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"
46 #include "./webmenc.h"
48 #include "./y4minput.h"
50 /* Swallow warnings about unused results of fread/fwrite */
51 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
53 return fread(ptr, size, nmemb, stream);
55 #define fread wrap_fread
57 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
59 return fwrite(ptr, size, nmemb, stream);
61 #define fwrite wrap_fwrite
64 static const char *exec_name;
66 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
67 const char *s, va_list ap) {
69 const char *detail = vpx_codec_error_detail(ctx);
71 vfprintf(stderr, s, ap);
72 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
75 fprintf(stderr, " %s\n", detail);
82 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
86 warn_or_exit_on_errorv(ctx, 1, s, ap);
90 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
95 warn_or_exit_on_errorv(ctx, fatal, s, ap);
99 int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
100 FILE *f = input_ctx->file;
101 y4m_input *y4m = &input_ctx->y4m;
104 if (input_ctx->file_type == FILE_TYPE_Y4M) {
105 if (y4m_input_fetch_frame(y4m, f, img) < 1)
108 shortread = read_yuv_frame(input_ctx, img);
114 int file_is_y4m(const char detect[4]) {
115 if (memcmp(detect, "YUV4", 4) == 0) {
121 int fourcc_is_ivf(const char detect[4]) {
122 if (memcmp(detect, "DKIF", 4) == 0) {
128 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
129 "Debug mode (makes output deterministic)");
130 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
132 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
133 "Input file is YV12 ");
134 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
135 "Input file is I420 (default)");
136 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
138 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
139 "Number of passes (1/2)");
140 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
141 "Pass to execute (1/2)");
142 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
143 "First pass statistics file name");
144 #if CONFIG_FP_MB_STATS
145 static const arg_def_t fpmbf_name = ARG_DEF(NULL, "fpmbf", 1,
146 "First pass block statistics file name");
148 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
149 "Stop encoding after n input frames");
150 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
151 "Skip the first n input frames");
152 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
153 "Deadline per frame (usec)");
154 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
155 "Use Best Quality Deadline");
156 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
157 "Use Good Quality Deadline");
158 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
159 "Use Realtime Quality Deadline");
160 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
161 "Do not print encode progress");
162 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
163 "Show encoder parameters");
164 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
165 "Show PSNR in status line");
167 static const struct arg_enum_list test_decode_enum[] = {
168 {"off", TEST_DECODE_OFF},
169 {"fatal", TEST_DECODE_FATAL},
170 {"warn", TEST_DECODE_WARN},
173 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
174 "Test encode/decode mismatch",
176 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
177 "Stream frame rate (rate/scale)");
178 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
179 "Output IVF (default is WebM if WebM IO is enabled)");
180 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
181 "Makes encoder output partitions. Requires IVF output!");
182 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
183 "Show quantizer histogram (n-buckets)");
184 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
185 "Show rate histogram (n-buckets)");
186 static const arg_def_t disable_warnings =
187 ARG_DEF(NULL, "disable-warnings", 0,
188 "Disable warnings about potentially incorrect encode settings.");
189 static const arg_def_t disable_warning_prompt =
190 ARG_DEF("y", "disable-warning-prompt", 0,
191 "Display warnings, but do not prompt user to continue.");
192 static const arg_def_t experimental_bitstream =
193 ARG_DEF(NULL, "experimental-bitstream", 0,
194 "Allow experimental bitstream features.");
197 static const arg_def_t *main_args[] = {
199 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
200 &deadline, &best_dl, &good_dl, &rt_dl,
201 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
202 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
206 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
207 "Usage profile number to use");
208 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
209 "Max number of threads to use");
210 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
211 "Bitstream profile number to use");
212 static const arg_def_t width = ARG_DEF("w", "width", 1,
214 static const arg_def_t height = ARG_DEF("h", "height", 1,
217 static const struct arg_enum_list stereo_mode_enum[] = {
218 {"mono", STEREO_FORMAT_MONO},
219 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
220 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
221 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
222 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
225 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
226 "Stereo 3D video format", stereo_mode_enum);
228 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
229 "Output timestamp precision (fractional seconds)");
230 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
231 "Enable error resiliency features");
232 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
233 "Max number of frames to lag");
235 static const arg_def_t *global_args[] = {
236 &use_yv12, &use_i420, &usage, &threads, &profile,
241 &timebase, &framerate,
246 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
247 "Temporal resampling threshold (buf %)");
248 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
249 "Spatial resampling enabled (bool)");
250 static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
251 "Width of encoded frame");
252 static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
253 "Height of encoded frame");
254 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
255 "Upscale threshold (buf %)");
256 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
257 "Downscale threshold (buf %)");
258 static const struct arg_enum_list end_usage_enum[] = {
265 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
266 "Rate control mode", end_usage_enum);
267 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
269 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
270 "Minimum (best) quantizer");
271 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
272 "Maximum (worst) quantizer");
273 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
274 "Datarate undershoot (min) target (%)");
275 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
276 "Datarate overshoot (max) target (%)");
277 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
278 "Client buffer size (ms)");
279 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
280 "Client initial buffer size (ms)");
281 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
282 "Client optimal buffer size (ms)");
283 static const arg_def_t *rc_args[] = {
284 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
285 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
286 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
287 &buf_initial_sz, &buf_optimal_sz, NULL
291 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
292 "CBR/VBR bias (0=CBR, 100=VBR)");
293 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
294 "GOP min bitrate (% of target)");
295 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
296 "GOP max bitrate (% of target)");
297 static const arg_def_t *rc_twopass_args[] = {
298 &bias_pct, &minsection_pct, &maxsection_pct, NULL
302 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
303 "Minimum keyframe interval (frames)");
304 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
305 "Maximum keyframe interval (frames)");
306 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
307 "Disable keyframe placement");
308 static const arg_def_t *kf_args[] = {
309 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
313 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
314 "Noise sensitivity (frames to blur)");
315 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
316 "Filter sharpness (0-7)");
317 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
318 "Motion detection threshold");
319 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
320 "CPU Used (-16..16)");
321 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
322 "Enable automatic alt reference frames");
323 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
324 "AltRef Max Frames");
325 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
327 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
329 static const struct arg_enum_list tuning_enum[] = {
330 {"psnr", VP8_TUNE_PSNR},
331 {"ssim", VP8_TUNE_SSIM},
334 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
335 "Material to favor", tuning_enum);
336 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
337 "Constant/Constrained Quality level");
338 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
339 "Max I-frame bitrate (pct)");
341 #if CONFIG_VP8_ENCODER
342 static const arg_def_t token_parts =
343 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
344 static const arg_def_t *vp8_args[] = {
345 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
346 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
347 &tune_ssim, &cq_level, &max_intra_rate_pct,
350 static const int vp8_arg_ctrl_map[] = {
351 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
352 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
353 VP8E_SET_TOKEN_PARTITIONS,
354 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
355 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
360 #if CONFIG_VP9_ENCODER
361 static const arg_def_t tile_cols =
362 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
363 static const arg_def_t tile_rows =
364 ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
365 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
366 static const arg_def_t frame_parallel_decoding = ARG_DEF(
367 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
368 static const arg_def_t aq_mode = ARG_DEF(
370 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
371 "3: cyclic refresh)");
372 static const arg_def_t frame_periodic_boost = ARG_DEF(
373 NULL, "frame_boost", 1,
374 "Enable frame periodic boost (0: off (default), 1: on)");
376 static const arg_def_t *vp9_args[] = {
377 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
378 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
379 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
380 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
383 static const int vp9_arg_ctrl_map[] = {
384 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
385 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
386 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
387 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
388 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
389 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
390 VP9E_SET_FRAME_PERIODIC_BOOST,
395 static const arg_def_t *no_args[] = { NULL };
400 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
403 fprintf(stderr, "\nOptions:\n");
404 arg_show_usage(stderr, main_args);
405 fprintf(stderr, "\nEncoder Global Options:\n");
406 arg_show_usage(stderr, global_args);
407 fprintf(stderr, "\nRate Control Options:\n");
408 arg_show_usage(stderr, rc_args);
409 fprintf(stderr, "\nTwopass Rate Control Options:\n");
410 arg_show_usage(stderr, rc_twopass_args);
411 fprintf(stderr, "\nKeyframe Placement Options:\n");
412 arg_show_usage(stderr, kf_args);
413 #if CONFIG_VP8_ENCODER
414 fprintf(stderr, "\nVP8 Specific Options:\n");
415 arg_show_usage(stderr, vp8_args);
417 #if CONFIG_VP9_ENCODER
418 fprintf(stderr, "\nVP9 Specific Options:\n");
419 arg_show_usage(stderr, vp9_args);
421 fprintf(stderr, "\nStream timebase (--timebase):\n"
422 " The desired precision of timestamps in the output, expressed\n"
423 " in fractional seconds. Default is 1/1000.\n");
424 fprintf(stderr, "\nIncluded encoders:\n\n");
426 for (i = 0; i < get_vpx_encoder_count(); ++i) {
427 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
428 fprintf(stderr, " %-6s - %s\n",
429 encoder->name, vpx_codec_iface_name(encoder->interface()));
435 #define mmin(a, b) ((a) < (b) ? (a) : (b))
436 static void find_mismatch(const vpx_image_t *const img1,
437 const vpx_image_t *const img2,
438 int yloc[4], int uloc[4], int vloc[4]) {
439 const uint32_t bsize = 64;
440 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
441 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
443 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
445 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
448 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
449 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
450 for (j = 0; match && j < img1->d_w; j += bsize) {
452 const int si = mmin(i + bsize, img1->d_h) - i;
453 const int sj = mmin(j + bsize, img1->d_w) - j;
454 for (k = 0; match && k < si; ++k) {
455 for (l = 0; match && l < sj; ++l) {
456 if (*(img1->planes[VPX_PLANE_Y] +
457 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
458 *(img2->planes[VPX_PLANE_Y] +
459 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
462 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
463 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
464 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
465 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
474 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
475 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
476 for (j = 0; match && j < c_w; j += bsizex) {
478 const int si = mmin(i + bsizey, c_h - i);
479 const int sj = mmin(j + bsizex, c_w - j);
480 for (k = 0; match && k < si; ++k) {
481 for (l = 0; match && l < sj; ++l) {
482 if (*(img1->planes[VPX_PLANE_U] +
483 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
484 *(img2->planes[VPX_PLANE_U] +
485 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
488 uloc[2] = *(img1->planes[VPX_PLANE_U] +
489 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
490 uloc[3] = *(img2->planes[VPX_PLANE_U] +
491 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
499 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
500 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
501 for (j = 0; match && j < c_w; j += bsizex) {
503 const int si = mmin(i + bsizey, c_h - i);
504 const int sj = mmin(j + bsizex, c_w - j);
505 for (k = 0; match && k < si; ++k) {
506 for (l = 0; match && l < sj; ++l) {
507 if (*(img1->planes[VPX_PLANE_V] +
508 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
509 *(img2->planes[VPX_PLANE_V] +
510 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
513 vloc[2] = *(img1->planes[VPX_PLANE_V] +
514 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
515 vloc[3] = *(img2->planes[VPX_PLANE_V] +
516 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
526 static int compare_img(const vpx_image_t *const img1,
527 const vpx_image_t *const img2) {
529 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
531 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
535 match &= (img1->fmt == img2->fmt);
536 match &= (img1->d_w == img2->d_w);
537 match &= (img1->d_h == img2->d_h);
539 for (i = 0; i < img1->d_h; ++i)
540 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
541 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
544 for (i = 0; i < c_h; ++i)
545 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
546 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
549 for (i = 0; i < c_h; ++i)
550 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
551 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
558 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
559 #define MAX(x,y) ((x)>(y)?(x):(y))
560 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
561 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
562 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
563 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
565 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
566 NELEMENTS(vp9_arg_ctrl_map))
570 typedef int stereo_format_t;
571 struct EbmlGlobal { int debug; };
574 /* Per-stream configuration */
575 struct stream_config {
576 struct vpx_codec_enc_cfg cfg;
578 const char *stats_fn;
579 #if CONFIG_FP_MB_STATS
580 const char *fpmb_stats_fn;
582 stereo_format_t stereo_fmt;
583 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
586 int have_kf_max_dist;
590 struct stream_state {
592 struct stream_state *next;
593 struct stream_config config;
595 struct rate_hist *rate_hist;
596 struct EbmlGlobal ebml;
597 uint64_t psnr_sse_total;
598 uint64_t psnr_samples_total;
599 double psnr_totals[4];
602 vpx_codec_ctx_t encoder;
603 unsigned int frames_out;
607 #if CONFIG_FP_MB_STATS
608 stats_io_t fpmb_stats;
610 struct vpx_image *img;
611 vpx_codec_ctx_t decoder;
616 void validate_positive_rational(const char *msg,
617 struct vpx_rational *rat) {
624 die("Error: %s must be positive\n", msg);
627 die("Error: %s has zero denominator\n", msg);
631 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
635 /* Initialize default parameters */
636 memset(global, 0, sizeof(*global));
637 global->codec = get_vpx_encoder_by_index(0);
639 global->use_i420 = 1;
640 /* Assign default deadline to good quality */
641 global->deadline = VPX_DL_GOOD_QUALITY;
643 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
646 if (arg_match(&arg, &codecarg, argi)) {
647 global->codec = get_vpx_encoder_by_name(arg.val);
649 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
650 } else if (arg_match(&arg, &passes, argi)) {
651 global->passes = arg_parse_uint(&arg);
653 if (global->passes < 1 || global->passes > 2)
654 die("Error: Invalid number of passes (%d)\n", global->passes);
655 } else if (arg_match(&arg, &pass_arg, argi)) {
656 global->pass = arg_parse_uint(&arg);
658 if (global->pass < 1 || global->pass > 2)
659 die("Error: Invalid pass selected (%d)\n",
661 } else if (arg_match(&arg, &usage, argi))
662 global->usage = arg_parse_uint(&arg);
663 else if (arg_match(&arg, &deadline, argi))
664 global->deadline = arg_parse_uint(&arg);
665 else if (arg_match(&arg, &best_dl, argi))
666 global->deadline = VPX_DL_BEST_QUALITY;
667 else if (arg_match(&arg, &good_dl, argi))
668 global->deadline = VPX_DL_GOOD_QUALITY;
669 else if (arg_match(&arg, &rt_dl, argi))
670 global->deadline = VPX_DL_REALTIME;
671 else if (arg_match(&arg, &use_yv12, argi))
672 global->use_i420 = 0;
673 else if (arg_match(&arg, &use_i420, argi))
674 global->use_i420 = 1;
675 else if (arg_match(&arg, &quietarg, argi))
677 else if (arg_match(&arg, &verbosearg, argi))
679 else if (arg_match(&arg, &limit, argi))
680 global->limit = arg_parse_uint(&arg);
681 else if (arg_match(&arg, &skip, argi))
682 global->skip_frames = arg_parse_uint(&arg);
683 else if (arg_match(&arg, &psnrarg, argi))
684 global->show_psnr = 1;
685 else if (arg_match(&arg, &recontest, argi))
686 global->test_decode = arg_parse_enum_or_int(&arg);
687 else if (arg_match(&arg, &framerate, argi)) {
688 global->framerate = arg_parse_rational(&arg);
689 validate_positive_rational(arg.name, &global->framerate);
690 global->have_framerate = 1;
691 } else if (arg_match(&arg, &out_part, argi))
692 global->out_part = 1;
693 else if (arg_match(&arg, &debugmode, argi))
695 else if (arg_match(&arg, &q_hist_n, argi))
696 global->show_q_hist_buckets = arg_parse_uint(&arg);
697 else if (arg_match(&arg, &rate_hist_n, argi))
698 global->show_rate_hist_buckets = arg_parse_uint(&arg);
699 else if (arg_match(&arg, &disable_warnings, argi))
700 global->disable_warnings = 1;
701 else if (arg_match(&arg, &disable_warning_prompt, argi))
702 global->disable_warning_prompt = 1;
703 else if (arg_match(&arg, &experimental_bitstream, argi))
704 global->experimental_bitstream = 1;
710 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
711 if (global->pass > global->passes) {
712 warn("Assuming --pass=%d implies --passes=%d\n",
713 global->pass, global->pass);
714 global->passes = global->pass;
717 /* Validate global config */
718 if (global->passes == 0) {
719 #if CONFIG_VP9_ENCODER
720 // Make default VP9 passes = 2 until there is a better quality 1-pass
722 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
723 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
729 if (global->deadline == VPX_DL_REALTIME &&
730 global->passes > 1) {
731 warn("Enforcing one-pass encoding in realtime mode\n");
737 void open_input_file(struct VpxInputContext *input) {
738 /* Parse certain options from the input file, if possible */
739 input->file = strcmp(input->filename, "-")
740 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
743 fatal("Failed to open input file");
745 if (!fseeko(input->file, 0, SEEK_END)) {
746 /* Input file is seekable. Figure out how long it is, so we can get
749 input->length = ftello(input->file);
753 /* For RAW input sources, these bytes will applied on the first frame
756 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
757 input->detect.position = 0;
759 if (input->detect.buf_read == 4
760 && file_is_y4m(input->detect.buf)) {
761 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
762 input->only_i420) >= 0) {
763 input->file_type = FILE_TYPE_Y4M;
764 input->width = input->y4m.pic_w;
765 input->height = input->y4m.pic_h;
766 input->framerate.numerator = input->y4m.fps_n;
767 input->framerate.denominator = input->y4m.fps_d;
768 input->fmt = input->y4m.vpx_fmt;
769 input->bit_depth = input->y4m.bit_depth;
771 fatal("Unsupported Y4M stream.");
772 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
773 fatal("IVF is not supported as input.");
775 input->file_type = FILE_TYPE_RAW;
780 static void close_input_file(struct VpxInputContext *input) {
782 if (input->file_type == FILE_TYPE_Y4M)
783 y4m_input_close(&input->y4m);
786 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
787 struct stream_state *prev) {
788 struct stream_state *stream;
790 stream = calloc(1, sizeof(*stream));
792 fatal("Failed to allocate new stream.");
794 memcpy(stream, prev, sizeof(*stream));
800 /* Populate encoder configuration */
801 res = vpx_codec_enc_config_default(global->codec->interface(),
805 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
807 /* Change the default timebase to a high enough value so that the
808 * encoder will always create strictly increasing timestamps.
810 stream->config.cfg.g_timebase.den = 1000;
812 /* Never use the library's default resolution, require it be parsed
813 * from the file or set on the command line.
815 stream->config.cfg.g_w = 0;
816 stream->config.cfg.g_h = 0;
818 /* Initialize remaining stream parameters */
819 stream->config.write_webm = 1;
821 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
822 stream->ebml.last_pts_ns = -1;
823 stream->ebml.writer = NULL;
824 stream->ebml.segment = NULL;
827 /* Allows removal of the application version from the EBML tags */
828 stream->ebml.debug = global->debug;
830 /* Default lag_in_frames is 0 in realtime mode */
831 if (global->deadline == VPX_DL_REALTIME)
832 stream->config.cfg.g_lag_in_frames = 0;
835 /* Output files must be specified for each stream */
836 stream->config.out_fn = NULL;
843 static int parse_stream_params(struct VpxEncoderConfig *global,
844 struct stream_state *stream,
848 static const arg_def_t **ctrl_args = no_args;
849 static const int *ctrl_args_map = NULL;
850 struct stream_config *config = &stream->config;
851 int eos_mark_found = 0;
853 // Handle codec specific options
855 #if CONFIG_VP8_ENCODER
856 } else if (strcmp(global->codec->name, "vp8") == 0) {
857 ctrl_args = vp8_args;
858 ctrl_args_map = vp8_arg_ctrl_map;
860 #if CONFIG_VP9_ENCODER
861 } else if (strcmp(global->codec->name, "vp9") == 0) {
862 ctrl_args = vp9_args;
863 ctrl_args_map = vp9_arg_ctrl_map;
867 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
870 /* Once we've found an end-of-stream marker (--) we want to continue
871 * shifting arguments but not consuming them.
873 if (eos_mark_found) {
876 } else if (!strcmp(*argj, "--")) {
882 } else if (arg_match(&arg, &outputfile, argi)) {
883 config->out_fn = arg.val;
884 } else if (arg_match(&arg, &fpf_name, argi)) {
885 config->stats_fn = arg.val;
886 #if CONFIG_FP_MB_STATS
887 } else if (arg_match(&arg, &fpmbf_name, argi)) {
888 config->fpmb_stats_fn = arg.val;
890 } else if (arg_match(&arg, &use_ivf, argi)) {
891 config->write_webm = 0;
892 } else if (arg_match(&arg, &threads, argi)) {
893 config->cfg.g_threads = arg_parse_uint(&arg);
894 } else if (arg_match(&arg, &profile, argi)) {
895 config->cfg.g_profile = arg_parse_uint(&arg);
896 } else if (arg_match(&arg, &width, argi)) {
897 config->cfg.g_w = arg_parse_uint(&arg);
898 } else if (arg_match(&arg, &height, argi)) {
899 config->cfg.g_h = arg_parse_uint(&arg);
901 } else if (arg_match(&arg, &stereo_mode, argi)) {
902 config->stereo_fmt = arg_parse_enum_or_int(&arg);
904 } else if (arg_match(&arg, &timebase, argi)) {
905 config->cfg.g_timebase = arg_parse_rational(&arg);
906 validate_positive_rational(arg.name, &config->cfg.g_timebase);
907 } else if (arg_match(&arg, &error_resilient, argi)) {
908 config->cfg.g_error_resilient = arg_parse_uint(&arg);
909 } else if (arg_match(&arg, &lag_in_frames, argi)) {
910 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
911 if (global->deadline == VPX_DL_REALTIME &&
912 config->cfg.g_lag_in_frames != 0) {
913 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
914 config->cfg.g_lag_in_frames = 0;
916 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
917 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
918 } else if (arg_match(&arg, &resize_allowed, argi)) {
919 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
920 } else if (arg_match(&arg, &resize_width, argi)) {
921 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
922 } else if (arg_match(&arg, &resize_height, argi)) {
923 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
924 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
925 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
926 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
927 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
928 } else if (arg_match(&arg, &end_usage, argi)) {
929 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
930 } else if (arg_match(&arg, &target_bitrate, argi)) {
931 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
932 } else if (arg_match(&arg, &min_quantizer, argi)) {
933 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
934 } else if (arg_match(&arg, &max_quantizer, argi)) {
935 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
936 } else if (arg_match(&arg, &undershoot_pct, argi)) {
937 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
938 } else if (arg_match(&arg, &overshoot_pct, argi)) {
939 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
940 } else if (arg_match(&arg, &buf_sz, argi)) {
941 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
942 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
943 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
944 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
945 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
946 } else if (arg_match(&arg, &bias_pct, argi)) {
947 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
948 if (global->passes < 2)
949 warn("option %s ignored in one-pass mode.\n", arg.name);
950 } else if (arg_match(&arg, &minsection_pct, argi)) {
951 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
953 if (global->passes < 2)
954 warn("option %s ignored in one-pass mode.\n", arg.name);
955 } else if (arg_match(&arg, &maxsection_pct, argi)) {
956 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
958 if (global->passes < 2)
959 warn("option %s ignored in one-pass mode.\n", arg.name);
960 } else if (arg_match(&arg, &kf_min_dist, argi)) {
961 config->cfg.kf_min_dist = arg_parse_uint(&arg);
962 } else if (arg_match(&arg, &kf_max_dist, argi)) {
963 config->cfg.kf_max_dist = arg_parse_uint(&arg);
964 config->have_kf_max_dist = 1;
965 } else if (arg_match(&arg, &kf_disabled, argi)) {
966 config->cfg.kf_mode = VPX_KF_DISABLED;
969 for (i = 0; ctrl_args[i]; i++) {
970 if (arg_match(&arg, ctrl_args[i], argi)) {
974 /* Point either to the next free element or the first
975 * instance of this control.
977 for (j = 0; j < config->arg_ctrl_cnt; j++)
978 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
982 assert(j < (int)ARG_CTRL_CNT_MAX);
983 if (j < (int)ARG_CTRL_CNT_MAX) {
984 config->arg_ctrls[j][0] = ctrl_args_map[i];
985 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
986 if (j == config->arg_ctrl_cnt)
987 config->arg_ctrl_cnt++;
996 return eos_mark_found;
1000 #define FOREACH_STREAM(func) \
1002 struct stream_state *stream; \
1003 for (stream = streams; stream; stream = stream->next) { \
1009 static void validate_stream_config(const struct stream_state *stream,
1010 const struct VpxEncoderConfig *global) {
1011 const struct stream_state *streami;
1013 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1014 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1015 " and --height (-h)", stream->index);
1017 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
1018 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
1019 stream->index, stream->config.cfg.g_profile,
1020 experimental_bitstream.long_name);
1023 for (streami = stream; streami; streami = streami->next) {
1024 /* All streams require output files */
1025 if (!streami->config.out_fn)
1026 fatal("Stream %d: Output file is required (specify with -o)",
1029 /* Check for two streams outputting to the same file */
1030 if (streami != stream) {
1031 const char *a = stream->config.out_fn;
1032 const char *b = streami->config.out_fn;
1033 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1034 fatal("Stream %d: duplicate output file (from stream %d)",
1035 streami->index, stream->index);
1038 /* Check for two streams sharing a stats file. */
1039 if (streami != stream) {
1040 const char *a = stream->config.stats_fn;
1041 const char *b = streami->config.stats_fn;
1042 if (a && b && !strcmp(a, b))
1043 fatal("Stream %d: duplicate stats file (from stream %d)",
1044 streami->index, stream->index);
1047 #if CONFIG_FP_MB_STATS
1048 /* Check for two streams sharing a mb stats file. */
1049 if (streami != stream) {
1050 const char *a = stream->config.fpmb_stats_fn;
1051 const char *b = streami->config.fpmb_stats_fn;
1052 if (a && b && !strcmp(a, b))
1053 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1054 streami->index, stream->index);
1061 static void set_stream_dimensions(struct stream_state *stream,
1064 if (!stream->config.cfg.g_w) {
1065 if (!stream->config.cfg.g_h)
1066 stream->config.cfg.g_w = w;
1068 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1070 if (!stream->config.cfg.g_h) {
1071 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1076 static void set_default_kf_interval(struct stream_state *stream,
1077 struct VpxEncoderConfig *global) {
1078 /* Use a max keyframe interval of 5 seconds, if none was
1079 * specified on the command line.
1081 if (!stream->config.have_kf_max_dist) {
1082 double framerate = (double)global->framerate.num / global->framerate.den;
1083 if (framerate > 0.0)
1084 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1088 static const char* file_type_to_string(enum VideoFileType t) {
1090 case FILE_TYPE_RAW: return "RAW";
1091 case FILE_TYPE_Y4M: return "Y4M";
1092 default: return "Other";
1096 static const char* image_format_to_string(vpx_img_fmt_t f) {
1098 case VPX_IMG_FMT_I420: return "I420";
1099 case VPX_IMG_FMT_I422: return "I422";
1100 case VPX_IMG_FMT_I444: return "I444";
1101 case VPX_IMG_FMT_YV12: return "YV12";
1102 default: return "Other";
1106 static void show_stream_config(struct stream_state *stream,
1107 struct VpxEncoderConfig *global,
1108 struct VpxInputContext *input) {
1110 #define SHOW(field) \
1111 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1113 if (stream->index == 0) {
1114 fprintf(stderr, "Codec: %s\n",
1115 vpx_codec_iface_name(global->codec->interface()));
1116 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1118 file_type_to_string(input->file_type),
1119 image_format_to_string(input->fmt));
1121 if (stream->next || stream->index)
1122 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1123 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1124 fprintf(stderr, "Encoder parameters:\n");
1131 SHOW(g_timebase.num);
1132 SHOW(g_timebase.den);
1133 SHOW(g_error_resilient);
1135 SHOW(g_lag_in_frames);
1136 SHOW(rc_dropframe_thresh);
1137 SHOW(rc_resize_allowed);
1138 SHOW(rc_scaled_width);
1139 SHOW(rc_scaled_height);
1140 SHOW(rc_resize_up_thresh);
1141 SHOW(rc_resize_down_thresh);
1143 SHOW(rc_target_bitrate);
1144 SHOW(rc_min_quantizer);
1145 SHOW(rc_max_quantizer);
1146 SHOW(rc_undershoot_pct);
1147 SHOW(rc_overshoot_pct);
1149 SHOW(rc_buf_initial_sz);
1150 SHOW(rc_buf_optimal_sz);
1151 SHOW(rc_2pass_vbr_bias_pct);
1152 SHOW(rc_2pass_vbr_minsection_pct);
1153 SHOW(rc_2pass_vbr_maxsection_pct);
1160 static void open_output_file(struct stream_state *stream,
1161 struct VpxEncoderConfig *global) {
1162 const char *fn = stream->config.out_fn;
1163 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1165 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1168 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1171 fatal("Failed to open output file");
1173 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1174 fatal("WebM output to pipes not supported.");
1177 if (stream->config.write_webm) {
1178 stream->ebml.stream = stream->file;
1179 write_webm_file_header(&stream->ebml, cfg,
1181 stream->config.stereo_fmt,
1182 global->codec->fourcc);
1186 if (!stream->config.write_webm) {
1187 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1192 static void close_output_file(struct stream_state *stream,
1193 unsigned int fourcc) {
1194 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1196 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1200 if (stream->config.write_webm) {
1201 write_webm_file_footer(&stream->ebml);
1205 if (!stream->config.write_webm) {
1206 if (!fseek(stream->file, 0, SEEK_SET))
1207 ivf_write_file_header(stream->file, &stream->config.cfg,
1209 stream->frames_out);
1212 fclose(stream->file);
1216 static void setup_pass(struct stream_state *stream,
1217 struct VpxEncoderConfig *global,
1219 if (stream->config.stats_fn) {
1220 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1222 fatal("Failed to open statistics store");
1224 if (!stats_open_mem(&stream->stats, pass))
1225 fatal("Failed to open statistics store");
1228 #if CONFIG_FP_MB_STATS
1229 if (stream->config.fpmb_stats_fn) {
1230 if (!stats_open_file(&stream->fpmb_stats,
1231 stream->config.fpmb_stats_fn, pass))
1232 fatal("Failed to open mb statistics store");
1234 if (!stats_open_mem(&stream->fpmb_stats, pass))
1235 fatal("Failed to open mb statistics store");
1239 stream->config.cfg.g_pass = global->passes == 2
1240 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1243 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1244 #if CONFIG_FP_MB_STATS
1245 stream->config.cfg.rc_firstpass_mb_stats_in =
1246 stats_get(&stream->fpmb_stats);
1250 stream->cx_time = 0;
1252 stream->frames_out = 0;
1256 static void initialize_encoder(struct stream_state *stream,
1257 struct VpxEncoderConfig *global) {
1261 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1262 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1264 /* Construct Encoder Context */
1265 vpx_codec_enc_init(&stream->encoder, global->codec->interface(),
1266 &stream->config.cfg, flags);
1267 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1269 /* Note that we bypass the vpx_codec_control wrapper macro because
1270 * we're being clever to store the control IDs in an array. Real
1271 * applications will want to make use of the enumerations directly
1273 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1274 int ctrl = stream->config.arg_ctrls[i][0];
1275 int value = stream->config.arg_ctrls[i][1];
1276 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1277 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1280 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1284 if (global->test_decode != TEST_DECODE_OFF) {
1285 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1286 vpx_codec_dec_init(&stream->decoder, decoder->interface(), NULL, 0);
1292 static void encode_frame(struct stream_state *stream,
1293 struct VpxEncoderConfig *global,
1294 struct vpx_image *img,
1295 unsigned int frames_in) {
1296 vpx_codec_pts_t frame_start, next_frame_start;
1297 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1298 struct vpx_usec_timer timer;
1300 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1301 * global->framerate.den)
1302 / cfg->g_timebase.num / global->framerate.num;
1303 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1304 * global->framerate.den)
1305 / cfg->g_timebase.num / global->framerate.num;
1307 /* Scale if necessary */
1308 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1309 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1310 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1315 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1316 cfg->g_w, cfg->g_h, 16);
1317 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1318 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1319 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1321 stream->img->planes[VPX_PLANE_Y],
1322 stream->img->stride[VPX_PLANE_Y],
1323 stream->img->planes[VPX_PLANE_U],
1324 stream->img->stride[VPX_PLANE_U],
1325 stream->img->planes[VPX_PLANE_V],
1326 stream->img->stride[VPX_PLANE_V],
1327 stream->img->d_w, stream->img->d_h,
1331 stream->encoder.err = 1;
1332 ctx_exit_on_error(&stream->encoder,
1333 "Stream %d: Failed to encode frame.\n"
1334 "Scaling disabled in this configuration. \n"
1335 "To enable, configure with --enable-libyuv\n",
1340 vpx_usec_timer_start(&timer);
1341 vpx_codec_encode(&stream->encoder, img, frame_start,
1342 (unsigned long)(next_frame_start - frame_start),
1343 0, global->deadline);
1344 vpx_usec_timer_mark(&timer);
1345 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1346 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1351 static void update_quantizer_histogram(struct stream_state *stream) {
1352 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1355 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1356 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1357 stream->counts[q]++;
1362 static void get_cx_data(struct stream_state *stream,
1363 struct VpxEncoderConfig *global,
1365 const vpx_codec_cx_pkt_t *pkt;
1366 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1367 vpx_codec_iter_t iter = NULL;
1370 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1371 static size_t fsize = 0;
1372 static int64_t ivf_header_pos = 0;
1374 switch (pkt->kind) {
1375 case VPX_CODEC_CX_FRAME_PKT:
1376 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1377 stream->frames_out++;
1380 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1382 update_rate_histogram(stream->rate_hist, cfg, pkt);
1384 if (stream->config.write_webm) {
1385 write_webm_block(&stream->ebml, cfg, pkt);
1388 if (!stream->config.write_webm) {
1389 if (pkt->data.frame.partition_id <= 0) {
1390 ivf_header_pos = ftello(stream->file);
1391 fsize = pkt->data.frame.sz;
1393 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1395 fsize += pkt->data.frame.sz;
1397 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1398 const int64_t currpos = ftello(stream->file);
1399 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1400 ivf_write_frame_size(stream->file, fsize);
1401 fseeko(stream->file, currpos, SEEK_SET);
1405 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1408 stream->nbytes += pkt->data.raw.sz;
1412 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1413 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1414 (unsigned int)pkt->data.frame.sz, NULL, 0);
1415 if (stream->decoder.err) {
1416 warn_or_exit_on_error(&stream->decoder,
1417 global->test_decode == TEST_DECODE_FATAL,
1418 "Failed to decode frame %d in stream %d",
1419 stream->frames_out + 1, stream->index);
1420 stream->mismatch_seen = stream->frames_out + 1;
1425 case VPX_CODEC_STATS_PKT:
1426 stream->frames_out++;
1427 stats_write(&stream->stats,
1428 pkt->data.twopass_stats.buf,
1429 pkt->data.twopass_stats.sz);
1430 stream->nbytes += pkt->data.raw.sz;
1432 #if CONFIG_FP_MB_STATS
1433 case VPX_CODEC_FPMB_STATS_PKT:
1434 stats_write(&stream->fpmb_stats,
1435 pkt->data.firstpass_mb_stats.buf,
1436 pkt->data.firstpass_mb_stats.sz);
1437 stream->nbytes += pkt->data.raw.sz;
1440 case VPX_CODEC_PSNR_PKT:
1442 if (global->show_psnr) {
1445 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1446 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1447 for (i = 0; i < 4; i++) {
1449 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1450 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1452 stream->psnr_count++;
1463 static void show_psnr(struct stream_state *stream) {
1467 if (!stream->psnr_count)
1470 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1471 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1472 (double)stream->psnr_sse_total);
1473 fprintf(stderr, " %.3f", ovpsnr);
1475 for (i = 0; i < 4; i++) {
1476 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1478 fprintf(stderr, "\n");
1482 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1483 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1487 static void test_decode(struct stream_state *stream,
1488 enum TestDecodeFatality fatal,
1489 const VpxInterface *codec) {
1490 vpx_image_t enc_img, dec_img;
1492 if (stream->mismatch_seen)
1495 /* Get the internal reference frame */
1496 if (strcmp(codec->name, "vp8") == 0) {
1497 struct vpx_ref_frame ref_enc, ref_dec;
1500 width = (stream->config.cfg.g_w + 15) & ~15;
1501 height = (stream->config.cfg.g_h + 15) & ~15;
1502 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1503 enc_img = ref_enc.img;
1504 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1505 dec_img = ref_dec.img;
1507 ref_enc.frame_type = VP8_LAST_FRAME;
1508 ref_dec.frame_type = VP8_LAST_FRAME;
1509 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1510 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1512 struct vp9_ref_frame ref;
1515 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
1517 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
1520 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1521 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1523 if (!compare_img(&enc_img, &dec_img)) {
1524 int y[4], u[4], v[4];
1525 find_mismatch(&enc_img, &dec_img, y, u, v);
1526 stream->decoder.err = 1;
1527 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1528 "Stream %d: Encode/decode mismatch on frame %d at"
1529 " Y[%d, %d] {%d/%d},"
1530 " U[%d, %d] {%d/%d},"
1531 " V[%d, %d] {%d/%d}",
1532 stream->index, stream->frames_out,
1533 y[0], y[1], y[2], y[3],
1534 u[0], u[1], u[2], u[3],
1535 v[0], v[1], v[2], v[3]);
1536 stream->mismatch_seen = stream->frames_out;
1539 vpx_img_free(&enc_img);
1540 vpx_img_free(&dec_img);
1544 static void print_time(const char *label, int64_t etl) {
1551 etl -= hours * 3600;
1556 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1557 label, hours, mins, secs);
1559 fprintf(stderr, "[%3s unknown] ", label);
1564 int main(int argc, const char **argv_) {
1567 int frame_avail, got_data;
1569 struct VpxInputContext input;
1570 struct VpxEncoderConfig global;
1571 struct stream_state *streams = NULL;
1572 char **argv, **argi;
1573 uint64_t cx_time = 0;
1577 memset(&input, 0, sizeof(input));
1578 exec_name = argv_[0];
1583 /* Setup default input stream settings */
1584 input.framerate.numerator = 30;
1585 input.framerate.denominator = 1;
1586 input.only_i420 = 1;
1587 input.bit_depth = 0;
1589 /* First parse the global configuration values, because we want to apply
1590 * other parameters on top of the default configuration provided by the
1593 argv = argv_dup(argc - 1, argv_ + 1);
1594 parse_global_config(&global, argv);
1596 input.fmt = global.use_i420 ? VPX_IMG_FMT_I420 : VPX_IMG_FMT_YV12;
1599 /* Now parse each stream's parameters. Using a local scope here
1600 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1603 struct stream_state *stream = NULL;
1606 stream = new_stream(&global, stream);
1610 } while (parse_stream_params(&global, stream, argv));
1613 /* Check for unrecognized options */
1614 for (argi = argv; *argi; argi++)
1615 if (argi[0][0] == '-' && argi[0][1])
1616 die("Error: Unrecognized option %s\n", *argi);
1618 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1619 &global, &stream->config.cfg););
1621 /* Handle non-option arguments */
1622 input.filename = argv[0];
1624 if (!input.filename)
1627 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1628 if (global.codec->fourcc == VP9_FOURCC)
1629 input.only_i420 = 0;
1631 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1632 int frames_in = 0, seen_frames = 0;
1633 int64_t estimated_time_left = -1;
1634 int64_t average_rate = -1;
1635 int64_t lagged_count = 0;
1637 open_input_file(&input);
1639 /* If the input file doesn't specify its w/h (raw files), try to get
1640 * the data from the first stream's configuration.
1642 if (!input.width || !input.height)
1644 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1645 input.width = stream->config.cfg.g_w;
1646 input.height = stream->config.cfg.g_h;
1651 /* Update stream configurations from the input file's parameters */
1652 if (!input.width || !input.height)
1653 fatal("Specify stream dimensions with --width (-w) "
1654 " and --height (-h)");
1655 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1656 FOREACH_STREAM(validate_stream_config(stream, &global));
1658 /* Ensure that --passes and --pass are consistent. If --pass is set and
1659 * --passes=2, ensure --fpf was set.
1661 if (global.pass && global.passes == 2)
1663 if (!stream->config.stats_fn)
1664 die("Stream %d: Must specify --fpf when --pass=%d"
1665 " and --passes=2\n", stream->index, global.pass);
1670 stream->config.write_webm = 0;
1671 warn("vpxenc was compiled without WebM container support."
1672 "Producing IVF output");
1676 /* Use the frame rate from the file only if none was specified
1677 * on the command-line.
1679 if (!global.have_framerate) {
1680 global.framerate.num = input.framerate.numerator;
1681 global.framerate.den = input.framerate.denominator;
1684 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1686 /* Show configuration */
1687 if (global.verbose && pass == 0)
1688 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1690 if (pass == (global.pass ? global.pass - 1 : 0)) {
1691 if (input.file_type == FILE_TYPE_Y4M)
1692 /*The Y4M reader does its own allocation.
1693 Just initialize this here to avoid problems if we never read any
1695 memset(&raw, 0, sizeof(raw));
1697 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1699 FOREACH_STREAM(stream->rate_hist =
1700 init_rate_histogram(&stream->config.cfg,
1701 &global.framerate));
1704 FOREACH_STREAM(setup_pass(stream, &global, pass));
1705 FOREACH_STREAM(open_output_file(stream, &global));
1706 FOREACH_STREAM(initialize_encoder(stream, &global));
1711 while (frame_avail || got_data) {
1712 struct vpx_usec_timer timer;
1714 if (!global.limit || frames_in < global.limit) {
1715 frame_avail = read_frame(&input, &raw);
1719 seen_frames = frames_in > global.skip_frames ?
1720 frames_in - global.skip_frames : 0;
1722 if (!global.quiet) {
1723 float fps = usec_to_fps(cx_time, seen_frames);
1724 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1726 if (stream_cnt == 1)
1728 "frame %4d/%-4d %7"PRId64"B ",
1729 frames_in, streams->frames_out, (int64_t)streams->nbytes);
1731 fprintf(stderr, "frame %4d ", frames_in);
1733 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
1734 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1735 cx_time > 9999999 ? "ms" : "us",
1736 fps >= 1.0 ? fps : fps * 60,
1737 fps >= 1.0 ? "fps" : "fpm");
1738 print_time("ETA", estimated_time_left);
1744 if (frames_in > global.skip_frames) {
1745 vpx_usec_timer_start(&timer);
1746 FOREACH_STREAM(encode_frame(stream, &global,
1747 frame_avail ? &raw : NULL,
1749 vpx_usec_timer_mark(&timer);
1750 cx_time += vpx_usec_timer_elapsed(&timer);
1752 FOREACH_STREAM(update_quantizer_histogram(stream));
1755 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1757 if (!got_data && input.length && !streams->frames_out) {
1758 lagged_count = global.limit ? seen_frames : ftello(input.file);
1759 } else if (input.length) {
1764 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1766 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1767 remaining = 1000 * (global.limit - global.skip_frames
1768 - seen_frames + lagged_count);
1770 const int64_t input_pos = ftello(input.file);
1771 const int64_t input_pos_lagged = input_pos - lagged_count;
1772 const int64_t limit = input.length;
1774 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1775 remaining = limit - input_pos + lagged_count;
1778 average_rate = (average_rate <= 0)
1780 : (average_rate * 7 + rate) / 8;
1781 estimated_time_left = average_rate ? remaining / average_rate : -1;
1784 if (got_data && global.test_decode != TEST_DECODE_OFF)
1785 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1790 fprintf(stderr, "\033[K");
1794 fprintf(stderr, "\n");
1797 FOREACH_STREAM(fprintf(
1799 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
1800 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
1801 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
1802 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
1803 seen_frames ? (int64_t)stream->nbytes * 8
1804 * (int64_t)global.framerate.num / global.framerate.den
1807 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1808 stream->cx_time > 9999999 ? "ms" : "us",
1809 usec_to_fps(stream->cx_time, seen_frames));
1812 if (global.show_psnr)
1813 FOREACH_STREAM(show_psnr(stream));
1815 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1817 if (global.test_decode != TEST_DECODE_OFF) {
1818 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1821 close_input_file(&input);
1823 if (global.test_decode == TEST_DECODE_FATAL) {
1824 FOREACH_STREAM(res |= stream->mismatch_seen);
1826 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1828 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1830 #if CONFIG_FP_MB_STATS
1831 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
1838 if (global.show_q_hist_buckets)
1839 FOREACH_STREAM(show_q_histogram(stream->counts,
1840 global.show_q_hist_buckets));
1842 if (global.show_rate_hist_buckets)
1843 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
1844 &stream->config.cfg,
1845 global.show_rate_hist_buckets));
1846 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1848 #if CONFIG_INTERNAL_STATS
1849 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1850 * to match some existing utilities.
1852 if (!(global.pass == 1 && global.passes == 2))
1854 FILE *f = fopen("opsnr.stt", "a");
1855 if (stream->mismatch_seen) {
1856 fprintf(f, "First mismatch occurred in frame %d\n",
1857 stream->mismatch_seen);
1859 fprintf(f, "No mismatch detected in recon buffers\n");
1868 return res ? EXIT_FAILURE : EXIT_SUCCESS;