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 use_i422 = ARG_DEF(NULL, "i422", 0,
137 "Input file is I422");
138 static const arg_def_t use_i444 = ARG_DEF(NULL, "i444", 0,
139 "Input file is I444");
140 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
142 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
143 "Number of passes (1/2)");
144 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
145 "Pass to execute (1/2)");
146 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
147 "First pass statistics file name");
148 #if CONFIG_FP_MB_STATS
149 static const arg_def_t fpmbf_name = ARG_DEF(NULL, "fpmbf", 1,
150 "First pass block statistics file name");
152 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
153 "Stop encoding after n input frames");
154 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
155 "Skip the first n input frames");
156 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
157 "Deadline per frame (usec)");
158 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
159 "Use Best Quality Deadline");
160 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
161 "Use Good Quality Deadline");
162 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
163 "Use Realtime Quality Deadline");
164 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
165 "Do not print encode progress");
166 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
167 "Show encoder parameters");
168 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
169 "Show PSNR in status line");
171 static const struct arg_enum_list test_decode_enum[] = {
172 {"off", TEST_DECODE_OFF},
173 {"fatal", TEST_DECODE_FATAL},
174 {"warn", TEST_DECODE_WARN},
177 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
178 "Test encode/decode mismatch",
180 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
181 "Stream frame rate (rate/scale)");
182 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
183 "Output IVF (default is WebM if WebM IO is enabled)");
184 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
185 "Makes encoder output partitions. Requires IVF output!");
186 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
187 "Show quantizer histogram (n-buckets)");
188 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
189 "Show rate histogram (n-buckets)");
190 static const arg_def_t disable_warnings =
191 ARG_DEF(NULL, "disable-warnings", 0,
192 "Disable warnings about potentially incorrect encode settings.");
193 static const arg_def_t disable_warning_prompt =
194 ARG_DEF("y", "disable-warning-prompt", 0,
195 "Display warnings, but do not prompt user to continue.");
196 static const arg_def_t experimental_bitstream =
197 ARG_DEF(NULL, "experimental-bitstream", 0,
198 "Allow experimental bitstream features.");
201 static const arg_def_t *main_args[] = {
203 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
204 &deadline, &best_dl, &good_dl, &rt_dl,
205 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
206 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
210 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
211 "Usage profile number to use");
212 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
213 "Max number of threads to use");
214 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
215 "Bitstream profile number to use");
216 static const arg_def_t width = ARG_DEF("w", "width", 1,
218 static const arg_def_t height = ARG_DEF("h", "height", 1,
221 static const struct arg_enum_list stereo_mode_enum[] = {
222 {"mono", STEREO_FORMAT_MONO},
223 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
224 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
225 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
226 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
229 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
230 "Stereo 3D video format", stereo_mode_enum);
232 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
233 "Output timestamp precision (fractional seconds)");
234 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
235 "Enable error resiliency features");
236 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
237 "Max number of frames to lag");
239 static const arg_def_t *global_args[] = {
240 &use_yv12, &use_i420, &use_i422, &use_i444,
241 &usage, &threads, &profile,
246 &timebase, &framerate,
251 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
252 "Temporal resampling threshold (buf %)");
253 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
254 "Spatial resampling enabled (bool)");
255 static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
256 "Width of encoded frame");
257 static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
258 "Height of encoded frame");
259 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
260 "Upscale threshold (buf %)");
261 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
262 "Downscale threshold (buf %)");
263 static const struct arg_enum_list end_usage_enum[] = {
270 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
271 "Rate control mode", end_usage_enum);
272 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
274 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
275 "Minimum (best) quantizer");
276 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
277 "Maximum (worst) quantizer");
278 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
279 "Datarate undershoot (min) target (%)");
280 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
281 "Datarate overshoot (max) target (%)");
282 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
283 "Client buffer size (ms)");
284 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
285 "Client initial buffer size (ms)");
286 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
287 "Client optimal buffer size (ms)");
288 static const arg_def_t *rc_args[] = {
289 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
290 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
291 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
292 &buf_initial_sz, &buf_optimal_sz, NULL
296 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
297 "CBR/VBR bias (0=CBR, 100=VBR)");
298 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
299 "GOP min bitrate (% of target)");
300 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
301 "GOP max bitrate (% of target)");
302 static const arg_def_t *rc_twopass_args[] = {
303 &bias_pct, &minsection_pct, &maxsection_pct, NULL
307 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
308 "Minimum keyframe interval (frames)");
309 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
310 "Maximum keyframe interval (frames)");
311 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
312 "Disable keyframe placement");
313 static const arg_def_t *kf_args[] = {
314 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
318 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
319 "Noise sensitivity (frames to blur)");
320 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
321 "Filter sharpness (0-7)");
322 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
323 "Motion detection threshold");
324 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
325 "CPU Used (-16..16)");
326 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
327 "Enable automatic alt reference frames");
328 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
329 "AltRef Max Frames");
330 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
332 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
334 static const struct arg_enum_list tuning_enum[] = {
335 {"psnr", VP8_TUNE_PSNR},
336 {"ssim", VP8_TUNE_SSIM},
339 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
340 "Material to favor", tuning_enum);
341 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
342 "Constant/Constrained Quality level");
343 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
344 "Max I-frame bitrate (pct)");
346 #if CONFIG_VP8_ENCODER
347 static const arg_def_t token_parts =
348 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
349 static const arg_def_t *vp8_args[] = {
350 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
351 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
352 &tune_ssim, &cq_level, &max_intra_rate_pct,
355 static const int vp8_arg_ctrl_map[] = {
356 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
357 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
358 VP8E_SET_TOKEN_PARTITIONS,
359 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
360 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
365 #if CONFIG_VP9_ENCODER
366 static const arg_def_t tile_cols =
367 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
368 static const arg_def_t tile_rows =
369 ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
370 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
371 static const arg_def_t frame_parallel_decoding = ARG_DEF(
372 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
373 static const arg_def_t aq_mode = ARG_DEF(
375 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
376 "3: cyclic refresh)");
377 static const arg_def_t frame_periodic_boost = ARG_DEF(
378 NULL, "frame_boost", 1,
379 "Enable frame periodic boost (0: off (default), 1: on)");
381 static const arg_def_t *vp9_args[] = {
382 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
383 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
384 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
385 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
388 static const int vp9_arg_ctrl_map[] = {
389 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
390 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
391 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
392 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
393 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
394 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
395 VP9E_SET_FRAME_PERIODIC_BOOST,
400 static const arg_def_t *no_args[] = { NULL };
405 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
408 fprintf(stderr, "\nOptions:\n");
409 arg_show_usage(stderr, main_args);
410 fprintf(stderr, "\nEncoder Global Options:\n");
411 arg_show_usage(stderr, global_args);
412 fprintf(stderr, "\nRate Control Options:\n");
413 arg_show_usage(stderr, rc_args);
414 fprintf(stderr, "\nTwopass Rate Control Options:\n");
415 arg_show_usage(stderr, rc_twopass_args);
416 fprintf(stderr, "\nKeyframe Placement Options:\n");
417 arg_show_usage(stderr, kf_args);
418 #if CONFIG_VP8_ENCODER
419 fprintf(stderr, "\nVP8 Specific Options:\n");
420 arg_show_usage(stderr, vp8_args);
422 #if CONFIG_VP9_ENCODER
423 fprintf(stderr, "\nVP9 Specific Options:\n");
424 arg_show_usage(stderr, vp9_args);
426 fprintf(stderr, "\nStream timebase (--timebase):\n"
427 " The desired precision of timestamps in the output, expressed\n"
428 " in fractional seconds. Default is 1/1000.\n");
429 fprintf(stderr, "\nIncluded encoders:\n\n");
431 for (i = 0; i < get_vpx_encoder_count(); ++i) {
432 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
433 fprintf(stderr, " %-6s - %s\n",
434 encoder->name, vpx_codec_iface_name(encoder->codec_interface()));
440 #define mmin(a, b) ((a) < (b) ? (a) : (b))
441 static void find_mismatch(const vpx_image_t *const img1,
442 const vpx_image_t *const img2,
443 int yloc[4], int uloc[4], int vloc[4]) {
444 const uint32_t bsize = 64;
445 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
446 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
448 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
450 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
453 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
454 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
455 for (j = 0; match && j < img1->d_w; j += bsize) {
457 const int si = mmin(i + bsize, img1->d_h) - i;
458 const int sj = mmin(j + bsize, img1->d_w) - j;
459 for (k = 0; match && k < si; ++k) {
460 for (l = 0; match && l < sj; ++l) {
461 if (*(img1->planes[VPX_PLANE_Y] +
462 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
463 *(img2->planes[VPX_PLANE_Y] +
464 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
467 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
468 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
469 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
470 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
479 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
480 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
481 for (j = 0; match && j < c_w; j += bsizex) {
483 const int si = mmin(i + bsizey, c_h - i);
484 const int sj = mmin(j + bsizex, c_w - j);
485 for (k = 0; match && k < si; ++k) {
486 for (l = 0; match && l < sj; ++l) {
487 if (*(img1->planes[VPX_PLANE_U] +
488 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
489 *(img2->planes[VPX_PLANE_U] +
490 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
493 uloc[2] = *(img1->planes[VPX_PLANE_U] +
494 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
495 uloc[3] = *(img2->planes[VPX_PLANE_U] +
496 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
504 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
505 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
506 for (j = 0; match && j < c_w; j += bsizex) {
508 const int si = mmin(i + bsizey, c_h - i);
509 const int sj = mmin(j + bsizex, c_w - j);
510 for (k = 0; match && k < si; ++k) {
511 for (l = 0; match && l < sj; ++l) {
512 if (*(img1->planes[VPX_PLANE_V] +
513 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
514 *(img2->planes[VPX_PLANE_V] +
515 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
518 vloc[2] = *(img1->planes[VPX_PLANE_V] +
519 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
520 vloc[3] = *(img2->planes[VPX_PLANE_V] +
521 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
531 static int compare_img(const vpx_image_t *const img1,
532 const vpx_image_t *const img2) {
534 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
536 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
540 match &= (img1->fmt == img2->fmt);
541 match &= (img1->d_w == img2->d_w);
542 match &= (img1->d_h == img2->d_h);
544 for (i = 0; i < img1->d_h; ++i)
545 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
546 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
549 for (i = 0; i < c_h; ++i)
550 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
551 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
554 for (i = 0; i < c_h; ++i)
555 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
556 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
563 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
564 #define MAX(x,y) ((x)>(y)?(x):(y))
565 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
566 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
567 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
568 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
570 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
571 NELEMENTS(vp9_arg_ctrl_map))
575 typedef int stereo_format_t;
576 struct EbmlGlobal { int debug; };
579 /* Per-stream configuration */
580 struct stream_config {
581 struct vpx_codec_enc_cfg cfg;
583 const char *stats_fn;
584 #if CONFIG_FP_MB_STATS
585 const char *fpmb_stats_fn;
587 stereo_format_t stereo_fmt;
588 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
591 int have_kf_max_dist;
595 struct stream_state {
597 struct stream_state *next;
598 struct stream_config config;
600 struct rate_hist *rate_hist;
601 struct EbmlGlobal ebml;
602 uint64_t psnr_sse_total;
603 uint64_t psnr_samples_total;
604 double psnr_totals[4];
607 vpx_codec_ctx_t encoder;
608 unsigned int frames_out;
612 #if CONFIG_FP_MB_STATS
613 stats_io_t fpmb_stats;
615 struct vpx_image *img;
616 vpx_codec_ctx_t decoder;
621 void validate_positive_rational(const char *msg,
622 struct vpx_rational *rat) {
629 die("Error: %s must be positive\n", msg);
632 die("Error: %s has zero denominator\n", msg);
636 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
640 /* Initialize default parameters */
641 memset(global, 0, sizeof(*global));
642 global->codec = get_vpx_encoder_by_index(0);
644 global->color_type = I420;
645 /* Assign default deadline to good quality */
646 global->deadline = VPX_DL_GOOD_QUALITY;
648 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
651 if (arg_match(&arg, &codecarg, argi)) {
652 global->codec = get_vpx_encoder_by_name(arg.val);
654 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
655 } else if (arg_match(&arg, &passes, argi)) {
656 global->passes = arg_parse_uint(&arg);
658 if (global->passes < 1 || global->passes > 2)
659 die("Error: Invalid number of passes (%d)\n", global->passes);
660 } else if (arg_match(&arg, &pass_arg, argi)) {
661 global->pass = arg_parse_uint(&arg);
663 if (global->pass < 1 || global->pass > 2)
664 die("Error: Invalid pass selected (%d)\n",
666 } else if (arg_match(&arg, &usage, argi))
667 global->usage = arg_parse_uint(&arg);
668 else if (arg_match(&arg, &deadline, argi))
669 global->deadline = arg_parse_uint(&arg);
670 else if (arg_match(&arg, &best_dl, argi))
671 global->deadline = VPX_DL_BEST_QUALITY;
672 else if (arg_match(&arg, &good_dl, argi))
673 global->deadline = VPX_DL_GOOD_QUALITY;
674 else if (arg_match(&arg, &rt_dl, argi))
675 global->deadline = VPX_DL_REALTIME;
676 else if (arg_match(&arg, &use_yv12, argi))
677 global->color_type = YV12;
678 else if (arg_match(&arg, &use_i420, argi))
679 global->color_type = I420;
680 else if (arg_match(&arg, &use_i422, argi))
681 global->color_type = I422;
682 else if (arg_match(&arg, &use_i444, argi))
683 global->color_type = I444;
684 else if (arg_match(&arg, &quietarg, argi))
686 else if (arg_match(&arg, &verbosearg, argi))
688 else if (arg_match(&arg, &limit, argi))
689 global->limit = arg_parse_uint(&arg);
690 else if (arg_match(&arg, &skip, argi))
691 global->skip_frames = arg_parse_uint(&arg);
692 else if (arg_match(&arg, &psnrarg, argi))
693 global->show_psnr = 1;
694 else if (arg_match(&arg, &recontest, argi))
695 global->test_decode = arg_parse_enum_or_int(&arg);
696 else if (arg_match(&arg, &framerate, argi)) {
697 global->framerate = arg_parse_rational(&arg);
698 validate_positive_rational(arg.name, &global->framerate);
699 global->have_framerate = 1;
700 } else if (arg_match(&arg, &out_part, argi))
701 global->out_part = 1;
702 else if (arg_match(&arg, &debugmode, argi))
704 else if (arg_match(&arg, &q_hist_n, argi))
705 global->show_q_hist_buckets = arg_parse_uint(&arg);
706 else if (arg_match(&arg, &rate_hist_n, argi))
707 global->show_rate_hist_buckets = arg_parse_uint(&arg);
708 else if (arg_match(&arg, &disable_warnings, argi))
709 global->disable_warnings = 1;
710 else if (arg_match(&arg, &disable_warning_prompt, argi))
711 global->disable_warning_prompt = 1;
712 else if (arg_match(&arg, &experimental_bitstream, argi))
713 global->experimental_bitstream = 1;
719 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
720 if (global->pass > global->passes) {
721 warn("Assuming --pass=%d implies --passes=%d\n",
722 global->pass, global->pass);
723 global->passes = global->pass;
726 /* Validate global config */
727 if (global->passes == 0) {
728 #if CONFIG_VP9_ENCODER
729 // Make default VP9 passes = 2 until there is a better quality 1-pass
731 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
732 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
738 if (global->deadline == VPX_DL_REALTIME &&
739 global->passes > 1) {
740 warn("Enforcing one-pass encoding in realtime mode\n");
746 void open_input_file(struct VpxInputContext *input) {
747 /* Parse certain options from the input file, if possible */
748 input->file = strcmp(input->filename, "-")
749 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
752 fatal("Failed to open input file");
754 if (!fseeko(input->file, 0, SEEK_END)) {
755 /* Input file is seekable. Figure out how long it is, so we can get
758 input->length = ftello(input->file);
762 /* For RAW input sources, these bytes will applied on the first frame
765 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
766 input->detect.position = 0;
768 if (input->detect.buf_read == 4
769 && file_is_y4m(input->detect.buf)) {
770 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
771 input->only_i420) >= 0) {
772 input->file_type = FILE_TYPE_Y4M;
773 input->width = input->y4m.pic_w;
774 input->height = input->y4m.pic_h;
775 input->framerate.numerator = input->y4m.fps_n;
776 input->framerate.denominator = input->y4m.fps_d;
777 input->fmt = input->y4m.vpx_fmt;
778 input->bit_depth = input->y4m.bit_depth;
780 fatal("Unsupported Y4M stream.");
781 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
782 fatal("IVF is not supported as input.");
784 input->file_type = FILE_TYPE_RAW;
789 static void close_input_file(struct VpxInputContext *input) {
791 if (input->file_type == FILE_TYPE_Y4M)
792 y4m_input_close(&input->y4m);
795 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
796 struct stream_state *prev) {
797 struct stream_state *stream;
799 stream = calloc(1, sizeof(*stream));
801 fatal("Failed to allocate new stream.");
803 memcpy(stream, prev, sizeof(*stream));
809 /* Populate encoder configuration */
810 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
814 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
816 /* Change the default timebase to a high enough value so that the
817 * encoder will always create strictly increasing timestamps.
819 stream->config.cfg.g_timebase.den = 1000;
821 /* Never use the library's default resolution, require it be parsed
822 * from the file or set on the command line.
824 stream->config.cfg.g_w = 0;
825 stream->config.cfg.g_h = 0;
827 /* Initialize remaining stream parameters */
828 stream->config.write_webm = 1;
830 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
831 stream->ebml.last_pts_ns = -1;
832 stream->ebml.writer = NULL;
833 stream->ebml.segment = NULL;
836 /* Allows removal of the application version from the EBML tags */
837 stream->ebml.debug = global->debug;
839 /* Default lag_in_frames is 0 in realtime mode */
840 if (global->deadline == VPX_DL_REALTIME)
841 stream->config.cfg.g_lag_in_frames = 0;
844 /* Output files must be specified for each stream */
845 stream->config.out_fn = NULL;
852 static int parse_stream_params(struct VpxEncoderConfig *global,
853 struct stream_state *stream,
857 static const arg_def_t **ctrl_args = no_args;
858 static const int *ctrl_args_map = NULL;
859 struct stream_config *config = &stream->config;
860 int eos_mark_found = 0;
862 // Handle codec specific options
864 #if CONFIG_VP8_ENCODER
865 } else if (strcmp(global->codec->name, "vp8") == 0) {
866 ctrl_args = vp8_args;
867 ctrl_args_map = vp8_arg_ctrl_map;
869 #if CONFIG_VP9_ENCODER
870 } else if (strcmp(global->codec->name, "vp9") == 0) {
871 ctrl_args = vp9_args;
872 ctrl_args_map = vp9_arg_ctrl_map;
876 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
879 /* Once we've found an end-of-stream marker (--) we want to continue
880 * shifting arguments but not consuming them.
882 if (eos_mark_found) {
885 } else if (!strcmp(*argj, "--")) {
891 } else if (arg_match(&arg, &outputfile, argi)) {
892 config->out_fn = arg.val;
893 } else if (arg_match(&arg, &fpf_name, argi)) {
894 config->stats_fn = arg.val;
895 #if CONFIG_FP_MB_STATS
896 } else if (arg_match(&arg, &fpmbf_name, argi)) {
897 config->fpmb_stats_fn = arg.val;
899 } else if (arg_match(&arg, &use_ivf, argi)) {
900 config->write_webm = 0;
901 } else if (arg_match(&arg, &threads, argi)) {
902 config->cfg.g_threads = arg_parse_uint(&arg);
903 } else if (arg_match(&arg, &profile, argi)) {
904 config->cfg.g_profile = arg_parse_uint(&arg);
905 } else if (arg_match(&arg, &width, argi)) {
906 config->cfg.g_w = arg_parse_uint(&arg);
907 } else if (arg_match(&arg, &height, argi)) {
908 config->cfg.g_h = arg_parse_uint(&arg);
910 } else if (arg_match(&arg, &stereo_mode, argi)) {
911 config->stereo_fmt = arg_parse_enum_or_int(&arg);
913 } else if (arg_match(&arg, &timebase, argi)) {
914 config->cfg.g_timebase = arg_parse_rational(&arg);
915 validate_positive_rational(arg.name, &config->cfg.g_timebase);
916 } else if (arg_match(&arg, &error_resilient, argi)) {
917 config->cfg.g_error_resilient = arg_parse_uint(&arg);
918 } else if (arg_match(&arg, &lag_in_frames, argi)) {
919 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
920 if (global->deadline == VPX_DL_REALTIME &&
921 config->cfg.g_lag_in_frames != 0) {
922 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
923 config->cfg.g_lag_in_frames = 0;
925 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
926 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
927 } else if (arg_match(&arg, &resize_allowed, argi)) {
928 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
929 } else if (arg_match(&arg, &resize_width, argi)) {
930 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
931 } else if (arg_match(&arg, &resize_height, argi)) {
932 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
933 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
934 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
935 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
936 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
937 } else if (arg_match(&arg, &end_usage, argi)) {
938 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
939 } else if (arg_match(&arg, &target_bitrate, argi)) {
940 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
941 } else if (arg_match(&arg, &min_quantizer, argi)) {
942 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
943 } else if (arg_match(&arg, &max_quantizer, argi)) {
944 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
945 } else if (arg_match(&arg, &undershoot_pct, argi)) {
946 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
947 } else if (arg_match(&arg, &overshoot_pct, argi)) {
948 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
949 } else if (arg_match(&arg, &buf_sz, argi)) {
950 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
951 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
952 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
953 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
954 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
955 } else if (arg_match(&arg, &bias_pct, argi)) {
956 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
957 if (global->passes < 2)
958 warn("option %s ignored in one-pass mode.\n", arg.name);
959 } else if (arg_match(&arg, &minsection_pct, argi)) {
960 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
962 if (global->passes < 2)
963 warn("option %s ignored in one-pass mode.\n", arg.name);
964 } else if (arg_match(&arg, &maxsection_pct, argi)) {
965 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
967 if (global->passes < 2)
968 warn("option %s ignored in one-pass mode.\n", arg.name);
969 } else if (arg_match(&arg, &kf_min_dist, argi)) {
970 config->cfg.kf_min_dist = arg_parse_uint(&arg);
971 } else if (arg_match(&arg, &kf_max_dist, argi)) {
972 config->cfg.kf_max_dist = arg_parse_uint(&arg);
973 config->have_kf_max_dist = 1;
974 } else if (arg_match(&arg, &kf_disabled, argi)) {
975 config->cfg.kf_mode = VPX_KF_DISABLED;
978 for (i = 0; ctrl_args[i]; i++) {
979 if (arg_match(&arg, ctrl_args[i], argi)) {
983 /* Point either to the next free element or the first
984 * instance of this control.
986 for (j = 0; j < config->arg_ctrl_cnt; j++)
987 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
991 assert(j < (int)ARG_CTRL_CNT_MAX);
992 if (j < (int)ARG_CTRL_CNT_MAX) {
993 config->arg_ctrls[j][0] = ctrl_args_map[i];
994 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
995 if (j == config->arg_ctrl_cnt)
996 config->arg_ctrl_cnt++;
1005 return eos_mark_found;
1009 #define FOREACH_STREAM(func) \
1011 struct stream_state *stream; \
1012 for (stream = streams; stream; stream = stream->next) { \
1018 static void validate_stream_config(const struct stream_state *stream,
1019 const struct VpxEncoderConfig *global) {
1020 const struct stream_state *streami;
1022 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1023 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1024 " and --height (-h)", stream->index);
1026 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
1027 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
1028 stream->index, stream->config.cfg.g_profile,
1029 experimental_bitstream.long_name);
1032 for (streami = stream; streami; streami = streami->next) {
1033 /* All streams require output files */
1034 if (!streami->config.out_fn)
1035 fatal("Stream %d: Output file is required (specify with -o)",
1038 /* Check for two streams outputting to the same file */
1039 if (streami != stream) {
1040 const char *a = stream->config.out_fn;
1041 const char *b = streami->config.out_fn;
1042 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1043 fatal("Stream %d: duplicate output file (from stream %d)",
1044 streami->index, stream->index);
1047 /* Check for two streams sharing a stats file. */
1048 if (streami != stream) {
1049 const char *a = stream->config.stats_fn;
1050 const char *b = streami->config.stats_fn;
1051 if (a && b && !strcmp(a, b))
1052 fatal("Stream %d: duplicate stats file (from stream %d)",
1053 streami->index, stream->index);
1056 #if CONFIG_FP_MB_STATS
1057 /* Check for two streams sharing a mb stats file. */
1058 if (streami != stream) {
1059 const char *a = stream->config.fpmb_stats_fn;
1060 const char *b = streami->config.fpmb_stats_fn;
1061 if (a && b && !strcmp(a, b))
1062 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1063 streami->index, stream->index);
1070 static void set_stream_dimensions(struct stream_state *stream,
1073 if (!stream->config.cfg.g_w) {
1074 if (!stream->config.cfg.g_h)
1075 stream->config.cfg.g_w = w;
1077 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1079 if (!stream->config.cfg.g_h) {
1080 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1085 static void set_default_kf_interval(struct stream_state *stream,
1086 struct VpxEncoderConfig *global) {
1087 /* Use a max keyframe interval of 5 seconds, if none was
1088 * specified on the command line.
1090 if (!stream->config.have_kf_max_dist) {
1091 double framerate = (double)global->framerate.num / global->framerate.den;
1092 if (framerate > 0.0)
1093 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1097 static const char* file_type_to_string(enum VideoFileType t) {
1099 case FILE_TYPE_RAW: return "RAW";
1100 case FILE_TYPE_Y4M: return "Y4M";
1101 default: return "Other";
1105 static const char* image_format_to_string(vpx_img_fmt_t f) {
1107 case VPX_IMG_FMT_I420: return "I420";
1108 case VPX_IMG_FMT_I422: return "I422";
1109 case VPX_IMG_FMT_I444: return "I444";
1110 case VPX_IMG_FMT_YV12: return "YV12";
1111 default: return "Other";
1115 static void show_stream_config(struct stream_state *stream,
1116 struct VpxEncoderConfig *global,
1117 struct VpxInputContext *input) {
1119 #define SHOW(field) \
1120 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1122 if (stream->index == 0) {
1123 fprintf(stderr, "Codec: %s\n",
1124 vpx_codec_iface_name(global->codec->codec_interface()));
1125 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1127 file_type_to_string(input->file_type),
1128 image_format_to_string(input->fmt));
1130 if (stream->next || stream->index)
1131 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1132 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1133 fprintf(stderr, "Encoder parameters:\n");
1140 SHOW(g_timebase.num);
1141 SHOW(g_timebase.den);
1142 SHOW(g_error_resilient);
1144 SHOW(g_lag_in_frames);
1145 SHOW(rc_dropframe_thresh);
1146 SHOW(rc_resize_allowed);
1147 SHOW(rc_scaled_width);
1148 SHOW(rc_scaled_height);
1149 SHOW(rc_resize_up_thresh);
1150 SHOW(rc_resize_down_thresh);
1152 SHOW(rc_target_bitrate);
1153 SHOW(rc_min_quantizer);
1154 SHOW(rc_max_quantizer);
1155 SHOW(rc_undershoot_pct);
1156 SHOW(rc_overshoot_pct);
1158 SHOW(rc_buf_initial_sz);
1159 SHOW(rc_buf_optimal_sz);
1160 SHOW(rc_2pass_vbr_bias_pct);
1161 SHOW(rc_2pass_vbr_minsection_pct);
1162 SHOW(rc_2pass_vbr_maxsection_pct);
1169 static void open_output_file(struct stream_state *stream,
1170 struct VpxEncoderConfig *global) {
1171 const char *fn = stream->config.out_fn;
1172 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1174 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1177 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1180 fatal("Failed to open output file");
1182 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1183 fatal("WebM output to pipes not supported.");
1186 if (stream->config.write_webm) {
1187 stream->ebml.stream = stream->file;
1188 write_webm_file_header(&stream->ebml, cfg,
1190 stream->config.stereo_fmt,
1191 global->codec->fourcc);
1195 if (!stream->config.write_webm) {
1196 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1201 static void close_output_file(struct stream_state *stream,
1202 unsigned int fourcc) {
1203 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1205 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1209 if (stream->config.write_webm) {
1210 write_webm_file_footer(&stream->ebml);
1214 if (!stream->config.write_webm) {
1215 if (!fseek(stream->file, 0, SEEK_SET))
1216 ivf_write_file_header(stream->file, &stream->config.cfg,
1218 stream->frames_out);
1221 fclose(stream->file);
1225 static void setup_pass(struct stream_state *stream,
1226 struct VpxEncoderConfig *global,
1228 if (stream->config.stats_fn) {
1229 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1231 fatal("Failed to open statistics store");
1233 if (!stats_open_mem(&stream->stats, pass))
1234 fatal("Failed to open statistics store");
1237 #if CONFIG_FP_MB_STATS
1238 if (stream->config.fpmb_stats_fn) {
1239 if (!stats_open_file(&stream->fpmb_stats,
1240 stream->config.fpmb_stats_fn, pass))
1241 fatal("Failed to open mb statistics store");
1243 if (!stats_open_mem(&stream->fpmb_stats, pass))
1244 fatal("Failed to open mb statistics store");
1248 stream->config.cfg.g_pass = global->passes == 2
1249 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1252 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1253 #if CONFIG_FP_MB_STATS
1254 stream->config.cfg.rc_firstpass_mb_stats_in =
1255 stats_get(&stream->fpmb_stats);
1259 stream->cx_time = 0;
1261 stream->frames_out = 0;
1265 static void initialize_encoder(struct stream_state *stream,
1266 struct VpxEncoderConfig *global) {
1270 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1271 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1273 /* Construct Encoder Context */
1274 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1275 &stream->config.cfg, flags);
1276 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1278 /* Note that we bypass the vpx_codec_control wrapper macro because
1279 * we're being clever to store the control IDs in an array. Real
1280 * applications will want to make use of the enumerations directly
1282 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1283 int ctrl = stream->config.arg_ctrls[i][0];
1284 int value = stream->config.arg_ctrls[i][1];
1285 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1286 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1289 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1293 if (global->test_decode != TEST_DECODE_OFF) {
1294 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1295 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1301 static void encode_frame(struct stream_state *stream,
1302 struct VpxEncoderConfig *global,
1303 struct vpx_image *img,
1304 unsigned int frames_in) {
1305 vpx_codec_pts_t frame_start, next_frame_start;
1306 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1307 struct vpx_usec_timer timer;
1309 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1310 * global->framerate.den)
1311 / cfg->g_timebase.num / global->framerate.num;
1312 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1313 * global->framerate.den)
1314 / cfg->g_timebase.num / global->framerate.num;
1316 /* Scale if necessary */
1317 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1318 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1319 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1324 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1325 cfg->g_w, cfg->g_h, 16);
1326 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1327 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1328 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1330 stream->img->planes[VPX_PLANE_Y],
1331 stream->img->stride[VPX_PLANE_Y],
1332 stream->img->planes[VPX_PLANE_U],
1333 stream->img->stride[VPX_PLANE_U],
1334 stream->img->planes[VPX_PLANE_V],
1335 stream->img->stride[VPX_PLANE_V],
1336 stream->img->d_w, stream->img->d_h,
1340 stream->encoder.err = 1;
1341 ctx_exit_on_error(&stream->encoder,
1342 "Stream %d: Failed to encode frame.\n"
1343 "Scaling disabled in this configuration. \n"
1344 "To enable, configure with --enable-libyuv\n",
1349 vpx_usec_timer_start(&timer);
1350 vpx_codec_encode(&stream->encoder, img, frame_start,
1351 (unsigned long)(next_frame_start - frame_start),
1352 0, global->deadline);
1353 vpx_usec_timer_mark(&timer);
1354 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1355 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1360 static void update_quantizer_histogram(struct stream_state *stream) {
1361 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1364 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1365 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1366 stream->counts[q]++;
1371 static void get_cx_data(struct stream_state *stream,
1372 struct VpxEncoderConfig *global,
1374 const vpx_codec_cx_pkt_t *pkt;
1375 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1376 vpx_codec_iter_t iter = NULL;
1379 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1380 static size_t fsize = 0;
1381 static int64_t ivf_header_pos = 0;
1383 switch (pkt->kind) {
1384 case VPX_CODEC_CX_FRAME_PKT:
1385 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1386 stream->frames_out++;
1389 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1391 update_rate_histogram(stream->rate_hist, cfg, pkt);
1393 if (stream->config.write_webm) {
1394 write_webm_block(&stream->ebml, cfg, pkt);
1397 if (!stream->config.write_webm) {
1398 if (pkt->data.frame.partition_id <= 0) {
1399 ivf_header_pos = ftello(stream->file);
1400 fsize = pkt->data.frame.sz;
1402 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1404 fsize += pkt->data.frame.sz;
1406 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1407 const int64_t currpos = ftello(stream->file);
1408 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1409 ivf_write_frame_size(stream->file, fsize);
1410 fseeko(stream->file, currpos, SEEK_SET);
1414 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1417 stream->nbytes += pkt->data.raw.sz;
1421 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1422 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1423 (unsigned int)pkt->data.frame.sz, NULL, 0);
1424 if (stream->decoder.err) {
1425 warn_or_exit_on_error(&stream->decoder,
1426 global->test_decode == TEST_DECODE_FATAL,
1427 "Failed to decode frame %d in stream %d",
1428 stream->frames_out + 1, stream->index);
1429 stream->mismatch_seen = stream->frames_out + 1;
1434 case VPX_CODEC_STATS_PKT:
1435 stream->frames_out++;
1436 stats_write(&stream->stats,
1437 pkt->data.twopass_stats.buf,
1438 pkt->data.twopass_stats.sz);
1439 stream->nbytes += pkt->data.raw.sz;
1441 #if CONFIG_FP_MB_STATS
1442 case VPX_CODEC_FPMB_STATS_PKT:
1443 stats_write(&stream->fpmb_stats,
1444 pkt->data.firstpass_mb_stats.buf,
1445 pkt->data.firstpass_mb_stats.sz);
1446 stream->nbytes += pkt->data.raw.sz;
1449 case VPX_CODEC_PSNR_PKT:
1451 if (global->show_psnr) {
1454 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1455 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1456 for (i = 0; i < 4; i++) {
1458 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1459 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1461 stream->psnr_count++;
1472 static void show_psnr(struct stream_state *stream) {
1476 if (!stream->psnr_count)
1479 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1480 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1481 (double)stream->psnr_sse_total);
1482 fprintf(stderr, " %.3f", ovpsnr);
1484 for (i = 0; i < 4; i++) {
1485 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1487 fprintf(stderr, "\n");
1491 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1492 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1496 static void test_decode(struct stream_state *stream,
1497 enum TestDecodeFatality fatal,
1498 const VpxInterface *codec) {
1499 vpx_image_t enc_img, dec_img;
1501 if (stream->mismatch_seen)
1504 /* Get the internal reference frame */
1505 if (strcmp(codec->name, "vp8") == 0) {
1506 struct vpx_ref_frame ref_enc, ref_dec;
1509 width = (stream->config.cfg.g_w + 15) & ~15;
1510 height = (stream->config.cfg.g_h + 15) & ~15;
1511 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1512 enc_img = ref_enc.img;
1513 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1514 dec_img = ref_dec.img;
1516 ref_enc.frame_type = VP8_LAST_FRAME;
1517 ref_dec.frame_type = VP8_LAST_FRAME;
1518 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1519 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1521 struct vp9_ref_frame ref;
1524 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
1526 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
1529 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1530 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1532 if (!compare_img(&enc_img, &dec_img)) {
1533 int y[4], u[4], v[4];
1534 find_mismatch(&enc_img, &dec_img, y, u, v);
1535 stream->decoder.err = 1;
1536 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1537 "Stream %d: Encode/decode mismatch on frame %d at"
1538 " Y[%d, %d] {%d/%d},"
1539 " U[%d, %d] {%d/%d},"
1540 " V[%d, %d] {%d/%d}",
1541 stream->index, stream->frames_out,
1542 y[0], y[1], y[2], y[3],
1543 u[0], u[1], u[2], u[3],
1544 v[0], v[1], v[2], v[3]);
1545 stream->mismatch_seen = stream->frames_out;
1548 vpx_img_free(&enc_img);
1549 vpx_img_free(&dec_img);
1553 static void print_time(const char *label, int64_t etl) {
1560 etl -= hours * 3600;
1565 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1566 label, hours, mins, secs);
1568 fprintf(stderr, "[%3s unknown] ", label);
1573 int main(int argc, const char **argv_) {
1576 int frame_avail, got_data;
1578 struct VpxInputContext input;
1579 struct VpxEncoderConfig global;
1580 struct stream_state *streams = NULL;
1581 char **argv, **argi;
1582 uint64_t cx_time = 0;
1586 memset(&input, 0, sizeof(input));
1587 exec_name = argv_[0];
1592 /* Setup default input stream settings */
1593 input.framerate.numerator = 30;
1594 input.framerate.denominator = 1;
1595 input.only_i420 = 1;
1596 input.bit_depth = 0;
1598 /* First parse the global configuration values, because we want to apply
1599 * other parameters on top of the default configuration provided by the
1602 argv = argv_dup(argc - 1, argv_ + 1);
1603 parse_global_config(&global, argv);
1605 switch (global.color_type) {
1607 input.fmt = VPX_IMG_FMT_I420;
1610 input.fmt = VPX_IMG_FMT_I422;
1613 input.fmt = VPX_IMG_FMT_I444;
1616 input.fmt = VPX_IMG_FMT_YV12;
1621 /* Now parse each stream's parameters. Using a local scope here
1622 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1625 struct stream_state *stream = NULL;
1628 stream = new_stream(&global, stream);
1632 } while (parse_stream_params(&global, stream, argv));
1635 /* Check for unrecognized options */
1636 for (argi = argv; *argi; argi++)
1637 if (argi[0][0] == '-' && argi[0][1])
1638 die("Error: Unrecognized option %s\n", *argi);
1640 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1641 &global, &stream->config.cfg););
1643 /* Handle non-option arguments */
1644 input.filename = argv[0];
1646 if (!input.filename)
1649 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1650 if (global.codec->fourcc == VP9_FOURCC)
1651 input.only_i420 = 0;
1653 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1654 int frames_in = 0, seen_frames = 0;
1655 int64_t estimated_time_left = -1;
1656 int64_t average_rate = -1;
1657 int64_t lagged_count = 0;
1659 open_input_file(&input);
1661 /* If the input file doesn't specify its w/h (raw files), try to get
1662 * the data from the first stream's configuration.
1664 if (!input.width || !input.height)
1666 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1667 input.width = stream->config.cfg.g_w;
1668 input.height = stream->config.cfg.g_h;
1673 /* Update stream configurations from the input file's parameters */
1674 if (!input.width || !input.height)
1675 fatal("Specify stream dimensions with --width (-w) "
1676 " and --height (-h)");
1677 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1678 FOREACH_STREAM(validate_stream_config(stream, &global));
1680 /* Ensure that --passes and --pass are consistent. If --pass is set and
1681 * --passes=2, ensure --fpf was set.
1683 if (global.pass && global.passes == 2)
1685 if (!stream->config.stats_fn)
1686 die("Stream %d: Must specify --fpf when --pass=%d"
1687 " and --passes=2\n", stream->index, global.pass);
1692 stream->config.write_webm = 0;
1693 warn("vpxenc was compiled without WebM container support."
1694 "Producing IVF output");
1698 /* Use the frame rate from the file only if none was specified
1699 * on the command-line.
1701 if (!global.have_framerate) {
1702 global.framerate.num = input.framerate.numerator;
1703 global.framerate.den = input.framerate.denominator;
1706 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1708 /* Show configuration */
1709 if (global.verbose && pass == 0)
1710 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1712 if (pass == (global.pass ? global.pass - 1 : 0)) {
1713 if (input.file_type == FILE_TYPE_Y4M)
1714 /*The Y4M reader does its own allocation.
1715 Just initialize this here to avoid problems if we never read any
1717 memset(&raw, 0, sizeof(raw));
1719 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1721 FOREACH_STREAM(stream->rate_hist =
1722 init_rate_histogram(&stream->config.cfg,
1723 &global.framerate));
1726 FOREACH_STREAM(setup_pass(stream, &global, pass));
1727 FOREACH_STREAM(open_output_file(stream, &global));
1728 FOREACH_STREAM(initialize_encoder(stream, &global));
1733 while (frame_avail || got_data) {
1734 struct vpx_usec_timer timer;
1736 if (!global.limit || frames_in < global.limit) {
1737 frame_avail = read_frame(&input, &raw);
1741 seen_frames = frames_in > global.skip_frames ?
1742 frames_in - global.skip_frames : 0;
1744 if (!global.quiet) {
1745 float fps = usec_to_fps(cx_time, seen_frames);
1746 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1748 if (stream_cnt == 1)
1750 "frame %4d/%-4d %7"PRId64"B ",
1751 frames_in, streams->frames_out, (int64_t)streams->nbytes);
1753 fprintf(stderr, "frame %4d ", frames_in);
1755 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
1756 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1757 cx_time > 9999999 ? "ms" : "us",
1758 fps >= 1.0 ? fps : fps * 60,
1759 fps >= 1.0 ? "fps" : "fpm");
1760 print_time("ETA", estimated_time_left);
1766 if (frames_in > global.skip_frames) {
1767 vpx_usec_timer_start(&timer);
1768 FOREACH_STREAM(encode_frame(stream, &global,
1769 frame_avail ? &raw : NULL,
1771 vpx_usec_timer_mark(&timer);
1772 cx_time += vpx_usec_timer_elapsed(&timer);
1774 FOREACH_STREAM(update_quantizer_histogram(stream));
1777 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1779 if (!got_data && input.length && !streams->frames_out) {
1780 lagged_count = global.limit ? seen_frames : ftello(input.file);
1781 } else if (input.length) {
1786 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1788 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1789 remaining = 1000 * (global.limit - global.skip_frames
1790 - seen_frames + lagged_count);
1792 const int64_t input_pos = ftello(input.file);
1793 const int64_t input_pos_lagged = input_pos - lagged_count;
1794 const int64_t limit = input.length;
1796 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1797 remaining = limit - input_pos + lagged_count;
1800 average_rate = (average_rate <= 0)
1802 : (average_rate * 7 + rate) / 8;
1803 estimated_time_left = average_rate ? remaining / average_rate : -1;
1806 if (got_data && global.test_decode != TEST_DECODE_OFF)
1807 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1812 fprintf(stderr, "\033[K");
1816 fprintf(stderr, "\n");
1819 FOREACH_STREAM(fprintf(
1821 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
1822 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
1823 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
1824 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
1825 seen_frames ? (int64_t)stream->nbytes * 8
1826 * (int64_t)global.framerate.num / global.framerate.den
1829 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1830 stream->cx_time > 9999999 ? "ms" : "us",
1831 usec_to_fps(stream->cx_time, seen_frames));
1834 if (global.show_psnr)
1835 FOREACH_STREAM(show_psnr(stream));
1837 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1839 if (global.test_decode != TEST_DECODE_OFF) {
1840 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1843 close_input_file(&input);
1845 if (global.test_decode == TEST_DECODE_FATAL) {
1846 FOREACH_STREAM(res |= stream->mismatch_seen);
1848 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1850 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1852 #if CONFIG_FP_MB_STATS
1853 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
1860 if (global.show_q_hist_buckets)
1861 FOREACH_STREAM(show_q_histogram(stream->counts,
1862 global.show_q_hist_buckets));
1864 if (global.show_rate_hist_buckets)
1865 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
1866 &stream->config.cfg,
1867 global.show_rate_hist_buckets));
1868 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1870 #if CONFIG_INTERNAL_STATS
1871 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1872 * to match some existing utilities.
1874 if (!(global.pass == 1 && global.passes == 2))
1876 FILE *f = fopen("opsnr.stt", "a");
1877 if (stream->mismatch_seen) {
1878 fprintf(f, "First mismatch occurred in frame %d\n",
1879 stream->mismatch_seen);
1881 fprintf(f, "No mismatch detected in recon buffers\n");
1890 return res ? EXIT_FAILURE : EXIT_SUCCESS;