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 "third_party/libyuv/include/libyuv/scale.h"
51 #endif
52
53 /* Swallow warnings about unused results of fread/fwrite */
wrap_fread(void * ptr,size_t size,size_t nmemb,FILE * stream)54 static 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
wrap_fwrite(const void * ptr,size_t size,size_t nmemb,FILE * stream)59 static 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
65 static const char *exec_name;
66
warn_or_exit_on_errorv(aom_codec_ctx_t * ctx,int fatal,const char * s,va_list ap)67 static 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) {
78 aom_codec_destroy(ctx);
79 exit(EXIT_FAILURE);
80 }
81 }
82 }
83
84 static AOM_TOOLS_FORMAT_PRINTF(2,
ctx_exit_on_error(aom_codec_ctx_t * ctx,const char * s,...)85 3) void ctx_exit_on_error(aom_codec_ctx_t *ctx,
86 const char *s, ...) {
87 va_list ap;
88
89 va_start(ap, s);
90 warn_or_exit_on_errorv(ctx, 1, s, ap);
91 va_end(ap);
92 }
93
warn_or_exit_on_error(aom_codec_ctx_t * ctx,int fatal,const char * s,...)94 static AOM_TOOLS_FORMAT_PRINTF(3, 4) void warn_or_exit_on_error(
95 aom_codec_ctx_t *ctx, int fatal, const char *s, ...) {
96 va_list ap;
97
98 va_start(ap, s);
99 warn_or_exit_on_errorv(ctx, fatal, s, ap);
100 va_end(ap);
101 }
102
read_frame(struct AvxInputContext * input_ctx,aom_image_t * img)103 static int read_frame(struct AvxInputContext *input_ctx, aom_image_t *img) {
104 FILE *f = input_ctx->file;
105 y4m_input *y4m = &input_ctx->y4m;
106 int shortread = 0;
107
108 if (input_ctx->file_type == FILE_TYPE_Y4M) {
109 if (y4m_input_fetch_frame(y4m, f, img) < 1) return 0;
110 } else {
111 shortread = read_yuv_frame(input_ctx, img);
112 }
113
114 return !shortread;
115 }
116
file_is_y4m(const char detect[4])117 static int file_is_y4m(const char detect[4]) {
118 if (memcmp(detect, "YUV4", 4) == 0) {
119 return 1;
120 }
121 return 0;
122 }
123
fourcc_is_ivf(const char detect[4])124 static int fourcc_is_ivf(const char detect[4]) {
125 if (memcmp(detect, "DKIF", 4) == 0) {
126 return 1;
127 }
128 return 0;
129 }
130
131 static const int av1_arg_ctrl_map[] = { AOME_SET_CPUUSED,
132 AOME_SET_ENABLEAUTOALTREF,
133 AOME_SET_SHARPNESS,
134 AOME_SET_STATIC_THRESHOLD,
135 AV1E_SET_ROW_MT,
136 AV1E_SET_FP_MT,
137 AV1E_SET_TILE_COLUMNS,
138 AV1E_SET_TILE_ROWS,
139 AV1E_SET_ENABLE_TPL_MODEL,
140 AV1E_SET_ENABLE_KEYFRAME_FILTERING,
141 AOME_SET_ARNR_MAXFRAMES,
142 AOME_SET_ARNR_STRENGTH,
143 AOME_SET_TUNING,
144 AOME_SET_CQ_LEVEL,
145 AOME_SET_MAX_INTRA_BITRATE_PCT,
146 AV1E_SET_MAX_INTER_BITRATE_PCT,
147 AV1E_SET_GF_CBR_BOOST_PCT,
148 AV1E_SET_LOSSLESS,
149 AV1E_SET_ENABLE_CDEF,
150 AV1E_SET_ENABLE_RESTORATION,
151 AV1E_SET_ENABLE_RECT_PARTITIONS,
152 AV1E_SET_ENABLE_AB_PARTITIONS,
153 AV1E_SET_ENABLE_1TO4_PARTITIONS,
154 AV1E_SET_MIN_PARTITION_SIZE,
155 AV1E_SET_MAX_PARTITION_SIZE,
156 AV1E_SET_ENABLE_DUAL_FILTER,
157 AV1E_SET_ENABLE_CHROMA_DELTAQ,
158 AV1E_SET_ENABLE_INTRA_EDGE_FILTER,
159 AV1E_SET_ENABLE_ORDER_HINT,
160 AV1E_SET_ENABLE_TX64,
161 AV1E_SET_ENABLE_FLIP_IDTX,
162 AV1E_SET_ENABLE_RECT_TX,
163 AV1E_SET_ENABLE_DIST_WTD_COMP,
164 AV1E_SET_ENABLE_MASKED_COMP,
165 AV1E_SET_ENABLE_ONESIDED_COMP,
166 AV1E_SET_ENABLE_INTERINTRA_COMP,
167 AV1E_SET_ENABLE_SMOOTH_INTERINTRA,
168 AV1E_SET_ENABLE_DIFF_WTD_COMP,
169 AV1E_SET_ENABLE_INTERINTER_WEDGE,
170 AV1E_SET_ENABLE_INTERINTRA_WEDGE,
171 AV1E_SET_ENABLE_GLOBAL_MOTION,
172 AV1E_SET_ENABLE_WARPED_MOTION,
173 AV1E_SET_ENABLE_FILTER_INTRA,
174 AV1E_SET_ENABLE_SMOOTH_INTRA,
175 AV1E_SET_ENABLE_PAETH_INTRA,
176 AV1E_SET_ENABLE_CFL_INTRA,
177 AV1E_SET_ENABLE_DIAGONAL_INTRA,
178 AV1E_SET_FORCE_VIDEO_MODE,
179 AV1E_SET_ENABLE_OBMC,
180 AV1E_SET_ENABLE_OVERLAY,
181 AV1E_SET_ENABLE_PALETTE,
182 AV1E_SET_ENABLE_INTRABC,
183 AV1E_SET_ENABLE_ANGLE_DELTA,
184 AV1E_SET_DISABLE_TRELLIS_QUANT,
185 AV1E_SET_ENABLE_QM,
186 AV1E_SET_QM_MIN,
187 AV1E_SET_QM_MAX,
188 AV1E_SET_REDUCED_TX_TYPE_SET,
189 AV1E_SET_INTRA_DCT_ONLY,
190 AV1E_SET_INTER_DCT_ONLY,
191 AV1E_SET_INTRA_DEFAULT_TX_ONLY,
192 AV1E_SET_QUANT_B_ADAPT,
193 AV1E_SET_COEFF_COST_UPD_FREQ,
194 AV1E_SET_MODE_COST_UPD_FREQ,
195 AV1E_SET_MV_COST_UPD_FREQ,
196 AV1E_SET_FRAME_PARALLEL_DECODING,
197 AV1E_SET_ERROR_RESILIENT_MODE,
198 AV1E_SET_AQ_MODE,
199 AV1E_SET_DELTAQ_MODE,
200 AV1E_SET_DELTAQ_STRENGTH,
201 AV1E_SET_DELTALF_MODE,
202 AV1E_SET_FRAME_PERIODIC_BOOST,
203 AV1E_SET_NOISE_SENSITIVITY,
204 AV1E_SET_TUNE_CONTENT,
205 AV1E_SET_CDF_UPDATE_MODE,
206 AV1E_SET_COLOR_PRIMARIES,
207 AV1E_SET_TRANSFER_CHARACTERISTICS,
208 AV1E_SET_MATRIX_COEFFICIENTS,
209 AV1E_SET_CHROMA_SAMPLE_POSITION,
210 AV1E_SET_MIN_GF_INTERVAL,
211 AV1E_SET_MAX_GF_INTERVAL,
212 AV1E_SET_GF_MIN_PYRAMID_HEIGHT,
213 AV1E_SET_GF_MAX_PYRAMID_HEIGHT,
214 AV1E_SET_SUPERBLOCK_SIZE,
215 AV1E_SET_NUM_TG,
216 AV1E_SET_MTU,
217 AV1E_SET_TIMING_INFO_TYPE,
218 AV1E_SET_FILM_GRAIN_TEST_VECTOR,
219 AV1E_SET_FILM_GRAIN_TABLE,
220 #if CONFIG_DENOISE
221 AV1E_SET_DENOISE_NOISE_LEVEL,
222 AV1E_SET_DENOISE_BLOCK_SIZE,
223 AV1E_SET_ENABLE_DNL_DENOISING,
224 #endif // CONFIG_DENOISE
225 AV1E_SET_MAX_REFERENCE_FRAMES,
226 AV1E_SET_REDUCED_REFERENCE_SET,
227 AV1E_SET_ENABLE_REF_FRAME_MVS,
228 AV1E_SET_TARGET_SEQ_LEVEL_IDX,
229 AV1E_SET_TIER_MASK,
230 AV1E_SET_MIN_CR,
231 AV1E_SET_VBR_CORPUS_COMPLEXITY_LAP,
232 AV1E_SET_CHROMA_SUBSAMPLING_X,
233 AV1E_SET_CHROMA_SUBSAMPLING_Y,
234 #if CONFIG_TUNE_VMAF
235 AV1E_SET_VMAF_MODEL_PATH,
236 #endif
237 AV1E_SET_DV_COST_UPD_FREQ,
238 AV1E_SET_PARTITION_INFO_PATH,
239 AV1E_SET_ENABLE_DIRECTIONAL_INTRA,
240 AV1E_SET_ENABLE_TX_SIZE_SEARCH,
241 AV1E_SET_LOOPFILTER_CONTROL,
242 AV1E_SET_AUTO_INTRA_TOOLS_OFF,
243 AV1E_ENABLE_RATE_GUIDE_DELTAQ,
244 AV1E_SET_RATE_DISTRIBUTION_INFO,
245 0 };
246
247 static const arg_def_t *const main_args[] = {
248 &g_av1_codec_arg_defs.help,
249 &g_av1_codec_arg_defs.use_cfg,
250 &g_av1_codec_arg_defs.debugmode,
251 &g_av1_codec_arg_defs.outputfile,
252 &g_av1_codec_arg_defs.codecarg,
253 &g_av1_codec_arg_defs.passes,
254 &g_av1_codec_arg_defs.pass_arg,
255 &g_av1_codec_arg_defs.fpf_name,
256 &g_av1_codec_arg_defs.limit,
257 &g_av1_codec_arg_defs.skip,
258 &g_av1_codec_arg_defs.good_dl,
259 &g_av1_codec_arg_defs.rt_dl,
260 &g_av1_codec_arg_defs.ai_dl,
261 &g_av1_codec_arg_defs.quietarg,
262 &g_av1_codec_arg_defs.verbosearg,
263 &g_av1_codec_arg_defs.psnrarg,
264 &g_av1_codec_arg_defs.use_webm,
265 &g_av1_codec_arg_defs.use_ivf,
266 &g_av1_codec_arg_defs.use_obu,
267 &g_av1_codec_arg_defs.q_hist_n,
268 &g_av1_codec_arg_defs.rate_hist_n,
269 &g_av1_codec_arg_defs.disable_warnings,
270 &g_av1_codec_arg_defs.disable_warning_prompt,
271 &g_av1_codec_arg_defs.recontest,
272 NULL
273 };
274
275 static const arg_def_t *const global_args[] = {
276 &g_av1_codec_arg_defs.use_nv12,
277 &g_av1_codec_arg_defs.use_yv12,
278 &g_av1_codec_arg_defs.use_i420,
279 &g_av1_codec_arg_defs.use_i422,
280 &g_av1_codec_arg_defs.use_i444,
281 &g_av1_codec_arg_defs.usage,
282 &g_av1_codec_arg_defs.threads,
283 &g_av1_codec_arg_defs.profile,
284 &g_av1_codec_arg_defs.width,
285 &g_av1_codec_arg_defs.height,
286 &g_av1_codec_arg_defs.forced_max_frame_width,
287 &g_av1_codec_arg_defs.forced_max_frame_height,
288 #if CONFIG_WEBM_IO
289 &g_av1_codec_arg_defs.stereo_mode,
290 #endif
291 &g_av1_codec_arg_defs.timebase,
292 &g_av1_codec_arg_defs.framerate,
293 &g_av1_codec_arg_defs.global_error_resilient,
294 &g_av1_codec_arg_defs.bitdeptharg,
295 &g_av1_codec_arg_defs.inbitdeptharg,
296 &g_av1_codec_arg_defs.lag_in_frames,
297 &g_av1_codec_arg_defs.large_scale_tile,
298 &g_av1_codec_arg_defs.monochrome,
299 &g_av1_codec_arg_defs.full_still_picture_hdr,
300 &g_av1_codec_arg_defs.use_16bit_internal,
301 &g_av1_codec_arg_defs.save_as_annexb,
302 NULL
303 };
304
305 static const arg_def_t *const rc_args[] = {
306 &g_av1_codec_arg_defs.dropframe_thresh,
307 &g_av1_codec_arg_defs.resize_mode,
308 &g_av1_codec_arg_defs.resize_denominator,
309 &g_av1_codec_arg_defs.resize_kf_denominator,
310 &g_av1_codec_arg_defs.superres_mode,
311 &g_av1_codec_arg_defs.superres_denominator,
312 &g_av1_codec_arg_defs.superres_kf_denominator,
313 &g_av1_codec_arg_defs.superres_qthresh,
314 &g_av1_codec_arg_defs.superres_kf_qthresh,
315 &g_av1_codec_arg_defs.end_usage,
316 &g_av1_codec_arg_defs.target_bitrate,
317 &g_av1_codec_arg_defs.min_quantizer,
318 &g_av1_codec_arg_defs.max_quantizer,
319 &g_av1_codec_arg_defs.undershoot_pct,
320 &g_av1_codec_arg_defs.overshoot_pct,
321 &g_av1_codec_arg_defs.buf_sz,
322 &g_av1_codec_arg_defs.buf_initial_sz,
323 &g_av1_codec_arg_defs.buf_optimal_sz,
324 &g_av1_codec_arg_defs.bias_pct,
325 &g_av1_codec_arg_defs.minsection_pct,
326 &g_av1_codec_arg_defs.maxsection_pct,
327 NULL
328 };
329
330 static const arg_def_t *const kf_args[] = {
331 &g_av1_codec_arg_defs.fwd_kf_enabled,
332 &g_av1_codec_arg_defs.kf_min_dist,
333 &g_av1_codec_arg_defs.kf_max_dist,
334 &g_av1_codec_arg_defs.kf_disabled,
335 &g_av1_codec_arg_defs.sframe_dist,
336 &g_av1_codec_arg_defs.sframe_mode,
337 NULL
338 };
339
340 // TODO(bohanli): Currently all options are supported by the key & value API.
341 // Consider removing the control ID usages?
342 static const arg_def_t *const av1_ctrl_args[] = {
343 &g_av1_codec_arg_defs.cpu_used_av1,
344 &g_av1_codec_arg_defs.auto_altref,
345 &g_av1_codec_arg_defs.sharpness,
346 &g_av1_codec_arg_defs.static_thresh,
347 &g_av1_codec_arg_defs.rowmtarg,
348 &g_av1_codec_arg_defs.fpmtarg,
349 &g_av1_codec_arg_defs.tile_cols,
350 &g_av1_codec_arg_defs.tile_rows,
351 &g_av1_codec_arg_defs.enable_tpl_model,
352 &g_av1_codec_arg_defs.enable_keyframe_filtering,
353 &g_av1_codec_arg_defs.arnr_maxframes,
354 &g_av1_codec_arg_defs.arnr_strength,
355 &g_av1_codec_arg_defs.tune_metric,
356 &g_av1_codec_arg_defs.cq_level,
357 &g_av1_codec_arg_defs.max_intra_rate_pct,
358 &g_av1_codec_arg_defs.max_inter_rate_pct,
359 &g_av1_codec_arg_defs.gf_cbr_boost_pct,
360 &g_av1_codec_arg_defs.lossless,
361 &g_av1_codec_arg_defs.enable_cdef,
362 &g_av1_codec_arg_defs.enable_restoration,
363 &g_av1_codec_arg_defs.enable_rect_partitions,
364 &g_av1_codec_arg_defs.enable_ab_partitions,
365 &g_av1_codec_arg_defs.enable_1to4_partitions,
366 &g_av1_codec_arg_defs.min_partition_size,
367 &g_av1_codec_arg_defs.max_partition_size,
368 &g_av1_codec_arg_defs.enable_dual_filter,
369 &g_av1_codec_arg_defs.enable_chroma_deltaq,
370 &g_av1_codec_arg_defs.enable_intra_edge_filter,
371 &g_av1_codec_arg_defs.enable_order_hint,
372 &g_av1_codec_arg_defs.enable_tx64,
373 &g_av1_codec_arg_defs.enable_flip_idtx,
374 &g_av1_codec_arg_defs.enable_rect_tx,
375 &g_av1_codec_arg_defs.enable_dist_wtd_comp,
376 &g_av1_codec_arg_defs.enable_masked_comp,
377 &g_av1_codec_arg_defs.enable_onesided_comp,
378 &g_av1_codec_arg_defs.enable_interintra_comp,
379 &g_av1_codec_arg_defs.enable_smooth_interintra,
380 &g_av1_codec_arg_defs.enable_diff_wtd_comp,
381 &g_av1_codec_arg_defs.enable_interinter_wedge,
382 &g_av1_codec_arg_defs.enable_interintra_wedge,
383 &g_av1_codec_arg_defs.enable_global_motion,
384 &g_av1_codec_arg_defs.enable_warped_motion,
385 &g_av1_codec_arg_defs.enable_filter_intra,
386 &g_av1_codec_arg_defs.enable_smooth_intra,
387 &g_av1_codec_arg_defs.enable_paeth_intra,
388 &g_av1_codec_arg_defs.enable_cfl_intra,
389 &g_av1_codec_arg_defs.enable_diagonal_intra,
390 &g_av1_codec_arg_defs.force_video_mode,
391 &g_av1_codec_arg_defs.enable_obmc,
392 &g_av1_codec_arg_defs.enable_overlay,
393 &g_av1_codec_arg_defs.enable_palette,
394 &g_av1_codec_arg_defs.enable_intrabc,
395 &g_av1_codec_arg_defs.enable_angle_delta,
396 &g_av1_codec_arg_defs.disable_trellis_quant,
397 &g_av1_codec_arg_defs.enable_qm,
398 &g_av1_codec_arg_defs.qm_min,
399 &g_av1_codec_arg_defs.qm_max,
400 &g_av1_codec_arg_defs.reduced_tx_type_set,
401 &g_av1_codec_arg_defs.use_intra_dct_only,
402 &g_av1_codec_arg_defs.use_inter_dct_only,
403 &g_av1_codec_arg_defs.use_intra_default_tx_only,
404 &g_av1_codec_arg_defs.quant_b_adapt,
405 &g_av1_codec_arg_defs.coeff_cost_upd_freq,
406 &g_av1_codec_arg_defs.mode_cost_upd_freq,
407 &g_av1_codec_arg_defs.mv_cost_upd_freq,
408 &g_av1_codec_arg_defs.frame_parallel_decoding,
409 &g_av1_codec_arg_defs.error_resilient_mode,
410 &g_av1_codec_arg_defs.aq_mode,
411 &g_av1_codec_arg_defs.deltaq_mode,
412 &g_av1_codec_arg_defs.deltaq_strength,
413 &g_av1_codec_arg_defs.deltalf_mode,
414 &g_av1_codec_arg_defs.frame_periodic_boost,
415 &g_av1_codec_arg_defs.noise_sens,
416 &g_av1_codec_arg_defs.tune_content,
417 &g_av1_codec_arg_defs.cdf_update_mode,
418 &g_av1_codec_arg_defs.input_color_primaries,
419 &g_av1_codec_arg_defs.input_transfer_characteristics,
420 &g_av1_codec_arg_defs.input_matrix_coefficients,
421 &g_av1_codec_arg_defs.input_chroma_sample_position,
422 &g_av1_codec_arg_defs.min_gf_interval,
423 &g_av1_codec_arg_defs.max_gf_interval,
424 &g_av1_codec_arg_defs.gf_min_pyr_height,
425 &g_av1_codec_arg_defs.gf_max_pyr_height,
426 &g_av1_codec_arg_defs.superblock_size,
427 &g_av1_codec_arg_defs.num_tg,
428 &g_av1_codec_arg_defs.mtu_size,
429 &g_av1_codec_arg_defs.timing_info,
430 &g_av1_codec_arg_defs.film_grain_test,
431 &g_av1_codec_arg_defs.film_grain_table,
432 #if CONFIG_DENOISE
433 &g_av1_codec_arg_defs.denoise_noise_level,
434 &g_av1_codec_arg_defs.denoise_block_size,
435 &g_av1_codec_arg_defs.enable_dnl_denoising,
436 #endif // CONFIG_DENOISE
437 &g_av1_codec_arg_defs.max_reference_frames,
438 &g_av1_codec_arg_defs.reduced_reference_set,
439 &g_av1_codec_arg_defs.enable_ref_frame_mvs,
440 &g_av1_codec_arg_defs.target_seq_level_idx,
441 &g_av1_codec_arg_defs.set_tier_mask,
442 &g_av1_codec_arg_defs.set_min_cr,
443 &g_av1_codec_arg_defs.vbr_corpus_complexity_lap,
444 &g_av1_codec_arg_defs.input_chroma_subsampling_x,
445 &g_av1_codec_arg_defs.input_chroma_subsampling_y,
446 #if CONFIG_TUNE_VMAF
447 &g_av1_codec_arg_defs.vmaf_model_path,
448 #endif
449 &g_av1_codec_arg_defs.dv_cost_upd_freq,
450 &g_av1_codec_arg_defs.partition_info_path,
451 &g_av1_codec_arg_defs.enable_directional_intra,
452 &g_av1_codec_arg_defs.enable_tx_size_search,
453 &g_av1_codec_arg_defs.loopfilter_control,
454 &g_av1_codec_arg_defs.auto_intra_tools_off,
455 &g_av1_codec_arg_defs.enable_rate_guide_deltaq,
456 &g_av1_codec_arg_defs.rate_distribution_info,
457 NULL,
458 };
459
460 static const arg_def_t *const av1_key_val_args[] = {
461 &g_av1_codec_arg_defs.passes,
462 &g_av1_codec_arg_defs.two_pass_output,
463 &g_av1_codec_arg_defs.second_pass_log,
464 &g_av1_codec_arg_defs.fwd_kf_dist,
465 &g_av1_codec_arg_defs.strict_level_conformance,
466 &g_av1_codec_arg_defs.sb_qp_sweep,
467 &g_av1_codec_arg_defs.dist_metric,
468 &g_av1_codec_arg_defs.kf_max_pyr_height,
469 &g_av1_codec_arg_defs.auto_tiles,
470 NULL,
471 };
472
473 static const arg_def_t *const no_args[] = { NULL };
474
show_help(FILE * fout,int shorthelp)475 static void show_help(FILE *fout, int shorthelp) {
476 fprintf(fout, "Usage: %s <options> -o dst_filename src_filename\n",
477 exec_name);
478
479 if (shorthelp) {
480 fprintf(fout, "Use --help to see the full list of options.\n");
481 return;
482 }
483
484 fprintf(fout, "\nOptions:\n");
485 arg_show_usage(fout, main_args);
486 fprintf(fout, "\nEncoder Global Options:\n");
487 arg_show_usage(fout, global_args);
488 fprintf(fout, "\nRate Control Options:\n");
489 arg_show_usage(fout, rc_args);
490 fprintf(fout, "\nKeyframe Placement Options:\n");
491 arg_show_usage(fout, kf_args);
492 #if CONFIG_AV1_ENCODER
493 fprintf(fout, "\nAV1 Specific Options:\n");
494 arg_show_usage(fout, av1_ctrl_args);
495 arg_show_usage(fout, av1_key_val_args);
496 #endif
497 fprintf(fout,
498 "\nStream timebase (--timebase):\n"
499 " The desired precision of timestamps in the output, expressed\n"
500 " in fractional seconds. Default is 1/1000.\n");
501 fprintf(fout, "\nIncluded encoders:\n\n");
502
503 const int num_encoder = get_aom_encoder_count();
504 for (int i = 0; i < num_encoder; ++i) {
505 aom_codec_iface_t *encoder = get_aom_encoder_by_index(i);
506 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
507 fprintf(fout, " %-6s - %s %s\n", get_short_name_by_aom_encoder(encoder),
508 aom_codec_iface_name(encoder), defstr);
509 }
510 fprintf(fout, "\n ");
511 fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
512 }
513
usage_exit(void)514 void usage_exit(void) {
515 show_help(stderr, 1);
516 exit(EXIT_FAILURE);
517 }
518
519 #if CONFIG_AV1_ENCODER
520 #define ARG_CTRL_CNT_MAX NELEMENTS(av1_arg_ctrl_map)
521 #define ARG_KEY_VAL_CNT_MAX NELEMENTS(av1_key_val_args)
522 #endif
523
524 #if !CONFIG_WEBM_IO
525 typedef int stereo_format_t;
526 struct WebmOutputContext {
527 int debug;
528 };
529 #endif
530
531 /* Per-stream configuration */
532 struct stream_config {
533 struct aom_codec_enc_cfg cfg;
534 const char *out_fn;
535 const char *stats_fn;
536 stereo_format_t stereo_fmt;
537 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
538 int arg_ctrl_cnt;
539 const char *arg_key_vals[ARG_KEY_VAL_CNT_MAX][2];
540 int arg_key_val_cnt;
541 int write_webm;
542 const char *film_grain_filename;
543 int write_ivf;
544 // whether to use 16bit internal buffers
545 int use_16bit_internal;
546 #if CONFIG_TUNE_VMAF
547 const char *vmaf_model_path;
548 #endif
549 const char *partition_info_path;
550 unsigned int enable_rate_guide_deltaq;
551 const char *rate_distribution_info;
552 aom_color_range_t color_range;
553 const char *two_pass_input;
554 const char *two_pass_output;
555 int two_pass_width;
556 int two_pass_height;
557 };
558
559 struct stream_state {
560 int index;
561 struct stream_state *next;
562 struct stream_config config;
563 FILE *file;
564 struct rate_hist *rate_hist;
565 struct WebmOutputContext webm_ctx;
566 uint64_t psnr_sse_total[2];
567 uint64_t psnr_samples_total[2];
568 double psnr_totals[2][4];
569 int psnr_count[2];
570 int counts[64];
571 aom_codec_ctx_t encoder;
572 unsigned int frames_out;
573 uint64_t cx_time;
574 size_t nbytes;
575 stats_io_t stats;
576 struct aom_image *img;
577 aom_codec_ctx_t decoder;
578 int mismatch_seen;
579 unsigned int chroma_subsampling_x;
580 unsigned int chroma_subsampling_y;
581 const char *orig_out_fn;
582 unsigned int orig_width;
583 unsigned int orig_height;
584 int orig_write_webm;
585 int orig_write_ivf;
586 char tmp_out_fn[1000];
587 };
588
validate_positive_rational(const char * msg,struct aom_rational * rat)589 static void validate_positive_rational(const char *msg,
590 struct aom_rational *rat) {
591 if (rat->den < 0) {
592 rat->num *= -1;
593 rat->den *= -1;
594 }
595
596 if (rat->num < 0) die("Error: %s must be positive\n", msg);
597
598 if (!rat->den) die("Error: %s has zero denominator\n", msg);
599 }
600
init_config(cfg_options_t * config)601 static void init_config(cfg_options_t *config) {
602 memset(config, 0, sizeof(cfg_options_t));
603 config->super_block_size = 0; // Dynamic
604 config->max_partition_size = 128;
605 config->min_partition_size = 4;
606 config->disable_trellis_quant = 3;
607 }
608
609 /* Parses global config arguments into the AvxEncoderConfig. Note that
610 * argv is modified and overwrites all parsed arguments.
611 */
parse_global_config(struct AvxEncoderConfig * global,char *** argv)612 static void parse_global_config(struct AvxEncoderConfig *global, char ***argv) {
613 char **argi, **argj;
614 struct arg arg;
615 const int num_encoder = get_aom_encoder_count();
616 char **argv_local = (char **)*argv;
617 if (num_encoder < 1) die("Error: no valid encoder available\n");
618
619 /* Initialize default parameters */
620 memset(global, 0, sizeof(*global));
621 global->codec = get_aom_encoder_by_index(num_encoder - 1);
622 global->passes = 0;
623 global->color_type = I420;
624 global->csp = AOM_CSP_UNKNOWN;
625 global->show_psnr = 0;
626
627 int cfg_included = 0;
628 init_config(&global->encoder_config);
629
630 for (argi = argj = argv_local; (*argj = *argi); argi += arg.argv_step) {
631 arg.argv_step = 1;
632
633 if (arg_match(&arg, &g_av1_codec_arg_defs.use_cfg, argi)) {
634 if (!cfg_included) {
635 parse_cfg(arg.val, &global->encoder_config);
636 cfg_included = 1;
637 }
638 } else if (arg_match(&arg, &g_av1_codec_arg_defs.help, argi)) {
639 show_help(stdout, 0);
640 exit(EXIT_SUCCESS);
641 } else if (arg_match(&arg, &g_av1_codec_arg_defs.codecarg, argi)) {
642 global->codec = get_aom_encoder_by_short_name(arg.val);
643 if (!global->codec)
644 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
645 } else if (arg_match(&arg, &g_av1_codec_arg_defs.passes, argi)) {
646 global->passes = arg_parse_uint(&arg);
647
648 if (global->passes < 1 || global->passes > 3)
649 die("Error: Invalid number of passes (%d)\n", global->passes);
650 } else if (arg_match(&arg, &g_av1_codec_arg_defs.pass_arg, argi)) {
651 global->pass = arg_parse_uint(&arg);
652
653 if (global->pass < 1 || global->pass > 3)
654 die("Error: Invalid pass selected (%d)\n", global->pass);
655 } else if (arg_match(&arg,
656 &g_av1_codec_arg_defs.input_chroma_sample_position,
657 argi)) {
658 global->csp = arg_parse_enum(&arg);
659 /* Flag is used by later code as well, preserve it. */
660 argj++;
661 } else if (arg_match(&arg, &g_av1_codec_arg_defs.usage, argi)) {
662 global->usage = arg_parse_uint(&arg);
663 } else if (arg_match(&arg, &g_av1_codec_arg_defs.good_dl, argi)) {
664 global->usage = AOM_USAGE_GOOD_QUALITY; // Good quality usage
665 } else if (arg_match(&arg, &g_av1_codec_arg_defs.rt_dl, argi)) {
666 global->usage = AOM_USAGE_REALTIME; // Real-time usage
667 } else if (arg_match(&arg, &g_av1_codec_arg_defs.ai_dl, argi)) {
668 global->usage = AOM_USAGE_ALL_INTRA; // All intra usage
669 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_nv12, argi)) {
670 global->color_type = NV12;
671 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_yv12, argi)) {
672 global->color_type = YV12;
673 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_i420, argi)) {
674 global->color_type = I420;
675 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_i422, argi)) {
676 global->color_type = I422;
677 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_i444, argi)) {
678 global->color_type = I444;
679 } else if (arg_match(&arg, &g_av1_codec_arg_defs.quietarg, argi)) {
680 global->quiet = 1;
681 } else if (arg_match(&arg, &g_av1_codec_arg_defs.verbosearg, argi)) {
682 global->verbose = 1;
683 } else if (arg_match(&arg, &g_av1_codec_arg_defs.limit, argi)) {
684 global->limit = arg_parse_uint(&arg);
685 } else if (arg_match(&arg, &g_av1_codec_arg_defs.skip, argi)) {
686 global->skip_frames = arg_parse_uint(&arg);
687 } else if (arg_match(&arg, &g_av1_codec_arg_defs.psnrarg, argi)) {
688 if (arg.val)
689 global->show_psnr = arg_parse_int(&arg);
690 else
691 global->show_psnr = 1;
692 } else if (arg_match(&arg, &g_av1_codec_arg_defs.recontest, argi)) {
693 global->test_decode = arg_parse_enum_or_int(&arg);
694 } else if (arg_match(&arg, &g_av1_codec_arg_defs.framerate, argi)) {
695 global->framerate = arg_parse_rational(&arg);
696 validate_positive_rational(arg.name, &global->framerate);
697 global->have_framerate = 1;
698 } else if (arg_match(&arg, &g_av1_codec_arg_defs.debugmode, argi)) {
699 global->debug = 1;
700 } else if (arg_match(&arg, &g_av1_codec_arg_defs.q_hist_n, argi)) {
701 global->show_q_hist_buckets = arg_parse_uint(&arg);
702 } else if (arg_match(&arg, &g_av1_codec_arg_defs.rate_hist_n, argi)) {
703 global->show_rate_hist_buckets = arg_parse_uint(&arg);
704 } else if (arg_match(&arg, &g_av1_codec_arg_defs.disable_warnings, argi)) {
705 global->disable_warnings = 1;
706 } else if (arg_match(&arg, &g_av1_codec_arg_defs.disable_warning_prompt,
707 argi)) {
708 global->disable_warning_prompt = 1;
709 } else {
710 argj++;
711 }
712 }
713
714 if (global->pass) {
715 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
716 if (global->pass > global->passes) {
717 aom_tools_warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
718 global->pass);
719 global->passes = global->pass;
720 }
721 }
722 /* Validate global config */
723 if (global->passes == 0) {
724 #if CONFIG_AV1_ENCODER
725 // Make default AV1 passes = 2 until there is a better quality 1-pass
726 // encoder
727 if (global->codec != NULL)
728 global->passes =
729 (strcmp(get_short_name_by_aom_encoder(global->codec), "av1") == 0 &&
730 global->usage != AOM_USAGE_REALTIME)
731 ? 2
732 : 1;
733 #else
734 global->passes = 1;
735 #endif
736 }
737
738 if (global->usage == AOM_USAGE_REALTIME && global->passes > 1) {
739 aom_tools_warn("Enforcing one-pass encoding in realtime mode\n");
740 if (global->pass > 1)
741 die("Error: Invalid --pass=%d for one-pass encoding\n", global->pass);
742 global->passes = 1;
743 }
744
745 if (global->usage == AOM_USAGE_ALL_INTRA && global->passes > 1) {
746 aom_tools_warn("Enforcing one-pass encoding in all intra mode\n");
747 global->passes = 1;
748 }
749 }
750
open_input_file(struct AvxInputContext * input,aom_chroma_sample_position_t csp)751 static void open_input_file(struct AvxInputContext *input,
752 aom_chroma_sample_position_t csp) {
753 /* Parse certain options from the input file, if possible */
754 input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
755 : set_binary_mode(stdin);
756
757 if (!input->file) fatal("Failed to open input file");
758
759 if (!fseeko(input->file, 0, SEEK_END)) {
760 /* Input file is seekable. Figure out how long it is, so we can get
761 * progress info.
762 */
763 input->length = ftello(input->file);
764 rewind(input->file);
765 }
766
767 /* Default to 1:1 pixel aspect ratio. */
768 input->pixel_aspect_ratio.numerator = 1;
769 input->pixel_aspect_ratio.denominator = 1;
770
771 /* For RAW input sources, these bytes will applied on the first frame
772 * in read_frame().
773 */
774 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
775 input->detect.position = 0;
776
777 if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
778 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4, csp,
779 input->only_i420) >= 0) {
780 input->file_type = FILE_TYPE_Y4M;
781 input->width = input->y4m.pic_w;
782 input->height = input->y4m.pic_h;
783 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
784 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
785 input->framerate.numerator = input->y4m.fps_n;
786 input->framerate.denominator = input->y4m.fps_d;
787 input->fmt = input->y4m.aom_fmt;
788 input->bit_depth = input->y4m.bit_depth;
789 input->color_range = input->y4m.color_range;
790 } else
791 fatal("Unsupported Y4M stream.");
792 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
793 fatal("IVF is not supported as input.");
794 } else {
795 input->file_type = FILE_TYPE_RAW;
796 }
797 }
798
close_input_file(struct AvxInputContext * input)799 static void close_input_file(struct AvxInputContext *input) {
800 fclose(input->file);
801 if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
802 }
803
new_stream(struct AvxEncoderConfig * global,struct stream_state * prev)804 static struct stream_state *new_stream(struct AvxEncoderConfig *global,
805 struct stream_state *prev) {
806 struct stream_state *stream;
807
808 stream = calloc(1, sizeof(*stream));
809 if (stream == NULL) {
810 fatal("Failed to allocate new stream.");
811 }
812
813 if (prev) {
814 memcpy(stream, prev, sizeof(*stream));
815 stream->index++;
816 prev->next = stream;
817 } else {
818 aom_codec_err_t res;
819
820 /* Populate encoder configuration */
821 res = aom_codec_enc_config_default(global->codec, &stream->config.cfg,
822 global->usage);
823 if (res) fatal("Failed to get config: %s\n", aom_codec_err_to_string(res));
824
825 /* Change the default timebase to a high enough value so that the
826 * encoder will always create strictly increasing timestamps.
827 */
828 stream->config.cfg.g_timebase.den = 1000;
829
830 /* Never use the library's default resolution, require it be parsed
831 * from the file or set on the command line.
832 */
833 stream->config.cfg.g_w = 0;
834 stream->config.cfg.g_h = 0;
835
836 /* Initialize remaining stream parameters */
837 stream->config.write_webm = 1;
838 stream->config.write_ivf = 0;
839
840 #if CONFIG_WEBM_IO
841 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
842 stream->webm_ctx.last_pts_ns = -1;
843 stream->webm_ctx.writer = NULL;
844 stream->webm_ctx.segment = NULL;
845 #endif
846
847 /* Allows removal of the application version from the EBML tags */
848 stream->webm_ctx.debug = global->debug;
849 memcpy(&stream->config.cfg.encoder_cfg, &global->encoder_config,
850 sizeof(stream->config.cfg.encoder_cfg));
851 }
852
853 /* Output files must be specified for each stream */
854 stream->config.out_fn = NULL;
855 stream->config.two_pass_input = NULL;
856 stream->config.two_pass_output = NULL;
857 stream->config.two_pass_width = 0;
858 stream->config.two_pass_height = 0;
859
860 stream->next = NULL;
861 return stream;
862 }
863
set_config_arg_ctrls(struct stream_config * config,int key,const struct arg * arg)864 static void set_config_arg_ctrls(struct stream_config *config, int key,
865 const struct arg *arg) {
866 int j;
867 if (key == AV1E_SET_FILM_GRAIN_TABLE) {
868 config->film_grain_filename = arg->val;
869 return;
870 }
871
872 // For target level, the settings should accumulate rather than overwrite,
873 // so we simply append it.
874 if (key == AV1E_SET_TARGET_SEQ_LEVEL_IDX) {
875 j = config->arg_ctrl_cnt;
876 assert(j < ARG_CTRL_CNT_MAX);
877 config->arg_ctrls[j][0] = key;
878 config->arg_ctrls[j][1] = arg_parse_enum_or_int(arg);
879 ++config->arg_ctrl_cnt;
880 return;
881 }
882
883 /* Point either to the next free element or the first instance of this
884 * control.
885 */
886 for (j = 0; j < config->arg_ctrl_cnt; j++)
887 if (config->arg_ctrls[j][0] == key) break;
888
889 /* Update/insert */
890 assert(j < ARG_CTRL_CNT_MAX);
891 config->arg_ctrls[j][0] = key;
892 config->arg_ctrls[j][1] = arg_parse_enum_or_int(arg);
893
894 if (key == AOME_SET_ENABLEAUTOALTREF && config->arg_ctrls[j][1] > 1) {
895 aom_tools_warn(
896 "auto-alt-ref > 1 is deprecated... setting auto-alt-ref=1\n");
897 config->arg_ctrls[j][1] = 1;
898 }
899
900 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
901 }
902
set_config_arg_key_vals(struct stream_config * config,const char * name,const struct arg * arg)903 static void set_config_arg_key_vals(struct stream_config *config,
904 const char *name, const struct arg *arg) {
905 int j;
906 const char *val = arg->val;
907 // For target level, the settings should accumulate rather than overwrite,
908 // so we simply append it.
909 if (strcmp(name, "target-seq-level-idx") == 0) {
910 j = config->arg_key_val_cnt;
911 assert(j < ARG_KEY_VAL_CNT_MAX);
912 config->arg_key_vals[j][0] = name;
913 config->arg_key_vals[j][1] = val;
914 ++config->arg_key_val_cnt;
915 return;
916 }
917
918 /* Point either to the next free element or the first instance of this
919 * option.
920 */
921 for (j = 0; j < config->arg_key_val_cnt; j++)
922 if (strcmp(name, config->arg_key_vals[j][0]) == 0) break;
923
924 /* Update/insert */
925 assert(j < ARG_KEY_VAL_CNT_MAX);
926 config->arg_key_vals[j][0] = name;
927 config->arg_key_vals[j][1] = val;
928
929 if (strcmp(name, g_av1_codec_arg_defs.auto_altref.long_name) == 0) {
930 int auto_altref = arg_parse_int(arg);
931 if (auto_altref > 1) {
932 aom_tools_warn(
933 "auto-alt-ref > 1 is deprecated... setting auto-alt-ref=1\n");
934 config->arg_key_vals[j][1] = "1";
935 }
936 }
937
938 if (j == config->arg_key_val_cnt) config->arg_key_val_cnt++;
939 }
940
parse_stream_params(struct AvxEncoderConfig * global,struct stream_state * stream,char ** argv)941 static int parse_stream_params(struct AvxEncoderConfig *global,
942 struct stream_state *stream, char **argv) {
943 char **argi, **argj;
944 struct arg arg;
945 const arg_def_t *const *ctrl_args = no_args;
946 const arg_def_t *const *key_val_args = no_args;
947 const int *ctrl_args_map = NULL;
948 struct stream_config *config = &stream->config;
949 int eos_mark_found = 0;
950 int webm_forced = 0;
951
952 // Handle codec specific options
953 if (0) {
954 #if CONFIG_AV1_ENCODER
955 } else if (strcmp(get_short_name_by_aom_encoder(global->codec), "av1") == 0) {
956 // TODO(jingning): Reuse AV1 specific encoder configuration parameters.
957 // Consider to expand this set for AV1 encoder control.
958 #if __STDC_VERSION__ >= 201112L
959 _Static_assert(NELEMENTS(av1_ctrl_args) == NELEMENTS(av1_arg_ctrl_map),
960 "The av1_ctrl_args and av1_arg_ctrl_map arrays must be of "
961 "the same size.");
962 #else
963 assert(NELEMENTS(av1_ctrl_args) == NELEMENTS(av1_arg_ctrl_map));
964 #endif
965 ctrl_args = av1_ctrl_args;
966 ctrl_args_map = av1_arg_ctrl_map;
967 key_val_args = av1_key_val_args;
968 #endif
969 }
970
971 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
972 arg.argv_step = 1;
973
974 /* Once we've found an end-of-stream marker (--) we want to continue
975 * shifting arguments but not consuming them.
976 */
977 if (eos_mark_found) {
978 argj++;
979 continue;
980 } else if (!strcmp(*argj, "--")) {
981 eos_mark_found = 1;
982 continue;
983 }
984
985 if (arg_match(&arg, &g_av1_codec_arg_defs.outputfile, argi)) {
986 config->out_fn = arg.val;
987 if (!webm_forced) {
988 const size_t out_fn_len = strlen(config->out_fn);
989 if (out_fn_len >= 4 &&
990 !strcmp(config->out_fn + out_fn_len - 4, ".ivf")) {
991 config->write_webm = 0;
992 config->write_ivf = 1;
993 } else if (out_fn_len >= 4 &&
994 !strcmp(config->out_fn + out_fn_len - 4, ".obu")) {
995 config->write_webm = 0;
996 config->write_ivf = 0;
997 }
998 }
999 } else if (arg_match(&arg, &g_av1_codec_arg_defs.fpf_name, argi)) {
1000 config->stats_fn = arg.val;
1001 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_webm, argi)) {
1002 #if CONFIG_WEBM_IO
1003 config->write_webm = 1;
1004 webm_forced = 1;
1005 #else
1006 die("Error: --webm specified but webm is disabled.");
1007 #endif
1008 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_ivf, argi)) {
1009 config->write_webm = 0;
1010 config->write_ivf = 1;
1011 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_obu, argi)) {
1012 config->write_webm = 0;
1013 config->write_ivf = 0;
1014 } else if (arg_match(&arg, &g_av1_codec_arg_defs.threads, argi)) {
1015 config->cfg.g_threads = arg_parse_uint(&arg);
1016 } else if (arg_match(&arg, &g_av1_codec_arg_defs.profile, argi)) {
1017 config->cfg.g_profile = arg_parse_uint(&arg);
1018 } else if (arg_match(&arg, &g_av1_codec_arg_defs.width, argi)) {
1019 config->cfg.g_w = arg_parse_uint(&arg);
1020 } else if (arg_match(&arg, &g_av1_codec_arg_defs.height, argi)) {
1021 config->cfg.g_h = arg_parse_uint(&arg);
1022 } else if (arg_match(&arg, &g_av1_codec_arg_defs.forced_max_frame_width,
1023 argi)) {
1024 config->cfg.g_forced_max_frame_width = arg_parse_uint(&arg);
1025 } else if (arg_match(&arg, &g_av1_codec_arg_defs.forced_max_frame_height,
1026 argi)) {
1027 config->cfg.g_forced_max_frame_height = arg_parse_uint(&arg);
1028 } else if (arg_match(&arg, &g_av1_codec_arg_defs.bitdeptharg, argi)) {
1029 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1030 } else if (arg_match(&arg, &g_av1_codec_arg_defs.inbitdeptharg, argi)) {
1031 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1032 } else if (arg_match(&arg, &g_av1_codec_arg_defs.input_chroma_subsampling_x,
1033 argi)) {
1034 stream->chroma_subsampling_x = arg_parse_uint(&arg);
1035 } else if (arg_match(&arg, &g_av1_codec_arg_defs.input_chroma_subsampling_y,
1036 argi)) {
1037 stream->chroma_subsampling_y = arg_parse_uint(&arg);
1038 #if CONFIG_WEBM_IO
1039 } else if (arg_match(&arg, &g_av1_codec_arg_defs.stereo_mode, argi)) {
1040 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1041 #endif
1042 } else if (arg_match(&arg, &g_av1_codec_arg_defs.timebase, argi)) {
1043 config->cfg.g_timebase = arg_parse_rational(&arg);
1044 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1045 } else if (arg_match(&arg, &g_av1_codec_arg_defs.global_error_resilient,
1046 argi)) {
1047 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1048 } else if (arg_match(&arg, &g_av1_codec_arg_defs.lag_in_frames, argi)) {
1049 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1050 } else if (arg_match(&arg, &g_av1_codec_arg_defs.large_scale_tile, argi)) {
1051 config->cfg.large_scale_tile = arg_parse_uint(&arg);
1052 if (config->cfg.large_scale_tile) {
1053 global->codec = get_aom_encoder_by_short_name("av1");
1054 }
1055 } else if (arg_match(&arg, &g_av1_codec_arg_defs.monochrome, argi)) {
1056 config->cfg.monochrome = 1;
1057 } else if (arg_match(&arg, &g_av1_codec_arg_defs.full_still_picture_hdr,
1058 argi)) {
1059 config->cfg.full_still_picture_hdr = 1;
1060 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_16bit_internal,
1061 argi)) {
1062 config->use_16bit_internal = CONFIG_AV1_HIGHBITDEPTH;
1063 if (!config->use_16bit_internal) {
1064 aom_tools_warn("%s option ignored with CONFIG_AV1_HIGHBITDEPTH=0.\n",
1065 arg.name);
1066 }
1067 } else if (arg_match(&arg, &g_av1_codec_arg_defs.dropframe_thresh, argi)) {
1068 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1069 } else if (arg_match(&arg, &g_av1_codec_arg_defs.resize_mode, argi)) {
1070 config->cfg.rc_resize_mode = arg_parse_uint(&arg);
1071 } else if (arg_match(&arg, &g_av1_codec_arg_defs.resize_denominator,
1072 argi)) {
1073 config->cfg.rc_resize_denominator = arg_parse_uint(&arg);
1074 } else if (arg_match(&arg, &g_av1_codec_arg_defs.resize_kf_denominator,
1075 argi)) {
1076 config->cfg.rc_resize_kf_denominator = arg_parse_uint(&arg);
1077 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_mode, argi)) {
1078 config->cfg.rc_superres_mode = arg_parse_uint(&arg);
1079 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_denominator,
1080 argi)) {
1081 config->cfg.rc_superres_denominator = arg_parse_uint(&arg);
1082 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_kf_denominator,
1083 argi)) {
1084 config->cfg.rc_superres_kf_denominator = arg_parse_uint(&arg);
1085 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_qthresh, argi)) {
1086 config->cfg.rc_superres_qthresh = arg_parse_uint(&arg);
1087 } else if (arg_match(&arg, &g_av1_codec_arg_defs.superres_kf_qthresh,
1088 argi)) {
1089 config->cfg.rc_superres_kf_qthresh = arg_parse_uint(&arg);
1090 } else if (arg_match(&arg, &g_av1_codec_arg_defs.end_usage, argi)) {
1091 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1092 } else if (arg_match(&arg, &g_av1_codec_arg_defs.target_bitrate, argi)) {
1093 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1094 } else if (arg_match(&arg, &g_av1_codec_arg_defs.min_quantizer, argi)) {
1095 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1096 } else if (arg_match(&arg, &g_av1_codec_arg_defs.max_quantizer, argi)) {
1097 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1098 } else if (arg_match(&arg, &g_av1_codec_arg_defs.undershoot_pct, argi)) {
1099 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1100 } else if (arg_match(&arg, &g_av1_codec_arg_defs.overshoot_pct, argi)) {
1101 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1102 } else if (arg_match(&arg, &g_av1_codec_arg_defs.buf_sz, argi)) {
1103 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1104 } else if (arg_match(&arg, &g_av1_codec_arg_defs.buf_initial_sz, argi)) {
1105 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1106 } else if (arg_match(&arg, &g_av1_codec_arg_defs.buf_optimal_sz, argi)) {
1107 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1108 } else if (arg_match(&arg, &g_av1_codec_arg_defs.bias_pct, argi)) {
1109 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1110 if (global->passes < 2)
1111 aom_tools_warn("option %s ignored in one-pass mode.\n", arg.name);
1112 } else if (arg_match(&arg, &g_av1_codec_arg_defs.minsection_pct, argi)) {
1113 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1114
1115 if (global->passes < 2)
1116 aom_tools_warn("option %s ignored in one-pass mode.\n", arg.name);
1117 } else if (arg_match(&arg, &g_av1_codec_arg_defs.maxsection_pct, argi)) {
1118 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1119
1120 if (global->passes < 2)
1121 aom_tools_warn("option %s ignored in one-pass mode.\n", arg.name);
1122 } else if (arg_match(&arg, &g_av1_codec_arg_defs.fwd_kf_enabled, argi)) {
1123 config->cfg.fwd_kf_enabled = arg_parse_uint(&arg);
1124 } else if (arg_match(&arg, &g_av1_codec_arg_defs.kf_min_dist, argi)) {
1125 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1126 } else if (arg_match(&arg, &g_av1_codec_arg_defs.kf_max_dist, argi)) {
1127 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1128 } else if (arg_match(&arg, &g_av1_codec_arg_defs.kf_disabled, argi)) {
1129 config->cfg.kf_mode = AOM_KF_DISABLED;
1130 } else if (arg_match(&arg, &g_av1_codec_arg_defs.sframe_dist, argi)) {
1131 config->cfg.sframe_dist = arg_parse_uint(&arg);
1132 } else if (arg_match(&arg, &g_av1_codec_arg_defs.sframe_mode, argi)) {
1133 config->cfg.sframe_mode = arg_parse_uint(&arg);
1134 } else if (arg_match(&arg, &g_av1_codec_arg_defs.save_as_annexb, argi)) {
1135 config->cfg.save_as_annexb = arg_parse_uint(&arg);
1136 } else if (arg_match(&arg, &g_av1_codec_arg_defs.tile_width, argi)) {
1137 config->cfg.tile_width_count =
1138 arg_parse_list(&arg, config->cfg.tile_widths, MAX_TILE_WIDTHS);
1139 } else if (arg_match(&arg, &g_av1_codec_arg_defs.tile_height, argi)) {
1140 config->cfg.tile_height_count =
1141 arg_parse_list(&arg, config->cfg.tile_heights, MAX_TILE_HEIGHTS);
1142 #if CONFIG_TUNE_VMAF
1143 } else if (arg_match(&arg, &g_av1_codec_arg_defs.vmaf_model_path, argi)) {
1144 config->vmaf_model_path = arg.val;
1145 #endif
1146 } else if (arg_match(&arg, &g_av1_codec_arg_defs.partition_info_path,
1147 argi)) {
1148 config->partition_info_path = arg.val;
1149 } else if (arg_match(&arg, &g_av1_codec_arg_defs.enable_rate_guide_deltaq,
1150 argi)) {
1151 config->enable_rate_guide_deltaq = arg_parse_uint(&arg);
1152 } else if (arg_match(&arg, &g_av1_codec_arg_defs.rate_distribution_info,
1153 argi)) {
1154 config->rate_distribution_info = arg.val;
1155 } else if (arg_match(&arg, &g_av1_codec_arg_defs.use_fixed_qp_offsets,
1156 argi)) {
1157 config->cfg.use_fixed_qp_offsets = arg_parse_uint(&arg);
1158 } else if (arg_match(&arg, &g_av1_codec_arg_defs.fixed_qp_offsets, argi)) {
1159 config->cfg.use_fixed_qp_offsets = 1;
1160 } else if (global->usage == AOM_USAGE_REALTIME &&
1161 arg_match(&arg, &g_av1_codec_arg_defs.enable_restoration,
1162 argi)) {
1163 if (arg_parse_uint(&arg) == 1) {
1164 aom_tools_warn("non-zero %s option ignored in realtime mode.\n",
1165 arg.name);
1166 }
1167 } else if (arg_match(&arg, &g_av1_codec_arg_defs.two_pass_input, argi)) {
1168 config->two_pass_input = arg.val;
1169 } else if (arg_match(&arg, &g_av1_codec_arg_defs.two_pass_output, argi)) {
1170 config->two_pass_output = arg.val;
1171 } else if (arg_match(&arg, &g_av1_codec_arg_defs.two_pass_width, argi)) {
1172 config->two_pass_width = arg_parse_int(&arg);
1173 } else if (arg_match(&arg, &g_av1_codec_arg_defs.two_pass_height, argi)) {
1174 config->two_pass_height = arg_parse_int(&arg);
1175 } else {
1176 int i, match = 0;
1177 // check if the control ID API supports this arg
1178 if (ctrl_args_map) {
1179 for (i = 0; ctrl_args[i]; i++) {
1180 if (arg_match(&arg, ctrl_args[i], argi)) {
1181 match = 1;
1182 set_config_arg_ctrls(config, ctrl_args_map[i], &arg);
1183 break;
1184 }
1185 }
1186 }
1187 if (!match) {
1188 // check if the key & value API supports this arg
1189 for (i = 0; key_val_args[i]; i++) {
1190 if (arg_match(&arg, key_val_args[i], argi)) {
1191 match = 1;
1192 set_config_arg_key_vals(config, key_val_args[i]->long_name, &arg);
1193 break;
1194 }
1195 }
1196 }
1197 if (!match) argj++;
1198 }
1199 }
1200 config->use_16bit_internal |= config->cfg.g_bit_depth > AOM_BITS_8;
1201
1202 if (global->usage == AOM_USAGE_REALTIME && config->cfg.g_lag_in_frames != 0) {
1203 aom_tools_warn("non-zero lag-in-frames option ignored in realtime mode.\n");
1204 config->cfg.g_lag_in_frames = 0;
1205 }
1206
1207 if (global->usage == AOM_USAGE_ALL_INTRA) {
1208 if (config->cfg.g_lag_in_frames != 0) {
1209 aom_tools_warn(
1210 "non-zero lag-in-frames option ignored in all intra mode.\n");
1211 config->cfg.g_lag_in_frames = 0;
1212 }
1213 if (config->cfg.kf_max_dist != 0) {
1214 aom_tools_warn(
1215 "non-zero max key frame distance option ignored in all intra "
1216 "mode.\n");
1217 config->cfg.kf_max_dist = 0;
1218 }
1219 }
1220
1221 // set the passes field using key & val API
1222 if (config->arg_key_val_cnt >= ARG_KEY_VAL_CNT_MAX) {
1223 die("Not enough buffer for the key & value API.");
1224 }
1225 config->arg_key_vals[config->arg_key_val_cnt][0] = "passes";
1226 switch (global->passes) {
1227 case 0: config->arg_key_vals[config->arg_key_val_cnt][1] = "0"; break;
1228 case 1: config->arg_key_vals[config->arg_key_val_cnt][1] = "1"; break;
1229 case 2: config->arg_key_vals[config->arg_key_val_cnt][1] = "2"; break;
1230 case 3: config->arg_key_vals[config->arg_key_val_cnt][1] = "3"; break;
1231 default: die("Invalid value of --passes.");
1232 }
1233 config->arg_key_val_cnt++;
1234
1235 // set the two_pass_output field
1236 if (!config->two_pass_output && global->passes == 3) {
1237 // If not specified, set the name of two_pass_output file here.
1238 snprintf(stream->tmp_out_fn, sizeof(stream->tmp_out_fn),
1239 "%.980s_pass2_%d.ivf", stream->config.out_fn, stream->index);
1240 stream->config.two_pass_output = stream->tmp_out_fn;
1241 }
1242 if (config->two_pass_output) {
1243 config->arg_key_vals[config->arg_key_val_cnt][0] = "two-pass-output";
1244 config->arg_key_vals[config->arg_key_val_cnt][1] = config->two_pass_output;
1245 config->arg_key_val_cnt++;
1246 }
1247
1248 return eos_mark_found;
1249 }
1250
1251 #define FOREACH_STREAM(iterator, list) \
1252 for (struct stream_state *iterator = list; iterator; \
1253 iterator = iterator->next)
1254
validate_stream_config(const struct stream_state * stream,const struct AvxEncoderConfig * global)1255 static void validate_stream_config(const struct stream_state *stream,
1256 const struct AvxEncoderConfig *global) {
1257 const struct stream_state *streami;
1258 (void)global;
1259
1260 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1261 fatal(
1262 "Stream %d: Specify stream dimensions with --width (-w) "
1263 " and --height (-h)",
1264 stream->index);
1265
1266 /* Even if bit depth is set on the command line flag to be lower,
1267 * it is upgraded to at least match the input bit depth.
1268 */
1269 assert(stream->config.cfg.g_input_bit_depth <=
1270 (unsigned int)stream->config.cfg.g_bit_depth);
1271
1272 for (streami = stream; streami; streami = streami->next) {
1273 /* All streams require output files */
1274 if (!streami->config.out_fn)
1275 fatal("Stream %d: Output file is required (specify with -o)",
1276 streami->index);
1277
1278 /* Check for two streams outputting to the same file */
1279 if (streami != stream) {
1280 const char *a = stream->config.out_fn;
1281 const char *b = streami->config.out_fn;
1282 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1283 fatal("Stream %d: duplicate output file (from stream %d)",
1284 streami->index, stream->index);
1285 }
1286
1287 /* Check for two streams sharing a stats file. */
1288 if (streami != stream) {
1289 const char *a = stream->config.stats_fn;
1290 const char *b = streami->config.stats_fn;
1291 if (a && b && !strcmp(a, b))
1292 fatal("Stream %d: duplicate stats file (from stream %d)",
1293 streami->index, stream->index);
1294 }
1295 }
1296 }
1297
set_stream_dimensions(struct stream_state * stream,unsigned int w,unsigned int h)1298 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1299 unsigned int h) {
1300 if (!stream->config.cfg.g_w) {
1301 if (!stream->config.cfg.g_h)
1302 stream->config.cfg.g_w = w;
1303 else
1304 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1305 }
1306 if (!stream->config.cfg.g_h) {
1307 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1308 }
1309 }
1310
file_type_to_string(enum VideoFileType t)1311 static const char *file_type_to_string(enum VideoFileType t) {
1312 switch (t) {
1313 case FILE_TYPE_RAW: return "RAW";
1314 case FILE_TYPE_Y4M: return "Y4M";
1315 default: return "Other";
1316 }
1317 }
1318
show_stream_config(struct stream_state * stream,struct AvxEncoderConfig * global,struct AvxInputContext * input)1319 static void show_stream_config(struct stream_state *stream,
1320 struct AvxEncoderConfig *global,
1321 struct AvxInputContext *input) {
1322 #define SHOW(field) \
1323 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1324
1325 if (stream->index == 0) {
1326 fprintf(stderr, "Codec: %s\n", aom_codec_iface_name(global->codec));
1327 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1328 input->filename, file_type_to_string(input->file_type),
1329 image_format_to_string(input->fmt));
1330 }
1331 if (stream->next || stream->index)
1332 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1333 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1334 fprintf(stderr, "Coding path: %s\n",
1335 stream->config.use_16bit_internal ? "HBD" : "LBD");
1336 fprintf(stderr, "Encoder parameters:\n");
1337
1338 SHOW(g_usage);
1339 SHOW(g_threads);
1340 SHOW(g_profile);
1341 SHOW(g_w);
1342 SHOW(g_h);
1343 SHOW(g_bit_depth);
1344 SHOW(g_input_bit_depth);
1345 SHOW(g_timebase.num);
1346 SHOW(g_timebase.den);
1347 SHOW(g_error_resilient);
1348 SHOW(g_pass);
1349 SHOW(g_lag_in_frames);
1350 SHOW(large_scale_tile);
1351 SHOW(rc_dropframe_thresh);
1352 SHOW(rc_resize_mode);
1353 SHOW(rc_resize_denominator);
1354 SHOW(rc_resize_kf_denominator);
1355 SHOW(rc_superres_mode);
1356 SHOW(rc_superres_denominator);
1357 SHOW(rc_superres_kf_denominator);
1358 SHOW(rc_superres_qthresh);
1359 SHOW(rc_superres_kf_qthresh);
1360 SHOW(rc_end_usage);
1361 SHOW(rc_target_bitrate);
1362 SHOW(rc_min_quantizer);
1363 SHOW(rc_max_quantizer);
1364 SHOW(rc_undershoot_pct);
1365 SHOW(rc_overshoot_pct);
1366 SHOW(rc_buf_sz);
1367 SHOW(rc_buf_initial_sz);
1368 SHOW(rc_buf_optimal_sz);
1369 SHOW(rc_2pass_vbr_bias_pct);
1370 SHOW(rc_2pass_vbr_minsection_pct);
1371 SHOW(rc_2pass_vbr_maxsection_pct);
1372 SHOW(fwd_kf_enabled);
1373 SHOW(kf_mode);
1374 SHOW(kf_min_dist);
1375 SHOW(kf_max_dist);
1376
1377 #define SHOW_PARAMS(field) \
1378 fprintf(stderr, " %-28s = %d\n", #field, \
1379 stream->config.cfg.encoder_cfg.field)
1380 if (global->encoder_config.init_by_cfg_file) {
1381 SHOW_PARAMS(super_block_size);
1382 SHOW_PARAMS(max_partition_size);
1383 SHOW_PARAMS(min_partition_size);
1384 SHOW_PARAMS(disable_ab_partition_type);
1385 SHOW_PARAMS(disable_rect_partition_type);
1386 SHOW_PARAMS(disable_1to4_partition_type);
1387 SHOW_PARAMS(disable_flip_idtx);
1388 SHOW_PARAMS(disable_cdef);
1389 SHOW_PARAMS(disable_lr);
1390 SHOW_PARAMS(disable_obmc);
1391 SHOW_PARAMS(disable_warp_motion);
1392 SHOW_PARAMS(disable_global_motion);
1393 SHOW_PARAMS(disable_dist_wtd_comp);
1394 SHOW_PARAMS(disable_diff_wtd_comp);
1395 SHOW_PARAMS(disable_inter_intra_comp);
1396 SHOW_PARAMS(disable_masked_comp);
1397 SHOW_PARAMS(disable_one_sided_comp);
1398 SHOW_PARAMS(disable_palette);
1399 SHOW_PARAMS(disable_intrabc);
1400 SHOW_PARAMS(disable_cfl);
1401 SHOW_PARAMS(disable_smooth_intra);
1402 SHOW_PARAMS(disable_filter_intra);
1403 SHOW_PARAMS(disable_dual_filter);
1404 SHOW_PARAMS(disable_intra_angle_delta);
1405 SHOW_PARAMS(disable_intra_edge_filter);
1406 SHOW_PARAMS(disable_tx_64x64);
1407 SHOW_PARAMS(disable_smooth_inter_intra);
1408 SHOW_PARAMS(disable_inter_inter_wedge);
1409 SHOW_PARAMS(disable_inter_intra_wedge);
1410 SHOW_PARAMS(disable_paeth_intra);
1411 SHOW_PARAMS(disable_trellis_quant);
1412 SHOW_PARAMS(disable_ref_frame_mv);
1413 SHOW_PARAMS(reduced_reference_set);
1414 SHOW_PARAMS(reduced_tx_type_set);
1415 }
1416 }
1417
open_output_file(struct stream_state * stream,struct AvxEncoderConfig * global,const struct AvxRational * pixel_aspect_ratio,const char * encoder_settings)1418 static void open_output_file(struct stream_state *stream,
1419 struct AvxEncoderConfig *global,
1420 const struct AvxRational *pixel_aspect_ratio,
1421 const char *encoder_settings) {
1422 const char *fn = stream->config.out_fn;
1423 const struct aom_codec_enc_cfg *const cfg = &stream->config.cfg;
1424
1425 if (cfg->g_pass == AOM_RC_FIRST_PASS) return;
1426
1427 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1428
1429 if (!stream->file) fatal("Failed to open output file");
1430
1431 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1432 fatal("WebM output to pipes not supported.");
1433
1434 #if CONFIG_WEBM_IO
1435 if (stream->config.write_webm) {
1436 stream->webm_ctx.stream = stream->file;
1437 if (write_webm_file_header(&stream->webm_ctx, &stream->encoder, cfg,
1438 stream->config.stereo_fmt,
1439 get_fourcc_by_aom_encoder(global->codec),
1440 pixel_aspect_ratio, encoder_settings) != 0) {
1441 fatal("WebM writer initialization failed.");
1442 }
1443 }
1444 #else
1445 (void)pixel_aspect_ratio;
1446 (void)encoder_settings;
1447 #endif
1448
1449 if (!stream->config.write_webm && stream->config.write_ivf) {
1450 ivf_write_file_header(stream->file, cfg,
1451 get_fourcc_by_aom_encoder(global->codec), 0);
1452 }
1453 }
1454
close_output_file(struct stream_state * stream,unsigned int fourcc)1455 static void close_output_file(struct stream_state *stream,
1456 unsigned int fourcc) {
1457 const struct aom_codec_enc_cfg *const cfg = &stream->config.cfg;
1458
1459 if (cfg->g_pass == AOM_RC_FIRST_PASS) return;
1460
1461 #if CONFIG_WEBM_IO
1462 if (stream->config.write_webm) {
1463 if (write_webm_file_footer(&stream->webm_ctx) != 0) {
1464 fatal("WebM writer finalization failed.");
1465 }
1466 }
1467 #endif
1468
1469 if (!stream->config.write_webm && stream->config.write_ivf) {
1470 if (!fseek(stream->file, 0, SEEK_SET))
1471 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1472 stream->frames_out);
1473 }
1474
1475 fclose(stream->file);
1476 }
1477
setup_pass(struct stream_state * stream,struct AvxEncoderConfig * global,int pass)1478 static void setup_pass(struct stream_state *stream,
1479 struct AvxEncoderConfig *global, int pass) {
1480 if (stream->config.stats_fn) {
1481 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1482 fatal("Failed to open statistics store");
1483 } else {
1484 if (!stats_open_mem(&stream->stats, pass))
1485 fatal("Failed to open statistics store");
1486 }
1487
1488 if (global->passes == 1) {
1489 stream->config.cfg.g_pass = AOM_RC_ONE_PASS;
1490 } else {
1491 switch (pass) {
1492 case 0: stream->config.cfg.g_pass = AOM_RC_FIRST_PASS; break;
1493 case 1: stream->config.cfg.g_pass = AOM_RC_SECOND_PASS; break;
1494 case 2: stream->config.cfg.g_pass = AOM_RC_THIRD_PASS; break;
1495 default: fatal("Failed to set pass");
1496 }
1497 }
1498
1499 if (pass) {
1500 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1501 }
1502
1503 stream->cx_time = 0;
1504 stream->nbytes = 0;
1505 stream->frames_out = 0;
1506 }
1507
initialize_encoder(struct stream_state * stream,struct AvxEncoderConfig * global)1508 static void initialize_encoder(struct stream_state *stream,
1509 struct AvxEncoderConfig *global) {
1510 int i;
1511 int flags = 0;
1512
1513 flags |= (global->show_psnr >= 1) ? AOM_CODEC_USE_PSNR : 0;
1514 flags |= stream->config.use_16bit_internal ? AOM_CODEC_USE_HIGHBITDEPTH : 0;
1515
1516 /* Construct Encoder Context */
1517 aom_codec_enc_init(&stream->encoder, global->codec, &stream->config.cfg,
1518 flags);
1519 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1520
1521 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1522 int ctrl = stream->config.arg_ctrls[i][0];
1523 int value = stream->config.arg_ctrls[i][1];
1524 if (aom_codec_control(&stream->encoder, ctrl, value))
1525 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1526
1527 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1528 }
1529
1530 for (i = 0; i < stream->config.arg_key_val_cnt; i++) {
1531 const char *name = stream->config.arg_key_vals[i][0];
1532 const char *val = stream->config.arg_key_vals[i][1];
1533 if (aom_codec_set_option(&stream->encoder, name, val))
1534 fprintf(stderr, "Error: Tried to set option %s = %s\n", name, val);
1535
1536 ctx_exit_on_error(&stream->encoder, "Failed to set codec option");
1537 }
1538
1539 #if CONFIG_TUNE_VMAF
1540 if (stream->config.vmaf_model_path) {
1541 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder, AV1E_SET_VMAF_MODEL_PATH,
1542 stream->config.vmaf_model_path);
1543 ctx_exit_on_error(&stream->encoder, "Failed to set vmaf model path");
1544 }
1545 #endif
1546 if (stream->config.partition_info_path) {
1547 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
1548 AV1E_SET_PARTITION_INFO_PATH,
1549 stream->config.partition_info_path);
1550 ctx_exit_on_error(&stream->encoder, "Failed to set partition info path");
1551 }
1552 if (stream->config.enable_rate_guide_deltaq) {
1553 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
1554 AV1E_ENABLE_RATE_GUIDE_DELTAQ,
1555 stream->config.enable_rate_guide_deltaq);
1556 ctx_exit_on_error(&stream->encoder, "Failed to enable rate guide deltaq");
1557 }
1558 if (stream->config.rate_distribution_info) {
1559 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
1560 AV1E_SET_RATE_DISTRIBUTION_INFO,
1561 stream->config.rate_distribution_info);
1562 ctx_exit_on_error(&stream->encoder, "Failed to set rate distribution info");
1563 }
1564
1565 if (stream->config.film_grain_filename) {
1566 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder, AV1E_SET_FILM_GRAIN_TABLE,
1567 stream->config.film_grain_filename);
1568 ctx_exit_on_error(&stream->encoder, "Failed to set film grain table");
1569 }
1570 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder, AV1E_SET_COLOR_RANGE,
1571 stream->config.color_range);
1572 ctx_exit_on_error(&stream->encoder, "Failed to set color range");
1573
1574 #if CONFIG_AV1_DECODER
1575 if (global->test_decode != TEST_DECODE_OFF) {
1576 aom_codec_iface_t *decoder = get_aom_decoder_by_short_name(
1577 get_short_name_by_aom_encoder(global->codec));
1578 aom_codec_dec_cfg_t cfg = { 0, 0, 0, !stream->config.use_16bit_internal };
1579 aom_codec_dec_init(&stream->decoder, decoder, &cfg, 0);
1580
1581 if (strcmp(get_short_name_by_aom_encoder(global->codec), "av1") == 0) {
1582 AOM_CODEC_CONTROL_TYPECHECKED(&stream->decoder, AV1_SET_TILE_MODE,
1583 stream->config.cfg.large_scale_tile);
1584 ctx_exit_on_error(&stream->decoder, "Failed to set decode_tile_mode");
1585
1586 AOM_CODEC_CONTROL_TYPECHECKED(&stream->decoder, AV1D_SET_IS_ANNEXB,
1587 stream->config.cfg.save_as_annexb);
1588 ctx_exit_on_error(&stream->decoder, "Failed to set is_annexb");
1589
1590 AOM_CODEC_CONTROL_TYPECHECKED(&stream->decoder, AV1_SET_DECODE_TILE_ROW,
1591 -1);
1592 ctx_exit_on_error(&stream->decoder, "Failed to set decode_tile_row");
1593
1594 AOM_CODEC_CONTROL_TYPECHECKED(&stream->decoder, AV1_SET_DECODE_TILE_COL,
1595 -1);
1596 ctx_exit_on_error(&stream->decoder, "Failed to set decode_tile_col");
1597 }
1598 }
1599 #endif
1600 }
1601
1602 // Convert the input image 'img' to a monochrome image. The Y plane of the
1603 // output image is a shallow copy of the Y plane of the input image, therefore
1604 // the input image must remain valid for the lifetime of the output image. The U
1605 // and V planes of the output image are set to null pointers. The output image
1606 // format is AOM_IMG_FMT_I420 because libaom does not have AOM_IMG_FMT_I400.
convert_image_to_monochrome(const struct aom_image * img,struct aom_image * monochrome_img)1607 static void convert_image_to_monochrome(const struct aom_image *img,
1608 struct aom_image *monochrome_img) {
1609 *monochrome_img = *img;
1610 monochrome_img->fmt = AOM_IMG_FMT_I420;
1611 if (img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
1612 monochrome_img->fmt |= AOM_IMG_FMT_HIGHBITDEPTH;
1613 }
1614 monochrome_img->monochrome = 1;
1615 monochrome_img->csp = AOM_CSP_UNKNOWN;
1616 monochrome_img->x_chroma_shift = 1;
1617 monochrome_img->y_chroma_shift = 1;
1618 monochrome_img->planes[AOM_PLANE_U] = NULL;
1619 monochrome_img->planes[AOM_PLANE_V] = NULL;
1620 monochrome_img->stride[AOM_PLANE_U] = 0;
1621 monochrome_img->stride[AOM_PLANE_V] = 0;
1622 monochrome_img->sz = 0;
1623 monochrome_img->bps = (img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 16 : 8;
1624 monochrome_img->img_data = NULL;
1625 monochrome_img->img_data_owner = 0;
1626 monochrome_img->self_allocd = 0;
1627 }
1628
encode_frame(struct stream_state * stream,struct AvxEncoderConfig * global,struct aom_image * img,unsigned int frames_in)1629 static void encode_frame(struct stream_state *stream,
1630 struct AvxEncoderConfig *global, struct aom_image *img,
1631 unsigned int frames_in) {
1632 aom_codec_pts_t frame_start, next_frame_start;
1633 struct aom_codec_enc_cfg *cfg = &stream->config.cfg;
1634 struct aom_usec_timer timer;
1635
1636 frame_start =
1637 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1638 cfg->g_timebase.num / global->framerate.num;
1639 next_frame_start =
1640 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1641 cfg->g_timebase.num / global->framerate.num;
1642
1643 /* Scale if necessary */
1644 if (img) {
1645 if ((img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) &&
1646 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1647 if (img->fmt != AOM_IMG_FMT_I42016) {
1648 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1649 exit(EXIT_FAILURE);
1650 }
1651 #if CONFIG_LIBYUV
1652 if (!stream->img) {
1653 stream->img =
1654 aom_img_alloc(NULL, AOM_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1655 }
1656 I420Scale_16(
1657 (uint16_t *)img->planes[AOM_PLANE_Y], img->stride[AOM_PLANE_Y] / 2,
1658 (uint16_t *)img->planes[AOM_PLANE_U], img->stride[AOM_PLANE_U] / 2,
1659 (uint16_t *)img->planes[AOM_PLANE_V], img->stride[AOM_PLANE_V] / 2,
1660 img->d_w, img->d_h, (uint16_t *)stream->img->planes[AOM_PLANE_Y],
1661 stream->img->stride[AOM_PLANE_Y] / 2,
1662 (uint16_t *)stream->img->planes[AOM_PLANE_U],
1663 stream->img->stride[AOM_PLANE_U] / 2,
1664 (uint16_t *)stream->img->planes[AOM_PLANE_V],
1665 stream->img->stride[AOM_PLANE_V] / 2, stream->img->d_w,
1666 stream->img->d_h, kFilterBox);
1667 img = stream->img;
1668 #else
1669 stream->encoder.err = 1;
1670 ctx_exit_on_error(&stream->encoder,
1671 "Stream %d: Failed to encode frame.\n"
1672 "libyuv is required for scaling but is currently "
1673 "disabled.\n"
1674 "Be sure to specify -DCONFIG_LIBYUV=1 when running "
1675 "cmake.\n",
1676 stream->index);
1677 #endif
1678 }
1679 }
1680 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1681 if (img->fmt != AOM_IMG_FMT_I420 && img->fmt != AOM_IMG_FMT_YV12) {
1682 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1683 exit(EXIT_FAILURE);
1684 }
1685 #if CONFIG_LIBYUV
1686 if (!stream->img)
1687 stream->img =
1688 aom_img_alloc(NULL, AOM_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1689 I420Scale(
1690 img->planes[AOM_PLANE_Y], img->stride[AOM_PLANE_Y],
1691 img->planes[AOM_PLANE_U], img->stride[AOM_PLANE_U],
1692 img->planes[AOM_PLANE_V], img->stride[AOM_PLANE_V], img->d_w, img->d_h,
1693 stream->img->planes[AOM_PLANE_Y], stream->img->stride[AOM_PLANE_Y],
1694 stream->img->planes[AOM_PLANE_U], stream->img->stride[AOM_PLANE_U],
1695 stream->img->planes[AOM_PLANE_V], stream->img->stride[AOM_PLANE_V],
1696 stream->img->d_w, stream->img->d_h, kFilterBox);
1697 img = stream->img;
1698 #else
1699 stream->encoder.err = 1;
1700 ctx_exit_on_error(&stream->encoder,
1701 "Stream %d: Failed to encode frame.\n"
1702 "Scaling disabled in this configuration. \n"
1703 "To enable, configure with --enable-libyuv\n",
1704 stream->index);
1705 #endif
1706 }
1707
1708 struct aom_image monochrome_img;
1709 if (img && cfg->monochrome) {
1710 convert_image_to_monochrome(img, &monochrome_img);
1711 img = &monochrome_img;
1712 }
1713
1714 aom_usec_timer_start(&timer);
1715 aom_codec_encode(&stream->encoder, img, frame_start,
1716 (uint32_t)(next_frame_start - frame_start), 0);
1717 aom_usec_timer_mark(&timer);
1718 stream->cx_time += aom_usec_timer_elapsed(&timer);
1719 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1720 stream->index);
1721 }
1722
update_quantizer_histogram(struct stream_state * stream)1723 static void update_quantizer_histogram(struct stream_state *stream) {
1724 if (stream->config.cfg.g_pass != AOM_RC_FIRST_PASS) {
1725 int q;
1726
1727 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder, AOME_GET_LAST_QUANTIZER_64,
1728 &q);
1729 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1730 stream->counts[q]++;
1731 }
1732 }
1733
get_cx_data(struct stream_state * stream,struct AvxEncoderConfig * global,int * got_data)1734 static void get_cx_data(struct stream_state *stream,
1735 struct AvxEncoderConfig *global, int *got_data) {
1736 const aom_codec_cx_pkt_t *pkt;
1737 const struct aom_codec_enc_cfg *cfg = &stream->config.cfg;
1738 aom_codec_iter_t iter = NULL;
1739
1740 *got_data = 0;
1741 while ((pkt = aom_codec_get_cx_data(&stream->encoder, &iter))) {
1742 static size_t fsize = 0;
1743 static FileOffset ivf_header_pos = 0;
1744
1745 switch (pkt->kind) {
1746 case AOM_CODEC_CX_FRAME_PKT:
1747 ++stream->frames_out;
1748 if (!global->quiet)
1749 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1750
1751 update_rate_histogram(stream->rate_hist, cfg, pkt);
1752 #if CONFIG_WEBM_IO
1753 if (stream->config.write_webm) {
1754 if (write_webm_block(&stream->webm_ctx, cfg, pkt) != 0) {
1755 fatal("WebM writer failed.");
1756 }
1757 }
1758 #endif
1759 if (!stream->config.write_webm) {
1760 if (stream->config.write_ivf) {
1761 if (pkt->data.frame.partition_id <= 0) {
1762 ivf_header_pos = ftello(stream->file);
1763 fsize = pkt->data.frame.sz;
1764
1765 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1766 } else {
1767 fsize += pkt->data.frame.sz;
1768
1769 const FileOffset currpos = ftello(stream->file);
1770 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1771 ivf_write_frame_size(stream->file, fsize);
1772 fseeko(stream->file, currpos, SEEK_SET);
1773 }
1774 }
1775
1776 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1777 stream->file);
1778 }
1779 stream->nbytes += pkt->data.raw.sz;
1780
1781 *got_data = 1;
1782 #if CONFIG_AV1_DECODER
1783 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1784 aom_codec_decode(&stream->decoder, pkt->data.frame.buf,
1785 pkt->data.frame.sz, NULL);
1786 if (stream->decoder.err) {
1787 warn_or_exit_on_error(&stream->decoder,
1788 global->test_decode == TEST_DECODE_FATAL,
1789 "Failed to decode frame %d in stream %d",
1790 stream->frames_out + 1, stream->index);
1791 stream->mismatch_seen = stream->frames_out + 1;
1792 }
1793 }
1794 #endif
1795 break;
1796 case AOM_CODEC_STATS_PKT:
1797 stream->frames_out++;
1798 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1799 pkt->data.twopass_stats.sz);
1800 stream->nbytes += pkt->data.raw.sz;
1801 break;
1802 case AOM_CODEC_PSNR_PKT:
1803
1804 if (global->show_psnr >= 1) {
1805 int i;
1806
1807 stream->psnr_sse_total[0] += pkt->data.psnr.sse[0];
1808 stream->psnr_samples_total[0] += pkt->data.psnr.samples[0];
1809 for (i = 0; i < 4; i++) {
1810 if (!global->quiet)
1811 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1812 stream->psnr_totals[0][i] += pkt->data.psnr.psnr[i];
1813 }
1814 stream->psnr_count[0]++;
1815
1816 #if CONFIG_AV1_HIGHBITDEPTH
1817 if (stream->config.cfg.g_input_bit_depth <
1818 (unsigned int)stream->config.cfg.g_bit_depth) {
1819 stream->psnr_sse_total[1] += pkt->data.psnr.sse_hbd[0];
1820 stream->psnr_samples_total[1] += pkt->data.psnr.samples_hbd[0];
1821 for (i = 0; i < 4; i++) {
1822 if (!global->quiet)
1823 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr_hbd[i]);
1824 stream->psnr_totals[1][i] += pkt->data.psnr.psnr_hbd[i];
1825 }
1826 stream->psnr_count[1]++;
1827 }
1828 #endif
1829 }
1830
1831 break;
1832 default: break;
1833 }
1834 }
1835 }
1836
show_psnr(struct stream_state * stream,double peak,int64_t bps)1837 static void show_psnr(struct stream_state *stream, double peak, int64_t bps) {
1838 int i;
1839 double ovpsnr;
1840
1841 if (!stream->psnr_count[0]) return;
1842
1843 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1844 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total[0], peak,
1845 (double)stream->psnr_sse_total[0]);
1846 fprintf(stderr, " %.3f", ovpsnr);
1847
1848 for (i = 0; i < 4; i++) {
1849 fprintf(stderr, " %.3f", stream->psnr_totals[0][i] / stream->psnr_count[0]);
1850 }
1851 if (bps > 0) {
1852 fprintf(stderr, " %7" PRId64 " bps", bps);
1853 }
1854 fprintf(stderr, " %7" PRId64 " ms", stream->cx_time / 1000);
1855 fprintf(stderr, "\n");
1856 }
1857
1858 #if CONFIG_AV1_HIGHBITDEPTH
show_psnr_hbd(struct stream_state * stream,double peak,int64_t bps)1859 static void show_psnr_hbd(struct stream_state *stream, double peak,
1860 int64_t bps) {
1861 int i;
1862 double ovpsnr;
1863 // Compute PSNR based on stream bit depth
1864 if (!stream->psnr_count[1]) return;
1865
1866 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1867 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total[1], peak,
1868 (double)stream->psnr_sse_total[1]);
1869 fprintf(stderr, " %.3f", ovpsnr);
1870
1871 for (i = 0; i < 4; i++) {
1872 fprintf(stderr, " %.3f", stream->psnr_totals[1][i] / stream->psnr_count[1]);
1873 }
1874 if (bps > 0) {
1875 fprintf(stderr, " %7" PRId64 " bps", bps);
1876 }
1877 fprintf(stderr, " %7" PRId64 " ms", stream->cx_time / 1000);
1878 fprintf(stderr, "\n");
1879 }
1880 #endif
1881
usec_to_fps(uint64_t usec,unsigned int frames)1882 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1883 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1884 }
1885
test_decode(struct stream_state * stream,enum TestDecodeFatality fatal)1886 static void test_decode(struct stream_state *stream,
1887 enum TestDecodeFatality fatal) {
1888 aom_image_t enc_img, dec_img;
1889
1890 if (stream->mismatch_seen) return;
1891
1892 /* Get the internal reference frame */
1893 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder, AV1_GET_NEW_FRAME_IMAGE,
1894 &enc_img);
1895 AOM_CODEC_CONTROL_TYPECHECKED(&stream->decoder, AV1_GET_NEW_FRAME_IMAGE,
1896 &dec_img);
1897
1898 if ((enc_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH) !=
1899 (dec_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH)) {
1900 if (enc_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
1901 aom_image_t enc_hbd_img;
1902 aom_img_alloc(&enc_hbd_img, enc_img.fmt - AOM_IMG_FMT_HIGHBITDEPTH,
1903 enc_img.d_w, enc_img.d_h, 16);
1904 aom_img_truncate_16_to_8(&enc_hbd_img, &enc_img);
1905 enc_img = enc_hbd_img;
1906 }
1907 if (dec_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
1908 aom_image_t dec_hbd_img;
1909 aom_img_alloc(&dec_hbd_img, dec_img.fmt - AOM_IMG_FMT_HIGHBITDEPTH,
1910 dec_img.d_w, dec_img.d_h, 16);
1911 aom_img_truncate_16_to_8(&dec_hbd_img, &dec_img);
1912 dec_img = dec_hbd_img;
1913 }
1914 }
1915
1916 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1917 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1918
1919 if (!aom_compare_img(&enc_img, &dec_img)) {
1920 int y[4], u[4], v[4];
1921 if (enc_img.fmt & AOM_IMG_FMT_HIGHBITDEPTH) {
1922 aom_find_mismatch_high(&enc_img, &dec_img, y, u, v);
1923 } else {
1924 aom_find_mismatch(&enc_img, &dec_img, y, u, v);
1925 }
1926 stream->decoder.err = 1;
1927 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1928 "Stream %d: Encode/decode mismatch on frame %d at"
1929 " Y[%d, %d] {%d/%d},"
1930 " U[%d, %d] {%d/%d},"
1931 " V[%d, %d] {%d/%d}",
1932 stream->index, stream->frames_out, y[0], y[1], y[2],
1933 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1934 stream->mismatch_seen = stream->frames_out;
1935 }
1936
1937 aom_img_free(&enc_img);
1938 aom_img_free(&dec_img);
1939 }
1940
print_time(const char * label,int64_t etl)1941 static void print_time(const char *label, int64_t etl) {
1942 int64_t hours;
1943 int64_t mins;
1944 int64_t secs;
1945
1946 if (etl >= 0) {
1947 hours = etl / 3600;
1948 etl -= hours * 3600;
1949 mins = etl / 60;
1950 etl -= mins * 60;
1951 secs = etl;
1952
1953 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1954 hours, mins, secs);
1955 } else {
1956 fprintf(stderr, "[%3s unknown] ", label);
1957 }
1958 }
1959
clear_stream_count_state(struct stream_state * stream)1960 static void clear_stream_count_state(struct stream_state *stream) {
1961 // PSNR counters
1962 for (int k = 0; k < 2; k++) {
1963 stream->psnr_sse_total[k] = 0;
1964 stream->psnr_samples_total[k] = 0;
1965 for (int i = 0; i < 4; i++) {
1966 stream->psnr_totals[k][i] = 0;
1967 }
1968 stream->psnr_count[k] = 0;
1969 }
1970 // q hist
1971 memset(stream->counts, 0, sizeof(stream->counts));
1972 }
1973
1974 // aomenc will downscale the second pass if:
1975 // 1. the specific pass is not given by commandline (aomenc will perform all
1976 // passes)
1977 // 2. there are more than 2 passes in total
1978 // 3. current pass is the second pass (the parameter pass starts with 0 so
1979 // pass == 1)
pass_need_downscale(int global_pass,int global_passes,int pass)1980 static int pass_need_downscale(int global_pass, int global_passes, int pass) {
1981 return !global_pass && global_passes > 2 && pass == 1;
1982 }
1983
main(int argc,const char ** argv_)1984 int main(int argc, const char **argv_) {
1985 int pass;
1986 aom_image_t raw;
1987 aom_image_t raw_shift;
1988 int allocated_raw_shift = 0;
1989 int do_16bit_internal = 0;
1990 int input_shift = 0;
1991 int frame_avail, got_data;
1992
1993 struct AvxInputContext input;
1994 struct AvxEncoderConfig global;
1995 struct stream_state *streams = NULL;
1996 char **argv, **argi;
1997 uint64_t cx_time = 0;
1998 int stream_cnt = 0;
1999 int res = 0;
2000 int profile_updated = 0;
2001
2002 memset(&input, 0, sizeof(input));
2003 memset(&raw, 0, sizeof(raw));
2004 exec_name = argv_[0];
2005
2006 /* Setup default input stream settings */
2007 input.framerate.numerator = 30;
2008 input.framerate.denominator = 1;
2009 input.only_i420 = 1;
2010 input.bit_depth = 0;
2011
2012 /* First parse the global configuration values, because we want to apply
2013 * other parameters on top of the default configuration provided by the
2014 * codec.
2015 */
2016 argv = argv_dup(argc - 1, argv_ + 1);
2017 if (!argv) {
2018 fprintf(stderr, "Error allocating argument list\n");
2019 return EXIT_FAILURE;
2020 }
2021 parse_global_config(&global, &argv);
2022
2023 if (argc < 2) usage_exit();
2024
2025 switch (global.color_type) {
2026 case I420: input.fmt = AOM_IMG_FMT_I420; break;
2027 case I422: input.fmt = AOM_IMG_FMT_I422; break;
2028 case I444: input.fmt = AOM_IMG_FMT_I444; break;
2029 case YV12: input.fmt = AOM_IMG_FMT_YV12; break;
2030 case NV12: input.fmt = AOM_IMG_FMT_NV12; break;
2031 }
2032
2033 {
2034 /* Now parse each stream's parameters. Using a local scope here
2035 * due to the use of 'stream' as loop variable in FOREACH_STREAM
2036 * loops
2037 */
2038 struct stream_state *stream = NULL;
2039
2040 do {
2041 stream = new_stream(&global, stream);
2042 stream_cnt++;
2043 if (!streams) streams = stream;
2044 } while (parse_stream_params(&global, stream, argv));
2045 }
2046
2047 /* Check for unrecognized options */
2048 for (argi = argv; *argi; argi++)
2049 if (argi[0][0] == '-' && argi[0][1])
2050 die("Error: Unrecognized option %s\n", *argi);
2051
2052 FOREACH_STREAM(stream, streams) {
2053 check_encoder_config(global.disable_warning_prompt, &global,
2054 &stream->config.cfg);
2055
2056 // If large_scale_tile = 1, only support to output to ivf format.
2057 if (stream->config.cfg.large_scale_tile && !stream->config.write_ivf)
2058 die("only support ivf output format while large-scale-tile=1\n");
2059 }
2060
2061 /* Handle non-option arguments */
2062 input.filename = argv[0];
2063 const char *orig_input_filename = input.filename;
2064 FOREACH_STREAM(stream, streams) {
2065 stream->orig_out_fn = stream->config.out_fn;
2066 stream->orig_width = stream->config.cfg.g_w;
2067 stream->orig_height = stream->config.cfg.g_h;
2068 stream->orig_write_ivf = stream->config.write_ivf;
2069 stream->orig_write_webm = stream->config.write_webm;
2070 }
2071
2072 if (!input.filename) {
2073 fprintf(stderr, "No input file specified!\n");
2074 usage_exit();
2075 }
2076
2077 /* Decide if other chroma subsamplings than 4:2:0 are supported */
2078 if (get_fourcc_by_aom_encoder(global.codec) == AV1_FOURCC)
2079 input.only_i420 = 0;
2080
2081 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2082 if (pass > 1) {
2083 FOREACH_STREAM(stream, streams) { clear_stream_count_state(stream); }
2084 }
2085
2086 int frames_in = 0, seen_frames = 0;
2087 int64_t estimated_time_left = -1;
2088 int64_t average_rate = -1;
2089 int64_t lagged_count = 0;
2090 const int need_downscale =
2091 pass_need_downscale(global.pass, global.passes, pass);
2092
2093 // Set the output to the specified two-pass output file, and
2094 // restore the width and height to the original values.
2095 FOREACH_STREAM(stream, streams) {
2096 if (need_downscale) {
2097 stream->config.out_fn = stream->config.two_pass_output;
2098 // Libaom currently only supports the ivf format for the third pass.
2099 stream->config.write_ivf = 1;
2100 stream->config.write_webm = 0;
2101 } else {
2102 stream->config.out_fn = stream->orig_out_fn;
2103 stream->config.write_ivf = stream->orig_write_ivf;
2104 stream->config.write_webm = stream->orig_write_webm;
2105 }
2106 stream->config.cfg.g_w = stream->orig_width;
2107 stream->config.cfg.g_h = stream->orig_height;
2108 }
2109
2110 // For second pass in three-pass encoding, set the input to
2111 // the given two-pass-input file if available. If the scaled input is not
2112 // given, we will attempt to re-scale the original input.
2113 input.filename = orig_input_filename;
2114 const char *two_pass_input = NULL;
2115 if (need_downscale) {
2116 FOREACH_STREAM(stream, streams) {
2117 if (stream->config.two_pass_input) {
2118 two_pass_input = stream->config.two_pass_input;
2119 input.filename = two_pass_input;
2120 break;
2121 }
2122 }
2123 }
2124
2125 open_input_file(&input, global.csp);
2126
2127 /* If the input file doesn't specify its w/h (raw files), try to get
2128 * the data from the first stream's configuration.
2129 */
2130 if (!input.width || !input.height) {
2131 if (two_pass_input) {
2132 FOREACH_STREAM(stream, streams) {
2133 if (stream->config.two_pass_width && stream->config.two_pass_height) {
2134 input.width = stream->config.two_pass_width;
2135 input.height = stream->config.two_pass_height;
2136 break;
2137 }
2138 }
2139 } else {
2140 FOREACH_STREAM(stream, streams) {
2141 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2142 input.width = stream->config.cfg.g_w;
2143 input.height = stream->config.cfg.g_h;
2144 break;
2145 }
2146 }
2147 }
2148 }
2149
2150 /* Update stream configurations from the input file's parameters */
2151 if (!input.width || !input.height) {
2152 if (two_pass_input) {
2153 fatal(
2154 "Specify downscaled stream dimensions with --two-pass-width "
2155 " and --two-pass-height");
2156 } else {
2157 fatal(
2158 "Specify stream dimensions with --width (-w) "
2159 " and --height (-h)");
2160 }
2161 }
2162
2163 if (need_downscale) {
2164 FOREACH_STREAM(stream, streams) {
2165 if (stream->config.two_pass_width && stream->config.two_pass_height) {
2166 stream->config.cfg.g_w = stream->config.two_pass_width;
2167 stream->config.cfg.g_h = stream->config.two_pass_height;
2168 } else if (two_pass_input) {
2169 stream->config.cfg.g_w = input.width;
2170 stream->config.cfg.g_h = input.height;
2171 } else if (stream->orig_width && stream->orig_height) {
2172 #if CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2173 stream->config.cfg.g_w = stream->orig_width;
2174 stream->config.cfg.g_h = stream->orig_height;
2175 #else // CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2176 stream->config.cfg.g_w = (stream->orig_width + 1) / 2;
2177 stream->config.cfg.g_h = (stream->orig_height + 1) / 2;
2178 #endif // CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2179 } else {
2180 #if CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2181 stream->config.cfg.g_w = input.width;
2182 stream->config.cfg.g_h = input.height;
2183 #else // CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2184 stream->config.cfg.g_w = (input.width + 1) / 2;
2185 stream->config.cfg.g_h = (input.height + 1) / 2;
2186 #endif // CONFIG_BITRATE_ACCURACY || CONFIG_BITRATE_ACCURACY_BL
2187 }
2188 }
2189 }
2190
2191 /* If input file does not specify bit-depth but input-bit-depth parameter
2192 * exists, assume that to be the input bit-depth. However, if the
2193 * input-bit-depth paramter does not exist, assume the input bit-depth
2194 * to be the same as the codec bit-depth.
2195 */
2196 if (!input.bit_depth) {
2197 FOREACH_STREAM(stream, streams) {
2198 if (stream->config.cfg.g_input_bit_depth)
2199 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2200 else
2201 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2202 (int)stream->config.cfg.g_bit_depth;
2203 }
2204 if (input.bit_depth > 8) input.fmt |= AOM_IMG_FMT_HIGHBITDEPTH;
2205 } else {
2206 FOREACH_STREAM(stream, streams) {
2207 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2208 }
2209 }
2210
2211 FOREACH_STREAM(stream, streams) {
2212 if (input.fmt != AOM_IMG_FMT_I420 && input.fmt != AOM_IMG_FMT_I42016 &&
2213 input.fmt != AOM_IMG_FMT_NV12) {
2214 /* Automatically upgrade if input is non-4:2:0 but a 4:2:0 profile
2215 was selected. */
2216 switch (stream->config.cfg.g_profile) {
2217 case 0:
2218 if (input.bit_depth < 12 && (input.fmt == AOM_IMG_FMT_I444 ||
2219 input.fmt == AOM_IMG_FMT_I44416)) {
2220 if (!stream->config.cfg.monochrome) {
2221 stream->config.cfg.g_profile = 1;
2222 profile_updated = 1;
2223 }
2224 } else if (input.bit_depth == 12 ||
2225 ((input.fmt == AOM_IMG_FMT_I422 ||
2226 input.fmt == AOM_IMG_FMT_I42216) &&
2227 !stream->config.cfg.monochrome)) {
2228 stream->config.cfg.g_profile = 2;
2229 profile_updated = 1;
2230 }
2231 break;
2232 case 1:
2233 if (input.bit_depth == 12 || input.fmt == AOM_IMG_FMT_I422 ||
2234 input.fmt == AOM_IMG_FMT_I42216) {
2235 stream->config.cfg.g_profile = 2;
2236 profile_updated = 1;
2237 } else if (input.bit_depth < 12 &&
2238 (input.fmt == AOM_IMG_FMT_I420 ||
2239 input.fmt == AOM_IMG_FMT_I42016)) {
2240 stream->config.cfg.g_profile = 0;
2241 profile_updated = 1;
2242 }
2243 break;
2244 case 2:
2245 if (input.bit_depth < 12 && (input.fmt == AOM_IMG_FMT_I444 ||
2246 input.fmt == AOM_IMG_FMT_I44416)) {
2247 stream->config.cfg.g_profile = 1;
2248 profile_updated = 1;
2249 } else if (input.bit_depth < 12 &&
2250 (input.fmt == AOM_IMG_FMT_I420 ||
2251 input.fmt == AOM_IMG_FMT_I42016)) {
2252 stream->config.cfg.g_profile = 0;
2253 profile_updated = 1;
2254 } else if (input.bit_depth == 12 &&
2255 input.file_type == FILE_TYPE_Y4M) {
2256 // Note that here the input file values for chroma subsampling
2257 // are used instead of those from the command line.
2258 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
2259 AV1E_SET_CHROMA_SUBSAMPLING_X,
2260 input.y4m.dst_c_dec_h >> 1);
2261 ctx_exit_on_error(&stream->encoder,
2262 "Failed to set chroma subsampling x");
2263 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
2264 AV1E_SET_CHROMA_SUBSAMPLING_Y,
2265 input.y4m.dst_c_dec_v >> 1);
2266 ctx_exit_on_error(&stream->encoder,
2267 "Failed to set chroma subsampling y");
2268 } else if (input.bit_depth == 12 &&
2269 input.file_type == FILE_TYPE_RAW) {
2270 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
2271 AV1E_SET_CHROMA_SUBSAMPLING_X,
2272 stream->chroma_subsampling_x);
2273 ctx_exit_on_error(&stream->encoder,
2274 "Failed to set chroma subsampling x");
2275 AOM_CODEC_CONTROL_TYPECHECKED(&stream->encoder,
2276 AV1E_SET_CHROMA_SUBSAMPLING_Y,
2277 stream->chroma_subsampling_y);
2278 ctx_exit_on_error(&stream->encoder,
2279 "Failed to set chroma subsampling y");
2280 }
2281 break;
2282 default: break;
2283 }
2284 }
2285 /* Automatically set the codec bit depth to match the input bit depth.
2286 * Upgrade the profile if required. */
2287 if (stream->config.cfg.g_input_bit_depth >
2288 (unsigned int)stream->config.cfg.g_bit_depth) {
2289 stream->config.cfg.g_bit_depth = stream->config.cfg.g_input_bit_depth;
2290 if (!global.quiet) {
2291 fprintf(stderr,
2292 "Warning: automatically updating bit depth to %d to "
2293 "match input format.\n",
2294 stream->config.cfg.g_input_bit_depth);
2295 }
2296 }
2297 #if !CONFIG_AV1_HIGHBITDEPTH
2298 if (stream->config.cfg.g_bit_depth > 8) {
2299 fatal("Unsupported bit-depth with CONFIG_AV1_HIGHBITDEPTH=0\n");
2300 }
2301 #endif // CONFIG_AV1_HIGHBITDEPTH
2302 if (stream->config.cfg.g_bit_depth > 10) {
2303 switch (stream->config.cfg.g_profile) {
2304 case 0:
2305 case 1:
2306 stream->config.cfg.g_profile = 2;
2307 profile_updated = 1;
2308 break;
2309 default: break;
2310 }
2311 }
2312 if (stream->config.cfg.g_bit_depth > 8) {
2313 stream->config.use_16bit_internal = 1;
2314 }
2315 if (profile_updated && !global.quiet) {
2316 fprintf(stderr,
2317 "Warning: automatically updating to profile %d to "
2318 "match input format.\n",
2319 stream->config.cfg.g_profile);
2320 }
2321 if ((global.show_psnr == 2) && (stream->config.cfg.g_input_bit_depth ==
2322 stream->config.cfg.g_bit_depth)) {
2323 fprintf(stderr,
2324 "Warning: --psnr==2 and --psnr==1 will provide same "
2325 "results when input bit-depth == stream bit-depth, "
2326 "falling back to default psnr value\n");
2327 global.show_psnr = 1;
2328 }
2329 if (global.show_psnr < 0 || global.show_psnr > 2) {
2330 fprintf(stderr,
2331 "Warning: --psnr can take only 0,1,2 as values,"
2332 "falling back to default psnr value\n");
2333 global.show_psnr = 1;
2334 }
2335 /* Set limit */
2336 stream->config.cfg.g_limit = global.limit;
2337 }
2338
2339 FOREACH_STREAM(stream, streams) {
2340 set_stream_dimensions(stream, input.width, input.height);
2341 stream->config.color_range = input.color_range;
2342 }
2343 FOREACH_STREAM(stream, streams) { validate_stream_config(stream, &global); }
2344
2345 /* Ensure that --passes and --pass are consistent. If --pass is set and
2346 * --passes >= 2, ensure --fpf was set.
2347 */
2348 if (global.pass > 0 && global.pass <= 3 && global.passes >= 2) {
2349 FOREACH_STREAM(stream, streams) {
2350 if (!stream->config.stats_fn)
2351 die("Stream %d: Must specify --fpf when --pass=%d"
2352 " and --passes=%d\n",
2353 stream->index, global.pass, global.passes);
2354 }
2355 }
2356
2357 #if !CONFIG_WEBM_IO
2358 FOREACH_STREAM(stream, streams) {
2359 if (stream->config.write_webm) {
2360 stream->config.write_webm = 0;
2361 stream->config.write_ivf = 0;
2362 aom_tools_warn("aomenc compiled w/o WebM support. Writing OBU stream.");
2363 }
2364 }
2365 #endif
2366
2367 /* Use the frame rate from the file only if none was specified
2368 * on the command-line.
2369 */
2370 if (!global.have_framerate) {
2371 global.framerate.num = input.framerate.numerator;
2372 global.framerate.den = input.framerate.denominator;
2373 }
2374 FOREACH_STREAM(stream, streams) {
2375 stream->config.cfg.g_timebase.den = global.framerate.num;
2376 stream->config.cfg.g_timebase.num = global.framerate.den;
2377 }
2378 /* Show configuration */
2379 if (global.verbose && pass == 0) {
2380 FOREACH_STREAM(stream, streams) {
2381 show_stream_config(stream, &global, &input);
2382 }
2383 }
2384
2385 if (pass == (global.pass ? global.pass - 1 : 0)) {
2386 // The Y4M reader does its own allocation.
2387 if (input.file_type != FILE_TYPE_Y4M) {
2388 aom_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2389 }
2390 FOREACH_STREAM(stream, streams) {
2391 stream->rate_hist =
2392 init_rate_histogram(&stream->config.cfg, &global.framerate);
2393 }
2394 }
2395
2396 FOREACH_STREAM(stream, streams) { setup_pass(stream, &global, pass); }
2397 FOREACH_STREAM(stream, streams) { initialize_encoder(stream, &global); }
2398 FOREACH_STREAM(stream, streams) {
2399 char *encoder_settings = NULL;
2400 #if CONFIG_WEBM_IO
2401 // Test frameworks may compare outputs from different versions, but only
2402 // wish to check for bitstream changes. The encoder-settings tag, however,
2403 // can vary if the version is updated, even if no encoder algorithm
2404 // changes were made. To work around this issue, do not output
2405 // the encoder-settings tag when --debug is enabled (which is the flag
2406 // that test frameworks should use, when they want deterministic output
2407 // from the container format).
2408 if (stream->config.write_webm && !stream->webm_ctx.debug) {
2409 encoder_settings = extract_encoder_settings(
2410 aom_codec_version_str(), argv_, argc, input.filename);
2411 if (encoder_settings == NULL) {
2412 fprintf(
2413 stderr,
2414 "Warning: unable to extract encoder settings. Continuing...\n");
2415 }
2416 }
2417 #endif
2418 open_output_file(stream, &global, &input.pixel_aspect_ratio,
2419 encoder_settings);
2420 free(encoder_settings);
2421 }
2422
2423 if (strcmp(get_short_name_by_aom_encoder(global.codec), "av1") == 0) {
2424 // Check to see if at least one stream uses 16 bit internal.
2425 // Currently assume that the bit_depths for all streams using
2426 // highbitdepth are the same.
2427 FOREACH_STREAM(stream, streams) {
2428 if (stream->config.use_16bit_internal) {
2429 do_16bit_internal = 1;
2430 }
2431 input_shift = (int)stream->config.cfg.g_bit_depth -
2432 stream->config.cfg.g_input_bit_depth;
2433 }
2434 }
2435
2436 frame_avail = 1;
2437 got_data = 0;
2438
2439 while (frame_avail || got_data) {
2440 struct aom_usec_timer timer;
2441
2442 if (!global.limit || frames_in < global.limit) {
2443 frame_avail = read_frame(&input, &raw);
2444
2445 if (frame_avail) frames_in++;
2446 seen_frames =
2447 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
2448
2449 if (!global.quiet) {
2450 float fps = usec_to_fps(cx_time, seen_frames);
2451 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2452
2453 if (stream_cnt == 1)
2454 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
2455 streams->frames_out, (int64_t)streams->nbytes);
2456 else
2457 fprintf(stderr, "frame %4d ", frames_in);
2458
2459 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
2460 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2461 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
2462 fps >= 1.0 ? "fps" : "fpm");
2463 print_time("ETA", estimated_time_left);
2464 // mingw-w64 gcc does not match msvc for stderr buffering behavior
2465 // and uses line buffering, thus the progress output is not
2466 // real-time. The fflush() is here to make sure the progress output
2467 // is sent out while the clip is being processed.
2468 fflush(stderr);
2469 }
2470
2471 } else {
2472 frame_avail = 0;
2473 }
2474
2475 if (frames_in > global.skip_frames) {
2476 aom_image_t *frame_to_encode;
2477 if (input_shift || (do_16bit_internal && input.bit_depth == 8)) {
2478 assert(do_16bit_internal);
2479 // Input bit depth and stream bit depth do not match, so up
2480 // shift frame to stream bit depth
2481 if (!allocated_raw_shift) {
2482 aom_img_alloc(&raw_shift, raw.fmt | AOM_IMG_FMT_HIGHBITDEPTH,
2483 input.width, input.height, 32);
2484 allocated_raw_shift = 1;
2485 }
2486 aom_img_upshift(&raw_shift, &raw, input_shift);
2487 frame_to_encode = &raw_shift;
2488 } else {
2489 frame_to_encode = &raw;
2490 }
2491 aom_usec_timer_start(&timer);
2492 if (do_16bit_internal) {
2493 assert(frame_to_encode->fmt & AOM_IMG_FMT_HIGHBITDEPTH);
2494 FOREACH_STREAM(stream, streams) {
2495 if (stream->config.use_16bit_internal)
2496 encode_frame(stream, &global,
2497 frame_avail ? frame_to_encode : NULL, frames_in);
2498 else
2499 assert(0);
2500 }
2501 } else {
2502 assert((frame_to_encode->fmt & AOM_IMG_FMT_HIGHBITDEPTH) == 0);
2503 FOREACH_STREAM(stream, streams) {
2504 encode_frame(stream, &global, frame_avail ? frame_to_encode : NULL,
2505 frames_in);
2506 }
2507 }
2508 aom_usec_timer_mark(&timer);
2509 cx_time += aom_usec_timer_elapsed(&timer);
2510
2511 FOREACH_STREAM(stream, streams) { update_quantizer_histogram(stream); }
2512
2513 got_data = 0;
2514 FOREACH_STREAM(stream, streams) {
2515 get_cx_data(stream, &global, &got_data);
2516 }
2517
2518 if (!got_data && input.length && streams != NULL &&
2519 !streams->frames_out) {
2520 lagged_count = global.limit ? seen_frames : ftello(input.file);
2521 } else if (input.length) {
2522 int64_t remaining;
2523 int64_t rate;
2524
2525 if (global.limit) {
2526 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2527
2528 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2529 remaining = 1000 * (global.limit - global.skip_frames -
2530 seen_frames + lagged_count);
2531 } else {
2532 const int64_t input_pos = ftello(input.file);
2533 const int64_t input_pos_lagged = input_pos - lagged_count;
2534 const int64_t input_limit = input.length;
2535
2536 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2537 remaining = input_limit - input_pos + lagged_count;
2538 }
2539
2540 average_rate =
2541 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
2542 estimated_time_left = average_rate ? remaining / average_rate : -1;
2543 }
2544
2545 if (got_data && global.test_decode != TEST_DECODE_OFF) {
2546 FOREACH_STREAM(stream, streams) {
2547 test_decode(stream, global.test_decode);
2548 }
2549 }
2550 }
2551
2552 fflush(stdout);
2553 if (!global.quiet) fprintf(stderr, "\033[K");
2554 }
2555
2556 if (stream_cnt > 1) fprintf(stderr, "\n");
2557
2558 if (!global.quiet) {
2559 FOREACH_STREAM(stream, streams) {
2560 const int64_t bpf =
2561 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0;
2562 const int64_t bps = bpf * global.framerate.num / global.framerate.den;
2563 fprintf(stderr,
2564 "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64
2565 "b/f %7" PRId64
2566 "b/s"
2567 " %7" PRId64 " %s (%.2f fps)\033[K\n",
2568 pass + 1, global.passes, frames_in, stream->frames_out,
2569 (int64_t)stream->nbytes, bpf, bps,
2570 stream->cx_time > 9999999 ? stream->cx_time / 1000
2571 : stream->cx_time,
2572 stream->cx_time > 9999999 ? "ms" : "us",
2573 usec_to_fps(stream->cx_time, seen_frames));
2574 // This instance of cr does not need fflush as it is followed by a
2575 // newline in the same string.
2576 }
2577 }
2578
2579 if (global.show_psnr >= 1) {
2580 if (get_fourcc_by_aom_encoder(global.codec) == AV1_FOURCC) {
2581 FOREACH_STREAM(stream, streams) {
2582 int64_t bps = 0;
2583 if (global.show_psnr == 1) {
2584 if (stream->psnr_count[0] && seen_frames && global.framerate.den) {
2585 bps = (int64_t)stream->nbytes * 8 *
2586 (int64_t)global.framerate.num / global.framerate.den /
2587 seen_frames;
2588 }
2589 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1,
2590 bps);
2591 }
2592 if (global.show_psnr == 2) {
2593 #if CONFIG_AV1_HIGHBITDEPTH
2594 if (stream->config.cfg.g_input_bit_depth <
2595 (unsigned int)stream->config.cfg.g_bit_depth)
2596 show_psnr_hbd(stream, (1 << stream->config.cfg.g_bit_depth) - 1,
2597 bps);
2598 #endif
2599 }
2600 }
2601 } else {
2602 FOREACH_STREAM(stream, streams) { show_psnr(stream, 255.0, 0); }
2603 }
2604 }
2605
2606 if (pass == global.passes - 1) {
2607 FOREACH_STREAM(stream, streams) {
2608 int num_operating_points;
2609 int levels[32];
2610 int target_levels[32];
2611 aom_codec_control(&stream->encoder, AV1E_GET_NUM_OPERATING_POINTS,
2612 &num_operating_points);
2613 aom_codec_control(&stream->encoder, AV1E_GET_SEQ_LEVEL_IDX, levels);
2614 aom_codec_control(&stream->encoder, AV1E_GET_TARGET_SEQ_LEVEL_IDX,
2615 target_levels);
2616
2617 for (int i = 0; i < num_operating_points; i++) {
2618 if (levels[i] > target_levels[i]) {
2619 if (levels[i] == 31) {
2620 aom_tools_warn(
2621 "Failed to encode to target level %d.%d for operating point "
2622 "%d. The output level is SEQ_LEVEL_MAX",
2623 2 + (target_levels[i] >> 2), target_levels[i] & 3, i);
2624 } else {
2625 aom_tools_warn(
2626 "Failed to encode to target level %d.%d for operating point "
2627 "%d. The output level is %d.%d",
2628 2 + (target_levels[i] >> 2), target_levels[i] & 3, i,
2629 2 + (levels[i] >> 2), levels[i] & 3);
2630 }
2631 }
2632 }
2633 }
2634 }
2635
2636 FOREACH_STREAM(stream, streams) { aom_codec_destroy(&stream->encoder); }
2637
2638 if (global.test_decode != TEST_DECODE_OFF) {
2639 FOREACH_STREAM(stream, streams) { aom_codec_destroy(&stream->decoder); }
2640 }
2641
2642 close_input_file(&input);
2643
2644 if (global.test_decode == TEST_DECODE_FATAL) {
2645 FOREACH_STREAM(stream, streams) { res |= stream->mismatch_seen; }
2646 }
2647 FOREACH_STREAM(stream, streams) {
2648 close_output_file(stream, get_fourcc_by_aom_encoder(global.codec));
2649 }
2650
2651 FOREACH_STREAM(stream, streams) {
2652 stats_close(&stream->stats, global.passes - 1);
2653 }
2654
2655 if (global.pass) break;
2656 }
2657
2658 if (global.show_q_hist_buckets) {
2659 FOREACH_STREAM(stream, streams) {
2660 show_q_histogram(stream->counts, global.show_q_hist_buckets);
2661 }
2662 }
2663
2664 if (global.show_rate_hist_buckets) {
2665 FOREACH_STREAM(stream, streams) {
2666 show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2667 global.show_rate_hist_buckets);
2668 }
2669 }
2670 FOREACH_STREAM(stream, streams) { destroy_rate_histogram(stream->rate_hist); }
2671
2672 #if CONFIG_INTERNAL_STATS
2673 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2674 * to match some existing utilities.
2675 */
2676 if (!(global.pass == 1 && global.passes == 2)) {
2677 FOREACH_STREAM(stream, streams) {
2678 FILE *f = fopen("opsnr.stt", "a");
2679 if (stream->mismatch_seen) {
2680 fprintf(f, "First mismatch occurred in frame %d\n",
2681 stream->mismatch_seen);
2682 } else {
2683 fprintf(f, "No mismatch detected in recon buffers\n");
2684 }
2685 fclose(f);
2686 }
2687 }
2688 #endif
2689
2690 if (allocated_raw_shift) aom_img_free(&raw_shift);
2691 aom_img_free(&raw);
2692 free(argv);
2693 free(streams);
2694 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2695 }
2696