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_ivf = ARG_DEF(
187 NULL, "ivf", 0, "Output IVF (default is WebM if WebM IO is enabled)");
188 static const arg_def_t out_part = ARG_DEF(
189 "P", "output-partitions", 0,
190 "Makes encoder output partitions. Requires IVF output!");
191 static const arg_def_t q_hist_n = ARG_DEF(
192 NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
193 static const arg_def_t rate_hist_n = ARG_DEF(
194 NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
195 static const arg_def_t disable_warnings = ARG_DEF(
196 NULL, "disable-warnings", 0,
197 "Disable warnings about potentially incorrect encode settings.");
198 static const arg_def_t disable_warning_prompt = ARG_DEF(
199 "y", "disable-warning-prompt", 0,
200 "Display warnings, but do not prompt user to continue.");
202 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
203 static const arg_def_t test16bitinternalarg = ARG_DEF(
204 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
207 static const arg_def_t *main_args[] = {
209 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
210 &deadline, &best_dl, &good_dl, &rt_dl,
211 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
212 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
216 static const arg_def_t usage = ARG_DEF(
217 "u", "usage", 1, "Usage profile number to use");
218 static const arg_def_t threads = ARG_DEF(
219 "t", "threads", 1, "Max number of threads to use");
220 static const arg_def_t profile = ARG_DEF(
221 NULL, "profile", 1, "Bitstream profile number to use");
222 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
223 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
225 static const struct arg_enum_list stereo_mode_enum[] = {
226 {"mono", STEREO_FORMAT_MONO},
227 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
228 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
229 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
230 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
233 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
234 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
236 static const arg_def_t timebase = ARG_DEF(
237 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
238 static const arg_def_t error_resilient = ARG_DEF(
239 NULL, "error-resilient", 1, "Enable error resiliency features");
240 static const arg_def_t lag_in_frames = ARG_DEF(
241 NULL, "lag-in-frames", 1, "Max number of frames to lag");
243 static const arg_def_t *global_args[] = {
244 &use_yv12, &use_i420, &use_i422, &use_i444, &use_i440,
245 &usage, &threads, &profile,
250 &timebase, &framerate,
252 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
253 &test16bitinternalarg,
258 static const arg_def_t dropframe_thresh = ARG_DEF(
259 NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
260 static const arg_def_t resize_allowed = ARG_DEF(
261 NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
262 static const arg_def_t resize_width = ARG_DEF(
263 NULL, "resize-width", 1, "Width of encoded frame");
264 static const arg_def_t resize_height = ARG_DEF(
265 NULL, "resize-height", 1, "Height of encoded frame");
266 static const arg_def_t resize_up_thresh = ARG_DEF(
267 NULL, "resize-up", 1, "Upscale threshold (buf %)");
268 static const arg_def_t resize_down_thresh = ARG_DEF(
269 NULL, "resize-down", 1, "Downscale threshold (buf %)");
270 static const struct arg_enum_list end_usage_enum[] = {
277 static const arg_def_t end_usage = ARG_DEF_ENUM(
278 NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
279 static const arg_def_t target_bitrate = ARG_DEF(
280 NULL, "target-bitrate", 1, "Bitrate (kbps)");
281 static const arg_def_t min_quantizer = ARG_DEF(
282 NULL, "min-q", 1, "Minimum (best) quantizer");
283 static const arg_def_t max_quantizer = ARG_DEF(
284 NULL, "max-q", 1, "Maximum (worst) quantizer");
285 static const arg_def_t undershoot_pct = ARG_DEF(
286 NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
287 static const arg_def_t overshoot_pct = ARG_DEF(
288 NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
289 static const arg_def_t buf_sz = ARG_DEF(
290 NULL, "buf-sz", 1, "Client buffer size (ms)");
291 static const arg_def_t buf_initial_sz = ARG_DEF(
292 NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
293 static const arg_def_t buf_optimal_sz = ARG_DEF(
294 NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
295 static const arg_def_t *rc_args[] = {
296 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
297 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
298 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
299 &buf_initial_sz, &buf_optimal_sz, NULL
303 static const arg_def_t bias_pct = ARG_DEF(
304 NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
305 static const arg_def_t minsection_pct = ARG_DEF(
306 NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
307 static const arg_def_t maxsection_pct = ARG_DEF(
308 NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
309 static const arg_def_t *rc_twopass_args[] = {
310 &bias_pct, &minsection_pct, &maxsection_pct, NULL
314 static const arg_def_t kf_min_dist = ARG_DEF(
315 NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
316 static const arg_def_t kf_max_dist = ARG_DEF(
317 NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
318 static const arg_def_t kf_disabled = ARG_DEF(
319 NULL, "disable-kf", 0, "Disable keyframe placement");
320 static const arg_def_t *kf_args[] = {
321 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
325 static const arg_def_t noise_sens = ARG_DEF(
326 NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
327 static const arg_def_t sharpness = ARG_DEF(
328 NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
329 static const arg_def_t static_thresh = ARG_DEF(
330 NULL, "static-thresh", 1, "Motion detection threshold");
331 static const arg_def_t cpu_used = ARG_DEF(
332 NULL, "cpu-used", 1, "CPU Used (-16..16)");
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)");
352 static const arg_def_t max_inter_rate_pct = ARG_DEF(
353 NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
355 #if CONFIG_VP8_ENCODER
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 *vp8_args[] = {
359 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
360 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
361 &tune_ssim, &cq_level, &max_intra_rate_pct,
364 static const int vp8_arg_ctrl_map[] = {
365 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
366 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
367 VP8E_SET_TOKEN_PARTITIONS,
368 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
369 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
374 #if CONFIG_VP9_ENCODER
375 static const arg_def_t tile_cols = ARG_DEF(
376 NULL, "tile-columns", 1, "Number of tile columns to use, log2");
377 static const arg_def_t tile_rows = ARG_DEF(
378 NULL, "tile-rows", 1, "Number of tile rows to use, log2");
379 static const arg_def_t lossless = ARG_DEF(
380 NULL, "lossless", 1, "Lossless mode");
381 static const arg_def_t frame_parallel_decoding = ARG_DEF(
382 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
383 static const arg_def_t aq_mode = ARG_DEF(
385 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
386 "3: cyclic refresh)");
387 static const arg_def_t frame_periodic_boost = ARG_DEF(
388 NULL, "frame-boost", 1,
389 "Enable frame periodic boost (0: off (default), 1: on)");
391 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
392 static const struct arg_enum_list bitdepth_enum[] = {
399 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
401 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
403 static const arg_def_t inbitdeptharg = ARG_DEF(
404 NULL, "input-bit-depth", 1, "Bit depth of input");
407 static const struct arg_enum_list tune_content_enum[] = {
408 {"default", VP9E_CONTENT_DEFAULT},
409 {"screen", VP9E_CONTENT_SCREEN},
413 static const arg_def_t tune_content = ARG_DEF_ENUM(
414 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
416 static const arg_def_t *vp9_args[] = {
417 &cpu_used, &auto_altref, &sharpness, &static_thresh,
418 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
419 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct, &lossless,
420 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
421 &noise_sens, &tune_content,
422 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
423 &bitdeptharg, &inbitdeptharg,
427 static const int vp9_arg_ctrl_map[] = {
428 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
429 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
430 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
431 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
432 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
433 VP8E_SET_MAX_INTER_BITRATE_PCT,
434 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
435 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
436 VP9E_SET_TUNE_CONTENT,
441 static const arg_def_t *no_args[] = { NULL };
446 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
449 fprintf(stderr, "\nOptions:\n");
450 arg_show_usage(stderr, main_args);
451 fprintf(stderr, "\nEncoder Global Options:\n");
452 arg_show_usage(stderr, global_args);
453 fprintf(stderr, "\nRate Control Options:\n");
454 arg_show_usage(stderr, rc_args);
455 fprintf(stderr, "\nTwopass Rate Control Options:\n");
456 arg_show_usage(stderr, rc_twopass_args);
457 fprintf(stderr, "\nKeyframe Placement Options:\n");
458 arg_show_usage(stderr, kf_args);
459 #if CONFIG_VP8_ENCODER
460 fprintf(stderr, "\nVP8 Specific Options:\n");
461 arg_show_usage(stderr, vp8_args);
463 #if CONFIG_VP9_ENCODER
464 fprintf(stderr, "\nVP9 Specific Options:\n");
465 arg_show_usage(stderr, vp9_args);
467 fprintf(stderr, "\nStream timebase (--timebase):\n"
468 " The desired precision of timestamps in the output, expressed\n"
469 " in fractional seconds. Default is 1/1000.\n");
470 fprintf(stderr, "\nIncluded encoders:\n\n");
472 for (i = 0; i < get_vpx_encoder_count(); ++i) {
473 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
474 fprintf(stderr, " %-6s - %s\n",
475 encoder->name, vpx_codec_iface_name(encoder->codec_interface()));
481 #define mmin(a, b) ((a) < (b) ? (a) : (b))
483 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
484 static void find_mismatch_high(const vpx_image_t *const img1,
485 const vpx_image_t *const img2,
486 int yloc[4], int uloc[4], int vloc[4]) {
487 uint16_t *plane1, *plane2;
488 uint32_t stride1, stride2;
489 const uint32_t bsize = 64;
490 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
491 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
493 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
495 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
498 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
499 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
500 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
501 stride1 = img1->stride[VPX_PLANE_Y]/2;
502 stride2 = img2->stride[VPX_PLANE_Y]/2;
503 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
504 for (j = 0; match && j < img1->d_w; j += bsize) {
506 const int si = mmin(i + bsize, img1->d_h) - i;
507 const int sj = mmin(j + bsize, img1->d_w) - j;
508 for (k = 0; match && k < si; ++k) {
509 for (l = 0; match && l < sj; ++l) {
510 if (*(plane1 + (i + k) * stride1 + j + l) !=
511 *(plane2 + (i + k) * stride2 + j + l)) {
514 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
515 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
524 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
525 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
526 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
527 stride1 = img1->stride[VPX_PLANE_U]/2;
528 stride2 = img2->stride[VPX_PLANE_U]/2;
529 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
530 for (j = 0; match && j < c_w; j += bsizex) {
532 const int si = mmin(i + bsizey, c_h - i);
533 const int sj = mmin(j + bsizex, c_w - j);
534 for (k = 0; match && k < si; ++k) {
535 for (l = 0; match && l < sj; ++l) {
536 if (*(plane1 + (i + k) * stride1 + j + l) !=
537 *(plane2 + (i + k) * stride2 + j + l)) {
540 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
541 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
550 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
551 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
552 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
553 stride1 = img1->stride[VPX_PLANE_V]/2;
554 stride2 = img2->stride[VPX_PLANE_V]/2;
555 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
556 for (j = 0; match && j < c_w; j += bsizex) {
558 const int si = mmin(i + bsizey, c_h - i);
559 const int sj = mmin(j + bsizex, c_w - j);
560 for (k = 0; match && k < si; ++k) {
561 for (l = 0; match && l < sj; ++l) {
562 if (*(plane1 + (i + k) * stride1 + j + l) !=
563 *(plane2 + (i + k) * stride2 + j + l)) {
566 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
567 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
578 static void find_mismatch(const vpx_image_t *const img1,
579 const vpx_image_t *const img2,
580 int yloc[4], int uloc[4], int vloc[4]) {
581 const uint32_t bsize = 64;
582 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
583 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
585 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
587 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
590 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
591 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
592 for (j = 0; match && j < img1->d_w; j += bsize) {
594 const int si = mmin(i + bsize, img1->d_h) - i;
595 const int sj = mmin(j + bsize, img1->d_w) - j;
596 for (k = 0; match && k < si; ++k) {
597 for (l = 0; match && l < sj; ++l) {
598 if (*(img1->planes[VPX_PLANE_Y] +
599 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
600 *(img2->planes[VPX_PLANE_Y] +
601 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
604 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
605 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
606 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
607 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
616 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
617 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
618 for (j = 0; match && j < c_w; j += bsizex) {
620 const int si = mmin(i + bsizey, c_h - i);
621 const int sj = mmin(j + bsizex, c_w - j);
622 for (k = 0; match && k < si; ++k) {
623 for (l = 0; match && l < sj; ++l) {
624 if (*(img1->planes[VPX_PLANE_U] +
625 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
626 *(img2->planes[VPX_PLANE_U] +
627 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
630 uloc[2] = *(img1->planes[VPX_PLANE_U] +
631 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
632 uloc[3] = *(img2->planes[VPX_PLANE_U] +
633 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
641 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
642 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
643 for (j = 0; match && j < c_w; j += bsizex) {
645 const int si = mmin(i + bsizey, c_h - i);
646 const int sj = mmin(j + bsizex, c_w - j);
647 for (k = 0; match && k < si; ++k) {
648 for (l = 0; match && l < sj; ++l) {
649 if (*(img1->planes[VPX_PLANE_V] +
650 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
651 *(img2->planes[VPX_PLANE_V] +
652 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
655 vloc[2] = *(img1->planes[VPX_PLANE_V] +
656 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
657 vloc[3] = *(img2->planes[VPX_PLANE_V] +
658 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
668 static int compare_img(const vpx_image_t *const img1,
669 const vpx_image_t *const img2) {
670 uint32_t l_w = img1->d_w;
672 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
674 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
678 match &= (img1->fmt == img2->fmt);
679 match &= (img1->d_w == img2->d_w);
680 match &= (img1->d_h == img2->d_h);
681 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
682 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
688 for (i = 0; i < img1->d_h; ++i)
689 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
690 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
693 for (i = 0; i < c_h; ++i)
694 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
695 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
698 for (i = 0; i < c_h; ++i)
699 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
700 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
707 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
708 #define MAX(x,y) ((x)>(y)?(x):(y))
709 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
710 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
711 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
712 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
714 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
715 NELEMENTS(vp9_arg_ctrl_map))
719 typedef int stereo_format_t;
720 struct EbmlGlobal { int debug; };
723 /* Per-stream configuration */
724 struct stream_config {
725 struct vpx_codec_enc_cfg cfg;
727 const char *stats_fn;
728 #if CONFIG_FP_MB_STATS
729 const char *fpmb_stats_fn;
731 stereo_format_t stereo_fmt;
732 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
735 int have_kf_max_dist;
736 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
737 // whether to use 16bit internal buffers
738 int use_16bit_internal;
743 struct stream_state {
745 struct stream_state *next;
746 struct stream_config config;
748 struct rate_hist *rate_hist;
749 struct EbmlGlobal ebml;
750 uint64_t psnr_sse_total;
751 uint64_t psnr_samples_total;
752 double psnr_totals[4];
755 vpx_codec_ctx_t encoder;
756 unsigned int frames_out;
760 #if CONFIG_FP_MB_STATS
761 stats_io_t fpmb_stats;
763 struct vpx_image *img;
764 vpx_codec_ctx_t decoder;
769 void validate_positive_rational(const char *msg,
770 struct vpx_rational *rat) {
777 die("Error: %s must be positive\n", msg);
780 die("Error: %s has zero denominator\n", msg);
784 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
788 /* Initialize default parameters */
789 memset(global, 0, sizeof(*global));
790 global->codec = get_vpx_encoder_by_index(0);
792 global->color_type = I420;
793 /* Assign default deadline to good quality */
794 global->deadline = VPX_DL_GOOD_QUALITY;
796 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
799 if (arg_match(&arg, &codecarg, argi)) {
800 global->codec = get_vpx_encoder_by_name(arg.val);
802 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
803 } else if (arg_match(&arg, &passes, argi)) {
804 global->passes = arg_parse_uint(&arg);
806 if (global->passes < 1 || global->passes > 2)
807 die("Error: Invalid number of passes (%d)\n", global->passes);
808 } else if (arg_match(&arg, &pass_arg, argi)) {
809 global->pass = arg_parse_uint(&arg);
811 if (global->pass < 1 || global->pass > 2)
812 die("Error: Invalid pass selected (%d)\n",
814 } else if (arg_match(&arg, &usage, argi))
815 global->usage = arg_parse_uint(&arg);
816 else if (arg_match(&arg, &deadline, argi))
817 global->deadline = arg_parse_uint(&arg);
818 else if (arg_match(&arg, &best_dl, argi))
819 global->deadline = VPX_DL_BEST_QUALITY;
820 else if (arg_match(&arg, &good_dl, argi))
821 global->deadline = VPX_DL_GOOD_QUALITY;
822 else if (arg_match(&arg, &rt_dl, argi))
823 global->deadline = VPX_DL_REALTIME;
824 else if (arg_match(&arg, &use_yv12, argi))
825 global->color_type = YV12;
826 else if (arg_match(&arg, &use_i420, argi))
827 global->color_type = I420;
828 else if (arg_match(&arg, &use_i422, argi))
829 global->color_type = I422;
830 else if (arg_match(&arg, &use_i444, argi))
831 global->color_type = I444;
832 else if (arg_match(&arg, &use_i440, argi))
833 global->color_type = I440;
834 else if (arg_match(&arg, &quietarg, argi))
836 else if (arg_match(&arg, &verbosearg, argi))
838 else if (arg_match(&arg, &limit, argi))
839 global->limit = arg_parse_uint(&arg);
840 else if (arg_match(&arg, &skip, argi))
841 global->skip_frames = arg_parse_uint(&arg);
842 else if (arg_match(&arg, &psnrarg, argi))
843 global->show_psnr = 1;
844 else if (arg_match(&arg, &recontest, argi))
845 global->test_decode = arg_parse_enum_or_int(&arg);
846 else if (arg_match(&arg, &framerate, argi)) {
847 global->framerate = arg_parse_rational(&arg);
848 validate_positive_rational(arg.name, &global->framerate);
849 global->have_framerate = 1;
850 } else if (arg_match(&arg, &out_part, argi))
851 global->out_part = 1;
852 else if (arg_match(&arg, &debugmode, argi))
854 else if (arg_match(&arg, &q_hist_n, argi))
855 global->show_q_hist_buckets = arg_parse_uint(&arg);
856 else if (arg_match(&arg, &rate_hist_n, argi))
857 global->show_rate_hist_buckets = arg_parse_uint(&arg);
858 else if (arg_match(&arg, &disable_warnings, argi))
859 global->disable_warnings = 1;
860 else if (arg_match(&arg, &disable_warning_prompt, argi))
861 global->disable_warning_prompt = 1;
867 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
868 if (global->pass > global->passes) {
869 warn("Assuming --pass=%d implies --passes=%d\n",
870 global->pass, global->pass);
871 global->passes = global->pass;
874 /* Validate global config */
875 if (global->passes == 0) {
876 #if CONFIG_VP9_ENCODER
877 // Make default VP9 passes = 2 until there is a better quality 1-pass
879 if (global->codec != NULL && global->codec->name != NULL)
880 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
881 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
887 if (global->deadline == VPX_DL_REALTIME &&
888 global->passes > 1) {
889 warn("Enforcing one-pass encoding in realtime mode\n");
895 void open_input_file(struct VpxInputContext *input) {
896 /* Parse certain options from the input file, if possible */
897 input->file = strcmp(input->filename, "-")
898 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
901 fatal("Failed to open input file");
903 if (!fseeko(input->file, 0, SEEK_END)) {
904 /* Input file is seekable. Figure out how long it is, so we can get
907 input->length = ftello(input->file);
911 /* For RAW input sources, these bytes will applied on the first frame
914 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
915 input->detect.position = 0;
917 if (input->detect.buf_read == 4
918 && file_is_y4m(input->detect.buf)) {
919 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
920 input->only_i420) >= 0) {
921 input->file_type = FILE_TYPE_Y4M;
922 input->width = input->y4m.pic_w;
923 input->height = input->y4m.pic_h;
924 input->framerate.numerator = input->y4m.fps_n;
925 input->framerate.denominator = input->y4m.fps_d;
926 input->fmt = input->y4m.vpx_fmt;
927 input->bit_depth = input->y4m.bit_depth;
929 fatal("Unsupported Y4M stream.");
930 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
931 fatal("IVF is not supported as input.");
933 input->file_type = FILE_TYPE_RAW;
938 static void close_input_file(struct VpxInputContext *input) {
940 if (input->file_type == FILE_TYPE_Y4M)
941 y4m_input_close(&input->y4m);
944 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
945 struct stream_state *prev) {
946 struct stream_state *stream;
948 stream = calloc(1, sizeof(*stream));
949 if (stream == NULL) {
950 fatal("Failed to allocate new stream.");
954 memcpy(stream, prev, sizeof(*stream));
960 /* Populate encoder configuration */
961 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
965 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
967 /* Change the default timebase to a high enough value so that the
968 * encoder will always create strictly increasing timestamps.
970 stream->config.cfg.g_timebase.den = 1000;
972 /* Never use the library's default resolution, require it be parsed
973 * from the file or set on the command line.
975 stream->config.cfg.g_w = 0;
976 stream->config.cfg.g_h = 0;
978 /* Initialize remaining stream parameters */
979 stream->config.write_webm = 1;
981 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
982 stream->ebml.last_pts_ns = -1;
983 stream->ebml.writer = NULL;
984 stream->ebml.segment = NULL;
987 /* Allows removal of the application version from the EBML tags */
988 stream->ebml.debug = global->debug;
990 /* Default lag_in_frames is 0 in realtime mode */
991 if (global->deadline == VPX_DL_REALTIME)
992 stream->config.cfg.g_lag_in_frames = 0;
995 /* Output files must be specified for each stream */
996 stream->config.out_fn = NULL;
1003 static int parse_stream_params(struct VpxEncoderConfig *global,
1004 struct stream_state *stream,
1006 char **argi, **argj;
1008 static const arg_def_t **ctrl_args = no_args;
1009 static const int *ctrl_args_map = NULL;
1010 struct stream_config *config = &stream->config;
1011 int eos_mark_found = 0;
1012 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1013 int test_16bit_internal = 0;
1016 // Handle codec specific options
1018 #if CONFIG_VP8_ENCODER
1019 } else if (strcmp(global->codec->name, "vp8") == 0) {
1020 ctrl_args = vp8_args;
1021 ctrl_args_map = vp8_arg_ctrl_map;
1023 #if CONFIG_VP9_ENCODER
1024 } else if (strcmp(global->codec->name, "vp9") == 0) {
1025 ctrl_args = vp9_args;
1026 ctrl_args_map = vp9_arg_ctrl_map;
1030 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1033 /* Once we've found an end-of-stream marker (--) we want to continue
1034 * shifting arguments but not consuming them.
1036 if (eos_mark_found) {
1039 } else if (!strcmp(*argj, "--")) {
1045 } else if (arg_match(&arg, &outputfile, argi)) {
1046 config->out_fn = arg.val;
1047 } else if (arg_match(&arg, &fpf_name, argi)) {
1048 config->stats_fn = arg.val;
1049 #if CONFIG_FP_MB_STATS
1050 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1051 config->fpmb_stats_fn = arg.val;
1053 } else if (arg_match(&arg, &use_ivf, argi)) {
1054 config->write_webm = 0;
1055 } else if (arg_match(&arg, &threads, argi)) {
1056 config->cfg.g_threads = arg_parse_uint(&arg);
1057 } else if (arg_match(&arg, &profile, argi)) {
1058 config->cfg.g_profile = arg_parse_uint(&arg);
1059 } else if (arg_match(&arg, &width, argi)) {
1060 config->cfg.g_w = arg_parse_uint(&arg);
1061 } else if (arg_match(&arg, &height, argi)) {
1062 config->cfg.g_h = arg_parse_uint(&arg);
1063 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1064 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1065 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1066 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1067 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1070 } else if (arg_match(&arg, &stereo_mode, argi)) {
1071 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1073 } else if (arg_match(&arg, &timebase, argi)) {
1074 config->cfg.g_timebase = arg_parse_rational(&arg);
1075 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1076 } else if (arg_match(&arg, &error_resilient, argi)) {
1077 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1078 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1079 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1080 if (global->deadline == VPX_DL_REALTIME &&
1081 config->cfg.g_lag_in_frames != 0) {
1082 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1083 config->cfg.g_lag_in_frames = 0;
1085 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1086 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1087 } else if (arg_match(&arg, &resize_allowed, argi)) {
1088 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1089 } else if (arg_match(&arg, &resize_width, argi)) {
1090 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1091 } else if (arg_match(&arg, &resize_height, argi)) {
1092 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1093 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1094 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1095 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1096 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1097 } else if (arg_match(&arg, &end_usage, argi)) {
1098 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1099 } else if (arg_match(&arg, &target_bitrate, argi)) {
1100 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1101 } else if (arg_match(&arg, &min_quantizer, argi)) {
1102 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1103 } else if (arg_match(&arg, &max_quantizer, argi)) {
1104 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1105 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1106 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1107 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1108 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1109 } else if (arg_match(&arg, &buf_sz, argi)) {
1110 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1111 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1112 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1113 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1114 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1115 } else if (arg_match(&arg, &bias_pct, argi)) {
1116 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1117 if (global->passes < 2)
1118 warn("option %s ignored in one-pass mode.\n", arg.name);
1119 } else if (arg_match(&arg, &minsection_pct, argi)) {
1120 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1122 if (global->passes < 2)
1123 warn("option %s ignored in one-pass mode.\n", arg.name);
1124 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1125 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1127 if (global->passes < 2)
1128 warn("option %s ignored in one-pass mode.\n", arg.name);
1129 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1130 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1131 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1132 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1133 config->have_kf_max_dist = 1;
1134 } else if (arg_match(&arg, &kf_disabled, argi)) {
1135 config->cfg.kf_mode = VPX_KF_DISABLED;
1136 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1137 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1138 if (strcmp(global->codec->name, "vp9") == 0) {
1139 test_16bit_internal = 1;
1144 for (i = 0; ctrl_args[i]; i++) {
1145 if (arg_match(&arg, ctrl_args[i], argi)) {
1149 /* Point either to the next free element or the first
1150 * instance of this control.
1152 for (j = 0; j < config->arg_ctrl_cnt; j++)
1153 if (ctrl_args_map != NULL &&
1154 config->arg_ctrls[j][0] == ctrl_args_map[i])
1158 assert(j < (int)ARG_CTRL_CNT_MAX);
1159 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1160 config->arg_ctrls[j][0] = ctrl_args_map[i];
1161 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1162 if (j == config->arg_ctrl_cnt)
1163 config->arg_ctrl_cnt++;
1171 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1172 if (strcmp(global->codec->name, "vp9") == 0) {
1173 config->use_16bit_internal = test_16bit_internal |
1174 (config->cfg.g_profile > 1);
1177 return eos_mark_found;
1181 #define FOREACH_STREAM(func) \
1183 struct stream_state *stream; \
1184 for (stream = streams; stream; stream = stream->next) { \
1190 static void validate_stream_config(const struct stream_state *stream,
1191 const struct VpxEncoderConfig *global) {
1192 const struct stream_state *streami;
1194 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1195 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1196 " and --height (-h)", stream->index);
1198 // Check that the codec bit depth is greater than the input bit depth.
1199 if (stream->config.cfg.g_input_bit_depth >
1200 (unsigned int)stream->config.cfg.g_bit_depth) {
1201 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1202 stream->index, (int)stream->config.cfg.g_bit_depth,
1203 stream->config.cfg.g_input_bit_depth);
1206 for (streami = stream; streami; streami = streami->next) {
1207 /* All streams require output files */
1208 if (!streami->config.out_fn)
1209 fatal("Stream %d: Output file is required (specify with -o)",
1212 /* Check for two streams outputting to the same file */
1213 if (streami != stream) {
1214 const char *a = stream->config.out_fn;
1215 const char *b = streami->config.out_fn;
1216 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1217 fatal("Stream %d: duplicate output file (from stream %d)",
1218 streami->index, stream->index);
1221 /* Check for two streams sharing a stats file. */
1222 if (streami != stream) {
1223 const char *a = stream->config.stats_fn;
1224 const char *b = streami->config.stats_fn;
1225 if (a && b && !strcmp(a, b))
1226 fatal("Stream %d: duplicate stats file (from stream %d)",
1227 streami->index, stream->index);
1230 #if CONFIG_FP_MB_STATS
1231 /* Check for two streams sharing a mb stats file. */
1232 if (streami != stream) {
1233 const char *a = stream->config.fpmb_stats_fn;
1234 const char *b = streami->config.fpmb_stats_fn;
1235 if (a && b && !strcmp(a, b))
1236 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1237 streami->index, stream->index);
1244 static void set_stream_dimensions(struct stream_state *stream,
1247 if (!stream->config.cfg.g_w) {
1248 if (!stream->config.cfg.g_h)
1249 stream->config.cfg.g_w = w;
1251 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1253 if (!stream->config.cfg.g_h) {
1254 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1259 static void set_default_kf_interval(struct stream_state *stream,
1260 struct VpxEncoderConfig *global) {
1261 /* Use a max keyframe interval of 5 seconds, if none was
1262 * specified on the command line.
1264 if (!stream->config.have_kf_max_dist) {
1265 double framerate = (double)global->framerate.num / global->framerate.den;
1266 if (framerate > 0.0)
1267 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1271 static const char* file_type_to_string(enum VideoFileType t) {
1273 case FILE_TYPE_RAW: return "RAW";
1274 case FILE_TYPE_Y4M: return "Y4M";
1275 default: return "Other";
1279 static const char* image_format_to_string(vpx_img_fmt_t f) {
1281 case VPX_IMG_FMT_I420: return "I420";
1282 case VPX_IMG_FMT_I422: return "I422";
1283 case VPX_IMG_FMT_I444: return "I444";
1284 case VPX_IMG_FMT_I440: return "I440";
1285 case VPX_IMG_FMT_YV12: return "YV12";
1286 case VPX_IMG_FMT_I42016: return "I42016";
1287 case VPX_IMG_FMT_I42216: return "I42216";
1288 case VPX_IMG_FMT_I44416: return "I44416";
1289 case VPX_IMG_FMT_I44016: return "I44016";
1290 default: return "Other";
1294 static void show_stream_config(struct stream_state *stream,
1295 struct VpxEncoderConfig *global,
1296 struct VpxInputContext *input) {
1298 #define SHOW(field) \
1299 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1301 if (stream->index == 0) {
1302 fprintf(stderr, "Codec: %s\n",
1303 vpx_codec_iface_name(global->codec->codec_interface()));
1304 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1306 file_type_to_string(input->file_type),
1307 image_format_to_string(input->fmt));
1309 if (stream->next || stream->index)
1310 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1311 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1312 fprintf(stderr, "Encoder parameters:\n");
1320 SHOW(g_input_bit_depth);
1321 SHOW(g_timebase.num);
1322 SHOW(g_timebase.den);
1323 SHOW(g_error_resilient);
1325 SHOW(g_lag_in_frames);
1326 SHOW(rc_dropframe_thresh);
1327 SHOW(rc_resize_allowed);
1328 SHOW(rc_scaled_width);
1329 SHOW(rc_scaled_height);
1330 SHOW(rc_resize_up_thresh);
1331 SHOW(rc_resize_down_thresh);
1333 SHOW(rc_target_bitrate);
1334 SHOW(rc_min_quantizer);
1335 SHOW(rc_max_quantizer);
1336 SHOW(rc_undershoot_pct);
1337 SHOW(rc_overshoot_pct);
1339 SHOW(rc_buf_initial_sz);
1340 SHOW(rc_buf_optimal_sz);
1341 SHOW(rc_2pass_vbr_bias_pct);
1342 SHOW(rc_2pass_vbr_minsection_pct);
1343 SHOW(rc_2pass_vbr_maxsection_pct);
1350 static void open_output_file(struct stream_state *stream,
1351 struct VpxEncoderConfig *global) {
1352 const char *fn = stream->config.out_fn;
1353 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1355 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1358 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1361 fatal("Failed to open output file");
1363 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1364 fatal("WebM output to pipes not supported.");
1367 if (stream->config.write_webm) {
1368 stream->ebml.stream = stream->file;
1369 write_webm_file_header(&stream->ebml, cfg,
1371 stream->config.stereo_fmt,
1372 global->codec->fourcc);
1376 if (!stream->config.write_webm) {
1377 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1382 static void close_output_file(struct stream_state *stream,
1383 unsigned int fourcc) {
1384 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1386 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1390 if (stream->config.write_webm) {
1391 write_webm_file_footer(&stream->ebml);
1395 if (!stream->config.write_webm) {
1396 if (!fseek(stream->file, 0, SEEK_SET))
1397 ivf_write_file_header(stream->file, &stream->config.cfg,
1399 stream->frames_out);
1402 fclose(stream->file);
1406 static void setup_pass(struct stream_state *stream,
1407 struct VpxEncoderConfig *global,
1409 if (stream->config.stats_fn) {
1410 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1412 fatal("Failed to open statistics store");
1414 if (!stats_open_mem(&stream->stats, pass))
1415 fatal("Failed to open statistics store");
1418 #if CONFIG_FP_MB_STATS
1419 if (stream->config.fpmb_stats_fn) {
1420 if (!stats_open_file(&stream->fpmb_stats,
1421 stream->config.fpmb_stats_fn, pass))
1422 fatal("Failed to open mb statistics store");
1424 if (!stats_open_mem(&stream->fpmb_stats, pass))
1425 fatal("Failed to open mb statistics store");
1429 stream->config.cfg.g_pass = global->passes == 2
1430 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1433 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1434 #if CONFIG_FP_MB_STATS
1435 stream->config.cfg.rc_firstpass_mb_stats_in =
1436 stats_get(&stream->fpmb_stats);
1440 stream->cx_time = 0;
1442 stream->frames_out = 0;
1446 static void initialize_encoder(struct stream_state *stream,
1447 struct VpxEncoderConfig *global) {
1451 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1452 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1453 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1454 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1457 /* Construct Encoder Context */
1458 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1459 &stream->config.cfg, flags);
1460 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1462 /* Note that we bypass the vpx_codec_control wrapper macro because
1463 * we're being clever to store the control IDs in an array. Real
1464 * applications will want to make use of the enumerations directly
1466 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1467 int ctrl = stream->config.arg_ctrls[i][0];
1468 int value = stream->config.arg_ctrls[i][1];
1469 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1470 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1473 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1477 if (global->test_decode != TEST_DECODE_OFF) {
1478 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1479 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1485 static void encode_frame(struct stream_state *stream,
1486 struct VpxEncoderConfig *global,
1487 struct vpx_image *img,
1488 unsigned int frames_in) {
1489 vpx_codec_pts_t frame_start, next_frame_start;
1490 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1491 struct vpx_usec_timer timer;
1493 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1494 * global->framerate.den)
1495 / cfg->g_timebase.num / global->framerate.num;
1496 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1497 * global->framerate.den)
1498 / cfg->g_timebase.num / global->framerate.num;
1500 /* Scale if necessary */
1501 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1503 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1504 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1505 if (img->fmt != VPX_IMG_FMT_I42016) {
1506 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1511 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1512 cfg->g_w, cfg->g_h, 16);
1514 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1515 img->stride[VPX_PLANE_Y]/2,
1516 (uint16*)img->planes[VPX_PLANE_U],
1517 img->stride[VPX_PLANE_U]/2,
1518 (uint16*)img->planes[VPX_PLANE_V],
1519 img->stride[VPX_PLANE_V]/2,
1521 (uint16*)stream->img->planes[VPX_PLANE_Y],
1522 stream->img->stride[VPX_PLANE_Y]/2,
1523 (uint16*)stream->img->planes[VPX_PLANE_U],
1524 stream->img->stride[VPX_PLANE_U]/2,
1525 (uint16*)stream->img->planes[VPX_PLANE_V],
1526 stream->img->stride[VPX_PLANE_V]/2,
1527 stream->img->d_w, stream->img->d_h,
1531 stream->encoder.err = 1;
1532 ctx_exit_on_error(&stream->encoder,
1533 "Stream %d: Failed to encode frame.\n"
1534 "Scaling disabled in this configuration. \n"
1535 "To enable, configure with --enable-libyuv\n",
1541 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1542 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1543 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1548 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1549 cfg->g_w, cfg->g_h, 16);
1550 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1551 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1552 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1554 stream->img->planes[VPX_PLANE_Y],
1555 stream->img->stride[VPX_PLANE_Y],
1556 stream->img->planes[VPX_PLANE_U],
1557 stream->img->stride[VPX_PLANE_U],
1558 stream->img->planes[VPX_PLANE_V],
1559 stream->img->stride[VPX_PLANE_V],
1560 stream->img->d_w, stream->img->d_h,
1564 stream->encoder.err = 1;
1565 ctx_exit_on_error(&stream->encoder,
1566 "Stream %d: Failed to encode frame.\n"
1567 "Scaling disabled in this configuration. \n"
1568 "To enable, configure with --enable-libyuv\n",
1573 vpx_usec_timer_start(&timer);
1574 vpx_codec_encode(&stream->encoder, img, frame_start,
1575 (unsigned long)(next_frame_start - frame_start),
1576 0, global->deadline);
1577 vpx_usec_timer_mark(&timer);
1578 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1579 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1584 static void update_quantizer_histogram(struct stream_state *stream) {
1585 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1588 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1589 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1590 stream->counts[q]++;
1595 static void get_cx_data(struct stream_state *stream,
1596 struct VpxEncoderConfig *global,
1598 const vpx_codec_cx_pkt_t *pkt;
1599 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1600 vpx_codec_iter_t iter = NULL;
1603 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1604 static size_t fsize = 0;
1605 static int64_t ivf_header_pos = 0;
1607 switch (pkt->kind) {
1608 case VPX_CODEC_CX_FRAME_PKT:
1609 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1610 stream->frames_out++;
1613 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1615 update_rate_histogram(stream->rate_hist, cfg, pkt);
1617 if (stream->config.write_webm) {
1618 write_webm_block(&stream->ebml, cfg, pkt);
1621 if (!stream->config.write_webm) {
1622 if (pkt->data.frame.partition_id <= 0) {
1623 ivf_header_pos = ftello(stream->file);
1624 fsize = pkt->data.frame.sz;
1626 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1628 fsize += pkt->data.frame.sz;
1630 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1631 const int64_t currpos = ftello(stream->file);
1632 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1633 ivf_write_frame_size(stream->file, fsize);
1634 fseeko(stream->file, currpos, SEEK_SET);
1638 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1641 stream->nbytes += pkt->data.raw.sz;
1645 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1646 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1647 (unsigned int)pkt->data.frame.sz, NULL, 0);
1648 if (stream->decoder.err) {
1649 warn_or_exit_on_error(&stream->decoder,
1650 global->test_decode == TEST_DECODE_FATAL,
1651 "Failed to decode frame %d in stream %d",
1652 stream->frames_out + 1, stream->index);
1653 stream->mismatch_seen = stream->frames_out + 1;
1658 case VPX_CODEC_STATS_PKT:
1659 stream->frames_out++;
1660 stats_write(&stream->stats,
1661 pkt->data.twopass_stats.buf,
1662 pkt->data.twopass_stats.sz);
1663 stream->nbytes += pkt->data.raw.sz;
1665 #if CONFIG_FP_MB_STATS
1666 case VPX_CODEC_FPMB_STATS_PKT:
1667 stats_write(&stream->fpmb_stats,
1668 pkt->data.firstpass_mb_stats.buf,
1669 pkt->data.firstpass_mb_stats.sz);
1670 stream->nbytes += pkt->data.raw.sz;
1673 case VPX_CODEC_PSNR_PKT:
1675 if (global->show_psnr) {
1678 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1679 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1680 for (i = 0; i < 4; i++) {
1682 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1683 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1685 stream->psnr_count++;
1696 static void show_psnr(struct stream_state *stream, double peak) {
1700 if (!stream->psnr_count)
1703 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1704 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1705 (double)stream->psnr_sse_total);
1706 fprintf(stderr, " %.3f", ovpsnr);
1708 for (i = 0; i < 4; i++) {
1709 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1711 fprintf(stderr, "\n");
1715 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1716 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1719 static void test_decode(struct stream_state *stream,
1720 enum TestDecodeFatality fatal,
1721 const VpxInterface *codec) {
1722 vpx_image_t enc_img, dec_img;
1724 if (stream->mismatch_seen)
1727 /* Get the internal reference frame */
1728 if (strcmp(codec->name, "vp8") == 0) {
1729 struct vpx_ref_frame ref_enc, ref_dec;
1732 width = (stream->config.cfg.g_w + 15) & ~15;
1733 height = (stream->config.cfg.g_h + 15) & ~15;
1734 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1735 enc_img = ref_enc.img;
1736 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1737 dec_img = ref_dec.img;
1739 ref_enc.frame_type = VP8_LAST_FRAME;
1740 ref_dec.frame_type = VP8_LAST_FRAME;
1741 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1742 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1744 struct vp9_ref_frame ref_enc, ref_dec;
1748 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1749 enc_img = ref_enc.img;
1750 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1751 dec_img = ref_dec.img;
1752 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1753 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1754 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1755 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1756 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1757 enc_img.d_w, enc_img.d_h, 16);
1758 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1760 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1761 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1762 dec_img.d_w, dec_img.d_h, 16);
1763 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1768 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1769 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1771 if (!compare_img(&enc_img, &dec_img)) {
1772 int y[4], u[4], v[4];
1773 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1774 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1775 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1777 find_mismatch(&enc_img, &dec_img, y, u, v);
1780 find_mismatch(&enc_img, &dec_img, y, u, v);
1782 stream->decoder.err = 1;
1783 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1784 "Stream %d: Encode/decode mismatch on frame %d at"
1785 " Y[%d, %d] {%d/%d},"
1786 " U[%d, %d] {%d/%d},"
1787 " V[%d, %d] {%d/%d}",
1788 stream->index, stream->frames_out,
1789 y[0], y[1], y[2], y[3],
1790 u[0], u[1], u[2], u[3],
1791 v[0], v[1], v[2], v[3]);
1792 stream->mismatch_seen = stream->frames_out;
1795 vpx_img_free(&enc_img);
1796 vpx_img_free(&dec_img);
1800 static void print_time(const char *label, int64_t etl) {
1807 etl -= hours * 3600;
1812 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1813 label, hours, mins, secs);
1815 fprintf(stderr, "[%3s unknown] ", label);
1820 int main(int argc, const char **argv_) {
1823 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1824 vpx_image_t raw_shift;
1825 int allocated_raw_shift = 0;
1826 int use_16bit_internal = 0;
1827 int input_shift = 0;
1829 int frame_avail, got_data;
1831 struct VpxInputContext input;
1832 struct VpxEncoderConfig global;
1833 struct stream_state *streams = NULL;
1834 char **argv, **argi;
1835 uint64_t cx_time = 0;
1839 memset(&input, 0, sizeof(input));
1840 exec_name = argv_[0];
1845 /* Setup default input stream settings */
1846 input.framerate.numerator = 30;
1847 input.framerate.denominator = 1;
1848 input.only_i420 = 1;
1849 input.bit_depth = 0;
1851 /* First parse the global configuration values, because we want to apply
1852 * other parameters on top of the default configuration provided by the
1855 argv = argv_dup(argc - 1, argv_ + 1);
1856 parse_global_config(&global, argv);
1858 switch (global.color_type) {
1860 input.fmt = VPX_IMG_FMT_I420;
1863 input.fmt = VPX_IMG_FMT_I422;
1866 input.fmt = VPX_IMG_FMT_I444;
1869 input.fmt = VPX_IMG_FMT_I440;
1872 input.fmt = VPX_IMG_FMT_YV12;
1877 /* Now parse each stream's parameters. Using a local scope here
1878 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1881 struct stream_state *stream = NULL;
1884 stream = new_stream(&global, stream);
1888 } while (parse_stream_params(&global, stream, argv));
1891 /* Check for unrecognized options */
1892 for (argi = argv; *argi; argi++)
1893 if (argi[0][0] == '-' && argi[0][1])
1894 die("Error: Unrecognized option %s\n", *argi);
1896 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1897 &global, &stream->config.cfg););
1899 /* Handle non-option arguments */
1900 input.filename = argv[0];
1902 if (!input.filename)
1905 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1906 if (global.codec->fourcc == VP9_FOURCC)
1907 input.only_i420 = 0;
1909 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1910 int frames_in = 0, seen_frames = 0;
1911 int64_t estimated_time_left = -1;
1912 int64_t average_rate = -1;
1913 int64_t lagged_count = 0;
1915 open_input_file(&input);
1917 /* If the input file doesn't specify its w/h (raw files), try to get
1918 * the data from the first stream's configuration.
1920 if (!input.width || !input.height) {
1922 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1923 input.width = stream->config.cfg.g_w;
1924 input.height = stream->config.cfg.g_h;
1930 /* Update stream configurations from the input file's parameters */
1931 if (!input.width || !input.height)
1932 fatal("Specify stream dimensions with --width (-w) "
1933 " and --height (-h)");
1935 /* If input file does not specify bit-depth but input-bit-depth parameter
1936 * exists, assume that to be the input bit-depth. However, if the
1937 * input-bit-depth paramter does not exist, assume the input bit-depth
1938 * to be the same as the codec bit-depth.
1940 if (!input.bit_depth) {
1942 if (stream->config.cfg.g_input_bit_depth)
1943 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1945 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1946 (int)stream->config.cfg.g_bit_depth;
1948 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1951 stream->config.cfg.g_input_bit_depth = input.bit_depth;
1955 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1956 FOREACH_STREAM(validate_stream_config(stream, &global));
1958 /* Ensure that --passes and --pass are consistent. If --pass is set and
1959 * --passes=2, ensure --fpf was set.
1961 if (global.pass && global.passes == 2)
1963 if (!stream->config.stats_fn)
1964 die("Stream %d: Must specify --fpf when --pass=%d"
1965 " and --passes=2\n", stream->index, global.pass);
1970 stream->config.write_webm = 0;
1971 warn("vpxenc was compiled without WebM container support."
1972 "Producing IVF output");
1976 /* Use the frame rate from the file only if none was specified
1977 * on the command-line.
1979 if (!global.have_framerate) {
1980 global.framerate.num = input.framerate.numerator;
1981 global.framerate.den = input.framerate.denominator;
1984 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1986 /* Show configuration */
1987 if (global.verbose && pass == 0)
1988 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1990 if (pass == (global.pass ? global.pass - 1 : 0)) {
1991 if (input.file_type == FILE_TYPE_Y4M)
1992 /*The Y4M reader does its own allocation.
1993 Just initialize this here to avoid problems if we never read any
1995 memset(&raw, 0, sizeof(raw));
1997 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1999 FOREACH_STREAM(stream->rate_hist =
2000 init_rate_histogram(&stream->config.cfg,
2001 &global.framerate));
2004 FOREACH_STREAM(setup_pass(stream, &global, pass));
2005 FOREACH_STREAM(open_output_file(stream, &global));
2006 FOREACH_STREAM(initialize_encoder(stream, &global));
2008 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2009 if (strcmp(global.codec->name, "vp9") == 0) {
2010 // Check to see if at least one stream uses 16 bit internal.
2011 // Currently assume that the bit_depths for all streams using
2012 // highbitdepth are the same.
2014 if (stream->config.use_16bit_internal) {
2015 use_16bit_internal = 1;
2017 if (stream->config.cfg.g_profile == 0) {
2020 input_shift = (int)stream->config.cfg.g_bit_depth -
2021 stream->config.cfg.g_input_bit_depth;
2030 while (frame_avail || got_data) {
2031 struct vpx_usec_timer timer;
2033 if (!global.limit || frames_in < global.limit) {
2034 frame_avail = read_frame(&input, &raw);
2038 seen_frames = frames_in > global.skip_frames ?
2039 frames_in - global.skip_frames : 0;
2041 if (!global.quiet) {
2042 float fps = usec_to_fps(cx_time, seen_frames);
2043 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2045 if (stream_cnt == 1)
2047 "frame %4d/%-4d %7"PRId64"B ",
2048 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2050 fprintf(stderr, "frame %4d ", frames_in);
2052 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2053 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2054 cx_time > 9999999 ? "ms" : "us",
2055 fps >= 1.0 ? fps : fps * 60,
2056 fps >= 1.0 ? "fps" : "fpm");
2057 print_time("ETA", estimated_time_left);
2063 if (frames_in > global.skip_frames) {
2064 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2065 vpx_image_t *frame_to_encode;
2066 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2067 assert(use_16bit_internal);
2068 // Input bit depth and stream bit depth do not match, so up
2069 // shift frame to stream bit depth
2070 if (!allocated_raw_shift) {
2071 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2072 input.width, input.height, 32);
2073 allocated_raw_shift = 1;
2075 vpx_img_upshift(&raw_shift, &raw, input_shift);
2076 frame_to_encode = &raw_shift;
2078 frame_to_encode = &raw;
2080 vpx_usec_timer_start(&timer);
2081 if (use_16bit_internal) {
2082 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2084 if (stream->config.use_16bit_internal)
2085 encode_frame(stream, &global,
2086 frame_avail ? frame_to_encode : NULL,
2092 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2093 FOREACH_STREAM(encode_frame(stream, &global,
2094 frame_avail ? frame_to_encode : NULL,
2098 vpx_usec_timer_start(&timer);
2099 FOREACH_STREAM(encode_frame(stream, &global,
2100 frame_avail ? &raw : NULL,
2103 vpx_usec_timer_mark(&timer);
2104 cx_time += vpx_usec_timer_elapsed(&timer);
2106 FOREACH_STREAM(update_quantizer_histogram(stream));
2109 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2111 if (!got_data && input.length && streams != NULL &&
2112 !streams->frames_out) {
2113 lagged_count = global.limit ? seen_frames : ftello(input.file);
2114 } else if (input.length) {
2119 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2121 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2122 remaining = 1000 * (global.limit - global.skip_frames
2123 - seen_frames + lagged_count);
2125 const int64_t input_pos = ftello(input.file);
2126 const int64_t input_pos_lagged = input_pos - lagged_count;
2127 const int64_t limit = input.length;
2129 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2130 remaining = limit - input_pos + lagged_count;
2133 average_rate = (average_rate <= 0)
2135 : (average_rate * 7 + rate) / 8;
2136 estimated_time_left = average_rate ? remaining / average_rate : -1;
2139 if (got_data && global.test_decode != TEST_DECODE_OFF)
2140 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2145 fprintf(stderr, "\033[K");
2149 fprintf(stderr, "\n");
2151 if (!global.quiet) {
2152 FOREACH_STREAM(fprintf(stderr,
2153 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2154 " %7"PRId64" %s (%.2f fps)\033[K\n",
2156 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2157 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2158 seen_frames ? (int64_t)stream->nbytes * 8 *
2159 (int64_t)global.framerate.num / global.framerate.den /
2161 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2162 stream->cx_time > 9999999 ? "ms" : "us",
2163 usec_to_fps(stream->cx_time, seen_frames)));
2166 if (global.show_psnr) {
2167 if (global.codec->fourcc == VP9_FOURCC) {
2169 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2171 FOREACH_STREAM(show_psnr(stream, 255.0));
2175 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2177 if (global.test_decode != TEST_DECODE_OFF) {
2178 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2181 close_input_file(&input);
2183 if (global.test_decode == TEST_DECODE_FATAL) {
2184 FOREACH_STREAM(res |= stream->mismatch_seen);
2186 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2188 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2190 #if CONFIG_FP_MB_STATS
2191 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2198 if (global.show_q_hist_buckets)
2199 FOREACH_STREAM(show_q_histogram(stream->counts,
2200 global.show_q_hist_buckets));
2202 if (global.show_rate_hist_buckets)
2203 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2204 &stream->config.cfg,
2205 global.show_rate_hist_buckets));
2206 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2208 #if CONFIG_INTERNAL_STATS
2209 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2210 * to match some existing utilities.
2212 if (!(global.pass == 1 && global.passes == 2))
2214 FILE *f = fopen("opsnr.stt", "a");
2215 if (stream->mismatch_seen) {
2216 fprintf(f, "First mismatch occurred in frame %d\n",
2217 stream->mismatch_seen);
2219 fprintf(f, "No mismatch detected in recon buffers\n");
2225 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2226 if (allocated_raw_shift)
2227 vpx_img_free(&raw_shift);
2232 return res ? EXIT_FAILURE : EXIT_SUCCESS;