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
36 #include "vpx/vp8cx.h"
38 #if CONFIG_VP8_DECODER || CONFIG_VP9_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 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 int file_is_y4m(const char detect[4]) {
118 if (memcmp(detect, "YUV4", 4) == 0) {
124 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 && 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,
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 && 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 cpu_used_vp8 = ARG_DEF(
334 NULL, "cpu-used", 1, "CPU Used (-16..16)");
335 static const arg_def_t cpu_used_vp9 = ARG_DEF(
336 NULL, "cpu-used", 1, "CPU Used (-8..8)");
337 static const arg_def_t auto_altref = ARG_DEF(
338 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
339 static const arg_def_t arnr_maxframes = ARG_DEF(
340 NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
341 static const arg_def_t arnr_strength = ARG_DEF(
342 NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
343 static const arg_def_t arnr_type = ARG_DEF(
344 NULL, "arnr-type", 1, "AltRef type");
345 static const struct arg_enum_list tuning_enum[] = {
346 {"psnr", VP8_TUNE_PSNR},
347 {"ssim", VP8_TUNE_SSIM},
350 static const arg_def_t tune_ssim = ARG_DEF_ENUM(
351 NULL, "tune", 1, "Material to favor", tuning_enum);
352 static const arg_def_t cq_level = ARG_DEF(
353 NULL, "cq-level", 1, "Constant/Constrained Quality level");
354 static const arg_def_t max_intra_rate_pct = ARG_DEF(
355 NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
356 static const arg_def_t max_inter_rate_pct = ARG_DEF(
357 NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
358 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
359 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
361 static const arg_def_t screen_content_mode = ARG_DEF(NULL, "screen-content-mode", 1,
362 "Screen content mode");
364 #if CONFIG_VP8_ENCODER
365 static const arg_def_t token_parts = ARG_DEF(
366 NULL, "token-parts", 1, "Number of token partitions to use, log2");
367 static const arg_def_t *vp8_args[] = {
368 &cpu_used_vp8, &auto_altref, &noise_sens, &sharpness, &static_thresh,
369 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
370 &tune_ssim, &cq_level, &max_intra_rate_pct, &screen_content_mode,
373 static const int vp8_arg_ctrl_map[] = {
374 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
375 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
376 VP8E_SET_TOKEN_PARTITIONS,
377 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
378 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
379 VP8E_SET_SCREEN_CONTENT_MODE,
384 #if CONFIG_VP9_ENCODER
385 static const arg_def_t tile_cols = ARG_DEF(
386 NULL, "tile-columns", 1, "Number of tile columns to use, log2");
387 static const arg_def_t tile_rows = ARG_DEF(
388 NULL, "tile-rows", 1, "Number of tile rows to use, log2");
389 static const arg_def_t lossless = ARG_DEF(
390 NULL, "lossless", 1, "Lossless mode");
391 static const arg_def_t frame_parallel_decoding = ARG_DEF(
392 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
393 static const arg_def_t aq_mode = ARG_DEF(
395 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
396 "3: cyclic refresh)");
397 static const arg_def_t frame_periodic_boost = ARG_DEF(
398 NULL, "frame-boost", 1,
399 "Enable frame periodic boost (0: off (default), 1: on)");
401 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
402 static const struct arg_enum_list bitdepth_enum[] = {
409 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
411 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
413 static const arg_def_t inbitdeptharg = ARG_DEF(
414 NULL, "input-bit-depth", 1, "Bit depth of input");
417 static const struct arg_enum_list tune_content_enum[] = {
418 {"default", VP9E_CONTENT_DEFAULT},
419 {"screen", VP9E_CONTENT_SCREEN},
423 static const arg_def_t tune_content = ARG_DEF_ENUM(
424 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
426 static const arg_def_t *vp9_args[] = {
427 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
428 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
429 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
430 &gf_cbr_boost_pct, &lossless,
431 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
432 &noise_sens, &tune_content,
433 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
434 &bitdeptharg, &inbitdeptharg,
438 static const int vp9_arg_ctrl_map[] = {
439 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
440 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
441 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
442 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
443 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
444 VP8E_SET_MAX_INTER_BITRATE_PCT, VP8E_SET_GF_CBR_BOOST_PCT,
445 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
446 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
447 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
452 static const arg_def_t *no_args[] = { NULL };
457 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
460 fprintf(stderr, "\nOptions:\n");
461 arg_show_usage(stderr, main_args);
462 fprintf(stderr, "\nEncoder Global Options:\n");
463 arg_show_usage(stderr, global_args);
464 fprintf(stderr, "\nRate Control Options:\n");
465 arg_show_usage(stderr, rc_args);
466 fprintf(stderr, "\nTwopass Rate Control Options:\n");
467 arg_show_usage(stderr, rc_twopass_args);
468 fprintf(stderr, "\nKeyframe Placement Options:\n");
469 arg_show_usage(stderr, kf_args);
470 #if CONFIG_VP8_ENCODER
471 fprintf(stderr, "\nVP8 Specific Options:\n");
472 arg_show_usage(stderr, vp8_args);
474 #if CONFIG_VP9_ENCODER
475 fprintf(stderr, "\nVP9 Specific Options:\n");
476 arg_show_usage(stderr, vp9_args);
478 fprintf(stderr, "\nStream timebase (--timebase):\n"
479 " The desired precision of timestamps in the output, expressed\n"
480 " in fractional seconds. Default is 1/1000.\n");
481 fprintf(stderr, "\nIncluded encoders:\n\n");
483 for (i = 0; i < get_vpx_encoder_count(); ++i) {
484 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
485 fprintf(stderr, " %-6s - %s\n",
486 encoder->name, vpx_codec_iface_name(encoder->codec_interface()));
492 #define mmin(a, b) ((a) < (b) ? (a) : (b))
494 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
495 static void find_mismatch_high(const vpx_image_t *const img1,
496 const vpx_image_t *const img2,
497 int yloc[4], int uloc[4], int vloc[4]) {
498 uint16_t *plane1, *plane2;
499 uint32_t stride1, stride2;
500 const uint32_t bsize = 64;
501 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
502 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
504 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
506 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
509 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
510 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
511 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
512 stride1 = img1->stride[VPX_PLANE_Y]/2;
513 stride2 = img2->stride[VPX_PLANE_Y]/2;
514 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
515 for (j = 0; match && j < img1->d_w; j += bsize) {
517 const int si = mmin(i + bsize, img1->d_h) - i;
518 const int sj = mmin(j + bsize, img1->d_w) - j;
519 for (k = 0; match && k < si; ++k) {
520 for (l = 0; match && l < sj; ++l) {
521 if (*(plane1 + (i + k) * stride1 + j + l) !=
522 *(plane2 + (i + k) * stride2 + j + l)) {
525 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
526 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
535 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
536 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
537 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
538 stride1 = img1->stride[VPX_PLANE_U]/2;
539 stride2 = img2->stride[VPX_PLANE_U]/2;
540 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
541 for (j = 0; match && j < c_w; j += bsizex) {
543 const int si = mmin(i + bsizey, c_h - i);
544 const int sj = mmin(j + bsizex, c_w - j);
545 for (k = 0; match && k < si; ++k) {
546 for (l = 0; match && l < sj; ++l) {
547 if (*(plane1 + (i + k) * stride1 + j + l) !=
548 *(plane2 + (i + k) * stride2 + j + l)) {
551 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
552 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
561 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
562 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
563 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
564 stride1 = img1->stride[VPX_PLANE_V]/2;
565 stride2 = img2->stride[VPX_PLANE_V]/2;
566 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
567 for (j = 0; match && j < c_w; j += bsizex) {
569 const int si = mmin(i + bsizey, c_h - i);
570 const int sj = mmin(j + bsizex, c_w - j);
571 for (k = 0; match && k < si; ++k) {
572 for (l = 0; match && l < sj; ++l) {
573 if (*(plane1 + (i + k) * stride1 + j + l) !=
574 *(plane2 + (i + k) * stride2 + j + l)) {
577 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
578 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
589 static void find_mismatch(const vpx_image_t *const img1,
590 const vpx_image_t *const img2,
591 int yloc[4], int uloc[4], int vloc[4]) {
592 const uint32_t bsize = 64;
593 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
594 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
596 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
598 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
601 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
602 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
603 for (j = 0; match && j < img1->d_w; j += bsize) {
605 const int si = mmin(i + bsize, img1->d_h) - i;
606 const int sj = mmin(j + bsize, img1->d_w) - j;
607 for (k = 0; match && k < si; ++k) {
608 for (l = 0; match && l < sj; ++l) {
609 if (*(img1->planes[VPX_PLANE_Y] +
610 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
611 *(img2->planes[VPX_PLANE_Y] +
612 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
615 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
616 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
617 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
618 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
627 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
628 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
629 for (j = 0; match && j < c_w; j += bsizex) {
631 const int si = mmin(i + bsizey, c_h - i);
632 const int sj = mmin(j + bsizex, c_w - j);
633 for (k = 0; match && k < si; ++k) {
634 for (l = 0; match && l < sj; ++l) {
635 if (*(img1->planes[VPX_PLANE_U] +
636 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
637 *(img2->planes[VPX_PLANE_U] +
638 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
641 uloc[2] = *(img1->planes[VPX_PLANE_U] +
642 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
643 uloc[3] = *(img2->planes[VPX_PLANE_U] +
644 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
652 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
653 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
654 for (j = 0; match && j < c_w; j += bsizex) {
656 const int si = mmin(i + bsizey, c_h - i);
657 const int sj = mmin(j + bsizex, c_w - j);
658 for (k = 0; match && k < si; ++k) {
659 for (l = 0; match && l < sj; ++l) {
660 if (*(img1->planes[VPX_PLANE_V] +
661 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
662 *(img2->planes[VPX_PLANE_V] +
663 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
666 vloc[2] = *(img1->planes[VPX_PLANE_V] +
667 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
668 vloc[3] = *(img2->planes[VPX_PLANE_V] +
669 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
679 static int compare_img(const vpx_image_t *const img1,
680 const vpx_image_t *const img2) {
681 uint32_t l_w = img1->d_w;
683 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
685 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
689 match &= (img1->fmt == img2->fmt);
690 match &= (img1->d_w == img2->d_w);
691 match &= (img1->d_h == img2->d_h);
692 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
693 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
699 for (i = 0; i < img1->d_h; ++i)
700 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
701 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
704 for (i = 0; i < c_h; ++i)
705 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
706 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
709 for (i = 0; i < c_h; ++i)
710 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
711 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
718 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
719 #define MAX(x,y) ((x)>(y)?(x):(y))
720 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
721 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
722 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
723 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
725 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
726 NELEMENTS(vp9_arg_ctrl_map))
730 typedef int stereo_format_t;
731 struct EbmlGlobal { int debug; };
734 /* Per-stream configuration */
735 struct stream_config {
736 struct vpx_codec_enc_cfg cfg;
738 const char *stats_fn;
739 #if CONFIG_FP_MB_STATS
740 const char *fpmb_stats_fn;
742 stereo_format_t stereo_fmt;
743 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
746 int have_kf_max_dist;
747 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
748 // whether to use 16bit internal buffers
749 int use_16bit_internal;
754 struct stream_state {
756 struct stream_state *next;
757 struct stream_config config;
759 struct rate_hist *rate_hist;
760 struct EbmlGlobal ebml;
761 uint64_t psnr_sse_total;
762 uint64_t psnr_samples_total;
763 double psnr_totals[4];
766 vpx_codec_ctx_t encoder;
767 unsigned int frames_out;
771 #if CONFIG_FP_MB_STATS
772 stats_io_t fpmb_stats;
774 struct vpx_image *img;
775 vpx_codec_ctx_t decoder;
780 void validate_positive_rational(const char *msg,
781 struct vpx_rational *rat) {
788 die("Error: %s must be positive\n", msg);
791 die("Error: %s has zero denominator\n", msg);
795 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
799 /* Initialize default parameters */
800 memset(global, 0, sizeof(*global));
801 global->codec = get_vpx_encoder_by_index(0);
803 global->color_type = I420;
804 /* Assign default deadline to good quality */
805 global->deadline = VPX_DL_GOOD_QUALITY;
807 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
810 if (arg_match(&arg, &codecarg, argi)) {
811 global->codec = get_vpx_encoder_by_name(arg.val);
813 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
814 } else if (arg_match(&arg, &passes, argi)) {
815 global->passes = arg_parse_uint(&arg);
817 if (global->passes < 1 || global->passes > 2)
818 die("Error: Invalid number of passes (%d)\n", global->passes);
819 } else if (arg_match(&arg, &pass_arg, argi)) {
820 global->pass = arg_parse_uint(&arg);
822 if (global->pass < 1 || global->pass > 2)
823 die("Error: Invalid pass selected (%d)\n",
825 } else if (arg_match(&arg, &usage, argi))
826 global->usage = arg_parse_uint(&arg);
827 else if (arg_match(&arg, &deadline, argi))
828 global->deadline = arg_parse_uint(&arg);
829 else if (arg_match(&arg, &best_dl, argi))
830 global->deadline = VPX_DL_BEST_QUALITY;
831 else if (arg_match(&arg, &good_dl, argi))
832 global->deadline = VPX_DL_GOOD_QUALITY;
833 else if (arg_match(&arg, &rt_dl, argi))
834 global->deadline = VPX_DL_REALTIME;
835 else if (arg_match(&arg, &use_yv12, argi))
836 global->color_type = YV12;
837 else if (arg_match(&arg, &use_i420, argi))
838 global->color_type = I420;
839 else if (arg_match(&arg, &use_i422, argi))
840 global->color_type = I422;
841 else if (arg_match(&arg, &use_i444, argi))
842 global->color_type = I444;
843 else if (arg_match(&arg, &use_i440, argi))
844 global->color_type = I440;
845 else if (arg_match(&arg, &quietarg, argi))
847 else if (arg_match(&arg, &verbosearg, argi))
849 else if (arg_match(&arg, &limit, argi))
850 global->limit = arg_parse_uint(&arg);
851 else if (arg_match(&arg, &skip, argi))
852 global->skip_frames = arg_parse_uint(&arg);
853 else if (arg_match(&arg, &psnrarg, argi))
854 global->show_psnr = 1;
855 else if (arg_match(&arg, &recontest, argi))
856 global->test_decode = arg_parse_enum_or_int(&arg);
857 else if (arg_match(&arg, &framerate, argi)) {
858 global->framerate = arg_parse_rational(&arg);
859 validate_positive_rational(arg.name, &global->framerate);
860 global->have_framerate = 1;
861 } else if (arg_match(&arg, &out_part, argi))
862 global->out_part = 1;
863 else if (arg_match(&arg, &debugmode, argi))
865 else if (arg_match(&arg, &q_hist_n, argi))
866 global->show_q_hist_buckets = arg_parse_uint(&arg);
867 else if (arg_match(&arg, &rate_hist_n, argi))
868 global->show_rate_hist_buckets = arg_parse_uint(&arg);
869 else if (arg_match(&arg, &disable_warnings, argi))
870 global->disable_warnings = 1;
871 else if (arg_match(&arg, &disable_warning_prompt, argi))
872 global->disable_warning_prompt = 1;
878 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
879 if (global->pass > global->passes) {
880 warn("Assuming --pass=%d implies --passes=%d\n",
881 global->pass, global->pass);
882 global->passes = global->pass;
885 /* Validate global config */
886 if (global->passes == 0) {
887 #if CONFIG_VP9_ENCODER
888 // Make default VP9 passes = 2 until there is a better quality 1-pass
890 if (global->codec != NULL && global->codec->name != NULL)
891 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
892 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
898 if (global->deadline == VPX_DL_REALTIME &&
899 global->passes > 1) {
900 warn("Enforcing one-pass encoding in realtime mode\n");
906 void open_input_file(struct VpxInputContext *input) {
907 /* Parse certain options from the input file, if possible */
908 input->file = strcmp(input->filename, "-")
909 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
912 fatal("Failed to open input file");
914 if (!fseeko(input->file, 0, SEEK_END)) {
915 /* Input file is seekable. Figure out how long it is, so we can get
918 input->length = ftello(input->file);
922 /* For RAW input sources, these bytes will applied on the first frame
925 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
926 input->detect.position = 0;
928 if (input->detect.buf_read == 4
929 && file_is_y4m(input->detect.buf)) {
930 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
931 input->only_i420) >= 0) {
932 input->file_type = FILE_TYPE_Y4M;
933 input->width = input->y4m.pic_w;
934 input->height = input->y4m.pic_h;
935 input->framerate.numerator = input->y4m.fps_n;
936 input->framerate.denominator = input->y4m.fps_d;
937 input->fmt = input->y4m.vpx_fmt;
938 input->bit_depth = input->y4m.bit_depth;
940 fatal("Unsupported Y4M stream.");
941 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
942 fatal("IVF is not supported as input.");
944 input->file_type = FILE_TYPE_RAW;
949 static void close_input_file(struct VpxInputContext *input) {
951 if (input->file_type == FILE_TYPE_Y4M)
952 y4m_input_close(&input->y4m);
955 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
956 struct stream_state *prev) {
957 struct stream_state *stream;
959 stream = calloc(1, sizeof(*stream));
960 if (stream == NULL) {
961 fatal("Failed to allocate new stream.");
965 memcpy(stream, prev, sizeof(*stream));
971 /* Populate encoder configuration */
972 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
976 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
978 /* Change the default timebase to a high enough value so that the
979 * encoder will always create strictly increasing timestamps.
981 stream->config.cfg.g_timebase.den = 1000;
983 /* Never use the library's default resolution, require it be parsed
984 * from the file or set on the command line.
986 stream->config.cfg.g_w = 0;
987 stream->config.cfg.g_h = 0;
989 /* Initialize remaining stream parameters */
990 stream->config.write_webm = 1;
992 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
993 stream->ebml.last_pts_ns = -1;
994 stream->ebml.writer = NULL;
995 stream->ebml.segment = NULL;
998 /* Allows removal of the application version from the EBML tags */
999 stream->ebml.debug = global->debug;
1001 /* Default lag_in_frames is 0 in realtime mode */
1002 if (global->deadline == VPX_DL_REALTIME)
1003 stream->config.cfg.g_lag_in_frames = 0;
1006 /* Output files must be specified for each stream */
1007 stream->config.out_fn = NULL;
1009 stream->next = NULL;
1014 static int parse_stream_params(struct VpxEncoderConfig *global,
1015 struct stream_state *stream,
1017 char **argi, **argj;
1019 static const arg_def_t **ctrl_args = no_args;
1020 static const int *ctrl_args_map = NULL;
1021 struct stream_config *config = &stream->config;
1022 int eos_mark_found = 0;
1023 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1024 int test_16bit_internal = 0;
1027 // Handle codec specific options
1029 #if CONFIG_VP8_ENCODER
1030 } else if (strcmp(global->codec->name, "vp8") == 0) {
1031 ctrl_args = vp8_args;
1032 ctrl_args_map = vp8_arg_ctrl_map;
1034 #if CONFIG_VP9_ENCODER
1035 } else if (strcmp(global->codec->name, "vp9") == 0) {
1036 ctrl_args = vp9_args;
1037 ctrl_args_map = vp9_arg_ctrl_map;
1041 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1044 /* Once we've found an end-of-stream marker (--) we want to continue
1045 * shifting arguments but not consuming them.
1047 if (eos_mark_found) {
1050 } else if (!strcmp(*argj, "--")) {
1056 } else if (arg_match(&arg, &outputfile, argi)) {
1057 config->out_fn = arg.val;
1058 } else if (arg_match(&arg, &fpf_name, argi)) {
1059 config->stats_fn = arg.val;
1060 #if CONFIG_FP_MB_STATS
1061 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1062 config->fpmb_stats_fn = arg.val;
1064 } else if (arg_match(&arg, &use_webm, argi)) {
1066 config->write_webm = 1;
1068 die("Error: --webm specified but webm is disabled.");
1070 } else if (arg_match(&arg, &use_ivf, argi)) {
1071 config->write_webm = 0;
1072 } else if (arg_match(&arg, &threads, argi)) {
1073 config->cfg.g_threads = arg_parse_uint(&arg);
1074 } else if (arg_match(&arg, &profile, argi)) {
1075 config->cfg.g_profile = arg_parse_uint(&arg);
1076 } else if (arg_match(&arg, &width, argi)) {
1077 config->cfg.g_w = arg_parse_uint(&arg);
1078 } else if (arg_match(&arg, &height, argi)) {
1079 config->cfg.g_h = arg_parse_uint(&arg);
1080 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1081 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1082 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1083 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1084 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1087 } else if (arg_match(&arg, &stereo_mode, argi)) {
1088 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1090 } else if (arg_match(&arg, &timebase, argi)) {
1091 config->cfg.g_timebase = arg_parse_rational(&arg);
1092 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1093 } else if (arg_match(&arg, &error_resilient, argi)) {
1094 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1095 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1096 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1097 if (global->deadline == VPX_DL_REALTIME &&
1098 config->cfg.g_lag_in_frames != 0) {
1099 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1100 config->cfg.g_lag_in_frames = 0;
1102 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1103 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1104 } else if (arg_match(&arg, &resize_allowed, argi)) {
1105 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1106 } else if (arg_match(&arg, &resize_width, argi)) {
1107 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1108 } else if (arg_match(&arg, &resize_height, argi)) {
1109 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1110 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1111 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1112 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1113 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1114 } else if (arg_match(&arg, &end_usage, argi)) {
1115 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1116 } else if (arg_match(&arg, &target_bitrate, argi)) {
1117 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1118 } else if (arg_match(&arg, &min_quantizer, argi)) {
1119 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1120 } else if (arg_match(&arg, &max_quantizer, argi)) {
1121 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1122 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1123 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1124 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1125 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1126 } else if (arg_match(&arg, &buf_sz, argi)) {
1127 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1128 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1129 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1130 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1131 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1132 } else if (arg_match(&arg, &bias_pct, argi)) {
1133 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1134 if (global->passes < 2)
1135 warn("option %s ignored in one-pass mode.\n", arg.name);
1136 } else if (arg_match(&arg, &minsection_pct, argi)) {
1137 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1139 if (global->passes < 2)
1140 warn("option %s ignored in one-pass mode.\n", arg.name);
1141 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1142 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1144 if (global->passes < 2)
1145 warn("option %s ignored in one-pass mode.\n", arg.name);
1146 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1147 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1148 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1149 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1150 config->have_kf_max_dist = 1;
1151 } else if (arg_match(&arg, &kf_disabled, argi)) {
1152 config->cfg.kf_mode = VPX_KF_DISABLED;
1153 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1154 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1155 if (strcmp(global->codec->name, "vp9") == 0) {
1156 test_16bit_internal = 1;
1161 for (i = 0; ctrl_args[i]; i++) {
1162 if (arg_match(&arg, ctrl_args[i], argi)) {
1166 /* Point either to the next free element or the first
1167 * instance of this control.
1169 for (j = 0; j < config->arg_ctrl_cnt; j++)
1170 if (ctrl_args_map != NULL &&
1171 config->arg_ctrls[j][0] == ctrl_args_map[i])
1175 assert(j < (int)ARG_CTRL_CNT_MAX);
1176 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1177 config->arg_ctrls[j][0] = ctrl_args_map[i];
1178 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1179 if (j == config->arg_ctrl_cnt)
1180 config->arg_ctrl_cnt++;
1188 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1189 if (strcmp(global->codec->name, "vp9") == 0) {
1190 config->use_16bit_internal = test_16bit_internal |
1191 (config->cfg.g_profile > 1);
1194 return eos_mark_found;
1198 #define FOREACH_STREAM(func) \
1200 struct stream_state *stream; \
1201 for (stream = streams; stream; stream = stream->next) { \
1207 static void validate_stream_config(const struct stream_state *stream,
1208 const struct VpxEncoderConfig *global) {
1209 const struct stream_state *streami;
1212 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1213 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1214 " and --height (-h)", stream->index);
1216 // Check that the codec bit depth is greater than the input bit depth.
1217 if (stream->config.cfg.g_input_bit_depth >
1218 (unsigned int)stream->config.cfg.g_bit_depth) {
1219 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1220 stream->index, (int)stream->config.cfg.g_bit_depth,
1221 stream->config.cfg.g_input_bit_depth);
1224 for (streami = stream; streami; streami = streami->next) {
1225 /* All streams require output files */
1226 if (!streami->config.out_fn)
1227 fatal("Stream %d: Output file is required (specify with -o)",
1230 /* Check for two streams outputting to the same file */
1231 if (streami != stream) {
1232 const char *a = stream->config.out_fn;
1233 const char *b = streami->config.out_fn;
1234 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1235 fatal("Stream %d: duplicate output file (from stream %d)",
1236 streami->index, stream->index);
1239 /* Check for two streams sharing a stats file. */
1240 if (streami != stream) {
1241 const char *a = stream->config.stats_fn;
1242 const char *b = streami->config.stats_fn;
1243 if (a && b && !strcmp(a, b))
1244 fatal("Stream %d: duplicate stats file (from stream %d)",
1245 streami->index, stream->index);
1248 #if CONFIG_FP_MB_STATS
1249 /* Check for two streams sharing a mb stats file. */
1250 if (streami != stream) {
1251 const char *a = stream->config.fpmb_stats_fn;
1252 const char *b = streami->config.fpmb_stats_fn;
1253 if (a && b && !strcmp(a, b))
1254 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1255 streami->index, stream->index);
1262 static void set_stream_dimensions(struct stream_state *stream,
1265 if (!stream->config.cfg.g_w) {
1266 if (!stream->config.cfg.g_h)
1267 stream->config.cfg.g_w = w;
1269 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1271 if (!stream->config.cfg.g_h) {
1272 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1277 static void set_default_kf_interval(struct stream_state *stream,
1278 struct VpxEncoderConfig *global) {
1279 /* Use a max keyframe interval of 5 seconds, if none was
1280 * specified on the command line.
1282 if (!stream->config.have_kf_max_dist) {
1283 double framerate = (double)global->framerate.num / global->framerate.den;
1284 if (framerate > 0.0)
1285 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1289 static const char* file_type_to_string(enum VideoFileType t) {
1291 case FILE_TYPE_RAW: return "RAW";
1292 case FILE_TYPE_Y4M: return "Y4M";
1293 default: return "Other";
1297 static const char* image_format_to_string(vpx_img_fmt_t f) {
1299 case VPX_IMG_FMT_I420: return "I420";
1300 case VPX_IMG_FMT_I422: return "I422";
1301 case VPX_IMG_FMT_I444: return "I444";
1302 case VPX_IMG_FMT_I440: return "I440";
1303 case VPX_IMG_FMT_YV12: return "YV12";
1304 case VPX_IMG_FMT_I42016: return "I42016";
1305 case VPX_IMG_FMT_I42216: return "I42216";
1306 case VPX_IMG_FMT_I44416: return "I44416";
1307 case VPX_IMG_FMT_I44016: return "I44016";
1308 default: return "Other";
1312 static void show_stream_config(struct stream_state *stream,
1313 struct VpxEncoderConfig *global,
1314 struct VpxInputContext *input) {
1316 #define SHOW(field) \
1317 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1319 if (stream->index == 0) {
1320 fprintf(stderr, "Codec: %s\n",
1321 vpx_codec_iface_name(global->codec->codec_interface()));
1322 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1324 file_type_to_string(input->file_type),
1325 image_format_to_string(input->fmt));
1327 if (stream->next || stream->index)
1328 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1329 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1330 fprintf(stderr, "Encoder parameters:\n");
1338 SHOW(g_input_bit_depth);
1339 SHOW(g_timebase.num);
1340 SHOW(g_timebase.den);
1341 SHOW(g_error_resilient);
1343 SHOW(g_lag_in_frames);
1344 SHOW(rc_dropframe_thresh);
1345 SHOW(rc_resize_allowed);
1346 SHOW(rc_scaled_width);
1347 SHOW(rc_scaled_height);
1348 SHOW(rc_resize_up_thresh);
1349 SHOW(rc_resize_down_thresh);
1351 SHOW(rc_target_bitrate);
1352 SHOW(rc_min_quantizer);
1353 SHOW(rc_max_quantizer);
1354 SHOW(rc_undershoot_pct);
1355 SHOW(rc_overshoot_pct);
1357 SHOW(rc_buf_initial_sz);
1358 SHOW(rc_buf_optimal_sz);
1359 SHOW(rc_2pass_vbr_bias_pct);
1360 SHOW(rc_2pass_vbr_minsection_pct);
1361 SHOW(rc_2pass_vbr_maxsection_pct);
1368 static void open_output_file(struct stream_state *stream,
1369 struct VpxEncoderConfig *global) {
1370 const char *fn = stream->config.out_fn;
1371 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1373 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1376 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1379 fatal("Failed to open output file");
1381 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1382 fatal("WebM output to pipes not supported.");
1385 if (stream->config.write_webm) {
1386 stream->ebml.stream = stream->file;
1387 write_webm_file_header(&stream->ebml, cfg,
1389 stream->config.stereo_fmt,
1390 global->codec->fourcc);
1394 if (!stream->config.write_webm) {
1395 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1400 static void close_output_file(struct stream_state *stream,
1401 unsigned int fourcc) {
1402 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1404 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1408 if (stream->config.write_webm) {
1409 write_webm_file_footer(&stream->ebml);
1413 if (!stream->config.write_webm) {
1414 if (!fseek(stream->file, 0, SEEK_SET))
1415 ivf_write_file_header(stream->file, &stream->config.cfg,
1417 stream->frames_out);
1420 fclose(stream->file);
1424 static void setup_pass(struct stream_state *stream,
1425 struct VpxEncoderConfig *global,
1427 if (stream->config.stats_fn) {
1428 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1430 fatal("Failed to open statistics store");
1432 if (!stats_open_mem(&stream->stats, pass))
1433 fatal("Failed to open statistics store");
1436 #if CONFIG_FP_MB_STATS
1437 if (stream->config.fpmb_stats_fn) {
1438 if (!stats_open_file(&stream->fpmb_stats,
1439 stream->config.fpmb_stats_fn, pass))
1440 fatal("Failed to open mb statistics store");
1442 if (!stats_open_mem(&stream->fpmb_stats, pass))
1443 fatal("Failed to open mb statistics store");
1447 stream->config.cfg.g_pass = global->passes == 2
1448 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1451 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1452 #if CONFIG_FP_MB_STATS
1453 stream->config.cfg.rc_firstpass_mb_stats_in =
1454 stats_get(&stream->fpmb_stats);
1458 stream->cx_time = 0;
1460 stream->frames_out = 0;
1464 static void initialize_encoder(struct stream_state *stream,
1465 struct VpxEncoderConfig *global) {
1469 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1470 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1471 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1472 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1475 /* Construct Encoder Context */
1476 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1477 &stream->config.cfg, flags);
1478 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1480 /* Note that we bypass the vpx_codec_control wrapper macro because
1481 * we're being clever to store the control IDs in an array. Real
1482 * applications will want to make use of the enumerations directly
1484 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1485 int ctrl = stream->config.arg_ctrls[i][0];
1486 int value = stream->config.arg_ctrls[i][1];
1487 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1488 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1491 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1495 if (global->test_decode != TEST_DECODE_OFF) {
1496 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1497 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1503 static void encode_frame(struct stream_state *stream,
1504 struct VpxEncoderConfig *global,
1505 struct vpx_image *img,
1506 unsigned int frames_in) {
1507 vpx_codec_pts_t frame_start, next_frame_start;
1508 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1509 struct vpx_usec_timer timer;
1511 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1512 * global->framerate.den)
1513 / cfg->g_timebase.num / global->framerate.num;
1514 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1515 * global->framerate.den)
1516 / cfg->g_timebase.num / global->framerate.num;
1518 /* Scale if necessary */
1519 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1521 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1522 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1523 if (img->fmt != VPX_IMG_FMT_I42016) {
1524 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1529 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1530 cfg->g_w, cfg->g_h, 16);
1532 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1533 img->stride[VPX_PLANE_Y]/2,
1534 (uint16*)img->planes[VPX_PLANE_U],
1535 img->stride[VPX_PLANE_U]/2,
1536 (uint16*)img->planes[VPX_PLANE_V],
1537 img->stride[VPX_PLANE_V]/2,
1539 (uint16*)stream->img->planes[VPX_PLANE_Y],
1540 stream->img->stride[VPX_PLANE_Y]/2,
1541 (uint16*)stream->img->planes[VPX_PLANE_U],
1542 stream->img->stride[VPX_PLANE_U]/2,
1543 (uint16*)stream->img->planes[VPX_PLANE_V],
1544 stream->img->stride[VPX_PLANE_V]/2,
1545 stream->img->d_w, stream->img->d_h,
1549 stream->encoder.err = 1;
1550 ctx_exit_on_error(&stream->encoder,
1551 "Stream %d: Failed to encode frame.\n"
1552 "Scaling disabled in this configuration. \n"
1553 "To enable, configure with --enable-libyuv\n",
1559 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1560 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1561 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1566 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1567 cfg->g_w, cfg->g_h, 16);
1568 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1569 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1570 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1572 stream->img->planes[VPX_PLANE_Y],
1573 stream->img->stride[VPX_PLANE_Y],
1574 stream->img->planes[VPX_PLANE_U],
1575 stream->img->stride[VPX_PLANE_U],
1576 stream->img->planes[VPX_PLANE_V],
1577 stream->img->stride[VPX_PLANE_V],
1578 stream->img->d_w, stream->img->d_h,
1582 stream->encoder.err = 1;
1583 ctx_exit_on_error(&stream->encoder,
1584 "Stream %d: Failed to encode frame.\n"
1585 "Scaling disabled in this configuration. \n"
1586 "To enable, configure with --enable-libyuv\n",
1591 vpx_usec_timer_start(&timer);
1592 vpx_codec_encode(&stream->encoder, img, frame_start,
1593 (unsigned long)(next_frame_start - frame_start),
1594 0, global->deadline);
1595 vpx_usec_timer_mark(&timer);
1596 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1597 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1602 static void update_quantizer_histogram(struct stream_state *stream) {
1603 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1606 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1607 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1608 stream->counts[q]++;
1613 static void get_cx_data(struct stream_state *stream,
1614 struct VpxEncoderConfig *global,
1616 const vpx_codec_cx_pkt_t *pkt;
1617 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1618 vpx_codec_iter_t iter = NULL;
1621 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1622 static size_t fsize = 0;
1623 static int64_t ivf_header_pos = 0;
1625 switch (pkt->kind) {
1626 case VPX_CODEC_CX_FRAME_PKT:
1627 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1628 stream->frames_out++;
1631 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1633 update_rate_histogram(stream->rate_hist, cfg, pkt);
1635 if (stream->config.write_webm) {
1636 write_webm_block(&stream->ebml, cfg, pkt);
1639 if (!stream->config.write_webm) {
1640 if (pkt->data.frame.partition_id <= 0) {
1641 ivf_header_pos = ftello(stream->file);
1642 fsize = pkt->data.frame.sz;
1644 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1646 fsize += pkt->data.frame.sz;
1648 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1649 const int64_t currpos = ftello(stream->file);
1650 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1651 ivf_write_frame_size(stream->file, fsize);
1652 fseeko(stream->file, currpos, SEEK_SET);
1656 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1659 stream->nbytes += pkt->data.raw.sz;
1663 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1664 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1665 (unsigned int)pkt->data.frame.sz, NULL, 0);
1666 if (stream->decoder.err) {
1667 warn_or_exit_on_error(&stream->decoder,
1668 global->test_decode == TEST_DECODE_FATAL,
1669 "Failed to decode frame %d in stream %d",
1670 stream->frames_out + 1, stream->index);
1671 stream->mismatch_seen = stream->frames_out + 1;
1676 case VPX_CODEC_STATS_PKT:
1677 stream->frames_out++;
1678 stats_write(&stream->stats,
1679 pkt->data.twopass_stats.buf,
1680 pkt->data.twopass_stats.sz);
1681 stream->nbytes += pkt->data.raw.sz;
1683 #if CONFIG_FP_MB_STATS
1684 case VPX_CODEC_FPMB_STATS_PKT:
1685 stats_write(&stream->fpmb_stats,
1686 pkt->data.firstpass_mb_stats.buf,
1687 pkt->data.firstpass_mb_stats.sz);
1688 stream->nbytes += pkt->data.raw.sz;
1691 case VPX_CODEC_PSNR_PKT:
1693 if (global->show_psnr) {
1696 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1697 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1698 for (i = 0; i < 4; i++) {
1700 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1701 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1703 stream->psnr_count++;
1714 static void show_psnr(struct stream_state *stream, double peak) {
1718 if (!stream->psnr_count)
1721 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1722 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1723 (double)stream->psnr_sse_total);
1724 fprintf(stderr, " %.3f", ovpsnr);
1726 for (i = 0; i < 4; i++) {
1727 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1729 fprintf(stderr, "\n");
1733 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1734 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1737 static void test_decode(struct stream_state *stream,
1738 enum TestDecodeFatality fatal,
1739 const VpxInterface *codec) {
1740 vpx_image_t enc_img, dec_img;
1742 if (stream->mismatch_seen)
1745 /* Get the internal reference frame */
1746 if (strcmp(codec->name, "vp8") == 0) {
1747 struct vpx_ref_frame ref_enc, ref_dec;
1750 width = (stream->config.cfg.g_w + 15) & ~15;
1751 height = (stream->config.cfg.g_h + 15) & ~15;
1752 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1753 enc_img = ref_enc.img;
1754 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1755 dec_img = ref_dec.img;
1757 ref_enc.frame_type = VP8_LAST_FRAME;
1758 ref_dec.frame_type = VP8_LAST_FRAME;
1759 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1760 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1762 struct vp9_ref_frame ref_enc, ref_dec;
1766 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1767 enc_img = ref_enc.img;
1768 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1769 dec_img = ref_dec.img;
1770 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1771 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1772 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1773 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1774 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1775 enc_img.d_w, enc_img.d_h, 16);
1776 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1778 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1779 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1780 dec_img.d_w, dec_img.d_h, 16);
1781 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1786 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1787 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1789 if (!compare_img(&enc_img, &dec_img)) {
1790 int y[4], u[4], v[4];
1791 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1792 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1793 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1795 find_mismatch(&enc_img, &dec_img, y, u, v);
1798 find_mismatch(&enc_img, &dec_img, y, u, v);
1800 stream->decoder.err = 1;
1801 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1802 "Stream %d: Encode/decode mismatch on frame %d at"
1803 " Y[%d, %d] {%d/%d},"
1804 " U[%d, %d] {%d/%d},"
1805 " V[%d, %d] {%d/%d}",
1806 stream->index, stream->frames_out,
1807 y[0], y[1], y[2], y[3],
1808 u[0], u[1], u[2], u[3],
1809 v[0], v[1], v[2], v[3]);
1810 stream->mismatch_seen = stream->frames_out;
1813 vpx_img_free(&enc_img);
1814 vpx_img_free(&dec_img);
1818 static void print_time(const char *label, int64_t etl) {
1825 etl -= hours * 3600;
1830 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1831 label, hours, mins, secs);
1833 fprintf(stderr, "[%3s unknown] ", label);
1838 int main(int argc, const char **argv_) {
1841 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1842 vpx_image_t raw_shift;
1843 int allocated_raw_shift = 0;
1844 int use_16bit_internal = 0;
1845 int input_shift = 0;
1847 int frame_avail, got_data;
1849 struct VpxInputContext input;
1850 struct VpxEncoderConfig global;
1851 struct stream_state *streams = NULL;
1852 char **argv, **argi;
1853 uint64_t cx_time = 0;
1857 memset(&input, 0, sizeof(input));
1858 exec_name = argv_[0];
1863 /* Setup default input stream settings */
1864 input.framerate.numerator = 30;
1865 input.framerate.denominator = 1;
1866 input.only_i420 = 1;
1867 input.bit_depth = 0;
1869 /* First parse the global configuration values, because we want to apply
1870 * other parameters on top of the default configuration provided by the
1873 argv = argv_dup(argc - 1, argv_ + 1);
1874 parse_global_config(&global, argv);
1876 switch (global.color_type) {
1878 input.fmt = VPX_IMG_FMT_I420;
1881 input.fmt = VPX_IMG_FMT_I422;
1884 input.fmt = VPX_IMG_FMT_I444;
1887 input.fmt = VPX_IMG_FMT_I440;
1890 input.fmt = VPX_IMG_FMT_YV12;
1895 /* Now parse each stream's parameters. Using a local scope here
1896 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1899 struct stream_state *stream = NULL;
1902 stream = new_stream(&global, stream);
1906 } while (parse_stream_params(&global, stream, argv));
1909 /* Check for unrecognized options */
1910 for (argi = argv; *argi; argi++)
1911 if (argi[0][0] == '-' && argi[0][1])
1912 die("Error: Unrecognized option %s\n", *argi);
1914 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1915 &global, &stream->config.cfg););
1917 /* Handle non-option arguments */
1918 input.filename = argv[0];
1920 if (!input.filename)
1923 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1924 if (global.codec->fourcc == VP9_FOURCC)
1925 input.only_i420 = 0;
1927 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1928 int frames_in = 0, seen_frames = 0;
1929 int64_t estimated_time_left = -1;
1930 int64_t average_rate = -1;
1931 int64_t lagged_count = 0;
1933 open_input_file(&input);
1935 /* If the input file doesn't specify its w/h (raw files), try to get
1936 * the data from the first stream's configuration.
1938 if (!input.width || !input.height) {
1940 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1941 input.width = stream->config.cfg.g_w;
1942 input.height = stream->config.cfg.g_h;
1948 /* Update stream configurations from the input file's parameters */
1949 if (!input.width || !input.height)
1950 fatal("Specify stream dimensions with --width (-w) "
1951 " and --height (-h)");
1953 /* If input file does not specify bit-depth but input-bit-depth parameter
1954 * exists, assume that to be the input bit-depth. However, if the
1955 * input-bit-depth paramter does not exist, assume the input bit-depth
1956 * to be the same as the codec bit-depth.
1958 if (!input.bit_depth) {
1960 if (stream->config.cfg.g_input_bit_depth)
1961 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1963 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1964 (int)stream->config.cfg.g_bit_depth;
1966 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1969 stream->config.cfg.g_input_bit_depth = input.bit_depth;
1973 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1974 FOREACH_STREAM(validate_stream_config(stream, &global));
1976 /* Ensure that --passes and --pass are consistent. If --pass is set and
1977 * --passes=2, ensure --fpf was set.
1979 if (global.pass && global.passes == 2)
1981 if (!stream->config.stats_fn)
1982 die("Stream %d: Must specify --fpf when --pass=%d"
1983 " and --passes=2\n", stream->index, global.pass);
1988 stream->config.write_webm = 0;
1989 warn("vpxenc was compiled without WebM container support."
1990 "Producing IVF output");
1994 /* Use the frame rate from the file only if none was specified
1995 * on the command-line.
1997 if (!global.have_framerate) {
1998 global.framerate.num = input.framerate.numerator;
1999 global.framerate.den = input.framerate.denominator;
2002 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2004 /* Show configuration */
2005 if (global.verbose && pass == 0)
2006 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2008 if (pass == (global.pass ? global.pass - 1 : 0)) {
2009 if (input.file_type == FILE_TYPE_Y4M)
2010 /*The Y4M reader does its own allocation.
2011 Just initialize this here to avoid problems if we never read any
2013 memset(&raw, 0, sizeof(raw));
2015 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2017 FOREACH_STREAM(stream->rate_hist =
2018 init_rate_histogram(&stream->config.cfg,
2019 &global.framerate));
2022 FOREACH_STREAM(setup_pass(stream, &global, pass));
2023 FOREACH_STREAM(open_output_file(stream, &global));
2024 FOREACH_STREAM(initialize_encoder(stream, &global));
2026 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2027 if (strcmp(global.codec->name, "vp9") == 0) {
2028 // Check to see if at least one stream uses 16 bit internal.
2029 // Currently assume that the bit_depths for all streams using
2030 // highbitdepth are the same.
2032 if (stream->config.use_16bit_internal) {
2033 use_16bit_internal = 1;
2035 if (stream->config.cfg.g_profile == 0) {
2038 input_shift = (int)stream->config.cfg.g_bit_depth -
2039 stream->config.cfg.g_input_bit_depth;
2048 while (frame_avail || got_data) {
2049 struct vpx_usec_timer timer;
2051 if (!global.limit || frames_in < global.limit) {
2052 frame_avail = read_frame(&input, &raw);
2056 seen_frames = frames_in > global.skip_frames ?
2057 frames_in - global.skip_frames : 0;
2059 if (!global.quiet) {
2060 float fps = usec_to_fps(cx_time, seen_frames);
2061 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2063 if (stream_cnt == 1)
2065 "frame %4d/%-4d %7"PRId64"B ",
2066 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2068 fprintf(stderr, "frame %4d ", frames_in);
2070 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2071 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2072 cx_time > 9999999 ? "ms" : "us",
2073 fps >= 1.0 ? fps : fps * 60,
2074 fps >= 1.0 ? "fps" : "fpm");
2075 print_time("ETA", estimated_time_left);
2081 if (frames_in > global.skip_frames) {
2082 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2083 vpx_image_t *frame_to_encode;
2084 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2085 assert(use_16bit_internal);
2086 // Input bit depth and stream bit depth do not match, so up
2087 // shift frame to stream bit depth
2088 if (!allocated_raw_shift) {
2089 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2090 input.width, input.height, 32);
2091 allocated_raw_shift = 1;
2093 vpx_img_upshift(&raw_shift, &raw, input_shift);
2094 frame_to_encode = &raw_shift;
2096 frame_to_encode = &raw;
2098 vpx_usec_timer_start(&timer);
2099 if (use_16bit_internal) {
2100 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2102 if (stream->config.use_16bit_internal)
2103 encode_frame(stream, &global,
2104 frame_avail ? frame_to_encode : NULL,
2110 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2111 FOREACH_STREAM(encode_frame(stream, &global,
2112 frame_avail ? frame_to_encode : NULL,
2116 vpx_usec_timer_start(&timer);
2117 FOREACH_STREAM(encode_frame(stream, &global,
2118 frame_avail ? &raw : NULL,
2121 vpx_usec_timer_mark(&timer);
2122 cx_time += vpx_usec_timer_elapsed(&timer);
2124 FOREACH_STREAM(update_quantizer_histogram(stream));
2127 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2129 if (!got_data && input.length && streams != NULL &&
2130 !streams->frames_out) {
2131 lagged_count = global.limit ? seen_frames : ftello(input.file);
2132 } else if (input.length) {
2137 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2139 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2140 remaining = 1000 * (global.limit - global.skip_frames
2141 - seen_frames + lagged_count);
2143 const int64_t input_pos = ftello(input.file);
2144 const int64_t input_pos_lagged = input_pos - lagged_count;
2145 const int64_t limit = input.length;
2147 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2148 remaining = limit - input_pos + lagged_count;
2151 average_rate = (average_rate <= 0)
2153 : (average_rate * 7 + rate) / 8;
2154 estimated_time_left = average_rate ? remaining / average_rate : -1;
2157 if (got_data && global.test_decode != TEST_DECODE_OFF)
2158 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2163 fprintf(stderr, "\033[K");
2167 fprintf(stderr, "\n");
2169 if (!global.quiet) {
2170 FOREACH_STREAM(fprintf(stderr,
2171 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2172 " %7"PRId64" %s (%.2f fps)\033[K\n",
2174 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2175 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2176 seen_frames ? (int64_t)stream->nbytes * 8 *
2177 (int64_t)global.framerate.num / global.framerate.den /
2179 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2180 stream->cx_time > 9999999 ? "ms" : "us",
2181 usec_to_fps(stream->cx_time, seen_frames)));
2184 if (global.show_psnr) {
2185 if (global.codec->fourcc == VP9_FOURCC) {
2187 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2189 FOREACH_STREAM(show_psnr(stream, 255.0));
2193 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2195 if (global.test_decode != TEST_DECODE_OFF) {
2196 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2199 close_input_file(&input);
2201 if (global.test_decode == TEST_DECODE_FATAL) {
2202 FOREACH_STREAM(res |= stream->mismatch_seen);
2204 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2206 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2208 #if CONFIG_FP_MB_STATS
2209 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2216 if (global.show_q_hist_buckets)
2217 FOREACH_STREAM(show_q_histogram(stream->counts,
2218 global.show_q_hist_buckets));
2220 if (global.show_rate_hist_buckets)
2221 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2222 &stream->config.cfg,
2223 global.show_rate_hist_buckets));
2224 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2226 #if CONFIG_INTERNAL_STATS
2227 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2228 * to match some existing utilities.
2230 if (!(global.pass == 1 && global.passes == 2))
2232 FILE *f = fopen("opsnr.stt", "a");
2233 if (stream->mismatch_seen) {
2234 fprintf(f, "First mismatch occurred in frame %d\n",
2235 stream->mismatch_seen);
2237 fprintf(f, "No mismatch detected in recon buffers\n");
2243 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2244 if (allocated_raw_shift)
2245 vpx_img_free(&raw_shift);
2250 return res ? EXIT_FAILURE : EXIT_SUCCESS;