1 /*
2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10
11 #include "./vpxenc.h"
12 #include "./vpx_config.h"
13
14 #include <assert.h>
15 #include <limits.h>
16 #include <math.h>
17 #include <stdarg.h>
18 #include <stdio.h>
19 #include <stdlib.h>
20 #include <string.h>
21
22 #if CONFIG_LIBYUV
23 #include "third_party/libyuv/include/libyuv/scale.h"
24 #endif
25
26 #include "vpx/vpx_encoder.h"
27 #if CONFIG_DECODERS
28 #include "vpx/vpx_decoder.h"
29 #endif
30
31 #include "./args.h"
32 #include "./ivfenc.h"
33 #include "./tools_common.h"
34
35 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
36 #include "vpx/vp8cx.h"
37 #endif
38 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
39 #include "vpx/vp8dx.h"
40 #endif
41
42 #include "vpx/vpx_integer.h"
43 #include "vpx_ports/mem_ops.h"
44 #include "vpx_ports/vpx_timer.h"
45 #include "./rate_hist.h"
46 #include "./vpxstats.h"
47 #include "./warnings.h"
48 #if CONFIG_WEBM_IO
49 #include "./webmenc.h"
50 #endif
51 #include "./y4minput.h"
52
wrap_fwrite(const void * ptr,size_t size,size_t nmemb,FILE * stream)53 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
54 FILE *stream) {
55 return fwrite(ptr, size, nmemb, stream);
56 }
57 #define fwrite wrap_fwrite
58
59 static const char *exec_name;
60
warn_or_exit_on_errorv(vpx_codec_ctx_t * ctx,int fatal,const char * s,va_list ap)61 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
62 const char *s, va_list ap) {
63 if (ctx->err) {
64 const char *detail = vpx_codec_error_detail(ctx);
65
66 vfprintf(stderr, s, ap);
67 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
68
69 if (detail) fprintf(stderr, " %s\n", detail);
70
71 if (fatal) exit(EXIT_FAILURE);
72 }
73 }
74
ctx_exit_on_error(vpx_codec_ctx_t * ctx,const char * s,...)75 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
76 va_list ap;
77
78 va_start(ap, s);
79 warn_or_exit_on_errorv(ctx, 1, s, ap);
80 va_end(ap);
81 }
82
warn_or_exit_on_error(vpx_codec_ctx_t * ctx,int fatal,const char * s,...)83 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
84 const char *s, ...) {
85 va_list ap;
86
87 va_start(ap, s);
88 warn_or_exit_on_errorv(ctx, fatal, s, ap);
89 va_end(ap);
90 }
91
92 static const arg_def_t help =
93 ARG_DEF(NULL, "help", 0, "Show usage options and exit");
94 static const arg_def_t debugmode =
95 ARG_DEF("D", "debug", 0, "Debug mode (makes output deterministic)");
96 static const arg_def_t outputfile =
97 ARG_DEF("o", "output", 1, "Output filename");
98 static const arg_def_t use_yv12 =
99 ARG_DEF(NULL, "yv12", 0, "Input file is YV12 ");
100 static const arg_def_t use_i420 =
101 ARG_DEF(NULL, "i420", 0, "Input file is I420 (default)");
102 static const arg_def_t use_i422 =
103 ARG_DEF(NULL, "i422", 0, "Input file is I422");
104 static const arg_def_t use_i444 =
105 ARG_DEF(NULL, "i444", 0, "Input file is I444");
106 static const arg_def_t use_i440 =
107 ARG_DEF(NULL, "i440", 0, "Input file is I440");
108 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, "Codec to use");
109 static const arg_def_t passes =
110 ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
111 static const arg_def_t pass_arg =
112 ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
113 static const arg_def_t fpf_name =
114 ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");
115 #if CONFIG_FP_MB_STATS
116 static const arg_def_t fpmbf_name =
117 ARG_DEF(NULL, "fpmbf", 1, "First pass block statistics file name");
118 #endif
119 static const arg_def_t limit =
120 ARG_DEF(NULL, "limit", 1, "Stop encoding after n input frames");
121 static const arg_def_t skip =
122 ARG_DEF(NULL, "skip", 1, "Skip the first n input frames");
123 static const arg_def_t deadline =
124 ARG_DEF("d", "deadline", 1, "Deadline per frame (usec)");
125 static const arg_def_t best_dl =
126 ARG_DEF(NULL, "best", 0, "Use Best Quality Deadline");
127 static const arg_def_t good_dl =
128 ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
129 static const arg_def_t rt_dl =
130 ARG_DEF(NULL, "rt", 0, "Use Realtime Quality Deadline");
131 static const arg_def_t quietarg =
132 ARG_DEF("q", "quiet", 0, "Do not print encode progress");
133 static const arg_def_t verbosearg =
134 ARG_DEF("v", "verbose", 0, "Show encoder parameters");
135 static const arg_def_t psnrarg =
136 ARG_DEF(NULL, "psnr", 0, "Show PSNR in status line");
137
138 static const struct arg_enum_list test_decode_enum[] = {
139 { "off", TEST_DECODE_OFF },
140 { "fatal", TEST_DECODE_FATAL },
141 { "warn", TEST_DECODE_WARN },
142 { NULL, 0 }
143 };
144 static const arg_def_t recontest = ARG_DEF_ENUM(
145 NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
146 static const arg_def_t framerate =
147 ARG_DEF(NULL, "fps", 1, "Stream frame rate (rate/scale)");
148 static const arg_def_t use_webm =
149 ARG_DEF(NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
150 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, "Output IVF");
151 static const arg_def_t out_part =
152 ARG_DEF("P", "output-partitions", 0,
153 "Makes encoder output partitions. Requires IVF output!");
154 static const arg_def_t q_hist_n =
155 ARG_DEF(NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
156 static const arg_def_t rate_hist_n =
157 ARG_DEF(NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
158 static const arg_def_t disable_warnings =
159 ARG_DEF(NULL, "disable-warnings", 0,
160 "Disable warnings about potentially incorrect encode settings.");
161 static const arg_def_t disable_warning_prompt =
162 ARG_DEF("y", "disable-warning-prompt", 0,
163 "Display warnings, but do not prompt user to continue.");
164
165 #if CONFIG_VP9_HIGHBITDEPTH
166 static const arg_def_t test16bitinternalarg = ARG_DEF(
167 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
168 #endif
169
170 static const arg_def_t *main_args[] = { &help,
171 &debugmode,
172 &outputfile,
173 &codecarg,
174 &passes,
175 &pass_arg,
176 &fpf_name,
177 &limit,
178 &skip,
179 &deadline,
180 &best_dl,
181 &good_dl,
182 &rt_dl,
183 &quietarg,
184 &verbosearg,
185 &psnrarg,
186 &use_webm,
187 &use_ivf,
188 &out_part,
189 &q_hist_n,
190 &rate_hist_n,
191 &disable_warnings,
192 &disable_warning_prompt,
193 &recontest,
194 NULL };
195
196 static const arg_def_t usage =
197 ARG_DEF("u", "usage", 1, "Usage profile number to use");
198 static const arg_def_t threads =
199 ARG_DEF("t", "threads", 1, "Max number of threads to use");
200 static const arg_def_t profile =
201 ARG_DEF(NULL, "profile", 1, "Bitstream profile number to use");
202 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
203 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
204 #if CONFIG_WEBM_IO
205 static const struct arg_enum_list stereo_mode_enum[] = {
206 { "mono", STEREO_FORMAT_MONO },
207 { "left-right", STEREO_FORMAT_LEFT_RIGHT },
208 { "bottom-top", STEREO_FORMAT_BOTTOM_TOP },
209 { "top-bottom", STEREO_FORMAT_TOP_BOTTOM },
210 { "right-left", STEREO_FORMAT_RIGHT_LEFT },
211 { NULL, 0 }
212 };
213 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
214 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
215 #endif
216 static const arg_def_t timebase = ARG_DEF(
217 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
218 static const arg_def_t error_resilient =
219 ARG_DEF(NULL, "error-resilient", 1, "Enable error resiliency features");
220 static const arg_def_t lag_in_frames =
221 ARG_DEF(NULL, "lag-in-frames", 1, "Max number of frames to lag");
222
223 static const arg_def_t *global_args[] = { &use_yv12,
224 &use_i420,
225 &use_i422,
226 &use_i444,
227 &use_i440,
228 &usage,
229 &threads,
230 &profile,
231 &width,
232 &height,
233 #if CONFIG_WEBM_IO
234 &stereo_mode,
235 #endif
236 &timebase,
237 &framerate,
238 &error_resilient,
239 #if CONFIG_VP9_HIGHBITDEPTH
240 &test16bitinternalarg,
241 #endif
242 &lag_in_frames,
243 NULL };
244
245 static const arg_def_t dropframe_thresh =
246 ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
247 static const arg_def_t resize_allowed =
248 ARG_DEF(NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
249 static const arg_def_t resize_width =
250 ARG_DEF(NULL, "resize-width", 1, "Width of encoded frame");
251 static const arg_def_t resize_height =
252 ARG_DEF(NULL, "resize-height", 1, "Height of encoded frame");
253 static const arg_def_t resize_up_thresh =
254 ARG_DEF(NULL, "resize-up", 1, "Upscale threshold (buf %)");
255 static const arg_def_t resize_down_thresh =
256 ARG_DEF(NULL, "resize-down", 1, "Downscale threshold (buf %)");
257 static const struct arg_enum_list end_usage_enum[] = { { "vbr", VPX_VBR },
258 { "cbr", VPX_CBR },
259 { "cq", VPX_CQ },
260 { "q", VPX_Q },
261 { NULL, 0 } };
262 static const arg_def_t end_usage =
263 ARG_DEF_ENUM(NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
264 static const arg_def_t target_bitrate =
265 ARG_DEF(NULL, "target-bitrate", 1, "Bitrate (kbps)");
266 static const arg_def_t min_quantizer =
267 ARG_DEF(NULL, "min-q", 1, "Minimum (best) quantizer");
268 static const arg_def_t max_quantizer =
269 ARG_DEF(NULL, "max-q", 1, "Maximum (worst) quantizer");
270 static const arg_def_t undershoot_pct =
271 ARG_DEF(NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
272 static const arg_def_t overshoot_pct =
273 ARG_DEF(NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
274 static const arg_def_t buf_sz =
275 ARG_DEF(NULL, "buf-sz", 1, "Client buffer size (ms)");
276 static const arg_def_t buf_initial_sz =
277 ARG_DEF(NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
278 static const arg_def_t buf_optimal_sz =
279 ARG_DEF(NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
280 static const arg_def_t *rc_args[] = {
281 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
282 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
283 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct,
284 &buf_sz, &buf_initial_sz, &buf_optimal_sz, NULL
285 };
286
287 static const arg_def_t bias_pct =
288 ARG_DEF(NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
289 static const arg_def_t minsection_pct =
290 ARG_DEF(NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
291 static const arg_def_t maxsection_pct =
292 ARG_DEF(NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
293 static const arg_def_t corpus_complexity =
294 ARG_DEF(NULL, "corpus-complexity", 1, "corpus vbr complexity midpoint");
295 static const arg_def_t *rc_twopass_args[] = { &bias_pct, &minsection_pct,
296 &maxsection_pct,
297 &corpus_complexity, NULL };
298
299 static const arg_def_t kf_min_dist =
300 ARG_DEF(NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
301 static const arg_def_t kf_max_dist =
302 ARG_DEF(NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
303 static const arg_def_t kf_disabled =
304 ARG_DEF(NULL, "disable-kf", 0, "Disable keyframe placement");
305 static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
306 NULL };
307
308 static const arg_def_t noise_sens =
309 ARG_DEF(NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
310 static const arg_def_t sharpness =
311 ARG_DEF(NULL, "sharpness", 1,
312 "Increase sharpness at the expense of lower PSNR. (0..7)");
313 static const arg_def_t static_thresh =
314 ARG_DEF(NULL, "static-thresh", 1, "Motion detection threshold");
315 static const arg_def_t arnr_maxframes =
316 ARG_DEF(NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
317 static const arg_def_t arnr_strength =
318 ARG_DEF(NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
319 static const arg_def_t arnr_type =
320 ARG_DEF(NULL, "arnr-type", 1, "AltRef filter type (1..3)");
321 static const struct arg_enum_list tuning_enum[] = { { "psnr", VP8_TUNE_PSNR },
322 { "ssim", VP8_TUNE_SSIM },
323 { NULL, 0 } };
324 static const arg_def_t tune_ssim =
325 ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
326 static const arg_def_t cq_level =
327 ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
328 static const arg_def_t max_intra_rate_pct =
329 ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
330 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
331 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
332
333 #if CONFIG_VP8_ENCODER
334 static const arg_def_t cpu_used_vp8 =
335 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
336 static const arg_def_t auto_altref_vp8 = ARG_DEF(
337 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames. (0..1)");
338 static const arg_def_t token_parts =
339 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
340 static const arg_def_t screen_content_mode =
341 ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
342 static const arg_def_t *vp8_args[] = { &cpu_used_vp8,
343 &auto_altref_vp8,
344 &noise_sens,
345 &sharpness,
346 &static_thresh,
347 &token_parts,
348 &arnr_maxframes,
349 &arnr_strength,
350 &arnr_type,
351 &tune_ssim,
352 &cq_level,
353 &max_intra_rate_pct,
354 &gf_cbr_boost_pct,
355 &screen_content_mode,
356 NULL };
357 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
358 VP8E_SET_ENABLEAUTOALTREF,
359 VP8E_SET_NOISE_SENSITIVITY,
360 VP8E_SET_SHARPNESS,
361 VP8E_SET_STATIC_THRESHOLD,
362 VP8E_SET_TOKEN_PARTITIONS,
363 VP8E_SET_ARNR_MAXFRAMES,
364 VP8E_SET_ARNR_STRENGTH,
365 VP8E_SET_ARNR_TYPE,
366 VP8E_SET_TUNING,
367 VP8E_SET_CQ_LEVEL,
368 VP8E_SET_MAX_INTRA_BITRATE_PCT,
369 VP8E_SET_GF_CBR_BOOST_PCT,
370 VP8E_SET_SCREEN_CONTENT_MODE,
371 0 };
372 #endif
373
374 #if CONFIG_VP9_ENCODER
375 static const arg_def_t cpu_used_vp9 =
376 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-9..9)");
377 static const arg_def_t auto_altref_vp9 = ARG_DEF(
378 NULL, "auto-alt-ref", 1,
379 "Enable automatic alt reference frames, 2+ enables multi-layer. (0..6)");
380 static const arg_def_t tile_cols =
381 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
382 static const arg_def_t tile_rows =
383 ARG_DEF(NULL, "tile-rows", 1,
384 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
385
386 static const arg_def_t enable_tpl_model =
387 ARG_DEF(NULL, "enable-tpl", 1, "Enable temporal dependency model");
388
389 static const arg_def_t lossless =
390 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
391 static const arg_def_t frame_parallel_decoding = ARG_DEF(
392 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
393 static const arg_def_t aq_mode = ARG_DEF(
394 NULL, "aq-mode", 1,
395 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
396 "3: cyclic refresh, 4: equator360)");
397 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
398 "Special adaptive quantization for "
399 "the alternate reference frames.");
400 static const arg_def_t frame_periodic_boost =
401 ARG_DEF(NULL, "frame-boost", 1,
402 "Enable frame periodic boost (0: off (default), 1: on)");
403 static const arg_def_t max_inter_rate_pct =
404 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
405 static const arg_def_t min_gf_interval = ARG_DEF(
406 NULL, "min-gf-interval", 1,
407 "min gf/arf frame interval (default 0, indicating in-built behavior)");
408 static const arg_def_t max_gf_interval = ARG_DEF(
409 NULL, "max-gf-interval", 1,
410 "max gf/arf frame interval (default 0, indicating in-built behavior)");
411
412 static const struct arg_enum_list color_space_enum[] = {
413 { "unknown", VPX_CS_UNKNOWN },
414 { "bt601", VPX_CS_BT_601 },
415 { "bt709", VPX_CS_BT_709 },
416 { "smpte170", VPX_CS_SMPTE_170 },
417 { "smpte240", VPX_CS_SMPTE_240 },
418 { "bt2020", VPX_CS_BT_2020 },
419 { "reserved", VPX_CS_RESERVED },
420 { "sRGB", VPX_CS_SRGB },
421 { NULL, 0 }
422 };
423
424 static const arg_def_t input_color_space =
425 ARG_DEF_ENUM(NULL, "color-space", 1,
426 "The color space of input content:", color_space_enum);
427
428 #if CONFIG_VP9_HIGHBITDEPTH
429 static const struct arg_enum_list bitdepth_enum[] = {
430 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
431 };
432
433 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
434 "b", "bit-depth", 1,
435 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
436 bitdepth_enum);
437 static const arg_def_t inbitdeptharg =
438 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
439 #endif
440
441 static const struct arg_enum_list tune_content_enum[] = {
442 { "default", VP9E_CONTENT_DEFAULT },
443 { "screen", VP9E_CONTENT_SCREEN },
444 { "film", VP9E_CONTENT_FILM },
445 { NULL, 0 }
446 };
447
448 static const arg_def_t tune_content = ARG_DEF_ENUM(
449 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
450
451 static const arg_def_t target_level = ARG_DEF(
452 NULL, "target-level", 1,
453 "Target level\n"
454 " 255: off (default)\n"
455 " 0: only keep level stats\n"
456 " 1: adaptively set alt-ref "
457 "distance and column tile limit based on picture size, and keep"
458 " level stats\n"
459 " 10: level 1.0 11: level 1.1 "
460 "... 62: level 6.2");
461
462 static const arg_def_t row_mt =
463 ARG_DEF(NULL, "row-mt", 1,
464 "Enable row based non-deterministic multi-threading in VP9");
465 #endif
466
467 #if CONFIG_VP9_ENCODER
468 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
469 &auto_altref_vp9,
470 &sharpness,
471 &static_thresh,
472 &tile_cols,
473 &tile_rows,
474 &enable_tpl_model,
475 &arnr_maxframes,
476 &arnr_strength,
477 &arnr_type,
478 &tune_ssim,
479 &cq_level,
480 &max_intra_rate_pct,
481 &max_inter_rate_pct,
482 &gf_cbr_boost_pct,
483 &lossless,
484 &frame_parallel_decoding,
485 &aq_mode,
486 &alt_ref_aq,
487 &frame_periodic_boost,
488 &noise_sens,
489 &tune_content,
490 &input_color_space,
491 &min_gf_interval,
492 &max_gf_interval,
493 &target_level,
494 &row_mt,
495 #if CONFIG_VP9_HIGHBITDEPTH
496 &bitdeptharg,
497 &inbitdeptharg,
498 #endif // CONFIG_VP9_HIGHBITDEPTH
499 NULL };
500 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
501 VP8E_SET_ENABLEAUTOALTREF,
502 VP8E_SET_SHARPNESS,
503 VP8E_SET_STATIC_THRESHOLD,
504 VP9E_SET_TILE_COLUMNS,
505 VP9E_SET_TILE_ROWS,
506 VP9E_SET_TPL,
507 VP8E_SET_ARNR_MAXFRAMES,
508 VP8E_SET_ARNR_STRENGTH,
509 VP8E_SET_ARNR_TYPE,
510 VP8E_SET_TUNING,
511 VP8E_SET_CQ_LEVEL,
512 VP8E_SET_MAX_INTRA_BITRATE_PCT,
513 VP9E_SET_MAX_INTER_BITRATE_PCT,
514 VP9E_SET_GF_CBR_BOOST_PCT,
515 VP9E_SET_LOSSLESS,
516 VP9E_SET_FRAME_PARALLEL_DECODING,
517 VP9E_SET_AQ_MODE,
518 VP9E_SET_ALT_REF_AQ,
519 VP9E_SET_FRAME_PERIODIC_BOOST,
520 VP9E_SET_NOISE_SENSITIVITY,
521 VP9E_SET_TUNE_CONTENT,
522 VP9E_SET_COLOR_SPACE,
523 VP9E_SET_MIN_GF_INTERVAL,
524 VP9E_SET_MAX_GF_INTERVAL,
525 VP9E_SET_TARGET_LEVEL,
526 VP9E_SET_ROW_MT,
527 0 };
528 #endif
529
530 static const arg_def_t *no_args[] = { NULL };
531
show_help(FILE * fout,int shorthelp)532 static void show_help(FILE *fout, int shorthelp) {
533 int i;
534 const int num_encoder = get_vpx_encoder_count();
535
536 fprintf(fout, "Usage: %s <options> -o dst_filename src_filename \n",
537 exec_name);
538
539 if (shorthelp) {
540 fprintf(fout, "Use --help to see the full list of options.\n");
541 return;
542 }
543
544 fprintf(fout, "\nOptions:\n");
545 arg_show_usage(fout, main_args);
546 fprintf(fout, "\nEncoder Global Options:\n");
547 arg_show_usage(fout, global_args);
548 fprintf(fout, "\nRate Control Options:\n");
549 arg_show_usage(fout, rc_args);
550 fprintf(fout, "\nTwopass Rate Control Options:\n");
551 arg_show_usage(fout, rc_twopass_args);
552 fprintf(fout, "\nKeyframe Placement Options:\n");
553 arg_show_usage(fout, kf_args);
554 #if CONFIG_VP8_ENCODER
555 fprintf(fout, "\nVP8 Specific Options:\n");
556 arg_show_usage(fout, vp8_args);
557 #endif
558 #if CONFIG_VP9_ENCODER
559 fprintf(fout, "\nVP9 Specific Options:\n");
560 arg_show_usage(fout, vp9_args);
561 #endif
562 fprintf(fout,
563 "\nStream timebase (--timebase):\n"
564 " The desired precision of timestamps in the output, expressed\n"
565 " in fractional seconds. Default is 1/1000.\n");
566 fprintf(fout, "\nIncluded encoders:\n\n");
567
568 for (i = 0; i < num_encoder; ++i) {
569 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
570 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
571 fprintf(fout, " %-6s - %s %s\n", encoder->name,
572 vpx_codec_iface_name(encoder->codec_interface()), defstr);
573 }
574 fprintf(fout, "\n ");
575 fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
576 }
577
usage_exit(void)578 void usage_exit(void) {
579 show_help(stderr, 1);
580 exit(EXIT_FAILURE);
581 }
582
583 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
584 #if CONFIG_VP9_ENCODER
585 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
586 #else
587 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
588 #endif
589
590 #if !CONFIG_WEBM_IO
591 typedef int stereo_format_t;
592 struct WebmOutputContext {
593 int debug;
594 };
595 #endif
596
597 /* Per-stream configuration */
598 struct stream_config {
599 struct vpx_codec_enc_cfg cfg;
600 const char *out_fn;
601 const char *stats_fn;
602 #if CONFIG_FP_MB_STATS
603 const char *fpmb_stats_fn;
604 #endif
605 stereo_format_t stereo_fmt;
606 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
607 int arg_ctrl_cnt;
608 int write_webm;
609 #if CONFIG_VP9_HIGHBITDEPTH
610 // whether to use 16bit internal buffers
611 int use_16bit_internal;
612 #endif
613 };
614
615 struct stream_state {
616 int index;
617 struct stream_state *next;
618 struct stream_config config;
619 FILE *file;
620 struct rate_hist *rate_hist;
621 struct WebmOutputContext webm_ctx;
622 uint64_t psnr_sse_total;
623 uint64_t psnr_samples_total;
624 double psnr_totals[4];
625 int psnr_count;
626 int counts[64];
627 vpx_codec_ctx_t encoder;
628 unsigned int frames_out;
629 uint64_t cx_time;
630 size_t nbytes;
631 stats_io_t stats;
632 #if CONFIG_FP_MB_STATS
633 stats_io_t fpmb_stats;
634 #endif
635 struct vpx_image *img;
636 vpx_codec_ctx_t decoder;
637 int mismatch_seen;
638 };
639
validate_positive_rational(const char * msg,struct vpx_rational * rat)640 static void validate_positive_rational(const char *msg,
641 struct vpx_rational *rat) {
642 if (rat->den < 0) {
643 rat->num *= -1;
644 rat->den *= -1;
645 }
646
647 if (rat->num < 0) die("Error: %s must be positive\n", msg);
648
649 if (!rat->den) die("Error: %s has zero denominator\n", msg);
650 }
651
parse_global_config(struct VpxEncoderConfig * global,char ** argv)652 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
653 char **argi, **argj;
654 struct arg arg;
655 const int num_encoder = get_vpx_encoder_count();
656
657 if (num_encoder < 1) die("Error: no valid encoder available\n");
658
659 /* Initialize default parameters */
660 memset(global, 0, sizeof(*global));
661 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
662 global->passes = 0;
663 global->color_type = I420;
664 /* Assign default deadline to good quality */
665 global->deadline = VPX_DL_GOOD_QUALITY;
666
667 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
668 arg.argv_step = 1;
669
670 if (arg_match(&arg, &help, argi)) {
671 show_help(stdout, 0);
672 exit(EXIT_SUCCESS);
673 } else if (arg_match(&arg, &codecarg, argi)) {
674 global->codec = get_vpx_encoder_by_name(arg.val);
675 if (!global->codec)
676 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
677 } else if (arg_match(&arg, &passes, argi)) {
678 global->passes = arg_parse_uint(&arg);
679
680 if (global->passes < 1 || global->passes > 2)
681 die("Error: Invalid number of passes (%d)\n", global->passes);
682 } else if (arg_match(&arg, &pass_arg, argi)) {
683 global->pass = arg_parse_uint(&arg);
684
685 if (global->pass < 1 || global->pass > 2)
686 die("Error: Invalid pass selected (%d)\n", global->pass);
687 } else if (arg_match(&arg, &usage, argi))
688 global->usage = arg_parse_uint(&arg);
689 else if (arg_match(&arg, &deadline, argi))
690 global->deadline = arg_parse_uint(&arg);
691 else if (arg_match(&arg, &best_dl, argi))
692 global->deadline = VPX_DL_BEST_QUALITY;
693 else if (arg_match(&arg, &good_dl, argi))
694 global->deadline = VPX_DL_GOOD_QUALITY;
695 else if (arg_match(&arg, &rt_dl, argi))
696 global->deadline = VPX_DL_REALTIME;
697 else if (arg_match(&arg, &use_yv12, argi))
698 global->color_type = YV12;
699 else if (arg_match(&arg, &use_i420, argi))
700 global->color_type = I420;
701 else if (arg_match(&arg, &use_i422, argi))
702 global->color_type = I422;
703 else if (arg_match(&arg, &use_i444, argi))
704 global->color_type = I444;
705 else if (arg_match(&arg, &use_i440, argi))
706 global->color_type = I440;
707 else if (arg_match(&arg, &quietarg, argi))
708 global->quiet = 1;
709 else if (arg_match(&arg, &verbosearg, argi))
710 global->verbose = 1;
711 else if (arg_match(&arg, &limit, argi))
712 global->limit = arg_parse_uint(&arg);
713 else if (arg_match(&arg, &skip, argi))
714 global->skip_frames = arg_parse_uint(&arg);
715 else if (arg_match(&arg, &psnrarg, argi))
716 global->show_psnr = 1;
717 else if (arg_match(&arg, &recontest, argi))
718 global->test_decode = arg_parse_enum_or_int(&arg);
719 else if (arg_match(&arg, &framerate, argi)) {
720 global->framerate = arg_parse_rational(&arg);
721 validate_positive_rational(arg.name, &global->framerate);
722 global->have_framerate = 1;
723 } else if (arg_match(&arg, &out_part, argi))
724 global->out_part = 1;
725 else if (arg_match(&arg, &debugmode, argi))
726 global->debug = 1;
727 else if (arg_match(&arg, &q_hist_n, argi))
728 global->show_q_hist_buckets = arg_parse_uint(&arg);
729 else if (arg_match(&arg, &rate_hist_n, argi))
730 global->show_rate_hist_buckets = arg_parse_uint(&arg);
731 else if (arg_match(&arg, &disable_warnings, argi))
732 global->disable_warnings = 1;
733 else if (arg_match(&arg, &disable_warning_prompt, argi))
734 global->disable_warning_prompt = 1;
735 else
736 argj++;
737 }
738
739 if (global->pass) {
740 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
741 if (global->pass > global->passes) {
742 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
743 global->pass);
744 global->passes = global->pass;
745 }
746 }
747 /* Validate global config */
748 if (global->passes == 0) {
749 #if CONFIG_VP9_ENCODER
750 // Make default VP9 passes = 2 until there is a better quality 1-pass
751 // encoder
752 if (global->codec != NULL && global->codec->name != NULL)
753 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
754 global->deadline != VPX_DL_REALTIME)
755 ? 2
756 : 1;
757 #else
758 global->passes = 1;
759 #endif
760 }
761
762 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
763 warn("Enforcing one-pass encoding in realtime mode\n");
764 global->passes = 1;
765 }
766 }
767
new_stream(struct VpxEncoderConfig * global,struct stream_state * prev)768 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
769 struct stream_state *prev) {
770 struct stream_state *stream;
771
772 stream = calloc(1, sizeof(*stream));
773 if (stream == NULL) {
774 fatal("Failed to allocate new stream.");
775 }
776
777 if (prev) {
778 memcpy(stream, prev, sizeof(*stream));
779 stream->index++;
780 prev->next = stream;
781 } else {
782 vpx_codec_err_t res;
783
784 /* Populate encoder configuration */
785 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
786 &stream->config.cfg, global->usage);
787 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
788
789 /* Change the default timebase to a high enough value so that the
790 * encoder will always create strictly increasing timestamps.
791 */
792 stream->config.cfg.g_timebase.den = 1000;
793
794 /* Never use the library's default resolution, require it be parsed
795 * from the file or set on the command line.
796 */
797 stream->config.cfg.g_w = 0;
798 stream->config.cfg.g_h = 0;
799
800 /* Initialize remaining stream parameters */
801 stream->config.write_webm = 1;
802 #if CONFIG_WEBM_IO
803 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
804 stream->webm_ctx.last_pts_ns = -1;
805 stream->webm_ctx.writer = NULL;
806 stream->webm_ctx.segment = NULL;
807 #endif
808
809 /* Allows removal of the application version from the EBML tags */
810 stream->webm_ctx.debug = global->debug;
811
812 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
813 if (global->deadline == VPX_DL_REALTIME &&
814 stream->config.cfg.rc_end_usage == 1)
815 stream->config.cfg.g_lag_in_frames = 0;
816 }
817
818 /* Output files must be specified for each stream */
819 stream->config.out_fn = NULL;
820
821 stream->next = NULL;
822 return stream;
823 }
824
parse_stream_params(struct VpxEncoderConfig * global,struct stream_state * stream,char ** argv)825 static int parse_stream_params(struct VpxEncoderConfig *global,
826 struct stream_state *stream, char **argv) {
827 char **argi, **argj;
828 struct arg arg;
829 static const arg_def_t **ctrl_args = no_args;
830 static const int *ctrl_args_map = NULL;
831 struct stream_config *config = &stream->config;
832 int eos_mark_found = 0;
833 #if CONFIG_VP9_HIGHBITDEPTH
834 int test_16bit_internal = 0;
835 #endif
836
837 // Handle codec specific options
838 if (0) {
839 #if CONFIG_VP8_ENCODER
840 } else if (strcmp(global->codec->name, "vp8") == 0) {
841 ctrl_args = vp8_args;
842 ctrl_args_map = vp8_arg_ctrl_map;
843 #endif
844 #if CONFIG_VP9_ENCODER
845 } else if (strcmp(global->codec->name, "vp9") == 0) {
846 ctrl_args = vp9_args;
847 ctrl_args_map = vp9_arg_ctrl_map;
848 #endif
849 }
850
851 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
852 arg.argv_step = 1;
853
854 /* Once we've found an end-of-stream marker (--) we want to continue
855 * shifting arguments but not consuming them.
856 */
857 if (eos_mark_found) {
858 argj++;
859 continue;
860 } else if (!strcmp(*argj, "--")) {
861 eos_mark_found = 1;
862 continue;
863 }
864
865 if (arg_match(&arg, &outputfile, argi)) {
866 config->out_fn = arg.val;
867 } else if (arg_match(&arg, &fpf_name, argi)) {
868 config->stats_fn = arg.val;
869 #if CONFIG_FP_MB_STATS
870 } else if (arg_match(&arg, &fpmbf_name, argi)) {
871 config->fpmb_stats_fn = arg.val;
872 #endif
873 } else if (arg_match(&arg, &use_webm, argi)) {
874 #if CONFIG_WEBM_IO
875 config->write_webm = 1;
876 #else
877 die("Error: --webm specified but webm is disabled.");
878 #endif
879 } else if (arg_match(&arg, &use_ivf, argi)) {
880 config->write_webm = 0;
881 } else if (arg_match(&arg, &threads, argi)) {
882 config->cfg.g_threads = arg_parse_uint(&arg);
883 } else if (arg_match(&arg, &profile, argi)) {
884 config->cfg.g_profile = arg_parse_uint(&arg);
885 } else if (arg_match(&arg, &width, argi)) {
886 config->cfg.g_w = arg_parse_uint(&arg);
887 } else if (arg_match(&arg, &height, argi)) {
888 config->cfg.g_h = arg_parse_uint(&arg);
889 #if CONFIG_VP9_HIGHBITDEPTH
890 } else if (arg_match(&arg, &bitdeptharg, argi)) {
891 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
892 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
893 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
894 #endif
895 #if CONFIG_WEBM_IO
896 } else if (arg_match(&arg, &stereo_mode, argi)) {
897 config->stereo_fmt = arg_parse_enum_or_int(&arg);
898 #endif
899 } else if (arg_match(&arg, &timebase, argi)) {
900 config->cfg.g_timebase = arg_parse_rational(&arg);
901 validate_positive_rational(arg.name, &config->cfg.g_timebase);
902 } else if (arg_match(&arg, &error_resilient, argi)) {
903 config->cfg.g_error_resilient = arg_parse_uint(&arg);
904 } else if (arg_match(&arg, &end_usage, argi)) {
905 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
906 } else if (arg_match(&arg, &lag_in_frames, argi)) {
907 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
908 if (global->deadline == VPX_DL_REALTIME &&
909 config->cfg.rc_end_usage == VPX_CBR &&
910 config->cfg.g_lag_in_frames != 0) {
911 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
912 config->cfg.g_lag_in_frames = 0;
913 }
914 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
915 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
916 } else if (arg_match(&arg, &resize_allowed, argi)) {
917 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
918 } else if (arg_match(&arg, &resize_width, argi)) {
919 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
920 } else if (arg_match(&arg, &resize_height, argi)) {
921 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
922 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
923 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
924 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
925 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
926 } else if (arg_match(&arg, &end_usage, argi)) {
927 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
928 } else if (arg_match(&arg, &target_bitrate, argi)) {
929 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
930 } else if (arg_match(&arg, &min_quantizer, argi)) {
931 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
932 } else if (arg_match(&arg, &max_quantizer, argi)) {
933 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
934 } else if (arg_match(&arg, &undershoot_pct, argi)) {
935 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
936 } else if (arg_match(&arg, &overshoot_pct, argi)) {
937 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
938 } else if (arg_match(&arg, &buf_sz, argi)) {
939 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
940 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
941 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
942 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
943 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
944 } else if (arg_match(&arg, &bias_pct, argi)) {
945 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
946 if (global->passes < 2)
947 warn("option %s ignored in one-pass mode.\n", arg.name);
948 } else if (arg_match(&arg, &minsection_pct, argi)) {
949 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
950
951 if (global->passes < 2)
952 warn("option %s ignored in one-pass mode.\n", arg.name);
953 } else if (arg_match(&arg, &maxsection_pct, argi)) {
954 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
955
956 if (global->passes < 2)
957 warn("option %s ignored in one-pass mode.\n", arg.name);
958 } else if (arg_match(&arg, &corpus_complexity, argi)) {
959 config->cfg.rc_2pass_vbr_corpus_complexity = arg_parse_uint(&arg);
960
961 if (global->passes < 2)
962 warn("option %s ignored in one-pass mode.\n", arg.name);
963 } else if (arg_match(&arg, &kf_min_dist, argi)) {
964 config->cfg.kf_min_dist = arg_parse_uint(&arg);
965 } else if (arg_match(&arg, &kf_max_dist, argi)) {
966 config->cfg.kf_max_dist = arg_parse_uint(&arg);
967 } else if (arg_match(&arg, &kf_disabled, argi)) {
968 config->cfg.kf_mode = VPX_KF_DISABLED;
969 #if CONFIG_VP9_HIGHBITDEPTH
970 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
971 if (strcmp(global->codec->name, "vp9") == 0) {
972 test_16bit_internal = 1;
973 }
974 #endif
975 } else {
976 int i, match = 0;
977 for (i = 0; ctrl_args[i]; i++) {
978 if (arg_match(&arg, ctrl_args[i], argi)) {
979 int j;
980 match = 1;
981
982 /* Point either to the next free element or the first
983 * instance of this control.
984 */
985 for (j = 0; j < config->arg_ctrl_cnt; j++)
986 if (ctrl_args_map != NULL &&
987 config->arg_ctrls[j][0] == ctrl_args_map[i])
988 break;
989
990 /* Update/insert */
991 assert(j < (int)ARG_CTRL_CNT_MAX);
992 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
993 config->arg_ctrls[j][0] = ctrl_args_map[i];
994 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
995 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
996 }
997 }
998 }
999 if (!match) argj++;
1000 }
1001 }
1002 #if CONFIG_VP9_HIGHBITDEPTH
1003 if (strcmp(global->codec->name, "vp9") == 0) {
1004 config->use_16bit_internal =
1005 test_16bit_internal | (config->cfg.g_profile > 1);
1006 }
1007 #endif
1008 return eos_mark_found;
1009 }
1010
1011 #define FOREACH_STREAM(func) \
1012 do { \
1013 struct stream_state *stream; \
1014 for (stream = streams; stream; stream = stream->next) { \
1015 func; \
1016 } \
1017 } while (0)
1018
validate_stream_config(const struct stream_state * stream,const struct VpxEncoderConfig * global)1019 static void validate_stream_config(const struct stream_state *stream,
1020 const struct VpxEncoderConfig *global) {
1021 const struct stream_state *streami;
1022 (void)global;
1023
1024 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1025 fatal(
1026 "Stream %d: Specify stream dimensions with --width (-w) "
1027 " and --height (-h)",
1028 stream->index);
1029
1030 // Check that the codec bit depth is greater than the input bit depth.
1031 if (stream->config.cfg.g_input_bit_depth >
1032 (unsigned int)stream->config.cfg.g_bit_depth) {
1033 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1034 stream->index, (int)stream->config.cfg.g_bit_depth,
1035 stream->config.cfg.g_input_bit_depth);
1036 }
1037
1038 for (streami = stream; streami; streami = streami->next) {
1039 /* All streams require output files */
1040 if (!streami->config.out_fn)
1041 fatal("Stream %d: Output file is required (specify with -o)",
1042 streami->index);
1043
1044 /* Check for two streams outputting to the same file */
1045 if (streami != stream) {
1046 const char *a = stream->config.out_fn;
1047 const char *b = streami->config.out_fn;
1048 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1049 fatal("Stream %d: duplicate output file (from stream %d)",
1050 streami->index, stream->index);
1051 }
1052
1053 /* Check for two streams sharing a stats file. */
1054 if (streami != stream) {
1055 const char *a = stream->config.stats_fn;
1056 const char *b = streami->config.stats_fn;
1057 if (a && b && !strcmp(a, b))
1058 fatal("Stream %d: duplicate stats file (from stream %d)",
1059 streami->index, stream->index);
1060 }
1061
1062 #if CONFIG_FP_MB_STATS
1063 /* Check for two streams sharing a mb stats file. */
1064 if (streami != stream) {
1065 const char *a = stream->config.fpmb_stats_fn;
1066 const char *b = streami->config.fpmb_stats_fn;
1067 if (a && b && !strcmp(a, b))
1068 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1069 streami->index, stream->index);
1070 }
1071 #endif
1072 }
1073 }
1074
set_stream_dimensions(struct stream_state * stream,unsigned int w,unsigned int h)1075 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1076 unsigned int h) {
1077 if (!stream->config.cfg.g_w) {
1078 if (!stream->config.cfg.g_h)
1079 stream->config.cfg.g_w = w;
1080 else
1081 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1082 }
1083 if (!stream->config.cfg.g_h) {
1084 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1085 }
1086 }
1087
file_type_to_string(enum VideoFileType t)1088 static const char *file_type_to_string(enum VideoFileType t) {
1089 switch (t) {
1090 case FILE_TYPE_RAW: return "RAW";
1091 case FILE_TYPE_Y4M: return "Y4M";
1092 default: return "Other";
1093 }
1094 }
1095
image_format_to_string(vpx_img_fmt_t f)1096 static const char *image_format_to_string(vpx_img_fmt_t f) {
1097 switch (f) {
1098 case VPX_IMG_FMT_I420: return "I420";
1099 case VPX_IMG_FMT_I422: return "I422";
1100 case VPX_IMG_FMT_I444: return "I444";
1101 case VPX_IMG_FMT_I440: return "I440";
1102 case VPX_IMG_FMT_YV12: return "YV12";
1103 case VPX_IMG_FMT_I42016: return "I42016";
1104 case VPX_IMG_FMT_I42216: return "I42216";
1105 case VPX_IMG_FMT_I44416: return "I44416";
1106 case VPX_IMG_FMT_I44016: return "I44016";
1107 default: return "Other";
1108 }
1109 }
1110
show_stream_config(struct stream_state * stream,struct VpxEncoderConfig * global,struct VpxInputContext * input)1111 static void show_stream_config(struct stream_state *stream,
1112 struct VpxEncoderConfig *global,
1113 struct VpxInputContext *input) {
1114 #define SHOW(field) \
1115 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1116
1117 if (stream->index == 0) {
1118 fprintf(stderr, "Codec: %s\n",
1119 vpx_codec_iface_name(global->codec->codec_interface()));
1120 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1121 input->filename, file_type_to_string(input->file_type),
1122 image_format_to_string(input->fmt));
1123 }
1124 if (stream->next || stream->index)
1125 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1126 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1127 fprintf(stderr, "Encoder parameters:\n");
1128
1129 SHOW(g_usage);
1130 SHOW(g_threads);
1131 SHOW(g_profile);
1132 SHOW(g_w);
1133 SHOW(g_h);
1134 SHOW(g_bit_depth);
1135 SHOW(g_input_bit_depth);
1136 SHOW(g_timebase.num);
1137 SHOW(g_timebase.den);
1138 SHOW(g_error_resilient);
1139 SHOW(g_pass);
1140 SHOW(g_lag_in_frames);
1141 SHOW(rc_dropframe_thresh);
1142 SHOW(rc_resize_allowed);
1143 SHOW(rc_scaled_width);
1144 SHOW(rc_scaled_height);
1145 SHOW(rc_resize_up_thresh);
1146 SHOW(rc_resize_down_thresh);
1147 SHOW(rc_end_usage);
1148 SHOW(rc_target_bitrate);
1149 SHOW(rc_min_quantizer);
1150 SHOW(rc_max_quantizer);
1151 SHOW(rc_undershoot_pct);
1152 SHOW(rc_overshoot_pct);
1153 SHOW(rc_buf_sz);
1154 SHOW(rc_buf_initial_sz);
1155 SHOW(rc_buf_optimal_sz);
1156 SHOW(rc_2pass_vbr_bias_pct);
1157 SHOW(rc_2pass_vbr_minsection_pct);
1158 SHOW(rc_2pass_vbr_maxsection_pct);
1159 SHOW(rc_2pass_vbr_corpus_complexity);
1160 SHOW(kf_mode);
1161 SHOW(kf_min_dist);
1162 SHOW(kf_max_dist);
1163 }
1164
open_output_file(struct stream_state * stream,struct VpxEncoderConfig * global,const struct VpxRational * pixel_aspect_ratio)1165 static void open_output_file(struct stream_state *stream,
1166 struct VpxEncoderConfig *global,
1167 const struct VpxRational *pixel_aspect_ratio) {
1168 const char *fn = stream->config.out_fn;
1169 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1170
1171 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1172
1173 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1174
1175 if (!stream->file) fatal("Failed to open output file");
1176
1177 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1178 fatal("WebM output to pipes not supported.");
1179
1180 #if CONFIG_WEBM_IO
1181 if (stream->config.write_webm) {
1182 stream->webm_ctx.stream = stream->file;
1183 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1184 global->codec->fourcc, pixel_aspect_ratio);
1185 }
1186 #else
1187 (void)pixel_aspect_ratio;
1188 #endif
1189
1190 if (!stream->config.write_webm) {
1191 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1192 }
1193 }
1194
close_output_file(struct stream_state * stream,unsigned int fourcc)1195 static void close_output_file(struct stream_state *stream,
1196 unsigned int fourcc) {
1197 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1198
1199 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1200
1201 #if CONFIG_WEBM_IO
1202 if (stream->config.write_webm) {
1203 write_webm_file_footer(&stream->webm_ctx);
1204 }
1205 #endif
1206
1207 if (!stream->config.write_webm) {
1208 if (!fseek(stream->file, 0, SEEK_SET))
1209 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1210 stream->frames_out);
1211 }
1212
1213 fclose(stream->file);
1214 }
1215
setup_pass(struct stream_state * stream,struct VpxEncoderConfig * global,int pass)1216 static void setup_pass(struct stream_state *stream,
1217 struct VpxEncoderConfig *global, int pass) {
1218 if (stream->config.stats_fn) {
1219 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1220 fatal("Failed to open statistics store");
1221 } else {
1222 if (!stats_open_mem(&stream->stats, pass))
1223 fatal("Failed to open statistics store");
1224 }
1225
1226 #if CONFIG_FP_MB_STATS
1227 if (stream->config.fpmb_stats_fn) {
1228 if (!stats_open_file(&stream->fpmb_stats, stream->config.fpmb_stats_fn,
1229 pass))
1230 fatal("Failed to open mb statistics store");
1231 } else {
1232 if (!stats_open_mem(&stream->fpmb_stats, pass))
1233 fatal("Failed to open mb statistics store");
1234 }
1235 #endif
1236
1237 stream->config.cfg.g_pass = global->passes == 2
1238 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1239 : VPX_RC_ONE_PASS;
1240 if (pass) {
1241 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1242 #if CONFIG_FP_MB_STATS
1243 stream->config.cfg.rc_firstpass_mb_stats_in =
1244 stats_get(&stream->fpmb_stats);
1245 #endif
1246 }
1247
1248 stream->cx_time = 0;
1249 stream->nbytes = 0;
1250 stream->frames_out = 0;
1251 }
1252
initialize_encoder(struct stream_state * stream,struct VpxEncoderConfig * global)1253 static void initialize_encoder(struct stream_state *stream,
1254 struct VpxEncoderConfig *global) {
1255 int i;
1256 int flags = 0;
1257
1258 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1259 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1260 #if CONFIG_VP9_HIGHBITDEPTH
1261 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1262 #endif
1263
1264 /* Construct Encoder Context */
1265 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1266 &stream->config.cfg, flags);
1267 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1268
1269 /* Note that we bypass the vpx_codec_control wrapper macro because
1270 * we're being clever to store the control IDs in an array. Real
1271 * applications will want to make use of the enumerations directly
1272 */
1273 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1274 int ctrl = stream->config.arg_ctrls[i][0];
1275 int value = stream->config.arg_ctrls[i][1];
1276 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1277 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1278
1279 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1280 }
1281
1282 #if CONFIG_DECODERS
1283 if (global->test_decode != TEST_DECODE_OFF) {
1284 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1285 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1286 }
1287 #endif
1288 }
1289
encode_frame(struct stream_state * stream,struct VpxEncoderConfig * global,struct vpx_image * img,unsigned int frames_in)1290 static void encode_frame(struct stream_state *stream,
1291 struct VpxEncoderConfig *global, struct vpx_image *img,
1292 unsigned int frames_in) {
1293 vpx_codec_pts_t frame_start, next_frame_start;
1294 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1295 struct vpx_usec_timer timer;
1296
1297 frame_start =
1298 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1299 cfg->g_timebase.num / global->framerate.num;
1300 next_frame_start =
1301 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1302 cfg->g_timebase.num / global->framerate.num;
1303
1304 /* Scale if necessary */
1305 #if CONFIG_VP9_HIGHBITDEPTH
1306 if (img) {
1307 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1308 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1309 if (img->fmt != VPX_IMG_FMT_I42016) {
1310 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1311 exit(EXIT_FAILURE);
1312 }
1313 #if CONFIG_LIBYUV
1314 if (!stream->img) {
1315 stream->img =
1316 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1317 }
1318 I420Scale_16(
1319 (uint16_t *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1320 (uint16_t *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1321 (uint16_t *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1322 img->d_w, img->d_h, (uint16_t *)stream->img->planes[VPX_PLANE_Y],
1323 stream->img->stride[VPX_PLANE_Y] / 2,
1324 (uint16_t *)stream->img->planes[VPX_PLANE_U],
1325 stream->img->stride[VPX_PLANE_U] / 2,
1326 (uint16_t *)stream->img->planes[VPX_PLANE_V],
1327 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1328 stream->img->d_h, kFilterBox);
1329 img = stream->img;
1330 #else
1331 stream->encoder.err = 1;
1332 ctx_exit_on_error(&stream->encoder,
1333 "Stream %d: Failed to encode frame.\n"
1334 "Scaling disabled in this configuration. \n"
1335 "To enable, configure with --enable-libyuv\n",
1336 stream->index);
1337 #endif
1338 }
1339 }
1340 #endif
1341 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1342 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1343 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1344 exit(EXIT_FAILURE);
1345 }
1346 #if CONFIG_LIBYUV
1347 if (!stream->img)
1348 stream->img =
1349 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1350 I420Scale(
1351 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1352 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1353 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1354 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1355 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1356 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1357 stream->img->d_w, stream->img->d_h, kFilterBox);
1358 img = stream->img;
1359 #else
1360 stream->encoder.err = 1;
1361 ctx_exit_on_error(&stream->encoder,
1362 "Stream %d: Failed to encode frame.\n"
1363 "Scaling disabled in this configuration. \n"
1364 "To enable, configure with --enable-libyuv\n",
1365 stream->index);
1366 #endif
1367 }
1368
1369 vpx_usec_timer_start(&timer);
1370 vpx_codec_encode(&stream->encoder, img, frame_start,
1371 (unsigned long)(next_frame_start - frame_start), 0,
1372 global->deadline);
1373 vpx_usec_timer_mark(&timer);
1374 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1375 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1376 stream->index);
1377 }
1378
update_quantizer_histogram(struct stream_state * stream)1379 static void update_quantizer_histogram(struct stream_state *stream) {
1380 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1381 int q;
1382
1383 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1384 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1385 stream->counts[q]++;
1386 }
1387 }
1388
get_cx_data(struct stream_state * stream,struct VpxEncoderConfig * global,int * got_data)1389 static void get_cx_data(struct stream_state *stream,
1390 struct VpxEncoderConfig *global, int *got_data) {
1391 const vpx_codec_cx_pkt_t *pkt;
1392 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1393 vpx_codec_iter_t iter = NULL;
1394
1395 *got_data = 0;
1396 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1397 static size_t fsize = 0;
1398 static FileOffset ivf_header_pos = 0;
1399
1400 switch (pkt->kind) {
1401 case VPX_CODEC_CX_FRAME_PKT:
1402 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1403 stream->frames_out++;
1404 }
1405 if (!global->quiet)
1406 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1407
1408 update_rate_histogram(stream->rate_hist, cfg, pkt);
1409 #if CONFIG_WEBM_IO
1410 if (stream->config.write_webm) {
1411 write_webm_block(&stream->webm_ctx, cfg, pkt);
1412 }
1413 #endif
1414 if (!stream->config.write_webm) {
1415 if (pkt->data.frame.partition_id <= 0) {
1416 ivf_header_pos = ftello(stream->file);
1417 fsize = pkt->data.frame.sz;
1418
1419 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1420 } else {
1421 fsize += pkt->data.frame.sz;
1422
1423 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1424 const FileOffset currpos = ftello(stream->file);
1425 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1426 ivf_write_frame_size(stream->file, fsize);
1427 fseeko(stream->file, currpos, SEEK_SET);
1428 }
1429 }
1430
1431 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1432 stream->file);
1433 }
1434 stream->nbytes += pkt->data.raw.sz;
1435
1436 *got_data = 1;
1437 #if CONFIG_DECODERS
1438 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1439 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1440 (unsigned int)pkt->data.frame.sz, NULL, 0);
1441 if (stream->decoder.err) {
1442 warn_or_exit_on_error(&stream->decoder,
1443 global->test_decode == TEST_DECODE_FATAL,
1444 "Failed to decode frame %d in stream %d",
1445 stream->frames_out + 1, stream->index);
1446 stream->mismatch_seen = stream->frames_out + 1;
1447 }
1448 }
1449 #endif
1450 break;
1451 case VPX_CODEC_STATS_PKT:
1452 stream->frames_out++;
1453 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1454 pkt->data.twopass_stats.sz);
1455 stream->nbytes += pkt->data.raw.sz;
1456 break;
1457 #if CONFIG_FP_MB_STATS
1458 case VPX_CODEC_FPMB_STATS_PKT:
1459 stats_write(&stream->fpmb_stats, pkt->data.firstpass_mb_stats.buf,
1460 pkt->data.firstpass_mb_stats.sz);
1461 stream->nbytes += pkt->data.raw.sz;
1462 break;
1463 #endif
1464 case VPX_CODEC_PSNR_PKT:
1465
1466 if (global->show_psnr) {
1467 int i;
1468
1469 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1470 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1471 for (i = 0; i < 4; i++) {
1472 if (!global->quiet)
1473 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1474 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1475 }
1476 stream->psnr_count++;
1477 }
1478
1479 break;
1480 default: break;
1481 }
1482 }
1483 }
1484
show_psnr(struct stream_state * stream,double peak)1485 static void show_psnr(struct stream_state *stream, double peak) {
1486 int i;
1487 double ovpsnr;
1488
1489 if (!stream->psnr_count) return;
1490
1491 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1492 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1493 (double)stream->psnr_sse_total);
1494 fprintf(stderr, " %.3f", ovpsnr);
1495
1496 for (i = 0; i < 4; i++) {
1497 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1498 }
1499 fprintf(stderr, "\n");
1500 }
1501
usec_to_fps(uint64_t usec,unsigned int frames)1502 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1503 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1504 }
1505
test_decode(struct stream_state * stream,enum TestDecodeFatality fatal,const VpxInterface * codec)1506 static void test_decode(struct stream_state *stream,
1507 enum TestDecodeFatality fatal,
1508 const VpxInterface *codec) {
1509 vpx_image_t enc_img, dec_img;
1510
1511 if (stream->mismatch_seen) return;
1512
1513 /* Get the internal reference frame */
1514 if (strcmp(codec->name, "vp8") == 0) {
1515 struct vpx_ref_frame ref_enc, ref_dec;
1516 int width, height;
1517
1518 width = (stream->config.cfg.g_w + 15) & ~15;
1519 height = (stream->config.cfg.g_h + 15) & ~15;
1520 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1521 enc_img = ref_enc.img;
1522 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1523 dec_img = ref_dec.img;
1524
1525 ref_enc.frame_type = VP8_LAST_FRAME;
1526 ref_dec.frame_type = VP8_LAST_FRAME;
1527 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1528 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1529 } else {
1530 struct vp9_ref_frame ref_enc, ref_dec;
1531
1532 ref_enc.idx = 0;
1533 ref_dec.idx = 0;
1534 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1535 enc_img = ref_enc.img;
1536 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1537 dec_img = ref_dec.img;
1538 #if CONFIG_VP9_HIGHBITDEPTH
1539 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1540 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1541 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1542 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1543 enc_img.d_w, enc_img.d_h, 16);
1544 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1545 }
1546 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1547 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1548 dec_img.d_w, dec_img.d_h, 16);
1549 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1550 }
1551 }
1552 #endif
1553 }
1554 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1555 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1556
1557 if (!compare_img(&enc_img, &dec_img)) {
1558 int y[4], u[4], v[4];
1559 #if CONFIG_VP9_HIGHBITDEPTH
1560 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1561 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1562 } else {
1563 find_mismatch(&enc_img, &dec_img, y, u, v);
1564 }
1565 #else
1566 find_mismatch(&enc_img, &dec_img, y, u, v);
1567 #endif
1568 stream->decoder.err = 1;
1569 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1570 "Stream %d: Encode/decode mismatch on frame %d at"
1571 " Y[%d, %d] {%d/%d},"
1572 " U[%d, %d] {%d/%d},"
1573 " V[%d, %d] {%d/%d}",
1574 stream->index, stream->frames_out, y[0], y[1], y[2],
1575 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1576 stream->mismatch_seen = stream->frames_out;
1577 }
1578
1579 vpx_img_free(&enc_img);
1580 vpx_img_free(&dec_img);
1581 }
1582
print_time(const char * label,int64_t etl)1583 static void print_time(const char *label, int64_t etl) {
1584 int64_t hours;
1585 int64_t mins;
1586 int64_t secs;
1587
1588 if (etl >= 0) {
1589 hours = etl / 3600;
1590 etl -= hours * 3600;
1591 mins = etl / 60;
1592 etl -= mins * 60;
1593 secs = etl;
1594
1595 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1596 hours, mins, secs);
1597 } else {
1598 fprintf(stderr, "[%3s unknown] ", label);
1599 }
1600 }
1601
main(int argc,const char ** argv_)1602 int main(int argc, const char **argv_) {
1603 int pass;
1604 vpx_image_t raw;
1605 #if CONFIG_VP9_HIGHBITDEPTH
1606 vpx_image_t raw_shift;
1607 int allocated_raw_shift = 0;
1608 int use_16bit_internal = 0;
1609 int input_shift = 0;
1610 #endif
1611 int frame_avail, got_data;
1612
1613 struct VpxInputContext input;
1614 struct VpxEncoderConfig global;
1615 struct stream_state *streams = NULL;
1616 char **argv, **argi;
1617 uint64_t cx_time = 0;
1618 int stream_cnt = 0;
1619 int res = 0;
1620
1621 memset(&input, 0, sizeof(input));
1622 exec_name = argv_[0];
1623
1624 /* Setup default input stream settings */
1625 input.framerate.numerator = 30;
1626 input.framerate.denominator = 1;
1627 input.only_i420 = 1;
1628 input.bit_depth = 0;
1629
1630 /* First parse the global configuration values, because we want to apply
1631 * other parameters on top of the default configuration provided by the
1632 * codec.
1633 */
1634 argv = argv_dup(argc - 1, argv_ + 1);
1635 parse_global_config(&global, argv);
1636
1637 if (argc < 3) usage_exit();
1638
1639 switch (global.color_type) {
1640 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1641 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1642 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1643 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1644 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1645 }
1646
1647 {
1648 /* Now parse each stream's parameters. Using a local scope here
1649 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1650 * loops
1651 */
1652 struct stream_state *stream = NULL;
1653
1654 do {
1655 stream = new_stream(&global, stream);
1656 stream_cnt++;
1657 if (!streams) streams = stream;
1658 } while (parse_stream_params(&global, stream, argv));
1659 }
1660
1661 /* Check for unrecognized options */
1662 for (argi = argv; *argi; argi++)
1663 if (argi[0][0] == '-' && argi[0][1])
1664 die("Error: Unrecognized option %s\n", *argi);
1665
1666 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1667 &stream->config.cfg););
1668
1669 /* Handle non-option arguments */
1670 input.filename = argv[0];
1671
1672 if (!input.filename) {
1673 fprintf(stderr, "No input file specified!\n");
1674 usage_exit();
1675 }
1676
1677 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1678 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1679
1680 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1681 int frames_in = 0, seen_frames = 0;
1682 int64_t estimated_time_left = -1;
1683 int64_t average_rate = -1;
1684 int64_t lagged_count = 0;
1685
1686 open_input_file(&input);
1687
1688 /* If the input file doesn't specify its w/h (raw files), try to get
1689 * the data from the first stream's configuration.
1690 */
1691 if (!input.width || !input.height) {
1692 FOREACH_STREAM({
1693 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1694 input.width = stream->config.cfg.g_w;
1695 input.height = stream->config.cfg.g_h;
1696 break;
1697 }
1698 });
1699 }
1700
1701 /* Update stream configurations from the input file's parameters */
1702 if (!input.width || !input.height)
1703 fatal(
1704 "Specify stream dimensions with --width (-w) "
1705 " and --height (-h)");
1706
1707 /* If input file does not specify bit-depth but input-bit-depth parameter
1708 * exists, assume that to be the input bit-depth. However, if the
1709 * input-bit-depth paramter does not exist, assume the input bit-depth
1710 * to be the same as the codec bit-depth.
1711 */
1712 if (!input.bit_depth) {
1713 FOREACH_STREAM({
1714 if (stream->config.cfg.g_input_bit_depth)
1715 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1716 else
1717 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1718 (int)stream->config.cfg.g_bit_depth;
1719 });
1720 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1721 } else {
1722 FOREACH_STREAM(
1723 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
1724 }
1725
1726 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1727 FOREACH_STREAM(validate_stream_config(stream, &global));
1728
1729 /* Ensure that --passes and --pass are consistent. If --pass is set and
1730 * --passes=2, ensure --fpf was set.
1731 */
1732 if (global.pass && global.passes == 2)
1733 FOREACH_STREAM({
1734 if (!stream->config.stats_fn)
1735 die("Stream %d: Must specify --fpf when --pass=%d"
1736 " and --passes=2\n",
1737 stream->index, global.pass);
1738 });
1739
1740 #if !CONFIG_WEBM_IO
1741 FOREACH_STREAM({
1742 if (stream->config.write_webm) {
1743 stream->config.write_webm = 0;
1744 warn(
1745 "vpxenc was compiled without WebM container support."
1746 "Producing IVF output");
1747 }
1748 });
1749 #endif
1750
1751 /* Use the frame rate from the file only if none was specified
1752 * on the command-line.
1753 */
1754 if (!global.have_framerate) {
1755 global.framerate.num = input.framerate.numerator;
1756 global.framerate.den = input.framerate.denominator;
1757 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
1758 stream->config.cfg.g_timebase.num = global.framerate.den);
1759 }
1760
1761 /* Show configuration */
1762 if (global.verbose && pass == 0)
1763 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1764
1765 if (pass == (global.pass ? global.pass - 1 : 0)) {
1766 if (input.file_type == FILE_TYPE_Y4M)
1767 /*The Y4M reader does its own allocation.
1768 Just initialize this here to avoid problems if we never read any
1769 frames.*/
1770 memset(&raw, 0, sizeof(raw));
1771 else
1772 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1773
1774 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
1775 &stream->config.cfg, &global.framerate));
1776 }
1777
1778 FOREACH_STREAM(setup_pass(stream, &global, pass));
1779 FOREACH_STREAM(
1780 open_output_file(stream, &global, &input.pixel_aspect_ratio));
1781 FOREACH_STREAM(initialize_encoder(stream, &global));
1782
1783 #if CONFIG_VP9_HIGHBITDEPTH
1784 if (strcmp(global.codec->name, "vp9") == 0) {
1785 // Check to see if at least one stream uses 16 bit internal.
1786 // Currently assume that the bit_depths for all streams using
1787 // highbitdepth are the same.
1788 FOREACH_STREAM({
1789 if (stream->config.use_16bit_internal) {
1790 use_16bit_internal = 1;
1791 }
1792 if (stream->config.cfg.g_profile == 0) {
1793 input_shift = 0;
1794 } else {
1795 input_shift = (int)stream->config.cfg.g_bit_depth -
1796 stream->config.cfg.g_input_bit_depth;
1797 }
1798 });
1799 }
1800 #endif
1801
1802 frame_avail = 1;
1803 got_data = 0;
1804
1805 while (frame_avail || got_data) {
1806 struct vpx_usec_timer timer;
1807
1808 if (!global.limit || frames_in < global.limit) {
1809 frame_avail = read_frame(&input, &raw);
1810
1811 if (frame_avail) frames_in++;
1812 seen_frames =
1813 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
1814
1815 if (!global.quiet) {
1816 float fps = usec_to_fps(cx_time, seen_frames);
1817 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1818
1819 if (stream_cnt == 1)
1820 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
1821 streams->frames_out, (int64_t)streams->nbytes);
1822 else
1823 fprintf(stderr, "frame %4d ", frames_in);
1824
1825 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
1826 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1827 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
1828 fps >= 1.0 ? "fps" : "fpm");
1829 print_time("ETA", estimated_time_left);
1830 }
1831
1832 } else
1833 frame_avail = 0;
1834
1835 if (frames_in > global.skip_frames) {
1836 #if CONFIG_VP9_HIGHBITDEPTH
1837 vpx_image_t *frame_to_encode;
1838 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
1839 assert(use_16bit_internal);
1840 // Input bit depth and stream bit depth do not match, so up
1841 // shift frame to stream bit depth
1842 if (!allocated_raw_shift) {
1843 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
1844 input.width, input.height, 32);
1845 allocated_raw_shift = 1;
1846 }
1847 vpx_img_upshift(&raw_shift, &raw, input_shift);
1848 frame_to_encode = &raw_shift;
1849 } else {
1850 frame_to_encode = &raw;
1851 }
1852 vpx_usec_timer_start(&timer);
1853 if (use_16bit_internal) {
1854 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
1855 FOREACH_STREAM({
1856 if (stream->config.use_16bit_internal)
1857 encode_frame(stream, &global,
1858 frame_avail ? frame_to_encode : NULL, frames_in);
1859 else
1860 assert(0);
1861 });
1862 } else {
1863 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
1864 FOREACH_STREAM(encode_frame(stream, &global,
1865 frame_avail ? frame_to_encode : NULL,
1866 frames_in));
1867 }
1868 #else
1869 vpx_usec_timer_start(&timer);
1870 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
1871 frames_in));
1872 #endif
1873 vpx_usec_timer_mark(&timer);
1874 cx_time += vpx_usec_timer_elapsed(&timer);
1875
1876 FOREACH_STREAM(update_quantizer_histogram(stream));
1877
1878 got_data = 0;
1879 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1880
1881 if (!got_data && input.length && streams != NULL &&
1882 !streams->frames_out) {
1883 lagged_count = global.limit ? seen_frames : ftello(input.file);
1884 } else if (input.length) {
1885 int64_t remaining;
1886 int64_t rate;
1887
1888 if (global.limit) {
1889 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1890
1891 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1892 remaining = 1000 * (global.limit - global.skip_frames -
1893 seen_frames + lagged_count);
1894 } else {
1895 const int64_t input_pos = ftello(input.file);
1896 const int64_t input_pos_lagged = input_pos - lagged_count;
1897 const int64_t limit = input.length;
1898
1899 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1900 remaining = limit - input_pos + lagged_count;
1901 }
1902
1903 average_rate =
1904 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
1905 estimated_time_left = average_rate ? remaining / average_rate : -1;
1906 }
1907
1908 if (got_data && global.test_decode != TEST_DECODE_OFF)
1909 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1910 }
1911
1912 fflush(stdout);
1913 if (!global.quiet) fprintf(stderr, "\033[K");
1914 }
1915
1916 if (stream_cnt > 1) fprintf(stderr, "\n");
1917
1918 if (!global.quiet) {
1919 FOREACH_STREAM(fprintf(
1920 stderr,
1921 "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64 "b/f %7" PRId64
1922 "b/s %7" PRId64 " %s (%.2f fps)\033[K\n",
1923 pass + 1, global.passes, frames_in, stream->frames_out,
1924 (int64_t)stream->nbytes,
1925 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
1926 seen_frames
1927 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
1928 global.framerate.den / seen_frames
1929 : 0,
1930 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1931 stream->cx_time > 9999999 ? "ms" : "us",
1932 usec_to_fps(stream->cx_time, seen_frames)));
1933 }
1934
1935 if (global.show_psnr) {
1936 if (global.codec->fourcc == VP9_FOURCC) {
1937 FOREACH_STREAM(
1938 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
1939 } else {
1940 FOREACH_STREAM(show_psnr(stream, 255.0));
1941 }
1942 }
1943
1944 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1945
1946 if (global.test_decode != TEST_DECODE_OFF) {
1947 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1948 }
1949
1950 close_input_file(&input);
1951
1952 if (global.test_decode == TEST_DECODE_FATAL) {
1953 FOREACH_STREAM(res |= stream->mismatch_seen);
1954 }
1955 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1956
1957 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1958
1959 #if CONFIG_FP_MB_STATS
1960 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
1961 #endif
1962
1963 if (global.pass) break;
1964 }
1965
1966 if (global.show_q_hist_buckets)
1967 FOREACH_STREAM(
1968 show_q_histogram(stream->counts, global.show_q_hist_buckets));
1969
1970 if (global.show_rate_hist_buckets)
1971 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
1972 global.show_rate_hist_buckets));
1973 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1974
1975 #if CONFIG_INTERNAL_STATS
1976 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1977 * to match some existing utilities.
1978 */
1979 if (!(global.pass == 1 && global.passes == 2))
1980 FOREACH_STREAM({
1981 FILE *f = fopen("opsnr.stt", "a");
1982 if (stream->mismatch_seen) {
1983 fprintf(f, "First mismatch occurred in frame %d\n",
1984 stream->mismatch_seen);
1985 } else {
1986 fprintf(f, "No mismatch detected in recon buffers\n");
1987 }
1988 fclose(f);
1989 });
1990 #endif
1991
1992 #if CONFIG_VP9_HIGHBITDEPTH
1993 if (allocated_raw_shift) vpx_img_free(&raw_shift);
1994 #endif
1995 vpx_img_free(&raw);
1996 free(argv);
1997 free(streams);
1998 return res ? EXIT_FAILURE : EXIT_SUCCESS;
1999 }
2000