2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
12 #include "./vpx_config.h"
22 #include "vpx/vpx_encoder.h"
24 #include "vpx/vpx_decoder.h"
27 #include "third_party/libyuv/include/libyuv/scale.h"
30 #include "./tools_common.h"
32 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
33 #include "vpx/vp8cx.h"
35 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
36 #include "vpx/vp8dx.h"
39 #include "vpx/vpx_integer.h"
40 #include "vpx_ports/mem_ops.h"
41 #include "vpx_ports/vpx_timer.h"
42 #include "./rate_hist.h"
43 #include "./vpxstats.h"
44 #include "./warnings.h"
45 #include "./webmenc.h"
46 #include "./y4minput.h"
48 /* Swallow warnings about unused results of fread/fwrite */
49 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
51 return fread(ptr, size, nmemb, stream);
53 #define fread wrap_fread
55 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
57 return fwrite(ptr, size, nmemb, stream);
59 #define fwrite wrap_fwrite
62 static const char *exec_name;
64 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
65 const char *s, va_list ap) {
67 const char *detail = vpx_codec_error_detail(ctx);
69 vfprintf(stderr, s, ap);
70 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
73 fprintf(stderr, " %s\n", detail);
80 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
84 warn_or_exit_on_errorv(ctx, 1, s, ap);
88 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
93 warn_or_exit_on_errorv(ctx, fatal, s, ap);
97 int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
98 FILE *f = input_ctx->file;
99 y4m_input *y4m = &input_ctx->y4m;
102 if (input_ctx->file_type == FILE_TYPE_Y4M) {
103 if (y4m_input_fetch_frame(y4m, f, img) < 1)
106 shortread = read_yuv_frame(input_ctx, img);
112 int file_is_y4m(const char detect[4]) {
113 if (memcmp(detect, "YUV4", 4) == 0) {
119 int fourcc_is_ivf(const char detect[4]) {
120 if (memcmp(detect, "DKIF", 4) == 0) {
126 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
127 "Debug mode (makes output deterministic)");
128 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
130 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
131 "Input file is YV12 ");
132 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
133 "Input file is I420 (default)");
134 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
136 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
137 "Number of passes (1/2)");
138 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
139 "Pass to execute (1/2)");
140 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
141 "First pass statistics file name");
142 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
143 "Stop encoding after n input frames");
144 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
145 "Skip the first n input frames");
146 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
147 "Deadline per frame (usec)");
148 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
149 "Use Best Quality Deadline");
150 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
151 "Use Good Quality Deadline");
152 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
153 "Use Realtime Quality Deadline");
154 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
155 "Do not print encode progress");
156 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
157 "Show encoder parameters");
158 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
159 "Show PSNR in status line");
161 static const struct arg_enum_list test_decode_enum[] = {
162 {"off", TEST_DECODE_OFF},
163 {"fatal", TEST_DECODE_FATAL},
164 {"warn", TEST_DECODE_WARN},
167 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
168 "Test encode/decode mismatch",
170 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
171 "Stream frame rate (rate/scale)");
172 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
173 "Output IVF (default is WebM if WebM IO is enabled)");
174 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
175 "Makes encoder output partitions. Requires IVF output!");
176 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
177 "Show quantizer histogram (n-buckets)");
178 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
179 "Show rate histogram (n-buckets)");
180 static const arg_def_t disable_warnings =
181 ARG_DEF(NULL, "disable-warnings", 0,
182 "Disable warnings about potentially incorrect encode settings.");
183 static const arg_def_t disable_warning_prompt =
184 ARG_DEF("y", "disable-warning-prompt", 0,
185 "Display warnings, but do not prompt user to continue.");
186 static const arg_def_t experimental_bitstream =
187 ARG_DEF(NULL, "experimental-bitstream", 0,
188 "Allow experimental bitstream features.");
191 static const arg_def_t *main_args[] = {
193 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
194 &deadline, &best_dl, &good_dl, &rt_dl,
195 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
196 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
200 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
201 "Usage profile number to use");
202 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
203 "Max number of threads to use");
204 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
205 "Bitstream profile number to use");
206 static const arg_def_t width = ARG_DEF("w", "width", 1,
208 static const arg_def_t height = ARG_DEF("h", "height", 1,
210 static const struct arg_enum_list stereo_mode_enum[] = {
211 {"mono", STEREO_FORMAT_MONO},
212 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
213 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
214 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
215 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
218 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
219 "Stereo 3D video format", stereo_mode_enum);
220 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
221 "Output timestamp precision (fractional seconds)");
222 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
223 "Enable error resiliency features");
224 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
225 "Max number of frames to lag");
227 static const arg_def_t *global_args[] = {
228 &use_yv12, &use_i420, &usage, &threads, &profile,
229 &width, &height, &stereo_mode, &timebase, &framerate,
234 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
235 "Temporal resampling threshold (buf %)");
236 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
237 "Spatial resampling enabled (bool)");
238 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
239 "Upscale threshold (buf %)");
240 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
241 "Downscale threshold (buf %)");
242 static const struct arg_enum_list end_usage_enum[] = {
249 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
250 "Rate control mode", end_usage_enum);
251 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
253 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
254 "Minimum (best) quantizer");
255 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
256 "Maximum (worst) quantizer");
257 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
258 "Datarate undershoot (min) target (%)");
259 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
260 "Datarate overshoot (max) target (%)");
261 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
262 "Client buffer size (ms)");
263 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
264 "Client initial buffer size (ms)");
265 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
266 "Client optimal buffer size (ms)");
267 static const arg_def_t *rc_args[] = {
268 &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
269 &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
270 &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
275 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
276 "CBR/VBR bias (0=CBR, 100=VBR)");
277 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
278 "GOP min bitrate (% of target)");
279 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
280 "GOP max bitrate (% of target)");
281 static const arg_def_t *rc_twopass_args[] = {
282 &bias_pct, &minsection_pct, &maxsection_pct, NULL
286 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
287 "Minimum keyframe interval (frames)");
288 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
289 "Maximum keyframe interval (frames)");
290 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
291 "Disable keyframe placement");
292 static const arg_def_t *kf_args[] = {
293 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
297 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
298 "Noise sensitivity (frames to blur)");
299 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
300 "Filter sharpness (0-7)");
301 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
302 "Motion detection threshold");
303 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
304 "CPU Used (-16..16)");
305 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
306 "Enable automatic alt reference frames");
307 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
308 "AltRef Max Frames");
309 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
311 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
313 static const struct arg_enum_list tuning_enum[] = {
314 {"psnr", VP8_TUNE_PSNR},
315 {"ssim", VP8_TUNE_SSIM},
318 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
319 "Material to favor", tuning_enum);
320 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
321 "Constant/Constrained Quality level");
322 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
323 "Max I-frame bitrate (pct)");
325 #if CONFIG_VP8_ENCODER
326 static const arg_def_t token_parts =
327 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
328 static const arg_def_t *vp8_args[] = {
329 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
330 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
331 &tune_ssim, &cq_level, &max_intra_rate_pct,
334 static const int vp8_arg_ctrl_map[] = {
335 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
336 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
337 VP8E_SET_TOKEN_PARTITIONS,
338 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
339 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
344 #if CONFIG_VP9_ENCODER
345 static const arg_def_t tile_cols =
346 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
347 static const arg_def_t tile_rows =
348 ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
349 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
350 static const arg_def_t frame_parallel_decoding = ARG_DEF(
351 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
352 static const arg_def_t aq_mode = ARG_DEF(
354 "Adaptive q mode (0: off (by default), 1: variance 2: complexity, "
355 "3: cyclic refresh)");
356 static const arg_def_t frame_periodic_boost = ARG_DEF(
357 NULL, "frame_boost", 1,
358 "Enable frame periodic boost (0: off (by default), 1: on)");
360 static const arg_def_t *vp9_args[] = {
361 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
362 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
363 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
364 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
367 static const int vp9_arg_ctrl_map[] = {
368 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
369 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
370 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
371 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
372 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
373 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
374 VP9E_SET_FRAME_PERIODIC_BOOST,
379 static const arg_def_t *no_args[] = { NULL };
384 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
387 fprintf(stderr, "\nOptions:\n");
388 arg_show_usage(stderr, main_args);
389 fprintf(stderr, "\nEncoder Global Options:\n");
390 arg_show_usage(stderr, global_args);
391 fprintf(stderr, "\nRate Control Options:\n");
392 arg_show_usage(stderr, rc_args);
393 fprintf(stderr, "\nTwopass Rate Control Options:\n");
394 arg_show_usage(stderr, rc_twopass_args);
395 fprintf(stderr, "\nKeyframe Placement Options:\n");
396 arg_show_usage(stderr, kf_args);
397 #if CONFIG_VP8_ENCODER
398 fprintf(stderr, "\nVP8 Specific Options:\n");
399 arg_show_usage(stderr, vp8_args);
401 #if CONFIG_VP9_ENCODER
402 fprintf(stderr, "\nVP9 Specific Options:\n");
403 arg_show_usage(stderr, vp9_args);
405 fprintf(stderr, "\nStream timebase (--timebase):\n"
406 " The desired precision of timestamps in the output, expressed\n"
407 " in fractional seconds. Default is 1/1000.\n");
408 fprintf(stderr, "\nIncluded encoders:\n\n");
410 for (i = 0; i < get_vpx_encoder_count(); ++i) {
411 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
412 fprintf(stderr, " %-6s - %s\n",
413 encoder->name, vpx_codec_iface_name(encoder->interface()));
419 #define mmin(a, b) ((a) < (b) ? (a) : (b))
420 static void find_mismatch(const vpx_image_t *const img1,
421 const vpx_image_t *const img2,
422 int yloc[4], int uloc[4], int vloc[4]) {
423 const uint32_t bsize = 64;
424 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
425 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
427 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
429 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
432 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
433 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
434 for (j = 0; match && j < img1->d_w; j += bsize) {
436 const int si = mmin(i + bsize, img1->d_h) - i;
437 const int sj = mmin(j + bsize, img1->d_w) - j;
438 for (k = 0; match && k < si; ++k) {
439 for (l = 0; match && l < sj; ++l) {
440 if (*(img1->planes[VPX_PLANE_Y] +
441 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
442 *(img2->planes[VPX_PLANE_Y] +
443 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
446 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
447 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
448 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
449 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
458 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
459 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
460 for (j = 0; match && j < c_w; j += bsizex) {
462 const int si = mmin(i + bsizey, c_h - i);
463 const int sj = mmin(j + bsizex, c_w - j);
464 for (k = 0; match && k < si; ++k) {
465 for (l = 0; match && l < sj; ++l) {
466 if (*(img1->planes[VPX_PLANE_U] +
467 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
468 *(img2->planes[VPX_PLANE_U] +
469 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
472 uloc[2] = *(img1->planes[VPX_PLANE_U] +
473 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
474 uloc[3] = *(img2->planes[VPX_PLANE_U] +
475 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
483 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
484 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
485 for (j = 0; match && j < c_w; j += bsizex) {
487 const int si = mmin(i + bsizey, c_h - i);
488 const int sj = mmin(j + bsizex, c_w - j);
489 for (k = 0; match && k < si; ++k) {
490 for (l = 0; match && l < sj; ++l) {
491 if (*(img1->planes[VPX_PLANE_V] +
492 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
493 *(img2->planes[VPX_PLANE_V] +
494 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
497 vloc[2] = *(img1->planes[VPX_PLANE_V] +
498 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
499 vloc[3] = *(img2->planes[VPX_PLANE_V] +
500 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
510 static int compare_img(const vpx_image_t *const img1,
511 const vpx_image_t *const img2) {
513 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
515 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
519 match &= (img1->fmt == img2->fmt);
520 match &= (img1->d_w == img2->d_w);
521 match &= (img1->d_h == img2->d_h);
523 for (i = 0; i < img1->d_h; ++i)
524 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
525 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
528 for (i = 0; i < c_h; ++i)
529 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
530 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
533 for (i = 0; i < c_h; ++i)
534 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
535 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
542 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
543 #define MAX(x,y) ((x)>(y)?(x):(y))
544 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
545 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
546 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
547 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
549 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
550 NELEMENTS(vp9_arg_ctrl_map))
553 /* Per-stream configuration */
554 struct stream_config {
555 struct vpx_codec_enc_cfg cfg;
557 const char *stats_fn;
558 stereo_format_t stereo_fmt;
559 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
562 int have_kf_max_dist;
566 struct stream_state {
568 struct stream_state *next;
569 struct stream_config config;
571 struct rate_hist *rate_hist;
572 struct EbmlGlobal ebml;
573 uint64_t psnr_sse_total;
574 uint64_t psnr_samples_total;
575 double psnr_totals[4];
578 vpx_codec_ctx_t encoder;
579 unsigned int frames_out;
583 struct vpx_image *img;
584 vpx_codec_ctx_t decoder;
589 void validate_positive_rational(const char *msg,
590 struct vpx_rational *rat) {
597 die("Error: %s must be positive\n", msg);
600 die("Error: %s has zero denominator\n", msg);
604 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
608 /* Initialize default parameters */
609 memset(global, 0, sizeof(*global));
610 global->codec = get_vpx_encoder_by_index(0);
612 global->use_i420 = 1;
613 /* Assign default deadline to good quality */
614 global->deadline = VPX_DL_GOOD_QUALITY;
616 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
619 if (arg_match(&arg, &codecarg, argi)) {
620 global->codec = get_vpx_encoder_by_name(arg.val);
622 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
623 } else if (arg_match(&arg, &passes, argi)) {
624 global->passes = arg_parse_uint(&arg);
626 if (global->passes < 1 || global->passes > 2)
627 die("Error: Invalid number of passes (%d)\n", global->passes);
628 } else if (arg_match(&arg, &pass_arg, argi)) {
629 global->pass = arg_parse_uint(&arg);
631 if (global->pass < 1 || global->pass > 2)
632 die("Error: Invalid pass selected (%d)\n",
634 } else if (arg_match(&arg, &usage, argi))
635 global->usage = arg_parse_uint(&arg);
636 else if (arg_match(&arg, &deadline, argi))
637 global->deadline = arg_parse_uint(&arg);
638 else if (arg_match(&arg, &best_dl, argi))
639 global->deadline = VPX_DL_BEST_QUALITY;
640 else if (arg_match(&arg, &good_dl, argi))
641 global->deadline = VPX_DL_GOOD_QUALITY;
642 else if (arg_match(&arg, &rt_dl, argi))
643 global->deadline = VPX_DL_REALTIME;
644 else if (arg_match(&arg, &use_yv12, argi))
645 global->use_i420 = 0;
646 else if (arg_match(&arg, &use_i420, argi))
647 global->use_i420 = 1;
648 else if (arg_match(&arg, &quietarg, argi))
650 else if (arg_match(&arg, &verbosearg, argi))
652 else if (arg_match(&arg, &limit, argi))
653 global->limit = arg_parse_uint(&arg);
654 else if (arg_match(&arg, &skip, argi))
655 global->skip_frames = arg_parse_uint(&arg);
656 else if (arg_match(&arg, &psnrarg, argi))
657 global->show_psnr = 1;
658 else if (arg_match(&arg, &recontest, argi))
659 global->test_decode = arg_parse_enum_or_int(&arg);
660 else if (arg_match(&arg, &framerate, argi)) {
661 global->framerate = arg_parse_rational(&arg);
662 validate_positive_rational(arg.name, &global->framerate);
663 global->have_framerate = 1;
664 } else if (arg_match(&arg, &out_part, argi))
665 global->out_part = 1;
666 else if (arg_match(&arg, &debugmode, argi))
668 else if (arg_match(&arg, &q_hist_n, argi))
669 global->show_q_hist_buckets = arg_parse_uint(&arg);
670 else if (arg_match(&arg, &rate_hist_n, argi))
671 global->show_rate_hist_buckets = arg_parse_uint(&arg);
672 else if (arg_match(&arg, &disable_warnings, argi))
673 global->disable_warnings = 1;
674 else if (arg_match(&arg, &disable_warning_prompt, argi))
675 global->disable_warning_prompt = 1;
676 else if (arg_match(&arg, &experimental_bitstream, argi))
677 global->experimental_bitstream = 1;
683 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
684 if (global->pass > global->passes) {
685 warn("Assuming --pass=%d implies --passes=%d\n",
686 global->pass, global->pass);
687 global->passes = global->pass;
690 /* Validate global config */
691 if (global->passes == 0) {
692 #if CONFIG_VP9_ENCODER
693 // Make default VP9 passes = 2 until there is a better quality 1-pass
695 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
696 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
702 if (global->deadline == VPX_DL_REALTIME &&
703 global->passes > 1) {
704 warn("Enforcing one-pass encoding in realtime mode\n");
710 void open_input_file(struct VpxInputContext *input) {
711 /* Parse certain options from the input file, if possible */
712 input->file = strcmp(input->filename, "-")
713 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
716 fatal("Failed to open input file");
718 if (!fseeko(input->file, 0, SEEK_END)) {
719 /* Input file is seekable. Figure out how long it is, so we can get
722 input->length = ftello(input->file);
726 /* For RAW input sources, these bytes will applied on the first frame
729 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
730 input->detect.position = 0;
732 if (input->detect.buf_read == 4
733 && file_is_y4m(input->detect.buf)) {
734 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
735 input->only_i420) >= 0) {
736 input->file_type = FILE_TYPE_Y4M;
737 input->width = input->y4m.pic_w;
738 input->height = input->y4m.pic_h;
739 input->framerate.numerator = input->y4m.fps_n;
740 input->framerate.denominator = input->y4m.fps_d;
743 fatal("Unsupported Y4M stream.");
744 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
745 fatal("IVF is not supported as input.");
747 input->file_type = FILE_TYPE_RAW;
752 static void close_input_file(struct VpxInputContext *input) {
754 if (input->file_type == FILE_TYPE_Y4M)
755 y4m_input_close(&input->y4m);
758 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
759 struct stream_state *prev) {
760 struct stream_state *stream;
762 stream = calloc(1, sizeof(*stream));
764 fatal("Failed to allocate new stream.");
766 memcpy(stream, prev, sizeof(*stream));
772 /* Populate encoder configuration */
773 res = vpx_codec_enc_config_default(global->codec->interface(),
777 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
779 /* Change the default timebase to a high enough value so that the
780 * encoder will always create strictly increasing timestamps.
782 stream->config.cfg.g_timebase.den = 1000;
784 /* Never use the library's default resolution, require it be parsed
785 * from the file or set on the command line.
787 stream->config.cfg.g_w = 0;
788 stream->config.cfg.g_h = 0;
790 /* Initialize remaining stream parameters */
791 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
792 stream->config.write_webm = 1;
794 stream->ebml.last_pts_ns = -1;
795 stream->ebml.writer = NULL;
796 stream->ebml.segment = NULL;
799 /* Allows removal of the application version from the EBML tags */
800 stream->ebml.debug = global->debug;
802 /* Default lag_in_frames is 0 in realtime mode */
803 if (global->deadline == VPX_DL_REALTIME)
804 stream->config.cfg.g_lag_in_frames = 0;
807 /* Output files must be specified for each stream */
808 stream->config.out_fn = NULL;
815 static int parse_stream_params(struct VpxEncoderConfig *global,
816 struct stream_state *stream,
820 static const arg_def_t **ctrl_args = no_args;
821 static const int *ctrl_args_map = NULL;
822 struct stream_config *config = &stream->config;
823 int eos_mark_found = 0;
825 // Handle codec specific options
827 #if CONFIG_VP8_ENCODER
828 } else if (strcmp(global->codec->name, "vp8") == 0) {
829 ctrl_args = vp8_args;
830 ctrl_args_map = vp8_arg_ctrl_map;
832 #if CONFIG_VP9_ENCODER
833 } else if (strcmp(global->codec->name, "vp9") == 0) {
834 ctrl_args = vp9_args;
835 ctrl_args_map = vp9_arg_ctrl_map;
839 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
842 /* Once we've found an end-of-stream marker (--) we want to continue
843 * shifting arguments but not consuming them.
845 if (eos_mark_found) {
848 } else if (!strcmp(*argj, "--")) {
854 } else if (arg_match(&arg, &outputfile, argi)) {
855 config->out_fn = arg.val;
856 } else if (arg_match(&arg, &fpf_name, argi)) {
857 config->stats_fn = arg.val;
858 } else if (arg_match(&arg, &use_ivf, argi)) {
859 config->write_webm = 0;
860 } else if (arg_match(&arg, &threads, argi)) {
861 config->cfg.g_threads = arg_parse_uint(&arg);
862 } else if (arg_match(&arg, &profile, argi)) {
863 config->cfg.g_profile = arg_parse_uint(&arg);
864 } else if (arg_match(&arg, &width, argi)) {
865 config->cfg.g_w = arg_parse_uint(&arg);
866 } else if (arg_match(&arg, &height, argi)) {
867 config->cfg.g_h = arg_parse_uint(&arg);
868 } else if (arg_match(&arg, &stereo_mode, argi)) {
869 config->stereo_fmt = arg_parse_enum_or_int(&arg);
870 } else if (arg_match(&arg, &timebase, argi)) {
871 config->cfg.g_timebase = arg_parse_rational(&arg);
872 validate_positive_rational(arg.name, &config->cfg.g_timebase);
873 } else if (arg_match(&arg, &error_resilient, argi)) {
874 config->cfg.g_error_resilient = arg_parse_uint(&arg);
875 } else if (arg_match(&arg, &lag_in_frames, argi)) {
876 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
877 if (global->deadline == VPX_DL_REALTIME &&
878 config->cfg.g_lag_in_frames != 0) {
879 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
880 config->cfg.g_lag_in_frames = 0;
882 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
883 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
884 } else if (arg_match(&arg, &resize_allowed, argi)) {
885 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
886 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
887 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
888 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
889 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
890 } else if (arg_match(&arg, &end_usage, argi)) {
891 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
892 } else if (arg_match(&arg, &target_bitrate, argi)) {
893 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
894 } else if (arg_match(&arg, &min_quantizer, argi)) {
895 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
896 } else if (arg_match(&arg, &max_quantizer, argi)) {
897 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
898 } else if (arg_match(&arg, &undershoot_pct, argi)) {
899 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
900 } else if (arg_match(&arg, &overshoot_pct, argi)) {
901 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
902 } else if (arg_match(&arg, &buf_sz, argi)) {
903 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
904 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
905 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
906 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
907 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
908 } else if (arg_match(&arg, &bias_pct, argi)) {
909 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
910 if (global->passes < 2)
911 warn("option %s ignored in one-pass mode.\n", arg.name);
912 } else if (arg_match(&arg, &minsection_pct, argi)) {
913 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
915 if (global->passes < 2)
916 warn("option %s ignored in one-pass mode.\n", arg.name);
917 } else if (arg_match(&arg, &maxsection_pct, argi)) {
918 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
920 if (global->passes < 2)
921 warn("option %s ignored in one-pass mode.\n", arg.name);
922 } else if (arg_match(&arg, &kf_min_dist, argi)) {
923 config->cfg.kf_min_dist = arg_parse_uint(&arg);
924 } else if (arg_match(&arg, &kf_max_dist, argi)) {
925 config->cfg.kf_max_dist = arg_parse_uint(&arg);
926 config->have_kf_max_dist = 1;
927 } else if (arg_match(&arg, &kf_disabled, argi)) {
928 config->cfg.kf_mode = VPX_KF_DISABLED;
931 for (i = 0; ctrl_args[i]; i++) {
932 if (arg_match(&arg, ctrl_args[i], argi)) {
936 /* Point either to the next free element or the first
937 * instance of this control.
939 for (j = 0; j < config->arg_ctrl_cnt; j++)
940 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
944 assert(j < ARG_CTRL_CNT_MAX);
945 if (j < ARG_CTRL_CNT_MAX) {
946 config->arg_ctrls[j][0] = ctrl_args_map[i];
947 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
948 if (j == config->arg_ctrl_cnt)
949 config->arg_ctrl_cnt++;
958 return eos_mark_found;
962 #define FOREACH_STREAM(func) \
964 struct stream_state *stream; \
965 for (stream = streams; stream; stream = stream->next) { \
971 static void validate_stream_config(const struct stream_state *stream,
972 const struct VpxEncoderConfig *global) {
973 const struct stream_state *streami;
975 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
976 fatal("Stream %d: Specify stream dimensions with --width (-w) "
977 " and --height (-h)", stream->index);
979 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
980 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
981 stream->index, stream->config.cfg.g_profile,
982 experimental_bitstream.long_name);
985 for (streami = stream; streami; streami = streami->next) {
986 /* All streams require output files */
987 if (!streami->config.out_fn)
988 fatal("Stream %d: Output file is required (specify with -o)",
991 /* Check for two streams outputting to the same file */
992 if (streami != stream) {
993 const char *a = stream->config.out_fn;
994 const char *b = streami->config.out_fn;
995 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
996 fatal("Stream %d: duplicate output file (from stream %d)",
997 streami->index, stream->index);
1000 /* Check for two streams sharing a stats file. */
1001 if (streami != stream) {
1002 const char *a = stream->config.stats_fn;
1003 const char *b = streami->config.stats_fn;
1004 if (a && b && !strcmp(a, b))
1005 fatal("Stream %d: duplicate stats file (from stream %d)",
1006 streami->index, stream->index);
1012 static void set_stream_dimensions(struct stream_state *stream,
1015 if (!stream->config.cfg.g_w) {
1016 if (!stream->config.cfg.g_h)
1017 stream->config.cfg.g_w = w;
1019 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1021 if (!stream->config.cfg.g_h) {
1022 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1027 static void set_default_kf_interval(struct stream_state *stream,
1028 struct VpxEncoderConfig *global) {
1029 /* Use a max keyframe interval of 5 seconds, if none was
1030 * specified on the command line.
1032 if (!stream->config.have_kf_max_dist) {
1033 double framerate = (double)global->framerate.num / global->framerate.den;
1034 if (framerate > 0.0)
1035 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1040 static void show_stream_config(struct stream_state *stream,
1041 struct VpxEncoderConfig *global,
1042 struct VpxInputContext *input) {
1044 #define SHOW(field) \
1045 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1047 if (stream->index == 0) {
1048 fprintf(stderr, "Codec: %s\n",
1049 vpx_codec_iface_name(global->codec->interface()));
1050 fprintf(stderr, "Source file: %s Format: %s\n", input->filename,
1051 input->use_i420 ? "I420" : "YV12");
1053 if (stream->next || stream->index)
1054 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1055 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1056 fprintf(stderr, "Encoder parameters:\n");
1063 SHOW(g_timebase.num);
1064 SHOW(g_timebase.den);
1065 SHOW(g_error_resilient);
1067 SHOW(g_lag_in_frames);
1068 SHOW(rc_dropframe_thresh);
1069 SHOW(rc_resize_allowed);
1070 SHOW(rc_resize_up_thresh);
1071 SHOW(rc_resize_down_thresh);
1073 SHOW(rc_target_bitrate);
1074 SHOW(rc_min_quantizer);
1075 SHOW(rc_max_quantizer);
1076 SHOW(rc_undershoot_pct);
1077 SHOW(rc_overshoot_pct);
1079 SHOW(rc_buf_initial_sz);
1080 SHOW(rc_buf_optimal_sz);
1081 SHOW(rc_2pass_vbr_bias_pct);
1082 SHOW(rc_2pass_vbr_minsection_pct);
1083 SHOW(rc_2pass_vbr_maxsection_pct);
1090 static void open_output_file(struct stream_state *stream,
1091 struct VpxEncoderConfig *global) {
1092 const char *fn = stream->config.out_fn;
1093 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1095 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1098 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1101 fatal("Failed to open output file");
1103 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1104 fatal("WebM output to pipes not supported.");
1107 if (stream->config.write_webm) {
1108 stream->ebml.stream = stream->file;
1109 write_webm_file_header(&stream->ebml, cfg,
1111 stream->config.stereo_fmt,
1112 global->codec->fourcc);
1116 if (!stream->config.write_webm) {
1117 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1122 static void close_output_file(struct stream_state *stream,
1123 unsigned int fourcc) {
1124 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1126 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1130 if (stream->config.write_webm) {
1131 write_webm_file_footer(&stream->ebml);
1135 if (!stream->config.write_webm) {
1136 if (!fseek(stream->file, 0, SEEK_SET))
1137 ivf_write_file_header(stream->file, &stream->config.cfg,
1139 stream->frames_out);
1142 fclose(stream->file);
1146 static void setup_pass(struct stream_state *stream,
1147 struct VpxEncoderConfig *global,
1149 if (stream->config.stats_fn) {
1150 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1152 fatal("Failed to open statistics store");
1154 if (!stats_open_mem(&stream->stats, pass))
1155 fatal("Failed to open statistics store");
1158 stream->config.cfg.g_pass = global->passes == 2
1159 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1162 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1164 stream->cx_time = 0;
1166 stream->frames_out = 0;
1170 static void initialize_encoder(struct stream_state *stream,
1171 struct VpxEncoderConfig *global) {
1175 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1176 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1178 /* Construct Encoder Context */
1179 vpx_codec_enc_init(&stream->encoder, global->codec->interface(),
1180 &stream->config.cfg, flags);
1181 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1183 /* Note that we bypass the vpx_codec_control wrapper macro because
1184 * we're being clever to store the control IDs in an array. Real
1185 * applications will want to make use of the enumerations directly
1187 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1188 int ctrl = stream->config.arg_ctrls[i][0];
1189 int value = stream->config.arg_ctrls[i][1];
1190 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1191 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1194 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1198 if (global->test_decode != TEST_DECODE_OFF) {
1199 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1200 vpx_codec_dec_init(&stream->decoder, decoder->interface(), NULL, 0);
1206 static void encode_frame(struct stream_state *stream,
1207 struct VpxEncoderConfig *global,
1208 struct vpx_image *img,
1209 unsigned int frames_in) {
1210 vpx_codec_pts_t frame_start, next_frame_start;
1211 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1212 struct vpx_usec_timer timer;
1214 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1215 * global->framerate.den)
1216 / cfg->g_timebase.num / global->framerate.num;
1217 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1218 * global->framerate.den)
1219 / cfg->g_timebase.num / global->framerate.num;
1221 /* Scale if necessary */
1222 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1224 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1225 cfg->g_w, cfg->g_h, 16);
1226 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1227 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1228 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1230 stream->img->planes[VPX_PLANE_Y],
1231 stream->img->stride[VPX_PLANE_Y],
1232 stream->img->planes[VPX_PLANE_U],
1233 stream->img->stride[VPX_PLANE_U],
1234 stream->img->planes[VPX_PLANE_V],
1235 stream->img->stride[VPX_PLANE_V],
1236 stream->img->d_w, stream->img->d_h,
1242 vpx_usec_timer_start(&timer);
1243 vpx_codec_encode(&stream->encoder, img, frame_start,
1244 (unsigned long)(next_frame_start - frame_start),
1245 0, global->deadline);
1246 vpx_usec_timer_mark(&timer);
1247 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1248 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1253 static void update_quantizer_histogram(struct stream_state *stream) {
1254 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1257 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1258 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1259 stream->counts[q]++;
1264 static void get_cx_data(struct stream_state *stream,
1265 struct VpxEncoderConfig *global,
1267 const vpx_codec_cx_pkt_t *pkt;
1268 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1269 vpx_codec_iter_t iter = NULL;
1272 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1273 static size_t fsize = 0;
1274 static off_t ivf_header_pos = 0;
1276 switch (pkt->kind) {
1277 case VPX_CODEC_CX_FRAME_PKT:
1278 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1279 stream->frames_out++;
1282 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1284 update_rate_histogram(stream->rate_hist, cfg, pkt);
1286 if (stream->config.write_webm) {
1287 write_webm_block(&stream->ebml, cfg, pkt);
1290 if (!stream->config.write_webm) {
1291 if (pkt->data.frame.partition_id <= 0) {
1292 ivf_header_pos = ftello(stream->file);
1293 fsize = pkt->data.frame.sz;
1295 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1297 fsize += pkt->data.frame.sz;
1299 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1300 off_t currpos = ftello(stream->file);
1301 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1302 ivf_write_frame_size(stream->file, fsize);
1303 fseeko(stream->file, currpos, SEEK_SET);
1307 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1310 stream->nbytes += pkt->data.raw.sz;
1314 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1315 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1316 (unsigned int)pkt->data.frame.sz, NULL, 0);
1317 if (stream->decoder.err) {
1318 warn_or_exit_on_error(&stream->decoder,
1319 global->test_decode == TEST_DECODE_FATAL,
1320 "Failed to decode frame %d in stream %d",
1321 stream->frames_out + 1, stream->index);
1322 stream->mismatch_seen = stream->frames_out + 1;
1327 case VPX_CODEC_STATS_PKT:
1328 stream->frames_out++;
1329 stats_write(&stream->stats,
1330 pkt->data.twopass_stats.buf,
1331 pkt->data.twopass_stats.sz);
1332 stream->nbytes += pkt->data.raw.sz;
1334 case VPX_CODEC_PSNR_PKT:
1336 if (global->show_psnr) {
1339 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1340 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1341 for (i = 0; i < 4; i++) {
1343 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1344 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1346 stream->psnr_count++;
1357 static void show_psnr(struct stream_state *stream) {
1361 if (!stream->psnr_count)
1364 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1365 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1366 (double)stream->psnr_sse_total);
1367 fprintf(stderr, " %.3f", ovpsnr);
1369 for (i = 0; i < 4; i++) {
1370 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1372 fprintf(stderr, "\n");
1376 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1377 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1381 static void test_decode(struct stream_state *stream,
1382 enum TestDecodeFatality fatal,
1383 const VpxInterface *codec) {
1384 vpx_image_t enc_img, dec_img;
1386 if (stream->mismatch_seen)
1389 /* Get the internal reference frame */
1390 if (strcmp(codec->name, "vp8") == 0) {
1391 struct vpx_ref_frame ref_enc, ref_dec;
1394 width = (stream->config.cfg.g_w + 15) & ~15;
1395 height = (stream->config.cfg.g_h + 15) & ~15;
1396 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1397 enc_img = ref_enc.img;
1398 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1399 dec_img = ref_dec.img;
1401 ref_enc.frame_type = VP8_LAST_FRAME;
1402 ref_dec.frame_type = VP8_LAST_FRAME;
1403 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1404 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1406 struct vp9_ref_frame ref;
1409 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
1411 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
1414 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1415 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1417 if (!compare_img(&enc_img, &dec_img)) {
1418 int y[4], u[4], v[4];
1419 find_mismatch(&enc_img, &dec_img, y, u, v);
1420 stream->decoder.err = 1;
1421 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1422 "Stream %d: Encode/decode mismatch on frame %d at"
1423 " Y[%d, %d] {%d/%d},"
1424 " U[%d, %d] {%d/%d},"
1425 " V[%d, %d] {%d/%d}",
1426 stream->index, stream->frames_out,
1427 y[0], y[1], y[2], y[3],
1428 u[0], u[1], u[2], u[3],
1429 v[0], v[1], v[2], v[3]);
1430 stream->mismatch_seen = stream->frames_out;
1433 vpx_img_free(&enc_img);
1434 vpx_img_free(&dec_img);
1438 static void print_time(const char *label, int64_t etl) {
1445 etl -= hours * 3600;
1450 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1451 label, hours, mins, secs);
1453 fprintf(stderr, "[%3s unknown] ", label);
1458 int main(int argc, const char **argv_) {
1461 int frame_avail, got_data;
1463 struct VpxInputContext input = {0};
1464 struct VpxEncoderConfig global;
1465 struct stream_state *streams = NULL;
1466 char **argv, **argi;
1467 uint64_t cx_time = 0;
1471 exec_name = argv_[0];
1476 /* Setup default input stream settings */
1477 input.framerate.numerator = 30;
1478 input.framerate.denominator = 1;
1480 input.only_i420 = 1;
1482 /* First parse the global configuration values, because we want to apply
1483 * other parameters on top of the default configuration provided by the
1486 argv = argv_dup(argc - 1, argv_ + 1);
1487 parse_global_config(&global, argv);
1491 /* Now parse each stream's parameters. Using a local scope here
1492 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1495 struct stream_state *stream = NULL;
1498 stream = new_stream(&global, stream);
1502 } while (parse_stream_params(&global, stream, argv));
1505 /* Check for unrecognized options */
1506 for (argi = argv; *argi; argi++)
1507 if (argi[0][0] == '-' && argi[0][1])
1508 die("Error: Unrecognized option %s\n", *argi);
1510 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1511 &global, &stream->config.cfg););
1513 /* Handle non-option arguments */
1514 input.filename = argv[0];
1516 if (!input.filename)
1519 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1520 if (global.codec->fourcc == VP9_FOURCC)
1521 input.only_i420 = 0;
1523 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1524 int frames_in = 0, seen_frames = 0;
1525 int64_t estimated_time_left = -1;
1526 int64_t average_rate = -1;
1527 off_t lagged_count = 0;
1529 open_input_file(&input);
1531 /* If the input file doesn't specify its w/h (raw files), try to get
1532 * the data from the first stream's configuration.
1534 if (!input.width || !input.height)
1536 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1537 input.width = stream->config.cfg.g_w;
1538 input.height = stream->config.cfg.g_h;
1543 /* Update stream configurations from the input file's parameters */
1544 if (!input.width || !input.height)
1545 fatal("Specify stream dimensions with --width (-w) "
1546 " and --height (-h)");
1547 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1548 FOREACH_STREAM(validate_stream_config(stream, &global));
1550 /* Ensure that --passes and --pass are consistent. If --pass is set and
1551 * --passes=2, ensure --fpf was set.
1553 if (global.pass && global.passes == 2)
1555 if (!stream->config.stats_fn)
1556 die("Stream %d: Must specify --fpf when --pass=%d"
1557 " and --passes=2\n", stream->index, global.pass);
1562 stream->config.write_webm = 0;
1563 warn("vpxenc was compiled without WebM container support."
1564 "Producing IVF output");
1568 /* Use the frame rate from the file only if none was specified
1569 * on the command-line.
1571 if (!global.have_framerate) {
1572 global.framerate.num = input.framerate.numerator;
1573 global.framerate.den = input.framerate.denominator;
1576 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1578 /* Show configuration */
1579 if (global.verbose && pass == 0)
1580 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1582 if (pass == (global.pass ? global.pass - 1 : 0)) {
1583 if (input.file_type == FILE_TYPE_Y4M)
1584 /*The Y4M reader does its own allocation.
1585 Just initialize this here to avoid problems if we never read any
1587 memset(&raw, 0, sizeof(raw));
1590 input.use_i420 ? VPX_IMG_FMT_I420
1592 input.width, input.height, 32);
1594 FOREACH_STREAM(stream->rate_hist =
1595 init_rate_histogram(&stream->config.cfg,
1596 &global.framerate));
1599 FOREACH_STREAM(setup_pass(stream, &global, pass));
1600 FOREACH_STREAM(open_output_file(stream, &global));
1601 FOREACH_STREAM(initialize_encoder(stream, &global));
1606 while (frame_avail || got_data) {
1607 struct vpx_usec_timer timer;
1609 if (!global.limit || frames_in < global.limit) {
1610 frame_avail = read_frame(&input, &raw);
1614 seen_frames = frames_in > global.skip_frames ?
1615 frames_in - global.skip_frames : 0;
1617 if (!global.quiet) {
1618 float fps = usec_to_fps(cx_time, seen_frames);
1619 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1621 if (stream_cnt == 1)
1623 "frame %4d/%-4d %7"PRId64"B ",
1624 frames_in, streams->frames_out, (int64_t)streams->nbytes);
1626 fprintf(stderr, "frame %4d ", frames_in);
1628 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
1629 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1630 cx_time > 9999999 ? "ms" : "us",
1631 fps >= 1.0 ? fps : fps * 60,
1632 fps >= 1.0 ? "fps" : "fpm");
1633 print_time("ETA", estimated_time_left);
1634 fprintf(stderr, "\033[K");
1640 if (frames_in > global.skip_frames) {
1641 vpx_usec_timer_start(&timer);
1642 FOREACH_STREAM(encode_frame(stream, &global,
1643 frame_avail ? &raw : NULL,
1645 vpx_usec_timer_mark(&timer);
1646 cx_time += vpx_usec_timer_elapsed(&timer);
1648 FOREACH_STREAM(update_quantizer_histogram(stream));
1651 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1653 if (!got_data && input.length && !streams->frames_out) {
1654 lagged_count = global.limit ? seen_frames : ftello(input.file);
1655 } else if (input.length) {
1660 off_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1662 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1663 remaining = 1000 * (global.limit - global.skip_frames
1664 - seen_frames + lagged_count);
1666 off_t input_pos = ftello(input.file);
1667 off_t input_pos_lagged = input_pos - lagged_count;
1668 int64_t limit = input.length;
1670 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1671 remaining = limit - input_pos + lagged_count;
1674 average_rate = (average_rate <= 0)
1676 : (average_rate * 7 + rate) / 8;
1677 estimated_time_left = average_rate ? remaining / average_rate : -1;
1680 if (got_data && global.test_decode != TEST_DECODE_OFF)
1681 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1688 fprintf(stderr, "\n");
1691 FOREACH_STREAM(fprintf(
1693 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
1694 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
1695 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
1696 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
1697 seen_frames ? (int64_t)stream->nbytes * 8
1698 * (int64_t)global.framerate.num / global.framerate.den
1701 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1702 stream->cx_time > 9999999 ? "ms" : "us",
1703 usec_to_fps(stream->cx_time, seen_frames));
1706 if (global.show_psnr)
1707 FOREACH_STREAM(show_psnr(stream));
1709 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1711 if (global.test_decode != TEST_DECODE_OFF) {
1712 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1715 close_input_file(&input);
1717 if (global.test_decode == TEST_DECODE_FATAL) {
1718 FOREACH_STREAM(res |= stream->mismatch_seen);
1720 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1722 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1728 if (global.show_q_hist_buckets)
1729 FOREACH_STREAM(show_q_histogram(stream->counts,
1730 global.show_q_hist_buckets));
1732 if (global.show_rate_hist_buckets)
1733 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
1734 &stream->config.cfg,
1735 global.show_rate_hist_buckets));
1736 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1738 #if CONFIG_INTERNAL_STATS
1739 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1740 * to match some existing utilities.
1742 if (!(global.pass == 1 && global.passes == 2))
1744 FILE *f = fopen("opsnr.stt", "a");
1745 if (stream->mismatch_seen) {
1746 fprintf(f, "First mismatch occurred in frame %d\n",
1747 stream->mismatch_seen);
1749 fprintf(f, "No mismatch detected in recon buffers\n");
1758 return res ? EXIT_FAILURE : EXIT_SUCCESS;