?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge "VP9: Do not call vp9_adjust_mask() in vp9_setup_mask()"
[aom.git] / vpxenc.c
blob50e7c7fc7947de7d1342a403d8af2d39cfd3c627
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 || CONFIG_VP10_ENCODER
36 #include "vpx/vp8cx.h"
37 #endif
38 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER || CONFIG_VP10_DECODER
39 #include "vpx/vp8dx.h"
40 #endif
41
42 #include "vpx/vpx_integer.h"
43 #include "vpx_ports/mem_ops.h"
44 #include "vpx_ports/vpx_timer.h"
45 #include "./rate_hist.h"
46 #include "./vpxstats.h"
47 #include "./warnings.h"
48 #if CONFIG_WEBM_IO
49 #include "./webmenc.h"
50 #endif
51 #include "./y4minput.h"
52
53 /* Swallow warnings about unused results of fread/fwrite */
54 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
55 FILE *stream) {
56 return fread(ptr, size, nmemb, stream);
57 }
58 #define fread wrap_fread
59
60 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
61 FILE *stream) {
62 return fwrite(ptr, size, nmemb, stream);
63 }
64 #define fwrite wrap_fwrite
65
66
67 static const char *exec_name;
68
69 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
70 const char *s, va_list ap) {
71 if (ctx->err) {
72 const char *detail = vpx_codec_error_detail(ctx);
73
74 vfprintf(stderr, s, ap);
75 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
76
77 if (detail)
78 fprintf(stderr, " %s\n", detail);
79
80 if (fatal)
81 exit(EXIT_FAILURE);
82 }
83 }
84
85 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
86 va_list ap;
87
88 va_start(ap, s);
89 warn_or_exit_on_errorv(ctx, 1, s, ap);
90 va_end(ap);
91 }
92
93 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
94 const char *s, ...) {
95 va_list ap;
96
97 va_start(ap, s);
98 warn_or_exit_on_errorv(ctx, fatal, s, ap);
99 va_end(ap);
100 }
101
102 static int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
103 FILE *f = input_ctx->file;
104 y4m_input *y4m = &input_ctx->y4m;
105 int shortread = 0;
106
107 if (input_ctx->file_type == FILE_TYPE_Y4M) {
108 if (y4m_input_fetch_frame(y4m, f, img) < 1)
109 return 0;
110 } else {
111 shortread = read_yuv_frame(input_ctx, img);
112 }
113
114 return !shortread;
115 }
116
117 static int file_is_y4m(const char detect[4]) {
118 if (memcmp(detect, "YUV4", 4) == 0) {
119 return 1;
120 }
121 return 0;
122 }
123
124 static int fourcc_is_ivf(const char detect[4]) {
125 if (memcmp(detect, "DKIF", 4) == 0) {
126 return 1;
127 }
128 return 0;
129 }
130
131 static const arg_def_t debugmode = ARG_DEF(
132 "D", "debug", 0, "Debug mode (makes output deterministic)");
133 static const arg_def_t outputfile = ARG_DEF(
134 "o", "output", 1, "Output filename");
135 static const arg_def_t use_yv12 = ARG_DEF(
136 NULL, "yv12", 0, "Input file is YV12 ");
137 static const arg_def_t use_i420 = ARG_DEF(
138 NULL, "i420", 0, "Input file is I420 (default)");
139 static const arg_def_t use_i422 = ARG_DEF(
140 NULL, "i422", 0, "Input file is I422");
141 static const arg_def_t use_i444 = ARG_DEF(
142 NULL, "i444", 0, "Input file is I444");
143 static const arg_def_t use_i440 = ARG_DEF(
144 NULL, "i440", 0, "Input file is I440");
145 static const arg_def_t codecarg = ARG_DEF(
146 NULL, "codec", 1, "Codec to use");
147 static const arg_def_t passes = ARG_DEF(
148 "p", "passes", 1, "Number of passes (1/2)");
149 static const arg_def_t pass_arg = ARG_DEF(
150 NULL, "pass", 1, "Pass to execute (1/2)");
151 static const arg_def_t fpf_name = ARG_DEF(
152 NULL, "fpf", 1, "First pass statistics file name");
153 #if CONFIG_FP_MB_STATS
154 static const arg_def_t fpmbf_name = ARG_DEF(
155 NULL, "fpmbf", 1, "First pass block statistics file name");
156 #endif
157 static const arg_def_t limit = ARG_DEF(
158 NULL, "limit", 1, "Stop encoding after n input frames");
159 static const arg_def_t skip = ARG_DEF(
160 NULL, "skip", 1, "Skip the first n input frames");
161 static const arg_def_t deadline = ARG_DEF(
162 "d", "deadline", 1, "Deadline per frame (usec)");
163 static const arg_def_t best_dl = ARG_DEF(
164 NULL, "best", 0, "Use Best Quality Deadline");
165 static const arg_def_t good_dl = ARG_DEF(
166 NULL, "good", 0, "Use Good Quality Deadline");
167 static const arg_def_t rt_dl = ARG_DEF(
168 NULL, "rt", 0, "Use Realtime Quality Deadline");
169 static const arg_def_t quietarg = ARG_DEF(
170 "q", "quiet", 0, "Do not print encode progress");
171 static const arg_def_t verbosearg = ARG_DEF(
172 "v", "verbose", 0, "Show encoder parameters");
173 static const arg_def_t psnrarg = ARG_DEF(
174 NULL, "psnr", 0, "Show PSNR in status line");
175
176 static const struct arg_enum_list test_decode_enum[] = {
177 {"off", TEST_DECODE_OFF},
178 {"fatal", TEST_DECODE_FATAL},
179 {"warn", TEST_DECODE_WARN},
180 {NULL, 0}
181 };
182 static const arg_def_t recontest = ARG_DEF_ENUM(
183 NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
184 static const arg_def_t framerate = ARG_DEF(
185 NULL, "fps", 1, "Stream frame rate (rate/scale)");
186 static const arg_def_t use_webm = ARG_DEF(
187 NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
188 static const arg_def_t use_ivf = ARG_DEF(
189 NULL, "ivf", 0, "Output IVF");
190 static const arg_def_t out_part = ARG_DEF(
191 "P", "output-partitions", 0,
192 "Makes encoder output partitions. Requires IVF output!");
193 static const arg_def_t q_hist_n = ARG_DEF(
194 NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
195 static const arg_def_t rate_hist_n = ARG_DEF(
196 NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
197 static const arg_def_t disable_warnings = ARG_DEF(
198 NULL, "disable-warnings", 0,
199 "Disable warnings about potentially incorrect encode settings.");
200 static const arg_def_t disable_warning_prompt = ARG_DEF(
201 "y", "disable-warning-prompt", 0,
202 "Display warnings, but do not prompt user to continue.");
203
204 #if CONFIG_VP9_HIGHBITDEPTH
205 static const arg_def_t test16bitinternalarg = ARG_DEF(
206 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
207 #endif
208
209 static const arg_def_t *main_args[] = {
210 &debugmode,
211 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
212 &deadline, &best_dl, &good_dl, &rt_dl,
213 &quietarg, &verbosearg, &psnrarg, &use_webm, &use_ivf, &out_part, &q_hist_n,
214 &rate_hist_n, &disable_warnings, &disable_warning_prompt, &recontest,
215 NULL
216 };
217
218 static const arg_def_t usage = ARG_DEF(
219 "u", "usage", 1, "Usage profile number to use");
220 static const arg_def_t threads = ARG_DEF(
221 "t", "threads", 1, "Max number of threads to use");
222 static const arg_def_t profile = ARG_DEF(
223 NULL, "profile", 1, "Bitstream profile number to use");
224 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
225 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
226 #if CONFIG_WEBM_IO
227 static const struct arg_enum_list stereo_mode_enum[] = {
228 {"mono", STEREO_FORMAT_MONO},
229 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
230 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
231 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
232 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
233 {NULL, 0}
234 };
235 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
236 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
237 #endif
238 static const arg_def_t timebase = ARG_DEF(
239 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
240 static const arg_def_t error_resilient = ARG_DEF(
241 NULL, "error-resilient", 1, "Enable error resiliency features");
242 static const arg_def_t lag_in_frames = ARG_DEF(
243 NULL, "lag-in-frames", 1, "Max number of frames to lag");
244
245 static const arg_def_t *global_args[] = {
246 &use_yv12, &use_i420, &use_i422, &use_i444, &use_i440,
247 &usage, &threads, &profile,
248 &width, &height,
249 #if CONFIG_WEBM_IO
250 &stereo_mode,
251 #endif
252 &timebase, &framerate,
253 &error_resilient,
254 #if CONFIG_VP9_HIGHBITDEPTH
255 &test16bitinternalarg,
256 #endif
257 &lag_in_frames, NULL
258 };
259
260 static const arg_def_t dropframe_thresh = ARG_DEF(
261 NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
262 static const arg_def_t resize_allowed = ARG_DEF(
263 NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
264 static const arg_def_t resize_width = ARG_DEF(
265 NULL, "resize-width", 1, "Width of encoded frame");
266 static const arg_def_t resize_height = ARG_DEF(
267 NULL, "resize-height", 1, "Height of encoded frame");
268 static const arg_def_t resize_up_thresh = ARG_DEF(
269 NULL, "resize-up", 1, "Upscale threshold (buf %)");
270 static const arg_def_t resize_down_thresh = ARG_DEF(
271 NULL, "resize-down", 1, "Downscale threshold (buf %)");
272 static const struct arg_enum_list end_usage_enum[] = {
273 {"vbr", VPX_VBR},
274 {"cbr", VPX_CBR},
275 {"cq", VPX_CQ},
276 {"q", VPX_Q},
277 {NULL, 0}
278 };
279 static const arg_def_t end_usage = ARG_DEF_ENUM(
280 NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
281 static const arg_def_t target_bitrate = ARG_DEF(
282 NULL, "target-bitrate", 1, "Bitrate (kbps)");
283 static const arg_def_t min_quantizer = ARG_DEF(
284 NULL, "min-q", 1, "Minimum (best) quantizer");
285 static const arg_def_t max_quantizer = ARG_DEF(
286 NULL, "max-q", 1, "Maximum (worst) quantizer");
287 static const arg_def_t undershoot_pct = ARG_DEF(
288 NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
289 static const arg_def_t overshoot_pct = ARG_DEF(
290 NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
291 static const arg_def_t buf_sz = ARG_DEF(
292 NULL, "buf-sz", 1, "Client buffer size (ms)");
293 static const arg_def_t buf_initial_sz = ARG_DEF(
294 NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
295 static const arg_def_t buf_optimal_sz = ARG_DEF(
296 NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
297 static const arg_def_t *rc_args[] = {
298 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
299 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
300 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
301 &buf_initial_sz, &buf_optimal_sz, NULL
302 };
303
304
305 static const arg_def_t bias_pct = ARG_DEF(
306 NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
307 static const arg_def_t minsection_pct = ARG_DEF(
308 NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
309 static const arg_def_t maxsection_pct = ARG_DEF(
310 NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
311 static const arg_def_t *rc_twopass_args[] = {
312 &bias_pct, &minsection_pct, &maxsection_pct, NULL
313 };
314
315
316 static const arg_def_t kf_min_dist = ARG_DEF(
317 NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
318 static const arg_def_t kf_max_dist = ARG_DEF(
319 NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
320 static const arg_def_t kf_disabled = ARG_DEF(
321 NULL, "disable-kf", 0, "Disable keyframe placement");
322 static const arg_def_t *kf_args[] = {
323 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
324 };
325
326
327 static const arg_def_t noise_sens = ARG_DEF(
328 NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
329 static const arg_def_t sharpness = ARG_DEF(
330 NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
331 static const arg_def_t static_thresh = ARG_DEF(
332 NULL, "static-thresh", 1, "Motion detection threshold");
333 static const arg_def_t auto_altref = ARG_DEF(
334 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
335 static const arg_def_t arnr_maxframes = ARG_DEF(
336 NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
337 static const arg_def_t arnr_strength = ARG_DEF(
338 NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
339 static const arg_def_t arnr_type = ARG_DEF(
340 NULL, "arnr-type", 1, "AltRef type");
341 static const struct arg_enum_list tuning_enum[] = {
342 {"psnr", VP8_TUNE_PSNR},
343 {"ssim", VP8_TUNE_SSIM},
344 {NULL, 0}
345 };
346 static const arg_def_t tune_ssim = ARG_DEF_ENUM(
347 NULL, "tune", 1, "Material to favor", tuning_enum);
348 static const arg_def_t cq_level = ARG_DEF(
349 NULL, "cq-level", 1, "Constant/Constrained Quality level");
350 static const arg_def_t max_intra_rate_pct = ARG_DEF(
351 NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
352
353 #if CONFIG_VP8_ENCODER
354 static const arg_def_t cpu_used_vp8 = ARG_DEF(
355 NULL, "cpu-used", 1, "CPU Used (-16..16)");
356 static const arg_def_t token_parts = ARG_DEF(
357 NULL, "token-parts", 1, "Number of token partitions to use, log2");
358 static const arg_def_t screen_content_mode = ARG_DEF(
359 NULL, "screen-content-mode", 1, "Screen content mode");
360 static const arg_def_t *vp8_args[] = {
361 &cpu_used_vp8, &auto_altref, &noise_sens, &sharpness, &static_thresh,
362 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
363 &tune_ssim, &cq_level, &max_intra_rate_pct, &screen_content_mode,
364 NULL
365 };
366 static const int vp8_arg_ctrl_map[] = {
367 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
368 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
369 VP8E_SET_TOKEN_PARTITIONS,
370 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
371 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
372 VP8E_SET_SCREEN_CONTENT_MODE,
373 0
374 };
375 #endif
376
377 #if CONFIG_VP9_ENCODER || CONFIG_VP10_ENCODER
378 static const arg_def_t cpu_used_vp9 = ARG_DEF(
379 NULL, "cpu-used", 1, "CPU Used (-8..8)");
380 static const arg_def_t tile_cols = ARG_DEF(
381 NULL, "tile-columns", 1, "Number of tile columns to use, log2");
382 static const arg_def_t tile_rows = ARG_DEF(
383 NULL, "tile-rows", 1,
384 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
385 static const arg_def_t lossless = ARG_DEF(
386 NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
387 static const arg_def_t frame_parallel_decoding = ARG_DEF(
388 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
389 static const arg_def_t aq_mode = ARG_DEF(
390 NULL, "aq-mode", 1,
391 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
392 "3: cyclic refresh, 4: equator360)");
393 static const arg_def_t frame_periodic_boost = ARG_DEF(
394 NULL, "frame-boost", 1,
395 "Enable frame periodic boost (0: off (default), 1: on)");
396 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
397 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
398 static const arg_def_t max_inter_rate_pct = ARG_DEF(
399 NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
400 static const arg_def_t min_gf_interval = ARG_DEF(
401 NULL, "min-gf-interval", 1,
402 "min gf/arf frame interval (default 0, indicating in-built behavior)");
403 static const arg_def_t max_gf_interval = ARG_DEF(
404 NULL, "max-gf-interval", 1,
405 "max gf/arf frame interval (default 0, indicating in-built behavior)");
406
407 static const struct arg_enum_list color_space_enum[] = {
408 { "unknown", VPX_CS_UNKNOWN },
409 { "bt601", VPX_CS_BT_601 },
410 { "bt709", VPX_CS_BT_709 },
411 { "smpte170", VPX_CS_SMPTE_170 },
412 { "smpte240", VPX_CS_SMPTE_240 },
413 { "bt2020", VPX_CS_BT_2020 },
414 { "reserved", VPX_CS_RESERVED },
415 { "sRGB", VPX_CS_SRGB },
416 { NULL, 0 }
417 };
418
419 static const arg_def_t input_color_space = ARG_DEF_ENUM(
420 NULL, "color-space", 1,
421 "The color space of input content:", color_space_enum);
422
423 #if CONFIG_VP9_HIGHBITDEPTH
424 static const struct arg_enum_list bitdepth_enum[] = {
425 {"8", VPX_BITS_8},
426 {"10", VPX_BITS_10},
427 {"12", VPX_BITS_12},
428 {NULL, 0}
429 };
430
431 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
432 "b", "bit-depth", 1,
433 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
434 bitdepth_enum);
435 static const arg_def_t inbitdeptharg = ARG_DEF(
436 NULL, "input-bit-depth", 1, "Bit depth of input");
437 #endif
438
439 static const struct arg_enum_list tune_content_enum[] = {
440 {"default", VP9E_CONTENT_DEFAULT},
441 {"screen", VP9E_CONTENT_SCREEN},
442 {NULL, 0}
443 };
444
445 static const arg_def_t tune_content = ARG_DEF_ENUM(
446 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
447 #endif
448
449 #if CONFIG_VP9_ENCODER
450 static const arg_def_t *vp9_args[] = {
451 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
452 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
453 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
454 &gf_cbr_boost_pct, &lossless,
455 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
456 &noise_sens, &tune_content, &input_color_space,
457 &min_gf_interval, &max_gf_interval,
458 #if CONFIG_VP9_HIGHBITDEPTH
459 &bitdeptharg, &inbitdeptharg,
460 #endif // CONFIG_VP9_HIGHBITDEPTH
461 NULL
462 };
463 static const int vp9_arg_ctrl_map[] = {
464 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
465 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
466 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
467 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
468 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
469 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
470 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
471 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
472 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
473 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL,
474 0
475 };
476 #endif
477
478 #if CONFIG_VP10_ENCODER
479 static const arg_def_t *vp10_args[] = {
480 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
481 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
482 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
483 &gf_cbr_boost_pct, &lossless,
484 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
485 &noise_sens, &tune_content, &input_color_space,
486 &min_gf_interval, &max_gf_interval,
487 #if CONFIG_VP9_HIGHBITDEPTH
488 &bitdeptharg, &inbitdeptharg,
489 #endif // CONFIG_VP9_HIGHBITDEPTH
490 NULL
491 };
492 static const int vp10_arg_ctrl_map[] = {
493 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
494 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
495 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
496 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
497 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
498 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
499 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
500 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
501 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
502 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL,
503 0
504 };
505 #endif
506
507 static const arg_def_t *no_args[] = { NULL };
508
509 void usage_exit(void) {
510 int i;
511 const int num_encoder = get_vpx_encoder_count();
512
513 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
514 exec_name);
515
516 fprintf(stderr, "\nOptions:\n");
517 arg_show_usage(stderr, main_args);
518 fprintf(stderr, "\nEncoder Global Options:\n");
519 arg_show_usage(stderr, global_args);
520 fprintf(stderr, "\nRate Control Options:\n");
521 arg_show_usage(stderr, rc_args);
522 fprintf(stderr, "\nTwopass Rate Control Options:\n");
523 arg_show_usage(stderr, rc_twopass_args);
524 fprintf(stderr, "\nKeyframe Placement Options:\n");
525 arg_show_usage(stderr, kf_args);
526 #if CONFIG_VP8_ENCODER
527 fprintf(stderr, "\nVP8 Specific Options:\n");
528 arg_show_usage(stderr, vp8_args);
529 #endif
530 #if CONFIG_VP9_ENCODER
531 fprintf(stderr, "\nVP9 Specific Options:\n");
532 arg_show_usage(stderr, vp9_args);
533 #endif
534 #if CONFIG_VP10_ENCODER
535 fprintf(stderr, "\nVP10 Specific Options:\n");
536 arg_show_usage(stderr, vp10_args);
537 #endif
538 fprintf(stderr, "\nStream timebase (--timebase):\n"
539 " The desired precision of timestamps in the output, expressed\n"
540 " in fractional seconds. Default is 1/1000.\n");
541 fprintf(stderr, "\nIncluded encoders:\n\n");
542
543 for (i = 0; i < num_encoder; ++i) {
544 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
545 const char* defstr = (i == (num_encoder - 1)) ? "(default)" : "";
546 fprintf(stderr, " %-6s - %s %s\n",
547 encoder->name, vpx_codec_iface_name(encoder->codec_interface()),
548 defstr);
549 }
550 fprintf(stderr, "\n ");
551 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
552
553 exit(EXIT_FAILURE);
554 }
555
556 #define mmin(a, b) ((a) < (b) ? (a) : (b))
557
558 #if CONFIG_VP9_HIGHBITDEPTH
559 static void find_mismatch_high(const vpx_image_t *const img1,
560 const vpx_image_t *const img2,
561 int yloc[4], int uloc[4], int vloc[4]) {
562 uint16_t *plane1, *plane2;
563 uint32_t stride1, stride2;
564 const uint32_t bsize = 64;
565 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
566 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
567 const uint32_t c_w =
568 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
569 const uint32_t c_h =
570 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
571 int match = 1;
572 uint32_t i, j;
573 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
574 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
575 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
576 stride1 = img1->stride[VPX_PLANE_Y]/2;
577 stride2 = img2->stride[VPX_PLANE_Y]/2;
578 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
579 for (j = 0; match && j < img1->d_w; j += bsize) {
580 int k, l;
581 const int si = mmin(i + bsize, img1->d_h) - i;
582 const int sj = mmin(j + bsize, img1->d_w) - j;
583 for (k = 0; match && k < si; ++k) {
584 for (l = 0; match && l < sj; ++l) {
585 if (*(plane1 + (i + k) * stride1 + j + l) !=
586 *(plane2 + (i + k) * stride2 + j + l)) {
587 yloc[0] = i + k;
588 yloc[1] = j + l;
589 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
590 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
591 match = 0;
592 break;
593 }
594 }
595 }
596 }
597 }
598
599 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
600 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
601 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
602 stride1 = img1->stride[VPX_PLANE_U]/2;
603 stride2 = img2->stride[VPX_PLANE_U]/2;
604 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
605 for (j = 0; match && j < c_w; j += bsizex) {
606 int k, l;
607 const int si = mmin(i + bsizey, c_h - i);
608 const int sj = mmin(j + bsizex, c_w - j);
609 for (k = 0; match && k < si; ++k) {
610 for (l = 0; match && l < sj; ++l) {
611 if (*(plane1 + (i + k) * stride1 + j + l) !=
612 *(plane2 + (i + k) * stride2 + j + l)) {
613 uloc[0] = i + k;
614 uloc[1] = j + l;
615 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
616 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
617 match = 0;
618 break;
619 }
620 }
621 }
622 }
623 }
624
625 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
626 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
627 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
628 stride1 = img1->stride[VPX_PLANE_V]/2;
629 stride2 = img2->stride[VPX_PLANE_V]/2;
630 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
631 for (j = 0; match && j < c_w; j += bsizex) {
632 int k, l;
633 const int si = mmin(i + bsizey, c_h - i);
634 const int sj = mmin(j + bsizex, c_w - j);
635 for (k = 0; match && k < si; ++k) {
636 for (l = 0; match && l < sj; ++l) {
637 if (*(plane1 + (i + k) * stride1 + j + l) !=
638 *(plane2 + (i + k) * stride2 + j + l)) {
639 vloc[0] = i + k;
640 vloc[1] = j + l;
641 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
642 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
643 match = 0;
644 break;
645 }
646 }
647 }
648 }
649 }
650 }
651 #endif
652
653 static void find_mismatch(const vpx_image_t *const img1,
654 const vpx_image_t *const img2,
655 int yloc[4], int uloc[4], int vloc[4]) {
656 const uint32_t bsize = 64;
657 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
658 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
659 const uint32_t c_w =
660 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
661 const uint32_t c_h =
662 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
663 int match = 1;
664 uint32_t i, j;
665 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
666 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
667 for (j = 0; match && j < img1->d_w; j += bsize) {
668 int k, l;
669 const int si = mmin(i + bsize, img1->d_h) - i;
670 const int sj = mmin(j + bsize, img1->d_w) - j;
671 for (k = 0; match && k < si; ++k) {
672 for (l = 0; match && l < sj; ++l) {
673 if (*(img1->planes[VPX_PLANE_Y] +
674 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
675 *(img2->planes[VPX_PLANE_Y] +
676 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
677 yloc[0] = i + k;
678 yloc[1] = j + l;
679 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
680 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
681 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
682 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
683 match = 0;
684 break;
685 }
686 }
687 }
688 }
689 }
690
691 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
692 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
693 for (j = 0; match && j < c_w; j += bsizex) {
694 int k, l;
695 const int si = mmin(i + bsizey, c_h - i);
696 const int sj = mmin(j + bsizex, c_w - j);
697 for (k = 0; match && k < si; ++k) {
698 for (l = 0; match && l < sj; ++l) {
699 if (*(img1->planes[VPX_PLANE_U] +
700 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
701 *(img2->planes[VPX_PLANE_U] +
702 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
703 uloc[0] = i + k;
704 uloc[1] = j + l;
705 uloc[2] = *(img1->planes[VPX_PLANE_U] +
706 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
707 uloc[3] = *(img2->planes[VPX_PLANE_U] +
708 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
709 match = 0;
710 break;
711 }
712 }
713 }
714 }
715 }
716 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
717 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
718 for (j = 0; match && j < c_w; j += bsizex) {
719 int k, l;
720 const int si = mmin(i + bsizey, c_h - i);
721 const int sj = mmin(j + bsizex, c_w - j);
722 for (k = 0; match && k < si; ++k) {
723 for (l = 0; match && l < sj; ++l) {
724 if (*(img1->planes[VPX_PLANE_V] +
725 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
726 *(img2->planes[VPX_PLANE_V] +
727 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
728 vloc[0] = i + k;
729 vloc[1] = j + l;
730 vloc[2] = *(img1->planes[VPX_PLANE_V] +
731 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
732 vloc[3] = *(img2->planes[VPX_PLANE_V] +
733 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
734 match = 0;
735 break;
736 }
737 }
738 }
739 }
740 }
741 }
742
743 static int compare_img(const vpx_image_t *const img1,
744 const vpx_image_t *const img2) {
745 uint32_t l_w = img1->d_w;
746 uint32_t c_w =
747 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
748 const uint32_t c_h =
749 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
750 uint32_t i;
751 int match = 1;
752
753 match &= (img1->fmt == img2->fmt);
754 match &= (img1->d_w == img2->d_w);
755 match &= (img1->d_h == img2->d_h);
756 #if CONFIG_VP9_HIGHBITDEPTH
757 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
758 l_w *= 2;
759 c_w *= 2;
760 }
761 #endif
762
763 for (i = 0; i < img1->d_h; ++i)
764 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
765 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
766 l_w) == 0);
767
768 for (i = 0; i < c_h; ++i)
769 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
770 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
771 c_w) == 0);
772
773 for (i = 0; i < c_h; ++i)
774 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
775 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
776 c_w) == 0);
777
778 return match;
779 }
780
781
782 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
783 #if CONFIG_VP10_ENCODER
784 #define ARG_CTRL_CNT_MAX NELEMENTS(vp10_arg_ctrl_map)
785 #elif CONFIG_VP9_ENCODER
786 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
787 #else
788 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
789 #endif
790
791 #if !CONFIG_WEBM_IO
792 typedef int stereo_format_t;
793 struct EbmlGlobal { int debug; };
794 #endif
795
796 /* Per-stream configuration */
797 struct stream_config {
798 struct vpx_codec_enc_cfg cfg;
799 const char *out_fn;
800 const char *stats_fn;
801 #if CONFIG_FP_MB_STATS
802 const char *fpmb_stats_fn;
803 #endif
804 stereo_format_t stereo_fmt;
805 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
806 int arg_ctrl_cnt;
807 int write_webm;
808 #if CONFIG_VP9_HIGHBITDEPTH
809 // whether to use 16bit internal buffers
810 int use_16bit_internal;
811 #endif
812 };
813
814
815 struct stream_state {
816 int index;
817 struct stream_state *next;
818 struct stream_config config;
819 FILE *file;
820 struct rate_hist *rate_hist;
821 struct EbmlGlobal ebml;
822 uint64_t psnr_sse_total;
823 uint64_t psnr_samples_total;
824 double psnr_totals[4];
825 int psnr_count;
826 int counts[64];
827 vpx_codec_ctx_t encoder;
828 unsigned int frames_out;
829 uint64_t cx_time;
830 size_t nbytes;
831 stats_io_t stats;
832 #if CONFIG_FP_MB_STATS
833 stats_io_t fpmb_stats;
834 #endif
835 struct vpx_image *img;
836 vpx_codec_ctx_t decoder;
837 int mismatch_seen;
838 };
839
840
841 static void validate_positive_rational(const char *msg,
842 struct vpx_rational *rat) {
843 if (rat->den < 0) {
844 rat->num *= -1;
845 rat->den *= -1;
846 }
847
848 if (rat->num < 0)
849 die("Error: %s must be positive\n", msg);
850
851 if (!rat->den)
852 die("Error: %s has zero denominator\n", msg);
853 }
854
855
856 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
857 char **argi, **argj;
858 struct arg arg;
859 const int num_encoder = get_vpx_encoder_count();
860
861 if (num_encoder < 1)
862 die("Error: no valid encoder available\n");
863
864 /* Initialize default parameters */
865 memset(global, 0, sizeof(*global));
866 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
867 global->passes = 0;
868 global->color_type = I420;
869 /* Assign default deadline to good quality */
870 global->deadline = VPX_DL_GOOD_QUALITY;
871
872 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
873 arg.argv_step = 1;
874
875 if (arg_match(&arg, &codecarg, argi)) {
876 global->codec = get_vpx_encoder_by_name(arg.val);
877 if (!global->codec)
878 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
879 } else if (arg_match(&arg, &passes, argi)) {
880 global->passes = arg_parse_uint(&arg);
881
882 if (global->passes < 1 || global->passes > 2)
883 die("Error: Invalid number of passes (%d)\n", global->passes);
884 } else if (arg_match(&arg, &pass_arg, argi)) {
885 global->pass = arg_parse_uint(&arg);
886
887 if (global->pass < 1 || global->pass > 2)
888 die("Error: Invalid pass selected (%d)\n",
889 global->pass);
890 } else if (arg_match(&arg, &usage, argi))
891 global->usage = arg_parse_uint(&arg);
892 else if (arg_match(&arg, &deadline, argi))
893 global->deadline = arg_parse_uint(&arg);
894 else if (arg_match(&arg, &best_dl, argi))
895 global->deadline = VPX_DL_BEST_QUALITY;
896 else if (arg_match(&arg, &good_dl, argi))
897 global->deadline = VPX_DL_GOOD_QUALITY;
898 else if (arg_match(&arg, &rt_dl, argi))
899 global->deadline = VPX_DL_REALTIME;
900 else if (arg_match(&arg, &use_yv12, argi))
901 global->color_type = YV12;
902 else if (arg_match(&arg, &use_i420, argi))
903 global->color_type = I420;
904 else if (arg_match(&arg, &use_i422, argi))
905 global->color_type = I422;
906 else if (arg_match(&arg, &use_i444, argi))
907 global->color_type = I444;
908 else if (arg_match(&arg, &use_i440, argi))
909 global->color_type = I440;
910 else if (arg_match(&arg, &quietarg, argi))
911 global->quiet = 1;
912 else if (arg_match(&arg, &verbosearg, argi))
913 global->verbose = 1;
914 else if (arg_match(&arg, &limit, argi))
915 global->limit = arg_parse_uint(&arg);
916 else if (arg_match(&arg, &skip, argi))
917 global->skip_frames = arg_parse_uint(&arg);
918 else if (arg_match(&arg, &psnrarg, argi))
919 global->show_psnr = 1;
920 else if (arg_match(&arg, &recontest, argi))
921 global->test_decode = arg_parse_enum_or_int(&arg);
922 else if (arg_match(&arg, &framerate, argi)) {
923 global->framerate = arg_parse_rational(&arg);
924 validate_positive_rational(arg.name, &global->framerate);
925 global->have_framerate = 1;
926 } else if (arg_match(&arg, &out_part, argi))
927 global->out_part = 1;
928 else if (arg_match(&arg, &debugmode, argi))
929 global->debug = 1;
930 else if (arg_match(&arg, &q_hist_n, argi))
931 global->show_q_hist_buckets = arg_parse_uint(&arg);
932 else if (arg_match(&arg, &rate_hist_n, argi))
933 global->show_rate_hist_buckets = arg_parse_uint(&arg);
934 else if (arg_match(&arg, &disable_warnings, argi))
935 global->disable_warnings = 1;
936 else if (arg_match(&arg, &disable_warning_prompt, argi))
937 global->disable_warning_prompt = 1;
938 else
939 argj++;
940 }
941
942 if (global->pass) {
943 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
944 if (global->pass > global->passes) {
945 warn("Assuming --pass=%d implies --passes=%d\n",
946 global->pass, global->pass);
947 global->passes = global->pass;
948 }
949 }
950 /* Validate global config */
951 if (global->passes == 0) {
952 #if CONFIG_VP9_ENCODER || CONFIG_VP10_ENCODER
953 // Make default VP9 passes = 2 until there is a better quality 1-pass
954 // encoder
955 if (global->codec != NULL && global->codec->name != NULL)
956 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
957 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
958 #else
959 global->passes = 1;
960 #endif
961 }
962
963 if (global->deadline == VPX_DL_REALTIME &&
964 global->passes > 1) {
965 warn("Enforcing one-pass encoding in realtime mode\n");
966 global->passes = 1;
967 }
968 }
969
970
971 static void open_input_file(struct VpxInputContext *input) {
972 /* Parse certain options from the input file, if possible */
973 input->file = strcmp(input->filename, "-")
974 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
975
976 if (!input->file)
977 fatal("Failed to open input file");
978
979 if (!fseeko(input->file, 0, SEEK_END)) {
980 /* Input file is seekable. Figure out how long it is, so we can get
981 * progress info.
982 */
983 input->length = ftello(input->file);
984 rewind(input->file);
985 }
986
987 /* Default to 1:1 pixel aspect ratio. */
988 input->pixel_aspect_ratio.numerator = 1;
989 input->pixel_aspect_ratio.denominator = 1;
990
991 /* For RAW input sources, these bytes will applied on the first frame
992 * in read_frame().
993 */
994 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
995 input->detect.position = 0;
996
997 if (input->detect.buf_read == 4
998 && file_is_y4m(input->detect.buf)) {
999 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1000 input->only_i420) >= 0) {
1001 input->file_type = FILE_TYPE_Y4M;
1002 input->width = input->y4m.pic_w;
1003 input->height = input->y4m.pic_h;
1004 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1005 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1006 input->framerate.numerator = input->y4m.fps_n;
1007 input->framerate.denominator = input->y4m.fps_d;
1008 input->fmt = input->y4m.vpx_fmt;
1009 input->bit_depth = input->y4m.bit_depth;
1010 } else
1011 fatal("Unsupported Y4M stream.");
1012 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1013 fatal("IVF is not supported as input.");
1014 } else {
1015 input->file_type = FILE_TYPE_RAW;
1016 }
1017 }
1018
1019
1020 static void close_input_file(struct VpxInputContext *input) {
1021 fclose(input->file);
1022 if (input->file_type == FILE_TYPE_Y4M)
1023 y4m_input_close(&input->y4m);
1024 }
1025
1026 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1027 struct stream_state *prev) {
1028 struct stream_state *stream;
1029
1030 stream = calloc(1, sizeof(*stream));
1031 if (stream == NULL) {
1032 fatal("Failed to allocate new stream.");
1033 }
1034
1035 if (prev) {
1036 memcpy(stream, prev, sizeof(*stream));
1037 stream->index++;
1038 prev->next = stream;
1039 } else {
1040 vpx_codec_err_t res;
1041
1042 /* Populate encoder configuration */
1043 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1044 &stream->config.cfg,
1045 global->usage);
1046 if (res)
1047 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1048
1049 /* Change the default timebase to a high enough value so that the
1050 * encoder will always create strictly increasing timestamps.
1051 */
1052 stream->config.cfg.g_timebase.den = 1000;
1053
1054 /* Never use the library's default resolution, require it be parsed
1055 * from the file or set on the command line.
1056 */
1057 stream->config.cfg.g_w = 0;
1058 stream->config.cfg.g_h = 0;
1059
1060 /* Initialize remaining stream parameters */
1061 stream->config.write_webm = 1;
1062 #if CONFIG_WEBM_IO
1063 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1064 stream->ebml.last_pts_ns = -1;
1065 stream->ebml.writer = NULL;
1066 stream->ebml.segment = NULL;
1067 #endif
1068
1069 /* Allows removal of the application version from the EBML tags */
1070 stream->ebml.debug = global->debug;
1071
1072 /* Default lag_in_frames is 0 in realtime mode */
1073 if (global->deadline == VPX_DL_REALTIME)
1074 stream->config.cfg.g_lag_in_frames = 0;
1075 }
1076
1077 /* Output files must be specified for each stream */
1078 stream->config.out_fn = NULL;
1079
1080 stream->next = NULL;
1081 return stream;
1082 }
1083
1084
1085 static int parse_stream_params(struct VpxEncoderConfig *global,
1086 struct stream_state *stream,
1087 char **argv) {
1088 char **argi, **argj;
1089 struct arg arg;
1090 static const arg_def_t **ctrl_args = no_args;
1091 static const int *ctrl_args_map = NULL;
1092 struct stream_config *config = &stream->config;
1093 int eos_mark_found = 0;
1094 #if CONFIG_VP9_HIGHBITDEPTH
1095 int test_16bit_internal = 0;
1096 #endif
1097
1098 // Handle codec specific options
1099 if (0) {
1100 #if CONFIG_VP8_ENCODER
1101 } else if (strcmp(global->codec->name, "vp8") == 0) {
1102 ctrl_args = vp8_args;
1103 ctrl_args_map = vp8_arg_ctrl_map;
1104 #endif
1105 #if CONFIG_VP9_ENCODER
1106 } else if (strcmp(global->codec->name, "vp9") == 0) {
1107 ctrl_args = vp9_args;
1108 ctrl_args_map = vp9_arg_ctrl_map;
1109 #endif
1110 #if CONFIG_VP10_ENCODER
1111 } else if (strcmp(global->codec->name, "vp10") == 0) {
1112 // TODO(jingning): Reuse VP9 specific encoder configuration parameters.
1113 // Consider to expand this set for VP10 encoder control.
1114 ctrl_args = vp10_args;
1115 ctrl_args_map = vp10_arg_ctrl_map;
1116 #endif
1117 }
1118
1119 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1120 arg.argv_step = 1;
1121
1122 /* Once we've found an end-of-stream marker (--) we want to continue
1123 * shifting arguments but not consuming them.
1124 */
1125 if (eos_mark_found) {
1126 argj++;
1127 continue;
1128 } else if (!strcmp(*argj, "--")) {
1129 eos_mark_found = 1;
1130 continue;
1131 }
1132
1133 if (arg_match(&arg, &outputfile, argi)) {
1134 config->out_fn = arg.val;
1135 } else if (arg_match(&arg, &fpf_name, argi)) {
1136 config->stats_fn = arg.val;
1137 #if CONFIG_FP_MB_STATS
1138 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1139 config->fpmb_stats_fn = arg.val;
1140 #endif
1141 } else if (arg_match(&arg, &use_webm, argi)) {
1142 #if CONFIG_WEBM_IO
1143 config->write_webm = 1;
1144 #else
1145 die("Error: --webm specified but webm is disabled.");
1146 #endif
1147 } else if (arg_match(&arg, &use_ivf, argi)) {
1148 config->write_webm = 0;
1149 } else if (arg_match(&arg, &threads, argi)) {
1150 config->cfg.g_threads = arg_parse_uint(&arg);
1151 } else if (arg_match(&arg, &profile, argi)) {
1152 config->cfg.g_profile = arg_parse_uint(&arg);
1153 } else if (arg_match(&arg, &width, argi)) {
1154 config->cfg.g_w = arg_parse_uint(&arg);
1155 } else if (arg_match(&arg, &height, argi)) {
1156 config->cfg.g_h = arg_parse_uint(&arg);
1157 #if CONFIG_VP9_HIGHBITDEPTH
1158 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1159 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1160 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1161 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1162 #endif
1163 #if CONFIG_WEBM_IO
1164 } else if (arg_match(&arg, &stereo_mode, argi)) {
1165 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1166 #endif
1167 } else if (arg_match(&arg, &timebase, argi)) {
1168 config->cfg.g_timebase = arg_parse_rational(&arg);
1169 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1170 } else if (arg_match(&arg, &error_resilient, argi)) {
1171 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1172 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1173 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1174 if (global->deadline == VPX_DL_REALTIME &&
1175 config->cfg.g_lag_in_frames != 0) {
1176 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1177 config->cfg.g_lag_in_frames = 0;
1178 }
1179 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1180 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1181 } else if (arg_match(&arg, &resize_allowed, argi)) {
1182 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1183 } else if (arg_match(&arg, &resize_width, argi)) {
1184 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1185 } else if (arg_match(&arg, &resize_height, argi)) {
1186 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1187 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1188 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1189 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1190 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1191 } else if (arg_match(&arg, &end_usage, argi)) {
1192 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1193 } else if (arg_match(&arg, &target_bitrate, argi)) {
1194 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1195 } else if (arg_match(&arg, &min_quantizer, argi)) {
1196 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1197 } else if (arg_match(&arg, &max_quantizer, argi)) {
1198 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1199 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1200 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1201 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1202 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1203 } else if (arg_match(&arg, &buf_sz, argi)) {
1204 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1205 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1206 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1207 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1208 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1209 } else if (arg_match(&arg, &bias_pct, argi)) {
1210 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1211 if (global->passes < 2)
1212 warn("option %s ignored in one-pass mode.\n", arg.name);
1213 } else if (arg_match(&arg, &minsection_pct, argi)) {
1214 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1215
1216 if (global->passes < 2)
1217 warn("option %s ignored in one-pass mode.\n", arg.name);
1218 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1219 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1220
1221 if (global->passes < 2)
1222 warn("option %s ignored in one-pass mode.\n", arg.name);
1223 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1224 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1225 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1226 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1227 } else if (arg_match(&arg, &kf_disabled, argi)) {
1228 config->cfg.kf_mode = VPX_KF_DISABLED;
1229 #if CONFIG_VP9_HIGHBITDEPTH
1230 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1231 if (strcmp(global->codec->name, "vp9") == 0 ||
1232 strcmp(global->codec->name, "vp10") == 0) {
1233 test_16bit_internal = 1;
1234 }
1235 #endif
1236 } else {
1237 int i, match = 0;
1238 for (i = 0; ctrl_args[i]; i++) {
1239 if (arg_match(&arg, ctrl_args[i], argi)) {
1240 int j;
1241 match = 1;
1242
1243 /* Point either to the next free element or the first
1244 * instance of this control.
1245 */
1246 for (j = 0; j < config->arg_ctrl_cnt; j++)
1247 if (ctrl_args_map != NULL &&
1248 config->arg_ctrls[j][0] == ctrl_args_map[i])
1249 break;
1250
1251 /* Update/insert */
1252 assert(j < (int)ARG_CTRL_CNT_MAX);
1253 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1254 config->arg_ctrls[j][0] = ctrl_args_map[i];
1255 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1256 if (j == config->arg_ctrl_cnt)
1257 config->arg_ctrl_cnt++;
1258 }
1259 }
1260 }
1261 if (!match)
1262 argj++;
1263 }
1264 }
1265 #if CONFIG_VP9_HIGHBITDEPTH
1266 if (strcmp(global->codec->name, "vp9") == 0 ||
1267 strcmp(global->codec->name, "vp10") == 0) {
1268 config->use_16bit_internal = test_16bit_internal |
1269 (config->cfg.g_profile > 1);
1270 }
1271 #endif
1272 return eos_mark_found;
1273 }
1274
1275
1276 #define FOREACH_STREAM(func) \
1277 do { \
1278 struct stream_state *stream; \
1279 for (stream = streams; stream; stream = stream->next) { \
1280 func; \
1281 } \
1282 } while (0)
1283
1284
1285 static void validate_stream_config(const struct stream_state *stream,
1286 const struct VpxEncoderConfig *global) {
1287 const struct stream_state *streami;
1288 (void)global;
1289
1290 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1291 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1292 " and --height (-h)", stream->index);
1293
1294 // Check that the codec bit depth is greater than the input bit depth.
1295 if (stream->config.cfg.g_input_bit_depth >
1296 (unsigned int)stream->config.cfg.g_bit_depth) {
1297 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1298 stream->index, (int)stream->config.cfg.g_bit_depth,
1299 stream->config.cfg.g_input_bit_depth);
1300 }
1301
1302 for (streami = stream; streami; streami = streami->next) {
1303 /* All streams require output files */
1304 if (!streami->config.out_fn)
1305 fatal("Stream %d: Output file is required (specify with -o)",
1306 streami->index);
1307
1308 /* Check for two streams outputting to the same file */
1309 if (streami != stream) {
1310 const char *a = stream->config.out_fn;
1311 const char *b = streami->config.out_fn;
1312 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1313 fatal("Stream %d: duplicate output file (from stream %d)",
1314 streami->index, stream->index);
1315 }
1316
1317 /* Check for two streams sharing a stats file. */
1318 if (streami != stream) {
1319 const char *a = stream->config.stats_fn;
1320 const char *b = streami->config.stats_fn;
1321 if (a && b && !strcmp(a, b))
1322 fatal("Stream %d: duplicate stats file (from stream %d)",
1323 streami->index, stream->index);
1324 }
1325
1326 #if CONFIG_FP_MB_STATS
1327 /* Check for two streams sharing a mb stats file. */
1328 if (streami != stream) {
1329 const char *a = stream->config.fpmb_stats_fn;
1330 const char *b = streami->config.fpmb_stats_fn;
1331 if (a && b && !strcmp(a, b))
1332 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1333 streami->index, stream->index);
1334 }
1335 #endif
1336 }
1337 }
1338
1339
1340 static void set_stream_dimensions(struct stream_state *stream,
1341 unsigned int w,
1342 unsigned int h) {
1343 if (!stream->config.cfg.g_w) {
1344 if (!stream->config.cfg.g_h)
1345 stream->config.cfg.g_w = w;
1346 else
1347 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1348 }
1349 if (!stream->config.cfg.g_h) {
1350 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1351 }
1352 }
1353
1354 static const char* file_type_to_string(enum VideoFileType t) {
1355 switch (t) {
1356 case FILE_TYPE_RAW: return "RAW";
1357 case FILE_TYPE_Y4M: return "Y4M";
1358 default: return "Other";
1359 }
1360 }
1361
1362 static const char* image_format_to_string(vpx_img_fmt_t f) {
1363 switch (f) {
1364 case VPX_IMG_FMT_I420: return "I420";
1365 case VPX_IMG_FMT_I422: return "I422";
1366 case VPX_IMG_FMT_I444: return "I444";
1367 case VPX_IMG_FMT_I440: return "I440";
1368 case VPX_IMG_FMT_YV12: return "YV12";
1369 case VPX_IMG_FMT_I42016: return "I42016";
1370 case VPX_IMG_FMT_I42216: return "I42216";
1371 case VPX_IMG_FMT_I44416: return "I44416";
1372 case VPX_IMG_FMT_I44016: return "I44016";
1373 default: return "Other";
1374 }
1375 }
1376
1377 static void show_stream_config(struct stream_state *stream,
1378 struct VpxEncoderConfig *global,
1379 struct VpxInputContext *input) {
1380
1381 #define SHOW(field) \
1382 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1383
1384 if (stream->index == 0) {
1385 fprintf(stderr, "Codec: %s\n",
1386 vpx_codec_iface_name(global->codec->codec_interface()));
1387 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1388 input->filename,
1389 file_type_to_string(input->file_type),
1390 image_format_to_string(input->fmt));
1391 }
1392 if (stream->next || stream->index)
1393 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1394 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1395 fprintf(stderr, "Encoder parameters:\n");
1396
1397 SHOW(g_usage);
1398 SHOW(g_threads);
1399 SHOW(g_profile);
1400 SHOW(g_w);
1401 SHOW(g_h);
1402 SHOW(g_bit_depth);
1403 SHOW(g_input_bit_depth);
1404 SHOW(g_timebase.num);
1405 SHOW(g_timebase.den);
1406 SHOW(g_error_resilient);
1407 SHOW(g_pass);
1408 SHOW(g_lag_in_frames);
1409 SHOW(rc_dropframe_thresh);
1410 SHOW(rc_resize_allowed);
1411 SHOW(rc_scaled_width);
1412 SHOW(rc_scaled_height);
1413 SHOW(rc_resize_up_thresh);
1414 SHOW(rc_resize_down_thresh);
1415 SHOW(rc_end_usage);
1416 SHOW(rc_target_bitrate);
1417 SHOW(rc_min_quantizer);
1418 SHOW(rc_max_quantizer);
1419 SHOW(rc_undershoot_pct);
1420 SHOW(rc_overshoot_pct);
1421 SHOW(rc_buf_sz);
1422 SHOW(rc_buf_initial_sz);
1423 SHOW(rc_buf_optimal_sz);
1424 SHOW(rc_2pass_vbr_bias_pct);
1425 SHOW(rc_2pass_vbr_minsection_pct);
1426 SHOW(rc_2pass_vbr_maxsection_pct);
1427 SHOW(kf_mode);
1428 SHOW(kf_min_dist);
1429 SHOW(kf_max_dist);
1430 }
1431
1432
1433 static void open_output_file(struct stream_state *stream,
1434 struct VpxEncoderConfig *global,
1435 const struct VpxRational *pixel_aspect_ratio) {
1436 const char *fn = stream->config.out_fn;
1437 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1438
1439 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1440 return;
1441
1442 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1443
1444 if (!stream->file)
1445 fatal("Failed to open output file");
1446
1447 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1448 fatal("WebM output to pipes not supported.");
1449
1450 #if CONFIG_WEBM_IO
1451 if (stream->config.write_webm) {
1452 stream->ebml.stream = stream->file;
1453 write_webm_file_header(&stream->ebml, cfg,
1454 &global->framerate,
1455 stream->config.stereo_fmt,
1456 global->codec->fourcc,
1457 pixel_aspect_ratio);
1458 }
1459 #else
1460 (void)pixel_aspect_ratio;
1461 #endif
1462
1463 if (!stream->config.write_webm) {
1464 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1465 }
1466 }
1467
1468
1469 static void close_output_file(struct stream_state *stream,
1470 unsigned int fourcc) {
1471 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1472
1473 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1474 return;
1475
1476 #if CONFIG_WEBM_IO
1477 if (stream->config.write_webm) {
1478 write_webm_file_footer(&stream->ebml);
1479 }
1480 #endif
1481
1482 if (!stream->config.write_webm) {
1483 if (!fseek(stream->file, 0, SEEK_SET))
1484 ivf_write_file_header(stream->file, &stream->config.cfg,
1485 fourcc,
1486 stream->frames_out);
1487 }
1488
1489 fclose(stream->file);
1490 }
1491
1492
1493 static void setup_pass(struct stream_state *stream,
1494 struct VpxEncoderConfig *global,
1495 int pass) {
1496 if (stream->config.stats_fn) {
1497 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1498 pass))
1499 fatal("Failed to open statistics store");
1500 } else {
1501 if (!stats_open_mem(&stream->stats, pass))
1502 fatal("Failed to open statistics store");
1503 }
1504
1505 #if CONFIG_FP_MB_STATS
1506 if (stream->config.fpmb_stats_fn) {
1507 if (!stats_open_file(&stream->fpmb_stats,
1508 stream->config.fpmb_stats_fn, pass))
1509 fatal("Failed to open mb statistics store");
1510 } else {
1511 if (!stats_open_mem(&stream->fpmb_stats, pass))
1512 fatal("Failed to open mb statistics store");
1513 }
1514 #endif
1515
1516 stream->config.cfg.g_pass = global->passes == 2
1517 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1518 : VPX_RC_ONE_PASS;
1519 if (pass) {
1520 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1521 #if CONFIG_FP_MB_STATS
1522 stream->config.cfg.rc_firstpass_mb_stats_in =
1523 stats_get(&stream->fpmb_stats);
1524 #endif
1525 }
1526
1527 stream->cx_time = 0;
1528 stream->nbytes = 0;
1529 stream->frames_out = 0;
1530 }
1531
1532
1533 static void initialize_encoder(struct stream_state *stream,
1534 struct VpxEncoderConfig *global) {
1535 int i;
1536 int flags = 0;
1537
1538 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1539 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1540 #if CONFIG_VP9_HIGHBITDEPTH
1541 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1542 #endif
1543
1544 /* Construct Encoder Context */
1545 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1546 &stream->config.cfg, flags);
1547 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1548
1549 /* Note that we bypass the vpx_codec_control wrapper macro because
1550 * we're being clever to store the control IDs in an array. Real
1551 * applications will want to make use of the enumerations directly
1552 */
1553 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1554 int ctrl = stream->config.arg_ctrls[i][0];
1555 int value = stream->config.arg_ctrls[i][1];
1556 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1557 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1558 ctrl, value);
1559
1560 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1561 }
1562
1563 #if CONFIG_DECODERS
1564 if (global->test_decode != TEST_DECODE_OFF) {
1565 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1566 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1567 }
1568 #endif
1569 }
1570
1571
1572 static void encode_frame(struct stream_state *stream,
1573 struct VpxEncoderConfig *global,
1574 struct vpx_image *img,
1575 unsigned int frames_in) {
1576 vpx_codec_pts_t frame_start, next_frame_start;
1577 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1578 struct vpx_usec_timer timer;
1579
1580 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1581 * global->framerate.den)
1582 / cfg->g_timebase.num / global->framerate.num;
1583 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1584 * global->framerate.den)
1585 / cfg->g_timebase.num / global->framerate.num;
1586
1587 /* Scale if necessary */
1588 #if CONFIG_VP9_HIGHBITDEPTH
1589 if (img) {
1590 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1591 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1592 if (img->fmt != VPX_IMG_FMT_I42016) {
1593 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1594 exit(EXIT_FAILURE);
1595 }
1596 #if CONFIG_LIBYUV
1597 if (!stream->img) {
1598 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1599 cfg->g_w, cfg->g_h, 16);
1600 }
1601 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1602 img->stride[VPX_PLANE_Y]/2,
1603 (uint16*)img->planes[VPX_PLANE_U],
1604 img->stride[VPX_PLANE_U]/2,
1605 (uint16*)img->planes[VPX_PLANE_V],
1606 img->stride[VPX_PLANE_V]/2,
1607 img->d_w, img->d_h,
1608 (uint16*)stream->img->planes[VPX_PLANE_Y],
1609 stream->img->stride[VPX_PLANE_Y]/2,
1610 (uint16*)stream->img->planes[VPX_PLANE_U],
1611 stream->img->stride[VPX_PLANE_U]/2,
1612 (uint16*)stream->img->planes[VPX_PLANE_V],
1613 stream->img->stride[VPX_PLANE_V]/2,
1614 stream->img->d_w, stream->img->d_h,
1615 kFilterBox);
1616 img = stream->img;
1617 #else
1618 stream->encoder.err = 1;
1619 ctx_exit_on_error(&stream->encoder,
1620 "Stream %d: Failed to encode frame.\n"
1621 "Scaling disabled in this configuration. \n"
1622 "To enable, configure with --enable-libyuv\n",
1623 stream->index);
1624 #endif
1625 }
1626 }
1627 #endif
1628 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1629 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1630 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1631 exit(EXIT_FAILURE);
1632 }
1633 #if CONFIG_LIBYUV
1634 if (!stream->img)
1635 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1636 cfg->g_w, cfg->g_h, 16);
1637 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1638 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1639 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1640 img->d_w, img->d_h,
1641 stream->img->planes[VPX_PLANE_Y],
1642 stream->img->stride[VPX_PLANE_Y],
1643 stream->img->planes[VPX_PLANE_U],
1644 stream->img->stride[VPX_PLANE_U],
1645 stream->img->planes[VPX_PLANE_V],
1646 stream->img->stride[VPX_PLANE_V],
1647 stream->img->d_w, stream->img->d_h,
1648 kFilterBox);
1649 img = stream->img;
1650 #else
1651 stream->encoder.err = 1;
1652 ctx_exit_on_error(&stream->encoder,
1653 "Stream %d: Failed to encode frame.\n"
1654 "Scaling disabled in this configuration. \n"
1655 "To enable, configure with --enable-libyuv\n",
1656 stream->index);
1657 #endif
1658 }
1659
1660 vpx_usec_timer_start(&timer);
1661 vpx_codec_encode(&stream->encoder, img, frame_start,
1662 (unsigned long)(next_frame_start - frame_start),
1663 0, global->deadline);
1664 vpx_usec_timer_mark(&timer);
1665 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1666 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1667 stream->index);
1668 }
1669
1670
1671 static void update_quantizer_histogram(struct stream_state *stream) {
1672 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1673 int q;
1674
1675 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1676 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1677 stream->counts[q]++;
1678 }
1679 }
1680
1681
1682 static void get_cx_data(struct stream_state *stream,
1683 struct VpxEncoderConfig *global,
1684 int *got_data) {
1685 const vpx_codec_cx_pkt_t *pkt;
1686 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1687 vpx_codec_iter_t iter = NULL;
1688
1689 *got_data = 0;
1690 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1691 static size_t fsize = 0;
1692 static int64_t ivf_header_pos = 0;
1693
1694 switch (pkt->kind) {
1695 case VPX_CODEC_CX_FRAME_PKT:
1696 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1697 stream->frames_out++;
1698 }
1699 if (!global->quiet)
1700 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1701
1702 update_rate_histogram(stream->rate_hist, cfg, pkt);
1703 #if CONFIG_WEBM_IO
1704 if (stream->config.write_webm) {
1705 write_webm_block(&stream->ebml, cfg, pkt);
1706 }
1707 #endif
1708 if (!stream->config.write_webm) {
1709 if (pkt->data.frame.partition_id <= 0) {
1710 ivf_header_pos = ftello(stream->file);
1711 fsize = pkt->data.frame.sz;
1712
1713 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1714 } else {
1715 fsize += pkt->data.frame.sz;
1716
1717 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1718 const int64_t currpos = ftello(stream->file);
1719 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1720 ivf_write_frame_size(stream->file, fsize);
1721 fseeko(stream->file, currpos, SEEK_SET);
1722 }
1723 }
1724
1725 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1726 stream->file);
1727 }
1728 stream->nbytes += pkt->data.raw.sz;
1729
1730 *got_data = 1;
1731 #if CONFIG_DECODERS
1732 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1733 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1734 (unsigned int)pkt->data.frame.sz, NULL, 0);
1735 if (stream->decoder.err) {
1736 warn_or_exit_on_error(&stream->decoder,
1737 global->test_decode == TEST_DECODE_FATAL,
1738 "Failed to decode frame %d in stream %d",
1739 stream->frames_out + 1, stream->index);
1740 stream->mismatch_seen = stream->frames_out + 1;
1741 }
1742 }
1743 #endif
1744 break;
1745 case VPX_CODEC_STATS_PKT:
1746 stream->frames_out++;
1747 stats_write(&stream->stats,
1748 pkt->data.twopass_stats.buf,
1749 pkt->data.twopass_stats.sz);
1750 stream->nbytes += pkt->data.raw.sz;
1751 break;
1752 #if CONFIG_FP_MB_STATS
1753 case VPX_CODEC_FPMB_STATS_PKT:
1754 stats_write(&stream->fpmb_stats,
1755 pkt->data.firstpass_mb_stats.buf,
1756 pkt->data.firstpass_mb_stats.sz);
1757 stream->nbytes += pkt->data.raw.sz;
1758 break;
1759 #endif
1760 case VPX_CODEC_PSNR_PKT:
1761
1762 if (global->show_psnr) {
1763 int i;
1764
1765 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1766 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1767 for (i = 0; i < 4; i++) {
1768 if (!global->quiet)
1769 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1770 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1771 }
1772 stream->psnr_count++;
1773 }
1774
1775 break;
1776 default:
1777 break;
1778 }
1779 }
1780 }
1781
1782
1783 static void show_psnr(struct stream_state *stream, double peak) {
1784 int i;
1785 double ovpsnr;
1786
1787 if (!stream->psnr_count)
1788 return;
1789
1790 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1791 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1792 (double)stream->psnr_sse_total);
1793 fprintf(stderr, " %.3f", ovpsnr);
1794
1795 for (i = 0; i < 4; i++) {
1796 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1797 }
1798 fprintf(stderr, "\n");
1799 }
1800
1801
1802 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1803 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1804 }
1805
1806 static void test_decode(struct stream_state *stream,
1807 enum TestDecodeFatality fatal,
1808 const VpxInterface *codec) {
1809 vpx_image_t enc_img, dec_img;
1810
1811 if (stream->mismatch_seen)
1812 return;
1813
1814 /* Get the internal reference frame */
1815 if (strcmp(codec->name, "vp8") == 0) {
1816 struct vpx_ref_frame ref_enc, ref_dec;
1817 int width, height;
1818
1819 width = (stream->config.cfg.g_w + 15) & ~15;
1820 height = (stream->config.cfg.g_h + 15) & ~15;
1821 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1822 enc_img = ref_enc.img;
1823 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1824 dec_img = ref_dec.img;
1825
1826 ref_enc.frame_type = VP8_LAST_FRAME;
1827 ref_dec.frame_type = VP8_LAST_FRAME;
1828 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1829 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1830 } else {
1831 struct vp9_ref_frame ref_enc, ref_dec;
1832
1833 ref_enc.idx = 0;
1834 ref_dec.idx = 0;
1835 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1836 enc_img = ref_enc.img;
1837 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1838 dec_img = ref_dec.img;
1839 #if CONFIG_VP9_HIGHBITDEPTH
1840 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1841 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1842 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1843 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1844 enc_img.d_w, enc_img.d_h, 16);
1845 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1846 }
1847 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1848 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1849 dec_img.d_w, dec_img.d_h, 16);
1850 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1851 }
1852 }
1853 #endif
1854 }
1855 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1856 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1857
1858 if (!compare_img(&enc_img, &dec_img)) {
1859 int y[4], u[4], v[4];
1860 #if CONFIG_VP9_HIGHBITDEPTH
1861 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1862 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1863 } else {
1864 find_mismatch(&enc_img, &dec_img, y, u, v);
1865 }
1866 #else
1867 find_mismatch(&enc_img, &dec_img, y, u, v);
1868 #endif
1869 stream->decoder.err = 1;
1870 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1871 "Stream %d: Encode/decode mismatch on frame %d at"
1872 " Y[%d, %d] {%d/%d},"
1873 " U[%d, %d] {%d/%d},"
1874 " V[%d, %d] {%d/%d}",
1875 stream->index, stream->frames_out,
1876 y[0], y[1], y[2], y[3],
1877 u[0], u[1], u[2], u[3],
1878 v[0], v[1], v[2], v[3]);
1879 stream->mismatch_seen = stream->frames_out;
1880 }
1881
1882 vpx_img_free(&enc_img);
1883 vpx_img_free(&dec_img);
1884 }
1885
1886
1887 static void print_time(const char *label, int64_t etl) {
1888 int64_t hours;
1889 int64_t mins;
1890 int64_t secs;
1891
1892 if (etl >= 0) {
1893 hours = etl / 3600;
1894 etl -= hours * 3600;
1895 mins = etl / 60;
1896 etl -= mins * 60;
1897 secs = etl;
1898
1899 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1900 label, hours, mins, secs);
1901 } else {
1902 fprintf(stderr, "[%3s unknown] ", label);
1903 }
1904 }
1905
1906
1907 int main(int argc, const char **argv_) {
1908 int pass;
1909 vpx_image_t raw;
1910 #if CONFIG_VP9_HIGHBITDEPTH
1911 vpx_image_t raw_shift;
1912 int allocated_raw_shift = 0;
1913 int use_16bit_internal = 0;
1914 int input_shift = 0;
1915 #endif
1916 int frame_avail, got_data;
1917
1918 struct VpxInputContext input;
1919 struct VpxEncoderConfig global;
1920 struct stream_state *streams = NULL;
1921 char **argv, **argi;
1922 uint64_t cx_time = 0;
1923 int stream_cnt = 0;
1924 int res = 0;
1925
1926 memset(&input, 0, sizeof(input));
1927 exec_name = argv_[0];
1928
1929 if (argc < 3)
1930 usage_exit();
1931
1932 /* Setup default input stream settings */
1933 input.framerate.numerator = 30;
1934 input.framerate.denominator = 1;
1935 input.only_i420 = 1;
1936 input.bit_depth = 0;
1937
1938 /* First parse the global configuration values, because we want to apply
1939 * other parameters on top of the default configuration provided by the
1940 * codec.
1941 */
1942 argv = argv_dup(argc - 1, argv_ + 1);
1943 parse_global_config(&global, argv);
1944
1945 switch (global.color_type) {
1946 case I420:
1947 input.fmt = VPX_IMG_FMT_I420;
1948 break;
1949 case I422:
1950 input.fmt = VPX_IMG_FMT_I422;
1951 break;
1952 case I444:
1953 input.fmt = VPX_IMG_FMT_I444;
1954 break;
1955 case I440:
1956 input.fmt = VPX_IMG_FMT_I440;
1957 break;
1958 case YV12:
1959 input.fmt = VPX_IMG_FMT_YV12;
1960 break;
1961 }
1962
1963 {
1964 /* Now parse each stream's parameters. Using a local scope here
1965 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1966 * loops
1967 */
1968 struct stream_state *stream = NULL;
1969
1970 do {
1971 stream = new_stream(&global, stream);
1972 stream_cnt++;
1973 if (!streams)
1974 streams = stream;
1975 } while (parse_stream_params(&global, stream, argv));
1976 }
1977
1978 /* Check for unrecognized options */
1979 for (argi = argv; *argi; argi++)
1980 if (argi[0][0] == '-' && argi[0][1])
1981 die("Error: Unrecognized option %s\n", *argi);
1982
1983 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1984 &global, &stream->config.cfg););
1985
1986 /* Handle non-option arguments */
1987 input.filename = argv[0];
1988
1989 if (!input.filename)
1990 usage_exit();
1991
1992 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1993 if (global.codec->fourcc == VP9_FOURCC || global.codec->fourcc == VP10_FOURCC)
1994 input.only_i420 = 0;
1995
1996 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1997 int frames_in = 0, seen_frames = 0;
1998 int64_t estimated_time_left = -1;
1999 int64_t average_rate = -1;
2000 int64_t lagged_count = 0;
2001
2002 open_input_file(&input);
2003
2004 /* If the input file doesn't specify its w/h (raw files), try to get
2005 * the data from the first stream's configuration.
2006 */
2007 if (!input.width || !input.height) {
2008 FOREACH_STREAM({
2009 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2010 input.width = stream->config.cfg.g_w;
2011 input.height = stream->config.cfg.g_h;
2012 break;
2013 }
2014 });
2015 }
2016
2017 /* Update stream configurations from the input file's parameters */
2018 if (!input.width || !input.height)
2019 fatal("Specify stream dimensions with --width (-w) "
2020 " and --height (-h)");
2021
2022 /* If input file does not specify bit-depth but input-bit-depth parameter
2023 * exists, assume that to be the input bit-depth. However, if the
2024 * input-bit-depth paramter does not exist, assume the input bit-depth
2025 * to be the same as the codec bit-depth.
2026 */
2027 if (!input.bit_depth) {
2028 FOREACH_STREAM({
2029 if (stream->config.cfg.g_input_bit_depth)
2030 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2031 else
2032 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2033 (int)stream->config.cfg.g_bit_depth;
2034 });
2035 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2036 } else {
2037 FOREACH_STREAM({
2038 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2039 });
2040 }
2041
2042 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2043 FOREACH_STREAM(validate_stream_config(stream, &global));
2044
2045 /* Ensure that --passes and --pass are consistent. If --pass is set and
2046 * --passes=2, ensure --fpf was set.
2047 */
2048 if (global.pass && global.passes == 2)
2049 FOREACH_STREAM( {
2050 if (!stream->config.stats_fn)
2051 die("Stream %d: Must specify --fpf when --pass=%d"
2052 " and --passes=2\n", stream->index, global.pass);
2053 });
2054
2055 #if !CONFIG_WEBM_IO
2056 FOREACH_STREAM({
2057 if (stream->config.write_webm) {
2058 stream->config.write_webm = 0;
2059 warn("vpxenc was compiled without WebM container support."
2060 "Producing IVF output");
2061 }
2062 });
2063 #endif
2064
2065 /* Use the frame rate from the file only if none was specified
2066 * on the command-line.
2067 */
2068 if (!global.have_framerate) {
2069 global.framerate.num = input.framerate.numerator;
2070 global.framerate.den = input.framerate.denominator;
2071 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2072 stream->config.cfg.g_timebase.num = global.framerate.den);
2073 }
2074
2075 /* Show configuration */
2076 if (global.verbose && pass == 0)
2077 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2078
2079 if (pass == (global.pass ? global.pass - 1 : 0)) {
2080 if (input.file_type == FILE_TYPE_Y4M)
2081 /*The Y4M reader does its own allocation.
2082 Just initialize this here to avoid problems if we never read any
2083 frames.*/
2084 memset(&raw, 0, sizeof(raw));
2085 else
2086 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2087
2088 FOREACH_STREAM(stream->rate_hist =
2089 init_rate_histogram(&stream->config.cfg,
2090 &global.framerate));
2091 }
2092
2093 FOREACH_STREAM(setup_pass(stream, &global, pass));
2094 FOREACH_STREAM(open_output_file(stream, &global,
2095 &input.pixel_aspect_ratio));
2096 FOREACH_STREAM(initialize_encoder(stream, &global));
2097
2098 #if CONFIG_VP9_HIGHBITDEPTH
2099 if (strcmp(global.codec->name, "vp9") == 0 ||
2100 strcmp(global.codec->name, "vp10") == 0) {
2101 // Check to see if at least one stream uses 16 bit internal.
2102 // Currently assume that the bit_depths for all streams using
2103 // highbitdepth are the same.
2104 FOREACH_STREAM({
2105 if (stream->config.use_16bit_internal) {
2106 use_16bit_internal = 1;
2107 }
2108 if (stream->config.cfg.g_profile == 0) {
2109 input_shift = 0;
2110 } else {
2111 input_shift = (int)stream->config.cfg.g_bit_depth -
2112 stream->config.cfg.g_input_bit_depth;
2113 }
2114 });
2115 }
2116 #endif
2117
2118 frame_avail = 1;
2119 got_data = 0;
2120
2121 while (frame_avail || got_data) {
2122 struct vpx_usec_timer timer;
2123
2124 if (!global.limit || frames_in < global.limit) {
2125 frame_avail = read_frame(&input, &raw);
2126
2127 if (frame_avail)
2128 frames_in++;
2129 seen_frames = frames_in > global.skip_frames ?
2130 frames_in - global.skip_frames : 0;
2131
2132 if (!global.quiet) {
2133 float fps = usec_to_fps(cx_time, seen_frames);
2134 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2135
2136 if (stream_cnt == 1)
2137 fprintf(stderr,
2138 "frame %4d/%-4d %7"PRId64"B ",
2139 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2140 else
2141 fprintf(stderr, "frame %4d ", frames_in);
2142
2143 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2144 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2145 cx_time > 9999999 ? "ms" : "us",
2146 fps >= 1.0 ? fps : fps * 60,
2147 fps >= 1.0 ? "fps" : "fpm");
2148 print_time("ETA", estimated_time_left);
2149 }
2150
2151 } else
2152 frame_avail = 0;
2153
2154 if (frames_in > global.skip_frames) {
2155 #if CONFIG_VP9_HIGHBITDEPTH
2156 vpx_image_t *frame_to_encode;
2157 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2158 assert(use_16bit_internal);
2159 // Input bit depth and stream bit depth do not match, so up
2160 // shift frame to stream bit depth
2161 if (!allocated_raw_shift) {
2162 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2163 input.width, input.height, 32);
2164 allocated_raw_shift = 1;
2165 }
2166 vpx_img_upshift(&raw_shift, &raw, input_shift);
2167 frame_to_encode = &raw_shift;
2168 } else {
2169 frame_to_encode = &raw;
2170 }
2171 vpx_usec_timer_start(&timer);
2172 if (use_16bit_internal) {
2173 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2174 FOREACH_STREAM({
2175 if (stream->config.use_16bit_internal)
2176 encode_frame(stream, &global,
2177 frame_avail ? frame_to_encode : NULL,
2178 frames_in);
2179 else
2180 assert(0);
2181 });
2182 } else {
2183 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2184 FOREACH_STREAM(encode_frame(stream, &global,
2185 frame_avail ? frame_to_encode : NULL,
2186 frames_in));
2187 }
2188 #else
2189 vpx_usec_timer_start(&timer);
2190 FOREACH_STREAM(encode_frame(stream, &global,
2191 frame_avail ? &raw : NULL,
2192 frames_in));
2193 #endif
2194 vpx_usec_timer_mark(&timer);
2195 cx_time += vpx_usec_timer_elapsed(&timer);
2196
2197 FOREACH_STREAM(update_quantizer_histogram(stream));
2198
2199 got_data = 0;
2200 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2201
2202 if (!got_data && input.length && streams != NULL &&
2203 !streams->frames_out) {
2204 lagged_count = global.limit ? seen_frames : ftello(input.file);
2205 } else if (input.length) {
2206 int64_t remaining;
2207 int64_t rate;
2208
2209 if (global.limit) {
2210 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2211
2212 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2213 remaining = 1000 * (global.limit - global.skip_frames
2214 - seen_frames + lagged_count);
2215 } else {
2216 const int64_t input_pos = ftello(input.file);
2217 const int64_t input_pos_lagged = input_pos - lagged_count;
2218 const int64_t limit = input.length;
2219
2220 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2221 remaining = limit - input_pos + lagged_count;
2222 }
2223
2224 average_rate = (average_rate <= 0)
2225 ? rate
2226 : (average_rate * 7 + rate) / 8;
2227 estimated_time_left = average_rate ? remaining / average_rate : -1;
2228 }
2229
2230 if (got_data && global.test_decode != TEST_DECODE_OFF)
2231 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2232 }
2233
2234 fflush(stdout);
2235 if (!global.quiet)
2236 fprintf(stderr, "\033[K");
2237 }
2238
2239 if (stream_cnt > 1)
2240 fprintf(stderr, "\n");
2241
2242 if (!global.quiet) {
2243 FOREACH_STREAM(fprintf(stderr,
2244 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2245 " %7"PRId64" %s (%.2f fps)\033[K\n",
2246 pass + 1,
2247 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2248 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2249 seen_frames ? (int64_t)stream->nbytes * 8 *
2250 (int64_t)global.framerate.num / global.framerate.den /
2251 seen_frames : 0,
2252 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2253 stream->cx_time > 9999999 ? "ms" : "us",
2254 usec_to_fps(stream->cx_time, seen_frames)));
2255 }
2256
2257 if (global.show_psnr) {
2258 if (global.codec->fourcc == VP9_FOURCC) {
2259 FOREACH_STREAM(
2260 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2261 } else {
2262 FOREACH_STREAM(show_psnr(stream, 255.0));
2263 }
2264 }
2265
2266 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2267
2268 if (global.test_decode != TEST_DECODE_OFF) {
2269 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2270 }
2271
2272 close_input_file(&input);
2273
2274 if (global.test_decode == TEST_DECODE_FATAL) {
2275 FOREACH_STREAM(res |= stream->mismatch_seen);
2276 }
2277 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2278
2279 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2280
2281 #if CONFIG_FP_MB_STATS
2282 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2283 #endif
2284
2285 if (global.pass)
2286 break;
2287 }
2288
2289 if (global.show_q_hist_buckets)
2290 FOREACH_STREAM(show_q_histogram(stream->counts,
2291 global.show_q_hist_buckets));
2292
2293 if (global.show_rate_hist_buckets)
2294 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2295 &stream->config.cfg,
2296 global.show_rate_hist_buckets));
2297 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2298
2299 #if CONFIG_INTERNAL_STATS
2300 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2301 * to match some existing utilities.
2302 */
2303 if (!(global.pass == 1 && global.passes == 2))
2304 FOREACH_STREAM({
2305 FILE *f = fopen("opsnr.stt", "a");
2306 if (stream->mismatch_seen) {
2307 fprintf(f, "First mismatch occurred in frame %d\n",
2308 stream->mismatch_seen);
2309 } else {
2310 fprintf(f, "No mismatch detected in recon buffers\n");
2311 }
2312 fclose(f);
2313 });
2314 #endif
2315
2316 #if CONFIG_VP9_HIGHBITDEPTH
2317 if (allocated_raw_shift)
2318 vpx_img_free(&raw_shift);
2319 #endif
2320 vpx_img_free(&raw);
2321 free(argv);
2322 free(streams);
2323 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2324 }
AV1 Codec Library