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, FILE *stream) {
55 return fread(ptr, size, nmemb, stream);
57 #define fread wrap_fread
59 static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
61 return fwrite(ptr, size, nmemb, stream);
63 #define fwrite wrap_fwrite
65 static const char *exec_name;
67 static void warn_or_exit_on_errorv(vpx_codec_ctx_t *ctx, int fatal,
68 const char *s, va_list ap) {
70 const char *detail = vpx_codec_error_detail(ctx);
72 vfprintf(stderr, s, ap);
73 fprintf(stderr, ": %s\n", vpx_codec_error(ctx));
75 if (detail) fprintf(stderr, " %s\n", detail);
77 if (fatal) exit(EXIT_FAILURE);
81 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s, ...) {
85 warn_or_exit_on_errorv(ctx, 1, s, ap);
89 static void warn_or_exit_on_error(vpx_codec_ctx_t *ctx, int fatal,
94 warn_or_exit_on_errorv(ctx, fatal, s, ap);
98 static int read_frame(struct VpxInputContext *input_ctx, vpx_image_t *img) {
99 FILE *f = input_ctx->file;
100 y4m_input *y4m = &input_ctx->y4m;
103 if (input_ctx->file_type == FILE_TYPE_Y4M) {
104 if (y4m_input_fetch_frame(y4m, f, img) < 1) return 0;
106 shortread = read_yuv_frame(input_ctx, img);
112 static int file_is_y4m(const char detect[4]) {
113 if (memcmp(detect, "YUV4", 4) == 0) {
119 static int fourcc_is_ivf(const char detect[4]) {
120 if (memcmp(detect, "DKIF", 4) == 0) {
126 static const arg_def_t help =
127 ARG_DEF(NULL, "help", 0, "Show usage options and exit");
128 static const arg_def_t debugmode =
129 ARG_DEF("D", "debug", 0, "Debug mode (makes output deterministic)");
130 static const arg_def_t outputfile =
131 ARG_DEF("o", "output", 1, "Output filename");
132 static const arg_def_t use_yv12 =
133 ARG_DEF(NULL, "yv12", 0, "Input file is YV12 ");
134 static const arg_def_t use_i420 =
135 ARG_DEF(NULL, "i420", 0, "Input file is I420 (default)");
136 static const arg_def_t use_i422 =
137 ARG_DEF(NULL, "i422", 0, "Input file is I422");
138 static const arg_def_t use_i444 =
139 ARG_DEF(NULL, "i444", 0, "Input file is I444");
140 static const arg_def_t use_i440 =
141 ARG_DEF(NULL, "i440", 0, "Input file is I440");
142 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, "Codec to use");
143 static const arg_def_t passes =
144 ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
145 static const arg_def_t pass_arg =
146 ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
147 static const arg_def_t fpf_name =
148 ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");
149 #if CONFIG_FP_MB_STATS
150 static const arg_def_t fpmbf_name =
151 ARG_DEF(NULL, "fpmbf", 1, "First pass block statistics file name");
153 static const arg_def_t limit =
154 ARG_DEF(NULL, "limit", 1, "Stop encoding after n input frames");
155 static const arg_def_t skip =
156 ARG_DEF(NULL, "skip", 1, "Skip the first n input frames");
157 static const arg_def_t deadline =
158 ARG_DEF("d", "deadline", 1, "Deadline per frame (usec)");
159 static const arg_def_t best_dl =
160 ARG_DEF(NULL, "best", 0, "Use Best Quality Deadline");
161 static const arg_def_t good_dl =
162 ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
163 static const arg_def_t rt_dl =
164 ARG_DEF(NULL, "rt", 0, "Use Realtime Quality Deadline");
165 static const arg_def_t quietarg =
166 ARG_DEF("q", "quiet", 0, "Do not print encode progress");
167 static const arg_def_t verbosearg =
168 ARG_DEF("v", "verbose", 0, "Show encoder parameters");
169 static const arg_def_t psnrarg =
170 ARG_DEF(NULL, "psnr", 0, "Show PSNR in status line");
172 static const struct arg_enum_list test_decode_enum[] = {
173 { "off", TEST_DECODE_OFF },
174 { "fatal", TEST_DECODE_FATAL },
175 { "warn", TEST_DECODE_WARN },
178 static const arg_def_t recontest = ARG_DEF_ENUM(
179 NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
180 static const arg_def_t framerate =
181 ARG_DEF(NULL, "fps", 1, "Stream frame rate (rate/scale)");
182 static const arg_def_t use_webm =
183 ARG_DEF(NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
184 static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, "Output IVF");
185 static const arg_def_t out_part =
186 ARG_DEF("P", "output-partitions", 0,
187 "Makes encoder output partitions. Requires IVF output!");
188 static const arg_def_t q_hist_n =
189 ARG_DEF(NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
190 static const arg_def_t rate_hist_n =
191 ARG_DEF(NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
192 static const arg_def_t disable_warnings =
193 ARG_DEF(NULL, "disable-warnings", 0,
194 "Disable warnings about potentially incorrect encode settings.");
195 static const arg_def_t disable_warning_prompt =
196 ARG_DEF("y", "disable-warning-prompt", 0,
197 "Display warnings, but do not prompt user to continue.");
199 #if CONFIG_VP9_HIGHBITDEPTH
200 static const arg_def_t test16bitinternalarg = ARG_DEF(
201 NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");
204 static const arg_def_t *main_args[] = { &help,
226 &disable_warning_prompt,
230 static const arg_def_t usage =
231 ARG_DEF("u", "usage", 1, "Usage profile number to use");
232 static const arg_def_t threads =
233 ARG_DEF("t", "threads", 1, "Max number of threads to use");
234 static const arg_def_t profile =
235 ARG_DEF(NULL, "profile", 1, "Bitstream profile number to use");
236 static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
237 static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
239 static const struct arg_enum_list stereo_mode_enum[] = {
240 { "mono", STEREO_FORMAT_MONO },
241 { "left-right", STEREO_FORMAT_LEFT_RIGHT },
242 { "bottom-top", STEREO_FORMAT_BOTTOM_TOP },
243 { "top-bottom", STEREO_FORMAT_TOP_BOTTOM },
244 { "right-left", STEREO_FORMAT_RIGHT_LEFT },
247 static const arg_def_t stereo_mode = ARG_DEF_ENUM(
248 NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
250 static const arg_def_t timebase = ARG_DEF(
251 NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
252 static const arg_def_t error_resilient =
253 ARG_DEF(NULL, "error-resilient", 1, "Enable error resiliency features");
254 static const arg_def_t lag_in_frames =
255 ARG_DEF(NULL, "lag-in-frames", 1, "Max number of frames to lag");
257 static const arg_def_t *global_args[] = { &use_yv12,
273 #if CONFIG_VP9_HIGHBITDEPTH
274 &test16bitinternalarg,
279 static const arg_def_t dropframe_thresh =
280 ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
281 static const arg_def_t resize_allowed =
282 ARG_DEF(NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
283 static const arg_def_t resize_width =
284 ARG_DEF(NULL, "resize-width", 1, "Width of encoded frame");
285 static const arg_def_t resize_height =
286 ARG_DEF(NULL, "resize-height", 1, "Height of encoded frame");
287 static const arg_def_t resize_up_thresh =
288 ARG_DEF(NULL, "resize-up", 1, "Upscale threshold (buf %)");
289 static const arg_def_t resize_down_thresh =
290 ARG_DEF(NULL, "resize-down", 1, "Downscale threshold (buf %)");
291 static const struct arg_enum_list end_usage_enum[] = { { "vbr", VPX_VBR },
296 static const arg_def_t end_usage =
297 ARG_DEF_ENUM(NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
298 static const arg_def_t target_bitrate =
299 ARG_DEF(NULL, "target-bitrate", 1, "Bitrate (kbps)");
300 static const arg_def_t min_quantizer =
301 ARG_DEF(NULL, "min-q", 1, "Minimum (best) quantizer");
302 static const arg_def_t max_quantizer =
303 ARG_DEF(NULL, "max-q", 1, "Maximum (worst) quantizer");
304 static const arg_def_t undershoot_pct =
305 ARG_DEF(NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
306 static const arg_def_t overshoot_pct =
307 ARG_DEF(NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
308 static const arg_def_t buf_sz =
309 ARG_DEF(NULL, "buf-sz", 1, "Client buffer size (ms)");
310 static const arg_def_t buf_initial_sz =
311 ARG_DEF(NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
312 static const arg_def_t buf_optimal_sz =
313 ARG_DEF(NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
314 static const arg_def_t *rc_args[] = {
315 &dropframe_thresh, &resize_allowed, &resize_width, &resize_height,
316 &resize_up_thresh, &resize_down_thresh, &end_usage, &target_bitrate,
317 &min_quantizer, &max_quantizer, &undershoot_pct, &overshoot_pct,
318 &buf_sz, &buf_initial_sz, &buf_optimal_sz, NULL
321 static const arg_def_t bias_pct =
322 ARG_DEF(NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
323 static const arg_def_t minsection_pct =
324 ARG_DEF(NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
325 static const arg_def_t maxsection_pct =
326 ARG_DEF(NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
327 static const arg_def_t corpus_complexity =
328 ARG_DEF(NULL, "corpus-complexity", 1, "corpus vbr complexity midpoint");
329 static const arg_def_t *rc_twopass_args[] = {
330 &bias_pct, &minsection_pct, &maxsection_pct, &corpus_complexity, NULL
333 static const arg_def_t kf_min_dist =
334 ARG_DEF(NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
335 static const arg_def_t kf_max_dist =
336 ARG_DEF(NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
337 static const arg_def_t kf_disabled =
338 ARG_DEF(NULL, "disable-kf", 0, "Disable keyframe placement");
339 static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
342 static const arg_def_t noise_sens =
343 ARG_DEF(NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
344 static const arg_def_t sharpness =
345 ARG_DEF(NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
346 static const arg_def_t static_thresh =
347 ARG_DEF(NULL, "static-thresh", 1, "Motion detection threshold");
348 static const arg_def_t auto_altref =
349 ARG_DEF(NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
350 static const arg_def_t arnr_maxframes =
351 ARG_DEF(NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
352 static const arg_def_t arnr_strength =
353 ARG_DEF(NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
354 static const arg_def_t arnr_type =
355 ARG_DEF(NULL, "arnr-type", 1, "AltRef filter type (1..3)");
356 static const struct arg_enum_list tuning_enum[] = {
357 { "psnr", VP8_TUNE_PSNR }, { "ssim", VP8_TUNE_SSIM }, { NULL, 0 }
359 static const arg_def_t tune_ssim =
360 ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
361 static const arg_def_t cq_level =
362 ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
363 static const arg_def_t max_intra_rate_pct =
364 ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
365 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
366 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
368 #if CONFIG_VP8_ENCODER
369 static const arg_def_t cpu_used_vp8 =
370 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
371 static const arg_def_t token_parts =
372 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
373 static const arg_def_t screen_content_mode =
374 ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
375 static const arg_def_t *vp8_args[] = { &cpu_used_vp8,
388 &screen_content_mode,
390 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
391 VP8E_SET_ENABLEAUTOALTREF,
392 VP8E_SET_NOISE_SENSITIVITY,
394 VP8E_SET_STATIC_THRESHOLD,
395 VP8E_SET_TOKEN_PARTITIONS,
396 VP8E_SET_ARNR_MAXFRAMES,
397 VP8E_SET_ARNR_STRENGTH,
401 VP8E_SET_MAX_INTRA_BITRATE_PCT,
402 VP8E_SET_GF_CBR_BOOST_PCT,
403 VP8E_SET_SCREEN_CONTENT_MODE,
407 #if CONFIG_VP9_ENCODER
408 static const arg_def_t cpu_used_vp9 =
409 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-8..8)");
410 static const arg_def_t tile_cols =
411 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
412 static const arg_def_t tile_rows =
413 ARG_DEF(NULL, "tile-rows", 1,
414 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
415 static const arg_def_t lossless =
416 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
417 static const arg_def_t frame_parallel_decoding = ARG_DEF(
418 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
419 static const arg_def_t aq_mode = ARG_DEF(
421 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
422 "3: cyclic refresh, 4: equator360)");
423 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
424 "Special adaptive quantization for "
425 "the alternate reference frames.");
426 static const arg_def_t frame_periodic_boost =
427 ARG_DEF(NULL, "frame-boost", 1,
428 "Enable frame periodic boost (0: off (default), 1: on)");
429 static const arg_def_t max_inter_rate_pct =
430 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
431 static const arg_def_t min_gf_interval = ARG_DEF(
432 NULL, "min-gf-interval", 1,
433 "min gf/arf frame interval (default 0, indicating in-built behavior)");
434 static const arg_def_t max_gf_interval = ARG_DEF(
435 NULL, "max-gf-interval", 1,
436 "max gf/arf frame interval (default 0, indicating in-built behavior)");
438 static const struct arg_enum_list color_space_enum[] = {
439 { "unknown", VPX_CS_UNKNOWN },
440 { "bt601", VPX_CS_BT_601 },
441 { "bt709", VPX_CS_BT_709 },
442 { "smpte170", VPX_CS_SMPTE_170 },
443 { "smpte240", VPX_CS_SMPTE_240 },
444 { "bt2020", VPX_CS_BT_2020 },
445 { "reserved", VPX_CS_RESERVED },
446 { "sRGB", VPX_CS_SRGB },
450 static const arg_def_t input_color_space =
451 ARG_DEF_ENUM(NULL, "color-space", 1,
452 "The color space of input content:", color_space_enum);
454 #if CONFIG_VP9_HIGHBITDEPTH
455 static const struct arg_enum_list bitdepth_enum[] = {
456 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
459 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
461 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
463 static const arg_def_t inbitdeptharg =
464 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
467 static const struct arg_enum_list tune_content_enum[] = {
468 { "default", VP9E_CONTENT_DEFAULT },
469 { "screen", VP9E_CONTENT_SCREEN },
470 { "film", VP9E_CONTENT_FILM },
474 static const arg_def_t tune_content = ARG_DEF_ENUM(
475 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
477 static const arg_def_t target_level = ARG_DEF(
478 NULL, "target-level", 1,
480 " 255: off (default)\n"
481 " 0: only keep level stats\n"
482 " 1: adaptively set alt-ref "
483 "distance and column tile limit based on picture size, and keep"
485 " 10: level 1.0 11: level 1.1 "
486 "... 62: level 6.2");
488 static const arg_def_t row_mt =
489 ARG_DEF(NULL, "row-mt", 1,
490 "Enable row based non-deterministic multi-threading in VP9");
493 #if CONFIG_VP9_ENCODER
494 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
509 &frame_parallel_decoding,
512 &frame_periodic_boost,
520 #if CONFIG_VP9_HIGHBITDEPTH
523 #endif // CONFIG_VP9_HIGHBITDEPTH
525 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
526 VP8E_SET_ENABLEAUTOALTREF,
528 VP8E_SET_STATIC_THRESHOLD,
529 VP9E_SET_TILE_COLUMNS,
531 VP8E_SET_ARNR_MAXFRAMES,
532 VP8E_SET_ARNR_STRENGTH,
536 VP8E_SET_MAX_INTRA_BITRATE_PCT,
537 VP9E_SET_MAX_INTER_BITRATE_PCT,
538 VP9E_SET_GF_CBR_BOOST_PCT,
540 VP9E_SET_FRAME_PARALLEL_DECODING,
543 VP9E_SET_FRAME_PERIODIC_BOOST,
544 VP9E_SET_NOISE_SENSITIVITY,
545 VP9E_SET_TUNE_CONTENT,
546 VP9E_SET_COLOR_SPACE,
547 VP9E_SET_MIN_GF_INTERVAL,
548 VP9E_SET_MAX_GF_INTERVAL,
549 VP9E_SET_TARGET_LEVEL,
554 static const arg_def_t *no_args[] = { NULL };
556 void show_help(FILE *fout, int shorthelp) {
558 const int num_encoder = get_vpx_encoder_count();
560 fprintf(fout, "Usage: %s <options> -o dst_filename src_filename \n",
564 fprintf(fout, "Use --help to see the full list of options.\n");
568 fprintf(fout, "\nOptions:\n");
569 arg_show_usage(fout, main_args);
570 fprintf(fout, "\nEncoder Global Options:\n");
571 arg_show_usage(fout, global_args);
572 fprintf(fout, "\nRate Control Options:\n");
573 arg_show_usage(fout, rc_args);
574 fprintf(fout, "\nTwopass Rate Control Options:\n");
575 arg_show_usage(fout, rc_twopass_args);
576 fprintf(fout, "\nKeyframe Placement Options:\n");
577 arg_show_usage(fout, kf_args);
578 #if CONFIG_VP8_ENCODER
579 fprintf(fout, "\nVP8 Specific Options:\n");
580 arg_show_usage(fout, vp8_args);
582 #if CONFIG_VP9_ENCODER
583 fprintf(fout, "\nVP9 Specific Options:\n");
584 arg_show_usage(fout, vp9_args);
587 "\nStream timebase (--timebase):\n"
588 " The desired precision of timestamps in the output, expressed\n"
589 " in fractional seconds. Default is 1/1000.\n");
590 fprintf(fout, "\nIncluded encoders:\n\n");
592 for (i = 0; i < num_encoder; ++i) {
593 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
594 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
595 fprintf(fout, " %-6s - %s %s\n", encoder->name,
596 vpx_codec_iface_name(encoder->codec_interface()), defstr);
598 fprintf(fout, "\n ");
599 fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
602 void usage_exit(void) {
603 show_help(stderr, 1);
607 #define mmin(a, b) ((a) < (b) ? (a) : (b))
609 #if CONFIG_VP9_HIGHBITDEPTH
610 static void find_mismatch_high(const vpx_image_t *const img1,
611 const vpx_image_t *const img2, int yloc[4],
612 int uloc[4], int vloc[4]) {
613 uint16_t *plane1, *plane2;
614 uint32_t stride1, stride2;
615 const uint32_t bsize = 64;
616 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
617 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
619 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
621 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
624 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
625 plane1 = (uint16_t *)img1->planes[VPX_PLANE_Y];
626 plane2 = (uint16_t *)img2->planes[VPX_PLANE_Y];
627 stride1 = img1->stride[VPX_PLANE_Y] / 2;
628 stride2 = img2->stride[VPX_PLANE_Y] / 2;
629 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
630 for (j = 0; match && j < img1->d_w; j += bsize) {
632 const int si = mmin(i + bsize, img1->d_h) - i;
633 const int sj = mmin(j + bsize, img1->d_w) - j;
634 for (k = 0; match && k < si; ++k) {
635 for (l = 0; match && l < sj; ++l) {
636 if (*(plane1 + (i + k) * stride1 + j + l) !=
637 *(plane2 + (i + k) * stride2 + j + l)) {
640 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
641 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
650 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
651 plane1 = (uint16_t *)img1->planes[VPX_PLANE_U];
652 plane2 = (uint16_t *)img2->planes[VPX_PLANE_U];
653 stride1 = img1->stride[VPX_PLANE_U] / 2;
654 stride2 = img2->stride[VPX_PLANE_U] / 2;
655 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
656 for (j = 0; match && j < c_w; j += bsizex) {
658 const int si = mmin(i + bsizey, c_h - i);
659 const int sj = mmin(j + bsizex, c_w - j);
660 for (k = 0; match && k < si; ++k) {
661 for (l = 0; match && l < sj; ++l) {
662 if (*(plane1 + (i + k) * stride1 + j + l) !=
663 *(plane2 + (i + k) * stride2 + j + l)) {
666 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
667 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
676 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
677 plane1 = (uint16_t *)img1->planes[VPX_PLANE_V];
678 plane2 = (uint16_t *)img2->planes[VPX_PLANE_V];
679 stride1 = img1->stride[VPX_PLANE_V] / 2;
680 stride2 = img2->stride[VPX_PLANE_V] / 2;
681 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
682 for (j = 0; match && j < c_w; j += bsizex) {
684 const int si = mmin(i + bsizey, c_h - i);
685 const int sj = mmin(j + bsizex, c_w - j);
686 for (k = 0; match && k < si; ++k) {
687 for (l = 0; match && l < sj; ++l) {
688 if (*(plane1 + (i + k) * stride1 + j + l) !=
689 *(plane2 + (i + k) * stride2 + j + l)) {
692 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
693 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
704 static void find_mismatch(const vpx_image_t *const img1,
705 const vpx_image_t *const img2, int yloc[4],
706 int uloc[4], int vloc[4]) {
707 const uint32_t bsize = 64;
708 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
709 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
711 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
713 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
716 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
717 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
718 for (j = 0; match && j < img1->d_w; j += bsize) {
720 const int si = mmin(i + bsize, img1->d_h) - i;
721 const int sj = mmin(j + bsize, img1->d_w) - j;
722 for (k = 0; match && k < si; ++k) {
723 for (l = 0; match && l < sj; ++l) {
724 if (*(img1->planes[VPX_PLANE_Y] +
725 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
726 *(img2->planes[VPX_PLANE_Y] +
727 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
730 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
731 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
732 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
733 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
742 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
743 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
744 for (j = 0; match && j < c_w; j += bsizex) {
746 const int si = mmin(i + bsizey, c_h - i);
747 const int sj = mmin(j + bsizex, c_w - j);
748 for (k = 0; match && k < si; ++k) {
749 for (l = 0; match && l < sj; ++l) {
750 if (*(img1->planes[VPX_PLANE_U] +
751 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
752 *(img2->planes[VPX_PLANE_U] +
753 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
756 uloc[2] = *(img1->planes[VPX_PLANE_U] +
757 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
758 uloc[3] = *(img2->planes[VPX_PLANE_U] +
759 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
767 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
768 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
769 for (j = 0; match && j < c_w; j += bsizex) {
771 const int si = mmin(i + bsizey, c_h - i);
772 const int sj = mmin(j + bsizex, c_w - j);
773 for (k = 0; match && k < si; ++k) {
774 for (l = 0; match && l < sj; ++l) {
775 if (*(img1->planes[VPX_PLANE_V] +
776 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
777 *(img2->planes[VPX_PLANE_V] +
778 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
781 vloc[2] = *(img1->planes[VPX_PLANE_V] +
782 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
783 vloc[3] = *(img2->planes[VPX_PLANE_V] +
784 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
794 static int compare_img(const vpx_image_t *const img1,
795 const vpx_image_t *const img2) {
796 uint32_t l_w = img1->d_w;
797 uint32_t c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
799 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
803 match &= (img1->fmt == img2->fmt);
804 match &= (img1->d_w == img2->d_w);
805 match &= (img1->d_h == img2->d_h);
806 #if CONFIG_VP9_HIGHBITDEPTH
807 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
813 for (i = 0; i < img1->d_h; ++i)
814 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
815 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
818 for (i = 0; i < c_h; ++i)
819 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
820 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
823 for (i = 0; i < c_h; ++i)
824 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
825 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
831 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
832 #if CONFIG_VP9_ENCODER
833 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
835 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
839 typedef int stereo_format_t;
840 struct WebmOutputContext {
845 /* Per-stream configuration */
846 struct stream_config {
847 struct vpx_codec_enc_cfg cfg;
849 const char *stats_fn;
850 #if CONFIG_FP_MB_STATS
851 const char *fpmb_stats_fn;
853 stereo_format_t stereo_fmt;
854 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
857 #if CONFIG_VP9_HIGHBITDEPTH
858 // whether to use 16bit internal buffers
859 int use_16bit_internal;
863 struct stream_state {
865 struct stream_state *next;
866 struct stream_config config;
868 struct rate_hist *rate_hist;
869 struct WebmOutputContext webm_ctx;
870 uint64_t psnr_sse_total;
871 uint64_t psnr_samples_total;
872 double psnr_totals[4];
875 vpx_codec_ctx_t encoder;
876 unsigned int frames_out;
880 #if CONFIG_FP_MB_STATS
881 stats_io_t fpmb_stats;
883 struct vpx_image *img;
884 vpx_codec_ctx_t decoder;
888 static void validate_positive_rational(const char *msg,
889 struct vpx_rational *rat) {
895 if (rat->num < 0) die("Error: %s must be positive\n", msg);
897 if (!rat->den) die("Error: %s has zero denominator\n", msg);
900 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
903 const int num_encoder = get_vpx_encoder_count();
905 if (num_encoder < 1) die("Error: no valid encoder available\n");
907 /* Initialize default parameters */
908 memset(global, 0, sizeof(*global));
909 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
911 global->color_type = I420;
912 /* Assign default deadline to good quality */
913 global->deadline = VPX_DL_GOOD_QUALITY;
915 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
918 if (arg_match(&arg, &help, argi)) {
919 show_help(stdout, 0);
921 } else if (arg_match(&arg, &codecarg, argi)) {
922 global->codec = get_vpx_encoder_by_name(arg.val);
924 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
925 } else if (arg_match(&arg, &passes, argi)) {
926 global->passes = arg_parse_uint(&arg);
928 if (global->passes < 1 || global->passes > 2)
929 die("Error: Invalid number of passes (%d)\n", global->passes);
930 } else if (arg_match(&arg, &pass_arg, argi)) {
931 global->pass = arg_parse_uint(&arg);
933 if (global->pass < 1 || global->pass > 2)
934 die("Error: Invalid pass selected (%d)\n", global->pass);
935 } else if (arg_match(&arg, &usage, argi))
936 global->usage = arg_parse_uint(&arg);
937 else if (arg_match(&arg, &deadline, argi))
938 global->deadline = arg_parse_uint(&arg);
939 else if (arg_match(&arg, &best_dl, argi))
940 global->deadline = VPX_DL_BEST_QUALITY;
941 else if (arg_match(&arg, &good_dl, argi))
942 global->deadline = VPX_DL_GOOD_QUALITY;
943 else if (arg_match(&arg, &rt_dl, argi))
944 global->deadline = VPX_DL_REALTIME;
945 else if (arg_match(&arg, &use_yv12, argi))
946 global->color_type = YV12;
947 else if (arg_match(&arg, &use_i420, argi))
948 global->color_type = I420;
949 else if (arg_match(&arg, &use_i422, argi))
950 global->color_type = I422;
951 else if (arg_match(&arg, &use_i444, argi))
952 global->color_type = I444;
953 else if (arg_match(&arg, &use_i440, argi))
954 global->color_type = I440;
955 else if (arg_match(&arg, &quietarg, argi))
957 else if (arg_match(&arg, &verbosearg, argi))
959 else if (arg_match(&arg, &limit, argi))
960 global->limit = arg_parse_uint(&arg);
961 else if (arg_match(&arg, &skip, argi))
962 global->skip_frames = arg_parse_uint(&arg);
963 else if (arg_match(&arg, &psnrarg, argi))
964 global->show_psnr = 1;
965 else if (arg_match(&arg, &recontest, argi))
966 global->test_decode = arg_parse_enum_or_int(&arg);
967 else if (arg_match(&arg, &framerate, argi)) {
968 global->framerate = arg_parse_rational(&arg);
969 validate_positive_rational(arg.name, &global->framerate);
970 global->have_framerate = 1;
971 } else if (arg_match(&arg, &out_part, argi))
972 global->out_part = 1;
973 else if (arg_match(&arg, &debugmode, argi))
975 else if (arg_match(&arg, &q_hist_n, argi))
976 global->show_q_hist_buckets = arg_parse_uint(&arg);
977 else if (arg_match(&arg, &rate_hist_n, argi))
978 global->show_rate_hist_buckets = arg_parse_uint(&arg);
979 else if (arg_match(&arg, &disable_warnings, argi))
980 global->disable_warnings = 1;
981 else if (arg_match(&arg, &disable_warning_prompt, argi))
982 global->disable_warning_prompt = 1;
988 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
989 if (global->pass > global->passes) {
990 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
992 global->passes = global->pass;
995 /* Validate global config */
996 if (global->passes == 0) {
997 #if CONFIG_VP9_ENCODER
998 // Make default VP9 passes = 2 until there is a better quality 1-pass
1000 if (global->codec != NULL && global->codec->name != NULL)
1001 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
1002 global->deadline != VPX_DL_REALTIME)
1010 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
1011 warn("Enforcing one-pass encoding in realtime mode\n");
1016 static void open_input_file(struct VpxInputContext *input) {
1017 /* Parse certain options from the input file, if possible */
1018 input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
1019 : set_binary_mode(stdin);
1021 if (!input->file) fatal("Failed to open input file");
1023 if (!fseeko(input->file, 0, SEEK_END)) {
1024 /* Input file is seekable. Figure out how long it is, so we can get
1027 input->length = ftello(input->file);
1028 rewind(input->file);
1031 /* Default to 1:1 pixel aspect ratio. */
1032 input->pixel_aspect_ratio.numerator = 1;
1033 input->pixel_aspect_ratio.denominator = 1;
1035 /* For RAW input sources, these bytes will applied on the first frame
1038 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1039 input->detect.position = 0;
1041 if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
1042 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1043 input->only_i420) >= 0) {
1044 input->file_type = FILE_TYPE_Y4M;
1045 input->width = input->y4m.pic_w;
1046 input->height = input->y4m.pic_h;
1047 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1048 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1049 input->framerate.numerator = input->y4m.fps_n;
1050 input->framerate.denominator = input->y4m.fps_d;
1051 input->fmt = input->y4m.vpx_fmt;
1052 input->bit_depth = input->y4m.bit_depth;
1054 fatal("Unsupported Y4M stream.");
1055 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1056 fatal("IVF is not supported as input.");
1058 input->file_type = FILE_TYPE_RAW;
1062 static void close_input_file(struct VpxInputContext *input) {
1063 fclose(input->file);
1064 if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
1067 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1068 struct stream_state *prev) {
1069 struct stream_state *stream;
1071 stream = calloc(1, sizeof(*stream));
1072 if (stream == NULL) {
1073 fatal("Failed to allocate new stream.");
1077 memcpy(stream, prev, sizeof(*stream));
1079 prev->next = stream;
1081 vpx_codec_err_t res;
1083 /* Populate encoder configuration */
1084 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1085 &stream->config.cfg, global->usage);
1086 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1088 /* Change the default timebase to a high enough value so that the
1089 * encoder will always create strictly increasing timestamps.
1091 stream->config.cfg.g_timebase.den = 1000;
1093 /* Never use the library's default resolution, require it be parsed
1094 * from the file or set on the command line.
1096 stream->config.cfg.g_w = 0;
1097 stream->config.cfg.g_h = 0;
1099 /* Initialize remaining stream parameters */
1100 stream->config.write_webm = 1;
1102 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1103 stream->webm_ctx.last_pts_ns = -1;
1104 stream->webm_ctx.writer = NULL;
1105 stream->webm_ctx.segment = NULL;
1108 /* Allows removal of the application version from the EBML tags */
1109 stream->webm_ctx.debug = global->debug;
1111 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
1112 if (global->deadline == VPX_DL_REALTIME &&
1113 stream->config.cfg.rc_end_usage == 1)
1114 stream->config.cfg.g_lag_in_frames = 0;
1117 /* Output files must be specified for each stream */
1118 stream->config.out_fn = NULL;
1120 stream->next = NULL;
1124 static int parse_stream_params(struct VpxEncoderConfig *global,
1125 struct stream_state *stream, char **argv) {
1126 char **argi, **argj;
1128 static const arg_def_t **ctrl_args = no_args;
1129 static const int *ctrl_args_map = NULL;
1130 struct stream_config *config = &stream->config;
1131 int eos_mark_found = 0;
1132 #if CONFIG_VP9_HIGHBITDEPTH
1133 int test_16bit_internal = 0;
1136 // Handle codec specific options
1138 #if CONFIG_VP8_ENCODER
1139 } else if (strcmp(global->codec->name, "vp8") == 0) {
1140 ctrl_args = vp8_args;
1141 ctrl_args_map = vp8_arg_ctrl_map;
1143 #if CONFIG_VP9_ENCODER
1144 } else if (strcmp(global->codec->name, "vp9") == 0) {
1145 ctrl_args = vp9_args;
1146 ctrl_args_map = vp9_arg_ctrl_map;
1150 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1153 /* Once we've found an end-of-stream marker (--) we want to continue
1154 * shifting arguments but not consuming them.
1156 if (eos_mark_found) {
1159 } else if (!strcmp(*argj, "--")) {
1164 if (arg_match(&arg, &outputfile, argi)) {
1165 config->out_fn = arg.val;
1166 } else if (arg_match(&arg, &fpf_name, argi)) {
1167 config->stats_fn = arg.val;
1168 #if CONFIG_FP_MB_STATS
1169 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1170 config->fpmb_stats_fn = arg.val;
1172 } else if (arg_match(&arg, &use_webm, argi)) {
1174 config->write_webm = 1;
1176 die("Error: --webm specified but webm is disabled.");
1178 } else if (arg_match(&arg, &use_ivf, argi)) {
1179 config->write_webm = 0;
1180 } else if (arg_match(&arg, &threads, argi)) {
1181 config->cfg.g_threads = arg_parse_uint(&arg);
1182 } else if (arg_match(&arg, &profile, argi)) {
1183 config->cfg.g_profile = arg_parse_uint(&arg);
1184 } else if (arg_match(&arg, &width, argi)) {
1185 config->cfg.g_w = arg_parse_uint(&arg);
1186 } else if (arg_match(&arg, &height, argi)) {
1187 config->cfg.g_h = arg_parse_uint(&arg);
1188 #if CONFIG_VP9_HIGHBITDEPTH
1189 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1190 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1191 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1192 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1195 } else if (arg_match(&arg, &stereo_mode, argi)) {
1196 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1198 } else if (arg_match(&arg, &timebase, argi)) {
1199 config->cfg.g_timebase = arg_parse_rational(&arg);
1200 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1201 } else if (arg_match(&arg, &error_resilient, argi)) {
1202 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1203 } else if (arg_match(&arg, &end_usage, argi)) {
1204 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1205 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1206 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1207 if (global->deadline == VPX_DL_REALTIME &&
1208 config->cfg.rc_end_usage == VPX_CBR &&
1209 config->cfg.g_lag_in_frames != 0) {
1210 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
1211 config->cfg.g_lag_in_frames = 0;
1213 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1214 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1215 } else if (arg_match(&arg, &resize_allowed, argi)) {
1216 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1217 } else if (arg_match(&arg, &resize_width, argi)) {
1218 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1219 } else if (arg_match(&arg, &resize_height, argi)) {
1220 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1221 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1222 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1223 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1224 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1225 } else if (arg_match(&arg, &end_usage, argi)) {
1226 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1227 } else if (arg_match(&arg, &target_bitrate, argi)) {
1228 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1229 } else if (arg_match(&arg, &min_quantizer, argi)) {
1230 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1231 } else if (arg_match(&arg, &max_quantizer, argi)) {
1232 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1233 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1234 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1235 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1236 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1237 } else if (arg_match(&arg, &buf_sz, argi)) {
1238 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1239 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1240 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1241 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1242 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1243 } else if (arg_match(&arg, &bias_pct, argi)) {
1244 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1245 if (global->passes < 2)
1246 warn("option %s ignored in one-pass mode.\n", arg.name);
1247 } else if (arg_match(&arg, &minsection_pct, argi)) {
1248 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1250 if (global->passes < 2)
1251 warn("option %s ignored in one-pass mode.\n", arg.name);
1252 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1253 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1255 if (global->passes < 2)
1256 warn("option %s ignored in one-pass mode.\n", arg.name);
1257 } else if (arg_match(&arg, &corpus_complexity, argi)) {
1258 config->cfg.rc_2pass_vbr_corpus_complexity = arg_parse_uint(&arg);
1260 if (global->passes < 2)
1261 warn("option %s ignored in one-pass mode.\n", arg.name);
1262 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1263 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1264 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1265 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1266 } else if (arg_match(&arg, &kf_disabled, argi)) {
1267 config->cfg.kf_mode = VPX_KF_DISABLED;
1268 #if CONFIG_VP9_HIGHBITDEPTH
1269 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1270 if (strcmp(global->codec->name, "vp9") == 0) {
1271 test_16bit_internal = 1;
1276 for (i = 0; ctrl_args[i]; i++) {
1277 if (arg_match(&arg, ctrl_args[i], argi)) {
1281 /* Point either to the next free element or the first
1282 * instance of this control.
1284 for (j = 0; j < config->arg_ctrl_cnt; j++)
1285 if (ctrl_args_map != NULL &&
1286 config->arg_ctrls[j][0] == ctrl_args_map[i])
1290 assert(j < (int)ARG_CTRL_CNT_MAX);
1291 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1292 config->arg_ctrls[j][0] = ctrl_args_map[i];
1293 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1294 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1301 #if CONFIG_VP9_HIGHBITDEPTH
1302 if (strcmp(global->codec->name, "vp9") == 0) {
1303 config->use_16bit_internal =
1304 test_16bit_internal | (config->cfg.g_profile > 1);
1307 return eos_mark_found;
1310 #define FOREACH_STREAM(func) \
1312 struct stream_state *stream; \
1313 for (stream = streams; stream; stream = stream->next) { \
1318 static void validate_stream_config(const struct stream_state *stream,
1319 const struct VpxEncoderConfig *global) {
1320 const struct stream_state *streami;
1323 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1325 "Stream %d: Specify stream dimensions with --width (-w) "
1326 " and --height (-h)",
1329 // Check that the codec bit depth is greater than the input bit depth.
1330 if (stream->config.cfg.g_input_bit_depth >
1331 (unsigned int)stream->config.cfg.g_bit_depth) {
1332 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1333 stream->index, (int)stream->config.cfg.g_bit_depth,
1334 stream->config.cfg.g_input_bit_depth);
1337 for (streami = stream; streami; streami = streami->next) {
1338 /* All streams require output files */
1339 if (!streami->config.out_fn)
1340 fatal("Stream %d: Output file is required (specify with -o)",
1343 /* Check for two streams outputting to the same file */
1344 if (streami != stream) {
1345 const char *a = stream->config.out_fn;
1346 const char *b = streami->config.out_fn;
1347 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1348 fatal("Stream %d: duplicate output file (from stream %d)",
1349 streami->index, stream->index);
1352 /* Check for two streams sharing a stats file. */
1353 if (streami != stream) {
1354 const char *a = stream->config.stats_fn;
1355 const char *b = streami->config.stats_fn;
1356 if (a && b && !strcmp(a, b))
1357 fatal("Stream %d: duplicate stats file (from stream %d)",
1358 streami->index, stream->index);
1361 #if CONFIG_FP_MB_STATS
1362 /* Check for two streams sharing a mb stats file. */
1363 if (streami != stream) {
1364 const char *a = stream->config.fpmb_stats_fn;
1365 const char *b = streami->config.fpmb_stats_fn;
1366 if (a && b && !strcmp(a, b))
1367 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1368 streami->index, stream->index);
1374 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1376 if (!stream->config.cfg.g_w) {
1377 if (!stream->config.cfg.g_h)
1378 stream->config.cfg.g_w = w;
1380 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1382 if (!stream->config.cfg.g_h) {
1383 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1387 static const char *file_type_to_string(enum VideoFileType t) {
1389 case FILE_TYPE_RAW: return "RAW";
1390 case FILE_TYPE_Y4M: return "Y4M";
1391 default: return "Other";
1395 static const char *image_format_to_string(vpx_img_fmt_t f) {
1397 case VPX_IMG_FMT_I420: return "I420";
1398 case VPX_IMG_FMT_I422: return "I422";
1399 case VPX_IMG_FMT_I444: return "I444";
1400 case VPX_IMG_FMT_I440: return "I440";
1401 case VPX_IMG_FMT_YV12: return "YV12";
1402 case VPX_IMG_FMT_I42016: return "I42016";
1403 case VPX_IMG_FMT_I42216: return "I42216";
1404 case VPX_IMG_FMT_I44416: return "I44416";
1405 case VPX_IMG_FMT_I44016: return "I44016";
1406 default: return "Other";
1410 static void show_stream_config(struct stream_state *stream,
1411 struct VpxEncoderConfig *global,
1412 struct VpxInputContext *input) {
1413 #define SHOW(field) \
1414 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1416 if (stream->index == 0) {
1417 fprintf(stderr, "Codec: %s\n",
1418 vpx_codec_iface_name(global->codec->codec_interface()));
1419 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1420 input->filename, file_type_to_string(input->file_type),
1421 image_format_to_string(input->fmt));
1423 if (stream->next || stream->index)
1424 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1425 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1426 fprintf(stderr, "Encoder parameters:\n");
1434 SHOW(g_input_bit_depth);
1435 SHOW(g_timebase.num);
1436 SHOW(g_timebase.den);
1437 SHOW(g_error_resilient);
1439 SHOW(g_lag_in_frames);
1440 SHOW(rc_dropframe_thresh);
1441 SHOW(rc_resize_allowed);
1442 SHOW(rc_scaled_width);
1443 SHOW(rc_scaled_height);
1444 SHOW(rc_resize_up_thresh);
1445 SHOW(rc_resize_down_thresh);
1447 SHOW(rc_target_bitrate);
1448 SHOW(rc_min_quantizer);
1449 SHOW(rc_max_quantizer);
1450 SHOW(rc_undershoot_pct);
1451 SHOW(rc_overshoot_pct);
1453 SHOW(rc_buf_initial_sz);
1454 SHOW(rc_buf_optimal_sz);
1455 SHOW(rc_2pass_vbr_bias_pct);
1456 SHOW(rc_2pass_vbr_minsection_pct);
1457 SHOW(rc_2pass_vbr_maxsection_pct);
1458 SHOW(rc_2pass_vbr_corpus_complexity);
1464 static void open_output_file(struct stream_state *stream,
1465 struct VpxEncoderConfig *global,
1466 const struct VpxRational *pixel_aspect_ratio) {
1467 const char *fn = stream->config.out_fn;
1468 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1470 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1472 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1474 if (!stream->file) fatal("Failed to open output file");
1476 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1477 fatal("WebM output to pipes not supported.");
1480 if (stream->config.write_webm) {
1481 stream->webm_ctx.stream = stream->file;
1482 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1483 global->codec->fourcc, pixel_aspect_ratio);
1486 (void)pixel_aspect_ratio;
1489 if (!stream->config.write_webm) {
1490 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1494 static void close_output_file(struct stream_state *stream,
1495 unsigned int fourcc) {
1496 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1498 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1501 if (stream->config.write_webm) {
1502 write_webm_file_footer(&stream->webm_ctx);
1506 if (!stream->config.write_webm) {
1507 if (!fseek(stream->file, 0, SEEK_SET))
1508 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1509 stream->frames_out);
1512 fclose(stream->file);
1515 static void setup_pass(struct stream_state *stream,
1516 struct VpxEncoderConfig *global, int pass) {
1517 if (stream->config.stats_fn) {
1518 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1519 fatal("Failed to open statistics store");
1521 if (!stats_open_mem(&stream->stats, pass))
1522 fatal("Failed to open statistics store");
1525 #if CONFIG_FP_MB_STATS
1526 if (stream->config.fpmb_stats_fn) {
1527 if (!stats_open_file(&stream->fpmb_stats, stream->config.fpmb_stats_fn,
1529 fatal("Failed to open mb statistics store");
1531 if (!stats_open_mem(&stream->fpmb_stats, pass))
1532 fatal("Failed to open mb statistics store");
1536 stream->config.cfg.g_pass = global->passes == 2
1537 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1540 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1541 #if CONFIG_FP_MB_STATS
1542 stream->config.cfg.rc_firstpass_mb_stats_in =
1543 stats_get(&stream->fpmb_stats);
1547 stream->cx_time = 0;
1549 stream->frames_out = 0;
1552 static void initialize_encoder(struct stream_state *stream,
1553 struct VpxEncoderConfig *global) {
1557 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1558 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1559 #if CONFIG_VP9_HIGHBITDEPTH
1560 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1563 /* Construct Encoder Context */
1564 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1565 &stream->config.cfg, flags);
1566 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1568 /* Note that we bypass the vpx_codec_control wrapper macro because
1569 * we're being clever to store the control IDs in an array. Real
1570 * applications will want to make use of the enumerations directly
1572 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1573 int ctrl = stream->config.arg_ctrls[i][0];
1574 int value = stream->config.arg_ctrls[i][1];
1575 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1576 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1578 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1582 if (global->test_decode != TEST_DECODE_OFF) {
1583 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1584 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1589 static void encode_frame(struct stream_state *stream,
1590 struct VpxEncoderConfig *global, struct vpx_image *img,
1591 unsigned int frames_in) {
1592 vpx_codec_pts_t frame_start, next_frame_start;
1593 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1594 struct vpx_usec_timer timer;
1597 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1598 cfg->g_timebase.num / global->framerate.num;
1600 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1601 cfg->g_timebase.num / global->framerate.num;
1603 /* Scale if necessary */
1604 #if CONFIG_VP9_HIGHBITDEPTH
1606 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1607 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1608 if (img->fmt != VPX_IMG_FMT_I42016) {
1609 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1615 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1618 (uint16 *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1619 (uint16 *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1620 (uint16 *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1621 img->d_w, img->d_h, (uint16 *)stream->img->planes[VPX_PLANE_Y],
1622 stream->img->stride[VPX_PLANE_Y] / 2,
1623 (uint16 *)stream->img->planes[VPX_PLANE_U],
1624 stream->img->stride[VPX_PLANE_U] / 2,
1625 (uint16 *)stream->img->planes[VPX_PLANE_V],
1626 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1627 stream->img->d_h, kFilterBox);
1630 stream->encoder.err = 1;
1631 ctx_exit_on_error(&stream->encoder,
1632 "Stream %d: Failed to encode frame.\n"
1633 "Scaling disabled in this configuration. \n"
1634 "To enable, configure with --enable-libyuv\n",
1640 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1641 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1642 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1648 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1650 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1651 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1652 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1653 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1654 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1655 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1656 stream->img->d_w, stream->img->d_h, kFilterBox);
1659 stream->encoder.err = 1;
1660 ctx_exit_on_error(&stream->encoder,
1661 "Stream %d: Failed to encode frame.\n"
1662 "Scaling disabled in this configuration. \n"
1663 "To enable, configure with --enable-libyuv\n",
1668 vpx_usec_timer_start(&timer);
1669 vpx_codec_encode(&stream->encoder, img, frame_start,
1670 (unsigned long)(next_frame_start - frame_start), 0,
1672 vpx_usec_timer_mark(&timer);
1673 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1674 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1678 static void update_quantizer_histogram(struct stream_state *stream) {
1679 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1682 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1683 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1684 stream->counts[q]++;
1688 static void get_cx_data(struct stream_state *stream,
1689 struct VpxEncoderConfig *global, int *got_data) {
1690 const vpx_codec_cx_pkt_t *pkt;
1691 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1692 vpx_codec_iter_t iter = NULL;
1695 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1696 static size_t fsize = 0;
1697 static FileOffset ivf_header_pos = 0;
1699 switch (pkt->kind) {
1700 case VPX_CODEC_CX_FRAME_PKT:
1701 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1702 stream->frames_out++;
1705 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1707 update_rate_histogram(stream->rate_hist, cfg, pkt);
1709 if (stream->config.write_webm) {
1710 write_webm_block(&stream->webm_ctx, cfg, pkt);
1713 if (!stream->config.write_webm) {
1714 if (pkt->data.frame.partition_id <= 0) {
1715 ivf_header_pos = ftello(stream->file);
1716 fsize = pkt->data.frame.sz;
1718 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1720 fsize += pkt->data.frame.sz;
1722 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1723 const FileOffset currpos = ftello(stream->file);
1724 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1725 ivf_write_frame_size(stream->file, fsize);
1726 fseeko(stream->file, currpos, SEEK_SET);
1730 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1733 stream->nbytes += pkt->data.raw.sz;
1737 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1738 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1739 (unsigned int)pkt->data.frame.sz, NULL, 0);
1740 if (stream->decoder.err) {
1741 warn_or_exit_on_error(&stream->decoder,
1742 global->test_decode == TEST_DECODE_FATAL,
1743 "Failed to decode frame %d in stream %d",
1744 stream->frames_out + 1, stream->index);
1745 stream->mismatch_seen = stream->frames_out + 1;
1750 case VPX_CODEC_STATS_PKT:
1751 stream->frames_out++;
1752 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1753 pkt->data.twopass_stats.sz);
1754 stream->nbytes += pkt->data.raw.sz;
1756 #if CONFIG_FP_MB_STATS
1757 case VPX_CODEC_FPMB_STATS_PKT:
1758 stats_write(&stream->fpmb_stats, pkt->data.firstpass_mb_stats.buf,
1759 pkt->data.firstpass_mb_stats.sz);
1760 stream->nbytes += pkt->data.raw.sz;
1763 case VPX_CODEC_PSNR_PKT:
1765 if (global->show_psnr) {
1768 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1769 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1770 for (i = 0; i < 4; i++) {
1772 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1773 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1775 stream->psnr_count++;
1784 static void show_psnr(struct stream_state *stream, double peak) {
1788 if (!stream->psnr_count) return;
1790 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1791 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1792 (double)stream->psnr_sse_total);
1793 fprintf(stderr, " %.3f", ovpsnr);
1795 for (i = 0; i < 4; i++) {
1796 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1798 fprintf(stderr, "\n");
1801 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1802 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1805 static void test_decode(struct stream_state *stream,
1806 enum TestDecodeFatality fatal,
1807 const VpxInterface *codec) {
1808 vpx_image_t enc_img, dec_img;
1810 if (stream->mismatch_seen) return;
1812 /* Get the internal reference frame */
1813 if (strcmp(codec->name, "vp8") == 0) {
1814 struct vpx_ref_frame ref_enc, ref_dec;
1817 width = (stream->config.cfg.g_w + 15) & ~15;
1818 height = (stream->config.cfg.g_h + 15) & ~15;
1819 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1820 enc_img = ref_enc.img;
1821 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1822 dec_img = ref_dec.img;
1824 ref_enc.frame_type = VP8_LAST_FRAME;
1825 ref_dec.frame_type = VP8_LAST_FRAME;
1826 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1827 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1829 struct vp9_ref_frame ref_enc, ref_dec;
1833 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1834 enc_img = ref_enc.img;
1835 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1836 dec_img = ref_dec.img;
1837 #if CONFIG_VP9_HIGHBITDEPTH
1838 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1839 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1840 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1841 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1842 enc_img.d_w, enc_img.d_h, 16);
1843 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1845 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1846 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1847 dec_img.d_w, dec_img.d_h, 16);
1848 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1853 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1854 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1856 if (!compare_img(&enc_img, &dec_img)) {
1857 int y[4], u[4], v[4];
1858 #if CONFIG_VP9_HIGHBITDEPTH
1859 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1860 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1862 find_mismatch(&enc_img, &dec_img, y, u, v);
1865 find_mismatch(&enc_img, &dec_img, y, u, v);
1867 stream->decoder.err = 1;
1868 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1869 "Stream %d: Encode/decode mismatch on frame %d at"
1870 " Y[%d, %d] {%d/%d},"
1871 " U[%d, %d] {%d/%d},"
1872 " V[%d, %d] {%d/%d}",
1873 stream->index, stream->frames_out, y[0], y[1], y[2],
1874 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1875 stream->mismatch_seen = stream->frames_out;
1878 vpx_img_free(&enc_img);
1879 vpx_img_free(&dec_img);
1882 static void print_time(const char *label, int64_t etl) {
1889 etl -= hours * 3600;
1894 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1897 fprintf(stderr, "[%3s unknown] ", label);
1901 int main(int argc, const char **argv_) {
1904 #if CONFIG_VP9_HIGHBITDEPTH
1905 vpx_image_t raw_shift;
1906 int allocated_raw_shift = 0;
1907 int use_16bit_internal = 0;
1908 int input_shift = 0;
1910 int frame_avail, got_data;
1912 struct VpxInputContext input;
1913 struct VpxEncoderConfig global;
1914 struct stream_state *streams = NULL;
1915 char **argv, **argi;
1916 uint64_t cx_time = 0;
1920 memset(&input, 0, sizeof(input));
1921 exec_name = argv_[0];
1923 /* Setup default input stream settings */
1924 input.framerate.numerator = 30;
1925 input.framerate.denominator = 1;
1926 input.only_i420 = 1;
1927 input.bit_depth = 0;
1929 /* First parse the global configuration values, because we want to apply
1930 * other parameters on top of the default configuration provided by the
1933 argv = argv_dup(argc - 1, argv_ + 1);
1934 parse_global_config(&global, argv);
1936 if (argc < 3) usage_exit();
1938 switch (global.color_type) {
1939 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1940 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1941 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1942 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1943 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1947 /* Now parse each stream's parameters. Using a local scope here
1948 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1951 struct stream_state *stream = NULL;
1954 stream = new_stream(&global, stream);
1956 if (!streams) streams = stream;
1957 } while (parse_stream_params(&global, stream, argv));
1960 /* Check for unrecognized options */
1961 for (argi = argv; *argi; argi++)
1962 if (argi[0][0] == '-' && argi[0][1])
1963 die("Error: Unrecognized option %s\n", *argi);
1965 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1966 &stream->config.cfg););
1968 /* Handle non-option arguments */
1969 input.filename = argv[0];
1971 if (!input.filename) {
1972 fprintf(stderr, "No input file specified!\n");
1976 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1977 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1979 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1980 int frames_in = 0, seen_frames = 0;
1981 int64_t estimated_time_left = -1;
1982 int64_t average_rate = -1;
1983 int64_t lagged_count = 0;
1985 open_input_file(&input);
1987 /* If the input file doesn't specify its w/h (raw files), try to get
1988 * the data from the first stream's configuration.
1990 if (!input.width || !input.height) {
1992 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1993 input.width = stream->config.cfg.g_w;
1994 input.height = stream->config.cfg.g_h;
2000 /* Update stream configurations from the input file's parameters */
2001 if (!input.width || !input.height)
2003 "Specify stream dimensions with --width (-w) "
2004 " and --height (-h)");
2006 /* If input file does not specify bit-depth but input-bit-depth parameter
2007 * exists, assume that to be the input bit-depth. However, if the
2008 * input-bit-depth paramter does not exist, assume the input bit-depth
2009 * to be the same as the codec bit-depth.
2011 if (!input.bit_depth) {
2013 if (stream->config.cfg.g_input_bit_depth)
2014 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2016 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2017 (int)stream->config.cfg.g_bit_depth;
2019 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2022 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
2025 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2026 FOREACH_STREAM(validate_stream_config(stream, &global));
2028 /* Ensure that --passes and --pass are consistent. If --pass is set and
2029 * --passes=2, ensure --fpf was set.
2031 if (global.pass && global.passes == 2)
2033 if (!stream->config.stats_fn)
2034 die("Stream %d: Must specify --fpf when --pass=%d"
2035 " and --passes=2\n",
2036 stream->index, global.pass);
2041 if (stream->config.write_webm) {
2042 stream->config.write_webm = 0;
2044 "vpxenc was compiled without WebM container support."
2045 "Producing IVF output");
2050 /* Use the frame rate from the file only if none was specified
2051 * on the command-line.
2053 if (!global.have_framerate) {
2054 global.framerate.num = input.framerate.numerator;
2055 global.framerate.den = input.framerate.denominator;
2056 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2057 stream->config.cfg.g_timebase.num = global.framerate.den);
2060 /* Show configuration */
2061 if (global.verbose && pass == 0)
2062 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2064 if (pass == (global.pass ? global.pass - 1 : 0)) {
2065 if (input.file_type == FILE_TYPE_Y4M)
2066 /*The Y4M reader does its own allocation.
2067 Just initialize this here to avoid problems if we never read any
2069 memset(&raw, 0, sizeof(raw));
2071 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2073 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
2074 &stream->config.cfg, &global.framerate));
2077 FOREACH_STREAM(setup_pass(stream, &global, pass));
2079 open_output_file(stream, &global, &input.pixel_aspect_ratio));
2080 FOREACH_STREAM(initialize_encoder(stream, &global));
2082 #if CONFIG_VP9_HIGHBITDEPTH
2083 if (strcmp(global.codec->name, "vp9") == 0) {
2084 // Check to see if at least one stream uses 16 bit internal.
2085 // Currently assume that the bit_depths for all streams using
2086 // highbitdepth are the same.
2088 if (stream->config.use_16bit_internal) {
2089 use_16bit_internal = 1;
2091 if (stream->config.cfg.g_profile == 0) {
2094 input_shift = (int)stream->config.cfg.g_bit_depth -
2095 stream->config.cfg.g_input_bit_depth;
2104 while (frame_avail || got_data) {
2105 struct vpx_usec_timer timer;
2107 if (!global.limit || frames_in < global.limit) {
2108 frame_avail = read_frame(&input, &raw);
2110 if (frame_avail) frames_in++;
2112 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
2114 if (!global.quiet) {
2115 float fps = usec_to_fps(cx_time, seen_frames);
2116 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2118 if (stream_cnt == 1)
2119 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
2120 streams->frames_out, (int64_t)streams->nbytes);
2122 fprintf(stderr, "frame %4d ", frames_in);
2124 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
2125 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2126 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
2127 fps >= 1.0 ? "fps" : "fpm");
2128 print_time("ETA", estimated_time_left);
2134 if (frames_in > global.skip_frames) {
2135 #if CONFIG_VP9_HIGHBITDEPTH
2136 vpx_image_t *frame_to_encode;
2137 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2138 assert(use_16bit_internal);
2139 // Input bit depth and stream bit depth do not match, so up
2140 // shift frame to stream bit depth
2141 if (!allocated_raw_shift) {
2142 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2143 input.width, input.height, 32);
2144 allocated_raw_shift = 1;
2146 vpx_img_upshift(&raw_shift, &raw, input_shift);
2147 frame_to_encode = &raw_shift;
2149 frame_to_encode = &raw;
2151 vpx_usec_timer_start(&timer);
2152 if (use_16bit_internal) {
2153 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2155 if (stream->config.use_16bit_internal)
2156 encode_frame(stream, &global,
2157 frame_avail ? frame_to_encode : NULL, frames_in);
2162 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2163 FOREACH_STREAM(encode_frame(stream, &global,
2164 frame_avail ? frame_to_encode : NULL,
2168 vpx_usec_timer_start(&timer);
2169 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
2172 vpx_usec_timer_mark(&timer);
2173 cx_time += vpx_usec_timer_elapsed(&timer);
2175 FOREACH_STREAM(update_quantizer_histogram(stream));
2178 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2180 if (!got_data && input.length && streams != NULL &&
2181 !streams->frames_out) {
2182 lagged_count = global.limit ? seen_frames : ftello(input.file);
2183 } else if (input.length) {
2188 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2190 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2191 remaining = 1000 * (global.limit - global.skip_frames -
2192 seen_frames + lagged_count);
2194 const int64_t input_pos = ftello(input.file);
2195 const int64_t input_pos_lagged = input_pos - lagged_count;
2196 const int64_t limit = input.length;
2198 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2199 remaining = limit - input_pos + lagged_count;
2203 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
2204 estimated_time_left = average_rate ? remaining / average_rate : -1;
2207 if (got_data && global.test_decode != TEST_DECODE_OFF)
2208 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2212 if (!global.quiet) fprintf(stderr, "\033[K");
2215 if (stream_cnt > 1) fprintf(stderr, "\n");
2217 if (!global.quiet) {
2218 FOREACH_STREAM(fprintf(
2220 "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64 "b/f %7" PRId64
2221 "b/s %7" PRId64 " %s (%.2f fps)\033[K\n",
2222 pass + 1, global.passes, frames_in, stream->frames_out,
2223 (int64_t)stream->nbytes,
2224 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2226 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
2227 global.framerate.den / seen_frames
2229 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2230 stream->cx_time > 9999999 ? "ms" : "us",
2231 usec_to_fps(stream->cx_time, seen_frames)));
2234 if (global.show_psnr) {
2235 if (global.codec->fourcc == VP9_FOURCC) {
2237 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2239 FOREACH_STREAM(show_psnr(stream, 255.0));
2243 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2245 if (global.test_decode != TEST_DECODE_OFF) {
2246 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2249 close_input_file(&input);
2251 if (global.test_decode == TEST_DECODE_FATAL) {
2252 FOREACH_STREAM(res |= stream->mismatch_seen);
2254 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2256 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2258 #if CONFIG_FP_MB_STATS
2259 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2262 if (global.pass) break;
2265 if (global.show_q_hist_buckets)
2267 show_q_histogram(stream->counts, global.show_q_hist_buckets));
2269 if (global.show_rate_hist_buckets)
2270 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2271 global.show_rate_hist_buckets));
2272 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2274 #if CONFIG_INTERNAL_STATS
2275 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2276 * to match some existing utilities.
2278 if (!(global.pass == 1 && global.passes == 2))
2280 FILE *f = fopen("opsnr.stt", "a");
2281 if (stream->mismatch_seen) {
2282 fprintf(f, "First mismatch occurred in frame %d\n",
2283 stream->mismatch_seen);
2285 fprintf(f, "No mismatch detected in recon buffers\n");
2291 #if CONFIG_VP9_HIGHBITDEPTH
2292 if (allocated_raw_shift) vpx_img_free(&raw_shift);
2297 return res ? EXIT_FAILURE : EXIT_SUCCESS;