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 = ARG_DEF(NULL, "arnr-type", 1, "AltRef type");
355 static const struct arg_enum_list tuning_enum[] = {
356 { "psnr", VP8_TUNE_PSNR }, { "ssim", VP8_TUNE_SSIM }, { NULL, 0 }
358 static const arg_def_t tune_ssim =
359 ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
360 static const arg_def_t cq_level =
361 ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
362 static const arg_def_t max_intra_rate_pct =
363 ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");
364 static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
365 NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
367 #if CONFIG_VP8_ENCODER
368 static const arg_def_t cpu_used_vp8 =
369 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-16..16)");
370 static const arg_def_t token_parts =
371 ARG_DEF(NULL, "token-parts", 1, "Number of token partitions to use, log2");
372 static const arg_def_t screen_content_mode =
373 ARG_DEF(NULL, "screen-content-mode", 1, "Screen content mode");
374 static const arg_def_t *vp8_args[] = { &cpu_used_vp8,
387 &screen_content_mode,
389 static const int vp8_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
390 VP8E_SET_ENABLEAUTOALTREF,
391 VP8E_SET_NOISE_SENSITIVITY,
393 VP8E_SET_STATIC_THRESHOLD,
394 VP8E_SET_TOKEN_PARTITIONS,
395 VP8E_SET_ARNR_MAXFRAMES,
396 VP8E_SET_ARNR_STRENGTH,
400 VP8E_SET_MAX_INTRA_BITRATE_PCT,
401 VP8E_SET_GF_CBR_BOOST_PCT,
402 VP8E_SET_SCREEN_CONTENT_MODE,
406 #if CONFIG_VP9_ENCODER
407 static const arg_def_t cpu_used_vp9 =
408 ARG_DEF(NULL, "cpu-used", 1, "CPU Used (-8..8)");
409 static const arg_def_t tile_cols =
410 ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
411 static const arg_def_t tile_rows =
412 ARG_DEF(NULL, "tile-rows", 1,
413 "Number of tile rows to use, log2 (set to 0 while threads > 1)");
414 static const arg_def_t lossless =
415 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
416 static const arg_def_t frame_parallel_decoding = ARG_DEF(
417 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
418 static const arg_def_t aq_mode = ARG_DEF(
420 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
421 "3: cyclic refresh, 4: equator360)");
422 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
423 "Special adaptive quantization for "
424 "the alternate reference frames.");
425 static const arg_def_t frame_periodic_boost =
426 ARG_DEF(NULL, "frame-boost", 1,
427 "Enable frame periodic boost (0: off (default), 1: on)");
428 static const arg_def_t max_inter_rate_pct =
429 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
430 static const arg_def_t min_gf_interval = ARG_DEF(
431 NULL, "min-gf-interval", 1,
432 "min gf/arf frame interval (default 0, indicating in-built behavior)");
433 static const arg_def_t max_gf_interval = ARG_DEF(
434 NULL, "max-gf-interval", 1,
435 "max gf/arf frame interval (default 0, indicating in-built behavior)");
437 static const struct arg_enum_list color_space_enum[] = {
438 { "unknown", VPX_CS_UNKNOWN },
439 { "bt601", VPX_CS_BT_601 },
440 { "bt709", VPX_CS_BT_709 },
441 { "smpte170", VPX_CS_SMPTE_170 },
442 { "smpte240", VPX_CS_SMPTE_240 },
443 { "bt2020", VPX_CS_BT_2020 },
444 { "reserved", VPX_CS_RESERVED },
445 { "sRGB", VPX_CS_SRGB },
449 static const arg_def_t input_color_space =
450 ARG_DEF_ENUM(NULL, "color-space", 1,
451 "The color space of input content:", color_space_enum);
453 #if CONFIG_VP9_HIGHBITDEPTH
454 static const struct arg_enum_list bitdepth_enum[] = {
455 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
458 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
460 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
462 static const arg_def_t inbitdeptharg =
463 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
466 static const struct arg_enum_list tune_content_enum[] = {
467 { "default", VP9E_CONTENT_DEFAULT },
468 { "screen", VP9E_CONTENT_SCREEN },
469 { "film", VP9E_CONTENT_FILM },
473 static const arg_def_t tune_content = ARG_DEF_ENUM(
474 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
476 static const arg_def_t target_level = ARG_DEF(
477 NULL, "target-level", 1,
479 " 255: off (default)\n"
480 " 0: only keep level stats\n"
481 " 1: adaptively set alt-ref "
482 "distance and column tile limit based on picture size, and keep"
484 " 10: level 1.0 11: level 1.1 "
485 "... 62: level 6.2");
487 static const arg_def_t row_mt =
488 ARG_DEF(NULL, "row-mt", 1,
489 "Enable row based non-deterministic multi-threading in VP9");
492 #if CONFIG_VP9_ENCODER
493 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
508 &frame_parallel_decoding,
511 &frame_periodic_boost,
519 #if CONFIG_VP9_HIGHBITDEPTH
522 #endif // CONFIG_VP9_HIGHBITDEPTH
524 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
525 VP8E_SET_ENABLEAUTOALTREF,
527 VP8E_SET_STATIC_THRESHOLD,
528 VP9E_SET_TILE_COLUMNS,
530 VP8E_SET_ARNR_MAXFRAMES,
531 VP8E_SET_ARNR_STRENGTH,
535 VP8E_SET_MAX_INTRA_BITRATE_PCT,
536 VP9E_SET_MAX_INTER_BITRATE_PCT,
537 VP9E_SET_GF_CBR_BOOST_PCT,
539 VP9E_SET_FRAME_PARALLEL_DECODING,
542 VP9E_SET_FRAME_PERIODIC_BOOST,
543 VP9E_SET_NOISE_SENSITIVITY,
544 VP9E_SET_TUNE_CONTENT,
545 VP9E_SET_COLOR_SPACE,
546 VP9E_SET_MIN_GF_INTERVAL,
547 VP9E_SET_MAX_GF_INTERVAL,
548 VP9E_SET_TARGET_LEVEL,
553 static const arg_def_t *no_args[] = { NULL };
555 void show_help(FILE *fout, int shorthelp) {
557 const int num_encoder = get_vpx_encoder_count();
559 fprintf(fout, "Usage: %s <options> -o dst_filename src_filename \n",
563 fprintf(fout, "Use --help to see the full list of options.\n");
567 fprintf(fout, "\nOptions:\n");
568 arg_show_usage(fout, main_args);
569 fprintf(fout, "\nEncoder Global Options:\n");
570 arg_show_usage(fout, global_args);
571 fprintf(fout, "\nRate Control Options:\n");
572 arg_show_usage(fout, rc_args);
573 fprintf(fout, "\nTwopass Rate Control Options:\n");
574 arg_show_usage(fout, rc_twopass_args);
575 fprintf(fout, "\nKeyframe Placement Options:\n");
576 arg_show_usage(fout, kf_args);
577 #if CONFIG_VP8_ENCODER
578 fprintf(fout, "\nVP8 Specific Options:\n");
579 arg_show_usage(fout, vp8_args);
581 #if CONFIG_VP9_ENCODER
582 fprintf(fout, "\nVP9 Specific Options:\n");
583 arg_show_usage(fout, vp9_args);
586 "\nStream timebase (--timebase):\n"
587 " The desired precision of timestamps in the output, expressed\n"
588 " in fractional seconds. Default is 1/1000.\n");
589 fprintf(fout, "\nIncluded encoders:\n\n");
591 for (i = 0; i < num_encoder; ++i) {
592 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
593 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
594 fprintf(fout, " %-6s - %s %s\n", encoder->name,
595 vpx_codec_iface_name(encoder->codec_interface()), defstr);
597 fprintf(fout, "\n ");
598 fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
601 void usage_exit(void) {
602 show_help(stderr, 1);
606 #define mmin(a, b) ((a) < (b) ? (a) : (b))
608 #if CONFIG_VP9_HIGHBITDEPTH
609 static void find_mismatch_high(const vpx_image_t *const img1,
610 const vpx_image_t *const img2, int yloc[4],
611 int uloc[4], int vloc[4]) {
612 uint16_t *plane1, *plane2;
613 uint32_t stride1, stride2;
614 const uint32_t bsize = 64;
615 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
616 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
618 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
620 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
623 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
624 plane1 = (uint16_t *)img1->planes[VPX_PLANE_Y];
625 plane2 = (uint16_t *)img2->planes[VPX_PLANE_Y];
626 stride1 = img1->stride[VPX_PLANE_Y] / 2;
627 stride2 = img2->stride[VPX_PLANE_Y] / 2;
628 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
629 for (j = 0; match && j < img1->d_w; j += bsize) {
631 const int si = mmin(i + bsize, img1->d_h) - i;
632 const int sj = mmin(j + bsize, img1->d_w) - j;
633 for (k = 0; match && k < si; ++k) {
634 for (l = 0; match && l < sj; ++l) {
635 if (*(plane1 + (i + k) * stride1 + j + l) !=
636 *(plane2 + (i + k) * stride2 + j + l)) {
639 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
640 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
649 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
650 plane1 = (uint16_t *)img1->planes[VPX_PLANE_U];
651 plane2 = (uint16_t *)img2->planes[VPX_PLANE_U];
652 stride1 = img1->stride[VPX_PLANE_U] / 2;
653 stride2 = img2->stride[VPX_PLANE_U] / 2;
654 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
655 for (j = 0; match && j < c_w; j += bsizex) {
657 const int si = mmin(i + bsizey, c_h - i);
658 const int sj = mmin(j + bsizex, c_w - j);
659 for (k = 0; match && k < si; ++k) {
660 for (l = 0; match && l < sj; ++l) {
661 if (*(plane1 + (i + k) * stride1 + j + l) !=
662 *(plane2 + (i + k) * stride2 + j + l)) {
665 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
666 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
675 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
676 plane1 = (uint16_t *)img1->planes[VPX_PLANE_V];
677 plane2 = (uint16_t *)img2->planes[VPX_PLANE_V];
678 stride1 = img1->stride[VPX_PLANE_V] / 2;
679 stride2 = img2->stride[VPX_PLANE_V] / 2;
680 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
681 for (j = 0; match && j < c_w; j += bsizex) {
683 const int si = mmin(i + bsizey, c_h - i);
684 const int sj = mmin(j + bsizex, c_w - j);
685 for (k = 0; match && k < si; ++k) {
686 for (l = 0; match && l < sj; ++l) {
687 if (*(plane1 + (i + k) * stride1 + j + l) !=
688 *(plane2 + (i + k) * stride2 + j + l)) {
691 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
692 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
703 static void find_mismatch(const vpx_image_t *const img1,
704 const vpx_image_t *const img2, int yloc[4],
705 int uloc[4], int vloc[4]) {
706 const uint32_t bsize = 64;
707 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
708 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
710 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
712 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
715 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
716 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
717 for (j = 0; match && j < img1->d_w; j += bsize) {
719 const int si = mmin(i + bsize, img1->d_h) - i;
720 const int sj = mmin(j + bsize, img1->d_w) - j;
721 for (k = 0; match && k < si; ++k) {
722 for (l = 0; match && l < sj; ++l) {
723 if (*(img1->planes[VPX_PLANE_Y] +
724 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
725 *(img2->planes[VPX_PLANE_Y] +
726 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
729 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
730 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
731 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
732 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
741 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
742 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
743 for (j = 0; match && j < c_w; j += bsizex) {
745 const int si = mmin(i + bsizey, c_h - i);
746 const int sj = mmin(j + bsizex, c_w - j);
747 for (k = 0; match && k < si; ++k) {
748 for (l = 0; match && l < sj; ++l) {
749 if (*(img1->planes[VPX_PLANE_U] +
750 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
751 *(img2->planes[VPX_PLANE_U] +
752 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
755 uloc[2] = *(img1->planes[VPX_PLANE_U] +
756 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
757 uloc[3] = *(img2->planes[VPX_PLANE_U] +
758 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
766 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
767 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
768 for (j = 0; match && j < c_w; j += bsizex) {
770 const int si = mmin(i + bsizey, c_h - i);
771 const int sj = mmin(j + bsizex, c_w - j);
772 for (k = 0; match && k < si; ++k) {
773 for (l = 0; match && l < sj; ++l) {
774 if (*(img1->planes[VPX_PLANE_V] +
775 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
776 *(img2->planes[VPX_PLANE_V] +
777 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
780 vloc[2] = *(img1->planes[VPX_PLANE_V] +
781 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
782 vloc[3] = *(img2->planes[VPX_PLANE_V] +
783 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
793 static int compare_img(const vpx_image_t *const img1,
794 const vpx_image_t *const img2) {
795 uint32_t l_w = img1->d_w;
796 uint32_t c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
798 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
802 match &= (img1->fmt == img2->fmt);
803 match &= (img1->d_w == img2->d_w);
804 match &= (img1->d_h == img2->d_h);
805 #if CONFIG_VP9_HIGHBITDEPTH
806 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
812 for (i = 0; i < img1->d_h; ++i)
813 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
814 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
817 for (i = 0; i < c_h; ++i)
818 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
819 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
822 for (i = 0; i < c_h; ++i)
823 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
824 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
830 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
831 #if CONFIG_VP9_ENCODER
832 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
834 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
838 typedef int stereo_format_t;
839 struct WebmOutputContext {
844 /* Per-stream configuration */
845 struct stream_config {
846 struct vpx_codec_enc_cfg cfg;
848 const char *stats_fn;
849 #if CONFIG_FP_MB_STATS
850 const char *fpmb_stats_fn;
852 stereo_format_t stereo_fmt;
853 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
856 #if CONFIG_VP9_HIGHBITDEPTH
857 // whether to use 16bit internal buffers
858 int use_16bit_internal;
862 struct stream_state {
864 struct stream_state *next;
865 struct stream_config config;
867 struct rate_hist *rate_hist;
868 struct WebmOutputContext webm_ctx;
869 uint64_t psnr_sse_total;
870 uint64_t psnr_samples_total;
871 double psnr_totals[4];
874 vpx_codec_ctx_t encoder;
875 unsigned int frames_out;
879 #if CONFIG_FP_MB_STATS
880 stats_io_t fpmb_stats;
882 struct vpx_image *img;
883 vpx_codec_ctx_t decoder;
887 static void validate_positive_rational(const char *msg,
888 struct vpx_rational *rat) {
894 if (rat->num < 0) die("Error: %s must be positive\n", msg);
896 if (!rat->den) die("Error: %s has zero denominator\n", msg);
899 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
902 const int num_encoder = get_vpx_encoder_count();
904 if (num_encoder < 1) die("Error: no valid encoder available\n");
906 /* Initialize default parameters */
907 memset(global, 0, sizeof(*global));
908 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
910 global->color_type = I420;
911 /* Assign default deadline to good quality */
912 global->deadline = VPX_DL_GOOD_QUALITY;
914 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
917 if (arg_match(&arg, &help, argi)) {
918 show_help(stdout, 0);
920 } else if (arg_match(&arg, &codecarg, argi)) {
921 global->codec = get_vpx_encoder_by_name(arg.val);
923 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
924 } else if (arg_match(&arg, &passes, argi)) {
925 global->passes = arg_parse_uint(&arg);
927 if (global->passes < 1 || global->passes > 2)
928 die("Error: Invalid number of passes (%d)\n", global->passes);
929 } else if (arg_match(&arg, &pass_arg, argi)) {
930 global->pass = arg_parse_uint(&arg);
932 if (global->pass < 1 || global->pass > 2)
933 die("Error: Invalid pass selected (%d)\n", global->pass);
934 } else if (arg_match(&arg, &usage, argi))
935 global->usage = arg_parse_uint(&arg);
936 else if (arg_match(&arg, &deadline, argi))
937 global->deadline = arg_parse_uint(&arg);
938 else if (arg_match(&arg, &best_dl, argi))
939 global->deadline = VPX_DL_BEST_QUALITY;
940 else if (arg_match(&arg, &good_dl, argi))
941 global->deadline = VPX_DL_GOOD_QUALITY;
942 else if (arg_match(&arg, &rt_dl, argi))
943 global->deadline = VPX_DL_REALTIME;
944 else if (arg_match(&arg, &use_yv12, argi))
945 global->color_type = YV12;
946 else if (arg_match(&arg, &use_i420, argi))
947 global->color_type = I420;
948 else if (arg_match(&arg, &use_i422, argi))
949 global->color_type = I422;
950 else if (arg_match(&arg, &use_i444, argi))
951 global->color_type = I444;
952 else if (arg_match(&arg, &use_i440, argi))
953 global->color_type = I440;
954 else if (arg_match(&arg, &quietarg, argi))
956 else if (arg_match(&arg, &verbosearg, argi))
958 else if (arg_match(&arg, &limit, argi))
959 global->limit = arg_parse_uint(&arg);
960 else if (arg_match(&arg, &skip, argi))
961 global->skip_frames = arg_parse_uint(&arg);
962 else if (arg_match(&arg, &psnrarg, argi))
963 global->show_psnr = 1;
964 else if (arg_match(&arg, &recontest, argi))
965 global->test_decode = arg_parse_enum_or_int(&arg);
966 else if (arg_match(&arg, &framerate, argi)) {
967 global->framerate = arg_parse_rational(&arg);
968 validate_positive_rational(arg.name, &global->framerate);
969 global->have_framerate = 1;
970 } else if (arg_match(&arg, &out_part, argi))
971 global->out_part = 1;
972 else if (arg_match(&arg, &debugmode, argi))
974 else if (arg_match(&arg, &q_hist_n, argi))
975 global->show_q_hist_buckets = arg_parse_uint(&arg);
976 else if (arg_match(&arg, &rate_hist_n, argi))
977 global->show_rate_hist_buckets = arg_parse_uint(&arg);
978 else if (arg_match(&arg, &disable_warnings, argi))
979 global->disable_warnings = 1;
980 else if (arg_match(&arg, &disable_warning_prompt, argi))
981 global->disable_warning_prompt = 1;
987 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
988 if (global->pass > global->passes) {
989 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
991 global->passes = global->pass;
994 /* Validate global config */
995 if (global->passes == 0) {
996 #if CONFIG_VP9_ENCODER
997 // Make default VP9 passes = 2 until there is a better quality 1-pass
999 if (global->codec != NULL && global->codec->name != NULL)
1000 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
1001 global->deadline != VPX_DL_REALTIME)
1009 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
1010 warn("Enforcing one-pass encoding in realtime mode\n");
1015 static void open_input_file(struct VpxInputContext *input) {
1016 /* Parse certain options from the input file, if possible */
1017 input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
1018 : set_binary_mode(stdin);
1020 if (!input->file) fatal("Failed to open input file");
1022 if (!fseeko(input->file, 0, SEEK_END)) {
1023 /* Input file is seekable. Figure out how long it is, so we can get
1026 input->length = ftello(input->file);
1027 rewind(input->file);
1030 /* Default to 1:1 pixel aspect ratio. */
1031 input->pixel_aspect_ratio.numerator = 1;
1032 input->pixel_aspect_ratio.denominator = 1;
1034 /* For RAW input sources, these bytes will applied on the first frame
1037 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1038 input->detect.position = 0;
1040 if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
1041 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1042 input->only_i420) >= 0) {
1043 input->file_type = FILE_TYPE_Y4M;
1044 input->width = input->y4m.pic_w;
1045 input->height = input->y4m.pic_h;
1046 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1047 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1048 input->framerate.numerator = input->y4m.fps_n;
1049 input->framerate.denominator = input->y4m.fps_d;
1050 input->fmt = input->y4m.vpx_fmt;
1051 input->bit_depth = input->y4m.bit_depth;
1053 fatal("Unsupported Y4M stream.");
1054 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1055 fatal("IVF is not supported as input.");
1057 input->file_type = FILE_TYPE_RAW;
1061 static void close_input_file(struct VpxInputContext *input) {
1062 fclose(input->file);
1063 if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
1066 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1067 struct stream_state *prev) {
1068 struct stream_state *stream;
1070 stream = calloc(1, sizeof(*stream));
1071 if (stream == NULL) {
1072 fatal("Failed to allocate new stream.");
1076 memcpy(stream, prev, sizeof(*stream));
1078 prev->next = stream;
1080 vpx_codec_err_t res;
1082 /* Populate encoder configuration */
1083 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1084 &stream->config.cfg, global->usage);
1085 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1087 /* Change the default timebase to a high enough value so that the
1088 * encoder will always create strictly increasing timestamps.
1090 stream->config.cfg.g_timebase.den = 1000;
1092 /* Never use the library's default resolution, require it be parsed
1093 * from the file or set on the command line.
1095 stream->config.cfg.g_w = 0;
1096 stream->config.cfg.g_h = 0;
1098 /* Initialize remaining stream parameters */
1099 stream->config.write_webm = 1;
1101 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1102 stream->webm_ctx.last_pts_ns = -1;
1103 stream->webm_ctx.writer = NULL;
1104 stream->webm_ctx.segment = NULL;
1107 /* Allows removal of the application version from the EBML tags */
1108 stream->webm_ctx.debug = global->debug;
1110 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
1111 if (global->deadline == VPX_DL_REALTIME &&
1112 stream->config.cfg.rc_end_usage == 1)
1113 stream->config.cfg.g_lag_in_frames = 0;
1116 /* Output files must be specified for each stream */
1117 stream->config.out_fn = NULL;
1119 stream->next = NULL;
1123 static int parse_stream_params(struct VpxEncoderConfig *global,
1124 struct stream_state *stream, char **argv) {
1125 char **argi, **argj;
1127 static const arg_def_t **ctrl_args = no_args;
1128 static const int *ctrl_args_map = NULL;
1129 struct stream_config *config = &stream->config;
1130 int eos_mark_found = 0;
1131 #if CONFIG_VP9_HIGHBITDEPTH
1132 int test_16bit_internal = 0;
1135 // Handle codec specific options
1137 #if CONFIG_VP8_ENCODER
1138 } else if (strcmp(global->codec->name, "vp8") == 0) {
1139 ctrl_args = vp8_args;
1140 ctrl_args_map = vp8_arg_ctrl_map;
1142 #if CONFIG_VP9_ENCODER
1143 } else if (strcmp(global->codec->name, "vp9") == 0) {
1144 ctrl_args = vp9_args;
1145 ctrl_args_map = vp9_arg_ctrl_map;
1149 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1152 /* Once we've found an end-of-stream marker (--) we want to continue
1153 * shifting arguments but not consuming them.
1155 if (eos_mark_found) {
1158 } else if (!strcmp(*argj, "--")) {
1163 if (arg_match(&arg, &outputfile, argi)) {
1164 config->out_fn = arg.val;
1165 } else if (arg_match(&arg, &fpf_name, argi)) {
1166 config->stats_fn = arg.val;
1167 #if CONFIG_FP_MB_STATS
1168 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1169 config->fpmb_stats_fn = arg.val;
1171 } else if (arg_match(&arg, &use_webm, argi)) {
1173 config->write_webm = 1;
1175 die("Error: --webm specified but webm is disabled.");
1177 } else if (arg_match(&arg, &use_ivf, argi)) {
1178 config->write_webm = 0;
1179 } else if (arg_match(&arg, &threads, argi)) {
1180 config->cfg.g_threads = arg_parse_uint(&arg);
1181 } else if (arg_match(&arg, &profile, argi)) {
1182 config->cfg.g_profile = arg_parse_uint(&arg);
1183 } else if (arg_match(&arg, &width, argi)) {
1184 config->cfg.g_w = arg_parse_uint(&arg);
1185 } else if (arg_match(&arg, &height, argi)) {
1186 config->cfg.g_h = arg_parse_uint(&arg);
1187 #if CONFIG_VP9_HIGHBITDEPTH
1188 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1189 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1190 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1191 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1194 } else if (arg_match(&arg, &stereo_mode, argi)) {
1195 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1197 } else if (arg_match(&arg, &timebase, argi)) {
1198 config->cfg.g_timebase = arg_parse_rational(&arg);
1199 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1200 } else if (arg_match(&arg, &error_resilient, argi)) {
1201 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1202 } else if (arg_match(&arg, &end_usage, argi)) {
1203 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1204 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1205 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1206 if (global->deadline == VPX_DL_REALTIME &&
1207 config->cfg.rc_end_usage == VPX_CBR &&
1208 config->cfg.g_lag_in_frames != 0) {
1209 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
1210 config->cfg.g_lag_in_frames = 0;
1212 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1213 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1214 } else if (arg_match(&arg, &resize_allowed, argi)) {
1215 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1216 } else if (arg_match(&arg, &resize_width, argi)) {
1217 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1218 } else if (arg_match(&arg, &resize_height, argi)) {
1219 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1220 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1221 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1222 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1223 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1224 } else if (arg_match(&arg, &end_usage, argi)) {
1225 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1226 } else if (arg_match(&arg, &target_bitrate, argi)) {
1227 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1228 } else if (arg_match(&arg, &min_quantizer, argi)) {
1229 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1230 } else if (arg_match(&arg, &max_quantizer, argi)) {
1231 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1232 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1233 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1234 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1235 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1236 } else if (arg_match(&arg, &buf_sz, argi)) {
1237 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1238 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1239 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1240 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1241 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1242 } else if (arg_match(&arg, &bias_pct, argi)) {
1243 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1244 if (global->passes < 2)
1245 warn("option %s ignored in one-pass mode.\n", arg.name);
1246 } else if (arg_match(&arg, &minsection_pct, argi)) {
1247 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1249 if (global->passes < 2)
1250 warn("option %s ignored in one-pass mode.\n", arg.name);
1251 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1252 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1254 if (global->passes < 2)
1255 warn("option %s ignored in one-pass mode.\n", arg.name);
1256 } else if (arg_match(&arg, &corpus_complexity, argi)) {
1257 config->cfg.rc_2pass_vbr_corpus_complexity = arg_parse_uint(&arg);
1259 if (global->passes < 2)
1260 warn("option %s ignored in one-pass mode.\n", arg.name);
1261 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1262 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1263 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1264 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1265 } else if (arg_match(&arg, &kf_disabled, argi)) {
1266 config->cfg.kf_mode = VPX_KF_DISABLED;
1267 #if CONFIG_VP9_HIGHBITDEPTH
1268 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1269 if (strcmp(global->codec->name, "vp9") == 0) {
1270 test_16bit_internal = 1;
1275 for (i = 0; ctrl_args[i]; i++) {
1276 if (arg_match(&arg, ctrl_args[i], argi)) {
1280 /* Point either to the next free element or the first
1281 * instance of this control.
1283 for (j = 0; j < config->arg_ctrl_cnt; j++)
1284 if (ctrl_args_map != NULL &&
1285 config->arg_ctrls[j][0] == ctrl_args_map[i])
1289 assert(j < (int)ARG_CTRL_CNT_MAX);
1290 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1291 config->arg_ctrls[j][0] = ctrl_args_map[i];
1292 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1293 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1300 #if CONFIG_VP9_HIGHBITDEPTH
1301 if (strcmp(global->codec->name, "vp9") == 0) {
1302 config->use_16bit_internal =
1303 test_16bit_internal | (config->cfg.g_profile > 1);
1306 return eos_mark_found;
1309 #define FOREACH_STREAM(func) \
1311 struct stream_state *stream; \
1312 for (stream = streams; stream; stream = stream->next) { \
1317 static void validate_stream_config(const struct stream_state *stream,
1318 const struct VpxEncoderConfig *global) {
1319 const struct stream_state *streami;
1322 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1324 "Stream %d: Specify stream dimensions with --width (-w) "
1325 " and --height (-h)",
1328 // Check that the codec bit depth is greater than the input bit depth.
1329 if (stream->config.cfg.g_input_bit_depth >
1330 (unsigned int)stream->config.cfg.g_bit_depth) {
1331 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1332 stream->index, (int)stream->config.cfg.g_bit_depth,
1333 stream->config.cfg.g_input_bit_depth);
1336 for (streami = stream; streami; streami = streami->next) {
1337 /* All streams require output files */
1338 if (!streami->config.out_fn)
1339 fatal("Stream %d: Output file is required (specify with -o)",
1342 /* Check for two streams outputting to the same file */
1343 if (streami != stream) {
1344 const char *a = stream->config.out_fn;
1345 const char *b = streami->config.out_fn;
1346 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1347 fatal("Stream %d: duplicate output file (from stream %d)",
1348 streami->index, stream->index);
1351 /* Check for two streams sharing a stats file. */
1352 if (streami != stream) {
1353 const char *a = stream->config.stats_fn;
1354 const char *b = streami->config.stats_fn;
1355 if (a && b && !strcmp(a, b))
1356 fatal("Stream %d: duplicate stats file (from stream %d)",
1357 streami->index, stream->index);
1360 #if CONFIG_FP_MB_STATS
1361 /* Check for two streams sharing a mb stats file. */
1362 if (streami != stream) {
1363 const char *a = stream->config.fpmb_stats_fn;
1364 const char *b = streami->config.fpmb_stats_fn;
1365 if (a && b && !strcmp(a, b))
1366 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1367 streami->index, stream->index);
1373 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1375 if (!stream->config.cfg.g_w) {
1376 if (!stream->config.cfg.g_h)
1377 stream->config.cfg.g_w = w;
1379 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1381 if (!stream->config.cfg.g_h) {
1382 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1386 static const char *file_type_to_string(enum VideoFileType t) {
1388 case FILE_TYPE_RAW: return "RAW";
1389 case FILE_TYPE_Y4M: return "Y4M";
1390 default: return "Other";
1394 static const char *image_format_to_string(vpx_img_fmt_t f) {
1396 case VPX_IMG_FMT_I420: return "I420";
1397 case VPX_IMG_FMT_I422: return "I422";
1398 case VPX_IMG_FMT_I444: return "I444";
1399 case VPX_IMG_FMT_I440: return "I440";
1400 case VPX_IMG_FMT_YV12: return "YV12";
1401 case VPX_IMG_FMT_I42016: return "I42016";
1402 case VPX_IMG_FMT_I42216: return "I42216";
1403 case VPX_IMG_FMT_I44416: return "I44416";
1404 case VPX_IMG_FMT_I44016: return "I44016";
1405 default: return "Other";
1409 static void show_stream_config(struct stream_state *stream,
1410 struct VpxEncoderConfig *global,
1411 struct VpxInputContext *input) {
1412 #define SHOW(field) \
1413 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1415 if (stream->index == 0) {
1416 fprintf(stderr, "Codec: %s\n",
1417 vpx_codec_iface_name(global->codec->codec_interface()));
1418 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1419 input->filename, file_type_to_string(input->file_type),
1420 image_format_to_string(input->fmt));
1422 if (stream->next || stream->index)
1423 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1424 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1425 fprintf(stderr, "Encoder parameters:\n");
1433 SHOW(g_input_bit_depth);
1434 SHOW(g_timebase.num);
1435 SHOW(g_timebase.den);
1436 SHOW(g_error_resilient);
1438 SHOW(g_lag_in_frames);
1439 SHOW(rc_dropframe_thresh);
1440 SHOW(rc_resize_allowed);
1441 SHOW(rc_scaled_width);
1442 SHOW(rc_scaled_height);
1443 SHOW(rc_resize_up_thresh);
1444 SHOW(rc_resize_down_thresh);
1446 SHOW(rc_target_bitrate);
1447 SHOW(rc_min_quantizer);
1448 SHOW(rc_max_quantizer);
1449 SHOW(rc_undershoot_pct);
1450 SHOW(rc_overshoot_pct);
1452 SHOW(rc_buf_initial_sz);
1453 SHOW(rc_buf_optimal_sz);
1454 SHOW(rc_2pass_vbr_bias_pct);
1455 SHOW(rc_2pass_vbr_minsection_pct);
1456 SHOW(rc_2pass_vbr_maxsection_pct);
1457 SHOW(rc_2pass_vbr_corpus_complexity);
1463 static void open_output_file(struct stream_state *stream,
1464 struct VpxEncoderConfig *global,
1465 const struct VpxRational *pixel_aspect_ratio) {
1466 const char *fn = stream->config.out_fn;
1467 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1469 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1471 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1473 if (!stream->file) fatal("Failed to open output file");
1475 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1476 fatal("WebM output to pipes not supported.");
1479 if (stream->config.write_webm) {
1480 stream->webm_ctx.stream = stream->file;
1481 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1482 global->codec->fourcc, pixel_aspect_ratio);
1485 (void)pixel_aspect_ratio;
1488 if (!stream->config.write_webm) {
1489 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1493 static void close_output_file(struct stream_state *stream,
1494 unsigned int fourcc) {
1495 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1497 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1500 if (stream->config.write_webm) {
1501 write_webm_file_footer(&stream->webm_ctx);
1505 if (!stream->config.write_webm) {
1506 if (!fseek(stream->file, 0, SEEK_SET))
1507 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1508 stream->frames_out);
1511 fclose(stream->file);
1514 static void setup_pass(struct stream_state *stream,
1515 struct VpxEncoderConfig *global, int pass) {
1516 if (stream->config.stats_fn) {
1517 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1518 fatal("Failed to open statistics store");
1520 if (!stats_open_mem(&stream->stats, pass))
1521 fatal("Failed to open statistics store");
1524 #if CONFIG_FP_MB_STATS
1525 if (stream->config.fpmb_stats_fn) {
1526 if (!stats_open_file(&stream->fpmb_stats, stream->config.fpmb_stats_fn,
1528 fatal("Failed to open mb statistics store");
1530 if (!stats_open_mem(&stream->fpmb_stats, pass))
1531 fatal("Failed to open mb statistics store");
1535 stream->config.cfg.g_pass = global->passes == 2
1536 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1539 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1540 #if CONFIG_FP_MB_STATS
1541 stream->config.cfg.rc_firstpass_mb_stats_in =
1542 stats_get(&stream->fpmb_stats);
1546 stream->cx_time = 0;
1548 stream->frames_out = 0;
1551 static void initialize_encoder(struct stream_state *stream,
1552 struct VpxEncoderConfig *global) {
1556 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1557 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1558 #if CONFIG_VP9_HIGHBITDEPTH
1559 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1562 /* Construct Encoder Context */
1563 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1564 &stream->config.cfg, flags);
1565 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1567 /* Note that we bypass the vpx_codec_control wrapper macro because
1568 * we're being clever to store the control IDs in an array. Real
1569 * applications will want to make use of the enumerations directly
1571 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1572 int ctrl = stream->config.arg_ctrls[i][0];
1573 int value = stream->config.arg_ctrls[i][1];
1574 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1575 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1577 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1581 if (global->test_decode != TEST_DECODE_OFF) {
1582 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1583 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1588 static void encode_frame(struct stream_state *stream,
1589 struct VpxEncoderConfig *global, struct vpx_image *img,
1590 unsigned int frames_in) {
1591 vpx_codec_pts_t frame_start, next_frame_start;
1592 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1593 struct vpx_usec_timer timer;
1596 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1597 cfg->g_timebase.num / global->framerate.num;
1599 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1600 cfg->g_timebase.num / global->framerate.num;
1602 /* Scale if necessary */
1603 #if CONFIG_VP9_HIGHBITDEPTH
1605 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1606 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1607 if (img->fmt != VPX_IMG_FMT_I42016) {
1608 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1614 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1617 (uint16 *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1618 (uint16 *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1619 (uint16 *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1620 img->d_w, img->d_h, (uint16 *)stream->img->planes[VPX_PLANE_Y],
1621 stream->img->stride[VPX_PLANE_Y] / 2,
1622 (uint16 *)stream->img->planes[VPX_PLANE_U],
1623 stream->img->stride[VPX_PLANE_U] / 2,
1624 (uint16 *)stream->img->planes[VPX_PLANE_V],
1625 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1626 stream->img->d_h, kFilterBox);
1629 stream->encoder.err = 1;
1630 ctx_exit_on_error(&stream->encoder,
1631 "Stream %d: Failed to encode frame.\n"
1632 "Scaling disabled in this configuration. \n"
1633 "To enable, configure with --enable-libyuv\n",
1639 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1640 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1641 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1647 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1649 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1650 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1651 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1652 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1653 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1654 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1655 stream->img->d_w, stream->img->d_h, kFilterBox);
1658 stream->encoder.err = 1;
1659 ctx_exit_on_error(&stream->encoder,
1660 "Stream %d: Failed to encode frame.\n"
1661 "Scaling disabled in this configuration. \n"
1662 "To enable, configure with --enable-libyuv\n",
1667 vpx_usec_timer_start(&timer);
1668 vpx_codec_encode(&stream->encoder, img, frame_start,
1669 (unsigned long)(next_frame_start - frame_start), 0,
1671 vpx_usec_timer_mark(&timer);
1672 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1673 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1677 static void update_quantizer_histogram(struct stream_state *stream) {
1678 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1681 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1682 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1683 stream->counts[q]++;
1687 static void get_cx_data(struct stream_state *stream,
1688 struct VpxEncoderConfig *global, int *got_data) {
1689 const vpx_codec_cx_pkt_t *pkt;
1690 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1691 vpx_codec_iter_t iter = NULL;
1694 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1695 static size_t fsize = 0;
1696 static FileOffset ivf_header_pos = 0;
1698 switch (pkt->kind) {
1699 case VPX_CODEC_CX_FRAME_PKT:
1700 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1701 stream->frames_out++;
1704 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1706 update_rate_histogram(stream->rate_hist, cfg, pkt);
1708 if (stream->config.write_webm) {
1709 write_webm_block(&stream->webm_ctx, cfg, pkt);
1712 if (!stream->config.write_webm) {
1713 if (pkt->data.frame.partition_id <= 0) {
1714 ivf_header_pos = ftello(stream->file);
1715 fsize = pkt->data.frame.sz;
1717 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1719 fsize += pkt->data.frame.sz;
1721 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1722 const FileOffset currpos = ftello(stream->file);
1723 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1724 ivf_write_frame_size(stream->file, fsize);
1725 fseeko(stream->file, currpos, SEEK_SET);
1729 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1732 stream->nbytes += pkt->data.raw.sz;
1736 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1737 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1738 (unsigned int)pkt->data.frame.sz, NULL, 0);
1739 if (stream->decoder.err) {
1740 warn_or_exit_on_error(&stream->decoder,
1741 global->test_decode == TEST_DECODE_FATAL,
1742 "Failed to decode frame %d in stream %d",
1743 stream->frames_out + 1, stream->index);
1744 stream->mismatch_seen = stream->frames_out + 1;
1749 case VPX_CODEC_STATS_PKT:
1750 stream->frames_out++;
1751 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1752 pkt->data.twopass_stats.sz);
1753 stream->nbytes += pkt->data.raw.sz;
1755 #if CONFIG_FP_MB_STATS
1756 case VPX_CODEC_FPMB_STATS_PKT:
1757 stats_write(&stream->fpmb_stats, pkt->data.firstpass_mb_stats.buf,
1758 pkt->data.firstpass_mb_stats.sz);
1759 stream->nbytes += pkt->data.raw.sz;
1762 case VPX_CODEC_PSNR_PKT:
1764 if (global->show_psnr) {
1767 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1768 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1769 for (i = 0; i < 4; i++) {
1771 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1772 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1774 stream->psnr_count++;
1783 static void show_psnr(struct stream_state *stream, double peak) {
1787 if (!stream->psnr_count) return;
1789 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1790 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1791 (double)stream->psnr_sse_total);
1792 fprintf(stderr, " %.3f", ovpsnr);
1794 for (i = 0; i < 4; i++) {
1795 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1797 fprintf(stderr, "\n");
1800 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1801 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1804 static void test_decode(struct stream_state *stream,
1805 enum TestDecodeFatality fatal,
1806 const VpxInterface *codec) {
1807 vpx_image_t enc_img, dec_img;
1809 if (stream->mismatch_seen) return;
1811 /* Get the internal reference frame */
1812 if (strcmp(codec->name, "vp8") == 0) {
1813 struct vpx_ref_frame ref_enc, ref_dec;
1816 width = (stream->config.cfg.g_w + 15) & ~15;
1817 height = (stream->config.cfg.g_h + 15) & ~15;
1818 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1819 enc_img = ref_enc.img;
1820 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1821 dec_img = ref_dec.img;
1823 ref_enc.frame_type = VP8_LAST_FRAME;
1824 ref_dec.frame_type = VP8_LAST_FRAME;
1825 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1826 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1828 struct vp9_ref_frame ref_enc, ref_dec;
1832 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1833 enc_img = ref_enc.img;
1834 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1835 dec_img = ref_dec.img;
1836 #if CONFIG_VP9_HIGHBITDEPTH
1837 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1838 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1839 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1840 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1841 enc_img.d_w, enc_img.d_h, 16);
1842 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1844 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1845 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1846 dec_img.d_w, dec_img.d_h, 16);
1847 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1852 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1853 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1855 if (!compare_img(&enc_img, &dec_img)) {
1856 int y[4], u[4], v[4];
1857 #if CONFIG_VP9_HIGHBITDEPTH
1858 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1859 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1861 find_mismatch(&enc_img, &dec_img, y, u, v);
1864 find_mismatch(&enc_img, &dec_img, y, u, v);
1866 stream->decoder.err = 1;
1867 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1868 "Stream %d: Encode/decode mismatch on frame %d at"
1869 " Y[%d, %d] {%d/%d},"
1870 " U[%d, %d] {%d/%d},"
1871 " V[%d, %d] {%d/%d}",
1872 stream->index, stream->frames_out, y[0], y[1], y[2],
1873 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1874 stream->mismatch_seen = stream->frames_out;
1877 vpx_img_free(&enc_img);
1878 vpx_img_free(&dec_img);
1881 static void print_time(const char *label, int64_t etl) {
1888 etl -= hours * 3600;
1893 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1896 fprintf(stderr, "[%3s unknown] ", label);
1900 int main(int argc, const char **argv_) {
1903 #if CONFIG_VP9_HIGHBITDEPTH
1904 vpx_image_t raw_shift;
1905 int allocated_raw_shift = 0;
1906 int use_16bit_internal = 0;
1907 int input_shift = 0;
1909 int frame_avail, got_data;
1911 struct VpxInputContext input;
1912 struct VpxEncoderConfig global;
1913 struct stream_state *streams = NULL;
1914 char **argv, **argi;
1915 uint64_t cx_time = 0;
1919 memset(&input, 0, sizeof(input));
1920 exec_name = argv_[0];
1922 /* Setup default input stream settings */
1923 input.framerate.numerator = 30;
1924 input.framerate.denominator = 1;
1925 input.only_i420 = 1;
1926 input.bit_depth = 0;
1928 /* First parse the global configuration values, because we want to apply
1929 * other parameters on top of the default configuration provided by the
1932 argv = argv_dup(argc - 1, argv_ + 1);
1933 parse_global_config(&global, argv);
1935 if (argc < 3) usage_exit();
1937 switch (global.color_type) {
1938 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1939 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1940 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1941 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1942 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1946 /* Now parse each stream's parameters. Using a local scope here
1947 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1950 struct stream_state *stream = NULL;
1953 stream = new_stream(&global, stream);
1955 if (!streams) streams = stream;
1956 } while (parse_stream_params(&global, stream, argv));
1959 /* Check for unrecognized options */
1960 for (argi = argv; *argi; argi++)
1961 if (argi[0][0] == '-' && argi[0][1])
1962 die("Error: Unrecognized option %s\n", *argi);
1964 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1965 &stream->config.cfg););
1967 /* Handle non-option arguments */
1968 input.filename = argv[0];
1970 if (!input.filename) {
1971 fprintf(stderr, "No input file specified!\n");
1975 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1976 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1978 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1979 int frames_in = 0, seen_frames = 0;
1980 int64_t estimated_time_left = -1;
1981 int64_t average_rate = -1;
1982 int64_t lagged_count = 0;
1984 open_input_file(&input);
1986 /* If the input file doesn't specify its w/h (raw files), try to get
1987 * the data from the first stream's configuration.
1989 if (!input.width || !input.height) {
1991 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1992 input.width = stream->config.cfg.g_w;
1993 input.height = stream->config.cfg.g_h;
1999 /* Update stream configurations from the input file's parameters */
2000 if (!input.width || !input.height)
2002 "Specify stream dimensions with --width (-w) "
2003 " and --height (-h)");
2005 /* If input file does not specify bit-depth but input-bit-depth parameter
2006 * exists, assume that to be the input bit-depth. However, if the
2007 * input-bit-depth paramter does not exist, assume the input bit-depth
2008 * to be the same as the codec bit-depth.
2010 if (!input.bit_depth) {
2012 if (stream->config.cfg.g_input_bit_depth)
2013 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2015 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2016 (int)stream->config.cfg.g_bit_depth;
2018 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2021 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
2024 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2025 FOREACH_STREAM(validate_stream_config(stream, &global));
2027 /* Ensure that --passes and --pass are consistent. If --pass is set and
2028 * --passes=2, ensure --fpf was set.
2030 if (global.pass && global.passes == 2)
2032 if (!stream->config.stats_fn)
2033 die("Stream %d: Must specify --fpf when --pass=%d"
2034 " and --passes=2\n",
2035 stream->index, global.pass);
2040 if (stream->config.write_webm) {
2041 stream->config.write_webm = 0;
2043 "vpxenc was compiled without WebM container support."
2044 "Producing IVF output");
2049 /* Use the frame rate from the file only if none was specified
2050 * on the command-line.
2052 if (!global.have_framerate) {
2053 global.framerate.num = input.framerate.numerator;
2054 global.framerate.den = input.framerate.denominator;
2055 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2056 stream->config.cfg.g_timebase.num = global.framerate.den);
2059 /* Show configuration */
2060 if (global.verbose && pass == 0)
2061 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2063 if (pass == (global.pass ? global.pass - 1 : 0)) {
2064 if (input.file_type == FILE_TYPE_Y4M)
2065 /*The Y4M reader does its own allocation.
2066 Just initialize this here to avoid problems if we never read any
2068 memset(&raw, 0, sizeof(raw));
2070 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2072 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
2073 &stream->config.cfg, &global.framerate));
2076 FOREACH_STREAM(setup_pass(stream, &global, pass));
2078 open_output_file(stream, &global, &input.pixel_aspect_ratio));
2079 FOREACH_STREAM(initialize_encoder(stream, &global));
2081 #if CONFIG_VP9_HIGHBITDEPTH
2082 if (strcmp(global.codec->name, "vp9") == 0) {
2083 // Check to see if at least one stream uses 16 bit internal.
2084 // Currently assume that the bit_depths for all streams using
2085 // highbitdepth are the same.
2087 if (stream->config.use_16bit_internal) {
2088 use_16bit_internal = 1;
2090 if (stream->config.cfg.g_profile == 0) {
2093 input_shift = (int)stream->config.cfg.g_bit_depth -
2094 stream->config.cfg.g_input_bit_depth;
2103 while (frame_avail || got_data) {
2104 struct vpx_usec_timer timer;
2106 if (!global.limit || frames_in < global.limit) {
2107 frame_avail = read_frame(&input, &raw);
2109 if (frame_avail) frames_in++;
2111 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
2113 if (!global.quiet) {
2114 float fps = usec_to_fps(cx_time, seen_frames);
2115 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2117 if (stream_cnt == 1)
2118 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
2119 streams->frames_out, (int64_t)streams->nbytes);
2121 fprintf(stderr, "frame %4d ", frames_in);
2123 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
2124 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2125 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
2126 fps >= 1.0 ? "fps" : "fpm");
2127 print_time("ETA", estimated_time_left);
2133 if (frames_in > global.skip_frames) {
2134 #if CONFIG_VP9_HIGHBITDEPTH
2135 vpx_image_t *frame_to_encode;
2136 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2137 assert(use_16bit_internal);
2138 // Input bit depth and stream bit depth do not match, so up
2139 // shift frame to stream bit depth
2140 if (!allocated_raw_shift) {
2141 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2142 input.width, input.height, 32);
2143 allocated_raw_shift = 1;
2145 vpx_img_upshift(&raw_shift, &raw, input_shift);
2146 frame_to_encode = &raw_shift;
2148 frame_to_encode = &raw;
2150 vpx_usec_timer_start(&timer);
2151 if (use_16bit_internal) {
2152 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2154 if (stream->config.use_16bit_internal)
2155 encode_frame(stream, &global,
2156 frame_avail ? frame_to_encode : NULL, frames_in);
2161 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2162 FOREACH_STREAM(encode_frame(stream, &global,
2163 frame_avail ? frame_to_encode : NULL,
2167 vpx_usec_timer_start(&timer);
2168 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
2171 vpx_usec_timer_mark(&timer);
2172 cx_time += vpx_usec_timer_elapsed(&timer);
2174 FOREACH_STREAM(update_quantizer_histogram(stream));
2177 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2179 if (!got_data && input.length && streams != NULL &&
2180 !streams->frames_out) {
2181 lagged_count = global.limit ? seen_frames : ftello(input.file);
2182 } else if (input.length) {
2187 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2189 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2190 remaining = 1000 * (global.limit - global.skip_frames -
2191 seen_frames + lagged_count);
2193 const int64_t input_pos = ftello(input.file);
2194 const int64_t input_pos_lagged = input_pos - lagged_count;
2195 const int64_t limit = input.length;
2197 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2198 remaining = limit - input_pos + lagged_count;
2202 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
2203 estimated_time_left = average_rate ? remaining / average_rate : -1;
2206 if (got_data && global.test_decode != TEST_DECODE_OFF)
2207 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2211 if (!global.quiet) fprintf(stderr, "\033[K");
2214 if (stream_cnt > 1) fprintf(stderr, "\n");
2216 if (!global.quiet) {
2217 FOREACH_STREAM(fprintf(
2219 "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64 "b/f %7" PRId64
2220 "b/s %7" PRId64 " %s (%.2f fps)\033[K\n",
2221 pass + 1, global.passes, frames_in, stream->frames_out,
2222 (int64_t)stream->nbytes,
2223 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2225 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
2226 global.framerate.den / seen_frames
2228 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2229 stream->cx_time > 9999999 ? "ms" : "us",
2230 usec_to_fps(stream->cx_time, seen_frames)));
2233 if (global.show_psnr) {
2234 if (global.codec->fourcc == VP9_FOURCC) {
2236 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2238 FOREACH_STREAM(show_psnr(stream, 255.0));
2242 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2244 if (global.test_decode != TEST_DECODE_OFF) {
2245 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2248 close_input_file(&input);
2250 if (global.test_decode == TEST_DECODE_FATAL) {
2251 FOREACH_STREAM(res |= stream->mismatch_seen);
2253 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2255 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2257 #if CONFIG_FP_MB_STATS
2258 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2261 if (global.pass) break;
2264 if (global.show_q_hist_buckets)
2266 show_q_histogram(stream->counts, global.show_q_hist_buckets));
2268 if (global.show_rate_hist_buckets)
2269 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2270 global.show_rate_hist_buckets));
2271 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2273 #if CONFIG_INTERNAL_STATS
2274 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2275 * to match some existing utilities.
2277 if (!(global.pass == 1 && global.passes == 2))
2279 FILE *f = fopen("opsnr.stt", "a");
2280 if (stream->mismatch_seen) {
2281 fprintf(f, "First mismatch occurred in frame %d\n",
2282 stream->mismatch_seen);
2284 fprintf(f, "No mismatch detected in recon buffers\n");
2290 #if CONFIG_VP9_HIGHBITDEPTH
2291 if (allocated_raw_shift) vpx_img_free(&raw_shift);
2296 return res ? EXIT_FAILURE : EXIT_SUCCESS;