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)");
416 static const arg_def_t enable_tpl_model =
417 ARG_DEF(NULL, "enable-tpl", 1, "Enable temporal dependency model");
419 static const arg_def_t lossless =
420 ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
421 static const arg_def_t frame_parallel_decoding = ARG_DEF(
422 NULL, "frame-parallel", 1, "Enable frame parallel decodability features");
423 static const arg_def_t aq_mode = ARG_DEF(
425 "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
426 "3: cyclic refresh, 4: equator360)");
427 static const arg_def_t alt_ref_aq = ARG_DEF(NULL, "alt-ref-aq", 1,
428 "Special adaptive quantization for "
429 "the alternate reference frames.");
430 static const arg_def_t frame_periodic_boost =
431 ARG_DEF(NULL, "frame-boost", 1,
432 "Enable frame periodic boost (0: off (default), 1: on)");
433 static const arg_def_t max_inter_rate_pct =
434 ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
435 static const arg_def_t min_gf_interval = ARG_DEF(
436 NULL, "min-gf-interval", 1,
437 "min gf/arf frame interval (default 0, indicating in-built behavior)");
438 static const arg_def_t max_gf_interval = ARG_DEF(
439 NULL, "max-gf-interval", 1,
440 "max gf/arf frame interval (default 0, indicating in-built behavior)");
442 static const struct arg_enum_list color_space_enum[] = {
443 { "unknown", VPX_CS_UNKNOWN },
444 { "bt601", VPX_CS_BT_601 },
445 { "bt709", VPX_CS_BT_709 },
446 { "smpte170", VPX_CS_SMPTE_170 },
447 { "smpte240", VPX_CS_SMPTE_240 },
448 { "bt2020", VPX_CS_BT_2020 },
449 { "reserved", VPX_CS_RESERVED },
450 { "sRGB", VPX_CS_SRGB },
454 static const arg_def_t input_color_space =
455 ARG_DEF_ENUM(NULL, "color-space", 1,
456 "The color space of input content:", color_space_enum);
458 #if CONFIG_VP9_HIGHBITDEPTH
459 static const struct arg_enum_list bitdepth_enum[] = {
460 { "8", VPX_BITS_8 }, { "10", VPX_BITS_10 }, { "12", VPX_BITS_12 }, { NULL, 0 }
463 static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
465 "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
467 static const arg_def_t inbitdeptharg =
468 ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
471 static const struct arg_enum_list tune_content_enum[] = {
472 { "default", VP9E_CONTENT_DEFAULT },
473 { "screen", VP9E_CONTENT_SCREEN },
474 { "film", VP9E_CONTENT_FILM },
478 static const arg_def_t tune_content = ARG_DEF_ENUM(
479 NULL, "tune-content", 1, "Tune content type", tune_content_enum);
481 static const arg_def_t target_level = ARG_DEF(
482 NULL, "target-level", 1,
484 " 255: off (default)\n"
485 " 0: only keep level stats\n"
486 " 1: adaptively set alt-ref "
487 "distance and column tile limit based on picture size, and keep"
489 " 10: level 1.0 11: level 1.1 "
490 "... 62: level 6.2");
492 static const arg_def_t row_mt =
493 ARG_DEF(NULL, "row-mt", 1,
494 "Enable row based non-deterministic multi-threading in VP9");
497 #if CONFIG_VP9_ENCODER
498 static const arg_def_t *vp9_args[] = { &cpu_used_vp9,
514 &frame_parallel_decoding,
517 &frame_periodic_boost,
525 #if CONFIG_VP9_HIGHBITDEPTH
528 #endif // CONFIG_VP9_HIGHBITDEPTH
530 static const int vp9_arg_ctrl_map[] = { VP8E_SET_CPUUSED,
531 VP8E_SET_ENABLEAUTOALTREF,
533 VP8E_SET_STATIC_THRESHOLD,
534 VP9E_SET_TILE_COLUMNS,
537 VP8E_SET_ARNR_MAXFRAMES,
538 VP8E_SET_ARNR_STRENGTH,
542 VP8E_SET_MAX_INTRA_BITRATE_PCT,
543 VP9E_SET_MAX_INTER_BITRATE_PCT,
544 VP9E_SET_GF_CBR_BOOST_PCT,
546 VP9E_SET_FRAME_PARALLEL_DECODING,
549 VP9E_SET_FRAME_PERIODIC_BOOST,
550 VP9E_SET_NOISE_SENSITIVITY,
551 VP9E_SET_TUNE_CONTENT,
552 VP9E_SET_COLOR_SPACE,
553 VP9E_SET_MIN_GF_INTERVAL,
554 VP9E_SET_MAX_GF_INTERVAL,
555 VP9E_SET_TARGET_LEVEL,
560 static const arg_def_t *no_args[] = { NULL };
562 void show_help(FILE *fout, int shorthelp) {
564 const int num_encoder = get_vpx_encoder_count();
566 fprintf(fout, "Usage: %s <options> -o dst_filename src_filename \n",
570 fprintf(fout, "Use --help to see the full list of options.\n");
574 fprintf(fout, "\nOptions:\n");
575 arg_show_usage(fout, main_args);
576 fprintf(fout, "\nEncoder Global Options:\n");
577 arg_show_usage(fout, global_args);
578 fprintf(fout, "\nRate Control Options:\n");
579 arg_show_usage(fout, rc_args);
580 fprintf(fout, "\nTwopass Rate Control Options:\n");
581 arg_show_usage(fout, rc_twopass_args);
582 fprintf(fout, "\nKeyframe Placement Options:\n");
583 arg_show_usage(fout, kf_args);
584 #if CONFIG_VP8_ENCODER
585 fprintf(fout, "\nVP8 Specific Options:\n");
586 arg_show_usage(fout, vp8_args);
588 #if CONFIG_VP9_ENCODER
589 fprintf(fout, "\nVP9 Specific Options:\n");
590 arg_show_usage(fout, vp9_args);
593 "\nStream timebase (--timebase):\n"
594 " The desired precision of timestamps in the output, expressed\n"
595 " in fractional seconds. Default is 1/1000.\n");
596 fprintf(fout, "\nIncluded encoders:\n\n");
598 for (i = 0; i < num_encoder; ++i) {
599 const VpxInterface *const encoder = get_vpx_encoder_by_index(i);
600 const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
601 fprintf(fout, " %-6s - %s %s\n", encoder->name,
602 vpx_codec_iface_name(encoder->codec_interface()), defstr);
604 fprintf(fout, "\n ");
605 fprintf(fout, "Use --codec to switch to a non-default encoder.\n\n");
608 void usage_exit(void) {
609 show_help(stderr, 1);
613 #define mmin(a, b) ((a) < (b) ? (a) : (b))
615 #if CONFIG_VP9_HIGHBITDEPTH
616 static void find_mismatch_high(const vpx_image_t *const img1,
617 const vpx_image_t *const img2, int yloc[4],
618 int uloc[4], int vloc[4]) {
619 uint16_t *plane1, *plane2;
620 uint32_t stride1, stride2;
621 const uint32_t bsize = 64;
622 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
623 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
625 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
627 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
630 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
631 plane1 = (uint16_t *)img1->planes[VPX_PLANE_Y];
632 plane2 = (uint16_t *)img2->planes[VPX_PLANE_Y];
633 stride1 = img1->stride[VPX_PLANE_Y] / 2;
634 stride2 = img2->stride[VPX_PLANE_Y] / 2;
635 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
636 for (j = 0; match && j < img1->d_w; j += bsize) {
638 const int si = mmin(i + bsize, img1->d_h) - i;
639 const int sj = mmin(j + bsize, img1->d_w) - j;
640 for (k = 0; match && k < si; ++k) {
641 for (l = 0; match && l < sj; ++l) {
642 if (*(plane1 + (i + k) * stride1 + j + l) !=
643 *(plane2 + (i + k) * stride2 + j + l)) {
646 yloc[2] = *(plane1 + (i + k) * stride1 + j + l);
647 yloc[3] = *(plane2 + (i + k) * stride2 + j + l);
656 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
657 plane1 = (uint16_t *)img1->planes[VPX_PLANE_U];
658 plane2 = (uint16_t *)img2->planes[VPX_PLANE_U];
659 stride1 = img1->stride[VPX_PLANE_U] / 2;
660 stride2 = img2->stride[VPX_PLANE_U] / 2;
661 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
662 for (j = 0; match && j < c_w; j += bsizex) {
664 const int si = mmin(i + bsizey, c_h - i);
665 const int sj = mmin(j + bsizex, c_w - j);
666 for (k = 0; match && k < si; ++k) {
667 for (l = 0; match && l < sj; ++l) {
668 if (*(plane1 + (i + k) * stride1 + j + l) !=
669 *(plane2 + (i + k) * stride2 + j + l)) {
672 uloc[2] = *(plane1 + (i + k) * stride1 + j + l);
673 uloc[3] = *(plane2 + (i + k) * stride2 + j + l);
682 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
683 plane1 = (uint16_t *)img1->planes[VPX_PLANE_V];
684 plane2 = (uint16_t *)img2->planes[VPX_PLANE_V];
685 stride1 = img1->stride[VPX_PLANE_V] / 2;
686 stride2 = img2->stride[VPX_PLANE_V] / 2;
687 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
688 for (j = 0; match && j < c_w; j += bsizex) {
690 const int si = mmin(i + bsizey, c_h - i);
691 const int sj = mmin(j + bsizex, c_w - j);
692 for (k = 0; match && k < si; ++k) {
693 for (l = 0; match && l < sj; ++l) {
694 if (*(plane1 + (i + k) * stride1 + j + l) !=
695 *(plane2 + (i + k) * stride2 + j + l)) {
698 vloc[2] = *(plane1 + (i + k) * stride1 + j + l);
699 vloc[3] = *(plane2 + (i + k) * stride2 + j + l);
710 static void find_mismatch(const vpx_image_t *const img1,
711 const vpx_image_t *const img2, int yloc[4],
712 int uloc[4], int vloc[4]) {
713 const uint32_t bsize = 64;
714 const uint32_t bsizey = bsize >> img1->y_chroma_shift;
715 const uint32_t bsizex = bsize >> img1->x_chroma_shift;
717 (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
719 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
722 yloc[0] = yloc[1] = yloc[2] = yloc[3] = -1;
723 for (i = 0, match = 1; match && i < img1->d_h; i += bsize) {
724 for (j = 0; match && j < img1->d_w; j += bsize) {
726 const int si = mmin(i + bsize, img1->d_h) - i;
727 const int sj = mmin(j + bsize, img1->d_w) - j;
728 for (k = 0; match && k < si; ++k) {
729 for (l = 0; match && l < sj; ++l) {
730 if (*(img1->planes[VPX_PLANE_Y] +
731 (i + k) * img1->stride[VPX_PLANE_Y] + j + l) !=
732 *(img2->planes[VPX_PLANE_Y] +
733 (i + k) * img2->stride[VPX_PLANE_Y] + j + l)) {
736 yloc[2] = *(img1->planes[VPX_PLANE_Y] +
737 (i + k) * img1->stride[VPX_PLANE_Y] + j + l);
738 yloc[3] = *(img2->planes[VPX_PLANE_Y] +
739 (i + k) * img2->stride[VPX_PLANE_Y] + j + l);
748 uloc[0] = uloc[1] = uloc[2] = uloc[3] = -1;
749 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
750 for (j = 0; match && j < c_w; j += bsizex) {
752 const int si = mmin(i + bsizey, c_h - i);
753 const int sj = mmin(j + bsizex, c_w - j);
754 for (k = 0; match && k < si; ++k) {
755 for (l = 0; match && l < sj; ++l) {
756 if (*(img1->planes[VPX_PLANE_U] +
757 (i + k) * img1->stride[VPX_PLANE_U] + j + l) !=
758 *(img2->planes[VPX_PLANE_U] +
759 (i + k) * img2->stride[VPX_PLANE_U] + j + l)) {
762 uloc[2] = *(img1->planes[VPX_PLANE_U] +
763 (i + k) * img1->stride[VPX_PLANE_U] + j + l);
764 uloc[3] = *(img2->planes[VPX_PLANE_U] +
765 (i + k) * img2->stride[VPX_PLANE_U] + j + l);
773 vloc[0] = vloc[1] = vloc[2] = vloc[3] = -1;
774 for (i = 0, match = 1; match && i < c_h; i += bsizey) {
775 for (j = 0; match && j < c_w; j += bsizex) {
777 const int si = mmin(i + bsizey, c_h - i);
778 const int sj = mmin(j + bsizex, c_w - j);
779 for (k = 0; match && k < si; ++k) {
780 for (l = 0; match && l < sj; ++l) {
781 if (*(img1->planes[VPX_PLANE_V] +
782 (i + k) * img1->stride[VPX_PLANE_V] + j + l) !=
783 *(img2->planes[VPX_PLANE_V] +
784 (i + k) * img2->stride[VPX_PLANE_V] + j + l)) {
787 vloc[2] = *(img1->planes[VPX_PLANE_V] +
788 (i + k) * img1->stride[VPX_PLANE_V] + j + l);
789 vloc[3] = *(img2->planes[VPX_PLANE_V] +
790 (i + k) * img2->stride[VPX_PLANE_V] + j + l);
800 static int compare_img(const vpx_image_t *const img1,
801 const vpx_image_t *const img2) {
802 uint32_t l_w = img1->d_w;
803 uint32_t c_w = (img1->d_w + img1->x_chroma_shift) >> img1->x_chroma_shift;
805 (img1->d_h + img1->y_chroma_shift) >> img1->y_chroma_shift;
809 match &= (img1->fmt == img2->fmt);
810 match &= (img1->d_w == img2->d_w);
811 match &= (img1->d_h == img2->d_h);
812 #if CONFIG_VP9_HIGHBITDEPTH
813 if (img1->fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
819 for (i = 0; i < img1->d_h; ++i)
820 match &= (memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
821 img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
824 for (i = 0; i < c_h; ++i)
825 match &= (memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
826 img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
829 for (i = 0; i < c_h; ++i)
830 match &= (memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
831 img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
837 #define NELEMENTS(x) (sizeof(x) / sizeof(x[0]))
838 #if CONFIG_VP9_ENCODER
839 #define ARG_CTRL_CNT_MAX NELEMENTS(vp9_arg_ctrl_map)
841 #define ARG_CTRL_CNT_MAX NELEMENTS(vp8_arg_ctrl_map)
845 typedef int stereo_format_t;
846 struct WebmOutputContext {
851 /* Per-stream configuration */
852 struct stream_config {
853 struct vpx_codec_enc_cfg cfg;
855 const char *stats_fn;
856 #if CONFIG_FP_MB_STATS
857 const char *fpmb_stats_fn;
859 stereo_format_t stereo_fmt;
860 int arg_ctrls[ARG_CTRL_CNT_MAX][2];
863 #if CONFIG_VP9_HIGHBITDEPTH
864 // whether to use 16bit internal buffers
865 int use_16bit_internal;
869 struct stream_state {
871 struct stream_state *next;
872 struct stream_config config;
874 struct rate_hist *rate_hist;
875 struct WebmOutputContext webm_ctx;
876 uint64_t psnr_sse_total;
877 uint64_t psnr_samples_total;
878 double psnr_totals[4];
881 vpx_codec_ctx_t encoder;
882 unsigned int frames_out;
886 #if CONFIG_FP_MB_STATS
887 stats_io_t fpmb_stats;
889 struct vpx_image *img;
890 vpx_codec_ctx_t decoder;
894 static void validate_positive_rational(const char *msg,
895 struct vpx_rational *rat) {
901 if (rat->num < 0) die("Error: %s must be positive\n", msg);
903 if (!rat->den) die("Error: %s has zero denominator\n", msg);
906 static void parse_global_config(struct VpxEncoderConfig *global, char **argv) {
909 const int num_encoder = get_vpx_encoder_count();
911 if (num_encoder < 1) die("Error: no valid encoder available\n");
913 /* Initialize default parameters */
914 memset(global, 0, sizeof(*global));
915 global->codec = get_vpx_encoder_by_index(num_encoder - 1);
917 global->color_type = I420;
918 /* Assign default deadline to good quality */
919 global->deadline = VPX_DL_GOOD_QUALITY;
921 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
924 if (arg_match(&arg, &help, argi)) {
925 show_help(stdout, 0);
927 } else if (arg_match(&arg, &codecarg, argi)) {
928 global->codec = get_vpx_encoder_by_name(arg.val);
930 die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
931 } else if (arg_match(&arg, &passes, argi)) {
932 global->passes = arg_parse_uint(&arg);
934 if (global->passes < 1 || global->passes > 2)
935 die("Error: Invalid number of passes (%d)\n", global->passes);
936 } else if (arg_match(&arg, &pass_arg, argi)) {
937 global->pass = arg_parse_uint(&arg);
939 if (global->pass < 1 || global->pass > 2)
940 die("Error: Invalid pass selected (%d)\n", global->pass);
941 } else if (arg_match(&arg, &usage, argi))
942 global->usage = arg_parse_uint(&arg);
943 else if (arg_match(&arg, &deadline, argi))
944 global->deadline = arg_parse_uint(&arg);
945 else if (arg_match(&arg, &best_dl, argi))
946 global->deadline = VPX_DL_BEST_QUALITY;
947 else if (arg_match(&arg, &good_dl, argi))
948 global->deadline = VPX_DL_GOOD_QUALITY;
949 else if (arg_match(&arg, &rt_dl, argi))
950 global->deadline = VPX_DL_REALTIME;
951 else if (arg_match(&arg, &use_yv12, argi))
952 global->color_type = YV12;
953 else if (arg_match(&arg, &use_i420, argi))
954 global->color_type = I420;
955 else if (arg_match(&arg, &use_i422, argi))
956 global->color_type = I422;
957 else if (arg_match(&arg, &use_i444, argi))
958 global->color_type = I444;
959 else if (arg_match(&arg, &use_i440, argi))
960 global->color_type = I440;
961 else if (arg_match(&arg, &quietarg, argi))
963 else if (arg_match(&arg, &verbosearg, argi))
965 else if (arg_match(&arg, &limit, argi))
966 global->limit = arg_parse_uint(&arg);
967 else if (arg_match(&arg, &skip, argi))
968 global->skip_frames = arg_parse_uint(&arg);
969 else if (arg_match(&arg, &psnrarg, argi))
970 global->show_psnr = 1;
971 else if (arg_match(&arg, &recontest, argi))
972 global->test_decode = arg_parse_enum_or_int(&arg);
973 else if (arg_match(&arg, &framerate, argi)) {
974 global->framerate = arg_parse_rational(&arg);
975 validate_positive_rational(arg.name, &global->framerate);
976 global->have_framerate = 1;
977 } else if (arg_match(&arg, &out_part, argi))
978 global->out_part = 1;
979 else if (arg_match(&arg, &debugmode, argi))
981 else if (arg_match(&arg, &q_hist_n, argi))
982 global->show_q_hist_buckets = arg_parse_uint(&arg);
983 else if (arg_match(&arg, &rate_hist_n, argi))
984 global->show_rate_hist_buckets = arg_parse_uint(&arg);
985 else if (arg_match(&arg, &disable_warnings, argi))
986 global->disable_warnings = 1;
987 else if (arg_match(&arg, &disable_warning_prompt, argi))
988 global->disable_warning_prompt = 1;
994 /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
995 if (global->pass > global->passes) {
996 warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
998 global->passes = global->pass;
1001 /* Validate global config */
1002 if (global->passes == 0) {
1003 #if CONFIG_VP9_ENCODER
1004 // Make default VP9 passes = 2 until there is a better quality 1-pass
1006 if (global->codec != NULL && global->codec->name != NULL)
1007 global->passes = (strcmp(global->codec->name, "vp9") == 0 &&
1008 global->deadline != VPX_DL_REALTIME)
1016 if (global->deadline == VPX_DL_REALTIME && global->passes > 1) {
1017 warn("Enforcing one-pass encoding in realtime mode\n");
1022 static void open_input_file(struct VpxInputContext *input) {
1023 /* Parse certain options from the input file, if possible */
1024 input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
1025 : set_binary_mode(stdin);
1027 if (!input->file) fatal("Failed to open input file");
1029 if (!fseeko(input->file, 0, SEEK_END)) {
1030 /* Input file is seekable. Figure out how long it is, so we can get
1033 input->length = ftello(input->file);
1034 rewind(input->file);
1037 /* Default to 1:1 pixel aspect ratio. */
1038 input->pixel_aspect_ratio.numerator = 1;
1039 input->pixel_aspect_ratio.denominator = 1;
1041 /* For RAW input sources, these bytes will applied on the first frame
1044 input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
1045 input->detect.position = 0;
1047 if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
1048 if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
1049 input->only_i420) >= 0) {
1050 input->file_type = FILE_TYPE_Y4M;
1051 input->width = input->y4m.pic_w;
1052 input->height = input->y4m.pic_h;
1053 input->pixel_aspect_ratio.numerator = input->y4m.par_n;
1054 input->pixel_aspect_ratio.denominator = input->y4m.par_d;
1055 input->framerate.numerator = input->y4m.fps_n;
1056 input->framerate.denominator = input->y4m.fps_d;
1057 input->fmt = input->y4m.vpx_fmt;
1058 input->bit_depth = input->y4m.bit_depth;
1060 fatal("Unsupported Y4M stream.");
1061 } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
1062 fatal("IVF is not supported as input.");
1064 input->file_type = FILE_TYPE_RAW;
1068 static void close_input_file(struct VpxInputContext *input) {
1069 fclose(input->file);
1070 if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
1073 static struct stream_state *new_stream(struct VpxEncoderConfig *global,
1074 struct stream_state *prev) {
1075 struct stream_state *stream;
1077 stream = calloc(1, sizeof(*stream));
1078 if (stream == NULL) {
1079 fatal("Failed to allocate new stream.");
1083 memcpy(stream, prev, sizeof(*stream));
1085 prev->next = stream;
1087 vpx_codec_err_t res;
1089 /* Populate encoder configuration */
1090 res = vpx_codec_enc_config_default(global->codec->codec_interface(),
1091 &stream->config.cfg, global->usage);
1092 if (res) fatal("Failed to get config: %s\n", vpx_codec_err_to_string(res));
1094 /* Change the default timebase to a high enough value so that the
1095 * encoder will always create strictly increasing timestamps.
1097 stream->config.cfg.g_timebase.den = 1000;
1099 /* Never use the library's default resolution, require it be parsed
1100 * from the file or set on the command line.
1102 stream->config.cfg.g_w = 0;
1103 stream->config.cfg.g_h = 0;
1105 /* Initialize remaining stream parameters */
1106 stream->config.write_webm = 1;
1108 stream->config.stereo_fmt = STEREO_FORMAT_MONO;
1109 stream->webm_ctx.last_pts_ns = -1;
1110 stream->webm_ctx.writer = NULL;
1111 stream->webm_ctx.segment = NULL;
1114 /* Allows removal of the application version from the EBML tags */
1115 stream->webm_ctx.debug = global->debug;
1117 /* Default lag_in_frames is 0 in realtime mode CBR mode*/
1118 if (global->deadline == VPX_DL_REALTIME &&
1119 stream->config.cfg.rc_end_usage == 1)
1120 stream->config.cfg.g_lag_in_frames = 0;
1123 /* Output files must be specified for each stream */
1124 stream->config.out_fn = NULL;
1126 stream->next = NULL;
1130 static int parse_stream_params(struct VpxEncoderConfig *global,
1131 struct stream_state *stream, char **argv) {
1132 char **argi, **argj;
1134 static const arg_def_t **ctrl_args = no_args;
1135 static const int *ctrl_args_map = NULL;
1136 struct stream_config *config = &stream->config;
1137 int eos_mark_found = 0;
1138 #if CONFIG_VP9_HIGHBITDEPTH
1139 int test_16bit_internal = 0;
1142 // Handle codec specific options
1144 #if CONFIG_VP8_ENCODER
1145 } else if (strcmp(global->codec->name, "vp8") == 0) {
1146 ctrl_args = vp8_args;
1147 ctrl_args_map = vp8_arg_ctrl_map;
1149 #if CONFIG_VP9_ENCODER
1150 } else if (strcmp(global->codec->name, "vp9") == 0) {
1151 ctrl_args = vp9_args;
1152 ctrl_args_map = vp9_arg_ctrl_map;
1156 for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
1159 /* Once we've found an end-of-stream marker (--) we want to continue
1160 * shifting arguments but not consuming them.
1162 if (eos_mark_found) {
1165 } else if (!strcmp(*argj, "--")) {
1170 if (arg_match(&arg, &outputfile, argi)) {
1171 config->out_fn = arg.val;
1172 } else if (arg_match(&arg, &fpf_name, argi)) {
1173 config->stats_fn = arg.val;
1174 #if CONFIG_FP_MB_STATS
1175 } else if (arg_match(&arg, &fpmbf_name, argi)) {
1176 config->fpmb_stats_fn = arg.val;
1178 } else if (arg_match(&arg, &use_webm, argi)) {
1180 config->write_webm = 1;
1182 die("Error: --webm specified but webm is disabled.");
1184 } else if (arg_match(&arg, &use_ivf, argi)) {
1185 config->write_webm = 0;
1186 } else if (arg_match(&arg, &threads, argi)) {
1187 config->cfg.g_threads = arg_parse_uint(&arg);
1188 } else if (arg_match(&arg, &profile, argi)) {
1189 config->cfg.g_profile = arg_parse_uint(&arg);
1190 } else if (arg_match(&arg, &width, argi)) {
1191 config->cfg.g_w = arg_parse_uint(&arg);
1192 } else if (arg_match(&arg, &height, argi)) {
1193 config->cfg.g_h = arg_parse_uint(&arg);
1194 #if CONFIG_VP9_HIGHBITDEPTH
1195 } else if (arg_match(&arg, &bitdeptharg, argi)) {
1196 config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
1197 } else if (arg_match(&arg, &inbitdeptharg, argi)) {
1198 config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
1201 } else if (arg_match(&arg, &stereo_mode, argi)) {
1202 config->stereo_fmt = arg_parse_enum_or_int(&arg);
1204 } else if (arg_match(&arg, &timebase, argi)) {
1205 config->cfg.g_timebase = arg_parse_rational(&arg);
1206 validate_positive_rational(arg.name, &config->cfg.g_timebase);
1207 } else if (arg_match(&arg, &error_resilient, argi)) {
1208 config->cfg.g_error_resilient = arg_parse_uint(&arg);
1209 } else if (arg_match(&arg, &end_usage, argi)) {
1210 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1211 } else if (arg_match(&arg, &lag_in_frames, argi)) {
1212 config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
1213 if (global->deadline == VPX_DL_REALTIME &&
1214 config->cfg.rc_end_usage == VPX_CBR &&
1215 config->cfg.g_lag_in_frames != 0) {
1216 warn("non-zero %s option ignored in realtime CBR mode.\n", arg.name);
1217 config->cfg.g_lag_in_frames = 0;
1219 } else if (arg_match(&arg, &dropframe_thresh, argi)) {
1220 config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
1221 } else if (arg_match(&arg, &resize_allowed, argi)) {
1222 config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
1223 } else if (arg_match(&arg, &resize_width, argi)) {
1224 config->cfg.rc_scaled_width = arg_parse_uint(&arg);
1225 } else if (arg_match(&arg, &resize_height, argi)) {
1226 config->cfg.rc_scaled_height = arg_parse_uint(&arg);
1227 } else if (arg_match(&arg, &resize_up_thresh, argi)) {
1228 config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
1229 } else if (arg_match(&arg, &resize_down_thresh, argi)) {
1230 config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
1231 } else if (arg_match(&arg, &end_usage, argi)) {
1232 config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
1233 } else if (arg_match(&arg, &target_bitrate, argi)) {
1234 config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
1235 } else if (arg_match(&arg, &min_quantizer, argi)) {
1236 config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
1237 } else if (arg_match(&arg, &max_quantizer, argi)) {
1238 config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
1239 } else if (arg_match(&arg, &undershoot_pct, argi)) {
1240 config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
1241 } else if (arg_match(&arg, &overshoot_pct, argi)) {
1242 config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
1243 } else if (arg_match(&arg, &buf_sz, argi)) {
1244 config->cfg.rc_buf_sz = arg_parse_uint(&arg);
1245 } else if (arg_match(&arg, &buf_initial_sz, argi)) {
1246 config->cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
1247 } else if (arg_match(&arg, &buf_optimal_sz, argi)) {
1248 config->cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
1249 } else if (arg_match(&arg, &bias_pct, argi)) {
1250 config->cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
1251 if (global->passes < 2)
1252 warn("option %s ignored in one-pass mode.\n", arg.name);
1253 } else if (arg_match(&arg, &minsection_pct, argi)) {
1254 config->cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
1256 if (global->passes < 2)
1257 warn("option %s ignored in one-pass mode.\n", arg.name);
1258 } else if (arg_match(&arg, &maxsection_pct, argi)) {
1259 config->cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
1261 if (global->passes < 2)
1262 warn("option %s ignored in one-pass mode.\n", arg.name);
1263 } else if (arg_match(&arg, &corpus_complexity, argi)) {
1264 config->cfg.rc_2pass_vbr_corpus_complexity = arg_parse_uint(&arg);
1266 if (global->passes < 2)
1267 warn("option %s ignored in one-pass mode.\n", arg.name);
1268 } else if (arg_match(&arg, &kf_min_dist, argi)) {
1269 config->cfg.kf_min_dist = arg_parse_uint(&arg);
1270 } else if (arg_match(&arg, &kf_max_dist, argi)) {
1271 config->cfg.kf_max_dist = arg_parse_uint(&arg);
1272 } else if (arg_match(&arg, &kf_disabled, argi)) {
1273 config->cfg.kf_mode = VPX_KF_DISABLED;
1274 #if CONFIG_VP9_HIGHBITDEPTH
1275 } else if (arg_match(&arg, &test16bitinternalarg, argi)) {
1276 if (strcmp(global->codec->name, "vp9") == 0) {
1277 test_16bit_internal = 1;
1282 for (i = 0; ctrl_args[i]; i++) {
1283 if (arg_match(&arg, ctrl_args[i], argi)) {
1287 /* Point either to the next free element or the first
1288 * instance of this control.
1290 for (j = 0; j < config->arg_ctrl_cnt; j++)
1291 if (ctrl_args_map != NULL &&
1292 config->arg_ctrls[j][0] == ctrl_args_map[i])
1296 assert(j < (int)ARG_CTRL_CNT_MAX);
1297 if (ctrl_args_map != NULL && j < (int)ARG_CTRL_CNT_MAX) {
1298 config->arg_ctrls[j][0] = ctrl_args_map[i];
1299 config->arg_ctrls[j][1] = arg_parse_enum_or_int(&arg);
1300 if (j == config->arg_ctrl_cnt) config->arg_ctrl_cnt++;
1307 #if CONFIG_VP9_HIGHBITDEPTH
1308 if (strcmp(global->codec->name, "vp9") == 0) {
1309 config->use_16bit_internal =
1310 test_16bit_internal | (config->cfg.g_profile > 1);
1313 return eos_mark_found;
1316 #define FOREACH_STREAM(func) \
1318 struct stream_state *stream; \
1319 for (stream = streams; stream; stream = stream->next) { \
1324 static void validate_stream_config(const struct stream_state *stream,
1325 const struct VpxEncoderConfig *global) {
1326 const struct stream_state *streami;
1329 if (!stream->config.cfg.g_w || !stream->config.cfg.g_h)
1331 "Stream %d: Specify stream dimensions with --width (-w) "
1332 " and --height (-h)",
1335 // Check that the codec bit depth is greater than the input bit depth.
1336 if (stream->config.cfg.g_input_bit_depth >
1337 (unsigned int)stream->config.cfg.g_bit_depth) {
1338 fatal("Stream %d: codec bit depth (%d) less than input bit depth (%d)",
1339 stream->index, (int)stream->config.cfg.g_bit_depth,
1340 stream->config.cfg.g_input_bit_depth);
1343 for (streami = stream; streami; streami = streami->next) {
1344 /* All streams require output files */
1345 if (!streami->config.out_fn)
1346 fatal("Stream %d: Output file is required (specify with -o)",
1349 /* Check for two streams outputting to the same file */
1350 if (streami != stream) {
1351 const char *a = stream->config.out_fn;
1352 const char *b = streami->config.out_fn;
1353 if (!strcmp(a, b) && strcmp(a, "/dev/null") && strcmp(a, ":nul"))
1354 fatal("Stream %d: duplicate output file (from stream %d)",
1355 streami->index, stream->index);
1358 /* Check for two streams sharing a stats file. */
1359 if (streami != stream) {
1360 const char *a = stream->config.stats_fn;
1361 const char *b = streami->config.stats_fn;
1362 if (a && b && !strcmp(a, b))
1363 fatal("Stream %d: duplicate stats file (from stream %d)",
1364 streami->index, stream->index);
1367 #if CONFIG_FP_MB_STATS
1368 /* Check for two streams sharing a mb stats file. */
1369 if (streami != stream) {
1370 const char *a = stream->config.fpmb_stats_fn;
1371 const char *b = streami->config.fpmb_stats_fn;
1372 if (a && b && !strcmp(a, b))
1373 fatal("Stream %d: duplicate mb stats file (from stream %d)",
1374 streami->index, stream->index);
1380 static void set_stream_dimensions(struct stream_state *stream, unsigned int w,
1382 if (!stream->config.cfg.g_w) {
1383 if (!stream->config.cfg.g_h)
1384 stream->config.cfg.g_w = w;
1386 stream->config.cfg.g_w = w * stream->config.cfg.g_h / h;
1388 if (!stream->config.cfg.g_h) {
1389 stream->config.cfg.g_h = h * stream->config.cfg.g_w / w;
1393 static const char *file_type_to_string(enum VideoFileType t) {
1395 case FILE_TYPE_RAW: return "RAW";
1396 case FILE_TYPE_Y4M: return "Y4M";
1397 default: return "Other";
1401 static const char *image_format_to_string(vpx_img_fmt_t f) {
1403 case VPX_IMG_FMT_I420: return "I420";
1404 case VPX_IMG_FMT_I422: return "I422";
1405 case VPX_IMG_FMT_I444: return "I444";
1406 case VPX_IMG_FMT_I440: return "I440";
1407 case VPX_IMG_FMT_YV12: return "YV12";
1408 case VPX_IMG_FMT_I42016: return "I42016";
1409 case VPX_IMG_FMT_I42216: return "I42216";
1410 case VPX_IMG_FMT_I44416: return "I44416";
1411 case VPX_IMG_FMT_I44016: return "I44016";
1412 default: return "Other";
1416 static void show_stream_config(struct stream_state *stream,
1417 struct VpxEncoderConfig *global,
1418 struct VpxInputContext *input) {
1419 #define SHOW(field) \
1420 fprintf(stderr, " %-28s = %d\n", #field, stream->config.cfg.field)
1422 if (stream->index == 0) {
1423 fprintf(stderr, "Codec: %s\n",
1424 vpx_codec_iface_name(global->codec->codec_interface()));
1425 fprintf(stderr, "Source file: %s File Type: %s Format: %s\n",
1426 input->filename, file_type_to_string(input->file_type),
1427 image_format_to_string(input->fmt));
1429 if (stream->next || stream->index)
1430 fprintf(stderr, "\nStream Index: %d\n", stream->index);
1431 fprintf(stderr, "Destination file: %s\n", stream->config.out_fn);
1432 fprintf(stderr, "Encoder parameters:\n");
1440 SHOW(g_input_bit_depth);
1441 SHOW(g_timebase.num);
1442 SHOW(g_timebase.den);
1443 SHOW(g_error_resilient);
1445 SHOW(g_lag_in_frames);
1446 SHOW(rc_dropframe_thresh);
1447 SHOW(rc_resize_allowed);
1448 SHOW(rc_scaled_width);
1449 SHOW(rc_scaled_height);
1450 SHOW(rc_resize_up_thresh);
1451 SHOW(rc_resize_down_thresh);
1453 SHOW(rc_target_bitrate);
1454 SHOW(rc_min_quantizer);
1455 SHOW(rc_max_quantizer);
1456 SHOW(rc_undershoot_pct);
1457 SHOW(rc_overshoot_pct);
1459 SHOW(rc_buf_initial_sz);
1460 SHOW(rc_buf_optimal_sz);
1461 SHOW(rc_2pass_vbr_bias_pct);
1462 SHOW(rc_2pass_vbr_minsection_pct);
1463 SHOW(rc_2pass_vbr_maxsection_pct);
1464 SHOW(rc_2pass_vbr_corpus_complexity);
1470 static void open_output_file(struct stream_state *stream,
1471 struct VpxEncoderConfig *global,
1472 const struct VpxRational *pixel_aspect_ratio) {
1473 const char *fn = stream->config.out_fn;
1474 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1476 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1478 stream->file = strcmp(fn, "-") ? fopen(fn, "wb") : set_binary_mode(stdout);
1480 if (!stream->file) fatal("Failed to open output file");
1482 if (stream->config.write_webm && fseek(stream->file, 0, SEEK_CUR))
1483 fatal("WebM output to pipes not supported.");
1486 if (stream->config.write_webm) {
1487 stream->webm_ctx.stream = stream->file;
1488 write_webm_file_header(&stream->webm_ctx, cfg, stream->config.stereo_fmt,
1489 global->codec->fourcc, pixel_aspect_ratio);
1492 (void)pixel_aspect_ratio;
1495 if (!stream->config.write_webm) {
1496 ivf_write_file_header(stream->file, cfg, global->codec->fourcc, 0);
1500 static void close_output_file(struct stream_state *stream,
1501 unsigned int fourcc) {
1502 const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
1504 if (cfg->g_pass == VPX_RC_FIRST_PASS) return;
1507 if (stream->config.write_webm) {
1508 write_webm_file_footer(&stream->webm_ctx);
1512 if (!stream->config.write_webm) {
1513 if (!fseek(stream->file, 0, SEEK_SET))
1514 ivf_write_file_header(stream->file, &stream->config.cfg, fourcc,
1515 stream->frames_out);
1518 fclose(stream->file);
1521 static void setup_pass(struct stream_state *stream,
1522 struct VpxEncoderConfig *global, int pass) {
1523 if (stream->config.stats_fn) {
1524 if (!stats_open_file(&stream->stats, stream->config.stats_fn, pass))
1525 fatal("Failed to open statistics store");
1527 if (!stats_open_mem(&stream->stats, pass))
1528 fatal("Failed to open statistics store");
1531 #if CONFIG_FP_MB_STATS
1532 if (stream->config.fpmb_stats_fn) {
1533 if (!stats_open_file(&stream->fpmb_stats, stream->config.fpmb_stats_fn,
1535 fatal("Failed to open mb statistics store");
1537 if (!stats_open_mem(&stream->fpmb_stats, pass))
1538 fatal("Failed to open mb statistics store");
1542 stream->config.cfg.g_pass = global->passes == 2
1543 ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
1546 stream->config.cfg.rc_twopass_stats_in = stats_get(&stream->stats);
1547 #if CONFIG_FP_MB_STATS
1548 stream->config.cfg.rc_firstpass_mb_stats_in =
1549 stats_get(&stream->fpmb_stats);
1553 stream->cx_time = 0;
1555 stream->frames_out = 0;
1558 static void initialize_encoder(struct stream_state *stream,
1559 struct VpxEncoderConfig *global) {
1563 flags |= global->show_psnr ? VPX_CODEC_USE_PSNR : 0;
1564 flags |= global->out_part ? VPX_CODEC_USE_OUTPUT_PARTITION : 0;
1565 #if CONFIG_VP9_HIGHBITDEPTH
1566 flags |= stream->config.use_16bit_internal ? VPX_CODEC_USE_HIGHBITDEPTH : 0;
1569 /* Construct Encoder Context */
1570 vpx_codec_enc_init(&stream->encoder, global->codec->codec_interface(),
1571 &stream->config.cfg, flags);
1572 ctx_exit_on_error(&stream->encoder, "Failed to initialize encoder");
1574 /* Note that we bypass the vpx_codec_control wrapper macro because
1575 * we're being clever to store the control IDs in an array. Real
1576 * applications will want to make use of the enumerations directly
1578 for (i = 0; i < stream->config.arg_ctrl_cnt; i++) {
1579 int ctrl = stream->config.arg_ctrls[i][0];
1580 int value = stream->config.arg_ctrls[i][1];
1581 if (vpx_codec_control_(&stream->encoder, ctrl, value))
1582 fprintf(stderr, "Error: Tried to set control %d = %d\n", ctrl, value);
1584 ctx_exit_on_error(&stream->encoder, "Failed to control codec");
1588 if (global->test_decode != TEST_DECODE_OFF) {
1589 const VpxInterface *decoder = get_vpx_decoder_by_name(global->codec->name);
1590 vpx_codec_dec_init(&stream->decoder, decoder->codec_interface(), NULL, 0);
1595 static void encode_frame(struct stream_state *stream,
1596 struct VpxEncoderConfig *global, struct vpx_image *img,
1597 unsigned int frames_in) {
1598 vpx_codec_pts_t frame_start, next_frame_start;
1599 struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1600 struct vpx_usec_timer timer;
1603 (cfg->g_timebase.den * (int64_t)(frames_in - 1) * global->framerate.den) /
1604 cfg->g_timebase.num / global->framerate.num;
1606 (cfg->g_timebase.den * (int64_t)(frames_in)*global->framerate.den) /
1607 cfg->g_timebase.num / global->framerate.num;
1609 /* Scale if necessary */
1610 #if CONFIG_VP9_HIGHBITDEPTH
1612 if ((img->fmt & VPX_IMG_FMT_HIGHBITDEPTH) &&
1613 (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1614 if (img->fmt != VPX_IMG_FMT_I42016) {
1615 fprintf(stderr, "%s can only scale 4:2:0 inputs\n", exec_name);
1621 vpx_img_alloc(NULL, VPX_IMG_FMT_I42016, cfg->g_w, cfg->g_h, 16);
1624 (uint16_t *)img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y] / 2,
1625 (uint16_t *)img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U] / 2,
1626 (uint16_t *)img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V] / 2,
1627 img->d_w, img->d_h, (uint16_t *)stream->img->planes[VPX_PLANE_Y],
1628 stream->img->stride[VPX_PLANE_Y] / 2,
1629 (uint16_t *)stream->img->planes[VPX_PLANE_U],
1630 stream->img->stride[VPX_PLANE_U] / 2,
1631 (uint16_t *)stream->img->planes[VPX_PLANE_V],
1632 stream->img->stride[VPX_PLANE_V] / 2, stream->img->d_w,
1633 stream->img->d_h, kFilterBox);
1636 stream->encoder.err = 1;
1637 ctx_exit_on_error(&stream->encoder,
1638 "Stream %d: Failed to encode frame.\n"
1639 "Scaling disabled in this configuration. \n"
1640 "To enable, configure with --enable-libyuv\n",
1646 if (img && (img->d_w != cfg->g_w || img->d_h != cfg->g_h)) {
1647 if (img->fmt != VPX_IMG_FMT_I420 && img->fmt != VPX_IMG_FMT_YV12) {
1648 fprintf(stderr, "%s can only scale 4:2:0 8bpp inputs\n", exec_name);
1654 vpx_img_alloc(NULL, VPX_IMG_FMT_I420, cfg->g_w, cfg->g_h, 16);
1656 img->planes[VPX_PLANE_Y], img->stride[VPX_PLANE_Y],
1657 img->planes[VPX_PLANE_U], img->stride[VPX_PLANE_U],
1658 img->planes[VPX_PLANE_V], img->stride[VPX_PLANE_V], img->d_w, img->d_h,
1659 stream->img->planes[VPX_PLANE_Y], stream->img->stride[VPX_PLANE_Y],
1660 stream->img->planes[VPX_PLANE_U], stream->img->stride[VPX_PLANE_U],
1661 stream->img->planes[VPX_PLANE_V], stream->img->stride[VPX_PLANE_V],
1662 stream->img->d_w, stream->img->d_h, kFilterBox);
1665 stream->encoder.err = 1;
1666 ctx_exit_on_error(&stream->encoder,
1667 "Stream %d: Failed to encode frame.\n"
1668 "Scaling disabled in this configuration. \n"
1669 "To enable, configure with --enable-libyuv\n",
1674 vpx_usec_timer_start(&timer);
1675 vpx_codec_encode(&stream->encoder, img, frame_start,
1676 (unsigned long)(next_frame_start - frame_start), 0,
1678 vpx_usec_timer_mark(&timer);
1679 stream->cx_time += vpx_usec_timer_elapsed(&timer);
1680 ctx_exit_on_error(&stream->encoder, "Stream %d: Failed to encode frame",
1684 static void update_quantizer_histogram(struct stream_state *stream) {
1685 if (stream->config.cfg.g_pass != VPX_RC_FIRST_PASS) {
1688 vpx_codec_control(&stream->encoder, VP8E_GET_LAST_QUANTIZER_64, &q);
1689 ctx_exit_on_error(&stream->encoder, "Failed to read quantizer");
1690 stream->counts[q]++;
1694 static void get_cx_data(struct stream_state *stream,
1695 struct VpxEncoderConfig *global, int *got_data) {
1696 const vpx_codec_cx_pkt_t *pkt;
1697 const struct vpx_codec_enc_cfg *cfg = &stream->config.cfg;
1698 vpx_codec_iter_t iter = NULL;
1701 while ((pkt = vpx_codec_get_cx_data(&stream->encoder, &iter))) {
1702 static size_t fsize = 0;
1703 static FileOffset ivf_header_pos = 0;
1705 switch (pkt->kind) {
1706 case VPX_CODEC_CX_FRAME_PKT:
1707 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1708 stream->frames_out++;
1711 fprintf(stderr, " %6luF", (unsigned long)pkt->data.frame.sz);
1713 update_rate_histogram(stream->rate_hist, cfg, pkt);
1715 if (stream->config.write_webm) {
1716 write_webm_block(&stream->webm_ctx, cfg, pkt);
1719 if (!stream->config.write_webm) {
1720 if (pkt->data.frame.partition_id <= 0) {
1721 ivf_header_pos = ftello(stream->file);
1722 fsize = pkt->data.frame.sz;
1724 ivf_write_frame_header(stream->file, pkt->data.frame.pts, fsize);
1726 fsize += pkt->data.frame.sz;
1728 if (!(pkt->data.frame.flags & VPX_FRAME_IS_FRAGMENT)) {
1729 const FileOffset currpos = ftello(stream->file);
1730 fseeko(stream->file, ivf_header_pos, SEEK_SET);
1731 ivf_write_frame_size(stream->file, fsize);
1732 fseeko(stream->file, currpos, SEEK_SET);
1736 (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
1739 stream->nbytes += pkt->data.raw.sz;
1743 if (global->test_decode != TEST_DECODE_OFF && !stream->mismatch_seen) {
1744 vpx_codec_decode(&stream->decoder, pkt->data.frame.buf,
1745 (unsigned int)pkt->data.frame.sz, NULL, 0);
1746 if (stream->decoder.err) {
1747 warn_or_exit_on_error(&stream->decoder,
1748 global->test_decode == TEST_DECODE_FATAL,
1749 "Failed to decode frame %d in stream %d",
1750 stream->frames_out + 1, stream->index);
1751 stream->mismatch_seen = stream->frames_out + 1;
1756 case VPX_CODEC_STATS_PKT:
1757 stream->frames_out++;
1758 stats_write(&stream->stats, pkt->data.twopass_stats.buf,
1759 pkt->data.twopass_stats.sz);
1760 stream->nbytes += pkt->data.raw.sz;
1762 #if CONFIG_FP_MB_STATS
1763 case VPX_CODEC_FPMB_STATS_PKT:
1764 stats_write(&stream->fpmb_stats, pkt->data.firstpass_mb_stats.buf,
1765 pkt->data.firstpass_mb_stats.sz);
1766 stream->nbytes += pkt->data.raw.sz;
1769 case VPX_CODEC_PSNR_PKT:
1771 if (global->show_psnr) {
1774 stream->psnr_sse_total += pkt->data.psnr.sse[0];
1775 stream->psnr_samples_total += pkt->data.psnr.samples[0];
1776 for (i = 0; i < 4; i++) {
1778 fprintf(stderr, "%.3f ", pkt->data.psnr.psnr[i]);
1779 stream->psnr_totals[i] += pkt->data.psnr.psnr[i];
1781 stream->psnr_count++;
1790 static void show_psnr(struct stream_state *stream, double peak) {
1794 if (!stream->psnr_count) return;
1796 fprintf(stderr, "Stream %d PSNR (Overall/Avg/Y/U/V)", stream->index);
1797 ovpsnr = sse_to_psnr((double)stream->psnr_samples_total, peak,
1798 (double)stream->psnr_sse_total);
1799 fprintf(stderr, " %.3f", ovpsnr);
1801 for (i = 0; i < 4; i++) {
1802 fprintf(stderr, " %.3f", stream->psnr_totals[i] / stream->psnr_count);
1804 fprintf(stderr, "\n");
1807 static float usec_to_fps(uint64_t usec, unsigned int frames) {
1808 return (float)(usec > 0 ? frames * 1000000.0 / (float)usec : 0);
1811 static void test_decode(struct stream_state *stream,
1812 enum TestDecodeFatality fatal,
1813 const VpxInterface *codec) {
1814 vpx_image_t enc_img, dec_img;
1816 if (stream->mismatch_seen) return;
1818 /* Get the internal reference frame */
1819 if (strcmp(codec->name, "vp8") == 0) {
1820 struct vpx_ref_frame ref_enc, ref_dec;
1823 width = (stream->config.cfg.g_w + 15) & ~15;
1824 height = (stream->config.cfg.g_h + 15) & ~15;
1825 vpx_img_alloc(&ref_enc.img, VPX_IMG_FMT_I420, width, height, 1);
1826 enc_img = ref_enc.img;
1827 vpx_img_alloc(&ref_dec.img, VPX_IMG_FMT_I420, width, height, 1);
1828 dec_img = ref_dec.img;
1830 ref_enc.frame_type = VP8_LAST_FRAME;
1831 ref_dec.frame_type = VP8_LAST_FRAME;
1832 vpx_codec_control(&stream->encoder, VP8_COPY_REFERENCE, &ref_enc);
1833 vpx_codec_control(&stream->decoder, VP8_COPY_REFERENCE, &ref_dec);
1835 struct vp9_ref_frame ref_enc, ref_dec;
1839 vpx_codec_control(&stream->encoder, VP9_GET_REFERENCE, &ref_enc);
1840 enc_img = ref_enc.img;
1841 vpx_codec_control(&stream->decoder, VP9_GET_REFERENCE, &ref_dec);
1842 dec_img = ref_dec.img;
1843 #if CONFIG_VP9_HIGHBITDEPTH
1844 if ((enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) !=
1845 (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH)) {
1846 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1847 vpx_img_alloc(&enc_img, enc_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1848 enc_img.d_w, enc_img.d_h, 16);
1849 vpx_img_truncate_16_to_8(&enc_img, &ref_enc.img);
1851 if (dec_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1852 vpx_img_alloc(&dec_img, dec_img.fmt - VPX_IMG_FMT_HIGHBITDEPTH,
1853 dec_img.d_w, dec_img.d_h, 16);
1854 vpx_img_truncate_16_to_8(&dec_img, &ref_dec.img);
1859 ctx_exit_on_error(&stream->encoder, "Failed to get encoder reference frame");
1860 ctx_exit_on_error(&stream->decoder, "Failed to get decoder reference frame");
1862 if (!compare_img(&enc_img, &dec_img)) {
1863 int y[4], u[4], v[4];
1864 #if CONFIG_VP9_HIGHBITDEPTH
1865 if (enc_img.fmt & VPX_IMG_FMT_HIGHBITDEPTH) {
1866 find_mismatch_high(&enc_img, &dec_img, y, u, v);
1868 find_mismatch(&enc_img, &dec_img, y, u, v);
1871 find_mismatch(&enc_img, &dec_img, y, u, v);
1873 stream->decoder.err = 1;
1874 warn_or_exit_on_error(&stream->decoder, fatal == TEST_DECODE_FATAL,
1875 "Stream %d: Encode/decode mismatch on frame %d at"
1876 " Y[%d, %d] {%d/%d},"
1877 " U[%d, %d] {%d/%d},"
1878 " V[%d, %d] {%d/%d}",
1879 stream->index, stream->frames_out, y[0], y[1], y[2],
1880 y[3], u[0], u[1], u[2], u[3], v[0], v[1], v[2], v[3]);
1881 stream->mismatch_seen = stream->frames_out;
1884 vpx_img_free(&enc_img);
1885 vpx_img_free(&dec_img);
1888 static void print_time(const char *label, int64_t etl) {
1895 etl -= hours * 3600;
1900 fprintf(stderr, "[%3s %2" PRId64 ":%02" PRId64 ":%02" PRId64 "] ", label,
1903 fprintf(stderr, "[%3s unknown] ", label);
1907 int main(int argc, const char **argv_) {
1910 #if CONFIG_VP9_HIGHBITDEPTH
1911 vpx_image_t raw_shift;
1912 int allocated_raw_shift = 0;
1913 int use_16bit_internal = 0;
1914 int input_shift = 0;
1916 int frame_avail, got_data;
1918 struct VpxInputContext input;
1919 struct VpxEncoderConfig global;
1920 struct stream_state *streams = NULL;
1921 char **argv, **argi;
1922 uint64_t cx_time = 0;
1926 memset(&input, 0, sizeof(input));
1927 exec_name = argv_[0];
1929 /* Setup default input stream settings */
1930 input.framerate.numerator = 30;
1931 input.framerate.denominator = 1;
1932 input.only_i420 = 1;
1933 input.bit_depth = 0;
1935 /* First parse the global configuration values, because we want to apply
1936 * other parameters on top of the default configuration provided by the
1939 argv = argv_dup(argc - 1, argv_ + 1);
1940 parse_global_config(&global, argv);
1942 if (argc < 3) usage_exit();
1944 switch (global.color_type) {
1945 case I420: input.fmt = VPX_IMG_FMT_I420; break;
1946 case I422: input.fmt = VPX_IMG_FMT_I422; break;
1947 case I444: input.fmt = VPX_IMG_FMT_I444; break;
1948 case I440: input.fmt = VPX_IMG_FMT_I440; break;
1949 case YV12: input.fmt = VPX_IMG_FMT_YV12; break;
1953 /* Now parse each stream's parameters. Using a local scope here
1954 * due to the use of 'stream' as loop variable in FOREACH_STREAM
1957 struct stream_state *stream = NULL;
1960 stream = new_stream(&global, stream);
1962 if (!streams) streams = stream;
1963 } while (parse_stream_params(&global, stream, argv));
1966 /* Check for unrecognized options */
1967 for (argi = argv; *argi; argi++)
1968 if (argi[0][0] == '-' && argi[0][1])
1969 die("Error: Unrecognized option %s\n", *argi);
1971 FOREACH_STREAM(check_encoder_config(global.disable_warning_prompt, &global,
1972 &stream->config.cfg););
1974 /* Handle non-option arguments */
1975 input.filename = argv[0];
1977 if (!input.filename) {
1978 fprintf(stderr, "No input file specified!\n");
1982 /* Decide if other chroma subsamplings than 4:2:0 are supported */
1983 if (global.codec->fourcc == VP9_FOURCC) input.only_i420 = 0;
1985 for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
1986 int frames_in = 0, seen_frames = 0;
1987 int64_t estimated_time_left = -1;
1988 int64_t average_rate = -1;
1989 int64_t lagged_count = 0;
1991 open_input_file(&input);
1993 /* If the input file doesn't specify its w/h (raw files), try to get
1994 * the data from the first stream's configuration.
1996 if (!input.width || !input.height) {
1998 if (stream->config.cfg.g_w && stream->config.cfg.g_h) {
1999 input.width = stream->config.cfg.g_w;
2000 input.height = stream->config.cfg.g_h;
2006 /* Update stream configurations from the input file's parameters */
2007 if (!input.width || !input.height)
2009 "Specify stream dimensions with --width (-w) "
2010 " and --height (-h)");
2012 /* If input file does not specify bit-depth but input-bit-depth parameter
2013 * exists, assume that to be the input bit-depth. However, if the
2014 * input-bit-depth paramter does not exist, assume the input bit-depth
2015 * to be the same as the codec bit-depth.
2017 if (!input.bit_depth) {
2019 if (stream->config.cfg.g_input_bit_depth)
2020 input.bit_depth = stream->config.cfg.g_input_bit_depth;
2022 input.bit_depth = stream->config.cfg.g_input_bit_depth =
2023 (int)stream->config.cfg.g_bit_depth;
2025 if (input.bit_depth > 8) input.fmt |= VPX_IMG_FMT_HIGHBITDEPTH;
2028 { stream->config.cfg.g_input_bit_depth = input.bit_depth; });
2031 FOREACH_STREAM(set_stream_dimensions(stream, input.width, input.height));
2032 FOREACH_STREAM(validate_stream_config(stream, &global));
2034 /* Ensure that --passes and --pass are consistent. If --pass is set and
2035 * --passes=2, ensure --fpf was set.
2037 if (global.pass && global.passes == 2)
2039 if (!stream->config.stats_fn)
2040 die("Stream %d: Must specify --fpf when --pass=%d"
2041 " and --passes=2\n",
2042 stream->index, global.pass);
2047 if (stream->config.write_webm) {
2048 stream->config.write_webm = 0;
2050 "vpxenc was compiled without WebM container support."
2051 "Producing IVF output");
2056 /* Use the frame rate from the file only if none was specified
2057 * on the command-line.
2059 if (!global.have_framerate) {
2060 global.framerate.num = input.framerate.numerator;
2061 global.framerate.den = input.framerate.denominator;
2062 FOREACH_STREAM(stream->config.cfg.g_timebase.den = global.framerate.num;
2063 stream->config.cfg.g_timebase.num = global.framerate.den);
2066 /* Show configuration */
2067 if (global.verbose && pass == 0)
2068 FOREACH_STREAM(show_stream_config(stream, &global, &input));
2070 if (pass == (global.pass ? global.pass - 1 : 0)) {
2071 if (input.file_type == FILE_TYPE_Y4M)
2072 /*The Y4M reader does its own allocation.
2073 Just initialize this here to avoid problems if we never read any
2075 memset(&raw, 0, sizeof(raw));
2077 vpx_img_alloc(&raw, input.fmt, input.width, input.height, 32);
2079 FOREACH_STREAM(stream->rate_hist = init_rate_histogram(
2080 &stream->config.cfg, &global.framerate));
2083 FOREACH_STREAM(setup_pass(stream, &global, pass));
2085 open_output_file(stream, &global, &input.pixel_aspect_ratio));
2086 FOREACH_STREAM(initialize_encoder(stream, &global));
2088 #if CONFIG_VP9_HIGHBITDEPTH
2089 if (strcmp(global.codec->name, "vp9") == 0) {
2090 // Check to see if at least one stream uses 16 bit internal.
2091 // Currently assume that the bit_depths for all streams using
2092 // highbitdepth are the same.
2094 if (stream->config.use_16bit_internal) {
2095 use_16bit_internal = 1;
2097 if (stream->config.cfg.g_profile == 0) {
2100 input_shift = (int)stream->config.cfg.g_bit_depth -
2101 stream->config.cfg.g_input_bit_depth;
2110 while (frame_avail || got_data) {
2111 struct vpx_usec_timer timer;
2113 if (!global.limit || frames_in < global.limit) {
2114 frame_avail = read_frame(&input, &raw);
2116 if (frame_avail) frames_in++;
2118 frames_in > global.skip_frames ? frames_in - global.skip_frames : 0;
2120 if (!global.quiet) {
2121 float fps = usec_to_fps(cx_time, seen_frames);
2122 fprintf(stderr, "\rPass %d/%d ", pass + 1, global.passes);
2124 if (stream_cnt == 1)
2125 fprintf(stderr, "frame %4d/%-4d %7" PRId64 "B ", frames_in,
2126 streams->frames_out, (int64_t)streams->nbytes);
2128 fprintf(stderr, "frame %4d ", frames_in);
2130 fprintf(stderr, "%7" PRId64 " %s %.2f %s ",
2131 cx_time > 9999999 ? cx_time / 1000 : cx_time,
2132 cx_time > 9999999 ? "ms" : "us", fps >= 1.0 ? fps : fps * 60,
2133 fps >= 1.0 ? "fps" : "fpm");
2134 print_time("ETA", estimated_time_left);
2140 if (frames_in > global.skip_frames) {
2141 #if CONFIG_VP9_HIGHBITDEPTH
2142 vpx_image_t *frame_to_encode;
2143 if (input_shift || (use_16bit_internal && input.bit_depth == 8)) {
2144 assert(use_16bit_internal);
2145 // Input bit depth and stream bit depth do not match, so up
2146 // shift frame to stream bit depth
2147 if (!allocated_raw_shift) {
2148 vpx_img_alloc(&raw_shift, raw.fmt | VPX_IMG_FMT_HIGHBITDEPTH,
2149 input.width, input.height, 32);
2150 allocated_raw_shift = 1;
2152 vpx_img_upshift(&raw_shift, &raw, input_shift);
2153 frame_to_encode = &raw_shift;
2155 frame_to_encode = &raw;
2157 vpx_usec_timer_start(&timer);
2158 if (use_16bit_internal) {
2159 assert(frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH);
2161 if (stream->config.use_16bit_internal)
2162 encode_frame(stream, &global,
2163 frame_avail ? frame_to_encode : NULL, frames_in);
2168 assert((frame_to_encode->fmt & VPX_IMG_FMT_HIGHBITDEPTH) == 0);
2169 FOREACH_STREAM(encode_frame(stream, &global,
2170 frame_avail ? frame_to_encode : NULL,
2174 vpx_usec_timer_start(&timer);
2175 FOREACH_STREAM(encode_frame(stream, &global, frame_avail ? &raw : NULL,
2178 vpx_usec_timer_mark(&timer);
2179 cx_time += vpx_usec_timer_elapsed(&timer);
2181 FOREACH_STREAM(update_quantizer_histogram(stream));
2184 FOREACH_STREAM(get_cx_data(stream, &global, &got_data));
2186 if (!got_data && input.length && streams != NULL &&
2187 !streams->frames_out) {
2188 lagged_count = global.limit ? seen_frames : ftello(input.file);
2189 } else if (input.length) {
2194 const int64_t frame_in_lagged = (seen_frames - lagged_count) * 1000;
2196 rate = cx_time ? frame_in_lagged * (int64_t)1000000 / cx_time : 0;
2197 remaining = 1000 * (global.limit - global.skip_frames -
2198 seen_frames + lagged_count);
2200 const int64_t input_pos = ftello(input.file);
2201 const int64_t input_pos_lagged = input_pos - lagged_count;
2202 const int64_t limit = input.length;
2204 rate = cx_time ? input_pos_lagged * (int64_t)1000000 / cx_time : 0;
2205 remaining = limit - input_pos + lagged_count;
2209 (average_rate <= 0) ? rate : (average_rate * 7 + rate) / 8;
2210 estimated_time_left = average_rate ? remaining / average_rate : -1;
2213 if (got_data && global.test_decode != TEST_DECODE_OFF)
2214 FOREACH_STREAM(test_decode(stream, global.test_decode, global.codec));
2218 if (!global.quiet) fprintf(stderr, "\033[K");
2221 if (stream_cnt > 1) fprintf(stderr, "\n");
2223 if (!global.quiet) {
2224 FOREACH_STREAM(fprintf(
2226 "\rPass %d/%d frame %4d/%-4d %7" PRId64 "B %7" PRId64 "b/f %7" PRId64
2227 "b/s %7" PRId64 " %s (%.2f fps)\033[K\n",
2228 pass + 1, global.passes, frames_in, stream->frames_out,
2229 (int64_t)stream->nbytes,
2230 seen_frames ? (int64_t)(stream->nbytes * 8 / seen_frames) : 0,
2232 ? (int64_t)stream->nbytes * 8 * (int64_t)global.framerate.num /
2233 global.framerate.den / seen_frames
2235 stream->cx_time > 9999999 ? stream->cx_time / 1000 : stream->cx_time,
2236 stream->cx_time > 9999999 ? "ms" : "us",
2237 usec_to_fps(stream->cx_time, seen_frames)));
2240 if (global.show_psnr) {
2241 if (global.codec->fourcc == VP9_FOURCC) {
2243 show_psnr(stream, (1 << stream->config.cfg.g_input_bit_depth) - 1));
2245 FOREACH_STREAM(show_psnr(stream, 255.0));
2249 FOREACH_STREAM(vpx_codec_destroy(&stream->encoder));
2251 if (global.test_decode != TEST_DECODE_OFF) {
2252 FOREACH_STREAM(vpx_codec_destroy(&stream->decoder));
2255 close_input_file(&input);
2257 if (global.test_decode == TEST_DECODE_FATAL) {
2258 FOREACH_STREAM(res |= stream->mismatch_seen);
2260 FOREACH_STREAM(close_output_file(stream, global.codec->fourcc));
2262 FOREACH_STREAM(stats_close(&stream->stats, global.passes - 1));
2264 #if CONFIG_FP_MB_STATS
2265 FOREACH_STREAM(stats_close(&stream->fpmb_stats, global.passes - 1));
2268 if (global.pass) break;
2271 if (global.show_q_hist_buckets)
2273 show_q_histogram(stream->counts, global.show_q_hist_buckets));
2275 if (global.show_rate_hist_buckets)
2276 FOREACH_STREAM(show_rate_histogram(stream->rate_hist, &stream->config.cfg,
2277 global.show_rate_hist_buckets));
2278 FOREACH_STREAM(destroy_rate_histogram(stream->rate_hist));
2280 #if CONFIG_INTERNAL_STATS
2281 /* TODO(jkoleszar): This doesn't belong in this executable. Do it for now,
2282 * to match some existing utilities.
2284 if (!(global.pass == 1 && global.passes == 2))
2286 FILE *f = fopen("opsnr.stt", "a");
2287 if (stream->mismatch_seen) {
2288 fprintf(f, "First mismatch occurred in frame %d\n",
2289 stream->mismatch_seen);
2291 fprintf(f, "No mismatch detected in recon buffers\n");
2297 #if CONFIG_VP9_HIGHBITDEPTH
2298 if (allocated_raw_shift) vpx_img_free(&raw_shift);
2303 return res ? EXIT_FAILURE : EXIT_SUCCESS;