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"
23 #include "third_party/libyuv/include/libyuv/scale.h"
26 #include "vpx/vpx_encoder.h"
28 #include "vpx/vpx_decoder.h"
33 #include "./tools_common.h"
35 #if CONFIG_VP8_ENCODER || CONFIG_VP9_ENCODER
36 #include "vpx/vp8cx.h"
38 #if CONFIG_VP8_DECODER || CONFIG_VP9_DECODER
39 #include "vpx/vp8dx.h"
42 #include "vpx/vpx_integer.h"
43 #include "vpx_ports/mem_ops.h"
44 #include "vpx_ports/vpx_timer.h"
45 #include "./rate_hist.h"
46 #include "./vpxstats.h"
47 #include "./warnings.h"
49 #include "./webmenc.h"
51 #include "./y4minput.h"
53 /* Swallow warnings about unused results of fread/fwrite */
54 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
56 return fread(ptr, size, nmemb, stream);
58 #define fread wrap_fread
60 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
62 return fwrite(ptr, size, nmemb, stream);
64 #define fwrite wrap_fwrite
67 static const char *exec_name;
69 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
70 const char *s, va_list ap) {
72 const char *detail = vpx_codec_error_detail(ctx);
74 vfprintf(stderr, s, ap);
75 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
78 fprintf(stderr, " %s\n", detail);
85 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
89 warn_or_exit_on_errorv(ctx, 1, s, ap);
93 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
98 warn_or_exit_on_errorv(ctx, fatal, s, ap);
102 int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
103 FILE *f = input_ctx->file;
104 y4m_input *y4m = &input_ctx->y4m;
107 if (input_ctx->file_type == FILE_TYPE_Y4M) {
108 if (y4m_input_fetch_frame(y4m, f, img) < 1)
111 shortread = read_yuv_frame(input_ctx, img);
117 int file_is_y4m(const char detect[4]) {
118 if (memcmp(detect, "YUV4", 4) == 0) {
124 int fourcc_is_ivf(const char detect[4]) {
125 if (memcmp(detect, "DKIF", 4) == 0) {
131 static const arg_def_t debugmode = ARG_DEF("D", "debug", 0,
132 "Debug mode (makes output deterministic)");
133 static const arg_def_t outputfile = ARG_DEF("o", "output", 1,
135 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
136 "Input file is YV12 ");
137 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
138 "Input file is I420 (default)");
139 static const arg_def_t use_i422 = ARG_DEF(NULL, "i422", 0,
140 "Input file is I422");
141 static const arg_def_t use_i444 = ARG_DEF(NULL, "i444", 0,
142 "Input file is I444");
143 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
145 static const arg_def_t passes = ARG_DEF("p", "passes", 1,
146 "Number of passes (1/2)");
147 static const arg_def_t pass_arg = ARG_DEF(NULL, "pass", 1,
148 "Pass to execute (1/2)");
149 static const arg_def_t fpf_name = ARG_DEF(NULL, "fpf", 1,
150 "First pass statistics file name");
151 #if CONFIG_FP_MB_STATS
152 static const arg_def_t fpmbf_name = ARG_DEF(NULL, "fpmbf", 1,
153 "First pass block statistics file name");
155 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
156 "Stop encoding after n input frames");
157 static const arg_def_t skip = ARG_DEF(NULL, "skip", 1,
158 "Skip the first n input frames");
159 static const arg_def_t deadline = ARG_DEF("d", "deadline", 1,
160 "Deadline per frame (usec)");
161 static const arg_def_t best_dl = ARG_DEF(NULL, "best", 0,
162 "Use Best Quality Deadline");
163 static const arg_def_t good_dl = ARG_DEF(NULL, "good", 0,
164 "Use Good Quality Deadline");
165 static const arg_def_t rt_dl = ARG_DEF(NULL, "rt", 0,
166 "Use Realtime Quality Deadline");
167 static const arg_def_t quietarg = ARG_DEF("q", "quiet", 0,
168 "Do not print encode progress");
169 static const arg_def_t verbosearg = ARG_DEF("v", "verbose", 0,
170 "Show encoder parameters");
171 static const arg_def_t psnrarg = ARG_DEF(NULL, "psnr", 0,
172 "Show PSNR in status line");
174 static const struct arg_enum_list test_decode_enum[] = {
175 {"off", TEST_DECODE_OFF},
176 {"fatal", TEST_DECODE_FATAL},
177 {"warn", TEST_DECODE_WARN},
180 static const arg_def_t recontest = ARG_DEF_ENUM(NULL, "test-decode", 1,
181 "Test encode/decode mismatch",
183 static const arg_def_t framerate = ARG_DEF(NULL, "fps", 1,
184 "Stream frame rate (rate/scale)");
185 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0,
186 "Output IVF (default is WebM if WebM IO is enabled)");
187 static const arg_def_t out_part = ARG_DEF("P", "output-partitions", 0,
188 "Makes encoder output partitions. Requires IVF output!");
189 static const arg_def_t q_hist_n = ARG_DEF(NULL, "q-hist", 1,
190 "Show quantizer histogram (n-buckets)");
191 static const arg_def_t rate_hist_n = ARG_DEF(NULL, "rate-hist", 1,
192 "Show rate histogram (n-buckets)");
193 static const arg_def_t disable_warnings =
194 ARG_DEF(NULL, "disable-warnings", 0,
195 "Disable warnings about potentially incorrect encode settings.");
196 static const arg_def_t disable_warning_prompt =
197 ARG_DEF("y", "disable-warning-prompt", 0,
198 "Display warnings, but do not prompt user to continue.");
199 static const arg_def_t experimental_bitstream =
200 ARG_DEF(NULL, "experimental-bitstream", 0,
201 "Allow experimental bitstream features.");
204 static const arg_def_t *main_args[] = {
206 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
207 &deadline, &best_dl, &good_dl, &rt_dl,
208 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
209 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
213 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
214 "Usage profile number to use");
215 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
216 "Max number of threads to use");
217 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
218 "Bitstream profile number to use");
219 static const arg_def_t width = ARG_DEF("w", "width", 1,
221 static const arg_def_t height = ARG_DEF("h", "height", 1,
224 static const struct arg_enum_list stereo_mode_enum[] = {
225 {"mono", STEREO_FORMAT_MONO},
226 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
227 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
228 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
229 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
232 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
233 "Stereo 3D video format", stereo_mode_enum);
235 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
236 "Output timestamp precision (fractional seconds)");
237 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
238 "Enable error resiliency features");
239 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
240 "Max number of frames to lag");
242 static const arg_def_t *global_args[] = {
243 &use_yv12, &use_i420, &use_i422, &use_i444,
244 &usage, &threads, &profile,
249 &timebase, &framerate,
254 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
255 "Temporal resampling threshold (buf %)");
256 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
257 "Spatial resampling enabled (bool)");
258 static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
259 "Width of encoded frame");
260 static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
261 "Height of encoded frame");
262 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
263 "Upscale threshold (buf %)");
264 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
265 "Downscale threshold (buf %)");
266 static const struct arg_enum_list end_usage_enum[] = {
273 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
274 "Rate control mode", end_usage_enum);
275 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
277 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
278 "Minimum (best) quantizer");
279 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
280 "Maximum (worst) quantizer");
281 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
282 "Datarate undershoot (min) target (%)");
283 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
284 "Datarate overshoot (max) target (%)");
285 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
286 "Client buffer size (ms)");
287 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
288 "Client initial buffer size (ms)");
289 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
290 "Client optimal buffer size (ms)");
291 static const arg_def_t *rc_args[] = {
292 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
293 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
294 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
295 &buf_initial_sz, &buf_optimal_sz, NULL
299 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
300 "CBR/VBR bias (0=CBR, 100=VBR)");
301 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
302 "GOP min bitrate (% of target)");
303 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
304 "GOP max bitrate (% of target)");
305 static const arg_def_t *rc_twopass_args[] = {
306 &bias_pct, &minsection_pct, &maxsection_pct, NULL
310 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
311 "Minimum keyframe interval (frames)");
312 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
313 "Maximum keyframe interval (frames)");
314 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
315 "Disable keyframe placement");
316 static const arg_def_t *kf_args[] = {
317 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
321 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
322 "Noise sensitivity (frames to blur)");
323 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
324 "Filter sharpness (0-7)");
325 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
326 "Motion detection threshold");
327 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
328 "CPU Used (-16..16)");
329 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
330 "Enable automatic alt reference frames");
331 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
332 "AltRef Max Frames");
333 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
335 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
337 static const struct arg_enum_list tuning_enum[] = {
338 {"psnr", VP8_TUNE_PSNR},
339 {"ssim", VP8_TUNE_SSIM},
342 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
343 "Material to favor", tuning_enum);
344 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
345 "Constant/Constrained Quality level");
346 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
347 "Max I-frame bitrate (pct)");
349 #if CONFIG_VP8_ENCODER
350 static const arg_def_t token_parts =
351 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
352 static const arg_def_t *vp8_args[] = {
353 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
354 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
355 &tune_ssim, &cq_level, &max_intra_rate_pct,
358 static const int vp8_arg_ctrl_map[] = {
359 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
360 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
361 VP8E_SET_TOKEN_PARTITIONS,
362 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
363 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
368 #if CONFIG_VP9_ENCODER
369 static const arg_def_t tile_cols =
370 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
371 static const arg_def_t tile_rows =
372 ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
373 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
374 static const arg_def_t frame_parallel_decoding = ARG_DEF(
375 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
376 static const arg_def_t aq_mode = ARG_DEF(
378 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
379 "3: cyclic refresh)");
380 static const arg_def_t frame_periodic_boost = ARG_DEF(
381 NULL, "frame_boost", 1,
382 "Enable frame periodic boost (0: off (default), 1: on)");
384 static const struct arg_enum_list tune_content_enum[] = {
385 {"default", VP9E_CONTENT_DEFAULT},
386 {"screen", VP9E_CONTENT_SCREEN},
390 static const arg_def_t tune_content = ARG_DEF_ENUM(
391 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
393 static const arg_def_t *vp9_args[] = {
394 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
395 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
396 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
397 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost, &tune_content,
400 static const int vp9_arg_ctrl_map[] = {
401 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
402 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
403 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
404 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
405 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
406 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
407 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_TUNE_CONTENT,
412 static const arg_def_t *no_args[] = { NULL };
417 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
420 fprintf(stderr, "\nOptions:\n");
421 arg_show_usage(stderr, main_args);
422 fprintf(stderr, "\nEncoder Global Options:\n");
423 arg_show_usage(stderr, global_args);
424 fprintf(stderr, "\nRate Control Options:\n");
425 arg_show_usage(stderr, rc_args);
426 fprintf(stderr, "\nTwopass Rate Control Options:\n");
427 arg_show_usage(stderr, rc_twopass_args);
428 fprintf(stderr, "\nKeyframe Placement Options:\n");
429 arg_show_usage(stderr, kf_args);
430 #if CONFIG_VP8_ENCODER
431 fprintf(stderr, "\nVP8 Specific Options:\n");
432 arg_show_usage(stderr, vp8_args);
434 #if CONFIG_VP9_ENCODER
435 fprintf(stderr, "\nVP9 Specific Options:\n");
436 arg_show_usage(stderr, vp9_args);
438 fprintf(stderr, "\nStream timebase (--timebase):\n"
439 " The desired precision of timestamps in the output, expressed\n"
440 " in fractional seconds. Default is 1/1000.\n");
441 fprintf(stderr, "\nIncluded encoders:\n\n");
443 for (i = 0; i < get_vpx_encoder_count(); ++i) {
444 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
445 fprintf(stderr, " %-6s - %s\n",
446 encoder->name, vpx_codec_iface_name(encoder->codec_interface()));
452 #define mmin(a, b) ((a) < (b) ? (a) : (b))
453 static void find_mismatch(const vpx_image_t *const img1,
454 const vpx_image_t *const img2,
455 int yloc[4], int uloc[4], int vloc[4]) {
456 const uint32_t bsize = 64;
457 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
458 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
460 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
462 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
465 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
466 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
467 for (j = 0; match && j < img1->d_w; j += bsize) {
469 const int si = mmin(i + bsize, img1->d_h) - i;
470 const int sj = mmin(j + bsize, img1->d_w) - j;
471 for (k = 0; match && k < si; ++k) {
472 for (l = 0; match && l < sj; ++l) {
473 if (*(img1->planes[VPX_PLANE_Y] +
474 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
475 *(img2->planes[VPX_PLANE_Y] +
476 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
479 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
480 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
481 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
482 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
491 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
492 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
493 for (j = 0; match && j < c_w; j += bsizex) {
495 const int si = mmin(i + bsizey, c_h - i);
496 const int sj = mmin(j + bsizex, c_w - j);
497 for (k = 0; match && k < si; ++k) {
498 for (l = 0; match && l < sj; ++l) {
499 if (*(img1->planes[VPX_PLANE_U] +
500 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
501 *(img2->planes[VPX_PLANE_U] +
502 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
505 uloc[2] = *(img1->planes[VPX_PLANE_U] +
506 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
507 uloc[3] = *(img2->planes[VPX_PLANE_U] +
508 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
516 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
517 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
518 for (j = 0; match && j < c_w; j += bsizex) {
520 const int si = mmin(i + bsizey, c_h - i);
521 const int sj = mmin(j + bsizex, c_w - j);
522 for (k = 0; match && k < si; ++k) {
523 for (l = 0; match && l < sj; ++l) {
524 if (*(img1->planes[VPX_PLANE_V] +
525 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
526 *(img2->planes[VPX_PLANE_V] +
527 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
530 vloc[2] = *(img1->planes[VPX_PLANE_V] +
531 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
532 vloc[3] = *(img2->planes[VPX_PLANE_V] +
533 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
543 static int compare_img(const vpx_image_t *const img1,
544 const vpx_image_t *const img2) {
546 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
548 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
552 match &= (img1->fmt == img2->fmt);
553 match &= (img1->d_w == img2->d_w);
554 match &= (img1->d_h == img2->d_h);
556 for (i = 0; i < img1->d_h; ++i)
557 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
558 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
561 for (i = 0; i < c_h; ++i)
562 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
563 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
566 for (i = 0; i < c_h; ++i)
567 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
568 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
575 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
576 #define MAX(x,y) ((x)>(y)?(x):(y))
577 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
578 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
579 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
580 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
582 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
583 NELEMENTS(vp9_arg_ctrl_map))
587 typedef int stereo_format_t;
588 struct EbmlGlobal { int debug; };
591 /* Per-stream configuration */
592 struct stream_config {
593 struct vpx_codec_enc_cfg cfg;
595 const char *stats_fn;
596 #if CONFIG_FP_MB_STATS
597 const char *fpmb_stats_fn;
599 stereo_format_t stereo_fmt;
600 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
603 int have_kf_max_dist;
607 struct stream_state {
609 struct stream_state *next;
610 struct stream_config config;
612 struct rate_hist *rate_hist;
613 struct EbmlGlobal ebml;
614 uint64_t psnr_sse_total;
615 uint64_t psnr_samples_total;
616 double psnr_totals[4];
619 vpx_codec_ctx_t encoder;
620 unsigned int frames_out;
624 #if CONFIG_FP_MB_STATS
625 stats_io_t fpmb_stats;
627 struct vpx_image *img;
628 vpx_codec_ctx_t decoder;
633 void validate_positive_rational(const char *msg,
634 struct vpx_rational *rat) {
641 die("Error: %s must be positive\n", msg);
644 die("Error: %s has zero denominator\n", msg);
648 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
652 /* Initialize default parameters */
653 memset(global, 0, sizeof(*global));
654 global->codec = get_vpx_encoder_by_index(0);
656 global->color_type = I420;
657 /* Assign default deadline to good quality */
658 global->deadline = VPX_DL_GOOD_QUALITY;
660 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
663 if (arg_match(&arg, &codecarg, argi)) {
664 global->codec = get_vpx_encoder_by_name(arg.val);
666 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
667 } else if (arg_match(&arg, &passes, argi)) {
668 global->passes = arg_parse_uint(&arg);
670 if (global->passes < 1 || global->passes > 2)
671 die("Error: Invalid number of passes (%d)\n", global->passes);
672 } else if (arg_match(&arg, &pass_arg, argi)) {
673 global->pass = arg_parse_uint(&arg);
675 if (global->pass < 1 || global->pass > 2)
676 die("Error: Invalid pass selected (%d)\n",
678 } else if (arg_match(&arg, &usage, argi))
679 global->usage = arg_parse_uint(&arg);
680 else if (arg_match(&arg, &deadline, argi))
681 global->deadline = arg_parse_uint(&arg);
682 else if (arg_match(&arg, &best_dl, argi))
683 global->deadline = VPX_DL_BEST_QUALITY;
684 else if (arg_match(&arg, &good_dl, argi))
685 global->deadline = VPX_DL_GOOD_QUALITY;
686 else if (arg_match(&arg, &rt_dl, argi))
687 global->deadline = VPX_DL_REALTIME;
688 else if (arg_match(&arg, &use_yv12, argi))
689 global->color_type = YV12;
690 else if (arg_match(&arg, &use_i420, argi))
691 global->color_type = I420;
692 else if (arg_match(&arg, &use_i422, argi))
693 global->color_type = I422;
694 else if (arg_match(&arg, &use_i444, argi))
695 global->color_type = I444;
696 else if (arg_match(&arg, &quietarg, argi))
698 else if (arg_match(&arg, &verbosearg, argi))
700 else if (arg_match(&arg, &limit, argi))
701 global->limit = arg_parse_uint(&arg);
702 else if (arg_match(&arg, &skip, argi))
703 global->skip_frames = arg_parse_uint(&arg);
704 else if (arg_match(&arg, &psnrarg, argi))
705 global->show_psnr = 1;
706 else if (arg_match(&arg, &recontest, argi))
707 global->test_decode = arg_parse_enum_or_int(&arg);
708 else if (arg_match(&arg, &framerate, argi)) {
709 global->framerate = arg_parse_rational(&arg);
710 validate_positive_rational(arg.name, &global->framerate);
711 global->have_framerate = 1;
712 } else if (arg_match(&arg, &out_part, argi))
713 global->out_part = 1;
714 else if (arg_match(&arg, &debugmode, argi))
716 else if (arg_match(&arg, &q_hist_n, argi))
717 global->show_q_hist_buckets = arg_parse_uint(&arg);
718 else if (arg_match(&arg, &rate_hist_n, argi))
719 global->show_rate_hist_buckets = arg_parse_uint(&arg);
720 else if (arg_match(&arg, &disable_warnings, argi))
721 global->disable_warnings = 1;
722 else if (arg_match(&arg, &disable_warning_prompt, argi))
723 global->disable_warning_prompt = 1;
724 else if (arg_match(&arg, &experimental_bitstream, argi))
725 global->experimental_bitstream = 1;
731 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
732 if (global->pass > global->passes) {
733 warn("Assuming --pass=%d implies --passes=%d\n",
734 global->pass, global->pass);
735 global->passes = global->pass;
738 /* Validate global config */
739 if (global->passes == 0) {
740 #if CONFIG_VP9_ENCODER
741 // Make default VP9 passes = 2 until there is a better quality 1-pass
743 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
744 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
750 if (global->deadline == VPX_DL_REALTIME &&
751 global->passes > 1) {
752 warn("Enforcing one-pass encoding in realtime mode\n");
758 void open_input_file(struct VpxInputContext *input) {
759 /* Parse certain options from the input file, if possible */
760 input->file = strcmp(input->filename, "-")
761 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
764 fatal("Failed to open input file");
766 if (!fseeko(input->file, 0, SEEK_END)) {
767 /* Input file is seekable. Figure out how long it is, so we can get
770 input->length = ftello(input->file);
774 /* For RAW input sources, these bytes will applied on the first frame
777 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
778 input->detect.position = 0;
780 if (input->detect.buf_read == 4
781 && file_is_y4m(input->detect.buf)) {
782 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
783 input->only_i420) >= 0) {
784 input->file_type = FILE_TYPE_Y4M;
785 input->width = input->y4m.pic_w;
786 input->height = input->y4m.pic_h;
787 input->framerate.numerator = input->y4m.fps_n;
788 input->framerate.denominator = input->y4m.fps_d;
789 input->fmt = input->y4m.vpx_fmt;
790 input->bit_depth = input->y4m.bit_depth;
792 fatal("Unsupported Y4M stream.");
793 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
794 fatal("IVF is not supported as input.");
796 input->file_type = FILE_TYPE_RAW;
801 static void close_input_file(struct VpxInputContext *input) {
803 if (input->file_type == FILE_TYPE_Y4M)
804 y4m_input_close(&input->y4m);
807 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
808 struct stream_state *prev) {
809 struct stream_state *stream;
811 stream = calloc(1, sizeof(*stream));
813 fatal("Failed to allocate new stream.");
815 memcpy(stream, prev, sizeof(*stream));
821 /* Populate encoder configuration */
822 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
826 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
828 /* Change the default timebase to a high enough value so that the
829 * encoder will always create strictly increasing timestamps.
831 stream->config.cfg.g_timebase.den = 1000;
833 /* Never use the library's default resolution, require it be parsed
834 * from the file or set on the command line.
836 stream->config.cfg.g_w = 0;
837 stream->config.cfg.g_h = 0;
839 /* Initialize remaining stream parameters */
840 stream->config.write_webm = 1;
842 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
843 stream->ebml.last_pts_ns = -1;
844 stream->ebml.writer = NULL;
845 stream->ebml.segment = NULL;
848 /* Allows removal of the application version from the EBML tags */
849 stream->ebml.debug = global->debug;
851 /* Default lag_in_frames is 0 in realtime mode */
852 if (global->deadline == VPX_DL_REALTIME)
853 stream->config.cfg.g_lag_in_frames = 0;
856 /* Output files must be specified for each stream */
857 stream->config.out_fn = NULL;
864 static int parse_stream_params(struct VpxEncoderConfig *global,
865 struct stream_state *stream,
869 static const arg_def_t **ctrl_args = no_args;
870 static const int *ctrl_args_map = NULL;
871 struct stream_config *config = &stream->config;
872 int eos_mark_found = 0;
874 // Handle codec specific options
876 #if CONFIG_VP8_ENCODER
877 } else if (strcmp(global->codec->name, "vp8") == 0) {
878 ctrl_args = vp8_args;
879 ctrl_args_map = vp8_arg_ctrl_map;
881 #if CONFIG_VP9_ENCODER
882 } else if (strcmp(global->codec->name, "vp9") == 0) {
883 ctrl_args = vp9_args;
884 ctrl_args_map = vp9_arg_ctrl_map;
888 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
891 /* Once we've found an end-of-stream marker (--) we want to continue
892 * shifting arguments but not consuming them.
894 if (eos_mark_found) {
897 } else if (!strcmp(*argj, "--")) {
903 } else if (arg_match(&arg, &outputfile, argi)) {
904 config->out_fn = arg.val;
905 } else if (arg_match(&arg, &fpf_name, argi)) {
906 config->stats_fn = arg.val;
907 #if CONFIG_FP_MB_STATS
908 } else if (arg_match(&arg, &fpmbf_name, argi)) {
909 config->fpmb_stats_fn = arg.val;
911 } else if (arg_match(&arg, &use_ivf, argi)) {
912 config->write_webm = 0;
913 } else if (arg_match(&arg, &threads, argi)) {
914 config->cfg.g_threads = arg_parse_uint(&arg);
915 } else if (arg_match(&arg, &profile, argi)) {
916 config->cfg.g_profile = arg_parse_uint(&arg);
917 } else if (arg_match(&arg, &width, argi)) {
918 config->cfg.g_w = arg_parse_uint(&arg);
919 } else if (arg_match(&arg, &height, argi)) {
920 config->cfg.g_h = arg_parse_uint(&arg);
922 } else if (arg_match(&arg, &stereo_mode, argi)) {
923 config->stereo_fmt = arg_parse_enum_or_int(&arg);
925 } else if (arg_match(&arg, &timebase, argi)) {
926 config->cfg.g_timebase = arg_parse_rational(&arg);
927 validate_positive_rational(arg.name, &config->cfg.g_timebase);
928 } else if (arg_match(&arg, &error_resilient, argi)) {
929 config->cfg.g_error_resilient = arg_parse_uint(&arg);
930 } else if (arg_match(&arg, &lag_in_frames, argi)) {
931 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
932 if (global->deadline == VPX_DL_REALTIME &&
933 config->cfg.g_lag_in_frames != 0) {
934 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
935 config->cfg.g_lag_in_frames = 0;
937 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
938 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
939 } else if (arg_match(&arg, &resize_allowed, argi)) {
940 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
941 } else if (arg_match(&arg, &resize_width, argi)) {
942 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
943 } else if (arg_match(&arg, &resize_height, argi)) {
944 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
945 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
946 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
947 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
948 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
949 } else if (arg_match(&arg, &end_usage, argi)) {
950 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
951 } else if (arg_match(&arg, &target_bitrate, argi)) {
952 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
953 } else if (arg_match(&arg, &min_quantizer, argi)) {
954 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
955 } else if (arg_match(&arg, &max_quantizer, argi)) {
956 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
957 } else if (arg_match(&arg, &undershoot_pct, argi)) {
958 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
959 } else if (arg_match(&arg, &overshoot_pct, argi)) {
960 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
961 } else if (arg_match(&arg, &buf_sz, argi)) {
962 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
963 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
964 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
965 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
966 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
967 } else if (arg_match(&arg, &bias_pct, argi)) {
968 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
969 if (global->passes < 2)
970 warn("option %s ignored in one-pass mode.\n", arg.name);
971 } else if (arg_match(&arg, &minsection_pct, argi)) {
972 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
974 if (global->passes < 2)
975 warn("option %s ignored in one-pass mode.\n", arg.name);
976 } else if (arg_match(&arg, &maxsection_pct, argi)) {
977 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
979 if (global->passes < 2)
980 warn("option %s ignored in one-pass mode.\n", arg.name);
981 } else if (arg_match(&arg, &kf_min_dist, argi)) {
982 config->cfg.kf_min_dist = arg_parse_uint(&arg);
983 } else if (arg_match(&arg, &kf_max_dist, argi)) {
984 config->cfg.kf_max_dist = arg_parse_uint(&arg);
985 config->have_kf_max_dist = 1;
986 } else if (arg_match(&arg, &kf_disabled, argi)) {
987 config->cfg.kf_mode = VPX_KF_DISABLED;
990 for (i = 0; ctrl_args[i]; i++) {
991 if (arg_match(&arg, ctrl_args[i], argi)) {
995 /* Point either to the next free element or the first
996 * instance of this control.
998 for (j = 0; j < config->arg_ctrl_cnt; j++)
999 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
1003 assert(j < (int)ARG_CTRL_CNT_MAX);
1004 if (j < (int)ARG_CTRL_CNT_MAX) {
1005 config->arg_ctrls[j][0] = ctrl_args_map[i];
1006 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1007 if (j == config->arg_ctrl_cnt)
1008 config->arg_ctrl_cnt++;
1017 return eos_mark_found;
1021 #define FOREACH_STREAM(func) \
1023 struct stream_state *stream; \
1024 for (stream = streams; stream; stream = stream->next) { \
1030 static void validate_stream_config(const struct stream_state *stream,
1031 const struct VpxEncoderConfig *global) {
1032 const struct stream_state *streami;
1034 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1035 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1036 " and --height (-h)", stream->index);
1038 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
1039 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
1040 stream->index, stream->config.cfg.g_profile,
1041 experimental_bitstream.long_name);
1044 for (streami = stream; streami; streami = streami->next) {
1045 /* All streams require output files */
1046 if (!streami->config.out_fn)
1047 fatal("Stream %d: Output file is required (specify with -o)",
1050 /* Check for two streams outputting to the same file */
1051 if (streami != stream) {
1052 const char *a = stream->config.out_fn;
1053 const char *b = streami->config.out_fn;
1054 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1055 fatal("Stream %d: duplicate output file (from stream %d)",
1056 streami->index, stream->index);
1059 /* Check for two streams sharing a stats file. */
1060 if (streami != stream) {
1061 const char *a = stream->config.stats_fn;
1062 const char *b = streami->config.stats_fn;
1063 if (a && b && !strcmp(a, b))
1064 fatal("Stream %d: duplicate stats file (from stream %d)",
1065 streami->index, stream->index);
1068 #if CONFIG_FP_MB_STATS
1069 /* Check for two streams sharing a mb stats file. */
1070 if (streami != stream) {
1071 const char *a = stream->config.fpmb_stats_fn;
1072 const char *b = streami->config.fpmb_stats_fn;
1073 if (a && b && !strcmp(a, b))
1074 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1075 streami->index, stream->index);
1082 static void set_stream_dimensions(struct stream_state *stream,
1085 if (!stream->config.cfg.g_w) {
1086 if (!stream->config.cfg.g_h)
1087 stream->config.cfg.g_w = w;
1089 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1091 if (!stream->config.cfg.g_h) {
1092 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1097 static void set_default_kf_interval(struct stream_state *stream,
1098 struct VpxEncoderConfig *global) {
1099 /* Use a max keyframe interval of 5 seconds, if none was
1100 * specified on the command line.
1102 if (!stream->config.have_kf_max_dist) {
1103 double framerate = (double)global->framerate.num / global->framerate.den;
1104 if (framerate > 0.0)
1105 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1109 static const char* file_type_to_string(enum VideoFileType t) {
1111 case FILE_TYPE_RAW: return "RAW";
1112 case FILE_TYPE_Y4M: return "Y4M";
1113 default: return "Other";
1117 static const char* image_format_to_string(vpx_img_fmt_t f) {
1119 case VPX_IMG_FMT_I420: return "I420";
1120 case VPX_IMG_FMT_I422: return "I422";
1121 case VPX_IMG_FMT_I444: return "I444";
1122 case VPX_IMG_FMT_YV12: return "YV12";
1123 default: return "Other";
1127 static void show_stream_config(struct stream_state *stream,
1128 struct VpxEncoderConfig *global,
1129 struct VpxInputContext *input) {
1131 #define SHOW(field) \
1132 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1134 if (stream->index == 0) {
1135 fprintf(stderr, "Codec: %s\n",
1136 vpx_codec_iface_name(global->codec->codec_interface()));
1137 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1139 file_type_to_string(input->file_type),
1140 image_format_to_string(input->fmt));
1142 if (stream->next || stream->index)
1143 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1144 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1145 fprintf(stderr, "Encoder parameters:\n");
1152 SHOW(g_timebase.num);
1153 SHOW(g_timebase.den);
1154 SHOW(g_error_resilient);
1156 SHOW(g_lag_in_frames);
1157 SHOW(rc_dropframe_thresh);
1158 SHOW(rc_resize_allowed);
1159 SHOW(rc_scaled_width);
1160 SHOW(rc_scaled_height);
1161 SHOW(rc_resize_up_thresh);
1162 SHOW(rc_resize_down_thresh);
1164 SHOW(rc_target_bitrate);
1165 SHOW(rc_min_quantizer);
1166 SHOW(rc_max_quantizer);
1167 SHOW(rc_undershoot_pct);
1168 SHOW(rc_overshoot_pct);
1170 SHOW(rc_buf_initial_sz);
1171 SHOW(rc_buf_optimal_sz);
1172 SHOW(rc_2pass_vbr_bias_pct);
1173 SHOW(rc_2pass_vbr_minsection_pct);
1174 SHOW(rc_2pass_vbr_maxsection_pct);
1181 static void open_output_file(struct stream_state *stream,
1182 struct VpxEncoderConfig *global) {
1183 const char *fn = stream->config.out_fn;
1184 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1186 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1189 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1192 fatal("Failed to open output file");
1194 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1195 fatal("WebM output to pipes not supported.");
1198 if (stream->config.write_webm) {
1199 stream->ebml.stream = stream->file;
1200 write_webm_file_header(&stream->ebml, cfg,
1202 stream->config.stereo_fmt,
1203 global->codec->fourcc);
1207 if (!stream->config.write_webm) {
1208 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1213 static void close_output_file(struct stream_state *stream,
1214 unsigned int fourcc) {
1215 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1217 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1221 if (stream->config.write_webm) {
1222 write_webm_file_footer(&stream->ebml);
1226 if (!stream->config.write_webm) {
1227 if (!fseek(stream->file, 0, SEEK_SET))
1228 ivf_write_file_header(stream->file, &stream->config.cfg,
1230 stream->frames_out);
1233 fclose(stream->file);
1237 static void setup_pass(struct stream_state *stream,
1238 struct VpxEncoderConfig *global,
1240 if (stream->config.stats_fn) {
1241 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1243 fatal("Failed to open statistics store");
1245 if (!stats_open_mem(&stream->stats, pass))
1246 fatal("Failed to open statistics store");
1249 #if CONFIG_FP_MB_STATS
1250 if (stream->config.fpmb_stats_fn) {
1251 if (!stats_open_file(&stream->fpmb_stats,
1252 stream->config.fpmb_stats_fn, pass))
1253 fatal("Failed to open mb statistics store");
1255 if (!stats_open_mem(&stream->fpmb_stats, pass))
1256 fatal("Failed to open mb statistics store");
1260 stream->config.cfg.g_pass = global->passes == 2
1261 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1264 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1265 #if CONFIG_FP_MB_STATS
1266 stream->config.cfg.rc_firstpass_mb_stats_in =
1267 stats_get(&stream->fpmb_stats);
1271 stream->cx_time = 0;
1273 stream->frames_out = 0;
1277 static void initialize_encoder(struct stream_state *stream,
1278 struct VpxEncoderConfig *global) {
1282 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1283 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1285 /* Construct Encoder Context */
1286 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1287 &stream->config.cfg, flags);
1288 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1290 /* Note that we bypass the vpx_codec_control wrapper macro because
1291 * we're being clever to store the control IDs in an array. Real
1292 * applications will want to make use of the enumerations directly
1294 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1295 int ctrl = stream->config.arg_ctrls[i][0];
1296 int value = stream->config.arg_ctrls[i][1];
1297 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1298 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1301 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1305 if (global->test_decode != TEST_DECODE_OFF) {
1306 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1307 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1313 static void encode_frame(struct stream_state *stream,
1314 struct VpxEncoderConfig *global,
1315 struct vpx_image *img,
1316 unsigned int frames_in) {
1317 vpx_codec_pts_t frame_start, next_frame_start;
1318 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1319 struct vpx_usec_timer timer;
1321 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1322 * global->framerate.den)
1323 / cfg->g_timebase.num / global->framerate.num;
1324 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1325 * global->framerate.den)
1326 / cfg->g_timebase.num / global->framerate.num;
1328 /* Scale if necessary */
1329 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1330 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1331 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1336 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1337 cfg->g_w, cfg->g_h, 16);
1338 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1339 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1340 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1342 stream->img->planes[VPX_PLANE_Y],
1343 stream->img->stride[VPX_PLANE_Y],
1344 stream->img->planes[VPX_PLANE_U],
1345 stream->img->stride[VPX_PLANE_U],
1346 stream->img->planes[VPX_PLANE_V],
1347 stream->img->stride[VPX_PLANE_V],
1348 stream->img->d_w, stream->img->d_h,
1352 stream->encoder.err = 1;
1353 ctx_exit_on_error(&stream->encoder,
1354 "Stream %d: Failed to encode frame.\n"
1355 "Scaling disabled in this configuration. \n"
1356 "To enable, configure with --enable-libyuv\n",
1361 vpx_usec_timer_start(&timer);
1362 vpx_codec_encode(&stream->encoder, img, frame_start,
1363 (unsigned long)(next_frame_start - frame_start),
1364 0, global->deadline);
1365 vpx_usec_timer_mark(&timer);
1366 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1367 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1372 static void update_quantizer_histogram(struct stream_state *stream) {
1373 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1376 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1377 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1378 stream->counts[q]++;
1383 static void get_cx_data(struct stream_state *stream,
1384 struct VpxEncoderConfig *global,
1386 const vpx_codec_cx_pkt_t *pkt;
1387 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1388 vpx_codec_iter_t iter = NULL;
1391 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1392 static size_t fsize = 0;
1393 static int64_t ivf_header_pos = 0;
1395 switch (pkt->kind) {
1396 case VPX_CODEC_CX_FRAME_PKT:
1397 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1398 stream->frames_out++;
1401 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1403 update_rate_histogram(stream->rate_hist, cfg, pkt);
1405 if (stream->config.write_webm) {
1406 write_webm_block(&stream->ebml, cfg, pkt);
1409 if (!stream->config.write_webm) {
1410 if (pkt->data.frame.partition_id <= 0) {
1411 ivf_header_pos = ftello(stream->file);
1412 fsize = pkt->data.frame.sz;
1414 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1416 fsize += pkt->data.frame.sz;
1418 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1419 const int64_t currpos = ftello(stream->file);
1420 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1421 ivf_write_frame_size(stream->file, fsize);
1422 fseeko(stream->file, currpos, SEEK_SET);
1426 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1429 stream->nbytes += pkt->data.raw.sz;
1433 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1434 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1435 (unsigned int)pkt->data.frame.sz, NULL, 0);
1436 if (stream->decoder.err) {
1437 warn_or_exit_on_error(&stream->decoder,
1438 global->test_decode == TEST_DECODE_FATAL,
1439 "Failed to decode frame %d in stream %d",
1440 stream->frames_out + 1, stream->index);
1441 stream->mismatch_seen = stream->frames_out + 1;
1446 case VPX_CODEC_STATS_PKT:
1447 stream->frames_out++;
1448 stats_write(&stream->stats,
1449 pkt->data.twopass_stats.buf,
1450 pkt->data.twopass_stats.sz);
1451 stream->nbytes += pkt->data.raw.sz;
1453 #if CONFIG_FP_MB_STATS
1454 case VPX_CODEC_FPMB_STATS_PKT:
1455 stats_write(&stream->fpmb_stats,
1456 pkt->data.firstpass_mb_stats.buf,
1457 pkt->data.firstpass_mb_stats.sz);
1458 stream->nbytes += pkt->data.raw.sz;
1461 case VPX_CODEC_PSNR_PKT:
1463 if (global->show_psnr) {
1466 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1467 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1468 for (i = 0; i < 4; i++) {
1470 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1471 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1473 stream->psnr_count++;
1484 static void show_psnr(struct stream_state *stream) {
1488 if (!stream->psnr_count)
1491 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1492 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1493 (double)stream->psnr_sse_total);
1494 fprintf(stderr, " %.3f", ovpsnr);
1496 for (i = 0; i < 4; i++) {
1497 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1499 fprintf(stderr, "\n");
1503 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1504 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1508 static void test_decode(struct stream_state *stream,
1509 enum TestDecodeFatality fatal,
1510 const VpxInterface *codec) {
1511 vpx_image_t enc_img, dec_img;
1513 if (stream->mismatch_seen)
1516 /* Get the internal reference frame */
1517 if (strcmp(codec->name, "vp8") == 0) {
1518 struct vpx_ref_frame ref_enc, ref_dec;
1521 width = (stream->config.cfg.g_w + 15) & ~15;
1522 height = (stream->config.cfg.g_h + 15) & ~15;
1523 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1524 enc_img = ref_enc.img;
1525 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1526 dec_img = ref_dec.img;
1528 ref_enc.frame_type = VP8_LAST_FRAME;
1529 ref_dec.frame_type = VP8_LAST_FRAME;
1530 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1531 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1533 struct vp9_ref_frame ref;
1536 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
1538 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
1541 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1542 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1544 if (!compare_img(&enc_img, &dec_img)) {
1545 int y[4], u[4], v[4];
1546 find_mismatch(&enc_img, &dec_img, y, u, v);
1547 stream->decoder.err = 1;
1548 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1549 "Stream %d: Encode/decode mismatch on frame %d at"
1550 " Y[%d, %d] {%d/%d},"
1551 " U[%d, %d] {%d/%d},"
1552 " V[%d, %d] {%d/%d}",
1553 stream->index, stream->frames_out,
1554 y[0], y[1], y[2], y[3],
1555 u[0], u[1], u[2], u[3],
1556 v[0], v[1], v[2], v[3]);
1557 stream->mismatch_seen = stream->frames_out;
1560 vpx_img_free(&enc_img);
1561 vpx_img_free(&dec_img);
1565 static void print_time(const char *label, int64_t etl) {
1572 etl -= hours * 3600;
1577 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1578 label, hours, mins, secs);
1580 fprintf(stderr, "[%3s unknown] ", label);
1585 int main(int argc, const char **argv_) {
1588 int frame_avail, got_data;
1590 struct VpxInputContext input;
1591 struct VpxEncoderConfig global;
1592 struct stream_state *streams = NULL;
1593 char **argv, **argi;
1594 uint64_t cx_time = 0;
1598 memset(&input, 0, sizeof(input));
1599 exec_name = argv_[0];
1604 /* Setup default input stream settings */
1605 input.framerate.numerator = 30;
1606 input.framerate.denominator = 1;
1607 input.only_i420 = 1;
1608 input.bit_depth = 0;
1610 /* First parse the global configuration values, because we want to apply
1611 * other parameters on top of the default configuration provided by the
1614 argv = argv_dup(argc - 1, argv_ + 1);
1615 parse_global_config(&global, argv);
1617 switch (global.color_type) {
1619 input.fmt = VPX_IMG_FMT_I420;
1622 input.fmt = VPX_IMG_FMT_I422;
1625 input.fmt = VPX_IMG_FMT_I444;
1628 input.fmt = VPX_IMG_FMT_YV12;
1633 /* Now parse each stream's parameters. Using a local scope here
1634 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1637 struct stream_state *stream = NULL;
1640 stream = new_stream(&global, stream);
1644 } while (parse_stream_params(&global, stream, argv));
1647 /* Check for unrecognized options */
1648 for (argi = argv; *argi; argi++)
1649 if (argi[0][0] == '-' && argi[0][1])
1650 die("Error: Unrecognized option %s\n", *argi);
1652 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1653 &global, &stream->config.cfg););
1655 /* Handle non-option arguments */
1656 input.filename = argv[0];
1658 if (!input.filename)
1661 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1662 if (global.codec->fourcc == VP9_FOURCC)
1663 input.only_i420 = 0;
1665 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1666 int frames_in = 0, seen_frames = 0;
1667 int64_t estimated_time_left = -1;
1668 int64_t average_rate = -1;
1669 int64_t lagged_count = 0;
1671 open_input_file(&input);
1673 /* If the input file doesn't specify its w/h (raw files), try to get
1674 * the data from the first stream's configuration.
1676 if (!input.width || !input.height)
1678 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1679 input.width = stream->config.cfg.g_w;
1680 input.height = stream->config.cfg.g_h;
1685 /* Update stream configurations from the input file's parameters */
1686 if (!input.width || !input.height)
1687 fatal("Specify stream dimensions with --width (-w) "
1688 " and --height (-h)");
1689 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1690 FOREACH_STREAM(validate_stream_config(stream, &global));
1692 /* Ensure that --passes and --pass are consistent. If --pass is set and
1693 * --passes=2, ensure --fpf was set.
1695 if (global.pass && global.passes == 2)
1697 if (!stream->config.stats_fn)
1698 die("Stream %d: Must specify --fpf when --pass=%d"
1699 " and --passes=2\n", stream->index, global.pass);
1704 stream->config.write_webm = 0;
1705 warn("vpxenc was compiled without WebM container support."
1706 "Producing IVF output");
1710 /* Use the frame rate from the file only if none was specified
1711 * on the command-line.
1713 if (!global.have_framerate) {
1714 global.framerate.num = input.framerate.numerator;
1715 global.framerate.den = input.framerate.denominator;
1718 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1720 /* Show configuration */
1721 if (global.verbose && pass == 0)
1722 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1724 if (pass == (global.pass ? global.pass - 1 : 0)) {
1725 if (input.file_type == FILE_TYPE_Y4M)
1726 /*The Y4M reader does its own allocation.
1727 Just initialize this here to avoid problems if we never read any
1729 memset(&raw, 0, sizeof(raw));
1731 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1733 FOREACH_STREAM(stream->rate_hist =
1734 init_rate_histogram(&stream->config.cfg,
1735 &global.framerate));
1738 FOREACH_STREAM(setup_pass(stream, &global, pass));
1739 FOREACH_STREAM(open_output_file(stream, &global));
1740 FOREACH_STREAM(initialize_encoder(stream, &global));
1745 while (frame_avail || got_data) {
1746 struct vpx_usec_timer timer;
1748 if (!global.limit || frames_in < global.limit) {
1749 frame_avail = read_frame(&input, &raw);
1753 seen_frames = frames_in > global.skip_frames ?
1754 frames_in - global.skip_frames : 0;
1756 if (!global.quiet) {
1757 float fps = usec_to_fps(cx_time, seen_frames);
1758 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1760 if (stream_cnt == 1)
1762 "frame %4d/%-4d %7"PRId64"B ",
1763 frames_in, streams->frames_out, (int64_t)streams->nbytes);
1765 fprintf(stderr, "frame %4d ", frames_in);
1767 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
1768 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1769 cx_time > 9999999 ? "ms" : "us",
1770 fps >= 1.0 ? fps : fps * 60,
1771 fps >= 1.0 ? "fps" : "fpm");
1772 print_time("ETA", estimated_time_left);
1778 if (frames_in > global.skip_frames) {
1779 vpx_usec_timer_start(&timer);
1780 FOREACH_STREAM(encode_frame(stream, &global,
1781 frame_avail ? &raw : NULL,
1783 vpx_usec_timer_mark(&timer);
1784 cx_time += vpx_usec_timer_elapsed(&timer);
1786 FOREACH_STREAM(update_quantizer_histogram(stream));
1789 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1791 if (!got_data && input.length && !streams->frames_out) {
1792 lagged_count = global.limit ? seen_frames : ftello(input.file);
1793 } else if (input.length) {
1798 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1800 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1801 remaining = 1000 * (global.limit - global.skip_frames
1802 - seen_frames + lagged_count);
1804 const int64_t input_pos = ftello(input.file);
1805 const int64_t input_pos_lagged = input_pos - lagged_count;
1806 const int64_t limit = input.length;
1808 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1809 remaining = limit - input_pos + lagged_count;
1812 average_rate = (average_rate <= 0)
1814 : (average_rate * 7 + rate) / 8;
1815 estimated_time_left = average_rate ? remaining / average_rate : -1;
1818 if (got_data && global.test_decode != TEST_DECODE_OFF)
1819 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1824 fprintf(stderr, "\033[K");
1828 fprintf(stderr, "\n");
1831 FOREACH_STREAM(fprintf(
1833 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
1834 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
1835 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
1836 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
1837 seen_frames ? (int64_t)stream->nbytes * 8
1838 * (int64_t)global.framerate.num / global.framerate.den
1841 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1842 stream->cx_time > 9999999 ? "ms" : "us",
1843 usec_to_fps(stream->cx_time, seen_frames));
1846 if (global.show_psnr)
1847 FOREACH_STREAM(show_psnr(stream));
1849 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1851 if (global.test_decode != TEST_DECODE_OFF) {
1852 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1855 close_input_file(&input);
1857 if (global.test_decode == TEST_DECODE_FATAL) {
1858 FOREACH_STREAM(res |= stream->mismatch_seen);
1860 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1862 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1864 #if CONFIG_FP_MB_STATS
1865 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
1872 if (global.show_q_hist_buckets)
1873 FOREACH_STREAM(show_q_histogram(stream->counts,
1874 global.show_q_hist_buckets));
1876 if (global.show_rate_hist_buckets)
1877 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
1878 &stream->config.cfg,
1879 global.show_rate_hist_buckets));
1880 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1882 #if CONFIG_INTERNAL_STATS
1883 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1884 * to match some existing utilities.
1886 if (!(global.pass == 1 && global.passes == 2))
1888 FILE *f = fopen("opsnr.stt", "a");
1889 if (stream->mismatch_seen) {
1890 fprintf(f, "First mismatch occurred in frame %d\n",
1891 stream->mismatch_seen);
1893 fprintf(f, "No mismatch detected in recon buffers\n");
1902 return res ? EXIT_FAILURE : EXIT_SUCCESS;