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_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 static const arg_def_t limit =
118 ARG_DEF(NULL, "limit", 1, "Stop encoding after n input frames");
119 static const arg_def_t skip =
120 ARG_DEF(NULL, "skip", 1, "Skip the first n input frames");
121 static const arg_def_t deadline =
122 ARG_DEF("d", "deadline", 1, "Deadline per frame (usec)");
123 static const arg_def_t best_dl =
124 ARG_DEF(NULL, "best", 0, "Use Best Quality Deadline");
125 static const arg_def_t good_dl =
126 ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
127 static const arg_def_t rt_dl =
128 ARG_DEF(NULL, "rt", 0, "Use Realtime Quality Deadline");
129 static const arg_def_t quietarg =
130 ARG_DEF("q", "quiet", 0, "Do not print encode progress");
131 static const arg_def_t verbosearg =
132 ARG_DEF("v", "verbose", 0, "Show encoder parameters");
133 static const arg_def_t psnrarg =
134 ARG_DEF(NULL, "psnr", 0, "Show PSNR in status line");
135
136 static const struct arg_enum_list test_decode_enum[] = {
137 { "off", TEST_DECODE_OFF },
138 { "fatal", TEST_DECODE_FATAL },
139 { "warn", TEST_DECODE_WARN },
140 { NULL, 0 }
141 };
142 static const arg_def_t recontest = ARG_DEF_ENUM(
143 NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
144 static const arg_def_t framerate =
145 ARG_DEF(NULL, "fps", 1, "Stream frame rate (rate/scale)");
146 static const arg_def_t use_webm =
147 ARG_DEF(NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
148 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, "Output IVF");
149 static const arg_def_t out_part =
150 ARG_DEF("P", "output-partitions", 0,
151 "Makes encoder output partitions. Requires IVF output!");
152 static const arg_def_t q_hist_n =
153 ARG_DEF(NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
154 static const arg_def_t rate_hist_n =
155 ARG_DEF(NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
156 static const arg_def_t disable_warnings =
157 ARG_DEF(NULL, "disable-warnings", 0,
158 "Disable warnings about potentially incorrect encode settings.");
159 static const arg_def_t disable_warning_prompt =
160 ARG_DEF("y", "disable-warning-prompt", 0,
161 "Display warnings, but do not prompt user to continue.");
162
163 #if CONFIG_VP9_HIGHBITDEPTH
164 static const arg_def_t test16bitinternalarg = ARG_DEF(
165 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
166 #endif
167
168 static const arg_def_t *main_args[] = { &help,
169 &debugmode,
170 &outputfile,
171 &codecarg,
172 &passes,
173 &pass_arg,
174 &fpf_name,
175 &limit,
176 &skip,
177 &deadline,
178 &best_dl,
179 &good_dl,
180 &rt_dl,
181 &quietarg,
182 &verbosearg,
183 &psnrarg,
184 &use_webm,
185 &use_ivf,
186 &out_part,
187 &q_hist_n,
188 &rate_hist_n,
189 &disable_warnings,
190 &disable_warning_prompt,
191 &recontest,
192 NULL };
193
194 static const arg_def_t usage =
195 ARG_DEF("u", "usage", 1, "Usage profile number to use");
196 static const arg_def_t threads =
197 ARG_DEF("t", "threads", 1, "Max number of threads to use");
198 static const arg_def_t profile =
199 ARG_DEF(NULL, "profile", 1, "Bitstream profile number to use");
200 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
201 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
202 #if CONFIG_WEBM_IO
203 static const struct arg_enum_list stereo_mode_enum[] = {
204 { "mono", STEREO_FORMAT_MONO },
205 { "left-right", STEREO_FORMAT_LEFT_RIGHT },
206 { "bottom-top", STEREO_FORMAT_BOTTOM_TOP },
207 { "top-bottom", STEREO_FORMAT_TOP_BOTTOM },
208 { "right-left", STEREO_FORMAT_RIGHT_LEFT },
209 { NULL, 0 }
210 };
211 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
212 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
213 #endif
214 static const arg_def_t timebase = ARG_DEF(
215 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
216 static const arg_def_t error_resilient =
217 ARG_DEF(NULL, "error-resilient", 1, "Enable error resiliency features");
218 static const arg_def_t lag_in_frames =
219 ARG_DEF(NULL, "lag-in-frames", 1, "Max number of frames to lag");
220
221 static const arg_def_t *global_args[] = { &use_nv12,
222 &use_yv12,
223 &use_i420,
224 &use_i422,
225 &use_i444,
226 &use_i440,
227 &usage,
228 &threads,
229 &profile,
230 &width,
231 &height,
232 #if CONFIG_WEBM_IO
233 &stereo_mode,
234 #endif
235 &timebase,
236 &framerate,
237 &error_resilient,
238 #if CONFIG_VP9_HIGHBITDEPTH
239 &test16bitinternalarg,
240 #endif
241 &lag_in_frames,
242 NULL };
243
244 static const arg_def_t dropframe_thresh =
245 ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
246 static const arg_def_t resize_allowed =
247 ARG_DEF(NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
248 static const arg_def_t resize_width =
249 ARG_DEF(NULL, "resize-width", 1, "Width of encoded frame");
250 static const arg_def_t resize_height =
251 ARG_DEF(NULL, "resize-height", 1, "Height of encoded frame");
252 static const arg_def_t resize_up_thresh =
253 ARG_DEF(NULL, "resize-up", 1, "Upscale threshold (buf %)");
254 static const arg_def_t resize_down_thresh =
255 ARG_DEF(NULL, "resize-down", 1, "Downscale threshold (buf %)");
256 static const struct arg_enum_list end_usage_enum[] = { { "vbr", VPX_VBR },
257 { "cbr", VPX_CBR },
258 { "cq", VPX_CQ },
259 { "q", VPX_Q },
260 { NULL, 0 } };
261 static const arg_def_t end_usage =
262 ARG_DEF_ENUM(NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
263 static const arg_def_t target_bitrate =
264 ARG_DEF(NULL, "target-bitrate", 1, "Bitrate (kbps)");
265 static const arg_def_t min_quantizer =
266 ARG_DEF(NULL, "min-q", 1, "Minimum (best) quantizer");
267 static const arg_def_t max_quantizer =
268 ARG_DEF(NULL, "max-q", 1, "Maximum (worst) quantizer");
269 static const arg_def_t undershoot_pct =
270 ARG_DEF(NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
271 static const arg_def_t overshoot_pct =
272 ARG_DEF(NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
273 static const arg_def_t buf_sz =
274 ARG_DEF(NULL, "buf-sz", 1, "Client buffer size (ms)");
275 static const arg_def_t buf_initial_sz =
276 ARG_DEF(NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
277 static const arg_def_t buf_optimal_sz =
278 ARG_DEF(NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
279 static const arg_def_t *rc_args[] = {
280 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
281 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
282 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct,
283 &buf_sz, &buf_initial_sz, &buf_optimal_sz, NULL
284 };
285
286 #if CONFIG_VP9_ENCODER
287 static const arg_def_t use_vizier_rc_params =
288 ARG_DEF(NULL, "use-vizier-rc-params", 1, "Use vizier rc params");
289 static const arg_def_t active_wq_factor =
290 ARG_DEF(NULL, "active-wq-factor", 1, "Active worst quality factor");
291 static const arg_def_t err_per_mb_factor =
292 ARG_DEF(NULL, "err-per-mb-factor", 1, "Error per macroblock factor");
293 static const arg_def_t sr_default_decay_limit = ARG_DEF(
294 NULL, "sr-default-decay-limit", 1, "Second reference default decay limit");
295 static const arg_def_t sr_diff_factor =
296 ARG_DEF(NULL, "sr-diff-factor", 1, "Second reference diff factor");
297 static const arg_def_t kf_err_per_mb_factor = ARG_DEF(
298 NULL, "kf-err-per-mb-factor", 1, "Keyframe error per macroblock factor");
299 static const arg_def_t kf_frame_min_boost_factor =
300 ARG_DEF(NULL, "kf-frame-min-boost-factor", 1, "Keyframe min boost");
301 static const arg_def_t kf_frame_max_boost_first_factor =
302 ARG_DEF(NULL, "kf-frame-max-boost-first-factor", 1,
303 "Max keyframe boost adjustment factor for first frame");
304 static const arg_def_t kf_frame_max_boost_subs_factor =
305 ARG_DEF(NULL, "kf-frame-max-boost-subs-factor", 1,
306 "Max boost adjustment factor for subsequent KFs");
307 static const arg_def_t kf_max_total_boost_factor = ARG_DEF(
308 NULL, "kf-max-total-boost-factor", 1, "Keyframe max total boost factor");
309 static const arg_def_t gf_max_total_boost_factor =
310 ARG_DEF(NULL, "gf-max-total-boost-factor", 1,
311 "Golden frame max total boost factor");
312 static const arg_def_t gf_frame_max_boost_factor =
313 ARG_DEF(NULL, "gf-frame-max-boost-factor", 1,
314 "Golden frame max per frame boost factor");
315 static const arg_def_t zm_factor =
316 ARG_DEF(NULL, "zm-factor", 1, "Zero motion power factor");
317 static const arg_def_t rd_mult_inter_qp_fac =
318 ARG_DEF(NULL, "rd-mult-inter-qp-fac", 1,
319 "RD multiplier adjustment for inter frames");
320 static const arg_def_t rd_mult_arf_qp_fac =
321 ARG_DEF(NULL, "rd-mult-arf-qp-fac", 1,
322 "RD multiplier adjustment for alt-ref frames");
323 static const arg_def_t rd_mult_key_qp_fac = ARG_DEF(
324 NULL, "rd-mult-key-qp-fac", 1, "RD multiplier adjustment for key frames");
325 static const arg_def_t *vizier_rc_args[] = { &use_vizier_rc_params,
326 &active_wq_factor,
327 &err_per_mb_factor,
328 &sr_default_decay_limit,
329 &sr_diff_factor,
330 &kf_err_per_mb_factor,
331 &kf_frame_min_boost_factor,
332 &kf_frame_max_boost_first_factor,
333 &kf_frame_max_boost_subs_factor,
334 &kf_max_total_boost_factor,
335 &gf_max_total_boost_factor,
336 &gf_frame_max_boost_factor,
337 &zm_factor,
338 &rd_mult_inter_qp_fac,
339 &rd_mult_arf_qp_fac,
340 &rd_mult_key_qp_fac,
341 NULL };
342 #endif
343
344 static const arg_def_t bias_pct =
345 ARG_DEF(NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
346 static const arg_def_t minsection_pct =
347 ARG_DEF(NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
348 static const arg_def_t maxsection_pct =
349 ARG_DEF(NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
350 static const arg_def_t corpus_complexity =
351 ARG_DEF(NULL, "corpus-complexity", 1, "corpus vbr complexity midpoint");
352 static const arg_def_t *rc_twopass_args[] = { &bias_pct, &minsection_pct,
353 &maxsection_pct,
354 &corpus_complexity, NULL };
355
356 static const arg_def_t kf_min_dist =
357 ARG_DEF(NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
358 static const arg_def_t kf_max_dist =
359 ARG_DEF(NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
360 static const arg_def_t kf_disabled =
361 ARG_DEF(NULL, "disable-kf", 0, "Disable keyframe placement");
362 static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
363 NULL };
364
365 static const arg_def_t noise_sens =
366 ARG_DEF(NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
367 static const arg_def_t sharpness =
368 ARG_DEF(NULL, "sharpness", 1,
369 "Increase sharpness at the expense of lower PSNR. (0..7)");
370 static const arg_def_t static_thresh =
371 ARG_DEF(NULL, "static-thresh", 1, "Motion detection threshold");
372 static const arg_def_t arnr_maxframes =
373 ARG_DEF(NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
374 static const arg_def_t arnr_strength =
375 ARG_DEF(NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
376 static const arg_def_t arnr_type =
377 ARG_DEF(NULL, "arnr-type", 1, "AltRef filter type (1..3)");
378 static const struct arg_enum_list tuning_enum[] = { { "psnr", VP8_TUNE_PSNR },
379 { "ssim", VP8_TUNE_SSIM },
380 { NULL, 0 } };
381 static const arg_def_t tune_ssim =
382 ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
383 static const arg_def_t cq_level =
384 ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
385 static const arg_def_t max_intra_rate_pct =
386 ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
387 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
388 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
389
390 #if CONFIG_VP8_ENCODER
391 static const arg_def_t cpu_used_vp8 =
392 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
393 static const arg_def_t auto_altref_vp8 = ARG_DEF(
394 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames. (0..1)");
395 static const arg_def_t token_parts =
396 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
397 static const arg_def_t screen_content_mode =
398 ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
399 static const arg_def_t *vp8_args[] = { &cpu_used_vp8,
400 &auto_altref_vp8,
401 &noise_sens,
402 &sharpness,
403 &static_thresh,
404 &token_parts,
405 &arnr_maxframes,
406 &arnr_strength,
407 &arnr_type,
408 &tune_ssim,
409 &cq_level,
410 &max_intra_rate_pct,
411 &gf_cbr_boost_pct,
412 &screen_content_mode,
413 NULL };
414 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
415 VP8E_SET_ENABLEAUTOALTREF,
416 VP8E_SET_NOISE_SENSITIVITY,
417 VP8E_SET_SHARPNESS,
418 VP8E_SET_STATIC_THRESHOLD,
419 VP8E_SET_TOKEN_PARTITIONS,
420 VP8E_SET_ARNR_MAXFRAMES,
421 VP8E_SET_ARNR_STRENGTH,
422 VP8E_SET_ARNR_TYPE,
423 VP8E_SET_TUNING,
424 VP8E_SET_CQ_LEVEL,
425 VP8E_SET_MAX_INTRA_BITRATE_PCT,
426 VP8E_SET_GF_CBR_BOOST_PCT,
427 VP8E_SET_SCREEN_CONTENT_MODE,
428 0 };
429 #endif
430
431 #if CONFIG_VP9_ENCODER
432 static const arg_def_t cpu_used_vp9 =
433 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-9..9)");
434 static const arg_def_t auto_altref_vp9 = ARG_DEF(
435 NULL, "auto-alt-ref", 1,
436 "Enable automatic alt reference frames, 2+ enables multi-layer. (0..6)");
437 static const arg_def_t tile_cols =
438 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
439 static const arg_def_t tile_rows =
440 ARG_DEF(NULL, "tile-rows", 1,
441 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
442
443 static const arg_def_t enable_tpl_model =
444 ARG_DEF(NULL, "enable-tpl", 1, "Enable temporal dependency model");
445
446 static const arg_def_t lossless =
447 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
448 static const arg_def_t frame_parallel_decoding = ARG_DEF(
449 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
450 static const arg_def_t aq_mode = ARG_DEF(
451 NULL, "aq-mode", 1,
452 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
453 "3: cyclic refresh, 4: equator360)");
454 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
455 "Special adaptive quantization for "
456 "the alternate reference frames.");
457 static const arg_def_t frame_periodic_boost =
458 ARG_DEF(NULL, "frame-boost", 1,
459 "Enable frame periodic boost (0: off (default), 1: on)");
460 static const arg_def_t max_inter_rate_pct =
461 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
462 static const arg_def_t min_gf_interval = ARG_DEF(
463 NULL, "min-gf-interval", 1,
464 "min gf/arf frame interval (default 0, indicating in-built behavior)");
465 static const arg_def_t max_gf_interval = ARG_DEF(
466 NULL, "max-gf-interval", 1,
467 "max gf/arf frame interval (default 0, indicating in-built behavior)");
468
469 static const struct arg_enum_list color_space_enum[] = {
470 { "unknown", VPX_CS_UNKNOWN },
471 { "bt601", VPX_CS_BT_601 },
472 { "bt709", VPX_CS_BT_709 },
473 { "smpte170", VPX_CS_SMPTE_170 },
474 { "smpte240", VPX_CS_SMPTE_240 },
475 { "bt2020", VPX_CS_BT_2020 },
476 { "reserved", VPX_CS_RESERVED },
477 { "sRGB", VPX_CS_SRGB },
478 { NULL, 0 }
479 };
480
481 static const arg_def_t input_color_space =
482 ARG_DEF_ENUM(NULL, "color-space", 1,
483 "The color space of input content:", color_space_enum);
484
485 #if CONFIG_VP9_HIGHBITDEPTH
486 static const struct arg_enum_list bitdepth_enum[] = {
487 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
488 };
489
490 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
491 "b", "bit-depth", 1,
492 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
493 bitdepth_enum);
494 static const arg_def_t inbitdeptharg =
495 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
496 #endif
497
498 static const struct arg_enum_list tune_content_enum[] = {
499 { "default", VP9E_CONTENT_DEFAULT },
500 { "screen", VP9E_CONTENT_SCREEN },
501 { "film", VP9E_CONTENT_FILM },
502 { NULL, 0 }
503 };
504
505 static const arg_def_t tune_content = ARG_DEF_ENUM(
506 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
507
508 static const arg_def_t target_level = ARG_DEF(
509 NULL, "target-level", 1,
510 "Target level\n"
511 " 255: off (default)\n"
512 " 0: only keep level stats\n"
513 " 1: adaptively set alt-ref "
514 "distance and column tile limit based on picture size, and keep"
515 " level stats\n"
516 " 10: level 1.0 11: level 1.1 "
517 "... 62: level 6.2");
518
519 static const arg_def_t row_mt =
520 ARG_DEF(NULL, "row-mt", 1,
521 "Enable row based non-deterministic multi-threading in VP9");
522
523 static const arg_def_t disable_loopfilter =
524 ARG_DEF(NULL, "disable-loopfilter", 1,
525 "Control Loopfilter in VP9\n"
526 "0: Loopfilter on for all frames (default)\n"
527 "1: Loopfilter off for non reference frames\n"
528 "2: Loopfilter off for all frames");
529 #endif
530
531 #if CONFIG_VP9_ENCODER
532 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
533 &auto_altref_vp9,
534 &sharpness,
535 &static_thresh,
536 &tile_cols,
537 &tile_rows,
538 &enable_tpl_model,
539 &arnr_maxframes,
540 &arnr_strength,
541 &arnr_type,
542 &tune_ssim,
543 &cq_level,
544 &max_intra_rate_pct,
545 &max_inter_rate_pct,
546 &gf_cbr_boost_pct,
547 &lossless,
548 &frame_parallel_decoding,
549 &aq_mode,
550 &alt_ref_aq,
551 &frame_periodic_boost,
552 &noise_sens,
553 &tune_content,
554 &input_color_space,
555 &min_gf_interval,
556 &max_gf_interval,
557 &target_level,
558 &row_mt,
559 &disable_loopfilter,
560 // NOTE: The entries above have a corresponding entry in vp9_arg_ctrl_map. The
561 // entries below do not have a corresponding entry in vp9_arg_ctrl_map. They
562 // must be listed at the end of vp9_args.
563 #if CONFIG_VP9_HIGHBITDEPTH
564 &bitdeptharg,
565 &inbitdeptharg,
566 #endif // CONFIG_VP9_HIGHBITDEPTH
567 NULL };
568 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
569 VP8E_SET_ENABLEAUTOALTREF,
570 VP8E_SET_SHARPNESS,
571 VP8E_SET_STATIC_THRESHOLD,
572 VP9E_SET_TILE_COLUMNS,
573 VP9E_SET_TILE_ROWS,
574 VP9E_SET_TPL,
575 VP8E_SET_ARNR_MAXFRAMES,
576 VP8E_SET_ARNR_STRENGTH,
577 VP8E_SET_ARNR_TYPE,
578 VP8E_SET_TUNING,
579 VP8E_SET_CQ_LEVEL,
580 VP8E_SET_MAX_INTRA_BITRATE_PCT,
581 VP9E_SET_MAX_INTER_BITRATE_PCT,
582 VP9E_SET_GF_CBR_BOOST_PCT,
583 VP9E_SET_LOSSLESS,
584 VP9E_SET_FRAME_PARALLEL_DECODING,
585 VP9E_SET_AQ_MODE,
586 VP9E_SET_ALT_REF_AQ,
587 VP9E_SET_FRAME_PERIODIC_BOOST,
588 VP9E_SET_NOISE_SENSITIVITY,
589 VP9E_SET_TUNE_CONTENT,
590 VP9E_SET_COLOR_SPACE,
591 VP9E_SET_MIN_GF_INTERVAL,
592 VP9E_SET_MAX_GF_INTERVAL,
593 VP9E_SET_TARGET_LEVEL,
594 VP9E_SET_ROW_MT,
595 VP9E_SET_DISABLE_LOOPFILTER,
596 0 };
597 #endif
598
599 static const arg_def_t *no_args[] = { NULL };
600
show_help(FILE * fout,int shorthelp)601 static void show_help(FILE *fout, int shorthelp) {
602 int i;
603 const int num_encoder = get_vpx_encoder_count();
604
605 fprintf(fout, "Usage: %s <options> -o dst_filename src_filename \n",
606 exec_name);
607
608 if (shorthelp) {
609 fprintf(fout, "Use --help to see the full list of options.\n");
610 return;
611 }
612
613 fprintf(fout, "\nOptions:\n");
614 arg_show_usage(fout, main_args);
615 fprintf(fout, "\nEncoder Global Options:\n");
616 arg_show_usage(fout, global_args);
617 fprintf(fout, "\nRate Control Options:\n");
618 arg_show_usage(fout, rc_args);
619 fprintf(fout, "\nTwopass Rate Control Options:\n");
620 arg_show_usage(fout, rc_twopass_args);
621 fprintf(fout, "\nKeyframe Placement Options:\n");
622 arg_show_usage(fout, kf_args);
623 #if CONFIG_VP8_ENCODER
624 fprintf(fout, "\nVP8 Specific Options:\n");
625 arg_show_usage(fout, vp8_args);
626 #endif
627 #if CONFIG_VP9_ENCODER
628 fprintf(fout, "\nVP9 Specific Options:\n");
629 arg_show_usage(fout, vp9_args);
630 fprintf(fout, "\nVizier Rate Control Options:\n");
631 arg_show_usage(fout, vizier_rc_args);
632 #endif
633 fprintf(fout,
634 "\nStream timebase (--timebase):\n"
635 " The desired precision of timestamps in the output, expressed\n"
636 " in fractional seconds. Default is 1/1000.\n");
637 fprintf(fout, "\nIncluded encoders:\n\n");
638
639 for (i = 0; i < num_encoder; ++i) {
640 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
641 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
642 fprintf(fout, " %-6s - %s %s\n", encoder->name,
643 vpx_codec_iface_name(encoder->codec_interface()), defstr);
644 }
645 fprintf(fout, "\n ");
646 fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
647 }
648
usage_exit(void)649 void usage_exit(void) {
650 show_help(stderr, 1);
651 exit(EXIT_FAILURE);
652 }
653
654 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
655 #if CONFIG_VP9_ENCODER
656 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
657 #else
658 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
659 #endif
660
661 #if !CONFIG_WEBM_IO
662 typedef int stereo_format_t;
663 struct WebmOutputContext {
664 int debug;
665 };
666 #endif
667
668 /* Per-stream configuration */
669 struct stream_config {
670 struct vpx_codec_enc_cfg cfg;
671 const char *out_fn;
672 const char *stats_fn;
673 stereo_format_t stereo_fmt;
674 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
675 int arg_ctrl_cnt;
676 int write_webm;
677 #if CONFIG_VP9_HIGHBITDEPTH
678 // whether to use 16bit internal buffers
679 int use_16bit_internal;
680 #endif
681 };
682
683 struct stream_state {
684 int index;
685 struct stream_state *next;
686 struct stream_config config;
687 FILE *file;
688 struct rate_hist *rate_hist;
689 struct WebmOutputContext webm_ctx;
690 uint64_t psnr_sse_total;
691 uint64_t psnr_samples_total;
692 double psnr_totals[4];
693 int psnr_count;
694 int counts[64];
695 vpx_codec_ctx_t encoder;
696 unsigned int frames_out;
697 uint64_t cx_time;
698 size_t nbytes;
699 stats_io_t stats;
700 struct vpx_image *img;
701 vpx_codec_ctx_t decoder;
702 int mismatch_seen;
703 };
704
validate_positive_rational(const char * msg,struct vpx_rational * rat)705 static void validate_positive_rational(const char *msg,
706 struct vpx_rational *rat) {
707 if (rat->den < 0) {
708 rat->num *= -1;
709 rat->den *= -1;
710 }
711
712 if (rat->num < 0) die("Error: %s must be positive\n", msg);
713
714 if (!rat->den) die("Error: %s has zero denominator\n", msg);
715 }
716
parse_global_config(struct VpxEncoderConfig * global,char ** argv)717 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
718 char **argi, **argj;
719 struct arg arg;
720 const int num_encoder = get_vpx_encoder_count();
721
722 if (num_encoder < 1) die("Error: no valid encoder available\n");
723
724 /* Initialize default parameters */
725 memset(global, 0, sizeof(*global));
726 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
727 global->passes = 0;
728 global->color_type = I420;
729 /* Assign default deadline to good quality */
730 global->deadline = VPX_DL_GOOD_QUALITY;
731
732 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
733 arg.argv_step = 1;
734
735 if (arg_match(&arg, &help, argi)) {
736 show_help(stdout, 0);
737 exit(EXIT_SUCCESS);
738 } else if (arg_match(&arg, &codecarg, argi)) {
739 global->codec = get_vpx_encoder_by_name(arg.val);
740 if (!global->codec)
741 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
742 } else if (arg_match(&arg, &passes, argi)) {
743 global->passes = arg_parse_uint(&arg);
744
745 if (global->passes < 1 || global->passes > 2)
746 die("Error: Invalid number of passes (%d)\n", global->passes);
747 } else if (arg_match(&arg, &pass_arg, argi)) {
748 global->pass = arg_parse_uint(&arg);
749
750 if (global->pass < 1 || global->pass > 2)
751 die("Error: Invalid pass selected (%d)\n", global->pass);
752 } else if (arg_match(&arg, &usage, argi))
753 global->usage = arg_parse_uint(&arg);
754 else if (arg_match(&arg, &deadline, argi))
755 global->deadline = arg_parse_uint(&arg);
756 else if (arg_match(&arg, &best_dl, argi))
757 global->deadline = VPX_DL_BEST_QUALITY;
758 else if (arg_match(&arg, &good_dl, argi))
759 global->deadline = VPX_DL_GOOD_QUALITY;
760 else if (arg_match(&arg, &rt_dl, argi))
761 global->deadline = VPX_DL_REALTIME;
762 else if (arg_match(&arg, &use_yv12, argi))
763 global->color_type = YV12;
764 else if (arg_match(&arg, &use_nv12, argi))
765 global->color_type = NV12;
766 else if (arg_match(&arg, &use_i420, argi))
767 global->color_type = I420;
768 else if (arg_match(&arg, &use_i422, argi))
769 global->color_type = I422;
770 else if (arg_match(&arg, &use_i444, argi))
771 global->color_type = I444;
772 else if (arg_match(&arg, &use_i440, argi))
773 global->color_type = I440;
774 else if (arg_match(&arg, &quietarg, argi))
775 global->quiet = 1;
776 else if (arg_match(&arg, &verbosearg, argi))
777 global->verbose = 1;
778 else if (arg_match(&arg, &limit, argi))
779 global->limit = arg_parse_uint(&arg);
780 else if (arg_match(&arg, &skip, argi))
781 global->skip_frames = arg_parse_uint(&arg);
782 else if (arg_match(&arg, &psnrarg, argi))
783 global->show_psnr = 1;
784 else if (arg_match(&arg, &recontest, argi))
785 global->test_decode = arg_parse_enum_or_int(&arg);
786 else if (arg_match(&arg, &framerate, argi)) {
787 global->framerate = arg_parse_rational(&arg);
788 validate_positive_rational(arg.name, &global->framerate);
789 global->have_framerate = 1;
790 } else if (arg_match(&arg, &out_part, argi))
791 global->out_part = 1;
792 else if (arg_match(&arg, &debugmode, argi))
793 global->debug = 1;
794 else if (arg_match(&arg, &q_hist_n, argi))
795 global->show_q_hist_buckets = arg_parse_uint(&arg);
796 else if (arg_match(&arg, &rate_hist_n, argi))
797 global->show_rate_hist_buckets = arg_parse_uint(&arg);
798 else if (arg_match(&arg, &disable_warnings, argi))
799 global->disable_warnings = 1;
800 else if (arg_match(&arg, &disable_warning_prompt, argi))
801 global->disable_warning_prompt = 1;
802 else
803 argj++;
804 }
805
806 if (global->pass) {
807 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
808 if (global->pass > global->passes) {
809 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
810 global->pass);
811 global->passes = global->pass;
812 }
813 }
814 /* Validate global config */
815 if (global->passes == 0) {
816 #if CONFIG_VP9_ENCODER
817 // Make default VP9 passes = 2 until there is a better quality 1-pass
818 // encoder
819 if (global->codec != NULL && global->codec->name != NULL)
820 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
821 global->deadline != VPX_DL_REALTIME)
822 ? 2
823 : 1;
824 #else
825 global->passes = 1;
826 #endif
827 }
828
829 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
830 warn("Enforcing one-pass encoding in realtime mode\n");
831 global->passes = 1;
832 }
833 }
834
new_stream(struct VpxEncoderConfig * global,struct stream_state * prev)835 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
836 struct stream_state *prev) {
837 struct stream_state *stream;
838
839 stream = calloc(1, sizeof(*stream));
840 if (stream == NULL) {
841 fatal("Failed to allocate new stream.");
842 }
843
844 if (prev) {
845 memcpy(stream, prev, sizeof(*stream));
846 stream->index++;
847 prev->next = stream;
848 } else {
849 vpx_codec_err_t res;
850
851 /* Populate encoder configuration */
852 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
853 &stream->config.cfg, global->usage);
854 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
855
856 /* Change the default timebase to a high enough value so that the
857 * encoder will always create strictly increasing timestamps.
858 */
859 stream->config.cfg.g_timebase.den = 1000;
860
861 /* Never use the library's default resolution, require it be parsed
862 * from the file or set on the command line.
863 */
864 stream->config.cfg.g_w = 0;
865 stream->config.cfg.g_h = 0;
866
867 /* Initialize remaining stream parameters */
868 stream->config.write_webm = 1;
869 #if CONFIG_WEBM_IO
870 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
871 stream->webm_ctx.last_pts_ns = -1;
872 stream->webm_ctx.writer = NULL;
873 stream->webm_ctx.segment = NULL;
874 #endif
875
876 /* Allows removal of the application version from the EBML tags */
877 stream->webm_ctx.debug = global->debug;
878
879 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
880 if (global->deadline == VPX_DL_REALTIME &&
881 stream->config.cfg.rc_end_usage == 1)
882 stream->config.cfg.g_lag_in_frames = 0;
883 }
884
885 /* Output files must be specified for each stream */
886 stream->config.out_fn = NULL;
887
888 stream->next = NULL;
889 return stream;
890 }
891
parse_stream_params(struct VpxEncoderConfig * global,struct stream_state * stream,char ** argv)892 static int parse_stream_params(struct VpxEncoderConfig *global,
893 struct stream_state *stream, char **argv) {
894 char **argi, **argj;
895 struct arg arg;
896 static const arg_def_t **ctrl_args = no_args;
897 static const int *ctrl_args_map = NULL;
898 struct stream_config *config = &stream->config;
899 int eos_mark_found = 0;
900 #if CONFIG_VP9_HIGHBITDEPTH
901 int test_16bit_internal = 0;
902 #endif
903
904 // Handle codec specific options
905 if (0) {
906 #if CONFIG_VP8_ENCODER
907 } else if (strcmp(global->codec->name, "vp8") == 0) {
908 ctrl_args = vp8_args;
909 ctrl_args_map = vp8_arg_ctrl_map;
910 #endif
911 #if CONFIG_VP9_ENCODER
912 } else if (strcmp(global->codec->name, "vp9") == 0) {
913 ctrl_args = vp9_args;
914 ctrl_args_map = vp9_arg_ctrl_map;
915 #endif
916 }
917
918 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
919 arg.argv_step = 1;
920
921 /* Once we've found an end-of-stream marker (--) we want to continue
922 * shifting arguments but not consuming them.
923 */
924 if (eos_mark_found) {
925 argj++;
926 continue;
927 } else if (!strcmp(*argj, "--")) {
928 eos_mark_found = 1;
929 continue;
930 }
931
932 if (arg_match(&arg, &outputfile, argi)) {
933 config->out_fn = arg.val;
934 } else if (arg_match(&arg, &fpf_name, argi)) {
935 config->stats_fn = arg.val;
936 } else if (arg_match(&arg, &use_webm, argi)) {
937 #if CONFIG_WEBM_IO
938 config->write_webm = 1;
939 #else
940 die("Error: --webm specified but webm is disabled.");
941 #endif
942 } else if (arg_match(&arg, &use_ivf, argi)) {
943 config->write_webm = 0;
944 } else if (arg_match(&arg, &threads, argi)) {
945 config->cfg.g_threads = arg_parse_uint(&arg);
946 } else if (arg_match(&arg, &profile, argi)) {
947 config->cfg.g_profile = arg_parse_uint(&arg);
948 } else if (arg_match(&arg, &width, argi)) {
949 config->cfg.g_w = arg_parse_uint(&arg);
950 } else if (arg_match(&arg, &height, argi)) {
951 config->cfg.g_h = arg_parse_uint(&arg);
952 #if CONFIG_VP9_HIGHBITDEPTH
953 } else if (arg_match(&arg, &bitdeptharg, argi)) {
954 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
955 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
956 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
957 #endif
958 #if CONFIG_WEBM_IO
959 } else if (arg_match(&arg, &stereo_mode, argi)) {
960 config->stereo_fmt = arg_parse_enum_or_int(&arg);
961 #endif
962 } else if (arg_match(&arg, &timebase, argi)) {
963 config->cfg.g_timebase = arg_parse_rational(&arg);
964 validate_positive_rational(arg.name, &config->cfg.g_timebase);
965 } else if (arg_match(&arg, &error_resilient, argi)) {
966 config->cfg.g_error_resilient = arg_parse_uint(&arg);
967 } else if (arg_match(&arg, &end_usage, argi)) {
968 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
969 } else if (arg_match(&arg, &lag_in_frames, argi)) {
970 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
971 if (global->deadline == VPX_DL_REALTIME &&
972 config->cfg.rc_end_usage == VPX_CBR &&
973 config->cfg.g_lag_in_frames != 0) {
974 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
975 config->cfg.g_lag_in_frames = 0;
976 }
977 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
978 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
979 } else if (arg_match(&arg, &resize_allowed, argi)) {
980 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
981 } else if (arg_match(&arg, &resize_width, argi)) {
982 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
983 } else if (arg_match(&arg, &resize_height, argi)) {
984 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
985 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
986 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
987 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
988 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
989 } else if (arg_match(&arg, &end_usage, argi)) {
990 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
991 } else if (arg_match(&arg, &target_bitrate, argi)) {
992 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
993 } else if (arg_match(&arg, &min_quantizer, argi)) {
994 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
995 } else if (arg_match(&arg, &max_quantizer, argi)) {
996 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
997 } else if (arg_match(&arg, &undershoot_pct, argi)) {
998 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
999 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1000 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1001 } else if (arg_match(&arg, &buf_sz, argi)) {
1002 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1003 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1004 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1005 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1006 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1007 } else if (arg_match(&arg, &bias_pct, argi)) {
1008 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1009 if (global->passes < 2)
1010 warn("option %s ignored in one-pass mode.\n", arg.name);
1011 } else if (arg_match(&arg, &minsection_pct, argi)) {
1012 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1013
1014 if (global->passes < 2)
1015 warn("option %s ignored in one-pass mode.\n", arg.name);
1016 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1017 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1018
1019 if (global->passes < 2)
1020 warn("option %s ignored in one-pass mode.\n", arg.name);
1021 } else if (arg_match(&arg, &corpus_complexity, argi)) {
1022 config->cfg.rc_2pass_vbr_corpus_complexity = arg_parse_uint(&arg);
1023
1024 if (global->passes < 2)
1025 warn("option %s ignored in one-pass mode.\n", arg.name);
1026 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1027 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1028 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1029 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1030 } else if (arg_match(&arg, &kf_disabled, argi)) {
1031 config->cfg.kf_mode = VPX_KF_DISABLED;
1032 #if CONFIG_VP9_ENCODER
1033 } else if (arg_match(&arg, &use_vizier_rc_params, argi)) {
1034 config->cfg.use_vizier_rc_params = arg_parse_int(&arg);
1035 } else if (arg_match(&arg, &active_wq_factor, argi)) {
1036 config->cfg.active_wq_factor = arg_parse_rational(&arg);
1037 } else if (arg_match(&arg, &err_per_mb_factor, argi)) {
1038 config->cfg.err_per_mb_factor = arg_parse_rational(&arg);
1039 } else if (arg_match(&arg, &sr_default_decay_limit, argi)) {
1040 config->cfg.sr_default_decay_limit = arg_parse_rational(&arg);
1041 } else if (arg_match(&arg, &sr_diff_factor, argi)) {
1042 config->cfg.sr_diff_factor = arg_parse_rational(&arg);
1043 } else if (arg_match(&arg, &kf_err_per_mb_factor, argi)) {
1044 config->cfg.kf_err_per_mb_factor = arg_parse_rational(&arg);
1045 } else if (arg_match(&arg, &kf_frame_min_boost_factor, argi)) {
1046 config->cfg.kf_frame_min_boost_factor = arg_parse_rational(&arg);
1047 } else if (arg_match(&arg, &kf_frame_max_boost_first_factor, argi)) {
1048 config->cfg.kf_frame_max_boost_first_factor = arg_parse_rational(&arg);
1049 } else if (arg_match(&arg, &kf_frame_max_boost_subs_factor, argi)) {
1050 config->cfg.kf_frame_max_boost_subs_factor = arg_parse_rational(&arg);
1051 } else if (arg_match(&arg, &kf_max_total_boost_factor, argi)) {
1052 config->cfg.kf_max_total_boost_factor = arg_parse_rational(&arg);
1053 } else if (arg_match(&arg, &gf_max_total_boost_factor, argi)) {
1054 config->cfg.gf_max_total_boost_factor = arg_parse_rational(&arg);
1055 } else if (arg_match(&arg, &gf_frame_max_boost_factor, argi)) {
1056 config->cfg.gf_frame_max_boost_factor = arg_parse_rational(&arg);
1057 } else if (arg_match(&arg, &zm_factor, argi)) {
1058 config->cfg.zm_factor = arg_parse_rational(&arg);
1059 } else if (arg_match(&arg, &rd_mult_inter_qp_fac, argi)) {
1060 config->cfg.rd_mult_inter_qp_fac = arg_parse_rational(&arg);
1061 } else if (arg_match(&arg, &rd_mult_arf_qp_fac, argi)) {
1062 config->cfg.rd_mult_arf_qp_fac = arg_parse_rational(&arg);
1063 } else if (arg_match(&arg, &rd_mult_key_qp_fac, argi)) {
1064 config->cfg.rd_mult_key_qp_fac = arg_parse_rational(&arg);
1065 #endif
1066 #if CONFIG_VP9_HIGHBITDEPTH
1067 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1068 if (strcmp(global->codec->name, "vp9") == 0) {
1069 test_16bit_internal = 1;
1070 }
1071 #endif
1072 } else {
1073 int i, match = 0;
1074 for (i = 0; ctrl_args[i]; i++) {
1075 if (arg_match(&arg, ctrl_args[i], argi)) {
1076 int j;
1077 match = 1;
1078
1079 /* Point either to the next free element or the first
1080 * instance of this control.
1081 */
1082 for (j = 0; j < config->arg_ctrl_cnt; j++)
1083 if (ctrl_args_map != NULL &&
1084 config->arg_ctrls[j][0] == ctrl_args_map[i])
1085 break;
1086
1087 /* Update/insert */
1088 assert(j < (int)ARG_CTRL_CNT_MAX);
1089 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1090 config->arg_ctrls[j][0] = ctrl_args_map[i];
1091 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1092 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1093 }
1094 }
1095 }
1096 if (!match) argj++;
1097 }
1098 }
1099 #if CONFIG_VP9_HIGHBITDEPTH
1100 if (strcmp(global->codec->name, "vp9") == 0) {
1101 config->use_16bit_internal =
1102 test_16bit_internal | (config->cfg.g_profile > 1);
1103 }
1104 #endif
1105 return eos_mark_found;
1106 }
1107
1108 #define FOREACH_STREAM(func) \
1109 do { \
1110 struct stream_state *stream; \
1111 for (stream = streams; stream; stream = stream->next) { \
1112 func; \
1113 } \
1114 } while (0)
1115
validate_stream_config(const struct stream_state * stream,const struct VpxEncoderConfig * global)1116 static void validate_stream_config(const struct stream_state *stream,
1117 const struct VpxEncoderConfig *global) {
1118 const struct stream_state *streami;
1119 (void)global;
1120
1121 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1122 fatal(
1123 "Stream %d: Specify stream dimensions with --width (-w) "
1124 " and --height (-h)",
1125 stream->index);
1126
1127 // Check that the codec bit depth is greater than the input bit depth.
1128 if (stream->config.cfg.g_input_bit_depth >
1129 (unsigned int)stream->config.cfg.g_bit_depth) {
1130 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1131 stream->index, (int)stream->config.cfg.g_bit_depth,
1132 stream->config.cfg.g_input_bit_depth);
1133 }
1134
1135 for (streami = stream; streami; streami = streami->next) {
1136 /* All streams require output files */
1137 if (!streami->config.out_fn)
1138 fatal("Stream %d: Output file is required (specify with -o)",
1139 streami->index);
1140
1141 /* Check for two streams outputting to the same file */
1142 if (streami != stream) {
1143 const char *a = stream->config.out_fn;
1144 const char *b = streami->config.out_fn;
1145 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1146 fatal("Stream %d: duplicate output file (from stream %d)",
1147 streami->index, stream->index);
1148 }
1149
1150 /* Check for two streams sharing a stats file. */
1151 if (streami != stream) {
1152 const char *a = stream->config.stats_fn;
1153 const char *b = streami->config.stats_fn;
1154 if (a && b && !strcmp(a, b))
1155 fatal("Stream %d: duplicate stats file (from stream %d)",
1156 streami->index, stream->index);
1157 }
1158 }
1159 }
1160
set_stream_dimensions(struct stream_state * stream,unsigned int w,unsigned int h)1161 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1162 unsigned int h) {
1163 if (!stream->config.cfg.g_w) {
1164 if (!stream->config.cfg.g_h)
1165 stream->config.cfg.g_w = w;
1166 else
1167 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1168 }
1169 if (!stream->config.cfg.g_h) {
1170 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1171 }
1172 }
1173
file_type_to_string(enum VideoFileType t)1174 static const char *file_type_to_string(enum VideoFileType t) {
1175 switch (t) {
1176 case FILE_TYPE_RAW: return "RAW";
1177 case FILE_TYPE_Y4M: return "Y4M";
1178 default: return "Other";
1179 }
1180 }
1181
image_format_to_string(vpx_img_fmt_t f)1182 static const char *image_format_to_string(vpx_img_fmt_t f) {
1183 switch (f) {
1184 case VPX_IMG_FMT_I420: return "I420";
1185 case VPX_IMG_FMT_I422: return "I422";
1186 case VPX_IMG_FMT_I444: return "I444";
1187 case VPX_IMG_FMT_I440: return "I440";
1188 case VPX_IMG_FMT_YV12: return "YV12";
1189 case VPX_IMG_FMT_I42016: return "I42016";
1190 case VPX_IMG_FMT_I42216: return "I42216";
1191 case VPX_IMG_FMT_I44416: return "I44416";
1192 case VPX_IMG_FMT_I44016: return "I44016";
1193 default: return "Other";
1194 }
1195 }
1196
show_stream_config(struct stream_state * stream,struct VpxEncoderConfig * global,struct VpxInputContext * input)1197 static void show_stream_config(struct stream_state *stream,
1198 struct VpxEncoderConfig *global,
1199 struct VpxInputContext *input) {
1200 #define SHOW(field) \
1201 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1202
1203 if (stream->index == 0) {
1204 fprintf(stderr, "Codec: %s\n",
1205 vpx_codec_iface_name(global->codec->codec_interface()));
1206 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1207 input->filename, file_type_to_string(input->file_type),
1208 image_format_to_string(input->fmt));
1209 }
1210 if (stream->next || stream->index)
1211 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1212 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1213 fprintf(stderr, "Encoder parameters:\n");
1214
1215 SHOW(g_usage);
1216 SHOW(g_threads);
1217 SHOW(g_profile);
1218 SHOW(g_w);
1219 SHOW(g_h);
1220 SHOW(g_bit_depth);
1221 SHOW(g_input_bit_depth);
1222 SHOW(g_timebase.num);
1223 SHOW(g_timebase.den);
1224 SHOW(g_error_resilient);
1225 SHOW(g_pass);
1226 SHOW(g_lag_in_frames);
1227 SHOW(rc_dropframe_thresh);
1228 SHOW(rc_resize_allowed);
1229 SHOW(rc_scaled_width);
1230 SHOW(rc_scaled_height);
1231 SHOW(rc_resize_up_thresh);
1232 SHOW(rc_resize_down_thresh);
1233 SHOW(rc_end_usage);
1234 SHOW(rc_target_bitrate);
1235 SHOW(rc_min_quantizer);
1236 SHOW(rc_max_quantizer);
1237 SHOW(rc_undershoot_pct);
1238 SHOW(rc_overshoot_pct);
1239 SHOW(rc_buf_sz);
1240 SHOW(rc_buf_initial_sz);
1241 SHOW(rc_buf_optimal_sz);
1242 SHOW(rc_2pass_vbr_bias_pct);
1243 SHOW(rc_2pass_vbr_minsection_pct);
1244 SHOW(rc_2pass_vbr_maxsection_pct);
1245 SHOW(rc_2pass_vbr_corpus_complexity);
1246 SHOW(kf_mode);
1247 SHOW(kf_min_dist);
1248 SHOW(kf_max_dist);
1249 // Temporary use for debug
1250 SHOW(use_vizier_rc_params);
1251 SHOW(active_wq_factor.num);
1252 SHOW(active_wq_factor.den);
1253 }
1254
open_output_file(struct stream_state * stream,struct VpxEncoderConfig * global,const struct VpxRational * pixel_aspect_ratio)1255 static void open_output_file(struct stream_state *stream,
1256 struct VpxEncoderConfig *global,
1257 const struct VpxRational *pixel_aspect_ratio) {
1258 const char *fn = stream->config.out_fn;
1259 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1260
1261 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1262
1263 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1264
1265 if (!stream->file) fatal("Failed to open output file");
1266
1267 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1268 fatal("WebM output to pipes not supported.");
1269
1270 #if CONFIG_WEBM_IO
1271 if (stream->config.write_webm) {
1272 stream->webm_ctx.stream = stream->file;
1273 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1274 global->codec->fourcc, pixel_aspect_ratio);
1275 }
1276 #else
1277 (void)pixel_aspect_ratio;
1278 #endif
1279
1280 if (!stream->config.write_webm) {
1281 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1282 }
1283 }
1284
close_output_file(struct stream_state * stream,unsigned int fourcc)1285 static void close_output_file(struct stream_state *stream,
1286 unsigned int fourcc) {
1287 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1288
1289 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1290
1291 #if CONFIG_WEBM_IO
1292 if (stream->config.write_webm) {
1293 write_webm_file_footer(&stream->webm_ctx);
1294 }
1295 #endif
1296
1297 if (!stream->config.write_webm) {
1298 if (!fseek(stream->file, 0, SEEK_SET))
1299 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1300 stream->frames_out);
1301 }
1302
1303 fclose(stream->file);
1304 }
1305
setup_pass(struct stream_state * stream,struct VpxEncoderConfig * global,int pass)1306 static void setup_pass(struct stream_state *stream,
1307 struct VpxEncoderConfig *global, int pass) {
1308 if (stream->config.stats_fn) {
1309 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1310 fatal("Failed to open statistics store");
1311 } else {
1312 if (!stats_open_mem(&stream->stats, pass))
1313 fatal("Failed to open statistics store");
1314 }
1315
1316 stream->config.cfg.g_pass = global->passes == 2
1317 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1318 : VPX_RC_ONE_PASS;
1319 if (pass) {
1320 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1321 }
1322
1323 stream->cx_time = 0;
1324 stream->nbytes = 0;
1325 stream->frames_out = 0;
1326 }
1327
initialize_encoder(struct stream_state * stream,struct VpxEncoderConfig * global)1328 static void initialize_encoder(struct stream_state *stream,
1329 struct VpxEncoderConfig *global) {
1330 int i;
1331 int flags = 0;
1332
1333 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1334 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1335 #if CONFIG_VP9_HIGHBITDEPTH
1336 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1337 #endif
1338
1339 /* Construct Encoder Context */
1340 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1341 &stream->config.cfg, flags);
1342 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1343
1344 /* Note that we bypass the vpx_codec_control wrapper macro because
1345 * we're being clever to store the control IDs in an array. Real
1346 * applications will want to make use of the enumerations directly
1347 */
1348 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1349 int ctrl = stream->config.arg_ctrls[i][0];
1350 int value = stream->config.arg_ctrls[i][1];
1351 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1352 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1353
1354 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1355 }
1356
1357 #if CONFIG_DECODERS
1358 if (global->test_decode != TEST_DECODE_OFF) {
1359 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1360 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1361 }
1362 #endif
1363 }
1364
encode_frame(struct stream_state * stream,struct VpxEncoderConfig * global,struct vpx_image * img,unsigned int frames_in)1365 static void encode_frame(struct stream_state *stream,
1366 struct VpxEncoderConfig *global, struct vpx_image *img,
1367 unsigned int frames_in) {
1368 vpx_codec_pts_t frame_start, next_frame_start;
1369 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1370 struct vpx_usec_timer timer;
1371
1372 frame_start =
1373 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1374 cfg->g_timebase.num / global->framerate.num;
1375 next_frame_start =
1376 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1377 cfg->g_timebase.num / global->framerate.num;
1378
1379 /* Scale if necessary */
1380 #if CONFIG_VP9_HIGHBITDEPTH
1381 if (img) {
1382 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1383 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1384 if (img->fmt != VPX_IMG_FMT_I42016) {
1385 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1386 exit(EXIT_FAILURE);
1387 }
1388 #if CONFIG_LIBYUV
1389 if (!stream->img) {
1390 stream->img =
1391 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1392 }
1393 I420Scale_16(
1394 (uint16_t *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1395 (uint16_t *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1396 (uint16_t *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1397 img->d_w, img->d_h, (uint16_t *)stream->img->planes[VPX_PLANE_Y],
1398 stream->img->stride[VPX_PLANE_Y] / 2,
1399 (uint16_t *)stream->img->planes[VPX_PLANE_U],
1400 stream->img->stride[VPX_PLANE_U] / 2,
1401 (uint16_t *)stream->img->planes[VPX_PLANE_V],
1402 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1403 stream->img->d_h, kFilterBox);
1404 img = stream->img;
1405 #else
1406 stream->encoder.err = 1;
1407 ctx_exit_on_error(&stream->encoder,
1408 "Stream %d: Failed to encode frame.\n"
1409 "Scaling disabled in this configuration. \n"
1410 "To enable, configure with --enable-libyuv\n",
1411 stream->index);
1412 #endif
1413 }
1414 }
1415 #endif
1416 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1417 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1418 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1419 exit(EXIT_FAILURE);
1420 }
1421 #if CONFIG_LIBYUV
1422 if (!stream->img)
1423 stream->img =
1424 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1425 I420Scale(
1426 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1427 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1428 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1429 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1430 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1431 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1432 stream->img->d_w, stream->img->d_h, kFilterBox);
1433 img = stream->img;
1434 #else
1435 stream->encoder.err = 1;
1436 ctx_exit_on_error(&stream->encoder,
1437 "Stream %d: Failed to encode frame.\n"
1438 "Scaling disabled in this configuration. \n"
1439 "To enable, configure with --enable-libyuv\n",
1440 stream->index);
1441 #endif
1442 }
1443
1444 vpx_usec_timer_start(&timer);
1445 vpx_codec_encode(&stream->encoder, img, frame_start,
1446 (unsigned long)(next_frame_start - frame_start), 0,
1447 global->deadline);
1448 vpx_usec_timer_mark(&timer);
1449 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1450 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1451 stream->index);
1452 }
1453
update_quantizer_histogram(struct stream_state * stream)1454 static void update_quantizer_histogram(struct stream_state *stream) {
1455 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1456 int q;
1457
1458 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1459 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1460 stream->counts[q]++;
1461 }
1462 }
1463
get_cx_data(struct stream_state * stream,struct VpxEncoderConfig * global,int * got_data)1464 static void get_cx_data(struct stream_state *stream,
1465 struct VpxEncoderConfig *global, int *got_data) {
1466 const vpx_codec_cx_pkt_t *pkt;
1467 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1468 vpx_codec_iter_t iter = NULL;
1469
1470 *got_data = 0;
1471 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1472 static size_t fsize = 0;
1473 static FileOffset ivf_header_pos = 0;
1474
1475 switch (pkt->kind) {
1476 case VPX_CODEC_CX_FRAME_PKT:
1477 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1478 stream->frames_out++;
1479 }
1480 if (!global->quiet)
1481 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1482
1483 update_rate_histogram(stream->rate_hist, cfg, pkt);
1484 #if CONFIG_WEBM_IO
1485 if (stream->config.write_webm) {
1486 write_webm_block(&stream->webm_ctx, cfg, pkt);
1487 }
1488 #endif
1489 if (!stream->config.write_webm) {
1490 if (pkt->data.frame.partition_id <= 0) {
1491 ivf_header_pos = ftello(stream->file);
1492 fsize = pkt->data.frame.sz;
1493
1494 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1495 } else {
1496 fsize += pkt->data.frame.sz;
1497
1498 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1499 const FileOffset currpos = ftello(stream->file);
1500 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1501 ivf_write_frame_size(stream->file, fsize);
1502 fseeko(stream->file, currpos, SEEK_SET);
1503 }
1504 }
1505
1506 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1507 stream->file);
1508 }
1509 stream->nbytes += pkt->data.raw.sz;
1510
1511 *got_data = 1;
1512 #if CONFIG_DECODERS
1513 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1514 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1515 (unsigned int)pkt->data.frame.sz, NULL, 0);
1516 if (stream->decoder.err) {
1517 warn_or_exit_on_error(&stream->decoder,
1518 global->test_decode == TEST_DECODE_FATAL,
1519 "Failed to decode frame %d in stream %d",
1520 stream->frames_out + 1, stream->index);
1521 stream->mismatch_seen = stream->frames_out + 1;
1522 }
1523 }
1524 #endif
1525 break;
1526 case VPX_CODEC_STATS_PKT:
1527 stream->frames_out++;
1528 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1529 pkt->data.twopass_stats.sz);
1530 stream->nbytes += pkt->data.raw.sz;
1531 break;
1532 case VPX_CODEC_PSNR_PKT:
1533
1534 if (global->show_psnr) {
1535 int i;
1536
1537 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1538 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1539 for (i = 0; i < 4; i++) {
1540 if (!global->quiet)
1541 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1542 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1543 }
1544 stream->psnr_count++;
1545 }
1546
1547 break;
1548 default: break;
1549 }
1550 }
1551 }
1552
show_psnr(struct stream_state * stream,double peak)1553 static void show_psnr(struct stream_state *stream, double peak) {
1554 int i;
1555 double ovpsnr;
1556
1557 if (!stream->psnr_count) return;
1558
1559 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1560 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1561 (double)stream->psnr_sse_total);
1562 fprintf(stderr, " %.3f", ovpsnr);
1563
1564 for (i = 0; i < 4; i++) {
1565 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1566 }
1567 fprintf(stderr, "\n");
1568 }
1569
usec_to_fps(uint64_t usec,unsigned int frames)1570 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1571 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1572 }
1573
test_decode(struct stream_state * stream,enum TestDecodeFatality fatal,const VpxInterface * codec)1574 static void test_decode(struct stream_state *stream,
1575 enum TestDecodeFatality fatal,
1576 const VpxInterface *codec) {
1577 vpx_image_t enc_img, dec_img;
1578
1579 if (stream->mismatch_seen) return;
1580
1581 /* Get the internal reference frame */
1582 if (strcmp(codec->name, "vp8") == 0) {
1583 struct vpx_ref_frame ref_enc, ref_dec;
1584 int width, height;
1585
1586 width = (stream->config.cfg.g_w + 15) & ~15;
1587 height = (stream->config.cfg.g_h + 15) & ~15;
1588 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1589 enc_img = ref_enc.img;
1590 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1591 dec_img = ref_dec.img;
1592
1593 ref_enc.frame_type = VP8_LAST_FRAME;
1594 ref_dec.frame_type = VP8_LAST_FRAME;
1595 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1596 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1597 } else {
1598 struct vp9_ref_frame ref_enc, ref_dec;
1599
1600 ref_enc.idx = 0;
1601 ref_dec.idx = 0;
1602 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1603 enc_img = ref_enc.img;
1604 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1605 dec_img = ref_dec.img;
1606 #if CONFIG_VP9_HIGHBITDEPTH
1607 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1608 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1609 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1610 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1611 enc_img.d_w, enc_img.d_h, 16);
1612 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1613 }
1614 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1615 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1616 dec_img.d_w, dec_img.d_h, 16);
1617 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1618 }
1619 }
1620 #endif
1621 }
1622 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1623 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1624
1625 if (!compare_img(&enc_img, &dec_img)) {
1626 int y[4], u[4], v[4];
1627 #if CONFIG_VP9_HIGHBITDEPTH
1628 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1629 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1630 } else {
1631 find_mismatch(&enc_img, &dec_img, y, u, v);
1632 }
1633 #else
1634 find_mismatch(&enc_img, &dec_img, y, u, v);
1635 #endif
1636 stream->decoder.err = 1;
1637 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1638 "Stream %d: Encode/decode mismatch on frame %d at"
1639 " Y[%d, %d] {%d/%d},"
1640 " U[%d, %d] {%d/%d},"
1641 " V[%d, %d] {%d/%d}",
1642 stream->index, stream->frames_out, y[0], y[1], y[2],
1643 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1644 stream->mismatch_seen = stream->frames_out;
1645 }
1646
1647 vpx_img_free(&enc_img);
1648 vpx_img_free(&dec_img);
1649 }
1650
print_time(const char * label,int64_t etl)1651 static void print_time(const char *label, int64_t etl) {
1652 int64_t hours;
1653 int64_t mins;
1654 int64_t secs;
1655
1656 if (etl >= 0) {
1657 hours = etl / 3600;
1658 etl -= hours * 3600;
1659 mins = etl / 60;
1660 etl -= mins * 60;
1661 secs = etl;
1662
1663 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1664 hours, mins, secs);
1665 } else {
1666 fprintf(stderr, "[%3s unknown] ", label);
1667 }
1668 }
1669
main(int argc,const char ** argv_)1670 int main(int argc, const char **argv_) {
1671 int pass;
1672 vpx_image_t raw;
1673 #if CONFIG_VP9_HIGHBITDEPTH
1674 vpx_image_t raw_shift;
1675 int allocated_raw_shift = 0;
1676 int use_16bit_internal = 0;
1677 int input_shift = 0;
1678 #endif
1679 int frame_avail, got_data;
1680
1681 struct VpxInputContext input;
1682 struct VpxEncoderConfig global;
1683 struct stream_state *streams = NULL;
1684 char **argv, **argi;
1685 uint64_t cx_time = 0;
1686 int stream_cnt = 0;
1687 int res = 0;
1688
1689 memset(&input, 0, sizeof(input));
1690 memset(&raw, 0, sizeof(raw));
1691 exec_name = argv_[0];
1692
1693 /* Setup default input stream settings */
1694 input.framerate.numerator = 30;
1695 input.framerate.denominator = 1;
1696 input.only_i420 = 1;
1697 input.bit_depth = 0;
1698
1699 /* First parse the global configuration values, because we want to apply
1700 * other parameters on top of the default configuration provided by the
1701 * codec.
1702 */
1703 argv = argv_dup(argc - 1, argv_ + 1);
1704 parse_global_config(&global, argv);
1705
1706 if (argc < 3) usage_exit();
1707
1708 switch (global.color_type) {
1709 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1710 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1711 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1712 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1713 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1714 case NV12: input.fmt = VPX_IMG_FMT_NV12; break;
1715 }
1716
1717 {
1718 /* Now parse each stream's parameters. Using a local scope here
1719 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1720 * loops
1721 */
1722 struct stream_state *stream = NULL;
1723
1724 do {
1725 stream = new_stream(&global, stream);
1726 stream_cnt++;
1727 if (!streams) streams = stream;
1728 } while (parse_stream_params(&global, stream, argv));
1729 }
1730
1731 /* Check for unrecognized options */
1732 for (argi = argv; *argi; argi++)
1733 if (argi[0][0] == '-' && argi[0][1])
1734 die("Error: Unrecognized option %s\n", *argi);
1735
1736 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1737 &stream->config.cfg););
1738
1739 /* Handle non-option arguments */
1740 input.filename = argv[0];
1741
1742 if (!input.filename) {
1743 fprintf(stderr, "No input file specified!\n");
1744 usage_exit();
1745 }
1746
1747 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1748 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1749
1750 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1751 int frames_in = 0, seen_frames = 0;
1752 int64_t estimated_time_left = -1;
1753 int64_t average_rate = -1;
1754 int64_t lagged_count = 0;
1755
1756 open_input_file(&input);
1757
1758 /* If the input file doesn't specify its w/h (raw files), try to get
1759 * the data from the first stream's configuration.
1760 */
1761 if (!input.width || !input.height) {
1762 FOREACH_STREAM({
1763 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1764 input.width = stream->config.cfg.g_w;
1765 input.height = stream->config.cfg.g_h;
1766 break;
1767 }
1768 });
1769 }
1770
1771 /* Update stream configurations from the input file's parameters */
1772 if (!input.width || !input.height)
1773 fatal(
1774 "Specify stream dimensions with --width (-w) "
1775 " and --height (-h)");
1776
1777 /* If input file does not specify bit-depth but input-bit-depth parameter
1778 * exists, assume that to be the input bit-depth. However, if the
1779 * input-bit-depth paramter does not exist, assume the input bit-depth
1780 * to be the same as the codec bit-depth.
1781 */
1782 if (!input.bit_depth) {
1783 FOREACH_STREAM({
1784 if (stream->config.cfg.g_input_bit_depth)
1785 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1786 else
1787 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1788 (int)stream->config.cfg.g_bit_depth;
1789 });
1790 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1791 } else {
1792 FOREACH_STREAM(
1793 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
1794 }
1795
1796 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1797 FOREACH_STREAM(validate_stream_config(stream, &global));
1798
1799 /* Ensure that --passes and --pass are consistent. If --pass is set and
1800 * --passes=2, ensure --fpf was set.
1801 */
1802 if (global.pass && global.passes == 2)
1803 FOREACH_STREAM({
1804 if (!stream->config.stats_fn)
1805 die("Stream %d: Must specify --fpf when --pass=%d"
1806 " and --passes=2\n",
1807 stream->index, global.pass);
1808 });
1809
1810 #if !CONFIG_WEBM_IO
1811 FOREACH_STREAM({
1812 if (stream->config.write_webm) {
1813 stream->config.write_webm = 0;
1814 warn(
1815 "vpxenc was compiled without WebM container support."
1816 "Producing IVF output");
1817 }
1818 });
1819 #endif
1820
1821 /* Use the frame rate from the file only if none was specified
1822 * on the command-line.
1823 */
1824 if (!global.have_framerate) {
1825 global.framerate.num = input.framerate.numerator;
1826 global.framerate.den = input.framerate.denominator;
1827 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
1828 stream->config.cfg.g_timebase.num = global.framerate.den);
1829 }
1830
1831 /* Show configuration */
1832 if (global.verbose && pass == 0)
1833 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1834
1835 if (pass == (global.pass ? global.pass - 1 : 0)) {
1836 // The Y4M reader does its own allocation.
1837 if (input.file_type != FILE_TYPE_Y4M) {
1838 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1839 }
1840 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
1841 &stream->config.cfg, &global.framerate));
1842 }
1843
1844 FOREACH_STREAM(setup_pass(stream, &global, pass));
1845 FOREACH_STREAM(
1846 open_output_file(stream, &global, &input.pixel_aspect_ratio));
1847 FOREACH_STREAM(initialize_encoder(stream, &global));
1848
1849 #if CONFIG_VP9_HIGHBITDEPTH
1850 if (strcmp(global.codec->name, "vp9") == 0) {
1851 // Check to see if at least one stream uses 16 bit internal.
1852 // Currently assume that the bit_depths for all streams using
1853 // highbitdepth are the same.
1854 FOREACH_STREAM({
1855 if (stream->config.use_16bit_internal) {
1856 use_16bit_internal = 1;
1857 }
1858 if (stream->config.cfg.g_profile == 0) {
1859 input_shift = 0;
1860 } else {
1861 input_shift = (int)stream->config.cfg.g_bit_depth -
1862 stream->config.cfg.g_input_bit_depth;
1863 }
1864 });
1865 }
1866 #endif
1867
1868 frame_avail = 1;
1869 got_data = 0;
1870
1871 while (frame_avail || got_data) {
1872 struct vpx_usec_timer timer;
1873
1874 if (!global.limit || frames_in < global.limit) {
1875 frame_avail = read_frame(&input, &raw);
1876
1877 if (frame_avail) frames_in++;
1878 seen_frames =
1879 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
1880
1881 if (!global.quiet) {
1882 float fps = usec_to_fps(cx_time, seen_frames);
1883 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1884
1885 if (stream_cnt == 1)
1886 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
1887 streams->frames_out, (int64_t)streams->nbytes);
1888 else
1889 fprintf(stderr, "frame %4d ", frames_in);
1890
1891 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
1892 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1893 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
1894 fps >= 1.0 ? "fps" : "fpm");
1895 print_time("ETA", estimated_time_left);
1896 }
1897
1898 } else
1899 frame_avail = 0;
1900
1901 if (frames_in > global.skip_frames) {
1902 #if CONFIG_VP9_HIGHBITDEPTH
1903 vpx_image_t *frame_to_encode;
1904 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
1905 assert(use_16bit_internal);
1906 // Input bit depth and stream bit depth do not match, so up
1907 // shift frame to stream bit depth
1908 if (!allocated_raw_shift) {
1909 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
1910 input.width, input.height, 32);
1911 allocated_raw_shift = 1;
1912 }
1913 vpx_img_upshift(&raw_shift, &raw, input_shift);
1914 frame_to_encode = &raw_shift;
1915 } else {
1916 frame_to_encode = &raw;
1917 }
1918 vpx_usec_timer_start(&timer);
1919 if (use_16bit_internal) {
1920 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
1921 FOREACH_STREAM({
1922 if (stream->config.use_16bit_internal)
1923 encode_frame(stream, &global,
1924 frame_avail ? frame_to_encode : NULL, frames_in);
1925 else
1926 assert(0);
1927 });
1928 } else {
1929 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
1930 FOREACH_STREAM(encode_frame(stream, &global,
1931 frame_avail ? frame_to_encode : NULL,
1932 frames_in));
1933 }
1934 #else
1935 vpx_usec_timer_start(&timer);
1936 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
1937 frames_in));
1938 #endif
1939 vpx_usec_timer_mark(&timer);
1940 cx_time += vpx_usec_timer_elapsed(&timer);
1941
1942 FOREACH_STREAM(update_quantizer_histogram(stream));
1943
1944 got_data = 0;
1945 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1946
1947 if (!got_data && input.length && streams != NULL &&
1948 !streams->frames_out) {
1949 lagged_count = global.limit ? seen_frames : ftello(input.file);
1950 } else if (input.length) {
1951 int64_t remaining;
1952 int64_t rate;
1953
1954 if (global.limit) {
1955 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1956
1957 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1958 remaining = 1000 * (global.limit - global.skip_frames -
1959 seen_frames + lagged_count);
1960 } else {
1961 const int64_t input_pos = ftello(input.file);
1962 const int64_t input_pos_lagged = input_pos - lagged_count;
1963 const int64_t limit = input.length;
1964
1965 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1966 remaining = limit - input_pos + lagged_count;
1967 }
1968
1969 average_rate =
1970 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
1971 estimated_time_left = average_rate ? remaining / average_rate : -1;
1972 }
1973
1974 if (got_data && global.test_decode != TEST_DECODE_OFF)
1975 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1976 }
1977
1978 fflush(stdout);
1979 if (!global.quiet) fprintf(stderr, "\033[K");
1980 }
1981
1982 if (stream_cnt > 1) fprintf(stderr, "\n");
1983
1984 if (!global.quiet) {
1985 FOREACH_STREAM(fprintf(
1986 stderr,
1987 "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64 "b/f %7" PRId64
1988 "b/s %7" PRId64 " %s (%.2f fps)\033[K\n",
1989 pass + 1, global.passes, frames_in, stream->frames_out,
1990 (int64_t)stream->nbytes,
1991 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
1992 seen_frames
1993 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
1994 global.framerate.den / seen_frames
1995 : 0,
1996 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1997 stream->cx_time > 9999999 ? "ms" : "us",
1998 usec_to_fps(stream->cx_time, seen_frames)));
1999 }
2000
2001 if (global.show_psnr) {
2002 if (global.codec->fourcc == VP9_FOURCC) {
2003 FOREACH_STREAM(
2004 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2005 } else {
2006 FOREACH_STREAM(show_psnr(stream, 255.0));
2007 }
2008 }
2009
2010 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2011
2012 if (global.test_decode != TEST_DECODE_OFF) {
2013 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2014 }
2015
2016 close_input_file(&input);
2017
2018 if (global.test_decode == TEST_DECODE_FATAL) {
2019 FOREACH_STREAM(res |= stream->mismatch_seen);
2020 }
2021 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2022
2023 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2024
2025 if (global.pass) break;
2026 }
2027
2028 if (global.show_q_hist_buckets)
2029 FOREACH_STREAM(
2030 show_q_histogram(stream->counts, global.show_q_hist_buckets));
2031
2032 if (global.show_rate_hist_buckets)
2033 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2034 global.show_rate_hist_buckets));
2035 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2036
2037 #if CONFIG_INTERNAL_STATS
2038 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2039 * to match some existing utilities.
2040 */
2041 if (!(global.pass == 1 && global.passes == 2))
2042 FOREACH_STREAM({
2043 FILE *f = fopen("opsnr.stt", "a");
2044 if (stream->mismatch_seen) {
2045 fprintf(f, "First mismatch occurred in frame %d\n",
2046 stream->mismatch_seen);
2047 } else {
2048 fprintf(f, "No mismatch detected in recon buffers\n");
2049 }
2050 fclose(f);
2051 });
2052 #endif
2053
2054 #if CONFIG_VP9_HIGHBITDEPTH
2055 if (allocated_raw_shift) vpx_img_free(&raw_shift);
2056 #endif
2057 vpx_img_free(&raw);
2058 free(argv);
2059 free(streams);
2060 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2061 }
2062