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.");
203 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
204 static const arg_def_t test16bitinternalarg = ARG_DEF(
205 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
208 static const arg_def_t *main_args[] = {
210 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
211 &deadline, &best_dl, &good_dl, &rt_dl,
212 &quietarg, &verbosearg, &psnrarg, &use_ivf, &out_part, &q_hist_n,
213 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
217 static const arg_def_t usage = ARG_DEF("u", "usage", 1,
218 "Usage profile number to use");
219 static const arg_def_t threads = ARG_DEF("t", "threads", 1,
220 "Max number of threads to use");
221 static const arg_def_t profile = ARG_DEF(NULL, "profile", 1,
222 "Bitstream profile number to use");
223 static const arg_def_t width = ARG_DEF("w", "width", 1,
225 static const arg_def_t height = ARG_DEF("h", "height", 1,
228 static const struct arg_enum_list stereo_mode_enum[] = {
229 {"mono", STEREO_FORMAT_MONO},
230 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
231 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
232 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
233 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
236 static const arg_def_t stereo_mode = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
237 "Stereo 3D video format", stereo_mode_enum);
239 static const arg_def_t timebase = ARG_DEF(NULL, "timebase", 1,
240 "Output timestamp precision (fractional seconds)");
241 static const arg_def_t error_resilient = ARG_DEF(NULL, "error-resilient", 1,
242 "Enable error resiliency features");
243 static const arg_def_t lag_in_frames = ARG_DEF(NULL, "lag-in-frames", 1,
244 "Max number of frames to lag");
246 static const arg_def_t *global_args[] = {
247 &use_yv12, &use_i420, &use_i422, &use_i444,
248 &usage, &threads, &profile,
253 &timebase, &framerate,
255 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
256 &test16bitinternalarg,
261 static const arg_def_t dropframe_thresh = ARG_DEF(NULL, "drop-frame", 1,
262 "Temporal resampling threshold (buf %)");
263 static const arg_def_t resize_allowed = ARG_DEF(NULL, "resize-allowed", 1,
264 "Spatial resampling enabled (bool)");
265 static const arg_def_t resize_width = ARG_DEF(NULL, "resize-width", 1,
266 "Width of encoded frame");
267 static const arg_def_t resize_height = ARG_DEF(NULL, "resize-height", 1,
268 "Height of encoded frame");
269 static const arg_def_t resize_up_thresh = ARG_DEF(NULL, "resize-up", 1,
270 "Upscale threshold (buf %)");
271 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
272 "Downscale threshold (buf %)");
273 static const struct arg_enum_list end_usage_enum[] = {
280 static const arg_def_t end_usage = ARG_DEF_ENUM(NULL, "end-usage", 1,
281 "Rate control mode", end_usage_enum);
282 static const arg_def_t target_bitrate = ARG_DEF(NULL, "target-bitrate", 1,
284 static const arg_def_t min_quantizer = ARG_DEF(NULL, "min-q", 1,
285 "Minimum (best) quantizer");
286 static const arg_def_t max_quantizer = ARG_DEF(NULL, "max-q", 1,
287 "Maximum (worst) quantizer");
288 static const arg_def_t undershoot_pct = ARG_DEF(NULL, "undershoot-pct", 1,
289 "Datarate undershoot (min) target (%)");
290 static const arg_def_t overshoot_pct = ARG_DEF(NULL, "overshoot-pct", 1,
291 "Datarate overshoot (max) target (%)");
292 static const arg_def_t buf_sz = ARG_DEF(NULL, "buf-sz", 1,
293 "Client buffer size (ms)");
294 static const arg_def_t buf_initial_sz = ARG_DEF(NULL, "buf-initial-sz", 1,
295 "Client initial buffer size (ms)");
296 static const arg_def_t buf_optimal_sz = ARG_DEF(NULL, "buf-optimal-sz", 1,
297 "Client optimal buffer size (ms)");
298 static const arg_def_t *rc_args[] = {
299 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
300 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
301 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
302 &buf_initial_sz, &buf_optimal_sz, NULL
306 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
307 "CBR/VBR bias (0=CBR, 100=VBR)");
308 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
309 "GOP min bitrate (% of target)");
310 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
311 "GOP max bitrate (% of target)");
312 static const arg_def_t *rc_twopass_args[] = {
313 &bias_pct, &minsection_pct, &maxsection_pct, NULL
317 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
318 "Minimum keyframe interval (frames)");
319 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
320 "Maximum keyframe interval (frames)");
321 static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
322 "Disable keyframe placement");
323 static const arg_def_t *kf_args[] = {
324 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
328 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
329 "Noise sensitivity (frames to blur)");
330 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
331 "Loop filter sharpness (0..7)");
332 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
333 "Motion detection threshold");
334 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
335 "CPU Used (-16..16)");
336 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
337 "Enable automatic alt reference frames");
338 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
339 "AltRef max frames (0..15)");
340 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
341 "AltRef filter strength (0..6)");
342 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
344 static const struct arg_enum_list tuning_enum[] = {
345 {"psnr", VP8_TUNE_PSNR},
346 {"ssim", VP8_TUNE_SSIM},
349 static const arg_def_t tune_ssim = ARG_DEF_ENUM(NULL, "tune", 1,
350 "Material to favor", tuning_enum);
351 static const arg_def_t cq_level = ARG_DEF(NULL, "cq-level", 1,
352 "Constant/Constrained Quality level");
353 static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
354 "Max I-frame bitrate (pct)");
356 #if CONFIG_VP8_ENCODER
357 static const arg_def_t token_parts =
358 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
359 static const arg_def_t *vp8_args[] = {
360 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
361 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
362 &tune_ssim, &cq_level, &max_intra_rate_pct,
365 static const int vp8_arg_ctrl_map[] = {
366 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
367 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
368 VP8E_SET_TOKEN_PARTITIONS,
369 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
370 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
375 #if CONFIG_VP9_ENCODER
376 static const arg_def_t tile_cols =
377 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
378 static const arg_def_t tile_rows =
379 ARG_DEF(NULL, "tile-rows", 1, "Number of tile rows to use, log2");
380 static const arg_def_t lossless = ARG_DEF(NULL, "lossless", 1, "Lossless mode");
381 static const arg_def_t frame_parallel_decoding = ARG_DEF(
382 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
383 static const arg_def_t aq_mode = ARG_DEF(
385 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
386 "3: cyclic refresh)");
387 static const arg_def_t frame_periodic_boost = ARG_DEF(
388 NULL, "frame-boost", 1,
389 "Enable frame periodic boost (0: off (default), 1: on)");
391 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
392 static const struct arg_enum_list bitdepth_enum[] = {
399 static const arg_def_t bitdeptharg = ARG_DEF_ENUM("b", "bit-depth", 1,
400 "Bit depth for codec "
401 "(8 for version <=1, "
402 "10 or 12 for version 2)",
404 static const arg_def_t inbitdeptharg = ARG_DEF(NULL, "input-bit-depth", 1,
405 "Bit depth of input");
408 static const struct arg_enum_list tune_content_enum[] = {
409 {"default", VP9E_CONTENT_DEFAULT},
410 {"screen", VP9E_CONTENT_SCREEN},
414 static const arg_def_t tune_content = ARG_DEF_ENUM(
415 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
417 static const arg_def_t *vp9_args[] = {
418 &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
419 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
420 &tune_ssim, &cq_level, &max_intra_rate_pct, &lossless,
421 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost, &tune_content,
422 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
423 &bitdeptharg, &inbitdeptharg,
427 static const int vp9_arg_ctrl_map[] = {
428 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
429 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
430 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
431 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
432 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
433 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
434 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_TUNE_CONTENT,
439 static const arg_def_t *no_args[] = { NULL };
444 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
447 fprintf(stderr, "\nOptions:\n");
448 arg_show_usage(stderr, main_args);
449 fprintf(stderr, "\nEncoder Global Options:\n");
450 arg_show_usage(stderr, global_args);
451 fprintf(stderr, "\nRate Control Options:\n");
452 arg_show_usage(stderr, rc_args);
453 fprintf(stderr, "\nTwopass Rate Control Options:\n");
454 arg_show_usage(stderr, rc_twopass_args);
455 fprintf(stderr, "\nKeyframe Placement Options:\n");
456 arg_show_usage(stderr, kf_args);
457 #if CONFIG_VP8_ENCODER
458 fprintf(stderr, "\nVP8 Specific Options:\n");
459 arg_show_usage(stderr, vp8_args);
461 #if CONFIG_VP9_ENCODER
462 fprintf(stderr, "\nVP9 Specific Options:\n");
463 arg_show_usage(stderr, vp9_args);
465 fprintf(stderr, "\nStream timebase (--timebase):\n"
466 " The desired precision of timestamps in the output, expressed\n"
467 " in fractional seconds. Default is 1/1000.\n");
468 fprintf(stderr, "\nIncluded encoders:\n\n");
470 for (i = 0; i < get_vpx_encoder_count(); ++i) {
471 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
472 fprintf(stderr, " %-6s - %s\n",
473 encoder->name, vpx_codec_iface_name(encoder->codec_interface()));
479 #define mmin(a, b) ((a) < (b) ? (a) : (b))
481 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
482 static void find_mismatch_high(const vpx_image_t *const img1,
483 const vpx_image_t *const img2,
484 int yloc[4], int uloc[4], int vloc[4]) {
485 uint16_t *plane1, *plane2;
486 uint32_t stride1, stride2;
487 const uint32_t bsize = 64;
488 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
489 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
491 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
493 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
496 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
497 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
498 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
499 stride1 = img1->stride[VPX_PLANE_Y]/2;
500 stride2 = img2->stride[VPX_PLANE_Y]/2;
501 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
502 for (j = 0; match && j < img1->d_w; j += bsize) {
504 const int si = mmin(i + bsize, img1->d_h) - i;
505 const int sj = mmin(j + bsize, img1->d_w) - j;
506 for (k = 0; match && k < si; ++k) {
507 for (l = 0; match && l < sj; ++l) {
508 if (*(plane1 + (i + k) * stride1 + j + l) !=
509 *(plane2 + (i + k) * stride2 + j + l)) {
512 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
513 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
522 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
523 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
524 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
525 stride1 = img1->stride[VPX_PLANE_U]/2;
526 stride2 = img2->stride[VPX_PLANE_U]/2;
527 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
528 for (j = 0; match && j < c_w; j += bsizex) {
530 const int si = mmin(i + bsizey, c_h - i);
531 const int sj = mmin(j + bsizex, c_w - j);
532 for (k = 0; match && k < si; ++k) {
533 for (l = 0; match && l < sj; ++l) {
534 if (*(plane1 + (i + k) * stride1 + j + l) !=
535 *(plane2 + (i + k) * stride2 + j + l)) {
538 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
539 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
548 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
549 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
550 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
551 stride1 = img1->stride[VPX_PLANE_V]/2;
552 stride2 = img2->stride[VPX_PLANE_V]/2;
553 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
554 for (j = 0; match && j < c_w; j += bsizex) {
556 const int si = mmin(i + bsizey, c_h - i);
557 const int sj = mmin(j + bsizex, c_w - j);
558 for (k = 0; match && k < si; ++k) {
559 for (l = 0; match && l < sj; ++l) {
560 if (*(plane1 + (i + k) * stride1 + j + l) !=
561 *(plane2 + (i + k) * stride2 + j + l)) {
564 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
565 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
576 static void find_mismatch(const vpx_image_t *const img1,
577 const vpx_image_t *const img2,
578 int yloc[4], int uloc[4], int vloc[4]) {
579 const uint32_t bsize = 64;
580 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
581 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
583 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
585 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
588 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
589 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
590 for (j = 0; match && j < img1->d_w; j += bsize) {
592 const int si = mmin(i + bsize, img1->d_h) - i;
593 const int sj = mmin(j + bsize, img1->d_w) - j;
594 for (k = 0; match && k < si; ++k) {
595 for (l = 0; match && l < sj; ++l) {
596 if (*(img1->planes[VPX_PLANE_Y] +
597 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
598 *(img2->planes[VPX_PLANE_Y] +
599 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
602 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
603 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
604 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
605 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
614 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
615 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
616 for (j = 0; match && j < c_w; j += bsizex) {
618 const int si = mmin(i + bsizey, c_h - i);
619 const int sj = mmin(j + bsizex, c_w - j);
620 for (k = 0; match && k < si; ++k) {
621 for (l = 0; match && l < sj; ++l) {
622 if (*(img1->planes[VPX_PLANE_U] +
623 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
624 *(img2->planes[VPX_PLANE_U] +
625 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
628 uloc[2] = *(img1->planes[VPX_PLANE_U] +
629 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
630 uloc[3] = *(img2->planes[VPX_PLANE_U] +
631 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
639 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
640 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
641 for (j = 0; match && j < c_w; j += bsizex) {
643 const int si = mmin(i + bsizey, c_h - i);
644 const int sj = mmin(j + bsizex, c_w - j);
645 for (k = 0; match && k < si; ++k) {
646 for (l = 0; match && l < sj; ++l) {
647 if (*(img1->planes[VPX_PLANE_V] +
648 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
649 *(img2->planes[VPX_PLANE_V] +
650 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
653 vloc[2] = *(img1->planes[VPX_PLANE_V] +
654 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
655 vloc[3] = *(img2->planes[VPX_PLANE_V] +
656 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
666 static int compare_img(const vpx_image_t *const img1,
667 const vpx_image_t *const img2) {
668 uint32_t l_w = img1->d_w;
670 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
672 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
676 match &= (img1->fmt == img2->fmt);
677 match &= (img1->d_w == img2->d_w);
678 match &= (img1->d_h == img2->d_h);
679 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
680 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
686 for (i = 0; i < img1->d_h; ++i)
687 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
688 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
691 for (i = 0; i < c_h; ++i)
692 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
693 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
696 for (i = 0; i < c_h; ++i)
697 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
698 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
705 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
706 #define MAX(x,y) ((x)>(y)?(x):(y))
707 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
708 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
709 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
710 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
712 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
713 NELEMENTS(vp9_arg_ctrl_map))
717 typedef int stereo_format_t;
718 struct EbmlGlobal { int debug; };
721 /* Per-stream configuration */
722 struct stream_config {
723 struct vpx_codec_enc_cfg cfg;
725 const char *stats_fn;
726 #if CONFIG_FP_MB_STATS
727 const char *fpmb_stats_fn;
729 stereo_format_t stereo_fmt;
730 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
733 int have_kf_max_dist;
734 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
735 // whether to use 16bit internal buffers
736 int use_16bit_internal;
741 struct stream_state {
743 struct stream_state *next;
744 struct stream_config config;
746 struct rate_hist *rate_hist;
747 struct EbmlGlobal ebml;
748 uint64_t psnr_sse_total;
749 uint64_t psnr_samples_total;
750 double psnr_totals[4];
753 vpx_codec_ctx_t encoder;
754 unsigned int frames_out;
758 #if CONFIG_FP_MB_STATS
759 stats_io_t fpmb_stats;
761 struct vpx_image *img;
762 vpx_codec_ctx_t decoder;
767 void validate_positive_rational(const char *msg,
768 struct vpx_rational *rat) {
775 die("Error: %s must be positive\n", msg);
778 die("Error: %s has zero denominator\n", msg);
782 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
786 /* Initialize default parameters */
787 memset(global, 0, sizeof(*global));
788 global->codec = get_vpx_encoder_by_index(0);
790 global->color_type = I420;
791 /* Assign default deadline to good quality */
792 global->deadline = VPX_DL_GOOD_QUALITY;
794 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
797 if (arg_match(&arg, &codecarg, argi)) {
798 global->codec = get_vpx_encoder_by_name(arg.val);
800 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
801 } else if (arg_match(&arg, &passes, argi)) {
802 global->passes = arg_parse_uint(&arg);
804 if (global->passes < 1 || global->passes > 2)
805 die("Error: Invalid number of passes (%d)\n", global->passes);
806 } else if (arg_match(&arg, &pass_arg, argi)) {
807 global->pass = arg_parse_uint(&arg);
809 if (global->pass < 1 || global->pass > 2)
810 die("Error: Invalid pass selected (%d)\n",
812 } else if (arg_match(&arg, &usage, argi))
813 global->usage = arg_parse_uint(&arg);
814 else if (arg_match(&arg, &deadline, argi))
815 global->deadline = arg_parse_uint(&arg);
816 else if (arg_match(&arg, &best_dl, argi))
817 global->deadline = VPX_DL_BEST_QUALITY;
818 else if (arg_match(&arg, &good_dl, argi))
819 global->deadline = VPX_DL_GOOD_QUALITY;
820 else if (arg_match(&arg, &rt_dl, argi))
821 global->deadline = VPX_DL_REALTIME;
822 else if (arg_match(&arg, &use_yv12, argi))
823 global->color_type = YV12;
824 else if (arg_match(&arg, &use_i420, argi))
825 global->color_type = I420;
826 else if (arg_match(&arg, &use_i422, argi))
827 global->color_type = I422;
828 else if (arg_match(&arg, &use_i444, argi))
829 global->color_type = I444;
830 else if (arg_match(&arg, &quietarg, argi))
832 else if (arg_match(&arg, &verbosearg, argi))
834 else if (arg_match(&arg, &limit, argi))
835 global->limit = arg_parse_uint(&arg);
836 else if (arg_match(&arg, &skip, argi))
837 global->skip_frames = arg_parse_uint(&arg);
838 else if (arg_match(&arg, &psnrarg, argi))
839 global->show_psnr = 1;
840 else if (arg_match(&arg, &recontest, argi))
841 global->test_decode = arg_parse_enum_or_int(&arg);
842 else if (arg_match(&arg, &framerate, argi)) {
843 global->framerate = arg_parse_rational(&arg);
844 validate_positive_rational(arg.name, &global->framerate);
845 global->have_framerate = 1;
846 } else if (arg_match(&arg, &out_part, argi))
847 global->out_part = 1;
848 else if (arg_match(&arg, &debugmode, argi))
850 else if (arg_match(&arg, &q_hist_n, argi))
851 global->show_q_hist_buckets = arg_parse_uint(&arg);
852 else if (arg_match(&arg, &rate_hist_n, argi))
853 global->show_rate_hist_buckets = arg_parse_uint(&arg);
854 else if (arg_match(&arg, &disable_warnings, argi))
855 global->disable_warnings = 1;
856 else if (arg_match(&arg, &disable_warning_prompt, argi))
857 global->disable_warning_prompt = 1;
858 else if (arg_match(&arg, &experimental_bitstream, argi))
859 global->experimental_bitstream = 1;
865 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
866 if (global->pass > global->passes) {
867 warn("Assuming --pass=%d implies --passes=%d\n",
868 global->pass, global->pass);
869 global->passes = global->pass;
872 /* Validate global config */
873 if (global->passes == 0) {
874 #if CONFIG_VP9_ENCODER
875 // Make default VP9 passes = 2 until there is a better quality 1-pass
877 if (global->codec != NULL && global->codec->name != NULL)
878 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
879 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
885 if (global->deadline == VPX_DL_REALTIME &&
886 global->passes > 1) {
887 warn("Enforcing one-pass encoding in realtime mode\n");
893 void open_input_file(struct VpxInputContext *input) {
894 /* Parse certain options from the input file, if possible */
895 input->file = strcmp(input->filename, "-")
896 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
899 fatal("Failed to open input file");
901 if (!fseeko(input->file, 0, SEEK_END)) {
902 /* Input file is seekable. Figure out how long it is, so we can get
905 input->length = ftello(input->file);
909 /* For RAW input sources, these bytes will applied on the first frame
912 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
913 input->detect.position = 0;
915 if (input->detect.buf_read == 4
916 && file_is_y4m(input->detect.buf)) {
917 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
918 input->only_i420) >= 0) {
919 input->file_type = FILE_TYPE_Y4M;
920 input->width = input->y4m.pic_w;
921 input->height = input->y4m.pic_h;
922 input->framerate.numerator = input->y4m.fps_n;
923 input->framerate.denominator = input->y4m.fps_d;
924 input->fmt = input->y4m.vpx_fmt;
925 input->bit_depth = input->y4m.bit_depth;
927 fatal("Unsupported Y4M stream.");
928 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
929 fatal("IVF is not supported as input.");
931 input->file_type = FILE_TYPE_RAW;
936 static void close_input_file(struct VpxInputContext *input) {
938 if (input->file_type == FILE_TYPE_Y4M)
939 y4m_input_close(&input->y4m);
942 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
943 struct stream_state *prev) {
944 struct stream_state *stream;
946 stream = calloc(1, sizeof(*stream));
947 if (stream == NULL) {
948 fatal("Failed to allocate new stream.");
952 memcpy(stream, prev, sizeof(*stream));
958 /* Populate encoder configuration */
959 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
963 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
965 /* Change the default timebase to a high enough value so that the
966 * encoder will always create strictly increasing timestamps.
968 stream->config.cfg.g_timebase.den = 1000;
970 /* Never use the library's default resolution, require it be parsed
971 * from the file or set on the command line.
973 stream->config.cfg.g_w = 0;
974 stream->config.cfg.g_h = 0;
976 /* Initialize remaining stream parameters */
977 stream->config.write_webm = 1;
979 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
980 stream->ebml.last_pts_ns = -1;
981 stream->ebml.writer = NULL;
982 stream->ebml.segment = NULL;
985 /* Allows removal of the application version from the EBML tags */
986 stream->ebml.debug = global->debug;
988 /* Default lag_in_frames is 0 in realtime mode */
989 if (global->deadline == VPX_DL_REALTIME)
990 stream->config.cfg.g_lag_in_frames = 0;
993 /* Output files must be specified for each stream */
994 stream->config.out_fn = NULL;
1001 static int parse_stream_params(struct VpxEncoderConfig *global,
1002 struct stream_state *stream,
1004 char **argi, **argj;
1006 static const arg_def_t **ctrl_args = no_args;
1007 static const int *ctrl_args_map = NULL;
1008 struct stream_config *config = &stream->config;
1009 int eos_mark_found = 0;
1010 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1011 int test_16bit_internal = 0;
1014 // Handle codec specific options
1016 #if CONFIG_VP8_ENCODER
1017 } else if (strcmp(global->codec->name, "vp8") == 0) {
1018 ctrl_args = vp8_args;
1019 ctrl_args_map = vp8_arg_ctrl_map;
1021 #if CONFIG_VP9_ENCODER
1022 } else if (strcmp(global->codec->name, "vp9") == 0) {
1023 ctrl_args = vp9_args;
1024 ctrl_args_map = vp9_arg_ctrl_map;
1028 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1031 /* Once we've found an end-of-stream marker (--) we want to continue
1032 * shifting arguments but not consuming them.
1034 if (eos_mark_found) {
1037 } else if (!strcmp(*argj, "--")) {
1043 } else if (arg_match(&arg, &outputfile, argi)) {
1044 config->out_fn = arg.val;
1045 } else if (arg_match(&arg, &fpf_name, argi)) {
1046 config->stats_fn = arg.val;
1047 #if CONFIG_FP_MB_STATS
1048 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1049 config->fpmb_stats_fn = arg.val;
1051 } else if (arg_match(&arg, &use_ivf, argi)) {
1052 config->write_webm = 0;
1053 } else if (arg_match(&arg, &threads, argi)) {
1054 config->cfg.g_threads = arg_parse_uint(&arg);
1055 } else if (arg_match(&arg, &profile, argi)) {
1056 config->cfg.g_profile = arg_parse_uint(&arg);
1057 } else if (arg_match(&arg, &width, argi)) {
1058 config->cfg.g_w = arg_parse_uint(&arg);
1059 } else if (arg_match(&arg, &height, argi)) {
1060 config->cfg.g_h = arg_parse_uint(&arg);
1061 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1062 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1063 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1064 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1065 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1068 } else if (arg_match(&arg, &stereo_mode, argi)) {
1069 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1071 } else if (arg_match(&arg, &timebase, argi)) {
1072 config->cfg.g_timebase = arg_parse_rational(&arg);
1073 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1074 } else if (arg_match(&arg, &error_resilient, argi)) {
1075 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1076 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1077 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1078 if (global->deadline == VPX_DL_REALTIME &&
1079 config->cfg.g_lag_in_frames != 0) {
1080 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1081 config->cfg.g_lag_in_frames = 0;
1083 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1084 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1085 } else if (arg_match(&arg, &resize_allowed, argi)) {
1086 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1087 } else if (arg_match(&arg, &resize_width, argi)) {
1088 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1089 } else if (arg_match(&arg, &resize_height, argi)) {
1090 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1091 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1092 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1093 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1094 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1095 } else if (arg_match(&arg, &end_usage, argi)) {
1096 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1097 } else if (arg_match(&arg, &target_bitrate, argi)) {
1098 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1099 } else if (arg_match(&arg, &min_quantizer, argi)) {
1100 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1101 } else if (arg_match(&arg, &max_quantizer, argi)) {
1102 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1103 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1104 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1105 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1106 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1107 } else if (arg_match(&arg, &buf_sz, argi)) {
1108 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1109 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1110 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1111 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1112 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1113 } else if (arg_match(&arg, &bias_pct, argi)) {
1114 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1115 if (global->passes < 2)
1116 warn("option %s ignored in one-pass mode.\n", arg.name);
1117 } else if (arg_match(&arg, &minsection_pct, argi)) {
1118 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1120 if (global->passes < 2)
1121 warn("option %s ignored in one-pass mode.\n", arg.name);
1122 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1123 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1125 if (global->passes < 2)
1126 warn("option %s ignored in one-pass mode.\n", arg.name);
1127 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1128 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1129 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1130 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1131 config->have_kf_max_dist = 1;
1132 } else if (arg_match(&arg, &kf_disabled, argi)) {
1133 config->cfg.kf_mode = VPX_KF_DISABLED;
1134 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1135 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1136 if (strcmp(global->codec->name, "vp9") == 0) {
1137 test_16bit_internal = 1;
1142 for (i = 0; ctrl_args[i]; i++) {
1143 if (arg_match(&arg, ctrl_args[i], argi)) {
1147 /* Point either to the next free element or the first
1148 * instance of this control.
1150 for (j = 0; j < config->arg_ctrl_cnt; j++)
1151 if (ctrl_args_map != NULL &&
1152 config->arg_ctrls[j][0] == ctrl_args_map[i])
1156 assert(j < (int)ARG_CTRL_CNT_MAX);
1157 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1158 config->arg_ctrls[j][0] = ctrl_args_map[i];
1159 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1160 if (j == config->arg_ctrl_cnt)
1161 config->arg_ctrl_cnt++;
1170 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1171 if (strcmp(global->codec->name, "vp9") == 0) {
1172 config->use_16bit_internal = test_16bit_internal |
1173 (config->cfg.g_profile > 1);
1176 return eos_mark_found;
1180 #define FOREACH_STREAM(func) \
1182 struct stream_state *stream; \
1183 for (stream = streams; stream; stream = stream->next) { \
1189 static void validate_stream_config(const struct stream_state *stream,
1190 const struct VpxEncoderConfig *global) {
1191 const struct stream_state *streami;
1193 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1194 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1195 " and --height (-h)", stream->index);
1197 if (stream->config.cfg.g_profile != 0 && !global->experimental_bitstream) {
1198 fatal("Stream %d: profile %d is experimental and requires the --%s flag",
1199 stream->index, stream->config.cfg.g_profile,
1200 experimental_bitstream.long_name);
1203 // Check that the codec bit depth is greater than the input bit depth.
1204 if (stream->config.cfg.g_input_bit_depth >
1205 (unsigned int)stream->config.cfg.g_bit_depth) {
1206 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1207 stream->index, (int)stream->config.cfg.g_bit_depth,
1208 stream->config.cfg.g_input_bit_depth);
1211 for (streami = stream; streami; streami = streami->next) {
1212 /* All streams require output files */
1213 if (!streami->config.out_fn)
1214 fatal("Stream %d: Output file is required (specify with -o)",
1217 /* Check for two streams outputting to the same file */
1218 if (streami != stream) {
1219 const char *a = stream->config.out_fn;
1220 const char *b = streami->config.out_fn;
1221 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1222 fatal("Stream %d: duplicate output file (from stream %d)",
1223 streami->index, stream->index);
1226 /* Check for two streams sharing a stats file. */
1227 if (streami != stream) {
1228 const char *a = stream->config.stats_fn;
1229 const char *b = streami->config.stats_fn;
1230 if (a && b && !strcmp(a, b))
1231 fatal("Stream %d: duplicate stats file (from stream %d)",
1232 streami->index, stream->index);
1235 #if CONFIG_FP_MB_STATS
1236 /* Check for two streams sharing a mb stats file. */
1237 if (streami != stream) {
1238 const char *a = stream->config.fpmb_stats_fn;
1239 const char *b = streami->config.fpmb_stats_fn;
1240 if (a && b && !strcmp(a, b))
1241 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1242 streami->index, stream->index);
1249 static void set_stream_dimensions(struct stream_state *stream,
1252 if (!stream->config.cfg.g_w) {
1253 if (!stream->config.cfg.g_h)
1254 stream->config.cfg.g_w = w;
1256 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1258 if (!stream->config.cfg.g_h) {
1259 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1264 static void set_default_kf_interval(struct stream_state *stream,
1265 struct VpxEncoderConfig *global) {
1266 /* Use a max keyframe interval of 5 seconds, if none was
1267 * specified on the command line.
1269 if (!stream->config.have_kf_max_dist) {
1270 double framerate = (double)global->framerate.num / global->framerate.den;
1271 if (framerate > 0.0)
1272 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1276 static const char* file_type_to_string(enum VideoFileType t) {
1278 case FILE_TYPE_RAW: return "RAW";
1279 case FILE_TYPE_Y4M: return "Y4M";
1280 default: return "Other";
1284 static const char* image_format_to_string(vpx_img_fmt_t f) {
1286 case VPX_IMG_FMT_I420: return "I420";
1287 case VPX_IMG_FMT_I422: return "I422";
1288 case VPX_IMG_FMT_I444: return "I444";
1289 case VPX_IMG_FMT_YV12: return "YV12";
1290 default: return "Other";
1294 static void show_stream_config(struct stream_state *stream,
1295 struct VpxEncoderConfig *global,
1296 struct VpxInputContext *input) {
1298 #define SHOW(field) \
1299 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1301 if (stream->index == 0) {
1302 fprintf(stderr, "Codec: %s\n",
1303 vpx_codec_iface_name(global->codec->codec_interface()));
1304 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1306 file_type_to_string(input->file_type),
1307 image_format_to_string(input->fmt));
1309 if (stream->next || stream->index)
1310 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1311 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1312 fprintf(stderr, "Encoder parameters:\n");
1320 SHOW(g_input_bit_depth);
1321 SHOW(g_timebase.num);
1322 SHOW(g_timebase.den);
1323 SHOW(g_error_resilient);
1325 SHOW(g_lag_in_frames);
1326 SHOW(rc_dropframe_thresh);
1327 SHOW(rc_resize_allowed);
1328 SHOW(rc_scaled_width);
1329 SHOW(rc_scaled_height);
1330 SHOW(rc_resize_up_thresh);
1331 SHOW(rc_resize_down_thresh);
1333 SHOW(rc_target_bitrate);
1334 SHOW(rc_min_quantizer);
1335 SHOW(rc_max_quantizer);
1336 SHOW(rc_undershoot_pct);
1337 SHOW(rc_overshoot_pct);
1339 SHOW(rc_buf_initial_sz);
1340 SHOW(rc_buf_optimal_sz);
1341 SHOW(rc_2pass_vbr_bias_pct);
1342 SHOW(rc_2pass_vbr_minsection_pct);
1343 SHOW(rc_2pass_vbr_maxsection_pct);
1350 static void open_output_file(struct stream_state *stream,
1351 struct VpxEncoderConfig *global) {
1352 const char *fn = stream->config.out_fn;
1353 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1355 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1358 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1361 fatal("Failed to open output file");
1363 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1364 fatal("WebM output to pipes not supported.");
1367 if (stream->config.write_webm) {
1368 stream->ebml.stream = stream->file;
1369 write_webm_file_header(&stream->ebml, cfg,
1371 stream->config.stereo_fmt,
1372 global->codec->fourcc);
1376 if (!stream->config.write_webm) {
1377 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1382 static void close_output_file(struct stream_state *stream,
1383 unsigned int fourcc) {
1384 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1386 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1390 if (stream->config.write_webm) {
1391 write_webm_file_footer(&stream->ebml);
1395 if (!stream->config.write_webm) {
1396 if (!fseek(stream->file, 0, SEEK_SET))
1397 ivf_write_file_header(stream->file, &stream->config.cfg,
1399 stream->frames_out);
1402 fclose(stream->file);
1406 static void setup_pass(struct stream_state *stream,
1407 struct VpxEncoderConfig *global,
1409 if (stream->config.stats_fn) {
1410 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1412 fatal("Failed to open statistics store");
1414 if (!stats_open_mem(&stream->stats, pass))
1415 fatal("Failed to open statistics store");
1418 #if CONFIG_FP_MB_STATS
1419 if (stream->config.fpmb_stats_fn) {
1420 if (!stats_open_file(&stream->fpmb_stats,
1421 stream->config.fpmb_stats_fn, pass))
1422 fatal("Failed to open mb statistics store");
1424 if (!stats_open_mem(&stream->fpmb_stats, pass))
1425 fatal("Failed to open mb statistics store");
1429 stream->config.cfg.g_pass = global->passes == 2
1430 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1433 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1434 #if CONFIG_FP_MB_STATS
1435 stream->config.cfg.rc_firstpass_mb_stats_in =
1436 stats_get(&stream->fpmb_stats);
1440 stream->cx_time = 0;
1442 stream->frames_out = 0;
1446 static void initialize_encoder(struct stream_state *stream,
1447 struct VpxEncoderConfig *global) {
1451 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1452 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1453 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1454 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1457 /* Construct Encoder Context */
1458 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1459 &stream->config.cfg, flags);
1460 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1462 /* Note that we bypass the vpx_codec_control wrapper macro because
1463 * we're being clever to store the control IDs in an array. Real
1464 * applications will want to make use of the enumerations directly
1466 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1467 int ctrl = stream->config.arg_ctrls[i][0];
1468 int value = stream->config.arg_ctrls[i][1];
1469 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1470 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1473 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1477 if (global->test_decode != TEST_DECODE_OFF) {
1478 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1479 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1485 static void encode_frame(struct stream_state *stream,
1486 struct VpxEncoderConfig *global,
1487 struct vpx_image *img,
1488 unsigned int frames_in) {
1489 vpx_codec_pts_t frame_start, next_frame_start;
1490 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1491 struct vpx_usec_timer timer;
1493 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1494 * global->framerate.den)
1495 / cfg->g_timebase.num / global->framerate.num;
1496 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1497 * global->framerate.den)
1498 / cfg->g_timebase.num / global->framerate.num;
1500 /* Scale if necessary */
1501 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1503 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1504 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1505 if (img->fmt != VPX_IMG_FMT_I42016) {
1506 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1511 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1512 cfg->g_w, cfg->g_h, 16);
1514 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1515 img->stride[VPX_PLANE_Y]/2,
1516 (uint16*)img->planes[VPX_PLANE_U],
1517 img->stride[VPX_PLANE_U]/2,
1518 (uint16*)img->planes[VPX_PLANE_V],
1519 img->stride[VPX_PLANE_V]/2,
1521 (uint16*)stream->img->planes[VPX_PLANE_Y],
1522 stream->img->stride[VPX_PLANE_Y]/2,
1523 (uint16*)stream->img->planes[VPX_PLANE_U],
1524 stream->img->stride[VPX_PLANE_U]/2,
1525 (uint16*)stream->img->planes[VPX_PLANE_V],
1526 stream->img->stride[VPX_PLANE_V]/2,
1527 stream->img->d_w, stream->img->d_h,
1531 stream->encoder.err = 1;
1532 ctx_exit_on_error(&stream->encoder,
1533 "Stream %d: Failed to encode frame.\n"
1534 "Scaling disabled in this configuration. \n"
1535 "To enable, configure with --enable-libyuv\n",
1541 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1542 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1543 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1548 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1549 cfg->g_w, cfg->g_h, 16);
1550 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1551 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1552 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1554 stream->img->planes[VPX_PLANE_Y],
1555 stream->img->stride[VPX_PLANE_Y],
1556 stream->img->planes[VPX_PLANE_U],
1557 stream->img->stride[VPX_PLANE_U],
1558 stream->img->planes[VPX_PLANE_V],
1559 stream->img->stride[VPX_PLANE_V],
1560 stream->img->d_w, stream->img->d_h,
1564 stream->encoder.err = 1;
1565 ctx_exit_on_error(&stream->encoder,
1566 "Stream %d: Failed to encode frame.\n"
1567 "Scaling disabled in this configuration. \n"
1568 "To enable, configure with --enable-libyuv\n",
1573 vpx_usec_timer_start(&timer);
1574 vpx_codec_encode(&stream->encoder, img, frame_start,
1575 (unsigned long)(next_frame_start - frame_start),
1576 0, global->deadline);
1577 vpx_usec_timer_mark(&timer);
1578 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1579 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1584 static void update_quantizer_histogram(struct stream_state *stream) {
1585 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1588 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1589 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1590 stream->counts[q]++;
1595 static void get_cx_data(struct stream_state *stream,
1596 struct VpxEncoderConfig *global,
1598 const vpx_codec_cx_pkt_t *pkt;
1599 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1600 vpx_codec_iter_t iter = NULL;
1603 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1604 static size_t fsize = 0;
1605 static int64_t ivf_header_pos = 0;
1607 switch (pkt->kind) {
1608 case VPX_CODEC_CX_FRAME_PKT:
1609 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1610 stream->frames_out++;
1613 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1615 update_rate_histogram(stream->rate_hist, cfg, pkt);
1617 if (stream->config.write_webm) {
1618 write_webm_block(&stream->ebml, cfg, pkt);
1621 if (!stream->config.write_webm) {
1622 if (pkt->data.frame.partition_id <= 0) {
1623 ivf_header_pos = ftello(stream->file);
1624 fsize = pkt->data.frame.sz;
1626 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1628 fsize += pkt->data.frame.sz;
1630 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1631 const int64_t currpos = ftello(stream->file);
1632 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1633 ivf_write_frame_size(stream->file, fsize);
1634 fseeko(stream->file, currpos, SEEK_SET);
1638 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1641 stream->nbytes += pkt->data.raw.sz;
1645 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1646 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1647 (unsigned int)pkt->data.frame.sz, NULL, 0);
1648 if (stream->decoder.err) {
1649 warn_or_exit_on_error(&stream->decoder,
1650 global->test_decode == TEST_DECODE_FATAL,
1651 "Failed to decode frame %d in stream %d",
1652 stream->frames_out + 1, stream->index);
1653 stream->mismatch_seen = stream->frames_out + 1;
1658 case VPX_CODEC_STATS_PKT:
1659 stream->frames_out++;
1660 stats_write(&stream->stats,
1661 pkt->data.twopass_stats.buf,
1662 pkt->data.twopass_stats.sz);
1663 stream->nbytes += pkt->data.raw.sz;
1665 #if CONFIG_FP_MB_STATS
1666 case VPX_CODEC_FPMB_STATS_PKT:
1667 stats_write(&stream->fpmb_stats,
1668 pkt->data.firstpass_mb_stats.buf,
1669 pkt->data.firstpass_mb_stats.sz);
1670 stream->nbytes += pkt->data.raw.sz;
1673 case VPX_CODEC_PSNR_PKT:
1675 if (global->show_psnr) {
1678 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1679 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1680 for (i = 0; i < 4; i++) {
1682 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1683 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1685 stream->psnr_count++;
1696 static void show_psnr(struct stream_state *stream) {
1700 if (!stream->psnr_count)
1703 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1704 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, 255.0,
1705 (double)stream->psnr_sse_total);
1706 fprintf(stderr, " %.3f", ovpsnr);
1708 for (i = 0; i < 4; i++) {
1709 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1711 fprintf(stderr, "\n");
1715 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1716 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1719 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1720 static void high_img_upshift(vpx_image_t *dst, vpx_image_t *src,
1722 // Note the offset is 1 less than half
1723 const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0;
1725 if (dst->w != src->w || dst->h != src->h ||
1726 dst->x_chroma_shift != src->x_chroma_shift ||
1727 dst->y_chroma_shift != src->y_chroma_shift ||
1728 dst->fmt != src->fmt || input_shift < 0) {
1729 fatal("Unsupported image conversion");
1732 case VPX_IMG_FMT_I42016:
1733 case VPX_IMG_FMT_I42216:
1734 case VPX_IMG_FMT_I44416:
1737 fatal("Unsupported image conversion");
1740 for (plane = 0; plane < 3; plane++) {
1745 w >>= src->x_chroma_shift;
1746 h >>= src->y_chroma_shift;
1748 for (y = 0; y < h; y++) {
1749 uint16_t *p_src = (uint16_t *)(src->planes[plane] +
1750 y * src->stride[plane]);
1751 uint16_t *p_dst = (uint16_t *)(dst->planes[plane] +
1752 y * dst->stride[plane]);
1753 for (x = 0; x < w; x++)
1754 *p_dst++ = (*p_src++ << input_shift) + offset;
1759 static void low_img_upshift(vpx_image_t *dst, vpx_image_t *src,
1761 // Note the offset is 1 less than half
1762 const int offset = input_shift > 0 ? (1 << (input_shift - 1)) - 1 : 0;
1764 if (dst->w != src->w || dst->h != src->h ||
1765 dst->x_chroma_shift != src->x_chroma_shift ||
1766 dst->y_chroma_shift != src->y_chroma_shift ||
1767 dst->fmt != src->fmt + VPX_IMG_FMT_HIGHBITDEPTH ||
1769 fatal("Unsupported image conversion");
1772 case VPX_IMG_FMT_I420:
1773 case VPX_IMG_FMT_I422:
1774 case VPX_IMG_FMT_I444:
1777 fatal("Unsupported image conversion");
1780 for (plane = 0; plane < 3; plane++) {
1785 w >>= src->x_chroma_shift;
1786 h >>= src->y_chroma_shift;
1788 for (y = 0; y < h; y++) {
1789 uint8_t *p_src = src->planes[plane] + y * src->stride[plane];
1790 uint16_t *p_dst = (uint16_t *)(dst->planes[plane] +
1791 y * dst->stride[plane]);
1792 for (x = 0; x < w; x++) {
1793 *p_dst++ = (*p_src++ << input_shift) + offset;
1799 static void img_upshift(vpx_image_t *dst, vpx_image_t *src,
1801 if (src->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1802 high_img_upshift(dst, src, input_shift);
1804 low_img_upshift(dst, src, input_shift);
1808 static void img_cast_16_to_8(vpx_image_t *dst, vpx_image_t *src) {
1810 if (dst->fmt + VPX_IMG_FMT_HIGHBITDEPTH != src->fmt ||
1811 dst->d_w != src->d_w || dst->d_h != src->d_h ||
1812 dst->x_chroma_shift != src->x_chroma_shift ||
1813 dst->y_chroma_shift != src->y_chroma_shift) {
1814 fatal("Unsupported image conversion");
1817 case VPX_IMG_FMT_I420:
1818 case VPX_IMG_FMT_I422:
1819 case VPX_IMG_FMT_I444:
1822 fatal("Unsupported image conversion");
1825 for (plane = 0; plane < 3; plane++) {
1830 w >>= src->x_chroma_shift;
1831 h >>= src->y_chroma_shift;
1833 for (y = 0; y < h; y++) {
1834 uint16_t *p_src = (uint16_t *)(src->planes[plane] +
1835 y * src->stride[plane]);
1836 uint8_t *p_dst = dst->planes[plane] + y * dst->stride[plane];
1837 for (x = 0; x < w; x++) {
1838 *p_dst++ = *p_src++;
1845 static void test_decode(struct stream_state *stream,
1846 enum TestDecodeFatality fatal,
1847 const VpxInterface *codec) {
1848 vpx_image_t enc_img, dec_img;
1850 if (stream->mismatch_seen)
1853 /* Get the internal reference frame */
1854 if (strcmp(codec->name, "vp8") == 0) {
1855 struct vpx_ref_frame ref_enc, ref_dec;
1858 width = (stream->config.cfg.g_w + 15) & ~15;
1859 height = (stream->config.cfg.g_h + 15) & ~15;
1860 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1861 enc_img = ref_enc.img;
1862 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1863 dec_img = ref_dec.img;
1865 ref_enc.frame_type = VP8_LAST_FRAME;
1866 ref_dec.frame_type = VP8_LAST_FRAME;
1867 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1868 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1870 struct vp9_ref_frame ref_enc, ref_dec;
1874 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1875 enc_img = ref_enc.img;
1876 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1877 dec_img = ref_dec.img;
1878 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1879 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1880 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1881 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1882 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1883 enc_img.d_w, enc_img.d_h, 16);
1884 img_cast_16_to_8(&enc_img, &ref_enc.img);
1886 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1887 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1888 dec_img.d_w, dec_img.d_h, 16);
1889 img_cast_16_to_8(&dec_img, &ref_dec.img);
1894 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1895 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1897 if (!compare_img(&enc_img, &dec_img)) {
1898 int y[4], u[4], v[4];
1899 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1900 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1901 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1903 find_mismatch(&enc_img, &dec_img, y, u, v);
1906 find_mismatch(&enc_img, &dec_img, y, u, v);
1908 stream->decoder.err = 1;
1909 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1910 "Stream %d: Encode/decode mismatch on frame %d at"
1911 " Y[%d, %d] {%d/%d},"
1912 " U[%d, %d] {%d/%d},"
1913 " V[%d, %d] {%d/%d}",
1914 stream->index, stream->frames_out,
1915 y[0], y[1], y[2], y[3],
1916 u[0], u[1], u[2], u[3],
1917 v[0], v[1], v[2], v[3]);
1918 stream->mismatch_seen = stream->frames_out;
1921 vpx_img_free(&enc_img);
1922 vpx_img_free(&dec_img);
1926 static void print_time(const char *label, int64_t etl) {
1933 etl -= hours * 3600;
1938 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1939 label, hours, mins, secs);
1941 fprintf(stderr, "[%3s unknown] ", label);
1946 int main(int argc, const char **argv_) {
1949 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1950 vpx_image_t raw_shift;
1951 int allocated_raw_shift = 0;
1952 int use_16bit_internal = 0;
1953 int input_shift = 0;
1955 int frame_avail, got_data;
1957 struct VpxInputContext input;
1958 struct VpxEncoderConfig global;
1959 struct stream_state *streams = NULL;
1960 char **argv, **argi;
1961 uint64_t cx_time = 0;
1965 memset(&input, 0, sizeof(input));
1966 exec_name = argv_[0];
1971 /* Setup default input stream settings */
1972 input.framerate.numerator = 30;
1973 input.framerate.denominator = 1;
1974 input.only_i420 = 1;
1975 input.bit_depth = 0;
1977 /* First parse the global configuration values, because we want to apply
1978 * other parameters on top of the default configuration provided by the
1981 argv = argv_dup(argc - 1, argv_ + 1);
1982 parse_global_config(&global, argv);
1984 switch (global.color_type) {
1986 input.fmt = VPX_IMG_FMT_I420;
1989 input.fmt = VPX_IMG_FMT_I422;
1992 input.fmt = VPX_IMG_FMT_I444;
1995 input.fmt = VPX_IMG_FMT_YV12;
2000 /* Now parse each stream's parameters. Using a local scope here
2001 * due to the use of 'stream' as loop variable in FOREACH_STREAM
2004 struct stream_state *stream = NULL;
2007 stream = new_stream(&global, stream);
2011 } while (parse_stream_params(&global, stream, argv));
2014 /* Check for unrecognized options */
2015 for (argi = argv; *argi; argi++)
2016 if (argi[0][0] == '-' && argi[0][1])
2017 die("Error: Unrecognized option %s\n", *argi);
2019 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
2020 &global, &stream->config.cfg););
2022 /* Handle non-option arguments */
2023 input.filename = argv[0];
2025 if (!input.filename)
2028 /* Decide if other chroma subsamplings than 4:2:0 are supported */
2029 if (global.codec->fourcc == VP9_FOURCC)
2030 input.only_i420 = 0;
2032 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
2033 int frames_in = 0, seen_frames = 0;
2034 int64_t estimated_time_left = -1;
2035 int64_t average_rate = -1;
2036 int64_t lagged_count = 0;
2038 open_input_file(&input);
2040 /* If the input file doesn't specify its w/h (raw files), try to get
2041 * the data from the first stream's configuration.
2043 if (!input.width || !input.height)
2045 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
2046 input.width = stream->config.cfg.g_w;
2047 input.height = stream->config.cfg.g_h;
2052 /* Update stream configurations from the input file's parameters */
2053 if (!input.width || !input.height)
2054 fatal("Specify stream dimensions with --width (-w) "
2055 " and --height (-h)");
2057 /* If input file does not specify bit-depth but input-bit-depth parameter
2058 * exists, assume that to be the input bit-depth. However, if the
2059 * input-bit-depth paramter does not exist, assume the input bit-depth
2060 * to be the same as the codec bit-depth.
2062 if (!input.bit_depth) {
2064 if (stream->config.cfg.g_input_bit_depth)
2065 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2067 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2068 (int)stream->config.cfg.g_bit_depth;
2070 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2073 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2077 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2078 FOREACH_STREAM(validate_stream_config(stream, &global));
2080 /* Ensure that --passes and --pass are consistent. If --pass is set and
2081 * --passes=2, ensure --fpf was set.
2083 if (global.pass && global.passes == 2)
2085 if (!stream->config.stats_fn)
2086 die("Stream %d: Must specify --fpf when --pass=%d"
2087 " and --passes=2\n", stream->index, global.pass);
2092 stream->config.write_webm = 0;
2093 warn("vpxenc was compiled without WebM container support."
2094 "Producing IVF output");
2098 /* Use the frame rate from the file only if none was specified
2099 * on the command-line.
2101 if (!global.have_framerate) {
2102 global.framerate.num = input.framerate.numerator;
2103 global.framerate.den = input.framerate.denominator;
2106 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2108 /* Show configuration */
2109 if (global.verbose && pass == 0)
2110 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2112 if (pass == (global.pass ? global.pass - 1 : 0)) {
2113 if (input.file_type == FILE_TYPE_Y4M)
2114 /*The Y4M reader does its own allocation.
2115 Just initialize this here to avoid problems if we never read any
2117 memset(&raw, 0, sizeof(raw));
2119 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2121 FOREACH_STREAM(stream->rate_hist =
2122 init_rate_histogram(&stream->config.cfg,
2123 &global.framerate));
2126 FOREACH_STREAM(setup_pass(stream, &global, pass));
2127 FOREACH_STREAM(open_output_file(stream, &global));
2128 FOREACH_STREAM(initialize_encoder(stream, &global));
2130 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2131 if (strcmp(global.codec->name, "vp9") == 0) {
2132 // Check to see if at least one stream uses 16 bit internal.
2133 // Currently assume that the bit_depths for all streams using
2134 // highbitdepth are the same.
2136 if (stream->config.use_16bit_internal) {
2137 use_16bit_internal = 1;
2139 if (stream->config.cfg.g_profile == 0) {
2142 input_shift = (int)stream->config.cfg.g_bit_depth -
2143 stream->config.cfg.g_input_bit_depth;
2152 while (frame_avail || got_data) {
2153 struct vpx_usec_timer timer;
2155 if (!global.limit || frames_in < global.limit) {
2156 frame_avail = read_frame(&input, &raw);
2160 seen_frames = frames_in > global.skip_frames ?
2161 frames_in - global.skip_frames : 0;
2163 if (!global.quiet) {
2164 float fps = usec_to_fps(cx_time, seen_frames);
2165 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2167 if (stream_cnt == 1)
2169 "frame %4d/%-4d %7"PRId64"B ",
2170 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2172 fprintf(stderr, "frame %4d ", frames_in);
2174 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2175 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2176 cx_time > 9999999 ? "ms" : "us",
2177 fps >= 1.0 ? fps : fps * 60,
2178 fps >= 1.0 ? "fps" : "fpm");
2179 print_time("ETA", estimated_time_left);
2185 if (frames_in > global.skip_frames) {
2186 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2187 vpx_image_t *frame_to_encode;
2188 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2189 assert(use_16bit_internal);
2190 // Input bit depth and stream bit depth do not match, so up
2191 // shift frame to stream bit depth
2192 if (!allocated_raw_shift) {
2193 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2194 input.width, input.height, 32);
2195 allocated_raw_shift = 1;
2197 img_upshift(&raw_shift, &raw, input_shift);
2198 frame_to_encode = &raw_shift;
2200 frame_to_encode = &raw;
2202 vpx_usec_timer_start(&timer);
2203 if (use_16bit_internal) {
2204 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2206 if (stream->config.use_16bit_internal)
2207 encode_frame(stream, &global,
2208 frame_avail ? frame_to_encode : NULL,
2214 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2215 FOREACH_STREAM(encode_frame(stream, &global,
2216 frame_avail ? frame_to_encode : NULL,
2220 vpx_usec_timer_start(&timer);
2221 FOREACH_STREAM(encode_frame(stream, &global,
2222 frame_avail ? &raw : NULL,
2225 vpx_usec_timer_mark(&timer);
2226 cx_time += vpx_usec_timer_elapsed(&timer);
2228 FOREACH_STREAM(update_quantizer_histogram(stream));
2231 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2233 if (!got_data && input.length && streams != NULL &&
2234 !streams->frames_out) {
2235 lagged_count = global.limit ? seen_frames : ftello(input.file);
2236 } else if (input.length) {
2241 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2243 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2244 remaining = 1000 * (global.limit - global.skip_frames
2245 - seen_frames + lagged_count);
2247 const int64_t input_pos = ftello(input.file);
2248 const int64_t input_pos_lagged = input_pos - lagged_count;
2249 const int64_t limit = input.length;
2251 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2252 remaining = limit - input_pos + lagged_count;
2255 average_rate = (average_rate <= 0)
2257 : (average_rate * 7 + rate) / 8;
2258 estimated_time_left = average_rate ? remaining / average_rate : -1;
2261 if (got_data && global.test_decode != TEST_DECODE_OFF)
2262 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2267 fprintf(stderr, "\033[K");
2271 fprintf(stderr, "\n");
2274 FOREACH_STREAM(fprintf(
2276 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7lub/f %7"PRId64"b/s"
2277 " %7"PRId64" %s (%.2f fps)\033[K\n", pass + 1,
2278 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2279 seen_frames ? (unsigned long)(stream->nbytes * 8 / seen_frames) : 0,
2280 seen_frames ? (int64_t)stream->nbytes * 8
2281 * (int64_t)global.framerate.num / global.framerate.den
2284 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2285 stream->cx_time > 9999999 ? "ms" : "us",
2286 usec_to_fps(stream->cx_time, seen_frames));
2289 if (global.show_psnr)
2290 FOREACH_STREAM(show_psnr(stream));
2292 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2294 if (global.test_decode != TEST_DECODE_OFF) {
2295 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2298 close_input_file(&input);
2300 if (global.test_decode == TEST_DECODE_FATAL) {
2301 FOREACH_STREAM(res |= stream->mismatch_seen);
2303 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2305 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2307 #if CONFIG_FP_MB_STATS
2308 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2315 if (global.show_q_hist_buckets)
2316 FOREACH_STREAM(show_q_histogram(stream->counts,
2317 global.show_q_hist_buckets));
2319 if (global.show_rate_hist_buckets)
2320 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2321 &stream->config.cfg,
2322 global.show_rate_hist_buckets));
2323 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2325 #if CONFIG_INTERNAL_STATS
2326 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2327 * to match some existing utilities.
2329 if (!(global.pass == 1 && global.passes == 2))
2331 FILE *f = fopen("opsnr.stt", "a");
2332 if (stream->mismatch_seen) {
2333 fprintf(f, "First mismatch occurred in frame %d\n",
2334 stream->mismatch_seen);
2336 fprintf(f, "No mismatch detected in recon buffers\n");
2342 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2343 if (allocated_raw_shift)
2344 vpx_img_free(&raw_shift);
2349 return res ? EXIT_FAILURE : EXIT_SUCCESS;