1 /*
2 * Copyright (c) 2010, Google, Inc.
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file
23 * AV1 encoder support via libaom
24 */
25
26 #define AOM_DISABLE_CTRL_TYPECHECKS 1
27 #include <aom/aom_encoder.h>
28 #include <aom/aomcx.h>
29
30 #include "libavutil/avassert.h"
31 #include "libavutil/base64.h"
32 #include "libavutil/common.h"
33 #include "libavutil/mathematics.h"
34 #include "libavutil/opt.h"
35 #include "libavutil/pixdesc.h"
36
37 #include "av1.h"
38 #include "avcodec.h"
39 #include "internal.h"
40 #include "packet_internal.h"
41 #include "profiles.h"
42
43 /*
44 * Portion of struct aom_codec_cx_pkt from aom_encoder.h.
45 * One encoded frame returned from the library.
46 */
47 struct FrameListData {
48 void *buf; /**< compressed data buffer */
49 size_t sz; /**< length of compressed data */
50 int64_t pts; /**< time stamp to show frame
51 (in timebase units) */
52 unsigned long duration; /**< duration to show frame
53 (in timebase units) */
54 uint32_t flags; /**< flags for this frame */
55 uint64_t sse[4];
56 int have_sse; /**< true if we have pending sse[] */
57 uint64_t frame_number;
58 struct FrameListData *next;
59 };
60
61 typedef struct AOMEncoderContext {
62 AVClass *class;
63 AVBSFContext *bsf;
64 struct aom_codec_ctx encoder;
65 struct aom_image rawimg;
66 struct aom_fixed_buf twopass_stats;
67 struct FrameListData *coded_frame_list;
68 int cpu_used;
69 int auto_alt_ref;
70 int arnr_max_frames;
71 int arnr_strength;
72 int aq_mode;
73 int lag_in_frames;
74 int error_resilient;
75 int crf;
76 int static_thresh;
77 int drop_threshold;
78 int denoise_noise_level;
79 int denoise_block_size;
80 uint64_t sse[4];
81 int have_sse; /**< true if we have pending sse[] */
82 uint64_t frame_number;
83 int rc_undershoot_pct;
84 int rc_overshoot_pct;
85 int minsection_pct;
86 int maxsection_pct;
87 int frame_parallel;
88 int tile_cols, tile_rows;
89 int tile_cols_log2, tile_rows_log2;
90 aom_superblock_size_t superblock_size;
91 int uniform_tiles;
92 int row_mt;
93 int enable_cdef;
94 int enable_global_motion;
95 int enable_intrabc;
96 int enable_restoration;
97 int usage;
98 int tune;
99 } AOMContext;
100
101 static const char *const ctlidstr[] = {
102 [AOME_SET_CPUUSED] = "AOME_SET_CPUUSED",
103 [AOME_SET_CQ_LEVEL] = "AOME_SET_CQ_LEVEL",
104 [AOME_SET_ENABLEAUTOALTREF] = "AOME_SET_ENABLEAUTOALTREF",
105 [AOME_SET_ARNR_MAXFRAMES] = "AOME_SET_ARNR_MAXFRAMES",
106 [AOME_SET_ARNR_STRENGTH] = "AOME_SET_ARNR_STRENGTH",
107 [AOME_SET_STATIC_THRESHOLD] = "AOME_SET_STATIC_THRESHOLD",
108 [AV1E_SET_COLOR_RANGE] = "AV1E_SET_COLOR_RANGE",
109 [AV1E_SET_COLOR_PRIMARIES] = "AV1E_SET_COLOR_PRIMARIES",
110 [AV1E_SET_MATRIX_COEFFICIENTS] = "AV1E_SET_MATRIX_COEFFICIENTS",
111 [AV1E_SET_TRANSFER_CHARACTERISTICS] = "AV1E_SET_TRANSFER_CHARACTERISTICS",
112 [AV1E_SET_AQ_MODE] = "AV1E_SET_AQ_MODE",
113 [AV1E_SET_FRAME_PARALLEL_DECODING] = "AV1E_SET_FRAME_PARALLEL_DECODING",
114 [AV1E_SET_SUPERBLOCK_SIZE] = "AV1E_SET_SUPERBLOCK_SIZE",
115 [AV1E_SET_TILE_COLUMNS] = "AV1E_SET_TILE_COLUMNS",
116 [AV1E_SET_TILE_ROWS] = "AV1E_SET_TILE_ROWS",
117 [AV1E_SET_ENABLE_RESTORATION] = "AV1E_SET_ENABLE_RESTORATION",
118 #ifdef AOM_CTRL_AV1E_SET_ROW_MT
119 [AV1E_SET_ROW_MT] = "AV1E_SET_ROW_MT",
120 #endif
121 #ifdef AOM_CTRL_AV1E_SET_DENOISE_NOISE_LEVEL
122 [AV1E_SET_DENOISE_NOISE_LEVEL] = "AV1E_SET_DENOISE_NOISE_LEVEL",
123 #endif
124 #ifdef AOM_CTRL_AV1E_SET_DENOISE_BLOCK_SIZE
125 [AV1E_SET_DENOISE_BLOCK_SIZE] = "AV1E_SET_DENOISE_BLOCK_SIZE",
126 #endif
127 #ifdef AOM_CTRL_AV1E_SET_MAX_REFERENCE_FRAMES
128 [AV1E_SET_MAX_REFERENCE_FRAMES] = "AV1E_SET_MAX_REFERENCE_FRAMES",
129 #endif
130 #ifdef AOM_CTRL_AV1E_SET_ENABLE_GLOBAL_MOTION
131 [AV1E_SET_ENABLE_GLOBAL_MOTION] = "AV1E_SET_ENABLE_GLOBAL_MOTION",
132 #endif
133 #ifdef AOM_CTRL_AV1E_SET_ENABLE_INTRABC
134 [AV1E_SET_ENABLE_INTRABC] = "AV1E_SET_ENABLE_INTRABC",
135 #endif
136 [AV1E_SET_ENABLE_CDEF] = "AV1E_SET_ENABLE_CDEF",
137 [AOME_SET_TUNING] = "AOME_SET_TUNING",
138 };
139
log_encoder_error(AVCodecContext * avctx,const char * desc)140 static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
141 {
142 AOMContext *ctx = avctx->priv_data;
143 const char *error = aom_codec_error(&ctx->encoder);
144 const char *detail = aom_codec_error_detail(&ctx->encoder);
145
146 av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
147 if (detail)
148 av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
149 }
150
dump_enc_cfg(AVCodecContext * avctx,const struct aom_codec_enc_cfg * cfg)151 static av_cold void dump_enc_cfg(AVCodecContext *avctx,
152 const struct aom_codec_enc_cfg *cfg)
153 {
154 int width = -30;
155 int level = AV_LOG_DEBUG;
156
157 av_log(avctx, level, "aom_codec_enc_cfg\n");
158 av_log(avctx, level, "generic settings\n"
159 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
160 " %*s%u\n %*s%u\n"
161 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
162 width, "g_usage:", cfg->g_usage,
163 width, "g_threads:", cfg->g_threads,
164 width, "g_profile:", cfg->g_profile,
165 width, "g_w:", cfg->g_w,
166 width, "g_h:", cfg->g_h,
167 width, "g_bit_depth:", cfg->g_bit_depth,
168 width, "g_input_bit_depth:", cfg->g_input_bit_depth,
169 width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
170 width, "g_error_resilient:", cfg->g_error_resilient,
171 width, "g_pass:", cfg->g_pass,
172 width, "g_lag_in_frames:", cfg->g_lag_in_frames);
173 av_log(avctx, level, "rate control settings\n"
174 " %*s%u\n %*s%d\n %*s%p(%"SIZE_SPECIFIER")\n %*s%u\n",
175 width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
176 width, "rc_end_usage:", cfg->rc_end_usage,
177 width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
178 width, "rc_target_bitrate:", cfg->rc_target_bitrate);
179 av_log(avctx, level, "quantizer settings\n"
180 " %*s%u\n %*s%u\n",
181 width, "rc_min_quantizer:", cfg->rc_min_quantizer,
182 width, "rc_max_quantizer:", cfg->rc_max_quantizer);
183 av_log(avctx, level, "bitrate tolerance\n"
184 " %*s%u\n %*s%u\n",
185 width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
186 width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
187 av_log(avctx, level, "decoder buffer model\n"
188 " %*s%u\n %*s%u\n %*s%u\n",
189 width, "rc_buf_sz:", cfg->rc_buf_sz,
190 width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
191 width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
192 av_log(avctx, level, "2 pass rate control settings\n"
193 " %*s%u\n %*s%u\n %*s%u\n",
194 width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
195 width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
196 width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
197 av_log(avctx, level, "keyframing settings\n"
198 " %*s%d\n %*s%u\n %*s%u\n",
199 width, "kf_mode:", cfg->kf_mode,
200 width, "kf_min_dist:", cfg->kf_min_dist,
201 width, "kf_max_dist:", cfg->kf_max_dist);
202 av_log(avctx, level, "tile settings\n"
203 " %*s%d\n %*s%d\n",
204 width, "tile_width_count:", cfg->tile_width_count,
205 width, "tile_height_count:", cfg->tile_height_count);
206 av_log(avctx, level, "\n");
207 }
208
coded_frame_add(void * list,struct FrameListData * cx_frame)209 static void coded_frame_add(void *list, struct FrameListData *cx_frame)
210 {
211 struct FrameListData **p = list;
212
213 while (*p)
214 p = &(*p)->next;
215 *p = cx_frame;
216 cx_frame->next = NULL;
217 }
218
free_coded_frame(struct FrameListData * cx_frame)219 static av_cold void free_coded_frame(struct FrameListData *cx_frame)
220 {
221 av_freep(&cx_frame->buf);
222 av_freep(&cx_frame);
223 }
224
free_frame_list(struct FrameListData * list)225 static av_cold void free_frame_list(struct FrameListData *list)
226 {
227 struct FrameListData *p = list;
228
229 while (p) {
230 list = list->next;
231 free_coded_frame(p);
232 p = list;
233 }
234 }
235
codecctl_int(AVCodecContext * avctx,aome_enc_control_id id,int val)236 static av_cold int codecctl_int(AVCodecContext *avctx,
237 #ifdef UENUM1BYTE
238 aome_enc_control_id id,
239 #else
240 enum aome_enc_control_id id,
241 #endif
242 int val)
243 {
244 AOMContext *ctx = avctx->priv_data;
245 char buf[80];
246 int width = -30;
247 int res;
248
249 snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
250 av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
251
252 res = aom_codec_control(&ctx->encoder, id, val);
253 if (res != AOM_CODEC_OK) {
254 snprintf(buf, sizeof(buf), "Failed to set %s codec control",
255 ctlidstr[id]);
256 log_encoder_error(avctx, buf);
257 return AVERROR(EINVAL);
258 }
259
260 return 0;
261 }
262
aom_free(AVCodecContext * avctx)263 static av_cold int aom_free(AVCodecContext *avctx)
264 {
265 AOMContext *ctx = avctx->priv_data;
266
267 aom_codec_destroy(&ctx->encoder);
268 av_freep(&ctx->twopass_stats.buf);
269 av_freep(&avctx->stats_out);
270 free_frame_list(ctx->coded_frame_list);
271 av_bsf_free(&ctx->bsf);
272 return 0;
273 }
274
set_pix_fmt(AVCodecContext * avctx,aom_codec_caps_t codec_caps,struct aom_codec_enc_cfg * enccfg,aom_codec_flags_t * flags,aom_img_fmt_t * img_fmt)275 static int set_pix_fmt(AVCodecContext *avctx, aom_codec_caps_t codec_caps,
276 struct aom_codec_enc_cfg *enccfg, aom_codec_flags_t *flags,
277 aom_img_fmt_t *img_fmt)
278 {
279 AOMContext av_unused *ctx = avctx->priv_data;
280 enccfg->g_bit_depth = enccfg->g_input_bit_depth = 8;
281 switch (avctx->pix_fmt) {
282 case AV_PIX_FMT_YUV420P:
283 enccfg->g_profile = FF_PROFILE_AV1_MAIN;
284 *img_fmt = AOM_IMG_FMT_I420;
285 return 0;
286 case AV_PIX_FMT_YUV422P:
287 enccfg->g_profile = FF_PROFILE_AV1_PROFESSIONAL;
288 *img_fmt = AOM_IMG_FMT_I422;
289 return 0;
290 case AV_PIX_FMT_YUV444P:
291 enccfg->g_profile = FF_PROFILE_AV1_HIGH;
292 *img_fmt = AOM_IMG_FMT_I444;
293 return 0;
294 case AV_PIX_FMT_YUV420P10:
295 case AV_PIX_FMT_YUV420P12:
296 if (codec_caps & AOM_CODEC_CAP_HIGHBITDEPTH) {
297 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
298 avctx->pix_fmt == AV_PIX_FMT_YUV420P10 ? 10 : 12;
299 enccfg->g_profile =
300 enccfg->g_bit_depth == 10 ? FF_PROFILE_AV1_MAIN : FF_PROFILE_AV1_PROFESSIONAL;
301 *img_fmt = AOM_IMG_FMT_I42016;
302 *flags |= AOM_CODEC_USE_HIGHBITDEPTH;
303 return 0;
304 }
305 break;
306 case AV_PIX_FMT_YUV422P10:
307 case AV_PIX_FMT_YUV422P12:
308 if (codec_caps & AOM_CODEC_CAP_HIGHBITDEPTH) {
309 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
310 avctx->pix_fmt == AV_PIX_FMT_YUV422P10 ? 10 : 12;
311 enccfg->g_profile = FF_PROFILE_AV1_PROFESSIONAL;
312 *img_fmt = AOM_IMG_FMT_I42216;
313 *flags |= AOM_CODEC_USE_HIGHBITDEPTH;
314 return 0;
315 }
316 break;
317 case AV_PIX_FMT_YUV444P10:
318 case AV_PIX_FMT_YUV444P12:
319 if (codec_caps & AOM_CODEC_CAP_HIGHBITDEPTH) {
320 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
321 avctx->pix_fmt == AV_PIX_FMT_YUV444P10 ? 10 : 12;
322 enccfg->g_profile =
323 enccfg->g_bit_depth == 10 ? FF_PROFILE_AV1_HIGH : FF_PROFILE_AV1_PROFESSIONAL;
324 *img_fmt = AOM_IMG_FMT_I44416;
325 *flags |= AOM_CODEC_USE_HIGHBITDEPTH;
326 return 0;
327 }
328 break;
329 default:
330 break;
331 }
332 av_log(avctx, AV_LOG_ERROR, "Unsupported pixel format.\n");
333 return AVERROR_INVALIDDATA;
334 }
335
set_color_range(AVCodecContext * avctx)336 static void set_color_range(AVCodecContext *avctx)
337 {
338 aom_color_range_t aom_cr;
339 switch (avctx->color_range) {
340 case AVCOL_RANGE_UNSPECIFIED:
341 case AVCOL_RANGE_MPEG: aom_cr = AOM_CR_STUDIO_RANGE; break;
342 case AVCOL_RANGE_JPEG: aom_cr = AOM_CR_FULL_RANGE; break;
343 default:
344 av_log(avctx, AV_LOG_WARNING, "Unsupported color range (%d)\n",
345 avctx->color_range);
346 return;
347 }
348
349 codecctl_int(avctx, AV1E_SET_COLOR_RANGE, aom_cr);
350 }
351
count_uniform_tiling(int dim,int sb_size,int tiles_log2)352 static int count_uniform_tiling(int dim, int sb_size, int tiles_log2)
353 {
354 int sb_dim = (dim + sb_size - 1) / sb_size;
355 int tile_dim = (sb_dim + (1 << tiles_log2) - 1) >> tiles_log2;
356 av_assert0(tile_dim > 0);
357 return (sb_dim + tile_dim - 1) / tile_dim;
358 }
359
choose_tiling(AVCodecContext * avctx,struct aom_codec_enc_cfg * enccfg)360 static int choose_tiling(AVCodecContext *avctx,
361 struct aom_codec_enc_cfg *enccfg)
362 {
363 AOMContext *ctx = avctx->priv_data;
364 int sb_128x128_possible, sb_size, sb_width, sb_height;
365 int uniform_rows, uniform_cols;
366 int uniform_64x64_possible, uniform_128x128_possible;
367 int tile_size, rounding, i;
368
369 if (ctx->tile_cols_log2 >= 0)
370 ctx->tile_cols = 1 << ctx->tile_cols_log2;
371 if (ctx->tile_rows_log2 >= 0)
372 ctx->tile_rows = 1 << ctx->tile_rows_log2;
373
374 if (ctx->tile_cols == 0) {
375 ctx->tile_cols = (avctx->width + AV1_MAX_TILE_WIDTH - 1) /
376 AV1_MAX_TILE_WIDTH;
377 if (ctx->tile_cols > 1) {
378 av_log(avctx, AV_LOG_DEBUG, "Automatically using %d tile "
379 "columns to fill width.\n", ctx->tile_cols);
380 }
381 }
382 av_assert0(ctx->tile_cols > 0);
383 if (ctx->tile_rows == 0) {
384 int max_tile_width =
385 FFALIGN((FFALIGN(avctx->width, 128) +
386 ctx->tile_cols - 1) / ctx->tile_cols, 128);
387 ctx->tile_rows =
388 (max_tile_width * FFALIGN(avctx->height, 128) +
389 AV1_MAX_TILE_AREA - 1) / AV1_MAX_TILE_AREA;
390 if (ctx->tile_rows > 1) {
391 av_log(avctx, AV_LOG_DEBUG, "Automatically using %d tile "
392 "rows to fill area.\n", ctx->tile_rows);
393 }
394 }
395 av_assert0(ctx->tile_rows > 0);
396
397 if ((avctx->width + 63) / 64 < ctx->tile_cols ||
398 (avctx->height + 63) / 64 < ctx->tile_rows) {
399 av_log(avctx, AV_LOG_ERROR, "Invalid tile sizing: frame not "
400 "large enough to fit specified tile arrangement.\n");
401 return AVERROR(EINVAL);
402 }
403 if (ctx->tile_cols > AV1_MAX_TILE_COLS ||
404 ctx->tile_rows > AV1_MAX_TILE_ROWS) {
405 av_log(avctx, AV_LOG_ERROR, "Invalid tile sizing: AV1 does "
406 "not allow more than %dx%d tiles.\n",
407 AV1_MAX_TILE_COLS, AV1_MAX_TILE_ROWS);
408 return AVERROR(EINVAL);
409 }
410 if (avctx->width / ctx->tile_cols > AV1_MAX_TILE_WIDTH) {
411 av_log(avctx, AV_LOG_ERROR, "Invalid tile sizing: AV1 does "
412 "not allow tiles of width greater than %d.\n",
413 AV1_MAX_TILE_WIDTH);
414 return AVERROR(EINVAL);
415 }
416
417 ctx->superblock_size = AOM_SUPERBLOCK_SIZE_DYNAMIC;
418
419 if (ctx->tile_cols == 1 && ctx->tile_rows == 1) {
420 av_log(avctx, AV_LOG_DEBUG, "Using a single tile.\n");
421 return 0;
422 }
423
424 sb_128x128_possible =
425 (avctx->width + 127) / 128 >= ctx->tile_cols &&
426 (avctx->height + 127) / 128 >= ctx->tile_rows;
427
428 ctx->tile_cols_log2 = ctx->tile_cols == 1 ? 0 :
429 av_log2(ctx->tile_cols - 1) + 1;
430 ctx->tile_rows_log2 = ctx->tile_rows == 1 ? 0 :
431 av_log2(ctx->tile_rows - 1) + 1;
432
433 uniform_cols = count_uniform_tiling(avctx->width,
434 64, ctx->tile_cols_log2);
435 uniform_rows = count_uniform_tiling(avctx->height,
436 64, ctx->tile_rows_log2);
437 av_log(avctx, AV_LOG_DEBUG, "Uniform with 64x64 superblocks "
438 "-> %dx%d tiles.\n", uniform_cols, uniform_rows);
439 uniform_64x64_possible = uniform_cols == ctx->tile_cols &&
440 uniform_rows == ctx->tile_rows;
441
442 if (sb_128x128_possible) {
443 uniform_cols = count_uniform_tiling(avctx->width,
444 128, ctx->tile_cols_log2);
445 uniform_rows = count_uniform_tiling(avctx->height,
446 128, ctx->tile_rows_log2);
447 av_log(avctx, AV_LOG_DEBUG, "Uniform with 128x128 superblocks "
448 "-> %dx%d tiles.\n", uniform_cols, uniform_rows);
449 uniform_128x128_possible = uniform_cols == ctx->tile_cols &&
450 uniform_rows == ctx->tile_rows;
451 } else {
452 av_log(avctx, AV_LOG_DEBUG, "128x128 superblocks not possible.\n");
453 uniform_128x128_possible = 0;
454 }
455
456 ctx->uniform_tiles = 1;
457 if (uniform_64x64_possible && uniform_128x128_possible) {
458 av_log(avctx, AV_LOG_DEBUG, "Using uniform tiling with dynamic "
459 "superblocks (tile_cols_log2 = %d, tile_rows_log2 = %d).\n",
460 ctx->tile_cols_log2, ctx->tile_rows_log2);
461 return 0;
462 }
463 if (uniform_64x64_possible && !sb_128x128_possible) {
464 av_log(avctx, AV_LOG_DEBUG, "Using uniform tiling with 64x64 "
465 "superblocks (tile_cols_log2 = %d, tile_rows_log2 = %d).\n",
466 ctx->tile_cols_log2, ctx->tile_rows_log2);
467 ctx->superblock_size = AOM_SUPERBLOCK_SIZE_64X64;
468 return 0;
469 }
470 if (uniform_128x128_possible) {
471 av_log(avctx, AV_LOG_DEBUG, "Using uniform tiling with 128x128 "
472 "superblocks (tile_cols_log2 = %d, tile_rows_log2 = %d).\n",
473 ctx->tile_cols_log2, ctx->tile_rows_log2);
474 ctx->superblock_size = AOM_SUPERBLOCK_SIZE_128X128;
475 return 0;
476 }
477 ctx->uniform_tiles = 0;
478
479 if (sb_128x128_possible) {
480 sb_size = 128;
481 ctx->superblock_size = AOM_SUPERBLOCK_SIZE_128X128;
482 } else {
483 sb_size = 64;
484 ctx->superblock_size = AOM_SUPERBLOCK_SIZE_64X64;
485 }
486 av_log(avctx, AV_LOG_DEBUG, "Using fixed tiling with %dx%d "
487 "superblocks (tile_cols = %d, tile_rows = %d).\n",
488 sb_size, sb_size, ctx->tile_cols, ctx->tile_rows);
489
490 enccfg->tile_width_count = ctx->tile_cols;
491 enccfg->tile_height_count = ctx->tile_rows;
492
493 sb_width = (avctx->width + sb_size - 1) / sb_size;
494 sb_height = (avctx->height + sb_size - 1) / sb_size;
495
496 tile_size = sb_width / ctx->tile_cols;
497 rounding = sb_width % ctx->tile_cols;
498 for (i = 0; i < ctx->tile_cols; i++) {
499 enccfg->tile_widths[i] = tile_size +
500 (i < rounding / 2 ||
501 i > ctx->tile_cols - 1 - (rounding + 1) / 2);
502 }
503
504 tile_size = sb_height / ctx->tile_rows;
505 rounding = sb_height % ctx->tile_rows;
506 for (i = 0; i < ctx->tile_rows; i++) {
507 enccfg->tile_heights[i] = tile_size +
508 (i < rounding / 2 ||
509 i > ctx->tile_rows - 1 - (rounding + 1) / 2);
510 }
511
512 return 0;
513 }
514
aom_init(AVCodecContext * avctx,const struct aom_codec_iface * iface)515 static av_cold int aom_init(AVCodecContext *avctx,
516 const struct aom_codec_iface *iface)
517 {
518 AOMContext *ctx = avctx->priv_data;
519 struct aom_codec_enc_cfg enccfg = { 0 };
520 #ifdef AOM_FRAME_IS_INTRAONLY
521 aom_codec_flags_t flags =
522 (avctx->flags & AV_CODEC_FLAG_PSNR) ? AOM_CODEC_USE_PSNR : 0;
523 #else
524 aom_codec_flags_t flags = 0;
525 #endif
526 AVCPBProperties *cpb_props;
527 int res;
528 aom_img_fmt_t img_fmt;
529 aom_codec_caps_t codec_caps = aom_codec_get_caps(iface);
530
531 av_log(avctx, AV_LOG_INFO, "%s\n", aom_codec_version_str());
532 av_log(avctx, AV_LOG_VERBOSE, "%s\n", aom_codec_build_config());
533
534 if ((res = aom_codec_enc_config_default(iface, &enccfg, 0)) != AOM_CODEC_OK) {
535 av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
536 aom_codec_err_to_string(res));
537 return AVERROR(EINVAL);
538 }
539
540 if (set_pix_fmt(avctx, codec_caps, &enccfg, &flags, &img_fmt))
541 return AVERROR(EINVAL);
542
543 if(!avctx->bit_rate)
544 if(avctx->rc_max_rate || avctx->rc_buffer_size || avctx->rc_initial_buffer_occupancy) {
545 av_log( avctx, AV_LOG_ERROR, "Rate control parameters set without a bitrate\n");
546 return AVERROR(EINVAL);
547 }
548
549 dump_enc_cfg(avctx, &enccfg);
550
551 enccfg.g_w = avctx->width;
552 enccfg.g_h = avctx->height;
553 enccfg.g_timebase.num = avctx->time_base.num;
554 enccfg.g_timebase.den = avctx->time_base.den;
555 enccfg.g_threads =
556 FFMIN(avctx->thread_count ? avctx->thread_count : av_cpu_count(), 64);
557
558 enccfg.g_usage = ctx->usage;
559
560 if (ctx->lag_in_frames >= 0)
561 enccfg.g_lag_in_frames = ctx->lag_in_frames;
562
563 if (avctx->flags & AV_CODEC_FLAG_PASS1)
564 enccfg.g_pass = AOM_RC_FIRST_PASS;
565 else if (avctx->flags & AV_CODEC_FLAG_PASS2)
566 enccfg.g_pass = AOM_RC_LAST_PASS;
567 else
568 enccfg.g_pass = AOM_RC_ONE_PASS;
569
570 if (avctx->rc_min_rate == avctx->rc_max_rate &&
571 avctx->rc_min_rate == avctx->bit_rate && avctx->bit_rate) {
572 enccfg.rc_end_usage = AOM_CBR;
573 } else if (ctx->crf >= 0) {
574 enccfg.rc_end_usage = AOM_CQ;
575 if (!avctx->bit_rate)
576 enccfg.rc_end_usage = AOM_Q;
577 }
578
579 if (avctx->bit_rate) {
580 enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
581 AV_ROUND_NEAR_INF);
582 } else if (enccfg.rc_end_usage != AOM_Q) {
583 enccfg.rc_end_usage = AOM_Q;
584 ctx->crf = 32;
585 av_log(avctx, AV_LOG_WARNING,
586 "Neither bitrate nor constrained quality specified, using default CRF of %d\n",
587 ctx->crf);
588 }
589
590 if (avctx->qmin >= 0)
591 enccfg.rc_min_quantizer = avctx->qmin;
592 if (avctx->qmax >= 0)
593 enccfg.rc_max_quantizer = avctx->qmax;
594
595 if (enccfg.rc_end_usage == AOM_CQ || enccfg.rc_end_usage == AOM_Q) {
596 if (ctx->crf < enccfg.rc_min_quantizer || ctx->crf > enccfg.rc_max_quantizer) {
597 av_log(avctx, AV_LOG_ERROR,
598 "CQ level %d must be between minimum and maximum quantizer value (%d-%d)\n",
599 ctx->crf, enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
600 return AVERROR(EINVAL);
601 }
602 }
603
604 enccfg.rc_dropframe_thresh = ctx->drop_threshold;
605
606 // 0-100 (0 => CBR, 100 => VBR)
607 enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
608 if (ctx->minsection_pct >= 0)
609 enccfg.rc_2pass_vbr_minsection_pct = ctx->minsection_pct;
610 else if (avctx->bit_rate)
611 enccfg.rc_2pass_vbr_minsection_pct =
612 avctx->rc_min_rate * 100LL / avctx->bit_rate;
613 if (ctx->maxsection_pct >= 0)
614 enccfg.rc_2pass_vbr_maxsection_pct = ctx->maxsection_pct;
615 else if (avctx->rc_max_rate)
616 enccfg.rc_2pass_vbr_maxsection_pct =
617 avctx->rc_max_rate * 100LL / avctx->bit_rate;
618
619 if (avctx->rc_buffer_size)
620 enccfg.rc_buf_sz =
621 avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
622 if (avctx->rc_initial_buffer_occupancy)
623 enccfg.rc_buf_initial_sz =
624 avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
625 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
626
627 if (ctx->rc_undershoot_pct >= 0)
628 enccfg.rc_undershoot_pct = ctx->rc_undershoot_pct;
629 if (ctx->rc_overshoot_pct >= 0)
630 enccfg.rc_overshoot_pct = ctx->rc_overshoot_pct;
631
632 // _enc_init() will balk if kf_min_dist differs from max w/AOM_KF_AUTO
633 if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
634 enccfg.kf_min_dist = avctx->keyint_min;
635 if (avctx->gop_size >= 0)
636 enccfg.kf_max_dist = avctx->gop_size;
637
638 if (enccfg.g_pass == AOM_RC_FIRST_PASS)
639 enccfg.g_lag_in_frames = 0;
640 else if (enccfg.g_pass == AOM_RC_LAST_PASS) {
641 int decode_size, ret;
642
643 if (!avctx->stats_in) {
644 av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
645 return AVERROR_INVALIDDATA;
646 }
647
648 ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
649 ret = av_reallocp(&ctx->twopass_stats.buf, ctx->twopass_stats.sz);
650 if (ret < 0) {
651 av_log(avctx, AV_LOG_ERROR,
652 "Stat buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
653 ctx->twopass_stats.sz);
654 ctx->twopass_stats.sz = 0;
655 return ret;
656 }
657 decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
658 ctx->twopass_stats.sz);
659 if (decode_size < 0) {
660 av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
661 return AVERROR_INVALIDDATA;
662 }
663
664 ctx->twopass_stats.sz = decode_size;
665 enccfg.rc_twopass_stats_in = ctx->twopass_stats;
666 }
667
668 /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
669 * complexity playback on low powered devices at the expense of encode
670 * quality. */
671 if (avctx->profile != FF_PROFILE_UNKNOWN)
672 enccfg.g_profile = avctx->profile;
673
674 enccfg.g_error_resilient = ctx->error_resilient;
675
676 res = choose_tiling(avctx, &enccfg);
677 if (res < 0)
678 return res;
679
680 dump_enc_cfg(avctx, &enccfg);
681 /* Construct Encoder Context */
682 res = aom_codec_enc_init(&ctx->encoder, iface, &enccfg, flags);
683 if (res != AOM_CODEC_OK) {
684 log_encoder_error(avctx, "Failed to initialize encoder");
685 return AVERROR(EINVAL);
686 }
687
688 // codec control failures are currently treated only as warnings
689 av_log(avctx, AV_LOG_DEBUG, "aom_codec_control\n");
690 codecctl_int(avctx, AOME_SET_CPUUSED, ctx->cpu_used);
691 if (ctx->auto_alt_ref >= 0)
692 codecctl_int(avctx, AOME_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
693 if (ctx->arnr_max_frames >= 0)
694 codecctl_int(avctx, AOME_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
695 if (ctx->arnr_strength >= 0)
696 codecctl_int(avctx, AOME_SET_ARNR_STRENGTH, ctx->arnr_strength);
697 if (ctx->enable_cdef >= 0)
698 codecctl_int(avctx, AV1E_SET_ENABLE_CDEF, ctx->enable_cdef);
699 if (ctx->enable_restoration >= 0)
700 codecctl_int(avctx, AV1E_SET_ENABLE_RESTORATION, ctx->enable_restoration);
701
702 codecctl_int(avctx, AOME_SET_STATIC_THRESHOLD, ctx->static_thresh);
703 if (ctx->crf >= 0)
704 codecctl_int(avctx, AOME_SET_CQ_LEVEL, ctx->crf);
705 if (ctx->tune >= 0)
706 codecctl_int(avctx, AOME_SET_TUNING, ctx->tune);
707
708 codecctl_int(avctx, AV1E_SET_COLOR_PRIMARIES, avctx->color_primaries);
709 codecctl_int(avctx, AV1E_SET_MATRIX_COEFFICIENTS, avctx->colorspace);
710 codecctl_int(avctx, AV1E_SET_TRANSFER_CHARACTERISTICS, avctx->color_trc);
711 if (ctx->aq_mode >= 0)
712 codecctl_int(avctx, AV1E_SET_AQ_MODE, ctx->aq_mode);
713 if (ctx->frame_parallel >= 0)
714 codecctl_int(avctx, AV1E_SET_FRAME_PARALLEL_DECODING, ctx->frame_parallel);
715 set_color_range(avctx);
716
717 codecctl_int(avctx, AV1E_SET_SUPERBLOCK_SIZE, ctx->superblock_size);
718 if (ctx->uniform_tiles) {
719 codecctl_int(avctx, AV1E_SET_TILE_COLUMNS, ctx->tile_cols_log2);
720 codecctl_int(avctx, AV1E_SET_TILE_ROWS, ctx->tile_rows_log2);
721 }
722
723 #ifdef AOM_CTRL_AV1E_SET_DENOISE_NOISE_LEVEL
724 if (ctx->denoise_noise_level >= 0)
725 codecctl_int(avctx, AV1E_SET_DENOISE_NOISE_LEVEL, ctx->denoise_noise_level);
726 #endif
727 #ifdef AOM_CTRL_AV1E_SET_DENOISE_BLOCK_SIZE
728 if (ctx->denoise_block_size >= 0)
729 codecctl_int(avctx, AV1E_SET_DENOISE_BLOCK_SIZE, ctx->denoise_block_size);
730 #endif
731 #ifdef AOM_CTRL_AV1E_SET_ENABLE_GLOBAL_MOTION
732 if (ctx->enable_global_motion >= 0)
733 codecctl_int(avctx, AV1E_SET_ENABLE_GLOBAL_MOTION, ctx->enable_global_motion);
734 #endif
735 #ifdef AOM_CTRL_AV1E_SET_MAX_REFERENCE_FRAMES
736 if (avctx->refs >= 3) {
737 codecctl_int(avctx, AV1E_SET_MAX_REFERENCE_FRAMES, avctx->refs);
738 }
739 #endif
740 #ifdef AOM_CTRL_AV1E_SET_ROW_MT
741 if (ctx->row_mt >= 0)
742 codecctl_int(avctx, AV1E_SET_ROW_MT, ctx->row_mt);
743 #endif
744 #ifdef AOM_CTRL_AV1E_SET_ENABLE_INTRABC
745 if (ctx->enable_intrabc >= 0)
746 codecctl_int(avctx, AV1E_SET_ENABLE_INTRABC, ctx->enable_intrabc);
747 #endif
748
749 // provide dummy value to initialize wrapper, values will be updated each _encode()
750 aom_img_wrap(&ctx->rawimg, img_fmt, avctx->width, avctx->height, 1,
751 (unsigned char*)1);
752
753 if (codec_caps & AOM_CODEC_CAP_HIGHBITDEPTH)
754 ctx->rawimg.bit_depth = enccfg.g_bit_depth;
755
756 cpb_props = ff_add_cpb_side_data(avctx);
757 if (!cpb_props)
758 return AVERROR(ENOMEM);
759
760 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
761 const AVBitStreamFilter *filter = av_bsf_get_by_name("extract_extradata");
762 int ret;
763
764 if (!filter) {
765 av_log(avctx, AV_LOG_ERROR, "extract_extradata bitstream filter "
766 "not found. This is a bug, please report it.\n");
767 return AVERROR_BUG;
768 }
769 ret = av_bsf_alloc(filter, &ctx->bsf);
770 if (ret < 0)
771 return ret;
772
773 ret = avcodec_parameters_from_context(ctx->bsf->par_in, avctx);
774 if (ret < 0)
775 return ret;
776
777 ret = av_bsf_init(ctx->bsf);
778 if (ret < 0)
779 return ret;
780 }
781
782 if (enccfg.rc_end_usage == AOM_CBR ||
783 enccfg.g_pass != AOM_RC_ONE_PASS) {
784 cpb_props->max_bitrate = avctx->rc_max_rate;
785 cpb_props->min_bitrate = avctx->rc_min_rate;
786 cpb_props->avg_bitrate = avctx->bit_rate;
787 }
788 cpb_props->buffer_size = avctx->rc_buffer_size;
789
790 return 0;
791 }
792
cx_pktcpy(AOMContext * ctx,struct FrameListData * dst,const struct aom_codec_cx_pkt * src)793 static inline void cx_pktcpy(AOMContext *ctx,
794 struct FrameListData *dst,
795 const struct aom_codec_cx_pkt *src)
796 {
797 dst->pts = src->data.frame.pts;
798 dst->duration = src->data.frame.duration;
799 dst->flags = src->data.frame.flags;
800 dst->sz = src->data.frame.sz;
801 dst->buf = src->data.frame.buf;
802 #ifdef AOM_FRAME_IS_INTRAONLY
803 dst->have_sse = 0;
804 dst->frame_number = ++ctx->frame_number;
805 dst->have_sse = ctx->have_sse;
806 if (ctx->have_sse) {
807 /* associate last-seen SSE to the frame. */
808 /* Transfers ownership from ctx to dst. */
809 memcpy(dst->sse, ctx->sse, sizeof(dst->sse));
810 ctx->have_sse = 0;
811 }
812 #endif
813 }
814
815 /**
816 * Store coded frame information in format suitable for return from encode2().
817 *
818 * Write information from @a cx_frame to @a pkt
819 * @return packet data size on success
820 * @return a negative AVERROR on error
821 */
storeframe(AVCodecContext * avctx,struct FrameListData * cx_frame,AVPacket * pkt)822 static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
823 AVPacket *pkt)
824 {
825 AOMContext *ctx = avctx->priv_data;
826 int av_unused pict_type;
827 int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz, 0);
828 if (ret < 0) {
829 av_log(avctx, AV_LOG_ERROR,
830 "Error getting output packet of size %"SIZE_SPECIFIER".\n", cx_frame->sz);
831 return ret;
832 }
833 memcpy(pkt->data, cx_frame->buf, pkt->size);
834 pkt->pts = pkt->dts = cx_frame->pts;
835
836 if (!!(cx_frame->flags & AOM_FRAME_IS_KEY)) {
837 pkt->flags |= AV_PKT_FLAG_KEY;
838 #ifdef AOM_FRAME_IS_INTRAONLY
839 pict_type = AV_PICTURE_TYPE_I;
840 } else if (cx_frame->flags & AOM_FRAME_IS_INTRAONLY) {
841 pict_type = AV_PICTURE_TYPE_I;
842 } else {
843 pict_type = AV_PICTURE_TYPE_P;
844 }
845
846 ff_side_data_set_encoder_stats(pkt, 0, cx_frame->sse + 1,
847 cx_frame->have_sse ? 3 : 0, pict_type);
848
849 if (cx_frame->have_sse) {
850 int i;
851 for (i = 0; i < 3; ++i) {
852 avctx->error[i] += cx_frame->sse[i + 1];
853 }
854 cx_frame->have_sse = 0;
855 #endif
856 }
857
858 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
859 ret = av_bsf_send_packet(ctx->bsf, pkt);
860 if (ret < 0) {
861 av_log(avctx, AV_LOG_ERROR, "extract_extradata filter "
862 "failed to send input packet\n");
863 return ret;
864 }
865 ret = av_bsf_receive_packet(ctx->bsf, pkt);
866
867 if (ret < 0) {
868 av_log(avctx, AV_LOG_ERROR, "extract_extradata filter "
869 "failed to receive output packet\n");
870 return ret;
871 }
872 }
873 return pkt->size;
874 }
875
876 /**
877 * Queue multiple output frames from the encoder, returning the front-most.
878 * In cases where aom_codec_get_cx_data() returns more than 1 frame append
879 * the frame queue. Return the head frame if available.
880 * @return Stored frame size
881 * @return AVERROR(EINVAL) on output size error
882 * @return AVERROR(ENOMEM) on coded frame queue data allocation error
883 */
queue_frames(AVCodecContext * avctx,AVPacket * pkt_out)884 static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out)
885 {
886 AOMContext *ctx = avctx->priv_data;
887 const struct aom_codec_cx_pkt *pkt;
888 const void *iter = NULL;
889 int size = 0;
890
891 if (ctx->coded_frame_list) {
892 struct FrameListData *cx_frame = ctx->coded_frame_list;
893 /* return the leading frame if we've already begun queueing */
894 size = storeframe(avctx, cx_frame, pkt_out);
895 if (size < 0)
896 return size;
897 ctx->coded_frame_list = cx_frame->next;
898 free_coded_frame(cx_frame);
899 }
900
901 /* consume all available output from the encoder before returning. buffers
902 * are only good through the next aom_codec call */
903 while ((pkt = aom_codec_get_cx_data(&ctx->encoder, &iter))) {
904 switch (pkt->kind) {
905 case AOM_CODEC_CX_FRAME_PKT:
906 if (!size) {
907 struct FrameListData cx_frame;
908
909 /* avoid storing the frame when the list is empty and we haven't yet
910 * provided a frame for output */
911 av_assert0(!ctx->coded_frame_list);
912 cx_pktcpy(ctx, &cx_frame, pkt);
913 size = storeframe(avctx, &cx_frame, pkt_out);
914 if (size < 0)
915 return size;
916 } else {
917 struct FrameListData *cx_frame =
918 av_malloc(sizeof(struct FrameListData));
919
920 if (!cx_frame) {
921 av_log(avctx, AV_LOG_ERROR,
922 "Frame queue element alloc failed\n");
923 return AVERROR(ENOMEM);
924 }
925 cx_pktcpy(ctx, cx_frame, pkt);
926 cx_frame->buf = av_malloc(cx_frame->sz);
927
928 if (!cx_frame->buf) {
929 av_log(avctx, AV_LOG_ERROR,
930 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
931 cx_frame->sz);
932 av_freep(&cx_frame);
933 return AVERROR(ENOMEM);
934 }
935 memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
936 coded_frame_add(&ctx->coded_frame_list, cx_frame);
937 }
938 break;
939 case AOM_CODEC_STATS_PKT:
940 {
941 struct aom_fixed_buf *stats = &ctx->twopass_stats;
942 int err;
943 if ((err = av_reallocp(&stats->buf,
944 stats->sz +
945 pkt->data.twopass_stats.sz)) < 0) {
946 stats->sz = 0;
947 av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
948 return err;
949 }
950 memcpy((uint8_t *)stats->buf + stats->sz,
951 pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
952 stats->sz += pkt->data.twopass_stats.sz;
953 break;
954 }
955 #ifdef AOM_FRAME_IS_INTRAONLY
956 case AOM_CODEC_PSNR_PKT:
957 {
958 av_assert0(!ctx->have_sse);
959 ctx->sse[0] = pkt->data.psnr.sse[0];
960 ctx->sse[1] = pkt->data.psnr.sse[1];
961 ctx->sse[2] = pkt->data.psnr.sse[2];
962 ctx->sse[3] = pkt->data.psnr.sse[3];
963 ctx->have_sse = 1;
964 break;
965 }
966 #endif
967 case AOM_CODEC_CUSTOM_PKT:
968 // ignore unsupported/unrecognized packet types
969 break;
970 }
971 }
972
973 return size;
974 }
975
aom_encode(AVCodecContext * avctx,AVPacket * pkt,const AVFrame * frame,int * got_packet)976 static int aom_encode(AVCodecContext *avctx, AVPacket *pkt,
977 const AVFrame *frame, int *got_packet)
978 {
979 AOMContext *ctx = avctx->priv_data;
980 struct aom_image *rawimg = NULL;
981 int64_t timestamp = 0;
982 int res, coded_size;
983 aom_enc_frame_flags_t flags = 0;
984
985 if (frame) {
986 rawimg = &ctx->rawimg;
987 rawimg->planes[AOM_PLANE_Y] = frame->data[0];
988 rawimg->planes[AOM_PLANE_U] = frame->data[1];
989 rawimg->planes[AOM_PLANE_V] = frame->data[2];
990 rawimg->stride[AOM_PLANE_Y] = frame->linesize[0];
991 rawimg->stride[AOM_PLANE_U] = frame->linesize[1];
992 rawimg->stride[AOM_PLANE_V] = frame->linesize[2];
993 timestamp = frame->pts;
994 switch (frame->color_range) {
995 case AVCOL_RANGE_MPEG:
996 rawimg->range = AOM_CR_STUDIO_RANGE;
997 break;
998 case AVCOL_RANGE_JPEG:
999 rawimg->range = AOM_CR_FULL_RANGE;
1000 break;
1001 }
1002
1003 if (frame->pict_type == AV_PICTURE_TYPE_I)
1004 flags |= AOM_EFLAG_FORCE_KF;
1005 }
1006
1007 res = aom_codec_encode(&ctx->encoder, rawimg, timestamp,
1008 avctx->ticks_per_frame, flags);
1009 if (res != AOM_CODEC_OK) {
1010 log_encoder_error(avctx, "Error encoding frame");
1011 return AVERROR_INVALIDDATA;
1012 }
1013 coded_size = queue_frames(avctx, pkt);
1014
1015 if (!frame && avctx->flags & AV_CODEC_FLAG_PASS1) {
1016 size_t b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
1017
1018 avctx->stats_out = av_malloc(b64_size);
1019 if (!avctx->stats_out) {
1020 av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
1021 b64_size);
1022 return AVERROR(ENOMEM);
1023 }
1024 av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
1025 ctx->twopass_stats.sz);
1026 }
1027
1028 *got_packet = !!coded_size;
1029 return 0;
1030 }
1031
1032 static const enum AVPixelFormat av1_pix_fmts[] = {
1033 AV_PIX_FMT_YUV420P,
1034 AV_PIX_FMT_YUV422P,
1035 AV_PIX_FMT_YUV444P,
1036 AV_PIX_FMT_NONE
1037 };
1038
1039 static const enum AVPixelFormat av1_pix_fmts_highbd[] = {
1040 AV_PIX_FMT_YUV420P,
1041 AV_PIX_FMT_YUV422P,
1042 AV_PIX_FMT_YUV444P,
1043 AV_PIX_FMT_YUV420P10,
1044 AV_PIX_FMT_YUV422P10,
1045 AV_PIX_FMT_YUV444P10,
1046 AV_PIX_FMT_YUV420P12,
1047 AV_PIX_FMT_YUV422P12,
1048 AV_PIX_FMT_YUV444P12,
1049 AV_PIX_FMT_NONE
1050 };
1051
av1_init_static(AVCodec * codec)1052 static av_cold void av1_init_static(AVCodec *codec)
1053 {
1054 aom_codec_caps_t codec_caps = aom_codec_get_caps(aom_codec_av1_cx());
1055 if (codec_caps & AOM_CODEC_CAP_HIGHBITDEPTH)
1056 codec->pix_fmts = av1_pix_fmts_highbd;
1057 else
1058 codec->pix_fmts = av1_pix_fmts;
1059
1060 if (aom_codec_version_major() < 2)
1061 codec->capabilities |= AV_CODEC_CAP_EXPERIMENTAL;
1062 }
1063
av1_init(AVCodecContext * avctx)1064 static av_cold int av1_init(AVCodecContext *avctx)
1065 {
1066 return aom_init(avctx, aom_codec_av1_cx());
1067 }
1068
1069 #define OFFSET(x) offsetof(AOMContext, x)
1070 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
1071 static const AVOption options[] = {
1072 { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 8, VE},
1073 { "auto-alt-ref", "Enable use of alternate reference "
1074 "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, VE},
1075 { "lag-in-frames", "Number of frames to look ahead at for "
1076 "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
1077 { "arnr-max-frames", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
1078 { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
1079 { "aq-mode", "adaptive quantization mode", OFFSET(aq_mode), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 4, VE, "aq_mode"},
1080 { "none", "Aq not used", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, 0, 0, VE, "aq_mode"},
1081 { "variance", "Variance based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "aq_mode"},
1082 { "complexity", "Complexity based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "aq_mode"},
1083 { "cyclic", "Cyclic Refresh Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "aq_mode"},
1084 { "error-resilience", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"},
1085 { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = AOM_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"},
1086 { "crf", "Select the quality for constant quality mode", offsetof(AOMContext, crf), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 63, VE },
1087 { "static-thresh", "A change threshold on blocks below which they will be skipped by the encoder", OFFSET(static_thresh), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
1088 { "drop-threshold", "Frame drop threshold", offsetof(AOMContext, drop_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, VE },
1089 { "denoise-noise-level", "Amount of noise to be removed", OFFSET(denoise_noise_level), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
1090 { "denoise-block-size", "Denoise block size ", OFFSET(denoise_block_size), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
1091 { "undershoot-pct", "Datarate undershoot (min) target (%)", OFFSET(rc_undershoot_pct), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 100, VE},
1092 { "overshoot-pct", "Datarate overshoot (max) target (%)", OFFSET(rc_overshoot_pct), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1000, VE},
1093 { "minsection-pct", "GOP min bitrate (% of target)", OFFSET(minsection_pct), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 100, VE},
1094 { "maxsection-pct", "GOP max bitrate (% of target)", OFFSET(maxsection_pct), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 5000, VE},
1095 { "frame-parallel", "Enable frame parallel decodability features", OFFSET(frame_parallel), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
1096 { "tiles", "Tile columns x rows", OFFSET(tile_cols), AV_OPT_TYPE_IMAGE_SIZE, { .str = NULL }, 0, 0, VE },
1097 { "tile-columns", "Log2 of number of tile columns to use", OFFSET(tile_cols_log2), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
1098 { "tile-rows", "Log2 of number of tile rows to use", OFFSET(tile_rows_log2), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
1099 { "row-mt", "Enable row based multi-threading", OFFSET(row_mt), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
1100 { "enable-cdef", "Enable CDEF filtering", OFFSET(enable_cdef), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
1101 { "enable-global-motion", "Enable global motion", OFFSET(enable_global_motion), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
1102 { "enable-intrabc", "Enable intra block copy prediction mode", OFFSET(enable_intrabc), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
1103 { "enable-restoration", "Enable Loop Restoration filtering", OFFSET(enable_restoration), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
1104 { "usage", "Quality and compression efficiency vs speed trade-off", OFFSET(usage), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, VE, "usage"},
1105 { "good", "Good quality", 0, AV_OPT_TYPE_CONST, {.i64 = 0 /* AOM_USAGE_GOOD_QUALITY */}, 0, 0, VE, "usage"},
1106 { "realtime", "Realtime encoding", 0, AV_OPT_TYPE_CONST, {.i64 = 1 /* AOM_USAGE_REALTIME */}, 0, 0, VE, "usage"},
1107 { "tune", "The metric that the encoder tunes for. Automatically chosen by the encoder by default", OFFSET(tune), AV_OPT_TYPE_INT, {.i64 = -1}, -1, AOM_TUNE_SSIM, VE, "tune"},
1108 { "psnr", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AOM_TUNE_PSNR}, 0, 0, VE, "tune"},
1109 { "ssim", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AOM_TUNE_SSIM}, 0, 0, VE, "tune"},
1110 { NULL },
1111 };
1112
1113 static const AVCodecDefault defaults[] = {
1114 { "b", "0" },
1115 { "qmin", "-1" },
1116 { "qmax", "-1" },
1117 { "g", "-1" },
1118 { "keyint_min", "-1" },
1119 { NULL },
1120 };
1121
1122 static const AVClass class_aom = {
1123 .class_name = "libaom-av1 encoder",
1124 .item_name = av_default_item_name,
1125 .option = options,
1126 .version = LIBAVUTIL_VERSION_INT,
1127 };
1128
1129 AVCodec ff_libaom_av1_encoder = {
1130 .name = "libaom-av1",
1131 .long_name = NULL_IF_CONFIG_SMALL("libaom AV1"),
1132 .type = AVMEDIA_TYPE_VIDEO,
1133 .id = AV_CODEC_ID_AV1,
1134 .priv_data_size = sizeof(AOMContext),
1135 .init = av1_init,
1136 .encode2 = aom_encode,
1137 .close = aom_free,
1138 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
1139 .profiles = NULL_IF_CONFIG_SMALL(ff_av1_profiles),
1140 .priv_class = &class_aom,
1141 .defaults = defaults,
1142 .init_static_data = av1_init_static,
1143 .wrapper_name = "libaom",
1144 };
1145