?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge "VP9: _get_pred_context_switchable_interp()"
[aom.git] / vpxenc.c
blobf2def54684d6405e43863525129d27e69b69060b
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
448 static const arg_def_t target_level = ARG_DEF(
449 NULL, "target-level", 1,
450 "Target level (255: off (default); 0: only keep level stats; 10: level 1.0;"
451 " 11: level 1.1; ... 62: level 6.2)");
452 #endif
453
454 #if CONFIG_VP9_ENCODER
455 static const arg_def_t *vp9_args[] = {
456 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
457 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
458 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
459 &gf_cbr_boost_pct, &lossless,
460 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
461 &noise_sens, &tune_content, &input_color_space,
462 &min_gf_interval, &max_gf_interval, &target_level,
463 #if CONFIG_VP9_HIGHBITDEPTH
464 &bitdeptharg, &inbitdeptharg,
465 #endif // CONFIG_VP9_HIGHBITDEPTH
466 NULL
467 };
468 static const int vp9_arg_ctrl_map[] = {
469 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
470 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
471 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
472 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
473 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
474 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
475 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
476 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
477 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
478 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL, VP9E_SET_TARGET_LEVEL,
479 0
480 };
481 #endif
482
483 #if CONFIG_VP10_ENCODER
484 static const arg_def_t *vp10_args[] = {
485 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
486 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
487 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
488 &gf_cbr_boost_pct, &lossless,
489 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
490 &noise_sens, &tune_content, &input_color_space,
491 &min_gf_interval, &max_gf_interval,
492 #if CONFIG_VP9_HIGHBITDEPTH
493 &bitdeptharg, &inbitdeptharg,
494 #endif // CONFIG_VP9_HIGHBITDEPTH
495 NULL
496 };
497 static const int vp10_arg_ctrl_map[] = {
498 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
499 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
500 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
501 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
502 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
503 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
504 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
505 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
506 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
507 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL,
508 0
509 };
510 #endif
511
512 static const arg_def_t *no_args[] = { NULL };
513
514 void usage_exit(void) {
515 int i;
516 const int num_encoder = get_vpx_encoder_count();
517
518 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
519 exec_name);
520
521 fprintf(stderr, "\nOptions:\n");
522 arg_show_usage(stderr, main_args);
523 fprintf(stderr, "\nEncoder Global Options:\n");
524 arg_show_usage(stderr, global_args);
525 fprintf(stderr, "\nRate Control Options:\n");
526 arg_show_usage(stderr, rc_args);
527 fprintf(stderr, "\nTwopass Rate Control Options:\n");
528 arg_show_usage(stderr, rc_twopass_args);
529 fprintf(stderr, "\nKeyframe Placement Options:\n");
530 arg_show_usage(stderr, kf_args);
531 #if CONFIG_VP8_ENCODER
532 fprintf(stderr, "\nVP8 Specific Options:\n");
533 arg_show_usage(stderr, vp8_args);
534 #endif
535 #if CONFIG_VP9_ENCODER
536 fprintf(stderr, "\nVP9 Specific Options:\n");
537 arg_show_usage(stderr, vp9_args);
538 #endif
539 #if CONFIG_VP10_ENCODER
540 fprintf(stderr, "\nVP10 Specific Options:\n");
541 arg_show_usage(stderr, vp10_args);
542 #endif
543 fprintf(stderr, "\nStream timebase (--timebase):\n"
544 " The desired precision of timestamps in the output, expressed\n"
545 " in fractional seconds. Default is 1/1000.\n");
546 fprintf(stderr, "\nIncluded encoders:\n\n");
547
548 for (i = 0; i < num_encoder; ++i) {
549 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
550 const char* defstr = (i == (num_encoder - 1)) ? "(default)" : "";
551 fprintf(stderr, " %-6s - %s %s\n",
552 encoder->name, vpx_codec_iface_name(encoder->codec_interface()),
553 defstr);
554 }
555 fprintf(stderr, "\n ");
556 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
557
558 exit(EXIT_FAILURE);
559 }
560
561 #define mmin(a, b) ((a) < (b) ? (a) : (b))
562
563 #if CONFIG_VP9_HIGHBITDEPTH
564 static void find_mismatch_high(const vpx_image_t *const img1,
565 const vpx_image_t *const img2,
566 int yloc[4], int uloc[4], int vloc[4]) {
567 uint16_t *plane1, *plane2;
568 uint32_t stride1, stride2;
569 const uint32_t bsize = 64;
570 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
571 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
572 const uint32_t c_w =
573 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
574 const uint32_t c_h =
575 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
576 int match = 1;
577 uint32_t i, j;
578 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
579 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
580 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
581 stride1 = img1->stride[VPX_PLANE_Y]/2;
582 stride2 = img2->stride[VPX_PLANE_Y]/2;
583 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
584 for (j = 0; match && j < img1->d_w; j += bsize) {
585 int k, l;
586 const int si = mmin(i + bsize, img1->d_h) - i;
587 const int sj = mmin(j + bsize, img1->d_w) - j;
588 for (k = 0; match && k < si; ++k) {
589 for (l = 0; match && l < sj; ++l) {
590 if (*(plane1 + (i + k) * stride1 + j + l) !=
591 *(plane2 + (i + k) * stride2 + j + l)) {
592 yloc[0] = i + k;
593 yloc[1] = j + l;
594 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
595 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
596 match = 0;
597 break;
598 }
599 }
600 }
601 }
602 }
603
604 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
605 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
606 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
607 stride1 = img1->stride[VPX_PLANE_U]/2;
608 stride2 = img2->stride[VPX_PLANE_U]/2;
609 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
610 for (j = 0; match && j < c_w; j += bsizex) {
611 int k, l;
612 const int si = mmin(i + bsizey, c_h - i);
613 const int sj = mmin(j + bsizex, c_w - j);
614 for (k = 0; match && k < si; ++k) {
615 for (l = 0; match && l < sj; ++l) {
616 if (*(plane1 + (i + k) * stride1 + j + l) !=
617 *(plane2 + (i + k) * stride2 + j + l)) {
618 uloc[0] = i + k;
619 uloc[1] = j + l;
620 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
621 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
622 match = 0;
623 break;
624 }
625 }
626 }
627 }
628 }
629
630 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
631 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
632 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
633 stride1 = img1->stride[VPX_PLANE_V]/2;
634 stride2 = img2->stride[VPX_PLANE_V]/2;
635 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
636 for (j = 0; match && j < c_w; j += bsizex) {
637 int k, l;
638 const int si = mmin(i + bsizey, c_h - i);
639 const int sj = mmin(j + bsizex, c_w - j);
640 for (k = 0; match && k < si; ++k) {
641 for (l = 0; match && l < sj; ++l) {
642 if (*(plane1 + (i + k) * stride1 + j + l) !=
643 *(plane2 + (i + k) * stride2 + j + l)) {
644 vloc[0] = i + k;
645 vloc[1] = j + l;
646 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
647 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
648 match = 0;
649 break;
650 }
651 }
652 }
653 }
654 }
655 }
656 #endif
657
658 static void find_mismatch(const vpx_image_t *const img1,
659 const vpx_image_t *const img2,
660 int yloc[4], int uloc[4], int vloc[4]) {
661 const uint32_t bsize = 64;
662 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
663 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
664 const uint32_t c_w =
665 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
666 const uint32_t c_h =
667 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
668 int match = 1;
669 uint32_t i, j;
670 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
671 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
672 for (j = 0; match && j < img1->d_w; j += bsize) {
673 int k, l;
674 const int si = mmin(i + bsize, img1->d_h) - i;
675 const int sj = mmin(j + bsize, img1->d_w) - j;
676 for (k = 0; match && k < si; ++k) {
677 for (l = 0; match && l < sj; ++l) {
678 if (*(img1->planes[VPX_PLANE_Y] +
679 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
680 *(img2->planes[VPX_PLANE_Y] +
681 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
682 yloc[0] = i + k;
683 yloc[1] = j + l;
684 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
685 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
686 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
687 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
688 match = 0;
689 break;
690 }
691 }
692 }
693 }
694 }
695
696 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
697 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
698 for (j = 0; match && j < c_w; j += bsizex) {
699 int k, l;
700 const int si = mmin(i + bsizey, c_h - i);
701 const int sj = mmin(j + bsizex, c_w - j);
702 for (k = 0; match && k < si; ++k) {
703 for (l = 0; match && l < sj; ++l) {
704 if (*(img1->planes[VPX_PLANE_U] +
705 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
706 *(img2->planes[VPX_PLANE_U] +
707 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
708 uloc[0] = i + k;
709 uloc[1] = j + l;
710 uloc[2] = *(img1->planes[VPX_PLANE_U] +
711 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
712 uloc[3] = *(img2->planes[VPX_PLANE_U] +
713 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
714 match = 0;
715 break;
716 }
717 }
718 }
719 }
720 }
721 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
722 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
723 for (j = 0; match && j < c_w; j += bsizex) {
724 int k, l;
725 const int si = mmin(i + bsizey, c_h - i);
726 const int sj = mmin(j + bsizex, c_w - j);
727 for (k = 0; match && k < si; ++k) {
728 for (l = 0; match && l < sj; ++l) {
729 if (*(img1->planes[VPX_PLANE_V] +
730 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
731 *(img2->planes[VPX_PLANE_V] +
732 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
733 vloc[0] = i + k;
734 vloc[1] = j + l;
735 vloc[2] = *(img1->planes[VPX_PLANE_V] +
736 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
737 vloc[3] = *(img2->planes[VPX_PLANE_V] +
738 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
739 match = 0;
740 break;
741 }
742 }
743 }
744 }
745 }
746 }
747
748 static int compare_img(const vpx_image_t *const img1,
749 const vpx_image_t *const img2) {
750 uint32_t l_w = img1->d_w;
751 uint32_t c_w =
752 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
753 const uint32_t c_h =
754 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
755 uint32_t i;
756 int match = 1;
757
758 match &= (img1->fmt == img2->fmt);
759 match &= (img1->d_w == img2->d_w);
760 match &= (img1->d_h == img2->d_h);
761 #if CONFIG_VP9_HIGHBITDEPTH
762 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
763 l_w *= 2;
764 c_w *= 2;
765 }
766 #endif
767
768 for (i = 0; i < img1->d_h; ++i)
769 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
770 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
771 l_w) == 0);
772
773 for (i = 0; i < c_h; ++i)
774 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
775 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
776 c_w) == 0);
777
778 for (i = 0; i < c_h; ++i)
779 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
780 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
781 c_w) == 0);
782
783 return match;
784 }
785
786
787 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
788 #if CONFIG_VP10_ENCODER
789 #define ARG_CTRL_CNT_MAX NELEMENTS(vp10_arg_ctrl_map)
790 #elif CONFIG_VP9_ENCODER
791 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
792 #else
793 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
794 #endif
795
796 #if !CONFIG_WEBM_IO
797 typedef int stereo_format_t;
798 struct WebmOutputContext { int debug; };
799 #endif
800
801 /* Per-stream configuration */
802 struct stream_config {
803 struct vpx_codec_enc_cfg cfg;
804 const char *out_fn;
805 const char *stats_fn;
806 #if CONFIG_FP_MB_STATS
807 const char *fpmb_stats_fn;
808 #endif
809 stereo_format_t stereo_fmt;
810 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
811 int arg_ctrl_cnt;
812 int write_webm;
813 #if CONFIG_VP9_HIGHBITDEPTH
814 // whether to use 16bit internal buffers
815 int use_16bit_internal;
816 #endif
817 };
818
819
820 struct stream_state {
821 int index;
822 struct stream_state *next;
823 struct stream_config config;
824 FILE *file;
825 struct rate_hist *rate_hist;
826 struct WebmOutputContext webm_ctx;
827 uint64_t psnr_sse_total;
828 uint64_t psnr_samples_total;
829 double psnr_totals[4];
830 int psnr_count;
831 int counts[64];
832 vpx_codec_ctx_t encoder;
833 unsigned int frames_out;
834 uint64_t cx_time;
835 size_t nbytes;
836 stats_io_t stats;
837 #if CONFIG_FP_MB_STATS
838 stats_io_t fpmb_stats;
839 #endif
840 struct vpx_image *img;
841 vpx_codec_ctx_t decoder;
842 int mismatch_seen;
843 };
844
845
846 static void validate_positive_rational(const char *msg,
847 struct vpx_rational *rat) {
848 if (rat->den < 0) {
849 rat->num *= -1;
850 rat->den *= -1;
851 }
852
853 if (rat->num < 0)
854 die("Error: %s must be positive\n", msg);
855
856 if (!rat->den)
857 die("Error: %s has zero denominator\n", msg);
858 }
859
860
861 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
862 char **argi, **argj;
863 struct arg arg;
864 const int num_encoder = get_vpx_encoder_count();
865
866 if (num_encoder < 1)
867 die("Error: no valid encoder available\n");
868
869 /* Initialize default parameters */
870 memset(global, 0, sizeof(*global));
871 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
872 global->passes = 0;
873 global->color_type = I420;
874 /* Assign default deadline to good quality */
875 global->deadline = VPX_DL_GOOD_QUALITY;
876
877 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
878 arg.argv_step = 1;
879
880 if (arg_match(&arg, &codecarg, argi)) {
881 global->codec = get_vpx_encoder_by_name(arg.val);
882 if (!global->codec)
883 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
884 } else if (arg_match(&arg, &passes, argi)) {
885 global->passes = arg_parse_uint(&arg);
886
887 if (global->passes < 1 || global->passes > 2)
888 die("Error: Invalid number of passes (%d)\n", global->passes);
889 } else if (arg_match(&arg, &pass_arg, argi)) {
890 global->pass = arg_parse_uint(&arg);
891
892 if (global->pass < 1 || global->pass > 2)
893 die("Error: Invalid pass selected (%d)\n",
894 global->pass);
895 } else if (arg_match(&arg, &usage, argi))
896 global->usage = arg_parse_uint(&arg);
897 else if (arg_match(&arg, &deadline, argi))
898 global->deadline = arg_parse_uint(&arg);
899 else if (arg_match(&arg, &best_dl, argi))
900 global->deadline = VPX_DL_BEST_QUALITY;
901 else if (arg_match(&arg, &good_dl, argi))
902 global->deadline = VPX_DL_GOOD_QUALITY;
903 else if (arg_match(&arg, &rt_dl, argi))
904 global->deadline = VPX_DL_REALTIME;
905 else if (arg_match(&arg, &use_yv12, argi))
906 global->color_type = YV12;
907 else if (arg_match(&arg, &use_i420, argi))
908 global->color_type = I420;
909 else if (arg_match(&arg, &use_i422, argi))
910 global->color_type = I422;
911 else if (arg_match(&arg, &use_i444, argi))
912 global->color_type = I444;
913 else if (arg_match(&arg, &use_i440, argi))
914 global->color_type = I440;
915 else if (arg_match(&arg, &quietarg, argi))
916 global->quiet = 1;
917 else if (arg_match(&arg, &verbosearg, argi))
918 global->verbose = 1;
919 else if (arg_match(&arg, &limit, argi))
920 global->limit = arg_parse_uint(&arg);
921 else if (arg_match(&arg, &skip, argi))
922 global->skip_frames = arg_parse_uint(&arg);
923 else if (arg_match(&arg, &psnrarg, argi))
924 global->show_psnr = 1;
925 else if (arg_match(&arg, &recontest, argi))
926 global->test_decode = arg_parse_enum_or_int(&arg);
927 else if (arg_match(&arg, &framerate, argi)) {
928 global->framerate = arg_parse_rational(&arg);
929 validate_positive_rational(arg.name, &global->framerate);
930 global->have_framerate = 1;
931 } else if (arg_match(&arg, &out_part, argi))
932 global->out_part = 1;
933 else if (arg_match(&arg, &debugmode, argi))
934 global->debug = 1;
935 else if (arg_match(&arg, &q_hist_n, argi))
936 global->show_q_hist_buckets = arg_parse_uint(&arg);
937 else if (arg_match(&arg, &rate_hist_n, argi))
938 global->show_rate_hist_buckets = arg_parse_uint(&arg);
939 else if (arg_match(&arg, &disable_warnings, argi))
940 global->disable_warnings = 1;
941 else if (arg_match(&arg, &disable_warning_prompt, argi))
942 global->disable_warning_prompt = 1;
943 else
944 argj++;
945 }
946
947 if (global->pass) {
948 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
949 if (global->pass > global->passes) {
950 warn("Assuming --pass=%d implies --passes=%d\n",
951 global->pass, global->pass);
952 global->passes = global->pass;
953 }
954 }
955 /* Validate global config */
956 if (global->passes == 0) {
957 #if CONFIG_VP9_ENCODER || CONFIG_VP10_ENCODER
958 // Make default VP9 passes = 2 until there is a better quality 1-pass
959 // encoder
960 if (global->codec != NULL && global->codec->name != NULL)
961 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
962 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
963 #else
964 global->passes = 1;
965 #endif
966 }
967
968 if (global->deadline == VPX_DL_REALTIME &&
969 global->passes > 1) {
970 warn("Enforcing one-pass encoding in realtime mode\n");
971 global->passes = 1;
972 }
973 }
974
975
976 static void open_input_file(struct VpxInputContext *input) {
977 /* Parse certain options from the input file, if possible */
978 input->file = strcmp(input->filename, "-")
979 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
980
981 if (!input->file)
982 fatal("Failed to open input file");
983
984 if (!fseeko(input->file, 0, SEEK_END)) {
985 /* Input file is seekable. Figure out how long it is, so we can get
986 * progress info.
987 */
988 input->length = ftello(input->file);
989 rewind(input->file);
990 }
991
992 /* Default to 1:1 pixel aspect ratio. */
993 input->pixel_aspect_ratio.numerator = 1;
994 input->pixel_aspect_ratio.denominator = 1;
995
996 /* For RAW input sources, these bytes will applied on the first frame
997 * in read_frame().
998 */
999 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1000 input->detect.position = 0;
1001
1002 if (input->detect.buf_read == 4
1003 && file_is_y4m(input->detect.buf)) {
1004 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1005 input->only_i420) >= 0) {
1006 input->file_type = FILE_TYPE_Y4M;
1007 input->width = input->y4m.pic_w;
1008 input->height = input->y4m.pic_h;
1009 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1010 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1011 input->framerate.numerator = input->y4m.fps_n;
1012 input->framerate.denominator = input->y4m.fps_d;
1013 input->fmt = input->y4m.vpx_fmt;
1014 input->bit_depth = input->y4m.bit_depth;
1015 } else
1016 fatal("Unsupported Y4M stream.");
1017 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1018 fatal("IVF is not supported as input.");
1019 } else {
1020 input->file_type = FILE_TYPE_RAW;
1021 }
1022 }
1023
1024
1025 static void close_input_file(struct VpxInputContext *input) {
1026 fclose(input->file);
1027 if (input->file_type == FILE_TYPE_Y4M)
1028 y4m_input_close(&input->y4m);
1029 }
1030
1031 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1032 struct stream_state *prev) {
1033 struct stream_state *stream;
1034
1035 stream = calloc(1, sizeof(*stream));
1036 if (stream == NULL) {
1037 fatal("Failed to allocate new stream.");
1038 }
1039
1040 if (prev) {
1041 memcpy(stream, prev, sizeof(*stream));
1042 stream->index++;
1043 prev->next = stream;
1044 } else {
1045 vpx_codec_err_t res;
1046
1047 /* Populate encoder configuration */
1048 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1049 &stream->config.cfg,
1050 global->usage);
1051 if (res)
1052 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1053
1054 /* Change the default timebase to a high enough value so that the
1055 * encoder will always create strictly increasing timestamps.
1056 */
1057 stream->config.cfg.g_timebase.den = 1000;
1058
1059 /* Never use the library's default resolution, require it be parsed
1060 * from the file or set on the command line.
1061 */
1062 stream->config.cfg.g_w = 0;
1063 stream->config.cfg.g_h = 0;
1064
1065 /* Initialize remaining stream parameters */
1066 stream->config.write_webm = 1;
1067 #if CONFIG_WEBM_IO
1068 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1069 stream->webm_ctx.last_pts_ns = -1;
1070 stream->webm_ctx.writer = NULL;
1071 stream->webm_ctx.segment = NULL;
1072 #endif
1073
1074 /* Allows removal of the application version from the EBML tags */
1075 stream->webm_ctx.debug = global->debug;
1076
1077 /* Default lag_in_frames is 0 in realtime mode */
1078 if (global->deadline == VPX_DL_REALTIME)
1079 stream->config.cfg.g_lag_in_frames = 0;
1080 }
1081
1082 /* Output files must be specified for each stream */
1083 stream->config.out_fn = NULL;
1084
1085 stream->next = NULL;
1086 return stream;
1087 }
1088
1089
1090 static int parse_stream_params(struct VpxEncoderConfig *global,
1091 struct stream_state *stream,
1092 char **argv) {
1093 char **argi, **argj;
1094 struct arg arg;
1095 static const arg_def_t **ctrl_args = no_args;
1096 static const int *ctrl_args_map = NULL;
1097 struct stream_config *config = &stream->config;
1098 int eos_mark_found = 0;
1099 #if CONFIG_VP9_HIGHBITDEPTH
1100 int test_16bit_internal = 0;
1101 #endif
1102
1103 // Handle codec specific options
1104 if (0) {
1105 #if CONFIG_VP8_ENCODER
1106 } else if (strcmp(global->codec->name, "vp8") == 0) {
1107 ctrl_args = vp8_args;
1108 ctrl_args_map = vp8_arg_ctrl_map;
1109 #endif
1110 #if CONFIG_VP9_ENCODER
1111 } else if (strcmp(global->codec->name, "vp9") == 0) {
1112 ctrl_args = vp9_args;
1113 ctrl_args_map = vp9_arg_ctrl_map;
1114 #endif
1115 #if CONFIG_VP10_ENCODER
1116 } else if (strcmp(global->codec->name, "vp10") == 0) {
1117 // TODO(jingning): Reuse VP9 specific encoder configuration parameters.
1118 // Consider to expand this set for VP10 encoder control.
1119 ctrl_args = vp10_args;
1120 ctrl_args_map = vp10_arg_ctrl_map;
1121 #endif
1122 }
1123
1124 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1125 arg.argv_step = 1;
1126
1127 /* Once we've found an end-of-stream marker (--) we want to continue
1128 * shifting arguments but not consuming them.
1129 */
1130 if (eos_mark_found) {
1131 argj++;
1132 continue;
1133 } else if (!strcmp(*argj, "--")) {
1134 eos_mark_found = 1;
1135 continue;
1136 }
1137
1138 if (arg_match(&arg, &outputfile, argi)) {
1139 config->out_fn = arg.val;
1140 } else if (arg_match(&arg, &fpf_name, argi)) {
1141 config->stats_fn = arg.val;
1142 #if CONFIG_FP_MB_STATS
1143 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1144 config->fpmb_stats_fn = arg.val;
1145 #endif
1146 } else if (arg_match(&arg, &use_webm, argi)) {
1147 #if CONFIG_WEBM_IO
1148 config->write_webm = 1;
1149 #else
1150 die("Error: --webm specified but webm is disabled.");
1151 #endif
1152 } else if (arg_match(&arg, &use_ivf, argi)) {
1153 config->write_webm = 0;
1154 } else if (arg_match(&arg, &threads, argi)) {
1155 config->cfg.g_threads = arg_parse_uint(&arg);
1156 } else if (arg_match(&arg, &profile, argi)) {
1157 config->cfg.g_profile = arg_parse_uint(&arg);
1158 } else if (arg_match(&arg, &width, argi)) {
1159 config->cfg.g_w = arg_parse_uint(&arg);
1160 } else if (arg_match(&arg, &height, argi)) {
1161 config->cfg.g_h = arg_parse_uint(&arg);
1162 #if CONFIG_VP9_HIGHBITDEPTH
1163 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1164 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1165 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1166 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1167 #endif
1168 #if CONFIG_WEBM_IO
1169 } else if (arg_match(&arg, &stereo_mode, argi)) {
1170 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1171 #endif
1172 } else if (arg_match(&arg, &timebase, argi)) {
1173 config->cfg.g_timebase = arg_parse_rational(&arg);
1174 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1175 } else if (arg_match(&arg, &error_resilient, argi)) {
1176 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1177 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1178 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1179 if (global->deadline == VPX_DL_REALTIME &&
1180 config->cfg.g_lag_in_frames != 0) {
1181 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1182 config->cfg.g_lag_in_frames = 0;
1183 }
1184 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1185 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1186 } else if (arg_match(&arg, &resize_allowed, argi)) {
1187 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1188 } else if (arg_match(&arg, &resize_width, argi)) {
1189 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1190 } else if (arg_match(&arg, &resize_height, argi)) {
1191 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1192 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1193 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1194 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1195 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1196 } else if (arg_match(&arg, &end_usage, argi)) {
1197 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1198 } else if (arg_match(&arg, &target_bitrate, argi)) {
1199 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1200 } else if (arg_match(&arg, &min_quantizer, argi)) {
1201 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1202 } else if (arg_match(&arg, &max_quantizer, argi)) {
1203 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1204 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1205 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1206 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1207 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1208 } else if (arg_match(&arg, &buf_sz, argi)) {
1209 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1210 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1211 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1212 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1213 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1214 } else if (arg_match(&arg, &bias_pct, argi)) {
1215 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1216 if (global->passes < 2)
1217 warn("option %s ignored in one-pass mode.\n", arg.name);
1218 } else if (arg_match(&arg, &minsection_pct, argi)) {
1219 config->cfg.rc_2pass_vbr_minsection_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, &maxsection_pct, argi)) {
1224 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1225
1226 if (global->passes < 2)
1227 warn("option %s ignored in one-pass mode.\n", arg.name);
1228 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1229 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1230 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1231 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1232 } else if (arg_match(&arg, &kf_disabled, argi)) {
1233 config->cfg.kf_mode = VPX_KF_DISABLED;
1234 #if CONFIG_VP9_HIGHBITDEPTH
1235 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1236 if (strcmp(global->codec->name, "vp9") == 0 ||
1237 strcmp(global->codec->name, "vp10") == 0) {
1238 test_16bit_internal = 1;
1239 }
1240 #endif
1241 } else {
1242 int i, match = 0;
1243 for (i = 0; ctrl_args[i]; i++) {
1244 if (arg_match(&arg, ctrl_args[i], argi)) {
1245 int j;
1246 match = 1;
1247
1248 /* Point either to the next free element or the first
1249 * instance of this control.
1250 */
1251 for (j = 0; j < config->arg_ctrl_cnt; j++)
1252 if (ctrl_args_map != NULL &&
1253 config->arg_ctrls[j][0] == ctrl_args_map[i])
1254 break;
1255
1256 /* Update/insert */
1257 assert(j < (int)ARG_CTRL_CNT_MAX);
1258 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1259 config->arg_ctrls[j][0] = ctrl_args_map[i];
1260 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1261 if (j == config->arg_ctrl_cnt)
1262 config->arg_ctrl_cnt++;
1263 }
1264 }
1265 }
1266 if (!match)
1267 argj++;
1268 }
1269 }
1270 #if CONFIG_VP9_HIGHBITDEPTH
1271 if (strcmp(global->codec->name, "vp9") == 0 ||
1272 strcmp(global->codec->name, "vp10") == 0) {
1273 config->use_16bit_internal = test_16bit_internal |
1274 (config->cfg.g_profile > 1);
1275 }
1276 #endif
1277 return eos_mark_found;
1278 }
1279
1280
1281 #define FOREACH_STREAM(func) \
1282 do { \
1283 struct stream_state *stream; \
1284 for (stream = streams; stream; stream = stream->next) { \
1285 func; \
1286 } \
1287 } while (0)
1288
1289
1290 static void validate_stream_config(const struct stream_state *stream,
1291 const struct VpxEncoderConfig *global) {
1292 const struct stream_state *streami;
1293 (void)global;
1294
1295 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1296 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1297 " and --height (-h)", stream->index);
1298
1299 // Check that the codec bit depth is greater than the input bit depth.
1300 if (stream->config.cfg.g_input_bit_depth >
1301 (unsigned int)stream->config.cfg.g_bit_depth) {
1302 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1303 stream->index, (int)stream->config.cfg.g_bit_depth,
1304 stream->config.cfg.g_input_bit_depth);
1305 }
1306
1307 for (streami = stream; streami; streami = streami->next) {
1308 /* All streams require output files */
1309 if (!streami->config.out_fn)
1310 fatal("Stream %d: Output file is required (specify with -o)",
1311 streami->index);
1312
1313 /* Check for two streams outputting to the same file */
1314 if (streami != stream) {
1315 const char *a = stream->config.out_fn;
1316 const char *b = streami->config.out_fn;
1317 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1318 fatal("Stream %d: duplicate output file (from stream %d)",
1319 streami->index, stream->index);
1320 }
1321
1322 /* Check for two streams sharing a stats file. */
1323 if (streami != stream) {
1324 const char *a = stream->config.stats_fn;
1325 const char *b = streami->config.stats_fn;
1326 if (a && b && !strcmp(a, b))
1327 fatal("Stream %d: duplicate stats file (from stream %d)",
1328 streami->index, stream->index);
1329 }
1330
1331 #if CONFIG_FP_MB_STATS
1332 /* Check for two streams sharing a mb stats file. */
1333 if (streami != stream) {
1334 const char *a = stream->config.fpmb_stats_fn;
1335 const char *b = streami->config.fpmb_stats_fn;
1336 if (a && b && !strcmp(a, b))
1337 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1338 streami->index, stream->index);
1339 }
1340 #endif
1341 }
1342 }
1343
1344
1345 static void set_stream_dimensions(struct stream_state *stream,
1346 unsigned int w,
1347 unsigned int h) {
1348 if (!stream->config.cfg.g_w) {
1349 if (!stream->config.cfg.g_h)
1350 stream->config.cfg.g_w = w;
1351 else
1352 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1353 }
1354 if (!stream->config.cfg.g_h) {
1355 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1356 }
1357 }
1358
1359 static const char* file_type_to_string(enum VideoFileType t) {
1360 switch (t) {
1361 case FILE_TYPE_RAW: return "RAW";
1362 case FILE_TYPE_Y4M: return "Y4M";
1363 default: return "Other";
1364 }
1365 }
1366
1367 static const char* image_format_to_string(vpx_img_fmt_t f) {
1368 switch (f) {
1369 case VPX_IMG_FMT_I420: return "I420";
1370 case VPX_IMG_FMT_I422: return "I422";
1371 case VPX_IMG_FMT_I444: return "I444";
1372 case VPX_IMG_FMT_I440: return "I440";
1373 case VPX_IMG_FMT_YV12: return "YV12";
1374 case VPX_IMG_FMT_I42016: return "I42016";
1375 case VPX_IMG_FMT_I42216: return "I42216";
1376 case VPX_IMG_FMT_I44416: return "I44416";
1377 case VPX_IMG_FMT_I44016: return "I44016";
1378 default: return "Other";
1379 }
1380 }
1381
1382 static void show_stream_config(struct stream_state *stream,
1383 struct VpxEncoderConfig *global,
1384 struct VpxInputContext *input) {
1385
1386 #define SHOW(field) \
1387 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1388
1389 if (stream->index == 0) {
1390 fprintf(stderr, "Codec: %s\n",
1391 vpx_codec_iface_name(global->codec->codec_interface()));
1392 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1393 input->filename,
1394 file_type_to_string(input->file_type),
1395 image_format_to_string(input->fmt));
1396 }
1397 if (stream->next || stream->index)
1398 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1399 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1400 fprintf(stderr, "Encoder parameters:\n");
1401
1402 SHOW(g_usage);
1403 SHOW(g_threads);
1404 SHOW(g_profile);
1405 SHOW(g_w);
1406 SHOW(g_h);
1407 SHOW(g_bit_depth);
1408 SHOW(g_input_bit_depth);
1409 SHOW(g_timebase.num);
1410 SHOW(g_timebase.den);
1411 SHOW(g_error_resilient);
1412 SHOW(g_pass);
1413 SHOW(g_lag_in_frames);
1414 SHOW(rc_dropframe_thresh);
1415 SHOW(rc_resize_allowed);
1416 SHOW(rc_scaled_width);
1417 SHOW(rc_scaled_height);
1418 SHOW(rc_resize_up_thresh);
1419 SHOW(rc_resize_down_thresh);
1420 SHOW(rc_end_usage);
1421 SHOW(rc_target_bitrate);
1422 SHOW(rc_min_quantizer);
1423 SHOW(rc_max_quantizer);
1424 SHOW(rc_undershoot_pct);
1425 SHOW(rc_overshoot_pct);
1426 SHOW(rc_buf_sz);
1427 SHOW(rc_buf_initial_sz);
1428 SHOW(rc_buf_optimal_sz);
1429 SHOW(rc_2pass_vbr_bias_pct);
1430 SHOW(rc_2pass_vbr_minsection_pct);
1431 SHOW(rc_2pass_vbr_maxsection_pct);
1432 SHOW(kf_mode);
1433 SHOW(kf_min_dist);
1434 SHOW(kf_max_dist);
1435 }
1436
1437
1438 static void open_output_file(struct stream_state *stream,
1439 struct VpxEncoderConfig *global,
1440 const struct VpxRational *pixel_aspect_ratio) {
1441 const char *fn = stream->config.out_fn;
1442 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1443
1444 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1445 return;
1446
1447 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1448
1449 if (!stream->file)
1450 fatal("Failed to open output file");
1451
1452 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1453 fatal("WebM output to pipes not supported.");
1454
1455 #if CONFIG_WEBM_IO
1456 if (stream->config.write_webm) {
1457 stream->webm_ctx.stream = stream->file;
1458 write_webm_file_header(&stream->webm_ctx, cfg,
1459 &global->framerate,
1460 stream->config.stereo_fmt,
1461 global->codec->fourcc,
1462 pixel_aspect_ratio);
1463 }
1464 #else
1465 (void)pixel_aspect_ratio;
1466 #endif
1467
1468 if (!stream->config.write_webm) {
1469 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1470 }
1471 }
1472
1473
1474 static void close_output_file(struct stream_state *stream,
1475 unsigned int fourcc) {
1476 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1477
1478 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1479 return;
1480
1481 #if CONFIG_WEBM_IO
1482 if (stream->config.write_webm) {
1483 write_webm_file_footer(&stream->webm_ctx);
1484 }
1485 #endif
1486
1487 if (!stream->config.write_webm) {
1488 if (!fseek(stream->file, 0, SEEK_SET))
1489 ivf_write_file_header(stream->file, &stream->config.cfg,
1490 fourcc,
1491 stream->frames_out);
1492 }
1493
1494 fclose(stream->file);
1495 }
1496
1497
1498 static void setup_pass(struct stream_state *stream,
1499 struct VpxEncoderConfig *global,
1500 int pass) {
1501 if (stream->config.stats_fn) {
1502 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1503 pass))
1504 fatal("Failed to open statistics store");
1505 } else {
1506 if (!stats_open_mem(&stream->stats, pass))
1507 fatal("Failed to open statistics store");
1508 }
1509
1510 #if CONFIG_FP_MB_STATS
1511 if (stream->config.fpmb_stats_fn) {
1512 if (!stats_open_file(&stream->fpmb_stats,
1513 stream->config.fpmb_stats_fn, pass))
1514 fatal("Failed to open mb statistics store");
1515 } else {
1516 if (!stats_open_mem(&stream->fpmb_stats, pass))
1517 fatal("Failed to open mb statistics store");
1518 }
1519 #endif
1520
1521 stream->config.cfg.g_pass = global->passes == 2
1522 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1523 : VPX_RC_ONE_PASS;
1524 if (pass) {
1525 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1526 #if CONFIG_FP_MB_STATS
1527 stream->config.cfg.rc_firstpass_mb_stats_in =
1528 stats_get(&stream->fpmb_stats);
1529 #endif
1530 }
1531
1532 stream->cx_time = 0;
1533 stream->nbytes = 0;
1534 stream->frames_out = 0;
1535 }
1536
1537
1538 static void initialize_encoder(struct stream_state *stream,
1539 struct VpxEncoderConfig *global) {
1540 int i;
1541 int flags = 0;
1542
1543 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1544 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1545 #if CONFIG_VP9_HIGHBITDEPTH
1546 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1547 #endif
1548
1549 /* Construct Encoder Context */
1550 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1551 &stream->config.cfg, flags);
1552 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1553
1554 /* Note that we bypass the vpx_codec_control wrapper macro because
1555 * we're being clever to store the control IDs in an array. Real
1556 * applications will want to make use of the enumerations directly
1557 */
1558 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1559 int ctrl = stream->config.arg_ctrls[i][0];
1560 int value = stream->config.arg_ctrls[i][1];
1561 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1562 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1563 ctrl, value);
1564
1565 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1566 }
1567
1568 #if CONFIG_DECODERS
1569 if (global->test_decode != TEST_DECODE_OFF) {
1570 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1571 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1572 }
1573 #endif
1574 }
1575
1576
1577 static void encode_frame(struct stream_state *stream,
1578 struct VpxEncoderConfig *global,
1579 struct vpx_image *img,
1580 unsigned int frames_in) {
1581 vpx_codec_pts_t frame_start, next_frame_start;
1582 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1583 struct vpx_usec_timer timer;
1584
1585 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1586 * global->framerate.den)
1587 / cfg->g_timebase.num / global->framerate.num;
1588 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1589 * global->framerate.den)
1590 / cfg->g_timebase.num / global->framerate.num;
1591
1592 /* Scale if necessary */
1593 #if CONFIG_VP9_HIGHBITDEPTH
1594 if (img) {
1595 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1596 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1597 if (img->fmt != VPX_IMG_FMT_I42016) {
1598 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1599 exit(EXIT_FAILURE);
1600 }
1601 #if CONFIG_LIBYUV
1602 if (!stream->img) {
1603 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1604 cfg->g_w, cfg->g_h, 16);
1605 }
1606 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1607 img->stride[VPX_PLANE_Y]/2,
1608 (uint16*)img->planes[VPX_PLANE_U],
1609 img->stride[VPX_PLANE_U]/2,
1610 (uint16*)img->planes[VPX_PLANE_V],
1611 img->stride[VPX_PLANE_V]/2,
1612 img->d_w, img->d_h,
1613 (uint16*)stream->img->planes[VPX_PLANE_Y],
1614 stream->img->stride[VPX_PLANE_Y]/2,
1615 (uint16*)stream->img->planes[VPX_PLANE_U],
1616 stream->img->stride[VPX_PLANE_U]/2,
1617 (uint16*)stream->img->planes[VPX_PLANE_V],
1618 stream->img->stride[VPX_PLANE_V]/2,
1619 stream->img->d_w, stream->img->d_h,
1620 kFilterBox);
1621 img = stream->img;
1622 #else
1623 stream->encoder.err = 1;
1624 ctx_exit_on_error(&stream->encoder,
1625 "Stream %d: Failed to encode frame.\n"
1626 "Scaling disabled in this configuration. \n"
1627 "To enable, configure with --enable-libyuv\n",
1628 stream->index);
1629 #endif
1630 }
1631 }
1632 #endif
1633 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1634 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1635 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1636 exit(EXIT_FAILURE);
1637 }
1638 #if CONFIG_LIBYUV
1639 if (!stream->img)
1640 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1641 cfg->g_w, cfg->g_h, 16);
1642 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1643 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1644 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1645 img->d_w, img->d_h,
1646 stream->img->planes[VPX_PLANE_Y],
1647 stream->img->stride[VPX_PLANE_Y],
1648 stream->img->planes[VPX_PLANE_U],
1649 stream->img->stride[VPX_PLANE_U],
1650 stream->img->planes[VPX_PLANE_V],
1651 stream->img->stride[VPX_PLANE_V],
1652 stream->img->d_w, stream->img->d_h,
1653 kFilterBox);
1654 img = stream->img;
1655 #else
1656 stream->encoder.err = 1;
1657 ctx_exit_on_error(&stream->encoder,
1658 "Stream %d: Failed to encode frame.\n"
1659 "Scaling disabled in this configuration. \n"
1660 "To enable, configure with --enable-libyuv\n",
1661 stream->index);
1662 #endif
1663 }
1664
1665 vpx_usec_timer_start(&timer);
1666 vpx_codec_encode(&stream->encoder, img, frame_start,
1667 (unsigned long)(next_frame_start - frame_start),
1668 0, global->deadline);
1669 vpx_usec_timer_mark(&timer);
1670 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1671 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1672 stream->index);
1673 }
1674
1675
1676 static void update_quantizer_histogram(struct stream_state *stream) {
1677 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1678 int q;
1679
1680 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1681 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1682 stream->counts[q]++;
1683 }
1684 }
1685
1686
1687 static void get_cx_data(struct stream_state *stream,
1688 struct VpxEncoderConfig *global,
1689 int *got_data) {
1690 const vpx_codec_cx_pkt_t *pkt;
1691 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1692 vpx_codec_iter_t iter = NULL;
1693
1694 *got_data = 0;
1695 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1696 static size_t fsize = 0;
1697 static int64_t ivf_header_pos = 0;
1698
1699 switch (pkt->kind) {
1700 case VPX_CODEC_CX_FRAME_PKT:
1701 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1702 stream->frames_out++;
1703 }
1704 if (!global->quiet)
1705 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1706
1707 update_rate_histogram(stream->rate_hist, cfg, pkt);
1708 #if CONFIG_WEBM_IO
1709 if (stream->config.write_webm) {
1710 write_webm_block(&stream->webm_ctx, cfg, pkt);
1711 }
1712 #endif
1713 if (!stream->config.write_webm) {
1714 if (pkt->data.frame.partition_id <= 0) {
1715 ivf_header_pos = ftello(stream->file);
1716 fsize = pkt->data.frame.sz;
1717
1718 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1719 } else {
1720 fsize += pkt->data.frame.sz;
1721
1722 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1723 const int64_t currpos = ftello(stream->file);
1724 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1725 ivf_write_frame_size(stream->file, fsize);
1726 fseeko(stream->file, currpos, SEEK_SET);
1727 }
1728 }
1729
1730 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1731 stream->file);
1732 }
1733 stream->nbytes += pkt->data.raw.sz;
1734
1735 *got_data = 1;
1736 #if CONFIG_DECODERS
1737 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1738 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1739 (unsigned int)pkt->data.frame.sz, NULL, 0);
1740 if (stream->decoder.err) {
1741 warn_or_exit_on_error(&stream->decoder,
1742 global->test_decode == TEST_DECODE_FATAL,
1743 "Failed to decode frame %d in stream %d",
1744 stream->frames_out + 1, stream->index);
1745 stream->mismatch_seen = stream->frames_out + 1;
1746 }
1747 }
1748 #endif
1749 break;
1750 case VPX_CODEC_STATS_PKT:
1751 stream->frames_out++;
1752 stats_write(&stream->stats,
1753 pkt->data.twopass_stats.buf,
1754 pkt->data.twopass_stats.sz);
1755 stream->nbytes += pkt->data.raw.sz;
1756 break;
1757 #if CONFIG_FP_MB_STATS
1758 case VPX_CODEC_FPMB_STATS_PKT:
1759 stats_write(&stream->fpmb_stats,
1760 pkt->data.firstpass_mb_stats.buf,
1761 pkt->data.firstpass_mb_stats.sz);
1762 stream->nbytes += pkt->data.raw.sz;
1763 break;
1764 #endif
1765 case VPX_CODEC_PSNR_PKT:
1766
1767 if (global->show_psnr) {
1768 int i;
1769
1770 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1771 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1772 for (i = 0; i < 4; i++) {
1773 if (!global->quiet)
1774 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1775 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1776 }
1777 stream->psnr_count++;
1778 }
1779
1780 break;
1781 default:
1782 break;
1783 }
1784 }
1785 }
1786
1787
1788 static void show_psnr(struct stream_state *stream, double peak) {
1789 int i;
1790 double ovpsnr;
1791
1792 if (!stream->psnr_count)
1793 return;
1794
1795 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1796 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1797 (double)stream->psnr_sse_total);
1798 fprintf(stderr, " %.3f", ovpsnr);
1799
1800 for (i = 0; i < 4; i++) {
1801 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1802 }
1803 fprintf(stderr, "\n");
1804 }
1805
1806
1807 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1808 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1809 }
1810
1811 static void test_decode(struct stream_state *stream,
1812 enum TestDecodeFatality fatal,
1813 const VpxInterface *codec) {
1814 vpx_image_t enc_img, dec_img;
1815
1816 if (stream->mismatch_seen)
1817 return;
1818
1819 /* Get the internal reference frame */
1820 if (strcmp(codec->name, "vp8") == 0) {
1821 struct vpx_ref_frame ref_enc, ref_dec;
1822 int width, height;
1823
1824 width = (stream->config.cfg.g_w + 15) & ~15;
1825 height = (stream->config.cfg.g_h + 15) & ~15;
1826 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1827 enc_img = ref_enc.img;
1828 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1829 dec_img = ref_dec.img;
1830
1831 ref_enc.frame_type = VP8_LAST_FRAME;
1832 ref_dec.frame_type = VP8_LAST_FRAME;
1833 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1834 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1835 } else {
1836 struct vp9_ref_frame ref_enc, ref_dec;
1837
1838 ref_enc.idx = 0;
1839 ref_dec.idx = 0;
1840 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1841 enc_img = ref_enc.img;
1842 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1843 dec_img = ref_dec.img;
1844 #if CONFIG_VP9_HIGHBITDEPTH
1845 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1846 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1847 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1848 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1849 enc_img.d_w, enc_img.d_h, 16);
1850 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1851 }
1852 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1853 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1854 dec_img.d_w, dec_img.d_h, 16);
1855 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1856 }
1857 }
1858 #endif
1859 }
1860 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1861 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1862
1863 if (!compare_img(&enc_img, &dec_img)) {
1864 int y[4], u[4], v[4];
1865 #if CONFIG_VP9_HIGHBITDEPTH
1866 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1867 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1868 } else {
1869 find_mismatch(&enc_img, &dec_img, y, u, v);
1870 }
1871 #else
1872 find_mismatch(&enc_img, &dec_img, y, u, v);
1873 #endif
1874 stream->decoder.err = 1;
1875 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1876 "Stream %d: Encode/decode mismatch on frame %d at"
1877 " Y[%d, %d] {%d/%d},"
1878 " U[%d, %d] {%d/%d},"
1879 " V[%d, %d] {%d/%d}",
1880 stream->index, stream->frames_out,
1881 y[0], y[1], y[2], y[3],
1882 u[0], u[1], u[2], u[3],
1883 v[0], v[1], v[2], v[3]);
1884 stream->mismatch_seen = stream->frames_out;
1885 }
1886
1887 vpx_img_free(&enc_img);
1888 vpx_img_free(&dec_img);
1889 }
1890
1891
1892 static void print_time(const char *label, int64_t etl) {
1893 int64_t hours;
1894 int64_t mins;
1895 int64_t secs;
1896
1897 if (etl >= 0) {
1898 hours = etl / 3600;
1899 etl -= hours * 3600;
1900 mins = etl / 60;
1901 etl -= mins * 60;
1902 secs = etl;
1903
1904 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1905 label, hours, mins, secs);
1906 } else {
1907 fprintf(stderr, "[%3s unknown] ", label);
1908 }
1909 }
1910
1911
1912 int main(int argc, const char **argv_) {
1913 int pass;
1914 vpx_image_t raw;
1915 #if CONFIG_VP9_HIGHBITDEPTH
1916 vpx_image_t raw_shift;
1917 int allocated_raw_shift = 0;
1918 int use_16bit_internal = 0;
1919 int input_shift = 0;
1920 #endif
1921 int frame_avail, got_data;
1922
1923 struct VpxInputContext input;
1924 struct VpxEncoderConfig global;
1925 struct stream_state *streams = NULL;
1926 char **argv, **argi;
1927 uint64_t cx_time = 0;
1928 int stream_cnt = 0;
1929 int res = 0;
1930
1931 memset(&input, 0, sizeof(input));
1932 exec_name = argv_[0];
1933
1934 if (argc < 3)
1935 usage_exit();
1936
1937 /* Setup default input stream settings */
1938 input.framerate.numerator = 30;
1939 input.framerate.denominator = 1;
1940 input.only_i420 = 1;
1941 input.bit_depth = 0;
1942
1943 /* First parse the global configuration values, because we want to apply
1944 * other parameters on top of the default configuration provided by the
1945 * codec.
1946 */
1947 argv = argv_dup(argc - 1, argv_ + 1);
1948 parse_global_config(&global, argv);
1949
1950 switch (global.color_type) {
1951 case I420:
1952 input.fmt = VPX_IMG_FMT_I420;
1953 break;
1954 case I422:
1955 input.fmt = VPX_IMG_FMT_I422;
1956 break;
1957 case I444:
1958 input.fmt = VPX_IMG_FMT_I444;
1959 break;
1960 case I440:
1961 input.fmt = VPX_IMG_FMT_I440;
1962 break;
1963 case YV12:
1964 input.fmt = VPX_IMG_FMT_YV12;
1965 break;
1966 }
1967
1968 {
1969 /* Now parse each stream's parameters. Using a local scope here
1970 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1971 * loops
1972 */
1973 struct stream_state *stream = NULL;
1974
1975 do {
1976 stream = new_stream(&global, stream);
1977 stream_cnt++;
1978 if (!streams)
1979 streams = stream;
1980 } while (parse_stream_params(&global, stream, argv));
1981 }
1982
1983 /* Check for unrecognized options */
1984 for (argi = argv; *argi; argi++)
1985 if (argi[0][0] == '-' && argi[0][1])
1986 die("Error: Unrecognized option %s\n", *argi);
1987
1988 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1989 &global, &stream->config.cfg););
1990
1991 /* Handle non-option arguments */
1992 input.filename = argv[0];
1993
1994 if (!input.filename)
1995 usage_exit();
1996
1997 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1998 if (global.codec->fourcc == VP9_FOURCC || global.codec->fourcc == VP10_FOURCC)
1999 input.only_i420 = 0;
2000
2001 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2002 int frames_in = 0, seen_frames = 0;
2003 int64_t estimated_time_left = -1;
2004 int64_t average_rate = -1;
2005 int64_t lagged_count = 0;
2006
2007 open_input_file(&input);
2008
2009 /* If the input file doesn't specify its w/h (raw files), try to get
2010 * the data from the first stream's configuration.
2011 */
2012 if (!input.width || !input.height) {
2013 FOREACH_STREAM({
2014 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2015 input.width = stream->config.cfg.g_w;
2016 input.height = stream->config.cfg.g_h;
2017 break;
2018 }
2019 });
2020 }
2021
2022 /* Update stream configurations from the input file's parameters */
2023 if (!input.width || !input.height)
2024 fatal("Specify stream dimensions with --width (-w) "
2025 " and --height (-h)");
2026
2027 /* If input file does not specify bit-depth but input-bit-depth parameter
2028 * exists, assume that to be the input bit-depth. However, if the
2029 * input-bit-depth paramter does not exist, assume the input bit-depth
2030 * to be the same as the codec bit-depth.
2031 */
2032 if (!input.bit_depth) {
2033 FOREACH_STREAM({
2034 if (stream->config.cfg.g_input_bit_depth)
2035 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2036 else
2037 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2038 (int)stream->config.cfg.g_bit_depth;
2039 });
2040 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2041 } else {
2042 FOREACH_STREAM({
2043 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2044 });
2045 }
2046
2047 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2048 FOREACH_STREAM(validate_stream_config(stream, &global));
2049
2050 /* Ensure that --passes and --pass are consistent. If --pass is set and
2051 * --passes=2, ensure --fpf was set.
2052 */
2053 if (global.pass && global.passes == 2)
2054 FOREACH_STREAM( {
2055 if (!stream->config.stats_fn)
2056 die("Stream %d: Must specify --fpf when --pass=%d"
2057 " and --passes=2\n", stream->index, global.pass);
2058 });
2059
2060 #if !CONFIG_WEBM_IO
2061 FOREACH_STREAM({
2062 if (stream->config.write_webm) {
2063 stream->config.write_webm = 0;
2064 warn("vpxenc was compiled without WebM container support."
2065 "Producing IVF output");
2066 }
2067 });
2068 #endif
2069
2070 /* Use the frame rate from the file only if none was specified
2071 * on the command-line.
2072 */
2073 if (!global.have_framerate) {
2074 global.framerate.num = input.framerate.numerator;
2075 global.framerate.den = input.framerate.denominator;
2076 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2077 stream->config.cfg.g_timebase.num = global.framerate.den);
2078 }
2079
2080 /* Show configuration */
2081 if (global.verbose && pass == 0)
2082 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2083
2084 if (pass == (global.pass ? global.pass - 1 : 0)) {
2085 if (input.file_type == FILE_TYPE_Y4M)
2086 /*The Y4M reader does its own allocation.
2087 Just initialize this here to avoid problems if we never read any
2088 frames.*/
2089 memset(&raw, 0, sizeof(raw));
2090 else
2091 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2092
2093 FOREACH_STREAM(stream->rate_hist =
2094 init_rate_histogram(&stream->config.cfg,
2095 &global.framerate));
2096 }
2097
2098 FOREACH_STREAM(setup_pass(stream, &global, pass));
2099 FOREACH_STREAM(open_output_file(stream, &global,
2100 &input.pixel_aspect_ratio));
2101 FOREACH_STREAM(initialize_encoder(stream, &global));
2102
2103 #if CONFIG_VP9_HIGHBITDEPTH
2104 if (strcmp(global.codec->name, "vp9") == 0 ||
2105 strcmp(global.codec->name, "vp10") == 0) {
2106 // Check to see if at least one stream uses 16 bit internal.
2107 // Currently assume that the bit_depths for all streams using
2108 // highbitdepth are the same.
2109 FOREACH_STREAM({
2110 if (stream->config.use_16bit_internal) {
2111 use_16bit_internal = 1;
2112 }
2113 if (stream->config.cfg.g_profile == 0) {
2114 input_shift = 0;
2115 } else {
2116 input_shift = (int)stream->config.cfg.g_bit_depth -
2117 stream->config.cfg.g_input_bit_depth;
2118 }
2119 });
2120 }
2121 #endif
2122
2123 frame_avail = 1;
2124 got_data = 0;
2125
2126 while (frame_avail || got_data) {
2127 struct vpx_usec_timer timer;
2128
2129 if (!global.limit || frames_in < global.limit) {
2130 frame_avail = read_frame(&input, &raw);
2131
2132 if (frame_avail)
2133 frames_in++;
2134 seen_frames = frames_in > global.skip_frames ?
2135 frames_in - global.skip_frames : 0;
2136
2137 if (!global.quiet) {
2138 float fps = usec_to_fps(cx_time, seen_frames);
2139 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2140
2141 if (stream_cnt == 1)
2142 fprintf(stderr,
2143 "frame %4d/%-4d %7"PRId64"B ",
2144 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2145 else
2146 fprintf(stderr, "frame %4d ", frames_in);
2147
2148 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2149 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2150 cx_time > 9999999 ? "ms" : "us",
2151 fps >= 1.0 ? fps : fps * 60,
2152 fps >= 1.0 ? "fps" : "fpm");
2153 print_time("ETA", estimated_time_left);
2154 }
2155
2156 } else
2157 frame_avail = 0;
2158
2159 if (frames_in > global.skip_frames) {
2160 #if CONFIG_VP9_HIGHBITDEPTH
2161 vpx_image_t *frame_to_encode;
2162 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2163 assert(use_16bit_internal);
2164 // Input bit depth and stream bit depth do not match, so up
2165 // shift frame to stream bit depth
2166 if (!allocated_raw_shift) {
2167 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2168 input.width, input.height, 32);
2169 allocated_raw_shift = 1;
2170 }
2171 vpx_img_upshift(&raw_shift, &raw, input_shift);
2172 frame_to_encode = &raw_shift;
2173 } else {
2174 frame_to_encode = &raw;
2175 }
2176 vpx_usec_timer_start(&timer);
2177 if (use_16bit_internal) {
2178 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2179 FOREACH_STREAM({
2180 if (stream->config.use_16bit_internal)
2181 encode_frame(stream, &global,
2182 frame_avail ? frame_to_encode : NULL,
2183 frames_in);
2184 else
2185 assert(0);
2186 });
2187 } else {
2188 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2189 FOREACH_STREAM(encode_frame(stream, &global,
2190 frame_avail ? frame_to_encode : NULL,
2191 frames_in));
2192 }
2193 #else
2194 vpx_usec_timer_start(&timer);
2195 FOREACH_STREAM(encode_frame(stream, &global,
2196 frame_avail ? &raw : NULL,
2197 frames_in));
2198 #endif
2199 vpx_usec_timer_mark(&timer);
2200 cx_time += vpx_usec_timer_elapsed(&timer);
2201
2202 FOREACH_STREAM(update_quantizer_histogram(stream));
2203
2204 got_data = 0;
2205 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2206
2207 if (!got_data && input.length && streams != NULL &&
2208 !streams->frames_out) {
2209 lagged_count = global.limit ? seen_frames : ftello(input.file);
2210 } else if (input.length) {
2211 int64_t remaining;
2212 int64_t rate;
2213
2214 if (global.limit) {
2215 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2216
2217 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2218 remaining = 1000 * (global.limit - global.skip_frames
2219 - seen_frames + lagged_count);
2220 } else {
2221 const int64_t input_pos = ftello(input.file);
2222 const int64_t input_pos_lagged = input_pos - lagged_count;
2223 const int64_t limit = input.length;
2224
2225 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2226 remaining = limit - input_pos + lagged_count;
2227 }
2228
2229 average_rate = (average_rate <= 0)
2230 ? rate
2231 : (average_rate * 7 + rate) / 8;
2232 estimated_time_left = average_rate ? remaining / average_rate : -1;
2233 }
2234
2235 if (got_data && global.test_decode != TEST_DECODE_OFF)
2236 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2237 }
2238
2239 fflush(stdout);
2240 if (!global.quiet)
2241 fprintf(stderr, "\033[K");
2242 }
2243
2244 if (stream_cnt > 1)
2245 fprintf(stderr, "\n");
2246
2247 if (!global.quiet) {
2248 FOREACH_STREAM(fprintf(stderr,
2249 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2250 " %7"PRId64" %s (%.2f fps)\033[K\n",
2251 pass + 1,
2252 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2253 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2254 seen_frames ? (int64_t)stream->nbytes * 8 *
2255 (int64_t)global.framerate.num / global.framerate.den /
2256 seen_frames : 0,
2257 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2258 stream->cx_time > 9999999 ? "ms" : "us",
2259 usec_to_fps(stream->cx_time, seen_frames)));
2260 }
2261
2262 if (global.show_psnr) {
2263 if (global.codec->fourcc == VP9_FOURCC) {
2264 FOREACH_STREAM(
2265 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2266 } else {
2267 FOREACH_STREAM(show_psnr(stream, 255.0));
2268 }
2269 }
2270
2271 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2272
2273 if (global.test_decode != TEST_DECODE_OFF) {
2274 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2275 }
2276
2277 close_input_file(&input);
2278
2279 if (global.test_decode == TEST_DECODE_FATAL) {
2280 FOREACH_STREAM(res |= stream->mismatch_seen);
2281 }
2282 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2283
2284 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2285
2286 #if CONFIG_FP_MB_STATS
2287 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2288 #endif
2289
2290 if (global.pass)
2291 break;
2292 }
2293
2294 if (global.show_q_hist_buckets)
2295 FOREACH_STREAM(show_q_histogram(stream->counts,
2296 global.show_q_hist_buckets));
2297
2298 if (global.show_rate_hist_buckets)
2299 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2300 &stream->config.cfg,
2301 global.show_rate_hist_buckets));
2302 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2303
2304 #if CONFIG_INTERNAL_STATS
2305 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2306 * to match some existing utilities.
2307 */
2308 if (!(global.pass == 1 && global.passes == 2))
2309 FOREACH_STREAM({
2310 FILE *f = fopen("opsnr.stt", "a");
2311 if (stream->mismatch_seen) {
2312 fprintf(f, "First mismatch occurred in frame %d\n",
2313 stream->mismatch_seen);
2314 } else {
2315 fprintf(f, "No mismatch detected in recon buffers\n");
2316 }
2317 fclose(f);
2318 });
2319 #endif
2320
2321 #if CONFIG_VP9_HIGHBITDEPTH
2322 if (allocated_raw_shift)
2323 vpx_img_free(&raw_shift);
2324 #endif
2325 vpx_img_free(&raw);
2326 free(argv);
2327 free(streams);
2328 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2329 }
AV1 Codec Library