iface_name(encoder), defstr);
493 }
494 fprintf(fout, "\n ");
495 fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
496}
497
498void usage_exit(void) {
499 show_help(stderr, 1);
500 exit(EXIT_FAILURE);
501}
502
503#if CONFIG_AV1_ENCODER
504#define ARG_CTRL_CNT_MAX NELEMENTS(av1_arg_ctrl_map)
505#define ARG_KEY_VAL_CNT_MAX NELEMENTS(av1_key_val_args)
506#endif
507
508#if !CONFIG_WEBM_IO
509typedef int stereo_format_t;
510struct WebmOutputContext {
511 int debug;
512};
513#endif
514
515/* Per-stream configuration */
516struct stream_config {
517 struct aom_codec_enc_cfg cfg;
518 const char *out_fn;
519 const char *stats_fn;
520 stereo_format_t stereo_fmt;
521 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
522 int arg_ctrl_cnt;
523 const char *arg_key_vals[ARG_KEY_VAL_CNT_MAX][2];
524 int arg_key_val_cnt;
525 int write_webm;
526 const char *film_grain_filename;
527 int write_ivf;
528 // whether to use 16bit internal buffers
529 int use_16bit_internal;
530#if CONFIG_TUNE_VMAF
531 const char *vmaf_model_path;
532#endif
533 const char *partition_info_path;
534 aom_color_range_t color_range;
535 const char *two_pass_input;
536 const char *two_pass_output;
537 int two_pass_width;
538 int two_pass_height;
539};
540
541struct stream_state {
542 int index;
543 struct stream_state *next;
544 struct stream_config config;
545 FILE *file;
546 struct rate_hist *rate_hist;
547 struct WebmOutputContext webm_ctx;
548 uint64_t psnr_sse_total[2];
549 uint64_t psnr_samples_total[2];
550 double psnr_totals[2][4];
551 int psnr_count[2];
552 int counts[64];
553 aom_codec_ctx_t encoder;
554 unsigned int frames_out;
555 uint64_t cx_time;
556 size_t nbytes;
557 stats_io_t stats;
558 struct aom_image *img;
559 aom_codec_ctx_t decoder;
560 int mismatch_seen;
561 unsigned int chroma_subsampling_x;
562 unsigned int chroma_subsampling_y;
563 const char *orig_out_fn;
564 unsigned int orig_width;
565 unsigned int orig_height;
566 int orig_write_webm;
567 int orig_write_ivf;
568 char tmp_out_fn[1000];
569};
570
571static void validate_positive_rational(const char *msg,
572 struct aom_rational *rat) {
573 if (rat->den < 0) {
574 rat->num *= -1;
575 rat->den *= -1;
576 }
577
578 if (rat->num < 0) die("Error: %s must be positive\n", msg);
579
580 if (!rat->den) die("Error: %s has zero denominator\n", msg);
581}
582
583static void init_config(cfg_options_t *config) {
584 memset(config, 0, sizeof(cfg_options_t));
585 config->super_block_size = 0; // Dynamic
586 config->max_partition_size = 128;
587 config->min_partition_size = 4;
588 config->disable_trellis_quant = 3;
589}
590
591/* Parses global config arguments into the AvxEncoderConfig. Note that
592 * argv is modified and overwrites all parsed arguments.
593 */
594static void parse_global_config(struct AvxEncoderConfig *global, char ***argv) {
595 char **argi, **argj;
596 struct arg arg;
597 const int num_encoder = get_aom_encoder_count();
598 char **argv_local = (char **)*argv;
599 if (num_encoder < 1) die("Error: no valid encoder available\n");
600
601 /* Initialize default parameters */
602 memset(global, 0, sizeof(*global));
603 global->codec = get_aom_encoder_by_index(num_encoder - 1);
604 global->passes = 0;
605 global->color_type = I420;
606 global->csp = AOM_CSP_UNKNOWN;
607 global->show_psnr = 0;
608
609 int cfg_included = 0;
610 init_config(&global->encoder_config);
611
612 for (argi = argj = argv_local; (*argj = *argi); argi += arg.argv_step) {
613 arg.argv_step = 1;
614
615 if (arg_match(&arg, &g_av1_codec_arg_defs.use_cfg, argi)) {
616 if (!cfg_included) {
617 parse_cfg(arg.val, &global->encoder_config);
618 cfg_included = 1;
619 }
620 } else if (arg_match(&arg, &g_av1_codec_arg_defs.help, argi)) {
621 show_help(stdout, 0);
622 exit(EXIT_SUCCESS);
623 } else if (arg_match(&arg, &g_av1_codec_arg_defs.codecarg, argi)) {
624 global->codec = get_aom_encoder_by_short_name(arg.val);
625 if (!global->codec)
626 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
627 } else if (arg_match(&arg, &g_av1_codec_arg_defs.passes, argi)) {
628 global->passes = arg_parse_uint(&arg);
629
630 if (global->passes < 1 || global->passes > 3)
631 die("Error: Invalid number of passes (%d)\n", global->passes);
632 } else if (arg_match(&arg, &g_av1_codec_arg_defs.pass_arg, argi)) {
633 global->pass = arg_parse_uint(&arg);
634
635 if (global->pass < 1 || global->pass > 3)
636 die("Error: Invalid pass selected (%d)\n", global->pass);
637 } else if (arg_match(&arg,
638 &g_av1_codec_arg_defs.input_chroma_sample_position,
639 argi)) {
640 global->csp = arg_parse_enum(&arg);
641 /* Flag is used by later code as well, preserve it. */
642 argj++;
643 } else if (arg_match(&arg, &g_av1_codec_arg_defs.usage, argi)) {
644 global->usage = arg_parse_uint(&arg);
645 } else if (arg_match(&arg, &g_av1_codec_arg_defs.good_dl, argi)) {
646 global->usage = AOM_USAGE_GOOD_QUALITY; // Good quality usage
647 } else if (arg_match(&arg, &g_av1_codec_arg_defs.rt_dl, argi)) {
648 global->usage = AOM_USAGE_REALTIME; // Real-time usage
649 } else if (arg_match(&arg, &g_av1_codec_arg_defs.ai_dl, argi)) {
650 global->usage = AOM_USAGE_ALL_INTRA; // All intra usage
651 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_nv12, argi)) {
652 global->color_type = NV12;
653 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_yv12, argi)) {
654 global->color_type = YV12;
655 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_i420, argi)) {
656 global->color_type = I420;
657 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_i422, argi)) {
658 global->color_type = I422;
659 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_i444, argi)) {
660 global->color_type = I444;
661 } else if (arg_match(&arg, &g_av1_codec_arg_defs.quietarg, argi)) {
662 global->quiet = 1;
663 } else if (arg_match(&arg, &g_av1_codec_arg_defs.verbosearg, argi)) {
664 global->verbose = 1;
665 } else if (arg_match(&arg, &g_av1_codec_arg_defs.limit, argi)) {
666 global->limit = arg_parse_uint(&arg);
667 } else if (arg_match(&arg, &g_av1_codec_arg_defs.skip, argi)) {
668 global->skip_frames = arg_parse_uint(&arg);
669 } else if (arg_match(&arg, &g_av1_codec_arg_defs.psnrarg, argi)) {
670 if (arg.val)
671 global->show_psnr = arg_parse_int(&arg);
672 else
673 global->show_psnr = 1;
674 } else if (arg_match(&arg, &g_av1_codec_arg_defs.recontest, argi)) {
675 global->test_decode = arg_parse_enum_or_int(&arg);
676 } else if (arg_match(&arg, &g_av1_codec_arg_defs.framerate, argi)) {
677 global->framerate = arg_parse_rational(&arg);
678 validate_positive_rational(arg.name, &global->framerate);
679 global->have_framerate = 1;
680 } else if (arg_match(&arg, &g_av1_codec_arg_defs.debugmode, argi)) {
681 global->debug = 1;
682 } else if (arg_match(&arg, &g_av1_codec_arg_defs.q_hist_n, argi)) {
683 global->show_q_hist_buckets = arg_parse_uint(&arg);
684 } else if (arg_match(&arg, &g_av1_codec_arg_defs.rate_hist_n, argi)) {
685 global->show_rate_hist_buckets = arg_parse_uint(&arg);
686 } else if (arg_match(&arg, &g_av1_codec_arg_defs.disable_warnings, argi)) {
687 global->disable_warnings = 1;
688 } else if (arg_match(&arg, &g_av1_codec_arg_defs.disable_warning_prompt,
689 argi)) {
690 global->disable_warning_prompt = 1;
691 } else {
692 argj++;
693 }
694 }
695
696 if (global->pass) {
697 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
698 if (global->pass > global->passes) {
699 aom_tools_warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
700 global->pass);
701 global->passes = global->pass;
702 }
703 }
704 /* Validate global config */
705 if (global->passes == 0) {
706#if CONFIG_AV1_ENCODER
707 // Make default AV1 passes = 2 until there is a better quality 1-pass
708 // encoder
709 if (global->codec != NULL)
710 global->passes =
711 (strcmp(get_short_name_by_aom_encoder(global->codec), "av1") == 0 &&
712 global->usage != AOM_USAGE_REALTIME)
713 ? 2
714 : 1;
715#else
716 global->passes = 1;
717#endif
718 }
719
720 if (global->usage == AOM_USAGE_REALTIME && global->passes > 1) {
721 aom_tools_warn("Enforcing one-pass encoding in realtime mode\n");
722 if (global->pass > 1)
723 die("Error: Invalid --pass=%d for one-pass encoding\n", global->pass);
724 global->passes = 1;
725 }
726
727 if (global->usage == AOM_USAGE_ALL_INTRA && global->passes > 1) {
728 aom_tools_warn("Enforcing one-pass encoding in all intra mode\n");
729 global->passes = 1;
730 }
731}
732
733static void open_input_file(struct AvxInputContext *input,
735 /* Parse certain options from the input file, if possible */
736 input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
737 : set_binary_mode(stdin);
738
739 if (!input->file) fatal("Failed to open input file");
740
741 if (!fseeko(input->file, 0, SEEK_END)) {
742 /* Input file is seekable. Figure out how long it is, so we can get
743 * progress info.
744 */
745 input->length = ftello(input->file);
746 rewind(input->file);
747 }
748
749 /* Default to 1:1 pixel aspect ratio. */
750 input->pixel_aspect_ratio.numerator = 1;
751 input->pixel_aspect_ratio.denominator = 1;
752
753 /* For RAW input sources, these bytes will applied on the first frame
754 * in read_frame().
755 */
756 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
757 input->detect.position = 0;
758
759 if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
760 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4, csp,
761 input->only_i420) >= 0) {
762 input->file_type = FILE_TYPE_Y4M;
763 input->width = input->y4m.pic_w;
764 input->height = input->y4m.pic_h;
765 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
766 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
767 input->framerate.numerator = input->y4m.fps_n;
768 input->framerate.denominator = input->y4m.fps_d;
769 input->fmt = input->y4m.aom_fmt;
770 input->bit_depth = input->y4m.bit_depth;
771 input->color_range = input->y4m.color_range;
772 } else
773 fatal("Unsupported Y4M stream.");
774 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
775 fatal("IVF is not supported as input.");
776 } else {
777 input->file_type = FILE_TYPE_RAW;
778 }
779}
780
781static void close_input_file(struct AvxInputContext *input) {
782 fclose(input->file);
783 if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
784}
785
786static struct stream_state *new_stream(struct AvxEncoderConfig *global,
787 struct stream_state *prev) {
788 struct stream_state *stream;
789
790 stream = calloc(1, sizeof(*stream));
791 if (stream == NULL) {
792 fatal("Failed to allocate new stream.");
793 }
794
795 if (prev) {
796 memcpy(stream, prev, sizeof(*stream));
797 stream->index++;
798 prev->next = stream;
799 } else {
800 aom_codec_err_t res;
801
802 /* Populate encoder configuration */
803 res = aom_codec_enc_config_default(global->codec, &stream->config.cfg,
804 global->usage);
805 if (res) fatal("Failed to get config: %s\n", aom_codec_err_to_string(res));
806
807 /* Change the default timebase to a high enough value so that the
808 * encoder will always create strictly increasing timestamps.
809 */
810 stream->config.cfg.g_timebase.den = 1000;
811
812 /* Never use the library's default resolution, require it be parsed
813 * from the file or set on the command line.
814 */
815 stream->config.cfg.g_w = 0;
816 stream->config.cfg.g_h = 0;
817
818 /* Initialize remaining stream parameters */
819 stream->config.write_webm = 1;
820 stream->config.write_ivf = 0;
821
822#if CONFIG_WEBM_IO
823 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
824 stream->webm_ctx.last_pts_ns = -1;
825 stream->webm_ctx.writer = NULL;
826 stream->webm_ctx.segment = NULL;
827#endif
828
829 /* Allows removal of the application version from the EBML tags */
830 stream->webm_ctx.debug = global->debug;
831 memcpy(&stream->config.cfg.encoder_cfg, &global->encoder_config,
832 sizeof(stream->config.cfg.encoder_cfg));
833 }
834
835 /* Output files must be specified for each stream */
836 stream->config.out_fn = NULL;
837 stream->config.two_pass_input = NULL;
838 stream->config.two_pass_output = NULL;
839 stream->config.two_pass_width = 0;
840 stream->config.two_pass_height = 0;
841
842 stream->next = NULL;
843 return stream;
844}
845
846static void set_config_arg_ctrls(struct stream_config *config, int key,
847 const struct arg *arg) {
848 int j;
849 if (key == AV1E_SET_FILM_GRAIN_TABLE) {
850 config->film_grain_filename = arg->val;
851 return;
852 }
853
854 // For target level, the settings should accumulate rather than overwrite,
855 // so we simply append it.
857 j = config->arg_ctrl_cnt;
858 assert(j < ARG_CTRL_CNT_MAX);
859 config->arg_ctrls[j][0] = key;
860 config->arg_ctrls[j][1] = arg_parse_enum_or_int(arg);
861 ++config->arg_ctrl_cnt;
862 return;
863 }
864
865 /* Point either to the next free element or the first instance of this
866 * control.
867 */
868 for (j = 0; j < config->arg_ctrl_cnt; j++)
869 if (config->arg_ctrls[j][0] == key) break;
870
871 /* Update/insert */
872 assert(j < ARG_CTRL_CNT_MAX);
873 config->arg_ctrls[j][0] = key;
874 config->arg_ctrls[j][1] = arg_parse_enum_or_int(arg);
875
876 if (key == AOME_SET_ENABLEAUTOALTREF && config->arg_ctrls[j][1] > 1) {
877 aom_tools_warn(
878 "auto-alt-ref > 1 is deprecated... setting auto-alt-ref=1\n");
879 config->arg_ctrls[j][1] = 1;
880 }
881
882 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
883}
884
885static void set_config_arg_key_vals(struct stream_config *config,
886 const char *name, const struct arg *arg) {
887 int j;
888 const char *val = arg->val;
889 // For target level, the settings should accumulate rather than overwrite,
890 // so we simply append it.
891 if (strcmp(name, "target-seq-level-idx") == 0) {
892 j = config->arg_key_val_cnt;
893 assert(j < ARG_KEY_VAL_CNT_MAX);
894 config->arg_key_vals[j][0] = name;
895 config->arg_key_vals[j][1] = val;
896 ++config->arg_key_val_cnt;
897 return;
898 }
899
900 /* Point either to the next free element or the first instance of this
901 * option.
902 */
903 for (j = 0; j < config->arg_key_val_cnt; j++)
904 if (strcmp(name, config->arg_key_vals[j][0]) == 0) break;
905
906 /* Update/insert */
907 assert(j < ARG_KEY_VAL_CNT_MAX);
908 config->arg_key_vals[j][0] = name;
909 config->arg_key_vals[j][1] = val;
910
911 if (strcmp(name, g_av1_codec_arg_defs.auto_altref.long_name) == 0) {
912 int auto_altref = arg_parse_int(arg);
913 if (auto_altref > 1) {
914 aom_tools_warn(
915 "auto-alt-ref > 1 is deprecated... setting auto-alt-ref=1\n");
916 config->arg_key_vals[j][1] = "1";
917 }
918 }
919
920 if (j == config->arg_key_val_cnt) config->arg_key_val_cnt++;
921}
922
923static int parse_stream_params(struct AvxEncoderConfig *global,
924 struct stream_state *stream, char **argv) {
925 char **argi, **argj;
926 struct arg arg;
927 static const arg_def_t **ctrl_args = no_args;
928 static const arg_def_t **key_val_args = no_args;
929 static const int *ctrl_args_map = NULL;
930 struct stream_config *config = &stream->config;
931 int eos_mark_found = 0;
932 int webm_forced = 0;
933
934 // Handle codec specific options
935 if (0) {
936#if CONFIG_AV1_ENCODER
937 } else if (strcmp(get_short_name_by_aom_encoder(global->codec), "av1") == 0) {
938 // TODO(jingning): Reuse AV1 specific encoder configuration parameters.
939 // Consider to expand this set for AV1 encoder control.
940#if __STDC_VERSION__ >= 201112L
941 _Static_assert(NELEMENTS(av1_ctrl_args) == NELEMENTS(av1_arg_ctrl_map),
942 "The av1_ctrl_args and av1_arg_ctrl_map arrays must be of "
943 "the same size.");
944#else
945 assert(NELEMENTS(av1_ctrl_args) == NELEMENTS(av1_arg_ctrl_map));
946#endif
947 ctrl_args = av1_ctrl_args;
948 ctrl_args_map = av1_arg_ctrl_map;
949 key_val_args = av1_key_val_args;
950#endif
951 }
952
953 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
954 arg.argv_step = 1;
955
956 /* Once we've found an end-of-stream marker (--) we want to continue
957 * shifting arguments but not consuming them.
958 */
959 if (eos_mark_found) {
960 argj++;
961 continue;
962 } else if (!strcmp(*argj, "--")) {
963 eos_mark_found = 1;
964 continue;
965 }
966
967 if (arg_match(&arg, &g_av1_codec_arg_defs.outputfile, argi)) {
968 config->out_fn = arg.val;
969 if (!webm_forced) {
970 const size_t out_fn_len = strlen(config->out_fn);
971 if (out_fn_len >= 4 &&
972 !strcmp(config->out_fn + out_fn_len - 4, ".ivf")) {
973 config->write_webm = 0;
974 config->write_ivf = 1;
975 } else if (out_fn_len >= 4 &&
976 !strcmp(config->out_fn + out_fn_len - 4, ".obu")) {
977 config->write_webm = 0;
978 config->write_ivf = 0;
979 }
980 }
981 } else if (arg_match(&arg, &g_av1_codec_arg_defs.fpf_name, argi)) {
982 config->stats_fn = arg.val;
983 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_webm, argi)) {
984#if CONFIG_WEBM_IO
985 config->write_webm = 1;
986 webm_forced = 1;
987#else
988 die("Error: --webm specified but webm is disabled.");
989#endif
990 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_ivf, argi)) {
991 config->write_webm = 0;
992 config->write_ivf = 1;
993 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_obu, argi)) {
994 config->write_webm = 0;
995 config->write_ivf = 0;
996 } else if (arg_match(&arg, &g_av1_codec_arg_defs.threads, argi)) {
997 config->cfg.g_threads = arg_parse_uint(&arg);
998 } else if (arg_match(&arg, &g_av1_codec_arg_defs.profile, argi)) {
999 config->cfg.g_profile = arg_parse_uint(&arg);
1000 } else if (arg_match(&arg, &g_av1_codec_arg_defs.width, argi)) {
1001 config->cfg.g_w = arg_parse_uint(&arg);
1002 } else if (arg_match(&arg, &g_av1_codec_arg_defs.height, argi)) {
1003 config->cfg.g_h = arg_parse_uint(&arg);
1004 } else if (arg_match(&arg, &g_av1_codec_arg_defs.forced_max_frame_width,
1005 argi)) {
1006 config->cfg.g_forced_max_frame_width = arg_parse_uint(&arg);
1007 } else if (arg_match(&arg, &g_av1_codec_arg_defs.forced_max_frame_height,
1008 argi)) {
1009 config->cfg.g_forced_max_frame_height = arg_parse_uint(&arg);
1010 } else if (arg_match(&arg, &g_av1_codec_arg_defs.bitdeptharg, argi)) {
1011 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1012 } else if (arg_match(&arg, &g_av1_codec_arg_defs.inbitdeptharg, argi)) {
1013 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1014 } else if (arg_match(&arg, &g_av1_codec_arg_defs.input_chroma_subsampling_x,
1015 argi)) {
1016 stream->chroma_subsampling_x = arg_parse_uint(&arg);
1017 } else if (arg_match(&arg, &g_av1_codec_arg_defs.input_chroma_subsampling_y,
1018 argi)) {
1019 stream->chroma_subsampling_y = arg_parse_uint(&arg);
1020#if CONFIG_WEBM_IO
1021 } else if (arg_match(&arg, &g_av1_codec_arg_defs.stereo_mode, argi)) {
1022 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1023#endif
1024 } else if (arg_match(&arg, &g_av1_codec_arg_defs.timebase, argi)) {
1025 config->cfg.g_timebase = arg_parse_rational(&arg);
1026 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1027 } else if (arg_match(&arg, &g_av1_codec_arg_defs.global_error_resilient,
1028 argi)) {
1029 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1030 } else if (arg_match(&arg, &g_av1_codec_arg_defs.lag_in_frames, argi)) {
1031 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1032 } else if (arg_match(&arg, &g_av1_codec_arg_defs.large_scale_tile, argi)) {
1033 config->cfg.large_scale_tile = arg_parse_uint(&arg);
1034 if (config->cfg.large_scale_tile) {
1035 global->codec = get_aom_encoder_by_short_name("av1");
1036 }
1037 } else if (arg_match(&arg, &g_av1_codec_arg_defs.monochrome, argi)) {
1038 config->cfg.monochrome = 1;
1039 } else if (arg_match(&arg, &g_av1_codec_arg_defs.full_still_picture_hdr,
1040 argi)) {
1041 config->cfg.full_still_picture_hdr = 1;
1042 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_16bit_internal,
1043 argi)) {
1044 config->use_16bit_internal = CONFIG_AV1_HIGHBITDEPTH;
1045 if (!config->use_16bit_internal) {
1046 aom_tools_warn("%s option ignored with CONFIG_AV1_HIGHBITDEPTH=0.\n",
1047 arg.name);
1048 }
1049 } else if (arg_match(&arg, &g_av1_codec_arg_defs.dropframe_thresh, argi)) {
1050 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1051 } else if (arg_match(&arg, &g_av1_codec_arg_defs.resize_mode, argi)) {
1052 config->cfg.rc_resize_mode = arg_parse_uint(&arg);
1053 } else if (arg_match(&arg, &g_av1_codec_arg_defs.resize_denominator,
1054 argi)) {
1055 config->cfg.rc_resize_denominator = arg_parse_uint(&arg);
1056 } else if (arg_match(&arg, &g_av1_codec_arg_defs.resize_kf_denominator,
1057 argi)) {
1058 config->cfg.rc_resize_kf_denominator = arg_parse_uint(&arg);
1059 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_mode, argi)) {
1060 config->cfg.rc_superres_mode = arg_parse_uint(&arg);
1061 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_denominator,
1062 argi)) {
1063 config->cfg.rc_superres_denominator = arg_parse_uint(&arg);
1064 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_kf_denominator,
1065 argi)) {
1066 config->cfg.rc_superres_kf_denominator = arg_parse_uint(&arg);
1067 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_qthresh, argi)) {
1068 config->cfg.rc_superres_qthresh = arg_parse_uint(&arg);
1069 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_kf_qthresh,
1070 argi)) {
1071 config->cfg.rc_superres_kf_qthresh = arg_parse_uint(&arg);
1072 } else if (arg_match(&arg, &g_av1_codec_arg_defs.end_usage, argi)) {
1073 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1074 } else if (arg_match(&arg, &g_av1_codec_arg_defs.target_bitrate, argi)) {
1075 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1076 } else if (arg_match(&arg, &g_av1_codec_arg_defs.min_quantizer, argi)) {
1077 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1078 } else if (arg_match(&arg, &g_av1_codec_arg_defs.max_quantizer, argi)) {
1079 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1080 } else if (arg_match(&arg, &g_av1_codec_arg_defs.undershoot_pct, argi)) {
1081 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1082 } else if (arg_match(&arg, &g_av1_codec_arg_defs.overshoot_pct, argi)) {
1083 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1084 } else if (arg_match(&arg, &g_av1_codec_arg_defs.buf_sz, argi)) {
1085 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1086 } else if (arg_match(&arg, &g_av1_codec_arg_defs.buf_initial_sz, argi)) {
1087 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1088 } else if (arg_match(&arg, &g_av1_codec_arg_defs.buf_optimal_sz, argi)) {
1089 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1090 } else if (arg_match(&arg, &g_av1_codec_arg_defs.bias_pct, argi)) {
1091 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1092 if (global->passes < 2)
1093 aom_tools_warn("option %s ignored in one-pass mode.\n", arg.name);
1094 } else if (arg_match(&arg, &g_av1_codec_arg_defs.minsection_pct, argi)) {
1095 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1096
1097 if (global->passes < 2)
1098 aom_tools_warn("option %s ignored in one-pass mode.\n", arg.name);
1099 } else if (arg_match(&arg, &g_av1_codec_arg_defs.maxsection_pct, argi)) {
1100 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1101
1102 if (global->passes < 2)
1103 aom_tools_warn("option %s ignored in one-pass mode.\n", arg.name);
1104 } else if (arg_match(&arg, &g_av1_codec_arg_defs.fwd_kf_enabled, argi)) {
1105 config->cfg.fwd_kf_enabled = arg_parse_uint(&arg);
1106 } else if (arg_match(&arg, &g_av1_codec_arg_defs.kf_min_dist, argi)) {
1107 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1108 } else if (arg_match(&arg, &g_av1_codec_arg_defs.kf_max_dist, argi)) {
1109 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1110 } else if (arg_match(&arg, &g_av1_codec_arg_defs.kf_disabled, argi)) {
1111 config->cfg.kf_mode = AOM_KF_DISABLED;
1112 } else if (arg_match(&arg, &g_av1_codec_arg_defs.sframe_dist, argi)) {
1113 config->cfg.sframe_dist = arg_parse_uint(&arg);
1114 } else if (arg_match(&arg, &g_av1_codec_arg_defs.sframe_mode, argi)) {
1115 config->cfg.sframe_mode = arg_parse_uint(&arg);
1116 } else if (arg_match(&arg, &g_av1_codec_arg_defs.save_as_annexb, argi)) {
1117 config->cfg.save_as_annexb = arg_parse_uint(&arg);
1118 } else if (arg_match(&arg, &g_av1_codec_arg_defs.tile_width, argi)) {
1119 config->cfg.tile_width_count =
1120 arg_parse_list(&arg, config->cfg.tile_widths, MAX_TILE_WIDTHS);
1121 } else if (arg_match(&arg, &g_av1_codec_arg_defs.tile_height, argi)) {
1122 config->cfg.tile_height_count =
1123 arg_parse_list(&arg, config->cfg.tile_heights, MAX_TILE_HEIGHTS);
1124#if CONFIG_TUNE_VMAF
1125 } else if (arg_match(&arg, &g_av1_codec_arg_defs.vmaf_model_path, argi)) {
1126 config->vmaf_model_path = arg.val;
1127#endif
1128 } else if (arg_match(&arg, &g_av1_codec_arg_defs.partition_info_path,
1129 argi)) {
1130 config->partition_info_path = arg.val;
1131 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_fixed_qp_offsets,
1132 argi)) {
1133 config->cfg.use_fixed_qp_offsets = arg_parse_uint(&arg);
1134 } else if (arg_match(&arg, &g_av1_codec_arg_defs.fixed_qp_offsets, argi)) {
1135 config->cfg.use_fixed_qp_offsets = 1;
1136 } else if (global->usage == AOM_USAGE_REALTIME &&
1137 arg_match(&arg, &g_av1_codec_arg_defs.enable_restoration,
1138 argi)) {
1139 if (arg_parse_uint(&arg) == 1) {
1140 aom_tools_warn("non-zero %s option ignored in realtime mode.\n",
1141 arg.name);
1142 }
1143 } else if (arg_match(&arg, &g_av1_codec_arg_defs.two_pass_input, argi)) {
1144 config->two_pass_input = arg.val;
1145 } else if (arg_match(&arg, &g_av1_codec_arg_defs.two_pass_output, argi)) {
1146 config->two_pass_output = arg.val;
1147 } else if (arg_match(&arg, &g_av1_codec_arg_defs.two_pass_width, argi)) {
1148 config->two_pass_width = arg_parse_int(&arg);
1149 } else if (arg_match(&arg, &g_av1_codec_arg_defs.two_pass_height, argi)) {
1150 config->two_pass_height = arg_parse_int(&arg);
1151 } else {
1152 int i, match = 0;
1153 // check if the control ID API supports this arg
1154 if (ctrl_args_map) {
1155 for (i = 0; ctrl_args[i]; i++) {
1156 if (arg_match(&arg, ctrl_args[i], argi)) {
1157 match = 1;
1158 set_config_arg_ctrls(config, ctrl_args_map[i], &arg);
1159 break;
1160 }
1161 }
1162 }
1163 if (!match) {
1164 // check if the key & value API supports this arg
1165 for (i = 0; key_val_args[i]; i++) {
1166 if (arg_match(&arg, key_val_args[i], argi)) {
1167 match = 1;
1168 set_config_arg_key_vals(config, key_val_args[i]->long_name, &arg);
1169 break;
1170 }
1171 }
1172 }
1173 if (!match) argj++;
1174 }
1175 }
1176 config->use_16bit_internal |= config->cfg.g_bit_depth > AOM_BITS_8;
1177
1178 if (global->usage == AOM_USAGE_REALTIME && config->cfg.g_lag_in_frames != 0) {
1179 aom_tools_warn("non-zero lag-in-frames option ignored in realtime mode.\n");
1180 config->cfg.g_lag_in_frames = 0;
1181 }
1182
1183 if (global->usage == AOM_USAGE_ALL_INTRA) {
1184 if (config->cfg.g_lag_in_frames != 0) {
1185 aom_tools_warn(
1186 "non-zero lag-in-frames option ignored in all intra mode.\n");
1187 config->cfg.g_lag_in_frames = 0;
1188 }
1189 if (config->cfg.kf_max_dist != 0) {
1190 aom_tools_warn(
1191 "non-zero max key frame distance option ignored in all intra "
1192 "mode.\n");
1193 config->cfg.kf_max_dist = 0;
1194 }
1195 }
1196
1197 // set the passes field using key & val API
1198 if (config->arg_key_val_cnt >= ARG_KEY_VAL_CNT_MAX) {
1199 die("Not enough buffer for the key & value API.");
1200 }
1201 config->arg_key_vals[config->arg_key_val_cnt][0] = "passes";
1202 switch (global->passes) {
1203 case 0: config->arg_key_vals[config->arg_key_val_cnt][1] = "0"; break;
1204 case 1: config->arg_key_vals[config->arg_key_val_cnt][1] = "1"; break;
1205 case 2: config->arg_key_vals[config->arg_key_val_cnt][1] = "2"; break;
1206 case 3: config->arg_key_vals[config->arg_key_val_cnt][1] = "3"; break;
1207 default: die("Invalid value of --passes.");
1208 }
1209 config->arg_key_val_cnt++;
1210
1211 // set the two_pass_output field
1212 if (!config->two_pass_output && global->passes == 3) {
1213 // If not specified, set the name of two_pass_output file here.
1214 snprintf(stream->tmp_out_fn, sizeof(stream->tmp_out_fn),
1215 "%.980s_pass2_%d.ivf", stream->config.out_fn, stream->index);
1216 stream->config.two_pass_output = stream->tmp_out_fn;
1217 }
1218 if (config->two_pass_output) {
1219 config->arg_key_vals[config->arg_key_val_cnt][0] = "two-pass-output";
1220 config->arg_key_vals[config->arg_key_val_cnt][1] = config->two_pass_output;
1221 config->arg_key_val_cnt++;
1222 }
1223
1224 return eos_mark_found;
1225}
1226
1227#define FOREACH_STREAM(iterator, list) \
1228 for (struct stream_state *iterator = list; iterator; \
1229 iterator = iterator->next)
1230
1231static void validate_stream_config(const struct stream_state *stream,
1232 const struct AvxEncoderConfig *global) {
1233 const struct stream_state *streami;
1234 (void)global;
1235
1236 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1237 fatal(
1238 "Stream %d: Specify stream dimensions with --width (-w) "
1239 " and --height (-h)",
1240 stream->index);
1241
1242 /* Even if bit depth is set on the command line flag to be lower,
1243 * it is upgraded to at least match the input bit depth.
1244 */
1245 assert(stream->config.cfg.g_input_bit_depth <=
1246 (unsigned int)stream->config.cfg.g_bit_depth);
1247
1248 for (streami = stream; streami; streami = streami->next) {
1249 /* All streams require output files */
1250 if (!streami->config.out_fn)
1251 fatal("Stream %d: Output file is required (specify with -o)",
1252 streami->index);
1253
1254 /* Check for two streams outputting to the same file */
1255 if (streami != stream) {
1256 const char *a = stream->config.out_fn;
1257 const char *b = streami->config.out_fn;
1258 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1259 fatal("Stream %d: duplicate output file (from stream %d)",
1260 streami->index, stream->index);
1261 }
1262
1263 /* Check for two streams sharing a stats file. */
1264 if (streami != stream) {
1265 const char *a = stream->config.stats_fn;
1266 const char *b = streami->config.stats_fn;
1267 if (a && b && !strcmp(a, b))
1268 fatal("Stream %d: duplicate stats file (from stream %d)",
1269 streami->index, stream->index);
1270 }
1271 }
1272}
1273
1274static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1275 unsigned int h) {
1276 if (!stream->config.cfg.g_w) {
1277 if (!stream->config.cfg.g_h)
1278 stream->config.cfg.g_w = w;
1279 else
1280 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1281 }
1282 if (!stream->config.cfg.g_h) {
1283 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1284 }
1285}
1286
1287static const char *file_type_to_string(enum VideoFileType t) {
1288 switch (t) {
1289 case FILE_TYPE_RAW: return "RAW";
1290 case FILE_TYPE_Y4M: return "Y4M";
1291 default: return "Other";
1292 }
1293}
1294
1295static const char *image_format_to_string(aom_img_fmt_t f) {
1296 switch (f) {
1297 case AOM_IMG_FMT_I420: return "I420";
1298 case AOM_IMG_FMT_I422: return "I422";
1299 case AOM_IMG_FMT_I444: return "I444";
1300 case AOM_IMG_FMT_YV12: return "YV12";
1301 case AOM_IMG_FMT_NV12: return "NV12";
1302 case AOM_IMG_FMT_YV1216: return "YV1216";
1303 case AOM_IMG_FMT_I42016: return "I42016";
1304 case AOM_IMG_FMT_I42216: return "I42216";
1305 case AOM_IMG_FMT_I44416: return "I44416";
1306 default: return "Other";
1307 }
1308}
1309
1310static void show_stream_config(struct stream_state *stream,
1311 struct AvxEncoderConfig *global,
1312 struct AvxInputContext *input) {
1313#define SHOW(field) \
1314 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1315
1316 if (stream->index == 0) {
1317 fprintf(stderr, "Codec: %s\n", aom_codec_iface_name(global->codec));
1318 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1319 input->filename, file_type_to_string(input->file_type),
1320 image_format_to_string(input->fmt));
1321 }
1322 if (stream->next || stream->index)
1323 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1324 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1325 fprintf(stderr, "Coding path: %s\n",
1326 stream->config.use_16bit_internal ? "HBD" : "LBD");
1327 fprintf(stderr, "Encoder parameters:\n");
1328
1329 SHOW(g_usage);
1330 SHOW(g_threads);
1331 SHOW(g_profile);
1332 SHOW(g_w);
1333 SHOW(g_h);
1334 SHOW(g_bit_depth);
1335 SHOW(g_input_bit_depth);
1336 SHOW(g_timebase.num);
1337 SHOW(g_timebase.den);
1338 SHOW(g_error_resilient);
1339 SHOW(g_pass);
1340 SHOW(g_lag_in_frames);
1341 SHOW(large_scale_tile);
1342 SHOW(rc_dropframe_thresh);
1343 SHOW(rc_resize_mode);
1344 SHOW(rc_resize_denominator);
1345 SHOW(rc_resize_kf_denominator);
1346 SHOW(rc_superres_mode);
1347 SHOW(rc_superres_denominator);
1348 SHOW(rc_superres_kf_denominator);
1349 SHOW(rc_superres_qthresh);
1350 SHOW(rc_superres_kf_qthresh);
1351 SHOW(rc_end_usage);
1352 SHOW(rc_target_bitrate);
1353 SHOW(rc_min_quantizer);
1354 SHOW(rc_max_quantizer);
1355 SHOW(rc_undershoot_pct);
1356 SHOW(rc_overshoot_pct);
1357 SHOW(rc_buf_sz);
1358 SHOW(rc_buf_initial_sz);
1359 SHOW(rc_buf_optimal_sz);
1360 SHOW(rc_2pass_vbr_bias_pct);
1361 SHOW(rc_2pass_vbr_minsection_pct);
1362 SHOW(rc_2pass_vbr_maxsection_pct);
1363 SHOW(fwd_kf_enabled);
1364 SHOW(kf_mode);
1365 SHOW(kf_min_dist);
1366 SHOW(kf_max_dist);
1367
1368#define SHOW_PARAMS(field) \
1369 fprintf(stderr, " %-28s = %d\n", #field, \
1370 stream->config.cfg.encoder_cfg.field)
1371 if (global->encoder_config.init_by_cfg_file) {
1372 SHOW_PARAMS(super_block_size);
1373 SHOW_PARAMS(max_partition_size);
1374 SHOW_PARAMS(min_partition_size);
1375 SHOW_PARAMS(disable_ab_partition_type);
1376 SHOW_PARAMS(disable_rect_partition_type);
1377 SHOW_PARAMS(disable_1to4_partition_type);
1378 SHOW_PARAMS(disable_flip_idtx);
1379 SHOW_PARAMS(disable_cdef);
1380 SHOW_PARAMS(disable_lr);
1381 SHOW_PARAMS(disable_obmc);
1382 SHOW_PARAMS(disable_warp_motion);
1383 SHOW_PARAMS(disable_global_motion);
1384 SHOW_PARAMS(disable_dist_wtd_comp);
1385 SHOW_PARAMS(disable_diff_wtd_comp);
1386 SHOW_PARAMS(disable_inter_intra_comp);
1387 SHOW_PARAMS(disable_masked_comp);
1388 SHOW_PARAMS(disable_one_sided_comp);
1389 SHOW_PARAMS(disable_palette);
1390 SHOW_PARAMS(disable_intrabc);
1391 SHOW_PARAMS(disable_cfl);
1392 SHOW_PARAMS(disable_smooth_intra);
1393 SHOW_PARAMS(disable_filter_intra);
1394 SHOW_PARAMS(disable_dual_filter);
1395 SHOW_PARAMS(disable_intra_angle_delta);
1396 SHOW_PARAMS(disable_intra_edge_filter);
1397 SHOW_PARAMS(disable_tx_64x64);
1398 SHOW_PARAMS(disable_smooth_inter_intra);
1399 SHOW_PARAMS(disable_inter_inter_wedge);
1400 SHOW_PARAMS(disable_inter_intra_wedge);
1401 SHOW_PARAMS(disable_paeth_intra);
1402 SHOW_PARAMS(disable_trellis_quant);
1403 SHOW_PARAMS(disable_ref_frame_mv);
1404 SHOW_PARAMS(reduced_reference_set);
1405 SHOW_PARAMS(reduced_tx_type_set);
1406 }
1407}
1408
1409static void open_output_file(struct stream_state *stream,
1410 struct AvxEncoderConfig *global,
1411 const struct AvxRational *pixel_aspect_ratio,
1412 const char *encoder_settings) {
1413 const char *fn = stream->config.out_fn;
1414 const struct aom_codec_enc_cfg *const cfg = &stream->config.cfg;
1415
1416 if (cfg->g_pass == AOM_RC_FIRST_PASS) return;
1417
1418 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1419
1420 if (!stream->file) fatal("Failed to open output file");
1421
1422 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1423 fatal("WebM output to pipes not supported.");
1424
1425#if CONFIG_WEBM_IO
1426 if (stream->config.write_webm) {
1427 stream->webm_ctx.stream = stream->file;
1428 if (write_webm_file_header(&stream->webm_ctx, &stream->encoder, cfg,
1429 stream->config.stereo_fmt,
1430 get_fourcc_by_aom_encoder(global->codec),
1431 pixel_aspect_ratio, encoder_settings) != 0) {
1432 fatal("WebM writer initialization failed.");
1433 }
1434 }
1435#else
1436 (void)pixel_aspect_ratio;
1437 (void)encoder_settings;
1438#endif
1439
1440 if (!stream->config.write_webm && stream->config.write_ivf) {
1441 ivf_write_file_header(stream->file, cfg,
1442 get_fourcc_by_aom_encoder(global->codec), 0);
1443 }
1444}
1445
1446static void close_output_file(struct stream_state *stream,
1447 unsigned int fourcc) {
1448 const struct aom_codec_enc_cfg *const cfg = &stream->config.cfg;
1449
1450 if (cfg->g_pass == AOM_RC_FIRST_PASS) return;
1451
1452#if CONFIG_WEBM_IO
1453 if (stream->config.write_webm) {
1454 if (write_webm_file_footer(&stream->webm_ctx) != 0) {
1455 fatal("WebM writer finalization failed.");
1456 }
1457 }
1458#endif
1459
1460 if (!stream->config.write_webm && stream->config.write_ivf) {
1461 if (!fseek(stream->file, 0, SEEK_SET))
1462 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1463 stream->frames_out);
1464 }
1465
1466 fclose(stream->file);
1467}
1468
1469static void setup_pass(struct stream_state *stream,
1470 struct AvxEncoderConfig *global, int pass) {
1471 if (stream->config.stats_fn) {
1472 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1473 fatal("Failed to open statistics store");
1474 } else {
1475 if (!stats_open_mem(&stream->stats, pass))
1476 fatal("Failed to open statistics store");
1477 }
1478
1479 if (global->passes == 1) {
1480 stream->config.cfg.g_pass = AOM_RC_ONE_PASS;
1481 } else {
1482 switch (pass) {
1483 case 0: stream->config.cfg.g_pass = AOM_RC_FIRST_PASS; break;
1484 case 1: stream->config.cfg.g_pass = AOM_RC_SECOND_PASS; break;
1485 case 2: stream->config.cfg.g_pass = AOM_RC_THIRD_PASS; break;
1486 default: fatal("Failed to set pass");
1487 }
1488 }
1489
1490 if (pass) {
1491 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1492 }
1493
1494 stream->cx_time = 0;
1495 stream->nbytes = 0;
1496 stream->frames_out = 0;
1497}
1498
1499static void initialize_encoder(struct stream_state *stream,
1500 struct AvxEncoderConfig *global) {
1501 int i;
1502 int flags = 0;
1503
1504 flags |= (global->show_psnr >= 1) ? AOM_CODEC_USE_PSNR : 0;
1505 flags |= stream->config.use_16bit_internal ? AOM_CODEC_USE_HIGHBITDEPTH : 0;
1506
1507 /* Construct Encoder Context */
1508 aom_codec_enc_init(&stream->encoder, global->codec, &stream->config.cfg,
1509 flags);
1510 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1511
1512 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1513 int ctrl = stream->config.arg_ctrls[i][0];
1514 int value = stream->config.arg_ctrls[i][1];
1515 if (aom_codec_control(&stream->encoder, ctrl, value))
1516 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1517
1518 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1519 }
1520
1521 for (i = 0; i < stream->config.arg_key_val_cnt; i++) {
1522 const char *name = stream->config.arg_key_vals[i][0];
1523 const char *val = stream->config.arg_key_vals[i][1];
1524 if (aom_codec_set_option(&stream->encoder, name, val))
1525 fprintf(stderr, "Error: Tried to set option %s = %s\n", name, val);
1526
1527 ctx_exit_on_error(&stream->encoder, "Failed to set codec option");
1528 }
1529
1530#if CONFIG_TUNE_VMAF
1531 if (stream->config.vmaf_model_path) {
1533 stream->config.vmaf_model_path);
1534 }
1535#endif
1536 if (stream->config.partition_info_path) {
1537 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
1539 stream->config.partition_info_path);
1540 }
1541
1542 if (stream->config.film_grain_filename) {
1544 stream->config.film_grain_filename);
1545 }
1547 stream->config.color_range);
1548
1549#if CONFIG_AV1_DECODER
1550 if (global->test_decode != TEST_DECODE_OFF) {
1551 aom_codec_iface_t *decoder = get_aom_decoder_by_short_name(
1552 get_short_name_by_aom_encoder(global->codec));
1553 aom_codec_dec_cfg_t cfg = { 0, 0, 0, !stream->config.use_16bit_internal };
1554 aom_codec_dec_init(&stream->decoder, decoder, &cfg, 0);
1555
1556 if (strcmp(get_short_name_by_aom_encoder(global->codec), "av1") == 0) {
1558 stream->config.cfg.large_scale_tile);
1559 ctx_exit_on_error(&stream->decoder, "Failed to set decode_tile_mode");
1560
1562 stream->config.cfg.save_as_annexb);
1563 ctx_exit_on_error(&stream->decoder, "Failed to set is_annexb");
1564
1566 -1);
1567 ctx_exit_on_error(&stream->decoder, "Failed to set decode_tile_row");
1568
1569 AOM_CODEC_CONTROL_TYPECHECKED(&stream->decoder, AV1_SET_DECODE_TILE_COL,
1570 -1);
1571 ctx_exit_on_error(&stream->decoder, "Failed to set decode_tile_col");
1572 }
1573 }
1574#endif
1575}
1576
1577// Convert the input image 'img' to a monochrome image. The Y plane of the
1578// output image is a shallow copy of the Y plane of the input image, therefore
1579// the input image must remain valid for the lifetime of the output image. The U
1580// and V planes of the output image are set to null pointers. The output image
1581// format is AOM_IMG_FMT_I420 because libaom does not have AOM_IMG_FMT_I400.
1582static void convert_image_to_monochrome(const struct aom_image *img,
1583 struct aom_image *monochrome_img) {
1584 *monochrome_img = *img;
1585 monochrome_img->fmt = AOM_IMG_FMT_I420;
1586 if (img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
1587 monochrome_img->fmt |= AOM_IMG_FMT_HIGHBITDEPTH;
1588 }
1589 monochrome_img->monochrome = 1;
1590 monochrome_img->csp = AOM_CSP_UNKNOWN;
1591 monochrome_img->x_chroma_shift = 1;
1592 monochrome_img->y_chroma_shift = 1;
1593 monochrome_img->planes[AOM_PLANE_U] = NULL;
1594 monochrome_img->planes[AOM_PLANE_V] = NULL;
1595 monochrome_img->stride[AOM_PLANE_U] = 0;
1596 monochrome_img->stride[AOM_PLANE_V] = 0;
1597 monochrome_img->sz = 0;
1598 monochrome_img->bps = (img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 16 : 8;
1599 monochrome_img->img_data = NULL;
1600 monochrome_img->img_data_owner = 0;
1601 monochrome_img->self_allocd = 0;
1602}
1603
1604static void encode_frame(struct stream_state *stream,
1605 struct AvxEncoderConfig *global, struct aom_image *img,
1606 unsigned int frames_in) {
1607 aom_codec_pts_t frame_start, next_frame_start;
1608 struct aom_codec_enc_cfg *cfg = &stream->config.cfg;
1609 struct aom_usec_timer timer;
1610
1611 frame_start =
1612 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1613 cfg->g_timebase.num / global->framerate.num;
1614 next_frame_start =
1615 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1616 cfg->g_timebase.num / global->framerate.num;
1617
1618 /* Scale if necessary */
1619 if (img) {
1620 if ((img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) &&
1621 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1622 if (img->fmt != AOM_IMG_FMT_I42016) {
1623 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1624 exit(EXIT_FAILURE);
1625 }
1626#if CONFIG_LIBYUV
1627 if (!stream->img) {
1628 stream->img =
1629 aom_img_alloc(NULL, AOM_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1630 }
1631 I420Scale_16(
1632 (uint16_t *)img->planes[AOM_PLANE_Y], img->stride[AOM_PLANE_Y] / 2,
1633 (uint16_t *)img->planes[AOM_PLANE_U], img->stride[AOM_PLANE_U] / 2,
1634 (uint16_t *)img->planes[AOM_PLANE_V], img->stride[AOM_PLANE_V] / 2,
1635 img->d_w, img->d_h, (uint16_t *)stream->img->planes[AOM_PLANE_Y],
1636 stream->img->stride[AOM_PLANE_Y] / 2,
1637 (uint16_t *)stream->img->planes[AOM_PLANE_U],
1638 stream->img->stride[AOM_PLANE_U] / 2,
1639 (uint16_t *)stream->img->planes[AOM_PLANE_V],
1640 stream->img->stride[AOM_PLANE_V] / 2, stream->img->d_w,
1641 stream->img->d_h, kFilterBox);
1642 img = stream->img;
1643#else
1644 stream->encoder.err = 1;
1645 ctx_exit_on_error(&stream->encoder,
1646 "Stream %d: Failed to encode frame.\n"
1647 "libyuv is required for scaling but is currently "
1648 "disabled.\n"
1649 "Be sure to specify -DCONFIG_LIBYUV=1 when running "
1650 "cmake.\n",
1651 stream->index);
1652#endif
1653 }
1654 }
1655 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1656 if (img->fmt != AOM_IMG_FMT_I420 && img->fmt != AOM_IMG_FMT_YV12) {
1657 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1658 exit(EXIT_FAILURE);
1659 }
1660#if CONFIG_LIBYUV
1661 if (!stream->img)
1662 stream->img =
1663 aom_img_alloc(NULL, AOM_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1664 I420Scale(
1665 img->planes[AOM_PLANE_Y], img->stride[AOM_PLANE_Y],
1666 img->planes[AOM_PLANE_U], img->stride[AOM_PLANE_U],
1667 img->planes[AOM_PLANE_V], img->stride[AOM_PLANE_V], img->d_w, img->d_h,
1668 stream->img->planes[AOM_PLANE_Y], stream->img->stride[AOM_PLANE_Y],
1669 stream->img->planes[AOM_PLANE_U], stream->img->stride[AOM_PLANE_U],
1670 stream->img->planes[AOM_PLANE_V], stream->img->stride[AOM_PLANE_V],
1671 stream->img->d_w, stream->img->d_h, kFilterBox);
1672 img = stream->img;
1673#else
1674 stream->encoder.err = 1;
1675 ctx_exit_on_error(&stream->encoder,
1676 "Stream %d: Failed to encode frame.\n"
1677 "Scaling disabled in this configuration. \n"
1678 "To enable, configure with --enable-libyuv\n",
1679 stream->index);
1680#endif
1681 }
1682
1683 struct aom_image monochrome_img;
1684 if (img && cfg->monochrome) {
1685 convert_image_to_monochrome(img, &monochrome_img);
1686 img = &monochrome_img;
1687 }
1688
1689 aom_usec_timer_start(&timer);
1690 aom_codec_encode(&stream->encoder, img, frame_start,
1691 (uint32_t)(next_frame_start - frame_start), 0);
1692 aom_usec_timer_mark(&timer);
1693 stream->cx_time += aom_usec_timer_elapsed(&timer);
1694 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1695 stream->index);
1696}
1697
1698static void update_quantizer_histogram(struct stream_state *stream) {
1699 if (stream->config.cfg.g_pass != AOM_RC_FIRST_PASS) {
1700 int q;
1701
1703 &q);
1704 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1705 stream->counts[q]++;
1706 }
1707}
1708
1709static void get_cx_data(struct stream_state *stream,
1710 struct AvxEncoderConfig *global, int *got_data) {
1711 const aom_codec_cx_pkt_t *pkt;
1712 const struct aom_codec_enc_cfg *cfg = &stream->config.cfg;
1713 aom_codec_iter_t iter = NULL;
1714
1715 *got_data = 0;
1716 while ((pkt = aom_codec_get_cx_data(&stream->encoder, &iter))) {
1717 static size_t fsize = 0;
1718 static FileOffset ivf_header_pos = 0;
1719
1720 switch (pkt->kind) {
1722 ++stream->frames_out;
1723 if (!global->quiet)
1724 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1725
1726 update_rate_histogram(stream->rate_hist, cfg, pkt);
1727#if CONFIG_WEBM_IO
1728 if (stream->config.write_webm) {
1729 if (write_webm_block(&stream->webm_ctx, cfg, pkt) != 0) {
1730 fatal("WebM writer failed.");
1731 }
1732 }
1733#endif
1734 if (!stream->config.write_webm) {
1735 if (stream->config.write_ivf) {
1736 if (pkt->data.frame.partition_id <= 0) {
1737 ivf_header_pos = ftello(stream->file);
1738 fsize = pkt->data.frame.sz;
1739
1740 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1741 } else {
1742 fsize += pkt->data.frame.sz;
1743
1744 const FileOffset currpos = ftello(stream->file);
1745 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1746 ivf_write_frame_size(stream->file, fsize);
1747 fseeko(stream->file, currpos, SEEK_SET);
1748 }
1749 }
1750
1751 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1752 stream->file);
1753 }
1754 stream->nbytes += pkt->data.raw.sz;
1755
1756 *got_data = 1;
1757#if CONFIG_AV1_DECODER
1758 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1759 aom_codec_decode(&stream->decoder, pkt->data.frame.buf,
1760 pkt->data.frame.sz, NULL);
1761 if (stream->decoder.err) {
1762 warn_or_exit_on_error(&stream->decoder,
1763 global->test_decode == TEST_DECODE_FATAL,
1764 "Failed to decode frame %d in stream %d",
1765 stream->frames_out + 1, stream->index);
1766 stream->mismatch_seen = stream->frames_out + 1;
1767 }
1768 }
1769#endif
1770 break;
1772 stream->frames_out++;
1773 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1774 pkt->data.twopass_stats.sz);
1775 stream->nbytes += pkt->data.raw.sz;
1776 break;
1777 case AOM_CODEC_PSNR_PKT:
1778
1779 if (global->show_psnr >= 1) {
1780 int i;
1781
1782 stream->psnr_sse_total[0] += pkt->data.psnr.sse[0];
1783 stream->psnr_samples_total[0] += pkt->data.psnr.samples[0];
1784 for (i = 0; i < 4; i++) {
1785 if (!global->quiet)
1786 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1787 stream->psnr_totals[0][i] += pkt->data.psnr.psnr[i];
1788 }
1789 stream->psnr_count[0]++;
1790
1791#if CONFIG_AV1_HIGHBITDEPTH
1792 if (stream->config.cfg.g_input_bit_depth <
1793 (unsigned int)stream->config.cfg.g_bit_depth) {
1794 stream->psnr_sse_total[1] += pkt->data.psnr.sse_hbd[0];
1795 stream->psnr_samples_total[1] += pkt->data.psnr.samples_hbd[0];
1796 for (i = 0; i < 4; i++) {
1797 if (!global->quiet)
1798 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr_hbd[i]);
1799 stream->psnr_totals[1][i] += pkt->data.psnr.psnr_hbd[i];
1800 }
1801 stream->psnr_count[1]++;
1802 }
1803#endif
1804 }
1805
1806 break;
1807 default: break;
1808 }
1809 }
1810}
1811
1812static void show_psnr(struct stream_state *stream, double peak, int64_t bps) {
1813 int i;
1814 double ovpsnr;
1815
1816 if (!stream->psnr_count[0]) return;
1817
1818 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1819 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total[0], peak,
1820 (double)stream->psnr_sse_total[0]);
1821 fprintf(stderr, " %.3f", ovpsnr);
1822
1823 for (i = 0; i < 4; i++) {
1824 fprintf(stderr, " %.3f", stream->psnr_totals[0][i] / stream->psnr_count[0]);
1825 }
1826 if (bps > 0) {
1827 fprintf(stderr, " %7" PRId64 " bps", bps);
1828 }
1829 fprintf(stderr, " %7" PRId64 " ms", stream->cx_time / 1000);
1830 fprintf(stderr, "\n");
1831}
1832
1833#if CONFIG_AV1_HIGHBITDEPTH
1834static void show_psnr_hbd(struct stream_state *stream, double peak,
1835 int64_t bps) {
1836 int i;
1837 double ovpsnr;
1838 // Compute PSNR based on stream bit depth
1839 if (!stream->psnr_count[1]) return;
1840
1841 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1842 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total[1], peak,
1843 (double)stream->psnr_sse_total[1]);
1844 fprintf(stderr, " %.3f", ovpsnr);
1845
1846 for (i = 0; i < 4; i++) {
1847 fprintf(stderr, " %.3f", stream->psnr_totals[1][i] / stream->psnr_count[1]);
1848 }
1849 if (bps > 0) {
1850 fprintf(stderr, " %7" PRId64 " bps", bps);
1851 }
1852 fprintf(stderr, " %7" PRId64 " ms", stream->cx_time / 1000);
1853 fprintf(stderr, "\n");
1854}
1855#endif
1856
1857static float usec_to_fps(uint64_t usec, unsigned int frames) {
1858 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1859}
1860
1861static void test_decode(struct stream_state *stream,
1862 enum TestDecodeFatality fatal) {
1863 aom_image_t enc_img, dec_img;
1864
1865 if (stream->mismatch_seen) return;
1866
1867 /* Get the internal reference frame */
1869 &enc_img);
1871 &dec_img);
1872
1873 if ((enc_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH) !=
1874 (dec_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH)) {
1875 if (enc_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
1876 aom_image_t enc_hbd_img;
1877 aom_img_alloc(&enc_hbd_img, enc_img.fmt - AOM_IMG_FMT_HIGHBITDEPTH,
1878 enc_img.d_w, enc_img.d_h, 16);
1879 aom_img_truncate_16_to_8(&enc_hbd_img, &enc_img);
1880 enc_img = enc_hbd_img;
1881 }
1882 if (dec_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
1883 aom_image_t dec_hbd_img;
1884 aom_img_alloc(&dec_hbd_img, dec_img.fmt - AOM_IMG_FMT_HIGHBITDEPTH,
1885 dec_img.d_w, dec_img.d_h, 16);
1886 aom_img_truncate_16_to_8(&dec_hbd_img, &dec_img);
1887 dec_img = dec_hbd_img;
1888 }
1889 }
1890
1891 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1892 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1893
1894 if (!aom_compare_img(&enc_img, &dec_img)) {
1895 int y[4], u[4], v[4];
1896 if (enc_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
1897 aom_find_mismatch_high(&enc_img, &dec_img, y, u, v);
1898 } else {
1899 aom_find_mismatch(&enc_img, &dec_img, y, u, v);
1900 }
1901 stream->decoder.err = 1;
1902 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1903 "Stream %d: Encode/decode mismatch on frame %d at"
1904 " Y[%d, %d] {%d/%d},"
1905 " U[%d, %d] {%d/%d},"
1906 " V[%d, %d] {%d/%d}",
1907 stream->index, stream->frames_out, y[0], y[1], y[2],
1908 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1909 stream->mismatch_seen = stream->frames_out;
1910 }
1911
1912 aom_img_free(&enc_img);
1913 aom_img_free(&dec_img);
1914}
1915
1916static void print_time(const char *label, int64_t etl) {
1917 int64_t hours;
1918 int64_t mins;
1919 int64_t secs;
1920
1921 if (etl >= 0) {
1922 hours = etl / 3600;
1923 etl -= hours * 3600;
1924 mins = etl / 60;
1925 etl -= mins * 60;
1926 secs = etl;
1927
1928 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1929 hours, mins, secs);
1930 } else {
1931 fprintf(stderr, "[%3s unknown] ", label);
1932 }
1933}
1934
1935static void clear_stream_count_state(struct stream_state *stream) {
1936 // PSNR counters
1937 for (int k = 0; k < 2; k++) {
1938 stream->psnr_sse_total[k] = 0;
1939 stream->psnr_samples_total[k] = 0;
1940 for (int i = 0; i < 4; i++) {
1941 stream->psnr_totals[k][i] = 0;
1942 }
1943 stream->psnr_count[k] = 0;
1944 }
1945 // q hist
1946 memset(stream->counts, 0, sizeof(stream->counts));
1947}
1948
1949// aomenc will downscale the second pass if:
1950// 1. the specific pass is not given by commandline (aomenc will perform all
1951// passes)
1952// 2. there are more than 2 passes in total
1953// 3. current pass is the second pass (the parameter pass starts with 0 so
1954// pass == 1)
1955static int pass_need_downscale(int global_pass, int global_passes, int pass) {
1956 return !global_pass && global_passes > 2 && pass == 1;
1957}
1958
1959int main(int argc, const char **argv_) {
1960 int pass;
1961 aom_image_t raw;
1962 aom_image_t raw_shift;
1963 int allocated_raw_shift = 0;
1964 int do_16bit_internal = 0;
1965 int input_shift = 0;
1966 int frame_avail, got_data;
1967
1968 struct AvxInputContext input;
1969 struct AvxEncoderConfig global;
1970 struct stream_state *streams = NULL;
1971 char **argv, **argi;
1972 uint64_t cx_time = 0;
1973 int stream_cnt = 0;
1974 int res = 0;
1975 int profile_updated = 0;
1976
1977 memset(&input, 0, sizeof(input));
1978 memset(&raw, 0, sizeof(raw));
1979 exec_name = argv_[0];
1980
1981 /* Setup default input stream settings */
1982 input.framerate.numerator = 30;
1983 input.framerate.denominator = 1;
1984 input.only_i420 = 1;
1985 input.bit_depth = 0;
1986
1987 /* First parse the global configuration values, because we want to apply
1988 * other parameters on top of the default configuration provided by the
1989 * codec.
1990 */
1991 argv = argv_dup(argc - 1, argv_ + 1);
1992 if (!argv) {
1993 fprintf(stderr, "Error allocating argument list\n");
1994 return EXIT_FAILURE;
1995 }
1996 parse_global_config(&global, &argv);
1997
1998 if (argc < 2) usage_exit();
1999
2000 switch (global.color_type) {
2001 case I420: input.fmt = AOM_IMG_FMT_I420; break;
2002 case I422: input.fmt = AOM_IMG_FMT_I422; break;
2003 case I444: input.fmt = AOM_IMG_FMT_I444; break;
2004 case YV12: input.fmt = AOM_IMG_FMT_YV12; break;
2005 case NV12: input.fmt = AOM_IMG_FMT_NV12; break;
2006 }
2007
2008 {
2009 /* Now parse each stream's parameters. Using a local scope here
2010 * due to the use of 'stream' as loop variable in FOREACH_STREAM
2011 * loops
2012 */
2013 struct stream_state *stream = NULL;
2014
2015 do {
2016 stream = new_stream(&global, stream);
2017 stream_cnt++;
2018 if (!streams) streams = stream;
2019 } while (parse_stream_params(&global, stream, argv));
2020 }
2021
2022 /* Check for unrecognized options */
2023 for (argi = argv; *argi; argi++)
2024 if (argi[0][0] == '-' && argi[0][1])
2025 die("Error: Unrecognized option %s\n", *argi);
2026
2027 FOREACH_STREAM(stream, streams) {
2028 check_encoder_config(global.disable_warning_prompt, &global,
2029 &stream->config.cfg);
2030
2031 // If large_scale_tile = 1, only support to output to ivf format.
2032 if (stream->config.cfg.large_scale_tile && !stream->config.write_ivf)
2033 die("only support ivf output format while large-scale-tile=1\n");
2034 }
2035
2036 /* Handle non-option arguments */
2037 input.filename = argv[0];
2038 const char *orig_input_filename = input.filename;
2039 FOREACH_STREAM(stream, streams) {
2040 stream->orig_out_fn = stream->config.out_fn;
2041 stream->orig_width = stream->config.cfg.g_w;
2042 stream->orig_height = stream->config.cfg.g_h;
2043 stream->orig_write_ivf = stream->config.write_ivf;
2044 stream->orig_write_webm = stream->config.write_webm;
2045 }
2046
2047 if (!input.filename) {
2048 fprintf(stderr, "No input file specified!\n");
2049 usage_exit();
2050 }
2051
2052 /* Decide if other chroma subsamplings than 4:2:0 are supported */
2053 if (get_fourcc_by_aom_encoder(global.codec) == AV1_FOURCC)
2054 input.only_i420 = 0;
2055
2056 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2057 if (pass > 1) {
2058 FOREACH_STREAM(stream, streams) { clear_stream_count_state(stream); }
2059 }
2060
2061 int frames_in = 0, seen_frames = 0;
2062 int64_t estimated_time_left = -1;
2063 int64_t average_rate = -1;
2064 int64_t lagged_count = 0;
2065 const int need_downscale =
2066 pass_need_downscale(global.pass, global.passes, pass);
2067
2068 // Set the output to the specified two-pass output file, and
2069 // restore the width and height to the original values.
2070 FOREACH_STREAM(stream, streams) {
2071 if (need_downscale) {
2072 stream->config.out_fn = stream->config.two_pass_output;
2073 // Libaom currently only supports the ivf format for the third pass.
2074 stream->config.write_ivf = 1;
2075 stream->config.write_webm = 0;
2076 } else {
2077 stream->config.out_fn = stream->orig_out_fn;
2078 stream->config.write_ivf = stream->orig_write_ivf;
2079 stream->config.write_webm = stream->orig_write_webm;
2080 }
2081 stream->config.cfg.g_w = stream->orig_width;
2082 stream->config.cfg.g_h = stream->orig_height;
2083 }
2084
2085 // For second pass in three-pass encoding, set the input to
2086 // the given two-pass-input file if available. If the scaled input is not
2087 // given, we will attempt to re-scale the original input.
2088 input.filename = orig_input_filename;
2089 const char *two_pass_input = NULL;
2090 if (need_downscale) {
2091 FOREACH_STREAM(stream, streams) {
2092 if (stream->config.two_pass_input) {
2093 two_pass_input = stream->config.two_pass_input;
2094 input.filename = two_pass_input;
2095 break;
2096 }
2097 }
2098 }
2099
2100 open_input_file(&input, global.csp);
2101
2102 /* If the input file doesn't specify its w/h (raw files), try to get
2103 * the data from the first stream's configuration.
2104 */
2105 if (!input.width || !input.height) {
2106 if (two_pass_input) {
2107 FOREACH_STREAM(stream, streams) {
2108 if (stream->config.two_pass_width && stream->config.two_pass_height) {
2109 input.width = stream->config.two_pass_width;
2110 input.height = stream->config.two_pass_height;
2111 break;
2112 }
2113 }
2114 } else {
2115 FOREACH_STREAM(stream, streams) {
2116 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2117 input.width = stream->config.cfg.g_w;
2118 input.height = stream->config.cfg.g_h;
2119 break;
2120 }
2121 }
2122 }
2123 }
2124
2125 /* Update stream configurations from the input file's parameters */
2126 if (!input.width || !input.height) {
2127 if (two_pass_input) {
2128 fatal(
2129 "Specify downscaled stream dimensions with --two-pass-width "
2130 " and --two-pass-height");
2131 } else {
2132 fatal(
2133 "Specify stream dimensions with --width (-w) "
2134 " and --height (-h)");
2135 }
2136 }
2137
2138 if (need_downscale) {
2139 FOREACH_STREAM(stream, streams) {
2140 if (stream->config.two_pass_width && stream->config.two_pass_height) {
2141 stream->config.cfg.g_w = stream->config.two_pass_width;
2142 stream->config.cfg.g_h = stream->config.two_pass_height;
2143 } else if (two_pass_input) {
2144 stream->config.cfg.g_w = input.width;
2145 stream->config.cfg.g_h = input.height;
2146 } else if (stream->orig_width && stream->orig_height) {
2147#if CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2148 stream->config.cfg.g_w = stream->orig_width;
2149 stream->config.cfg.g_h = stream->orig_height;
2150#else // CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2151 stream->config.cfg.g_w = (stream->orig_width + 1) / 2;
2152 stream->config.cfg.g_h = (stream->orig_height + 1) / 2;
2153#endif // CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2154 } else {
2155#if CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2156 stream->config.cfg.g_w = input.width;
2157 stream->config.cfg.g_h = input.height;
2158#else // CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2159 stream->config.cfg.g_w = (input.width + 1) / 2;
2160 stream->config.cfg.g_h = (input.height + 1) / 2;
2161#endif // CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2162 }
2163 }
2164 }
2165
2166 /* If input file does not specify bit-depth but input-bit-depth parameter
2167 * exists, assume that to be the input bit-depth. However, if the
2168 * input-bit-depth paramter does not exist, assume the input bit-depth
2169 * to be the same as the codec bit-depth.
2170 */
2171 if (!input.bit_depth) {
2172 FOREACH_STREAM(stream, streams) {
2173 if (stream->config.cfg.g_input_bit_depth)
2174 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2175 else
2176 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2177 (int)stream->config.cfg.g_bit_depth;
2178 }
2179 if (input.bit_depth > 8) input.fmt |= AOM_IMG_FMT_HIGHBITDEPTH;
2180 } else {
2181 FOREACH_STREAM(stream, streams) {
2182 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2183 }
2184 }
2185
2186 FOREACH_STREAM(stream, streams) {
2187 if (input.fmt != AOM_IMG_FMT_I420 && input.fmt != AOM_IMG_FMT_I42016 &&
2188 input.fmt != AOM_IMG_FMT_NV12) {
2189 /* Automatically upgrade if input is non-4:2:0 but a 4:2:0 profile
2190 was selected. */
2191 switch (stream->config.cfg.g_profile) {
2192 case 0:
2193 if (input.bit_depth < 12 && (input.fmt == AOM_IMG_FMT_I444 ||
2194 input.fmt == AOM_IMG_FMT_I44416)) {
2195 if (!stream->config.cfg.monochrome) {
2196 stream->config.cfg.g_profile = 1;
2197 profile_updated = 1;
2198 }
2199 } else if (input.bit_depth == 12 ||
2200 ((input.fmt == AOM_IMG_FMT_I422 ||
2201 input.fmt == AOM_IMG_FMT_I42216) &&
2202 !stream->config.cfg.monochrome)) {
2203 stream->config.cfg.g_profile = 2;
2204 profile_updated = 1;
2205 }
2206 break;
2207 case 1:
2208 if (input.bit_depth == 12 || input.fmt == AOM_IMG_FMT_I422 ||
2209 input.fmt == AOM_IMG_FMT_I42216) {
2210 stream->config.cfg.g_profile = 2;
2211 profile_updated = 1;
2212 } else if (input.bit_depth < 12 &&
2213 (input.fmt == AOM_IMG_FMT_I420 ||
2214 input.fmt == AOM_IMG_FMT_I42016)) {
2215 stream->config.cfg.g_profile = 0;
2216 profile_updated = 1;
2217 }
2218 break;
2219 case 2:
2220 if (input.bit_depth < 12 && (input.fmt == AOM_IMG_FMT_I444 ||
2221 input.fmt == AOM_IMG_FMT_I44416)) {
2222 stream->config.cfg.g_profile = 1;
2223 profile_updated = 1;
2224 } else if (input.bit_depth < 12 &&
2225 (input.fmt == AOM_IMG_FMT_I420 ||
2226 input.fmt == AOM_IMG_FMT_I42016)) {
2227 stream->config.cfg.g_profile = 0;
2228 profile_updated = 1;
2229 } else if (input.bit_depth == 12 &&
2230 input.file_type == FILE_TYPE_Y4M) {
2231 // Note that here the input file values for chroma subsampling
2232 // are used instead of those from the command line.
2233 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
2235 input.y4m.dst_c_dec_h >> 1);
2236 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
2238 input.y4m.dst_c_dec_v >> 1);
2239 } else if (input.bit_depth == 12 &&
2240 input.file_type == FILE_TYPE_RAW) {
2241 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
2243 stream->chroma_subsampling_x);
2244 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
2246 stream->chroma_subsampling_y);
2247 }
2248 break;
2249 default: break;
2250 }
2251 }
2252 /* Automatically set the codec bit depth to match the input bit depth.
2253 * Upgrade the profile if required. */
2254 if (stream->config.cfg.g_input_bit_depth >
2255 (unsigned int)stream->config.cfg.g_bit_depth) {
2256 stream->config.cfg.g_bit_depth = stream->config.cfg.g_input_bit_depth;
2257 if (!global.quiet) {
2258 fprintf(stderr,
2259 "Warning: automatically updating bit depth to %d to "
2260 "match input format.\n",
2261 stream->config.cfg.g_input_bit_depth);
2262 }
2263 }
2264#if !CONFIG_AV1_HIGHBITDEPTH
2265 if (stream->config.cfg.g_bit_depth > 8) {
2266 fatal("Unsupported bit-depth with CONFIG_AV1_HIGHBITDEPTH=0\n");
2267 }
2268#endif // CONFIG_AV1_HIGHBITDEPTH
2269 if (stream->config.cfg.g_bit_depth > 10) {
2270 switch (stream->config.cfg.g_profile) {
2271 case 0:
2272 case 1:
2273 stream->config.cfg.g_profile = 2;
2274 profile_updated = 1;
2275 break;
2276 default: break;
2277 }
2278 }
2279 if (stream->config.cfg.g_bit_depth > 8) {
2280 stream->config.use_16bit_internal = 1;
2281 }
2282 if (profile_updated && !global.quiet) {
2283 fprintf(stderr,
2284 "Warning: automatically updating to profile %d to "
2285 "match input format.\n",
2286 stream->config.cfg.g_profile);
2287 }
2288 if ((global.show_psnr == 2) && (stream->config.cfg.g_input_bit_depth ==
2289 stream->config.cfg.g_bit_depth)) {
2290 fprintf(stderr,
2291 "Warning: --psnr==2 and --psnr==1 will provide same "
2292 "results when input bit-depth == stream bit-depth, "
2293 "falling back to default psnr value\n");
2294 global.show_psnr = 1;
2295 }
2296 if (global.show_psnr < 0 || global.show_psnr > 2) {
2297 fprintf(stderr,
2298 "Warning: --psnr can take only 0,1,2 as values,"
2299 "falling back to default psnr value\n");
2300 global.show_psnr = 1;
2301 }
2302 /* Set limit */
2303 stream->config.cfg.g_limit = global.limit;
2304 }
2305
2306 FOREACH_STREAM(stream, streams) {
2307 set_stream_dimensions(stream, input.width, input.height);
2308 stream->config.color_range = input.color_range;
2309 }
2310 FOREACH_STREAM(stream, streams) { validate_stream_config(stream, &global); }
2311
2312 /* Ensure that --passes and --pass are consistent. If --pass is set and
2313 * --passes >= 2, ensure --fpf was set.
2314 */
2315 if (global.pass > 0 && global.pass <= 3 && global.passes >= 2) {
2316 FOREACH_STREAM(stream, streams) {
2317 if (!stream->config.stats_fn)
2318 die("Stream %d: Must specify --fpf when --pass=%d"
2319 " and --passes=%d\n",
2320 stream->index, global.pass, global.passes);
2321 }
2322 }
2323
2324#if !CONFIG_WEBM_IO
2325 FOREACH_STREAM(stream, streams) {
2326 if (stream->config.write_webm) {
2327 stream->config.write_webm = 0;
2328 stream->config.write_ivf = 0;
2329 aom_tools_warn("aomenc compiled w/o WebM support. Writing OBU stream.");
2330 }
2331 }
2332#endif
2333
2334 /* Use the frame rate from the file only if none was specified
2335 * on the command-line.
2336 */
2337 if (!global.have_framerate) {
2338 global.framerate.num = input.framerate.numerator;
2339 global.framerate.den = input.framerate.denominator;
2340 }
2341 FOREACH_STREAM(stream, streams) {
2342 stream->config.cfg.g_timebase.den = global.framerate.num;
2343 stream->config.cfg.g_timebase.num = global.framerate.den;
2344 }
2345 /* Show configuration */
2346 if (global.verbose && pass == 0) {
2347 FOREACH_STREAM(stream, streams) {
2348 show_stream_config(stream, &global, &input);
2349 }
2350 }
2351
2352 if (pass == (global.pass ? global.pass - 1 : 0)) {
2353 // The Y4M reader does its own allocation.
2354 if (input.file_type != FILE_TYPE_Y4M) {
2355 aom_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2356 }
2357 FOREACH_STREAM(stream, streams) {
2358 stream->rate_hist =
2359 init_rate_histogram(&stream->config.cfg, &global.framerate);
2360 }
2361 }
2362
2363 FOREACH_STREAM(stream, streams) { setup_pass(stream, &global, pass); }
2364 FOREACH_STREAM(stream, streams) { initialize_encoder(stream, &global); }
2365 FOREACH_STREAM(stream, streams) {
2366 char *encoder_settings = NULL;
2367#if CONFIG_WEBM_IO
2368 // Test frameworks may compare outputs from different versions, but only
2369 // wish to check for bitstream changes. The encoder-settings tag, however,
2370 // can vary if the version is updated, even if no encoder algorithm
2371 // changes were made. To work around this issue, do not output
2372 // the encoder-settings tag when --debug is enabled (which is the flag
2373 // that test frameworks should use, when they want deterministic output
2374 // from the container format).
2375 if (stream->config.write_webm && !stream->webm_ctx.debug) {
2376 encoder_settings = extract_encoder_settings(
2377 aom_codec_version_str(), argv_, argc, input.filename);
2378 if (encoder_settings == NULL) {
2379 fprintf(
2380 stderr,
2381 "Warning: unable to extract encoder settings. Continuing...\n");
2382 }
2383 }
2384#endif
2385 open_output_file(stream, &global, &input.pixel_aspect_ratio,
2386 encoder_settings);
2387 free(encoder_settings);
2388 }
2389
2390 if (strcmp(get_short_name_by_aom_encoder(global.codec), "av1") == 0) {
2391 // Check to see if at least one stream uses 16 bit internal.
2392 // Currently assume that the bit_depths for all streams using
2393 // highbitdepth are the same.
2394 FOREACH_STREAM(stream, streams) {
2395 if (stream->config.use_16bit_internal) {
2396 do_16bit_internal = 1;
2397 }
2398 input_shift = (int)stream->config.cfg.g_bit_depth -
2399 stream->config.cfg.g_input_bit_depth;
2400 }
2401 }
2402
2403 frame_avail = 1;
2404 got_data = 0;
2405
2406 while (frame_avail || got_data) {
2407 struct aom_usec_timer timer;
2408
2409 if (!global.limit || frames_in < global.limit) {
2410 frame_avail = read_frame(&input, &raw);
2411
2412 if (frame_avail) frames_in++;
2413 seen_frames =
2414 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
2415
2416 if (!global.quiet) {
2417 float fps = usec_to_fps(cx_time, seen_frames);
2418 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2419
2420 if (stream_cnt == 1)
2421 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
2422 streams->frames_out, (int64_t)streams->nbytes);
2423 else
2424 fprintf(stderr, "frame %4d ", frames_in);
2425
2426 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
2427 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2428 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
2429 fps >= 1.0 ? "fps" : "fpm");
2430 print_time("ETA", estimated_time_left);
2431 // mingw-w64 gcc does not match msvc for stderr buffering behavior
2432 // and uses line buffering, thus the progress output is not
2433 // real-time. The fflush() is here to make sure the progress output
2434 // is sent out while the clip is being processed.
2435 fflush(stderr);
2436 }
2437
2438 } else {
2439 frame_avail = 0;
2440 }
2441
2442 if (frames_in > global.skip_frames) {
2443 aom_image_t *frame_to_encode;
2444 if (input_shift || (do_16bit_internal && input.bit_depth == 8)) {
2445 assert(do_16bit_internal);
2446 // Input bit depth and stream bit depth do not match, so up
2447 // shift frame to stream bit depth
2448 if (!allocated_raw_shift) {
2449 aom_img_alloc(&raw_shift, raw.fmt | AOM_IMG_FMT_HIGHBITDEPTH,
2450 input.width, input.height, 32);
2451 allocated_raw_shift = 1;
2452 }
2453 aom_img_upshift(&raw_shift, &raw, input_shift);
2454 frame_to_encode = &raw_shift;
2455 } else {
2456 frame_to_encode = &raw;
2457 }
2458 aom_usec_timer_start(&timer);
2459 if (do_16bit_internal) {
2460 assert(frame_to_encode->fmt & AOM_IMG_FMT_HIGHBITDEPTH);
2461 FOREACH_STREAM(stream, streams) {
2462 if (stream->config.use_16bit_internal)
2463 encode_frame(stream, &global,
2464 frame_avail ? frame_to_encode : NULL, frames_in);
2465 else
2466 assert(0);
2467 }
2468 } else {
2469 assert((frame_to_encode->fmt & AOM_IMG_FMT_HIGHBITDEPTH) == 0);
2470 FOREACH_STREAM(stream, streams) {
2471 encode_frame(stream, &global, frame_avail ? frame_to_encode : NULL,
2472 frames_in);
2473 }
2474 }
2475 aom_usec_timer_mark(&timer);
2476 cx_time += aom_usec_timer_elapsed(&timer);
2477
2478 FOREACH_STREAM(stream, streams) { update_quantizer_histogram(stream); }
2479
2480 got_data = 0;
2481 FOREACH_STREAM(stream, streams) {
2482 get_cx_data(stream, &global, &got_data);
2483 }
2484
2485 if (!got_data && input.length && streams != NULL &&
2486 !streams->frames_out) {
2487 lagged_count = global.limit ? seen_frames : ftello(input.file);
2488 } else if (input.length) {
2489 int64_t remaining;
2490 int64_t rate;
2491
2492 if (global.limit) {
2493 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2494
2495 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2496 remaining = 1000 * (global.limit - global.skip_frames -
2497 seen_frames + lagged_count);
2498 } else {
2499 const int64_t input_pos = ftello(input.file);
2500 const int64_t input_pos_lagged = input_pos - lagged_count;
2501 const int64_t input_limit = input.length;
2502
2503 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2504 remaining = input_limit - input_pos + lagged_count;
2505 }
2506
2507 average_rate =
2508 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
2509 estimated_time_left = average_rate ? remaining / average_rate : -1;
2510 }
2511
2512 if (got_data && global.test_decode != TEST_DECODE_OFF) {
2513 FOREACH_STREAM(stream, streams) {
2514 test_decode(stream, global.test_decode);
2515 }
2516 }
2517 }
2518
2519 fflush(stdout);
2520 if (!global.quiet) fprintf(stderr, "\033[K");
2521 }
2522
2523 if (stream_cnt > 1) fprintf(stderr, "\n");
2524
2525 if (!global.quiet) {
2526 FOREACH_STREAM(stream, streams) {
2527 const int64_t bpf =
2528 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0;
2529 const int64_t bps = bpf * global.framerate.num / global.framerate.den;
2530 fprintf(stderr,
2531 "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64
2532 "b/f %7" PRId64
2533 "b/s"
2534 " %7" PRId64 " %s (%.2f fps)\033[K\n",
2535 pass + 1, global.passes, frames_in, stream->frames_out,
2536 (int64_t)stream->nbytes, bpf, bps,
2537 stream->cx_time > 9999999 ? stream->cx_time / 1000
2538 : stream->cx_time,
2539 stream->cx_time > 9999999 ? "ms" : "us",
2540 usec_to_fps(stream->cx_time, seen_frames));
2541 // This instance of cr does not need fflush as it is followed by a
2542 // newline in the same string.
2543 }
2544 }
2545
2546 if (global.show_psnr >= 1) {
2547 if (get_fourcc_by_aom_encoder(global.codec) == AV1_FOURCC) {
2548 FOREACH_STREAM(stream, streams) {
2549 int64_t bps = 0;
2550 if (global.show_psnr == 1) {
2551 if (stream->psnr_count[0] && seen_frames && global.framerate.den) {
2552 bps = (int64_t)stream->nbytes * 8 *
2553 (int64_t)global.framerate.num / global.framerate.den /
2554 seen_frames;
2555 }
2556 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1,
2557 bps);
2558 }
2559 if (global.show_psnr == 2) {
2560#if CONFIG_AV1_HIGHBITDEPTH
2561 if (stream->config.cfg.g_input_bit_depth <
2562 (unsigned int)stream->config.cfg.g_bit_depth)
2563 show_psnr_hbd(stream, (1 << stream->config.cfg.g_bit_depth) - 1,
2564 bps);
2565#endif
2566 }
2567 }
2568 } else {
2569 FOREACH_STREAM(stream, streams) { show_psnr(stream, 255.0, 0); }
2570 }
2571 }
2572
2573 if (pass == global.passes - 1) {
2574 FOREACH_STREAM(stream, streams) {
2575 int num_operating_points;
2576 int levels[32];
2577 int target_levels[32];
2579 &num_operating_points);
2580 aom_codec_control(&stream->encoder, AV1E_GET_SEQ_LEVEL_IDX, levels);
2582 target_levels);
2583
2584 for (int i = 0; i < num_operating_points; i++) {
2585 if (levels[i] > target_levels[i]) {
2586 aom_tools_warn(
2587 "Failed to encode to target level %d.%d for operating point "
2588 "%d. The output level is %d.%d",
2589 2 + (target_levels[i] >> 2), target_levels[i] & 3, i,
2590 2 + (levels[i] >> 2), levels[i] & 3);
2591 }
2592 }
2593 }
2594 }
2595
2596 FOREACH_STREAM(stream, streams) { aom_codec_destroy(&stream->encoder); }
2597
2598 if (global.test_decode != TEST_DECODE_OFF) {
2599 FOREACH_STREAM(stream, streams) { aom_codec_destroy(&stream->decoder); }
2600 }
2601
2602 close_input_file(&input);
2603
2604 if (global.test_decode == TEST_DECODE_FATAL) {
2605 FOREACH_STREAM(stream, streams) { res |= stream->mismatch_seen; }
2606 }
2607 FOREACH_STREAM(stream, streams) {
2608 close_output_file(stream, get_fourcc_by_aom_encoder(global.codec));
2609 }
2610
2611 FOREACH_STREAM(stream, streams) {
2612 stats_close(&stream->stats, global.passes - 1);
2613 }
2614
2615 if (global.pass) break;
2616 }
2617
2618 if (global.show_q_hist_buckets) {
2619 FOREACH_STREAM(stream, streams) {
2620 show_q_histogram(stream->counts, global.show_q_hist_buckets);
2621 }
2622 }
2623
2624 if (global.show_rate_hist_buckets) {
2625 FOREACH_STREAM(stream, streams) {
2626 show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2627 global.show_rate_hist_buckets);
2628 }
2629 }
2630 FOREACH_STREAM(stream, streams) { destroy_rate_histogram(stream->rate_hist); }
2631
2632#if CONFIG_INTERNAL_STATS
2633 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2634 * to match some existing utilities.
2635 */
2636 if (!(global.pass == 1 && global.passes == 2)) {
2637 FOREACH_STREAM(stream, streams) {
2638 FILE *f = fopen("opsnr.stt", "a");
2639 if (stream->mismatch_seen) {
2640 fprintf(f, "First mismatch occurred in frame %d\n",
2641 stream->mismatch_seen);
2642 } else {
2643 fprintf(f, "No mismatch detected in recon buffers\n");
2644 }
2645 fclose(f);
2646 }
2647 }
2648#endif
2649
2650 if (allocated_raw_shift) aom_img_free(&raw_shift);
2651 aom_img_free(&raw);
2652 free(argv);
2653 free(streams);
2654 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2655}
Describes the decoder algorithm interface to applications.
Describes the encoder algorithm interface to applications.
#define MAX_TILE_WIDTHS
Maximum number of tile widths in tile widths array.
Definition: aom_encoder.h:855
#define MAX_TILE_HEIGHTS
Maximum number of tile heights in tile heights array.
Definition: aom_encoder.h:868
#define AOM_PLANE_U
Definition: aom_image.h:209
@ AOM_CSP_UNKNOWN
Definition: aom_image.h:142
enum aom_chroma_sample_position aom_chroma_sample_position_t
List of chroma sample positions.
#define AOM_PLANE_Y
Definition: aom_image.h:208
#define AOM_PLANE_V
Definition: aom_image.h:210
enum aom_color_range aom_color_range_t
List of supported color range.
#define AOM_IMG_FMT_HIGHBITDEPTH
Definition: aom_image.h:38
aom_image_t * aom_img_alloc(aom_image_t *img, aom_img_fmt_t fmt, unsigned int d_w, unsigned int d_h, unsigned int align)
Open a descriptor, allocating storage for the underlying image.
@ AOM_IMG_FMT_I42216
Definition: aom_image.h:58
@ AOM_IMG_FMT_I42016
Definition: aom_image.h:56
@ AOM_IMG_FMT_YV1216
Definition: aom_image.h:57
@ AOM_IMG_FMT_I444
Definition: aom_image.h:50
@ AOM_IMG_FMT_I422
Definition: aom_image.h:49
@ AOM_IMG_FMT_I44416
Definition: aom_image.h:59
@ AOM_IMG_FMT_I420
Definition: aom_image.h:45
@ AOM_IMG_FMT_NV12
Definition: aom_image.h:54
@ AOM_IMG_FMT_YV12
Definition: aom_image.h:43
enum aom_img_fmt aom_img_fmt_t
List of supported image formats.
void aom_img_free(aom_image_t *img)
Close an image descriptor.
Provides definitions for using AOM or AV1 encoder algorithm within the aom Codec Interface.
Provides definitions for using AOM or AV1 within the aom Decoder interface.
@ AV1_SET_TILE_MODE
Codec control function to set the tile coding mode, unsigned int parameter.
Definition: aomdx.h:313
@ AV1D_SET_IS_ANNEXB
Codec control function to indicate whether bitstream is in Annex-B format, unsigned int parameter.
Definition: aomdx.h:349
@ AV1_SET_DECODE_TILE_ROW
Codec control function to set the range of tile decoding, int parameter.
Definition: aomdx.h:304
@ AV1E_SET_MATRIX_COEFFICIENTS
Codec control function to set transfer function info, int parameter.
Definition: aomcx.h:573
@ AV1E_SET_ENABLE_INTERINTER_WEDGE
Codec control function to turn on / off interinter wedge compound, int parameter.
Definition: aomcx.h:1006
@ AV1E_SET_ENABLE_DIAGONAL_INTRA
Codec control function to turn on / off D45 to D203 intra mode usage, int parameter.
Definition: aomcx.h:1344
@ AV1E_SET_MAX_GF_INTERVAL
Codec control function to set minimum interval between GF/ARF frames, unsigned int parameter.
Definition: aomcx.h:594
@ AV1E_SET_ROW_MT
Codec control function to enable the row based multi-threading of the encoder, unsigned int parameter...
Definition: aomcx.h:361
@ AV1E_SET_ENABLE_SMOOTH_INTRA
Codec control function to turn on / off smooth intra modes usage, int parameter.
Definition: aomcx.h:1066
@ AOME_SET_SHARPNESS
Codec control function to set the sharpness parameter, unsigned int parameter.
Definition: aomcx.h:241
@ AV1E_GET_TARGET_SEQ_LEVEL_IDX
Codec control function to get the target sequence level index for each operating point....
Definition: aomcx.h:1451
@ AV1E_SET_ENABLE_TPL_MODEL
Codec control function to enable RDO modulated by frame temporal dependency, unsigned int parameter.
Definition: aomcx.h:408
@ AOME_GET_LAST_QUANTIZER_64
Codec control function to get last quantizer chosen by the encoder, int* parameter.
Definition: aomcx.h:263
@ AV1E_SET_AQ_MODE
Codec control function to set adaptive quantization mode, unsigned int parameter.
Definition: aomcx.h:468
@ AV1E_SET_REDUCED_REFERENCE_SET
Control to use reduced set of single and compound references, int parameter.
Definition: aomcx.h:1220
@ AV1E_GET_NUM_OPERATING_POINTS
Codec control function to get the number of operating points. int* parameter.
Definition: aomcx.h:1456
@ AV1E_SET_GF_MIN_PYRAMID_HEIGHT
Control to select minimum height for the GF group pyramid structure, unsigned int parameter.
Definition: aomcx.h:1316
@ AV1E_SET_ENABLE_PAETH_INTRA
Codec control function to turn on / off Paeth intra mode usage, int parameter.
Definition: aomcx.h:1074
@ AV1E_SET_TUNE_CONTENT
Codec control function to set content type, aom_tune_content parameter.
Definition: aomcx.h:497
@ AV1E_SET_CDF_UPDATE_MODE
Codec control function to set CDF update mode, unsigned int parameter.
Definition: aomcx.h:506
@ AV1E_SET_CHROMA_SUBSAMPLING_X
Sets the chroma subsampling x value, unsigned int parameter.
Definition: aomcx.h:1183
@ AV1E_SET_COLOR_RANGE
Codec control function to set color range bit, int parameter.
Definition: aomcx.h:606
@ AV1E_SET_ENABLE_RESTORATION
Codec control function to encode with Loop Restoration Filter, unsigned int parameter.
Definition: aomcx.h:676
@ AV1E_SET_ENABLE_ANGLE_DELTA
Codec control function to turn on/off intra angle delta, int parameter.
Definition: aomcx.h:1113
@ AV1E_SET_MIN_GF_INTERVAL
Codec control function to set minimum interval between GF/ARF frames, unsigned int parameter.
Definition: aomcx.h:587
@ AOME_SET_ARNR_MAXFRAMES
Codec control function to set the max no of frames to create arf, unsigned int parameter.
Definition: aomcx.h:268
@ AV1E_SET_MV_COST_UPD_FREQ
Control to set frequency of the cost updates for motion vectors, unsigned int parameter.
Definition: aomcx.h:1250
@ AV1E_SET_INTRA_DEFAULT_TX_ONLY
Control to use default tx type only for intra modes, int parameter.
Definition: aomcx.h:1199
@ AV1E_SET_TRANSFER_CHARACTERISTICS
Codec control function to set transfer function info, int parameter.
Definition: aomcx.h:552
@ AV1E_SET_MTU
Codec control function to set an MTU size for a tile group, unsigned int parameter.
Definition: aomcx.h:796
@ AV1E_SET_DISABLE_TRELLIS_QUANT
Codec control function to encode without trellis quantization, unsigned int parameter.
Definition: aomcx.h:703
@ AV1E_SET_ENABLE_INTRABC
Codec control function to turn on/off intra block copy mode, int parameter.
Definition: aomcx.h:1109
@ AV1E_SET_ENABLE_AB_PARTITIONS
Codec control function to enable/disable AB partitions, int parameter.
Definition: aomcx.h:814
@ AV1E_SET_ENABLE_INTERINTRA_COMP
Codec control function to turn on / off interintra compound for a sequence, int parameter.
Definition: aomcx.h:982
@ AV1E_SET_FILM_GRAIN_TEST_VECTOR
Codec control function to add film grain parameters (one of several preset types) info in the bitstre...
Definition: aomcx.h:1169
@ AV1E_SET_ENABLE_CHROMA_DELTAQ
Codec control function to turn on / off delta quantization in chroma planes for a sequence,...
Definition: aomcx.h:958
@ AV1E_SET_ENABLE_DUAL_FILTER
Codec control function to turn on / off dual interpolation filter for a sequence, int parameter.
Definition: aomcx.h:950
@ AV1E_SET_FRAME_PARALLEL_DECODING
Codec control function to enable frame parallel decoding feature, unsigned int parameter.
Definition: aomcx.h:431
@ AV1E_SET_MIN_PARTITION_SIZE
Codec control function to set min partition size, int parameter.
Definition: aomcx.h:833
@ AV1E_SET_ENABLE_WARPED_MOTION
Codec control function to turn on / off warped motion usage at sequence level, int parameter.
Definition: aomcx.h:1034
@ AV1E_SET_FORCE_VIDEO_MODE
Codec control function to force video mode, unsigned int parameter.
Definition: aomcx.h:683
@ AV1E_SET_CHROMA_SUBSAMPLING_Y
Sets the chroma subsampling y value, unsigned int parameter.
Definition: aomcx.h:1186
@ AV1E_SET_ENABLE_INTRA_EDGE_FILTER
Codec control function to turn on / off intra edge filter at sequence level, int parameter.
Definition: aomcx.h:852
@ AV1E_SET_COEFF_COST_UPD_FREQ
Control to set frequency of the cost updates for coefficients, unsigned int parameter.
Definition: aomcx.h:1230
@ AV1E_SET_ENABLE_DIRECTIONAL_INTRA
Codec control function to turn on / off directional intra mode usage, int parameter.
Definition: aomcx.h:1373
@ AV1E_SET_MAX_INTER_BITRATE_PCT
Codec control function to set max data rate for inter frames, unsigned int parameter.
Definition: aomcx.h:325
@ AV1E_SET_DENOISE_NOISE_LEVEL
Sets the noise level, int parameter.
Definition: aomcx.h:1177
@ AV1E_SET_INTRA_DCT_ONLY
Control to use dct only for intra modes, int parameter.
Definition: aomcx.h:1192
@ AV1E_SET_TILE_ROWS
Codec control function to set number of tile rows, unsigned int parameter.
Definition: aomcx.h:398
@ AV1E_SET_ENABLE_REF_FRAME_MVS
Codec control function to turn on / off ref frame mvs (mfmv) usage at sequence level,...
Definition: aomcx.h:931
@ AV1E_SET_FP_MT
Codec control function to enable frame parallel multi-threading of the encoder, unsigned int paramete...
Definition: aomcx.h:1431
@ AV1E_SET_ENABLE_MASKED_COMP
Codec control function to turn on / off masked compound usage (wedge and diff-wtd compound modes) for...
Definition: aomcx.h:966
@ AV1E_SET_VBR_CORPUS_COMPLEXITY_LAP
Control to set average complexity of the corpus in the case of single pass vbr based on LAP,...
Definition: aomcx.h:1321
@ AV1E_SET_GF_MAX_PYRAMID_HEIGHT
Control to select maximum height for the GF group pyramid structure, unsigned int parameter.
Definition: aomcx.h:1209
@ AV1E_SET_ENABLE_CDEF
Codec control function to encode with CDEF, unsigned int parameter.
Definition: aomcx.h:666
@ AV1E_SET_ENABLE_FLIP_IDTX
Codec control function to turn on / off flip and identity transforms, int parameter.
Definition: aomcx.h:896
@ AV1E_GET_SEQ_LEVEL_IDX
Codec control function to get sequence level index for each operating point. int* parameter....
Definition: aomcx.h:639
@ AV1E_SET_FRAME_PERIODIC_BOOST
Codec control function to enable/disable periodic Q boost, unsigned int parameter.
Definition: aomcx.h:480
@ AV1E_SET_DV_COST_UPD_FREQ
Control to set frequency of the cost updates for intrabc motion vectors, unsigned int parameter.
Definition: aomcx.h:1354
@ AV1E_SET_AUTO_INTRA_TOOLS_OFF
Codec control to automatically turn off several intra coding tools, unsigned int parameter.
Definition: aomcx.h:1415
@ AV1E_SET_ENABLE_RECT_TX
Codec control function to turn on / off rectangular transforms, int parameter.
Definition: aomcx.h:908
@ AV1E_SET_ENABLE_DIST_WTD_COMP
Codec control function to turn on / off dist-wtd compound mode at sequence level, int parameter.
Definition: aomcx.h:920
@ AV1E_SET_TIMING_INFO_TYPE
Codec control function to signal picture timing info in the bitstream, aom_timing_info_type_t paramet...
Definition: aomcx.h:1162
@ AV1E_SET_SUPERBLOCK_SIZE
Codec control function to set intended superblock size, unsigned int parameter.
Definition: aomcx.h:647
@ AV1E_SET_TIER_MASK
Control to set bit mask that specifies which tier each of the 32 possible operating points conforms t...
Definition: aomcx.h:1258
@ AV1E_SET_ENABLE_INTERINTRA_WEDGE
Codec control function to turn on / off interintra wedge compound, int parameter.
Definition: aomcx.h:1014
@ AV1E_SET_NOISE_SENSITIVITY
Codec control function to set noise sensitivity, unsigned int parameter.
Definition: aomcx.h:488
@ AV1E_SET_ENABLE_DIFF_WTD_COMP
Codec control function to turn on / off difference weighted compound, int parameter.
Definition: aomcx.h:998
@ AV1E_SET_QUANT_B_ADAPT
Control to use adaptive quantize_b, int parameter.
Definition: aomcx.h:1202
@ AV1E_SET_ENABLE_FILTER_INTRA
Codec control function to turn on / off filter intra usage at sequence level, int parameter.
Definition: aomcx.h:1055
@ AV1E_SET_ENABLE_PALETTE
Codec control function to turn on/off palette mode, int parameter.
Definition: aomcx.h:1105
@ AV1E_SET_ENABLE_CFL_INTRA
Codec control function to turn on / off CFL uv intra mode usage, int parameter.
Definition: aomcx.h:1084
@ AV1E_SET_ENABLE_KEYFRAME_FILTERING
Codec control function to enable temporal filtering on key frame, unsigned int parameter.
Definition: aomcx.h:417
@ AV1E_SET_NUM_TG
Codec control function to set a maximum number of tile groups, unsigned int parameter.
Definition: aomcx.h:785
@ AOME_SET_MAX_INTRA_BITRATE_PCT
Codec control function to set max data rate for intra frames, unsigned int parameter.
Definition: aomcx.h:306
@ AV1E_SET_ERROR_RESILIENT_MODE
Codec control function to enable error_resilient_mode, int parameter.
Definition: aomcx.h:442
@ AV1E_SET_ENABLE_SMOOTH_INTERINTRA
Codec control function to turn on / off smooth inter-intra mode for a sequence, int parameter.
Definition: aomcx.h:990
@ AOME_SET_STATIC_THRESHOLD
Codec control function to set the threshold for MBs treated static, unsigned int parameter.
Definition: aomcx.h:246
@ AV1E_SET_ENABLE_OBMC
Codec control function to predict with OBMC mode, unsigned int parameter.
Definition: aomcx.h:693
@ AV1E_SET_PARTITION_INFO_PATH
Codec control to set the path for partition stats read and write. const char * parameter.
Definition: aomcx.h:1359
@ AV1E_SET_MAX_PARTITION_SIZE
Codec control function to set max partition size, int parameter.
Definition: aomcx.h:844
@ AV1E_SET_ENABLE_1TO4_PARTITIONS
Codec control function to enable/disable 1:4 and 4:1 partitions, int parameter.
Definition: aomcx.h:822
@ AV1E_SET_DELTALF_MODE
Codec control function to turn on/off loopfilter modulation when delta q modulation is enabled,...
Definition: aomcx.h:1135
@ AV1E_SET_ENABLE_TX64
Codec control function to turn on / off 64-length transforms, int parameter.
Definition: aomcx.h:872
@ AOME_SET_TUNING
Codec control function to set visual tuning, aom_tune_metric (int) parameter.
Definition: aomcx.h:282
@ AV1E_SET_TARGET_SEQ_LEVEL_IDX
Control to set target sequence level index for a certain operating point (OP), int parameter Possible...
Definition: aomcx.h:632
@ AV1E_SET_CHROMA_SAMPLE_POSITION
Codec control function to set chroma 4:2:0 sample position info, aom_chroma_sample_position_t paramet...
Definition: aomcx.h:580
@ AV1E_SET_REDUCED_TX_TYPE_SET
Control to use a reduced tx type set, int parameter.
Definition: aomcx.h:1189
@ AV1E_SET_DELTAQ_STRENGTH
Set –deltaq-mode strength.
Definition: aomcx.h:1394
@ AV1E_SET_INTER_DCT_ONLY
Control to use dct only for inter modes, int parameter.
Definition: aomcx.h:1195
@ AV1E_SET_LOOPFILTER_CONTROL
Codec control to control loop filter.
Definition: aomcx.h:1403
@ AOME_SET_ENABLEAUTOALTREF
Codec control function to enable automatic set and use alf frames, unsigned int parameter.
Definition: aomcx.h:228
@ AV1E_SET_TILE_COLUMNS
Codec control function to set number of tile columns. unsigned int parameter.
Definition: aomcx.h:380
@ AV1E_SET_ENABLE_ORDER_HINT
Codec control function to turn on / off frame order hint (int parameter). Affects: joint compound mod...
Definition: aomcx.h:861
@ AV1E_SET_DELTAQ_MODE
Codec control function to set the delta q mode, unsigned int parameter.
Definition: aomcx.h:1127
@ AV1E_SET_ENABLE_GLOBAL_MOTION
Codec control function to turn on / off global motion usage for a sequence, int parameter.
Definition: aomcx.h:1024
@ AV1E_SET_FILM_GRAIN_TABLE
Codec control function to set the path to the film grain parameters, const char* parameter.
Definition: aomcx.h:1174
@ AV1E_SET_QM_MAX
Codec control function to set the max quant matrix flatness, unsigned int parameter.
Definition: aomcx.h:739
@ AV1E_SET_MAX_REFERENCE_FRAMES
Control to select maximum reference frames allowed per frame, int parameter.
Definition: aomcx.h:1216
@ AOME_SET_CPUUSED
Codec control function to set encoder internal speed settings, int parameter.
Definition: aomcx.h:220
@ AV1E_SET_GF_CBR_BOOST_PCT
Boost percentage for Golden Frame in CBR mode, unsigned int parameter.
Definition: aomcx.h:339
@ AV1E_SET_ENABLE_ONESIDED_COMP
Codec control function to turn on / off one sided compound usage for a sequence, int parameter.
Definition: aomcx.h:974
@ AV1E_SET_DENOISE_BLOCK_SIZE
Sets the denoisers block size, unsigned int parameter.
Definition: aomcx.h:1180
@ AV1E_SET_VMAF_MODEL_PATH
Codec control function to set the path to the VMAF model used when tuning the encoder for VMAF,...
Definition: aomcx.h:1288
@ AV1E_SET_QM_MIN
Codec control function to set the min quant matrix flatness, unsigned int parameter.
Definition: aomcx.h:727
@ AV1E_SET_ENABLE_QM
Codec control function to encode with quantisation matrices, unsigned int parameter.
Definition: aomcx.h:714
@ AV1E_SET_ENABLE_OVERLAY
Codec control function to turn on / off overlay frames for filtered ALTREF frames,...
Definition: aomcx.h:1102
@ AV1E_SET_ENABLE_RECT_PARTITIONS
Codec control function to enable/disable rectangular partitions, int parameter.
Definition: aomcx.h:806
@ AV1E_SET_COLOR_PRIMARIES
Codec control function to set color space info, int parameter.
Definition: aomcx.h:527
@ AOME_SET_CQ_LEVEL
Codec control function to set constrained / constant quality level, unsigned int parameter.
Definition: aomcx.h:292
@ AV1E_SET_ENABLE_TX_SIZE_SEARCH
Control to turn on / off transform size search. Note: it can not work with non RD pick mode in real-t...
Definition: aomcx.h:1383
@ AV1E_SET_MODE_COST_UPD_FREQ
Control to set frequency of the cost updates for mode, unsigned int parameter.
Definition: aomcx.h:1240
@ AV1E_SET_MIN_CR
Control to set minimum compression ratio, unsigned int parameter Take integer values....
Definition: aomcx.h:1265
@ AV1E_SET_LOSSLESS
Codec control function to set lossless encoding mode, unsigned int parameter.
Definition: aomcx.h:353
@ AOME_SET_ARNR_STRENGTH
Codec control function to set the filter strength for the arf, unsigned int parameter.
Definition: aomcx.h:273
@ AV1_GET_NEW_FRAME_IMAGE
Codec control function to get a pointer to the new frame.
Definition: aom.h:70
const char * aom_codec_iface_name(aom_codec_iface_t *iface)
Return the name for a given interface.
aom_codec_err_t aom_codec_control(aom_codec_ctx_t *ctx, int ctrl_id,...)
Algorithm Control.
const char * aom_codec_error_detail(aom_codec_ctx_t *ctx)
Retrieve detailed error information for codec context.
const struct aom_codec_iface aom_codec_iface_t
Codec interface structure.
Definition: aom_codec.h:254
const char * aom_codec_version_str(void)
Return the version information (as a string)
const char * aom_codec_error(aom_codec_ctx_t *ctx)
Retrieve error synopsis for codec context.
aom_codec_err_t aom_codec_set_option(aom_codec_ctx_t *ctx, const char *name, const char *value)
Key & Value API.
int64_t aom_codec_pts_t
Time Stamp Type.
Definition: aom_codec.h:235
aom_codec_err_t aom_codec_destroy(aom_codec_ctx_t *ctx)
Destroy a codec instance.
const char * aom_codec_err_to_string(aom_codec_err_t err)
Convert error number to printable string.
aom_codec_err_t
Algorithm return codes.
Definition: aom_codec.h:155
#define AOM_CODEC_CONTROL_TYPECHECKED(ctx, id, data)
aom_codec_control wrapper macro (adds type-checking, less flexible)
Definition: aom_codec.h:521
const void * aom_codec_iter_t
Iterator.
Definition: aom_codec.h:288
@ AOM_BITS_8
Definition: aom_codec.h:319
aom_codec_err_t aom_codec_decode(aom_codec_ctx_t *ctx, const uint8_t *data, size_t data_sz, void *user_priv)
Decode data.
#define aom_codec_dec_init(ctx, iface, cfg, flags)
Convenience macro for aom_codec_dec_init_ver()
Definition: aom_decoder.h:129
#define AOM_USAGE_GOOD_QUALITY
usage parameter analogous to AV1 GOOD QUALITY mode.
Definition: aom_encoder.h:1000
#define AOM_USAGE_ALL_INTRA
usage parameter analogous to AV1 all intra mode.
Definition: aom_encoder.h:1004
const aom_codec_cx_pkt_t * aom_codec_get_cx_data(aom_codec_ctx_t *ctx, aom_codec_iter_t *iter)
Encoded data iterator.
aom_codec_err_t aom_codec_encode(aom_codec_ctx_t *ctx, const aom_image_t *img, aom_codec_pts_t pts, unsigned long duration, aom_enc_frame_flags_t flags)
Encode a frame.
#define aom_codec_enc_init(ctx, iface, cfg, flags)
Convenience macro for aom_codec_enc_init_ver()
Definition: aom_encoder.h:929
aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface, aom_codec_enc_cfg_t *cfg, unsigned int usage)
Get the default configuration for a usage.
#define AOM_USAGE_REALTIME
usage parameter analogous to AV1 REALTIME mode.
Definition: aom_encoder.h:1002
#define AOM_CODEC_USE_HIGHBITDEPTH
Make the encoder output one partition at a time.
Definition: aom_encoder.h:81
#define AOM_CODEC_USE_PSNR
Initialization-time Feature Enabling.
Definition: aom_encoder.h:79
@ AOM_RC_ONE_PASS
Definition: aom_encoder.h:176
@ AOM_RC_SECOND_PASS
Definition: aom_encoder.h:178
@ AOM_RC_THIRD_PASS
Definition: aom_encoder.h:179
@ AOM_RC_FIRST_PASS
Definition: aom_encoder.h:177
@ AOM_KF_DISABLED
Definition: aom_encoder.h:202
@ AOM_CODEC_PSNR_PKT
Definition: aom_encoder.h:112
@ AOM_CODEC_CX_FRAME_PKT
Definition: aom_encoder.h:109
@ AOM_CODEC_STATS_PKT
Definition: aom_encoder.h:110
Codec context structure.
Definition: aom_codec.h:298
Encoder output packet.
Definition: aom_encoder.h:121
size_t sz
Definition: aom_encoder.h:126
enum aom_codec_cx_pkt_kind kind
Definition: aom_encoder.h:122
double psnr[4]
Definition: aom_encoder.h:144
aom_fixed_buf_t twopass_stats
Definition: aom_encoder.h:139
aom_fixed_buf_t raw
Definition: aom_encoder.h:155
union aom_codec_cx_pkt::@1 data
aom_codec_pts_t pts
time stamp to show frame (in timebase units)
Definition: aom_encoder.h:128
struct aom_codec_cx_pkt::@1::@2 frame
int partition_id
the partition id defines the decoding order of the partitions. Only applicable when "output partition...
Definition: aom_encoder.h:135
void * buf
Definition: aom_encoder.h:125
Initialization Configurations.
Definition: aom_decoder.h:91
Encoder configuration structure.
Definition: aom_encoder.h:386
struct aom_rational g_timebase
Stream timebase units.
Definition: aom_encoder.h:483
unsigned int g_h
Height of the frame.
Definition: aom_encoder.h:434
unsigned int monochrome
Monochrome mode.
Definition: aom_encoder.h:816
unsigned int g_w
Width of the frame.
Definition: aom_encoder.h:425
enum aom_enc_pass g_pass
Multi-pass Encoding Mode.
Definition: aom_encoder.h:498
size_t sz
Definition: aom_encoder.h:89
void * buf
Definition: aom_encoder.h:88
Image Descriptor.
Definition: aom_image.h:180
aom_chroma_sample_position_t csp
Definition: aom_image.h:186
unsigned int y_chroma_shift
Definition: aom_image.h:204
aom_img_fmt_t fmt
Definition: aom_image.h:181
int stride[3]
Definition: aom_image.h:214
unsigned char * img_data
Definition: aom_image.h:228
unsigned int x_chroma_shift
Definition: aom_image.h:203
unsigned int d_w
Definition: aom_image.h:195
int bps
Definition: aom_image.h:217
int monochrome
Definition: aom_image.h:185
unsigned int d_h
Definition: aom_image.h:196
unsigned char * planes[3]
Definition: aom_image.h:213
int img_data_owner
Definition: aom_image.h:229
int self_allocd
Definition: aom_image.h:230
size_t sz
Definition: aom_image.h:215
Rational Number.
Definition: aom_encoder.h:163
int num
Definition: aom_encoder.h:164
int den
Definition: aom_encoder.h:165
Encoder Config Options.
Definition: aom_encoder.h:226
unsigned int min_partition_size
min partition size 8, 16, 32, 64, 128
Definition: aom_encoder.h:242
unsigned int max_partition_size
max partition size 8, 16, 32, 64, 128
Definition: aom_encoder.h:238
unsigned int disable_trellis_quant
disable trellis quantization
Definition: aom_encoder.h:354
unsigned int super_block_size
Superblock size 0, 64 or 128.
Definition: aom_encoder.h:234