?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge "VP9: Do not set up/left available in vp9_first_pass()"
[aom.git] / vpxenc.c
blob8798e69184a612cea1131a411541ba3253d9c8d2
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, "Number of tile rows to use, log2");
384 static const arg_def_t lossless = ARG_DEF(
385 NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
386 static const arg_def_t frame_parallel_decoding = ARG_DEF(
387 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
388 static const arg_def_t aq_mode = ARG_DEF(
389 NULL, "aq-mode", 1,
390 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
391 "3: cyclic refresh, 4: equator360)");
392 static const arg_def_t frame_periodic_boost = ARG_DEF(
393 NULL, "frame-boost", 1,
394 "Enable frame periodic boost (0: off (default), 1: on)");
395 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
396 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
397 static const arg_def_t max_inter_rate_pct = ARG_DEF(
398 NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
399 static const arg_def_t min_gf_interval = ARG_DEF(
400 NULL, "min-gf-interval", 1,
401 "min gf/arf frame interval (default 0, indicating in-built behavior)");
402 static const arg_def_t max_gf_interval = ARG_DEF(
403 NULL, "max-gf-interval", 1,
404 "max gf/arf frame interval (default 0, indicating in-built behavior)");
405
406 static const struct arg_enum_list color_space_enum[] = {
407 { "unknown", VPX_CS_UNKNOWN },
408 { "bt601", VPX_CS_BT_601 },
409 { "bt709", VPX_CS_BT_709 },
410 { "smpte170", VPX_CS_SMPTE_170 },
411 { "smpte240", VPX_CS_SMPTE_240 },
412 { "bt2020", VPX_CS_BT_2020 },
413 { "reserved", VPX_CS_RESERVED },
414 { "sRGB", VPX_CS_SRGB },
415 { NULL, 0 }
416 };
417
418 static const arg_def_t input_color_space = ARG_DEF_ENUM(
419 NULL, "color-space", 1,
420 "The color space of input content:", color_space_enum);
421
422 #if CONFIG_VP9_HIGHBITDEPTH
423 static const struct arg_enum_list bitdepth_enum[] = {
424 {"8", VPX_BITS_8},
425 {"10", VPX_BITS_10},
426 {"12", VPX_BITS_12},
427 {NULL, 0}
428 };
429
430 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
431 "b", "bit-depth", 1,
432 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
433 bitdepth_enum);
434 static const arg_def_t inbitdeptharg = ARG_DEF(
435 NULL, "input-bit-depth", 1, "Bit depth of input");
436 #endif
437
438 static const struct arg_enum_list tune_content_enum[] = {
439 {"default", VP9E_CONTENT_DEFAULT},
440 {"screen", VP9E_CONTENT_SCREEN},
441 {NULL, 0}
442 };
443
444 static const arg_def_t tune_content = ARG_DEF_ENUM(
445 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
446 #endif
447
448 #if CONFIG_VP9_ENCODER
449 static const arg_def_t *vp9_args[] = {
450 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
451 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
452 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
453 &gf_cbr_boost_pct, &lossless,
454 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
455 &noise_sens, &tune_content, &input_color_space,
456 &min_gf_interval, &max_gf_interval,
457 NULL
458 };
459 static const int vp9_arg_ctrl_map[] = {
460 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
461 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
462 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
463 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
464 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
465 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
466 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
467 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
468 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
469 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL,
470 0
471 };
472 #endif
473
474 #if CONFIG_VP10_ENCODER
475 static const arg_def_t *vp10_args[] = {
476 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
477 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
478 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
479 &gf_cbr_boost_pct, &lossless,
480 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
481 &noise_sens, &tune_content, &input_color_space,
482 &min_gf_interval, &max_gf_interval,
483 NULL
484 };
485 static const int vp10_arg_ctrl_map[] = {
486 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
487 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
488 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
489 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
490 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
491 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
492 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
493 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
494 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
495 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL,
496 0
497 };
498 #endif
499
500 static const arg_def_t *no_args[] = { NULL };
501
502 void usage_exit(void) {
503 int i;
504 const int num_encoder = get_vpx_encoder_count();
505
506 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
507 exec_name);
508
509 fprintf(stderr, "\nOptions:\n");
510 arg_show_usage(stderr, main_args);
511 fprintf(stderr, "\nEncoder Global Options:\n");
512 arg_show_usage(stderr, global_args);
513 fprintf(stderr, "\nRate Control Options:\n");
514 arg_show_usage(stderr, rc_args);
515 fprintf(stderr, "\nTwopass Rate Control Options:\n");
516 arg_show_usage(stderr, rc_twopass_args);
517 fprintf(stderr, "\nKeyframe Placement Options:\n");
518 arg_show_usage(stderr, kf_args);
519 #if CONFIG_VP8_ENCODER
520 fprintf(stderr, "\nVP8 Specific Options:\n");
521 arg_show_usage(stderr, vp8_args);
522 #endif
523 #if CONFIG_VP9_ENCODER
524 fprintf(stderr, "\nVP9 Specific Options:\n");
525 arg_show_usage(stderr, vp9_args);
526 #endif
527 #if CONFIG_VP10_ENCODER
528 fprintf(stderr, "\nVP10 Specific Options:\n");
529 arg_show_usage(stderr, vp10_args);
530 #endif
531 fprintf(stderr, "\nStream timebase (--timebase):\n"
532 " The desired precision of timestamps in the output, expressed\n"
533 " in fractional seconds. Default is 1/1000.\n");
534 fprintf(stderr, "\nIncluded encoders:\n\n");
535
536 for (i = 0; i < num_encoder; ++i) {
537 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
538 const char* defstr = (i == (num_encoder - 1)) ? "(default)" : "";
539 fprintf(stderr, " %-6s - %s %s\n",
540 encoder->name, vpx_codec_iface_name(encoder->codec_interface()),
541 defstr);
542 }
543 fprintf(stderr, "\n ");
544 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
545
546 exit(EXIT_FAILURE);
547 }
548
549 #define mmin(a, b) ((a) < (b) ? (a) : (b))
550
551 #if CONFIG_VP9_HIGHBITDEPTH
552 static void find_mismatch_high(const vpx_image_t *const img1,
553 const vpx_image_t *const img2,
554 int yloc[4], int uloc[4], int vloc[4]) {
555 uint16_t *plane1, *plane2;
556 uint32_t stride1, stride2;
557 const uint32_t bsize = 64;
558 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
559 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
560 const uint32_t c_w =
561 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
562 const uint32_t c_h =
563 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
564 int match = 1;
565 uint32_t i, j;
566 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
567 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
568 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
569 stride1 = img1->stride[VPX_PLANE_Y]/2;
570 stride2 = img2->stride[VPX_PLANE_Y]/2;
571 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
572 for (j = 0; match && j < img1->d_w; j += bsize) {
573 int k, l;
574 const int si = mmin(i + bsize, img1->d_h) - i;
575 const int sj = mmin(j + bsize, img1->d_w) - j;
576 for (k = 0; match && k < si; ++k) {
577 for (l = 0; match && l < sj; ++l) {
578 if (*(plane1 + (i + k) * stride1 + j + l) !=
579 *(plane2 + (i + k) * stride2 + j + l)) {
580 yloc[0] = i + k;
581 yloc[1] = j + l;
582 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
583 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
584 match = 0;
585 break;
586 }
587 }
588 }
589 }
590 }
591
592 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
593 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
594 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
595 stride1 = img1->stride[VPX_PLANE_U]/2;
596 stride2 = img2->stride[VPX_PLANE_U]/2;
597 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
598 for (j = 0; match && j < c_w; j += bsizex) {
599 int k, l;
600 const int si = mmin(i + bsizey, c_h - i);
601 const int sj = mmin(j + bsizex, c_w - j);
602 for (k = 0; match && k < si; ++k) {
603 for (l = 0; match && l < sj; ++l) {
604 if (*(plane1 + (i + k) * stride1 + j + l) !=
605 *(plane2 + (i + k) * stride2 + j + l)) {
606 uloc[0] = i + k;
607 uloc[1] = j + l;
608 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
609 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
610 match = 0;
611 break;
612 }
613 }
614 }
615 }
616 }
617
618 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
619 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
620 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
621 stride1 = img1->stride[VPX_PLANE_V]/2;
622 stride2 = img2->stride[VPX_PLANE_V]/2;
623 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
624 for (j = 0; match && j < c_w; j += bsizex) {
625 int k, l;
626 const int si = mmin(i + bsizey, c_h - i);
627 const int sj = mmin(j + bsizex, c_w - j);
628 for (k = 0; match && k < si; ++k) {
629 for (l = 0; match && l < sj; ++l) {
630 if (*(plane1 + (i + k) * stride1 + j + l) !=
631 *(plane2 + (i + k) * stride2 + j + l)) {
632 vloc[0] = i + k;
633 vloc[1] = j + l;
634 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
635 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
636 match = 0;
637 break;
638 }
639 }
640 }
641 }
642 }
643 }
644 #endif
645
646 static void find_mismatch(const vpx_image_t *const img1,
647 const vpx_image_t *const img2,
648 int yloc[4], int uloc[4], int vloc[4]) {
649 const uint32_t bsize = 64;
650 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
651 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
652 const uint32_t c_w =
653 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
654 const uint32_t c_h =
655 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
656 int match = 1;
657 uint32_t i, j;
658 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
659 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
660 for (j = 0; match && j < img1->d_w; j += bsize) {
661 int k, l;
662 const int si = mmin(i + bsize, img1->d_h) - i;
663 const int sj = mmin(j + bsize, img1->d_w) - j;
664 for (k = 0; match && k < si; ++k) {
665 for (l = 0; match && l < sj; ++l) {
666 if (*(img1->planes[VPX_PLANE_Y] +
667 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
668 *(img2->planes[VPX_PLANE_Y] +
669 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
670 yloc[0] = i + k;
671 yloc[1] = j + l;
672 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
673 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
674 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
675 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
676 match = 0;
677 break;
678 }
679 }
680 }
681 }
682 }
683
684 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
685 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
686 for (j = 0; match && j < c_w; j += bsizex) {
687 int k, l;
688 const int si = mmin(i + bsizey, c_h - i);
689 const int sj = mmin(j + bsizex, c_w - j);
690 for (k = 0; match && k < si; ++k) {
691 for (l = 0; match && l < sj; ++l) {
692 if (*(img1->planes[VPX_PLANE_U] +
693 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
694 *(img2->planes[VPX_PLANE_U] +
695 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
696 uloc[0] = i + k;
697 uloc[1] = j + l;
698 uloc[2] = *(img1->planes[VPX_PLANE_U] +
699 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
700 uloc[3] = *(img2->planes[VPX_PLANE_U] +
701 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
702 match = 0;
703 break;
704 }
705 }
706 }
707 }
708 }
709 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
710 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
711 for (j = 0; match && j < c_w; j += bsizex) {
712 int k, l;
713 const int si = mmin(i + bsizey, c_h - i);
714 const int sj = mmin(j + bsizex, c_w - j);
715 for (k = 0; match && k < si; ++k) {
716 for (l = 0; match && l < sj; ++l) {
717 if (*(img1->planes[VPX_PLANE_V] +
718 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
719 *(img2->planes[VPX_PLANE_V] +
720 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
721 vloc[0] = i + k;
722 vloc[1] = j + l;
723 vloc[2] = *(img1->planes[VPX_PLANE_V] +
724 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
725 vloc[3] = *(img2->planes[VPX_PLANE_V] +
726 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
727 match = 0;
728 break;
729 }
730 }
731 }
732 }
733 }
734 }
735
736 static int compare_img(const vpx_image_t *const img1,
737 const vpx_image_t *const img2) {
738 uint32_t l_w = img1->d_w;
739 uint32_t c_w =
740 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
741 const uint32_t c_h =
742 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
743 uint32_t i;
744 int match = 1;
745
746 match &= (img1->fmt == img2->fmt);
747 match &= (img1->d_w == img2->d_w);
748 match &= (img1->d_h == img2->d_h);
749 #if CONFIG_VP9_HIGHBITDEPTH
750 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
751 l_w *= 2;
752 c_w *= 2;
753 }
754 #endif
755
756 for (i = 0; i < img1->d_h; ++i)
757 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
758 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
759 l_w) == 0);
760
761 for (i = 0; i < c_h; ++i)
762 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
763 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
764 c_w) == 0);
765
766 for (i = 0; i < c_h; ++i)
767 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
768 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
769 c_w) == 0);
770
771 return match;
772 }
773
774
775 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
776 #if CONFIG_VP10_ENCODER
777 #define ARG_CTRL_CNT_MAX NELEMENTS(vp10_arg_ctrl_map)
778 #elif CONFIG_VP9_ENCODER
779 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
780 #else
781 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
782 #endif
783
784 #if !CONFIG_WEBM_IO
785 typedef int stereo_format_t;
786 struct EbmlGlobal { int debug; };
787 #endif
788
789 /* Per-stream configuration */
790 struct stream_config {
791 struct vpx_codec_enc_cfg cfg;
792 const char *out_fn;
793 const char *stats_fn;
794 #if CONFIG_FP_MB_STATS
795 const char *fpmb_stats_fn;
796 #endif
797 stereo_format_t stereo_fmt;
798 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
799 int arg_ctrl_cnt;
800 int write_webm;
801 int have_kf_max_dist;
802 #if CONFIG_VP9_HIGHBITDEPTH
803 // whether to use 16bit internal buffers
804 int use_16bit_internal;
805 #endif
806 };
807
808
809 struct stream_state {
810 int index;
811 struct stream_state *next;
812 struct stream_config config;
813 FILE *file;
814 struct rate_hist *rate_hist;
815 struct EbmlGlobal ebml;
816 uint64_t psnr_sse_total;
817 uint64_t psnr_samples_total;
818 double psnr_totals[4];
819 int psnr_count;
820 int counts[64];
821 vpx_codec_ctx_t encoder;
822 unsigned int frames_out;
823 uint64_t cx_time;
824 size_t nbytes;
825 stats_io_t stats;
826 #if CONFIG_FP_MB_STATS
827 stats_io_t fpmb_stats;
828 #endif
829 struct vpx_image *img;
830 vpx_codec_ctx_t decoder;
831 int mismatch_seen;
832 };
833
834
835 static void validate_positive_rational(const char *msg,
836 struct vpx_rational *rat) {
837 if (rat->den < 0) {
838 rat->num *= -1;
839 rat->den *= -1;
840 }
841
842 if (rat->num < 0)
843 die("Error: %s must be positive\n", msg);
844
845 if (!rat->den)
846 die("Error: %s has zero denominator\n", msg);
847 }
848
849
850 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
851 char **argi, **argj;
852 struct arg arg;
853 const int num_encoder = get_vpx_encoder_count();
854
855 if (num_encoder < 1)
856 die("Error: no valid encoder available\n");
857
858 /* Initialize default parameters */
859 memset(global, 0, sizeof(*global));
860 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
861 global->passes = 0;
862 global->color_type = I420;
863 /* Assign default deadline to good quality */
864 global->deadline = VPX_DL_GOOD_QUALITY;
865
866 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
867 arg.argv_step = 1;
868
869 if (arg_match(&arg, &codecarg, argi)) {
870 global->codec = get_vpx_encoder_by_name(arg.val);
871 if (!global->codec)
872 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
873 } else if (arg_match(&arg, &passes, argi)) {
874 global->passes = arg_parse_uint(&arg);
875
876 if (global->passes < 1 || global->passes > 2)
877 die("Error: Invalid number of passes (%d)\n", global->passes);
878 } else if (arg_match(&arg, &pass_arg, argi)) {
879 global->pass = arg_parse_uint(&arg);
880
881 if (global->pass < 1 || global->pass > 2)
882 die("Error: Invalid pass selected (%d)\n",
883 global->pass);
884 } else if (arg_match(&arg, &usage, argi))
885 global->usage = arg_parse_uint(&arg);
886 else if (arg_match(&arg, &deadline, argi))
887 global->deadline = arg_parse_uint(&arg);
888 else if (arg_match(&arg, &best_dl, argi))
889 global->deadline = VPX_DL_BEST_QUALITY;
890 else if (arg_match(&arg, &good_dl, argi))
891 global->deadline = VPX_DL_GOOD_QUALITY;
892 else if (arg_match(&arg, &rt_dl, argi))
893 global->deadline = VPX_DL_REALTIME;
894 else if (arg_match(&arg, &use_yv12, argi))
895 global->color_type = YV12;
896 else if (arg_match(&arg, &use_i420, argi))
897 global->color_type = I420;
898 else if (arg_match(&arg, &use_i422, argi))
899 global->color_type = I422;
900 else if (arg_match(&arg, &use_i444, argi))
901 global->color_type = I444;
902 else if (arg_match(&arg, &use_i440, argi))
903 global->color_type = I440;
904 else if (arg_match(&arg, &quietarg, argi))
905 global->quiet = 1;
906 else if (arg_match(&arg, &verbosearg, argi))
907 global->verbose = 1;
908 else if (arg_match(&arg, &limit, argi))
909 global->limit = arg_parse_uint(&arg);
910 else if (arg_match(&arg, &skip, argi))
911 global->skip_frames = arg_parse_uint(&arg);
912 else if (arg_match(&arg, &psnrarg, argi))
913 global->show_psnr = 1;
914 else if (arg_match(&arg, &recontest, argi))
915 global->test_decode = arg_parse_enum_or_int(&arg);
916 else if (arg_match(&arg, &framerate, argi)) {
917 global->framerate = arg_parse_rational(&arg);
918 validate_positive_rational(arg.name, &global->framerate);
919 global->have_framerate = 1;
920 } else if (arg_match(&arg, &out_part, argi))
921 global->out_part = 1;
922 else if (arg_match(&arg, &debugmode, argi))
923 global->debug = 1;
924 else if (arg_match(&arg, &q_hist_n, argi))
925 global->show_q_hist_buckets = arg_parse_uint(&arg);
926 else if (arg_match(&arg, &rate_hist_n, argi))
927 global->show_rate_hist_buckets = arg_parse_uint(&arg);
928 else if (arg_match(&arg, &disable_warnings, argi))
929 global->disable_warnings = 1;
930 else if (arg_match(&arg, &disable_warning_prompt, argi))
931 global->disable_warning_prompt = 1;
932 else
933 argj++;
934 }
935
936 if (global->pass) {
937 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
938 if (global->pass > global->passes) {
939 warn("Assuming --pass=%d implies --passes=%d\n",
940 global->pass, global->pass);
941 global->passes = global->pass;
942 }
943 }
944 /* Validate global config */
945 if (global->passes == 0) {
946 #if CONFIG_VP9_ENCODER || CONFIG_VP10_ENCODER
947 // Make default VP9 passes = 2 until there is a better quality 1-pass
948 // encoder
949 if (global->codec != NULL && global->codec->name != NULL)
950 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
951 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
952 #else
953 global->passes = 1;
954 #endif
955 }
956
957 if (global->deadline == VPX_DL_REALTIME &&
958 global->passes > 1) {
959 warn("Enforcing one-pass encoding in realtime mode\n");
960 global->passes = 1;
961 }
962 }
963
964
965 static void open_input_file(struct VpxInputContext *input) {
966 /* Parse certain options from the input file, if possible */
967 input->file = strcmp(input->filename, "-")
968 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
969
970 if (!input->file)
971 fatal("Failed to open input file");
972
973 if (!fseeko(input->file, 0, SEEK_END)) {
974 /* Input file is seekable. Figure out how long it is, so we can get
975 * progress info.
976 */
977 input->length = ftello(input->file);
978 rewind(input->file);
979 }
980
981 /* Default to 1:1 pixel aspect ratio. */
982 input->pixel_aspect_ratio.numerator = 1;
983 input->pixel_aspect_ratio.denominator = 1;
984
985 /* For RAW input sources, these bytes will applied on the first frame
986 * in read_frame().
987 */
988 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
989 input->detect.position = 0;
990
991 if (input->detect.buf_read == 4
992 && file_is_y4m(input->detect.buf)) {
993 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
994 input->only_i420) >= 0) {
995 input->file_type = FILE_TYPE_Y4M;
996 input->width = input->y4m.pic_w;
997 input->height = input->y4m.pic_h;
998 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
999 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1000 input->framerate.numerator = input->y4m.fps_n;
1001 input->framerate.denominator = input->y4m.fps_d;
1002 input->fmt = input->y4m.vpx_fmt;
1003 input->bit_depth = input->y4m.bit_depth;
1004 } else
1005 fatal("Unsupported Y4M stream.");
1006 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1007 fatal("IVF is not supported as input.");
1008 } else {
1009 input->file_type = FILE_TYPE_RAW;
1010 }
1011 }
1012
1013
1014 static void close_input_file(struct VpxInputContext *input) {
1015 fclose(input->file);
1016 if (input->file_type == FILE_TYPE_Y4M)
1017 y4m_input_close(&input->y4m);
1018 }
1019
1020 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1021 struct stream_state *prev) {
1022 struct stream_state *stream;
1023
1024 stream = calloc(1, sizeof(*stream));
1025 if (stream == NULL) {
1026 fatal("Failed to allocate new stream.");
1027 }
1028
1029 if (prev) {
1030 memcpy(stream, prev, sizeof(*stream));
1031 stream->index++;
1032 prev->next = stream;
1033 } else {
1034 vpx_codec_err_t res;
1035
1036 /* Populate encoder configuration */
1037 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1038 &stream->config.cfg,
1039 global->usage);
1040 if (res)
1041 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1042
1043 /* Change the default timebase to a high enough value so that the
1044 * encoder will always create strictly increasing timestamps.
1045 */
1046 stream->config.cfg.g_timebase.den = 1000;
1047
1048 /* Never use the library's default resolution, require it be parsed
1049 * from the file or set on the command line.
1050 */
1051 stream->config.cfg.g_w = 0;
1052 stream->config.cfg.g_h = 0;
1053
1054 /* Initialize remaining stream parameters */
1055 stream->config.write_webm = 1;
1056 #if CONFIG_WEBM_IO
1057 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1058 stream->ebml.last_pts_ns = -1;
1059 stream->ebml.writer = NULL;
1060 stream->ebml.segment = NULL;
1061 #endif
1062
1063 /* Allows removal of the application version from the EBML tags */
1064 stream->ebml.debug = global->debug;
1065
1066 /* Default lag_in_frames is 0 in realtime mode */
1067 if (global->deadline == VPX_DL_REALTIME)
1068 stream->config.cfg.g_lag_in_frames = 0;
1069 }
1070
1071 /* Output files must be specified for each stream */
1072 stream->config.out_fn = NULL;
1073
1074 stream->next = NULL;
1075 return stream;
1076 }
1077
1078
1079 static int parse_stream_params(struct VpxEncoderConfig *global,
1080 struct stream_state *stream,
1081 char **argv) {
1082 char **argi, **argj;
1083 struct arg arg;
1084 static const arg_def_t **ctrl_args = no_args;
1085 static const int *ctrl_args_map = NULL;
1086 struct stream_config *config = &stream->config;
1087 int eos_mark_found = 0;
1088 #if CONFIG_VP9_HIGHBITDEPTH
1089 int test_16bit_internal = 0;
1090 #endif
1091
1092 // Handle codec specific options
1093 if (0) {
1094 #if CONFIG_VP8_ENCODER
1095 } else if (strcmp(global->codec->name, "vp8") == 0) {
1096 ctrl_args = vp8_args;
1097 ctrl_args_map = vp8_arg_ctrl_map;
1098 #endif
1099 #if CONFIG_VP9_ENCODER
1100 } else if (strcmp(global->codec->name, "vp9") == 0) {
1101 ctrl_args = vp9_args;
1102 ctrl_args_map = vp9_arg_ctrl_map;
1103 #endif
1104 #if CONFIG_VP10_ENCODER
1105 } else if (strcmp(global->codec->name, "vp10") == 0) {
1106 // TODO(jingning): Reuse VP9 specific encoder configuration parameters.
1107 // Consider to expand this set for VP10 encoder control.
1108 ctrl_args = vp10_args;
1109 ctrl_args_map = vp10_arg_ctrl_map;
1110 #endif
1111 }
1112
1113 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1114 arg.argv_step = 1;
1115
1116 /* Once we've found an end-of-stream marker (--) we want to continue
1117 * shifting arguments but not consuming them.
1118 */
1119 if (eos_mark_found) {
1120 argj++;
1121 continue;
1122 } else if (!strcmp(*argj, "--")) {
1123 eos_mark_found = 1;
1124 continue;
1125 }
1126
1127 if (arg_match(&arg, &outputfile, argi)) {
1128 config->out_fn = arg.val;
1129 } else if (arg_match(&arg, &fpf_name, argi)) {
1130 config->stats_fn = arg.val;
1131 #if CONFIG_FP_MB_STATS
1132 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1133 config->fpmb_stats_fn = arg.val;
1134 #endif
1135 } else if (arg_match(&arg, &use_webm, argi)) {
1136 #if CONFIG_WEBM_IO
1137 config->write_webm = 1;
1138 #else
1139 die("Error: --webm specified but webm is disabled.");
1140 #endif
1141 } else if (arg_match(&arg, &use_ivf, argi)) {
1142 config->write_webm = 0;
1143 } else if (arg_match(&arg, &threads, argi)) {
1144 config->cfg.g_threads = arg_parse_uint(&arg);
1145 } else if (arg_match(&arg, &profile, argi)) {
1146 config->cfg.g_profile = arg_parse_uint(&arg);
1147 } else if (arg_match(&arg, &width, argi)) {
1148 config->cfg.g_w = arg_parse_uint(&arg);
1149 } else if (arg_match(&arg, &height, argi)) {
1150 config->cfg.g_h = arg_parse_uint(&arg);
1151 #if CONFIG_VP9_HIGHBITDEPTH
1152 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1153 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1154 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1155 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1156 #endif
1157 #if CONFIG_WEBM_IO
1158 } else if (arg_match(&arg, &stereo_mode, argi)) {
1159 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1160 #endif
1161 } else if (arg_match(&arg, &timebase, argi)) {
1162 config->cfg.g_timebase = arg_parse_rational(&arg);
1163 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1164 } else if (arg_match(&arg, &error_resilient, argi)) {
1165 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1166 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1167 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1168 if (global->deadline == VPX_DL_REALTIME &&
1169 config->cfg.g_lag_in_frames != 0) {
1170 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1171 config->cfg.g_lag_in_frames = 0;
1172 }
1173 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1174 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1175 } else if (arg_match(&arg, &resize_allowed, argi)) {
1176 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1177 } else if (arg_match(&arg, &resize_width, argi)) {
1178 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1179 } else if (arg_match(&arg, &resize_height, argi)) {
1180 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1181 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1182 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1183 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1184 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1185 } else if (arg_match(&arg, &end_usage, argi)) {
1186 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1187 } else if (arg_match(&arg, &target_bitrate, argi)) {
1188 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1189 } else if (arg_match(&arg, &min_quantizer, argi)) {
1190 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1191 } else if (arg_match(&arg, &max_quantizer, argi)) {
1192 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1193 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1194 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1195 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1196 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1197 } else if (arg_match(&arg, &buf_sz, argi)) {
1198 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1199 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1200 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1201 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1202 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1203 } else if (arg_match(&arg, &bias_pct, argi)) {
1204 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1205 if (global->passes < 2)
1206 warn("option %s ignored in one-pass mode.\n", arg.name);
1207 } else if (arg_match(&arg, &minsection_pct, argi)) {
1208 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1209
1210 if (global->passes < 2)
1211 warn("option %s ignored in one-pass mode.\n", arg.name);
1212 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1213 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1214
1215 if (global->passes < 2)
1216 warn("option %s ignored in one-pass mode.\n", arg.name);
1217 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1218 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1219 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1220 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1221 config->have_kf_max_dist = 1;
1222 } else if (arg_match(&arg, &kf_disabled, argi)) {
1223 config->cfg.kf_mode = VPX_KF_DISABLED;
1224 #if CONFIG_VP9_HIGHBITDEPTH
1225 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1226 if (strcmp(global->codec->name, "vp9") == 0 ||
1227 strcmp(global->codec->name, "vp10") == 0) {
1228 test_16bit_internal = 1;
1229 }
1230 #endif
1231 } else {
1232 int i, match = 0;
1233 for (i = 0; ctrl_args[i]; i++) {
1234 if (arg_match(&arg, ctrl_args[i], argi)) {
1235 int j;
1236 match = 1;
1237
1238 /* Point either to the next free element or the first
1239 * instance of this control.
1240 */
1241 for (j = 0; j < config->arg_ctrl_cnt; j++)
1242 if (ctrl_args_map != NULL &&
1243 config->arg_ctrls[j][0] == ctrl_args_map[i])
1244 break;
1245
1246 /* Update/insert */
1247 assert(j < (int)ARG_CTRL_CNT_MAX);
1248 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1249 config->arg_ctrls[j][0] = ctrl_args_map[i];
1250 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1251 if (j == config->arg_ctrl_cnt)
1252 config->arg_ctrl_cnt++;
1253 }
1254 }
1255 }
1256 if (!match)
1257 argj++;
1258 }
1259 }
1260 #if CONFIG_VP9_HIGHBITDEPTH
1261 if (strcmp(global->codec->name, "vp9") == 0 ||
1262 strcmp(global->codec->name, "vp10") == 0) {
1263 config->use_16bit_internal = test_16bit_internal |
1264 (config->cfg.g_profile > 1);
1265 }
1266 #endif
1267 return eos_mark_found;
1268 }
1269
1270
1271 #define FOREACH_STREAM(func) \
1272 do { \
1273 struct stream_state *stream; \
1274 for (stream = streams; stream; stream = stream->next) { \
1275 func; \
1276 } \
1277 } while (0)
1278
1279
1280 static void validate_stream_config(const struct stream_state *stream,
1281 const struct VpxEncoderConfig *global) {
1282 const struct stream_state *streami;
1283 (void)global;
1284
1285 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1286 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1287 " and --height (-h)", stream->index);
1288
1289 // Check that the codec bit depth is greater than the input bit depth.
1290 if (stream->config.cfg.g_input_bit_depth >
1291 (unsigned int)stream->config.cfg.g_bit_depth) {
1292 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1293 stream->index, (int)stream->config.cfg.g_bit_depth,
1294 stream->config.cfg.g_input_bit_depth);
1295 }
1296
1297 for (streami = stream; streami; streami = streami->next) {
1298 /* All streams require output files */
1299 if (!streami->config.out_fn)
1300 fatal("Stream %d: Output file is required (specify with -o)",
1301 streami->index);
1302
1303 /* Check for two streams outputting to the same file */
1304 if (streami != stream) {
1305 const char *a = stream->config.out_fn;
1306 const char *b = streami->config.out_fn;
1307 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1308 fatal("Stream %d: duplicate output file (from stream %d)",
1309 streami->index, stream->index);
1310 }
1311
1312 /* Check for two streams sharing a stats file. */
1313 if (streami != stream) {
1314 const char *a = stream->config.stats_fn;
1315 const char *b = streami->config.stats_fn;
1316 if (a && b && !strcmp(a, b))
1317 fatal("Stream %d: duplicate stats file (from stream %d)",
1318 streami->index, stream->index);
1319 }
1320
1321 #if CONFIG_FP_MB_STATS
1322 /* Check for two streams sharing a mb stats file. */
1323 if (streami != stream) {
1324 const char *a = stream->config.fpmb_stats_fn;
1325 const char *b = streami->config.fpmb_stats_fn;
1326 if (a && b && !strcmp(a, b))
1327 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1328 streami->index, stream->index);
1329 }
1330 #endif
1331 }
1332 }
1333
1334
1335 static void set_stream_dimensions(struct stream_state *stream,
1336 unsigned int w,
1337 unsigned int h) {
1338 if (!stream->config.cfg.g_w) {
1339 if (!stream->config.cfg.g_h)
1340 stream->config.cfg.g_w = w;
1341 else
1342 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1343 }
1344 if (!stream->config.cfg.g_h) {
1345 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1346 }
1347 }
1348
1349
1350 static void set_default_kf_interval(struct stream_state *stream,
1351 struct VpxEncoderConfig *global) {
1352 /* Use a max keyframe interval of 5 seconds, if none was
1353 * specified on the command line.
1354 */
1355 if (!stream->config.have_kf_max_dist) {
1356 double framerate = (double)global->framerate.num / global->framerate.den;
1357 if (framerate > 0.0)
1358 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1359 }
1360 }
1361
1362 static const char* file_type_to_string(enum VideoFileType t) {
1363 switch (t) {
1364 case FILE_TYPE_RAW: return "RAW";
1365 case FILE_TYPE_Y4M: return "Y4M";
1366 default: return "Other";
1367 }
1368 }
1369
1370 static const char* image_format_to_string(vpx_img_fmt_t f) {
1371 switch (f) {
1372 case VPX_IMG_FMT_I420: return "I420";
1373 case VPX_IMG_FMT_I422: return "I422";
1374 case VPX_IMG_FMT_I444: return "I444";
1375 case VPX_IMG_FMT_I440: return "I440";
1376 case VPX_IMG_FMT_YV12: return "YV12";
1377 case VPX_IMG_FMT_I42016: return "I42016";
1378 case VPX_IMG_FMT_I42216: return "I42216";
1379 case VPX_IMG_FMT_I44416: return "I44416";
1380 case VPX_IMG_FMT_I44016: return "I44016";
1381 default: return "Other";
1382 }
1383 }
1384
1385 static void show_stream_config(struct stream_state *stream,
1386 struct VpxEncoderConfig *global,
1387 struct VpxInputContext *input) {
1388
1389 #define SHOW(field) \
1390 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1391
1392 if (stream->index == 0) {
1393 fprintf(stderr, "Codec: %s\n",
1394 vpx_codec_iface_name(global->codec->codec_interface()));
1395 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1396 input->filename,
1397 file_type_to_string(input->file_type),
1398 image_format_to_string(input->fmt));
1399 }
1400 if (stream->next || stream->index)
1401 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1402 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1403 fprintf(stderr, "Encoder parameters:\n");
1404
1405 SHOW(g_usage);
1406 SHOW(g_threads);
1407 SHOW(g_profile);
1408 SHOW(g_w);
1409 SHOW(g_h);
1410 SHOW(g_bit_depth);
1411 SHOW(g_input_bit_depth);
1412 SHOW(g_timebase.num);
1413 SHOW(g_timebase.den);
1414 SHOW(g_error_resilient);
1415 SHOW(g_pass);
1416 SHOW(g_lag_in_frames);
1417 SHOW(rc_dropframe_thresh);
1418 SHOW(rc_resize_allowed);
1419 SHOW(rc_scaled_width);
1420 SHOW(rc_scaled_height);
1421 SHOW(rc_resize_up_thresh);
1422 SHOW(rc_resize_down_thresh);
1423 SHOW(rc_end_usage);
1424 SHOW(rc_target_bitrate);
1425 SHOW(rc_min_quantizer);
1426 SHOW(rc_max_quantizer);
1427 SHOW(rc_undershoot_pct);
1428 SHOW(rc_overshoot_pct);
1429 SHOW(rc_buf_sz);
1430 SHOW(rc_buf_initial_sz);
1431 SHOW(rc_buf_optimal_sz);
1432 SHOW(rc_2pass_vbr_bias_pct);
1433 SHOW(rc_2pass_vbr_minsection_pct);
1434 SHOW(rc_2pass_vbr_maxsection_pct);
1435 SHOW(kf_mode);
1436 SHOW(kf_min_dist);
1437 SHOW(kf_max_dist);
1438 }
1439
1440
1441 static void open_output_file(struct stream_state *stream,
1442 struct VpxEncoderConfig *global,
1443 const struct VpxRational *pixel_aspect_ratio) {
1444 const char *fn = stream->config.out_fn;
1445 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1446
1447 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1448 return;
1449
1450 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1451
1452 if (!stream->file)
1453 fatal("Failed to open output file");
1454
1455 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1456 fatal("WebM output to pipes not supported.");
1457
1458 #if CONFIG_WEBM_IO
1459 if (stream->config.write_webm) {
1460 stream->ebml.stream = stream->file;
1461 write_webm_file_header(&stream->ebml, cfg,
1462 &global->framerate,
1463 stream->config.stereo_fmt,
1464 global->codec->fourcc,
1465 pixel_aspect_ratio);
1466 }
1467 #endif
1468
1469 if (!stream->config.write_webm) {
1470 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1471 }
1472 }
1473
1474
1475 static void close_output_file(struct stream_state *stream,
1476 unsigned int fourcc) {
1477 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1478
1479 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1480 return;
1481
1482 #if CONFIG_WEBM_IO
1483 if (stream->config.write_webm) {
1484 write_webm_file_footer(&stream->ebml);
1485 }
1486 #endif
1487
1488 if (!stream->config.write_webm) {
1489 if (!fseek(stream->file, 0, SEEK_SET))
1490 ivf_write_file_header(stream->file, &stream->config.cfg,
1491 fourcc,
1492 stream->frames_out);
1493 }
1494
1495 fclose(stream->file);
1496 }
1497
1498
1499 static void setup_pass(struct stream_state *stream,
1500 struct VpxEncoderConfig *global,
1501 int pass) {
1502 if (stream->config.stats_fn) {
1503 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1504 pass))
1505 fatal("Failed to open statistics store");
1506 } else {
1507 if (!stats_open_mem(&stream->stats, pass))
1508 fatal("Failed to open statistics store");
1509 }
1510
1511 #if CONFIG_FP_MB_STATS
1512 if (stream->config.fpmb_stats_fn) {
1513 if (!stats_open_file(&stream->fpmb_stats,
1514 stream->config.fpmb_stats_fn, pass))
1515 fatal("Failed to open mb statistics store");
1516 } else {
1517 if (!stats_open_mem(&stream->fpmb_stats, pass))
1518 fatal("Failed to open mb statistics store");
1519 }
1520 #endif
1521
1522 stream->config.cfg.g_pass = global->passes == 2
1523 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1524 : VPX_RC_ONE_PASS;
1525 if (pass) {
1526 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1527 #if CONFIG_FP_MB_STATS
1528 stream->config.cfg.rc_firstpass_mb_stats_in =
1529 stats_get(&stream->fpmb_stats);
1530 #endif
1531 }
1532
1533 stream->cx_time = 0;
1534 stream->nbytes = 0;
1535 stream->frames_out = 0;
1536 }
1537
1538
1539 static void initialize_encoder(struct stream_state *stream,
1540 struct VpxEncoderConfig *global) {
1541 int i;
1542 int flags = 0;
1543
1544 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1545 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1546 #if CONFIG_VP9_HIGHBITDEPTH
1547 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1548 #endif
1549
1550 /* Construct Encoder Context */
1551 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1552 &stream->config.cfg, flags);
1553 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1554
1555 /* Note that we bypass the vpx_codec_control wrapper macro because
1556 * we're being clever to store the control IDs in an array. Real
1557 * applications will want to make use of the enumerations directly
1558 */
1559 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1560 int ctrl = stream->config.arg_ctrls[i][0];
1561 int value = stream->config.arg_ctrls[i][1];
1562 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1563 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1564 ctrl, value);
1565
1566 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1567 }
1568
1569 #if CONFIG_DECODERS
1570 if (global->test_decode != TEST_DECODE_OFF) {
1571 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1572 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1573 }
1574 #endif
1575 }
1576
1577
1578 static void encode_frame(struct stream_state *stream,
1579 struct VpxEncoderConfig *global,
1580 struct vpx_image *img,
1581 unsigned int frames_in) {
1582 vpx_codec_pts_t frame_start, next_frame_start;
1583 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1584 struct vpx_usec_timer timer;
1585
1586 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1587 * global->framerate.den)
1588 / cfg->g_timebase.num / global->framerate.num;
1589 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1590 * global->framerate.den)
1591 / cfg->g_timebase.num / global->framerate.num;
1592
1593 /* Scale if necessary */
1594 #if CONFIG_VP9_HIGHBITDEPTH
1595 if (img) {
1596 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1597 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1598 if (img->fmt != VPX_IMG_FMT_I42016) {
1599 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1600 exit(EXIT_FAILURE);
1601 }
1602 #if CONFIG_LIBYUV
1603 if (!stream->img) {
1604 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1605 cfg->g_w, cfg->g_h, 16);
1606 }
1607 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1608 img->stride[VPX_PLANE_Y]/2,
1609 (uint16*)img->planes[VPX_PLANE_U],
1610 img->stride[VPX_PLANE_U]/2,
1611 (uint16*)img->planes[VPX_PLANE_V],
1612 img->stride[VPX_PLANE_V]/2,
1613 img->d_w, img->d_h,
1614 (uint16*)stream->img->planes[VPX_PLANE_Y],
1615 stream->img->stride[VPX_PLANE_Y]/2,
1616 (uint16*)stream->img->planes[VPX_PLANE_U],
1617 stream->img->stride[VPX_PLANE_U]/2,
1618 (uint16*)stream->img->planes[VPX_PLANE_V],
1619 stream->img->stride[VPX_PLANE_V]/2,
1620 stream->img->d_w, stream->img->d_h,
1621 kFilterBox);
1622 img = stream->img;
1623 #else
1624 stream->encoder.err = 1;
1625 ctx_exit_on_error(&stream->encoder,
1626 "Stream %d: Failed to encode frame.\n"
1627 "Scaling disabled in this configuration. \n"
1628 "To enable, configure with --enable-libyuv\n",
1629 stream->index);
1630 #endif
1631 }
1632 }
1633 #endif
1634 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1635 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1636 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1637 exit(EXIT_FAILURE);
1638 }
1639 #if CONFIG_LIBYUV
1640 if (!stream->img)
1641 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1642 cfg->g_w, cfg->g_h, 16);
1643 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1644 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1645 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1646 img->d_w, img->d_h,
1647 stream->img->planes[VPX_PLANE_Y],
1648 stream->img->stride[VPX_PLANE_Y],
1649 stream->img->planes[VPX_PLANE_U],
1650 stream->img->stride[VPX_PLANE_U],
1651 stream->img->planes[VPX_PLANE_V],
1652 stream->img->stride[VPX_PLANE_V],
1653 stream->img->d_w, stream->img->d_h,
1654 kFilterBox);
1655 img = stream->img;
1656 #else
1657 stream->encoder.err = 1;
1658 ctx_exit_on_error(&stream->encoder,
1659 "Stream %d: Failed to encode frame.\n"
1660 "Scaling disabled in this configuration. \n"
1661 "To enable, configure with --enable-libyuv\n",
1662 stream->index);
1663 #endif
1664 }
1665
1666 vpx_usec_timer_start(&timer);
1667 vpx_codec_encode(&stream->encoder, img, frame_start,
1668 (unsigned long)(next_frame_start - frame_start),
1669 0, global->deadline);
1670 vpx_usec_timer_mark(&timer);
1671 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1672 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1673 stream->index);
1674 }
1675
1676
1677 static void update_quantizer_histogram(struct stream_state *stream) {
1678 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1679 int q;
1680
1681 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1682 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1683 stream->counts[q]++;
1684 }
1685 }
1686
1687
1688 static void get_cx_data(struct stream_state *stream,
1689 struct VpxEncoderConfig *global,
1690 int *got_data) {
1691 const vpx_codec_cx_pkt_t *pkt;
1692 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1693 vpx_codec_iter_t iter = NULL;
1694
1695 *got_data = 0;
1696 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1697 static size_t fsize = 0;
1698 static int64_t ivf_header_pos = 0;
1699
1700 switch (pkt->kind) {
1701 case VPX_CODEC_CX_FRAME_PKT:
1702 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1703 stream->frames_out++;
1704 }
1705 if (!global->quiet)
1706 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1707
1708 update_rate_histogram(stream->rate_hist, cfg, pkt);
1709 #if CONFIG_WEBM_IO
1710 if (stream->config.write_webm) {
1711 write_webm_block(&stream->ebml, cfg, pkt);
1712 }
1713 #endif
1714 if (!stream->config.write_webm) {
1715 if (pkt->data.frame.partition_id <= 0) {
1716 ivf_header_pos = ftello(stream->file);
1717 fsize = pkt->data.frame.sz;
1718
1719 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1720 } else {
1721 fsize += pkt->data.frame.sz;
1722
1723 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1724 const int64_t currpos = ftello(stream->file);
1725 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1726 ivf_write_frame_size(stream->file, fsize);
1727 fseeko(stream->file, currpos, SEEK_SET);
1728 }
1729 }
1730
1731 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1732 stream->file);
1733 }
1734 stream->nbytes += pkt->data.raw.sz;
1735
1736 *got_data = 1;
1737 #if CONFIG_DECODERS
1738 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1739 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1740 (unsigned int)pkt->data.frame.sz, NULL, 0);
1741 if (stream->decoder.err) {
1742 warn_or_exit_on_error(&stream->decoder,
1743 global->test_decode == TEST_DECODE_FATAL,
1744 "Failed to decode frame %d in stream %d",
1745 stream->frames_out + 1, stream->index);
1746 stream->mismatch_seen = stream->frames_out + 1;
1747 }
1748 }
1749 #endif
1750 break;
1751 case VPX_CODEC_STATS_PKT:
1752 stream->frames_out++;
1753 stats_write(&stream->stats,
1754 pkt->data.twopass_stats.buf,
1755 pkt->data.twopass_stats.sz);
1756 stream->nbytes += pkt->data.raw.sz;
1757 break;
1758 #if CONFIG_FP_MB_STATS
1759 case VPX_CODEC_FPMB_STATS_PKT:
1760 stats_write(&stream->fpmb_stats,
1761 pkt->data.firstpass_mb_stats.buf,
1762 pkt->data.firstpass_mb_stats.sz);
1763 stream->nbytes += pkt->data.raw.sz;
1764 break;
1765 #endif
1766 case VPX_CODEC_PSNR_PKT:
1767
1768 if (global->show_psnr) {
1769 int i;
1770
1771 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1772 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1773 for (i = 0; i < 4; i++) {
1774 if (!global->quiet)
1775 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1776 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1777 }
1778 stream->psnr_count++;
1779 }
1780
1781 break;
1782 default:
1783 break;
1784 }
1785 }
1786 }
1787
1788
1789 static void show_psnr(struct stream_state *stream, double peak) {
1790 int i;
1791 double ovpsnr;
1792
1793 if (!stream->psnr_count)
1794 return;
1795
1796 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1797 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1798 (double)stream->psnr_sse_total);
1799 fprintf(stderr, " %.3f", ovpsnr);
1800
1801 for (i = 0; i < 4; i++) {
1802 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1803 }
1804 fprintf(stderr, "\n");
1805 }
1806
1807
1808 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1809 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1810 }
1811
1812 static void test_decode(struct stream_state *stream,
1813 enum TestDecodeFatality fatal,
1814 const VpxInterface *codec) {
1815 vpx_image_t enc_img, dec_img;
1816
1817 if (stream->mismatch_seen)
1818 return;
1819
1820 /* Get the internal reference frame */
1821 if (strcmp(codec->name, "vp8") == 0) {
1822 struct vpx_ref_frame ref_enc, ref_dec;
1823 int width, height;
1824
1825 width = (stream->config.cfg.g_w + 15) & ~15;
1826 height = (stream->config.cfg.g_h + 15) & ~15;
1827 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1828 enc_img = ref_enc.img;
1829 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1830 dec_img = ref_dec.img;
1831
1832 ref_enc.frame_type = VP8_LAST_FRAME;
1833 ref_dec.frame_type = VP8_LAST_FRAME;
1834 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1835 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1836 } else {
1837 struct vp9_ref_frame ref_enc, ref_dec;
1838
1839 ref_enc.idx = 0;
1840 ref_dec.idx = 0;
1841 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1842 enc_img = ref_enc.img;
1843 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1844 dec_img = ref_dec.img;
1845 #if CONFIG_VP9_HIGHBITDEPTH
1846 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1847 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1848 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1849 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1850 enc_img.d_w, enc_img.d_h, 16);
1851 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1852 }
1853 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1854 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1855 dec_img.d_w, dec_img.d_h, 16);
1856 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1857 }
1858 }
1859 #endif
1860 }
1861 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1862 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1863
1864 if (!compare_img(&enc_img, &dec_img)) {
1865 int y[4], u[4], v[4];
1866 #if CONFIG_VP9_HIGHBITDEPTH
1867 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1868 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1869 } else {
1870 find_mismatch(&enc_img, &dec_img, y, u, v);
1871 }
1872 #else
1873 find_mismatch(&enc_img, &dec_img, y, u, v);
1874 #endif
1875 stream->decoder.err = 1;
1876 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1877 "Stream %d: Encode/decode mismatch on frame %d at"
1878 " Y[%d, %d] {%d/%d},"
1879 " U[%d, %d] {%d/%d},"
1880 " V[%d, %d] {%d/%d}",
1881 stream->index, stream->frames_out,
1882 y[0], y[1], y[2], y[3],
1883 u[0], u[1], u[2], u[3],
1884 v[0], v[1], v[2], v[3]);
1885 stream->mismatch_seen = stream->frames_out;
1886 }
1887
1888 vpx_img_free(&enc_img);
1889 vpx_img_free(&dec_img);
1890 }
1891
1892
1893 static void print_time(const char *label, int64_t etl) {
1894 int64_t hours;
1895 int64_t mins;
1896 int64_t secs;
1897
1898 if (etl >= 0) {
1899 hours = etl / 3600;
1900 etl -= hours * 3600;
1901 mins = etl / 60;
1902 etl -= mins * 60;
1903 secs = etl;
1904
1905 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1906 label, hours, mins, secs);
1907 } else {
1908 fprintf(stderr, "[%3s unknown] ", label);
1909 }
1910 }
1911
1912
1913 int main(int argc, const char **argv_) {
1914 int pass;
1915 vpx_image_t raw;
1916 #if CONFIG_VP9_HIGHBITDEPTH
1917 vpx_image_t raw_shift;
1918 int allocated_raw_shift = 0;
1919 int use_16bit_internal = 0;
1920 int input_shift = 0;
1921 #endif
1922 int frame_avail, got_data;
1923
1924 struct VpxInputContext input;
1925 struct VpxEncoderConfig global;
1926 struct stream_state *streams = NULL;
1927 char **argv, **argi;
1928 uint64_t cx_time = 0;
1929 int stream_cnt = 0;
1930 int res = 0;
1931
1932 memset(&input, 0, sizeof(input));
1933 exec_name = argv_[0];
1934
1935 if (argc < 3)
1936 usage_exit();
1937
1938 /* Setup default input stream settings */
1939 input.framerate.numerator = 30;
1940 input.framerate.denominator = 1;
1941 input.only_i420 = 1;
1942 input.bit_depth = 0;
1943
1944 /* First parse the global configuration values, because we want to apply
1945 * other parameters on top of the default configuration provided by the
1946 * codec.
1947 */
1948 argv = argv_dup(argc - 1, argv_ + 1);
1949 parse_global_config(&global, argv);
1950
1951 switch (global.color_type) {
1952 case I420:
1953 input.fmt = VPX_IMG_FMT_I420;
1954 break;
1955 case I422:
1956 input.fmt = VPX_IMG_FMT_I422;
1957 break;
1958 case I444:
1959 input.fmt = VPX_IMG_FMT_I444;
1960 break;
1961 case I440:
1962 input.fmt = VPX_IMG_FMT_I440;
1963 break;
1964 case YV12:
1965 input.fmt = VPX_IMG_FMT_YV12;
1966 break;
1967 }
1968
1969 {
1970 /* Now parse each stream's parameters. Using a local scope here
1971 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1972 * loops
1973 */
1974 struct stream_state *stream = NULL;
1975
1976 do {
1977 stream = new_stream(&global, stream);
1978 stream_cnt++;
1979 if (!streams)
1980 streams = stream;
1981 } while (parse_stream_params(&global, stream, argv));
1982 }
1983
1984 /* Check for unrecognized options */
1985 for (argi = argv; *argi; argi++)
1986 if (argi[0][0] == '-' && argi[0][1])
1987 die("Error: Unrecognized option %s\n", *argi);
1988
1989 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1990 &global, &stream->config.cfg););
1991
1992 /* Handle non-option arguments */
1993 input.filename = argv[0];
1994
1995 if (!input.filename)
1996 usage_exit();
1997
1998 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1999 if (global.codec->fourcc == VP9_FOURCC || global.codec->fourcc == VP10_FOURCC)
2000 input.only_i420 = 0;
2001
2002 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2003 int frames_in = 0, seen_frames = 0;
2004 int64_t estimated_time_left = -1;
2005 int64_t average_rate = -1;
2006 int64_t lagged_count = 0;
2007
2008 open_input_file(&input);
2009
2010 /* If the input file doesn't specify its w/h (raw files), try to get
2011 * the data from the first stream's configuration.
2012 */
2013 if (!input.width || !input.height) {
2014 FOREACH_STREAM({
2015 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2016 input.width = stream->config.cfg.g_w;
2017 input.height = stream->config.cfg.g_h;
2018 break;
2019 }
2020 });
2021 }
2022
2023 /* Update stream configurations from the input file's parameters */
2024 if (!input.width || !input.height)
2025 fatal("Specify stream dimensions with --width (-w) "
2026 " and --height (-h)");
2027
2028 /* If input file does not specify bit-depth but input-bit-depth parameter
2029 * exists, assume that to be the input bit-depth. However, if the
2030 * input-bit-depth paramter does not exist, assume the input bit-depth
2031 * to be the same as the codec bit-depth.
2032 */
2033 if (!input.bit_depth) {
2034 FOREACH_STREAM({
2035 if (stream->config.cfg.g_input_bit_depth)
2036 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2037 else
2038 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2039 (int)stream->config.cfg.g_bit_depth;
2040 });
2041 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2042 } else {
2043 FOREACH_STREAM({
2044 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2045 });
2046 }
2047
2048 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2049 FOREACH_STREAM(validate_stream_config(stream, &global));
2050
2051 /* Ensure that --passes and --pass are consistent. If --pass is set and
2052 * --passes=2, ensure --fpf was set.
2053 */
2054 if (global.pass && global.passes == 2)
2055 FOREACH_STREAM( {
2056 if (!stream->config.stats_fn)
2057 die("Stream %d: Must specify --fpf when --pass=%d"
2058 " and --passes=2\n", stream->index, global.pass);
2059 });
2060
2061 #if !CONFIG_WEBM_IO
2062 FOREACH_STREAM({
2063 if (stream->config.write_webm) {
2064 stream->config.write_webm = 0;
2065 warn("vpxenc was compiled without WebM container support."
2066 "Producing IVF output");
2067 }
2068 });
2069 #endif
2070
2071 /* Use the frame rate from the file only if none was specified
2072 * on the command-line.
2073 */
2074 if (!global.have_framerate) {
2075 global.framerate.num = input.framerate.numerator;
2076 global.framerate.den = input.framerate.denominator;
2077 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2078 stream->config.cfg.g_timebase.num = global.framerate.den);
2079 }
2080
2081 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2082
2083 /* Show configuration */
2084 if (global.verbose && pass == 0)
2085 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2086
2087 if (pass == (global.pass ? global.pass - 1 : 0)) {
2088 if (input.file_type == FILE_TYPE_Y4M)
2089 /*The Y4M reader does its own allocation.
2090 Just initialize this here to avoid problems if we never read any
2091 frames.*/
2092 memset(&raw, 0, sizeof(raw));
2093 else
2094 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2095
2096 FOREACH_STREAM(stream->rate_hist =
2097 init_rate_histogram(&stream->config.cfg,
2098 &global.framerate));
2099 }
2100
2101 FOREACH_STREAM(setup_pass(stream, &global, pass));
2102 FOREACH_STREAM(open_output_file(stream, &global,
2103 &input.pixel_aspect_ratio));
2104 FOREACH_STREAM(initialize_encoder(stream, &global));
2105
2106 #if CONFIG_VP9_HIGHBITDEPTH
2107 if (strcmp(global.codec->name, "vp9") == 0 ||
2108 strcmp(global.codec->name, "vp10") == 0) {
2109 // Check to see if at least one stream uses 16 bit internal.
2110 // Currently assume that the bit_depths for all streams using
2111 // highbitdepth are the same.
2112 FOREACH_STREAM({
2113 if (stream->config.use_16bit_internal) {
2114 use_16bit_internal = 1;
2115 }
2116 if (stream->config.cfg.g_profile == 0) {
2117 input_shift = 0;
2118 } else {
2119 input_shift = (int)stream->config.cfg.g_bit_depth -
2120 stream->config.cfg.g_input_bit_depth;
2121 }
2122 });
2123 }
2124 #endif
2125
2126 frame_avail = 1;
2127 got_data = 0;
2128
2129 while (frame_avail || got_data) {
2130 struct vpx_usec_timer timer;
2131
2132 if (!global.limit || frames_in < global.limit) {
2133 frame_avail = read_frame(&input, &raw);
2134
2135 if (frame_avail)
2136 frames_in++;
2137 seen_frames = frames_in > global.skip_frames ?
2138 frames_in - global.skip_frames : 0;
2139
2140 if (!global.quiet) {
2141 float fps = usec_to_fps(cx_time, seen_frames);
2142 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2143
2144 if (stream_cnt == 1)
2145 fprintf(stderr,
2146 "frame %4d/%-4d %7"PRId64"B ",
2147 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2148 else
2149 fprintf(stderr, "frame %4d ", frames_in);
2150
2151 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2152 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2153 cx_time > 9999999 ? "ms" : "us",
2154 fps >= 1.0 ? fps : fps * 60,
2155 fps >= 1.0 ? "fps" : "fpm");
2156 print_time("ETA", estimated_time_left);
2157 }
2158
2159 } else
2160 frame_avail = 0;
2161
2162 if (frames_in > global.skip_frames) {
2163 #if CONFIG_VP9_HIGHBITDEPTH
2164 vpx_image_t *frame_to_encode;
2165 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2166 assert(use_16bit_internal);
2167 // Input bit depth and stream bit depth do not match, so up
2168 // shift frame to stream bit depth
2169 if (!allocated_raw_shift) {
2170 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2171 input.width, input.height, 32);
2172 allocated_raw_shift = 1;
2173 }
2174 vpx_img_upshift(&raw_shift, &raw, input_shift);
2175 frame_to_encode = &raw_shift;
2176 } else {
2177 frame_to_encode = &raw;
2178 }
2179 vpx_usec_timer_start(&timer);
2180 if (use_16bit_internal) {
2181 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2182 FOREACH_STREAM({
2183 if (stream->config.use_16bit_internal)
2184 encode_frame(stream, &global,
2185 frame_avail ? frame_to_encode : NULL,
2186 frames_in);
2187 else
2188 assert(0);
2189 });
2190 } else {
2191 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2192 FOREACH_STREAM(encode_frame(stream, &global,
2193 frame_avail ? frame_to_encode : NULL,
2194 frames_in));
2195 }
2196 #else
2197 vpx_usec_timer_start(&timer);
2198 FOREACH_STREAM(encode_frame(stream, &global,
2199 frame_avail ? &raw : NULL,
2200 frames_in));
2201 #endif
2202 vpx_usec_timer_mark(&timer);
2203 cx_time += vpx_usec_timer_elapsed(&timer);
2204
2205 FOREACH_STREAM(update_quantizer_histogram(stream));
2206
2207 got_data = 0;
2208 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2209
2210 if (!got_data && input.length && streams != NULL &&
2211 !streams->frames_out) {
2212 lagged_count = global.limit ? seen_frames : ftello(input.file);
2213 } else if (input.length) {
2214 int64_t remaining;
2215 int64_t rate;
2216
2217 if (global.limit) {
2218 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2219
2220 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2221 remaining = 1000 * (global.limit - global.skip_frames
2222 - seen_frames + lagged_count);
2223 } else {
2224 const int64_t input_pos = ftello(input.file);
2225 const int64_t input_pos_lagged = input_pos - lagged_count;
2226 const int64_t limit = input.length;
2227
2228 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2229 remaining = limit - input_pos + lagged_count;
2230 }
2231
2232 average_rate = (average_rate <= 0)
2233 ? rate
2234 : (average_rate * 7 + rate) / 8;
2235 estimated_time_left = average_rate ? remaining / average_rate : -1;
2236 }
2237
2238 if (got_data && global.test_decode != TEST_DECODE_OFF)
2239 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2240 }
2241
2242 fflush(stdout);
2243 if (!global.quiet)
2244 fprintf(stderr, "\033[K");
2245 }
2246
2247 if (stream_cnt > 1)
2248 fprintf(stderr, "\n");
2249
2250 if (!global.quiet) {
2251 FOREACH_STREAM(fprintf(stderr,
2252 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2253 " %7"PRId64" %s (%.2f fps)\033[K\n",
2254 pass + 1,
2255 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2256 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2257 seen_frames ? (int64_t)stream->nbytes * 8 *
2258 (int64_t)global.framerate.num / global.framerate.den /
2259 seen_frames : 0,
2260 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2261 stream->cx_time > 9999999 ? "ms" : "us",
2262 usec_to_fps(stream->cx_time, seen_frames)));
2263 }
2264
2265 if (global.show_psnr) {
2266 if (global.codec->fourcc == VP9_FOURCC) {
2267 FOREACH_STREAM(
2268 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2269 } else {
2270 FOREACH_STREAM(show_psnr(stream, 255.0));
2271 }
2272 }
2273
2274 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2275
2276 if (global.test_decode != TEST_DECODE_OFF) {
2277 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2278 }
2279
2280 close_input_file(&input);
2281
2282 if (global.test_decode == TEST_DECODE_FATAL) {
2283 FOREACH_STREAM(res |= stream->mismatch_seen);
2284 }
2285 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2286
2287 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2288
2289 #if CONFIG_FP_MB_STATS
2290 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2291 #endif
2292
2293 if (global.pass)
2294 break;
2295 }
2296
2297 if (global.show_q_hist_buckets)
2298 FOREACH_STREAM(show_q_histogram(stream->counts,
2299 global.show_q_hist_buckets));
2300
2301 if (global.show_rate_hist_buckets)
2302 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2303 &stream->config.cfg,
2304 global.show_rate_hist_buckets));
2305 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2306
2307 #if CONFIG_INTERNAL_STATS
2308 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2309 * to match some existing utilities.
2310 */
2311 if (!(global.pass == 1 && global.passes == 2))
2312 FOREACH_STREAM({
2313 FILE *f = fopen("opsnr.stt", "a");
2314 if (stream->mismatch_seen) {
2315 fprintf(f, "First mismatch occurred in frame %d\n",
2316 stream->mismatch_seen);
2317 } else {
2318 fprintf(f, "No mismatch detected in recon buffers\n");
2319 }
2320 fclose(f);
2321 });
2322 #endif
2323
2324 #if CONFIG_VP9_HIGHBITDEPTH
2325 if (allocated_raw_shift)
2326 vpx_img_free(&raw_shift);
2327 #endif
2328 vpx_img_free(&raw);
2329 free(argv);
2330 free(streams);
2331 return res ? EXIT_FAILURE : EXIT_SUCCESS;
2332 }
AV1 Codec Library