2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
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.
12 #include "./vpx_config.h"
23 #include "third_party/libyuv/include/libyuv/scale.h"
26 #include "vpx/vpx_encoder.h"
28 #include "vpx/vpx_decoder.h"
33 #include "./tools_common.h"
35 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER || CONFIG_VP10_ENCODER
36 #include "vpx/vp8cx.h"
38 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER || CONFIG_VP10_DECODER
39 #include "vpx/vp8dx.h"
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"
49 #include "./webmenc.h"
51 #include "./y4minput.h"
53 /* Swallow warnings about unused results of fread/fwrite */
54 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
56 return fread(ptr, size, nmemb, stream);
58 #define fread wrap_fread
60 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
62 return fwrite(ptr, size, nmemb, stream);
64 #define fwrite wrap_fwrite
67 static const char *exec_name;
69 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
70 const char *s, va_list ap) {
72 const char *detail = vpx_codec_error_detail(ctx);
74 vfprintf(stderr, s, ap);
75 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
78 fprintf(stderr, " %s\n", detail);
85 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
89 warn_or_exit_on_errorv(ctx, 1, s, ap);
93 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
98 warn_or_exit_on_errorv(ctx, fatal, s, ap);
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;
107 if (input_ctx->file_type == FILE_TYPE_Y4M) {
108 if (y4m_input_fetch_frame(y4m, f, img) < 1)
111 shortread = read_yuv_frame(input_ctx, img);
117 static int file_is_y4m(const char detect[4]) {
118 if (memcmp(detect, "YUV4", 4) == 0) {
124 static int fourcc_is_ivf(const char detect[4]) {
125 if (memcmp(detect, "DKIF", 4) == 0) {
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");
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");
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},
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.");
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");
209 static const arg_def_t *main_args[] = {
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,
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");
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},
235 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
236 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
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");
245 static const arg_def_t *global_args[] = {
246 &use_yv12, &use_i420, &use_i422, &use_i444, &use_i440,
247 &usage, &threads, &profile,
252 &timebase, &framerate,
254 #if CONFIG_VP9_HIGHBITDEPTH
255 &test16bitinternalarg,
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[] = {
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
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
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
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},
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)");
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,
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,
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(
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)");
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 },
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);
422 #if CONFIG_VP9_HIGHBITDEPTH
423 static const struct arg_enum_list bitdepth_enum[] = {
430 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
432 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
434 static const arg_def_t inbitdeptharg = ARG_DEF(
435 NULL, "input-bit-depth", 1, "Bit depth of input");
438 static const struct arg_enum_list tune_content_enum[] = {
439 {"default", VP9E_CONTENT_DEFAULT},
440 {"screen", VP9E_CONTENT_SCREEN},
444 static const arg_def_t tune_content = ARG_DEF_ENUM(
445 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
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 #if CONFIG_VP9_HIGHBITDEPTH
458 &bitdeptharg, &inbitdeptharg,
459 #endif // CONFIG_VP9_HIGHBITDEPTH
462 static const int vp9_arg_ctrl_map[] = {
463 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
464 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
465 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
466 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
467 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
468 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
469 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
470 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
471 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
472 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL,
477 #if CONFIG_VP10_ENCODER
478 static const arg_def_t *vp10_args[] = {
479 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
480 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
481 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
482 &gf_cbr_boost_pct, &lossless,
483 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
484 &noise_sens, &tune_content, &input_color_space,
485 &min_gf_interval, &max_gf_interval,
486 #if CONFIG_VP9_HIGHBITDEPTH
487 &bitdeptharg, &inbitdeptharg,
488 #endif // CONFIG_VP9_HIGHBITDEPTH
491 static const int vp10_arg_ctrl_map[] = {
492 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
493 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
494 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
495 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
496 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
497 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
498 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
499 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
500 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
501 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL,
506 static const arg_def_t *no_args[] = { NULL };
508 void usage_exit(void) {
510 const int num_encoder = get_vpx_encoder_count();
512 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
515 fprintf(stderr, "\nOptions:\n");
516 arg_show_usage(stderr, main_args);
517 fprintf(stderr, "\nEncoder Global Options:\n");
518 arg_show_usage(stderr, global_args);
519 fprintf(stderr, "\nRate Control Options:\n");
520 arg_show_usage(stderr, rc_args);
521 fprintf(stderr, "\nTwopass Rate Control Options:\n");
522 arg_show_usage(stderr, rc_twopass_args);
523 fprintf(stderr, "\nKeyframe Placement Options:\n");
524 arg_show_usage(stderr, kf_args);
525 #if CONFIG_VP8_ENCODER
526 fprintf(stderr, "\nVP8 Specific Options:\n");
527 arg_show_usage(stderr, vp8_args);
529 #if CONFIG_VP9_ENCODER
530 fprintf(stderr, "\nVP9 Specific Options:\n");
531 arg_show_usage(stderr, vp9_args);
533 #if CONFIG_VP10_ENCODER
534 fprintf(stderr, "\nVP10 Specific Options:\n");
535 arg_show_usage(stderr, vp10_args);
537 fprintf(stderr, "\nStream timebase (--timebase):\n"
538 " The desired precision of timestamps in the output, expressed\n"
539 " in fractional seconds. Default is 1/1000.\n");
540 fprintf(stderr, "\nIncluded encoders:\n\n");
542 for (i = 0; i < num_encoder; ++i) {
543 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
544 const char* defstr = (i == (num_encoder - 1)) ? "(default)" : "";
545 fprintf(stderr, " %-6s - %s %s\n",
546 encoder->name, vpx_codec_iface_name(encoder->codec_interface()),
549 fprintf(stderr, "\n ");
550 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
555 #define mmin(a, b) ((a) < (b) ? (a) : (b))
557 #if CONFIG_VP9_HIGHBITDEPTH
558 static void find_mismatch_high(const vpx_image_t *const img1,
559 const vpx_image_t *const img2,
560 int yloc[4], int uloc[4], int vloc[4]) {
561 uint16_t *plane1, *plane2;
562 uint32_t stride1, stride2;
563 const uint32_t bsize = 64;
564 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
565 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
567 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
569 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
572 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
573 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
574 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
575 stride1 = img1->stride[VPX_PLANE_Y]/2;
576 stride2 = img2->stride[VPX_PLANE_Y]/2;
577 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
578 for (j = 0; match && j < img1->d_w; j += bsize) {
580 const int si = mmin(i + bsize, img1->d_h) - i;
581 const int sj = mmin(j + bsize, img1->d_w) - j;
582 for (k = 0; match && k < si; ++k) {
583 for (l = 0; match && l < sj; ++l) {
584 if (*(plane1 + (i + k) * stride1 + j + l) !=
585 *(plane2 + (i + k) * stride2 + j + l)) {
588 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
589 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
598 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
599 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
600 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
601 stride1 = img1->stride[VPX_PLANE_U]/2;
602 stride2 = img2->stride[VPX_PLANE_U]/2;
603 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
604 for (j = 0; match && j < c_w; j += bsizex) {
606 const int si = mmin(i + bsizey, c_h - i);
607 const int sj = mmin(j + bsizex, c_w - j);
608 for (k = 0; match && k < si; ++k) {
609 for (l = 0; match && l < sj; ++l) {
610 if (*(plane1 + (i + k) * stride1 + j + l) !=
611 *(plane2 + (i + k) * stride2 + j + l)) {
614 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
615 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
624 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
625 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
626 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
627 stride1 = img1->stride[VPX_PLANE_V]/2;
628 stride2 = img2->stride[VPX_PLANE_V]/2;
629 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
630 for (j = 0; match && j < c_w; j += bsizex) {
632 const int si = mmin(i + bsizey, c_h - i);
633 const int sj = mmin(j + bsizex, c_w - j);
634 for (k = 0; match && k < si; ++k) {
635 for (l = 0; match && l < sj; ++l) {
636 if (*(plane1 + (i + k) * stride1 + j + l) !=
637 *(plane2 + (i + k) * stride2 + j + l)) {
640 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
641 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
652 static void find_mismatch(const vpx_image_t *const img1,
653 const vpx_image_t *const img2,
654 int yloc[4], int uloc[4], int vloc[4]) {
655 const uint32_t bsize = 64;
656 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
657 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
659 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
661 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
664 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
665 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
666 for (j = 0; match && j < img1->d_w; j += bsize) {
668 const int si = mmin(i + bsize, img1->d_h) - i;
669 const int sj = mmin(j + bsize, img1->d_w) - j;
670 for (k = 0; match && k < si; ++k) {
671 for (l = 0; match && l < sj; ++l) {
672 if (*(img1->planes[VPX_PLANE_Y] +
673 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
674 *(img2->planes[VPX_PLANE_Y] +
675 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
678 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
679 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
680 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
681 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
690 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
691 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
692 for (j = 0; match && j < c_w; j += bsizex) {
694 const int si = mmin(i + bsizey, c_h - i);
695 const int sj = mmin(j + bsizex, c_w - j);
696 for (k = 0; match && k < si; ++k) {
697 for (l = 0; match && l < sj; ++l) {
698 if (*(img1->planes[VPX_PLANE_U] +
699 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
700 *(img2->planes[VPX_PLANE_U] +
701 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
704 uloc[2] = *(img1->planes[VPX_PLANE_U] +
705 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
706 uloc[3] = *(img2->planes[VPX_PLANE_U] +
707 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
715 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
716 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
717 for (j = 0; match && j < c_w; j += bsizex) {
719 const int si = mmin(i + bsizey, c_h - i);
720 const int sj = mmin(j + bsizex, c_w - j);
721 for (k = 0; match && k < si; ++k) {
722 for (l = 0; match && l < sj; ++l) {
723 if (*(img1->planes[VPX_PLANE_V] +
724 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
725 *(img2->planes[VPX_PLANE_V] +
726 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
729 vloc[2] = *(img1->planes[VPX_PLANE_V] +
730 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
731 vloc[3] = *(img2->planes[VPX_PLANE_V] +
732 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
742 static int compare_img(const vpx_image_t *const img1,
743 const vpx_image_t *const img2) {
744 uint32_t l_w = img1->d_w;
746 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
748 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
752 match &= (img1->fmt == img2->fmt);
753 match &= (img1->d_w == img2->d_w);
754 match &= (img1->d_h == img2->d_h);
755 #if CONFIG_VP9_HIGHBITDEPTH
756 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
762 for (i = 0; i < img1->d_h; ++i)
763 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
764 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
767 for (i = 0; i < c_h; ++i)
768 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
769 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
772 for (i = 0; i < c_h; ++i)
773 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
774 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
781 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
782 #if CONFIG_VP10_ENCODER
783 #define ARG_CTRL_CNT_MAX NELEMENTS(vp10_arg_ctrl_map)
784 #elif CONFIG_VP9_ENCODER
785 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
787 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
791 typedef int stereo_format_t;
792 struct EbmlGlobal { int debug; };
795 /* Per-stream configuration */
796 struct stream_config {
797 struct vpx_codec_enc_cfg cfg;
799 const char *stats_fn;
800 #if CONFIG_FP_MB_STATS
801 const char *fpmb_stats_fn;
803 stereo_format_t stereo_fmt;
804 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
807 #if CONFIG_VP9_HIGHBITDEPTH
808 // whether to use 16bit internal buffers
809 int use_16bit_internal;
814 struct stream_state {
816 struct stream_state *next;
817 struct stream_config config;
819 struct rate_hist *rate_hist;
820 struct EbmlGlobal ebml;
821 uint64_t psnr_sse_total;
822 uint64_t psnr_samples_total;
823 double psnr_totals[4];
826 vpx_codec_ctx_t encoder;
827 unsigned int frames_out;
831 #if CONFIG_FP_MB_STATS
832 stats_io_t fpmb_stats;
834 struct vpx_image *img;
835 vpx_codec_ctx_t decoder;
840 static void validate_positive_rational(const char *msg,
841 struct vpx_rational *rat) {
848 die("Error: %s must be positive\n", msg);
851 die("Error: %s has zero denominator\n", msg);
855 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
858 const int num_encoder = get_vpx_encoder_count();
861 die("Error: no valid encoder available\n");
863 /* Initialize default parameters */
864 memset(global, 0, sizeof(*global));
865 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
867 global->color_type = I420;
868 /* Assign default deadline to good quality */
869 global->deadline = VPX_DL_GOOD_QUALITY;
871 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
874 if (arg_match(&arg, &codecarg, argi)) {
875 global->codec = get_vpx_encoder_by_name(arg.val);
877 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
878 } else if (arg_match(&arg, &passes, argi)) {
879 global->passes = arg_parse_uint(&arg);
881 if (global->passes < 1 || global->passes > 2)
882 die("Error: Invalid number of passes (%d)\n", global->passes);
883 } else if (arg_match(&arg, &pass_arg, argi)) {
884 global->pass = arg_parse_uint(&arg);
886 if (global->pass < 1 || global->pass > 2)
887 die("Error: Invalid pass selected (%d)\n",
889 } else if (arg_match(&arg, &usage, argi))
890 global->usage = arg_parse_uint(&arg);
891 else if (arg_match(&arg, &deadline, argi))
892 global->deadline = arg_parse_uint(&arg);
893 else if (arg_match(&arg, &best_dl, argi))
894 global->deadline = VPX_DL_BEST_QUALITY;
895 else if (arg_match(&arg, &good_dl, argi))
896 global->deadline = VPX_DL_GOOD_QUALITY;
897 else if (arg_match(&arg, &rt_dl, argi))
898 global->deadline = VPX_DL_REALTIME;
899 else if (arg_match(&arg, &use_yv12, argi))
900 global->color_type = YV12;
901 else if (arg_match(&arg, &use_i420, argi))
902 global->color_type = I420;
903 else if (arg_match(&arg, &use_i422, argi))
904 global->color_type = I422;
905 else if (arg_match(&arg, &use_i444, argi))
906 global->color_type = I444;
907 else if (arg_match(&arg, &use_i440, argi))
908 global->color_type = I440;
909 else if (arg_match(&arg, &quietarg, argi))
911 else if (arg_match(&arg, &verbosearg, argi))
913 else if (arg_match(&arg, &limit, argi))
914 global->limit = arg_parse_uint(&arg);
915 else if (arg_match(&arg, &skip, argi))
916 global->skip_frames = arg_parse_uint(&arg);
917 else if (arg_match(&arg, &psnrarg, argi))
918 global->show_psnr = 1;
919 else if (arg_match(&arg, &recontest, argi))
920 global->test_decode = arg_parse_enum_or_int(&arg);
921 else if (arg_match(&arg, &framerate, argi)) {
922 global->framerate = arg_parse_rational(&arg);
923 validate_positive_rational(arg.name, &global->framerate);
924 global->have_framerate = 1;
925 } else if (arg_match(&arg, &out_part, argi))
926 global->out_part = 1;
927 else if (arg_match(&arg, &debugmode, argi))
929 else if (arg_match(&arg, &q_hist_n, argi))
930 global->show_q_hist_buckets = arg_parse_uint(&arg);
931 else if (arg_match(&arg, &rate_hist_n, argi))
932 global->show_rate_hist_buckets = arg_parse_uint(&arg);
933 else if (arg_match(&arg, &disable_warnings, argi))
934 global->disable_warnings = 1;
935 else if (arg_match(&arg, &disable_warning_prompt, argi))
936 global->disable_warning_prompt = 1;
942 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
943 if (global->pass > global->passes) {
944 warn("Assuming --pass=%d implies --passes=%d\n",
945 global->pass, global->pass);
946 global->passes = global->pass;
949 /* Validate global config */
950 if (global->passes == 0) {
951 #if CONFIG_VP9_ENCODER || CONFIG_VP10_ENCODER
952 // Make default VP9 passes = 2 until there is a better quality 1-pass
954 if (global->codec != NULL && global->codec->name != NULL)
955 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
956 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
962 if (global->deadline == VPX_DL_REALTIME &&
963 global->passes > 1) {
964 warn("Enforcing one-pass encoding in realtime mode\n");
970 static void open_input_file(struct VpxInputContext *input) {
971 /* Parse certain options from the input file, if possible */
972 input->file = strcmp(input->filename, "-")
973 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
976 fatal("Failed to open input file");
978 if (!fseeko(input->file, 0, SEEK_END)) {
979 /* Input file is seekable. Figure out how long it is, so we can get
982 input->length = ftello(input->file);
986 /* Default to 1:1 pixel aspect ratio. */
987 input->pixel_aspect_ratio.numerator = 1;
988 input->pixel_aspect_ratio.denominator = 1;
990 /* For RAW input sources, these bytes will applied on the first frame
993 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
994 input->detect.position = 0;
996 if (input->detect.buf_read == 4
997 && file_is_y4m(input->detect.buf)) {
998 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
999 input->only_i420) >= 0) {
1000 input->file_type = FILE_TYPE_Y4M;
1001 input->width = input->y4m.pic_w;
1002 input->height = input->y4m.pic_h;
1003 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1004 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1005 input->framerate.numerator = input->y4m.fps_n;
1006 input->framerate.denominator = input->y4m.fps_d;
1007 input->fmt = input->y4m.vpx_fmt;
1008 input->bit_depth = input->y4m.bit_depth;
1010 fatal("Unsupported Y4M stream.");
1011 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1012 fatal("IVF is not supported as input.");
1014 input->file_type = FILE_TYPE_RAW;
1019 static void close_input_file(struct VpxInputContext *input) {
1020 fclose(input->file);
1021 if (input->file_type == FILE_TYPE_Y4M)
1022 y4m_input_close(&input->y4m);
1025 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1026 struct stream_state *prev) {
1027 struct stream_state *stream;
1029 stream = calloc(1, sizeof(*stream));
1030 if (stream == NULL) {
1031 fatal("Failed to allocate new stream.");
1035 memcpy(stream, prev, sizeof(*stream));
1037 prev->next = stream;
1039 vpx_codec_err_t res;
1041 /* Populate encoder configuration */
1042 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1043 &stream->config.cfg,
1046 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1048 /* Change the default timebase to a high enough value so that the
1049 * encoder will always create strictly increasing timestamps.
1051 stream->config.cfg.g_timebase.den = 1000;
1053 /* Never use the library's default resolution, require it be parsed
1054 * from the file or set on the command line.
1056 stream->config.cfg.g_w = 0;
1057 stream->config.cfg.g_h = 0;
1059 /* Initialize remaining stream parameters */
1060 stream->config.write_webm = 1;
1062 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1063 stream->ebml.last_pts_ns = -1;
1064 stream->ebml.writer = NULL;
1065 stream->ebml.segment = NULL;
1068 /* Allows removal of the application version from the EBML tags */
1069 stream->ebml.debug = global->debug;
1071 /* Default lag_in_frames is 0 in realtime mode */
1072 if (global->deadline == VPX_DL_REALTIME)
1073 stream->config.cfg.g_lag_in_frames = 0;
1076 /* Output files must be specified for each stream */
1077 stream->config.out_fn = NULL;
1079 stream->next = NULL;
1084 static int parse_stream_params(struct VpxEncoderConfig *global,
1085 struct stream_state *stream,
1087 char **argi, **argj;
1089 static const arg_def_t **ctrl_args = no_args;
1090 static const int *ctrl_args_map = NULL;
1091 struct stream_config *config = &stream->config;
1092 int eos_mark_found = 0;
1093 #if CONFIG_VP9_HIGHBITDEPTH
1094 int test_16bit_internal = 0;
1097 // Handle codec specific options
1099 #if CONFIG_VP8_ENCODER
1100 } else if (strcmp(global->codec->name, "vp8") == 0) {
1101 ctrl_args = vp8_args;
1102 ctrl_args_map = vp8_arg_ctrl_map;
1104 #if CONFIG_VP9_ENCODER
1105 } else if (strcmp(global->codec->name, "vp9") == 0) {
1106 ctrl_args = vp9_args;
1107 ctrl_args_map = vp9_arg_ctrl_map;
1109 #if CONFIG_VP10_ENCODER
1110 } else if (strcmp(global->codec->name, "vp10") == 0) {
1111 // TODO(jingning): Reuse VP9 specific encoder configuration parameters.
1112 // Consider to expand this set for VP10 encoder control.
1113 ctrl_args = vp10_args;
1114 ctrl_args_map = vp10_arg_ctrl_map;
1118 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1121 /* Once we've found an end-of-stream marker (--) we want to continue
1122 * shifting arguments but not consuming them.
1124 if (eos_mark_found) {
1127 } else if (!strcmp(*argj, "--")) {
1132 if (arg_match(&arg, &outputfile, argi)) {
1133 config->out_fn = arg.val;
1134 } else if (arg_match(&arg, &fpf_name, argi)) {
1135 config->stats_fn = arg.val;
1136 #if CONFIG_FP_MB_STATS
1137 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1138 config->fpmb_stats_fn = arg.val;
1140 } else if (arg_match(&arg, &use_webm, argi)) {
1142 config->write_webm = 1;
1144 die("Error: --webm specified but webm is disabled.");
1146 } else if (arg_match(&arg, &use_ivf, argi)) {
1147 config->write_webm = 0;
1148 } else if (arg_match(&arg, &threads, argi)) {
1149 config->cfg.g_threads = arg_parse_uint(&arg);
1150 } else if (arg_match(&arg, &profile, argi)) {
1151 config->cfg.g_profile = arg_parse_uint(&arg);
1152 } else if (arg_match(&arg, &width, argi)) {
1153 config->cfg.g_w = arg_parse_uint(&arg);
1154 } else if (arg_match(&arg, &height, argi)) {
1155 config->cfg.g_h = arg_parse_uint(&arg);
1156 #if CONFIG_VP9_HIGHBITDEPTH
1157 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1158 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1159 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1160 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1163 } else if (arg_match(&arg, &stereo_mode, argi)) {
1164 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1166 } else if (arg_match(&arg, &timebase, argi)) {
1167 config->cfg.g_timebase = arg_parse_rational(&arg);
1168 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1169 } else if (arg_match(&arg, &error_resilient, argi)) {
1170 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1171 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1172 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1173 if (global->deadline == VPX_DL_REALTIME &&
1174 config->cfg.g_lag_in_frames != 0) {
1175 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1176 config->cfg.g_lag_in_frames = 0;
1178 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1179 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1180 } else if (arg_match(&arg, &resize_allowed, argi)) {
1181 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1182 } else if (arg_match(&arg, &resize_width, argi)) {
1183 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1184 } else if (arg_match(&arg, &resize_height, argi)) {
1185 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1186 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1187 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1188 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1189 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1190 } else if (arg_match(&arg, &end_usage, argi)) {
1191 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1192 } else if (arg_match(&arg, &target_bitrate, argi)) {
1193 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1194 } else if (arg_match(&arg, &min_quantizer, argi)) {
1195 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1196 } else if (arg_match(&arg, &max_quantizer, argi)) {
1197 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1198 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1199 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1200 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1201 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1202 } else if (arg_match(&arg, &buf_sz, argi)) {
1203 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1204 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1205 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1206 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1207 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1208 } else if (arg_match(&arg, &bias_pct, argi)) {
1209 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1210 if (global->passes < 2)
1211 warn("option %s ignored in one-pass mode.\n", arg.name);
1212 } else if (arg_match(&arg, &minsection_pct, argi)) {
1213 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1215 if (global->passes < 2)
1216 warn("option %s ignored in one-pass mode.\n", arg.name);
1217 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1218 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1220 if (global->passes < 2)
1221 warn("option %s ignored in one-pass mode.\n", arg.name);
1222 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1223 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1224 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1225 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1226 } else if (arg_match(&arg, &kf_disabled, argi)) {
1227 config->cfg.kf_mode = VPX_KF_DISABLED;
1228 #if CONFIG_VP9_HIGHBITDEPTH
1229 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1230 if (strcmp(global->codec->name, "vp9") == 0 ||
1231 strcmp(global->codec->name, "vp10") == 0) {
1232 test_16bit_internal = 1;
1237 for (i = 0; ctrl_args[i]; i++) {
1238 if (arg_match(&arg, ctrl_args[i], argi)) {
1242 /* Point either to the next free element or the first
1243 * instance of this control.
1245 for (j = 0; j < config->arg_ctrl_cnt; j++)
1246 if (ctrl_args_map != NULL &&
1247 config->arg_ctrls[j][0] == ctrl_args_map[i])
1251 assert(j < (int)ARG_CTRL_CNT_MAX);
1252 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1253 config->arg_ctrls[j][0] = ctrl_args_map[i];
1254 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1255 if (j == config->arg_ctrl_cnt)
1256 config->arg_ctrl_cnt++;
1264 #if CONFIG_VP9_HIGHBITDEPTH
1265 if (strcmp(global->codec->name, "vp9") == 0 ||
1266 strcmp(global->codec->name, "vp10") == 0) {
1267 config->use_16bit_internal = test_16bit_internal |
1268 (config->cfg.g_profile > 1);
1271 return eos_mark_found;
1275 #define FOREACH_STREAM(func) \
1277 struct stream_state *stream; \
1278 for (stream = streams; stream; stream = stream->next) { \
1284 static void validate_stream_config(const struct stream_state *stream,
1285 const struct VpxEncoderConfig *global) {
1286 const struct stream_state *streami;
1289 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1290 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1291 " and --height (-h)", stream->index);
1293 // Check that the codec bit depth is greater than the input bit depth.
1294 if (stream->config.cfg.g_input_bit_depth >
1295 (unsigned int)stream->config.cfg.g_bit_depth) {
1296 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1297 stream->index, (int)stream->config.cfg.g_bit_depth,
1298 stream->config.cfg.g_input_bit_depth);
1301 for (streami = stream; streami; streami = streami->next) {
1302 /* All streams require output files */
1303 if (!streami->config.out_fn)
1304 fatal("Stream %d: Output file is required (specify with -o)",
1307 /* Check for two streams outputting to the same file */
1308 if (streami != stream) {
1309 const char *a = stream->config.out_fn;
1310 const char *b = streami->config.out_fn;
1311 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1312 fatal("Stream %d: duplicate output file (from stream %d)",
1313 streami->index, stream->index);
1316 /* Check for two streams sharing a stats file. */
1317 if (streami != stream) {
1318 const char *a = stream->config.stats_fn;
1319 const char *b = streami->config.stats_fn;
1320 if (a && b && !strcmp(a, b))
1321 fatal("Stream %d: duplicate stats file (from stream %d)",
1322 streami->index, stream->index);
1325 #if CONFIG_FP_MB_STATS
1326 /* Check for two streams sharing a mb stats file. */
1327 if (streami != stream) {
1328 const char *a = stream->config.fpmb_stats_fn;
1329 const char *b = streami->config.fpmb_stats_fn;
1330 if (a && b && !strcmp(a, b))
1331 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1332 streami->index, stream->index);
1339 static void set_stream_dimensions(struct stream_state *stream,
1342 if (!stream->config.cfg.g_w) {
1343 if (!stream->config.cfg.g_h)
1344 stream->config.cfg.g_w = w;
1346 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1348 if (!stream->config.cfg.g_h) {
1349 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1353 static const char* file_type_to_string(enum VideoFileType t) {
1355 case FILE_TYPE_RAW: return "RAW";
1356 case FILE_TYPE_Y4M: return "Y4M";
1357 default: return "Other";
1361 static const char* image_format_to_string(vpx_img_fmt_t f) {
1363 case VPX_IMG_FMT_I420: return "I420";
1364 case VPX_IMG_FMT_I422: return "I422";
1365 case VPX_IMG_FMT_I444: return "I444";
1366 case VPX_IMG_FMT_I440: return "I440";
1367 case VPX_IMG_FMT_YV12: return "YV12";
1368 case VPX_IMG_FMT_I42016: return "I42016";
1369 case VPX_IMG_FMT_I42216: return "I42216";
1370 case VPX_IMG_FMT_I44416: return "I44416";
1371 case VPX_IMG_FMT_I44016: return "I44016";
1372 default: return "Other";
1376 static void show_stream_config(struct stream_state *stream,
1377 struct VpxEncoderConfig *global,
1378 struct VpxInputContext *input) {
1380 #define SHOW(field) \
1381 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1383 if (stream->index == 0) {
1384 fprintf(stderr, "Codec: %s\n",
1385 vpx_codec_iface_name(global->codec->codec_interface()));
1386 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1388 file_type_to_string(input->file_type),
1389 image_format_to_string(input->fmt));
1391 if (stream->next || stream->index)
1392 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1393 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1394 fprintf(stderr, "Encoder parameters:\n");
1402 SHOW(g_input_bit_depth);
1403 SHOW(g_timebase.num);
1404 SHOW(g_timebase.den);
1405 SHOW(g_error_resilient);
1407 SHOW(g_lag_in_frames);
1408 SHOW(rc_dropframe_thresh);
1409 SHOW(rc_resize_allowed);
1410 SHOW(rc_scaled_width);
1411 SHOW(rc_scaled_height);
1412 SHOW(rc_resize_up_thresh);
1413 SHOW(rc_resize_down_thresh);
1415 SHOW(rc_target_bitrate);
1416 SHOW(rc_min_quantizer);
1417 SHOW(rc_max_quantizer);
1418 SHOW(rc_undershoot_pct);
1419 SHOW(rc_overshoot_pct);
1421 SHOW(rc_buf_initial_sz);
1422 SHOW(rc_buf_optimal_sz);
1423 SHOW(rc_2pass_vbr_bias_pct);
1424 SHOW(rc_2pass_vbr_minsection_pct);
1425 SHOW(rc_2pass_vbr_maxsection_pct);
1432 static void open_output_file(struct stream_state *stream,
1433 struct VpxEncoderConfig *global,
1434 const struct VpxRational *pixel_aspect_ratio) {
1435 const char *fn = stream->config.out_fn;
1436 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1438 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1441 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1444 fatal("Failed to open output file");
1446 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1447 fatal("WebM output to pipes not supported.");
1450 if (stream->config.write_webm) {
1451 stream->ebml.stream = stream->file;
1452 write_webm_file_header(&stream->ebml, cfg,
1454 stream->config.stereo_fmt,
1455 global->codec->fourcc,
1456 pixel_aspect_ratio);
1459 (void)pixel_aspect_ratio;
1462 if (!stream->config.write_webm) {
1463 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1468 static void close_output_file(struct stream_state *stream,
1469 unsigned int fourcc) {
1470 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1472 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1476 if (stream->config.write_webm) {
1477 write_webm_file_footer(&stream->ebml);
1481 if (!stream->config.write_webm) {
1482 if (!fseek(stream->file, 0, SEEK_SET))
1483 ivf_write_file_header(stream->file, &stream->config.cfg,
1485 stream->frames_out);
1488 fclose(stream->file);
1492 static void setup_pass(struct stream_state *stream,
1493 struct VpxEncoderConfig *global,
1495 if (stream->config.stats_fn) {
1496 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1498 fatal("Failed to open statistics store");
1500 if (!stats_open_mem(&stream->stats, pass))
1501 fatal("Failed to open statistics store");
1504 #if CONFIG_FP_MB_STATS
1505 if (stream->config.fpmb_stats_fn) {
1506 if (!stats_open_file(&stream->fpmb_stats,
1507 stream->config.fpmb_stats_fn, pass))
1508 fatal("Failed to open mb statistics store");
1510 if (!stats_open_mem(&stream->fpmb_stats, pass))
1511 fatal("Failed to open mb statistics store");
1515 stream->config.cfg.g_pass = global->passes == 2
1516 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1519 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1520 #if CONFIG_FP_MB_STATS
1521 stream->config.cfg.rc_firstpass_mb_stats_in =
1522 stats_get(&stream->fpmb_stats);
1526 stream->cx_time = 0;
1528 stream->frames_out = 0;
1532 static void initialize_encoder(struct stream_state *stream,
1533 struct VpxEncoderConfig *global) {
1537 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1538 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1539 #if CONFIG_VP9_HIGHBITDEPTH
1540 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1543 /* Construct Encoder Context */
1544 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1545 &stream->config.cfg, flags);
1546 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1548 /* Note that we bypass the vpx_codec_control wrapper macro because
1549 * we're being clever to store the control IDs in an array. Real
1550 * applications will want to make use of the enumerations directly
1552 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1553 int ctrl = stream->config.arg_ctrls[i][0];
1554 int value = stream->config.arg_ctrls[i][1];
1555 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1556 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1559 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1563 if (global->test_decode != TEST_DECODE_OFF) {
1564 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1565 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1571 static void encode_frame(struct stream_state *stream,
1572 struct VpxEncoderConfig *global,
1573 struct vpx_image *img,
1574 unsigned int frames_in) {
1575 vpx_codec_pts_t frame_start, next_frame_start;
1576 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1577 struct vpx_usec_timer timer;
1579 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1580 * global->framerate.den)
1581 / cfg->g_timebase.num / global->framerate.num;
1582 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1583 * global->framerate.den)
1584 / cfg->g_timebase.num / global->framerate.num;
1586 /* Scale if necessary */
1587 #if CONFIG_VP9_HIGHBITDEPTH
1589 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1590 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1591 if (img->fmt != VPX_IMG_FMT_I42016) {
1592 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1597 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1598 cfg->g_w, cfg->g_h, 16);
1600 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1601 img->stride[VPX_PLANE_Y]/2,
1602 (uint16*)img->planes[VPX_PLANE_U],
1603 img->stride[VPX_PLANE_U]/2,
1604 (uint16*)img->planes[VPX_PLANE_V],
1605 img->stride[VPX_PLANE_V]/2,
1607 (uint16*)stream->img->planes[VPX_PLANE_Y],
1608 stream->img->stride[VPX_PLANE_Y]/2,
1609 (uint16*)stream->img->planes[VPX_PLANE_U],
1610 stream->img->stride[VPX_PLANE_U]/2,
1611 (uint16*)stream->img->planes[VPX_PLANE_V],
1612 stream->img->stride[VPX_PLANE_V]/2,
1613 stream->img->d_w, stream->img->d_h,
1617 stream->encoder.err = 1;
1618 ctx_exit_on_error(&stream->encoder,
1619 "Stream %d: Failed to encode frame.\n"
1620 "Scaling disabled in this configuration. \n"
1621 "To enable, configure with --enable-libyuv\n",
1627 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1628 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1629 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1634 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1635 cfg->g_w, cfg->g_h, 16);
1636 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1637 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1638 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1640 stream->img->planes[VPX_PLANE_Y],
1641 stream->img->stride[VPX_PLANE_Y],
1642 stream->img->planes[VPX_PLANE_U],
1643 stream->img->stride[VPX_PLANE_U],
1644 stream->img->planes[VPX_PLANE_V],
1645 stream->img->stride[VPX_PLANE_V],
1646 stream->img->d_w, stream->img->d_h,
1650 stream->encoder.err = 1;
1651 ctx_exit_on_error(&stream->encoder,
1652 "Stream %d: Failed to encode frame.\n"
1653 "Scaling disabled in this configuration. \n"
1654 "To enable, configure with --enable-libyuv\n",
1659 vpx_usec_timer_start(&timer);
1660 vpx_codec_encode(&stream->encoder, img, frame_start,
1661 (unsigned long)(next_frame_start - frame_start),
1662 0, global->deadline);
1663 vpx_usec_timer_mark(&timer);
1664 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1665 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1670 static void update_quantizer_histogram(struct stream_state *stream) {
1671 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1674 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1675 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1676 stream->counts[q]++;
1681 static void get_cx_data(struct stream_state *stream,
1682 struct VpxEncoderConfig *global,
1684 const vpx_codec_cx_pkt_t *pkt;
1685 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1686 vpx_codec_iter_t iter = NULL;
1689 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1690 static size_t fsize = 0;
1691 static int64_t ivf_header_pos = 0;
1693 switch (pkt->kind) {
1694 case VPX_CODEC_CX_FRAME_PKT:
1695 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1696 stream->frames_out++;
1699 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1701 update_rate_histogram(stream->rate_hist, cfg, pkt);
1703 if (stream->config.write_webm) {
1704 write_webm_block(&stream->ebml, cfg, pkt);
1707 if (!stream->config.write_webm) {
1708 if (pkt->data.frame.partition_id <= 0) {
1709 ivf_header_pos = ftello(stream->file);
1710 fsize = pkt->data.frame.sz;
1712 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1714 fsize += pkt->data.frame.sz;
1716 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1717 const int64_t currpos = ftello(stream->file);
1718 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1719 ivf_write_frame_size(stream->file, fsize);
1720 fseeko(stream->file, currpos, SEEK_SET);
1724 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1727 stream->nbytes += pkt->data.raw.sz;
1731 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1732 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1733 (unsigned int)pkt->data.frame.sz, NULL, 0);
1734 if (stream->decoder.err) {
1735 warn_or_exit_on_error(&stream->decoder,
1736 global->test_decode == TEST_DECODE_FATAL,
1737 "Failed to decode frame %d in stream %d",
1738 stream->frames_out + 1, stream->index);
1739 stream->mismatch_seen = stream->frames_out + 1;
1744 case VPX_CODEC_STATS_PKT:
1745 stream->frames_out++;
1746 stats_write(&stream->stats,
1747 pkt->data.twopass_stats.buf,
1748 pkt->data.twopass_stats.sz);
1749 stream->nbytes += pkt->data.raw.sz;
1751 #if CONFIG_FP_MB_STATS
1752 case VPX_CODEC_FPMB_STATS_PKT:
1753 stats_write(&stream->fpmb_stats,
1754 pkt->data.firstpass_mb_stats.buf,
1755 pkt->data.firstpass_mb_stats.sz);
1756 stream->nbytes += pkt->data.raw.sz;
1759 case VPX_CODEC_PSNR_PKT:
1761 if (global->show_psnr) {
1764 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1765 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1766 for (i = 0; i < 4; i++) {
1768 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1769 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1771 stream->psnr_count++;
1782 static void show_psnr(struct stream_state *stream, double peak) {
1786 if (!stream->psnr_count)
1789 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1790 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1791 (double)stream->psnr_sse_total);
1792 fprintf(stderr, " %.3f", ovpsnr);
1794 for (i = 0; i < 4; i++) {
1795 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1797 fprintf(stderr, "\n");
1801 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1802 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1805 static void test_decode(struct stream_state *stream,
1806 enum TestDecodeFatality fatal,
1807 const VpxInterface *codec) {
1808 vpx_image_t enc_img, dec_img;
1810 if (stream->mismatch_seen)
1813 /* Get the internal reference frame */
1814 if (strcmp(codec->name, "vp8") == 0) {
1815 struct vpx_ref_frame ref_enc, ref_dec;
1818 width = (stream->config.cfg.g_w + 15) & ~15;
1819 height = (stream->config.cfg.g_h + 15) & ~15;
1820 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1821 enc_img = ref_enc.img;
1822 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1823 dec_img = ref_dec.img;
1825 ref_enc.frame_type = VP8_LAST_FRAME;
1826 ref_dec.frame_type = VP8_LAST_FRAME;
1827 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1828 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1830 struct vp9_ref_frame ref_enc, ref_dec;
1834 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1835 enc_img = ref_enc.img;
1836 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1837 dec_img = ref_dec.img;
1838 #if CONFIG_VP9_HIGHBITDEPTH
1839 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1840 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1841 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1842 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1843 enc_img.d_w, enc_img.d_h, 16);
1844 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1846 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1847 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1848 dec_img.d_w, dec_img.d_h, 16);
1849 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1854 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1855 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1857 if (!compare_img(&enc_img, &dec_img)) {
1858 int y[4], u[4], v[4];
1859 #if CONFIG_VP9_HIGHBITDEPTH
1860 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1861 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1863 find_mismatch(&enc_img, &dec_img, y, u, v);
1866 find_mismatch(&enc_img, &dec_img, y, u, v);
1868 stream->decoder.err = 1;
1869 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1870 "Stream %d: Encode/decode mismatch on frame %d at"
1871 " Y[%d, %d] {%d/%d},"
1872 " U[%d, %d] {%d/%d},"
1873 " V[%d, %d] {%d/%d}",
1874 stream->index, stream->frames_out,
1875 y[0], y[1], y[2], y[3],
1876 u[0], u[1], u[2], u[3],
1877 v[0], v[1], v[2], v[3]);
1878 stream->mismatch_seen = stream->frames_out;
1881 vpx_img_free(&enc_img);
1882 vpx_img_free(&dec_img);
1886 static void print_time(const char *label, int64_t etl) {
1893 etl -= hours * 3600;
1898 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1899 label, hours, mins, secs);
1901 fprintf(stderr, "[%3s unknown] ", label);
1906 int main(int argc, const char **argv_) {
1909 #if CONFIG_VP9_HIGHBITDEPTH
1910 vpx_image_t raw_shift;
1911 int allocated_raw_shift = 0;
1912 int use_16bit_internal = 0;
1913 int input_shift = 0;
1915 int frame_avail, got_data;
1917 struct VpxInputContext input;
1918 struct VpxEncoderConfig global;
1919 struct stream_state *streams = NULL;
1920 char **argv, **argi;
1921 uint64_t cx_time = 0;
1925 memset(&input, 0, sizeof(input));
1926 exec_name = argv_[0];
1931 /* Setup default input stream settings */
1932 input.framerate.numerator = 30;
1933 input.framerate.denominator = 1;
1934 input.only_i420 = 1;
1935 input.bit_depth = 0;
1937 /* First parse the global configuration values, because we want to apply
1938 * other parameters on top of the default configuration provided by the
1941 argv = argv_dup(argc - 1, argv_ + 1);
1942 parse_global_config(&global, argv);
1944 switch (global.color_type) {
1946 input.fmt = VPX_IMG_FMT_I420;
1949 input.fmt = VPX_IMG_FMT_I422;
1952 input.fmt = VPX_IMG_FMT_I444;
1955 input.fmt = VPX_IMG_FMT_I440;
1958 input.fmt = VPX_IMG_FMT_YV12;
1963 /* Now parse each stream's parameters. Using a local scope here
1964 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1967 struct stream_state *stream = NULL;
1970 stream = new_stream(&global, stream);
1974 } while (parse_stream_params(&global, stream, argv));
1977 /* Check for unrecognized options */
1978 for (argi = argv; *argi; argi++)
1979 if (argi[0][0] == '-' && argi[0][1])
1980 die("Error: Unrecognized option %s\n", *argi);
1982 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1983 &global, &stream->config.cfg););
1985 /* Handle non-option arguments */
1986 input.filename = argv[0];
1988 if (!input.filename)
1991 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1992 if (global.codec->fourcc == VP9_FOURCC || global.codec->fourcc == VP10_FOURCC)
1993 input.only_i420 = 0;
1995 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1996 int frames_in = 0, seen_frames = 0;
1997 int64_t estimated_time_left = -1;
1998 int64_t average_rate = -1;
1999 int64_t lagged_count = 0;
2001 open_input_file(&input);
2003 /* If the input file doesn't specify its w/h (raw files), try to get
2004 * the data from the first stream's configuration.
2006 if (!input.width || !input.height) {
2008 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2009 input.width = stream->config.cfg.g_w;
2010 input.height = stream->config.cfg.g_h;
2016 /* Update stream configurations from the input file's parameters */
2017 if (!input.width || !input.height)
2018 fatal("Specify stream dimensions with --width (-w) "
2019 " and --height (-h)");
2021 /* If input file does not specify bit-depth but input-bit-depth parameter
2022 * exists, assume that to be the input bit-depth. However, if the
2023 * input-bit-depth paramter does not exist, assume the input bit-depth
2024 * to be the same as the codec bit-depth.
2026 if (!input.bit_depth) {
2028 if (stream->config.cfg.g_input_bit_depth)
2029 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2031 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2032 (int)stream->config.cfg.g_bit_depth;
2034 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2037 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2041 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2042 FOREACH_STREAM(validate_stream_config(stream, &global));
2044 /* Ensure that --passes and --pass are consistent. If --pass is set and
2045 * --passes=2, ensure --fpf was set.
2047 if (global.pass && global.passes == 2)
2049 if (!stream->config.stats_fn)
2050 die("Stream %d: Must specify --fpf when --pass=%d"
2051 " and --passes=2\n", stream->index, global.pass);
2056 if (stream->config.write_webm) {
2057 stream->config.write_webm = 0;
2058 warn("vpxenc was compiled without WebM container support."
2059 "Producing IVF output");
2064 /* Use the frame rate from the file only if none was specified
2065 * on the command-line.
2067 if (!global.have_framerate) {
2068 global.framerate.num = input.framerate.numerator;
2069 global.framerate.den = input.framerate.denominator;
2070 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2071 stream->config.cfg.g_timebase.num = global.framerate.den);
2074 /* Show configuration */
2075 if (global.verbose && pass == 0)
2076 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2078 if (pass == (global.pass ? global.pass - 1 : 0)) {
2079 if (input.file_type == FILE_TYPE_Y4M)
2080 /*The Y4M reader does its own allocation.
2081 Just initialize this here to avoid problems if we never read any
2083 memset(&raw, 0, sizeof(raw));
2085 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2087 FOREACH_STREAM(stream->rate_hist =
2088 init_rate_histogram(&stream->config.cfg,
2089 &global.framerate));
2092 FOREACH_STREAM(setup_pass(stream, &global, pass));
2093 FOREACH_STREAM(open_output_file(stream, &global,
2094 &input.pixel_aspect_ratio));
2095 FOREACH_STREAM(initialize_encoder(stream, &global));
2097 #if CONFIG_VP9_HIGHBITDEPTH
2098 if (strcmp(global.codec->name, "vp9") == 0 ||
2099 strcmp(global.codec->name, "vp10") == 0) {
2100 // Check to see if at least one stream uses 16 bit internal.
2101 // Currently assume that the bit_depths for all streams using
2102 // highbitdepth are the same.
2104 if (stream->config.use_16bit_internal) {
2105 use_16bit_internal = 1;
2107 if (stream->config.cfg.g_profile == 0) {
2110 input_shift = (int)stream->config.cfg.g_bit_depth -
2111 stream->config.cfg.g_input_bit_depth;
2120 while (frame_avail || got_data) {
2121 struct vpx_usec_timer timer;
2123 if (!global.limit || frames_in < global.limit) {
2124 frame_avail = read_frame(&input, &raw);
2128 seen_frames = frames_in > global.skip_frames ?
2129 frames_in - global.skip_frames : 0;
2131 if (!global.quiet) {
2132 float fps = usec_to_fps(cx_time, seen_frames);
2133 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2135 if (stream_cnt == 1)
2137 "frame %4d/%-4d %7"PRId64"B ",
2138 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2140 fprintf(stderr, "frame %4d ", frames_in);
2142 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2143 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2144 cx_time > 9999999 ? "ms" : "us",
2145 fps >= 1.0 ? fps : fps * 60,
2146 fps >= 1.0 ? "fps" : "fpm");
2147 print_time("ETA", estimated_time_left);
2153 if (frames_in > global.skip_frames) {
2154 #if CONFIG_VP9_HIGHBITDEPTH
2155 vpx_image_t *frame_to_encode;
2156 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2157 assert(use_16bit_internal);
2158 // Input bit depth and stream bit depth do not match, so up
2159 // shift frame to stream bit depth
2160 if (!allocated_raw_shift) {
2161 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2162 input.width, input.height, 32);
2163 allocated_raw_shift = 1;
2165 vpx_img_upshift(&raw_shift, &raw, input_shift);
2166 frame_to_encode = &raw_shift;
2168 frame_to_encode = &raw;
2170 vpx_usec_timer_start(&timer);
2171 if (use_16bit_internal) {
2172 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2174 if (stream->config.use_16bit_internal)
2175 encode_frame(stream, &global,
2176 frame_avail ? frame_to_encode : NULL,
2182 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2183 FOREACH_STREAM(encode_frame(stream, &global,
2184 frame_avail ? frame_to_encode : NULL,
2188 vpx_usec_timer_start(&timer);
2189 FOREACH_STREAM(encode_frame(stream, &global,
2190 frame_avail ? &raw : NULL,
2193 vpx_usec_timer_mark(&timer);
2194 cx_time += vpx_usec_timer_elapsed(&timer);
2196 FOREACH_STREAM(update_quantizer_histogram(stream));
2199 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2201 if (!got_data && input.length && streams != NULL &&
2202 !streams->frames_out) {
2203 lagged_count = global.limit ? seen_frames : ftello(input.file);
2204 } else if (input.length) {
2209 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2211 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2212 remaining = 1000 * (global.limit - global.skip_frames
2213 - seen_frames + lagged_count);
2215 const int64_t input_pos = ftello(input.file);
2216 const int64_t input_pos_lagged = input_pos - lagged_count;
2217 const int64_t limit = input.length;
2219 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2220 remaining = limit - input_pos + lagged_count;
2223 average_rate = (average_rate <= 0)
2225 : (average_rate * 7 + rate) / 8;
2226 estimated_time_left = average_rate ? remaining / average_rate : -1;
2229 if (got_data && global.test_decode != TEST_DECODE_OFF)
2230 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2235 fprintf(stderr, "\033[K");
2239 fprintf(stderr, "\n");
2241 if (!global.quiet) {
2242 FOREACH_STREAM(fprintf(stderr,
2243 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2244 " %7"PRId64" %s (%.2f fps)\033[K\n",
2246 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2247 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2248 seen_frames ? (int64_t)stream->nbytes * 8 *
2249 (int64_t)global.framerate.num / global.framerate.den /
2251 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2252 stream->cx_time > 9999999 ? "ms" : "us",
2253 usec_to_fps(stream->cx_time, seen_frames)));
2256 if (global.show_psnr) {
2257 if (global.codec->fourcc == VP9_FOURCC) {
2259 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2261 FOREACH_STREAM(show_psnr(stream, 255.0));
2265 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2267 if (global.test_decode != TEST_DECODE_OFF) {
2268 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2271 close_input_file(&input);
2273 if (global.test_decode == TEST_DECODE_FATAL) {
2274 FOREACH_STREAM(res |= stream->mismatch_seen);
2276 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2278 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2280 #if CONFIG_FP_MB_STATS
2281 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2288 if (global.show_q_hist_buckets)
2289 FOREACH_STREAM(show_q_histogram(stream->counts,
2290 global.show_q_hist_buckets));
2292 if (global.show_rate_hist_buckets)
2293 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2294 &stream->config.cfg,
2295 global.show_rate_hist_buckets));
2296 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2298 #if CONFIG_INTERNAL_STATS
2299 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2300 * to match some existing utilities.
2302 if (!(global.pass == 1 && global.passes == 2))
2304 FILE *f = fopen("opsnr.stt", "a");
2305 if (stream->mismatch_seen) {
2306 fprintf(f, "First mismatch occurred in frame %d\n",
2307 stream->mismatch_seen);
2309 fprintf(f, "No mismatch detected in recon buffers\n");
2315 #if CONFIG_VP9_HIGHBITDEPTH
2316 if (allocated_raw_shift)
2317 vpx_img_free(&raw_shift);
2322 return res ? EXIT_FAILURE : EXIT_SUCCESS;