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 struct arg_enum_list tune_content_enum[] = {
382 {"default", VP9E_CONTENT_DEFAULT},
383 {"screen", VP9E_CONTENT_SCREEN},
387 static const arg_def_t tune_content = ARG_DEF_ENUM(
388 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
390 static const arg_def_t *vp9_args[] = {
391 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
392 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
393 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
394 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost, &tune_content,
397 static const int vp9_arg_ctrl_map[] = {
398 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
399 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
400 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
401 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
402 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
403 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
404 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_TUNE_CONTENT,
409 static const arg_def_t *no_args[] = { NULL };
414 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
417 fprintf(stderr, "\nOptions:\n");
418 arg_show_usage(stderr, main_args);
419 fprintf(stderr, "\nEncoder Global Options:\n");
420 arg_show_usage(stderr, global_args);
421 fprintf(stderr, "\nRate Control Options:\n");
422 arg_show_usage(stderr, rc_args);
423 fprintf(stderr, "\nTwopass Rate Control Options:\n");
424 arg_show_usage(stderr, rc_twopass_args);
425 fprintf(stderr, "\nKeyframe Placement Options:\n");
426 arg_show_usage(stderr, kf_args);
427 #if CONFIG_VP8_ENCODER
428 fprintf(stderr, "\nVP8 Specific Options:\n");
429 arg_show_usage(stderr, vp8_args);
431 #if CONFIG_VP9_ENCODER
432 fprintf(stderr, "\nVP9 Specific Options:\n");
433 arg_show_usage(stderr, vp9_args);
435 fprintf(stderr, "\nStream timebase (--timebase):\n"
436 " The desired precision of timestamps in the output, expressed\n"
437 " in fractional seconds. Default is 1/1000.\n");
438 fprintf(stderr, "\nIncluded encoders:\n\n");
440 for (i = 0; i < get_vpx_encoder_count(); ++i) {
441 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
442 fprintf(stderr, " %-6s - %s\n",
443 encoder->name, vpx_codec_iface_name(encoder->codec_interface()));
449 #define mmin(a, b) ((a) < (b) ? (a) : (b))
450 static void find_mismatch(const vpx_image_t *const img1,
451 const vpx_image_t *const img2,
452 int yloc[4], int uloc[4], int vloc[4]) {
453 const uint32_t bsize = 64;
454 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
455 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
457 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
459 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
462 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
463 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
464 for (j = 0; match && j < img1->d_w; j += bsize) {
466 const int si = mmin(i + bsize, img1->d_h) - i;
467 const int sj = mmin(j + bsize, img1->d_w) - j;
468 for (k = 0; match && k < si; ++k) {
469 for (l = 0; match && l < sj; ++l) {
470 if (*(img1->planes[VPX_PLANE_Y] +
471 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
472 *(img2->planes[VPX_PLANE_Y] +
473 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
476 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
477 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
478 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
479 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
488 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
489 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
490 for (j = 0; match && j < c_w; j += bsizex) {
492 const int si = mmin(i + bsizey, c_h - i);
493 const int sj = mmin(j + bsizex, c_w - j);
494 for (k = 0; match && k < si; ++k) {
495 for (l = 0; match && l < sj; ++l) {
496 if (*(img1->planes[VPX_PLANE_U] +
497 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
498 *(img2->planes[VPX_PLANE_U] +
499 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
502 uloc[2] = *(img1->planes[VPX_PLANE_U] +
503 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
504 uloc[3] = *(img2->planes[VPX_PLANE_U] +
505 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
513 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
514 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
515 for (j = 0; match && j < c_w; j += bsizex) {
517 const int si = mmin(i + bsizey, c_h - i);
518 const int sj = mmin(j + bsizex, c_w - j);
519 for (k = 0; match && k < si; ++k) {
520 for (l = 0; match && l < sj; ++l) {
521 if (*(img1->planes[VPX_PLANE_V] +
522 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
523 *(img2->planes[VPX_PLANE_V] +
524 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
527 vloc[2] = *(img1->planes[VPX_PLANE_V] +
528 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
529 vloc[3] = *(img2->planes[VPX_PLANE_V] +
530 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
540 static int compare_img(const vpx_image_t *const img1,
541 const vpx_image_t *const img2) {
543 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
545 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
549 match &= (img1->fmt == img2->fmt);
550 match &= (img1->d_w == img2->d_w);
551 match &= (img1->d_h == img2->d_h);
553 for (i = 0; i < img1->d_h; ++i)
554 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
555 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
558 for (i = 0; i < c_h; ++i)
559 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
560 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
563 for (i = 0; i < c_h; ++i)
564 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
565 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
572 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
573 #define MAX(x,y) ((x)>(y)?(x):(y))
574 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
575 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
576 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
577 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
579 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
580 NELEMENTS(vp9_arg_ctrl_map))
584 typedef int stereo_format_t;
585 struct EbmlGlobal { int debug; };
588 /* Per-stream configuration */
589 struct stream_config {
590 struct vpx_codec_enc_cfg cfg;
592 const char *stats_fn;
593 #if CONFIG_FP_MB_STATS
594 const char *fpmb_stats_fn;
596 stereo_format_t stereo_fmt;
597 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
600 int have_kf_max_dist;
604 struct stream_state {
606 struct stream_state *next;
607 struct stream_config config;
609 struct rate_hist *rate_hist;
610 struct EbmlGlobal ebml;
611 uint64_t psnr_sse_total;
612 uint64_t psnr_samples_total;
613 double psnr_totals[4];
616 vpx_codec_ctx_t encoder;
617 unsigned int frames_out;
621 #if CONFIG_FP_MB_STATS
622 stats_io_t fpmb_stats;
624 struct vpx_image *img;
625 vpx_codec_ctx_t decoder;
630 void validate_positive_rational(const char *msg,
631 struct vpx_rational *rat) {
638 die("Error: %s must be positive\n", msg);
641 die("Error: %s has zero denominator\n", msg);
645 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
649 /* Initialize default parameters */
650 memset(global, 0, sizeof(*global));
651 global->codec = get_vpx_encoder_by_index(0);
653 global->color_type = I420;
654 /* Assign default deadline to good quality */
655 global->deadline = VPX_DL_GOOD_QUALITY;
657 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
660 if (arg_match(&arg, &codecarg, argi)) {
661 global->codec = get_vpx_encoder_by_name(arg.val);
663 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
664 } else if (arg_match(&arg, &passes, argi)) {
665 global->passes = arg_parse_uint(&arg);
667 if (global->passes < 1 || global->passes > 2)
668 die("Error: Invalid number of passes (%d)\n", global->passes);
669 } else if (arg_match(&arg, &pass_arg, argi)) {
670 global->pass = arg_parse_uint(&arg);
672 if (global->pass < 1 || global->pass > 2)
673 die("Error: Invalid pass selected (%d)\n",
675 } else if (arg_match(&arg, &usage, argi))
676 global->usage = arg_parse_uint(&arg);
677 else if (arg_match(&arg, &deadline, argi))
678 global->deadline = arg_parse_uint(&arg);
679 else if (arg_match(&arg, &best_dl, argi))
680 global->deadline = VPX_DL_BEST_QUALITY;
681 else if (arg_match(&arg, &good_dl, argi))
682 global->deadline = VPX_DL_GOOD_QUALITY;
683 else if (arg_match(&arg, &rt_dl, argi))
684 global->deadline = VPX_DL_REALTIME;
685 else if (arg_match(&arg, &use_yv12, argi))
686 global->color_type = YV12;
687 else if (arg_match(&arg, &use_i420, argi))
688 global->color_type = I420;
689 else if (arg_match(&arg, &use_i422, argi))
690 global->color_type = I422;
691 else if (arg_match(&arg, &use_i444, argi))
692 global->color_type = I444;
693 else if (arg_match(&arg, &quietarg, argi))
695 else if (arg_match(&arg, &verbosearg, argi))
697 else if (arg_match(&arg, &limit, argi))
698 global->limit = arg_parse_uint(&arg);
699 else if (arg_match(&arg, &skip, argi))
700 global->skip_frames = arg_parse_uint(&arg);
701 else if (arg_match(&arg, &psnrarg, argi))
702 global->show_psnr = 1;
703 else if (arg_match(&arg, &recontest, argi))
704 global->test_decode = arg_parse_enum_or_int(&arg);
705 else if (arg_match(&arg, &framerate, argi)) {
706 global->framerate = arg_parse_rational(&arg);
707 validate_positive_rational(arg.name, &global->framerate);
708 global->have_framerate = 1;
709 } else if (arg_match(&arg, &out_part, argi))
710 global->out_part = 1;
711 else if (arg_match(&arg, &debugmode, argi))
713 else if (arg_match(&arg, &q_hist_n, argi))
714 global->show_q_hist_buckets = arg_parse_uint(&arg);
715 else if (arg_match(&arg, &rate_hist_n, argi))
716 global->show_rate_hist_buckets = arg_parse_uint(&arg);
717 else if (arg_match(&arg, &disable_warnings, argi))
718 global->disable_warnings = 1;
719 else if (arg_match(&arg, &disable_warning_prompt, argi))
720 global->disable_warning_prompt = 1;
721 else if (arg_match(&arg, &experimental_bitstream, argi))
722 global->experimental_bitstream = 1;
728 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
729 if (global->pass > global->passes) {
730 warn("Assuming --pass=%d implies --passes=%d\n",
731 global->pass, global->pass);
732 global->passes = global->pass;
735 /* Validate global config */
736 if (global->passes == 0) {
737 #if CONFIG_VP9_ENCODER
738 // Make default VP9 passes = 2 until there is a better quality 1-pass
740 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
741 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
747 if (global->deadline == VPX_DL_REALTIME &&
748 global->passes > 1) {
749 warn("Enforcing one-pass encoding in realtime mode\n");
755 void open_input_file(struct VpxInputContext *input) {
756 /* Parse certain options from the input file, if possible */
757 input->file = strcmp(input->filename, "-")
758 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
761 fatal("Failed to open input file");
763 if (!fseeko(input->file, 0, SEEK_END)) {
764 /* Input file is seekable. Figure out how long it is, so we can get
767 input->length = ftello(input->file);
771 /* For RAW input sources, these bytes will applied on the first frame
774 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
775 input->detect.position = 0;
777 if (input->detect.buf_read == 4
778 && file_is_y4m(input->detect.buf)) {
779 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
780 input->only_i420) >= 0) {
781 input->file_type = FILE_TYPE_Y4M;
782 input->width = input->y4m.pic_w;
783 input->height = input->y4m.pic_h;
784 input->framerate.numerator = input->y4m.fps_n;
785 input->framerate.denominator = input->y4m.fps_d;
786 input->fmt = input->y4m.vpx_fmt;
787 input->bit_depth = input->y4m.bit_depth;
789 fatal("Unsupported Y4M stream.");
790 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
791 fatal("IVF is not supported as input.");
793 input->file_type = FILE_TYPE_RAW;
798 static void close_input_file(struct VpxInputContext *input) {
800 if (input->file_type == FILE_TYPE_Y4M)
801 y4m_input_close(&input->y4m);
804 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
805 struct stream_state *prev) {
806 struct stream_state *stream;
808 stream = calloc(1, sizeof(*stream));
810 fatal("Failed to allocate new stream.");
812 memcpy(stream, prev, sizeof(*stream));
818 /* Populate encoder configuration */
819 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
823 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
825 /* Change the default timebase to a high enough value so that the
826 * encoder will always create strictly increasing timestamps.
828 stream->config.cfg.g_timebase.den = 1000;
830 /* Never use the library's default resolution, require it be parsed
831 * from the file or set on the command line.
833 stream->config.cfg.g_w = 0;
834 stream->config.cfg.g_h = 0;
836 /* Initialize remaining stream parameters */
837 stream->config.write_webm = 1;
839 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
840 stream->ebml.last_pts_ns = -1;
841 stream->ebml.writer = NULL;
842 stream->ebml.segment = NULL;
845 /* Allows removal of the application version from the EBML tags */
846 stream->ebml.debug = global->debug;
848 /* Default lag_in_frames is 0 in realtime mode */
849 if (global->deadline == VPX_DL_REALTIME)
850 stream->config.cfg.g_lag_in_frames = 0;
853 /* Output files must be specified for each stream */
854 stream->config.out_fn = NULL;
861 static int parse_stream_params(struct VpxEncoderConfig *global,
862 struct stream_state *stream,
866 static const arg_def_t **ctrl_args = no_args;
867 static const int *ctrl_args_map = NULL;
868 struct stream_config *config = &stream->config;
869 int eos_mark_found = 0;
871 // Handle codec specific options
873 #if CONFIG_VP8_ENCODER
874 } else if (strcmp(global->codec->name, "vp8") == 0) {
875 ctrl_args = vp8_args;
876 ctrl_args_map = vp8_arg_ctrl_map;
878 #if CONFIG_VP9_ENCODER
879 } else if (strcmp(global->codec->name, "vp9") == 0) {
880 ctrl_args = vp9_args;
881 ctrl_args_map = vp9_arg_ctrl_map;
885 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
888 /* Once we've found an end-of-stream marker (--) we want to continue
889 * shifting arguments but not consuming them.
891 if (eos_mark_found) {
894 } else if (!strcmp(*argj, "--")) {
900 } else if (arg_match(&arg, &outputfile, argi)) {
901 config->out_fn = arg.val;
902 } else if (arg_match(&arg, &fpf_name, argi)) {
903 config->stats_fn = arg.val;
904 #if CONFIG_FP_MB_STATS
905 } else if (arg_match(&arg, &fpmbf_name, argi)) {
906 config->fpmb_stats_fn = arg.val;
908 } else if (arg_match(&arg, &use_ivf, argi)) {
909 config->write_webm = 0;
910 } else if (arg_match(&arg, &threads, argi)) {
911 config->cfg.g_threads = arg_parse_uint(&arg);
912 } else if (arg_match(&arg, &profile, argi)) {
913 config->cfg.g_profile = arg_parse_uint(&arg);
914 } else if (arg_match(&arg, &width, argi)) {
915 config->cfg.g_w = arg_parse_uint(&arg);
916 } else if (arg_match(&arg, &height, argi)) {
917 config->cfg.g_h = arg_parse_uint(&arg);
919 } else if (arg_match(&arg, &stereo_mode, argi)) {
920 config->stereo_fmt = arg_parse_enum_or_int(&arg);
922 } else if (arg_match(&arg, &timebase, argi)) {
923 config->cfg.g_timebase = arg_parse_rational(&arg);
924 validate_positive_rational(arg.name, &config->cfg.g_timebase);
925 } else if (arg_match(&arg, &error_resilient, argi)) {
926 config->cfg.g_error_resilient = arg_parse_uint(&arg);
927 } else if (arg_match(&arg, &lag_in_frames, argi)) {
928 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
929 if (global->deadline == VPX_DL_REALTIME &&
930 config->cfg.g_lag_in_frames != 0) {
931 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
932 config->cfg.g_lag_in_frames = 0;
934 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
935 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
936 } else if (arg_match(&arg, &resize_allowed, argi)) {
937 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
938 } else if (arg_match(&arg, &resize_width, argi)) {
939 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
940 } else if (arg_match(&arg, &resize_height, argi)) {
941 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
942 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
943 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
944 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
945 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
946 } else if (arg_match(&arg, &end_usage, argi)) {
947 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
948 } else if (arg_match(&arg, &target_bitrate, argi)) {
949 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
950 } else if (arg_match(&arg, &min_quantizer, argi)) {
951 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
952 } else if (arg_match(&arg, &max_quantizer, argi)) {
953 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
954 } else if (arg_match(&arg, &undershoot_pct, argi)) {
955 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
956 } else if (arg_match(&arg, &overshoot_pct, argi)) {
957 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
958 } else if (arg_match(&arg, &buf_sz, argi)) {
959 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
960 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
961 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
962 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
963 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
964 } else if (arg_match(&arg, &bias_pct, argi)) {
965 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
966 if (global->passes < 2)
967 warn("option %s ignored in one-pass mode.\n", arg.name);
968 } else if (arg_match(&arg, &minsection_pct, argi)) {
969 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
971 if (global->passes < 2)
972 warn("option %s ignored in one-pass mode.\n", arg.name);
973 } else if (arg_match(&arg, &maxsection_pct, argi)) {
974 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
976 if (global->passes < 2)
977 warn("option %s ignored in one-pass mode.\n", arg.name);
978 } else if (arg_match(&arg, &kf_min_dist, argi)) {
979 config->cfg.kf_min_dist = arg_parse_uint(&arg);
980 } else if (arg_match(&arg, &kf_max_dist, argi)) {
981 config->cfg.kf_max_dist = arg_parse_uint(&arg);
982 config->have_kf_max_dist = 1;
983 } else if (arg_match(&arg, &kf_disabled, argi)) {
984 config->cfg.kf_mode = VPX_KF_DISABLED;
987 for (i = 0; ctrl_args[i]; i++) {
988 if (arg_match(&arg, ctrl_args[i], argi)) {
992 /* Point either to the next free element or the first
993 * instance of this control.
995 for (j = 0; j < config->arg_ctrl_cnt; j++)
996 if (config->arg_ctrls[j][0] == ctrl_args_map[i])
1000 assert(j < (int)ARG_CTRL_CNT_MAX);
1001 if (j < (int)ARG_CTRL_CNT_MAX) {
1002 config->arg_ctrls[j][0] = ctrl_args_map[i];
1003 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1004 if (j == config->arg_ctrl_cnt)
1005 config->arg_ctrl_cnt++;
1014 return eos_mark_found;
1018 #define FOREACH_STREAM(func) \
1020 struct stream_state *stream; \
1021 for (stream = streams; stream; stream = stream->next) { \
1027 static void validate_stream_config(const struct stream_state *stream,
1028 const struct VpxEncoderConfig *global) {
1029 const struct stream_state *streami;
1031 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1032 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1033 " and --height (-h)", stream->index);
1035 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
1036 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
1037 stream->index, stream->config.cfg.g_profile,
1038 experimental_bitstream.long_name);
1041 for (streami = stream; streami; streami = streami->next) {
1042 /* All streams require output files */
1043 if (!streami->config.out_fn)
1044 fatal("Stream %d: Output file is required (specify with -o)",
1047 /* Check for two streams outputting to the same file */
1048 if (streami != stream) {
1049 const char *a = stream->config.out_fn;
1050 const char *b = streami->config.out_fn;
1051 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1052 fatal("Stream %d: duplicate output file (from stream %d)",
1053 streami->index, stream->index);
1056 /* Check for two streams sharing a stats file. */
1057 if (streami != stream) {
1058 const char *a = stream->config.stats_fn;
1059 const char *b = streami->config.stats_fn;
1060 if (a && b && !strcmp(a, b))
1061 fatal("Stream %d: duplicate stats file (from stream %d)",
1062 streami->index, stream->index);
1065 #if CONFIG_FP_MB_STATS
1066 /* Check for two streams sharing a mb stats file. */
1067 if (streami != stream) {
1068 const char *a = stream->config.fpmb_stats_fn;
1069 const char *b = streami->config.fpmb_stats_fn;
1070 if (a && b && !strcmp(a, b))
1071 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1072 streami->index, stream->index);
1079 static void set_stream_dimensions(struct stream_state *stream,
1082 if (!stream->config.cfg.g_w) {
1083 if (!stream->config.cfg.g_h)
1084 stream->config.cfg.g_w = w;
1086 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1088 if (!stream->config.cfg.g_h) {
1089 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1094 static void set_default_kf_interval(struct stream_state *stream,
1095 struct VpxEncoderConfig *global) {
1096 /* Use a max keyframe interval of 5 seconds, if none was
1097 * specified on the command line.
1099 if (!stream->config.have_kf_max_dist) {
1100 double framerate = (double)global->framerate.num / global->framerate.den;
1101 if (framerate > 0.0)
1102 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1106 static const char* file_type_to_string(enum VideoFileType t) {
1108 case FILE_TYPE_RAW: return "RAW";
1109 case FILE_TYPE_Y4M: return "Y4M";
1110 default: return "Other";
1114 static const char* image_format_to_string(vpx_img_fmt_t f) {
1116 case VPX_IMG_FMT_I420: return "I420";
1117 case VPX_IMG_FMT_I422: return "I422";
1118 case VPX_IMG_FMT_I444: return "I444";
1119 case VPX_IMG_FMT_YV12: return "YV12";
1120 default: return "Other";
1124 static void show_stream_config(struct stream_state *stream,
1125 struct VpxEncoderConfig *global,
1126 struct VpxInputContext *input) {
1128 #define SHOW(field) \
1129 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1131 if (stream->index == 0) {
1132 fprintf(stderr, "Codec: %s\n",
1133 vpx_codec_iface_name(global->codec->codec_interface()));
1134 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1136 file_type_to_string(input->file_type),
1137 image_format_to_string(input->fmt));
1139 if (stream->next || stream->index)
1140 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1141 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1142 fprintf(stderr, "Encoder parameters:\n");
1149 SHOW(g_timebase.num);
1150 SHOW(g_timebase.den);
1151 SHOW(g_error_resilient);
1153 SHOW(g_lag_in_frames);
1154 SHOW(rc_dropframe_thresh);
1155 SHOW(rc_resize_allowed);
1156 SHOW(rc_scaled_width);
1157 SHOW(rc_scaled_height);
1158 SHOW(rc_resize_up_thresh);
1159 SHOW(rc_resize_down_thresh);
1161 SHOW(rc_target_bitrate);
1162 SHOW(rc_min_quantizer);
1163 SHOW(rc_max_quantizer);
1164 SHOW(rc_undershoot_pct);
1165 SHOW(rc_overshoot_pct);
1167 SHOW(rc_buf_initial_sz);
1168 SHOW(rc_buf_optimal_sz);
1169 SHOW(rc_2pass_vbr_bias_pct);
1170 SHOW(rc_2pass_vbr_minsection_pct);
1171 SHOW(rc_2pass_vbr_maxsection_pct);
1178 static void open_output_file(struct stream_state *stream,
1179 struct VpxEncoderConfig *global) {
1180 const char *fn = stream->config.out_fn;
1181 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1183 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1186 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1189 fatal("Failed to open output file");
1191 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1192 fatal("WebM output to pipes not supported.");
1195 if (stream->config.write_webm) {
1196 stream->ebml.stream = stream->file;
1197 write_webm_file_header(&stream->ebml, cfg,
1199 stream->config.stereo_fmt,
1200 global->codec->fourcc);
1204 if (!stream->config.write_webm) {
1205 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1210 static void close_output_file(struct stream_state *stream,
1211 unsigned int fourcc) {
1212 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1214 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1218 if (stream->config.write_webm) {
1219 write_webm_file_footer(&stream->ebml);
1223 if (!stream->config.write_webm) {
1224 if (!fseek(stream->file, 0, SEEK_SET))
1225 ivf_write_file_header(stream->file, &stream->config.cfg,
1227 stream->frames_out);
1230 fclose(stream->file);
1234 static void setup_pass(struct stream_state *stream,
1235 struct VpxEncoderConfig *global,
1237 if (stream->config.stats_fn) {
1238 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1240 fatal("Failed to open statistics store");
1242 if (!stats_open_mem(&stream->stats, pass))
1243 fatal("Failed to open statistics store");
1246 #if CONFIG_FP_MB_STATS
1247 if (stream->config.fpmb_stats_fn) {
1248 if (!stats_open_file(&stream->fpmb_stats,
1249 stream->config.fpmb_stats_fn, pass))
1250 fatal("Failed to open mb statistics store");
1252 if (!stats_open_mem(&stream->fpmb_stats, pass))
1253 fatal("Failed to open mb statistics store");
1257 stream->config.cfg.g_pass = global->passes == 2
1258 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1261 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1262 #if CONFIG_FP_MB_STATS
1263 stream->config.cfg.rc_firstpass_mb_stats_in =
1264 stats_get(&stream->fpmb_stats);
1268 stream->cx_time = 0;
1270 stream->frames_out = 0;
1274 static void initialize_encoder(struct stream_state *stream,
1275 struct VpxEncoderConfig *global) {
1279 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1280 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1282 /* Construct Encoder Context */
1283 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1284 &stream->config.cfg, flags);
1285 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1287 /* Note that we bypass the vpx_codec_control wrapper macro because
1288 * we're being clever to store the control IDs in an array. Real
1289 * applications will want to make use of the enumerations directly
1291 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1292 int ctrl = stream->config.arg_ctrls[i][0];
1293 int value = stream->config.arg_ctrls[i][1];
1294 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1295 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1298 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1302 if (global->test_decode != TEST_DECODE_OFF) {
1303 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1304 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1310 static void encode_frame(struct stream_state *stream,
1311 struct VpxEncoderConfig *global,
1312 struct vpx_image *img,
1313 unsigned int frames_in) {
1314 vpx_codec_pts_t frame_start, next_frame_start;
1315 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1316 struct vpx_usec_timer timer;
1318 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1319 * global->framerate.den)
1320 / cfg->g_timebase.num / global->framerate.num;
1321 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1322 * global->framerate.den)
1323 / cfg->g_timebase.num / global->framerate.num;
1325 /* Scale if necessary */
1326 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1327 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1328 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1333 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1334 cfg->g_w, cfg->g_h, 16);
1335 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1336 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1337 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1339 stream->img->planes[VPX_PLANE_Y],
1340 stream->img->stride[VPX_PLANE_Y],
1341 stream->img->planes[VPX_PLANE_U],
1342 stream->img->stride[VPX_PLANE_U],
1343 stream->img->planes[VPX_PLANE_V],
1344 stream->img->stride[VPX_PLANE_V],
1345 stream->img->d_w, stream->img->d_h,
1349 stream->encoder.err = 1;
1350 ctx_exit_on_error(&stream->encoder,
1351 "Stream %d: Failed to encode frame.\n"
1352 "Scaling disabled in this configuration. \n"
1353 "To enable, configure with --enable-libyuv\n",
1358 vpx_usec_timer_start(&timer);
1359 vpx_codec_encode(&stream->encoder, img, frame_start,
1360 (unsigned long)(next_frame_start - frame_start),
1361 0, global->deadline);
1362 vpx_usec_timer_mark(&timer);
1363 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1364 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1369 static void update_quantizer_histogram(struct stream_state *stream) {
1370 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1373 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1374 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1375 stream->counts[q]++;
1380 static void get_cx_data(struct stream_state *stream,
1381 struct VpxEncoderConfig *global,
1383 const vpx_codec_cx_pkt_t *pkt;
1384 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1385 vpx_codec_iter_t iter = NULL;
1388 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1389 static size_t fsize = 0;
1390 static int64_t ivf_header_pos = 0;
1392 switch (pkt->kind) {
1393 case VPX_CODEC_CX_FRAME_PKT:
1394 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1395 stream->frames_out++;
1398 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1400 update_rate_histogram(stream->rate_hist, cfg, pkt);
1402 if (stream->config.write_webm) {
1403 write_webm_block(&stream->ebml, cfg, pkt);
1406 if (!stream->config.write_webm) {
1407 if (pkt->data.frame.partition_id <= 0) {
1408 ivf_header_pos = ftello(stream->file);
1409 fsize = pkt->data.frame.sz;
1411 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1413 fsize += pkt->data.frame.sz;
1415 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1416 const int64_t currpos = ftello(stream->file);
1417 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1418 ivf_write_frame_size(stream->file, fsize);
1419 fseeko(stream->file, currpos, SEEK_SET);
1423 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1426 stream->nbytes += pkt->data.raw.sz;
1430 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1431 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1432 (unsigned int)pkt->data.frame.sz, NULL, 0);
1433 if (stream->decoder.err) {
1434 warn_or_exit_on_error(&stream->decoder,
1435 global->test_decode == TEST_DECODE_FATAL,
1436 "Failed to decode frame %d in stream %d",
1437 stream->frames_out + 1, stream->index);
1438 stream->mismatch_seen = stream->frames_out + 1;
1443 case VPX_CODEC_STATS_PKT:
1444 stream->frames_out++;
1445 stats_write(&stream->stats,
1446 pkt->data.twopass_stats.buf,
1447 pkt->data.twopass_stats.sz);
1448 stream->nbytes += pkt->data.raw.sz;
1450 #if CONFIG_FP_MB_STATS
1451 case VPX_CODEC_FPMB_STATS_PKT:
1452 stats_write(&stream->fpmb_stats,
1453 pkt->data.firstpass_mb_stats.buf,
1454 pkt->data.firstpass_mb_stats.sz);
1455 stream->nbytes += pkt->data.raw.sz;
1458 case VPX_CODEC_PSNR_PKT:
1460 if (global->show_psnr) {
1463 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1464 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1465 for (i = 0; i < 4; i++) {
1467 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1468 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1470 stream->psnr_count++;
1481 static void show_psnr(struct stream_state *stream) {
1485 if (!stream->psnr_count)
1488 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1489 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1490 (double)stream->psnr_sse_total);
1491 fprintf(stderr, " %.3f", ovpsnr);
1493 for (i = 0; i < 4; i++) {
1494 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1496 fprintf(stderr, "\n");
1500 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1501 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1505 static void test_decode(struct stream_state *stream,
1506 enum TestDecodeFatality fatal,
1507 const VpxInterface *codec) {
1508 vpx_image_t enc_img, dec_img;
1510 if (stream->mismatch_seen)
1513 /* Get the internal reference frame */
1514 if (strcmp(codec->name, "vp8") == 0) {
1515 struct vpx_ref_frame ref_enc, ref_dec;
1518 width = (stream->config.cfg.g_w + 15) & ~15;
1519 height = (stream->config.cfg.g_h + 15) & ~15;
1520 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1521 enc_img = ref_enc.img;
1522 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1523 dec_img = ref_dec.img;
1525 ref_enc.frame_type = VP8_LAST_FRAME;
1526 ref_dec.frame_type = VP8_LAST_FRAME;
1527 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1528 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1530 struct vp9_ref_frame ref;
1533 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref);
1535 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref);
1538 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1539 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1541 if (!compare_img(&enc_img, &dec_img)) {
1542 int y[4], u[4], v[4];
1543 find_mismatch(&enc_img, &dec_img, y, u, v);
1544 stream->decoder.err = 1;
1545 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1546 "Stream %d: Encode/decode mismatch on frame %d at"
1547 " Y[%d, %d] {%d/%d},"
1548 " U[%d, %d] {%d/%d},"
1549 " V[%d, %d] {%d/%d}",
1550 stream->index, stream->frames_out,
1551 y[0], y[1], y[2], y[3],
1552 u[0], u[1], u[2], u[3],
1553 v[0], v[1], v[2], v[3]);
1554 stream->mismatch_seen = stream->frames_out;
1557 vpx_img_free(&enc_img);
1558 vpx_img_free(&dec_img);
1562 static void print_time(const char *label, int64_t etl) {
1569 etl -= hours * 3600;
1574 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1575 label, hours, mins, secs);
1577 fprintf(stderr, "[%3s unknown] ", label);
1582 int main(int argc, const char **argv_) {
1585 int frame_avail, got_data;
1587 struct VpxInputContext input;
1588 struct VpxEncoderConfig global;
1589 struct stream_state *streams = NULL;
1590 char **argv, **argi;
1591 uint64_t cx_time = 0;
1595 memset(&input, 0, sizeof(input));
1596 exec_name = argv_[0];
1601 /* Setup default input stream settings */
1602 input.framerate.numerator = 30;
1603 input.framerate.denominator = 1;
1604 input.only_i420 = 1;
1605 input.bit_depth = 0;
1607 /* First parse the global configuration values, because we want to apply
1608 * other parameters on top of the default configuration provided by the
1611 argv = argv_dup(argc - 1, argv_ + 1);
1612 parse_global_config(&global, argv);
1614 switch (global.color_type) {
1616 input.fmt = VPX_IMG_FMT_I420;
1619 input.fmt = VPX_IMG_FMT_I422;
1622 input.fmt = VPX_IMG_FMT_I444;
1625 input.fmt = VPX_IMG_FMT_YV12;
1630 /* Now parse each stream's parameters. Using a local scope here
1631 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1634 struct stream_state *stream = NULL;
1637 stream = new_stream(&global, stream);
1641 } while (parse_stream_params(&global, stream, argv));
1644 /* Check for unrecognized options */
1645 for (argi = argv; *argi; argi++)
1646 if (argi[0][0] == '-' && argi[0][1])
1647 die("Error: Unrecognized option %s\n", *argi);
1649 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1650 &global, &stream->config.cfg););
1652 /* Handle non-option arguments */
1653 input.filename = argv[0];
1655 if (!input.filename)
1658 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1659 if (global.codec->fourcc == VP9_FOURCC)
1660 input.only_i420 = 0;
1662 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1663 int frames_in = 0, seen_frames = 0;
1664 int64_t estimated_time_left = -1;
1665 int64_t average_rate = -1;
1666 int64_t lagged_count = 0;
1668 open_input_file(&input);
1670 /* If the input file doesn't specify its w/h (raw files), try to get
1671 * the data from the first stream's configuration.
1673 if (!input.width || !input.height)
1675 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1676 input.width = stream->config.cfg.g_w;
1677 input.height = stream->config.cfg.g_h;
1682 /* Update stream configurations from the input file's parameters */
1683 if (!input.width || !input.height)
1684 fatal("Specify stream dimensions with --width (-w) "
1685 " and --height (-h)");
1686 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1687 FOREACH_STREAM(validate_stream_config(stream, &global));
1689 /* Ensure that --passes and --pass are consistent. If --pass is set and
1690 * --passes=2, ensure --fpf was set.
1692 if (global.pass && global.passes == 2)
1694 if (!stream->config.stats_fn)
1695 die("Stream %d: Must specify --fpf when --pass=%d"
1696 " and --passes=2\n", stream->index, global.pass);
1701 stream->config.write_webm = 0;
1702 warn("vpxenc was compiled without WebM container support."
1703 "Producing IVF output");
1707 /* Use the frame rate from the file only if none was specified
1708 * on the command-line.
1710 if (!global.have_framerate) {
1711 global.framerate.num = input.framerate.numerator;
1712 global.framerate.den = input.framerate.denominator;
1715 FOREACH_STREAM(set_default_kf_interval(stream, &global));
1717 /* Show configuration */
1718 if (global.verbose && pass == 0)
1719 FOREACH_STREAM(show_stream_config(stream, &global, &input));
1721 if (pass == (global.pass ? global.pass - 1 : 0)) {
1722 if (input.file_type == FILE_TYPE_Y4M)
1723 /*The Y4M reader does its own allocation.
1724 Just initialize this here to avoid problems if we never read any
1726 memset(&raw, 0, sizeof(raw));
1728 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
1730 FOREACH_STREAM(stream->rate_hist =
1731 init_rate_histogram(&stream->config.cfg,
1732 &global.framerate));
1735 FOREACH_STREAM(setup_pass(stream, &global, pass));
1736 FOREACH_STREAM(open_output_file(stream, &global));
1737 FOREACH_STREAM(initialize_encoder(stream, &global));
1742 while (frame_avail || got_data) {
1743 struct vpx_usec_timer timer;
1745 if (!global.limit || frames_in < global.limit) {
1746 frame_avail = read_frame(&input, &raw);
1750 seen_frames = frames_in > global.skip_frames ?
1751 frames_in - global.skip_frames : 0;
1753 if (!global.quiet) {
1754 float fps = usec_to_fps(cx_time, seen_frames);
1755 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
1757 if (stream_cnt == 1)
1759 "frame %4d/%-4d %7"PRId64"B ",
1760 frames_in, streams->frames_out, (int64_t)streams->nbytes);
1762 fprintf(stderr, "frame %4d ", frames_in);
1764 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
1765 cx_time > 9999999 ? cx_time / 1000 : cx_time,
1766 cx_time > 9999999 ? "ms" : "us",
1767 fps >= 1.0 ? fps : fps * 60,
1768 fps >= 1.0 ? "fps" : "fpm");
1769 print_time("ETA", estimated_time_left);
1775 if (frames_in > global.skip_frames) {
1776 vpx_usec_timer_start(&timer);
1777 FOREACH_STREAM(encode_frame(stream, &global,
1778 frame_avail ? &raw : NULL,
1780 vpx_usec_timer_mark(&timer);
1781 cx_time += vpx_usec_timer_elapsed(&timer);
1783 FOREACH_STREAM(update_quantizer_histogram(stream));
1786 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
1788 if (!got_data && input.length && !streams->frames_out) {
1789 lagged_count = global.limit ? seen_frames : ftello(input.file);
1790 } else if (input.length) {
1795 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
1797 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
1798 remaining = 1000 * (global.limit - global.skip_frames
1799 - seen_frames + lagged_count);
1801 const int64_t input_pos = ftello(input.file);
1802 const int64_t input_pos_lagged = input_pos - lagged_count;
1803 const int64_t limit = input.length;
1805 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
1806 remaining = limit - input_pos + lagged_count;
1809 average_rate = (average_rate <= 0)
1811 : (average_rate * 7 + rate) / 8;
1812 estimated_time_left = average_rate ? remaining / average_rate : -1;
1815 if (got_data && global.test_decode != TEST_DECODE_OFF)
1816 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
1821 fprintf(stderr, "\033[K");
1825 fprintf(stderr, "\n");
1828 FOREACH_STREAM(fprintf(
1830 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
1831 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
1832 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
1833 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
1834 seen_frames ? (int64_t)stream->nbytes * 8
1835 * (int64_t)global.framerate.num / global.framerate.den
1838 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
1839 stream->cx_time > 9999999 ? "ms" : "us",
1840 usec_to_fps(stream->cx_time, seen_frames));
1843 if (global.show_psnr)
1844 FOREACH_STREAM(show_psnr(stream));
1846 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
1848 if (global.test_decode != TEST_DECODE_OFF) {
1849 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
1852 close_input_file(&input);
1854 if (global.test_decode == TEST_DECODE_FATAL) {
1855 FOREACH_STREAM(res |= stream->mismatch_seen);
1857 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
1859 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
1861 #if CONFIG_FP_MB_STATS
1862 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
1869 if (global.show_q_hist_buckets)
1870 FOREACH_STREAM(show_q_histogram(stream->counts,
1871 global.show_q_hist_buckets));
1873 if (global.show_rate_hist_buckets)
1874 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
1875 &stream->config.cfg,
1876 global.show_rate_hist_buckets));
1877 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
1879 #if CONFIG_INTERNAL_STATS
1880 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
1881 * to match some existing utilities.
1883 if (!(global.pass == 1 && global.passes == 2))
1885 FILE *f = fopen("opsnr.stt", "a");
1886 if (stream->mismatch_seen) {
1887 fprintf(f, "First mismatch occurred in frame %d\n",
1888 stream->mismatch_seen);
1890 fprintf(f, "No mismatch detected in recon buffers\n");
1899 return res ? EXIT_FAILURE : EXIT_SUCCESS;