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 VPX_TOOLS_FORMAT_PRINTF(3, 0) void warn_or_exit_on_errorv(
62 vpx_codec_ctx_t *ctx, int fatal, 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 VPX_TOOLS_FORMAT_PRINTF(2,
ctx_exit_on_error(vpx_codec_ctx_t * ctx,const char * s,...)76 3) void ctx_exit_on_error(vpx_codec_ctx_t *ctx,
77 const char *s, ...) {
78 va_list ap;
79
80 va_start(ap, s);
81 warn_or_exit_on_errorv(ctx, 1, s, ap);
82 va_end(ap);
83 }
84
warn_or_exit_on_error(vpx_codec_ctx_t * ctx,int fatal,const char * s,...)85 static VPX_TOOLS_FORMAT_PRINTF(3, 4) void warn_or_exit_on_error(
86 vpx_codec_ctx_t *ctx, int fatal, const char *s, ...) {
87 va_list ap;
88
89 va_start(ap, s);
90 warn_or_exit_on_errorv(ctx, fatal, s, ap);
91 va_end(ap);
92 }
93
94 static const arg_def_t help =
95 ARG_DEF(NULL, "help", 0, "Show usage options and exit");
96 static const arg_def_t debugmode =
97 ARG_DEF("D", "debug", 0, "Debug mode (makes output deterministic)");
98 static const arg_def_t outputfile =
99 ARG_DEF("o", "output", 1, "Output filename");
100 static const arg_def_t use_nv12 =
101 ARG_DEF(NULL, "nv12", 0, "Input file is NV12 ");
102 static const arg_def_t use_yv12 =
103 ARG_DEF(NULL, "yv12", 0, "Input file is YV12 ");
104 static const arg_def_t use_i420 =
105 ARG_DEF(NULL, "i420", 0, "Input file is I420 (default)");
106 static const arg_def_t use_i422 =
107 ARG_DEF(NULL, "i422", 0, "Input file is I422");
108 static const arg_def_t use_i444 =
109 ARG_DEF(NULL, "i444", 0, "Input file is I444");
110 static const arg_def_t use_i440 =
111 ARG_DEF(NULL, "i440", 0, "Input file is I440");
112 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, "Codec to use");
113 static const arg_def_t passes =
114 ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
115 static const arg_def_t pass_arg =
116 ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
117 static const arg_def_t fpf_name =
118 ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");
119 static const arg_def_t limit =
120 ARG_DEF(NULL, "limit", 1, "Stop encoding after n input frames");
121 static const arg_def_t skip =
122 ARG_DEF(NULL, "skip", 1, "Skip the first n input frames");
123 static const arg_def_t deadline =
124 ARG_DEF("d", "deadline", 1, "Deadline per frame (usec)");
125 static const arg_def_t best_dl =
126 ARG_DEF(NULL, "best", 0, "Use Best Quality Deadline");
127 static const arg_def_t good_dl =
128 ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
129 static const arg_def_t rt_dl =
130 ARG_DEF(NULL, "rt", 0, "Use Realtime Quality Deadline");
131 static const arg_def_t quietarg =
132 ARG_DEF("q", "quiet", 0, "Do not print encode progress");
133 static const arg_def_t verbosearg =
134 ARG_DEF("v", "verbose", 0, "Show encoder parameters");
135 static const arg_def_t psnrarg =
136 ARG_DEF(NULL, "psnr", 0, "Show PSNR in status line");
137
138 static const struct arg_enum_list test_decode_enum[] = {
139 { "off", TEST_DECODE_OFF },
140 { "fatal", TEST_DECODE_FATAL },
141 { "warn", TEST_DECODE_WARN },
142 { NULL, 0 }
143 };
144 static const arg_def_t recontest = ARG_DEF_ENUM(
145 NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
146 static const arg_def_t framerate =
147 ARG_DEF(NULL, "fps", 1, "Stream frame rate (rate/scale)");
148 static const arg_def_t use_webm =
149 ARG_DEF(NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
150 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, "Output IVF");
151 static const arg_def_t out_part =
152 ARG_DEF("P", "output-partitions", 0,
153 "Makes encoder output partitions. Requires IVF output!");
154 static const arg_def_t q_hist_n =
155 ARG_DEF(NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
156 static const arg_def_t rate_hist_n =
157 ARG_DEF(NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
158 static const arg_def_t disable_warnings =
159 ARG_DEF(NULL, "disable-warnings", 0,
160 "Disable warnings about potentially incorrect encode settings.");
161 static const arg_def_t disable_warning_prompt =
162 ARG_DEF("y", "disable-warning-prompt", 0,
163 "Display warnings, but do not prompt user to continue.");
164
165 #if CONFIG_VP9_HIGHBITDEPTH
166 static const arg_def_t test16bitinternalarg = ARG_DEF(
167 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
168 #endif
169
170 static const arg_def_t *main_args[] = { &help,
171 &debugmode,
172 &outputfile,
173 &codecarg,
174 &passes,
175 &pass_arg,
176 &fpf_name,
177 &limit,
178 &skip,
179 &deadline,
180 &best_dl,
181 &good_dl,
182 &rt_dl,
183 &quietarg,
184 &verbosearg,
185 &psnrarg,
186 &use_webm,
187 &use_ivf,
188 &out_part,
189 &q_hist_n,
190 &rate_hist_n,
191 &disable_warnings,
192 &disable_warning_prompt,
193 &recontest,
194 NULL };
195
196 static const arg_def_t usage =
197 ARG_DEF("u", "usage", 1, "Usage profile number to use");
198 static const arg_def_t threads =
199 ARG_DEF("t", "threads", 1, "Max number of threads to use");
200 static const arg_def_t profile =
201 ARG_DEF(NULL, "profile", 1, "Bitstream profile number to use");
202 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
203 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
204 #if CONFIG_WEBM_IO
205 static const struct arg_enum_list stereo_mode_enum[] = {
206 { "mono", STEREO_FORMAT_MONO },
207 { "left-right", STEREO_FORMAT_LEFT_RIGHT },
208 { "bottom-top", STEREO_FORMAT_BOTTOM_TOP },
209 { "top-bottom", STEREO_FORMAT_TOP_BOTTOM },
210 { "right-left", STEREO_FORMAT_RIGHT_LEFT },
211 { NULL, 0 }
212 };
213 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
214 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
215 #endif
216 static const arg_def_t timebase = ARG_DEF(
217 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
218 static const arg_def_t error_resilient =
219 ARG_DEF(NULL, "error-resilient", 1, "Enable error resiliency features");
220 static const arg_def_t lag_in_frames =
221 ARG_DEF(NULL, "lag-in-frames", 1, "Max number of frames to lag");
222
223 static const arg_def_t *global_args[] = { &use_nv12,
224 &use_yv12,
225 &use_i420,
226 &use_i422,
227 &use_i444,
228 &use_i440,
229 &usage,
230 &threads,
231 &profile,
232 &width,
233 &height,
234 #if CONFIG_WEBM_IO
235 &stereo_mode,
236 #endif
237 &timebase,
238 &framerate,
239 &error_resilient,
240 #if CONFIG_VP9_HIGHBITDEPTH
241 &test16bitinternalarg,
242 #endif
243 &lag_in_frames,
244 NULL };
245
246 static const arg_def_t dropframe_thresh =
247 ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
248 static const arg_def_t resize_allowed =
249 ARG_DEF(NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
250 static const arg_def_t resize_width =
251 ARG_DEF(NULL, "resize-width", 1, "Width of encoded frame");
252 static const arg_def_t resize_height =
253 ARG_DEF(NULL, "resize-height", 1, "Height of encoded frame");
254 static const arg_def_t resize_up_thresh =
255 ARG_DEF(NULL, "resize-up", 1, "Upscale threshold (buf %)");
256 static const arg_def_t resize_down_thresh =
257 ARG_DEF(NULL, "resize-down", 1, "Downscale threshold (buf %)");
258 static const struct arg_enum_list end_usage_enum[] = { { "vbr", VPX_VBR },
259 { "cbr", VPX_CBR },
260 { "cq", VPX_CQ },
261 { "q", VPX_Q },
262 { NULL, 0 } };
263 static const arg_def_t end_usage =
264 ARG_DEF_ENUM(NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
265 static const arg_def_t target_bitrate =
266 ARG_DEF(NULL, "target-bitrate", 1, "Bitrate (kbps)");
267 static const arg_def_t min_quantizer =
268 ARG_DEF(NULL, "min-q", 1, "Minimum (best) quantizer");
269 static const arg_def_t max_quantizer =
270 ARG_DEF(NULL, "max-q", 1, "Maximum (worst) quantizer");
271 static const arg_def_t undershoot_pct =
272 ARG_DEF(NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
273 static const arg_def_t overshoot_pct =
274 ARG_DEF(NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
275 static const arg_def_t buf_sz =
276 ARG_DEF(NULL, "buf-sz", 1, "Client buffer size (ms)");
277 static const arg_def_t buf_initial_sz =
278 ARG_DEF(NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
279 static const arg_def_t buf_optimal_sz =
280 ARG_DEF(NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
281 static const arg_def_t *rc_args[] = {
282 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
283 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
284 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct,
285 &buf_sz, &buf_initial_sz, &buf_optimal_sz, NULL
286 };
287
288 #if CONFIG_VP9_ENCODER
289 static const arg_def_t use_vizier_rc_params =
290 ARG_DEF(NULL, "use-vizier-rc-params", 1, "Use vizier rc params");
291 static const arg_def_t active_wq_factor =
292 ARG_DEF(NULL, "active-wq-factor", 1, "Active worst quality factor");
293 static const arg_def_t err_per_mb_factor =
294 ARG_DEF(NULL, "err-per-mb-factor", 1, "Error per macroblock factor");
295 static const arg_def_t sr_default_decay_limit = ARG_DEF(
296 NULL, "sr-default-decay-limit", 1, "Second reference default decay limit");
297 static const arg_def_t sr_diff_factor =
298 ARG_DEF(NULL, "sr-diff-factor", 1, "Second reference diff factor");
299 static const arg_def_t kf_err_per_mb_factor = ARG_DEF(
300 NULL, "kf-err-per-mb-factor", 1, "Keyframe error per macroblock factor");
301 static const arg_def_t kf_frame_min_boost_factor =
302 ARG_DEF(NULL, "kf-frame-min-boost-factor", 1, "Keyframe min boost");
303 static const arg_def_t kf_frame_max_boost_first_factor =
304 ARG_DEF(NULL, "kf-frame-max-boost-first-factor", 1,
305 "Max keyframe boost adjustment factor for first frame");
306 static const arg_def_t kf_frame_max_boost_subs_factor =
307 ARG_DEF(NULL, "kf-frame-max-boost-subs-factor", 1,
308 "Max boost adjustment factor for subsequent KFs");
309 static const arg_def_t kf_max_total_boost_factor = ARG_DEF(
310 NULL, "kf-max-total-boost-factor", 1, "Keyframe max total boost factor");
311 static const arg_def_t gf_max_total_boost_factor =
312 ARG_DEF(NULL, "gf-max-total-boost-factor", 1,
313 "Golden frame max total boost factor");
314 static const arg_def_t gf_frame_max_boost_factor =
315 ARG_DEF(NULL, "gf-frame-max-boost-factor", 1,
316 "Golden frame max per frame boost factor");
317 static const arg_def_t zm_factor =
318 ARG_DEF(NULL, "zm-factor", 1, "Zero motion power factor");
319 static const arg_def_t rd_mult_inter_qp_fac =
320 ARG_DEF(NULL, "rd-mult-inter-qp-fac", 1,
321 "RD multiplier adjustment for inter frames");
322 static const arg_def_t rd_mult_arf_qp_fac =
323 ARG_DEF(NULL, "rd-mult-arf-qp-fac", 1,
324 "RD multiplier adjustment for alt-ref frames");
325 static const arg_def_t rd_mult_key_qp_fac = ARG_DEF(
326 NULL, "rd-mult-key-qp-fac", 1, "RD multiplier adjustment for key frames");
327 static const arg_def_t *vizier_rc_args[] = { &use_vizier_rc_params,
328 &active_wq_factor,
329 &err_per_mb_factor,
330 &sr_default_decay_limit,
331 &sr_diff_factor,
332 &kf_err_per_mb_factor,
333 &kf_frame_min_boost_factor,
334 &kf_frame_max_boost_first_factor,
335 &kf_frame_max_boost_subs_factor,
336 &kf_max_total_boost_factor,
337 &gf_max_total_boost_factor,
338 &gf_frame_max_boost_factor,
339 &zm_factor,
340 &rd_mult_inter_qp_fac,
341 &rd_mult_arf_qp_fac,
342 &rd_mult_key_qp_fac,
343 NULL };
344 #endif
345
346 static const arg_def_t bias_pct =
347 ARG_DEF(NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
348 static const arg_def_t minsection_pct =
349 ARG_DEF(NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
350 static const arg_def_t maxsection_pct =
351 ARG_DEF(NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
352 static const arg_def_t corpus_complexity =
353 ARG_DEF(NULL, "corpus-complexity", 1, "corpus vbr complexity midpoint");
354 static const arg_def_t *rc_twopass_args[] = { &bias_pct, &minsection_pct,
355 &maxsection_pct,
356 &corpus_complexity, NULL };
357
358 static const arg_def_t kf_min_dist =
359 ARG_DEF(NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
360 static const arg_def_t kf_max_dist =
361 ARG_DEF(NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
362 static const arg_def_t kf_disabled =
363 ARG_DEF(NULL, "disable-kf", 0, "Disable keyframe placement");
364 static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
365 NULL };
366
367 static const arg_def_t noise_sens =
368 ARG_DEF(NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
369 static const arg_def_t sharpness =
370 ARG_DEF(NULL, "sharpness", 1,
371 "Increase sharpness at the expense of lower PSNR. (0..7)");
372 static const arg_def_t static_thresh =
373 ARG_DEF(NULL, "static-thresh", 1, "Motion detection threshold");
374 static const arg_def_t arnr_maxframes =
375 ARG_DEF(NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
376 static const arg_def_t arnr_strength =
377 ARG_DEF(NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
378 static const arg_def_t arnr_type =
379 ARG_DEF(NULL, "arnr-type", 1, "AltRef filter type (1..3)");
380 static const struct arg_enum_list tuning_enum[] = { { "psnr", VP8_TUNE_PSNR },
381 { "ssim", VP8_TUNE_SSIM },
382 { NULL, 0 } };
383 static const arg_def_t tune_ssim =
384 ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
385 static const arg_def_t cq_level =
386 ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
387 static const arg_def_t max_intra_rate_pct =
388 ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
389 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
390 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
391
392 #if CONFIG_VP8_ENCODER
393 static const arg_def_t cpu_used_vp8 =
394 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
395 static const arg_def_t auto_altref_vp8 = ARG_DEF(
396 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames. (0..1)");
397 static const arg_def_t token_parts =
398 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
399 static const arg_def_t screen_content_mode =
400 ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
401 static const arg_def_t *vp8_args[] = { &cpu_used_vp8,
402 &auto_altref_vp8,
403 &noise_sens,
404 &sharpness,
405 &static_thresh,
406 &token_parts,
407 &arnr_maxframes,
408 &arnr_strength,
409 &arnr_type,
410 &tune_ssim,
411 &cq_level,
412 &max_intra_rate_pct,
413 &gf_cbr_boost_pct,
414 &screen_content_mode,
415 NULL };
416 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
417 VP8E_SET_ENABLEAUTOALTREF,
418 VP8E_SET_NOISE_SENSITIVITY,
419 VP8E_SET_SHARPNESS,
420 VP8E_SET_STATIC_THRESHOLD,
421 VP8E_SET_TOKEN_PARTITIONS,
422 VP8E_SET_ARNR_MAXFRAMES,
423 VP8E_SET_ARNR_STRENGTH,
424 VP8E_SET_ARNR_TYPE,
425 VP8E_SET_TUNING,
426 VP8E_SET_CQ_LEVEL,
427 VP8E_SET_MAX_INTRA_BITRATE_PCT,
428 VP8E_SET_GF_CBR_BOOST_PCT,
429 VP8E_SET_SCREEN_CONTENT_MODE,
430 0 };
431 #endif
432
433 #if CONFIG_VP9_ENCODER
434 static const arg_def_t cpu_used_vp9 =
435 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-9..9)");
436 static const arg_def_t auto_altref_vp9 = ARG_DEF(
437 NULL, "auto-alt-ref", 1,
438 "Enable automatic alt reference frames, 2+ enables multi-layer. (0..6)");
439 static const arg_def_t tile_cols =
440 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
441 static const arg_def_t tile_rows =
442 ARG_DEF(NULL, "tile-rows", 1,
443 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
444
445 static const arg_def_t enable_tpl_model =
446 ARG_DEF(NULL, "enable-tpl", 1, "Enable temporal dependency model");
447
448 static const arg_def_t lossless =
449 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
450 static const arg_def_t frame_parallel_decoding = ARG_DEF(
451 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
452 static const arg_def_t aq_mode = ARG_DEF(
453 NULL, "aq-mode", 1,
454 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
455 "3: cyclic refresh, 4: equator360)");
456 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
457 "Special adaptive quantization for "
458 "the alternate reference frames.");
459 static const arg_def_t frame_periodic_boost =
460 ARG_DEF(NULL, "frame-boost", 1,
461 "Enable frame periodic boost (0: off (default), 1: on)");
462 static const arg_def_t max_inter_rate_pct =
463 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
464 static const arg_def_t min_gf_interval = ARG_DEF(
465 NULL, "min-gf-interval", 1,
466 "min gf/arf frame interval (default 0, indicating in-built behavior)");
467 static const arg_def_t max_gf_interval = ARG_DEF(
468 NULL, "max-gf-interval", 1,
469 "max gf/arf frame interval (default 0, indicating in-built behavior)");
470
471 static const struct arg_enum_list color_space_enum[] = {
472 { "unknown", VPX_CS_UNKNOWN },
473 { "bt601", VPX_CS_BT_601 },
474 { "bt709", VPX_CS_BT_709 },
475 { "smpte170", VPX_CS_SMPTE_170 },
476 { "smpte240", VPX_CS_SMPTE_240 },
477 { "bt2020", VPX_CS_BT_2020 },
478 { "reserved", VPX_CS_RESERVED },
479 { "sRGB", VPX_CS_SRGB },
480 { NULL, 0 }
481 };
482
483 static const arg_def_t input_color_space =
484 ARG_DEF_ENUM(NULL, "color-space", 1,
485 "The color space of input content:", color_space_enum);
486
487 #if CONFIG_VP9_HIGHBITDEPTH
488 static const struct arg_enum_list bitdepth_enum[] = {
489 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
490 };
491
492 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
493 "b", "bit-depth", 1,
494 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
495 bitdepth_enum);
496 static const arg_def_t inbitdeptharg =
497 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
498 #endif
499
500 static const struct arg_enum_list tune_content_enum[] = {
501 { "default", VP9E_CONTENT_DEFAULT },
502 { "screen", VP9E_CONTENT_SCREEN },
503 { "film", VP9E_CONTENT_FILM },
504 { NULL, 0 }
505 };
506
507 static const arg_def_t tune_content = ARG_DEF_ENUM(
508 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
509
510 static const arg_def_t target_level = ARG_DEF(
511 NULL, "target-level", 1,
512 "Target level\n"
513 " 255: off (default)\n"
514 " 0: only keep level stats\n"
515 " 1: adaptively set alt-ref "
516 "distance and column tile limit based on picture size, and keep"
517 " level stats\n"
518 " 10: level 1.0 11: level 1.1 "
519 "... 62: level 6.2");
520
521 static const arg_def_t row_mt =
522 ARG_DEF(NULL, "row-mt", 1,
523 "Enable row based non-deterministic multi-threading in VP9");
524
525 static const arg_def_t disable_loopfilter =
526 ARG_DEF(NULL, "disable-loopfilter", 1,
527 "Control Loopfilter in VP9:\n"
528 " "
529 "0: Loopfilter on for all frames (default)\n"
530 " "
531 "1: Loopfilter off for non reference frames\n"
532 " "
533 "2: Loopfilter off for all frames");
534 #endif
535
536 #if CONFIG_VP9_ENCODER
537 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
538 &auto_altref_vp9,
539 &sharpness,
540 &static_thresh,
541 &tile_cols,
542 &tile_rows,
543 &enable_tpl_model,
544 &arnr_maxframes,
545 &arnr_strength,
546 &arnr_type,
547 &tune_ssim,
548 &cq_level,
549 &max_intra_rate_pct,
550 &max_inter_rate_pct,
551 &gf_cbr_boost_pct,
552 &lossless,
553 &frame_parallel_decoding,
554 &aq_mode,
555 &alt_ref_aq,
556 &frame_periodic_boost,
557 &noise_sens,
558 &tune_content,
559 &input_color_space,
560 &min_gf_interval,
561 &max_gf_interval,
562 &target_level,
563 &row_mt,
564 &disable_loopfilter,
565 // NOTE: The entries above have a corresponding entry in vp9_arg_ctrl_map. The
566 // entries below do not have a corresponding entry in vp9_arg_ctrl_map. They
567 // must be listed at the end of vp9_args.
568 #if CONFIG_VP9_HIGHBITDEPTH
569 &bitdeptharg,
570 &inbitdeptharg,
571 #endif // CONFIG_VP9_HIGHBITDEPTH
572 NULL };
573 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
574 VP8E_SET_ENABLEAUTOALTREF,
575 VP8E_SET_SHARPNESS,
576 VP8E_SET_STATIC_THRESHOLD,
577 VP9E_SET_TILE_COLUMNS,
578 VP9E_SET_TILE_ROWS,
579 VP9E_SET_TPL,
580 VP8E_SET_ARNR_MAXFRAMES,
581 VP8E_SET_ARNR_STRENGTH,
582 VP8E_SET_ARNR_TYPE,
583 VP8E_SET_TUNING,
584 VP8E_SET_CQ_LEVEL,
585 VP8E_SET_MAX_INTRA_BITRATE_PCT,
586 VP9E_SET_MAX_INTER_BITRATE_PCT,
587 VP9E_SET_GF_CBR_BOOST_PCT,
588 VP9E_SET_LOSSLESS,
589 VP9E_SET_FRAME_PARALLEL_DECODING,
590 VP9E_SET_AQ_MODE,
591 VP9E_SET_ALT_REF_AQ,
592 VP9E_SET_FRAME_PERIODIC_BOOST,
593 VP9E_SET_NOISE_SENSITIVITY,
594 VP9E_SET_TUNE_CONTENT,
595 VP9E_SET_COLOR_SPACE,
596 VP9E_SET_MIN_GF_INTERVAL,
597 VP9E_SET_MAX_GF_INTERVAL,
598 VP9E_SET_TARGET_LEVEL,
599 VP9E_SET_ROW_MT,
600 VP9E_SET_DISABLE_LOOPFILTER,
601 0 };
602 #endif
603
604 static const arg_def_t *no_args[] = { NULL };
605
show_help(FILE * fout,int shorthelp)606 static void show_help(FILE *fout, int shorthelp) {
607 int i;
608 const int num_encoder = get_vpx_encoder_count();
609
610 fprintf(fout, "Usage: %s <options> -o dst_filename src_filename \n",
611 exec_name);
612
613 if (shorthelp) {
614 fprintf(fout, "Use --help to see the full list of options.\n");
615 return;
616 }
617
618 fprintf(fout, "\nOptions:\n");
619 arg_show_usage(fout, main_args);
620 fprintf(fout, "\nEncoder Global Options:\n");
621 arg_show_usage(fout, global_args);
622 fprintf(fout, "\nRate Control Options:\n");
623 arg_show_usage(fout, rc_args);
624 fprintf(fout, "\nTwopass Rate Control Options:\n");
625 arg_show_usage(fout, rc_twopass_args);
626 fprintf(fout, "\nKeyframe Placement Options:\n");
627 arg_show_usage(fout, kf_args);
628 #if CONFIG_VP8_ENCODER
629 fprintf(fout, "\nVP8 Specific Options:\n");
630 arg_show_usage(fout, vp8_args);
631 #endif
632 #if CONFIG_VP9_ENCODER
633 fprintf(fout, "\nVP9 Specific Options:\n");
634 arg_show_usage(fout, vp9_args);
635 fprintf(fout, "\nVizier Rate Control Options:\n");
636 arg_show_usage(fout, vizier_rc_args);
637 #endif
638 fprintf(fout,
639 "\nStream timebase (--timebase):\n"
640 " The desired precision of timestamps in the output, expressed\n"
641 " in fractional seconds. Default is 1/1000.\n");
642 fprintf(fout, "\nIncluded encoders:\n\n");
643
644 for (i = 0; i < num_encoder; ++i) {
645 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
646 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
647 fprintf(fout, " %-6s - %s %s\n", encoder->name,
648 vpx_codec_iface_name(encoder->codec_interface()), defstr);
649 }
650 fprintf(fout, "\n ");
651 fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
652 }
653
usage_exit(void)654 void usage_exit(void) {
655 show_help(stderr, 1);
656 exit(EXIT_FAILURE);
657 }
658
659 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
660 #if CONFIG_VP9_ENCODER
661 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
662 #else
663 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
664 #endif
665
666 #if !CONFIG_WEBM_IO
667 typedef int stereo_format_t;
668 struct WebmOutputContext {
669 int debug;
670 };
671 #endif
672
673 /* Per-stream configuration */
674 struct stream_config {
675 struct vpx_codec_enc_cfg cfg;
676 const char *out_fn;
677 const char *stats_fn;
678 stereo_format_t stereo_fmt;
679 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
680 int arg_ctrl_cnt;
681 int write_webm;
682 #if CONFIG_VP9_HIGHBITDEPTH
683 // whether to use 16bit internal buffers
684 int use_16bit_internal;
685 #endif
686 };
687
688 struct stream_state {
689 int index;
690 struct stream_state *next;
691 struct stream_config config;
692 FILE *file;
693 struct rate_hist *rate_hist;
694 struct WebmOutputContext webm_ctx;
695 uint64_t psnr_sse_total;
696 uint64_t psnr_samples_total;
697 double psnr_totals[4];
698 int psnr_count;
699 int counts[64];
700 vpx_codec_ctx_t encoder;
701 unsigned int frames_out;
702 uint64_t cx_time;
703 size_t nbytes;
704 stats_io_t stats;
705 struct vpx_image *img;
706 vpx_codec_ctx_t decoder;
707 int mismatch_seen;
708 };
709
validate_positive_rational(const char * msg,struct vpx_rational * rat)710 static void validate_positive_rational(const char *msg,
711 struct vpx_rational *rat) {
712 if (rat->den < 0) {
713 rat->num *= -1;
714 rat->den *= -1;
715 }
716
717 if (rat->num < 0) die("Error: %s must be positive\n", msg);
718
719 if (!rat->den) die("Error: %s has zero denominator\n", msg);
720 }
721
parse_global_config(struct VpxEncoderConfig * global,char ** argv)722 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
723 char **argi, **argj;
724 struct arg arg;
725 const int num_encoder = get_vpx_encoder_count();
726
727 if (num_encoder < 1) die("Error: no valid encoder available\n");
728
729 /* Initialize default parameters */
730 memset(global, 0, sizeof(*global));
731 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
732 global->passes = 0;
733 global->color_type = I420;
734 /* Assign default deadline to good quality */
735 global->deadline = VPX_DL_GOOD_QUALITY;
736
737 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
738 arg.argv_step = 1;
739
740 if (arg_match(&arg, &help, argi)) {
741 show_help(stdout, 0);
742 exit(EXIT_SUCCESS);
743 } else if (arg_match(&arg, &codecarg, argi)) {
744 global->codec = get_vpx_encoder_by_name(arg.val);
745 if (!global->codec)
746 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
747 } else if (arg_match(&arg, &passes, argi)) {
748 global->passes = arg_parse_uint(&arg);
749
750 if (global->passes < 1 || global->passes > 2)
751 die("Error: Invalid number of passes (%d)\n", global->passes);
752 } else if (arg_match(&arg, &pass_arg, argi)) {
753 global->pass = arg_parse_uint(&arg);
754
755 if (global->pass < 1 || global->pass > 2)
756 die("Error: Invalid pass selected (%d)\n", global->pass);
757 } else if (arg_match(&arg, &usage, argi))
758 global->usage = arg_parse_uint(&arg);
759 else if (arg_match(&arg, &deadline, argi))
760 global->deadline = arg_parse_uint(&arg);
761 else if (arg_match(&arg, &best_dl, argi))
762 global->deadline = VPX_DL_BEST_QUALITY;
763 else if (arg_match(&arg, &good_dl, argi))
764 global->deadline = VPX_DL_GOOD_QUALITY;
765 else if (arg_match(&arg, &rt_dl, argi))
766 global->deadline = VPX_DL_REALTIME;
767 else if (arg_match(&arg, &use_yv12, argi))
768 global->color_type = YV12;
769 else if (arg_match(&arg, &use_nv12, argi))
770 global->color_type = NV12;
771 else if (arg_match(&arg, &use_i420, argi))
772 global->color_type = I420;
773 else if (arg_match(&arg, &use_i422, argi))
774 global->color_type = I422;
775 else if (arg_match(&arg, &use_i444, argi))
776 global->color_type = I444;
777 else if (arg_match(&arg, &use_i440, argi))
778 global->color_type = I440;
779 else if (arg_match(&arg, &quietarg, argi))
780 global->quiet = 1;
781 else if (arg_match(&arg, &verbosearg, argi))
782 global->verbose = 1;
783 else if (arg_match(&arg, &limit, argi))
784 global->limit = arg_parse_uint(&arg);
785 else if (arg_match(&arg, &skip, argi))
786 global->skip_frames = arg_parse_uint(&arg);
787 else if (arg_match(&arg, &psnrarg, argi))
788 global->show_psnr = 1;
789 else if (arg_match(&arg, &recontest, argi))
790 global->test_decode = arg_parse_enum_or_int(&arg);
791 else if (arg_match(&arg, &framerate, argi)) {
792 global->framerate = arg_parse_rational(&arg);
793 validate_positive_rational(arg.name, &global->framerate);
794 global->have_framerate = 1;
795 } else if (arg_match(&arg, &out_part, argi))
796 global->out_part = 1;
797 else if (arg_match(&arg, &debugmode, argi))
798 global->debug = 1;
799 else if (arg_match(&arg, &q_hist_n, argi))
800 global->show_q_hist_buckets = arg_parse_uint(&arg);
801 else if (arg_match(&arg, &rate_hist_n, argi))
802 global->show_rate_hist_buckets = arg_parse_uint(&arg);
803 else if (arg_match(&arg, &disable_warnings, argi))
804 global->disable_warnings = 1;
805 else if (arg_match(&arg, &disable_warning_prompt, argi))
806 global->disable_warning_prompt = 1;
807 else
808 argj++;
809 }
810
811 if (global->pass) {
812 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
813 if (global->pass > global->passes) {
814 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
815 global->pass);
816 global->passes = global->pass;
817 }
818 }
819 /* Validate global config */
820 if (global->passes == 0) {
821 #if CONFIG_VP9_ENCODER
822 // Make default VP9 passes = 2 until there is a better quality 1-pass
823 // encoder
824 if (global->codec != NULL && global->codec->name != NULL)
825 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
826 global->deadline != VPX_DL_REALTIME)
827 ? 2
828 : 1;
829 #else
830 global->passes = 1;
831 #endif
832 }
833
834 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
835 warn("Enforcing one-pass encoding in realtime mode\n");
836 global->passes = 1;
837 }
838 }
839
new_stream(struct VpxEncoderConfig * global,struct stream_state * prev)840 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
841 struct stream_state *prev) {
842 struct stream_state *stream;
843
844 stream = calloc(1, sizeof(*stream));
845 if (stream == NULL) {
846 fatal("Failed to allocate new stream.");
847 }
848
849 if (prev) {
850 memcpy(stream, prev, sizeof(*stream));
851 stream->index++;
852 prev->next = stream;
853 } else {
854 vpx_codec_err_t res;
855
856 /* Populate encoder configuration */
857 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
858 &stream->config.cfg, global->usage);
859 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
860
861 /* Change the default timebase to a high enough value so that the
862 * encoder will always create strictly increasing timestamps.
863 */
864 stream->config.cfg.g_timebase.den = 1000;
865
866 /* Never use the library's default resolution, require it be parsed
867 * from the file or set on the command line.
868 */
869 stream->config.cfg.g_w = 0;
870 stream->config.cfg.g_h = 0;
871
872 /* Initialize remaining stream parameters */
873 stream->config.write_webm = 1;
874 #if CONFIG_WEBM_IO
875 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
876 stream->webm_ctx.last_pts_ns = -1;
877 stream->webm_ctx.writer = NULL;
878 stream->webm_ctx.segment = NULL;
879 #endif
880
881 /* Allows removal of the application version from the EBML tags */
882 stream->webm_ctx.debug = global->debug;
883
884 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
885 if (global->deadline == VPX_DL_REALTIME &&
886 stream->config.cfg.rc_end_usage == VPX_CBR)
887 stream->config.cfg.g_lag_in_frames = 0;
888 }
889
890 /* Output files must be specified for each stream */
891 stream->config.out_fn = NULL;
892
893 stream->next = NULL;
894 return stream;
895 }
896
parse_stream_params(struct VpxEncoderConfig * global,struct stream_state * stream,char ** argv)897 static int parse_stream_params(struct VpxEncoderConfig *global,
898 struct stream_state *stream, char **argv) {
899 char **argi, **argj;
900 struct arg arg;
901 static const arg_def_t **ctrl_args = no_args;
902 static const int *ctrl_args_map = NULL;
903 struct stream_config *config = &stream->config;
904 int eos_mark_found = 0;
905 #if CONFIG_VP9_HIGHBITDEPTH
906 int test_16bit_internal = 0;
907 #endif
908
909 // Handle codec specific options
910 if (0) {
911 #if CONFIG_VP8_ENCODER
912 } else if (strcmp(global->codec->name, "vp8") == 0) {
913 ctrl_args = vp8_args;
914 ctrl_args_map = vp8_arg_ctrl_map;
915 #endif
916 #if CONFIG_VP9_ENCODER
917 } else if (strcmp(global->codec->name, "vp9") == 0) {
918 ctrl_args = vp9_args;
919 ctrl_args_map = vp9_arg_ctrl_map;
920 #endif
921 }
922
923 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
924 arg.argv_step = 1;
925
926 /* Once we've found an end-of-stream marker (--) we want to continue
927 * shifting arguments but not consuming them.
928 */
929 if (eos_mark_found) {
930 argj++;
931 continue;
932 } else if (!strcmp(*argj, "--")) {
933 eos_mark_found = 1;
934 continue;
935 }
936
937 if (arg_match(&arg, &outputfile, argi)) {
938 config->out_fn = arg.val;
939 } else if (arg_match(&arg, &fpf_name, argi)) {
940 config->stats_fn = arg.val;
941 } else if (arg_match(&arg, &use_webm, argi)) {
942 #if CONFIG_WEBM_IO
943 config->write_webm = 1;
944 #else
945 die("Error: --webm specified but webm is disabled.");
946 #endif
947 } else if (arg_match(&arg, &use_ivf, argi)) {
948 config->write_webm = 0;
949 } else if (arg_match(&arg, &threads, argi)) {
950 config->cfg.g_threads = arg_parse_uint(&arg);
951 } else if (arg_match(&arg, &profile, argi)) {
952 config->cfg.g_profile = arg_parse_uint(&arg);
953 } else if (arg_match(&arg, &width, argi)) {
954 config->cfg.g_w = arg_parse_uint(&arg);
955 } else if (arg_match(&arg, &height, argi)) {
956 config->cfg.g_h = arg_parse_uint(&arg);
957 #if CONFIG_VP9_HIGHBITDEPTH
958 } else if (arg_match(&arg, &bitdeptharg, argi)) {
959 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
960 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
961 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
962 #endif
963 #if CONFIG_WEBM_IO
964 } else if (arg_match(&arg, &stereo_mode, argi)) {
965 config->stereo_fmt = arg_parse_enum_or_int(&arg);
966 #endif
967 } else if (arg_match(&arg, &timebase, argi)) {
968 config->cfg.g_timebase = arg_parse_rational(&arg);
969 validate_positive_rational(arg.name, &config->cfg.g_timebase);
970 } else if (arg_match(&arg, &error_resilient, argi)) {
971 config->cfg.g_error_resilient = arg_parse_uint(&arg);
972 } else if (arg_match(&arg, &end_usage, argi)) {
973 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
974 } else if (arg_match(&arg, &lag_in_frames, argi)) {
975 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
976 if (global->deadline == VPX_DL_REALTIME &&
977 config->cfg.rc_end_usage == VPX_CBR &&
978 config->cfg.g_lag_in_frames != 0) {
979 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
980 config->cfg.g_lag_in_frames = 0;
981 }
982 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
983 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
984 } else if (arg_match(&arg, &resize_allowed, argi)) {
985 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
986 } else if (arg_match(&arg, &resize_width, argi)) {
987 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
988 } else if (arg_match(&arg, &resize_height, argi)) {
989 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
990 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
991 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
992 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
993 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
994 } else if (arg_match(&arg, &end_usage, argi)) {
995 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
996 } else if (arg_match(&arg, &target_bitrate, argi)) {
997 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
998 } else if (arg_match(&arg, &min_quantizer, argi)) {
999 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1000 } else if (arg_match(&arg, &max_quantizer, argi)) {
1001 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1002 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1003 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1004 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1005 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1006 } else if (arg_match(&arg, &buf_sz, argi)) {
1007 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1008 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1009 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1010 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1011 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1012 } else if (arg_match(&arg, &bias_pct, argi)) {
1013 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1014 if (global->passes < 2)
1015 warn("option %s ignored in one-pass mode.\n", arg.name);
1016 } else if (arg_match(&arg, &minsection_pct, argi)) {
1017 config->cfg.rc_2pass_vbr_minsection_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, &maxsection_pct, argi)) {
1022 config->cfg.rc_2pass_vbr_maxsection_pct = 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, &corpus_complexity, argi)) {
1027 config->cfg.rc_2pass_vbr_corpus_complexity = arg_parse_uint(&arg);
1028
1029 if (global->passes < 2)
1030 warn("option %s ignored in one-pass mode.\n", arg.name);
1031 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1032 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1033 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1034 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1035 } else if (arg_match(&arg, &kf_disabled, argi)) {
1036 config->cfg.kf_mode = VPX_KF_DISABLED;
1037 #if CONFIG_VP9_ENCODER
1038 } else if (arg_match(&arg, &use_vizier_rc_params, argi)) {
1039 config->cfg.use_vizier_rc_params = arg_parse_int(&arg);
1040 } else if (arg_match(&arg, &active_wq_factor, argi)) {
1041 config->cfg.active_wq_factor = arg_parse_rational(&arg);
1042 } else if (arg_match(&arg, &err_per_mb_factor, argi)) {
1043 config->cfg.err_per_mb_factor = arg_parse_rational(&arg);
1044 } else if (arg_match(&arg, &sr_default_decay_limit, argi)) {
1045 config->cfg.sr_default_decay_limit = arg_parse_rational(&arg);
1046 } else if (arg_match(&arg, &sr_diff_factor, argi)) {
1047 config->cfg.sr_diff_factor = arg_parse_rational(&arg);
1048 } else if (arg_match(&arg, &kf_err_per_mb_factor, argi)) {
1049 config->cfg.kf_err_per_mb_factor = arg_parse_rational(&arg);
1050 } else if (arg_match(&arg, &kf_frame_min_boost_factor, argi)) {
1051 config->cfg.kf_frame_min_boost_factor = arg_parse_rational(&arg);
1052 } else if (arg_match(&arg, &kf_frame_max_boost_first_factor, argi)) {
1053 config->cfg.kf_frame_max_boost_first_factor = arg_parse_rational(&arg);
1054 } else if (arg_match(&arg, &kf_frame_max_boost_subs_factor, argi)) {
1055 config->cfg.kf_frame_max_boost_subs_factor = arg_parse_rational(&arg);
1056 } else if (arg_match(&arg, &kf_max_total_boost_factor, argi)) {
1057 config->cfg.kf_max_total_boost_factor = arg_parse_rational(&arg);
1058 } else if (arg_match(&arg, &gf_max_total_boost_factor, argi)) {
1059 config->cfg.gf_max_total_boost_factor = arg_parse_rational(&arg);
1060 } else if (arg_match(&arg, &gf_frame_max_boost_factor, argi)) {
1061 config->cfg.gf_frame_max_boost_factor = arg_parse_rational(&arg);
1062 } else if (arg_match(&arg, &zm_factor, argi)) {
1063 config->cfg.zm_factor = arg_parse_rational(&arg);
1064 } else if (arg_match(&arg, &rd_mult_inter_qp_fac, argi)) {
1065 config->cfg.rd_mult_inter_qp_fac = arg_parse_rational(&arg);
1066 } else if (arg_match(&arg, &rd_mult_arf_qp_fac, argi)) {
1067 config->cfg.rd_mult_arf_qp_fac = arg_parse_rational(&arg);
1068 } else if (arg_match(&arg, &rd_mult_key_qp_fac, argi)) {
1069 config->cfg.rd_mult_key_qp_fac = arg_parse_rational(&arg);
1070 #endif
1071 #if CONFIG_VP9_HIGHBITDEPTH
1072 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1073 if (strcmp(global->codec->name, "vp9") == 0) {
1074 test_16bit_internal = 1;
1075 }
1076 #endif
1077 } else {
1078 int i, match = 0;
1079 for (i = 0; ctrl_args[i]; i++) {
1080 if (arg_match(&arg, ctrl_args[i], argi)) {
1081 int j;
1082 match = 1;
1083
1084 /* Point either to the next free element or the first
1085 * instance of this control.
1086 */
1087 for (j = 0; j < config->arg_ctrl_cnt; j++)
1088 if (ctrl_args_map != NULL &&
1089 config->arg_ctrls[j][0] == ctrl_args_map[i])
1090 break;
1091
1092 /* Update/insert */
1093 assert(j < (int)ARG_CTRL_CNT_MAX);
1094 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1095 config->arg_ctrls[j][0] = ctrl_args_map[i];
1096 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1097 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1098 }
1099 }
1100 }
1101 if (!match) argj++;
1102 }
1103 }
1104 #if CONFIG_VP9_HIGHBITDEPTH
1105 if (strcmp(global->codec->name, "vp9") == 0) {
1106 config->use_16bit_internal =
1107 test_16bit_internal | (config->cfg.g_profile > 1);
1108 }
1109 #endif
1110 return eos_mark_found;
1111 }
1112
1113 #define FOREACH_STREAM(func) \
1114 do { \
1115 struct stream_state *stream; \
1116 for (stream = streams; stream; stream = stream->next) { \
1117 func; \
1118 } \
1119 } while (0)
1120
validate_stream_config(const struct stream_state * stream,const struct VpxEncoderConfig * global)1121 static void validate_stream_config(const struct stream_state *stream,
1122 const struct VpxEncoderConfig *global) {
1123 const struct stream_state *streami;
1124 (void)global;
1125
1126 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1127 fatal(
1128 "Stream %d: Specify stream dimensions with --width (-w) "
1129 " and --height (-h)",
1130 stream->index);
1131
1132 // Check that the codec bit depth is greater than the input bit depth.
1133 if (stream->config.cfg.g_input_bit_depth >
1134 (unsigned int)stream->config.cfg.g_bit_depth) {
1135 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1136 stream->index, (int)stream->config.cfg.g_bit_depth,
1137 stream->config.cfg.g_input_bit_depth);
1138 }
1139
1140 for (streami = stream; streami; streami = streami->next) {
1141 /* All streams require output files */
1142 if (!streami->config.out_fn)
1143 fatal("Stream %d: Output file is required (specify with -o)",
1144 streami->index);
1145
1146 /* Check for two streams outputting to the same file */
1147 if (streami != stream) {
1148 const char *a = stream->config.out_fn;
1149 const char *b = streami->config.out_fn;
1150 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1151 fatal("Stream %d: duplicate output file (from stream %d)",
1152 streami->index, stream->index);
1153 }
1154
1155 /* Check for two streams sharing a stats file. */
1156 if (streami != stream) {
1157 const char *a = stream->config.stats_fn;
1158 const char *b = streami->config.stats_fn;
1159 if (a && b && !strcmp(a, b))
1160 fatal("Stream %d: duplicate stats file (from stream %d)",
1161 streami->index, stream->index);
1162 }
1163 }
1164 }
1165
set_stream_dimensions(struct stream_state * stream,unsigned int w,unsigned int h)1166 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1167 unsigned int h) {
1168 if (!stream->config.cfg.g_w) {
1169 if (!stream->config.cfg.g_h)
1170 stream->config.cfg.g_w = w;
1171 else
1172 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1173 }
1174 if (!stream->config.cfg.g_h) {
1175 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1176 }
1177 }
1178
file_type_to_string(enum VideoFileType t)1179 static const char *file_type_to_string(enum VideoFileType t) {
1180 switch (t) {
1181 case FILE_TYPE_RAW: return "RAW";
1182 case FILE_TYPE_Y4M: return "Y4M";
1183 default: return "Other";
1184 }
1185 }
1186
image_format_to_string(vpx_img_fmt_t f)1187 static const char *image_format_to_string(vpx_img_fmt_t f) {
1188 switch (f) {
1189 case VPX_IMG_FMT_I420: return "I420";
1190 case VPX_IMG_FMT_I422: return "I422";
1191 case VPX_IMG_FMT_I444: return "I444";
1192 case VPX_IMG_FMT_I440: return "I440";
1193 case VPX_IMG_FMT_YV12: return "YV12";
1194 case VPX_IMG_FMT_I42016: return "I42016";
1195 case VPX_IMG_FMT_I42216: return "I42216";
1196 case VPX_IMG_FMT_I44416: return "I44416";
1197 case VPX_IMG_FMT_I44016: return "I44016";
1198 default: return "Other";
1199 }
1200 }
1201
show_stream_config(struct stream_state * stream,struct VpxEncoderConfig * global,struct VpxInputContext * input)1202 static void show_stream_config(struct stream_state *stream,
1203 struct VpxEncoderConfig *global,
1204 struct VpxInputContext *input) {
1205 #define SHOW(field) \
1206 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1207
1208 if (stream->index == 0) {
1209 fprintf(stderr, "Codec: %s\n",
1210 vpx_codec_iface_name(global->codec->codec_interface()));
1211 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1212 input->filename, file_type_to_string(input->file_type),
1213 image_format_to_string(input->fmt));
1214 }
1215 if (stream->next || stream->index)
1216 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1217 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1218 fprintf(stderr, "Encoder parameters:\n");
1219
1220 SHOW(g_usage);
1221 SHOW(g_threads);
1222 SHOW(g_profile);
1223 SHOW(g_w);
1224 SHOW(g_h);
1225 SHOW(g_bit_depth);
1226 SHOW(g_input_bit_depth);
1227 SHOW(g_timebase.num);
1228 SHOW(g_timebase.den);
1229 SHOW(g_error_resilient);
1230 SHOW(g_pass);
1231 SHOW(g_lag_in_frames);
1232 SHOW(rc_dropframe_thresh);
1233 SHOW(rc_resize_allowed);
1234 SHOW(rc_scaled_width);
1235 SHOW(rc_scaled_height);
1236 SHOW(rc_resize_up_thresh);
1237 SHOW(rc_resize_down_thresh);
1238 SHOW(rc_end_usage);
1239 SHOW(rc_target_bitrate);
1240 SHOW(rc_min_quantizer);
1241 SHOW(rc_max_quantizer);
1242 SHOW(rc_undershoot_pct);
1243 SHOW(rc_overshoot_pct);
1244 SHOW(rc_buf_sz);
1245 SHOW(rc_buf_initial_sz);
1246 SHOW(rc_buf_optimal_sz);
1247 SHOW(rc_2pass_vbr_bias_pct);
1248 SHOW(rc_2pass_vbr_minsection_pct);
1249 SHOW(rc_2pass_vbr_maxsection_pct);
1250 SHOW(rc_2pass_vbr_corpus_complexity);
1251 SHOW(kf_mode);
1252 SHOW(kf_min_dist);
1253 SHOW(kf_max_dist);
1254 // Temporary use for debug
1255 SHOW(use_vizier_rc_params);
1256 SHOW(active_wq_factor.num);
1257 SHOW(active_wq_factor.den);
1258 }
1259
open_output_file(struct stream_state * stream,struct VpxEncoderConfig * global,const struct VpxRational * pixel_aspect_ratio)1260 static void open_output_file(struct stream_state *stream,
1261 struct VpxEncoderConfig *global,
1262 const struct VpxRational *pixel_aspect_ratio) {
1263 const char *fn = stream->config.out_fn;
1264 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1265
1266 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1267
1268 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1269
1270 if (!stream->file) fatal("Failed to open output file");
1271
1272 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1273 fatal("WebM output to pipes not supported.");
1274
1275 #if CONFIG_WEBM_IO
1276 if (stream->config.write_webm) {
1277 stream->webm_ctx.stream = stream->file;
1278 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1279 global->codec->fourcc, pixel_aspect_ratio);
1280 }
1281 #else
1282 (void)pixel_aspect_ratio;
1283 #endif
1284
1285 if (!stream->config.write_webm) {
1286 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1287 }
1288 }
1289
close_output_file(struct stream_state * stream,unsigned int fourcc)1290 static void close_output_file(struct stream_state *stream,
1291 unsigned int fourcc) {
1292 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1293
1294 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1295
1296 #if CONFIG_WEBM_IO
1297 if (stream->config.write_webm) {
1298 write_webm_file_footer(&stream->webm_ctx);
1299 }
1300 #endif
1301
1302 if (!stream->config.write_webm) {
1303 if (!fseek(stream->file, 0, SEEK_SET))
1304 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1305 stream->frames_out);
1306 }
1307
1308 fclose(stream->file);
1309 }
1310
setup_pass(struct stream_state * stream,struct VpxEncoderConfig * global,int pass)1311 static void setup_pass(struct stream_state *stream,
1312 struct VpxEncoderConfig *global, int pass) {
1313 if (stream->config.stats_fn) {
1314 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1315 fatal("Failed to open statistics store");
1316 } else {
1317 if (!stats_open_mem(&stream->stats, pass))
1318 fatal("Failed to open statistics store");
1319 }
1320
1321 stream->config.cfg.g_pass = global->passes == 2
1322 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1323 : VPX_RC_ONE_PASS;
1324 if (pass) {
1325 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1326 }
1327
1328 stream->cx_time = 0;
1329 stream->nbytes = 0;
1330 stream->frames_out = 0;
1331 }
1332
initialize_encoder(struct stream_state * stream,struct VpxEncoderConfig * global)1333 static void initialize_encoder(struct stream_state *stream,
1334 struct VpxEncoderConfig *global) {
1335 int i;
1336 int flags = 0;
1337
1338 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1339 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1340 #if CONFIG_VP9_HIGHBITDEPTH
1341 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1342 #endif
1343
1344 /* Construct Encoder Context */
1345 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1346 &stream->config.cfg, flags);
1347 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1348
1349 /* Note that we bypass the vpx_codec_control wrapper macro because
1350 * we're being clever to store the control IDs in an array. Real
1351 * applications will want to make use of the enumerations directly
1352 */
1353 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1354 int ctrl = stream->config.arg_ctrls[i][0];
1355 int value = stream->config.arg_ctrls[i][1];
1356 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1357 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1358
1359 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1360 }
1361
1362 #if CONFIG_DECODERS
1363 if (global->test_decode != TEST_DECODE_OFF) {
1364 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1365 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1366 }
1367 #endif
1368 }
1369
encode_frame(struct stream_state * stream,struct VpxEncoderConfig * global,struct vpx_image * img,unsigned int frames_in)1370 static void encode_frame(struct stream_state *stream,
1371 struct VpxEncoderConfig *global, struct vpx_image *img,
1372 unsigned int frames_in) {
1373 vpx_codec_pts_t frame_start, next_frame_start;
1374 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1375 struct vpx_usec_timer timer;
1376
1377 frame_start =
1378 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1379 cfg->g_timebase.num / global->framerate.num;
1380 next_frame_start =
1381 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1382 cfg->g_timebase.num / global->framerate.num;
1383
1384 /* Scale if necessary */
1385 #if CONFIG_VP9_HIGHBITDEPTH
1386 if (img) {
1387 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1388 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1389 if (img->fmt != VPX_IMG_FMT_I42016) {
1390 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1391 exit(EXIT_FAILURE);
1392 }
1393 #if CONFIG_LIBYUV
1394 if (!stream->img) {
1395 stream->img =
1396 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1397 }
1398 I420Scale_16(
1399 (uint16_t *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1400 (uint16_t *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1401 (uint16_t *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1402 img->d_w, img->d_h, (uint16_t *)stream->img->planes[VPX_PLANE_Y],
1403 stream->img->stride[VPX_PLANE_Y] / 2,
1404 (uint16_t *)stream->img->planes[VPX_PLANE_U],
1405 stream->img->stride[VPX_PLANE_U] / 2,
1406 (uint16_t *)stream->img->planes[VPX_PLANE_V],
1407 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1408 stream->img->d_h, kFilterBox);
1409 img = stream->img;
1410 #else
1411 stream->encoder.err = 1;
1412 ctx_exit_on_error(&stream->encoder,
1413 "Stream %d: Failed to encode frame.\n"
1414 "Scaling disabled in this configuration. \n"
1415 "To enable, configure with --enable-libyuv\n",
1416 stream->index);
1417 #endif
1418 }
1419 }
1420 #endif
1421 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1422 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1423 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1424 exit(EXIT_FAILURE);
1425 }
1426 #if CONFIG_LIBYUV
1427 if (!stream->img)
1428 stream->img =
1429 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1430 I420Scale(
1431 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1432 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1433 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1434 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1435 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1436 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1437 stream->img->d_w, stream->img->d_h, kFilterBox);
1438 img = stream->img;
1439 #else
1440 stream->encoder.err = 1;
1441 ctx_exit_on_error(&stream->encoder,
1442 "Stream %d: Failed to encode frame.\n"
1443 "Scaling disabled in this configuration. \n"
1444 "To enable, configure with --enable-libyuv\n",
1445 stream->index);
1446 #endif
1447 }
1448
1449 vpx_usec_timer_start(&timer);
1450 vpx_codec_encode(&stream->encoder, img, frame_start,
1451 (unsigned long)(next_frame_start - frame_start), 0,
1452 global->deadline);
1453 vpx_usec_timer_mark(&timer);
1454 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1455 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1456 stream->index);
1457 }
1458
update_quantizer_histogram(struct stream_state * stream)1459 static void update_quantizer_histogram(struct stream_state *stream) {
1460 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1461 int q;
1462
1463 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1464 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1465 stream->counts[q]++;
1466 }
1467 }
1468
get_cx_data(struct stream_state * stream,struct VpxEncoderConfig * global,int * got_data)1469 static void get_cx_data(struct stream_state *stream,
1470 struct VpxEncoderConfig *global, int *got_data) {
1471 const vpx_codec_cx_pkt_t *pkt;
1472 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1473 vpx_codec_iter_t iter = NULL;
1474
1475 *got_data = 0;
1476 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1477 static size_t fsize = 0;
1478 static FileOffset ivf_header_pos = 0;
1479
1480 switch (pkt->kind) {
1481 case VPX_CODEC_CX_FRAME_PKT:
1482 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1483 stream->frames_out++;
1484 }
1485 if (!global->quiet)
1486 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1487
1488 update_rate_histogram(stream->rate_hist, cfg, pkt);
1489 #if CONFIG_WEBM_IO
1490 if (stream->config.write_webm) {
1491 write_webm_block(&stream->webm_ctx, cfg, pkt);
1492 }
1493 #endif
1494 if (!stream->config.write_webm) {
1495 if (pkt->data.frame.partition_id <= 0) {
1496 ivf_header_pos = ftello(stream->file);
1497 fsize = pkt->data.frame.sz;
1498
1499 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1500 } else {
1501 fsize += pkt->data.frame.sz;
1502
1503 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1504 const FileOffset currpos = ftello(stream->file);
1505 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1506 ivf_write_frame_size(stream->file, fsize);
1507 fseeko(stream->file, currpos, SEEK_SET);
1508 }
1509 }
1510
1511 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1512 stream->file);
1513 }
1514 stream->nbytes += pkt->data.raw.sz;
1515
1516 *got_data = 1;
1517 #if CONFIG_DECODERS
1518 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1519 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1520 (unsigned int)pkt->data.frame.sz, NULL, 0);
1521 if (stream->decoder.err) {
1522 warn_or_exit_on_error(&stream->decoder,
1523 global->test_decode == TEST_DECODE_FATAL,
1524 "Failed to decode frame %d in stream %d",
1525 stream->frames_out + 1, stream->index);
1526 stream->mismatch_seen = stream->frames_out + 1;
1527 }
1528 }
1529 #endif
1530 break;
1531 case VPX_CODEC_STATS_PKT:
1532 stream->frames_out++;
1533 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1534 pkt->data.twopass_stats.sz);
1535 stream->nbytes += pkt->data.raw.sz;
1536 break;
1537 case VPX_CODEC_PSNR_PKT:
1538
1539 if (global->show_psnr) {
1540 int i;
1541
1542 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1543 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1544 for (i = 0; i < 4; i++) {
1545 if (!global->quiet)
1546 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1547 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1548 }
1549 stream->psnr_count++;
1550 }
1551
1552 break;
1553 default: break;
1554 }
1555 }
1556 }
1557
show_psnr(struct stream_state * stream,double peak)1558 static void show_psnr(struct stream_state *stream, double peak) {
1559 int i;
1560 double ovpsnr;
1561
1562 if (!stream->psnr_count) return;
1563
1564 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1565 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1566 (double)stream->psnr_sse_total);
1567 fprintf(stderr, " %.3f", ovpsnr);
1568
1569 for (i = 0; i < 4; i++) {
1570 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1571 }
1572 fprintf(stderr, "\n");
1573 }
1574
usec_to_fps(uint64_t usec,unsigned int frames)1575 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1576 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1577 }
1578
test_decode(struct stream_state * stream,enum TestDecodeFatality fatal,const VpxInterface * codec)1579 static void test_decode(struct stream_state *stream,
1580 enum TestDecodeFatality fatal,
1581 const VpxInterface *codec) {
1582 vpx_image_t enc_img, dec_img;
1583
1584 if (stream->mismatch_seen) return;
1585
1586 /* Get the internal reference frame */
1587 if (strcmp(codec->name, "vp8") == 0) {
1588 struct vpx_ref_frame ref_enc, ref_dec;
1589 int aligned_width = (stream->config.cfg.g_w + 15) & ~15;
1590 int aligned_height = (stream->config.cfg.g_h + 15) & ~15;
1591
1592 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, aligned_width, aligned_height,
1593 1);
1594 enc_img = ref_enc.img;
1595 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, aligned_width, aligned_height,
1596 1);
1597 dec_img = ref_dec.img;
1598
1599 ref_enc.frame_type = VP8_LAST_FRAME;
1600 ref_dec.frame_type = VP8_LAST_FRAME;
1601 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1602 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1603 } else {
1604 struct vp9_ref_frame ref_enc, ref_dec;
1605
1606 ref_enc.idx = 0;
1607 ref_dec.idx = 0;
1608 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1609 enc_img = ref_enc.img;
1610 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1611 dec_img = ref_dec.img;
1612 #if CONFIG_VP9_HIGHBITDEPTH
1613 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1614 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1615 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1616 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1617 enc_img.d_w, enc_img.d_h, 16);
1618 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1619 }
1620 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1621 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1622 dec_img.d_w, dec_img.d_h, 16);
1623 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1624 }
1625 }
1626 #endif
1627 }
1628 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1629 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1630
1631 if (!compare_img(&enc_img, &dec_img)) {
1632 int y[4], u[4], v[4];
1633 #if CONFIG_VP9_HIGHBITDEPTH
1634 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1635 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1636 } else {
1637 find_mismatch(&enc_img, &dec_img, y, u, v);
1638 }
1639 #else
1640 find_mismatch(&enc_img, &dec_img, y, u, v);
1641 #endif
1642 stream->decoder.err = 1;
1643 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1644 "Stream %d: Encode/decode mismatch on frame %d at"
1645 " Y[%d, %d] {%d/%d},"
1646 " U[%d, %d] {%d/%d},"
1647 " V[%d, %d] {%d/%d}",
1648 stream->index, stream->frames_out, y[0], y[1], y[2],
1649 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1650 stream->mismatch_seen = stream->frames_out;
1651 }
1652
1653 vpx_img_free(&enc_img);
1654 vpx_img_free(&dec_img);
1655 }
1656
print_time(const char * label,int64_t etl)1657 static void print_time(const char *label, int64_t etl) {
1658 int64_t hours;
1659 int64_t mins;
1660 int64_t secs;
1661
1662 if (etl >= 0) {
1663 hours = etl / 3600;
1664 etl -= hours * 3600;
1665 mins = etl / 60;
1666 etl -= mins * 60;
1667 secs = etl;
1668
1669 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1670 hours, mins, secs);
1671 } else {
1672 fprintf(stderr, "[%3s unknown] ", label);
1673 }
1674 }
1675
main(int argc,const char ** argv_)1676 int main(int argc, const char **argv_) {
1677 int pass;
1678 vpx_image_t raw;
1679 #if CONFIG_VP9_HIGHBITDEPTH
1680 vpx_image_t raw_shift;
1681 int allocated_raw_shift = 0;
1682 int use_16bit_internal = 0;
1683 int input_shift = 0;
1684 #endif
1685 int frame_avail, got_data;
1686
1687 struct VpxInputContext input;
1688 struct VpxEncoderConfig global;
1689 struct stream_state *streams = NULL;
1690 char **argv, **argi;
1691 uint64_t cx_time = 0;
1692 int stream_cnt = 0;
1693 int res = 0;
1694
1695 memset(&input, 0, sizeof(input));
1696 memset(&raw, 0, sizeof(raw));
1697 exec_name = argv_[0];
1698
1699 /* Setup default input stream settings */
1700 input.framerate.numerator = 30;
1701 input.framerate.denominator = 1;
1702 input.only_i420 = 1;
1703 input.bit_depth = 0;
1704
1705 /* First parse the global configuration values, because we want to apply
1706 * other parameters on top of the default configuration provided by the
1707 * codec.
1708 */
1709 argv = argv_dup(argc - 1, argv_ + 1);
1710 if (!argv) {
1711 fprintf(stderr, "Error allocating argument list\n");
1712 return EXIT_FAILURE;
1713 }
1714 parse_global_config(&global, argv);
1715
1716 if (argc < 3) usage_exit();
1717
1718 switch (global.color_type) {
1719 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1720 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1721 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1722 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1723 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1724 case NV12: input.fmt = VPX_IMG_FMT_NV12; break;
1725 }
1726
1727 {
1728 /* Now parse each stream's parameters. Using a local scope here
1729 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1730 * loops
1731 */
1732 struct stream_state *stream = NULL;
1733
1734 do {
1735 stream = new_stream(&global, stream);
1736 stream_cnt++;
1737 if (!streams) streams = stream;
1738 } while (parse_stream_params(&global, stream, argv));
1739 }
1740
1741 /* Check for unrecognized options */
1742 for (argi = argv; *argi; argi++)
1743 if (argi[0][0] == '-' && argi[0][1])
1744 die("Error: Unrecognized option %s\n", *argi);
1745
1746 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1747 &stream->config.cfg););
1748
1749 /* Handle non-option arguments */
1750 input.filename = argv[0];
1751
1752 if (!input.filename) {
1753 fprintf(stderr, "No input file specified!\n");
1754 usage_exit();
1755 }
1756
1757 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1758 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1759
1760 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1761 int frames_in = 0, seen_frames = 0;
1762 int64_t estimated_time_left = -1;
1763 int64_t average_rate = -1;
1764 int64_t lagged_count = 0;
1765
1766 open_input_file(&input);
1767
1768 /* If the input file doesn't specify its w/h (raw files), try to get
1769 * the data from the first stream's configuration.
1770 */
1771 if (!input.width || !input.height) {
1772 FOREACH_STREAM({
1773 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1774 input.width = stream->config.cfg.g_w;
1775 input.height = stream->config.cfg.g_h;
1776 break;
1777 }
1778 });
1779 }
1780
1781 /* Update stream configurations from the input file's parameters */
1782 if (!input.width || !input.height)
1783 fatal(
1784 "Specify stream dimensions with --width (-w) "
1785 " and --height (-h)");
1786
1787 /* If input file does not specify bit-depth but input-bit-depth parameter
1788 * exists, assume that to be the input bit-depth. However, if the
1789 * input-bit-depth paramter does not exist, assume the input bit-depth
1790 * to be the same as the codec bit-depth.
1791 */
1792 if (!input.bit_depth) {
1793 FOREACH_STREAM({
1794 if (stream->config.cfg.g_input_bit_depth)
1795 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1796 else
1797 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1798 (int)stream->config.cfg.g_bit_depth;
1799 });
1800 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1801 } else {
1802 FOREACH_STREAM(
1803 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
1804 }
1805
1806 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1807 FOREACH_STREAM(validate_stream_config(stream, &global));
1808
1809 /* Ensure that --passes and --pass are consistent. If --pass is set and
1810 * --passes=2, ensure --fpf was set.
1811 */
1812 if (global.pass && global.passes == 2)
1813 FOREACH_STREAM({
1814 if (!stream->config.stats_fn)
1815 die("Stream %d: Must specify --fpf when --pass=%d"
1816 " and --passes=2\n",
1817 stream->index, global.pass);
1818 });
1819
1820 #if !CONFIG_WEBM_IO
1821 FOREACH_STREAM({
1822 if (stream->config.write_webm) {
1823 stream->config.write_webm = 0;
1824 warn(
1825 "vpxenc was compiled without WebM container support."
1826 "Producing IVF output");
1827 }
1828 });
1829 #endif
1830
1831 /* Use the frame rate from the file only if none was specified
1832 * on the command-line.
1833 */
1834 if (!global.have_framerate) {
1835 global.framerate.num = input.framerate.numerator;
1836 global.framerate.den = input.framerate.denominator;
1837 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
1838 stream->config.cfg.g_timebase.num = global.framerate.den);
1839 }
1840
1841 /* Show configuration */
1842 if (global.verbose && pass == 0)
1843 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1844
1845 if (pass == (global.pass ? global.pass - 1 : 0)) {
1846 // The Y4M reader does its own allocation.
1847 if (input.file_type != FILE_TYPE_Y4M) {
1848 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1849 }
1850 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
1851 &stream->config.cfg, &global.framerate));
1852 }
1853
1854 FOREACH_STREAM(setup_pass(stream, &global, pass));
1855 FOREACH_STREAM(
1856 open_output_file(stream, &global, &input.pixel_aspect_ratio));
1857 FOREACH_STREAM(initialize_encoder(stream, &global));
1858
1859 #if CONFIG_VP9_HIGHBITDEPTH
1860 if (strcmp(global.codec->name, "vp9") == 0) {
1861 // Check to see if at least one stream uses 16 bit internal.
1862 // Currently assume that the bit_depths for all streams using
1863 // highbitdepth are the same.
1864 FOREACH_STREAM({
1865 if (stream->config.use_16bit_internal) {
1866 use_16bit_internal = 1;
1867 }
1868 if (stream->config.cfg.g_profile == 0) {
1869 input_shift = 0;
1870 } else {
1871 input_shift = (int)stream->config.cfg.g_bit_depth -
1872 stream->config.cfg.g_input_bit_depth;
1873 }
1874 });
1875 }
1876 #endif
1877
1878 frame_avail = 1;
1879 got_data = 0;
1880
1881 while (frame_avail || got_data) {
1882 struct vpx_usec_timer timer;
1883
1884 if (!global.limit || frames_in < global.limit) {
1885 frame_avail = read_frame(&input, &raw);
1886
1887 if (frame_avail) frames_in++;
1888 seen_frames =
1889 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
1890
1891 if (!global.quiet) {
1892 float fps = usec_to_fps(cx_time, seen_frames);
1893 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1894
1895 if (stream_cnt == 1)
1896 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
1897 streams->frames_out, (int64_t)streams->nbytes);
1898 else
1899 fprintf(stderr, "frame %4d ", frames_in);
1900
1901 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
1902 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1903 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
1904 fps >= 1.0 ? "fps" : "fpm");
1905 print_time("ETA", estimated_time_left);
1906 }
1907
1908 } else
1909 frame_avail = 0;
1910
1911 if (frames_in > global.skip_frames) {
1912 #if CONFIG_VP9_HIGHBITDEPTH
1913 vpx_image_t *frame_to_encode;
1914 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
1915 assert(use_16bit_internal);
1916 // Input bit depth and stream bit depth do not match, so up
1917 // shift frame to stream bit depth
1918 if (!allocated_raw_shift) {
1919 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
1920 input.width, input.height, 32);
1921 allocated_raw_shift = 1;
1922 }
1923 vpx_img_upshift(&raw_shift, &raw, input_shift);
1924 frame_to_encode = &raw_shift;
1925 } else {
1926 frame_to_encode = &raw;
1927 }
1928 vpx_usec_timer_start(&timer);
1929 if (use_16bit_internal) {
1930 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
1931 FOREACH_STREAM({
1932 if (stream->config.use_16bit_internal)
1933 encode_frame(stream, &global,
1934 frame_avail ? frame_to_encode : NULL, frames_in);
1935 else
1936 assert(0);
1937 });
1938 } else {
1939 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
1940 FOREACH_STREAM(encode_frame(stream, &global,
1941 frame_avail ? frame_to_encode : NULL,
1942 frames_in));
1943 }
1944 #else
1945 vpx_usec_timer_start(&timer);
1946 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
1947 frames_in));
1948 #endif
1949 vpx_usec_timer_mark(&timer);
1950 cx_time += vpx_usec_timer_elapsed(&timer);
1951
1952 FOREACH_STREAM(update_quantizer_histogram(stream));
1953
1954 got_data = 0;
1955 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1956
1957 if (!got_data && input.length && streams != NULL &&
1958 !streams->frames_out) {
1959 lagged_count = global.limit ? seen_frames : ftello(input.file);
1960 } else if (input.length) {
1961 int64_t remaining;
1962 int64_t rate;
1963
1964 if (global.limit) {
1965 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1966
1967 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1968 remaining = 1000 * (global.limit - global.skip_frames -
1969 seen_frames + lagged_count);
1970 } else {
1971 const int64_t input_pos = ftello(input.file);
1972 const int64_t input_pos_lagged = input_pos - lagged_count;
1973
1974 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1975 remaining = input.length - input_pos + lagged_count;
1976 }
1977
1978 average_rate =
1979 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
1980 estimated_time_left = average_rate ? remaining / average_rate : -1;
1981 }
1982
1983 if (got_data && global.test_decode != TEST_DECODE_OFF)
1984 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1985 }
1986
1987 fflush(stdout);
1988 if (!global.quiet) fprintf(stderr, "\033[K");
1989 }
1990
1991 if (stream_cnt > 1) fprintf(stderr, "\n");
1992
1993 if (!global.quiet) {
1994 FOREACH_STREAM(fprintf(
1995 stderr,
1996 "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64 "b/f %7" PRId64
1997 "b/s %7" PRId64 " %s (%.2f fps)\033[K\n",
1998 pass + 1, global.passes, frames_in, stream->frames_out,
1999 (int64_t)stream->nbytes,
2000 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2001 seen_frames
2002 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
2003 global.framerate.den / seen_frames
2004 : 0,
2005 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2006 stream->cx_time > 9999999 ? "ms" : "us",
2007 usec_to_fps(stream->cx_time, seen_frames)));
2008 }
2009
2010 if (global.show_psnr) {
2011 if (global.codec->fourcc == VP9_FOURCC) {
2012 FOREACH_STREAM(
2013 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2014 } else {
2015 FOREACH_STREAM(show_psnr(stream, 255.0));
2016 }
2017 }
2018
2019 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2020
2021 if (global.test_decode != TEST_DECODE_OFF) {
2022 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2023 }
2024
2025 close_input_file(&input);
2026
2027 if (global.test_decode == TEST_DECODE_FATAL) {
2028 FOREACH_STREAM(res |= stream->mismatch_seen);
2029 }
2030 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2031
2032 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2033
2034 if (global.pass) break;
2035 }
2036
2037 if (global.show_q_hist_buckets)
2038 FOREACH_STREAM(
2039 show_q_histogram(stream->counts, global.show_q_hist_buckets));
2040
2041 if (global.show_rate_hist_buckets)
2042 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2043 global.show_rate_hist_buckets));
2044 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2045
2046 #if CONFIG_INTERNAL_STATS
2047 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2048 * to match some existing utilities.
2049 */
2050 if (!(global.pass == 1 && global.passes == 2))
2051 FOREACH_STREAM({
2052 FILE *f = fopen("opsnr.stt", "a");
2053 if (stream->mismatch_seen) {
2054 fprintf(f, "First mismatch occurred in frame %d\n",
2055 stream->mismatch_seen);
2056 } else {
2057 fprintf(f, "No mismatch detected in recon buffers\n");
2058 }
2059 fclose(f);
2060 });
2061 #endif
2062
2063 #if CONFIG_VP9_HIGHBITDEPTH
2064 if (allocated_raw_shift) vpx_img_free(&raw_shift);
2065 #endif
2066 vpx_img_free(&raw);
2067 free(argv);
2068 free(streams);
2069 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2070 }
2071