WebM Codec SDK
vpxenc
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 
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 
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 
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 
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_nv12 =
99  ARG_DEF(NULL, "nv12", 0, "Input file is NV12 ");
100 static const arg_def_t use_yv12 =
101  ARG_DEF(NULL, "yv12", 0, "Input file is YV12 ");
102 static const arg_def_t use_i420 =
103  ARG_DEF(NULL, "i420", 0, "Input file is I420 (default)");
104 static const arg_def_t use_i422 =
105  ARG_DEF(NULL, "i422", 0, "Input file is I422");
106 static const arg_def_t use_i444 =
107  ARG_DEF(NULL, "i444", 0, "Input file is I444");
108 static const arg_def_t use_i440 =
109  ARG_DEF(NULL, "i440", 0, "Input file is I440");
110 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, "Codec to use");
111 static const arg_def_t passes =
112  ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
113 static const arg_def_t pass_arg =
114  ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
115 static const arg_def_t fpf_name =
116  ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");
117 #if CONFIG_FP_MB_STATS
118 static const arg_def_t fpmbf_name =
119  ARG_DEF(NULL, "fpmbf", 1, "First pass block statistics file name");
120 #endif
121 static const arg_def_t limit =
122  ARG_DEF(NULL, "limit", 1, "Stop encoding after n input frames");
123 static const arg_def_t skip =
124  ARG_DEF(NULL, "skip", 1, "Skip the first n input frames");
125 static const arg_def_t deadline =
126  ARG_DEF("d", "deadline", 1, "Deadline per frame (usec)");
127 static const arg_def_t best_dl =
128  ARG_DEF(NULL, "best", 0, "Use Best Quality Deadline");
129 static const arg_def_t good_dl =
130  ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
131 static const arg_def_t rt_dl =
132  ARG_DEF(NULL, "rt", 0, "Use Realtime Quality Deadline");
133 static const arg_def_t quietarg =
134  ARG_DEF("q", "quiet", 0, "Do not print encode progress");
135 static const arg_def_t verbosearg =
136  ARG_DEF("v", "verbose", 0, "Show encoder parameters");
137 static const arg_def_t psnrarg =
138  ARG_DEF(NULL, "psnr", 0, "Show PSNR in status line");
139 
140 static const struct arg_enum_list test_decode_enum[] = {
141  { "off", TEST_DECODE_OFF },
142  { "fatal", TEST_DECODE_FATAL },
143  { "warn", TEST_DECODE_WARN },
144  { NULL, 0 }
145 };
146 static const arg_def_t recontest = ARG_DEF_ENUM(
147  NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
148 static const arg_def_t framerate =
149  ARG_DEF(NULL, "fps", 1, "Stream frame rate (rate/scale)");
150 static const arg_def_t use_webm =
151  ARG_DEF(NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
152 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, "Output IVF");
153 static const arg_def_t out_part =
154  ARG_DEF("P", "output-partitions", 0,
155  "Makes encoder output partitions. Requires IVF output!");
156 static const arg_def_t q_hist_n =
157  ARG_DEF(NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
158 static const arg_def_t rate_hist_n =
159  ARG_DEF(NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
160 static const arg_def_t disable_warnings =
161  ARG_DEF(NULL, "disable-warnings", 0,
162  "Disable warnings about potentially incorrect encode settings.");
163 static const arg_def_t disable_warning_prompt =
164  ARG_DEF("y", "disable-warning-prompt", 0,
165  "Display warnings, but do not prompt user to continue.");
166 
167 #if CONFIG_VP9_HIGHBITDEPTH
168 static const arg_def_t test16bitinternalarg = ARG_DEF(
169  NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
170 #endif
171 
172 static const arg_def_t *main_args[] = { &help,
173  &debugmode,
174  &outputfile,
175  &codecarg,
176  &passes,
177  &pass_arg,
178  &fpf_name,
179  &limit,
180  &skip,
181  &deadline,
182  &best_dl,
183  &good_dl,
184  &rt_dl,
185  &quietarg,
186  &verbosearg,
187  &psnrarg,
188  &use_webm,
189  &use_ivf,
190  &out_part,
191  &q_hist_n,
192  &rate_hist_n,
193  &disable_warnings,
194  &disable_warning_prompt,
195  &recontest,
196  NULL };
197 
198 static const arg_def_t usage =
199  ARG_DEF("u", "usage", 1, "Usage profile number to use");
200 static const arg_def_t threads =
201  ARG_DEF("t", "threads", 1, "Max number of threads to use");
202 static const arg_def_t profile =
203  ARG_DEF(NULL, "profile", 1, "Bitstream profile number to use");
204 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
205 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
206 #if CONFIG_WEBM_IO
207 static const struct arg_enum_list stereo_mode_enum[] = {
208  { "mono", STEREO_FORMAT_MONO },
209  { "left-right", STEREO_FORMAT_LEFT_RIGHT },
210  { "bottom-top", STEREO_FORMAT_BOTTOM_TOP },
211  { "top-bottom", STEREO_FORMAT_TOP_BOTTOM },
212  { "right-left", STEREO_FORMAT_RIGHT_LEFT },
213  { NULL, 0 }
214 };
215 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
216  NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
217 #endif
218 static const arg_def_t timebase = ARG_DEF(
219  NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
220 static const arg_def_t error_resilient =
221  ARG_DEF(NULL, "error-resilient", 1, "Enable error resiliency features");
222 static const arg_def_t lag_in_frames =
223  ARG_DEF(NULL, "lag-in-frames", 1, "Max number of frames to lag");
224 
225 static const arg_def_t *global_args[] = { &use_nv12,
226  &use_yv12,
227  &use_i420,
228  &use_i422,
229  &use_i444,
230  &use_i440,
231  &usage,
232  &threads,
233  &profile,
234  &width,
235  &height,
236 #if CONFIG_WEBM_IO
237  &stereo_mode,
238 #endif
239  &timebase,
240  &framerate,
241  &error_resilient,
242 #if CONFIG_VP9_HIGHBITDEPTH
243  &test16bitinternalarg,
244 #endif
245  &lag_in_frames,
246  NULL };
247 
248 static const arg_def_t dropframe_thresh =
249  ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
250 static const arg_def_t resize_allowed =
251  ARG_DEF(NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
252 static const arg_def_t resize_width =
253  ARG_DEF(NULL, "resize-width", 1, "Width of encoded frame");
254 static const arg_def_t resize_height =
255  ARG_DEF(NULL, "resize-height", 1, "Height of encoded frame");
256 static const arg_def_t resize_up_thresh =
257  ARG_DEF(NULL, "resize-up", 1, "Upscale threshold (buf %)");
258 static const arg_def_t resize_down_thresh =
259  ARG_DEF(NULL, "resize-down", 1, "Downscale threshold (buf %)");
260 static const struct arg_enum_list end_usage_enum[] = { { "vbr", VPX_VBR },
261  { "cbr", VPX_CBR },
262  { "cq", VPX_CQ },
263  { "q", VPX_Q },
264  { NULL, 0 } };
265 static const arg_def_t end_usage =
266  ARG_DEF_ENUM(NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
267 static const arg_def_t target_bitrate =
268  ARG_DEF(NULL, "target-bitrate", 1, "Bitrate (kbps)");
269 static const arg_def_t min_quantizer =
270  ARG_DEF(NULL, "min-q", 1, "Minimum (best) quantizer");
271 static const arg_def_t max_quantizer =
272  ARG_DEF(NULL, "max-q", 1, "Maximum (worst) quantizer");
273 static const arg_def_t undershoot_pct =
274  ARG_DEF(NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
275 static const arg_def_t overshoot_pct =
276  ARG_DEF(NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
277 static const arg_def_t buf_sz =
278  ARG_DEF(NULL, "buf-sz", 1, "Client buffer size (ms)");
279 static const arg_def_t buf_initial_sz =
280  ARG_DEF(NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
281 static const arg_def_t buf_optimal_sz =
282  ARG_DEF(NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
283 static const arg_def_t *rc_args[] = {
284  &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
285  &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
286  &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct,
287  &buf_sz, &buf_initial_sz, &buf_optimal_sz, NULL
288 };
289 
290 static const arg_def_t bias_pct =
291  ARG_DEF(NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
292 static const arg_def_t minsection_pct =
293  ARG_DEF(NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
294 static const arg_def_t maxsection_pct =
295  ARG_DEF(NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
296 static const arg_def_t corpus_complexity =
297  ARG_DEF(NULL, "corpus-complexity", 1, "corpus vbr complexity midpoint");
298 static const arg_def_t *rc_twopass_args[] = { &bias_pct, &minsection_pct,
299  &maxsection_pct,
300  &corpus_complexity, NULL };
301 
302 static const arg_def_t kf_min_dist =
303  ARG_DEF(NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
304 static const arg_def_t kf_max_dist =
305  ARG_DEF(NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
306 static const arg_def_t kf_disabled =
307  ARG_DEF(NULL, "disable-kf", 0, "Disable keyframe placement");
308 static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
309  NULL };
310 
311 static const arg_def_t noise_sens =
312  ARG_DEF(NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
313 static const arg_def_t sharpness =
314  ARG_DEF(NULL, "sharpness", 1,
315  "Increase sharpness at the expense of lower PSNR. (0..7)");
316 static const arg_def_t static_thresh =
317  ARG_DEF(NULL, "static-thresh", 1, "Motion detection threshold");
318 static const arg_def_t arnr_maxframes =
319  ARG_DEF(NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
320 static const arg_def_t arnr_strength =
321  ARG_DEF(NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
322 static const arg_def_t arnr_type =
323  ARG_DEF(NULL, "arnr-type", 1, "AltRef filter type (1..3)");
324 static const struct arg_enum_list tuning_enum[] = { { "psnr", VP8_TUNE_PSNR },
325  { "ssim", VP8_TUNE_SSIM },
326  { NULL, 0 } };
327 static const arg_def_t tune_ssim =
328  ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
329 static const arg_def_t cq_level =
330  ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
331 static const arg_def_t max_intra_rate_pct =
332  ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
333 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
334  NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
335 
336 #if CONFIG_VP8_ENCODER
337 static const arg_def_t cpu_used_vp8 =
338  ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
339 static const arg_def_t auto_altref_vp8 = ARG_DEF(
340  NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames. (0..1)");
341 static const arg_def_t token_parts =
342  ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
343 static const arg_def_t screen_content_mode =
344  ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
345 static const arg_def_t *vp8_args[] = { &cpu_used_vp8,
346  &auto_altref_vp8,
347  &noise_sens,
348  &sharpness,
349  &static_thresh,
350  &token_parts,
351  &arnr_maxframes,
352  &arnr_strength,
353  &arnr_type,
354  &tune_ssim,
355  &cq_level,
356  &max_intra_rate_pct,
357  &gf_cbr_boost_pct,
358  &screen_content_mode,
359  NULL };
360 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
374  0 };
375 #endif
376 
377 #if CONFIG_VP9_ENCODER
378 static const arg_def_t cpu_used_vp9 =
379  ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-9..9)");
380 static const arg_def_t auto_altref_vp9 = ARG_DEF(
381  NULL, "auto-alt-ref", 1,
382  "Enable automatic alt reference frames, 2+ enables multi-layer. (0..6)");
383 static const arg_def_t tile_cols =
384  ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
385 static const arg_def_t tile_rows =
386  ARG_DEF(NULL, "tile-rows", 1,
387  "Number of tile rows to use, log2 (set to 0 while threads > 1)");
388 
389 static const arg_def_t enable_tpl_model =
390  ARG_DEF(NULL, "enable-tpl", 1, "Enable temporal dependency model");
391 
392 static const arg_def_t lossless =
393  ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
394 static const arg_def_t frame_parallel_decoding = ARG_DEF(
395  NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
396 static const arg_def_t aq_mode = ARG_DEF(
397  NULL, "aq-mode", 1,
398  "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
399  "3: cyclic refresh, 4: equator360)");
400 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
401  "Special adaptive quantization for "
402  "the alternate reference frames.");
403 static const arg_def_t frame_periodic_boost =
404  ARG_DEF(NULL, "frame-boost", 1,
405  "Enable frame periodic boost (0: off (default), 1: on)");
406 static const arg_def_t max_inter_rate_pct =
407  ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
408 static const arg_def_t min_gf_interval = ARG_DEF(
409  NULL, "min-gf-interval", 1,
410  "min gf/arf frame interval (default 0, indicating in-built behavior)");
411 static const arg_def_t max_gf_interval = ARG_DEF(
412  NULL, "max-gf-interval", 1,
413  "max gf/arf frame interval (default 0, indicating in-built behavior)");
414 
415 static const struct arg_enum_list color_space_enum[] = {
416  { "unknown", VPX_CS_UNKNOWN },
417  { "bt601", VPX_CS_BT_601 },
418  { "bt709", VPX_CS_BT_709 },
419  { "smpte170", VPX_CS_SMPTE_170 },
420  { "smpte240", VPX_CS_SMPTE_240 },
421  { "bt2020", VPX_CS_BT_2020 },
422  { "reserved", VPX_CS_RESERVED },
423  { "sRGB", VPX_CS_SRGB },
424  { NULL, 0 }
425 };
426 
427 static const arg_def_t input_color_space =
428  ARG_DEF_ENUM(NULL, "color-space", 1,
429  "The color space of input content:", color_space_enum);
430 
431 #if CONFIG_VP9_HIGHBITDEPTH
432 static const struct arg_enum_list bitdepth_enum[] = {
433  { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
434 };
435 
436 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
437  "b", "bit-depth", 1,
438  "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
439  bitdepth_enum);
440 static const arg_def_t inbitdeptharg =
441  ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
442 #endif
443 
444 static const struct arg_enum_list tune_content_enum[] = {
445  { "default", VP9E_CONTENT_DEFAULT },
446  { "screen", VP9E_CONTENT_SCREEN },
447  { "film", VP9E_CONTENT_FILM },
448  { NULL, 0 }
449 };
450 
451 static const arg_def_t tune_content = ARG_DEF_ENUM(
452  NULL, "tune-content", 1, "Tune content type", tune_content_enum);
453 
454 static const arg_def_t target_level = ARG_DEF(
455  NULL, "target-level", 1,
456  "Target level\n"
457  " 255: off (default)\n"
458  " 0: only keep level stats\n"
459  " 1: adaptively set alt-ref "
460  "distance and column tile limit based on picture size, and keep"
461  " level stats\n"
462  " 10: level 1.0 11: level 1.1 "
463  "... 62: level 6.2");
464 
465 static const arg_def_t row_mt =
466  ARG_DEF(NULL, "row-mt", 1,
467  "Enable row based non-deterministic multi-threading in VP9");
468 #endif
469 
470 #if CONFIG_VP9_ENCODER
471 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
472  &auto_altref_vp9,
473  &sharpness,
474  &static_thresh,
475  &tile_cols,
476  &tile_rows,
477  &enable_tpl_model,
478  &arnr_maxframes,
479  &arnr_strength,
480  &arnr_type,
481  &tune_ssim,
482  &cq_level,
483  &max_intra_rate_pct,
484  &max_inter_rate_pct,
485  &gf_cbr_boost_pct,
486  &lossless,
487  &frame_parallel_decoding,
488  &aq_mode,
489  &alt_ref_aq,
490  &frame_periodic_boost,
491  &noise_sens,
492  &tune_content,
493  &input_color_space,
494  &min_gf_interval,
495  &max_gf_interval,
496  &target_level,
497  &row_mt,
498 #if CONFIG_VP9_HIGHBITDEPTH
499  &bitdeptharg,
500  &inbitdeptharg,
501 #endif // CONFIG_VP9_HIGHBITDEPTH
502  NULL };
503 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
509  VP9E_SET_TPL,
530  0 };
531 #endif
532 
533 static const arg_def_t *no_args[] = { NULL };
534 
535 static void show_help(FILE *fout, int shorthelp) {
536  int i;
537  const int num_encoder = get_vpx_encoder_count();
538 
539  fprintf(fout, "Usage: %s <options> -o dst_filename src_filename \n",
540  exec_name);
541 
542  if (shorthelp) {
543  fprintf(fout, "Use --help to see the full list of options.\n");
544  return;
545  }
546 
547  fprintf(fout, "\nOptions:\n");
548  arg_show_usage(fout, main_args);
549  fprintf(fout, "\nEncoder Global Options:\n");
550  arg_show_usage(fout, global_args);
551  fprintf(fout, "\nRate Control Options:\n");
552  arg_show_usage(fout, rc_args);
553  fprintf(fout, "\nTwopass Rate Control Options:\n");
554  arg_show_usage(fout, rc_twopass_args);
555  fprintf(fout, "\nKeyframe Placement Options:\n");
556  arg_show_usage(fout, kf_args);
557 #if CONFIG_VP8_ENCODER
558  fprintf(fout, "\nVP8 Specific Options:\n");
559  arg_show_usage(fout, vp8_args);
560 #endif
561 #if CONFIG_VP9_ENCODER
562  fprintf(fout, "\nVP9 Specific Options:\n");
563  arg_show_usage(fout, vp9_args);
564 #endif
565  fprintf(fout,
566  "\nStream timebase (--timebase):\n"
567  " The desired precision of timestamps in the output, expressed\n"
568  " in fractional seconds. Default is 1/1000.\n");
569  fprintf(fout, "\nIncluded encoders:\n\n");
570 
571  for (i = 0; i < num_encoder; ++i) {
572  const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
573  const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
574  fprintf(fout, " %-6s - %s %s\n", encoder->name,
575  vpx_codec_iface_name(encoder->codec_interface()), defstr);
576  }
577  fprintf(fout, "\n ");
578  fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
579 }
580 
581 void usage_exit(void) {
582  show_help(stderr, 1);
583  exit(EXIT_FAILURE);
584 }
585 
586 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
587 #if CONFIG_VP9_ENCODER
588 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
589 #else
590 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
591 #endif
592 
593 #if !CONFIG_WEBM_IO
594 typedef int stereo_format_t;
595 struct WebmOutputContext {
596  int debug;
597 };
598 #endif
599 
600 /* Per-stream configuration */
601 struct stream_config {
602  struct vpx_codec_enc_cfg cfg;
603  const char *out_fn;
604  const char *stats_fn;
605 #if CONFIG_FP_MB_STATS
606  const char *fpmb_stats_fn;
607 #endif
608  stereo_format_t stereo_fmt;
609  int arg_ctrls[ARG_CTRL_CNT_MAX][2];
610  int arg_ctrl_cnt;
611  int write_webm;
612 #if CONFIG_VP9_HIGHBITDEPTH
613  // whether to use 16bit internal buffers
614  int use_16bit_internal;
615 #endif
616 };
617 
618 struct stream_state {
619  int index;
620  struct stream_state *next;
621  struct stream_config config;
622  FILE *file;
623  struct rate_hist *rate_hist;
624  struct WebmOutputContext webm_ctx;
625  uint64_t psnr_sse_total;
626  uint64_t psnr_samples_total;
627  double psnr_totals[4];
628  int psnr_count;
629  int counts[64];
630  vpx_codec_ctx_t encoder;
631  unsigned int frames_out;
632  uint64_t cx_time;
633  size_t nbytes;
634  stats_io_t stats;
635 #if CONFIG_FP_MB_STATS
636  stats_io_t fpmb_stats;
637 #endif
638  struct vpx_image *img;
639  vpx_codec_ctx_t decoder;
640  int mismatch_seen;
641 };
642 
643 static void validate_positive_rational(const char *msg,
644  struct vpx_rational *rat) {
645  if (rat->den < 0) {
646  rat->num *= -1;
647  rat->den *= -1;
648  }
649 
650  if (rat->num < 0) die("Error: %s must be positive\n", msg);
651 
652  if (!rat->den) die("Error: %s has zero denominator\n", msg);
653 }
654 
655 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
656  char **argi, **argj;
657  struct arg arg;
658  const int num_encoder = get_vpx_encoder_count();
659 
660  if (num_encoder < 1) die("Error: no valid encoder available\n");
661 
662  /* Initialize default parameters */
663  memset(global, 0, sizeof(*global));
664  global->codec = get_vpx_encoder_by_index(num_encoder - 1);
665  global->passes = 0;
666  global->color_type = I420;
667  /* Assign default deadline to good quality */
668  global->deadline = VPX_DL_GOOD_QUALITY;
669 
670  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
671  arg.argv_step = 1;
672 
673  if (arg_match(&arg, &help, argi)) {
674  show_help(stdout, 0);
675  exit(EXIT_SUCCESS);
676  } else if (arg_match(&arg, &codecarg, argi)) {
677  global->codec = get_vpx_encoder_by_name(arg.val);
678  if (!global->codec)
679  die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
680  } else if (arg_match(&arg, &passes, argi)) {
681  global->passes = arg_parse_uint(&arg);
682 
683  if (global->passes < 1 || global->passes > 2)
684  die("Error: Invalid number of passes (%d)\n", global->passes);
685  } else if (arg_match(&arg, &pass_arg, argi)) {
686  global->pass = arg_parse_uint(&arg);
687 
688  if (global->pass < 1 || global->pass > 2)
689  die("Error: Invalid pass selected (%d)\n", global->pass);
690  } else if (arg_match(&arg, &usage, argi))
691  global->usage = arg_parse_uint(&arg);
692  else if (arg_match(&arg, &deadline, argi))
693  global->deadline = arg_parse_uint(&arg);
694  else if (arg_match(&arg, &best_dl, argi))
695  global->deadline = VPX_DL_BEST_QUALITY;
696  else if (arg_match(&arg, &good_dl, argi))
697  global->deadline = VPX_DL_GOOD_QUALITY;
698  else if (arg_match(&arg, &rt_dl, argi))
699  global->deadline = VPX_DL_REALTIME;
700  else if (arg_match(&arg, &use_yv12, argi))
701  global->color_type = YV12;
702  else if (arg_match(&arg, &use_nv12, argi))
703  global->color_type = NV12;
704  else if (arg_match(&arg, &use_i420, argi))
705  global->color_type = I420;
706  else if (arg_match(&arg, &use_i422, argi))
707  global->color_type = I422;
708  else if (arg_match(&arg, &use_i444, argi))
709  global->color_type = I444;
710  else if (arg_match(&arg, &use_i440, argi))
711  global->color_type = I440;
712  else if (arg_match(&arg, &quietarg, argi))
713  global->quiet = 1;
714  else if (arg_match(&arg, &verbosearg, argi))
715  global->verbose = 1;
716  else if (arg_match(&arg, &limit, argi))
717  global->limit = arg_parse_uint(&arg);
718  else if (arg_match(&arg, &skip, argi))
719  global->skip_frames = arg_parse_uint(&arg);
720  else if (arg_match(&arg, &psnrarg, argi))
721  global->show_psnr = 1;
722  else if (arg_match(&arg, &recontest, argi))
723  global->test_decode = arg_parse_enum_or_int(&arg);
724  else if (arg_match(&arg, &framerate, argi)) {
725  global->framerate = arg_parse_rational(&arg);
726  validate_positive_rational(arg.name, &global->framerate);
727  global->have_framerate = 1;
728  } else if (arg_match(&arg, &out_part, argi))
729  global->out_part = 1;
730  else if (arg_match(&arg, &debugmode, argi))
731  global->debug = 1;
732  else if (arg_match(&arg, &q_hist_n, argi))
733  global->show_q_hist_buckets = arg_parse_uint(&arg);
734  else if (arg_match(&arg, &rate_hist_n, argi))
735  global->show_rate_hist_buckets = arg_parse_uint(&arg);
736  else if (arg_match(&arg, &disable_warnings, argi))
737  global->disable_warnings = 1;
738  else if (arg_match(&arg, &disable_warning_prompt, argi))
739  global->disable_warning_prompt = 1;
740  else
741  argj++;
742  }
743 
744  if (global->pass) {
745  /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
746  if (global->pass > global->passes) {
747  warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
748  global->pass);
749  global->passes = global->pass;
750  }
751  }
752  /* Validate global config */
753  if (global->passes == 0) {
754 #if CONFIG_VP9_ENCODER
755  // Make default VP9 passes = 2 until there is a better quality 1-pass
756  // encoder
757  if (global->codec != NULL && global->codec->name != NULL)
758  global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
759  global->deadline != VPX_DL_REALTIME)
760  ? 2
761  : 1;
762 #else
763  global->passes = 1;
764 #endif
765  }
766 
767  if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
768  warn("Enforcing one-pass encoding in realtime mode\n");
769  global->passes = 1;
770  }
771 }
772 
773 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
774  struct stream_state *prev) {
775  struct stream_state *stream;
776 
777  stream = calloc(1, sizeof(*stream));
778  if (stream == NULL) {
779  fatal("Failed to allocate new stream.");
780  }
781 
782  if (prev) {
783  memcpy(stream, prev, sizeof(*stream));
784  stream->index++;
785  prev->next = stream;
786  } else {
787  vpx_codec_err_t res;
788 
789  /* Populate encoder configuration */
790  res = vpx_codec_enc_config_default(global->codec->codec_interface(),
791  &stream->config.cfg, global->usage);
792  if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
793 
794  /* Change the default timebase to a high enough value so that the
795  * encoder will always create strictly increasing timestamps.
796  */
797  stream->config.cfg.g_timebase.den = 1000;
798 
799  /* Never use the library's default resolution, require it be parsed
800  * from the file or set on the command line.
801  */
802  stream->config.cfg.g_w = 0;
803  stream->config.cfg.g_h = 0;
804 
805  /* Initialize remaining stream parameters */
806  stream->config.write_webm = 1;
807 #if CONFIG_WEBM_IO
808  stream->config.stereo_fmt = STEREO_FORMAT_MONO;
809  stream->webm_ctx.last_pts_ns = -1;
810  stream->webm_ctx.writer = NULL;
811  stream->webm_ctx.segment = NULL;
812 #endif
813 
814  /* Allows removal of the application version from the EBML tags */
815  stream->webm_ctx.debug = global->debug;
816 
817  /* Default lag_in_frames is 0 in realtime mode CBR mode*/
818  if (global->deadline == VPX_DL_REALTIME &&
819  stream->config.cfg.rc_end_usage == 1)
820  stream->config.cfg.g_lag_in_frames = 0;
821  }
822 
823  /* Output files must be specified for each stream */
824  stream->config.out_fn = NULL;
825 
826  stream->next = NULL;
827  return stream;
828 }
829 
830 static int parse_stream_params(struct VpxEncoderConfig *global,
831  struct stream_state *stream, char **argv) {
832  char **argi, **argj;
833  struct arg arg;
834  static const arg_def_t **ctrl_args = no_args;
835  static const int *ctrl_args_map = NULL;
836  struct stream_config *config = &stream->config;
837  int eos_mark_found = 0;
838 #if CONFIG_VP9_HIGHBITDEPTH
839  int test_16bit_internal = 0;
840 #endif
841 
842  // Handle codec specific options
843  if (0) {
844 #if CONFIG_VP8_ENCODER
845  } else if (strcmp(global->codec->name, "vp8") == 0) {
846  ctrl_args = vp8_args;
847  ctrl_args_map = vp8_arg_ctrl_map;
848 #endif
849 #if CONFIG_VP9_ENCODER
850  } else if (strcmp(global->codec->name, "vp9") == 0) {
851  ctrl_args = vp9_args;
852  ctrl_args_map = vp9_arg_ctrl_map;
853 #endif
854  }
855 
856  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
857  arg.argv_step = 1;
858 
859  /* Once we've found an end-of-stream marker (--) we want to continue
860  * shifting arguments but not consuming them.
861  */
862  if (eos_mark_found) {
863  argj++;
864  continue;
865  } else if (!strcmp(*argj, "--")) {
866  eos_mark_found = 1;
867  continue;
868  }
869 
870  if (arg_match(&arg, &outputfile, argi)) {
871  config->out_fn = arg.val;
872  } else if (arg_match(&arg, &fpf_name, argi)) {
873  config->stats_fn = arg.val;
874 #if CONFIG_FP_MB_STATS
875  } else if (arg_match(&arg, &fpmbf_name, argi)) {
876  config->fpmb_stats_fn = arg.val;
877 #endif
878  } else if (arg_match(&arg, &use_webm, argi)) {
879 #if CONFIG_WEBM_IO
880  config->write_webm = 1;
881 #else
882  die("Error: --webm specified but webm is disabled.");
883 #endif
884  } else if (arg_match(&arg, &use_ivf, argi)) {
885  config->write_webm = 0;
886  } else if (arg_match(&arg, &threads, argi)) {
887  config->cfg.g_threads = arg_parse_uint(&arg);
888  } else if (arg_match(&arg, &profile, argi)) {
889  config->cfg.g_profile = arg_parse_uint(&arg);
890  } else if (arg_match(&arg, &width, argi)) {
891  config->cfg.g_w = arg_parse_uint(&arg);
892  } else if (arg_match(&arg, &height, argi)) {
893  config->cfg.g_h = arg_parse_uint(&arg);
894 #if CONFIG_VP9_HIGHBITDEPTH
895  } else if (arg_match(&arg, &bitdeptharg, argi)) {
896  config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
897  } else if (arg_match(&arg, &inbitdeptharg, argi)) {
898  config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
899 #endif
900 #if CONFIG_WEBM_IO
901  } else if (arg_match(&arg, &stereo_mode, argi)) {
902  config->stereo_fmt = arg_parse_enum_or_int(&arg);
903 #endif
904  } else if (arg_match(&arg, &timebase, argi)) {
905  config->cfg.g_timebase = arg_parse_rational(&arg);
906  validate_positive_rational(arg.name, &config->cfg.g_timebase);
907  } else if (arg_match(&arg, &error_resilient, argi)) {
908  config->cfg.g_error_resilient = arg_parse_uint(&arg);
909  } else if (arg_match(&arg, &end_usage, argi)) {
910  config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
911  } else if (arg_match(&arg, &lag_in_frames, argi)) {
912  config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
913  if (global->deadline == VPX_DL_REALTIME &&
914  config->cfg.rc_end_usage == VPX_CBR &&
915  config->cfg.g_lag_in_frames != 0) {
916  warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
917  config->cfg.g_lag_in_frames = 0;
918  }
919  } else if (arg_match(&arg, &dropframe_thresh, argi)) {
920  config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
921  } else if (arg_match(&arg, &resize_allowed, argi)) {
922  config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
923  } else if (arg_match(&arg, &resize_width, argi)) {
924  config->cfg.rc_scaled_width = arg_parse_uint(&arg);
925  } else if (arg_match(&arg, &resize_height, argi)) {
926  config->cfg.rc_scaled_height = arg_parse_uint(&arg);
927  } else if (arg_match(&arg, &resize_up_thresh, argi)) {
928  config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
929  } else if (arg_match(&arg, &resize_down_thresh, argi)) {
930  config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
931  } else if (arg_match(&arg, &end_usage, argi)) {
932  config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
933  } else if (arg_match(&arg, &target_bitrate, argi)) {
934  config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
935  } else if (arg_match(&arg, &min_quantizer, argi)) {
936  config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
937  } else if (arg_match(&arg, &max_quantizer, argi)) {
938  config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
939  } else if (arg_match(&arg, &undershoot_pct, argi)) {
940  config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
941  } else if (arg_match(&arg, &overshoot_pct, argi)) {
942  config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
943  } else if (arg_match(&arg, &buf_sz, argi)) {
944  config->cfg.rc_buf_sz = arg_parse_uint(&arg);
945  } else if (arg_match(&arg, &buf_initial_sz, argi)) {
946  config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
947  } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
948  config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
949  } else if (arg_match(&arg, &bias_pct, argi)) {
950  config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
951  if (global->passes < 2)
952  warn("option %s ignored in one-pass mode.\n", arg.name);
953  } else if (arg_match(&arg, &minsection_pct, argi)) {
954  config->cfg.rc_2pass_vbr_minsection_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, &maxsection_pct, argi)) {
959  config->cfg.rc_2pass_vbr_maxsection_pct = 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, &corpus_complexity, argi)) {
964  config->cfg.rc_2pass_vbr_corpus_complexity = arg_parse_uint(&arg);
965 
966  if (global->passes < 2)
967  warn("option %s ignored in one-pass mode.\n", arg.name);
968  } else if (arg_match(&arg, &kf_min_dist, argi)) {
969  config->cfg.kf_min_dist = arg_parse_uint(&arg);
970  } else if (arg_match(&arg, &kf_max_dist, argi)) {
971  config->cfg.kf_max_dist = arg_parse_uint(&arg);
972  } else if (arg_match(&arg, &kf_disabled, argi)) {
973  config->cfg.kf_mode = VPX_KF_DISABLED;
974 #if CONFIG_VP9_HIGHBITDEPTH
975  } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
976  if (strcmp(global->codec->name, "vp9") == 0) {
977  test_16bit_internal = 1;
978  }
979 #endif
980  } else {
981  int i, match = 0;
982  for (i = 0; ctrl_args[i]; i++) {
983  if (arg_match(&arg, ctrl_args[i], argi)) {
984  int j;
985  match = 1;
986 
987  /* Point either to the next free element or the first
988  * instance of this control.
989  */
990  for (j = 0; j < config->arg_ctrl_cnt; j++)
991  if (ctrl_args_map != NULL &&
992  config->arg_ctrls[j][0] == ctrl_args_map[i])
993  break;
994 
995  /* Update/insert */
996  assert(j < (int)ARG_CTRL_CNT_MAX);
997  if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
998  config->arg_ctrls[j][0] = ctrl_args_map[i];
999  config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1000  if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1001  }
1002  }
1003  }
1004  if (!match) argj++;
1005  }
1006  }
1007 #if CONFIG_VP9_HIGHBITDEPTH
1008  if (strcmp(global->codec->name, "vp9") == 0) {
1009  config->use_16bit_internal =
1010  test_16bit_internal | (config->cfg.g_profile > 1);
1011  }
1012 #endif
1013  return eos_mark_found;
1014 }
1015 
1016 #define FOREACH_STREAM(func) \
1017  do { \
1018  struct stream_state *stream; \
1019  for (stream = streams; stream; stream = stream->next) { \
1020  func; \
1021  } \
1022  } while (0)
1023 
1024 static void validate_stream_config(const struct stream_state *stream,
1025  const struct VpxEncoderConfig *global) {
1026  const struct stream_state *streami;
1027  (void)global;
1028 
1029  if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1030  fatal(
1031  "Stream %d: Specify stream dimensions with --width (-w) "
1032  " and --height (-h)",
1033  stream->index);
1034 
1035  // Check that the codec bit depth is greater than the input bit depth.
1036  if (stream->config.cfg.g_input_bit_depth >
1037  (unsigned int)stream->config.cfg.g_bit_depth) {
1038  fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1039  stream->index, (int)stream->config.cfg.g_bit_depth,
1040  stream->config.cfg.g_input_bit_depth);
1041  }
1042 
1043  for (streami = stream; streami; streami = streami->next) {
1044  /* All streams require output files */
1045  if (!streami->config.out_fn)
1046  fatal("Stream %d: Output file is required (specify with -o)",
1047  streami->index);
1048 
1049  /* Check for two streams outputting to the same file */
1050  if (streami != stream) {
1051  const char *a = stream->config.out_fn;
1052  const char *b = streami->config.out_fn;
1053  if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1054  fatal("Stream %d: duplicate output file (from stream %d)",
1055  streami->index, stream->index);
1056  }
1057 
1058  /* Check for two streams sharing a stats file. */
1059  if (streami != stream) {
1060  const char *a = stream->config.stats_fn;
1061  const char *b = streami->config.stats_fn;
1062  if (a && b && !strcmp(a, b))
1063  fatal("Stream %d: duplicate stats file (from stream %d)",
1064  streami->index, stream->index);
1065  }
1066 
1067 #if CONFIG_FP_MB_STATS
1068  /* Check for two streams sharing a mb stats file. */
1069  if (streami != stream) {
1070  const char *a = stream->config.fpmb_stats_fn;
1071  const char *b = streami->config.fpmb_stats_fn;
1072  if (a && b && !strcmp(a, b))
1073  fatal("Stream %d: duplicate mb stats file (from stream %d)",
1074  streami->index, stream->index);
1075  }
1076 #endif
1077  }
1078 }
1079 
1080 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1081  unsigned int h) {
1082  if (!stream->config.cfg.g_w) {
1083  if (!stream->config.cfg.g_h)
1084  stream->config.cfg.g_w = w;
1085  else
1086  stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1087  }
1088  if (!stream->config.cfg.g_h) {
1089  stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1090  }
1091 }
1092 
1093 static const char *file_type_to_string(enum VideoFileType t) {
1094  switch (t) {
1095  case FILE_TYPE_RAW: return "RAW";
1096  case FILE_TYPE_Y4M: return "Y4M";
1097  default: return "Other";
1098  }
1099 }
1100 
1101 static const char *image_format_to_string(vpx_img_fmt_t f) {
1102  switch (f) {
1103  case VPX_IMG_FMT_I420: return "I420";
1104  case VPX_IMG_FMT_I422: return "I422";
1105  case VPX_IMG_FMT_I444: return "I444";
1106  case VPX_IMG_FMT_I440: return "I440";
1107  case VPX_IMG_FMT_YV12: return "YV12";
1108  case VPX_IMG_FMT_I42016: return "I42016";
1109  case VPX_IMG_FMT_I42216: return "I42216";
1110  case VPX_IMG_FMT_I44416: return "I44416";
1111  case VPX_IMG_FMT_I44016: return "I44016";
1112  default: return "Other";
1113  }
1114 }
1115 
1116 static void show_stream_config(struct stream_state *stream,
1117  struct VpxEncoderConfig *global,
1118  struct VpxInputContext *input) {
1119 #define SHOW(field) \
1120  fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1121 
1122  if (stream->index == 0) {
1123  fprintf(stderr, "Codec: %s\n",
1124  vpx_codec_iface_name(global->codec->codec_interface()));
1125  fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1126  input->filename, file_type_to_string(input->file_type),
1127  image_format_to_string(input->fmt));
1128  }
1129  if (stream->next || stream->index)
1130  fprintf(stderr, "\nStream Index: %d\n", stream->index);
1131  fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1132  fprintf(stderr, "Encoder parameters:\n");
1133 
1134  SHOW(g_usage);
1135  SHOW(g_threads);
1136  SHOW(g_profile);
1137  SHOW(g_w);
1138  SHOW(g_h);
1139  SHOW(g_bit_depth);
1140  SHOW(g_input_bit_depth);
1141  SHOW(g_timebase.num);
1142  SHOW(g_timebase.den);
1143  SHOW(g_error_resilient);
1144  SHOW(g_pass);
1145  SHOW(g_lag_in_frames);
1146  SHOW(rc_dropframe_thresh);
1147  SHOW(rc_resize_allowed);
1148  SHOW(rc_scaled_width);
1149  SHOW(rc_scaled_height);
1150  SHOW(rc_resize_up_thresh);
1151  SHOW(rc_resize_down_thresh);
1152  SHOW(rc_end_usage);
1153  SHOW(rc_target_bitrate);
1154  SHOW(rc_min_quantizer);
1155  SHOW(rc_max_quantizer);
1156  SHOW(rc_undershoot_pct);
1157  SHOW(rc_overshoot_pct);
1158  SHOW(rc_buf_sz);
1159  SHOW(rc_buf_initial_sz);
1160  SHOW(rc_buf_optimal_sz);
1161  SHOW(rc_2pass_vbr_bias_pct);
1162  SHOW(rc_2pass_vbr_minsection_pct);
1163  SHOW(rc_2pass_vbr_maxsection_pct);
1164  SHOW(rc_2pass_vbr_corpus_complexity);
1165  SHOW(kf_mode);
1166  SHOW(kf_min_dist);
1167  SHOW(kf_max_dist);
1168 }
1169 
1170 static void open_output_file(struct stream_state *stream,
1171  struct VpxEncoderConfig *global,
1172  const struct VpxRational *pixel_aspect_ratio) {
1173  const char *fn = stream->config.out_fn;
1174  const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1175 
1176  if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1177 
1178  stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1179 
1180  if (!stream->file) fatal("Failed to open output file");
1181 
1182  if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1183  fatal("WebM output to pipes not supported.");
1184 
1185 #if CONFIG_WEBM_IO
1186  if (stream->config.write_webm) {
1187  stream->webm_ctx.stream = stream->file;
1188  write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1189  global->codec->fourcc, pixel_aspect_ratio);
1190  }
1191 #else
1192  (void)pixel_aspect_ratio;
1193 #endif
1194 
1195  if (!stream->config.write_webm) {
1196  ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1197  }
1198 }
1199 
1200 static void close_output_file(struct stream_state *stream,
1201  unsigned int fourcc) {
1202  const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1203 
1204  if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1205 
1206 #if CONFIG_WEBM_IO
1207  if (stream->config.write_webm) {
1208  write_webm_file_footer(&stream->webm_ctx);
1209  }
1210 #endif
1211 
1212  if (!stream->config.write_webm) {
1213  if (!fseek(stream->file, 0, SEEK_SET))
1214  ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1215  stream->frames_out);
1216  }
1217 
1218  fclose(stream->file);
1219 }
1220 
1221 static void setup_pass(struct stream_state *stream,
1222  struct VpxEncoderConfig *global, int pass) {
1223  if (stream->config.stats_fn) {
1224  if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1225  fatal("Failed to open statistics store");
1226  } else {
1227  if (!stats_open_mem(&stream->stats, pass))
1228  fatal("Failed to open statistics store");
1229  }
1230 
1231 #if CONFIG_FP_MB_STATS
1232  if (stream->config.fpmb_stats_fn) {
1233  if (!stats_open_file(&stream->fpmb_stats, stream->config.fpmb_stats_fn,
1234  pass))
1235  fatal("Failed to open mb statistics store");
1236  } else {
1237  if (!stats_open_mem(&stream->fpmb_stats, pass))
1238  fatal("Failed to open mb statistics store");
1239  }
1240 #endif
1241 
1242  stream->config.cfg.g_pass = global->passes == 2
1244  : VPX_RC_ONE_PASS;
1245  if (pass) {
1246  stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1247 #if CONFIG_FP_MB_STATS
1248  stream->config.cfg.rc_firstpass_mb_stats_in =
1249  stats_get(&stream->fpmb_stats);
1250 #endif
1251  }
1252 
1253  stream->cx_time = 0;
1254  stream->nbytes = 0;
1255  stream->frames_out = 0;
1256 }
1257 
1258 static void initialize_encoder(struct stream_state *stream,
1259  struct VpxEncoderConfig *global) {
1260  int i;
1261  int flags = 0;
1262 
1263  flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1264  flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1265 #if CONFIG_VP9_HIGHBITDEPTH
1266  flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1267 #endif
1268 
1269  /* Construct Encoder Context */
1270  vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1271  &stream->config.cfg, flags);
1272  ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1273 
1274  /* Note that we bypass the vpx_codec_control wrapper macro because
1275  * we're being clever to store the control IDs in an array. Real
1276  * applications will want to make use of the enumerations directly
1277  */
1278  for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1279  int ctrl = stream->config.arg_ctrls[i][0];
1280  int value = stream->config.arg_ctrls[i][1];
1281  if (vpx_codec_control_(&stream->encoder, ctrl, value))
1282  fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1283 
1284  ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1285  }
1286 
1287 #if CONFIG_DECODERS
1288  if (global->test_decode != TEST_DECODE_OFF) {
1289  const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1290  vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1291  }
1292 #endif
1293 }
1294 
1295 static void encode_frame(struct stream_state *stream,
1296  struct VpxEncoderConfig *global, struct vpx_image *img,
1297  unsigned int frames_in) {
1298  vpx_codec_pts_t frame_start, next_frame_start;
1299  struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1300  struct vpx_usec_timer timer;
1301 
1302  frame_start =
1303  (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1304  cfg->g_timebase.num / global->framerate.num;
1305  next_frame_start =
1306  (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1307  cfg->g_timebase.num / global->framerate.num;
1308 
1309 /* Scale if necessary */
1310 #if CONFIG_VP9_HIGHBITDEPTH
1311  if (img) {
1312  if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1313  (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1314  if (img->fmt != VPX_IMG_FMT_I42016) {
1315  fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1316  exit(EXIT_FAILURE);
1317  }
1318 #if CONFIG_LIBYUV
1319  if (!stream->img) {
1320  stream->img =
1321  vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1322  }
1323  I420Scale_16(
1324  (uint16_t *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1325  (uint16_t *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1326  (uint16_t *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1327  img->d_w, img->d_h, (uint16_t *)stream->img->planes[VPX_PLANE_Y],
1328  stream->img->stride[VPX_PLANE_Y] / 2,
1329  (uint16_t *)stream->img->planes[VPX_PLANE_U],
1330  stream->img->stride[VPX_PLANE_U] / 2,
1331  (uint16_t *)stream->img->planes[VPX_PLANE_V],
1332  stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1333  stream->img->d_h, kFilterBox);
1334  img = stream->img;
1335 #else
1336  stream->encoder.err = 1;
1337  ctx_exit_on_error(&stream->encoder,
1338  "Stream %d: Failed to encode frame.\n"
1339  "Scaling disabled in this configuration. \n"
1340  "To enable, configure with --enable-libyuv\n",
1341  stream->index);
1342 #endif
1343  }
1344  }
1345 #endif
1346  if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1347  if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1348  fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1349  exit(EXIT_FAILURE);
1350  }
1351 #if CONFIG_LIBYUV
1352  if (!stream->img)
1353  stream->img =
1354  vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1355  I420Scale(
1356  img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1357  img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1358  img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1359  stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1360  stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1361  stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1362  stream->img->d_w, stream->img->d_h, kFilterBox);
1363  img = stream->img;
1364 #else
1365  stream->encoder.err = 1;
1366  ctx_exit_on_error(&stream->encoder,
1367  "Stream %d: Failed to encode frame.\n"
1368  "Scaling disabled in this configuration. \n"
1369  "To enable, configure with --enable-libyuv\n",
1370  stream->index);
1371 #endif
1372  }
1373 
1374  vpx_usec_timer_start(&timer);
1375  vpx_codec_encode(&stream->encoder, img, frame_start,
1376  (unsigned long)(next_frame_start - frame_start), 0,
1377  global->deadline);
1378  vpx_usec_timer_mark(&timer);
1379  stream->cx_time += vpx_usec_timer_elapsed(&timer);
1380  ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1381  stream->index);
1382 }
1383 
1384 static void update_quantizer_histogram(struct stream_state *stream) {
1385  if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1386  int q;
1387 
1388  vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1389  ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1390  stream->counts[q]++;
1391  }
1392 }
1393 
1394 static void get_cx_data(struct stream_state *stream,
1395  struct VpxEncoderConfig *global, int *got_data) {
1396  const vpx_codec_cx_pkt_t *pkt;
1397  const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1398  vpx_codec_iter_t iter = NULL;
1399 
1400  *got_data = 0;
1401  while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1402  static size_t fsize = 0;
1403  static FileOffset ivf_header_pos = 0;
1404 
1405  switch (pkt->kind) {
1407  if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1408  stream->frames_out++;
1409  }
1410  if (!global->quiet)
1411  fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1412 
1413  update_rate_histogram(stream->rate_hist, cfg, pkt);
1414 #if CONFIG_WEBM_IO
1415  if (stream->config.write_webm) {
1416  write_webm_block(&stream->webm_ctx, cfg, pkt);
1417  }
1418 #endif
1419  if (!stream->config.write_webm) {
1420  if (pkt->data.frame.partition_id <= 0) {
1421  ivf_header_pos = ftello(stream->file);
1422  fsize = pkt->data.frame.sz;
1423 
1424  ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1425  } else {
1426  fsize += pkt->data.frame.sz;
1427 
1428  if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1429  const FileOffset currpos = ftello(stream->file);
1430  fseeko(stream->file, ivf_header_pos, SEEK_SET);
1431  ivf_write_frame_size(stream->file, fsize);
1432  fseeko(stream->file, currpos, SEEK_SET);
1433  }
1434  }
1435 
1436  (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1437  stream->file);
1438  }
1439  stream->nbytes += pkt->data.raw.sz;
1440 
1441  *got_data = 1;
1442 #if CONFIG_DECODERS
1443  if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1444  vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1445  (unsigned int)pkt->data.frame.sz, NULL, 0);
1446  if (stream->decoder.err) {
1447  warn_or_exit_on_error(&stream->decoder,
1448  global->test_decode == TEST_DECODE_FATAL,
1449  "Failed to decode frame %d in stream %d",
1450  stream->frames_out + 1, stream->index);
1451  stream->mismatch_seen = stream->frames_out + 1;
1452  }
1453  }
1454 #endif
1455  break;
1456  case VPX_CODEC_STATS_PKT:
1457  stream->frames_out++;
1458  stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1459  pkt->data.twopass_stats.sz);
1460  stream->nbytes += pkt->data.raw.sz;
1461  break;
1462 #if CONFIG_FP_MB_STATS
1464  stats_write(&stream->fpmb_stats, pkt->data.firstpass_mb_stats.buf,
1465  pkt->data.firstpass_mb_stats.sz);
1466  stream->nbytes += pkt->data.raw.sz;
1467  break;
1468 #endif
1469  case VPX_CODEC_PSNR_PKT:
1470 
1471  if (global->show_psnr) {
1472  int i;
1473 
1474  stream->psnr_sse_total += pkt->data.psnr.sse[0];
1475  stream->psnr_samples_total += pkt->data.psnr.samples[0];
1476  for (i = 0; i < 4; i++) {
1477  if (!global->quiet)
1478  fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1479  stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1480  }
1481  stream->psnr_count++;
1482  }
1483 
1484  break;
1485  default: break;
1486  }
1487  }
1488 }
1489 
1490 static void show_psnr(struct stream_state *stream, double peak) {
1491  int i;
1492  double ovpsnr;
1493 
1494  if (!stream->psnr_count) return;
1495 
1496  fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1497  ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1498  (double)stream->psnr_sse_total);
1499  fprintf(stderr, " %.3f", ovpsnr);
1500 
1501  for (i = 0; i < 4; i++) {
1502  fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1503  }
1504  fprintf(stderr, "\n");
1505 }
1506 
1507 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1508  return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1509 }
1510 
1511 static void test_decode(struct stream_state *stream,
1512  enum TestDecodeFatality fatal,
1513  const VpxInterface *codec) {
1514  vpx_image_t enc_img, dec_img;
1515 
1516  if (stream->mismatch_seen) return;
1517 
1518  /* Get the internal reference frame */
1519  if (strcmp(codec->name, "vp8") == 0) {
1520  struct vpx_ref_frame ref_enc, ref_dec;
1521  int width, height;
1522 
1523  width = (stream->config.cfg.g_w + 15) & ~15;
1524  height = (stream->config.cfg.g_h + 15) & ~15;
1525  vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1526  enc_img = ref_enc.img;
1527  vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1528  dec_img = ref_dec.img;
1529 
1530  ref_enc.frame_type = VP8_LAST_FRAME;
1531  ref_dec.frame_type = VP8_LAST_FRAME;
1532  vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1533  vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1534  } else {
1535  struct vp9_ref_frame ref_enc, ref_dec;
1536 
1537  ref_enc.idx = 0;
1538  ref_dec.idx = 0;
1539  vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1540  enc_img = ref_enc.img;
1541  vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1542  dec_img = ref_dec.img;
1543 #if CONFIG_VP9_HIGHBITDEPTH
1544  if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1545  (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1546  if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1547  vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1548  enc_img.d_w, enc_img.d_h, 16);
1549  vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1550  }
1551  if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1552  vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1553  dec_img.d_w, dec_img.d_h, 16);
1554  vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1555  }
1556  }
1557 #endif
1558  }
1559  ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1560  ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1561 
1562  if (!compare_img(&enc_img, &dec_img)) {
1563  int y[4], u[4], v[4];
1564 #if CONFIG_VP9_HIGHBITDEPTH
1565  if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1566  find_mismatch_high(&enc_img, &dec_img, y, u, v);
1567  } else {
1568  find_mismatch(&enc_img, &dec_img, y, u, v);
1569  }
1570 #else
1571  find_mismatch(&enc_img, &dec_img, y, u, v);
1572 #endif
1573  stream->decoder.err = 1;
1574  warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1575  "Stream %d: Encode/decode mismatch on frame %d at"
1576  " Y[%d, %d] {%d/%d},"
1577  " U[%d, %d] {%d/%d},"
1578  " V[%d, %d] {%d/%d}",
1579  stream->index, stream->frames_out, y[0], y[1], y[2],
1580  y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1581  stream->mismatch_seen = stream->frames_out;
1582  }
1583 
1584  vpx_img_free(&enc_img);
1585  vpx_img_free(&dec_img);
1586 }
1587 
1588 static void print_time(const char *label, int64_t etl) {
1589  int64_t hours;
1590  int64_t mins;
1591  int64_t secs;
1592 
1593  if (etl >= 0) {
1594  hours = etl / 3600;
1595  etl -= hours * 3600;
1596  mins = etl / 60;
1597  etl -= mins * 60;
1598  secs = etl;
1599 
1600  fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1601  hours, mins, secs);
1602  } else {
1603  fprintf(stderr, "[%3s unknown] ", label);
1604  }
1605 }
1606 
1607 int main(int argc, const char **argv_) {
1608  int pass;
1609  vpx_image_t raw;
1610 #if CONFIG_VP9_HIGHBITDEPTH
1611  vpx_image_t raw_shift;
1612  int allocated_raw_shift = 0;
1613  int use_16bit_internal = 0;
1614  int input_shift = 0;
1615 #endif
1616  int frame_avail, got_data;
1617 
1618  struct VpxInputContext input;
1619  struct VpxEncoderConfig global;
1620  struct stream_state *streams = NULL;
1621  char **argv, **argi;
1622  uint64_t cx_time = 0;
1623  int stream_cnt = 0;
1624  int res = 0;
1625 
1626  memset(&input, 0, sizeof(input));
1627  exec_name = argv_[0];
1628 
1629  /* Setup default input stream settings */
1630  input.framerate.numerator = 30;
1631  input.framerate.denominator = 1;
1632  input.only_i420 = 1;
1633  input.bit_depth = 0;
1634 
1635  /* First parse the global configuration values, because we want to apply
1636  * other parameters on top of the default configuration provided by the
1637  * codec.
1638  */
1639  argv = argv_dup(argc - 1, argv_ + 1);
1640  parse_global_config(&global, argv);
1641 
1642  if (argc < 3) usage_exit();
1643 
1644  switch (global.color_type) {
1645  case I420: input.fmt = VPX_IMG_FMT_I420; break;
1646  case I422: input.fmt = VPX_IMG_FMT_I422; break;
1647  case I444: input.fmt = VPX_IMG_FMT_I444; break;
1648  case I440: input.fmt = VPX_IMG_FMT_I440; break;
1649  case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1650  case NV12: input.fmt = VPX_IMG_FMT_NV12; break;
1651  }
1652 
1653  {
1654  /* Now parse each stream's parameters. Using a local scope here
1655  * due to the use of 'stream' as loop variable in FOREACH_STREAM
1656  * loops
1657  */
1658  struct stream_state *stream = NULL;
1659 
1660  do {
1661  stream = new_stream(&global, stream);
1662  stream_cnt++;
1663  if (!streams) streams = stream;
1664  } while (parse_stream_params(&global, stream, argv));
1665  }
1666 
1667  /* Check for unrecognized options */
1668  for (argi = argv; *argi; argi++)
1669  if (argi[0][0] == '-' && argi[0][1])
1670  die("Error: Unrecognized option %s\n", *argi);
1671 
1672  FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1673  &stream->config.cfg););
1674 
1675  /* Handle non-option arguments */
1676  input.filename = argv[0];
1677 
1678  if (!input.filename) {
1679  fprintf(stderr, "No input file specified!\n");
1680  usage_exit();
1681  }
1682 
1683  /* Decide if other chroma subsamplings than 4:2:0 are supported */
1684  if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1685 
1686  for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1687  int frames_in = 0, seen_frames = 0;
1688  int64_t estimated_time_left = -1;
1689  int64_t average_rate = -1;
1690  int64_t lagged_count = 0;
1691 
1692  open_input_file(&input);
1693 
1694  /* If the input file doesn't specify its w/h (raw files), try to get
1695  * the data from the first stream's configuration.
1696  */
1697  if (!input.width || !input.height) {
1698  FOREACH_STREAM({
1699  if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1700  input.width = stream->config.cfg.g_w;
1701  input.height = stream->config.cfg.g_h;
1702  break;
1703  }
1704  });
1705  }
1706 
1707  /* Update stream configurations from the input file's parameters */
1708  if (!input.width || !input.height)
1709  fatal(
1710  "Specify stream dimensions with --width (-w) "
1711  " and --height (-h)");
1712 
1713  /* If input file does not specify bit-depth but input-bit-depth parameter
1714  * exists, assume that to be the input bit-depth. However, if the
1715  * input-bit-depth paramter does not exist, assume the input bit-depth
1716  * to be the same as the codec bit-depth.
1717  */
1718  if (!input.bit_depth) {
1719  FOREACH_STREAM({
1720  if (stream->config.cfg.g_input_bit_depth)
1721  input.bit_depth = stream->config.cfg.g_input_bit_depth;
1722  else
1723  input.bit_depth = stream->config.cfg.g_input_bit_depth =
1724  (int)stream->config.cfg.g_bit_depth;
1725  });
1726  if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1727  } else {
1728  FOREACH_STREAM(
1729  { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
1730  }
1731 
1732  FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1733  FOREACH_STREAM(validate_stream_config(stream, &global));
1734 
1735  /* Ensure that --passes and --pass are consistent. If --pass is set and
1736  * --passes=2, ensure --fpf was set.
1737  */
1738  if (global.pass && global.passes == 2)
1739  FOREACH_STREAM({
1740  if (!stream->config.stats_fn)
1741  die("Stream %d: Must specify --fpf when --pass=%d"
1742  " and --passes=2\n",
1743  stream->index, global.pass);
1744  });
1745 
1746 #if !CONFIG_WEBM_IO
1747  FOREACH_STREAM({
1748  if (stream->config.write_webm) {
1749  stream->config.write_webm = 0;
1750  warn(
1751  "vpxenc was compiled without WebM container support."
1752  "Producing IVF output");
1753  }
1754  });
1755 #endif
1756 
1757  /* Use the frame rate from the file only if none was specified
1758  * on the command-line.
1759  */
1760  if (!global.have_framerate) {
1761  global.framerate.num = input.framerate.numerator;
1762  global.framerate.den = input.framerate.denominator;
1763  FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
1764  stream->config.cfg.g_timebase.num = global.framerate.den);
1765  }
1766 
1767  /* Show configuration */
1768  if (global.verbose && pass == 0)
1769  FOREACH_STREAM(show_stream_config(stream, &global, &input));
1770 
1771  if (pass == (global.pass ? global.pass - 1 : 0)) {
1772  if (input.file_type == FILE_TYPE_Y4M)
1773  /*The Y4M reader does its own allocation.
1774  Just initialize this here to avoid problems if we never read any
1775  frames.*/
1776  memset(&raw, 0, sizeof(raw));
1777  else
1778  vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1779 
1780  FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
1781  &stream->config.cfg, &global.framerate));
1782  }
1783 
1784  FOREACH_STREAM(setup_pass(stream, &global, pass));
1785  FOREACH_STREAM(
1786  open_output_file(stream, &global, &input.pixel_aspect_ratio));
1787  FOREACH_STREAM(initialize_encoder(stream, &global));
1788 
1789 #if CONFIG_VP9_HIGHBITDEPTH
1790  if (strcmp(global.codec->name, "vp9") == 0) {
1791  // Check to see if at least one stream uses 16 bit internal.
1792  // Currently assume that the bit_depths for all streams using
1793  // highbitdepth are the same.
1794  FOREACH_STREAM({
1795  if (stream->config.use_16bit_internal) {
1796  use_16bit_internal = 1;
1797  }
1798  if (stream->config.cfg.g_profile == 0) {
1799  input_shift = 0;
1800  } else {
1801  input_shift = (int)stream->config.cfg.g_bit_depth -
1802  stream->config.cfg.g_input_bit_depth;
1803  }
1804  });
1805  }
1806 #endif
1807 
1808  frame_avail = 1;
1809  got_data = 0;
1810 
1811  while (frame_avail || got_data) {
1812  struct vpx_usec_timer timer;
1813 
1814  if (!global.limit || frames_in < global.limit) {
1815  frame_avail = read_frame(&input, &raw);
1816 
1817  if (frame_avail) frames_in++;
1818  seen_frames =
1819  frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
1820 
1821  if (!global.quiet) {
1822  float fps = usec_to_fps(cx_time, seen_frames);
1823  fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1824 
1825  if (stream_cnt == 1)
1826  fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
1827  streams->frames_out, (int64_t)streams->nbytes);
1828  else
1829  fprintf(stderr, "frame %4d ", frames_in);
1830 
1831  fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
1832  cx_time > 9999999 ? cx_time / 1000 : cx_time,
1833  cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
1834  fps >= 1.0 ? "fps" : "fpm");
1835  print_time("ETA", estimated_time_left);
1836  }
1837 
1838  } else
1839  frame_avail = 0;
1840 
1841  if (frames_in > global.skip_frames) {
1842 #if CONFIG_VP9_HIGHBITDEPTH
1843  vpx_image_t *frame_to_encode;
1844  if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
1845  assert(use_16bit_internal);
1846  // Input bit depth and stream bit depth do not match, so up
1847  // shift frame to stream bit depth
1848  if (!allocated_raw_shift) {
1849  vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
1850  input.width, input.height, 32);
1851  allocated_raw_shift = 1;
1852  }
1853  vpx_img_upshift(&raw_shift, &raw, input_shift);
1854  frame_to_encode = &raw_shift;
1855  } else {
1856  frame_to_encode = &raw;
1857  }
1858  vpx_usec_timer_start(&timer);
1859  if (use_16bit_internal) {
1860  assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
1861  FOREACH_STREAM({
1862  if (stream->config.use_16bit_internal)
1863  encode_frame(stream, &global,
1864  frame_avail ? frame_to_encode : NULL, frames_in);
1865  else
1866  assert(0);
1867  });
1868  } else {
1869  assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
1870  FOREACH_STREAM(encode_frame(stream, &global,
1871  frame_avail ? frame_to_encode : NULL,
1872  frames_in));
1873  }
1874 #else
1875  vpx_usec_timer_start(&timer);
1876  FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
1877  frames_in));
1878 #endif
1879  vpx_usec_timer_mark(&timer);
1880  cx_time += vpx_usec_timer_elapsed(&timer);
1881 
1882  FOREACH_STREAM(update_quantizer_histogram(stream));
1883 
1884  got_data = 0;
1885  FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1886 
1887  if (!got_data && input.length && streams != NULL &&
1888  !streams->frames_out) {
1889  lagged_count = global.limit ? seen_frames : ftello(input.file);
1890  } else if (input.length) {
1891  int64_t remaining;
1892  int64_t rate;
1893 
1894  if (global.limit) {
1895  const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1896 
1897  rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1898  remaining = 1000 * (global.limit - global.skip_frames -
1899  seen_frames + lagged_count);
1900  } else {
1901  const int64_t input_pos = ftello(input.file);
1902  const int64_t input_pos_lagged = input_pos - lagged_count;
1903  const int64_t limit = input.length;
1904 
1905  rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1906  remaining = limit - input_pos + lagged_count;
1907  }
1908 
1909  average_rate =
1910  (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
1911  estimated_time_left = average_rate ? remaining / average_rate : -1;
1912  }
1913 
1914  if (got_data && global.test_decode != TEST_DECODE_OFF)
1915  FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1916  }
1917 
1918  fflush(stdout);
1919  if (!global.quiet) fprintf(stderr, "\033[K");
1920  }
1921 
1922  if (stream_cnt > 1) fprintf(stderr, "\n");
1923 
1924  if (!global.quiet) {
1925  FOREACH_STREAM(fprintf(
1926  stderr,
1927  "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64 "b/f %7" PRId64
1928  "b/s %7" PRId64 " %s (%.2f fps)\033[K\n",
1929  pass + 1, global.passes, frames_in, stream->frames_out,
1930  (int64_t)stream->nbytes,
1931  seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
1932  seen_frames
1933  ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
1934  global.framerate.den / seen_frames
1935  : 0,
1936  stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1937  stream->cx_time > 9999999 ? "ms" : "us",
1938  usec_to_fps(stream->cx_time, seen_frames)));
1939  }
1940 
1941  if (global.show_psnr) {
1942  if (global.codec->fourcc == VP9_FOURCC) {
1943  FOREACH_STREAM(
1944  show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
1945  } else {
1946  FOREACH_STREAM(show_psnr(stream, 255.0));
1947  }
1948  }
1949 
1950  FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1951 
1952  if (global.test_decode != TEST_DECODE_OFF) {
1953  FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1954  }
1955 
1956  close_input_file(&input);
1957 
1958  if (global.test_decode == TEST_DECODE_FATAL) {
1959  FOREACH_STREAM(res |= stream->mismatch_seen);
1960  }
1961  FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1962 
1963  FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1964 
1965 #if CONFIG_FP_MB_STATS
1966  FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
1967 #endif
1968 
1969  if (global.pass) break;
1970  }
1971 
1972  if (global.show_q_hist_buckets)
1973  FOREACH_STREAM(
1974  show_q_histogram(stream->counts, global.show_q_hist_buckets));
1975 
1976  if (global.show_rate_hist_buckets)
1977  FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
1978  global.show_rate_hist_buckets));
1979  FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1980 
1981 #if CONFIG_INTERNAL_STATS
1982  /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1983  * to match some existing utilities.
1984  */
1985  if (!(global.pass == 1 && global.passes == 2))
1986  FOREACH_STREAM({
1987  FILE *f = fopen("opsnr.stt", "a");
1988  if (stream->mismatch_seen) {
1989  fprintf(f, "First mismatch occurred in frame %d\n",
1990  stream->mismatch_seen);
1991  } else {
1992  fprintf(f, "No mismatch detected in recon buffers\n");
1993  }
1994  fclose(f);
1995  });
1996 #endif
1997 
1998 #if CONFIG_VP9_HIGHBITDEPTH
1999  if (allocated_raw_shift) vpx_img_free(&raw_shift);
2000 #endif
2001  vpx_img_free(&raw);
2002  free(argv);
2003  free(streams);
2004  return res ? EXIT_FAILURE : EXIT_SUCCESS;
2005 }
Rational Number.
Definition: vpx_encoder.h:218
Definition: vpx_image.h:50
vpx_fixed_buf_t twopass_stats
Definition: vpx_encoder.h:182
Codec control function to set encoder internal speed settings.
Definition: vp8cx.h:155
Definition: vpx_encoder.h:235
Image Descriptor.
Definition: vpx_image.h:72
Describes the decoder algorithm interface to applications.
Describes the encoder algorithm interface to applications.
const char * vpx_codec_iface_name(vpx_codec_iface_t *iface)
Return the name for a given interface.
Definition: vpx_image.h:57
Codec control function to enable temporal dependency model.
Definition: vp8cx.h:668
Definition: vpx_image.h:40
const char * vpx_codec_err_to_string(vpx_codec_err_t err)
Convert error number to printable string.
Codec control function to set content type.
Definition: vp8cx.h:463
struct vpx_rational g_timebase
Stream timebase units.
Definition: vpx_encoder.h:343
Definition: vpx_encoder.h:233
Codec control function to set noise sensitivity.
Definition: vp8cx.h:421
Definition: vpx_image.h:61
Definition: vpx_image.h:62
Definition: vpx_image.h:46
Definition: vpx_image.h:59
int den
Definition: vpx_encoder.h:220
vpx_codec_err_t vpx_codec_encode(vpx_codec_ctx_t *ctx, const vpx_image_t *img, vpx_codec_pts_t pts, unsigned long duration, vpx_enc_frame_flags_t flags, unsigned long deadline)
Encode a frame.
Definition: vpx_encoder.h:148
Provides definitions for using VP8 or VP9 within the vpx Decoder interface.
Encoder configuration structure.
Definition: vpx_encoder.h:268
Codec control function to set visual tuning.
Definition: vp8cx.h:232
Codec control function to set constrained / constant quality level.
Definition: vp8cx.h:242
Definition: vp8cx.h:226
Definition: vpx_encoder.h:150
Codec control function to set row level multi-threading.
Definition: vp8cx.h:570
#define VPX_PLANE_Y
Definition: vpx_image.h:96
Codec control function to set Max data rate for Intra frames.
Definition: vp8cx.h:257
#define VPX_CODEC_USE_HIGHBITDEPTH
Definition: vpx_encoder.h:90
Encoder output packet.
Definition: vpx_encoder.h:159
void * buf
Definition: vpx_encoder.h:97
Definition: vpx_encoder.h:226
Definition: vpx_encoder.h:227
Codec control function to set number of tile columns.
Definition: vp8cx.h:351
#define VPX_IMG_FMT_HIGHBITDEPTH
Definition: vpx_image.h:35
struct vpx_codec_cx_pkt::@1::@2 frame
Definition: vp8.h:48
Definition: vpx_image.h:49
vpx_image_t * vpx_img_alloc(vpx_image_t *img, vpx_img_fmt_t fmt, unsigned int d_w, unsigned int d_h, unsigned int align)
Open a descriptor, allocating storage for the underlying image.
Definition: vpx_image.h:42
unsigned int d_w
Definition: vpx_image.h:83
#define vpx_codec_dec_init(ctx, iface, cfg, flags)
Convenience macro for vpx_codec_dec_init_ver()
Definition: vpx_decoder.h:143
Codec control function to set target level.
Definition: vp8cx.h:562
unsigned int g_w
Width of the frame.
Definition: vpx_encoder.h:304
Definition: vpx_image.h:48
Codec control function to set adaptive quantization mode.
Definition: vp8cx.h:398
Codec control function to set color space info.
Definition: vp8cx.h:494
vpx_codec_err_t vpx_codec_decode(vpx_codec_ctx_t *ctx, const uint8_t *data, unsigned int data_sz, void *user_priv, long deadline)
Decode data.
unsigned int g_h
Height of the frame.
Definition: vpx_encoder.h:313
enum vpx_img_fmt vpx_img_fmt_t
List of supported image formats.
int stride[4]
Definition: vpx_image.h:101
enum vpx_codec_cx_pkt_kind kind
Definition: vpx_encoder.h:160
Codec control function to set lossless encoding mode.
Definition: vp8cx.h:327
vpx_fixed_buf_t raw
Definition: vpx_encoder.h:189
Codec control function to get last quantizer chosen by the encoder.
Definition: vp8cx.h:211
Definition: vpx_image.h:58
Boost percentage for Golden Frame in CBR mode.
Definition: vp8cx.h:601
void vpx_img_free(vpx_image_t *img)
Close an image descriptor.
#define VPX_CODEC_USE_OUTPUT_PARTITION
Make the encoder output one partition at a time.
Definition: vpx_encoder.h:89
vpx_img_fmt_t fmt
Definition: vpx_image.h:73
Definition: vpx_image.h:43
unsigned char * planes[4]
Definition: vpx_image.h:100
Definition: vpx_image.h:56
Codec control function to set the number of token partitions.
Definition: vp8cx.h:194
#define VPX_DL_REALTIME
deadline parameter analogous to VPx REALTIME mode.
Definition: vpx_encoder.h:830
int num
Definition: vpx_encoder.h:219
control function to set noise sensitivity
Definition: vp8cx.h:173
Definition: vpx_codec.h:223
Boost percentage for Golden Frame in CBR mode.
Definition: vp8cx.h:293
#define VPX_DL_BEST_QUALITY
deadline parameter analogous to VPx BEST QUALITY mode.
Definition: vpx_encoder.h:834
vpx_codec_err_t vpx_codec_enc_config_default(vpx_codec_iface_t *iface, vpx_codec_enc_cfg_t *cfg, unsigned int usage)
Get a default configuration.
Definition: vpx_encoder.h:232
enum vpx_enc_pass g_pass
Multi-pass Encoding Mode.
Definition: vpx_encoder.h:358
double psnr[4]
Definition: vpx_encoder.h:187
#define VPX_CODEC_USE_PSNR
Initialization-time Feature Enabling.
Definition: vpx_encoder.h:87
#define VPX_DL_GOOD_QUALITY
deadline parameter analogous to VPx GOOD QUALITY mode.
Definition: vpx_encoder.h:832
vpx_fixed_buf_t firstpass_mb_stats
Definition: vpx_encoder.h:183
const char * vpx_codec_error_detail(vpx_codec_ctx_t *ctx)
Retrieve detailed error information for codec context.
Provides definitions for using VP8 or VP9 encoder algorithm within the vpx Codec Interface.
#define vpx_codec_enc_init(ctx, iface, cfg, flags)
Convenience macro for vpx_codec_enc_init_ver()
Definition: vpx_encoder.h:741
Codec control function to set encoder screen content mode.
Definition: vp8cx.h:312
vpx_codec_err_t
Algorithm return codes.
Definition: vpx_codec.h:93
const vpx_codec_cx_pkt_t * vpx_codec_get_cx_data(vpx_codec_ctx_t *ctx, vpx_codec_iter_t *iter)
Encoded data iterator.
union vpx_codec_cx_pkt::@1 data
Codec control function to set the max no of frames to create arf.
Definition: vp8cx.h:217
VP9 specific reference frame data struct.
Definition: vp8.h:110
Definition: vpx_encoder.h:249
Definition: vpx_image.h:47
Codec control function to set the filter strength for the arf.
Definition: vp8cx.h:223
Codec control function to enable/disable periodic Q boost.
Definition: vp8cx.h:413
Definition: vpx_encoder.h:149
int64_t vpx_codec_pts_t
Time Stamp Type.
Definition: vpx_encoder.h:106
Definition: vpx_image.h:44
Codec control function to enable automatic use of arf frames.
Definition: vp8cx.h:164
vpx_codec_err_t vpx_codec_control_(vpx_codec_ctx_t *ctx, int ctrl_id,...)
Control algorithm.
Codec control function to set minimum interval between GF/ARF frames.
Definition: vp8cx.h:514
reference frame data struct
Definition: vp8.h:101
Codec control function to set minimum interval between GF/ARF frames.
Definition: vp8cx.h:522
Definition: vpx_encoder.h:234
int idx
Definition: vp8.h:111
#define vpx_codec_control(ctx, id, data)
vpx_codec_control wrapper macro
Definition: vpx_codec.h:407
vpx_codec_err_t vpx_codec_destroy(vpx_codec_ctx_t *ctx)
Destroy a codec instance.
Codec control function to enable frame parallel decoding feature.
Definition: vp8cx.h:385
unsigned int d_h
Definition: vpx_image.h:84
Definition: vpx_image.h:55
size_t sz
Definition: vpx_encoder.h:98
Codec control function to set max data rate for Inter frames.
Definition: vp8cx.h:278
Definition: vpx_codec.h:221
vpx_codec_err_t err
Definition: vpx_codec.h:203
Definition: vp8.h:55
Codec control function to set the threshold for MBs treated static.
Definition: vp8cx.h:188
const char * vpx_codec_error(vpx_codec_ctx_t *ctx)
Retrieve error synopsis for codec context.
Definition: vpx_image.h:60
Definition: vpx_image.h:45
Definition: vpx_codec.h:222
Codec control function to set number of tile rows.
Definition: vp8cx.h:371
const void * vpx_codec_iter_t
Iterator.
Definition: vpx_codec.h:190
Codec control function to set higher sharpness at the expense of a lower PSNR.
Definition: vp8cx.h:182
Definition: vpx_encoder.h:147
Codec control function to enable/disable special mode for altref adaptive quantization. You can use it with –aq-mode concurrently.
Definition: vp8cx.h:586
#define VPX_FRAME_IS_FRAGMENT
this is a fragment of the encoded frame
Definition: vpx_encoder.h:123
Definition: vpx_encoder.h:225
Codec context structure.
Definition: vpx_codec.h:200