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"
22 #include "vpx/vpx_encoder.h"
24 #include "vpx/vpx_decoder.h"
27 #include "third_party/libyuv/include/libyuv/scale.h"
30 #include "./tools_common.h"
32 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
33 #include "vpx/vp8cx.h"
35 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
36 #include "vpx/vp8dx.h"
39 #include "vpx/vpx_integer.h"
40 #include "vpx_ports/mem_ops.h"
41 #include "vpx_ports/vpx_timer.h"
42 #include "./rate_hist.h"
43 #include "./vpxstats.h"
44 #include "./warnings.h"
45 #include "./webmenc.h"
46 #include "./y4minput.h"
48 /* Swallow warnings about unused results of fread/fwrite */
49 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
51 return fread(ptr, size, nmemb, stream);
53 #define fread wrap_fread
55 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
57 return fwrite(ptr, size, nmemb, stream);
59 #define fwrite wrap_fwrite
62 static const char *exec_name;
64 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
65 const char *s, va_list ap) {
67 const char *detail = vpx_codec_error_detail(ctx);
69 vfprintf(stderr, s, ap);
70 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
73 fprintf(stderr, " %s\n", detail);
80 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
84 warn_or_exit_on_errorv(ctx, 1, s, ap);
88 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
93 warn_or_exit_on_errorv(ctx, fatal, s, ap);
97 int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
98 FILE *f = input_ctx->file;
99 y4m_input *y4m = &input_ctx->y4m;
102 if (input_ctx->file_type == FILE_TYPE_Y4M) {
103 if (y4m_input_fetch_frame(y4m, f, img) < 1)
106 shortread = read_yuv_frame(input_ctx, img);
112 int file_is_y4m(const char detect[4]) {
113 if (memcmp(detect, "YUV4", 4) == 0) {
119 int fourcc_is_ivf(const char detect[4]) {
120 if (memcmp(detect, "DKIF", 4) == 0) {
126 /* Murmur hash derived from public domain reference implementation at
127 * http:// sites.google.com/site/murmurhash/
129 static unsigned int murmur(const void *key, int len, unsigned int seed) {
130 const unsigned int m = 0x5bd1e995;
133 unsigned int h = seed ^ len;
135 const unsigned char *data = (const unsigned char *)key;
140 k = (unsigned int)data[0];
141 k |= (unsigned int)data[1] << 8;
142 k |= (unsigned int)data[2] << 16;
143 k |= (unsigned int)data[3] << 24;
174 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
175 "Debug mode (makes output deterministic)");
176 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
178 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
179 "Input file is YV12 ");
180 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
181 "Input file is I420 (default)");
182 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
184 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
185 "Number of passes (1/2)");
186 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
187 "Pass to execute (1/2)");
188 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
189 "First pass statistics file name");
190 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
191 "Stop encoding after n input frames");
192 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
193 "Skip the first n input frames");
194 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
195 "Deadline per frame (usec)");
196 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
197 "Use Best Quality Deadline");
198 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
199 "Use Good Quality Deadline");
200 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
201 "Use Realtime Quality Deadline");
202 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
203 "Do not print encode progress");
204 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
205 "Show encoder parameters");
206 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
207 "Show PSNR in status line");
209 static const struct arg_enum_list test_decode_enum[] = {
210 {"off", TEST_DECODE_OFF},
211 {"fatal", TEST_DECODE_FATAL},
212 {"warn", TEST_DECODE_WARN},
215 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
216 "Test encode/decode mismatch",
218 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
219 "Stream frame rate (rate/scale)");
220 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
221 "Output IVF (default is WebM)");
222 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
223 "Makes encoder output partitions. Requires IVF output!");
224 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
225 "Show quantizer histogram (n-buckets)");
226 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
227 "Show rate histogram (n-buckets)");
228 static const arg_def_t disable_warnings =
229 ARG_DEF(NULL, "disable-warnings", 0,
230 "Disable warnings about potentially incorrect encode settings.");
231 static const arg_def_t disable_warning_prompt =
232 ARG_DEF("y", "disable-warning-prompt", 0,
233 "Display warnings, but do not prompt user to continue.");
234 static const arg_def_t experimental_bitstream =
235 ARG_DEF(NULL, "experimental-bitstream", 0,
236 "Allow experimental bitstream features.");
239 static const arg_def_t *main_args[] = {
241 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
242 &deadline, &best_dl, &good_dl, &rt_dl,
243 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
244 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
248 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
249 "Usage profile number to use");
250 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
251 "Max number of threads to use");
252 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
253 "Bitstream profile number to use");
254 static const arg_def_t width = ARG_DEF("w", "width", 1,
256 static const arg_def_t height = ARG_DEF("h", "height", 1,
258 static const struct arg_enum_list stereo_mode_enum[] = {
259 {"mono", STEREO_FORMAT_MONO},
260 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
261 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
262 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
263 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
266 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
267 "Stereo 3D video format", stereo_mode_enum);
268 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
269 "Output timestamp precision (fractional seconds)");
270 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
271 "Enable error resiliency features");
272 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
273 "Max number of frames to lag");
275 static const arg_def_t *global_args[] = {
276 &use_yv12, &use_i420, &usage, &threads, &profile,
277 &width, &height, &stereo_mode, &timebase, &framerate,
282 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
283 "Temporal resampling threshold (buf %)");
284 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
285 "Spatial resampling enabled (bool)");
286 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
287 "Upscale threshold (buf %)");
288 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
289 "Downscale threshold (buf %)");
290 static const struct arg_enum_list end_usage_enum[] = {
297 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
298 "Rate control mode", end_usage_enum);
299 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
301 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
302 "Minimum (best) quantizer");
303 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
304 "Maximum (worst) quantizer");
305 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
306 "Datarate undershoot (min) target (%)");
307 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
308 "Datarate overshoot (max) target (%)");
309 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
310 "Client buffer size (ms)");
311 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
312 "Client initial buffer size (ms)");
313 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
314 "Client optimal buffer size (ms)");
315 static const arg_def_t *rc_args[] = {
316 &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
317 &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
318 &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
323 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
324 "CBR/VBR bias (0=CBR, 100=VBR)");
325 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
326 "GOP min bitrate (% of target)");
327 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
328 "GOP max bitrate (% of target)");
329 static const arg_def_t *rc_twopass_args[] = {
330 &bias_pct, &minsection_pct, &maxsection_pct, NULL
334 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
335 "Minimum keyframe interval (frames)");
336 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
337 "Maximum keyframe interval (frames)");
338 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
339 "Disable keyframe placement");
340 static const arg_def_t *kf_args[] = {
341 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
345 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
346 "Noise sensitivity (frames to blur)");
347 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
348 "Filter sharpness (0-7)");
349 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
350 "Motion detection threshold");
351 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
352 "CPU Used (-16..16)");
353 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
354 "Enable automatic alt reference frames");
355 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
356 "AltRef Max Frames");
357 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
359 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
361 static const struct arg_enum_list tuning_enum[] = {
362 {"psnr", VP8_TUNE_PSNR},
363 {"ssim", VP8_TUNE_SSIM},
366 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
367 "Material to favor", tuning_enum);
368 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
369 "Constant/Constrained Quality level");
370 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
371 "Max I-frame bitrate (pct)");
373 #if CONFIG_VP8_ENCODER
374 static const arg_def_t token_parts =
375 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
376 static const arg_def_t *vp8_args[] = {
377 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
378 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
379 &tune_ssim, &cq_level, &max_intra_rate_pct,
382 static const int vp8_arg_ctrl_map[] = {
383 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
384 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
385 VP8E_SET_TOKEN_PARTITIONS,
386 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
387 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
392 #if CONFIG_VP9_ENCODER
393 static const arg_def_t tile_cols =
394 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
395 static const arg_def_t tile_rows =
396 ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
397 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
398 static const arg_def_t frame_parallel_decoding = ARG_DEF(
399 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
400 static const arg_def_t aq_mode = ARG_DEF(
402 "Adaptive q mode (0: off (by default), 1: variance 2: complexity)");
404 static const arg_def_t *vp9_args[] = {
405 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
406 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
407 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
408 &frame_parallel_decoding, &aq_mode,
411 static const int vp9_arg_ctrl_map[] = {
412 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
413 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
414 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
415 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
416 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
417 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
422 static const arg_def_t *no_args[] = { NULL };
427 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
430 fprintf(stderr, "\nOptions:\n");
431 arg_show_usage(stderr, main_args);
432 fprintf(stderr, "\nEncoder Global Options:\n");
433 arg_show_usage(stderr, global_args);
434 fprintf(stderr, "\nRate Control Options:\n");
435 arg_show_usage(stderr, rc_args);
436 fprintf(stderr, "\nTwopass Rate Control Options:\n");
437 arg_show_usage(stderr, rc_twopass_args);
438 fprintf(stderr, "\nKeyframe Placement Options:\n");
439 arg_show_usage(stderr, kf_args);
440 #if CONFIG_VP8_ENCODER
441 fprintf(stderr, "\nVP8 Specific Options:\n");
442 arg_show_usage(stderr, vp8_args);
444 #if CONFIG_VP9_ENCODER
445 fprintf(stderr, "\nVP9 Specific Options:\n");
446 arg_show_usage(stderr, vp9_args);
448 fprintf(stderr, "\nStream timebase (--timebase):\n"
449 " The desired precision of timestamps in the output, expressed\n"
450 " in fractional seconds. Default is 1/1000.\n");
451 fprintf(stderr, "\nIncluded encoders:\n\n");
453 for (i = 0; i < get_vpx_encoder_count(); ++i) {
454 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
455 fprintf(stderr, " %-6s - %s\n",
456 encoder->name, vpx_codec_iface_name(encoder->interface()));
462 #define mmin(a, b) ((a) < (b) ? (a) : (b))
463 static void find_mismatch(const vpx_image_t *const img1,
464 const vpx_image_t *const img2,
465 int yloc[4], int uloc[4], int vloc[4]) {
466 const uint32_t bsize = 64;
467 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
468 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
470 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
472 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
475 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
476 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
477 for (j = 0; match && j < img1->d_w; j += bsize) {
479 const int si = mmin(i + bsize, img1->d_h) - i;
480 const int sj = mmin(j + bsize, img1->d_w) - j;
481 for (k = 0; match && k < si; ++k) {
482 for (l = 0; match && l < sj; ++l) {
483 if (*(img1->planes[VPX_PLANE_Y] +
484 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
485 *(img2->planes[VPX_PLANE_Y] +
486 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
489 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
490 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
491 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
492 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
501 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
502 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
503 for (j = 0; match && j < c_w; j += bsizex) {
505 const int si = mmin(i + bsizey, c_h - i);
506 const int sj = mmin(j + bsizex, c_w - j);
507 for (k = 0; match && k < si; ++k) {
508 for (l = 0; match && l < sj; ++l) {
509 if (*(img1->planes[VPX_PLANE_U] +
510 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
511 *(img2->planes[VPX_PLANE_U] +
512 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
515 uloc[2] = *(img1->planes[VPX_PLANE_U] +
516 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
517 uloc[3] = *(img2->planes[VPX_PLANE_U] +
518 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
526 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
527 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
528 for (j = 0; match && j < c_w; j += bsizex) {
530 const int si = mmin(i + bsizey, c_h - i);
531 const int sj = mmin(j + bsizex, c_w - j);
532 for (k = 0; match && k < si; ++k) {
533 for (l = 0; match && l < sj; ++l) {
534 if (*(img1->planes[VPX_PLANE_V] +
535 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
536 *(img2->planes[VPX_PLANE_V] +
537 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
540 vloc[2] = *(img1->planes[VPX_PLANE_V] +
541 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
542 vloc[3] = *(img2->planes[VPX_PLANE_V] +
543 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
553 static int compare_img(const vpx_image_t *const img1,
554 const vpx_image_t *const img2) {
556 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
558 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
562 match &= (img1->fmt == img2->fmt);
563 match &= (img1->d_w == img2->d_w);
564 match &= (img1->d_h == img2->d_h);
566 for (i = 0; i < img1->d_h; ++i)
567 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
568 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
571 for (i = 0; i < c_h; ++i)
572 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
573 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
576 for (i = 0; i < c_h; ++i)
577 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
578 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
585 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
586 #define MAX(x,y) ((x)>(y)?(x):(y))
587 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
588 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
589 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
590 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
592 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
593 NELEMENTS(vp9_arg_ctrl_map))
596 /* Per-stream configuration */
597 struct stream_config {
598 struct vpx_codec_enc_cfg cfg;
600 const char *stats_fn;
601 stereo_format_t stereo_fmt;
602 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
605 int have_kf_max_dist;
609 struct stream_state {
611 struct stream_state *next;
612 struct stream_config config;
614 struct rate_hist *rate_hist;
615 struct EbmlGlobal ebml;
617 uint64_t psnr_sse_total;
618 uint64_t psnr_samples_total;
619 double psnr_totals[4];
622 vpx_codec_ctx_t encoder;
623 unsigned int frames_out;
627 struct vpx_image *img;
628 vpx_codec_ctx_t decoder;
633 void validate_positive_rational(const char *msg,
634 struct vpx_rational *rat) {
641 die("Error: %s must be positive\n", msg);
644 die("Error: %s has zero denominator\n", msg);
648 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
652 /* Initialize default parameters */
653 memset(global, 0, sizeof(*global));
654 global->codec = get_vpx_encoder_by_index(0);
656 global->use_i420 = 1;
657 /* Assign default deadline to good quality */
658 global->deadline = VPX_DL_GOOD_QUALITY;
660 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
663 if (arg_match(&arg, &codecarg, argi)) {
664 global->codec = get_vpx_encoder_by_name(arg.val);
666 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
667 } else if (arg_match(&arg, &passes, argi)) {
668 global->passes = arg_parse_uint(&arg);
670 if (global->passes < 1 || global->passes > 2)
671 die("Error: Invalid number of passes (%d)\n", global->passes);
672 } else if (arg_match(&arg, &pass_arg, argi)) {
673 global->pass = arg_parse_uint(&arg);
675 if (global->pass < 1 || global->pass > 2)
676 die("Error: Invalid pass selected (%d)\n",
678 } else if (arg_match(&arg, &usage, argi))
679 global->usage = arg_parse_uint(&arg);
680 else if (arg_match(&arg, &deadline, argi))
681 global->deadline = arg_parse_uint(&arg);
682 else if (arg_match(&arg, &best_dl, argi))
683 global->deadline = VPX_DL_BEST_QUALITY;
684 else if (arg_match(&arg, &good_dl, argi))
685 global->deadline = VPX_DL_GOOD_QUALITY;
686 else if (arg_match(&arg, &rt_dl, argi))
687 global->deadline = VPX_DL_REALTIME;
688 else if (arg_match(&arg, &use_yv12, argi))
689 global->use_i420 = 0;
690 else if (arg_match(&arg, &use_i420, argi))
691 global->use_i420 = 1;
692 else if (arg_match(&arg, &quietarg, argi))
694 else if (arg_match(&arg, &verbosearg, argi))
696 else if (arg_match(&arg, &limit, argi))
697 global->limit = arg_parse_uint(&arg);
698 else if (arg_match(&arg, &skip, argi))
699 global->skip_frames = arg_parse_uint(&arg);
700 else if (arg_match(&arg, &psnrarg, argi))
701 global->show_psnr = 1;
702 else if (arg_match(&arg, &recontest, argi))
703 global->test_decode = arg_parse_enum_or_int(&arg);
704 else if (arg_match(&arg, &framerate, argi)) {
705 global->framerate = arg_parse_rational(&arg);
706 validate_positive_rational(arg.name, &global->framerate);
707 global->have_framerate = 1;
708 } else if (arg_match(&arg, &out_part, argi))
709 global->out_part = 1;
710 else if (arg_match(&arg, &debugmode, argi))
712 else if (arg_match(&arg, &q_hist_n, argi))
713 global->show_q_hist_buckets = arg_parse_uint(&arg);
714 else if (arg_match(&arg, &rate_hist_n, argi))
715 global->show_rate_hist_buckets = arg_parse_uint(&arg);
716 else if (arg_match(&arg, &disable_warnings, argi))
717 global->disable_warnings = 1;
718 else if (arg_match(&arg, &disable_warning_prompt, argi))
719 global->disable_warning_prompt = 1;
720 else if (arg_match(&arg, &experimental_bitstream, argi))
721 global->experimental_bitstream = 1;
727 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
728 if (global->pass > global->passes) {
729 warn("Assuming --pass=%d implies --passes=%d\n",
730 global->pass, global->pass);
731 global->passes = global->pass;
734 /* Validate global config */
735 if (global->passes == 0) {
736 #if CONFIG_VP9_ENCODER
737 // Make default VP9 passes = 2 until there is a better quality 1-pass
739 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
740 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
746 if (global->deadline == VPX_DL_REALTIME &&
747 global->passes > 1) {
748 warn("Enforcing one-pass encoding in realtime mode\n");
754 void open_input_file(struct VpxInputContext *input) {
755 /* Parse certain options from the input file, if possible */
756 input->file = strcmp(input->filename, "-")
757 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
760 fatal("Failed to open input file");
762 if (!fseeko(input->file, 0, SEEK_END)) {
763 /* Input file is seekable. Figure out how long it is, so we can get
766 input->length = ftello(input->file);
770 /* For RAW input sources, these bytes will applied on the first frame
773 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
774 input->detect.position = 0;
776 if (input->detect.buf_read == 4
777 && file_is_y4m(input->detect.buf)) {
778 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
779 input->only_i420) >= 0) {
780 input->file_type = FILE_TYPE_Y4M;
781 input->width = input->y4m.pic_w;
782 input->height = input->y4m.pic_h;
783 input->framerate.numerator = input->y4m.fps_n;
784 input->framerate.denominator = input->y4m.fps_d;
787 fatal("Unsupported Y4M stream.");
788 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
789 fatal("IVF is not supported as input.");
791 input->file_type = FILE_TYPE_RAW;
796 static void close_input_file(struct VpxInputContext *input) {
798 if (input->file_type == FILE_TYPE_Y4M)
799 y4m_input_close(&input->y4m);
802 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
803 struct stream_state *prev) {
804 struct stream_state *stream;
806 stream = calloc(1, sizeof(*stream));
808 fatal("Failed to allocate new stream.");
810 memcpy(stream, prev, sizeof(*stream));
816 /* Populate encoder configuration */
817 res = vpx_codec_enc_config_default(global->codec->interface(),
821 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
823 /* Change the default timebase to a high enough value so that the
824 * encoder will always create strictly increasing timestamps.
826 stream->config.cfg.g_timebase.den = 1000;
828 /* Never use the library's default resolution, require it be parsed
829 * from the file or set on the command line.
831 stream->config.cfg.g_w = 0;
832 stream->config.cfg.g_h = 0;
834 /* Initialize remaining stream parameters */
835 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
836 stream->config.write_webm = 1;
837 stream->ebml.last_pts_ms = -1;
839 /* Allows removal of the application version from the EBML tags */
840 stream->ebml.debug = global->debug;
842 /* Default lag_in_frames is 0 in realtime mode */
843 if (global->deadline == VPX_DL_REALTIME)
844 stream->config.cfg.g_lag_in_frames = 0;
847 /* Output files must be specified for each stream */
848 stream->config.out_fn = NULL;
855 static int parse_stream_params(struct VpxEncoderConfig *global,
856 struct stream_state *stream,
860 static const arg_def_t **ctrl_args = no_args;
861 static const int *ctrl_args_map = NULL;
862 struct stream_config *config = &stream->config;
863 int eos_mark_found = 0;
865 // Handle codec specific options
867 #if CONFIG_VP8_ENCODER
868 } else if (strcmp(global->codec->name, "vp8") == 0) {
869 ctrl_args = vp8_args;
870 ctrl_args_map = vp8_arg_ctrl_map;
872 #if CONFIG_VP9_ENCODER
873 } else if (strcmp(global->codec->name, "vp9") == 0) {
874 ctrl_args = vp9_args;
875 ctrl_args_map = vp9_arg_ctrl_map;
879 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
882 /* Once we've found an end-of-stream marker (--) we want to continue
883 * shifting arguments but not consuming them.
885 if (eos_mark_found) {
888 } else if (!strcmp(*argj, "--")) {
894 } else if (arg_match(&arg, &outputfile, argi)) {
895 config->out_fn = arg.val;
896 } else if (arg_match(&arg, &fpf_name, argi)) {
897 config->stats_fn = arg.val;
898 } else if (arg_match(&arg, &use_ivf, argi)) {
899 config->write_webm = 0;
900 } else if (arg_match(&arg, &threads, argi)) {
901 config->cfg.g_threads = arg_parse_uint(&arg);
902 } else if (arg_match(&arg, &profile, argi)) {
903 config->cfg.g_profile = arg_parse_uint(&arg);
904 } else if (arg_match(&arg, &width, argi)) {
905 config->cfg.g_w = arg_parse_uint(&arg);
906 } else if (arg_match(&arg, &height, argi)) {
907 config->cfg.g_h = arg_parse_uint(&arg);
908 } else if (arg_match(&arg, &stereo_mode, argi)) {
909 config->stereo_fmt = arg_parse_enum_or_int(&arg);
910 } else if (arg_match(&arg, &timebase, argi)) {
911 config->cfg.g_timebase = arg_parse_rational(&arg);
912 validate_positive_rational(arg.name, &config->cfg.g_timebase);
913 } else if (arg_match(&arg, &error_resilient, argi)) {
914 config->cfg.g_error_resilient = arg_parse_uint(&arg);
915 } else if (arg_match(&arg, &lag_in_frames, argi)) {
916 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
917 if (global->deadline == VPX_DL_REALTIME &&
918 config->cfg.g_lag_in_frames != 0) {
919 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
920 config->cfg.g_lag_in_frames = 0;
922 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
923 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
924 } else if (arg_match(&arg, &resize_allowed, argi)) {
925 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
926 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
927 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
928 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
929 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
930 } else if (arg_match(&arg, &end_usage, argi)) {
931 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
932 } else if (arg_match(&arg, &target_bitrate, argi)) {
933 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
934 } else if (arg_match(&arg, &min_quantizer, argi)) {
935 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
936 } else if (arg_match(&arg, &max_quantizer, argi)) {
937 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
938 } else if (arg_match(&arg, &undershoot_pct, argi)) {
939 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
940 } else if (arg_match(&arg, &overshoot_pct, argi)) {
941 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
942 } else if (arg_match(&arg, &buf_sz, argi)) {
943 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
944 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
945 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
946 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
947 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
948 } else if (arg_match(&arg, &bias_pct, argi)) {
949 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
950 if (global->passes < 2)
951 warn("option %s ignored in one-pass mode.\n", arg.name);
952 } else if (arg_match(&arg, &minsection_pct, argi)) {
953 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
955 if (global->passes < 2)
956 warn("option %s ignored in one-pass mode.\n", arg.name);
957 } else if (arg_match(&arg, &maxsection_pct, argi)) {
958 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
960 if (global->passes < 2)
961 warn("option %s ignored in one-pass mode.\n", arg.name);
962 } else if (arg_match(&arg, &kf_min_dist, argi)) {
963 config->cfg.kf_min_dist = arg_parse_uint(&arg);
964 } else if (arg_match(&arg, &kf_max_dist, argi)) {
965 config->cfg.kf_max_dist = arg_parse_uint(&arg);
966 config->have_kf_max_dist = 1;
967 } else if (arg_match(&arg, &kf_disabled, argi)) {
968 config->cfg.kf_mode = VPX_KF_DISABLED;
971 for (i = 0; ctrl_args[i]; i++) {
972 if (arg_match(&arg, ctrl_args[i], argi)) {
976 /* Point either to the next free element or the first
977 * instance of this control.
979 for (j = 0; j < config->arg_ctrl_cnt; j++)
980 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
984 assert(j < ARG_CTRL_CNT_MAX);
985 if (j < ARG_CTRL_CNT_MAX) {
986 config->arg_ctrls[j][0] = ctrl_args_map[i];
987 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
988 if (j == config->arg_ctrl_cnt)
989 config->arg_ctrl_cnt++;
998 return eos_mark_found;
1002 #define FOREACH_STREAM(func) \
1004 struct stream_state *stream; \
1005 for (stream = streams; stream; stream = stream->next) { \
1011 static void validate_stream_config(const struct stream_state *stream,
1012 const struct VpxEncoderConfig *global) {
1013 const struct stream_state *streami;
1015 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1016 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1017 " and --height (-h)", stream->index);
1019 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
1020 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
1021 stream->index, stream->config.cfg.g_profile,
1022 experimental_bitstream.long_name);
1025 for (streami = stream; streami; streami = streami->next) {
1026 /* All streams require output files */
1027 if (!streami->config.out_fn)
1028 fatal("Stream %d: Output file is required (specify with -o)",
1031 /* Check for two streams outputting to the same file */
1032 if (streami != stream) {
1033 const char *a = stream->config.out_fn;
1034 const char *b = streami->config.out_fn;
1035 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1036 fatal("Stream %d: duplicate output file (from stream %d)",
1037 streami->index, stream->index);
1040 /* Check for two streams sharing a stats file. */
1041 if (streami != stream) {
1042 const char *a = stream->config.stats_fn;
1043 const char *b = streami->config.stats_fn;
1044 if (a && b && !strcmp(a, b))
1045 fatal("Stream %d: duplicate stats file (from stream %d)",
1046 streami->index, stream->index);
1052 static void set_stream_dimensions(struct stream_state *stream,
1055 if (!stream->config.cfg.g_w) {
1056 if (!stream->config.cfg.g_h)
1057 stream->config.cfg.g_w = w;
1059 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1061 if (!stream->config.cfg.g_h) {
1062 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1067 static void set_default_kf_interval(struct stream_state *stream,
1068 struct VpxEncoderConfig *global) {
1069 /* Use a max keyframe interval of 5 seconds, if none was
1070 * specified on the command line.
1072 if (!stream->config.have_kf_max_dist) {
1073 double framerate = (double)global->framerate.num / global->framerate.den;
1074 if (framerate > 0.0)
1075 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1080 static void show_stream_config(struct stream_state *stream,
1081 struct VpxEncoderConfig *global,
1082 struct VpxInputContext *input) {
1084 #define SHOW(field) \
1085 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1087 if (stream->index == 0) {
1088 fprintf(stderr, "Codec: %s\n",
1089 vpx_codec_iface_name(global->codec->interface()));
1090 fprintf(stderr, "Source file: %s Format: %s\n", input->filename,
1091 input->use_i420 ? "I420" : "YV12");
1093 if (stream->next || stream->index)
1094 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1095 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1096 fprintf(stderr, "Encoder parameters:\n");
1103 SHOW(g_timebase.num);
1104 SHOW(g_timebase.den);
1105 SHOW(g_error_resilient);
1107 SHOW(g_lag_in_frames);
1108 SHOW(rc_dropframe_thresh);
1109 SHOW(rc_resize_allowed);
1110 SHOW(rc_resize_up_thresh);
1111 SHOW(rc_resize_down_thresh);
1113 SHOW(rc_target_bitrate);
1114 SHOW(rc_min_quantizer);
1115 SHOW(rc_max_quantizer);
1116 SHOW(rc_undershoot_pct);
1117 SHOW(rc_overshoot_pct);
1119 SHOW(rc_buf_initial_sz);
1120 SHOW(rc_buf_optimal_sz);
1121 SHOW(rc_2pass_vbr_bias_pct);
1122 SHOW(rc_2pass_vbr_minsection_pct);
1123 SHOW(rc_2pass_vbr_maxsection_pct);
1130 static void open_output_file(struct stream_state *stream,
1131 struct VpxEncoderConfig *global) {
1132 const char *fn = stream->config.out_fn;
1133 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1135 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1138 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1141 fatal("Failed to open output file");
1143 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1144 fatal("WebM output to pipes not supported.");
1146 if (stream->config.write_webm) {
1147 stream->ebml.stream = stream->file;
1148 write_webm_file_header(&stream->ebml, cfg,
1150 stream->config.stereo_fmt,
1151 global->codec->fourcc);
1153 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1158 static void close_output_file(struct stream_state *stream,
1159 unsigned int fourcc) {
1160 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1162 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1165 if (stream->config.write_webm) {
1166 write_webm_file_footer(&stream->ebml, stream->hash);
1167 free(stream->ebml.cue_list);
1168 stream->ebml.cue_list = NULL;
1170 if (!fseek(stream->file, 0, SEEK_SET))
1171 ivf_write_file_header(stream->file, &stream->config.cfg,
1173 stream->frames_out);
1176 fclose(stream->file);
1180 static void setup_pass(struct stream_state *stream,
1181 struct VpxEncoderConfig *global,
1183 if (stream->config.stats_fn) {
1184 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1186 fatal("Failed to open statistics store");
1188 if (!stats_open_mem(&stream->stats, pass))
1189 fatal("Failed to open statistics store");
1192 stream->config.cfg.g_pass = global->passes == 2
1193 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1196 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1198 stream->cx_time = 0;
1200 stream->frames_out = 0;
1204 static void initialize_encoder(struct stream_state *stream,
1205 struct VpxEncoderConfig *global) {
1209 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1210 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1212 /* Construct Encoder Context */
1213 vpx_codec_enc_init(&stream->encoder, global->codec->interface(),
1214 &stream->config.cfg, flags);
1215 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1217 /* Note that we bypass the vpx_codec_control wrapper macro because
1218 * we're being clever to store the control IDs in an array. Real
1219 * applications will want to make use of the enumerations directly
1221 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1222 int ctrl = stream->config.arg_ctrls[i][0];
1223 int value = stream->config.arg_ctrls[i][1];
1224 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1225 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1228 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1232 if (global->test_decode != TEST_DECODE_OFF) {
1233 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1234 vpx_codec_dec_init(&stream->decoder, decoder->interface(), NULL, 0);
1240 static void encode_frame(struct stream_state *stream,
1241 struct VpxEncoderConfig *global,
1242 struct vpx_image *img,
1243 unsigned int frames_in) {
1244 vpx_codec_pts_t frame_start, next_frame_start;
1245 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1246 struct vpx_usec_timer timer;
1248 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1249 * global->framerate.den)
1250 / cfg->g_timebase.num / global->framerate.num;
1251 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1252 * global->framerate.den)
1253 / cfg->g_timebase.num / global->framerate.num;
1255 /* Scale if necessary */
1256 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1258 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1259 cfg->g_w, cfg->g_h, 16);
1260 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1261 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1262 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1264 stream->img->planes[VPX_PLANE_Y],
1265 stream->img->stride[VPX_PLANE_Y],
1266 stream->img->planes[VPX_PLANE_U],
1267 stream->img->stride[VPX_PLANE_U],
1268 stream->img->planes[VPX_PLANE_V],
1269 stream->img->stride[VPX_PLANE_V],
1270 stream->img->d_w, stream->img->d_h,
1276 vpx_usec_timer_start(&timer);
1277 vpx_codec_encode(&stream->encoder, img, frame_start,
1278 (unsigned long)(next_frame_start - frame_start),
1279 0, global->deadline);
1280 vpx_usec_timer_mark(&timer);
1281 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1282 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1287 static void update_quantizer_histogram(struct stream_state *stream) {
1288 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1291 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1292 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1293 stream->counts[q]++;
1298 static void get_cx_data(struct stream_state *stream,
1299 struct VpxEncoderConfig *global,
1301 const vpx_codec_cx_pkt_t *pkt;
1302 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1303 vpx_codec_iter_t iter = NULL;
1306 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1307 static size_t fsize = 0;
1308 static off_t ivf_header_pos = 0;
1310 switch (pkt->kind) {
1311 case VPX_CODEC_CX_FRAME_PKT:
1312 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1313 stream->frames_out++;
1316 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1318 update_rate_histogram(stream->rate_hist, cfg, pkt);
1319 if (stream->config.write_webm) {
1320 /* Update the hash */
1321 if (!stream->ebml.debug)
1322 stream->hash = murmur(pkt->data.frame.buf,
1323 (int)pkt->data.frame.sz,
1326 write_webm_block(&stream->ebml, cfg, pkt);
1328 if (pkt->data.frame.partition_id <= 0) {
1329 ivf_header_pos = ftello(stream->file);
1330 fsize = pkt->data.frame.sz;
1332 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1334 fsize += pkt->data.frame.sz;
1336 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1337 off_t currpos = ftello(stream->file);
1338 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1339 ivf_write_frame_size(stream->file, fsize);
1340 fseeko(stream->file, currpos, SEEK_SET);
1344 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1347 stream->nbytes += pkt->data.raw.sz;
1351 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1352 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1353 (unsigned int)pkt->data.frame.sz, NULL, 0);
1354 if (stream->decoder.err) {
1355 warn_or_exit_on_error(&stream->decoder,
1356 global->test_decode == TEST_DECODE_FATAL,
1357 "Failed to decode frame %d in stream %d",
1358 stream->frames_out + 1, stream->index);
1359 stream->mismatch_seen = stream->frames_out + 1;
1364 case VPX_CODEC_STATS_PKT:
1365 stream->frames_out++;
1366 stats_write(&stream->stats,
1367 pkt->data.twopass_stats.buf,
1368 pkt->data.twopass_stats.sz);
1369 stream->nbytes += pkt->data.raw.sz;
1371 case VPX_CODEC_PSNR_PKT:
1373 if (global->show_psnr) {
1376 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1377 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1378 for (i = 0; i < 4; i++) {
1380 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1381 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1383 stream->psnr_count++;
1394 static void show_psnr(struct stream_state *stream) {
1398 if (!stream->psnr_count)
1401 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1402 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1403 (double)stream->psnr_sse_total);
1404 fprintf(stderr, " %.3f", ovpsnr);
1406 for (i = 0; i < 4; i++) {
1407 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1409 fprintf(stderr, "\n");
1413 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1414 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1418 static void test_decode(struct stream_state *stream,
1419 enum TestDecodeFatality fatal,
1420 const VpxInterface *codec) {
1421 vpx_image_t enc_img, dec_img;
1423 if (stream->mismatch_seen)
1426 /* Get the internal reference frame */
1427 if (strcmp(codec->name, "vp8") == 0) {
1428 struct vpx_ref_frame ref_enc, ref_dec;
1431 width = (stream->config.cfg.g_w + 15) & ~15;
1432 height = (stream->config.cfg.g_h + 15) & ~15;
1433 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1434 enc_img = ref_enc.img;
1435 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1436 dec_img = ref_dec.img;
1438 ref_enc.frame_type = VP8_LAST_FRAME;
1439 ref_dec.frame_type = VP8_LAST_FRAME;
1440 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1441 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1443 struct vp9_ref_frame ref;
1446 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
1448 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
1451 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1452 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1454 if (!compare_img(&enc_img, &dec_img)) {
1455 int y[4], u[4], v[4];
1456 find_mismatch(&enc_img, &dec_img, y, u, v);
1457 stream->decoder.err = 1;
1458 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1459 "Stream %d: Encode/decode mismatch on frame %d at"
1460 " Y[%d, %d] {%d/%d},"
1461 " U[%d, %d] {%d/%d},"
1462 " V[%d, %d] {%d/%d}",
1463 stream->index, stream->frames_out,
1464 y[0], y[1], y[2], y[3],
1465 u[0], u[1], u[2], u[3],
1466 v[0], v[1], v[2], v[3]);
1467 stream->mismatch_seen = stream->frames_out;
1470 vpx_img_free(&enc_img);
1471 vpx_img_free(&dec_img);
1475 static void print_time(const char *label, int64_t etl) {
1482 etl -= hours * 3600;
1487 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64": % 02"PRId64"] ",
1488 label, hours, mins, secs);
1490 fprintf(stderr, "[%3s unknown] ", label);
1495 int main(int argc, const char **argv_) {
1498 int frame_avail, got_data;
1500 struct VpxInputContext input = {0};
1501 struct VpxEncoderConfig global;
1502 struct stream_state *streams = NULL;
1503 char **argv, **argi;
1504 uint64_t cx_time = 0;
1508 exec_name = argv_[0];
1513 /* Setup default input stream settings */
1514 input.framerate.numerator = 30;
1515 input.framerate.denominator = 1;
1517 input.only_i420 = 1;
1519 /* First parse the global configuration values, because we want to apply
1520 * other parameters on top of the default configuration provided by the
1523 argv = argv_dup(argc - 1, argv_ + 1);
1524 parse_global_config(&global, argv);
1528 /* Now parse each stream's parameters. Using a local scope here
1529 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1532 struct stream_state *stream = NULL;
1535 stream = new_stream(&global, stream);
1539 } while (parse_stream_params(&global, stream, argv));
1542 /* Check for unrecognized options */
1543 for (argi = argv; *argi; argi++)
1544 if (argi[0][0] == '-' && argi[0][1])
1545 die("Error: Unrecognized option %s\n", *argi);
1547 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1548 &global, &stream->config.cfg););
1550 /* Handle non-option arguments */
1551 input.filename = argv[0];
1553 if (!input.filename)
1556 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1557 if (global.codec->fourcc == VP9_FOURCC)
1558 input.only_i420 = 0;
1560 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1561 int frames_in = 0, seen_frames = 0;
1562 int64_t estimated_time_left = -1;
1563 int64_t average_rate = -1;
1564 off_t lagged_count = 0;
1566 open_input_file(&input);
1568 /* If the input file doesn't specify its w/h (raw files), try to get
1569 * the data from the first stream's configuration.
1571 if (!input.width || !input.height)
1573 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1574 input.width = stream->config.cfg.g_w;
1575 input.height = stream->config.cfg.g_h;
1580 /* Update stream configurations from the input file's parameters */
1581 if (!input.width || !input.height)
1582 fatal("Specify stream dimensions with --width (-w) "
1583 " and --height (-h)");
1584 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1585 FOREACH_STREAM(validate_stream_config(stream, &global));
1587 /* Ensure that --passes and --pass are consistent. If --pass is set and
1588 * --passes=2, ensure --fpf was set.
1590 if (global.pass && global.passes == 2)
1592 if (!stream->config.stats_fn)
1593 die("Stream %d: Must specify --fpf when --pass=%d"
1594 " and --passes=2\n", stream->index, global.pass);
1597 /* Use the frame rate from the file only if none was specified
1598 * on the command-line.
1600 if (!global.have_framerate) {
1601 global.framerate.num = input.framerate.numerator;
1602 global.framerate.den = input.framerate.denominator;
1605 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1607 /* Show configuration */
1608 if (global.verbose && pass == 0)
1609 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1611 if (pass == (global.pass ? global.pass - 1 : 0)) {
1612 if (input.file_type == FILE_TYPE_Y4M)
1613 /*The Y4M reader does its own allocation.
1614 Just initialize this here to avoid problems if we never read any
1616 memset(&raw, 0, sizeof(raw));
1619 input.use_i420 ? VPX_IMG_FMT_I420
1621 input.width, input.height, 32);
1623 FOREACH_STREAM(stream->rate_hist =
1624 init_rate_histogram(&stream->config.cfg,
1625 &global.framerate));
1628 FOREACH_STREAM(setup_pass(stream, &global, pass));
1629 FOREACH_STREAM(open_output_file(stream, &global));
1630 FOREACH_STREAM(initialize_encoder(stream, &global));
1635 while (frame_avail || got_data) {
1636 struct vpx_usec_timer timer;
1638 if (!global.limit || frames_in < global.limit) {
1639 frame_avail = read_frame(&input, &raw);
1643 seen_frames = frames_in > global.skip_frames ?
1644 frames_in - global.skip_frames : 0;
1646 if (!global.quiet) {
1647 float fps = usec_to_fps(cx_time, seen_frames);
1648 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1650 if (stream_cnt == 1)
1652 "frame %4d/%-4d %7"PRId64"B ",
1653 frames_in, streams->frames_out, (int64_t)streams->nbytes);
1655 fprintf(stderr, "frame %4d ", frames_in);
1657 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
1658 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1659 cx_time > 9999999 ? "ms" : "us",
1660 fps >= 1.0 ? fps : fps * 60,
1661 fps >= 1.0 ? "fps" : "fpm");
1662 print_time("ETA", estimated_time_left);
1663 fprintf(stderr, "\033[K");
1669 if (frames_in > global.skip_frames) {
1670 vpx_usec_timer_start(&timer);
1671 FOREACH_STREAM(encode_frame(stream, &global,
1672 frame_avail ? &raw : NULL,
1674 vpx_usec_timer_mark(&timer);
1675 cx_time += vpx_usec_timer_elapsed(&timer);
1677 FOREACH_STREAM(update_quantizer_histogram(stream));
1680 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1682 if (!got_data && input.length && !streams->frames_out) {
1683 lagged_count = global.limit ? seen_frames : ftello(input.file);
1684 } else if (input.length) {
1689 off_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1691 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1692 remaining = 1000 * (global.limit - global.skip_frames
1693 - seen_frames + lagged_count);
1695 off_t input_pos = ftello(input.file);
1696 off_t input_pos_lagged = input_pos - lagged_count;
1697 int64_t limit = input.length;
1699 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1700 remaining = limit - input_pos + lagged_count;
1703 average_rate = (average_rate <= 0)
1705 : (average_rate * 7 + rate) / 8;
1706 estimated_time_left = average_rate ? remaining / average_rate : -1;
1709 if (got_data && global.test_decode != TEST_DECODE_OFF)
1710 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1717 fprintf(stderr, "\n");
1720 FOREACH_STREAM(fprintf(
1722 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
1723 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
1724 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
1725 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
1726 seen_frames ? (int64_t)stream->nbytes * 8
1727 * (int64_t)global.framerate.num / global.framerate.den
1730 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1731 stream->cx_time > 9999999 ? "ms" : "us",
1732 usec_to_fps(stream->cx_time, seen_frames));
1735 if (global.show_psnr)
1736 FOREACH_STREAM(show_psnr(stream));
1738 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1740 if (global.test_decode != TEST_DECODE_OFF) {
1741 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1744 close_input_file(&input);
1746 if (global.test_decode == TEST_DECODE_FATAL) {
1747 FOREACH_STREAM(res |= stream->mismatch_seen);
1749 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1751 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1757 if (global.show_q_hist_buckets)
1758 FOREACH_STREAM(show_q_histogram(stream->counts,
1759 global.show_q_hist_buckets));
1761 if (global.show_rate_hist_buckets)
1762 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
1763 &stream->config.cfg,
1764 global.show_rate_hist_buckets));
1765 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1767 #if CONFIG_INTERNAL_STATS
1768 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1769 * to match some existing utilities.
1771 if (!(global.pass == 1 && global.passes == 2))
1773 FILE *f = fopen("opsnr.stt", "a");
1774 if (stream->mismatch_seen) {
1775 fprintf(f, "First mismatch occurred in frame %d\n",
1776 stream->mismatch_seen);
1778 fprintf(f, "No mismatch detected in recon buffers\n");
1787 return res ? EXIT_FAILURE : EXIT_SUCCESS;