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 static 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 static int file_is_y4m(const char detect[4]) {
118 if (memcmp(detect, "YUV4", 4) == 0) {
124 static 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_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, &recontest,
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_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 auto_altref = ARG_DEF(
334 NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
335 static const arg_def_t arnr_maxframes = ARG_DEF(
336 NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
337 static const arg_def_t arnr_strength = ARG_DEF(
338 NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
339 static const arg_def_t arnr_type = ARG_DEF(
340 NULL, "arnr-type", 1, "AltRef type");
341 static const struct arg_enum_list tuning_enum[] = {
342 {"psnr", VP8_TUNE_PSNR},
343 {"ssim", VP8_TUNE_SSIM},
346 static const arg_def_t tune_ssim = ARG_DEF_ENUM(
347 NULL, "tune", 1, "Material to favor", tuning_enum);
348 static const arg_def_t cq_level = ARG_DEF(
349 NULL, "cq-level", 1, "Constant/Constrained Quality level");
350 static const arg_def_t max_intra_rate_pct = ARG_DEF(
351 NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
353 #if CONFIG_VP8_ENCODER
354 static const arg_def_t cpu_used_vp8 = ARG_DEF(
355 NULL, "cpu-used", 1, "CPU Used (-16..16)");
356 static const arg_def_t token_parts = ARG_DEF(
357 NULL, "token-parts", 1, "Number of token partitions to use, log2");
358 static const arg_def_t screen_content_mode = ARG_DEF(
359 NULL, "screen-content-mode", 1, "Screen content mode");
360 static const arg_def_t *vp8_args[] = {
361 &cpu_used_vp8, &auto_altref, &noise_sens, &sharpness, &static_thresh,
362 &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type,
363 &tune_ssim, &cq_level, &max_intra_rate_pct, &screen_content_mode,
366 static const int vp8_arg_ctrl_map[] = {
367 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
368 VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
369 VP8E_SET_TOKEN_PARTITIONS,
370 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
371 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
372 VP8E_SET_SCREEN_CONTENT_MODE,
377 #if CONFIG_VP9_ENCODER
378 static const arg_def_t cpu_used_vp9 = ARG_DEF(
379 NULL, "cpu-used", 1, "CPU Used (-8..8)");
380 static const arg_def_t tile_cols = ARG_DEF(
381 NULL, "tile-columns", 1, "Number of tile columns to use, log2");
382 static const arg_def_t tile_rows = ARG_DEF(
383 NULL, "tile-rows", 1,
384 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
385 static const arg_def_t lossless = ARG_DEF(
386 NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
387 static const arg_def_t frame_parallel_decoding = ARG_DEF(
388 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
389 static const arg_def_t aq_mode = ARG_DEF(
391 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
392 "3: cyclic refresh, 4: equator360)");
393 static const arg_def_t frame_periodic_boost = ARG_DEF(
394 NULL, "frame-boost", 1,
395 "Enable frame periodic boost (0: off (default), 1: on)");
396 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
397 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
398 static const arg_def_t max_inter_rate_pct = ARG_DEF(
399 NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
400 static const arg_def_t min_gf_interval = ARG_DEF(
401 NULL, "min-gf-interval", 1,
402 "min gf/arf frame interval (default 0, indicating in-built behavior)");
403 static const arg_def_t max_gf_interval = ARG_DEF(
404 NULL, "max-gf-interval", 1,
405 "max gf/arf frame interval (default 0, indicating in-built behavior)");
407 static const struct arg_enum_list color_space_enum[] = {
408 { "unknown", VPX_CS_UNKNOWN },
409 { "bt601", VPX_CS_BT_601 },
410 { "bt709", VPX_CS_BT_709 },
411 { "smpte170", VPX_CS_SMPTE_170 },
412 { "smpte240", VPX_CS_SMPTE_240 },
413 { "bt2020", VPX_CS_BT_2020 },
414 { "reserved", VPX_CS_RESERVED },
415 { "sRGB", VPX_CS_SRGB },
419 static const arg_def_t input_color_space = ARG_DEF_ENUM(
420 NULL, "color-space", 1,
421 "The color space of input content:", color_space_enum);
423 #if CONFIG_VP9_HIGHBITDEPTH
424 static const struct arg_enum_list bitdepth_enum[] = {
431 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
433 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
435 static const arg_def_t inbitdeptharg = ARG_DEF(
436 NULL, "input-bit-depth", 1, "Bit depth of input");
439 static const struct arg_enum_list tune_content_enum[] = {
440 {"default", VP9E_CONTENT_DEFAULT},
441 {"screen", VP9E_CONTENT_SCREEN},
445 static const arg_def_t tune_content = ARG_DEF_ENUM(
446 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
448 static const arg_def_t target_level = ARG_DEF(
449 NULL, "target-level", 1,
450 "Target level (255: off (default); 0: only keep level stats; 10: level 1.0;"
451 " 11: level 1.1; ... 62: level 6.2)");
454 #if CONFIG_VP9_ENCODER
455 static const arg_def_t *vp9_args[] = {
456 &cpu_used_vp9, &auto_altref, &sharpness, &static_thresh,
457 &tile_cols, &tile_rows, &arnr_maxframes, &arnr_strength, &arnr_type,
458 &tune_ssim, &cq_level, &max_intra_rate_pct, &max_inter_rate_pct,
459 &gf_cbr_boost_pct, &lossless,
460 &frame_parallel_decoding, &aq_mode, &frame_periodic_boost,
461 &noise_sens, &tune_content, &input_color_space,
462 &min_gf_interval, &max_gf_interval, &target_level,
463 #if CONFIG_VP9_HIGHBITDEPTH
464 &bitdeptharg, &inbitdeptharg,
465 #endif // CONFIG_VP9_HIGHBITDEPTH
468 static const int vp9_arg_ctrl_map[] = {
469 VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
470 VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
471 VP9E_SET_TILE_COLUMNS, VP9E_SET_TILE_ROWS,
472 VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH, VP8E_SET_ARNR_TYPE,
473 VP8E_SET_TUNING, VP8E_SET_CQ_LEVEL, VP8E_SET_MAX_INTRA_BITRATE_PCT,
474 VP9E_SET_MAX_INTER_BITRATE_PCT, VP9E_SET_GF_CBR_BOOST_PCT,
475 VP9E_SET_LOSSLESS, VP9E_SET_FRAME_PARALLEL_DECODING, VP9E_SET_AQ_MODE,
476 VP9E_SET_FRAME_PERIODIC_BOOST, VP9E_SET_NOISE_SENSITIVITY,
477 VP9E_SET_TUNE_CONTENT, VP9E_SET_COLOR_SPACE,
478 VP9E_SET_MIN_GF_INTERVAL, VP9E_SET_MAX_GF_INTERVAL, VP9E_SET_TARGET_LEVEL,
483 static const arg_def_t *no_args[] = { NULL };
485 void usage_exit(void) {
487 const int num_encoder = get_vpx_encoder_count();
489 fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
492 fprintf(stderr, "\nOptions:\n");
493 arg_show_usage(stderr, main_args);
494 fprintf(stderr, "\nEncoder Global Options:\n");
495 arg_show_usage(stderr, global_args);
496 fprintf(stderr, "\nRate Control Options:\n");
497 arg_show_usage(stderr, rc_args);
498 fprintf(stderr, "\nTwopass Rate Control Options:\n");
499 arg_show_usage(stderr, rc_twopass_args);
500 fprintf(stderr, "\nKeyframe Placement Options:\n");
501 arg_show_usage(stderr, kf_args);
502 #if CONFIG_VP8_ENCODER
503 fprintf(stderr, "\nVP8 Specific Options:\n");
504 arg_show_usage(stderr, vp8_args);
506 #if CONFIG_VP9_ENCODER
507 fprintf(stderr, "\nVP9 Specific Options:\n");
508 arg_show_usage(stderr, vp9_args);
510 fprintf(stderr, "\nStream timebase (--timebase):\n"
511 " The desired precision of timestamps in the output, expressed\n"
512 " in fractional seconds. Default is 1/1000.\n");
513 fprintf(stderr, "\nIncluded encoders:\n\n");
515 for (i = 0; i < num_encoder; ++i) {
516 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
517 const char* defstr = (i == (num_encoder - 1)) ? "(default)" : "";
518 fprintf(stderr, " %-6s - %s %s\n",
519 encoder->name, vpx_codec_iface_name(encoder->codec_interface()),
522 fprintf(stderr, "\n ");
523 fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");
528 #define mmin(a, b) ((a) < (b) ? (a) : (b))
530 #if CONFIG_VP9_HIGHBITDEPTH
531 static void find_mismatch_high(const vpx_image_t *const img1,
532 const vpx_image_t *const img2,
533 int yloc[4], int uloc[4], int vloc[4]) {
534 uint16_t *plane1, *plane2;
535 uint32_t stride1, stride2;
536 const uint32_t bsize = 64;
537 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
538 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
540 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
542 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
545 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
546 plane1 = (uint16_t*)img1->planes[VPX_PLANE_Y];
547 plane2 = (uint16_t*)img2->planes[VPX_PLANE_Y];
548 stride1 = img1->stride[VPX_PLANE_Y]/2;
549 stride2 = img2->stride[VPX_PLANE_Y]/2;
550 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
551 for (j = 0; match && j < img1->d_w; j += bsize) {
553 const int si = mmin(i + bsize, img1->d_h) - i;
554 const int sj = mmin(j + bsize, img1->d_w) - j;
555 for (k = 0; match && k < si; ++k) {
556 for (l = 0; match && l < sj; ++l) {
557 if (*(plane1 + (i + k) * stride1 + j + l) !=
558 *(plane2 + (i + k) * stride2 + j + l)) {
561 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
562 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
571 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
572 plane1 = (uint16_t*)img1->planes[VPX_PLANE_U];
573 plane2 = (uint16_t*)img2->planes[VPX_PLANE_U];
574 stride1 = img1->stride[VPX_PLANE_U]/2;
575 stride2 = img2->stride[VPX_PLANE_U]/2;
576 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
577 for (j = 0; match && j < c_w; j += bsizex) {
579 const int si = mmin(i + bsizey, c_h - i);
580 const int sj = mmin(j + bsizex, c_w - j);
581 for (k = 0; match && k < si; ++k) {
582 for (l = 0; match && l < sj; ++l) {
583 if (*(plane1 + (i + k) * stride1 + j + l) !=
584 *(plane2 + (i + k) * stride2 + j + l)) {
587 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
588 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
597 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
598 plane1 = (uint16_t*)img1->planes[VPX_PLANE_V];
599 plane2 = (uint16_t*)img2->planes[VPX_PLANE_V];
600 stride1 = img1->stride[VPX_PLANE_V]/2;
601 stride2 = img2->stride[VPX_PLANE_V]/2;
602 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
603 for (j = 0; match && j < c_w; j += bsizex) {
605 const int si = mmin(i + bsizey, c_h - i);
606 const int sj = mmin(j + bsizex, c_w - j);
607 for (k = 0; match && k < si; ++k) {
608 for (l = 0; match && l < sj; ++l) {
609 if (*(plane1 + (i + k) * stride1 + j + l) !=
610 *(plane2 + (i + k) * stride2 + j + l)) {
613 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
614 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
625 static void find_mismatch(const vpx_image_t *const img1,
626 const vpx_image_t *const img2,
627 int yloc[4], int uloc[4], int vloc[4]) {
628 const uint32_t bsize = 64;
629 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
630 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
632 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
634 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
637 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
638 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
639 for (j = 0; match && j < img1->d_w; j += bsize) {
641 const int si = mmin(i + bsize, img1->d_h) - i;
642 const int sj = mmin(j + bsize, img1->d_w) - j;
643 for (k = 0; match && k < si; ++k) {
644 for (l = 0; match && l < sj; ++l) {
645 if (*(img1->planes[VPX_PLANE_Y] +
646 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
647 *(img2->planes[VPX_PLANE_Y] +
648 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
651 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
652 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
653 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
654 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
663 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
664 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
665 for (j = 0; match && j < c_w; j += bsizex) {
667 const int si = mmin(i + bsizey, c_h - i);
668 const int sj = mmin(j + bsizex, c_w - j);
669 for (k = 0; match && k < si; ++k) {
670 for (l = 0; match && l < sj; ++l) {
671 if (*(img1->planes[VPX_PLANE_U] +
672 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
673 *(img2->planes[VPX_PLANE_U] +
674 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
677 uloc[2] = *(img1->planes[VPX_PLANE_U] +
678 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
679 uloc[3] = *(img2->planes[VPX_PLANE_U] +
680 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
688 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
689 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
690 for (j = 0; match && j < c_w; j += bsizex) {
692 const int si = mmin(i + bsizey, c_h - i);
693 const int sj = mmin(j + bsizex, c_w - j);
694 for (k = 0; match && k < si; ++k) {
695 for (l = 0; match && l < sj; ++l) {
696 if (*(img1->planes[VPX_PLANE_V] +
697 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
698 *(img2->planes[VPX_PLANE_V] +
699 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
702 vloc[2] = *(img1->planes[VPX_PLANE_V] +
703 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
704 vloc[3] = *(img2->planes[VPX_PLANE_V] +
705 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
715 static int compare_img(const vpx_image_t *const img1,
716 const vpx_image_t *const img2) {
717 uint32_t l_w = img1->d_w;
719 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
721 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
725 match &= (img1->fmt == img2->fmt);
726 match &= (img1->d_w == img2->d_w);
727 match &= (img1->d_h == img2->d_h);
728 #if CONFIG_VP9_HIGHBITDEPTH
729 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
735 for (i = 0; i < img1->d_h; ++i)
736 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
737 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
740 for (i = 0; i < c_h; ++i)
741 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
742 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
745 for (i = 0; i < c_h; ++i)
746 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
747 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
754 #define NELEMENTS(x) (sizeof(x)/sizeof(x[0]))
755 #if CONFIG_VP9_ENCODER
756 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
758 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
762 typedef int stereo_format_t;
763 struct WebmOutputContext { int debug; };
766 /* Per-stream configuration */
767 struct stream_config {
768 struct vpx_codec_enc_cfg cfg;
770 const char *stats_fn;
771 #if CONFIG_FP_MB_STATS
772 const char *fpmb_stats_fn;
774 stereo_format_t stereo_fmt;
775 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
778 #if CONFIG_VP9_HIGHBITDEPTH
779 // whether to use 16bit internal buffers
780 int use_16bit_internal;
785 struct stream_state {
787 struct stream_state *next;
788 struct stream_config config;
790 struct rate_hist *rate_hist;
791 struct WebmOutputContext webm_ctx;
792 uint64_t psnr_sse_total;
793 uint64_t psnr_samples_total;
794 double psnr_totals[4];
797 vpx_codec_ctx_t encoder;
798 unsigned int frames_out;
802 #if CONFIG_FP_MB_STATS
803 stats_io_t fpmb_stats;
805 struct vpx_image *img;
806 vpx_codec_ctx_t decoder;
811 static void validate_positive_rational(const char *msg,
812 struct vpx_rational *rat) {
819 die("Error: %s must be positive\n", msg);
822 die("Error: %s has zero denominator\n", msg);
826 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
829 const int num_encoder = get_vpx_encoder_count();
832 die("Error: no valid encoder available\n");
834 /* Initialize default parameters */
835 memset(global, 0, sizeof(*global));
836 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
838 global->color_type = I420;
839 /* Assign default deadline to good quality */
840 global->deadline = VPX_DL_GOOD_QUALITY;
842 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
845 if (arg_match(&arg, &codecarg, argi)) {
846 global->codec = get_vpx_encoder_by_name(arg.val);
848 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
849 } else if (arg_match(&arg, &passes, argi)) {
850 global->passes = arg_parse_uint(&arg);
852 if (global->passes < 1 || global->passes > 2)
853 die("Error: Invalid number of passes (%d)\n", global->passes);
854 } else if (arg_match(&arg, &pass_arg, argi)) {
855 global->pass = arg_parse_uint(&arg);
857 if (global->pass < 1 || global->pass > 2)
858 die("Error: Invalid pass selected (%d)\n",
860 } else if (arg_match(&arg, &usage, argi))
861 global->usage = arg_parse_uint(&arg);
862 else if (arg_match(&arg, &deadline, argi))
863 global->deadline = arg_parse_uint(&arg);
864 else if (arg_match(&arg, &best_dl, argi))
865 global->deadline = VPX_DL_BEST_QUALITY;
866 else if (arg_match(&arg, &good_dl, argi))
867 global->deadline = VPX_DL_GOOD_QUALITY;
868 else if (arg_match(&arg, &rt_dl, argi))
869 global->deadline = VPX_DL_REALTIME;
870 else if (arg_match(&arg, &use_yv12, argi))
871 global->color_type = YV12;
872 else if (arg_match(&arg, &use_i420, argi))
873 global->color_type = I420;
874 else if (arg_match(&arg, &use_i422, argi))
875 global->color_type = I422;
876 else if (arg_match(&arg, &use_i444, argi))
877 global->color_type = I444;
878 else if (arg_match(&arg, &use_i440, argi))
879 global->color_type = I440;
880 else if (arg_match(&arg, &quietarg, argi))
882 else if (arg_match(&arg, &verbosearg, argi))
884 else if (arg_match(&arg, &limit, argi))
885 global->limit = arg_parse_uint(&arg);
886 else if (arg_match(&arg, &skip, argi))
887 global->skip_frames = arg_parse_uint(&arg);
888 else if (arg_match(&arg, &psnrarg, argi))
889 global->show_psnr = 1;
890 else if (arg_match(&arg, &recontest, argi))
891 global->test_decode = arg_parse_enum_or_int(&arg);
892 else if (arg_match(&arg, &framerate, argi)) {
893 global->framerate = arg_parse_rational(&arg);
894 validate_positive_rational(arg.name, &global->framerate);
895 global->have_framerate = 1;
896 } else if (arg_match(&arg, &out_part, argi))
897 global->out_part = 1;
898 else if (arg_match(&arg, &debugmode, argi))
900 else if (arg_match(&arg, &q_hist_n, argi))
901 global->show_q_hist_buckets = arg_parse_uint(&arg);
902 else if (arg_match(&arg, &rate_hist_n, argi))
903 global->show_rate_hist_buckets = arg_parse_uint(&arg);
904 else if (arg_match(&arg, &disable_warnings, argi))
905 global->disable_warnings = 1;
906 else if (arg_match(&arg, &disable_warning_prompt, argi))
907 global->disable_warning_prompt = 1;
913 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
914 if (global->pass > global->passes) {
915 warn("Assuming --pass=%d implies --passes=%d\n",
916 global->pass, global->pass);
917 global->passes = global->pass;
920 /* Validate global config */
921 if (global->passes == 0) {
922 #if CONFIG_VP9_ENCODER
923 // Make default VP9 passes = 2 until there is a better quality 1-pass
925 if (global->codec != NULL && global->codec->name != NULL)
926 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
927 global->deadline != VPX_DL_REALTIME) ? 2 : 1;
933 if (global->deadline == VPX_DL_REALTIME &&
934 global->passes > 1) {
935 warn("Enforcing one-pass encoding in realtime mode\n");
941 static void open_input_file(struct VpxInputContext *input) {
942 /* Parse certain options from the input file, if possible */
943 input->file = strcmp(input->filename, "-")
944 ? fopen(input->filename, "rb") : set_binary_mode(stdin);
947 fatal("Failed to open input file");
949 if (!fseeko(input->file, 0, SEEK_END)) {
950 /* Input file is seekable. Figure out how long it is, so we can get
953 input->length = ftello(input->file);
957 /* Default to 1:1 pixel aspect ratio. */
958 input->pixel_aspect_ratio.numerator = 1;
959 input->pixel_aspect_ratio.denominator = 1;
961 /* For RAW input sources, these bytes will applied on the first frame
964 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
965 input->detect.position = 0;
967 if (input->detect.buf_read == 4
968 && file_is_y4m(input->detect.buf)) {
969 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
970 input->only_i420) >= 0) {
971 input->file_type = FILE_TYPE_Y4M;
972 input->width = input->y4m.pic_w;
973 input->height = input->y4m.pic_h;
974 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
975 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
976 input->framerate.numerator = input->y4m.fps_n;
977 input->framerate.denominator = input->y4m.fps_d;
978 input->fmt = input->y4m.vpx_fmt;
979 input->bit_depth = input->y4m.bit_depth;
981 fatal("Unsupported Y4M stream.");
982 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
983 fatal("IVF is not supported as input.");
985 input->file_type = FILE_TYPE_RAW;
990 static void close_input_file(struct VpxInputContext *input) {
992 if (input->file_type == FILE_TYPE_Y4M)
993 y4m_input_close(&input->y4m);
996 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
997 struct stream_state *prev) {
998 struct stream_state *stream;
1000 stream = calloc(1, sizeof(*stream));
1001 if (stream == NULL) {
1002 fatal("Failed to allocate new stream.");
1006 memcpy(stream, prev, sizeof(*stream));
1008 prev->next = stream;
1010 vpx_codec_err_t res;
1012 /* Populate encoder configuration */
1013 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1014 &stream->config.cfg,
1017 fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1019 /* Change the default timebase to a high enough value so that the
1020 * encoder will always create strictly increasing timestamps.
1022 stream->config.cfg.g_timebase.den = 1000;
1024 /* Never use the library's default resolution, require it be parsed
1025 * from the file or set on the command line.
1027 stream->config.cfg.g_w = 0;
1028 stream->config.cfg.g_h = 0;
1030 /* Initialize remaining stream parameters */
1031 stream->config.write_webm = 1;
1033 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1034 stream->webm_ctx.last_pts_ns = -1;
1035 stream->webm_ctx.writer = NULL;
1036 stream->webm_ctx.segment = NULL;
1039 /* Allows removal of the application version from the EBML tags */
1040 stream->webm_ctx.debug = global->debug;
1042 /* Default lag_in_frames is 0 in realtime mode */
1043 if (global->deadline == VPX_DL_REALTIME)
1044 stream->config.cfg.g_lag_in_frames = 0;
1047 /* Output files must be specified for each stream */
1048 stream->config.out_fn = NULL;
1050 stream->next = NULL;
1055 static int parse_stream_params(struct VpxEncoderConfig *global,
1056 struct stream_state *stream,
1058 char **argi, **argj;
1060 static const arg_def_t **ctrl_args = no_args;
1061 static const int *ctrl_args_map = NULL;
1062 struct stream_config *config = &stream->config;
1063 int eos_mark_found = 0;
1064 #if CONFIG_VP9_HIGHBITDEPTH
1065 int test_16bit_internal = 0;
1068 // Handle codec specific options
1070 #if CONFIG_VP8_ENCODER
1071 } else if (strcmp(global->codec->name, "vp8") == 0) {
1072 ctrl_args = vp8_args;
1073 ctrl_args_map = vp8_arg_ctrl_map;
1075 #if CONFIG_VP9_ENCODER
1076 } else if (strcmp(global->codec->name, "vp9") == 0) {
1077 ctrl_args = vp9_args;
1078 ctrl_args_map = vp9_arg_ctrl_map;
1082 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1085 /* Once we've found an end-of-stream marker (--) we want to continue
1086 * shifting arguments but not consuming them.
1088 if (eos_mark_found) {
1091 } else if (!strcmp(*argj, "--")) {
1096 if (arg_match(&arg, &outputfile, argi)) {
1097 config->out_fn = arg.val;
1098 } else if (arg_match(&arg, &fpf_name, argi)) {
1099 config->stats_fn = arg.val;
1100 #if CONFIG_FP_MB_STATS
1101 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1102 config->fpmb_stats_fn = arg.val;
1104 } else if (arg_match(&arg, &use_webm, argi)) {
1106 config->write_webm = 1;
1108 die("Error: --webm specified but webm is disabled.");
1110 } else if (arg_match(&arg, &use_ivf, argi)) {
1111 config->write_webm = 0;
1112 } else if (arg_match(&arg, &threads, argi)) {
1113 config->cfg.g_threads = arg_parse_uint(&arg);
1114 } else if (arg_match(&arg, &profile, argi)) {
1115 config->cfg.g_profile = arg_parse_uint(&arg);
1116 } else if (arg_match(&arg, &width, argi)) {
1117 config->cfg.g_w = arg_parse_uint(&arg);
1118 } else if (arg_match(&arg, &height, argi)) {
1119 config->cfg.g_h = arg_parse_uint(&arg);
1120 #if CONFIG_VP9_HIGHBITDEPTH
1121 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1122 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1123 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1124 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1127 } else if (arg_match(&arg, &stereo_mode, argi)) {
1128 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1130 } else if (arg_match(&arg, &timebase, argi)) {
1131 config->cfg.g_timebase = arg_parse_rational(&arg);
1132 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1133 } else if (arg_match(&arg, &error_resilient, argi)) {
1134 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1135 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1136 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1137 if (global->deadline == VPX_DL_REALTIME &&
1138 config->cfg.g_lag_in_frames != 0) {
1139 warn("non-zero %s option ignored in realtime mode.\n", arg.name);
1140 config->cfg.g_lag_in_frames = 0;
1142 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1143 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1144 } else if (arg_match(&arg, &resize_allowed, argi)) {
1145 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1146 } else if (arg_match(&arg, &resize_width, argi)) {
1147 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1148 } else if (arg_match(&arg, &resize_height, argi)) {
1149 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1150 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1151 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1152 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1153 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1154 } else if (arg_match(&arg, &end_usage, argi)) {
1155 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1156 } else if (arg_match(&arg, &target_bitrate, argi)) {
1157 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1158 } else if (arg_match(&arg, &min_quantizer, argi)) {
1159 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1160 } else if (arg_match(&arg, &max_quantizer, argi)) {
1161 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1162 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1163 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1164 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1165 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1166 } else if (arg_match(&arg, &buf_sz, argi)) {
1167 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1168 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1169 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1170 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1171 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1172 } else if (arg_match(&arg, &bias_pct, argi)) {
1173 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1174 if (global->passes < 2)
1175 warn("option %s ignored in one-pass mode.\n", arg.name);
1176 } else if (arg_match(&arg, &minsection_pct, argi)) {
1177 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1179 if (global->passes < 2)
1180 warn("option %s ignored in one-pass mode.\n", arg.name);
1181 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1182 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1184 if (global->passes < 2)
1185 warn("option %s ignored in one-pass mode.\n", arg.name);
1186 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1187 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1188 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1189 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1190 } else if (arg_match(&arg, &kf_disabled, argi)) {
1191 config->cfg.kf_mode = VPX_KF_DISABLED;
1192 #if CONFIG_VP9_HIGHBITDEPTH
1193 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1194 if (strcmp(global->codec->name, "vp9") == 0) {
1195 test_16bit_internal = 1;
1200 for (i = 0; ctrl_args[i]; i++) {
1201 if (arg_match(&arg, ctrl_args[i], argi)) {
1205 /* Point either to the next free element or the first
1206 * instance of this control.
1208 for (j = 0; j < config->arg_ctrl_cnt; j++)
1209 if (ctrl_args_map != NULL &&
1210 config->arg_ctrls[j][0] == ctrl_args_map[i])
1214 assert(j < (int)ARG_CTRL_CNT_MAX);
1215 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1216 config->arg_ctrls[j][0] = ctrl_args_map[i];
1217 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1218 if (j == config->arg_ctrl_cnt)
1219 config->arg_ctrl_cnt++;
1227 #if CONFIG_VP9_HIGHBITDEPTH
1228 if (strcmp(global->codec->name, "vp9") == 0) {
1229 config->use_16bit_internal = test_16bit_internal |
1230 (config->cfg.g_profile > 1);
1233 return eos_mark_found;
1237 #define FOREACH_STREAM(func) \
1239 struct stream_state *stream; \
1240 for (stream = streams; stream; stream = stream->next) { \
1246 static void validate_stream_config(const struct stream_state *stream,
1247 const struct VpxEncoderConfig *global) {
1248 const struct stream_state *streami;
1251 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1252 fatal("Stream %d: Specify stream dimensions with --width (-w) "
1253 " and --height (-h)", stream->index);
1255 // Check that the codec bit depth is greater than the input bit depth.
1256 if (stream->config.cfg.g_input_bit_depth >
1257 (unsigned int)stream->config.cfg.g_bit_depth) {
1258 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1259 stream->index, (int)stream->config.cfg.g_bit_depth,
1260 stream->config.cfg.g_input_bit_depth);
1263 for (streami = stream; streami; streami = streami->next) {
1264 /* All streams require output files */
1265 if (!streami->config.out_fn)
1266 fatal("Stream %d: Output file is required (specify with -o)",
1269 /* Check for two streams outputting to the same file */
1270 if (streami != stream) {
1271 const char *a = stream->config.out_fn;
1272 const char *b = streami->config.out_fn;
1273 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1274 fatal("Stream %d: duplicate output file (from stream %d)",
1275 streami->index, stream->index);
1278 /* Check for two streams sharing a stats file. */
1279 if (streami != stream) {
1280 const char *a = stream->config.stats_fn;
1281 const char *b = streami->config.stats_fn;
1282 if (a && b && !strcmp(a, b))
1283 fatal("Stream %d: duplicate stats file (from stream %d)",
1284 streami->index, stream->index);
1287 #if CONFIG_FP_MB_STATS
1288 /* Check for two streams sharing a mb stats file. */
1289 if (streami != stream) {
1290 const char *a = stream->config.fpmb_stats_fn;
1291 const char *b = streami->config.fpmb_stats_fn;
1292 if (a && b && !strcmp(a, b))
1293 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1294 streami->index, stream->index);
1301 static void set_stream_dimensions(struct stream_state *stream,
1304 if (!stream->config.cfg.g_w) {
1305 if (!stream->config.cfg.g_h)
1306 stream->config.cfg.g_w = w;
1308 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1310 if (!stream->config.cfg.g_h) {
1311 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1315 static const char* file_type_to_string(enum VideoFileType t) {
1317 case FILE_TYPE_RAW: return "RAW";
1318 case FILE_TYPE_Y4M: return "Y4M";
1319 default: return "Other";
1323 static const char* image_format_to_string(vpx_img_fmt_t f) {
1325 case VPX_IMG_FMT_I420: return "I420";
1326 case VPX_IMG_FMT_I422: return "I422";
1327 case VPX_IMG_FMT_I444: return "I444";
1328 case VPX_IMG_FMT_I440: return "I440";
1329 case VPX_IMG_FMT_YV12: return "YV12";
1330 case VPX_IMG_FMT_I42016: return "I42016";
1331 case VPX_IMG_FMT_I42216: return "I42216";
1332 case VPX_IMG_FMT_I44416: return "I44416";
1333 case VPX_IMG_FMT_I44016: return "I44016";
1334 default: return "Other";
1338 static void show_stream_config(struct stream_state *stream,
1339 struct VpxEncoderConfig *global,
1340 struct VpxInputContext *input) {
1342 #define SHOW(field) \
1343 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1345 if (stream->index == 0) {
1346 fprintf(stderr, "Codec: %s\n",
1347 vpx_codec_iface_name(global->codec->codec_interface()));
1348 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1350 file_type_to_string(input->file_type),
1351 image_format_to_string(input->fmt));
1353 if (stream->next || stream->index)
1354 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1355 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1356 fprintf(stderr, "Encoder parameters:\n");
1364 SHOW(g_input_bit_depth);
1365 SHOW(g_timebase.num);
1366 SHOW(g_timebase.den);
1367 SHOW(g_error_resilient);
1369 SHOW(g_lag_in_frames);
1370 SHOW(rc_dropframe_thresh);
1371 SHOW(rc_resize_allowed);
1372 SHOW(rc_scaled_width);
1373 SHOW(rc_scaled_height);
1374 SHOW(rc_resize_up_thresh);
1375 SHOW(rc_resize_down_thresh);
1377 SHOW(rc_target_bitrate);
1378 SHOW(rc_min_quantizer);
1379 SHOW(rc_max_quantizer);
1380 SHOW(rc_undershoot_pct);
1381 SHOW(rc_overshoot_pct);
1383 SHOW(rc_buf_initial_sz);
1384 SHOW(rc_buf_optimal_sz);
1385 SHOW(rc_2pass_vbr_bias_pct);
1386 SHOW(rc_2pass_vbr_minsection_pct);
1387 SHOW(rc_2pass_vbr_maxsection_pct);
1394 static void open_output_file(struct stream_state *stream,
1395 struct VpxEncoderConfig *global,
1396 const struct VpxRational *pixel_aspect_ratio) {
1397 const char *fn = stream->config.out_fn;
1398 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1400 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1403 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1406 fatal("Failed to open output file");
1408 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1409 fatal("WebM output to pipes not supported.");
1412 if (stream->config.write_webm) {
1413 stream->webm_ctx.stream = stream->file;
1414 write_webm_file_header(&stream->webm_ctx, cfg,
1416 stream->config.stereo_fmt,
1417 global->codec->fourcc,
1418 pixel_aspect_ratio);
1421 (void)pixel_aspect_ratio;
1424 if (!stream->config.write_webm) {
1425 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1430 static void close_output_file(struct stream_state *stream,
1431 unsigned int fourcc) {
1432 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1434 if (cfg->g_pass == VPX_RC_FIRST_PASS)
1438 if (stream->config.write_webm) {
1439 write_webm_file_footer(&stream->webm_ctx);
1443 if (!stream->config.write_webm) {
1444 if (!fseek(stream->file, 0, SEEK_SET))
1445 ivf_write_file_header(stream->file, &stream->config.cfg,
1447 stream->frames_out);
1450 fclose(stream->file);
1454 static void setup_pass(struct stream_state *stream,
1455 struct VpxEncoderConfig *global,
1457 if (stream->config.stats_fn) {
1458 if (!stats_open_file(&stream->stats, stream->config.stats_fn,
1460 fatal("Failed to open statistics store");
1462 if (!stats_open_mem(&stream->stats, pass))
1463 fatal("Failed to open statistics store");
1466 #if CONFIG_FP_MB_STATS
1467 if (stream->config.fpmb_stats_fn) {
1468 if (!stats_open_file(&stream->fpmb_stats,
1469 stream->config.fpmb_stats_fn, pass))
1470 fatal("Failed to open mb statistics store");
1472 if (!stats_open_mem(&stream->fpmb_stats, pass))
1473 fatal("Failed to open mb statistics store");
1477 stream->config.cfg.g_pass = global->passes == 2
1478 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1481 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1482 #if CONFIG_FP_MB_STATS
1483 stream->config.cfg.rc_firstpass_mb_stats_in =
1484 stats_get(&stream->fpmb_stats);
1488 stream->cx_time = 0;
1490 stream->frames_out = 0;
1494 static void initialize_encoder(struct stream_state *stream,
1495 struct VpxEncoderConfig *global) {
1499 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1500 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1501 #if CONFIG_VP9_HIGHBITDEPTH
1502 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1505 /* Construct Encoder Context */
1506 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1507 &stream->config.cfg, flags);
1508 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1510 /* Note that we bypass the vpx_codec_control wrapper macro because
1511 * we're being clever to store the control IDs in an array. Real
1512 * applications will want to make use of the enumerations directly
1514 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1515 int ctrl = stream->config.arg_ctrls[i][0];
1516 int value = stream->config.arg_ctrls[i][1];
1517 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1518 fprintf(stderr, "Error: Tried to set control %d = %d\n",
1521 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1525 if (global->test_decode != TEST_DECODE_OFF) {
1526 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1527 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1533 static void encode_frame(struct stream_state *stream,
1534 struct VpxEncoderConfig *global,
1535 struct vpx_image *img,
1536 unsigned int frames_in) {
1537 vpx_codec_pts_t frame_start, next_frame_start;
1538 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1539 struct vpx_usec_timer timer;
1541 frame_start = (cfg->g_timebase.den * (int64_t)(frames_in - 1)
1542 * global->framerate.den)
1543 / cfg->g_timebase.num / global->framerate.num;
1544 next_frame_start = (cfg->g_timebase.den * (int64_t)(frames_in)
1545 * global->framerate.den)
1546 / cfg->g_timebase.num / global->framerate.num;
1548 /* Scale if necessary */
1549 #if CONFIG_VP9_HIGHBITDEPTH
1551 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1552 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1553 if (img->fmt != VPX_IMG_FMT_I42016) {
1554 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1559 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I42016,
1560 cfg->g_w, cfg->g_h, 16);
1562 I420Scale_16((uint16*)img->planes[VPX_PLANE_Y],
1563 img->stride[VPX_PLANE_Y]/2,
1564 (uint16*)img->planes[VPX_PLANE_U],
1565 img->stride[VPX_PLANE_U]/2,
1566 (uint16*)img->planes[VPX_PLANE_V],
1567 img->stride[VPX_PLANE_V]/2,
1569 (uint16*)stream->img->planes[VPX_PLANE_Y],
1570 stream->img->stride[VPX_PLANE_Y]/2,
1571 (uint16*)stream->img->planes[VPX_PLANE_U],
1572 stream->img->stride[VPX_PLANE_U]/2,
1573 (uint16*)stream->img->planes[VPX_PLANE_V],
1574 stream->img->stride[VPX_PLANE_V]/2,
1575 stream->img->d_w, stream->img->d_h,
1579 stream->encoder.err = 1;
1580 ctx_exit_on_error(&stream->encoder,
1581 "Stream %d: Failed to encode frame.\n"
1582 "Scaling disabled in this configuration. \n"
1583 "To enable, configure with --enable-libyuv\n",
1589 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1590 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1591 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1596 stream->img = vpx_img_alloc(NULL, VPX_IMG_FMT_I420,
1597 cfg->g_w, cfg->g_h, 16);
1598 I420Scale(img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1599 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1600 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V],
1602 stream->img->planes[VPX_PLANE_Y],
1603 stream->img->stride[VPX_PLANE_Y],
1604 stream->img->planes[VPX_PLANE_U],
1605 stream->img->stride[VPX_PLANE_U],
1606 stream->img->planes[VPX_PLANE_V],
1607 stream->img->stride[VPX_PLANE_V],
1608 stream->img->d_w, stream->img->d_h,
1612 stream->encoder.err = 1;
1613 ctx_exit_on_error(&stream->encoder,
1614 "Stream %d: Failed to encode frame.\n"
1615 "Scaling disabled in this configuration. \n"
1616 "To enable, configure with --enable-libyuv\n",
1621 vpx_usec_timer_start(&timer);
1622 vpx_codec_encode(&stream->encoder, img, frame_start,
1623 (unsigned long)(next_frame_start - frame_start),
1624 0, global->deadline);
1625 vpx_usec_timer_mark(&timer);
1626 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1627 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1632 static void update_quantizer_histogram(struct stream_state *stream) {
1633 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1636 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1637 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1638 stream->counts[q]++;
1643 static void get_cx_data(struct stream_state *stream,
1644 struct VpxEncoderConfig *global,
1646 const vpx_codec_cx_pkt_t *pkt;
1647 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1648 vpx_codec_iter_t iter = NULL;
1651 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1652 static size_t fsize = 0;
1653 static int64_t ivf_header_pos = 0;
1655 switch (pkt->kind) {
1656 case VPX_CODEC_CX_FRAME_PKT:
1657 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1658 stream->frames_out++;
1661 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1663 update_rate_histogram(stream->rate_hist, cfg, pkt);
1665 if (stream->config.write_webm) {
1666 write_webm_block(&stream->webm_ctx, cfg, pkt);
1669 if (!stream->config.write_webm) {
1670 if (pkt->data.frame.partition_id <= 0) {
1671 ivf_header_pos = ftello(stream->file);
1672 fsize = pkt->data.frame.sz;
1674 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1676 fsize += pkt->data.frame.sz;
1678 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1679 const int64_t currpos = ftello(stream->file);
1680 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1681 ivf_write_frame_size(stream->file, fsize);
1682 fseeko(stream->file, currpos, SEEK_SET);
1686 (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1689 stream->nbytes += pkt->data.raw.sz;
1693 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1694 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1695 (unsigned int)pkt->data.frame.sz, NULL, 0);
1696 if (stream->decoder.err) {
1697 warn_or_exit_on_error(&stream->decoder,
1698 global->test_decode == TEST_DECODE_FATAL,
1699 "Failed to decode frame %d in stream %d",
1700 stream->frames_out + 1, stream->index);
1701 stream->mismatch_seen = stream->frames_out + 1;
1706 case VPX_CODEC_STATS_PKT:
1707 stream->frames_out++;
1708 stats_write(&stream->stats,
1709 pkt->data.twopass_stats.buf,
1710 pkt->data.twopass_stats.sz);
1711 stream->nbytes += pkt->data.raw.sz;
1713 #if CONFIG_FP_MB_STATS
1714 case VPX_CODEC_FPMB_STATS_PKT:
1715 stats_write(&stream->fpmb_stats,
1716 pkt->data.firstpass_mb_stats.buf,
1717 pkt->data.firstpass_mb_stats.sz);
1718 stream->nbytes += pkt->data.raw.sz;
1721 case VPX_CODEC_PSNR_PKT:
1723 if (global->show_psnr) {
1726 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1727 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1728 for (i = 0; i < 4; i++) {
1730 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1731 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1733 stream->psnr_count++;
1744 static void show_psnr(struct stream_state *stream, double peak) {
1748 if (!stream->psnr_count)
1751 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1752 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1753 (double)stream->psnr_sse_total);
1754 fprintf(stderr, " %.3f", ovpsnr);
1756 for (i = 0; i < 4; i++) {
1757 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1759 fprintf(stderr, "\n");
1763 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1764 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1767 static void test_decode(struct stream_state *stream,
1768 enum TestDecodeFatality fatal,
1769 const VpxInterface *codec) {
1770 vpx_image_t enc_img, dec_img;
1772 if (stream->mismatch_seen)
1775 /* Get the internal reference frame */
1776 if (strcmp(codec->name, "vp8") == 0) {
1777 struct vpx_ref_frame ref_enc, ref_dec;
1780 width = (stream->config.cfg.g_w + 15) & ~15;
1781 height = (stream->config.cfg.g_h + 15) & ~15;
1782 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1783 enc_img = ref_enc.img;
1784 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1785 dec_img = ref_dec.img;
1787 ref_enc.frame_type = VP8_LAST_FRAME;
1788 ref_dec.frame_type = VP8_LAST_FRAME;
1789 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1790 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1792 struct vp9_ref_frame ref_enc, ref_dec;
1796 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1797 enc_img = ref_enc.img;
1798 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1799 dec_img = ref_dec.img;
1800 #if CONFIG_VP9_HIGHBITDEPTH
1801 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1802 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1803 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1804 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1805 enc_img.d_w, enc_img.d_h, 16);
1806 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1808 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1809 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1810 dec_img.d_w, dec_img.d_h, 16);
1811 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1816 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1817 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1819 if (!compare_img(&enc_img, &dec_img)) {
1820 int y[4], u[4], v[4];
1821 #if CONFIG_VP9_HIGHBITDEPTH
1822 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1823 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1825 find_mismatch(&enc_img, &dec_img, y, u, v);
1828 find_mismatch(&enc_img, &dec_img, y, u, v);
1830 stream->decoder.err = 1;
1831 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1832 "Stream %d: Encode/decode mismatch on frame %d at"
1833 " Y[%d, %d] {%d/%d},"
1834 " U[%d, %d] {%d/%d},"
1835 " V[%d, %d] {%d/%d}",
1836 stream->index, stream->frames_out,
1837 y[0], y[1], y[2], y[3],
1838 u[0], u[1], u[2], u[3],
1839 v[0], v[1], v[2], v[3]);
1840 stream->mismatch_seen = stream->frames_out;
1843 vpx_img_free(&enc_img);
1844 vpx_img_free(&dec_img);
1848 static void print_time(const char *label, int64_t etl) {
1855 etl -= hours * 3600;
1860 fprintf(stderr, "[%3s %2"PRId64":%02"PRId64":%02"PRId64"] ",
1861 label, hours, mins, secs);
1863 fprintf(stderr, "[%3s unknown] ", label);
1868 int main(int argc, const char **argv_) {
1871 #if CONFIG_VP9_HIGHBITDEPTH
1872 vpx_image_t raw_shift;
1873 int allocated_raw_shift = 0;
1874 int use_16bit_internal = 0;
1875 int input_shift = 0;
1877 int frame_avail, got_data;
1879 struct VpxInputContext input;
1880 struct VpxEncoderConfig global;
1881 struct stream_state *streams = NULL;
1882 char **argv, **argi;
1883 uint64_t cx_time = 0;
1887 memset(&input, 0, sizeof(input));
1888 exec_name = argv_[0];
1893 /* Setup default input stream settings */
1894 input.framerate.numerator = 30;
1895 input.framerate.denominator = 1;
1896 input.only_i420 = 1;
1897 input.bit_depth = 0;
1899 /* First parse the global configuration values, because we want to apply
1900 * other parameters on top of the default configuration provided by the
1903 argv = argv_dup(argc - 1, argv_ + 1);
1904 parse_global_config(&global, argv);
1906 switch (global.color_type) {
1908 input.fmt = VPX_IMG_FMT_I420;
1911 input.fmt = VPX_IMG_FMT_I422;
1914 input.fmt = VPX_IMG_FMT_I444;
1917 input.fmt = VPX_IMG_FMT_I440;
1920 input.fmt = VPX_IMG_FMT_YV12;
1925 /* Now parse each stream's parameters. Using a local scope here
1926 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1929 struct stream_state *stream = NULL;
1932 stream = new_stream(&global, stream);
1936 } while (parse_stream_params(&global, stream, argv));
1939 /* Check for unrecognized options */
1940 for (argi = argv; *argi; argi++)
1941 if (argi[0][0] == '-' && argi[0][1])
1942 die("Error: Unrecognized option %s\n", *argi);
1944 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt,
1945 &global, &stream->config.cfg););
1947 /* Handle non-option arguments */
1948 input.filename = argv[0];
1950 if (!input.filename)
1953 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1954 if (global.codec->fourcc == VP9_FOURCC)
1955 input.only_i420 = 0;
1957 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1958 int frames_in = 0, seen_frames = 0;
1959 int64_t estimated_time_left = -1;
1960 int64_t average_rate = -1;
1961 int64_t lagged_count = 0;
1963 open_input_file(&input);
1965 /* If the input file doesn't specify its w/h (raw files), try to get
1966 * the data from the first stream's configuration.
1968 if (!input.width || !input.height) {
1970 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1971 input.width = stream->config.cfg.g_w;
1972 input.height = stream->config.cfg.g_h;
1978 /* Update stream configurations from the input file's parameters */
1979 if (!input.width || !input.height)
1980 fatal("Specify stream dimensions with --width (-w) "
1981 " and --height (-h)");
1983 /* If input file does not specify bit-depth but input-bit-depth parameter
1984 * exists, assume that to be the input bit-depth. However, if the
1985 * input-bit-depth paramter does not exist, assume the input bit-depth
1986 * to be the same as the codec bit-depth.
1988 if (!input.bit_depth) {
1990 if (stream->config.cfg.g_input_bit_depth)
1991 input.bit_depth = stream->config.cfg.g_input_bit_depth;
1993 input.bit_depth = stream->config.cfg.g_input_bit_depth =
1994 (int)stream->config.cfg.g_bit_depth;
1996 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
1999 stream->config.cfg.g_input_bit_depth = input.bit_depth;
2003 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2004 FOREACH_STREAM(validate_stream_config(stream, &global));
2006 /* Ensure that --passes and --pass are consistent. If --pass is set and
2007 * --passes=2, ensure --fpf was set.
2009 if (global.pass && global.passes == 2)
2011 if (!stream->config.stats_fn)
2012 die("Stream %d: Must specify --fpf when --pass=%d"
2013 " and --passes=2\n", stream->index, global.pass);
2018 if (stream->config.write_webm) {
2019 stream->config.write_webm = 0;
2020 warn("vpxenc was compiled without WebM container support."
2021 "Producing IVF output");
2026 /* Use the frame rate from the file only if none was specified
2027 * on the command-line.
2029 if (!global.have_framerate) {
2030 global.framerate.num = input.framerate.numerator;
2031 global.framerate.den = input.framerate.denominator;
2032 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2033 stream->config.cfg.g_timebase.num = global.framerate.den);
2036 /* Show configuration */
2037 if (global.verbose && pass == 0)
2038 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2040 if (pass == (global.pass ? global.pass - 1 : 0)) {
2041 if (input.file_type == FILE_TYPE_Y4M)
2042 /*The Y4M reader does its own allocation.
2043 Just initialize this here to avoid problems if we never read any
2045 memset(&raw, 0, sizeof(raw));
2047 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2049 FOREACH_STREAM(stream->rate_hist =
2050 init_rate_histogram(&stream->config.cfg,
2051 &global.framerate));
2054 FOREACH_STREAM(setup_pass(stream, &global, pass));
2055 FOREACH_STREAM(open_output_file(stream, &global,
2056 &input.pixel_aspect_ratio));
2057 FOREACH_STREAM(initialize_encoder(stream, &global));
2059 #if CONFIG_VP9_HIGHBITDEPTH
2060 if (strcmp(global.codec->name, "vp9") == 0) {
2061 // Check to see if at least one stream uses 16 bit internal.
2062 // Currently assume that the bit_depths for all streams using
2063 // highbitdepth are the same.
2065 if (stream->config.use_16bit_internal) {
2066 use_16bit_internal = 1;
2068 if (stream->config.cfg.g_profile == 0) {
2071 input_shift = (int)stream->config.cfg.g_bit_depth -
2072 stream->config.cfg.g_input_bit_depth;
2081 while (frame_avail || got_data) {
2082 struct vpx_usec_timer timer;
2084 if (!global.limit || frames_in < global.limit) {
2085 frame_avail = read_frame(&input, &raw);
2089 seen_frames = frames_in > global.skip_frames ?
2090 frames_in - global.skip_frames : 0;
2092 if (!global.quiet) {
2093 float fps = usec_to_fps(cx_time, seen_frames);
2094 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2096 if (stream_cnt == 1)
2098 "frame %4d/%-4d %7"PRId64"B ",
2099 frames_in, streams->frames_out, (int64_t)streams->nbytes);
2101 fprintf(stderr, "frame %4d ", frames_in);
2103 fprintf(stderr, "%7"PRId64" %s %.2f %s ",
2104 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2105 cx_time > 9999999 ? "ms" : "us",
2106 fps >= 1.0 ? fps : fps * 60,
2107 fps >= 1.0 ? "fps" : "fpm");
2108 print_time("ETA", estimated_time_left);
2114 if (frames_in > global.skip_frames) {
2115 #if CONFIG_VP9_HIGHBITDEPTH
2116 vpx_image_t *frame_to_encode;
2117 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2118 assert(use_16bit_internal);
2119 // Input bit depth and stream bit depth do not match, so up
2120 // shift frame to stream bit depth
2121 if (!allocated_raw_shift) {
2122 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2123 input.width, input.height, 32);
2124 allocated_raw_shift = 1;
2126 vpx_img_upshift(&raw_shift, &raw, input_shift);
2127 frame_to_encode = &raw_shift;
2129 frame_to_encode = &raw;
2131 vpx_usec_timer_start(&timer);
2132 if (use_16bit_internal) {
2133 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2135 if (stream->config.use_16bit_internal)
2136 encode_frame(stream, &global,
2137 frame_avail ? frame_to_encode : NULL,
2143 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2144 FOREACH_STREAM(encode_frame(stream, &global,
2145 frame_avail ? frame_to_encode : NULL,
2149 vpx_usec_timer_start(&timer);
2150 FOREACH_STREAM(encode_frame(stream, &global,
2151 frame_avail ? &raw : NULL,
2154 vpx_usec_timer_mark(&timer);
2155 cx_time += vpx_usec_timer_elapsed(&timer);
2157 FOREACH_STREAM(update_quantizer_histogram(stream));
2160 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2162 if (!got_data && input.length && streams != NULL &&
2163 !streams->frames_out) {
2164 lagged_count = global.limit ? seen_frames : ftello(input.file);
2165 } else if (input.length) {
2170 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2172 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2173 remaining = 1000 * (global.limit - global.skip_frames
2174 - seen_frames + lagged_count);
2176 const int64_t input_pos = ftello(input.file);
2177 const int64_t input_pos_lagged = input_pos - lagged_count;
2178 const int64_t limit = input.length;
2180 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2181 remaining = limit - input_pos + lagged_count;
2184 average_rate = (average_rate <= 0)
2186 : (average_rate * 7 + rate) / 8;
2187 estimated_time_left = average_rate ? remaining / average_rate : -1;
2190 if (got_data && global.test_decode != TEST_DECODE_OFF)
2191 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2196 fprintf(stderr, "\033[K");
2200 fprintf(stderr, "\n");
2202 if (!global.quiet) {
2203 FOREACH_STREAM(fprintf(stderr,
2204 "\rPass %d/%d frame %4d/%-4d %7"PRId64"B %7"PRId64"b/f %7"PRId64"b/s"
2205 " %7"PRId64" %s (%.2f fps)\033[K\n",
2207 global.passes, frames_in, stream->frames_out, (int64_t)stream->nbytes,
2208 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2209 seen_frames ? (int64_t)stream->nbytes * 8 *
2210 (int64_t)global.framerate.num / global.framerate.den /
2212 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2213 stream->cx_time > 9999999 ? "ms" : "us",
2214 usec_to_fps(stream->cx_time, seen_frames)));
2217 if (global.show_psnr) {
2218 if (global.codec->fourcc == VP9_FOURCC) {
2220 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2222 FOREACH_STREAM(show_psnr(stream, 255.0));
2226 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2228 if (global.test_decode != TEST_DECODE_OFF) {
2229 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2232 close_input_file(&input);
2234 if (global.test_decode == TEST_DECODE_FATAL) {
2235 FOREACH_STREAM(res |= stream->mismatch_seen);
2237 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2239 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2241 #if CONFIG_FP_MB_STATS
2242 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2249 if (global.show_q_hist_buckets)
2250 FOREACH_STREAM(show_q_histogram(stream->counts,
2251 global.show_q_hist_buckets));
2253 if (global.show_rate_hist_buckets)
2254 FOREACH_STREAM(show_rate_histogram(stream->rate_hist,
2255 &stream->config.cfg,
2256 global.show_rate_hist_buckets));
2257 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2259 #if CONFIG_INTERNAL_STATS
2260 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2261 * to match some existing utilities.
2263 if (!(global.pass == 1 && global.passes == 2))
2265 FILE *f = fopen("opsnr.stt", "a");
2266 if (stream->mismatch_seen) {
2267 fprintf(f, "First mismatch occurred in frame %d\n",
2268 stream->mismatch_seen);
2270 fprintf(f, "No mismatch detected in recon buffers\n");
2276 #if CONFIG_VP9_HIGHBITDEPTH
2277 if (allocated_raw_shift)
2278 vpx_img_free(&raw_shift);
2283 return res ? EXIT_FAILURE : EXIT_SUCCESS;