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(
132 "D", "debug", 0, "Debug mode (makes output deterministic)");
133 static const arg_def_t outputfile = ARG_DEF(
134 "o", "output", 1, "Output filename");
135 static const arg_def_t use_yv12 = ARG_DEF(
136 NULL, "yv12", 0, "Input file is YV12 ");
137 static const arg_def_t use_i420 = ARG_DEF(
138 NULL, "i420", 0, "Input file is I420 (default)");
139 static const arg_def_t use_i422 = ARG_DEF(
140 NULL, "i422", 0, "Input file is I422");
141 static const arg_def_t use_i444 = ARG_DEF(
142 NULL, "i444", 0, "Input file is I444");
143 static const arg_def_t use_i440 = ARG_DEF(
144 NULL, "i440", 0, "Input file is I440");
145 static const arg_def_t codecarg = ARG_DEF(
146 NULL, "codec", 1, "Codec to use");
147 static const arg_def_t passes = ARG_DEF(
148 "p", "passes", 1, "Number of passes (1/2)");
149 static const arg_def_t pass_arg = ARG_DEF(
150 NULL, "pass", 1, "Pass to execute (1/2)");
151 static const arg_def_t fpf_name = ARG_DEF(
152 NULL, "fpf", 1, "First pass statistics file name");
153 #if CONFIG_FP_MB_STATS
154 static const arg_def_t fpmbf_name = ARG_DEF(
155 NULL, "fpmbf", 1, "First pass block statistics file name");
157 static const arg_def_t limit = ARG_DEF(
158 NULL, "limit", 1, "Stop encoding after n input frames");
159 static const arg_def_t skip = ARG_DEF(
160 NULL, "skip", 1, "Skip the first n input frames");
161 static const arg_def_t deadline = ARG_DEF(
162 "d", "deadline", 1, "Deadline per frame (usec)");
163 static const arg_def_t best_dl = ARG_DEF(
164 NULL, "best", 0, "Use Best Quality Deadline");
165 static const arg_def_t good_dl = ARG_DEF(
166 NULL, "good", 0, "Use Good Quality Deadline");
167 static const arg_def_t rt_dl = ARG_DEF(
168 NULL, "rt", 0, "Use Realtime Quality Deadline");
169 static const arg_def_t quietarg = ARG_DEF(
170 "q", "quiet", 0, "Do not print encode progress");
171 static const arg_def_t verbosearg = ARG_DEF(
172 "v", "verbose", 0, "Show encoder parameters");
173 static const arg_def_t psnrarg = ARG_DEF(
174 NULL, "psnr", 0, "Show PSNR in status line");
176 static const struct arg_enum_list test_decode_enum[] = {
177 {"off", TEST_DECODE_OFF},
178 {"fatal", TEST_DECODE_FATAL},
179 {"warn", TEST_DECODE_WARN},
182 static const arg_def_t recontest = ARG_DEF_ENUM(
183 NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
184 static const arg_def_t framerate = ARG_DEF(
185 NULL, "fps", 1, "Stream frame rate (rate/scale)");
186 static const arg_def_t use_webm = ARG_DEF(
187 NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
188 static const arg_def_t use_ivf = ARG_DEF(
189 NULL, "ivf", 0, "Output IVF");
190 static const arg_def_t out_part = ARG_DEF(
191 "P", "output-partitions", 0,
192 "Makes encoder output partitions. Requires IVF output!");
193 static const arg_def_t q_hist_n = ARG_DEF(
194 NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
195 static const arg_def_t rate_hist_n = ARG_DEF(
196 NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
197 static const arg_def_t disable_warnings = ARG_DEF(
198 NULL, "disable-warnings", 0,
199 "Disable warnings about potentially incorrect encode settings.");
200 static const arg_def_t disable_warning_prompt = ARG_DEF(
201 "y", "disable-warning-prompt", 0,
202 "Display warnings, but do not prompt user to continue.");
204 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
205 static const arg_def_t test16bitinternalarg = ARG_DEF(
206 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
209 static const arg_def_t *main_args[] = {
211 &outputfile, &codecarg, &passes, &pass_arg, &fpf_name, &limit, &skip,
212 &deadline, &best_dl, &good_dl, &rt_dl,
213 &quietarg, &verbosearg, &psnrarg, &use_webm, &use_ivf, &out_part, &q_hist_n,
214 &rate_hist_n, &disable_warnings, &disable_warning_prompt,
218 static const arg_def_t usage = ARG_DEF(
219 "u", "usage", 1, "Usage profile number to use");
220 static const arg_def_t threads = ARG_DEF(
221 "t", "threads", 1, "Max number of threads to use");
222 static const arg_def_t profile = ARG_DEF(
223 NULL, "profile", 1, "Bitstream profile number to use");
224 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
225 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
227 static const struct arg_enum_list stereo_mode_enum[] = {
228 {"mono", STEREO_FORMAT_MONO},
229 {"left-right", STEREO_FORMAT_LEFT_RIGHT},
230 {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
231 {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
232 {"right-left", STEREO_FORMAT_RIGHT_LEFT},
235 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
236 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
238 static const arg_def_t timebase = ARG_DEF(
239 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
240 static const arg_def_t error_resilient = ARG_DEF(
241 NULL, "error-resilient", 1, "Enable error resiliency features");
242 static const arg_def_t lag_in_frames = ARG_DEF(
243 NULL, "lag-in-frames", 1, "Max number of frames to lag");
245 static const arg_def_t *global_args[] = {
246 &use_yv12, &use_i420, &use_i422, &use_i444, &use_i440,
247 &usage, &threads, &profile,
252 &timebase, &framerate,
254 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
255 &test16bitinternalarg,
260 static const arg_def_t dropframe_thresh = ARG_DEF(
261 NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
262 static const arg_def_t resize_allowed = ARG_DEF(
263 NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
264 static const arg_def_t resize_width = ARG_DEF(
265 NULL, "resize-width", 1, "Width of encoded frame");
266 static const arg_def_t resize_height = ARG_DEF(
267 NULL, "resize-height", 1, "Height of encoded frame");
268 static const arg_def_t resize_up_thresh = ARG_DEF(
269 NULL, "resize-up", 1, "Upscale threshold (buf %)");
270 static const arg_def_t resize_down_thresh = ARG_DEF(
271 NULL, "resize-down", 1, "Downscale threshold (buf %)");
272 static const struct arg_enum_list end_usage_enum[] = {
279 static const arg_def_t end_usage = ARG_DEF_ENUM(
280 NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
281 static const arg_def_t target_bitrate = ARG_DEF(
282 NULL, "target-bitrate", 1, "Bitrate (kbps)");
283 static const arg_def_t min_quantizer = ARG_DEF(
284 NULL, "min-q", 1, "Minimum (best) quantizer");
285 static const arg_def_t max_quantizer = ARG_DEF(
286 NULL, "max-q", 1, "Maximum (worst) quantizer");
287 static const arg_def_t undershoot_pct = ARG_DEF(
288 NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
289 static const arg_def_t overshoot_pct = ARG_DEF(
290 NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
291 static const arg_def_t buf_sz = ARG_DEF(
292 NULL, "buf-sz", 1, "Client buffer size (ms)");
293 static const arg_def_t buf_initial_sz = ARG_DEF(
294 NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
295 static const arg_def_t buf_optimal_sz = ARG_DEF(
296 NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
297 static const arg_def_t *rc_args[] = {
298 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
299 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
300 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct, &buf_sz,
301 &buf_initial_sz, &buf_optimal_sz, NULL
305 static const arg_def_t bias_pct = ARG_DEF(
306 NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
307 static const arg_def_t minsection_pct = ARG_DEF(
308 NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
309 static const arg_def_t maxsection_pct = ARG_DEF(
310 NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
311 static const arg_def_t *rc_twopass_args[] = {
312 &bias_pct, &minsection_pct, &maxsection_pct, NULL
316 static const arg_def_t kf_min_dist = ARG_DEF(
317 NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
318 static const arg_def_t kf_max_dist = ARG_DEF(
319 NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
320 static const arg_def_t kf_disabled = ARG_DEF(
321 NULL, "disable-kf", 0, "Disable keyframe placement");
322 static const arg_def_t *kf_args[] = {
323 &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
327 static const arg_def_t noise_sens = ARG_DEF(
328 NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
329 static const arg_def_t sharpness = ARG_DEF(
330 NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
331 static const arg_def_t static_thresh = ARG_DEF(
332 NULL, "static-thresh", 1, "Motion detection threshold");
333 static const arg_def_t cpu_used_vp8 = ARG_DEF(
334 NULL, "cpu-used", 1, "CPU Used (-16..16)");
335 static const arg_def_t cpu_used_vp9 = ARG_DEF(
336 NULL, "cpu-used", 1, "CPU Used (-8..8)");
337 static const arg_def_t auto_altref = ARG_DEF(
338 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
339 static const arg_def_t arnr_maxframes = ARG_DEF(
340 NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
341 static const arg_def_t arnr_strength = ARG_DEF(
342 NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
343 static const arg_def_t arnr_type = ARG_DEF(
344 NULL, "arnr-type", 1, "AltRef type");
345 static const struct arg_enum_list tuning_enum[] = {
346 {"psnr", VP8_TUNE_PSNR},
347 {"ssim", VP8_TUNE_SSIM},
350 static const arg_def_t tune_ssim = ARG_DEF_ENUM(
351 NULL, "tune", 1, "Material to favor", tuning_enum);
352 static const arg_def_t cq_level = ARG_DEF(
353 NULL, "cq-level", 1, "Constant/Constrained Quality level");
354 static const arg_def_t max_intra_rate_pct = ARG_DEF(
355 NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
356 static const arg_def_t max_inter_rate_pct = ARG_DEF(
357 NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
358 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
359 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
361 static const arg_def_t screen_content_mode = ARG_DEF(NULL, "screen-content-mode", 1,
362 "Screen content mode");
364 #if CONFIG_VP8_ENCODER
365 static const arg_def_t token_parts = ARG_DEF(
366 NULL, "token-parts", 1, "Number of token partitions to use, log2");
367 static const arg_def_t *vp8_args[] = {
368 &cpu_used_vp8, &auto_altref, &noise_sens, &sharpness, &static_thresh,
369 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
370 &tune_ssim, &cq_level, &max_intra_rate_pct, &screen_content_mode,
373 static const int vp8_arg_ctrl_map[] = {
374 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
375 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
376 VP8E_SET_TOKEN_PARTITIONS,
377 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
378 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
379 VP8E_SET_SCREEN_CONTENT_MODE,
384 #if CONFIG_VP9_ENCODER
385 static const arg_def_t tile_cols = ARG_DEF(
386 NULL, "tile-columns", 1, "Number of tile columns to use, log2");
387 static const arg_def_t tile_rows = ARG_DEF(
388 NULL, "tile-rows", 1, "Number of tile rows to use, log2");
389 static const arg_def_t lossless = ARG_DEF(
390 NULL, "lossless", 1, "Lossless mode");
391 static const arg_def_t frame_parallel_decoding = ARG_DEF(
392 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
393 static const arg_def_t aq_mode = ARG_DEF(
395 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
396 "3: cyclic refresh)");
397 static const arg_def_t frame_periodic_boost = ARG_DEF(
398 NULL, "frame-boost", 1,
399 "Enable frame periodic boost (0: off (default), 1: on)");
401 static const struct arg_enum_list color_space_enum[] = {
402 { "unknown", VPX_CS_UNKNOWN },
403 { "bt601", VPX_CS_BT_601 },
404 { "bt709", VPX_CS_BT_709 },
405 { "smpte170", VPX_CS_SMPTE_170 },
406 { "smpte240", VPX_CS_SMPTE_240 },
407 { "bt2020", VPX_CS_BT_2020 },
408 { "reserved", VPX_CS_RESERVED },
409 { "sRGB", VPX_CS_SRGB },
413 static const arg_def_t input_color_space = ARG_DEF_ENUM(
414 NULL, "color-space", 1,
415 "The color space of input content:", color_space_enum);
417 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
418 static const struct arg_enum_list bitdepth_enum[] = {
425 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
427 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
429 static const arg_def_t inbitdeptharg = ARG_DEF(
430 NULL, "input-bit-depth", 1, "Bit depth of input");
433 static const struct arg_enum_list tune_content_enum[] = {
434 {"default", VP9E_CONTENT_DEFAULT},
435 {"screen", VP9E_CONTENT_SCREEN},
439 static const arg_def_t tune_content = ARG_DEF_ENUM(
440 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
442 static const arg_def_t *vp9_args[] = {
443 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
444 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
445 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
446 &gf_cbr_boost_pct, &lossless,
447 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
448 &noise_sens, &tune_content, &input_color_space,
449 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
450 &bitdeptharg, &inbitdeptharg,
454 static const int vp9_arg_ctrl_map[] = {
455 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
456 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
457 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
458 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
459 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
460 VP8E_SET_MAX_INTER_BITRATE_PCT, VP8E_SET_GF_CBR_BOOST_PCT,
461 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
462 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
463 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
468 static const arg_def_t *no_args[] = { NULL };
473 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
476 fprintf(stderr, "\nOptions:\n");
477 arg_show_usage(stderr, main_args);
478 fprintf(stderr, "\nEncoder Global Options:\n");
479 arg_show_usage(stderr, global_args);
480 fprintf(stderr, "\nRate Control Options:\n");
481 arg_show_usage(stderr, rc_args);
482 fprintf(stderr, "\nTwopass Rate Control Options:\n");
483 arg_show_usage(stderr, rc_twopass_args);
484 fprintf(stderr, "\nKeyframe Placement Options:\n");
485 arg_show_usage(stderr, kf_args);
486 #if CONFIG_VP8_ENCODER
487 fprintf(stderr, "\nVP8 Specific Options:\n");
488 arg_show_usage(stderr, vp8_args);
490 #if CONFIG_VP9_ENCODER
491 fprintf(stderr, "\nVP9 Specific Options:\n");
492 arg_show_usage(stderr, vp9_args);
494 fprintf(stderr, "\nStream timebase (--timebase):\n"
495 " The desired precision of timestamps in the output, expressed\n"
496 " in fractional seconds. Default is 1/1000.\n");
497 fprintf(stderr, "\nIncluded encoders:\n\n");
499 for (i = 0; i < get_vpx_encoder_count(); ++i) {
500 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
501 fprintf(stderr, " %-6s - %s\n",
502 encoder->name, vpx_codec_iface_name(encoder->codec_interface()));
508 #define mmin(a, b) ((a) < (b) ? (a) : (b))
510 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
511 static void find_mismatch_high(const vpx_image_t *const img1,
512 const vpx_image_t *const img2,
513 int yloc[4], int uloc[4], int vloc[4]) {
514 uint16_t *plane1, *plane2;
515 uint32_t stride1, stride2;
516 const uint32_t bsize = 64;
517 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
518 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
520 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
522 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
525 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
526 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
527 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
528 stride1 = img1->stride[VPX_PLANE_Y]/2;
529 stride2 = img2->stride[VPX_PLANE_Y]/2;
530 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
531 for (j = 0; match && j < img1->d_w; j += bsize) {
533 const int si = mmin(i + bsize, img1->d_h) - i;
534 const int sj = mmin(j + bsize, img1->d_w) - j;
535 for (k = 0; match && k < si; ++k) {
536 for (l = 0; match && l < sj; ++l) {
537 if (*(plane1 + (i + k) * stride1 + j + l) !=
538 *(plane2 + (i + k) * stride2 + j + l)) {
541 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
542 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
551 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
552 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
553 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
554 stride1 = img1->stride[VPX_PLANE_U]/2;
555 stride2 = img2->stride[VPX_PLANE_U]/2;
556 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
557 for (j = 0; match && j < c_w; j += bsizex) {
559 const int si = mmin(i + bsizey, c_h - i);
560 const int sj = mmin(j + bsizex, c_w - j);
561 for (k = 0; match && k < si; ++k) {
562 for (l = 0; match && l < sj; ++l) {
563 if (*(plane1 + (i + k) * stride1 + j + l) !=
564 *(plane2 + (i + k) * stride2 + j + l)) {
567 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
568 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
577 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
578 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
579 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
580 stride1 = img1->stride[VPX_PLANE_V]/2;
581 stride2 = img2->stride[VPX_PLANE_V]/2;
582 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
583 for (j = 0; match && j < c_w; j += bsizex) {
585 const int si = mmin(i + bsizey, c_h - i);
586 const int sj = mmin(j + bsizex, c_w - j);
587 for (k = 0; match && k < si; ++k) {
588 for (l = 0; match && l < sj; ++l) {
589 if (*(plane1 + (i + k) * stride1 + j + l) !=
590 *(plane2 + (i + k) * stride2 + j + l)) {
593 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
594 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
605 static void find_mismatch(const vpx_image_t *const img1,
606 const vpx_image_t *const img2,
607 int yloc[4], int uloc[4], int vloc[4]) {
608 const uint32_t bsize = 64;
609 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
610 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
612 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
614 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
617 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
618 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
619 for (j = 0; match && j < img1->d_w; j += bsize) {
621 const int si = mmin(i + bsize, img1->d_h) - i;
622 const int sj = mmin(j + bsize, img1->d_w) - j;
623 for (k = 0; match && k < si; ++k) {
624 for (l = 0; match && l < sj; ++l) {
625 if (*(img1->planes[VPX_PLANE_Y] +
626 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
627 *(img2->planes[VPX_PLANE_Y] +
628 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
631 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
632 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
633 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
634 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
643 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
644 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
645 for (j = 0; match && j < c_w; j += bsizex) {
647 const int si = mmin(i + bsizey, c_h - i);
648 const int sj = mmin(j + bsizex, c_w - j);
649 for (k = 0; match && k < si; ++k) {
650 for (l = 0; match && l < sj; ++l) {
651 if (*(img1->planes[VPX_PLANE_U] +
652 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
653 *(img2->planes[VPX_PLANE_U] +
654 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
657 uloc[2] = *(img1->planes[VPX_PLANE_U] +
658 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
659 uloc[3] = *(img2->planes[VPX_PLANE_U] +
660 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
668 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
669 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
670 for (j = 0; match && j < c_w; j += bsizex) {
672 const int si = mmin(i + bsizey, c_h - i);
673 const int sj = mmin(j + bsizex, c_w - j);
674 for (k = 0; match && k < si; ++k) {
675 for (l = 0; match && l < sj; ++l) {
676 if (*(img1->planes[VPX_PLANE_V] +
677 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
678 *(img2->planes[VPX_PLANE_V] +
679 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
682 vloc[2] = *(img1->planes[VPX_PLANE_V] +
683 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
684 vloc[3] = *(img2->planes[VPX_PLANE_V] +
685 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
695 static int compare_img(const vpx_image_t *const img1,
696 const vpx_image_t *const img2) {
697 uint32_t l_w = img1->d_w;
699 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
701 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
705 match &= (img1->fmt == img2->fmt);
706 match &= (img1->d_w == img2->d_w);
707 match &= (img1->d_h == img2->d_h);
708 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
709 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
715 for (i = 0; i < img1->d_h; ++i)
716 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
717 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
720 for (i = 0; i < c_h; ++i)
721 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
722 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
725 for (i = 0; i < c_h; ++i)
726 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
727 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
734 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
735 #define MAX(x,y) ((x)>(y)?(x):(y))
736 #if CONFIG_VP8_ENCODER && !CONFIG_VP9_ENCODER
737 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
738 #elif !CONFIG_VP8_ENCODER && CONFIG_VP9_ENCODER
739 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
741 #define ARG_CTRL_CNT_MAX MAX(NELEMENTS(vp8_arg_ctrl_map), \
742 NELEMENTS(vp9_arg_ctrl_map))
746 typedef int stereo_format_t;
747 struct EbmlGlobal { int debug; };
750 /* Per-stream configuration */
751 struct stream_config {
752 struct vpx_codec_enc_cfg cfg;
754 const char *stats_fn;
755 #if CONFIG_FP_MB_STATS
756 const char *fpmb_stats_fn;
758 stereo_format_t stereo_fmt;
759 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
762 int have_kf_max_dist;
763 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
764 // whether to use 16bit internal buffers
765 int use_16bit_internal;
770 struct stream_state {
772 struct stream_state *next;
773 struct stream_config config;
775 struct rate_hist *rate_hist;
776 struct EbmlGlobal ebml;
777 uint64_t psnr_sse_total;
778 uint64_t psnr_samples_total;
779 double psnr_totals[4];
782 vpx_codec_ctx_t encoder;
783 unsigned int frames_out;
787 #if CONFIG_FP_MB_STATS
788 stats_io_t fpmb_stats;
790 struct vpx_image *img;
791 vpx_codec_ctx_t decoder;
796 void validate_positive_rational(const char *msg,
797 struct vpx_rational *rat) {
804 die("Error: %s must be positive\n", msg);
807 die("Error: %s has zero denominator\n", msg);
811 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
815 /* Initialize default parameters */
816 memset(global, 0, sizeof(*global));
817 global->codec = get_vpx_encoder_by_index(0);
819 global->color_type = I420;
820 /* Assign default deadline to good quality */
821 global->deadline = VPX_DL_GOOD_QUALITY;
823 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
826 if (arg_match(&arg, &codecarg, argi)) {
827 global->codec = get_vpx_encoder_by_name(arg.val);
829 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
830 } else if (arg_match(&arg, &passes, argi)) {
831 global->passes = arg_parse_uint(&arg);
833 if (global->passes < 1 || global->passes > 2)
834 die("Error: Invalid number of passes (%d)\n", global->passes);
835 } else if (arg_match(&arg, &pass_arg, argi)) {
836 global->pass = arg_parse_uint(&arg);
838 if (global->pass < 1 || global->pass > 2)
839 die("Error: Invalid pass selected (%d)\n",
841 } else if (arg_match(&arg, &usage, argi))
842 global->usage = arg_parse_uint(&arg);
843 else if (arg_match(&arg, &deadline, argi))
844 global->deadline = arg_parse_uint(&arg);
845 else if (arg_match(&arg, &best_dl, argi))
846 global->deadline = VPX_DL_BEST_QUALITY;
847 else if (arg_match(&arg, &good_dl, argi))
848 global->deadline = VPX_DL_GOOD_QUALITY;
849 else if (arg_match(&arg, &rt_dl, argi))
850 global->deadline = VPX_DL_REALTIME;
851 else if (arg_match(&arg, &use_yv12, argi))
852 global->color_type = YV12;
853 else if (arg_match(&arg, &use_i420, argi))
854 global->color_type = I420;
855 else if (arg_match(&arg, &use_i422, argi))
856 global->color_type = I422;
857 else if (arg_match(&arg, &use_i444, argi))
858 global->color_type = I444;
859 else if (arg_match(&arg, &use_i440, argi))
860 global->color_type = I440;
861 else if (arg_match(&arg, &quietarg, argi))
863 else if (arg_match(&arg, &verbosearg, argi))
865 else if (arg_match(&arg, &limit, argi))
866 global->limit = arg_parse_uint(&arg);
867 else if (arg_match(&arg, &skip, argi))
868 global->skip_frames = arg_parse_uint(&arg);
869 else if (arg_match(&arg, &psnrarg, argi))
870 global->show_psnr = 1;
871 else if (arg_match(&arg, &recontest, argi))
872 global->test_decode = arg_parse_enum_or_int(&arg);
873 else if (arg_match(&arg, &framerate, argi)) {
874 global->framerate = arg_parse_rational(&arg);
875 validate_positive_rational(arg.name, &global->framerate);
876 global->have_framerate = 1;
877 } else if (arg_match(&arg, &out_part, argi))
878 global->out_part = 1;
879 else if (arg_match(&arg, &debugmode, argi))
881 else if (arg_match(&arg, &q_hist_n, argi))
882 global->show_q_hist_buckets = arg_parse_uint(&arg);
883 else if (arg_match(&arg, &rate_hist_n, argi))
884 global->show_rate_hist_buckets = arg_parse_uint(&arg);
885 else if (arg_match(&arg, &disable_warnings, argi))
886 global->disable_warnings = 1;
887 else if (arg_match(&arg, &disable_warning_prompt, argi))
888 global->disable_warning_prompt = 1;
894 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
895 if (global->pass > global->passes) {
896 warn("Assuming --pass=%d implies --passes=%d\n",
897 global->pass, global->pass);
898 global->passes = global->pass;
901 /* Validate global config */
902 if (global->passes == 0) {
903 #if CONFIG_VP9_ENCODER
904 // Make default VP9 passes = 2 until there is a better quality 1-pass
906 if (global->codec != NULL && global->codec->name != NULL)
907 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
908 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
914 if (global->deadline == VPX_DL_REALTIME &&
915 global->passes > 1) {
916 warn("Enforcing one-pass encoding in realtime mode\n");
922 void open_input_file(struct VpxInputContext *input) {
923 /* Parse certain options from the input file, if possible */
924 input->file = strcmp(input->filename, "-")
925 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
928 fatal("Failed to open input file");
930 if (!fseeko(input->file, 0, SEEK_END)) {
931 /* Input file is seekable. Figure out how long it is, so we can get
934 input->length = ftello(input->file);
938 /* For RAW input sources, these bytes will applied on the first frame
941 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
942 input->detect.position = 0;
944 if (input->detect.buf_read == 4
945 && file_is_y4m(input->detect.buf)) {
946 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
947 input->only_i420) >= 0) {
948 input->file_type = FILE_TYPE_Y4M;
949 input->width = input->y4m.pic_w;
950 input->height = input->y4m.pic_h;
951 input->framerate.numerator = input->y4m.fps_n;
952 input->framerate.denominator = input->y4m.fps_d;
953 input->fmt = input->y4m.vpx_fmt;
954 input->bit_depth = input->y4m.bit_depth;
956 fatal("Unsupported Y4M stream.");
957 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
958 fatal("IVF is not supported as input.");
960 input->file_type = FILE_TYPE_RAW;
965 static void close_input_file(struct VpxInputContext *input) {
967 if (input->file_type == FILE_TYPE_Y4M)
968 y4m_input_close(&input->y4m);
971 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
972 struct stream_state *prev) {
973 struct stream_state *stream;
975 stream = calloc(1, sizeof(*stream));
976 if (stream == NULL) {
977 fatal("Failed to allocate new stream.");
981 memcpy(stream, prev, sizeof(*stream));
987 /* Populate encoder configuration */
988 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
992 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
994 /* Change the default timebase to a high enough value so that the
995 * encoder will always create strictly increasing timestamps.
997 stream->config.cfg.g_timebase.den = 1000;
999 /* Never use the library's default resolution, require it be parsed
1000 * from the file or set on the command line.
1002 stream->config.cfg.g_w = 0;
1003 stream->config.cfg.g_h = 0;
1005 /* Initialize remaining stream parameters */
1006 stream->config.write_webm = 1;
1008 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1009 stream->ebml.last_pts_ns = -1;
1010 stream->ebml.writer = NULL;
1011 stream->ebml.segment = NULL;
1014 /* Allows removal of the application version from the EBML tags */
1015 stream->ebml.debug = global->debug;
1017 /* Default lag_in_frames is 0 in realtime mode */
1018 if (global->deadline == VPX_DL_REALTIME)
1019 stream->config.cfg.g_lag_in_frames = 0;
1022 /* Output files must be specified for each stream */
1023 stream->config.out_fn = NULL;
1025 stream->next = NULL;
1030 static int parse_stream_params(struct VpxEncoderConfig *global,
1031 struct stream_state *stream,
1033 char **argi, **argj;
1035 static const arg_def_t **ctrl_args = no_args;
1036 static const int *ctrl_args_map = NULL;
1037 struct stream_config *config = &stream->config;
1038 int eos_mark_found = 0;
1039 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1040 int test_16bit_internal = 0;
1043 // Handle codec specific options
1045 #if CONFIG_VP8_ENCODER
1046 } else if (strcmp(global->codec->name, "vp8") == 0) {
1047 ctrl_args = vp8_args;
1048 ctrl_args_map = vp8_arg_ctrl_map;
1050 #if CONFIG_VP9_ENCODER
1051 } else if (strcmp(global->codec->name, "vp9") == 0) {
1052 ctrl_args = vp9_args;
1053 ctrl_args_map = vp9_arg_ctrl_map;
1057 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1060 /* Once we've found an end-of-stream marker (--) we want to continue
1061 * shifting arguments but not consuming them.
1063 if (eos_mark_found) {
1066 } else if (!strcmp(*argj, "--")) {
1072 } else if (arg_match(&arg, &outputfile, argi)) {
1073 config->out_fn = arg.val;
1074 } else if (arg_match(&arg, &fpf_name, argi)) {
1075 config->stats_fn = arg.val;
1076 #if CONFIG_FP_MB_STATS
1077 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1078 config->fpmb_stats_fn = arg.val;
1080 } else if (arg_match(&arg, &use_webm, argi)) {
1082 config->write_webm = 1;
1084 die("Error: --webm specified but webm is disabled.");
1086 } else if (arg_match(&arg, &use_ivf, argi)) {
1087 config->write_webm = 0;
1088 } else if (arg_match(&arg, &threads, argi)) {
1089 config->cfg.g_threads = arg_parse_uint(&arg);
1090 } else if (arg_match(&arg, &profile, argi)) {
1091 config->cfg.g_profile = arg_parse_uint(&arg);
1092 } else if (arg_match(&arg, &width, argi)) {
1093 config->cfg.g_w = arg_parse_uint(&arg);
1094 } else if (arg_match(&arg, &height, argi)) {
1095 config->cfg.g_h = arg_parse_uint(&arg);
1096 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1097 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1098 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1099 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1100 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1103 } else if (arg_match(&arg, &stereo_mode, argi)) {
1104 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1106 } else if (arg_match(&arg, &timebase, argi)) {
1107 config->cfg.g_timebase = arg_parse_rational(&arg);
1108 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1109 } else if (arg_match(&arg, &error_resilient, argi)) {
1110 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1111 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1112 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1113 if (global->deadline == VPX_DL_REALTIME &&
1114 config->cfg.g_lag_in_frames != 0) {
1115 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1116 config->cfg.g_lag_in_frames = 0;
1118 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1119 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1120 } else if (arg_match(&arg, &resize_allowed, argi)) {
1121 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1122 } else if (arg_match(&arg, &resize_width, argi)) {
1123 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1124 } else if (arg_match(&arg, &resize_height, argi)) {
1125 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1126 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1127 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1128 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1129 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1130 } else if (arg_match(&arg, &end_usage, argi)) {
1131 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1132 } else if (arg_match(&arg, &target_bitrate, argi)) {
1133 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1134 } else if (arg_match(&arg, &min_quantizer, argi)) {
1135 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1136 } else if (arg_match(&arg, &max_quantizer, argi)) {
1137 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1138 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1139 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1140 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1141 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1142 } else if (arg_match(&arg, &buf_sz, argi)) {
1143 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1144 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1145 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1146 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1147 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1148 } else if (arg_match(&arg, &bias_pct, argi)) {
1149 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1150 if (global->passes < 2)
1151 warn("option %s ignored in one-pass mode.\n", arg.name);
1152 } else if (arg_match(&arg, &minsection_pct, argi)) {
1153 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1155 if (global->passes < 2)
1156 warn("option %s ignored in one-pass mode.\n", arg.name);
1157 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1158 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1160 if (global->passes < 2)
1161 warn("option %s ignored in one-pass mode.\n", arg.name);
1162 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1163 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1164 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1165 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1166 config->have_kf_max_dist = 1;
1167 } else if (arg_match(&arg, &kf_disabled, argi)) {
1168 config->cfg.kf_mode = VPX_KF_DISABLED;
1169 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1170 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1171 if (strcmp(global->codec->name, "vp9") == 0) {
1172 test_16bit_internal = 1;
1177 for (i = 0; ctrl_args[i]; i++) {
1178 if (arg_match(&arg, ctrl_args[i], argi)) {
1182 /* Point either to the next free element or the first
1183 * instance of this control.
1185 for (j = 0; j < config->arg_ctrl_cnt; j++)
1186 if (ctrl_args_map != NULL &&
1187 config->arg_ctrls[j][0] == ctrl_args_map[i])
1191 assert(j < (int)ARG_CTRL_CNT_MAX);
1192 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1193 config->arg_ctrls[j][0] = ctrl_args_map[i];
1194 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1195 if (j == config->arg_ctrl_cnt)
1196 config->arg_ctrl_cnt++;
1204 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1205 if (strcmp(global->codec->name, "vp9") == 0) {
1206 config->use_16bit_internal = test_16bit_internal |
1207 (config->cfg.g_profile > 1);
1210 return eos_mark_found;
1214 #define FOREACH_STREAM(func) \
1216 struct stream_state *stream; \
1217 for (stream = streams; stream; stream = stream->next) { \
1223 static void validate_stream_config(const struct stream_state *stream,
1224 const struct VpxEncoderConfig *global) {
1225 const struct stream_state *streami;
1228 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1229 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1230 " and --height (-h)", stream->index);
1232 // Check that the codec bit depth is greater than the input bit depth.
1233 if (stream->config.cfg.g_input_bit_depth >
1234 (unsigned int)stream->config.cfg.g_bit_depth) {
1235 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1236 stream->index, (int)stream->config.cfg.g_bit_depth,
1237 stream->config.cfg.g_input_bit_depth);
1240 for (streami = stream; streami; streami = streami->next) {
1241 /* All streams require output files */
1242 if (!streami->config.out_fn)
1243 fatal("Stream %d: Output file is required (specify with -o)",
1246 /* Check for two streams outputting to the same file */
1247 if (streami != stream) {
1248 const char *a = stream->config.out_fn;
1249 const char *b = streami->config.out_fn;
1250 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1251 fatal("Stream %d: duplicate output file (from stream %d)",
1252 streami->index, stream->index);
1255 /* Check for two streams sharing a stats file. */
1256 if (streami != stream) {
1257 const char *a = stream->config.stats_fn;
1258 const char *b = streami->config.stats_fn;
1259 if (a && b && !strcmp(a, b))
1260 fatal("Stream %d: duplicate stats file (from stream %d)",
1261 streami->index, stream->index);
1264 #if CONFIG_FP_MB_STATS
1265 /* Check for two streams sharing a mb stats file. */
1266 if (streami != stream) {
1267 const char *a = stream->config.fpmb_stats_fn;
1268 const char *b = streami->config.fpmb_stats_fn;
1269 if (a && b && !strcmp(a, b))
1270 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1271 streami->index, stream->index);
1278 static void set_stream_dimensions(struct stream_state *stream,
1281 if (!stream->config.cfg.g_w) {
1282 if (!stream->config.cfg.g_h)
1283 stream->config.cfg.g_w = w;
1285 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1287 if (!stream->config.cfg.g_h) {
1288 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1293 static void set_default_kf_interval(struct stream_state *stream,
1294 struct VpxEncoderConfig *global) {
1295 /* Use a max keyframe interval of 5 seconds, if none was
1296 * specified on the command line.
1298 if (!stream->config.have_kf_max_dist) {
1299 double framerate = (double)global->framerate.num / global->framerate.den;
1300 if (framerate > 0.0)
1301 stream->config.cfg.kf_max_dist = (unsigned int)(5.0 * framerate);
1305 static const char* file_type_to_string(enum VideoFileType t) {
1307 case FILE_TYPE_RAW: return "RAW";
1308 case FILE_TYPE_Y4M: return "Y4M";
1309 default: return "Other";
1313 static const char* image_format_to_string(vpx_img_fmt_t f) {
1315 case VPX_IMG_FMT_I420: return "I420";
1316 case VPX_IMG_FMT_I422: return "I422";
1317 case VPX_IMG_FMT_I444: return "I444";
1318 case VPX_IMG_FMT_I440: return "I440";
1319 case VPX_IMG_FMT_YV12: return "YV12";
1320 case VPX_IMG_FMT_I42016: return "I42016";
1321 case VPX_IMG_FMT_I42216: return "I42216";
1322 case VPX_IMG_FMT_I44416: return "I44416";
1323 case VPX_IMG_FMT_I44016: return "I44016";
1324 default: return "Other";
1328 static void show_stream_config(struct stream_state *stream,
1329 struct VpxEncoderConfig *global,
1330 struct VpxInputContext *input) {
1332 #define SHOW(field) \
1333 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1335 if (stream->index == 0) {
1336 fprintf(stderr, "Codec: %s\n",
1337 vpx_codec_iface_name(global->codec->codec_interface()));
1338 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1340 file_type_to_string(input->file_type),
1341 image_format_to_string(input->fmt));
1343 if (stream->next || stream->index)
1344 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1345 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1346 fprintf(stderr, "Encoder parameters:\n");
1354 SHOW(g_input_bit_depth);
1355 SHOW(g_timebase.num);
1356 SHOW(g_timebase.den);
1357 SHOW(g_error_resilient);
1359 SHOW(g_lag_in_frames);
1360 SHOW(rc_dropframe_thresh);
1361 SHOW(rc_resize_allowed);
1362 SHOW(rc_scaled_width);
1363 SHOW(rc_scaled_height);
1364 SHOW(rc_resize_up_thresh);
1365 SHOW(rc_resize_down_thresh);
1367 SHOW(rc_target_bitrate);
1368 SHOW(rc_min_quantizer);
1369 SHOW(rc_max_quantizer);
1370 SHOW(rc_undershoot_pct);
1371 SHOW(rc_overshoot_pct);
1373 SHOW(rc_buf_initial_sz);
1374 SHOW(rc_buf_optimal_sz);
1375 SHOW(rc_2pass_vbr_bias_pct);
1376 SHOW(rc_2pass_vbr_minsection_pct);
1377 SHOW(rc_2pass_vbr_maxsection_pct);
1384 static void open_output_file(struct stream_state *stream,
1385 struct VpxEncoderConfig *global) {
1386 const char *fn = stream->config.out_fn;
1387 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1389 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1392 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1395 fatal("Failed to open output file");
1397 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1398 fatal("WebM output to pipes not supported.");
1401 if (stream->config.write_webm) {
1402 stream->ebml.stream = stream->file;
1403 write_webm_file_header(&stream->ebml, cfg,
1405 stream->config.stereo_fmt,
1406 global->codec->fourcc);
1410 if (!stream->config.write_webm) {
1411 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1416 static void close_output_file(struct stream_state *stream,
1417 unsigned int fourcc) {
1418 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1420 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1424 if (stream->config.write_webm) {
1425 write_webm_file_footer(&stream->ebml);
1429 if (!stream->config.write_webm) {
1430 if (!fseek(stream->file, 0, SEEK_SET))
1431 ivf_write_file_header(stream->file, &stream->config.cfg,
1433 stream->frames_out);
1436 fclose(stream->file);
1440 static void setup_pass(struct stream_state *stream,
1441 struct VpxEncoderConfig *global,
1443 if (stream->config.stats_fn) {
1444 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1446 fatal("Failed to open statistics store");
1448 if (!stats_open_mem(&stream->stats, pass))
1449 fatal("Failed to open statistics store");
1452 #if CONFIG_FP_MB_STATS
1453 if (stream->config.fpmb_stats_fn) {
1454 if (!stats_open_file(&stream->fpmb_stats,
1455 stream->config.fpmb_stats_fn, pass))
1456 fatal("Failed to open mb statistics store");
1458 if (!stats_open_mem(&stream->fpmb_stats, pass))
1459 fatal("Failed to open mb statistics store");
1463 stream->config.cfg.g_pass = global->passes == 2
1464 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1467 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1468 #if CONFIG_FP_MB_STATS
1469 stream->config.cfg.rc_firstpass_mb_stats_in =
1470 stats_get(&stream->fpmb_stats);
1474 stream->cx_time = 0;
1476 stream->frames_out = 0;
1480 static void initialize_encoder(struct stream_state *stream,
1481 struct VpxEncoderConfig *global) {
1485 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1486 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1487 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1488 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1491 /* Construct Encoder Context */
1492 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1493 &stream->config.cfg, flags);
1494 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1496 /* Note that we bypass the vpx_codec_control wrapper macro because
1497 * we're being clever to store the control IDs in an array. Real
1498 * applications will want to make use of the enumerations directly
1500 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1501 int ctrl = stream->config.arg_ctrls[i][0];
1502 int value = stream->config.arg_ctrls[i][1];
1503 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1504 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1507 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1511 if (global->test_decode != TEST_DECODE_OFF) {
1512 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1513 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1519 static void encode_frame(struct stream_state *stream,
1520 struct VpxEncoderConfig *global,
1521 struct vpx_image *img,
1522 unsigned int frames_in) {
1523 vpx_codec_pts_t frame_start, next_frame_start;
1524 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1525 struct vpx_usec_timer timer;
1527 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1528 * global->framerate.den)
1529 / cfg->g_timebase.num / global->framerate.num;
1530 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1531 * global->framerate.den)
1532 / cfg->g_timebase.num / global->framerate.num;
1534 /* Scale if necessary */
1535 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1537 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1538 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1539 if (img->fmt != VPX_IMG_FMT_I42016) {
1540 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1545 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1546 cfg->g_w, cfg->g_h, 16);
1548 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1549 img->stride[VPX_PLANE_Y]/2,
1550 (uint16*)img->planes[VPX_PLANE_U],
1551 img->stride[VPX_PLANE_U]/2,
1552 (uint16*)img->planes[VPX_PLANE_V],
1553 img->stride[VPX_PLANE_V]/2,
1555 (uint16*)stream->img->planes[VPX_PLANE_Y],
1556 stream->img->stride[VPX_PLANE_Y]/2,
1557 (uint16*)stream->img->planes[VPX_PLANE_U],
1558 stream->img->stride[VPX_PLANE_U]/2,
1559 (uint16*)stream->img->planes[VPX_PLANE_V],
1560 stream->img->stride[VPX_PLANE_V]/2,
1561 stream->img->d_w, stream->img->d_h,
1565 stream->encoder.err = 1;
1566 ctx_exit_on_error(&stream->encoder,
1567 "Stream %d: Failed to encode frame.\n"
1568 "Scaling disabled in this configuration. \n"
1569 "To enable, configure with --enable-libyuv\n",
1575 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1576 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1577 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1582 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1583 cfg->g_w, cfg->g_h, 16);
1584 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1585 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1586 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1588 stream->img->planes[VPX_PLANE_Y],
1589 stream->img->stride[VPX_PLANE_Y],
1590 stream->img->planes[VPX_PLANE_U],
1591 stream->img->stride[VPX_PLANE_U],
1592 stream->img->planes[VPX_PLANE_V],
1593 stream->img->stride[VPX_PLANE_V],
1594 stream->img->d_w, stream->img->d_h,
1598 stream->encoder.err = 1;
1599 ctx_exit_on_error(&stream->encoder,
1600 "Stream %d: Failed to encode frame.\n"
1601 "Scaling disabled in this configuration. \n"
1602 "To enable, configure with --enable-libyuv\n",
1607 vpx_usec_timer_start(&timer);
1608 vpx_codec_encode(&stream->encoder, img, frame_start,
1609 (unsigned long)(next_frame_start - frame_start),
1610 0, global->deadline);
1611 vpx_usec_timer_mark(&timer);
1612 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1613 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1618 static void update_quantizer_histogram(struct stream_state *stream) {
1619 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1622 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1623 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1624 stream->counts[q]++;
1629 static void get_cx_data(struct stream_state *stream,
1630 struct VpxEncoderConfig *global,
1632 const vpx_codec_cx_pkt_t *pkt;
1633 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1634 vpx_codec_iter_t iter = NULL;
1637 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1638 static size_t fsize = 0;
1639 static int64_t ivf_header_pos = 0;
1641 switch (pkt->kind) {
1642 case VPX_CODEC_CX_FRAME_PKT:
1643 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1644 stream->frames_out++;
1647 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1649 update_rate_histogram(stream->rate_hist, cfg, pkt);
1651 if (stream->config.write_webm) {
1652 write_webm_block(&stream->ebml, cfg, pkt);
1655 if (!stream->config.write_webm) {
1656 if (pkt->data.frame.partition_id <= 0) {
1657 ivf_header_pos = ftello(stream->file);
1658 fsize = pkt->data.frame.sz;
1660 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1662 fsize += pkt->data.frame.sz;
1664 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1665 const int64_t currpos = ftello(stream->file);
1666 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1667 ivf_write_frame_size(stream->file, fsize);
1668 fseeko(stream->file, currpos, SEEK_SET);
1672 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1675 stream->nbytes += pkt->data.raw.sz;
1679 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1680 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1681 (unsigned int)pkt->data.frame.sz, NULL, 0);
1682 if (stream->decoder.err) {
1683 warn_or_exit_on_error(&stream->decoder,
1684 global->test_decode == TEST_DECODE_FATAL,
1685 "Failed to decode frame %d in stream %d",
1686 stream->frames_out + 1, stream->index);
1687 stream->mismatch_seen = stream->frames_out + 1;
1692 case VPX_CODEC_STATS_PKT:
1693 stream->frames_out++;
1694 stats_write(&stream->stats,
1695 pkt->data.twopass_stats.buf,
1696 pkt->data.twopass_stats.sz);
1697 stream->nbytes += pkt->data.raw.sz;
1699 #if CONFIG_FP_MB_STATS
1700 case VPX_CODEC_FPMB_STATS_PKT:
1701 stats_write(&stream->fpmb_stats,
1702 pkt->data.firstpass_mb_stats.buf,
1703 pkt->data.firstpass_mb_stats.sz);
1704 stream->nbytes += pkt->data.raw.sz;
1707 case VPX_CODEC_PSNR_PKT:
1709 if (global->show_psnr) {
1712 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1713 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1714 for (i = 0; i < 4; i++) {
1716 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1717 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1719 stream->psnr_count++;
1730 static void show_psnr(struct stream_state *stream, double peak) {
1734 if (!stream->psnr_count)
1737 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1738 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1739 (double)stream->psnr_sse_total);
1740 fprintf(stderr, " %.3f", ovpsnr);
1742 for (i = 0; i < 4; i++) {
1743 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1745 fprintf(stderr, "\n");
1749 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1750 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1753 static void test_decode(struct stream_state *stream,
1754 enum TestDecodeFatality fatal,
1755 const VpxInterface *codec) {
1756 vpx_image_t enc_img, dec_img;
1758 if (stream->mismatch_seen)
1761 /* Get the internal reference frame */
1762 if (strcmp(codec->name, "vp8") == 0) {
1763 struct vpx_ref_frame ref_enc, ref_dec;
1766 width = (stream->config.cfg.g_w + 15) & ~15;
1767 height = (stream->config.cfg.g_h + 15) & ~15;
1768 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1769 enc_img = ref_enc.img;
1770 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1771 dec_img = ref_dec.img;
1773 ref_enc.frame_type = VP8_LAST_FRAME;
1774 ref_dec.frame_type = VP8_LAST_FRAME;
1775 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1776 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1778 struct vp9_ref_frame ref_enc, ref_dec;
1782 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1783 enc_img = ref_enc.img;
1784 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1785 dec_img = ref_dec.img;
1786 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1787 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1788 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1789 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1790 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1791 enc_img.d_w, enc_img.d_h, 16);
1792 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1794 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1795 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1796 dec_img.d_w, dec_img.d_h, 16);
1797 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1802 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1803 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1805 if (!compare_img(&enc_img, &dec_img)) {
1806 int y[4], u[4], v[4];
1807 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1808 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1809 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1811 find_mismatch(&enc_img, &dec_img, y, u, v);
1814 find_mismatch(&enc_img, &dec_img, y, u, v);
1816 stream->decoder.err = 1;
1817 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1818 "Stream %d: Encode/decode mismatch on frame %d at"
1819 " Y[%d, %d] {%d/%d},"
1820 " U[%d, %d] {%d/%d},"
1821 " V[%d, %d] {%d/%d}",
1822 stream->index, stream->frames_out,
1823 y[0], y[1], y[2], y[3],
1824 u[0], u[1], u[2], u[3],
1825 v[0], v[1], v[2], v[3]);
1826 stream->mismatch_seen = stream->frames_out;
1829 vpx_img_free(&enc_img);
1830 vpx_img_free(&dec_img);
1834 static void print_time(const char *label, int64_t etl) {
1841 etl -= hours * 3600;
1846 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1847 label, hours, mins, secs);
1849 fprintf(stderr, "[%3s unknown] ", label);
1854 int main(int argc, const char **argv_) {
1857 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
1858 vpx_image_t raw_shift;
1859 int allocated_raw_shift = 0;
1860 int use_16bit_internal = 0;
1861 int input_shift = 0;
1863 int frame_avail, got_data;
1865 struct VpxInputContext input;
1866 struct VpxEncoderConfig global;
1867 struct stream_state *streams = NULL;
1868 char **argv, **argi;
1869 uint64_t cx_time = 0;
1873 memset(&input, 0, sizeof(input));
1874 exec_name = argv_[0];
1879 /* Setup default input stream settings */
1880 input.framerate.numerator = 30;
1881 input.framerate.denominator = 1;
1882 input.only_i420 = 1;
1883 input.bit_depth = 0;
1885 /* First parse the global configuration values, because we want to apply
1886 * other parameters on top of the default configuration provided by the
1889 argv = argv_dup(argc - 1, argv_ + 1);
1890 parse_global_config(&global, argv);
1892 switch (global.color_type) {
1894 input.fmt = VPX_IMG_FMT_I420;
1897 input.fmt = VPX_IMG_FMT_I422;
1900 input.fmt = VPX_IMG_FMT_I444;
1903 input.fmt = VPX_IMG_FMT_I440;
1906 input.fmt = VPX_IMG_FMT_YV12;
1911 /* Now parse each stream's parameters. Using a local scope here
1912 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1915 struct stream_state *stream = NULL;
1918 stream = new_stream(&global, stream);
1922 } while (parse_stream_params(&global, stream, argv));
1925 /* Check for unrecognized options */
1926 for (argi = argv; *argi; argi++)
1927 if (argi[0][0] == '-' && argi[0][1])
1928 die("Error: Unrecognized option %s\n", *argi);
1930 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1931 &global, &stream->config.cfg););
1933 /* Handle non-option arguments */
1934 input.filename = argv[0];
1936 if (!input.filename)
1939 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1940 if (global.codec->fourcc == VP9_FOURCC)
1941 input.only_i420 = 0;
1943 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1944 int frames_in = 0, seen_frames = 0;
1945 int64_t estimated_time_left = -1;
1946 int64_t average_rate = -1;
1947 int64_t lagged_count = 0;
1949 open_input_file(&input);
1951 /* If the input file doesn't specify its w/h (raw files), try to get
1952 * the data from the first stream's configuration.
1954 if (!input.width || !input.height) {
1956 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1957 input.width = stream->config.cfg.g_w;
1958 input.height = stream->config.cfg.g_h;
1964 /* Update stream configurations from the input file's parameters */
1965 if (!input.width || !input.height)
1966 fatal("Specify stream dimensions with --width (-w) "
1967 " and --height (-h)");
1969 /* If input file does not specify bit-depth but input-bit-depth parameter
1970 * exists, assume that to be the input bit-depth. However, if the
1971 * input-bit-depth paramter does not exist, assume the input bit-depth
1972 * to be the same as the codec bit-depth.
1974 if (!input.bit_depth) {
1976 if (stream->config.cfg.g_input_bit_depth)
1977 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1979 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1980 (int)stream->config.cfg.g_bit_depth;
1982 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1985 stream->config.cfg.g_input_bit_depth = input.bit_depth;
1989 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
1990 FOREACH_STREAM(validate_stream_config(stream, &global));
1992 /* Ensure that --passes and --pass are consistent. If --pass is set and
1993 * --passes=2, ensure --fpf was set.
1995 if (global.pass && global.passes == 2)
1997 if (!stream->config.stats_fn)
1998 die("Stream %d: Must specify --fpf when --pass=%d"
1999 " and --passes=2\n", stream->index, global.pass);
2004 stream->config.write_webm = 0;
2005 warn("vpxenc was compiled without WebM container support."
2006 "Producing IVF output");
2010 /* Use the frame rate from the file only if none was specified
2011 * on the command-line.
2013 if (!global.have_framerate) {
2014 global.framerate.num = input.framerate.numerator;
2015 global.framerate.den = input.framerate.denominator;
2018 FOREACH_STREAM(set_default_kf_interval(stream, &global));
2020 /* Show configuration */
2021 if (global.verbose && pass == 0)
2022 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2024 if (pass == (global.pass ? global.pass - 1 : 0)) {
2025 if (input.file_type == FILE_TYPE_Y4M)
2026 /*The Y4M reader does its own allocation.
2027 Just initialize this here to avoid problems if we never read any
2029 memset(&raw, 0, sizeof(raw));
2031 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2033 FOREACH_STREAM(stream->rate_hist =
2034 init_rate_histogram(&stream->config.cfg,
2035 &global.framerate));
2038 FOREACH_STREAM(setup_pass(stream, &global, pass));
2039 FOREACH_STREAM(open_output_file(stream, &global));
2040 FOREACH_STREAM(initialize_encoder(stream, &global));
2042 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2043 if (strcmp(global.codec->name, "vp9") == 0) {
2044 // Check to see if at least one stream uses 16 bit internal.
2045 // Currently assume that the bit_depths for all streams using
2046 // highbitdepth are the same.
2048 if (stream->config.use_16bit_internal) {
2049 use_16bit_internal = 1;
2051 if (stream->config.cfg.g_profile == 0) {
2054 input_shift = (int)stream->config.cfg.g_bit_depth -
2055 stream->config.cfg.g_input_bit_depth;
2064 while (frame_avail || got_data) {
2065 struct vpx_usec_timer timer;
2067 if (!global.limit || frames_in < global.limit) {
2068 frame_avail = read_frame(&input, &raw);
2072 seen_frames = frames_in > global.skip_frames ?
2073 frames_in - global.skip_frames : 0;
2075 if (!global.quiet) {
2076 float fps = usec_to_fps(cx_time, seen_frames);
2077 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2079 if (stream_cnt == 1)
2081 "frame %4d/%-4d %7"PRId64"B ",
2082 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2084 fprintf(stderr, "frame %4d ", frames_in);
2086 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2087 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2088 cx_time > 9999999 ? "ms" : "us",
2089 fps >= 1.0 ? fps : fps * 60,
2090 fps >= 1.0 ? "fps" : "fpm");
2091 print_time("ETA", estimated_time_left);
2097 if (frames_in > global.skip_frames) {
2098 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2099 vpx_image_t *frame_to_encode;
2100 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2101 assert(use_16bit_internal);
2102 // Input bit depth and stream bit depth do not match, so up
2103 // shift frame to stream bit depth
2104 if (!allocated_raw_shift) {
2105 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2106 input.width, input.height, 32);
2107 allocated_raw_shift = 1;
2109 vpx_img_upshift(&raw_shift, &raw, input_shift);
2110 frame_to_encode = &raw_shift;
2112 frame_to_encode = &raw;
2114 vpx_usec_timer_start(&timer);
2115 if (use_16bit_internal) {
2116 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2118 if (stream->config.use_16bit_internal)
2119 encode_frame(stream, &global,
2120 frame_avail ? frame_to_encode : NULL,
2126 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2127 FOREACH_STREAM(encode_frame(stream, &global,
2128 frame_avail ? frame_to_encode : NULL,
2132 vpx_usec_timer_start(&timer);
2133 FOREACH_STREAM(encode_frame(stream, &global,
2134 frame_avail ? &raw : NULL,
2137 vpx_usec_timer_mark(&timer);
2138 cx_time += vpx_usec_timer_elapsed(&timer);
2140 FOREACH_STREAM(update_quantizer_histogram(stream));
2143 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2145 if (!got_data && input.length && streams != NULL &&
2146 !streams->frames_out) {
2147 lagged_count = global.limit ? seen_frames : ftello(input.file);
2148 } else if (input.length) {
2153 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2155 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2156 remaining = 1000 * (global.limit - global.skip_frames
2157 - seen_frames + lagged_count);
2159 const int64_t input_pos = ftello(input.file);
2160 const int64_t input_pos_lagged = input_pos - lagged_count;
2161 const int64_t limit = input.length;
2163 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2164 remaining = limit - input_pos + lagged_count;
2167 average_rate = (average_rate <= 0)
2169 : (average_rate * 7 + rate) / 8;
2170 estimated_time_left = average_rate ? remaining / average_rate : -1;
2173 if (got_data && global.test_decode != TEST_DECODE_OFF)
2174 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2179 fprintf(stderr, "\033[K");
2183 fprintf(stderr, "\n");
2185 if (!global.quiet) {
2186 FOREACH_STREAM(fprintf(stderr,
2187 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2188 " %7"PRId64" %s (%.2f fps)\033[K\n",
2190 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2191 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2192 seen_frames ? (int64_t)stream->nbytes * 8 *
2193 (int64_t)global.framerate.num / global.framerate.den /
2195 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2196 stream->cx_time > 9999999 ? "ms" : "us",
2197 usec_to_fps(stream->cx_time, seen_frames)));
2200 if (global.show_psnr) {
2201 if (global.codec->fourcc == VP9_FOURCC) {
2203 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2205 FOREACH_STREAM(show_psnr(stream, 255.0));
2209 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2211 if (global.test_decode != TEST_DECODE_OFF) {
2212 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2215 close_input_file(&input);
2217 if (global.test_decode == TEST_DECODE_FATAL) {
2218 FOREACH_STREAM(res |= stream->mismatch_seen);
2220 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2222 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2224 #if CONFIG_FP_MB_STATS
2225 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2232 if (global.show_q_hist_buckets)
2233 FOREACH_STREAM(show_q_histogram(stream->counts,
2234 global.show_q_hist_buckets));
2236 if (global.show_rate_hist_buckets)
2237 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2238 &stream->config.cfg,
2239 global.show_rate_hist_buckets));
2240 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2242 #if CONFIG_INTERNAL_STATS
2243 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2244 * to match some existing utilities.
2246 if (!(global.pass == 1 && global.passes == 2))
2248 FILE *f = fopen("opsnr.stt", "a");
2249 if (stream->mismatch_seen) {
2250 fprintf(f, "First mismatch occurred in frame %d\n",
2251 stream->mismatch_seen);
2253 fprintf(f, "No mismatch detected in recon buffers\n");
2259 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
2260 if (allocated_raw_shift)
2261 vpx_img_free(&raw_shift);
2266 return res ? EXIT_FAILURE : EXIT_SUCCESS;