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 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
145 "Stop encoding after n input frames");
146 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
147 "Skip the first n input frames");
148 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
149 "Deadline per frame (usec)");
150 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
151 "Use Best Quality Deadline");
152 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
153 "Use Good Quality Deadline");
154 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
155 "Use Realtime Quality Deadline");
156 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
157 "Do not print encode progress");
158 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
159 "Show encoder parameters");
160 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
161 "Show PSNR in status line");
163 static const struct arg_enum_list test_decode_enum[] = {
164 {"off", TEST_DECODE_OFF},
165 {"fatal", TEST_DECODE_FATAL},
166 {"warn", TEST_DECODE_WARN},
169 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
170 "Test encode/decode mismatch",
172 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
173 "Stream frame rate (rate/scale)");
174 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
175 "Output IVF (default is WebM if WebM IO is enabled)");
176 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
177 "Makes encoder output partitions. Requires IVF output!");
178 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
179 "Show quantizer histogram (n-buckets)");
180 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
181 "Show rate histogram (n-buckets)");
182 static const arg_def_t disable_warnings =
183 ARG_DEF(NULL, "disable-warnings", 0,
184 "Disable warnings about potentially incorrect encode settings.");
185 static const arg_def_t disable_warning_prompt =
186 ARG_DEF("y", "disable-warning-prompt", 0,
187 "Display warnings, but do not prompt user to continue.");
188 static const arg_def_t experimental_bitstream =
189 ARG_DEF(NULL, "experimental-bitstream", 0,
190 "Allow experimental bitstream features.");
193 static const arg_def_t *main_args[] = {
195 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
196 &deadline, &best_dl, &good_dl, &rt_dl,
197 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
198 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
202 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
203 "Usage profile number to use");
204 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
205 "Max number of threads to use");
206 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
207 "Bitstream profile number to use");
208 static const arg_def_t width = ARG_DEF("w", "width", 1,
210 static const arg_def_t height = ARG_DEF("h", "height", 1,
213 static const struct arg_enum_list stereo_mode_enum[] = {
214 {"mono", STEREO_FORMAT_MONO},
215 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
216 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
217 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
218 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
221 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
222 "Stereo 3D video format", stereo_mode_enum);
224 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
225 "Output timestamp precision (fractional seconds)");
226 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
227 "Enable error resiliency features");
228 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
229 "Max number of frames to lag");
231 static const arg_def_t *global_args[] = {
232 &use_yv12, &use_i420, &usage, &threads, &profile,
237 &timebase, &framerate,
242 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
243 "Temporal resampling threshold (buf %)");
244 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
245 "Spatial resampling enabled (bool)");
246 static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
247 "Width of encoded frame");
248 static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
249 "Height of encoded frame");
250 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
251 "Upscale threshold (buf %)");
252 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
253 "Downscale threshold (buf %)");
254 static const struct arg_enum_list end_usage_enum[] = {
261 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
262 "Rate control mode", end_usage_enum);
263 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
265 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
266 "Minimum (best) quantizer");
267 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
268 "Maximum (worst) quantizer");
269 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
270 "Datarate undershoot (min) target (%)");
271 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
272 "Datarate overshoot (max) target (%)");
273 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
274 "Client buffer size (ms)");
275 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
276 "Client initial buffer size (ms)");
277 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
278 "Client optimal buffer size (ms)");
279 static const arg_def_t *rc_args[] = {
280 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
281 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
282 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
283 &buf_initial_sz, &buf_optimal_sz, NULL
287 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
288 "CBR/VBR bias (0=CBR, 100=VBR)");
289 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
290 "GOP min bitrate (% of target)");
291 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
292 "GOP max bitrate (% of target)");
293 static const arg_def_t *rc_twopass_args[] = {
294 &bias_pct, &minsection_pct, &maxsection_pct, NULL
298 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
299 "Minimum keyframe interval (frames)");
300 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
301 "Maximum keyframe interval (frames)");
302 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
303 "Disable keyframe placement");
304 static const arg_def_t *kf_args[] = {
305 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
309 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
310 "Noise sensitivity (frames to blur)");
311 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
312 "Filter sharpness (0-7)");
313 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
314 "Motion detection threshold");
315 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
316 "CPU Used (-16..16)");
317 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
318 "Enable automatic alt reference frames");
319 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
320 "AltRef Max Frames");
321 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
323 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
325 static const struct arg_enum_list tuning_enum[] = {
326 {"psnr", VP8_TUNE_PSNR},
327 {"ssim", VP8_TUNE_SSIM},
330 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
331 "Material to favor", tuning_enum);
332 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
333 "Constant/Constrained Quality level");
334 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
335 "Max I-frame bitrate (pct)");
337 #if CONFIG_VP8_ENCODER
338 static const arg_def_t token_parts =
339 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
340 static const arg_def_t *vp8_args[] = {
341 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
342 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
343 &tune_ssim, &cq_level, &max_intra_rate_pct,
346 static const int vp8_arg_ctrl_map[] = {
347 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
348 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
349 VP8E_SET_TOKEN_PARTITIONS,
350 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
351 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
356 #if CONFIG_VP9_ENCODER
357 static const arg_def_t tile_cols =
358 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
359 static const arg_def_t tile_rows =
360 ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
361 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
362 static const arg_def_t frame_parallel_decoding = ARG_DEF(
363 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
364 static const arg_def_t aq_mode = ARG_DEF(
366 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
367 "3: cyclic refresh)");
368 static const arg_def_t frame_periodic_boost = ARG_DEF(
369 NULL, "frame_boost", 1,
370 "Enable frame periodic boost (0: off (default), 1: on)");
372 static const arg_def_t *vp9_args[] = {
373 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
374 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
375 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
376 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
379 static const int vp9_arg_ctrl_map[] = {
380 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
381 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
382 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
383 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
384 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
385 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
386 VP9E_SET_FRAME_PERIODIC_BOOST,
391 static const arg_def_t *no_args[] = { NULL };
396 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
399 fprintf(stderr, "\nOptions:\n");
400 arg_show_usage(stderr, main_args);
401 fprintf(stderr, "\nEncoder Global Options:\n");
402 arg_show_usage(stderr, global_args);
403 fprintf(stderr, "\nRate Control Options:\n");
404 arg_show_usage(stderr, rc_args);
405 fprintf(stderr, "\nTwopass Rate Control Options:\n");
406 arg_show_usage(stderr, rc_twopass_args);
407 fprintf(stderr, "\nKeyframe Placement Options:\n");
408 arg_show_usage(stderr, kf_args);
409 #if CONFIG_VP8_ENCODER
410 fprintf(stderr, "\nVP8 Specific Options:\n");
411 arg_show_usage(stderr, vp8_args);
413 #if CONFIG_VP9_ENCODER
414 fprintf(stderr, "\nVP9 Specific Options:\n");
415 arg_show_usage(stderr, vp9_args);
417 fprintf(stderr, "\nStream timebase (--timebase):\n"
418 " The desired precision of timestamps in the output, expressed\n"
419 " in fractional seconds. Default is 1/1000.\n");
420 fprintf(stderr, "\nIncluded encoders:\n\n");
422 for (i = 0; i < get_vpx_encoder_count(); ++i) {
423 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
424 fprintf(stderr, " %-6s - %s\n",
425 encoder->name, vpx_codec_iface_name(encoder->interface()));
431 #define mmin(a, b) ((a) < (b) ? (a) : (b))
432 static void find_mismatch(const vpx_image_t *const img1,
433 const vpx_image_t *const img2,
434 int yloc[4], int uloc[4], int vloc[4]) {
435 const uint32_t bsize = 64;
436 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
437 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
439 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
441 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
444 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
445 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
446 for (j = 0; match && j < img1->d_w; j += bsize) {
448 const int si = mmin(i + bsize, img1->d_h) - i;
449 const int sj = mmin(j + bsize, img1->d_w) - j;
450 for (k = 0; match && k < si; ++k) {
451 for (l = 0; match && l < sj; ++l) {
452 if (*(img1->planes[VPX_PLANE_Y] +
453 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
454 *(img2->planes[VPX_PLANE_Y] +
455 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
458 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
459 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
460 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
461 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
470 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
471 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
472 for (j = 0; match && j < c_w; j += bsizex) {
474 const int si = mmin(i + bsizey, c_h - i);
475 const int sj = mmin(j + bsizex, c_w - j);
476 for (k = 0; match && k < si; ++k) {
477 for (l = 0; match && l < sj; ++l) {
478 if (*(img1->planes[VPX_PLANE_U] +
479 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
480 *(img2->planes[VPX_PLANE_U] +
481 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
484 uloc[2] = *(img1->planes[VPX_PLANE_U] +
485 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
486 uloc[3] = *(img2->planes[VPX_PLANE_U] +
487 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
495 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
496 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
497 for (j = 0; match && j < c_w; j += bsizex) {
499 const int si = mmin(i + bsizey, c_h - i);
500 const int sj = mmin(j + bsizex, c_w - j);
501 for (k = 0; match && k < si; ++k) {
502 for (l = 0; match && l < sj; ++l) {
503 if (*(img1->planes[VPX_PLANE_V] +
504 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
505 *(img2->planes[VPX_PLANE_V] +
506 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
509 vloc[2] = *(img1->planes[VPX_PLANE_V] +
510 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
511 vloc[3] = *(img2->planes[VPX_PLANE_V] +
512 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
522 static int compare_img(const vpx_image_t *const img1,
523 const vpx_image_t *const img2) {
525 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
527 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
531 match &= (img1->fmt == img2->fmt);
532 match &= (img1->d_w == img2->d_w);
533 match &= (img1->d_h == img2->d_h);
535 for (i = 0; i < img1->d_h; ++i)
536 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
537 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
540 for (i = 0; i < c_h; ++i)
541 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
542 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
545 for (i = 0; i < c_h; ++i)
546 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
547 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
554 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
555 #define MAX(x,y) ((x)>(y)?(x):(y))
556 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
557 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
558 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
559 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
561 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
562 NELEMENTS(vp9_arg_ctrl_map))
566 typedef int stereo_format_t;
567 struct EbmlGlobal { int debug; };
570 /* Per-stream configuration */
571 struct stream_config {
572 struct vpx_codec_enc_cfg cfg;
574 const char *stats_fn;
575 stereo_format_t stereo_fmt;
576 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
579 int have_kf_max_dist;
583 struct stream_state {
585 struct stream_state *next;
586 struct stream_config config;
588 struct rate_hist *rate_hist;
589 struct EbmlGlobal ebml;
590 uint64_t psnr_sse_total;
591 uint64_t psnr_samples_total;
592 double psnr_totals[4];
595 vpx_codec_ctx_t encoder;
596 unsigned int frames_out;
600 struct vpx_image *img;
601 vpx_codec_ctx_t decoder;
606 void validate_positive_rational(const char *msg,
607 struct vpx_rational *rat) {
614 die("Error: %s must be positive\n", msg);
617 die("Error: %s has zero denominator\n", msg);
621 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
625 /* Initialize default parameters */
626 memset(global, 0, sizeof(*global));
627 global->codec = get_vpx_encoder_by_index(0);
629 global->use_i420 = 1;
630 /* Assign default deadline to good quality */
631 global->deadline = VPX_DL_GOOD_QUALITY;
633 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
636 if (arg_match(&arg, &codecarg, argi)) {
637 global->codec = get_vpx_encoder_by_name(arg.val);
639 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
640 } else if (arg_match(&arg, &passes, argi)) {
641 global->passes = arg_parse_uint(&arg);
643 if (global->passes < 1 || global->passes > 2)
644 die("Error: Invalid number of passes (%d)\n", global->passes);
645 } else if (arg_match(&arg, &pass_arg, argi)) {
646 global->pass = arg_parse_uint(&arg);
648 if (global->pass < 1 || global->pass > 2)
649 die("Error: Invalid pass selected (%d)\n",
651 } else if (arg_match(&arg, &usage, argi))
652 global->usage = arg_parse_uint(&arg);
653 else if (arg_match(&arg, &deadline, argi))
654 global->deadline = arg_parse_uint(&arg);
655 else if (arg_match(&arg, &best_dl, argi))
656 global->deadline = VPX_DL_BEST_QUALITY;
657 else if (arg_match(&arg, &good_dl, argi))
658 global->deadline = VPX_DL_GOOD_QUALITY;
659 else if (arg_match(&arg, &rt_dl, argi))
660 global->deadline = VPX_DL_REALTIME;
661 else if (arg_match(&arg, &use_yv12, argi))
662 global->use_i420 = 0;
663 else if (arg_match(&arg, &use_i420, argi))
664 global->use_i420 = 1;
665 else if (arg_match(&arg, &quietarg, argi))
667 else if (arg_match(&arg, &verbosearg, argi))
669 else if (arg_match(&arg, &limit, argi))
670 global->limit = arg_parse_uint(&arg);
671 else if (arg_match(&arg, &skip, argi))
672 global->skip_frames = arg_parse_uint(&arg);
673 else if (arg_match(&arg, &psnrarg, argi))
674 global->show_psnr = 1;
675 else if (arg_match(&arg, &recontest, argi))
676 global->test_decode = arg_parse_enum_or_int(&arg);
677 else if (arg_match(&arg, &framerate, argi)) {
678 global->framerate = arg_parse_rational(&arg);
679 validate_positive_rational(arg.name, &global->framerate);
680 global->have_framerate = 1;
681 } else if (arg_match(&arg, &out_part, argi))
682 global->out_part = 1;
683 else if (arg_match(&arg, &debugmode, argi))
685 else if (arg_match(&arg, &q_hist_n, argi))
686 global->show_q_hist_buckets = arg_parse_uint(&arg);
687 else if (arg_match(&arg, &rate_hist_n, argi))
688 global->show_rate_hist_buckets = arg_parse_uint(&arg);
689 else if (arg_match(&arg, &disable_warnings, argi))
690 global->disable_warnings = 1;
691 else if (arg_match(&arg, &disable_warning_prompt, argi))
692 global->disable_warning_prompt = 1;
693 else if (arg_match(&arg, &experimental_bitstream, argi))
694 global->experimental_bitstream = 1;
700 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
701 if (global->pass > global->passes) {
702 warn("Assuming --pass=%d implies --passes=%d\n",
703 global->pass, global->pass);
704 global->passes = global->pass;
707 /* Validate global config */
708 if (global->passes == 0) {
709 #if CONFIG_VP9_ENCODER
710 // Make default VP9 passes = 2 until there is a better quality 1-pass
712 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
713 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
719 if (global->deadline == VPX_DL_REALTIME &&
720 global->passes > 1) {
721 warn("Enforcing one-pass encoding in realtime mode\n");
727 void open_input_file(struct VpxInputContext *input) {
728 /* Parse certain options from the input file, if possible */
729 input->file = strcmp(input->filename, "-")
730 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
733 fatal("Failed to open input file");
735 if (!fseeko(input->file, 0, SEEK_END)) {
736 /* Input file is seekable. Figure out how long it is, so we can get
739 input->length = ftello(input->file);
743 /* For RAW input sources, these bytes will applied on the first frame
746 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
747 input->detect.position = 0;
749 if (input->detect.buf_read == 4
750 && file_is_y4m(input->detect.buf)) {
751 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
752 input->only_i420) >= 0) {
753 input->file_type = FILE_TYPE_Y4M;
754 input->width = input->y4m.pic_w;
755 input->height = input->y4m.pic_h;
756 input->framerate.numerator = input->y4m.fps_n;
757 input->framerate.denominator = input->y4m.fps_d;
758 input->fmt = input->y4m.vpx_fmt;
760 fatal("Unsupported Y4M stream.");
761 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
762 fatal("IVF is not supported as input.");
764 input->file_type = FILE_TYPE_RAW;
769 static void close_input_file(struct VpxInputContext *input) {
771 if (input->file_type == FILE_TYPE_Y4M)
772 y4m_input_close(&input->y4m);
775 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
776 struct stream_state *prev) {
777 struct stream_state *stream;
779 stream = calloc(1, sizeof(*stream));
781 fatal("Failed to allocate new stream.");
783 memcpy(stream, prev, sizeof(*stream));
789 /* Populate encoder configuration */
790 res = vpx_codec_enc_config_default(global->codec->interface(),
794 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
796 /* Change the default timebase to a high enough value so that the
797 * encoder will always create strictly increasing timestamps.
799 stream->config.cfg.g_timebase.den = 1000;
801 /* Never use the library's default resolution, require it be parsed
802 * from the file or set on the command line.
804 stream->config.cfg.g_w = 0;
805 stream->config.cfg.g_h = 0;
807 /* Initialize remaining stream parameters */
808 stream->config.write_webm = 1;
810 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
811 stream->ebml.last_pts_ns = -1;
812 stream->ebml.writer = NULL;
813 stream->ebml.segment = NULL;
816 /* Allows removal of the application version from the EBML tags */
817 stream->ebml.debug = global->debug;
819 /* Default lag_in_frames is 0 in realtime mode */
820 if (global->deadline == VPX_DL_REALTIME)
821 stream->config.cfg.g_lag_in_frames = 0;
824 /* Output files must be specified for each stream */
825 stream->config.out_fn = NULL;
832 static int parse_stream_params(struct VpxEncoderConfig *global,
833 struct stream_state *stream,
837 static const arg_def_t **ctrl_args = no_args;
838 static const int *ctrl_args_map = NULL;
839 struct stream_config *config = &stream->config;
840 int eos_mark_found = 0;
842 // Handle codec specific options
844 #if CONFIG_VP8_ENCODER
845 } else if (strcmp(global->codec->name, "vp8") == 0) {
846 ctrl_args = vp8_args;
847 ctrl_args_map = vp8_arg_ctrl_map;
849 #if CONFIG_VP9_ENCODER
850 } else if (strcmp(global->codec->name, "vp9") == 0) {
851 ctrl_args = vp9_args;
852 ctrl_args_map = vp9_arg_ctrl_map;
856 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
859 /* Once we've found an end-of-stream marker (--) we want to continue
860 * shifting arguments but not consuming them.
862 if (eos_mark_found) {
865 } else if (!strcmp(*argj, "--")) {
871 } else if (arg_match(&arg, &outputfile, argi)) {
872 config->out_fn = arg.val;
873 } else if (arg_match(&arg, &fpf_name, argi)) {
874 config->stats_fn = arg.val;
875 } else if (arg_match(&arg, &use_ivf, argi)) {
876 config->write_webm = 0;
877 } else if (arg_match(&arg, &threads, argi)) {
878 config->cfg.g_threads = arg_parse_uint(&arg);
879 } else if (arg_match(&arg, &profile, argi)) {
880 config->cfg.g_profile = arg_parse_uint(&arg);
881 } else if (arg_match(&arg, &width, argi)) {
882 config->cfg.g_w = arg_parse_uint(&arg);
883 } else if (arg_match(&arg, &height, argi)) {
884 config->cfg.g_h = arg_parse_uint(&arg);
886 } else if (arg_match(&arg, &stereo_mode, argi)) {
887 config->stereo_fmt = arg_parse_enum_or_int(&arg);
889 } else if (arg_match(&arg, &timebase, argi)) {
890 config->cfg.g_timebase = arg_parse_rational(&arg);
891 validate_positive_rational(arg.name, &config->cfg.g_timebase);
892 } else if (arg_match(&arg, &error_resilient, argi)) {
893 config->cfg.g_error_resilient = arg_parse_uint(&arg);
894 } else if (arg_match(&arg, &lag_in_frames, argi)) {
895 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
896 if (global->deadline == VPX_DL_REALTIME &&
897 config->cfg.g_lag_in_frames != 0) {
898 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
899 config->cfg.g_lag_in_frames = 0;
901 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
902 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
903 } else if (arg_match(&arg, &resize_allowed, argi)) {
904 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
905 } else if (arg_match(&arg, &resize_width, argi)) {
906 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
907 } else if (arg_match(&arg, &resize_height, argi)) {
908 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
909 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
910 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
911 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
912 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
913 } else if (arg_match(&arg, &end_usage, argi)) {
914 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
915 } else if (arg_match(&arg, &target_bitrate, argi)) {
916 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
917 } else if (arg_match(&arg, &min_quantizer, argi)) {
918 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
919 } else if (arg_match(&arg, &max_quantizer, argi)) {
920 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
921 } else if (arg_match(&arg, &undershoot_pct, argi)) {
922 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
923 } else if (arg_match(&arg, &overshoot_pct, argi)) {
924 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
925 } else if (arg_match(&arg, &buf_sz, argi)) {
926 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
927 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
928 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
929 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
930 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
931 } else if (arg_match(&arg, &bias_pct, argi)) {
932 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
933 if (global->passes < 2)
934 warn("option %s ignored in one-pass mode.\n", arg.name);
935 } else if (arg_match(&arg, &minsection_pct, argi)) {
936 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
938 if (global->passes < 2)
939 warn("option %s ignored in one-pass mode.\n", arg.name);
940 } else if (arg_match(&arg, &maxsection_pct, argi)) {
941 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
943 if (global->passes < 2)
944 warn("option %s ignored in one-pass mode.\n", arg.name);
945 } else if (arg_match(&arg, &kf_min_dist, argi)) {
946 config->cfg.kf_min_dist = arg_parse_uint(&arg);
947 } else if (arg_match(&arg, &kf_max_dist, argi)) {
948 config->cfg.kf_max_dist = arg_parse_uint(&arg);
949 config->have_kf_max_dist = 1;
950 } else if (arg_match(&arg, &kf_disabled, argi)) {
951 config->cfg.kf_mode = VPX_KF_DISABLED;
954 for (i = 0; ctrl_args[i]; i++) {
955 if (arg_match(&arg, ctrl_args[i], argi)) {
959 /* Point either to the next free element or the first
960 * instance of this control.
962 for (j = 0; j < config->arg_ctrl_cnt; j++)
963 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
967 assert(j < ARG_CTRL_CNT_MAX);
968 if (j < ARG_CTRL_CNT_MAX) {
969 config->arg_ctrls[j][0] = ctrl_args_map[i];
970 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
971 if (j == config->arg_ctrl_cnt)
972 config->arg_ctrl_cnt++;
981 return eos_mark_found;
985 #define FOREACH_STREAM(func) \
987 struct stream_state *stream; \
988 for (stream = streams; stream; stream = stream->next) { \
994 static void validate_stream_config(const struct stream_state *stream,
995 const struct VpxEncoderConfig *global) {
996 const struct stream_state *streami;
998 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
999 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1000 " and --height (-h)", stream->index);
1002 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
1003 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
1004 stream->index, stream->config.cfg.g_profile,
1005 experimental_bitstream.long_name);
1008 for (streami = stream; streami; streami = streami->next) {
1009 /* All streams require output files */
1010 if (!streami->config.out_fn)
1011 fatal("Stream %d: Output file is required (specify with -o)",
1014 /* Check for two streams outputting to the same file */
1015 if (streami != stream) {
1016 const char *a = stream->config.out_fn;
1017 const char *b = streami->config.out_fn;
1018 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1019 fatal("Stream %d: duplicate output file (from stream %d)",
1020 streami->index, stream->index);
1023 /* Check for two streams sharing a stats file. */
1024 if (streami != stream) {
1025 const char *a = stream->config.stats_fn;
1026 const char *b = streami->config.stats_fn;
1027 if (a && b && !strcmp(a, b))
1028 fatal("Stream %d: duplicate stats file (from stream %d)",
1029 streami->index, stream->index);
1035 static void set_stream_dimensions(struct stream_state *stream,
1038 if (!stream->config.cfg.g_w) {
1039 if (!stream->config.cfg.g_h)
1040 stream->config.cfg.g_w = w;
1042 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1044 if (!stream->config.cfg.g_h) {
1045 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1050 static void set_default_kf_interval(struct stream_state *stream,
1051 struct VpxEncoderConfig *global) {
1052 /* Use a max keyframe interval of 5 seconds, if none was
1053 * specified on the command line.
1055 if (!stream->config.have_kf_max_dist) {
1056 double framerate = (double)global->framerate.num / global->framerate.den;
1057 if (framerate > 0.0)
1058 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1062 static const char* file_type_to_string(enum VideoFileType t) {
1064 case FILE_TYPE_RAW: return "RAW";
1065 case FILE_TYPE_Y4M: return "Y4M";
1066 default: return "Other";
1070 static const char* image_format_to_string(vpx_img_fmt_t f) {
1072 case VPX_IMG_FMT_I420: return "I420";
1073 case VPX_IMG_FMT_I422: return "I422";
1074 case VPX_IMG_FMT_I444: return "I444";
1075 case VPX_IMG_FMT_YV12: return "YV12";
1076 default: return "Other";
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 File Type: %s Format: %s\n",
1092 file_type_to_string(input->file_type),
1093 image_format_to_string(input->fmt));
1095 if (stream->next || stream->index)
1096 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1097 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1098 fprintf(stderr, "Encoder parameters:\n");
1105 SHOW(g_timebase.num);
1106 SHOW(g_timebase.den);
1107 SHOW(g_error_resilient);
1109 SHOW(g_lag_in_frames);
1110 SHOW(rc_dropframe_thresh);
1111 SHOW(rc_resize_allowed);
1112 SHOW(rc_scaled_width);
1113 SHOW(rc_scaled_height);
1114 SHOW(rc_resize_up_thresh);
1115 SHOW(rc_resize_down_thresh);
1117 SHOW(rc_target_bitrate);
1118 SHOW(rc_min_quantizer);
1119 SHOW(rc_max_quantizer);
1120 SHOW(rc_undershoot_pct);
1121 SHOW(rc_overshoot_pct);
1123 SHOW(rc_buf_initial_sz);
1124 SHOW(rc_buf_optimal_sz);
1125 SHOW(rc_2pass_vbr_bias_pct);
1126 SHOW(rc_2pass_vbr_minsection_pct);
1127 SHOW(rc_2pass_vbr_maxsection_pct);
1134 static void open_output_file(struct stream_state *stream,
1135 struct VpxEncoderConfig *global) {
1136 const char *fn = stream->config.out_fn;
1137 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1139 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1142 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1145 fatal("Failed to open output file");
1147 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1148 fatal("WebM output to pipes not supported.");
1151 if (stream->config.write_webm) {
1152 stream->ebml.stream = stream->file;
1153 write_webm_file_header(&stream->ebml, cfg,
1155 stream->config.stereo_fmt,
1156 global->codec->fourcc);
1160 if (!stream->config.write_webm) {
1161 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1166 static void close_output_file(struct stream_state *stream,
1167 unsigned int fourcc) {
1168 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1170 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1174 if (stream->config.write_webm) {
1175 write_webm_file_footer(&stream->ebml);
1179 if (!stream->config.write_webm) {
1180 if (!fseek(stream->file, 0, SEEK_SET))
1181 ivf_write_file_header(stream->file, &stream->config.cfg,
1183 stream->frames_out);
1186 fclose(stream->file);
1190 static void setup_pass(struct stream_state *stream,
1191 struct VpxEncoderConfig *global,
1193 if (stream->config.stats_fn) {
1194 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1196 fatal("Failed to open statistics store");
1198 if (!stats_open_mem(&stream->stats, pass))
1199 fatal("Failed to open statistics store");
1202 stream->config.cfg.g_pass = global->passes == 2
1203 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1206 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1208 stream->cx_time = 0;
1210 stream->frames_out = 0;
1214 static void initialize_encoder(struct stream_state *stream,
1215 struct VpxEncoderConfig *global) {
1219 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1220 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1222 /* Construct Encoder Context */
1223 vpx_codec_enc_init(&stream->encoder, global->codec->interface(),
1224 &stream->config.cfg, flags);
1225 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1227 /* Note that we bypass the vpx_codec_control wrapper macro because
1228 * we're being clever to store the control IDs in an array. Real
1229 * applications will want to make use of the enumerations directly
1231 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1232 int ctrl = stream->config.arg_ctrls[i][0];
1233 int value = stream->config.arg_ctrls[i][1];
1234 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1235 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1238 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1242 if (global->test_decode != TEST_DECODE_OFF) {
1243 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1244 vpx_codec_dec_init(&stream->decoder, decoder->interface(), NULL, 0);
1250 static void encode_frame(struct stream_state *stream,
1251 struct VpxEncoderConfig *global,
1252 struct vpx_image *img,
1253 unsigned int frames_in) {
1254 vpx_codec_pts_t frame_start, next_frame_start;
1255 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1256 struct vpx_usec_timer timer;
1258 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1259 * global->framerate.den)
1260 / cfg->g_timebase.num / global->framerate.num;
1261 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1262 * global->framerate.den)
1263 / cfg->g_timebase.num / global->framerate.num;
1265 /* Scale if necessary */
1266 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1267 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1268 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1272 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1273 cfg->g_w, cfg->g_h, 16);
1274 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1275 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1276 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1278 stream->img->planes[VPX_PLANE_Y],
1279 stream->img->stride[VPX_PLANE_Y],
1280 stream->img->planes[VPX_PLANE_U],
1281 stream->img->stride[VPX_PLANE_U],
1282 stream->img->planes[VPX_PLANE_V],
1283 stream->img->stride[VPX_PLANE_V],
1284 stream->img->d_w, stream->img->d_h,
1290 vpx_usec_timer_start(&timer);
1291 vpx_codec_encode(&stream->encoder, img, frame_start,
1292 (unsigned long)(next_frame_start - frame_start),
1293 0, global->deadline);
1294 vpx_usec_timer_mark(&timer);
1295 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1296 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1301 static void update_quantizer_histogram(struct stream_state *stream) {
1302 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1305 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1306 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1307 stream->counts[q]++;
1312 static void get_cx_data(struct stream_state *stream,
1313 struct VpxEncoderConfig *global,
1315 const vpx_codec_cx_pkt_t *pkt;
1316 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1317 vpx_codec_iter_t iter = NULL;
1320 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1321 static size_t fsize = 0;
1322 static int64_t ivf_header_pos = 0;
1324 switch (pkt->kind) {
1325 case VPX_CODEC_CX_FRAME_PKT:
1326 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1327 stream->frames_out++;
1330 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1332 update_rate_histogram(stream->rate_hist, cfg, pkt);
1334 if (stream->config.write_webm) {
1335 write_webm_block(&stream->ebml, cfg, pkt);
1338 if (!stream->config.write_webm) {
1339 if (pkt->data.frame.partition_id <= 0) {
1340 ivf_header_pos = ftello(stream->file);
1341 fsize = pkt->data.frame.sz;
1343 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1345 fsize += pkt->data.frame.sz;
1347 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1348 const int64_t currpos = ftello(stream->file);
1349 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1350 ivf_write_frame_size(stream->file, fsize);
1351 fseeko(stream->file, currpos, SEEK_SET);
1355 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1358 stream->nbytes += pkt->data.raw.sz;
1362 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1363 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1364 (unsigned int)pkt->data.frame.sz, NULL, 0);
1365 if (stream->decoder.err) {
1366 warn_or_exit_on_error(&stream->decoder,
1367 global->test_decode == TEST_DECODE_FATAL,
1368 "Failed to decode frame %d in stream %d",
1369 stream->frames_out + 1, stream->index);
1370 stream->mismatch_seen = stream->frames_out + 1;
1375 case VPX_CODEC_STATS_PKT:
1376 stream->frames_out++;
1377 stats_write(&stream->stats,
1378 pkt->data.twopass_stats.buf,
1379 pkt->data.twopass_stats.sz);
1380 stream->nbytes += pkt->data.raw.sz;
1382 case VPX_CODEC_PSNR_PKT:
1384 if (global->show_psnr) {
1387 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1388 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1389 for (i = 0; i < 4; i++) {
1391 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1392 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1394 stream->psnr_count++;
1405 static void show_psnr(struct stream_state *stream) {
1409 if (!stream->psnr_count)
1412 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1413 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1414 (double)stream->psnr_sse_total);
1415 fprintf(stderr, " %.3f", ovpsnr);
1417 for (i = 0; i < 4; i++) {
1418 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1420 fprintf(stderr, "\n");
1424 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1425 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1429 static void test_decode(struct stream_state *stream,
1430 enum TestDecodeFatality fatal,
1431 const VpxInterface *codec) {
1432 vpx_image_t enc_img, dec_img;
1434 if (stream->mismatch_seen)
1437 /* Get the internal reference frame */
1438 if (strcmp(codec->name, "vp8") == 0) {
1439 struct vpx_ref_frame ref_enc, ref_dec;
1442 width = (stream->config.cfg.g_w + 15) & ~15;
1443 height = (stream->config.cfg.g_h + 15) & ~15;
1444 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1445 enc_img = ref_enc.img;
1446 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1447 dec_img = ref_dec.img;
1449 ref_enc.frame_type = VP8_LAST_FRAME;
1450 ref_dec.frame_type = VP8_LAST_FRAME;
1451 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1452 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1454 struct vp9_ref_frame ref;
1457 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
1459 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
1462 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1463 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1465 if (!compare_img(&enc_img, &dec_img)) {
1466 int y[4], u[4], v[4];
1467 find_mismatch(&enc_img, &dec_img, y, u, v);
1468 stream->decoder.err = 1;
1469 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1470 "Stream %d: Encode/decode mismatch on frame %d at"
1471 " Y[%d, %d] {%d/%d},"
1472 " U[%d, %d] {%d/%d},"
1473 " V[%d, %d] {%d/%d}",
1474 stream->index, stream->frames_out,
1475 y[0], y[1], y[2], y[3],
1476 u[0], u[1], u[2], u[3],
1477 v[0], v[1], v[2], v[3]);
1478 stream->mismatch_seen = stream->frames_out;
1481 vpx_img_free(&enc_img);
1482 vpx_img_free(&dec_img);
1486 static void print_time(const char *label, int64_t etl) {
1493 etl -= hours * 3600;
1498 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1499 label, hours, mins, secs);
1501 fprintf(stderr, "[%3s unknown] ", label);
1506 int main(int argc, const char **argv_) {
1509 int frame_avail, got_data;
1511 struct VpxInputContext input = {0};
1512 struct VpxEncoderConfig global;
1513 struct stream_state *streams = NULL;
1514 char **argv, **argi;
1515 uint64_t cx_time = 0;
1519 exec_name = argv_[0];
1524 /* Setup default input stream settings */
1525 input.framerate.numerator = 30;
1526 input.framerate.denominator = 1;
1527 input.only_i420 = 1;
1529 /* First parse the global configuration values, because we want to apply
1530 * other parameters on top of the default configuration provided by the
1533 argv = argv_dup(argc - 1, argv_ + 1);
1534 parse_global_config(&global, argv);
1536 input.fmt = global.use_i420 ? VPX_IMG_FMT_I420 : VPX_IMG_FMT_YV12;
1539 /* Now parse each stream's parameters. Using a local scope here
1540 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1543 struct stream_state *stream = NULL;
1546 stream = new_stream(&global, stream);
1550 } while (parse_stream_params(&global, stream, argv));
1553 /* Check for unrecognized options */
1554 for (argi = argv; *argi; argi++)
1555 if (argi[0][0] == '-' && argi[0][1])
1556 die("Error: Unrecognized option %s\n", *argi);
1558 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1559 &global, &stream->config.cfg););
1561 /* Handle non-option arguments */
1562 input.filename = argv[0];
1564 if (!input.filename)
1567 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1568 if (global.codec->fourcc == VP9_FOURCC)
1569 input.only_i420 = 0;
1571 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1572 int frames_in = 0, seen_frames = 0;
1573 int64_t estimated_time_left = -1;
1574 int64_t average_rate = -1;
1575 int64_t lagged_count = 0;
1577 open_input_file(&input);
1579 /* If the input file doesn't specify its w/h (raw files), try to get
1580 * the data from the first stream's configuration.
1582 if (!input.width || !input.height)
1584 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1585 input.width = stream->config.cfg.g_w;
1586 input.height = stream->config.cfg.g_h;
1591 /* Update stream configurations from the input file's parameters */
1592 if (!input.width || !input.height)
1593 fatal("Specify stream dimensions with --width (-w) "
1594 " and --height (-h)");
1595 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1596 FOREACH_STREAM(validate_stream_config(stream, &global));
1598 /* Ensure that --passes and --pass are consistent. If --pass is set and
1599 * --passes=2, ensure --fpf was set.
1601 if (global.pass && global.passes == 2)
1603 if (!stream->config.stats_fn)
1604 die("Stream %d: Must specify --fpf when --pass=%d"
1605 " and --passes=2\n", stream->index, global.pass);
1610 stream->config.write_webm = 0;
1611 warn("vpxenc was compiled without WebM container support."
1612 "Producing IVF output");
1616 /* Use the frame rate from the file only if none was specified
1617 * on the command-line.
1619 if (!global.have_framerate) {
1620 global.framerate.num = input.framerate.numerator;
1621 global.framerate.den = input.framerate.denominator;
1624 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1626 /* Show configuration */
1627 if (global.verbose && pass == 0)
1628 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1630 if (pass == (global.pass ? global.pass - 1 : 0)) {
1631 if (input.file_type == FILE_TYPE_Y4M)
1632 /*The Y4M reader does its own allocation.
1633 Just initialize this here to avoid problems if we never read any
1635 memset(&raw, 0, sizeof(raw));
1637 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1639 FOREACH_STREAM(stream->rate_hist =
1640 init_rate_histogram(&stream->config.cfg,
1641 &global.framerate));
1644 FOREACH_STREAM(setup_pass(stream, &global, pass));
1645 FOREACH_STREAM(open_output_file(stream, &global));
1646 FOREACH_STREAM(initialize_encoder(stream, &global));
1651 while (frame_avail || got_data) {
1652 struct vpx_usec_timer timer;
1654 if (!global.limit || frames_in < global.limit) {
1655 frame_avail = read_frame(&input, &raw);
1659 seen_frames = frames_in > global.skip_frames ?
1660 frames_in - global.skip_frames : 0;
1662 if (!global.quiet) {
1663 float fps = usec_to_fps(cx_time, seen_frames);
1664 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1666 if (stream_cnt == 1)
1668 "frame %4d/%-4d %7"PRId64"B ",
1669 frames_in, streams->frames_out, (int64_t)streams->nbytes);
1671 fprintf(stderr, "frame %4d ", frames_in);
1673 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
1674 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1675 cx_time > 9999999 ? "ms" : "us",
1676 fps >= 1.0 ? fps : fps * 60,
1677 fps >= 1.0 ? "fps" : "fpm");
1678 print_time("ETA", estimated_time_left);
1679 fprintf(stderr, "\033[K");
1685 if (frames_in > global.skip_frames) {
1686 vpx_usec_timer_start(&timer);
1687 FOREACH_STREAM(encode_frame(stream, &global,
1688 frame_avail ? &raw : NULL,
1690 vpx_usec_timer_mark(&timer);
1691 cx_time += vpx_usec_timer_elapsed(&timer);
1693 FOREACH_STREAM(update_quantizer_histogram(stream));
1696 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1698 if (!got_data && input.length && !streams->frames_out) {
1699 lagged_count = global.limit ? seen_frames : ftello(input.file);
1700 } else if (input.length) {
1705 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1707 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1708 remaining = 1000 * (global.limit - global.skip_frames
1709 - seen_frames + lagged_count);
1711 const int64_t input_pos = ftello(input.file);
1712 const int64_t input_pos_lagged = input_pos - lagged_count;
1713 const int64_t limit = input.length;
1715 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1716 remaining = limit - input_pos + lagged_count;
1719 average_rate = (average_rate <= 0)
1721 : (average_rate * 7 + rate) / 8;
1722 estimated_time_left = average_rate ? remaining / average_rate : -1;
1725 if (got_data && global.test_decode != TEST_DECODE_OFF)
1726 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1733 fprintf(stderr, "\n");
1736 FOREACH_STREAM(fprintf(
1738 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
1739 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
1740 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
1741 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
1742 seen_frames ? (int64_t)stream->nbytes * 8
1743 * (int64_t)global.framerate.num / global.framerate.den
1746 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1747 stream->cx_time > 9999999 ? "ms" : "us",
1748 usec_to_fps(stream->cx_time, seen_frames));
1751 if (global.show_psnr)
1752 FOREACH_STREAM(show_psnr(stream));
1754 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1756 if (global.test_decode != TEST_DECODE_OFF) {
1757 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1760 close_input_file(&input);
1762 if (global.test_decode == TEST_DECODE_FATAL) {
1763 FOREACH_STREAM(res |= stream->mismatch_seen);
1765 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1767 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1773 if (global.show_q_hist_buckets)
1774 FOREACH_STREAM(show_q_histogram(stream->counts,
1775 global.show_q_hist_buckets));
1777 if (global.show_rate_hist_buckets)
1778 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
1779 &stream->config.cfg,
1780 global.show_rate_hist_buckets));
1781 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1783 #if CONFIG_INTERNAL_STATS
1784 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1785 * to match some existing utilities.
1787 if (!(global.pass == 1 && global.passes == 2))
1789 FILE *f = fopen("opsnr.stt", "a");
1790 if (stream->mismatch_seen) {
1791 fprintf(f, "First mismatch occurred in frame %d\n",
1792 stream->mismatch_seen);
1794 fprintf(f, "No mismatch detected in recon buffers\n");
1803 return res ? EXIT_FAILURE : EXIT_SUCCESS;